αTAT1 controls longitudinal spreading of acetylation marks from open microtubules extremities

Acetylation of the lysine 40 of α-tubulin (K40) is a post-translational modification occurring in the lumen of microtubules (MTs) and is controlled by the α-tubulin acetyl-transferase αTAT1. How αTAT1 accesses the lumen and acetylates α-tubulin there has been an open question. Here, we report that acetylation starts at open ends of MTs and progressively spreads longitudinally from there. We observed acetylation marks at the open ends of in vivo MTs re-growing after a Nocodazole block, and acetylated segments growing in length with time. Bias for MTs extremities was even more pronounced when using non-dynamic MTs extracted from HeLa cells. In contrast, K40 acetylation was mostly uniform along the length of MTs reconstituted from purified tubulin in vitro. Quantitative modelling of luminal diffusion of αTAT1 suggested that the uniform acetylation pattern observed in vitro is consistent with defects in the MT lattice providing lateral access to the lumen. Indeed, we observed that in vitro MTs are permeable to macromolecules along their shaft while cellular MTs are not. Our results demonstrate αTAT1 enters the lumen from open extremities and spreads K40 acetylation marks longitudinally along cellular MTs. This mode of tip-directed microtubule acetylation may allow for selective acetylation of subsets of microtubules.


Results and Discussion
Microtubules (MTs) are dynamic polymers composed of α β -tubulin dimers that assembled into hollow tubes. In most eukaryotic cells, MTs can undergo post-translational modifications (PTMs) that modify their properties and functions 1 . Acetylation of the lysine 40 of α -tubulin (K40) is a common PTM that is catalysed by the α -tubulin acetyl-transferase α TAT1 and is associated with stable, long-lived MTs [2][3][4] . Remarkably, K40 acetylation occurs in the lumen of MTs 5,6 and is the only such PTM that we know of ref. 1. Supporting this, Szyk et al. recently used in vitro approaches to demonstrate that α TAT1 enters into and diffuses within the MT lumen 7 . However, Szyk et al. also suggested that fast diffusivity of α TAT1 leads to stochastic acetylation that occurs uniformly along the length of MTs. This was in marked contrast with earlier in vivo observations of discrete acetylated segments along MTs 8-10 progressively elongating with time 11 . More recently, several groups have reported that the acetylated segments were predominately associated with the ends of MTs in vivo 3,12 . Thus, reported observations in vivo do not match the proposed model of uniformly distributed acetylated K40 marks based on experiments performed with in vitro MTs 7 .
To understand how acetylated K40 marks spreading occurs in vivo, we first analyzed acetylation dynamics in HeLa cells. In order to synchronize acetylation events, HeLa cells were subjected to complete MT depolymerisation by a prolonged treatment with Nocodazole before being allowed to reassemble MTs after washing out the Our observation of a preference for acetylation at MT extremities agrees with earlier reports in vivo 3,12 and suggests that α TAT1 accesses the lumen from the open ends of MTs and starts acetylating at those extremities. However, we also observed acetylation segments far from in vivo microtubule ends (Fig. 1C). While this might denote alternative lateral α TAT1 entry sites, we hypothesize instead that the extremely dynamic nature of microtubules in vivo allows unacetylated extensions to grow past acetylated K40-positive extremities, leaving acetylated segments behind. Indeed, microtubule polymerization is much faster than acetylation spreading (in the order of ~10 μ m/min while acetylated segments elongate only 2 μ m in 8 min as measured here). In this case, non-dynamic MTs should exhibit more acetylation segments at MT ends.
To test this hypothesis, we used α TAT1-knockdown HeLa cells extracted by a brief immersion in a MT stabilizing buffer containing Triton. In these conditions, Taxol-stabilized ex vivo cellular MTs were readily visible and were negative for anti-acetylated K40 staining ( Fig. 2A, upper panels). Ex vivo MTs were then incubated with 4 μ M of a recombinant catalytic domain of mouse α TAT1 (residues  in the presence of Acetyl-CoA for different time periods. Acetylated segments of MTs became visible as early as 30 s after addition of the recombinant enzyme and these segments were found to grow longer with time ( Supplementary Figure 1), similarly to our in vivo observations. We then analysed the distribution of acetylated segments along MTs and observed that virtually all acetylated segments detected after a 2 min incubation period with the enzyme were located at the ends of individual MTs, with no detectable staining in other MT regions ( Fig. 2A,   Measurements of fluorescence intensity showed that while total tubulin staining was constant along the length of ex vivo MTs (Supplementary Figure 3A), a strong bias for the extremities was observed for the acetylated K40 signal (Fig. 2B). We also noted a time-dependent progressive increase of staining intensity near the tips along with a progressive longitudinal (axial) spreading of acetylated K40 marks (Fig. 2B). We obtained similar results when using ex vivo MTs that were not stabilized with Taxol (Supplementary Figure 3B). In these latter conditions, the maximum intensity of acetylated K40 staining at the extremity was lower than in Taxol-stabilized ex vivo MTs. This may be due to a slow depolymerisation of MT ends leading to a partial loss of acetylation signal. Also, we performed acetylation assays on Taxol-stabilized ex vivo MTs by using a lower concentration (0.4 μ M) of recombinant α TAT1. While the kinetics of acetylation was reduced, we observed the same preference for MT extremities and a progressive spreading of acetylated K40 marks from the ends (Supplementary Figure 3C). Together, our analysis unambiguously demonstrates that acetylation spreads progressively, longitudinally along ex vivo MTs, from their open extremities. Our data also suggest that acetylated segments observed far from MT ends in vivo most likely results from the dynamicity of MTs.
Because this conclusion disagrees with reports of uniform acetylation in vitro 7 , we next decided to analyse acetylation dynamics in Taxol-stabilized MTs assembled in vitro from purified HeLa cells tubulin dimers. Because HeLa tubulin dimers are poorly acetylated, no acetylated K40 marks were observed immediately after polymerization in vitro (Fig. 3A, left panel). However, after a 2 minutes incubation period in the presence of 4 μ M recombinant α TAT1, we observed a punctuated acetylated K40 staining with an approximately random distribution along the length of MTs (Fig. 3A, middle panel). After a 4 minutes incubation period with the enzyme, the acetylated K40 signal was more intense but still distributed randomly along the length of the MTs. Careful examination revealed that extremities were often positive for acetylated K40 marks (Fig. 3A), though this was less striking than in vivo (Fig. 1C). The average fluorescence intensity distribution also showed a slight enrichment of acetylated K40 marks at open extremities as compare to other regions of MTs (Fig. 3C). We obtained similar results when using MTs assembled in vitro from purified bovine brain tubulin (not shown). In agreement with observations by Szyk et al., we concluded that acetylation occurs randomly (uniformly) along the length of in vitro MTs 7 . This is in marked contrast with our in vivo and ex vivo observations and therefore suggests that in vitro and cellular MTs may differ in their intrinsic properties.
Structural defects in the MT lattice, such as missing tubulin dimers or abrupt variations of protofilament number have been reported in in vitro assembled MTs 13,14 . It was recently proposed that stress accumulation at these lattice defects can results in the formation of larger damages that self-repair rapidly 14,15 . Indeed, transient opening of the MT lattice resulting from separation between protofilaments have been hypothesized to provide direct access to the MT lumen in a proposed "breathing model" 16,17 . To explore the hypothesis that the differences we observed were due to transient holes or defects on the sides of in vitro MTs, we have developed a mathematical model of α TAT1 diffusion within MTs followed by local acetylation by α TAT1. We have included both lateral access (due to holes), and access from MT extremities, followed by longitudinal diffusion as characterized by Szyk et al. 7 (see Experimental Procedures section for details). When lateral access was not allowed, we observed that acetylation marks spreads longitudinally from the open MT extremities (Fig. 3D). This is in qualitative agreement with our ex vivo results. However, when lateral access was implemented in the model using α TAT1 residence times determined by Szyk et al. 7 , acetylation was uniform along the length of MTs (Fig. 3E). Thus, our modelling captures the qualitative difference between the mostly uniform acetylation in vitro and the progressive acetylation ex vivo and suggests that the shaft of in vitro reconstituted MTs is permeable to α TAT1 while the one of cellular MTs is not. With our quantitative model of longitudinal diffusion, local acetylation, and possible lateral entry we observed that tip-directed acetylation is consistent with the measured luminal diffusivity of α TAT1 7 , together with a moderate acetylation rate and vanishing lateral entry. Conversely, we showed that uniform acetylation is consistent with rapid lateral entry (with respect to the times at which measurements are taken) independent of the other parameters. Are other parameterizations, perhaps without lateral entry, consistent with the observed phenomenology? Without lateral entry, tip-directed acetylation should be observed at earlier times in longer MT. The earliest reported uniform acetylation was at 15s in 10 μ m in vitro MT 7 , which we found (data not shown) requires luminal diffusivities at least 5 times larger than measured by Szyk et al. if uniformity of α TAT1 within 20% of the maximum is required. For 20 μ m long MT, at least 20 times larger diffusivities are needed. However, such large luminal diffusivities preclude the tip-directed acetylation we observe in vivo. Since we have no reason to expect luminal diffusivities to hugely differ between in vivo and in vitro MT, we conclude that uniform acetylation observed for in vitro MT requires rapid lateral access of α TAT1.
If defects or holes in lattice allow α TAT1 to access the lumen of MTs assembled in vitro, the same could be true for other macromolecules. It has been reported that antibodies targeting luminal epitopes cannot access to the lumen of unfixed cellular MTs 18 . Indeed, while fixed, Triton-extracted HeLa cell MTs showed a classical anti-acetylated K40 staining, only a background staining was detectable in unfixed conditions (Fig. 4A). In sharp contrast, anti-acetylated K40 staining was visible along both fixed and unfixed in vitro MTs (Fig. 4B). We concluded that, in unfixed conditions, the anti-acetylated K40 antibody can access the lumen of in vitro but not cellular MTs. Our data suggest that the shaft of in vitro MTs is leaky and allows lateral access of macromolecules to the lumen. This agrees with both our modelling, and the observed uniform acetylation patterns of in vitro MTs.
It has been suggested that α TAT1 could bind to the outside of MTs and that this could facilitate targeting the enzyme to the lumen 19 . To conclude, our data suggest that in vitro assembled MTs presents holes or defects along their shaft that allow lateral entry of α TAT1 to the lumen. The nature of such holes and why in vitro MTs would display more holes as compare to cellular MTs is not clear. One possibility is that MTs growing in the cellular context would be protected from developing or acquiring lateral access-promoting defects. Along this line, dislocation defects due to abrupt transition in the number of protofilaments have been reported to be twice as common in vitro as compare to in vivo conditions 13 . This raises the possibility that growth conditions for in vitro MT could be adjusted to reduce the number of holes or defects. Alternatively, microtubule-associated proteins (MAPs) might block lateral holes in cellular MTs, preventing α TAT1 access to the lumen at these sites. In this case, appropriate MAPs would need to be included in in vitro acetylation assays to recapitulate in vivo results.
Our results demonstrate that in vitro reconstituted MTs do not recapitulate the properties of cellular MTs and we urge great caution in studying the effects of MT acetylation with such MTs. Based on the apparent stochastic acetylation pattern in vitro, it was proposed that α TAT1 controls a selective, age-dependent MT acetylation 7 . In contrast to this model, the major finding of our work is that α TAT1 controls the longitudinal spreading of acetylation marks from open MT extremities in physiological conditions. In our model, the selective acetylation of a subset of MTs that is observed in vivo is not primarily controlled by the stability of this subset but rather by the chance that a given MT would acquire α TAT1 at its extremities. MTs have been shown to contact α TAT1-rich structures like clathrin-coated pits and focal adhesions, two types of structures that mostly accumulate at the front of migrating cells 12,20,21 . We propose that multiple contacts between MTs and these structures would enhance α TAT1 acquisition by front-oriented MTs leading to the progressive acetylation of this MT subset from their open extremities. This selective, tip-oriented acetylation mechanism has important consequences since cell-front oriented acetylated MTs are instrumental in controlling directional cell migration 12 .

Methods
Cell culture. HeLa cells (a gift from A. Dautry, Institut Pasteur, Paris, France) were grown in DMEM, high glucose, GlutaMax ™ supplement media, supplemented with 10% fetal calf serum at 37 °C in 5% CO 2 .
Antibodies and reagents. Mouse monoclonal anti-K40 acetylated tubulin (clone 6-11B1) was purchased from Sigma. Recombinant humanised anti-α -tubulin (clone F2C-hFc2) and Cy3-conjugated anti-human antibodies were purchased from the antibody platform of the Institut Curie (Paris, France). DyLight488-conjugated anti-mouse antibodies were from Jackson Immunoresearch. Nocodazole was purchased from Sigma. of mouse α TAT1 was obtained by PCR by using cDNA of C-terminally GFP-tagged mMEC17 as a template (Montagnac et al., Nature, 2013). PCR fragment with engineered flanking restriction sites was subcloned into the multi-cloning sites of pGEX4T1 (Amersham Pharmacia Biotech) to encode in-frame fusion proteins with the amino-terminal GST-tag. The construct was verified by double-stranded DNA sequencing. Purification of the recombinant protein was performed using standard protocols as described 12 . HeLa cell tubulin and bovine brain tubulin were purified as described 22 . Indirect immunofluorescence. For analysis of acetylation dynamics with in vivo microtubules, HeLa cells plated onto coverslips were incubated in the presence of 10 μ M Nocodazole for 5 hours to induce microtubule depolymerisation. Cells were then washed three times in complete medium to washout Nocodazole and allow microtubules to re-grow. Cells were then fixed at indicated time points after washout in ice-cold methanol and processed for immunofluorescence microscopy by using anti-α -tubulin and anti-acetylated K40 antibodies.
For ex vivo acetylation assay, cells were washed in PEM buffer (80 mM PIPES, 1 mM EGTA, 1 mM MgCl 2 , pH 6, 8) before microtubules extraction in PEM buffer containing 1% Triton X-100 and 10 μ M Taxol. Acetylation reactions were performed in PEM buffer containing 10 μ M Taxol and the indicated concentration of recombinant α TAT1 in presence of 50 μ M acetyl coenzyme A (Sigma) at 37 °C and stopped at indicated time points by fixation with ice cold methanol before being processed for immunofluorescence microscopy by using anti-α -tubulin and anti-acetylated K40 antibodies.
For the in vitro microtubule acetylation assay, 25 μ M HeLa cell-purified tubulin was incubated in PEM buffer supplemented with 15% glycerol and 1 mM GTP, for 40 min at 37 °C to promote polymerization. Microtubules were then diluted in PEM progressively supplemented with 10 μ M taxol (Sigma). Acetylation reactions were performed in PEM-Taxol buffer with 4 μ M recombinant α TAT, in presence of 50 μ M acetyl coenzyme A (Sigma) at 37 °C and stopped at indicated time points by fixation with 2% glutaraldehyde. Microtubules were adsorbed on poly-L-lysine coated coverslips and processed for immunofluorescence microscopy by using anti-α -tubulin and anti-acetylated K40 antibodies.
For experiments with unfixed microtubules, Triton-extracted HeLa cell microtubules or microtubules assembled in vitro from purified bovine brain tubulin were prepared as described above and incubated with the anti-acetylated K40 antibody for 30 min at room temperature. For in vitro microtubules, this incubation step was performed in suspension before microtubules were washed and spotted on a glass coverslip. Microtubules were then processed for immunofluorescence microscopy as described above except that microtubules were not fixed.
Cells or microtubules were imaged with the 100× objective of a wide-field microscope DM6000 the solution 23 is We model the acetylation as a local density of available sites a(x, t) that are acetylated by locally available α TAT1, so that where Γ is our acetylation rate. This is easily solvable in closed form, and defining ≡ a a a / 0 and ≡ z x Dt / 4 we obtain To model the effects of lateral entry of α TAT1 into the MT lumen, through transient or static holes, we assume uniform entry so that where the rate γ captures the size, density, frequency, and lifetime of holes that allow α TAT1 entry. The closed-form solution is We see that as γ → ∞ , we immediately impose the steady-state luminal concentration ρ = ρ 0 . When γ > 0, we solve for the time-dependent acetylation numerically to obtain â x t ( , )-note that the spatial-dependence of â will no longer be simply in terms of ≡ z x Dt / 4 . To evaluate our analytical and numerical results, we need to determine the diffusivity D, the lateral entry rate γ, and the combination ρ Γ 0 . Since acetylation is reported in arbitrary units, we do not need an explicit value of a 0 -i.e. we plot ≡ a a a / 0 . Szyk et al. determined D = 0.27 μm 2 /s from tracking of individual fluorescently-labeled α TAT1 in the MT lumen 7 . D is approximately 100x smaller than the geometrically hindered diffusivity D hindered = 26.3 μm 2 /s expected for a particle the size of α TAT1 in the tube-like MT geometry 7 . This difference is due to the reversible binding of α TAT1 to the MT 24 , with D = α D hindered . The correction factor α is the duty-factor, or the proportion of time that α TAT1 is unbound and diffusing. The 100x reduced D results from α TAT1 being almost always immobilized with α ≈ 0.01. Crucially, a key assumption is that the binding kinetics are "faster than diffusion transport rates" 24 . From the video microscopy of Szyk et al., no immobilized α TAT1 is apparent by eye. This implies very fast binding/unbinding within the lumen, with  Fig. 5c). However, τ cannot be a transition between D and D hindered , since D = 0.27 μm 2 /s already includes the effects of binding/ unbinding and we have already seen that binding and unbinding are relatively fast. Instead, we interpret τ as a lifetime of residence in the MT lumen. We believe τ represents the rate of escape from the MT lumen, and so also determines the lateral entry rate γ = = τ min 40/ 1 . Finally, Shida et al. determined k cat = 615± 34× 10 −6 s −1 for 1 μM bulk α TAT1 4 . We take Γ ρ 0 = 4 k cat = 0.15/min, corresponding to our 4 μM bulk α TAT1.
Atomic Force Microscopy. As described above microtubules polymerization was promoted by incubating tubulin in PEM buffer supplemented with 15% glycerol and 1 mM GTP, for 40 min at 37 °C to. Microtubules were then diluted in PEM supplemented with 10 μ M taxol. Experiments with α TAT were carried out in PEM-Taxol buffer with 400 nM recombinant α TAT1 at 37 °C.
Scientific RepoRts | 6:35624 | DOI: 10.1038/srep35624 5 μ l of microtubules aliquot (with and without α TAT1) were deposited on mica surface pretrated with 100 μ M spermidine 25 and incubated for typically 20 s, which allows the diffusion of microtubules on the surface and their adsorption 26 . The mica surface was then plunged for 30s in uranyl acetate solution (0.02% w/v) for microtubules fixation on mica surface. Finally, the sample was rinsed with pure water (Millipore) and dried with a filter paper. AFM imaging was carried out in Tapping Mode, with a Multimode system (Bruker) operating with a Nanoscope V controller (Bruker). We used silicon AC200TS cantilevers (Olympus) with resonance frequencies around 150 kHz. All images were collected at a scan frequency of 1 Hz and a resolution of 2048 × 2048 pixels. Images were analysed with Nanoscope V software, and a third-order polynomial function was used to remove the background.