Research Article

Detection of GTP-Tubulin Conformation in Vivo Reveals a Role for GTP Remnants in Microtubule Rescues

See allHide authors and affiliations

Science  28 Nov 2008:
Vol. 322, Issue 5906, pp. 1353-1356
DOI: 10.1126/science.1165401


Microtubules display dynamic instability, with alternating phases of growth and shrinkage separated by catastrophe and rescue events. The guanosine triphosphate (GTP) cap at the growing end of microtubules, whose presence is essential to prevent microtubule catastrophes in vitro, has been difficult to observe in vivo. We selected a recombinant antibody that specifically recognizes GTP-bound tubulin in microtubules and found that GTP-tubulin was indeed present at the plus end of growing microtubules. Unexpectedly, GTP-tubulin remnants were also present in older parts of microtubules, which suggests that GTP hydrolysis is sometimes incomplete during polymerization. Observations in living cells suggested that these GTP remnants may be responsible for the rescue events in which microtubules recover from catastrophe.

Microtubules are highly dynamic tubulin polymers that are essential for intracellular organization and cell division. They display a dynamic instability, with alternating phases of growth and shrinkage separated by catastrophe and rescue events (1, 2). Tubulin polymerizes in a guanosine triphosphate (GTP)–bound form and hydrolyzes GTP in the polymer with a slight delay. This creates a GTP cap at the growing end of microtubules (24). Loss of the GTP cap promotes catastrophic events, whereas microtubule rescues result from uncharacterized stochastic events.

Even though the characteristics of the GTP cap have been well studied in vitro, the evidence that such a cap exists in vivo is lacking, essentially because no antibodies specific for the GTP-bound conformation of tubulin are available. The GTP-bound tubulin dimer is in a straighter conformation than the guanosine diphosphate (GDP)–bound dimer (5), and even when constrained in the lattice, GDP-tubulin does not have the same conformation as GTP-tubulin (6, 7). This suggests that it should be possible to make conformational antibodies that specifically recognize GTP-bound tubulin in the polymer. Conformational antibodies specific for GTP-bound Rab6 were selected in vitro by antibody phage display (8). Here, we selected a recombinant antibody specific for the GTP-bound conformation of tubulin in the polymer. We used this antibody to localize GTP-tubulin in cellular microtubules.

Selection of a recombinant antibody specific for the GTP-bound conformation of tubulin. We screened a phage display library of recombinant scFv (single-chain fragment variable) against guanosine 5′-O-(3′-thiotriphosphate) (GTP-γ-S)–loaded tubulin and selected a series of recombinant antibodies to tubulin (9) (fig. S1). One scFv, named hMB11 (scFv MB11 fused to the Fc domain of human immunoglobulin G), was found to be conformation-specific. It did not recognize denatured tubulin by immunoblotting and seemed not to bind to native nonpolymerized tubulin. However, hMB11 cosedimented specifically with microtubules polymerized in the presence of guanylyl 5′-(β,γ-methylenediphosphonate) (GMPCPP), a nonhydrolyzable GTP analog, and not with control microtubules assembled in the presence of GTP (Fig. 1A). In this experiment, low concentrations of taxol (0.1 to 1 μM) were used to prevent depolymerization of control microtubules. When a higher concentration of taxol was used, hMB11 bound to both control and GMPCPP microtubules (Fig. 1, B and C), which suggests that it recognized a conformation and not the nucleotide itself.

Fig. 1.

Conformational detection of microtubules by the antibody hMB11. (A) Microtubules were polymerized in the presence of GTP (control) or GMPCPP. After stabilization with taxol, they were incubated with hMB11, anti-tsg101 (hTSB), or anti-tubulin (hF2C). After centrifugation, antibodies in pellets (lanes 1 to 4) and supernatants (lanes 5 to 8) were analyzed by immunoblot. Only hMB11 cosedimented specifically with GMPCPP microtubules. (B and C) Cosedimentation experiments were carried out as in (A) but in the presence of 10 μM taxol. Binding of hMB11 to control microtubules depended on taxol concentration (C). Data are means ± SEM of four experiments. (D) Fluorescent microtubules assembled separately with GMPCPP (red) or GTP (blue) were stabilized with 1 μM taxol, mixed together, and stained with hMB11 (green). (E) Quantification of hMB11 staining (mean ± SD, N = 585, two experiments).

We then used hMB11 to stain by immunofluorescence a mixture of microtubules polymerized from pure tubulin in the presence of GTP or GMPCPP (Fig. 1D). Under these conditions, hMB11 stained only GMPCPP microtubules [representing 68.6 ± 17.3% (SD) of MB11-positive microtubules] and not control microtubules (1.8 ± 0.9%). The remaining 29.7 ± 16.6% were bundles of both GMPCPP and control microtubules. Despite varying experimental conditions, not all GMPCPP-containing microtubules were stained by MB11, which suggests that some microtubules possessed conformational defects under these conditions.

Detection of tubulin in GTP conformation in cellular microtubules. We next used hMB11 to localize GTP-tubulin in cellular microtubules by immunofluorescence. Because of its conformational binding, hMB11 staining was very sensitive to structural alterations occurring after fixation (10). It was best to use unfixed cells permeabilized in the presence of glycerol and/or low taxol concentration to prevent microtubule depolymerization. In three representative cell lines (HeLa, Ptk2, and MDA-MB231), hMB11 stained the tips of only a fraction of microtubules (Fig. 2, white arrowheads representing 63 ± 4.5% of visible ends), whereas other microtubule ends were unstained (Fig. 2, open arrowheads). This was expected because the GTP-cap model proposes that only microtubules growing at the time of staining should be capped with GTP-tubulin. The observed proportion was very close to the 60% of growing microtubules identified in interphase cells (11).

Fig. 2.

Detection of GTP-tubulin by immunofluorescence in mammalian cells. Cultured cells were processed for hMB11 immunostaining (9) and microtubules were stained with the hF2C antibody (MDA-MB231 cells) or by GFP-tubulin expression (HeLa and Ptk2 cells). Boxed regions are shown enlarged (×5) below. Some microtubule ends were stained by hMB11 (white arrowheads); others were not (open arrowheads). hMB11 also detected GTP-tubulin dots inside the polymers (open arrows). In HeLa or Ptk2, extended stretches corresponding to microtubule bundling were also strongly stained (white arrows). Scale bar, 10 μm.

In addition to the microtubule tip staining predicted by the GTP-cap model, we also observed an unexpected GTP-tubulin staining. First, hMB11 labeled long internal stretches in areas where microtubules formed bundles (Fig. 2, white arrows), although not all bundles were positive. The occurrences of these stretches depended on the cell line used. It is not known whether the GTP domains of microtubules are prone to bundling (as observed upon long incubation with taxol; see fig. S2) or whether microtubules retain a GTP conformation due to bundling and/or to specific binding proteins. Second, hMB11 detected dots along individual microtubules, which we have termed “GTP remnants,” that looked randomly distributed (Fig. 2, open arrows). GTP caps and GTP remnants were also detected in mitotic cells and were more abundant in spindle than in astral microtubules (fig. S3).

Microtubules polymerized in vitro from GTP-tubulin were similarly stained by hMB11 at some of their ends and on discrete internal regions (Fig. 3A). To determine whether labeled ends could correspond to GTP caps, we stained microtubule asters that had polymerized from centrosomes for a short period of time. As predicted by the GTP-cap model, the majority of microtubule plus ends (73% of 226 microtubules in 22 asters) were labeled (Fig. 3B, arrows). Intriguingly, and as shown above, a few discrete internal microtubule regions were also decorated. One possibility is that hMB11 may be directed against a domain in tubulin that would face the lumen of the tube and thus only be accessible at plus ends and on random structural defects along microtubules. This seems unlikely, however, because hMB11 decorated microtubules all along their length when expressed intracellularly while fused to mCherry (fig. S4).

Fig. 3.

Staining of in vitro polymerized microtubules by hMB11. (A) Microtubules were assembled in vitro as in Fig. 1D in the presence of GTP and directly stained with hMB11 before being diluted in taxol and observed by fluorescence microscopy. In these conditions, hMB11 labeled discrete dots along polymerized microtubules. The arrows show microtubule ends stained by hMB11. Scale bar, 10 μm. (B) Microtubules were polymerized for a short period of time (15 min, 37°C) from centrosomes incubated with purified tubulin and labeled with hMB11 (red) and hF2C (green). Note that in addition to internal dots, hMB11 stained the majority of microtubule plus ends (arrows). Scale bar, 10 μm.

We propose that hMB11 stains GTP-bound or GDP–inorganic phosphate (GDP-Pi)–bound tubulin dimers that have been trapped in small regions of the microtubules. A molecular mechanical model indeed predicted that the presence of GTP dimers in the lattice would only locally perturb the microtubule structure (12). Experimentally, GTP or GDP-Pi tubulin have been detected in microtubules (1316). However, more recent studies have failed to detect GTP-bound or GDP-Pi–bound subunits in microtubules, and the presence of very small caps has been proposed (3, 4, 17, 18), although this has recently been challenged (19). In any case, only a small fraction of GTP-tubulin is present inside the polymer.

Coincidence of GTP remnants with microtubule rescue domains. The presence of GTP-tubulin conformation in microtubules suggests a model for dynamic instability (Fig. 4A) that would provide some mechanistic basis to the seemingly stochastic rescue events. In this model, GTP hydrolysis is not always complete and some tubulin dimers persist in a GTP conformation in the polymer. Upon depolymerization, these GTP remnants will become exposed. If GTP hydrolysis does not resume, any remnant as small as a single tubulin layer (4) may behave as a polymerization-prone GTP cap, thereby promoting microtubule rescue. The GTP remnants may explain the frequent rescue events observed when polymerizing microtubules experience shortening (19). Note that growing GTP caps are structurally shaped as open sheets, whereas uncovered internal GTP remnants may exhibit blunt ends.

Fig. 4.

A GTP-remnant model for microtubule dynamic instability. (A) Model for microtubule dynamics showing GTP-tubulin (red) in a GTP cap during polymerization (P) and in inner microtubule regions. Upon cap loss, the probability of catastrophe (C) increases and the microtubule depolymerizes (D) until its end reaches a GTP-tubulin remnant. A GTP end is restored and the probability of microtubule polymerization increases, allowing its rescue (R). (B) Ptk2 cells stably expressing GFP-tubulin were imaged at the indicated times. Rescue events (colored arrows) and the tip of a growing microtubule (arrowhead) are indicated. After cytosol extraction, cells were stained with hMB11 (red) and imaged again, often showing GTP-tubulin remnants at rescue locations. Scale bar, 10 μm. The two kymographs show the dynamics of the microtubules highlighted in red and yellow (top) aligned with hMB11 staining (bottom). Note the good coincidence of rescue position and GTP remnants. (C) Quantification of experiments done as in (A), showing the proportion of polymerizing microtubules stained by hMB11 at their plus ends and the proportions of GTP-tubulin remnants that colocalized with rescue locations in Ptk2 cells (means ± SEM). The proportion that would be expected in stochastic conditions is shown for reference at the right (Monte Carlo simulation, table S1) (9). The table shows that the rescue frequency varies with the distribution of GTP-tubulin remnants (means ± SEM, comparison of Ptk2 and RPE1 cells) (9) (table S1).

To test our model, we analyzed the dynamic behavior of microtubules in Ptk2 cells stably expressing a GFP (green fluorescent protein)–tubulin fusion protein and performed retrospective staining of GTP remnants. Figure 4B and movie S1 show such a sequence in which various events can be identified in particular microtubule rescues (arrows). The polymerizing microtubule exhibited a GTP cap (Fig. 4B, white arrowhead), as did more than 80% of the microtubules that were growing at the time of cell extraction (Fig. 4C). A large fraction of the rescue events recorded [88.8 ± 7.8% (SEM); 38 rescues, 35 microtubules, eight cells] occurred at locations where GTP remnants were retrospectively identified, thus supporting the GTP remnant model (see kymographs, Fig. 4B). A Monte Carlo simulation predicted that only 7.77 ± 1.53% coincidence would be expected to occur by chance (9) (Fig. 4C and table S1). GTP remnant distribution was roughly proportional to rescue frequency (see the comparison between RPE1 and Ptk2 cell lines, Fig. 4C), even though only one-third of GTP remnants seemed to rescue microtubules efficiently. In addition, GTP remnants could be found in newly polymerized portions of microtubules that had never encountered a rescue event (fig. S5), which suggests that the GTP remnants are most probably the cause rather than the consequence of rescue.

On the basis of these findings, we wrote simulation software to visualize the different models of microtubule dynamic instability (9) (MTsimul v1.4; fig. S6 and movie S2). According to the GTP-cap model, rescue depends on the probability of GDP tubulin present at the tip of the depolymerizing microtubule to start polymerizing again. Under the GTP-remnant model, rescues are linked to the probability of GTP hydrolysis. This implies that rescue locations are memorized in the polymer during the seconds or minutes before actual rescues, allowing cells to predetermine their microtubule life span. Factors may exist that would regulate GTP-remnant frequency and thus microtubule stability.

Supporting Online Material

Materials and Methods

Figs. S1 to S6

Table S1

Movies S1 to S3

Simulation software MTsimul 1.4 (for Mac OS X and Windows)

References and Notes

View Abstract

Navigate This Article