|Home | About | Journals | Submit | Contact Us | Français|
Microtubules (MTs) serve important roles in cell trafficking, division and signaling. MTs polymerize via net addition of GTP-tubulin subunits to the MT plus end, which subsequently hydrolyze to GDP-tubulin after incorporation into the MT lattice. The relatively stable GTP-tubulin subunits create a “GTP cap” at the growing MT plus end that protects the MT from catastrophe. Therefore, to understand MT assembly regulation we need to understand GTP hydrolysis reaction kinetics and the size of the GTP cap. In vitro, the GTP cap has been estimated to be as small as one layer [1-3](13 subunits) or as large as 100-200 subunits . GTP cap size estimates in vivo have not yet been reported. Using EB1-EGFP as a marker for GTP-tubulin in LLCPK1 epithelial cells, we find on average: (1) 270 EB1 dimers bound to growing MT plus ends, and (2) a GTP cap size of ~750 tubulin subunits. Thus, in vivo, the GTP cap size is far larger than previous estimates in vitro, and ~60-fold larger than a single layer cap. Consistent with these findings, we also find the tail region of a large GTP cap, 0.5-2.0 μm from the tip, promotes MT rescue and suppresses shortening. We speculate that a large GTP cap provides a locally concentrated (~100 μM) scaffold for tip-tracking proteins, and confers persistence to assembly in the face of physical barriers such as the cell cortex.
In vitro MTs assemble readily from GTP-tubulin, but do not assemble from GDP-tubulin [5-8]. However, when probed, only GDP-tubulin is detectable within the MT lattice [9, 10]. Therefore, GTP hydrolysis must be occurring within the lattice. After addition but before hydrolysis, the GTP-tubulin subunits at the growing tip protect the MT from catastrophe by constituting the so-called GTP cap . The majority of the in vitro estimates of the MT GTP cap size range from about 1-3 layers (13-40 tubulin subunits; Table S1), but to our knowledge in vivo estimates of the GTP cap size have not yet been reported. The questions that we address here are, how large is the MT GTP cap in vivo and what functional role, if any, does it play in suppressing MT disassembly post-catastrophe?
Multiple recent works have demonstrated that EB1, a MT plus-end-tracking protein (+TIP), recognizes microtubule lattices formed from analogs of GTP-tubulin, such as GMPCPP, GTPgS, and GDP/BeF3− in preference to lattices composed of GDP-tubulin [12-14]. These results, combined with the fact that EB1 binding strongly correlates with the growth state of the MT, indicate that EB1 recognizes the tubulin nucleotide state . Therefore, the size of the GTP cap in vivo can potentially be estimated by measuring the number of EB1-EGFP molecules at the growing MT tip and the fractional occupancy of EB1 binding to GTP-tubulin.
To use EB1-EGFP as a quantitative readout of GTP-tubulin at growing MT plus ends, we first determined whether EB1 exists as a monomer or dimer in vivo. In LLCPK1 epithelial cells, as in a wide range of cell types, EB1 binds to and rapidly turns over on the MT lattice and is especially concentrated at growing MT plus ends, forming a comet-like distribution behind the polymerizing MT tip (Figure 1A; [15-18]). EB1 is a ~30 kDa monomer that is thought to exist as a homodimer [19, 20], however its dimerization has not been confirmed in a living cell.
To test whether EB1 diffuses as a monomer or dimer in vivo, we used a combination of FRAP experiments and 3D Brownian dynamics simulations [21, 22] to determine the diffusion coefficients of 2xEGFP, EGFP, EB1-EGFP and EGFP-α-tubulin in the cytoplasm of LLCPK1 cells (Figure 1B). The FRAP experiments were then simulated with varying diffusion coefficients. For each fluorescent species, simulated and experimental halftimes of recovery were quantitatively compared to determine the underlying diffusion coefficient (Supplemental Experimental Procedures; ). Once the diffusion coefficients of all four species were determined, then the diffusion coefficient of EB1-EGFP was compared to that of the other three species based on molecular weight.
From the Stokes-Einstein-Sutherland Relationship, the diffusion coefficient of a spherical particle is predicted to decrease with the inverse cubed root of the molecular weight,
where D is the diffusion coefficient, kB is Boltzmann’s constant, T is temperature, η is the viscosity of the solution, rH is the hydrodynamic radius, NA is Avogadro’s Number, MW is the molecular weight and ρ is the density of the particle. As shown in Figure 1C, a 2-fold increase in molecular weight, from EGFP (~30 kDa) to 2xEGFP (~60 kDa), yields the theoretically predicted decrease in diffusion coefficient from 2.3 ± 0.30 μm2/s to 1.8 ± 0.15 μm2/s (mean ± s.e.m., n=20; Table S2). The diffusion coefficient of EGFP-α-tubulin (140 kDa; assumed to be in a heterodimeric complex with b-tubulin) is also consistent with theoretical predictions based on its molecular weight (1.4 ± 0.18 μm2/s; n=21). When the diffusion coefficient of EB1-EGFP (105 kDa brightness-corrected effective MW for dimer, see Supplemental Experimental Procedures) was measured (1.4 ± 0.25 μm2/s; n=18), it was found to be similar to that of EGFP-α-tubulin and significantly different from that of EGFP and from that of 2xEGFP (Figure 1D). We conclude that the diffusion of EB1-EGFP in the cytoplasm of LLCPK1 cells is consistent with a dimer, but not with a monomer, indicating that EB1-EGFP exists as a dimer in vivo.
To set a lower bound on the GTP cap size, we then determined the number of EB1 dimers in an EB1 comet at the tip of a growing MT. First, the brightness of a single EGFP molecule was calibrated using the known packing density of ab-tubulin in the MT (Figure 2), ~1625 tubulin subunits/μm of MT [23, 24]. In the stable EGFP-α-tubulin LLCPK1 pig epithelial cell line (LLCPK1α) it has been previously determined, via both quantitative Western blotting and fluorescence speckle analysis, that 17% of the α-tubulin molecules are labeled with EGFP [25, 26]. The integrated fluorescence intensity (FI), measured directly as digital camera electron counts, from a background-subtracted image of a MT, was calculated (Figure 2A) and then converted into counts per EGFP per exposure (Figure 2B). For our imaging conditions 1 EGFP molecule yielded 44 ± 0.15 counts·EGFP−1·exposure−1 (μ±sem, n=106 MTs), where one exposure was 300 ms. As long as the percentage of labeled tubulin is known, the packing density of tubulin in the MT lattice provides a convenient standard for in vivo calibration of EGFP fluorophore brightness.
The EGFP brightness was then used to calculate the number of EB1 dimers in an EB1-EGFP comet at the MT tip (Figure 2C). For the EB1 comet analysis the region of interest size was fixed to 24 pixels by 5 pixels (1 μm by 500 nm; Supplemental Experimental Procedures). In an EB1 comet we found there are 270 ± 18 EB1 dimers (mean±s.e.m., n=63 comets). Since comet lengths are ~1 μm, this tip-associated signal rises significantly above the background lattice intensity of 36 ± 5.2 EB1 dimers per μm of MT lattice (mean±s.e.m., n=16 MTs, 11 cells; Figure S1). This value of 270 EB1 dimers represents a lower bound on the GTP cap size, which would require that EB1 completely saturate the GTP-tubulin binding sites at the growing MT tip. This lower bound estimate of 270 GTP-tubulins, or equivalently at least 20 layers, indicates that the GTP cap in vivo is much larger than a single layer of tubulin (13 subunits) and even larger than the most extreme upper limit estimate from the literature of 200 GTP-tubulin subunits for MTs in vitro . Thus, our lower bound estimate of the GTP-cap size in vivo exceeds the previous upper bound estimate for cap size in vitro, and is 20-fold larger than previous estimates of a single-layer cap.
After establishing a lower bound for the GTP cap of 270 GTP-tubulin subunits, we wanted to determine the average cap size in vivo. To estimate the average GTP cap size we used the brightness of the EB1 comet and the known tubulin packing in a MT to establish the percent occupancy of GTP-tubulin sites by EB1 at the tip of the MT (Figure 3A). On average, there were 22 ± 1.6 EB1 dimers (mean±s.e.m., n=36) at the brightest pixel (42 nm/pixel=68.25 subunits/pixel) on the comet, which means the GTP cap is 32% saturated with EB1 dimers (assuming that the tip is composed of nearly 100% GTP-tubulin; see Supplemental Information for justification). With 270 EB1 dimers occupying the GTP cap at 32% of saturation, this corresponds to 270/0.32 ≈ 800 tubulin subunits in the GTP cap (95% confidence interval of 650-1000 GTP-tubulin subunits), which corresponds to a >65 layer distributed GTP cap that decays exponentially in vivo (Figure 3B).
To independently estimate the average GTP cap size we used the EB1 comet decay length (λ) and the known tubulin packing in a MT to calculate the number of tubulin subunits in the GTP cap (Figure 3C). On average, an EB1 comet had a half-length (L1/2) of 310 ± 18 nm (mean±s.e.m., n=36), which is related to the exponential decay length (λ=L1/2/ln2) of 440 nm. The exponential decay was then integrated from the growing tip toward the minus end of the MT, under the assumption that the growing tip is composed exclusively of GTP-tubulin whose concentration drops exponentially at a rate given by the EB1 comet decay length. Using this method, we estimate that there are ~700 GTP-tubulin subunits in a GTP cap (95% confidence interval of 620-780 GTP-tubulin subunits), which corresponds to a >50 layer distributed GTP cap that decays exponentially in vivo. This independent estimate of the in vivo MT GTP cap size corroborates our large GTP cap estimate, which utilized the EGFP fluorophore brightness calibration.
As a check on our GTP cap size estimate, we then compared the KD of EB1 binding to previous estimates of MAP binding to MTs in vivo . We estimated the KD of EB1 binding to GTP-tubulin in the MT lattice, assuming equilibrium, via
where [EB1free]eq is the free EB1 concentration at equilibrium, [GTP-Tubulin]eq is total concentration of available MT GTP-tubulin binding sites at equilibrium, and [EB1bound]eq is the concentration of MT-bound EB1 at equilibrium. We also measured the free EB1 dimer concentration to be 1.2 μM using the EGFP brightness calibration (Figure 2B). From the EB1 dimer free concentration and the fractional occupancy the KD was calculated to be 3.8 μM. This corresponds to a moderate binding affinity in vivo, which has been previously reported for in vitro EB1-tubulin binding to taxol-stabilized MTs . From this we can also estimate koff, assuming a typical kon values of 1-10 μM−1s−1. We estimate koff to be 3.8-0.38 s−1, which is consistent with previous fluorescence recovery after photobleaching (FRAP) experiments . The binding affinity of EB1-EGFP to the MT tip is comparable to the apparent affinity of ensconsin for MTs in vivo, estimated to be KD=11 μM . Therefore, our KD for EB1 in vivo is comparable to another MAP in vivo and consistent with previous FRAP experiments for EB3 and CLIP-170 .
Comparing our in vivo estimate of the MT GTP cap size to previous in vitro estimates (Table S1), we find that our in vivo estimates of both the lower bound and average MT GTP cap size are much larger than any in vitro estimates. However, in EB1-EGFP expressing LLCPK1 epithelial cells we found that MTs grow at 156 ± 13 nm/s (μ±sem, n=29) with 6.9 μM tubulin (measured using the EGFP brightness calibration from Figure 2B) whereas MTs grow at ~20 nm/s in vitro at nearly the same tubulin concentration (~5-10 μM GTP-tubulin; [29-31]). This ~7-fold disparity in MT growth rate from in vitro to in vivo conditions predicts a ~7-fold larger GTP cap size in vivo. Upon further examination of the previous in vitro estimates of the GTP cap size (Table S1), we find that while the majority of the estimates were too small, since they estimated a 1-3 layer cap, the largest in vitro estimate from Walker et al. (1991) scales properly to predict a consistent in vivo GTP cap size estimate. A 7-fold increase in the upper-bound estimate from Walker et al. (1991) would predict a GTP cap size range of 700-1400 GTP subunits, which is consistent with our average in vivo estimate of ~750 GTP-subunits (95% confidence interval of 670-830 GTP-tubulin subunits; further discussion in Supplemental Information). Therefore, while our estimate is 65-fold larger than the single layer cap theory, it is still consistent with the largest in vitro prediction.
There have been mechanisms proposed for EB1 tip tracking alternative to GTP-tubulin recognition, including MT seam binding and recognition of an unknown tubulin conformation such as a sheet [32, 33]. The EB1 homolog, Mal3, was reported to preferentially bind along the seam of the MT . However, if EB1 were to only bind along the MT seam, this would result in a uniformly dim signal along the length of the MT rather than a comet-like signal, as the seam is not confined to the MT plus end. Also, a seam-binding model does not explain the difference in binding affinities for the growing MT tip vs. the MT lattice. If EB1 binds to an unknown conformation confined to the tip of the MT, we would predict the EB1-EGFP comet decay length to be small, on the order of our recent estimates of MT tip structures in LLCPK1 cells , which have an average length of ~180 nm, much less than the 1/e decay length of EB1-comets we measured here to be ~440 nm. Given these considerations, and the recent demonstrations of EB1 preference for GTP-tubulin analogs over GDP-tubulin in the MT lattice [12-14], the overall conclusion from previous work is simply that EB1 recognizes the GTP-tubulin cap in growing MTs. Based on this assumption, we conclude that the GTP cap in LLCPK1 cells is comprised of ~750 subunits on average.
One prediction of a large GTP cap is that GTP-tubulin will exist far from the tip (i.e. more than one decay length, >440 nm). In this case, a post-catastrophic MT tip must first shorten through the tail of the GTP cap. A key prediction is that the shortening rate in the GTP cap tail region will be slower than in more proximal GDP-rich regions. To test this prediction, we divided MTs into two sections, the first was 500-2000 nm away from the MT tip and the second was >2000 nm from the MT tip, and compared the MT shortening rates within these two sections. We did not use the first 500 nm from the MT tip because we wanted to be very conservative in our estimate of the limit of detectability of a catastrophe event (see Supplemental Experimental Procedures). As predicted, we found that MTs shorten significantly slower near the MT tip as compared to regions >2000 nm away from the MT tip (p=0.003). Near the MT tip MTs shorten on average 610 ± 34 nm/s as compared to 790 ± 54 nm/s (μ ± s.e.m., n=31 MTs) >2000 nm away from the MT tip (Figure 4A).
Another prediction of a large GTP cap is that rescues should be common in the tail of the cap, and less frequent in more proximal regions. As shown in Figure 4B, we found that rescues were largely limited to the first 2000 nm of shortening, and that more proximal rescues were not detected. For 34 observed rescues, we found rescues occurred on average 1200 ± 95 nm (μ±sem, n=34) or ~150 tubulin layers from the MT catastrophe site, based on the mean tip position (Figure 4B). The minimum distance from the MT tip that a rescue event could be confidently resolved was 500 nm. However, this 500 nm threshold is larger than two point spread functions and is ~12 pixels in length, which is greater than a single layer and about double our lower bound estimate for the GTP cap size. Together, these data demonstrate that MT rescues occurred in the tail regions of the cap, located on average 150 tubulin layers away from the MT plus end, and very rarely in more proximal regions of the MT.
Furthermore, the large GTP cap predicts that the rescues should be due to residual low levels of local GTP-tubulin, and those rescue sites should therefore have slightly higher EB1-EGFP signal than more proximal regions. To test this prediction, we compared the EB1-EGFP fluorescence intensity (FI) at rescue sites to regions that were proximal and distal to the rescue site (Figure 4C). Important for determining the rescue site, we established EB1-EGFP as an accurate basis for tracking a growing MT tip in vivo. We found that our single time point accuracy of MT tip tracking via digital image analysis of EB1-EGFP is 13 nm, or about 1.6 dimer layers (1 layer=8 nm). This now establishes a method for near-molecular resolution tracking of MT growth in vivo, and eclipses our previous method for high-accuracy tracking (36 nm via EGFP-tubulin in LLCPK1 cells using the same microscope ) by three-fold. Thus, we can track microtubule tip dynamics at 5 Hz and l/40 nm, demonstrating powerful super-resolution analysis in vivo using conventional microscopy and conventional digital image analysis. As shown in Figure 4D, rescue sites within the MT lattice have a statistically 2.0-fold brighter EB1-EGFP FI than the EB1-EGFP signal on more proximal MT regions. However, rescue sites were not statistically brighter than more distal regions, which is consistent with a large distributed MT GTP cap but not consistent with an island or remnant of GTP-tubulin . These results indicate that a large GTP cap, decaying over 1 μm, influences post-catastrophe shortening dynamics.
If we assume the MT tip is completely saturated with GTP-tubulin subunits (see Supplemental Information for further discussion) and use the measured EB1-EGFP comet decay length, we can calculate the percentage of tubulin subunits at the mean rescue site that are GTP. We measured the EB1-EGFP comet decay length (1/e) in EB1-EGFP expressing LLCPK1 epithelial cells to be 440 ± 83 nm (μ±sem, n=26 MTs; Figure 3C), with an equivalent EB1 comet halflength (L1/2) of 310 nm (=440 nm * ln(2)). Based on the measured EB1-EGFP comet decay length, we predict the average rescue site location to be 6.5% saturated with GTP-tubulin subunits, or about one GTP-tubulin per layer. It is remarkable that such a small fraction of GTP subunits is sufficient to influence MT dynamics. Again, this data is consistent with a large GTP cap in vivo where multiple opportunities for rescue are available in the tail of the GTP cap but is not consistent with a small GTP cap that is only a few layers deep.
Collectively, these in vivo estimates of the MT GTP cap size indicate the presence of a very large GTP cap that is on the order of 750 tubulin subunits, with a lower bound of 270 subunits, and decays with increasing distance from the tip. Furthermore, stabilizing features of the cap can be detected at distances greater than 1 μm (150 layers from tip; Figure 4), which agrees with observed EB1 comets lengths. These results indicate the GTP cap size in vivo is 4- to 65-fold larger than published in vitro estimates and functions to promote rescue and suppress shortening.
A large GTP-tubulin cap on growing MTs has important potential implications for the cell. First, the large GTP cap provides multiple opportunities for depolymerizing MTs to rescue in the tail of the GTP cap, >100 layers away from the MT plus end. As previously reported , and as we have observed in LLCPK1 cells (Movie S1), MTs that reach the cell periphery undergo multiple cycles of growth and shortening before finally depolymerizing back towards the cell center. Thus, a MT is more persistent in the face of barriers that tend to suppress net assembly , because the long tail of the large GTP cap serves as a back-up mechanism to rescue a failing MT. As a result, a MT has multiple opportunities to deliver cargo or find a binding partner, rather than a single attempt as would be expected from a single-layer (or small) GTP cap. If the MT were to participate in signaling pathways, by either delivering cargo necessary for signaling , or binding to the actin cortex , then the MT will benefit from an extended GTP cap, with GTP-tubulin subunits >2 μm away from the MT plus end, to provide additional opportunities for rescue in the GTP cap tail. This “tail rescue” phenomenon will prolong the duration of interaction at the cortex, which can then amplify a MT-mediated signal.
A large GTP cap, extending over several hundred GTP-tubulin subunits in vivo, can potentially serve as a highly concentrated point source for biochemical activity. We estimate that the large GTP cap in vivo allows EB1 dimers in a comet to be highly concentrated (>100 μM) locally, far above the cytoplasmic free concentration of 1.2 μM. In this way, a growing MT plus end with a high concentration of proteins bound to the large GTP cap through EB1 can act as a potent localized signal and more effectively participate in signaling cascades than would be possible with a small (e.g. single-layer) GTP cap. An example of this type of signaling occurs during Drosophila embryogenesis where DRhoGEF2 is delivered to a localized region of the cell membrane to affect myosin-mediated cell contractility . In general, the large GTP cap makes MTs more robust to catastrophe in face of barriers, and potentially allows for more potent +TIP-mediated signaling.
Thanks to M. K. Gardner for technical assistance with the MT growth simulation, P. Wadsworth for the LLCPK1α cell line, L. Cassimeris for the EB1-EGFP LLCPK1 cell line, J. Mueller for the p2xEGFP construct and R. Y. Tsien for the pmCherry-α-tubulin construct. Funding support provided by NIH Grants GM-071522, GM-076177, and NSF Grant MCB-0615568.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.