Life cycle of MTs: persistent growth in the cell interior, asymmetric transition frequencies and effects of the cell boundary

Rapid remodeling of the microtubule (MT) array is important in large motile cells because such cells are constantly in the process of changing their shape. In fibroblast-like cells, MTs form a radial array emerging from the centrosome with minus ends tethered at the centrosome and plus ends extending towards the cell periphery (reviewed byDesai and Mitchison, 1997;Keating and Borisy, 1999). Although many studies have dealt with aspects of MT dynamics, a full understanding of the MT life cycle has yet to be achieved. In particular, the first phases, namely formation at the centrosome and growth in the cell interior have not been analyzed directly because of technical limitations imposed by the high density of MTs in the centrosomal region.

In previous studies, the dynamics and turnover of MTs in cultured cell lines have been analyzed primarily using fluorescently labeled tubulin. The picture that has emerged from these studies is that MTs in vivo undergo frequent transitions between phases of growth and shortening(Sammak et al., 1987;Cassimeris et al., 1988;Shelden and Wadsworth, 1993;Waterman-Storer and Salmon,1997; Yvon and Wadsworth,1997).

However, several considerations suggest that this picture is incomplete. One basic point is that almost all observations of individual MT dynamics in vivo have been made near the cell margin. This is because the cell is thinner and the MT density is lower near the edge of the cell, thus permitting better visualization of individual MTs. Consequently, MT dynamics in the cell interior have remained essentially unexplored and the formal possibility exists that they are different from that at the cell margin. A second point is that MT dynamics in vivo show complexity not seen in vitro. In vitro, MT dynamics are characterized by transitions between well defined phases of growth and shortening (Mitchison and Kirschner, 1984a; Mitchison and Kirschner, 1984b; Horio and Hotani, 1986; Walker et al., 1988). In contrast, MTs in many cell types do not show well defined phases. In addition, MTs in vivo frequently are quiescent, neither growing nor shortening, a behavior termed pause. Growth and shortening are highly variable in terms of velocity, duration and extent but, in general,tend to be brief. For example, the mean growth length has been reported as 1.3μm in PtK₁ cells and 3.2 μm in CHO cells(Shelden and Wadsworth,1993).

The difficulty in clearly defining phases of growth and shortening in vivo led us (Vorobjev et al., 1999;Vorobjev et al., 1997) to introduce an alternative description of MT dynamics. MT dynamics were considered as a 1-dimensional random walk of their plus ends along the cell radius and their overall properties were characterized by two parameters— a diffusion coefficient and a drift coefficient. The diffusion coefficient is a measure of the amplitude (squared) of growth and shortening excursions per unit time, while the drift parameter represents the imbalance of growth and shortening excursions over time. This framework provided an analysis of dynamics not dependent upon detailed assumptions of growth or shortening behavior. The diffusion and drift parameters permitted prediction of the steady state length distribution of MTs and the time for turnover of the MT population. However, turnover times predicted from reported dynamic instability parameters were substantially greater than experimental determinations (Vorobjev et al.,1997). A similar conclusion was drawn from Monte Carlo analysis using the dynamic instability model(Gliksman et al., 1993). Proceeding from the assumption that MT plus ends undergo stochastic excursions such as the random walk observed in the lamellar region of PtK₁cells (Vorobjev et al., 1997)or in fish melanophores (Vorobjev et al.,1999), the time for a nascent MT plus end starting from the centrosome to grow to the cell margin (or to shorten from the cell margin back to the centrosome) will be a few hours for a typical cultured cell of radius 25 μm. Such a long time seems incompatible with the requirements for rapid cytoskeletal remodeling in cell motility behavior. This disparity between theoretical and experimental analyses of MT turnover is a third point suggesting the incompleteness of our understanding of MT dynamics in vivo. Thus, we considered that the dynamics of MTs in the cell interior may somehow be different from that near the cell margin. Supporting this view, a few studies have reported the capacity of MTs in vivo to show behavior other than rapid fluctuations between shortening and growth. MTs have been observed to persistently grow into the lamellipodia of newt lung cells(Waterman-Storer and Salmon,1997) and to continuously elongate into newly formed protrusions of HGF-stimulated PtK₁ cells(Wadsworth, 1999).

These considerations taken together motivated us to reinvestigate the life cycle of MTs by procedures designed to evaluate their behavior in the cell interior and, in particular, in the vicinity of the centrosome. By employing a combination of novel approaches, we found that MT behavior in the cell interior indeed differed from that near the cell margin. Remarkably, nascent MTs freshly nucleated at the centrosome grew persistently until they approached the cell margin. Only then did they display the frequent fluctuations between growth and shortening that are considered to be the hallmark of dynamic instability. We suggest a revised view of MT dynamics in vivo in which the `default' condition of a nascent MT is persistent growth. In this view, `dynamic instability' at the cell margin results from the behavior of the MT system operating under the constraint of a `boundary condition'.

CHO-K1 and NRK cells were grown in F-10 medium supplemented with 10% fetal bovine serum and antibiotics on coverslips with photoetched locator grids(Bellco Glass, Vineland, NJ). For observation of MT dynamics in vivo, cells were cultured for 2 days after plating and were microinjected with Cy3-tagged tubulin at a needle concentration of 10 mg/ml. Cells were kept on the microscope stage at 36-37°C during observation and temperature was measured before and after each experiment. Cells injected with Cy3-tubulin were treated with the oxygen-depleting preparation, Oxyrase (Oxyrase, Ashland,OH), to reduce photodamage and photobleaching(Mikhailov and Gundersen,1995). Injected cells were observed on a Nikon Diaphot 300 inverted microscope equipped with a Plan 100×, 1.25 NA objective using a Cy3 filter set for observations of Cy3-labeled MTs and a GFP filter set for observation of GFP-CLIP-170 in transiently transfected cells. Images of 16-bit depth were collected with a CH350 slow scan, cooled CCD camera (Photometrics,Tucson, AZ) driven by Metamorph imaging software (Universal Imaging,Westchester, PA). The image was projected onto the CCD chip at a magnification of 250×, which corresponded to a resolution of 1 pixel=0.09 μm (11.1 pixels per μm). Time-lapse series of 50-200 images were collected at 3-5 second intervals. 16-bit images were processed and rescaled with Metamorph software and 8-bit images were prepared for presentation with Adobe PhotoShop(Adobe Systems, Mountain View, CA). To highlight MTs of interest in some figures, color overlays were painted with opacity of 15% within the color mode layer in Adobe PhotoShop.

Photobleaching was performed on a Zeiss IM-35 inverted microscope using a 3 W argon ion laser as described elsewhere(Keating et al., 1997), except that the cells were incubated at 37°C. The laser beam was shaped into an approximately 20×3 μm bar using a cylindrical lens and a Neofluar 100×, 1.3 NA objective. The zone was placed to photobleach a path across a cell with the centrosome at its center.

Cytoplasts were prepared by a modification of a described method(Karsenti et al., 1984). Briefly, 2 days after plating onto coverslips cells were treated with nocodazole (1 μg/ml) and cytochalasin D (1.5 μg/ml) for 90 minutes. Coverslips were then placed `cells-down' into centrifuge tubes containing culture medium with drugs and were centrifuged at 10,000 g for 25 minutes to enucleate the cells. Enucleation resulted in about equal numbers of cytoplasts containing or lacking the centrosome. Coverslips were washed with fresh medium to remove drugs and incubated for 2 hours for complete recovery of MTs in the cytoplasts. For observation of MTs, cells were microinjected with Cy3-tagged tubulin before enucleation. For simultaneous observation of MTs and CLIP-170 (see next section), cells were first transfected with and allowed to express GFP-CLIP-170, GFP-CLIP-170-positive cells were microinjected with Cy3-tubulin and then cytoplasts were prepared.

Cells were transfected by glass capillary microinjection of DNA into the nuclei following a previously described procedure(Perez et al., 1999) with slight modification and were observed after 10-20 hours. Briefly,pCB6-myc-GFP-CLIP-170 was used at a needle concentration of 20 μg/ml, which gave sufficiently low expression levels of recombinant protein without addition of cycloheximide. For visualization of the centrosome in some experiments, HsCen-2-GFP DNA (human centrin) was mixed with pCB6-myc-GFP-CLIP-170 DNA that had the same promoter. The final needle concentration of each vector was the same, -20 μg/ml.

The position of active ends of MTs and the length of growth or shortening episodes were analyzed by subtraction of sequential images(I_n—I_n+1) in time-lapse series as described previously (Vorobjev et al.,1999). The resultant difference images identified the extent of growth or shortening as black or white domains, respectively, at the ends of individual MTs. The threshold (minimal length) for determination of shortening or growth in subtraction analysis was set to 0.45 μm — that is, 5 pixels in the digital image. Lateral displacements of MTs were distinguished from growth and shortening episodes because they gave a parallel arrangement of long black and white segments, whereas growth or shortening gave single short segments that were either black or white but not both.

Mean velocity of MT growth was determined using direct observation of Cy3-labeled MTs and the CLIP-170 approach. For the direct observation approach, MTs were monitored until they reached 90% of cell radius. The velocity of apparent CLIP-170 movement was determined by tracking the head of the individual CLIP-170-positive structure from the centrosome until it disappeared within the cytoplasm or near the plasma membrane.

The extent of persistent growth was expressed both as an absolute distance and as a percentage of the local cell radius. For the latter calculation, the local cell radius was determined as the straight line distance from the centrosome to the cell margin in the direction of MT plus end or CLIP displacement.

Two problems have previously hindered the direct observation of MT nucleation at the centrosome. First, MT density at the centrosome is high,precluding visualization of individual MTs. Second, the cell is generally thick in the region of the centrosome, which results in substantial out-of-focus fluorescence that degrades imaging of individual MTs. To overcome these obstacles, we have employed two approaches. In the first approach, we photobleached a path across the centrosomal region to diminish the fluorescence density of pre-existing MTs. This permitted observation of nascent MTs growing from the centrosome in the direction of the path. In the second approach, we generated cytoplasts that, because they lack the nucleus,are much thinner than whole cells, thus facilitating direct observation of individual MTs while still close to the centrosome.

Cy3-tubulin was microinjected into cells and allowed to incorporate into MTs. The MTs in CHO cells are arranged in a predominantly radial pattern with the centrosome at its focus. An argon ion laser and an optical track with a cylindrical lens were used to photobleach a path across the cell with the centrosome in its center. MTs that were nucleated at the centrosome after the photobleaching and which grew in the direction of the path could then be visualized continuously beginning within seconds of their birth(Fig. 1A).

The nascent MTs displayed remarkable behavior not previously reported in animal cells. They showed negligible, if any, dynamic instability. Typically,their leading (presumably plus) ends grew persistently until they were at the cell periphery (Fig. 1B). As an operational definition, we consider the cell periphery to be a zone of∼15% of the cell radius (3 μm). On average, 68% of the MTs reached the cell margin without transition to a shortening phase and continuous MT growth phases sometimes exceeded 20 μm. The growth rate derived from the frequency histogram of instantaneous rates (Fig. 1C) was 17.8±13.8 μm/minute (n=33 MTs; 10 cells; total time of observation=1228 seconds). The histogram confirmed that episodes of shortening (8.4%) or pause (0.6%) comprised a minor fraction of nascent MT behavior. At the same time that nascent MTs were growing persistently in the interior, life history plots of MTs at the periphery of the cell showed that they displayed characteristic dynamic instability, both before and after photobleaching (data not shown). Therefore, the experimental protocol did not detectably affect MT behavior. Thus, observations after path photobleaching suggest that nascent MTs in the cell interior have different behavior than those at the periphery. However, a limitation of this approach was that only those MTs that were oriented along the path (usually 2-4 per cell) could be followed, limiting the size of the data set. To obtain a more comprehensive picture we examined cytoplasts where multiple MTs could be tracked starting close to the centrosome.

Enucleation of cells results in cytoplasts that are smaller and flatter than whole cells. The cytoplasts generally contained a reduced number of MTs(∼50%) compared with those in intact cells, suggesting that some nucleating material was lost or damaged during enucleation. Nevertheless,centrosome-containing cytoplasts retained a strong radial array of MTs with the centrosome at its center. The combination of thinness and reduced number of MTs allowed the centrosome to be clearly visible and MTs growing from it to be readily traceable (Fig. 2A). Thus, it was possible to identify an individual MT shortly after it was nucleated and to monitor its elongation in cytoplasts without the need for path photobleaching. MTs grew with rare pauses or shortening episodes(Fig. 2B) at a rate of 15.8±5.9 μm/minute (50 MTs; 8 cytoplasts), which was similar to that in whole cells after path photobleaching. Thus, the property of persistent MT growth was retained after enucleation.

As a measure of MT persistence, we estimated the frequency of transition from growth to shortening. Times were recorded for the growth of nascent MTs from the moment they could be detected near the centrosome until they had a`catastrophe' (shortening episode >0.5 μm) or arrived at the cell periphery, whichever came first. For MTs that arrived at the cell periphery without a detectable shortening episode, we logged their times but did not log a catastrophe. Since MTs in whole cells and centrosome-containing cytoplasts behaved similarly, we combined data from both protocols. For 52 MTs analyzed,only 12 transitions to shortening were logged in 2337 seconds of observation time for a catastrophe frequency of 0.005 second^-1.

As MTs approached the cell margin, their behavior changed. Continued observation of MTs for at least 3 minutes, starting from the centrosome(Fig. 2B) or selecting them near the plasma membrane (Fig. 2C) showed that the majority of plus ends displayed typical dynamic instability near the cell margin. MT behavior within the 3 μm zone from the cell margin was characterized by an apparent catastrophe frequency of 0.08 second^-1 (207 events for 2710 seconds, 61 MTs analyzed in 17 cells and cytoplasts), that is, 16-fold higher than that in the cell interior. Thus, as in the path photobleaching experiments, dynamic instability behavior was observed only at the cell periphery.

The apparent difference in MT behavior in the cell interior as opposed to the cell periphery represents a significant departure from our understanding of MT dynamics. Therefore, it seemed advisable to seek confirmation of these conclusions by an independent approach. Ideally, one would want to visualize MT growing ends without the background contributed by the rest of the MTs. Such selective visualization seemed possible through the use of CLIP-170 as a marker for MT elongation (Perez et al.,1999). Although Perez et al. showed that CLIP-170 co-localized with the ends of some growing MTs and was not targeted to stationary or shortening MTs, they were not able to determine whether CLIP-170 was targeted to all growing MTs or only to a certain subset. Thus, we first needed to evaluate this point. To do so, we compared MT growth with GFP-CLIP-170 behavior using a double labeling protocol for cytoplasts. Intact cells were transfected by microinjection of GFP-tagged CLIP-170 DNA into the nuclei. To avoid potential problems with overexpression, we selected cells at 10 hours after transfection expressing low levels of fluorescent GFP-CLIP-170. These GFP-positive cells were then injected with Cy3-labeled tubulin and incubated for 60-120 minutes prior to cytoplast preparation. Colocalization of GFP-CLIP-170 (red) and ends of MTs (green) in a centrosome-containing cytoplast is shown in Fig. 3A. Displacement of the CLIP-170 label in successive frames was identical to the growth of MT plus ends (Fig. 3B). CLIP-170 labeling disappeared within 5 seconds (time resolution of our double channel time-lapse series, 2.5 seconds per channel)when a MT paused or underwent a transition from growth to shortening(Fig. 3C). Several typical CLIP-170 and MT histories are plotted inFig. 3D. Invariably(n>100), CLIP-170 tracks were coincident with MT tip displacement,thus indicating that CLIP-170 was targeted to all growing ends and validating their use as a tool for selectively visualizing MT growth.

A prediction of persistent growth of MTs starting from the centrosome is that CLIP-170 will be associated with a MT tip almost continuously until the MT plus end approaches the cell margin. Therefore, we used the movement of CLIP-170 labeled zones as a tool to evaluate MT growth in the cell interior. For visualization of the centrosome in some experiments, GFP-Hscen2 DNA (human centrin expressing construct) was mixed with GFP-CLIP-170 DNA and they were used together for transfection.

We evaluated MT growth both in intact cells and in cytoplasts containing a centrosome. As predicted, GFP-CLIP-170-labeled zones displayed long excursions over time (continuous duration frequently >1 minute) through the cell interior (Fig. 4A). Track diagrams were built from successive positions of CLIP-170-labeled zones on individual MTs. The tracks showed that CLIP-170 zones moved radially outward from the centrosome. The mean unbroken length of GFP-CLIP-170 tracks was 17.3±4.8 μm (n=78) in whole cells(Fig. 4B) and 15.8±5.8μm/minute (n=40) in cytoplasts containing a centrosome. The mean velocity of CLIP-170 zones was 16.5±6.0 μm/minute (n=110)for whole cells and 18.3±4.8 μm/minute (n=39) for the cytoplasts. The velocity of CLIP-170 movement was similar to that determined for MT growth by direct observation, both in whole cells and cytoplasts. We may conclude from these measurements both that CLIP-170 movement is a good proxy for MT growth and that cytoplasts are a good proxy for intact cells. Normalizing the length distribution of CLIP tracks against the distance between the centrosome and the cell margin(Fig. 4C) shows that the majority of MTs nucleated at the centrosome reached ∼85% (84±16%)of the cell radius without catastrophes or long pauses. This result signifies that most MTs persistently grow from their time of birth at the centrosome until their plus end nears the cell margin.

Because new CLIP-170 labeled zones continuously appeared near the centrosome, it was possible to use their appearance to estimate the frequency of MT formation at the centrosome, defined operationally as a circle of 2μm radius drawn about the focus of the MT array. With this criterion, we estimated the rate of nucleation of MTs by the centrosome in CHO cells to be 5.6±2.3 per minute (178 CLIP-170 zones; 7 cells; 1997 seconds).

The only previously reported instance of persistent plus end growth in mammalian cells is the MT treadmilling observed in vivo in the absence of a centrosome (Rodionov and Borisy,1997; Rodionov et al.,1999). In our previous study(Rodionov et al., 1999) we suggested that treadmilling in cytoplasts lacking a centrosome resulted from exposed MT minus ends whose depolymerization caused an elevated tubulin pool that suppressed transitions from growth to shortening phase at the MT plus end. Contrary to that suggestion, this study has uncovered persistent growth of MT plus ends in the absence of substantial numbers of free minus ends. Therefore, we reevaluated the question of whether the presence of the centrosome had any significant effect on the rate of plus end growth. With our improved procedures, we compared cytoplasts containing and lacking the centrosome.

In centrosome-free cytoplasts, MTs spontaneously appeared in the cytoplasm,then translocated by means of treadmilling and eventually arrived at a cell margin (Fig. 5A). After reaching the cell margin, plus ends often became paused and MTs disassembled because of minus end depolymerization. The rate of growth of the plus end during treadmilling was measured in the same way as for MTs growing from the centrosome in intact cells or centrosome-containing cytoplasts. From time-lapse observations of Cy3-labeled MTs, the rate of plus end growth during treadmilling was 20.3±8.3 μm/minute (n=66; 11 cytoplasts). In steady-state treadmilling, growth of the plus ends must be balanced by minus end shortening. Consistent with this expectation, the rate of minus end shortening in the cytoplasts lacking a centrosome was 18.2±5.8μm/minute (n=66; 11 cytoplasts).

Our measurement by direct observation of mean plus end growth rate includes pauses and shortening episodes. To compare the real growth potential of the plus ends these episodes have to be excluded. The CLIP-170 approach provides a natural realization of this algorithm since CLIP disappears from the plus ends when they stop. The tracks of CLIP-170 in cytoplasts lacking a centrosome(Fig. 5B) gave the MT plus end growth rate as 22.0±10.1 μm/minute (n=75; 7 cytoplasts),while in centrosome-containing cytoplasts, the measured rate was 18.3±4.8 μm/minute (n=39; 5 cytoplasts), a difference of 20%. Thus, by both measures, MT growth was slightly faster in cytoplasts lacking a centrosome, consistent with an elevated tubulin pool. However, since persistent growth is a regular feature of MTs in intact cells and centrosome-containing cytoplasts, the creation of free minus ends is clearly not a prerequisite for this phenomenon.

Treadmilling MTs permitted an additional test of the localization of CLIP-170 exclusively at growing ends. Since one end of a treadmilling MT is shortening while the other is growing, CLIP-170 is expected to be present at the leading (plus) end but absent from the trailing (minus) end. This prediction was confirmed by direct dual label observation(Fig. 5C). Thus, CLIP-170 can be used as a polarity marker for the growing MT plus end.

Polymer balance of treadmilling MTs comes at the level of individual MTs with shortening from the minus end being equal, on average, to growth at the plus end. However, in intact cells or centrosome-containing cytoplasts, minus ends are tethered at the centrosome. Since release from the centrosome and minus end shortening is infrequent(Keating and Borisy, 1999;Waterman-Storer and Salmon,1997; Vorobjev et al.,1999), most of the balance must come from the dynamics of the plus end. Therefore, the persistent growth of MT plus ends in the cell interior must be balanced primarily by shortening of other MT plus ends.

The steady state requirement of balance of growth and shortening holds true for every position in the cell, including the centrosome. Since nascent MTs are born at the centrosome at the rate of 5.6±2.3 per minute, we predicted that an equivalent rate of MT shortening back to the centrosome should be observed. To evaluate this issue, we randomly selected MTs having their plus ends close to the margin and monitored their fate(Fig. 6A). In contrast to growth, shortening was not persistent. Shortening occurred at a velocity of 28.8±14.1 μm/minute (n=474; 10 cells)(Fig. 6B,C). Frequency histograms of the distances shortened showed an approximately exponential decay (Fig. 6D) with the mean distance being 2.9±3.4 μm. These properties are consistent with a first-order process of transition back to the growing state (rescue). The rescue frequency was determined to be 0.12 second^-1 (91 transitions, 100 MTs, 735 seconds of observation) by tracking MTs shortening back from the margin until they began to grow. Thus, rescue as opposed to catastrophe was frequent leading to small shortening excursions being common and long ones rare. Nevertheless, shortening back to the centrosome could be directly observed, albeit at low frequency. Since MTs that shortened long distances were generally lost as their plus ends neared the congested area around the centrosome, we scored, as a more reliable estimate, the number of MTs that shortened by two-thirds of the cell radius. Out of all episodes of shortening, we obtained 6.4±2.7 per minute that shortened more than two-thirds of the distance to the centrosome.

Besides shortening from the plus end, infrequent releases of MTs from the centrosome also occurred. Released MTs had a short lifespan and rapidly depolymerized from the minus end (Fig. 6E). In cytoplasts, the frequency of releases was 1.0±0.5 MT per centrosome per minute (58 releases, 11 cytoplasts). Thus, plus end death (6.4/minute) + minus end release (1.0/minute) was approximately equal to plus end birth (5.6 MTs/minute). This rough equality confirms that the MT system in CHO cells is indeed in steady state and it indicates that balance at the centrosome in CHO cells comes primarily (∼85%) from the plus end pathway with the minus end pathway making a minor (∼15%) but significant contribution.

The persistent growth of nascent MTs is an apparently paradoxical result because it is not clear how this behavior can be reconciled with the dynamic instability behavior observed near the cell margin. To address this question,we employed the conceptual framework of diffusion plus drift and the procedure of sequential subtraction analysis to obtain the necessary data.

Sequential subtraction images (Fig. 7A) permit identification of growing and shortening MT ends as black or white line segments, respectively, even in regions of high MT density, which otherwise preclude direct observation of individual MTs. MTs that do not undergo a displacement (i.e. are stable or paused) between successive frames become canceled out and do not appear in the subtraction image.

The velocities of growing and shortening were determined from the length of the line segments in the differential images divided by the time between acquisitions of the images. Frequency histograms of the velocities were bimodal with growth events slower than shortening but more frequent. An example histogram is shown in Fig. 7B for the region between 0.6-0.9 of the cell radius. For the same region, the fraction of growing MTs, f_g, was 0.70 and their velocity, v_g, was 17.0±5.9 μm/minute, n=372,while the fraction of shortening MTs, f_s, was 0.30 and their velocity, v_s, was 29.7±13.4 μm/minute, n=156. The drift coefficient, v_d, reflects the imbalance of growth over shortening and is calculated either as

The diffusion coefficient represents fluctuations around the behavior predicted from drift and is calculated as the variance of the velocity frequency histogram. Calculation gave the apparent diffusion coefficient as 13.1±1.3 μm²/minute for the inner, 15.1±1.1μm²/minute for the middle and 13.1±1.3μm²/minute for the outer thirds of the cell, respectively. The ratio of coefficients, diffusion/drift was ∼3 μm and did not change significantly along the cell radius. As shown previously(Vorobjev et al., 1999),positive drift predicts a non-random distribution of MT ends.

To determine the distribution of MT ends along the cell radius, the whole cell was divided into five radial zones. Counts of black or white line segments showed that MT ends, both growing and shortening, were rare near the centrosome and increased exponentially towards the cell margin(Fig. 7D). Thus, in CHO cells,the length distribution of dynamic MT ends is an ascending one and most MTs are long. From an exponential curve fit to the data inFig. 7D, the ratio of the coefficients was independently estimated as 5.8 μm, which is within a factor of 2 of the measured value. Thus, the diffusion plus drift framework is sufficient to account for the observed distribution of MT ends.

Since persistent growth in CHO cells seemed a remarkable property, it was important to test its generality. Consequently, we examined the MT life cycle in NRK cells and cytoplasts obtained from them. For these cells, we observed persistent growth of MTs towards the plasma membrane at a velocity of 17.4±5.7 μm/minute (data not shown). Sequential subtraction analysis of NRK cytoplasts gave a bimodal distribution similar to that for CHO cells(data not shown) with a drift coefficient in the cell interior of 4.0±1.1 μm/minute (n=488, 5 cytoplasts). Additional support for this conclusion was obtained by calculation of the drift coefficient from data available in the literature. Using the expression presented earlier,

In this study, we employed a combination of novel approaches to examine MT behavior in the cell interior. The results provide an updated picture of the MT life cycle in cells with a radial array originating from the centrosome(Fig. 8A). A MT is born by nucleation on a centrosomal template, it grows through persistent elongation by addition of subunits at the plus end; in the vicinity of the cell margin,the MT undergoes fluctuations in length through episodes of plus end shortening and growing; and, finally, it dies when one of the stochastic episodes of shortening is of sufficient length to depolymerize it all the way back to the centrosome. An alternative path of death is release from the centrosome followed by persistent shortening from the minus end. In CHO and other fibroblasts, these properties combine to generate an apparent imbalance of growth over shortening (drift) for individual MTs in the cell interior that leads in the steady-state to a distribution of MT plus ends that steeply climbs as a function of position along the cell radius. This view of the MT life cycle and steady state differs from previous work in three ways: (1) that the default behavior of MTs in the cell interior is to grow persistently; (2)that MTs at the cell periphery show `dynamic instability' type behavior because of a boundary effect; and (3) that positive `drift' generates an ascending distribution of MT lengths.

The frequency of transition from growth to shortening of a MT in the cell interior was determined to be very low, 0.005 second^-1, which equates to a half-time for the growing phase of 140 seconds. Assuming a radius of 20 μm for a typical CHO cell and a growth velocity for the MT plus end of 17 μm/minute, a MT would require only ∼70 seconds to grow the distance of the cell radius, which means that a nascent MT typically arrives at the cell margin before it has had a chance to shorten. Assuming a first-order process of transition to shortening, we calculate that an average of 72% of nascent MTs starting at the centrosome would still be in the growth phase by the time they reached the cell margin. This result is essentially the same as the experimentally observed value of 68% for cells. Cytoplasts are smaller than cells and therefore they have a higher percentage of MTs that arrive at the margin without shortening. Thus, persistent growth of a nascent MT can be understood as resulting from a combination of high elongation velocity and low catastrophe frequency. This remarkable feature of the MT life cycle was not previously recognized. The major explanation for this oversight is most likely that observations of MT dynamics in vivo have generally been made where it is technically easier to obtain the data — namely, at the cell periphery, where the cytoplasm is thin and MT density is relatively low,thus facilitating visualization of individual MTs. An implicit and untested assumption has been that MT dynamics parameters evaluated at the cell periphery are valid throughout the cell. Our results demonstrate that, at least in CHO and NRK cells, this is not the case.

Data consistent with persistent growth can be found in recent papers focusing on GFP-labeled plus end markers(Perez et al., 1999;Mimori-Kiyosue et al., 2000a). Although these reports did not emphasize the persistent behavior of MT growth,they did contain data from which such conclusions could be drawn. In Vero cells, GFP-CLIP-170 tracks of 10 μm length were common(Perez et al., 1999) and a similar picture emerges from movies of EB-1-GFP tracks in Xenopus A6 cells(Mimori-Kiyosue et al., 2000a;supplementary material). Persistent growth as a property of MTs was also described in the fission yeast, S. pombe, where the majority of MTs grew until they reached the cell ends(Brunner and Nurse, 2000). Taking these results together with our observations, we conclude that persistent growth of MTs in the cell interior is not an exceptional property of CHO cells. Rather, it may be an evolutionarily selected and widespread characteristic that has previously escaped notice because it is seen primarily in the cell interior.

As MTs neared the cell margin, their behavior changed to frequent alternation of brief phases of shortening and growth. Such behavior is how dynamic instability has come to be characterized in vivo. For example, growth and shortening lengths in PtK₁ cells have been reported as 1.3 and 1.6 μm and in CHO cells as 3.2 and 4.3 μm, respectively(Shelden and Wadsworth, 1993). We obtained similar values for CHO cells for MTs near the cell edge. Thus, the small duration of growth seen for MTs near the cell margin differed from the persistent growth in the cell interior.

What could be the explanation for this difference in behavior? Are MT dynamics intrinsically different in the cell interior as opposed to the cell margin or only apparently so? Closer inspection of MT behavior revealed that not all parameters depended on position in the cell. Velocities of growth and shortening (17 and 30 μm/minute, respectively) did not. We obtained the same values independently of whether the determinations were made at the cell margin or in the cell interior and they were similar to those obtained previously at the cell periphery (Shelden and Wadsworth, 1993). Likewise, the rescue frequency was essentially constant. In the cell interior, the mean distance of shortening was small (2.9±3.4 μm) and the rescue frequency was high (0.12 second^-1). Shortening of MTs back from the cell margin followed an exponential decay function, signifying that the probability of transition to the growing phase did not depend detectably upon the extent of shortening or position in the cell. The values of shortening length and rescue frequency were essentially the same as those obtained previously at the cell periphery(4.3±3.3 μm and 0.13 second^-1, respectively)(Shelden and Wadsworth, 1993). Finally, when we evaluated the apparent catastrophe frequency near the cell margin, we obtained a value (0.08 second^-1), similar to that previously reported (0.06 second^-1) (Sheldon and Wadsworth, 1993). Thus, the major difference in MT behavior is the apparent high frequency of catastrophe seen at the cell margin versus the low frequency (0.005 second^-1) that we find in the cell interior.

We propose that the difference in catastrophe frequency is only apparent and can be accounted for by a `boundary effect' of the cell margin. The boundary prevents growth from continuing, inducing a pause and/or catastrophe. It follows that observations restricted to the cell periphery underestimate intrinsic growth duration and overestimate the catastrophe frequency. A high rescue frequency prevents MTs from shortening all the way back to the centrosome. Instead, shortening MTs are quickly converted to growing ones,which are then driven back to the cell membrane. Thus, the asymmetry of transition frequencies and the boundary effect combine to produce persistent growth in the cell interior but many small fluctuations of MT length near the cell margin.

What is the nature of the boundary? The simplest possibility is that the membrane itself or the dense actin network that typically underlies it presents a physical obstacle to growth. Alternatively, a chemical signal emitted from the cell surface could change the state of the MT. Growing MTs are marked by the presence of CLIP-170-family proteins (CLIP-170 and CLIP-115 for mammalian cells) (Perez et al.,1999; Hoogenraad et al.,2000), CLASP (Akhmanova et al.,2001), EB-1/APC complex(Mimori-Kiyosue et al., 2000a;Mimori-Kiyosue et al., 2000b)and dynactin complex (Vaughan et al.,1999) at their plus ends. The relationship between the presence of these proteins and MT dynamics has yet to be established for mammalian cells. CLIP-170 family proteins could be good candidates for ensuring MT growth in the cell interior, as the fission yeast ortholog of CLIP-170, tip1p,stabilizes MT growth by preventing premature catastrophe(Brunner and Nurse, 2000). EB1/APC complex, which promotes MT polymerization in vitro(Nakamura et al., 2001), and increases MT stability in vitro and in vivo(Zumbrunn et al., 2001), might also prevent growing MT tips from depolymerization. If any or all of these proteins were required to maintain a low catastrophe frequency or a high rescue frequency and if a factor at the cell surface caused them to dissociate from the MT plus end, the result would be an induced transition to the shortening phase.

The simplest criterion of steady state is that the number and amount of MT polymer does not change over time. An apparently paradoxical result is that in the steady state some MTs persistently grow. Where does the shortening come from to balance the growth? Fig. 8A provides a qualitative explanation. Neglecting the minor contribution that the minus end pathway plays in CHO cells, the growth of a nascent MT is essentially compensated by the complete shortening of another MT from the plus end. Fluctuations of growth and shortening near the cell margin,on average, equate to each other because persistent growth up to the cell boundary restores polymer lost by shortening from the boundary. Thus, cellular levels of MT polymer remain constant in the presence of persistent growth of individual MTs.

A deeper paradox is the ascending distribution of MT ends. Why is the distribution not uniform in the cell? The answer is again related to a boundary effect. Persistent growth and non-persistent shortening of MTs is equivalent to stating that the catastrophe frequency is small and the rescue frequency is high. In CHO cells, the catastrophe frequency is so small that nascent MTs typically arrive at the cell margin before having had a chance to shorten. In contrast, MTs typically shorten for only a small distance before being driven back to the cell margin. Since, by definition, a boundary imposes a limit to growth, the asymmetry in transition frequencies will cause MT ends to spend more time and thus to accumulate near the cell margin(Fig. 8B).

A quantitative evaluation of the MT length distribution is provided by the conceptual framework of treating MT ends as dynamic elements undergoing 1-dimensional diffusion and drift. A diffusion plus drift treatment has been used previously to evaluate MT dynamics in melanophores(Vorobjev et al., 1999) and has been shown theoretically to be a valid approximation of MT dynamic instability (Maly, 2001). In this framework, steady state is attained when fluxes of MT ends due to drift and diffusion are equal and opposite. This equality may be expressed as

The diffusion plus drift framework permits calculation of the distribution of MT plus ends in the steady-state according to the equation

The asymmetry of MT dynamics in vivo stands in contrast to MTs in vitro,where both growth and shortening tend to be persistent(Mitchison and Kirschner,1984b; Horio and Hotani,1986; Walker et al.,1988). The difference between parameters observed in vitro and in vivo suggest that there is a regulatory mechanism at the MT plus end. Several proteins have been suggested to increase the catastrophe frequency including XKCM1 (Walczak et al., 1996;Desai et al., 1999;Tournebize et al., 2000) and stathmin/Op18 (Belmont and Mitchison,1996; Howell et al.,1999). Some MT-associated proteins (MAP4, Xenopus XMAP 215 and its human homolog TOGp) can play an opposite role, rescuing MTs by stabilization and/or assembly-promotion(Tournebize et al., 2000;Spittle et al., 2000;Popov et al., 2001) (for reviews, see Andersen, 2000;Heald, 2000). In addition, plus end binding proteins such as CLIP-170(Perez et al., 1999) or EB1(Mimori-Kiyosue et al., 2000b)represent potential regulatory factors, although mechanistic studies demonstrating how they might affect MT dynamics are lacking.

Growth of nascent MTs in CHO cells was not only persistent but rapid. The velocity we obtained, 15-20 μm/minute, was higher than most values reported for MT growth in other cells, cell extracts or for purified tubulin (reviewed by Odde and Buettner, 1995). However, similar instantaneous values were reported for CHO cells(Shelden and Wadsworth, 1993),GFP-CLIP-170 tracks in Vero cells (Perez et al., 1999) and EB-1-GFP tracks in Xenopus A6 cells(Mimori-Kiyosue et al.,2000a). It is not clear what factors are responsible for the differences in reported rates but our study and the several reports cited indicate that MT growth in vivo can be astonishingly rapid. The rates in vivo are almost an order of magnitude higher than those predicted by in vitro studies of pure tubulin (Walker et al.,1988). This consideration suggests that the rate constant for MT growth in vivo is higher than that determined in vitro or that the cell has a mechanism to maintain the free tubulin pool substantially above the critical concentration.

The life cycle of MTs in CHO and NRK cells displays several characteristic features including persistent growth in the cell interior, asymmetric transition frequencies and effects of the cell boundary. What might be the biological significance of this set of properties? We suggest that the answer is to provide a mechanism by which the MT array can rapidly `sense' changes of the cell periphery. Rapid and persistent growth allows nascent MTs to elongate from the centrosome to the cell boundary in a short time. This would be advantageous for a locomoting cell in which the cell margin is advancing in that the MT cytoskeleton could rapidly accommodate to any protrusion or change in cell shape. Such accommodation could also be important for maintenance of membrane trafficking within the cell, including transport of vesicles either to or away from the cell surface. Finally, efficient sensing of the periphery by MTs may be important for coordinating motile activities around the cell perimeter to achieve directional motility.

The new insights into MT dynamics obtained from this analysis of the MT life cycle in CHO and NRK cells suggest that similar comprehensive analyses in other cell types will be rewarding. It will be important to determine whether different cell types such as fibroblasts, epithelial or neuronal cells have similar or distinctive MT `lifestyles'. In any event, this study points out the critical necessity of carrying out determinations of MT behavior in the cell interior and in recognizing that the cell is a finite system in which its boundary exerts important effects.

We thank I. Maly, T. Svitkina, E. Taylor and G. Albrecht-Buehler for helpful discussions and critical reading of the manuscript, V. Rodionov for help with double-label experiments. This work was supported by NIH Grant GM 25062 to G.G.B., Fogarty International Research Collaboration Award TW-00748 and U.S. Civilian Research and Development Foundation Award RB1-2025 to G.G.B. and I.A.V.