Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://cellmotility.genebee.msu.ru/html/articles/jcs.pdf Äàòà èçìåíåíèÿ: Sun Jan 11 15:51:04 2004 Äàòà èíäåêñèðîâàíèÿ: Mon Oct 1 19:46:04 2012 Êîäèðîâêà:

Journal of Cell Science 110, 2635-2645 (1997) Printed in Great Britain © The Company of Biologists Limited 1997 JCS4467

2635

Cytoplasmic assembly of microtubules in cultured cells
Ivan A. Vorobjev1, Tatyana M. Svitkina2 and Gary G. Borisy2,*
1Laborator 2Laborator

y of Cell Motility, A. N. Belozersky Institute, Moscow State University y of Molecular Biology, University of Wisconsin, 1525 Linden Drive, Madison, WI 53706, USA

*Author for correspondence (e-mail: ggborisy@facstaff.wisc.edu)

SUMMARY The origin of non-centrosomal microtubules was investigated in a variety of animal cells in culture by means of time-lapse digital fluorescence microscopy. A previous study (Keating et al. (1997) Proc. Nat. Acad. Sci. USA 94, 5078-5083) demonstrated a pathway for formation of noncentrosomal microtubules by release from the centrosome. Here we show a parallel pathway not dependent upon the centrosome. Correlative immunostaining with anti-tubulin antibodies and electron microscopy established that apparent free microtubules observed in vivo were not growing ends of long stable microtubules. Free microtubules appeared spontaneously in the cytoplasm and occasionally by breakage of long microtubules. Estimates of the frequencies of free microtubule formation suggest that it can be a relatively common rather than exceptional event in PtK1 cells and may represent a significant source of noncentrosomal microtubules. The observation of free microtubules permitted analysis of the microtubule minus end. Unlike the plus end which showed dynamic instability, the minus end was stable or depolymerized. Breakage of long microtubules generated nascent plus and minus ends; the nascent minus end was generally stable while the plus end was always dynamic. The stability of microtubule minus ends in vivo apparently provides the necessary condition for free microtubule formation in the cytoplasm. Parameters of the dynamic instability of plus ends of free microtubules were similar to those for the distal ends of long microtubules, indicating that the free microtubules were not exceptional in their dynamic behavior. Random walk analysis of microtubule end dynamics gave apparent diffusion coefficients for free and long microtubules which permitted an estimate of turnover half-times. The results support the concept that, in PtK1 cells, a pathway other than plus end dynamics is needed to account for the rapidity of microtubule turnover.

Key words: Microtubule, Dynamic, Centrosome, Cultured cell, Fluorescence microscopy

INTRODUCTION The centrosome is generally thought to be the major site of nucleation of cytoplasmic microtubules (MTs) in animal cells. Nonetheless, many specialized cell types contain numerous MTs not associated with the centrosome. Non-centrosomal MTs are particularly evident in axonal and dendritic processes of neurons (Bray and Bunge, 1981; Chalfie and Thomson, 1979), marginal bands of avian erythrocytes (Murphy et al., 1986), and polarized epithelial cells (Mogensen et al., 1997). In animal cells in culture, tubulin immunofluorescence reveals an extensive array of MTs which fills the cytoplasm. The great density of the MTs in the perinuclear region has precluded a determination of the proportion of MTs in fact connected to the centrosome. However, the general impression given by the tubulin immunofluorescence images is that many MTs are probably not connected. The capacity of cytoplasm to nucleate and maintain MTs independent of the centrosome has been demonstrated in microsurgically obtained fragments (McNiven and Porter, 1988; Maniotis and Schliwa, 1991; Rodionov and Borisy, 1997), and in cytoplasts lacking centrosomes (Karsenti et al., 1984). Thus, taken together, substantial evidence exists for the formation and presence of non-centrosomal MTs in a

variety of cell types. However, the mode of formation of noncentrosomal MTs in intact cells which normally contain a centrosome remains unclear. The formation of non-centrosomal (or, free) MTs could conceivably be dependent or independent of the centrosome. A dependent pathway would consist of nucleation at the centrosome followed by release of the MT into the cytoplasm. Direct evidence for this model, originally proposed as the `conveyer hypothesis' (Vorobjev and Chentsov, 1983; Vorobjev and Nadezhdina, 1987) was recently obtained for epithelial cells in culture (Keating et al., 1997). A centrosome independent pathway would consist of nucleation or self-assembly of MTs in the cytoplasm. In our study of MT release from the centrosome (Keating et al., 1997), we did note that new MTs not visibly connected to the centrosome did appear in the cytoplasm. These could have resulted from self-assembly in the cytoplasm. However, another possibility was that they represented growth from the plus ends of pre-existing stable MTs that did not incorporate labeled subunits during the incubation period. One objective of the current study was to distinguish between these possibilities. An additional mechanism for the formation of non-centrosomal MTs might be the severing or breakage of MTs. Severing activity has been reported in


2636 I. A. Vorobjev, T. M. Svitkina and G. G. Borisy extracts of egg cytoplasm (McNally and Vale, 1993) and such a mechanism, if general, could operate on MTs formed by either the centrosome dependent or independent pathways. In our previous study, breakage was seen only rarely, but our observations were focussed primarily on the centrosomal region. Here, we re-evaluate this possibility by more comprehensive observations throughout the cytoplasm. Whatever pathways operate for the formation of non-centrosomal MTs, an additional question is the cellular role of these MTs. One aspect of MT behavior likely to be related to whether MTs are connected to the centrosome is turnover. The majority of MTs in cultured interphase cells turn over with a half-time of about 5 minutes (reviewed by Gelfand and Bershadsky, 1991). The length distribution of MTs in cultured mammalian cells in interphase has not been precisely determined but many MTs are thought to exceed 20 µm. In melanophores, where the turnover half-times are also of the order of 5 minutes (Rodionov et al., 1994), MTs may be up to 50 µm in length. The question is whether dynamic instability (Mitchison and Kirschner, 1984) can account for the turnover of such long MTs. MTs both in vitro (Hotani and Horio, 1988; Walker et al., 1988) and in vivo (Sammak and Borisy, 1988; Schulze and Kirschner, 1988) do not generally shorten to zero length; rather, they are `rescued' and return to the growing phase. This `tempered' dynamic instability presents a paradox for the MT system of the interphase cell, the limited dynamic instability of individual MTs seems to contradict the rapid turnover of the MT array as a whole. For example, from the observed dynamic instability parameters in PtK1 cells (Shelden and Wadsworth, 1993) a Monte Carlo calculation estimated the half-time for turnover of 20 µm MTs to be greater than 1 hour (Gliksman et al., 1993). A priori, the discrepancy between instability and rapid turnover may be overcome by assuming either that a mechanism additional to plus end dynamics is involved in the turnover, or that MTs on average are much shorter than believed. A rapid turnover pathway could come from the depolymerization of the minus end. Observations on minus end behavior are scarce in the literature and contradictory. Minus ends of MTs polymerized in vitro, although intrinsically dynamic, seem to be more stable than plus ends (Walker et al., 1989). Observations made in vivo after laser transection suggested that cytoplasmic MTs depolymerized primarily from nascent plus ends (Tao et al., 1988). On the other hand, recent observations of MT treadmilling in melanophore fragments (Rodionov and Borisy, 1997) and MT release from the centrosome followed by frequent depolymerization of the minus end (Keating et al., 1997) demonstrate the potential lability of the minus end. These results prompt a more thorough examination of minus end behavior in vivo. We report here observations on the cytoplasmic origin of free (non-centrosomal) MTs in a variety of cultured cells, we describe the behavior of plus and minus ends of the free MTs and we discuss the implications of the results for the turnover of MTs in the cell. MATERIALS AND METHODS
Cell cultures PtK1 epithelial cells (American Type Culture Collection, Rockville, MD) were cultured in F-10 medium (Gibco, Grand Island, NY) supplemented with 10% FBS (Hyclone Laboratories, Logan, UT), 20 mM Hepes, pH 7.2, and antibiotics (100 i.u./ml penicillin, 0.1 mg/ml streptomycin), and kept at 37°C in 5% CO2. Cells were plated 3-5 days before experiments at low concentration into observation chambers made by attaching a coverslip with Sylgard silicone elastomer (Dow Corning, Midland, MI) over a hole drilled in a 35 mm culture dish. Mouse 3T3 cells (American Type Culture Collection, Rockville, MD) and REF-52 cells (immortalized rat embryo fibroblasts) were cultured as for PtK1 cells but were plated at high concentration. After 3-4 days, a part of the monolayer was scratched with a razor blade and 6 hours later or on the next day, cells at the edge of the `experimental wound' were microinjected. Black tetra (Gymnocorymbus ternetzi) keratocytes were obtained from fish scales as described elsewhere (Svitkina et al., 1997) and cultured in DMEM (Hepes modification) (Sigma Chemical Co., St Louis, MO), pH 7.2, supplemented with 20% fetal bovine serum (HyClone Labs, Logan, UT) and antibiotics (100 i.u./ml penicillin, 0.1 mg/ml streptomycin, 0.1 mg/ml gentamycin). Cells in colonies were used for microinjection on the next day after plating. Preparation of Cy3-tubulin Cy3-labelled porcine brain tubulin was prepared as described elsewhere (Keating et al., 1997) and stored in 10 µl aliquots in liquid nitrogen. Prior to microinjection, a 10 µl aliquot of Cy3-tubulin was diluted with 5 µl PM buffer (0.1 M Pipes, 1 mM MgCl2, pH 6.9), centrifuged at 200,000 g for 10 minutes at 4°C to remove particulate material and minimize micropipette clogging, and stored on ice until the time of injection. Imaging and data analysis Cells injected with Cy3-tubulin were transferred to indicator-less 199 culture medium containing 10 mM Hepes and were treated with the oxygen-depleting enzyme Oxyrase (Oxyrase, Inc., Ashland, OH) to reduce photodamage and photobleaching (Mikhailov and Gundersen, 1995). Oxyrase was added to the observation chamber at a final dilution of 2-3% (v/v) of the original stock, along with lactic acid at a final concentration of 20 mM. The Oxyrase-treated dishes were then covered with a layer of mineral oil (Squibb and Sons, Princeton, NJ). Injected cells were observed on a Nikon Diaphot 300 inverted microscope equipped with a Plan â100, 1.25 NA objective using a rhodamine filter set. Images of 16-bit depth were collected with a CH250 slow scan, cooled CCD camera (Photometrics Ltd, Tucson, AZ) driven by Metamorph imaging software (Universal Imaging Corp., Westchester, PA). The image was projected onto the CCD chip at a magnification of 0.106 µm/pixel (9.6 pixels per micron). Exposure times were 0.2-0.3 seconds, and images were collected at 2.4-5.0 second intervals. Mammalian cells were kept at 37°C during observation, fish keratocytes at 28°C. A typical series comprised 120-150 frames covering a period of 5-10 minutes. 16-bit images were rescaled with Metamorph software or NIHImage software (National Institute of Health) and 8-bit images processed for presentation with Adobe Photoshop (Adobe Systems, Mountain View, CA). To highlight the MT of interest in some figures, red and yellow overlays were painted over the MT in Adobe Photoshop with opacity set to 30%. Photobleaching of labeled MTs Photobleaching of labeled MTs was performed on a Zeiss IM-35 inverted microscope using an Argon ion laser as described elsewhere (Keating et al., 1997), except that the cells were incubated at 37°C. Analysis of microtubule dynamics Position of MT ends and MT length were traced by mouse-driven cursor under NIH-Image software, and further data analysis performed using Sigma Plot (Jandel Scientific Corp., San Rafael, CA). All values given in the text are mean ± s.d. For free MTs, three parameters were determined, position of both ends and the entire length.


Free microtubules in the cytoplasm 2637
For comparison, long MTs coming from the central parts of cells were examined in the same sequences. Determination of growth and depolymerization rates was first performed as described elsewhere (Shelden and Wadsworth, 1993) with a threshold difference in length (for free MTs) or position of the end (for long MTs) of 0.3 µm between subsequent images. The minimal rate that could be determined using this threshold was 7 µm/minute. Slower rates could be determined by averaging several subsequent images (see Dhamodharan and Wadsworth, 1995, for details), but then the relative frequency of small rates becomes artificially lowered. As an alternative approach, we used a one-dimensional random walk model (Berg, 1993) to characterize the dynamics of MTs, resulting in a single `diffusion' constant. The magnitude of end displacements from their initial position were taken each 30 seconds (up to 5 minutes). The diffusion coefficient, D, was determined using linear regression analysis of mean square displacement versus time. Immunofluorescence staining Cells were fixed in a culture dish with glutaraldehyde (at a final concentration of 1%), permeabilized in 1% Triton X-100, rinsed in PBS (phosphate buffer saline, pH 7.3), treated with sodium borohydride (2% in PBS), and stained with monoclonal anti-tubulin antibodies (Amersham), then FITC-conjugated anti-mouse antibodies (Sigma). Images were acquired and processed as described above, then in Adobe Photoshop set to the green channel and superimposed onto a live cell image set to the red channel. Electron microscopy Correlative light and electron microscopy of the cytoskeleton was performed as described (Svitkina et al., 1995). Briefly, cells on coverslips were lysed for 3-5 minutes at room temperature with 1% Triton X-100 in a cytoskeleton buffer (100 mM Pipes, pH 6.9, 1 mM MgCl2, 1 mM EGTA) containing 4% polyethylene glycol (Mr 40,000), treated with recombinant gelsolin N-terminal domain, fixed with glutaraldehyde, tannic acid and uranyl acetate, critical point dried, and coated with platinum and carbon. To preserve microtubules, 10 µg/ml of taxol was added to all solutions before fixation.

RESULTS Visualization of free microtubules Close observation in a variety of cell types revealed the presence of apparently free MTs among the general population of long MTs (Fig. 1). The apparently free MTs grew and shrank in the same way as the long MTs in the same cellular regions. Remarkably, some apparently free MTs were born, that is, appeared during the observation period, while others died, that is, shortened to zero length. These observations suggested a pathway for MT formation independent of the centrosome. An alternative explanation for the appearance of apparently free MTs was addition of labeled subunits to the ends of long, stable and, thus, unlabeled MTs. This possibility did not seem very likely because our fluorescence observations were made between 1 and 2 hours after microinjection, a time sufficient for tubulin to equilibrate between the subunit and polymer pools. Nevertheless, a small subset of MTs has been demonstrated to survive in cultured mammalian cells for up to a cell generation (Webster et al., 1987). To test this possibility, we monitored the dynamics of apparently free MTs and then carried out correlative immunofluorescence microscopy of the same cells (Fig. 2A-C). Except for slight differences at the ends of MTs, which may represent growth or shortening between the acquisition of the last live cell image and the moment of

fixation, the tubulin immunofluorescence staining pattern proved virtually identical to the fluorescence images of the living cells. No domains were seen by tubulin immunostaining at the end of fluorescently labeled domains. Rather, the apparently free MTs were localized at the same position and had the same length both in vivo and after immunostaining. This correspondence shows that the apparently free MTs are not simply extensions at the end of long, unlabeled MTs. Another possibility was that the apparently free MTs were in fact some unusual aggregates of injected tubulin, such as sheets of protofilaments, rather than real MTs. To test this possibility, we performed correlative electron microscopy of MTs previously recorded in live cells injected with labeled tubulin. In intact cytoskeletons, individual MTs are difficult to trace because of abundant actin filaments, but depletion of actin filaments by treatment with the actin-severing protein, gelsolin, allows for better visualization of MTs (Svitkina et al., 1995). Fluorescence images of MTs taken from the same cells at different stages of preparation for electron microscopy were identical except for slight growth or shrinkage at MT ends, which might have occurred after acquiring the last fluorescence image and before cell lysis (unpublished results). The same pattern of MT distribution was observed by electron microscopy. In platinum replicas, long MTs were visualized as gently curving fibers about 28 nm thick. The apparently free MTs were localized at the same positions as in live cells and were undistinguishable by morphology from regular, long MTs (Fig. 2D). Thus, both immunofluorescence and electron microscopic studies confirm that the short MTs observed at steadystate were not labeled fragments growing from the ends of stable MTs. We conclude that the short MTs are indeed free and of cytoplasmic origin. The dynamics of MTs were cell type specific. In epithelial cells, PtK1 and keratocytes, growth and shortening excursions were mainly short, while in fibroblasts, 3T3 and REF-52, they were longer. In fibroblasts, free MTs often grew or shrank out of the field of view (or into an area of high MT density), thus hindering their continuous observation. In contrast, the greater stability of free MTs in the cells of epithelial origin permitted sufficient observation to characterize the dynamics at both of their ends. Cytoplasmic formation of free microtubules Detailed examination was performed on free MTs in PtK1 cells. To facilitate observation, we selected for microinjection and analysis cells in culture satisfying either of two criteria. First, cells were selected which were at the edge of large islets and had thin lamellae. Second, cells were selected which were stretched flat by their neighbors. These cells generally had thin layers of cytoplasm under their nuclei. In stable lamellae, short MTs continually appeared, grew and shrank, and disappeared. The lifespan of MTs that grew to about 1-2 µm was frequently less than 1 minute. MTs that grew up to 6-8 µm had an average lifespan exceeding our observation window (10 minutes). In cells with a free edge, we sometimes found large lamellar regions with few MTs invading them. In such regions, numerous free MTs generally appeared and grew extensively in a short period of time, filling the lamellae to a normal MT density within 20-30 minutes (Fig. 3). Individual free MTs appeared frequently and, after excursions of growth and shrinking, many of them depolymerized completely. Never-


2638 I. A. Vorobjev, T. M. Svitkina and G. G. Borisy

Fig. 1. Apparent free MTs in cultured cells. (A) 3T3 fibroblast. A free MT is born just before 0 time and rapidly grows. Its minus end remains at the same position throughout the series. (B) PtK1 cell. One free MT grows across preexisting MTs (0-186 seconds), and then becomes stable (442 seconds). Its minus end at 442 seconds is slightly shifted from its initial position. A short MT is stable (0-186 seconds), then slightly shortens (442). (C) Black tetra keratocyte. One free MT shortens (0-23 seconds), then grows (53 seconds) and disappears (123 seconds). Another free MT is stable (0-53 seconds), then moves laterally (123 seconds). Annotated MTs are overlaid with red. Growth and shortening of long MTs can be seen in each cell. Time after the first frame shown at lower left in seconds. Bar, 5 µm.

theless, in terms of both number and length, net growth prevailed against shortening. In addition to cytoplasmic nucleation, free MTs could be formed by release of MTs from the centrosome (Keating et al., 1997) and/or MT breakage. We attempted to assess the frequencies of these other routes relative to free MT formation by making direct observations in the same cell populations and under the same conditions. The frequency of MT release was approximately 1% of centrosomal MTs per minute (20 centrosomes, 380 MTs, 3,678 MT-minutes of observation, 27 release events). This is a lower value than reported previously (~6%) but the current results were obtained wholly at steady-state whereas most of the previous results were obtained shortly after microinjection. In 5

favorable cases where the centrosome was located under the center of the nucleus, all centrosomal MTs could be counted and gave an average of 57±28 MTs per centrosome. Thus, the number of free MTs produced by centrosomal release was approximately 0.5 per minute. Breakage was also a relatively infrequent event. In the course of analyzing many sequences, instances of spontaneous breakage of MTs were encountered. These were not the result of photodamage as they occurred at the beginning of an image sequence as well as toward the end. Overall, we observed 21 MT breakage events in 34 series analysed where each series comprised 100 to 200 images over a period of 5 to10 minutes. This gives a breakage rate of 0.08 per viewing area (20 µm â 25 µm) per minute. Because of the density of MTs, their


Free microtubules in the cytoplasm 2639

Fig. 2. Apparent free MTs are not growing ends of long stable MTs. Correlative fluorescence and electron microscopy of the same MT. (A) Cy3 MTs in a living PtK1 cell just before fixation; (B) tubulin immunofluorescence staining; (C) merged image of Cy-3 tubulin in red and tubulin antibody staining in green, showing essential identity. (D) electron microscopy of a free MT in rat embryo fibroblast visualized as a platinum replica after actin removal by gelsolin treatment. Inset: Cy-3 MTs in same cell just before extraction and fixation. Boxed region of inset shown in D. Red overlay of MTs in the EM image shows exact correspondence with the fluorescence image. Bars: (A-C) 3 µm; (D) 1 µm.

number was difficult to determine precisely and ranged from 30 to 70. On the assumption that a typical image contained approximately 50 MTs, the total data set therefore consisted of approximately 13,000 MT-minutes of observation, giving a breakage frequency of 0.16% per cytoplasmic MT per minute. The same 34 series of images were analysed for the appearance of free MTs. The formation of very short MTs (<2.5 µm) was quite frequent but many of them were ephemeral (lifetime <1-2 minutes). To avoid biassing the results by the behavior of these ephemeral MTs, analysis was restricted to free MTs which grew to >2.5 µm. With this criterion, 166 cases of MT formation were observed giving a birth frequency of 0.65 per viewing area per minute or 1.3% per cytoplasmic MT per minute. These estimates suggest that formation of free MTs is a common, not exceptional event and occurs more frequently than by release or breakage. However, the estimates involve

comparisons of disparate events and could contain unrecognized sources of error. Another way to assess the relative importance of pathways contributing to free MTs is to consider just the relative frequency of birth to death of free MTs. In the steady-state, the proportion of free MTs must hold constant; all pathways leading to formation of free MTs must be balanced by an equivalent loss, that is, release + breakage + birth = death. Consequently, if release from the centrosome and breakage were major mechanisms for the generation of free MTs, one would expect the death frequency to be significantly higher than their birth frequency. In fact, in the same data set where we recorded 166 cases of birth, we observed 131 cases of death. Of these numbers, 65 of the MTs both appeared and disappeared during the observation period. The imbalance of birth over death may reflect a preference for birth in the lamellum where most obser-


2640 I. A. Vorobjev, T. M. Svitkina and G. G. Borisy

Fig. 3. Cytoplasmic formation of free MTs. Time sequence shows free MTs which appear, grow and fill the nascent lamellum of a PtK1 cell. Almost no long MTs grew into the lamellum area from the central part of the cell. About 40 free MTs appeared in this series and their orientation seems nearly random. Time after first frame shown in lower left of each frame in seconds. Bar, 10 µm.

vations were made. The high frequency of birth is consistent with the idea that formation of free MTs in the cytoplasm is a significant cellular process. Behavior of plus and minus ends of free microtubules The visualization of free MTs permitted analysis of the dynamics of their ends. This was of special interest for the minus end because little information is available on its kinetic properties in vivo. Free MTs showed considerable heterogeneity of behavior (Fig. 4). Typically, one end was relatively quiet and the other end active. Although an independent polarity marker was not employed in this study, the active and quiet ends were considered to be plus and minus, respectively, on the basis of their dynamics. The active (plus) end fluctuated about a rather steady length (Fig. 4A), grew extensively (Fig. 4B) or shortened (Fig. 4C); sometimes, the plus end was quiet (Fig.

4D) and sometimes the minus shortened (Fig. 4E). Occasionally, both ends changed position in parallel (Fig. 4F), suggesting either simultaneous growth at the plus end and shortening at the minus end (treadmilling) or axial transport of the MT as a whole. Photobleach marking experiments (data not presented) showed that free MTs both treadmilled and were transported, in varying proportion, as reported previously for MTs released from the centrosome (Keating et al., 1997). Axial transport (not treadmilling) could in some instances result in displacement of the minus end which could be mistaken for growth. However, in the photobleach marking experiments, we never observed a clear instance of growth of the minus end. In some cases MTs were transported non-axially (laterally) towards the cell center as has been described (Mikhailov and Gundersen, 1995). The non-axial transport seemed to be the result of an independent process because the end dynamics of such MTs were not distinguishable from stationary MTs.


Free microtubules in the cytoplasm 2641 increase the probability of breakage. In any event, breakage produced a nascent plus end at the distal end of the remaining long MT and a nascent minus end at the proximal end of the newly formed free MT. The nascent ends displayed kinetic properties similar to pre-existing ends, that is, nascent plus ends were dynamic while nascent minus ends were not. Of 21 cases of breakage observed, the minus end remained stable until the end of the image series (17 cases), slowly shortened (4 cases), but never grew. In 3 of the 4 cases, shortening became paused and in 1 case shortening continued to completion. This behavior was similar to that for spontaneously nucleated MTs where, out of 37 cases examined, we found stability in 33 cases and shortening from the minus end in 4 cases. Thus, cellular breakage of MTs did not produce minus ends that were more labile than the ends of free MTs. This result suggests one of two possibilities: stabilizing factors are not limited to the minus end of an MT or they are able to rapidly associate with a nascent minus end. Analysis of microtubule dynamics: microtubule plus ends undergo a one-dimensional random walk Life history plots of free MTs taken at random show remarkable heterogeneity, raising the question of how to characterize the behavior. Using a procedure of analysis involving threshholding similar to that described elsewhere (Shelden and Wadsworth, 1993; Dhamodharan and Wadsworth, 1995), we determined a growth rate of 11.0±3.7 µm/minute and a shortening rate of 14.8±7.5 µm/minute similar to what was reported. However, these average rates did not reflect a typical behavior for the population of MTs. Analysis of instantaneous rates of either growth or shortening gave exponentially declining distributions of values; that is, lower rates were more frequent than higher rates. Also, the duration of growth and shortening periods varied and, in 90% of cases, the duration was a single interval between sequential images. Modification of the threshold for determining growth or shortening did not change this basic property, suggesting that the life history plots of MTs were fractal in nature. The fractal-like character of MT dynamics suggested that a random-walk analysis might be appropriate. Although MTs tend to shrink faster than they grow, the time spent shrinking is less than the time spent growing and, so, the `step' size in either direction is the same; further, transitions between growth and shrinking are stochastic. Thus, displacement of the end of a MT from its initial position may be treated as a one-dimensional random walk. The mathematical expression for analysis of a random walk is that the mean square displacement is proportional to time, = 2Dt-, where x is displacement, D is the `diffusion coefficient', t is time and is a `persistence time'. Random walk analysis was applied to displacement data for free MTs and for the ends of long MTs and the results are shown in Fig. 6. Linear regression analysis indicated that both data sets could be fitted by a straight line and that the `persistence' (intercept) was slight or insignificant. Thus, over the time window analyzed (300 seconds), both free MTs and the ends of long MTs may be considered as random walkers. The calculated diffusion coefficient for the plus ends (D+) of free MTs was 2.6±1.4 µm2/minute (mean ± s.d.; n=37) and for the ends of long MTs was 2.1±1.5 µm2/minute (mean ± s.d.; n=33); these values were, within the error of the measurements, essentially the same (P>0.25 in a paired t-test). Thus, the

Fig. 4. Representative life history plots of free MTs. Upper curve in each plot, plus end; lower curve, minus end. The reference point for each curve was the position of the minus end at zero time. Plus ends express different behavior: (A) dynamic instability; (B) growth; (C) growth and shortening that brings the MT to complete depolymerization; (D) stability with minor oscillations. Minus ends are usually stable (A-D), but in some cases they slowly depolymerize (E). (F) An example of a MT that polymerizes and then moves along its axis with plus end leading and minus end trailing. All plots shown are from PtK1 cells. MT origin (birth) and complete depolymerization (death) are indicated by arrows.

Overall, the behavior of the minus end of a free MT was characterized either by stability or shortening. The intrinsic behavior of MTs in vitro is to display dynamic activity at both minus and plus ends (Walker et al., 1988). Thus, the relative stability of the minus end in vivo suggested the presence of a cellular regulatory factor. Observations of preexisting or newly formed free MTs were consistent with them forming from a stable nucleating complex. The possibility that the stability of a free MT depended upon a structure limited to the minus end could be tested by breaking a MT along its length. Although the cause of spontaneous MT breakage is not known, these events nevertheless offered an opportunity to assess the kinetic properties of the newly created minus and plus ends. An example of a long MT whose plus end was displaying dynamic instability is shown in Fig. 5. The tubule became bent with reverse curvature shortly before breakage as if it was placed under compression from its plus end. A plus end motor acting on the MT could have such an effect and thus


2642 I. A. Vorobjev, T. M. Svitkina and G. G. Borisy

Fig. 5. MT breakage. Distal end of a MT (red) in a PtK1 cell shows dynamic instability, it grows, shortens and starts growing again (0148); breakage (150) produces a nascent plus end of the remaining long MT segment (red) which shortens (153-285), and a nascent minus end of a newly formed free MT (yellow) which remains stable (153-285). The plus end of the free MT grows. Time shown in upper right corner of each frame in seconds. Bar, 5 µm.

Fig. 6. Random walk analysis of MTs end behavior. (A) Plus and minus ends of free MTs; (B) ends of long MTs. Mean square displacement in µm2 versus time. The results for plus ends for both free and long MTs are closely approximated by linear regression (dashed lines, 95% confidence intervals) and are statistically indistinguishable from each other. Minus ends on average show negligible displacement activity.

DISCUSSION Cytoplasmic assembly of free MTs Self-assembly of MTs occurs spontaneously in vitro provided the tubulin subunit concentration is above a critical value. The formation of MTs in non-centrosomal fragments of cytoplasm (McNiven and Porter, 1988; Maniotis and Schliwa, 1991; Rodionov and Borisy, 1997) and in cytoplasts lacking centrosomes (Karsenti et al., 1984) indicates that interphase cytoplasm does not contain factors that suppress self-assembly.

kinetic behavior of the plus ends of free MTs was not distinguishable from that of long MTs. The diffusion coefficient for the minus ends (D-) of free MTs was 0.19±0.43 µm2/minute (mean ± s.d.; n=37), which was only slightly greater than zero, reflecting the relative stability of the minus end.


Free microtubules in the cytoplasm 2643 The biological question of interest is whether MTs can selfassemble in cytoplasm of cells containing a centrosome. The presence or absence of the centrosome is thought to be critical because of its function not only to nucleate MT formation but to cap and anchor the minus end. If the minus end is indeed capped by structures at the centrosome, such as -tubulin ring complexes (Moritz et al., 1995; Zheng et al., 1995), then the steady-state concentration of tubulin becomes determined by the properties of the plus end. The intrinsic structural polarity of the MT is reflected in different assembly properties at its two ends. If the steady-state concentration at the plus end is lower than at the minus end, treadmilling will result (Margolis and Wilson, 1978; Bergen and Borisy, 1980; Hotani and Horio, 1988). The coexistence of dynamic instability (Mitchison and Kirschner, 1984) and treadmilling has been demonstrated in vitro (Farrell et al., 1987) and treadmilling has recently been demonstrated in vivo (Rodionov and Borisy, 1997). At a steady-state determined solely by plus ends, a MT presenting both plus and minus ends would be unstable because its exposed minus end would depolymerize. In a cytoplasm dominated by centrosomal MTs and, in the absence of a mechanism stabilizing the minus end, non-centrosomal (free) MTs could exist only transiently. Thus, the occurrence of free MT formation is inextricably linked to the concept of minus end stabilization. Minus ends of free MTs in PtK1 cells are stable The production of free minus ends of MTs in PtK1 cells has now been observed by three pathways: self-assembly in the cytoplasm, release from the centrosome (Keating et al., 1997) and breakage of MTs. Independent of the pathway of production, the behavior of the minus end was similar. The minus end was never seen to grow; it was either stable or shortened. In vitro studies of purified tubulin showed that stability of the minus end is not an intrinsic property of tubulin polymer, although transitions between growth and shortening phases at the minus end are less frequent than at the plus end (Walker et al., 1988). Thus, it seems likely that stability of the minus end is a cellular property of PtK1 cells. The identity of the stabilizing factor(s) remains to be established. One possibility is that -tubulin ring complexes (Zheng et al., 1995; Moritz et al., 1995) in the cytoplasm are capable of nucleating and stabilizing MTs. Such complexes could associate reversibly with the minus end. Dissociation of the complex would result in minusend shortening. Alternatively, nascent minus ends formed by the occasional breakage event might associate with a ring complex and become stabilized. The fact that 80% of nascent minus ends appearing after MT breakage were stable indicates that stability does not depend upon a single nucleation structure limited to the original minus end and is consistent with a reversible stabilization mechanism. Stabilization of minus ends in vivo carries with it several implications. First, it permits the survival of free MTs in the cytoplasm. To the extent that these MTs are capped at their minus end, they are kinetically the same as centrosomal MTs. That is, the tubulin concentration in the cytoplasm will equilibrate at the steady-state value for the plus end. Consequently, uncapped free MTs will spontaneously depolymerize. Second, the competitive advantage enjoyed by centrosomal MTs over free (uncapped) MTs is abrogated. A consequence is that nucleation at the centrosome is no longer as favored over cytoplasmic formation. Thus, under minus end stabilization conditions, many non-centrosomal MTs are predicted to arise. Third, the reversibility of the minus end cap is also important. The uncapping of the minus end with the resultant loss of MT stability, provides a minus end pathway for MT depolymerization and turnover. This process is similar to the minus end pathway proposed for turnover of MTs released from the centrosome (Keating et al., 1997). Free MT assembly versus MT release from the centrosome The number of MTs in a mammalian cell in culture, centrosomal and non-centrosomal is difficult to determine with precision. Our observations in PtK1 cells suggest that the number of centrosomal MTs in interphase is highly variable and ranged from 25 to 100. Electron microscopic observations on other epithelial cells in culture also gave small numbers, from 20 to 50 MTs per centrosome (Alieva et al., 1992; Alieva and Vorobjev, 1994). We have not found in the literature a determination of the total number of MTs in an interphase PtK1 cell. Fluorescence microscopy does not permit a precise determination because of overlap of MTs and inability to track many MTs reliably. However, our observations give a rough estimate of 500 MTs for the cell as a whole, suggesting that the proportion of centrosomal MTs is about 10-20%. The observed frequency of MT release from the centrosome was 0.5-3 per minute. We observed cytoplasmic assembly in different regions of a cell (at the periphery and close to the centrosome), suggesting that cytoplasmic assembly of MTs occurs with similar probability throughout the cell. The observed frequency of MT formation in the cytoplasm was approximately 0.65 per minute per viewing area (20 µm â 25 µm). Taking into account that we always observed very thin layers of cytoplasm, corresponding to a minor part of the overall cell volume, we estimate that tens of MT appear every minute as a result of cytoplasmic assembly. Thus cytoplasmic assembly of MTs is not an exception, but rather a common event and, in PtK1 cells, is comparable if not a more frequent occurrence than release from the centrosome. The significance of the self-assembled MTs remains to be determined. Many of the newly born MTs are ephemeral but others survive and grow. The proportion of self-assembled MTs that grow to appreciable lengths, say 10 µm, has yet to be determined precisely but may be of the order of 5 to 10%. Even a proportion of 5% would make the self-assembly pathway of significance for MT formation in the cell. Finally, we note that the explosion of free MT formation in nascent lamellae (Fig. 3) suggests the possibility that the frequency of cytoplasmic assembly of MTs might be regulated. Dynamics of MT plus ends as a one-dimensional random walk Random walk analysis characterizes MT behavior in a single parameter D, the diffusion coefficient. It is important to note certain limitations of this analysis. Random walk analysis is appropriate only at steady-state and when MT ends convert from growth to shortening and vice versa stochastically. On the other hand, when MTs show long periods of growth or shortening and infrequent conversions between them, their behavior is better described in terms of characteristic rates. Long term behavior of MTs which would accompany locomotion or sub-


2644 I. A. Vorobjev, T. M. Svitkina and G. G. Borisy stantial shape change is likely to have a systematic component and is also not appropriate for a random walk analysis. Our observations, carried out at steady-state and within a 5 minute window, seemed to reasonably well fit random walk behavior. Given the validity of this analysis, the `diffusion coefficient' obtained permits one to evaluate overall MT dynamics and to estimate turnover times. The similarity of D values for free MTs and long MTs indicates that the free MTs were not kinetically exceptional and suggests that their behavior is the same as that for any other MT. They are born short but may grow long and then become indistinguishable from other long MTs. We have to this point not categorized the long MTs as centrosomal or non-centrosomal. Without tracking the long MTs to their minus end termination, one cannot be certain of their identity. However, the relatively small number of MTs seen to clearly radiate from the centrosome suggests that the bulk of the long MTs are noncentrosomal. As previously discussed, some of these may have arisen by release from the centrosome but most may arise in PtK1 cells by formation in the cytoplasm. The value of the diffusion coefficient for plus ends, 2.4 µm2/minute, signifies that plus end dynamics would require approximately 70 minutes to turn over a MT 20 µm in length. This conclusion is similar to that drawn from a Monte Carlo analysis (Gliksman et al., 1993) and is far greater than the 5 minutes usually reported for turnover half-times of interphase mammalian cells (Saxton et al., 1984). For melanophore MTs which may be up to 50 µm long, over 7 hours would be required. These calculations suggest that a pathway other than plus end dynamics contributes to turnover in epithelial cells. Previously, we suggested that MT release from the centrosome followed by depolymerization from the minus end could provide such a pathway (Keating et al., 1997). Our current results suggest that a minus end pathway also exists for MTs formed free in the cytoplasm. Although most free MTs at any given time show a stable minus end, approximately 20% show depolymerization. At an average depolymerization rate of 5 µm/minute (Keating et al., 1997), a 20 µm MT would require only 4 minutes to completely depolymerize. If capped minus ends became uncapped stochastically, all MTs could be turned over by the minus end pathway. The half-time for turnover would depend on the length distribution of MTs which becomes an important parameter to determine. For a mean length in the range 10 to 15 µm, the half-time would be 10 to 15 minutes, close to reported values. The apparent role for a minus end pathway in MT turnover in epithelial cells does not necessarily signify a similar role for other types of cells. Plus end dynamics would be sufficient for achieving rapid turnover in a cell with small MTs such as yeast. An alternative strategy would be for MTs to undergo longer and more frequent shortening excursions. Nevertheless, in all large cells, a general problem is how the centrosome can manage cytoplasm at a distance. The formation of free MTs in the cytoplasm may offer possibilities for local control and selforganization which might relieve the centrosome of some of its burden.
We thank Tom Keating, Vladimir Rodionov, Andrew Hope and John Peloquin for stimulating discussions and a critical reading of the manuscript and John Peloquin for preparation of Cy-3 tubulin. This work was supported by National Institutes of Health Grant GM 25062 to G.G.B., Fogarty International Research Collaboration Award TW00748 to G.G.B. and I.A.V., U.S. Civilian Research and Development Foundation Award RB1-168 to G.G.B. and I.A.V.

REFERENCES
Alieva, I. B., Vaisberg, E. A., Nadezhdina, E. S. and Vorobjev, I. A. (1992). Microtubule and intermediate filament patterns around the centrosome in interphase cells. In The centrosome (ed. V. C. Kalnins), pp. 103-129. Academic Press, San Diego. Alieva, I. B. and Vorobjev, I. A. (1994). Centrosome behavior under the action of uncoupler and the effect of disruption of cytoskeleton elements on the uncoupler-induced alterations of the centrosome. J. Struct. Biol. 113, 217224. Berg, H. C. (1993). Random Walks in Biology. Princeton University Press, Princeton, NJ. Bergen, L. G. and Borisy, G. G. (1980). Head-to-tail polymerization of microtubules in vitro. J. Cell Biol. 84, 141-150. Bray, D. and Bunge, M. B. (1981). Serial analysis of microtubules in cultured rat sensory axons. J. Neurocytol. 10, 589-605. Chalfie, M. and Thomson, J. N. (1979). Organization of neuronal microtubules in the nematode Caenorhabditis elegans. J. Cell Biol. 82, 278289. Dhamodharan, R. and Wadsworth, P. (1995). Modulation of microtubule dynamic instability in vivo by brain microtubule associated proteins. J. Cell Sci. 108, 1679-1689. Farrell, K. W., Jordan, M. A., Miller, H. P. and Wilson, L. (1987). Phase dynamics at microtubule ends: the coexistence of microtubule length changes and treadmilling. J. Cell Biol. 104, 1035-1046. Gelfand, V. I. and Bershadsky, A. D. (1991). Microtubule dynamics: mechanism, regulation and function. Annu. Rev. Cell Biol. 7, 93-116. Gliksman, N. R., Skibbens, R. V. and Salmon, E. D. (1993). How the transition frequencies of microtubule dynamic instability (nucleation, catastrophe, and rescue) regulate microtubule dynamics in interphase and mitosis: analysis using a Monte Carlo computer simulation. Mol. Biol. Cell 4, 1035-1050. Hotani, H. and Horio, T. (1988). Dynamics of microtubules visualized by darkfield microscopy: treadmilling and dynamic instability. Cell Motil. Cytoskel. 10, 229-236. Karsenti, E., Kobayashi, S., Mitchison, T. and Kirschner M. W. (1984). The role of centrosome in organizing of microtubule array: properties of cytoplasts containing or lacking centrosomes. J. Cell Biol. 98, 1763-1776. Keating, T. J., Peloquin, J. G., Rodionov, V. I., Momcilovic, D. and Borisy G. G. (1997). Microtubule release from the centrosome. Proc. Nat. Acad. Sci. USA 94, 5078-5083. Maniotis, A. and Schliwa, M. (1991). Microsurgical removal of centrosomes blocks cell reproduction and centriole generation in BSC-1 cells. Cell 67, 495-504. Margolis, R. L. and Wilson, L. (1978). Opposite end assembly and disassembly of microtubules at steady state in vitro Cell 13, 1-8. McNally, F. J. and Vale, R. D. (1993). Identification of katanin, an ATPase that severs and disassembles stable microtubules. Cell 75, 419-429. McNiven, M. A. and Porter, K. (1988). Organization of microtubules in centrosome-free cytoplasm. J. Cell Biol. 106, 1593-1605. Mikhailov, A. V. and Gundersen, G. G. (1995). Centripetal transport of microtubules in motile cells. Cell Motil. Cytoskel. 32, 173-186. Mitchison, T. and Kirschner, M. W. (1984). Dynamic instability of microtubule growth. Nature 312, 237-242. Mogensen, M. M., Mackie, J. B., Doxsey, S. J., Stearns, T. and Tucker, J. B. (1997). Centrosomal deployment of -tubulin and pericentrin: evidence for a microtubule nucleating domain and a minus-end docking domain in certain mouse epithelial cells. Cell Motil. Cytoskel. 36, 276-290. Moritz, M., Braunfeld, M. B., Sedat, J. W., Alberts, B. and Agard, D. A. (1995). Microtubule nucleation by gamma-tubulin-containing rings in the centrosome. Nature 378, 638-640. Murphy, D. B., Grasser, W. A. and Wallis, K. T. (1986). Immunofluorescence examination of beta tubulin expression and marginal band formation in development of chicken erythroblasts. J. Cell Biol. 102, 625-638. Rodionov, V. I., Lim, S.-S., Gelfand, V. I. and Borisy, G. G. (1994). Microtubule dynamics in fish melanophores. J. Cell Biol. 126, 1455-1464. Rodionov, V. I. and Borisy, G. G. (1997). Microtubule treadmilling in vivo. Science 275, 215-218.


Free microtubules in the cytoplasm 2645
Sammak, P. J. and Borisy, G. G. (1988). Direct observation of microtubule dynamics in living cells. Nature 332, 724-726. Saxton, W. M., Stemple, D. L., Leslie, R. J., Salmon, E. D., Zavortink, M. and McIntosh, J. R. (1984). Tubulin dynamics in cultured mammalian cells. J. Cell Biol. 99, 2175-2186. Schulze, E. and Kirschner, M. (1988). New features of microtubule behaviour observed in vivo. Nature 334, 356-359. Shelden, E. and Wadsworth, P. (1993). Observation and quantification of individual microtubule behavior in vivo: microtubule dynamics are cell-type specific. J. Cell Biol. 120, 935-945. Svitkina, T. M., Verkhovsky, A. B. and Borisy, G. G. (1995). Improved procedures for electron microscopic visualization of the cytoskeleton of cultured cells. J. Struct. Biol. 115, 290-303. Svitkina, T. M., Verkhovsky, A. B., McQuade, K. M. and Borisy, G. G. (1997). Analysis of the actin-myosin II system in fish epidermal keratocytes: mechanism of cell body translocation. J. Cell Biol. 139 (in press). Tao, W., Walter, R. J. and Berns, M. W. (1988). Laser-transected microtubules exhibit individuality of regrowth, however most free new ends of the microtubules are stable. J. Cell Biol. 107, 1025-1035. Vorobjev, I. A. and Chentsov, Yu. S. (1983). The dynamics of reconstitution of microtubules around the cell center after cooling. Eur. J. Cell Biol. 30, 139143. Vorobjev, I. A. and Nadezhdina, E. S. (1987). The centrosome and its role in the organization of microtubules. Int. Rev. Cytol. 106, 227-293 Walker, R. A., O'Brien, E. T., Pryer, N. K., Soboeiro, M. F., Voter, W. A., Erikson, H. P. and Salmon, E. D. (1988). Dynamic instability of individual microtubules analyzed by video light microscopy: rate constants and transition frequences. J. Cell Biol. 107, 1437-1448. Walker, R. A., Inoue, S. and Salmon, E. D. (1989). Asymmetric behavior of severed microtubule ends after ultraviolet microbeam irradiation of individual microtubule in vitro. J. Cell Biol. 108, 931-937. Webster, D. R., Gundersen, G. G., Bulinski, J. C. and Borisy, G. G. (1987). Differential turnover of tyrosinated and detyrosinated microtubules. Proc. Nat. Acad. Sci. USA 84, 9040-9044. Zheng, Y., Wong, M. L., Alberts, B. and Mitchison, T. (1995). Nucleation of microtubule assembly by a gamma-tubulin-containing ring complex. Nature 378, 578-583. (Received 3 July 1997 - Accepted 28 August 1997)

Note added in proof While this work was in progress, a related study appeared by A. C. Yvon and P. Wadsworth (J. Cell Science (1997) 110(19) 2391-2401), drawing similar conclusions about the existence of free microtubules in cytoplasm and the relative stability of the minus end.