βIII-Tubulin Induces Paclitaxel Resistance in Association with Reduced Effects on Microtubule Dynamic Instability*

The development of resistance to paclitaxel in tumors is one of the most significant obstacles to successful therapy. Overexpression of the βIII-tubulin isotype has been associated with paclitaxel resistance in a number of cancer cell lines and in tumors, but the mechanism of resistance has remained unclear. Paclitaxel inhibits cancer cell proliferation by binding to the β-subunit of tubulin in microtubules and suppressing microtubule dynamic instability, leading to mitotic arrest and cell death. We hypothesized that βIII-tubulin overexpression induces resistance to paclitaxel either by constitutively enhancing microtubule dynamic instability in resistant cells or by rendering the microtubules less sensitive to the suppression of dynamics by paclitaxel. Using Chinese hamster ovary cells that inducibly overexpress either βI- or βIII-tubulin, we analyzed microtubule dynamic instability during interphase by microinjection of rhodamine-labeled tubulin and time-lapse fluorescence microscopy. In the absence of paclitaxel, there were no differences in any aspect of dynamic instability between the two β-tubulin-overexpressing cell types. However, in the presence of 150 nm paclitaxel, dynamic instability was suppressed to a significantly lesser extent (suppressed only 12%) in cells overexpressing βIII-tubulin than in cells overexpressing similar levels of βI-tubulin (suppressed 47%). The results suggest that overexpression of βIII-tubulin induces paclitaxel resistance by reducing the ability of paclitaxel to suppress microtubule dynamics. The results also suggest that endogenous regulators of microtubule dynamics may differentially interact with individual tubulin isotypes, supporting the idea that differential expression of tubulin isotypes has functional consequences in cells.

Microtubules are dynamic polymers composed of ␣␤-tubulin heterodimers that, both in vitro and in living cells, can continuously grow and shorten through tubulin dimer addition and loss at the microtubule ends. Dynamic microtubules are required for many processes in cells including cell migration, cell signaling, and mitosis (1). Mitosis is particularly sensitive to changes in microtubule dynamics, and mitotic progression depends upon the maintenance of microtubule dynamics and microtubule polymer levels within a narrow range (2)(3)(4)(5)(6). Paclitaxel is an extremely effective microtubule-targeted antican-cer drug used to treat a wide range of tumor types (7). The binding of paclitaxel to tubulin in microtubules arrests cells in mitosis, leading to cell death (8). Acquired resistance to paclitaxel is one of the most significant reasons for its failure in chemotherapy (3). Determining the molecular mechanisms of paclitaxel resistance is of great clinical value both in the design of chemotherapeutic treatment strategies and in the development of drugs to avoid or overcome resistance.
The antimitotic and antiproliferative effects of paclitaxel are attributed to its ability to suppress microtubule dynamics and to induce microtubule polymerization and bundling (9 -11), driving dynamics and polymer levels outside of an acceptable range. Because the lowest concentrations of paclitaxel that effectively inhibit cell proliferation and block mitosis suppress microtubule dynamics (6,12,13) without significantly increasing microtubule polymer levels, suppression of microtubule dynamics appears to be its most potent mechanism of mitotic arrest (14).
Paclitaxel binds to the ␤-subunit of tubulin, of which at least seven isotypes exist at the protein level in humans: ␤I, ␤II, ␤III, ␤IVa, ␤IVb, ␤V, and ␤VI. The ␤-isotypes differ primarily within the C-terminal 15-20 amino acids, a region of the protein that lies on the exterior of the microtubule and is the putative binding site for several microtubule-associated proteins (MAPs) 1 (15)(16)(17). Expression of some tubulin isotypes is restricted to specific tissues, whereas other isotypes are constitutively expressed, resulting in a unique pattern of expression for each tissue. In non-neuronal cells, ␤I is often the predominant tubulin isotype, whereas ␤III-tubulin is generally expressed at very low levels. Tumor cells often express a different complement of ␤-tubulin isotypes than their normal counterparts (18). The functional significance of variations in tubulin isotype expression in both normal and tumor cells is not known.
Overexpression of ␤III-tubulin has been associated with paclitaxel resistance in cell lines and in tumors (19). Kavallaris et al. (20) showed that paclitaxel-resistant A549 cells overexpressed ␤III-tubulin compared with their sensitive counterparts and that partial sensitivity to paclitaxel was regained by down-regulation of ␤III-tubulin in these cells. Hari et al. (21) recently showed that overexpression of ␤III-tubulin conferred 1.5-2-fold resistance to paclitaxel in CHO cells; however, the mechanism of resistance remains unclear.
We hypothesized that ␤III-tubulin induces paclitaxel resistance by generating inherently more dynamic microtubules or by rendering the microtubules less responsive to the suppressive effects of paclitaxel on microtubule dynamics, thus allowing cells to maintain sufficiently rapid microtubule dynamics in the presence of paclitaxel to complete mitosis. To test this hypothesis, we analyzed microtubule dynamic instability in CHO cells induced to overexpress ␤III-tubulin or ␤I-tubulin, in the absence and presence of paclitaxel. We found that increased levels of ␤IIItubulin did not affect microtubule dynamic instability in the absence of paclitaxel. However, in the presence of paclitaxel, microtubules in ␤III-tubulin-overexpressing cells were significantly more dynamic than in either ␤I-tubulin-overexpressing cells or uninduced controls. Our results support the hypothesis that increased levels of ␤III-tubulin directly induce paclitaxel resistance by rendering microtubules less sensitive to the effects of paclitaxel. In addition, our results suggest that tubulin isotype composition can regulate dynamic instability through the differential effects of endogenous regulators on microtubules of different isotype compositions.

MATERIALS AND METHODS
All materials were purchased from Sigma unless otherwise noted. Cell Culture, Microinjection, and Analysis of Microtubule Dynamics-CHO cells were stably transfected with C-terminal HA-tagged ␤Ior ␤III-tubulin under the control of a tetracycline-regulatable promoter (21) such that in the presence of tetracycline, expression of the transfected ␤-tubulin was repressed ("tet-off ") and expression of the protein was induced when tetracycline was removed from the medium. In normal CHO cells, ␤I-tubulin is the predominant ␤-isotype comprising ϳ70% of the total ␤-tubulin pool. CHO cells additionally express lower levels of ␤IVb-tubulin (25%) and ␤V-tubulin (5%) (22,23). ␤III-tubulin is normally not expressed. After induction of HA-␤I-or HA-␤III-tubulin expression, the percentages of each ␤-tubulin isotype were 80, 14, 5, and 1% for the HA-␤-tubulin, endogenous ␤I, ␤IVb, and ␤V, respectively (21).
Cells were maintained in ␣ modification of minimal essential medium supplemented with 5% fetal bovine serum (Hyclone, Logan, UT), non-essential amino acids, 0.1% penicillin/streptomycin, 2 mg/ml G418 (BioWhittaker, Walkersville, MD), and 2 g/ml tetracycline. Cells were seeded at a density of 8 ϫ 10 4 cells/2 ml in a 6-well plate onto poly-Llysine-treated, gridded glass CELLocate coverslips (Eppendorf, Westbury, NY) with/without tetracycline. The cells were washed in phosphate-buffered saline after 8 h to remove any residual tetracycline and incubated for an additional 16 h, followed by a 24-h incubation in medium containing a reduced concentration of fetal bovine serum (2%) to promote cell flattening.
Rhodamine tubulin was prepared by carboxyrhodamine labeling (Molecular Probes, Eugene, OR) of microtubules assembled from phosphocellulose-purified bovine brain tubulin (24). To visualize microtubule dynamic instability, cells were microinjected with rhodaminelabeled tubulin (2.5 mg/ml) as described previously (2). Paclitaxel was dissolved in dimethyl sulfoxide and stored at Ϫ80°C until use. For experiments involving paclitaxel, cells were incubated with 150 nM paclitaxel for an additional 5 h prior to imaging and transferred to recording medium lacking paclitaxel during imaging (15 min-2 h).
Time-lapse Microscopy, Image Acquisition, and Analysis of Dynamic Instability-Microscopy and analysis of dynamic instability have been described elsewhere (2,12). Briefly, cells were sealed in a Rose chamber in recording medium containing 30 l/ml Oxyrase (Oxyrase Inc., Mansfield, OH). 31-46 images were captured at 4-s intervals. The growth and shortening dynamics of individual microtubules were tracked using the track points function of Metamorph, converted to life history plots, and analyzed using real time measurement software (25). Dynamicity and time-and length-based catastrophe and rescue frequencies were calculated per microtubule as described previously (12).
Mitotic Index-Cells were plated at a concentration of 8 ϫ 10 4 cells/2 ml into 6-well plates. After 48 h, cells were incubated in the absence or presence of paclitaxel at a range of concentrations (0.03-10 M) for 6 h. Fixation, staining, and determination of mitotic indices were performed on all cells, both floating and attached, as described in Ref. 12. Results are the mean and S.D. of five experiments, in each of which 500 cells were counted per drug concentration.

RESULTS
We used paclitaxel-resistant CHO cells overexpressing HAtagged ␤III-tubulin and paclitaxel-sensitive HA-␤I-tubulinoverexpressing cells to determine whether ␤III-tubulin expression differentially affects microtubule dynamics in the presence or absence of paclitaxel. In normal CHO cells, ␤I-tubulin is the predominant ␤-tubulin isotype. ␤III-tubulin is normally not expressed (21). Because we analyzed microtubule dynamics in individual cells, it was important to confirm that all the cells expressed similar levels of HA-␤I-or HA-␤III-tubulin after inducing expression under the conditions used to analyze dynamic instability. Cells were incubated in the presence or absence of tetracycline for 48 h and processed for immunofluorescence microscopy (see "Materials and Methods"). In the absence of tetracycline (Fig. 1, A, C, E, G), all of the cells expressed HA-␤I-(A, E) or HA-␤III-(C, G) tubulin at approximately equivalent levels, whereas in the presence of tetracycline, expression of the transfected ␤-tubulin was repressed and levels were negligible (B, D, F, H). In Fig. 1, E-H, cells were incubated in the presence or absence of tetracycline (to repress or induce HA-␤-tubulin expression, respectively), and cells were labeled with an antibody to HA (green) and tubulin (red) and the images were overlaid. In induced cells, HA-␤I-(E) or HA-␤IIItubulin (G) (green) coassembled with endogenous tubulin (red) producing yellow microtubules. Uninduced cells (F, H) expressed extremely low levels of HA-␤-tubulin, indicated by red microtubules.
To determine the effects of overexpressed ␤III-tubulin on microtubule dynamic instability, cells overexpressing ␤IIIor ␤I-tubulin were microinjected with rhodamine-labeled tubulin. The dynamic instability of individual microtubules in the peripheral regions of the two cell types was observed and recorded by time-lapse fluorescence microscopy and quantified (see "Materials and Methods"). As expected from previous studies of microtubule dynamic instability, in ␤I-overexpressing cells many of the individual microtubules underwent periods of slow growth and more rapid shortening within the 3-min observation period. For example, as shown in Table I, in ␤Ioverexpressing cells, microtubules grew at a mean rate of 14.4 Ϯ 6.8 m/min and shortened approximately twice as fast, at a mean rate of 33.2 Ϯ 12.8 m/min. The parameters of microtubule dynamics in ␤III-overexpressing cells were virtually identical to those of ␤1-overexpressing cells (Table I, compare data columns 1 and 2). Thus, by itself, overexpression of ␤III-tubulin in CHO cells does not significantly affect the dynamic instability of interphase microtubules.
To determine the paclitaxel concentration to use in microtubule dynamics experiments, we first determined the mitotic index after incubating cells for 5 h with a range of concentrations of paclitaxel (see "Materials and Methods"). As shown in Fig. 2, in the absence of paclitaxel 3.2% of the cells were in mitosis. The mitotic index increased to 4.8% at 0.1 M paclitaxel, and arrest was maximal by 1 M (ϳ16%). We analyzed microtubule dynamic instability in cells incubated with 150 nM paclitaxel, at which the mitotic index was ϳ6%. At this concentration, the microtubules remained somewhat dynamic, but the effects of paclitaxel were clearly detectable.
Paclitaxel had relatively minor effects on dynamic instability in ␤III-overexpressing cells, whereas the drug significantly suppressed many parameters of microtubule dynamics in ␤Ioverexpressing cells (Table I, compare data columns 1 and 3). The major effects of paclitaxel in ␤I-overexpressing cells are summarized in data column 1 of Table II; the growth rate was suppressed by 31%, the growth length by 45%, the shortening rate by 54%, the shortening length by 55%, and the overall microtubule dynamicity by 47%.
By contrast, only two parameters of microtubule dynamic instability were significantly affected by 150 nM paclitaxel in ␤III-overexpressing cells, the shortening rate (Ϫ32%) and the shortening length (Ϫ34%), and these changes resulted in only an insignificant change in overall dynamicity (Ϫ12%) (see data columns 2 and 4 in Table I and the summary in data column 2 in Table II).
To rule out the possibility that the effects we observed were due to other additional genetic changes that had occurred in the ␤III-overexpressing cells as a result of transfection or extensive passaging, we analyzed dynamic instability in ␤III-overexpressing cells under non-inducing conditions in the presence of paclitaxel. These cells expressed very low levels of ␤III-tubulin (Fig. 1, D, H). Paclitaxel (150 nM) suppressed microtubule dynamics to levels in ␤III-uninduced cells nearly identical to those in ␤I cells. (Compare data columns 3 and 5 in Table I and compare percentages in data columns 1 and 3 in Table II.) DISCUSSION We found that ␤III-tubulin overexpression induces paclitaxel resistance by decreasing the efficacy of paclitaxel binding to ␤III-tubulin, resulting in a weaker suppressive effect on microtubule dynamics. The results are also the first direct demonstration that drugs can differentially interact with different tubulin isotypes in cells, suggesting that microtubule dynamic instability might be regulated by the differential interaction of endogenous regulators with the individual isotypes as well.
Overexpression of ␤III-Tubulin Does Not Cause an Inherent Increase in Dynamic Instability as Compared with Overexpression of ␤I-Tubulin-One mechanism by which ␤III-tubulin has been proposed to mediate resistance to paclitaxel is to constitutively increase microtubule dynamics, such that in the presence of paclitaxel microtubules would remain sufficiently dynamic to complete mitosis (2). In living interphase CHO cells in the absence of paclitaxel, overexpression of ␤III-tubulin did not significantly alter any parameters of microtubule dynamic instability. Thus, in these cells a high level of ␤III-tubulin overexpression does not cause an inherent increase in dynamic instability, indicating that this is not the mechanism of ␤IIItubulin-induced resistance. Hari et al. (21) showed previously that in the ␤III-tubulin-overexpressing CHO cells polymer mass was decreased by ϳ30%. Taken together, these results suggest that a reduction in polymer mass does not necessitate a change in cellular microtubule dynamics. This result also suggests that in our previous study of paclitaxel-resistant and -dependent A549 cells (2), the increased microtubule dynamics resulted not from the increased levels of ␤III-tubulin but rather from other changes associated with resistance (a mutation in the putative stathmin and MAP4 binding site in conjunction with increased levels of unphosphorylated (active) stathmin and phosphorylated (inactive) MAP4 (26)).
The lack of a difference in microtubule dynamics parameters in the absence of paclitaxel observed in the present study in cells overexpressing either ␤Ior ␤III-tubulin is perhaps surprising. In two previous in vitro studies, the microtubules assembled from ␣␤III-tubulin were significantly more dynamic than microtubules assembled from ␣␤IV-tubulin. For example, comparison of their dynamicities indicated that ␣␤III-microtubules were 2.2-fold more dynamic than ␣␤IV-microtubules in one study (27) and 1.7-fold more dynamic in a second study (28). The sequence of ␤IV resembles that of ␤I; thus one might expect that microtubules in cells overexpressing ␤III-tubulin might be more dynamic than those in cells overexpressing ␤I-tubulin. In cells, tubulin undergoes post-translational modifications and interacts with a large number of microtubule regulatory proteins (ranging from proteins that induce catastrophe to those that stabilize microtubule dynamics) (11); thus the dynamics of individual isotypes may be significantly altered in cells.
Paclitaxel Has a Weaker Effect on Dynamic Instability in ␤III-Tubulin-overexpressing Cells-Microtubules in cells overexpressing ␤III-tubulin were significantly less susceptible to the suppressive effects of paclitaxel than in control cells. In controls (␤I-overexpressing cells or uninduced ␤III-tubulintransfected cells), paclitaxel significantly reduced the mean growth and shortening rates and lengths and dynamicity. In ␤III-overexpressing cells, only the mean shortening rate and length were reduced, and these parameters were affected to a significantly lesser extent than in controls (Tables I and II and Fig. 3).
Dynamicity is a calculated value representing the overall dynamic instability of a population of microtubules. In previous studies, dynamicity was reduced by 25-64% in conjunction with drug-induced inhibition of cell proliferation or mitotic arrest in A498, CaOV3, A549, and MCF7 cancer cell lines (6,12,13). Consistent with these results, in ␤I-overexpressing CHO cells, dynamicity was reduced by 47%. However, in ␤IIIoverexpressing paclitaxel-resistant cells, the dynamicity was reduced by only a statistically insignificant 12%. The lack of an effect of paclitaxel on dynamic instability in ␤III-overexpressing cells is also consistent with the in vitro study of Derry et al. (28), who found that microtubules composed exclusively of ␣␤III-tubulin were 7-fold less sensitive to suppression by paclitaxel than microtubules composed of ␣␤IIor unfractionated tubulin. Thus, both in vitro and in living cells, increased levels of ␤III-tubulin reduce the ability of paclitaxel to suppress microtubule dynamics.
The possibility that the C-terminal HA tag on the transfected ␤-tubulin artifactually produced the observed results is highly unlikely. First, previous studies have shown that the presence of the HA tag does not affect microtubule assembly or endogenous MAP4 binding nor does it change the growth or drug resistance properties of the cells when compared with its transfected untagged ␤-tubulin counterpart (21,29). Second, the HA tag was present on both overexpressed isotypes. Finally, transfection of ␤III-tubulin with or without the HA tag induced the same degree of paclitaxel resistance, indicating that the ␤IIItubulin alone is responsible for resistance (21). Suppression of microtubule dynamics correlates with inhibition of proliferation and sensitivity to paclitaxel (2, 6); thus, the lack of suppression of microtubule dynamics is the most likely explanation for the resistance phenotype.
Whether differential tubulin isotype expression has functional significance in cells has been a long-standing, unresolved question (30). Results presented here indicate that ␤-tubulin isotypes could potentially play a role in the control of microtubule dynamics in cells by differentially interacting with endogenous microtubule regulatory proteins. Microtubule dynamics are exquisitely controlled in cells by a large number of proteins, including the specific plus-end and minus-end microtubulebinding proteins and those that bind along microtubule surfaces (1,31). It is likely that many of the binding sites for drugs on microtubules are also binding sites for microtubule regulatory proteins. The microtubule-associated protein tau, for example, may bind to the same or an overlapping site on microtubules as paclitaxel (32), and, like paclitaxel, suppresses microtubule dynamic instability in cells (33). Thus it is plausible that the tubulin isotype composition of microtubules may function to determine which regulatory proteins bind to and act on microtubules.
What Is the Mechanism of ␤III-Tubulin-mediated Paclitaxel Resistance?-Our results indicate that the weaker effect of paclitaxel on microtubules in ␤III-tubulin-overexpressing cells is due to reduced drug binding or reduced drug efficacy (i.e. reduced ability of the drug to induce the conformational change required to suppress microtubule dynamics). Other microtubule-targeted drugs, including estramustine and colchicine, interact more weakly with ␤III-tubulin than with several other isotypes in vitro (34,35). In addition, colchicine binding studies and cysteine cross-linking studies suggest that ␤III-tubulin has a more rigid conformation than the other ␤-isotypes examined (18,36). Thus, microtubules composed largely of ␤III-tubulin might resist the paclitaxel-induced conformational changes that suppress microtubule dynamic instability. Tubulin dimers are thought to undergo a conformational change upon addition to the microtubule end (15). The rigidity of ␤III-tubulin might also cause it to resist this conformational change, which may account for the reduced microtubule polymer in ␤III-overexpressing cells. It is also conceivable that paclitaxel binding is prevented by a MAP with higher affinity for ␤III-tubulin than the other isotypes. However, this possibility is less likely be-  cause the in vitro studies indicating altered interactions between paclitaxel and the ␤-tubulin isotypes were performed in the absence of MAPs (28), suggesting a weaker or less efficacious interaction between paclitaxel and ␤III-tubulin.
One important question is how the levels of ␤III-tubulin overexpression in the paclitaxel-resistant CHO cells used here (80% of total tubulin) (21) relate to levels of isotype overexpression in cell lines and tumor samples that are resistant to paclitaxel. Quantitation of individual tubulin isotypes in cell lines and tissues is problematic, and adequate data are not available to answer this question. Early attempts to determine the relative abundance of tubulin isotypes in cells relied upon relative mRNA levels, which have since been shown to be poor indicators of tubulin expression (37,38). Current methods to quantitate ␤-tubulin isotypes in cells have generated different results depending upon the methodology (39,40). In tumor samples, these studies are further complicated by the presence of normal cells as well as many other cell types (39).
Development of more sophisticated methods to measure tubulin isotype expression will be necessary to understand what level of overexpression of ␤III-tubulin significantly contributes to paclitaxel resistance. In addition, understanding how ␤IIItubulin expression is regulated will be important in preventing or overcoming this type of drug resistance.
In summary, we have shown directly in living human tumor cells that increasing the expression of ␤III-tubulin can induce paclitaxel resistance by rendering microtubules relatively insensitive to the effects of paclitaxel. Clinical determination and regulation of ␤III-tubulin expression levels in tumors may lead to significantly improved prediction of tumor response to paclitaxel and ultimately to overcoming one source of clinical resistance to paclitaxel. FIG. 3. Histograms indicating the differential effects of 150 nM paclitaxel on growth rate (a), shortening rate (b), growth length (c), shortening length (d), and dynamicity (e) in cells induced to express ␤Ior ␤III-tubulin. Black bars, no paclitaxel; gray bars, 150 nM paclitaxel. *, values significantly differ from ␤I controls (no paclitaxel) at Ն99% confidence interval.