Paclitaxel-resistant Human Ovarian Cancer Cells Have Mutant β-Tubulins That Exhibit Impaired Paclitaxel-driven Polymerization*

Acquired resistance to paclitaxel can be mediated by P-glycoprotein or by alterations involving tubulin. We report two paclitaxel-resistant sublines derived from 1A9 human ovarian carcinoma cells. Single-step paclitaxel selection with verapamil yielded two clones that are resistant to paclitaxel and collaterally sensitive to vinblastine. The resistant sublines are not paclitaxel-dependent, and resistance remained stable after 3 years of drug-free culture. All cell lines accumulate [3H]paclitaxel equally, and no MDR-1mRNA was detected by polymerase chain reaction following reverse transcription. Total tubulin content is similar, but the polymerized fraction increased in parental but not in resistant cells following the paclitaxel addition. Purified tubulin from parental cells demonstrated paclitaxel-driven increased polymerization, in contrast to resistant cell tubulin, which did not polymerize under identical conditions. In contrast, epothilone B, an agent to which the resistant cells retained sensitivity, increased assembly. Comparable expression of β-tubulin isotypes was found in parental and resistant cells, with predominant expression of the M40 and β2 isotypes. Sequence analysis demonstrated acquired mutations in the M40 isotype at nucleotide 810 (T → G; Phe270 → Val) in 1A9PTX10 cells and nucleotide 1092 (G → A; Ala364 → Thr) in 1A9PTX22 cells. These results identify residues β270 and β364 as important modulators of paclitaxel’s interaction with tubulin.

Paclitaxel (PTX) 1 is a complex diterpene (1), active against a broad range of human tumors, including ovarian and breast carcinomas (2)(3)(4). The primary target of PTX is the microtubule (MT), which is vital for mitosis, motility, secretion, and proliferation (5). PTX stabilizes MTs by disrupting the dynamic equilibrium between soluble tubulin dimers and their polymerized form (6) and is a potent inhibitor of chromosome replication, blocking cells in the late G 2 or mitotic phases of the cell cycle (7)(8)(9).
Resistance to PTX has been shown to be mediated by the drug efflux pump, P-glycoprotein, but could also result from alterations in its intracellular target, tubulin. Although precise molecular alterations in MTs have not been identified, several mechanisms have been proposed. These include decreased tubulin (10), point mutations as evidenced by altered migration of ␣or ␤-tubulin on two-dimensional SDS-polyacrylamide gel electrophoresis (11)(12)(13), differential expression of ␤-tubulin isotypes (14 -17), and acetylation of ␣-tubulin (18).
To further elucidate the mechanism of action of PTX and to investigate mechanisms of PTX resistance other than those mediated by Pgp, we have isolated two PTX-resistant sublines derived from a human ovarian carcinoma cell line. Selected in the presence of the Pgp antagonist, verapamil, resistant cells do not express Pgp, but they exhibit a stable PTX-resistant phenotype. Compared with parental cells, resistant cells display defective PTX-driven tubulin polymerization both in intact cells and in vitro using purified tubulin. No differences are observed in the amount of total tubulin or in the relative level of expression of the ␤-tubulin isotypes. However, two different point mutations in ␤-tubulin were identified in the resistant cells. These results provide the first demonstration of tubulin point mutations in PTX-resistant cell lines and introduce a novel molecular mechanism of PTX resistance.

EXPERIMENTAL PROCEDURES
Materials-PTX, its C-2 analog 2-debenzoyl-2-meta-azidobenzoylpaclitaxel (19), and epothilone B (EPO B) were obtained from the Drug Synthesis and Chemistry Branch of NCI, National Institutes of Health. Vinblastine was from Sigma, and verapamil was from Knoll pharmaceuticals. Mouse monoclonal IgG antibodies against chick brain ␣-tubulin or rat brain ␤-tubulin were from Sigma. Horseradish peroxidaseconjugated sheep anti-mouse Ig antibody was from Amersham Corp. ECL (enhanced chemiluminescence) Western blotting detection reagents were from Amersham. The Taq DyeDeoxy terminator Cycle Sequencing kit was from Applied Biosystems, Inc., the PCR Select-III spin columns were from 5 Prime 3 3 Prime, Inc., and the Centri-Sep spin purification columns were from Princeton Separations. T 4 polynucleotide kinase was from Life Technologies, Inc., and [␥-32 P]ATP was from Amersham. All other chemicals were of reagent grade and were from Sigma.
Cell Culture-The 1A9 cell line is a clone of the human ovarian carcinoma cell line, A2780 (20). The two resistant sublines, 1A9PTX10 and 1A9PTX22, were isolated as individual clones in a single step selection, by exposing 1A9 cells to 5 ng/ml PTX in the presence of 5 g/ml verapamil, a Pgp antagonist. After an initial expansion in 5 ng/ml PTX, the concentration of PTX was gradually increased to 15 ng/ml. The cells utilized in these studies were maintained in 15 ng/ml PTX and 5 g/ml verapamil continuously. Drug was removed for 5-7 days prior to an experiment.
Cytotoxicity Assay-Cytotoxicity assays were performed in 96-well plates as described previously (21), by seeding 500 cells/well and incubating with cytotoxic agents for 4 days. The IC 50 was defined as the dose of drug required to inhibit cell growth by 50%.
Tubulin Polymerization Assay-To quantitate tubulin polymerization, a simple assay was developed by modifying a method originally described by Minotti (10). Cells grown to confluency in 24-well plates were washed twice with phosphate-buffered saline and lysed at 37°C for 5 min in the dark, with 100 l of hypotonic buffer (1 mM MgCl 2 , 2 mM EGTA, 0.5% Nonidet P-40, 2 mM phenylmethylsulfonyl fluoride, 200 units/ml aprotinin, 100 g/ml soybean trypsin inhibitor, 5.0 mM ⑀-amino caproic acid, 1 mM benzamidine, and 20 mM Tris-HCl, pH 6.8) without or with 0.04 -4 g/ml PTX. The presence of PTX in the hypotonic buffer allowed the assay to be performed on a crude tubulin extract, eliminating factors such as drug uptake and metabolism as experimental variables. The lysates were transferred to 1.5-ml Eppendorf tubes, and each well was rinsed with 100 l of hypotonic buffer. Following a brief but vigorous vortex, the samples were centrifuged at 14,000 rpm for 10 min at room temperature. The 200-l supernatants containing soluble (cytosolic) tubulin were transferred to another Eppendorf tube separating them from the pellets containing polymerized (cytoskeletal) tubulin. The pellets were resuspended in 200 l of hypotonic buffer.
The cytosolic and cytoskeletal fractions were each mixed with 70 l of 4 ϫ SDS-polyacrylamide gel electrophoresis sample buffer (45% glycerol, 20% ␤-mercaptoethanol, 9.2% SDS, 0.04% bromphenol blue, and 0.3 M Tris-HCl, pH 6.8) and heated at 95°C for 5-10 min. Twenty microliters of each sample were analyzed by immunoblotting, and ECL was quantitated by densitometry. The percentage of polymerized tubulin was determined by dividing the densitometry value of polymerized tubulin by the total tubulin content (the sum of the densitometry values of soluble and polymerized tubulin). The distribution of tubulin between the soluble and polymerized fractions could be altered to some extent by experimental conditions. Increased temperature during harvesting increased the polymerized fraction, so care was taken to maintain consistency, and each experiment was performed with appropriate controls.
Mixing Experiment-For the mixing experiment, parental 1A9 and clone 1A9PTX10 cells were lysed in hypotonic buffer, without or with 40 g/ml PTX, for 5 min at 37°C. Whole cell lysate from parental 1A9 cells was added to clone 1A9PTX10 lysate at varying ratios ranging from 1:1 to 1:19 (parental:resistant) at room temperature. The mixed lysates were incubated at 37°C for another 5 min, and then the extent of tubulin polymerization was assessed using the tubulin polymerization assay described above.
Tubulin in Vitro Polymerization Assay-Tubulin was purified from the cell lines as described previously (22). The final preparation of polymerization-competent tubulin was Ͼ95% pure by SDS gel analysis. In vitro polymerization of pure tubulin was assayed as described previously (23). The concentration of tubulin was 7.5 M, the concentration of GTP was 0.1 mM, and the concentration of PTX and epothilone B was 1 M in all cases.
Electron Microscopy-Fifty-microliter aliquots were removed from polymerization reactions at steady state using pipette tips with a wide orifice to avoid shearing. These samples were applied without fixation to carbon/formvar-coated grids and allowed to adsorb for 30 s. After draining, the grids were negative stained with 1% uranyl acetate and air-dried.
Quantitative PCR-Quantitative PCR for ␤-tubulin isotypes was performed as described previously (24). Specific primers for each ␤-tubulin isotype: M40 (class I), ␤9 (class II), ␤4 (class III), 5␤ (class IVa), ␤2 (class IVb), were utilized as described previously (16) (arabic numerals refer to the human gene, and roman numerals refer to the protein isotype class (25)). In the case of the ␤4 (class III) forward primer, the bases CAC were added at the 3Ј-end. The primers utilized in the quantitative PCR, with the exception of the ␤9 primers, span one or two introns. This allows for discrimination between RNA and DNA products. For ␤9, PCR was performed without reverse transcription, and no product was obtained, confirming the absence of DNA contamination.
Sequence Analysis of ␤-Tubulin-For PCR amplification and sequencing of ␤2 or M40 isotypes, four overlapping sets of primers for each isotype were used, as summarized in Table I. Despite the sequence identity of ␤2 and M40, the primers were designed to be specific for each isotype, using published sequence data for M40 and ␤2 (26). The Gen-Bank TM accession numbers for ␤2 and M40 isotypes are X02344 and J00314, respectively. PCR-amplified cDNA was purified with PCR Select-III spin columns and directly sequenced with the Taq DyeDeoxy terminator cycle sequencing kit following the manufacturer's instructions (Applied Biosystems, Inc.). The primers used for sequencing were the same primers used for PCR amplification. The reaction products were purified with Centri-Sep spin purification columns, electrophoresed on 48-cm/4.75% polyacrylamide/urea gels, and analyzed by an automated DNA sequencing system (model 373A; Applied Biosystems, Inc.).
Oligonucleotide Hybridization-Oligonucleotide hybridization was performed using probes specific for the mutations found in the resistant sublines. PCR amplification of 1 g of RNA after reverse transcription or PCR amplification of 1 g of DNA from the three cell lines was performed. Equal amounts of the products were applied to nitrocellulose membrane in each of two adjacent slots. For the ␤270 mutation (nucleotide 810) of the M40 isotype, set 3 primers were used, as described in Table I. For the ␤364 mutation (nucleotide 1092) of the M40 isotype, the forward primer was the primer of set 4 (Table I), and the reverse primer was 5Ј-1248 CGACCTCGTCTCTGAGTATC 1267 -3Ј.
To avoid the possibility of DNA contamination in the RNA, the RNA was treated with DNase (RNase-free) for 1 h at 37°C and 5 min at 75°C, prior to PCR. Subsequently, the absence of DNA contamination was confirmed by PCR amplification without the reverse transcription step.
Oligonucleotide probes specific for each of the mutations were 5Јphosphorylated with [␥-32 P]ATP and T 4 polynucleotide kinase. These probes were 801 ATGCCTGGCXTTGCCCCTCT 819 (where X represents T for the wild type and G for the mutant) and 1080 GGCCTCAAGATGX-CAGTCACC 1098 (where X represents G for the wild type and A for the mutant). Fifty and five nanograms of oligonucleotides complementary to each probe were slotted below the PCR products as controls. Hybridizations and washes were carried out as described by Mullis et al. (27). Table II. The resistant sublines were 24-fold more resistant to PTX and collaterally sensitive to vinblastine. They also exhibited slight cross-resistance (1.4 -3-fold) to the PTX analog, 2-debenzoyl-2-meta-azidobenzoylpaclitaxel and to EPO B, a natural product with a mechanism of action similar to that of PTX but with a different chemical structure (28).

Resistance Profile of 1A9 and PTX-resistant Cells-The drug sensitivity of parental 1A9 and the PTX-resistant cells is shown in
Effect of PTX on Tubulin Polymerization-Initial experiments showed that the two PTX-resistant sublines do not express MDR-1, that their intracellular PTX accumulation is comparable with that of the parental cell line, and that the total tubulin levels are similar to the levels observed in the parental cell line (data not shown). These findings led us to examine the effects of PTX on the relative levels of polymerized and soluble tubulin in parental and resistant cells. Cells were lysed for 5 min at 37°C with a hypotonic buffer without or with AAGAGGAGGCCTAGAGCCTT a Residues with identity between ␤2 and M40 ␤-tubulin isotype, are designated by a hyphen. b UTR, untranslated region. The coding sequence for the reverse primer for the ␤2 isotype ends at nucleotide 1335, and for the M40 isotype it ends at nucleotide 1332. 0.04 -4 g/ml PTX, and the levels of polymerized and soluble tubulin were assessed (Fig. 1A). To maximize the sensitivity for detecting polymerization, we chose experimental conditions that gave a minimal basal level of polymerization. In the absence of PTX, the majority of tubulin was found in the soluble form (10), and the ratios of polymerized (P) and soluble (S) tubulin were similar for all three cell lines. With increasing doses of PTX (0.04 -4 g/ml), tubulin polymerization increased in the parental 1A9 cells but not in the two resistant sublines.
Stability of PTX Resistance Phenotype-To examine the stability of the phenotype, the resistant clones were cultured in the absence of PTX and verapamil in an attempt to obtain drug-sensitive revertants. However, after 3 (PTX10) and 2 (PTX22) years of growth in drug-free media, reversal of the resistant phenotype had not occurred, as assessed by the polymerization analysis shown in Fig. 1B. Increasing PTX concentrations (0.04 -4 g/ml) resulted in a dose-dependent increase of tubulin polymerization in the parental, but not in the 1A9PTX10 and 1A9PTX22 cells that had been maintained drug-free for 3 and 2 years, respectively.
Mixing Experiment-To examine whether a soluble factor needed for tubulin polymerization was missing in the resistant sublines, we performed a mixing experiment. Whole cell lysate from parental 1A9 cells was added to lysate from resistant cells at varying ratios ranging from 1:1 to 1:19 (parental:resistant), to examine if parental cell lysate could induce polymerization in the resistant clone lysate. The data presented are from duplicate samples (Fig. 2). The upper panel shows the results when harvesting, mixing of lysates, and incubation were performed in the absence of PTX. Only a very small fraction of tubulin was polymerized (P lanes). The lower panel shows the results in the presence of PTX. As was demonstrated in Fig. 1A, in parental 1A9 cells, most of the tubulin is present in the polymerized form, in contrast to 1A9PTX10 cells, where the majority of tubulin is in the soluble form. Furthermore, the addition of lysate from parental cells did not induce polymerization of tubulin from the PTX-resistant clone. In each combination, the amount of polymerized tubulin in the P lanes reflects the contribution only from parental cells, as evidenced by the reduction in the amount of polymerized tubulin observed as the fraction of parental cell lysate decreases. Thus, there is no induction of polymerization or recruitment of tubulin from the resistant clone, even in the presence of some polymerized tubulin. Similar results were also found for the clone 1A9PTX22 (data not shown).
In Vitro Tubulin Polymerization with PTX and Epothilone B-To corroborate and extend the biochemical evidence presented thus far, tubulin was purified from parental and resistant cells to confirm that the differences were secondary to specific alterations in the tubulin molecule. As shown in Fig.  3A, 1 M PTX stimulated polymerization of parental cell tubulin to about the same extent as rat brain tubulin. In contrast, tubulin from the resistant cells exhibited impaired PTX-driven polymerization. Tubulin from clone 1A9PTX10 polymerized to a level less than half of parental, and tubulin from clone 1A9PTX22 failed to polymerize at all. Furthermore, EPO B polymerized tubulin from the two resistant clones to the same extent as parental cell tubulin (Fig. 3B), consistent with the cytotoxicity results (Table II), demonstrating low to absent cross-resistance to EPO B. Fig. 3C demonstrates that the turbidity increase shown in Fig. 3A results from normal microtubule structures that exhibit typical morphology and dimensions. This is illustrated by electron microscopy of a negatively stained preparation of parental cell tubulin following the addition of PTX. It is important to emphasize that the purified tubulins were unimpaired in their intrinsic ability to polymerize, since the final step in purification isolated protein that polymerized at 37°C and depolymerized on ice.
Expression of ␤-Tubulin Isotypes in Parental and Resistant Cells by Quantitative PCR-Our studies suggested alterations in tubulin as the cause for the defect in PTX-driven tubulin FIG. 1. Effect of PTX on tubulin polymerization in 1A9 and PTX-resistant cells. A, drug-sensitive parental 1A9 and the resistant clones PTX10 and PTX22 were lysed with a hypotonic buffer without (0) or with 0.04 -4 g/ml PTX, for 5 min at 37°C. Following cell lysis the polymerized (P) and the soluble (S) protein fractions were separated by centrifugation, and each fraction was resolved on adjacent lanes by electrophoresis. After transfer, the filter was probed with an antibody against ␣-tubulin. The resistant cells were maintained in drug-free medium for 1 week prior to the experiments. B, the same experiment was performed with the two resistant clones that had been maintained in drug-free medium for 3 (PTX10(Ϫ)) and 2 years (PTX22(Ϫ)), respectively. The percentage of polymerized tubulin (%P) was determined by dividing the densitometric value of polymerized tubulin by the total tubulin content (the sum of P plus S). polymerization. These alterations could be secondary to differences in the coding sequence or post-translational modifications.
The expression of five ␤-tubulin isotypes was examined by quantitative PCR (Fig. 4A). These isotypes were M40 (class I), ␤2 (class IVb), ␤9 (class II), ␤4 (class III), and 5␤ (class IVa). PCR amplification was performed for 30 cycles, with different dilutions used for each isotype to ensure that amplification was in the linear range. Tubulin isotype levels were normalized to the input RNA for each cell line, and their expression relative to that of parental cells was determined (Fig. 4B). Similar results were obtained when the isotype levels were normalized to ␤2-microglobulin (data not shown). A 1.4 -1.8-fold increase of M40 was observed in the two resistant clones, while a 20 -30fold decrease of 5␤ was observed for clones 1A9PTX10 and 1A9PTX22, respectively. The very low levels of 5␤ isotype precluded accurate densitometry, and argue against a significant contribution of this isotype in the tubulin behavior observed in these cells. Fig. 4, C and D, presents the relative expression of the five isotypes in parental 1A9 cells. Panel C presents the results of PCR amplification for 30 cycles, using different amounts of RNA as indicated for each isotype. The M40 isotype was detected with the lowest amount of RNA, indicating it is the most prevalent isotype. The ␤2 isotype was detected with the second lowest amount of RNA, indicating that its expression is lower than M40 but higher than the other three isotypes. By quantitating the products by densitometry and correcting for product size, the relative levels of expression of the five isotypes were calculated. This is an approximation, since the primers utilized could amplify with different efficiencies. The results are presented in Fig. 4D, where the relative percentage of each isotype in 1A9 cells is shown. Assuming comparable efficiencies of translation and protein stability, the percentage of each isotype in parental cells is as follows: M40, 84%; ␤2, 11%; ␤9, 4.2%; ␤4, 0.5%; and 5␤, 0.3%. Although these calculations are approximations, they suggest that the predominant isotypes are M40 and ␤2.
Sequencing of ␤2 and M40 ␤-Tubulin Isotypes-Since the PTX phenotype could not be explained by changes in the relative levels of ␤-tubulin isotypes, sequencing of the two predom-inant isotypes, M40 and ␤2, was performed. We reasoned that a mutation in a predominant isotype would be more likely to confer resistance. The cDNA sequence of the ␤2 isotype was "wild type" in all cell lines, both by automated sequencing and by subcloning and manual sequencing of three or four individual colonies for each cell line and each set of primers. In contrast, two distinct point mutations were identified in the cDNA of the M40 isotype, one in each of the two PTX-resistant sublines (Fig. 5). In clone 1A9PTX10, a T 3 G substitution in nucleotide 810 changed amino acid 270 from phenylalanine (TTT) to valine (GTT). In clone 1A9PTX22, a G 3 A substitution in nucleotide 1092 changed amino acid 364 from alanine (GCA) to threonine (ACA). Both substitutions were each present as a single peak in the sequence analysis (not shown), indicating expression of mRNA with wild type sequence could not be detected. Oligonucleotide Hybridization-Oligonucleotide hybridization using two different probes spanning the regions containing the mutations was performed (Fig. 5). To examine both RNA and DNA, PCR products from both RNA (converted to cDNA) and DNA were obtained by PCR using primers flanking the mutations. The products were slotted and hybridized under identical conditions. As shown in Fig. 5, clone 1A9PTX10 expresses only the RNA with the acquired ␤F270V mutation, although it is heterozygous at the DNA level. A similar result for the ␤A364T mutation was observed for the 1A9PTX22 clone. These results confirm the presence of tubulin mutations in the PTX-resistant human ovarian carcinoma cell lines. DISCUSSION Drug resistance can originate through multiple mechanisms, such as changes in cellular drug uptake, metabolic drug deactivation, structural changes in the drug target, or changes in other cellular components that interact with the target (29). Some of these have been implicated in the emergence of resist-ance to paclitaxel. To identify mechanisms of PTX resistance other that those mediated by Pgp, we isolated PTX-resistant human ovarian carcinoma cell lines. Two resistant cell lines, derived from a subclone of the human ovarian carcinoma line, A2780, were characterized. These sublines were isolated in a single step by exposure to PTX in the presence of the Pgp antagonist, verapamil. Previous studies have shown that verapamil can negate the appearance of Pgp, resulting in the selection of cells that exhibit other mechanisms of drug resistance (30). The resistant cells are 24-fold more resistant to PTX, 1.4 -3-fold more cross-resistant to either 2-debenzoyl-2-metaazidobenzoylpaclitaxel or EPO B, and collaterally sensitive to vinblastine. Expression of MDR-1 was not detectable, nor was PTX accumulation decreased in the resistant cells (data not shown). Thus, these cells provide a valid model of non-Pgpmediated PTX resistance.
In contrast to other PTX selected cell lines that have been reported, the resistant cells described in this study are not FIG. 4. Expression of ␤-tubulin isotypes. A, the expression of ␤-tubulin isotypes was examined in 1A9 and resistant cells by quantitative PCR, following reverse transcription. Different amounts of RNA were utilized as indicated to ensure that samples were in the exponential range of amplification. B, bar graph representation of the results displayed in A. The level of each ␤-tubulin isotype in the 1A9 parental cells is arbitrarily given the value of 10, to compare the expression level in the two resistant sublines relative to that of the parental. These values were derived from the results in A, using 28 S as the standard. C, the expression of the five ␤-tubulin isotypes in 1A9 parental cells by quantitative PCR following reverse transcription. Different amounts of RNA were utilized, as indicated. D, bar graph representation of the results displayed in C. The total ␤-tubulin percentage of each of the five isotypes in parental 1A9 cells is presented. The percentage was calculated after quantitation of each PCR product by densitometry and correction for the product size, as described under "Results." PTX-dependent for growth (18,31). Their growth rate is similar in the absence and presence of PTX (data not shown). This indicates that the resistant phenotype protects MT function from PTX effects without altering MT function in the absence of PTX. Furthermore, the resistant phenotype remained stable in the absence of drug for 3 years.
Multiple changes in the PTX target, tubulin, have been reported in PTX-resistant cell lines. These include reduced total tubulin levels (10,24), increased tubulin content (15), differential migration of ␣or ␤-tubulin on two-dimensional polyacrylamide gel electrophoresis (11,13), acetylation of ␣-tubulin (18), and increased expression of specific tubulin isotypes (14 -17). In the present study, the total tubulin content of the resistant cells was similar to that of the parental cells (data not shown), and no differences in the acetylation of purified ␣-tubulin were observed (data not shown). Thus, we sought other explanation(s) for the origin of PTX resistance in these cells.
Since PTX promotes tubulin polymerization, we compared the relative ratios of soluble and polymerized tubulin in the resistant cells. In a first approximation, immunofluorescent microscopy revealed no difference in the amount of MTs between the 1A9 parental cells and the resistant cells (data not shown). These results were confirmed and extended using the tubulin polymerization assay. Although this method of evaluation was designed to measure the response to added PTX rather than steady-state polymerization, similar ratios of polymerized and soluble tubulin were obtained for untreated cells for each cell line. In contrast, in the presence of PTX, a dosedependent tubulin polymerization was observed only in 1A9 parental cells. In addition, mixing cell lysates from parental and resistant cells did not induce polymerization of tubulin in the resistant cell lysates, nor did parental cell-derived polymers recruit tubulin from the resistant cell lysates. Moreover, the degree of PTX-induced polymerization of tubulin purified from the resistant clones was less than that of the tubulin from the parental cell or from rat brain. These data suggested that the primary defect was one of tubulin structure or composition.
Human genes coding for ␣or ␤-tubulin constitute a multigene family of 15-20 members, several of which are pseudogenes (32,33). Although there is some tissue specificity of isotype expression, the role of isotypes remains unclear (34). It has been reported that isotype composition can regulate MT dynamics in vitro (35); however, transfection experiments have shown that exogenous isotypes are incorporated into microtubule structures and appear to function interchangeably (36). There is also evidence that in the absence of a particular isotype, other isotypes are substituted without affecting normal function (36).
The demonstration of differential expression of tubulin isotypes in cell lines resistant to MT active agents (14 -17) has prompted some authors to postulate this as a mechanism of PTX resistance. However, mutations in a tubulin isotype have not been excluded. Indeed, if an acquired mutation confers an advantage, prolonged selection might lead to its overexpression. This could manifest as altered isotype expression, although the selective advantage was conferred by the mutation and not by the native isotype.
In the present study, M40 isotype expression was increased about 1.5-fold; however, this data appeared inadequate to explain the 24-fold PTX resistance, although the isotype expression levels might contribute to the drug-resistant phenotype. The polymerization results using either cell lysates (Fig. 1) or purified tubulin (Fig. 3) indicate that parental tubulin and hence the mixture of isotypes present in parental tubulin polymerizes in response to added PTX. Therefore, it seemed unlikely that a 1.5-fold change or less in the level of any of these PTX-responsive isotypes would result in PTX resistance. Consistent with this, point mutations in ␤-tubulin were found in the class I isotype, M40, which represented the majority of total ␤-tubulin mRNA. These two mutations in independently selected cells were not identical. The M40 ␤270 mutation in clone 1A9PTX10, a T 3 G substitution in nucleotide 810, changed amino acid 270 from phenylalanine (TTT) to valine (GTT). The M40 ␤364 mutation in clone 1A9PTX22, a G 3 A substitution in nucleotide 1092, changed amino acid 364 from alanine (GCA) to threonine (ACA). The remainder of the M40 nucleotide sequence in both clones was consistently found to be identical to the parental sequence. These changes in ␤-tubulin most likely contribute to PTX resistance in these cells.
The oligonucleotide hybridization studies confirmed the expression of mutant ␤270 and ␤364 tubulins in the respective cell lines and demonstrated at the DNA level both wild type and mutant sequences. Interestingly, although an allele with wild type sequence was present in the DNA from both resistant sublines, only the mutant allele was expressed. This pattern of expression is under current investigation.
Two models of resistance mechanisms might explain the effects of these mutations. It is possible that the mutations alter the PTX binding site such that PTX binding is inhibited but MT function is unaffected in the absence of PTX. Alternatively, it is possible that the mutations could alter a region of tubulin important for dimer-dimer contacts in the MT polymer, weakening the interaction and producing less stable MT (37). The addition of PTX up to some threshold level would then produce more normal MT rather than the "hyperstabilized" MT produced with wild type tubulin.
We favor the first mechanism for several reasons. First, the resistant clones do not exhibit any MT defect in the absence of PTX, their growth rate is the same in the absence or presence of PTX, and they have doubling times similar to parental cells. It seems unlikely that the MT of these cells are hypostable. Second, hypostable MT should be made more normal by any MT-stabilizing drug, and these should confer similar relative FIG. 5. Oligonucleotide hybridization. Independent experiments were performed for each mutation identified in ␤-tubulin. For each one, PCR products were obtained using primers spanning the mutation. The template was either RNA (converted to cDNA) or DNA. Equal amounts of PCR product from the three cell lines were slotted side by side on a nitrocellulose membrane. The membrane was divided into two equal halves, and each half was hybridized with an oligonucleotide probe containing either wild type (wt) or mutant (mut) sequences. After washing, the halves were brought together for the autoradiogram. Below the PCR products, 50 ng of primers complementary to each probe were slotted as controls. For each of the two probes, the nucleotide substitutions between the wild type and mutant sequences are shown in parentheses.
resistance, but that was not observed. Rather, small structural changes in PTX altered the extent of resistance. For example, relative resistance to taxotere is less than half that to PTX (data not shown). The relative resistance to 2-debenzoyl-2meta-azidobenzoylpaclitaxel, is even less. Moreover, the relative resistance to EPO-B, a MT-stabilizing agent structurally distinct from PTX that shares at least part of the same binding site, is less than 2-fold. Finally, the effects observed on cell growth paralleled an in vitro assay of drug-induced polymerization, using purified tubulin. PTX-induced polymerization of resistant tubulin was significantly decreased, but polymerization induced by EPO-B (Fig. 3) or 2-debenzoyl-2-meta-azidobenzoylpaclitaxel (38) was comparable with that obtained with parental or rat brain tubulin. Taken together, these results favor a mechanism based on weakening the interaction with the drug rather than weakening tubulin-tubulin interactions important to polymerization in the absence of drug. Fig. 6 compares the sites of the mutations with other known drug interaction sites and maps the residues or peptides within ␤-tubulin shown to interact with PTX, colchicine, and vinblastine/rhizoxin. Vinblastine and rhizoxin share a binding site that is distinct from that for colchicine or PTX (39,40). These sites/peptides were identified either by photoaffinity labeling or by induced structural changes. The sites identified by the latter, while not necessarily proximal to the binding site, are clearly affected by drug binding. Additionally, Fig. 6 shows the location of the two mutations identified in our study and the location of a loop exposed on the surface of the dimer. These features cluster in three regions of the sequence, indicated by brackets at the bottom of Fig. 6: the N-terminal 30 or 40 residues, a central region centered around residue 250, and a more C-terminal region containing residues 350 -390. It is notable that these same regions comprise the sites of binding and action of agents that stabilize MT and enhance assembly, such as PTX, as well as of agents that inhibit assembly such as colchicine, vinblastine, and rhizoxin. These regions will be considered in turn.
The N-terminal region can be photolabeled by PTX and by colchicine. PTX, with the photoreactive group on the C-13 side chain, labels a peptide containing residues 1-31 (41). Photolabeling with colchicine also identifies an N-terminal peptide of 1-36 (42). Additionally, this peptide contains Cys 12 , which is labeled by GTP (43). This residue, and the other cysteine res-idues indicated, will be discussed below.
The central region around residue 250 contains peptides labeled by all three groups of drugs, a loop sequence affected by drug binding, and the site of mutation in PTX10. PTX, with the photoreactive group on the C-2 side chain, labels a peptide containing residues 217-231 (44). Photolabeling with colchicine again labels an overlapping peptide, containing residues 214 -241 (42). Vinblastine binding labels a site contained in residues 175-213 (45). This region also contains a surfaceexposed, drug-sensitive loop sequence that contains the sole site of chymotrypsin cleavage in the native tubulin dimer, located after Tyr 281 (46,47). Exposure of this loop and consequent chymotryptic cleavage is enhanced by binding of colchicine, rhizoxin, or especially vinblastine (48) but is nearly completely suppressed by PTX (46). The finding that the PTX10 mutation locates to residue 270 in the center of this region and just N-terminal proximal to the loop sequence reinforces the importance of this region of the sequence to the binding of polymerization-modulating drugs indicated by previous data. It is notable that the PTX analog that most strongly reversed the resistant phenotype, 2-debenzoyl-2-meta-azidobenzoylpaclitaxel, is identical to the compound used to label this region (44).
The third, most C-terminal region, contains sites labeled by rhizoxin and colchicine and an amphipathic helical region strongly affected by drug binding. Photolabeling with a colchicine analog demonstrates that the A-ring of colchicine interacts with Cys 354 (49). Rhizoxin photolabels a peptide containing residues 363-379 (50). This peptide overlaps a drug bindingsensitive amphipathic helical region containing residues ϳ370 -390, whose carboxyl-terminal end is unfolded and made chymotrypsin-sensitive by colchicine (51). Unfolding is antagonized by vinblastine or rhizoxin, which promotes a form more tightly folded and protease-resistant at this site than the drugfree dimer (48). The location of the PTX22 mutation at residue 364 indicates that the PTX binding site contains contributions from this region of the sequence as well as the other two regions, as does the colchicine site.
In order for sequences from these three regions, very distant in linear sequence from each other, to contribute to the PTX binding site on ␤-tubulin, they clearly must be in close proximity in the folded form of native tubulin. Several lines of evidence, independent of data regarding PTX binding, allow us to draw this conclusion. In addition to the clustering of data from multiple probes, the data from direct photolabeling with a single probe, colchicine, require that these sequences be close. Since colchicine labels peptides from all three regions, the contact sites within these peptides cannot be farther apart than the dimensions of the colchicine molecule, ϳ11 Å (42,49). Additional compelling evidence comes from studies of cysteine cross-linking (52,53). Using a bifunctional cross-linker with maximum span of ϳ9 Å, two cross-links could be formed in ␤-tubulin. One cross-link is between Cys 12 (in the N-terminal region of Fig. 6) and Cys 201 or Cys 211 (in the central region of Fig. 6). The other is between Cys 239 (in the central region) and Cys 354 (in the C-terminal region). These cross-links are sensitive to drug binding; colchicine binding inhibits the Cys 239 -Cys 354 cross-link, whereas vinblastine enhances it while inhibiting the Cys 12 -Cys 201 /Cys 211 cross-link. Thus, data independent of PTX binding indicate that these three regions of the ␤-tubulin sequence are close in the native protein and that the interactions between them are altered in subtle but important ways by the binding of MT-active drugs, including both assembly inhibiting and assembly-promoting agents. The identification in the present work of residues ␤270 and ␤364 as important modulators of the interaction of PTX with tubulin add to our understanding of this important binding site.
In summary, the present study describes the isolation and characterization of two PTX-resistant human ovarian carcinoma cell lines. This study is the first to identify specific point mutations in tubulin that result in PTX resistance in human cells. Our data have identified regions of ␤-tubulin whose alteration abrogates PTX action in vitro and confers cellular PTX resistance.