Expression of Oncogenic Epidermal Growth Factor Receptor Family Kinases Induces Paclitaxel Resistance and Alters β-Tubulin Isotype Expression*

Oncogenic transformation confers resistance to chemotherapy through a variety of mechanisms, including suppression of apoptosis, increased drug metabolism, and modification of target proteins. Oncogenic epidermal growth factor receptor family members, including EGFRvIII and HER2, are expressed in a broad spectrum of human malignancies. Cell lines transfected with EGFRvIII andHER2 are more resistant to paclitaxel-mediated cytotoxicity, and tubulin polymerization induced by paclitaxel is suppressed compared with cells expressing wild type epidermal growth factor receptor. Because differential expression of β-tubulin isotypes has been proposed to modulate paclitaxel resistance, we analyzed β-tubulin isotypes expressed in cell lines transfected with different oncogenes. EGFRvIII- and HER2-expressing cells demonstrated equivalent total β-tubulin protein compared with cells transfected with wild type receptor or untransfected controls. EGFRvIII-expressing cells demonstrated increases in class IVa (2.5-fold) and IVb (3.1-fold) mRNA, and HER2-expressing cells showed increases in class IVa (2.95-fold) mRNA. Expression of oncogenic Ha-Ras did not change class IV RNA levels significantly. Inhibition of EGFRvIII kinase activity using a mutant allele with an inactivating mutation in the kinase domain decreased expression of class IVa by 50% and partially reversed resistance to paclitaxel. Expression of oncogenic epidermal growth factor receptor family members is associated with modulation of both β-tubulin isotype expression and paclitaxel resistance in cells transformed by expression of the receptor. This effect on tubulin expression may modulate drug resistance in human malignancies that express these oncogenes.

oncology, stabilize microtubules by binding to the ␤-tubulin component of ␣/␤-tubulin heterodimers, blocking cells in G 2 /M leading to cell death (1,2). Modification of ␤-tubulin by mutation or differential expression of isotypes have been proposed as mechanisms for either de novo or acquired taxane resistance (3,4). At least six ␤-tubulin isotypes have been identified, constituting five evolutionarily conserved isotype classes. These differ primarily in the carboxyl-terminal 15 amino acids (5). The conservation of distinct isotype classes has suggested that the different ␤-tubulins contribute unique functional properties to cell structure and function. Recent evidence suggests that these isotype classes vary in their paclitaxel binding affinities. Microtubules selectively enriched for isotype class III and IV ␤-tubulin are significantly more resistant to paclitaxel suppression of microtubule dynamics than are microtubules composed of unfractionated tubulin (6). Tumor cell lines rendered paclitaxel-resistant by exposure to high concentrations of drug also demonstrate substantial increases in class III and IV ␤-tubulin (7)(8)(9). Prostate carcinoma lines resistant to the tubulin-active agent estramustine demonstrate selective increases in class III and IV tubulin isotypes (10). Clearly, one response of tumor cells to in vitro selection with microtubule-stabilizing agents is to increase the proportion of taxane-resistant tubulin. Evidence for modulation of isotype expression by paclitaxel exposure is also found in patients treated with taxanes. Paired samples from patients with advanced ovarian cancer that developed clinical paclitaxel resistance showed increases in ␤-tubulin isotype classes I (3.6-fold), III (4.4-fold), and IVa (7.6-fold) (8).
The epidermal growth factor receptor (EGFR 1 or erbB) family of receptor tyrosine kinases plays a role in oncogenic transformation in tissues where it is overexpressed or coexpressed with one of its ligands (11,12). Transformation occurs by mutation, overexpression, and synergistic activation of different family members. Gene amplification and overexpression of intact or wild type (wt)EGFR have been found in a variety of epithelial and neuronal neoplasms including squamous cell carcinomas, melanoma, and glioblastoma multiforme and portend a poor prognosis for patients with these tumors (13)(14)(15)(16). Overexpression of EGFR has been associated with a relative resistance to chemotherapy in cell lines and in patients with breast, ovarian, and esophageal cancer (17)(18)(19)(20)(21). HER2 is a closely related member of the EGFR family which is oncogenic when overexpressed in several cell contexts as a result of transactivation of HER2 complexes (22). HER2 plays a role in oncogenesis in breast, ovarian, gastric, and lung carcinoma (23). A mutant EGFR with constitutive kinase activity has been identified in a variety of human tumors (24). This mutant, referred to as EGFRvIII (also EGFRm, ⌬ EGFR, de2-7 EGFR), activates signaling through multiple pathways and acts as a dominant oncogene when overexpressed (25)(26)(27)(28)(29). EGFRvIII also reduces spontaneous apoptosis when transfected into glioblastoma cell lines, suggesting that it may play a significant role in protecting tumor cells from drug-induced cytotoxicity (30). Activation of mutant EGFRvIII is ligand-independent and recapitulates the chronic autocrine or paracrine stimulation of EGFR in tissues that overexpress EGFR. Because of its broad prevalence and likely clinical significance in human tumors, we wished to determine the effects of EGFRvIII and the related oncoprotein HER2 on chemotherapy-induced cytotoxicity and potential mechanisms of resistance to these agents. This report demonstrates that expression of constitutively active, oncogenic EGFR and HER2 induce paclitaxel resistance and increase class IVa and IVb ␤-tubulin RNA. This modulation of ␤-tubulin isotype expression may demonstrate one mechanism by which oncogenic transformation mediates intrinsic resistance to paclitaxel in tumors that overexpress these oncogenes.

EXPERIMENTAL PROCEDURES
Materials-Antibodies to EGFRvIII were kindly provided by Albert Wong, Kimmel Cancer Center, and were prepared as described (31). Enhanced chemiluminescence (ECL) reagents and pan-␤-tubulin antibody (N357) were from Amersham Pharmacia Biotech. Antibodies to p21 were from Santa Cruz Biotechnology. Antibody to HER2 was from Transduction Laboratories (E19420). Precast SDS-PAGE gels were from NOVEX. GeneAmp PCR amplification kits were from Perkin-Elmer, and Taq polymerase was from Promega. Paclitaxel and ␣-tubulin antibody were from Sigma.
Paclitaxel Cytotoxicity Assay-2 ϫ 10 5 cells were plated on day 0, grown for 1 day, and exposed to 0 -100 nM paclitaxel for 4 days. Cells were trypsinized and assayed for viability by trypan blue staining. Triplicate flasks were counted for each concentration.
Tubulin Polymerization Assay-Quantitation of tubulin polymerization was carried out essentially as described previously with minor modifications (4). Cells were grown to confluence in 25-cm 2 flasks, and monolayers were washed twice with phosphate-buffered saline and lysed at 37°C in the dark with 300 l of 20 mM Tris-HCl (pH 6.8), 1 mM MgCl 2 , 2 mM EGTA, 0.5% Nonidet P-40, 1 mM phenylmethylsulfonyl fluoride, 10 g/ml aprotinin, leupeptin, and soybean trypsin inhibitor, 5 mM aminocaproic acid, 1 mM benzamidine, and paclitaxel (0 -1 g/ml). Lysates were then vortexed and centrifuged at 15,000 rpm for 10 min. Pellets were resuspended in 300 l of lysis buffer. The supernatants and pellets were mixed with equal amounts of 2ϫ SDS sample buffer and boiled for 5 min. 30 l of lysate was then separated on 7.5% SDS-PAGE, transferred to nitrocellulose filters, and probed with anti-␣-tubulin antibodies.
␤-Tubulin Isotype Analysis by RT-PCR-The gene-specific oligonucleotide primers for isotype classes I-IVb, ␤ 2 -microglobulin, and glyceraldehyde-3-phosphate dehydrogenase were exactly as published by Haber et al. (34) (Table I). The primers for 18 S rRNA were 5Ј-AT-GCTCTTAGCTGAGTGTCC-3Ј and 5Ј-AACTACGACGGTATCTGATC-3Ј. The primer sequences and expected product sizes are shown in Table  II. Total RNA isolated from transfected NIH3T3 cells was treated with RNase-free DNase to remove contaminating genomic DNA (35). Reverse transcription of 2 g of total RNA was performed to synthesize cDNA using specific primer pairs. PCRs were carried out with 50 ng of cDNA, and products were amplified for 35 cycles in a programmable thermocycler. Effective removal of genomic DNA was verified by amplification of intron-spanning glyceraldehyde-3-phosphate dehydrogenase primers. Isotypes classes I and II were coamplified with ␤ 2 -micro-globulin to allow normalization of products. Isotypes III, IVa, and IVb could not be coamplified with ␤ 2 -microglobulin and instead were quantified by comparing intensity of PCR products generated in the linear range of amplification to ␤ 2 -microglobulin and 18 S RNA (from parallel samples). PCR products were electrophoresed on 6% TBE gels and visualized by ethidium bromide staining. Densitometric measurements were performed on negatives of the gels. The band of interest was divided by the control band (18 S RNA, ␤ 2 -microglobulin). Data for the figures were expressed as the signal intensity ratios compared with ␤ 2 -microglobulin. Because isotype III levels were consistently very low, the results were not plotted. The ratios of 18 S and ␤ 2 -microglobulin products with tubulin isotypes were within 15% of each other for the different cell lines. The assays were performed in triplicate and standard deviations determined.
Western Analysis-Cells were lysed in 50 mM HEPES (pH 7.4), 1% Triton X-100, 10% glycerol, 1 mM sodium orthovanadate, 10 g/ml leupeptin/aprotinin/soybean trypsin inhibitor, normalized for protein concentrations, and resolved by SDS-PAGE on 8 -16% gels. Proteins were transferred to nitrocellulose and incubated with 5% nonfat dry milk for 1 h followed by immunoblotting antibody in the same solution for 4 h. Filters were extensively washed in TBS, 0.05% Tween 20, and bound antibodies were detected by addition of conjugated protein Ahorseradish peroxidase at 1:5000 for 1 h followed by ECL following the manufacturer's directions.
Experiments were performed at least 3 times and representative data are shown.

Oncogenic EGFR Expression Induces Resistance to Paclitaxel
Cytotoxicity-HER2 and EGFRvIII undergo ligand-independent autophosphorylation, mediating oncogenic signaling and transformation. Growth factor receptor stimulation may circumvent drug-induced apoptosis and cytotoxicity. As a model system that mimics these effects, we determined the effect of EGFR expression on sensitivity to paclitaxel in transfected fibroblast cell lines. Cell lines were exposed to 0 -100 nM paclitaxel, and viability was determined after 4 days of exposure. Table I demonstrates the effect of the different EGFR constructs on sensitivity to paclitaxel. Cells expressing the EGFR mutant (HC2) and HER2 were 2-3-fold more resistant to paclitaxel compared with the parental cell line. This level of resistance is modest compared with that induced by selection in tubulin-active agents (8,10,34), but it correlates well with other parameters of oncogenicity associated with expression of EGFRvIII, such growth in serum-free medium (27,36).
Paclitaxel-induced Tubulin Polymerization Is Reduced in Cells Expressing EGFRvIII and HER2-The major mechanism of paclitaxel cytotoxicity is stabilization of microtubule arrays that blocks effective cell division and growth. Because both EGFRvIII and HER2 induce resistance to paclitaxel, we measured paclitaxel-induced polymerization activity using a cellfree assay system. This assay separates measurement of polymerization from cell-mediated functions such as membrane transport (4). In the absence of paclitaxel, tubulin is relatively evenly distributed between soluble and insoluble forms in extracts derived from all cell lines (Fig. 1). With the addition of TABLE I Paclitaxel cytotoxicity for transfected NIH3T3 cells NIH3T3, CO12 (wt EGFR-expressing 3T3), HER2, and HC2 (EGFR-vIII-t-expressing 3T3) cells were grown for 1 day and exposed to paclitaxel (0 -100 nM) for 4 days. Cell viability was quantitated by trypan blue staining of cells. Data were expressed as IC 50 , the drug concentration that inhibits cell number by 50% after 4 days. Average cell counts at 4 days for controls were NIH3T3, 2.3 ϫ 10 6 , CO12-2.51 ϫ 10 6 , HER2, 1.73 ϫ 10 6 , HC2, 2.94 ϫ 10 6 . paclitaxel at 0.2-0.4 g/ml, the amount of soluble tubulin decreases substantially with a proportionate increase in insoluble tubulin in extracts from cells expressing wtEGFR. In contrast, there was no apparent increase in tubulin polymerization in cells expressing EGFRvIII or HER2, demonstrating diminished sensitivity to paclitaxel microtubule stabilization. ␤-Tubulin Isotype Expression Is Modulated by Expression of Oncogenes-The observed resistance to paclitaxel-induced polymerization in cells expressing EGFRvIII could be due to changes in total tubulin levels, to mutation in the tubulin molecule itself, or to changes in the relative proportions of paclitaxel-sensitive ␤-tubulin isotypes (3,7,37). To distinguish among these possibilities, total ␤-tubulin protein was analyzed from NIH3T3 cell lines expressing wtEGFR, EGFRvIII, Ras, and HER2. Total ␤-tubulin levels differed no more than 15% among the different cell lines (Fig. 2). In order to estimate the quantitative contribution of distinct ␤-tubulin isotypes to the total protein levels, we performed quantitative PCR assays. Relative RNA levels for class I-IVb were determined using RT-PCR with primers previously described by Haber et al. (34). Tubulin isotype RNA levels were normalized to PCR products for ␤ 2 -microglobulin and 18 S RNA in each gel. Relative ratios of amplified products to each of these standards were similar between the different cell lines. The parental NIH3T3-and wtEGFR-expressing cells showed similar amounts of ␤-tubulin I-IVb with very low levels of class III (Fig. 3, data for type III not shown). EGFRvIII-expressing cells showed a 2.5-fold increase in class IVa and a 3.1-fold increase in class IVb compared with cells expressing wtEGFR. HER2 expression resulted in a 2.95-fold increase in IVa and 2.1-fold increase in IVb. Ras expression induced insignificant changes compared with the parental line.
Expression of EGFRvIII and HER2 Does Not Affect p21 Cip1 Expression-Yu et al. (38) have recently shown that taxane resistance of HER2-expressing cells correlates with increased expression of p21 Cip1 . Suppression of p21 production blocked resistance to taxane, suggesting that this may be a general mechanism of taxane resistance in HER2-expressing cells. To test this hypothesis in our experimental system, we determined levels of p21 by immunostaining of NIH3T3 cell lysates expressing EGFR family members and Ras proteins. As shown in Fig. 4 and in contrast to the results of Yu et al. (38), expression of oncogenic EGFR family members did not alter p21 Cip1 expression in this specific cell type.
Inhibition of EGFRvIII Kinase Activity Suppresses Class IV Tubulin Production-EGFR signaling is not mediated exclusively by tyrosine kinase activity since various downstream kinases may be activated despite functional inactivation of the EGFR kinase domain (39). To determine whether kinase activity is important to induction of taxane resistance, HC2 cells, expressing EGFRvIII, were transfected with a HER2/neu construct, T691stop, which contains an inactivating mutation in the kinase domain (32,33). This mutant HER2 construct inhibits EGFRvIII signaling and kinase activity as a result of dimerization of T691stop with EGFRvIII (33). The resultant doubly transfected cells, designated H(T691), were analyzed for expression of EGFRvIII, the 115-kDa T691 protein construct, and for effects on EGFRvIII autophosphorylation (Fig. 5A). H(T691) cells maintained their expression of EGFRvIII and demonstrated significant expression of the T691 protein; however, autophosphorylation of EGFRvIII was suppressed, reflecting decreased kinase activity. H(T691) cells were then tested for paclitaxel sensitivity in the same fashion as was used to generate results shown in Table I. In these assays the ED 50 for HC2 was 40 nM, whereas the ED 50 for H(T691) was 20 nM. H(T691) cells demonstrate decreased resistance to paclitaxel and decreased expression of the paclitaxel-resistant ␤-tubulin class IVa (Fig. 5B), although levels of IVb were not affected (data not shown). This suggests that induction of tubulin isotype IVa is a direct result of kinase activity, and induction of isotype IVb is not kinase-dependent. The constitutive kinase activity of EGFRvIII is responsible for induction of paclitaxel resistance and increase in class IV tubulin expression.  1. Paclitaxel effect on tubulin polymerization. CO12, HER2, and HC2 were grown to confluence and lysed in 400 l in the presence and absence of paclitaxel (0 -0.4 g/ml) at room temperature in the dark. Lysates were then vortexed and centrifuged at 15,000 rpm in a minifuge for 10 min. Pellets were resuspended in 300 l of lysis buffer. The supernatants (SN) and pellets were mixed with equal amounts of sample buffer and boiled, and 30 l of lysate was separated on 7.5% SDS-PAGE, transferred to nitrocellulose, and probed with anti-␣-tubulin antibodies.

DISCUSSION
Resistance to antineoplastic agents can be mediated by many different mechanisms, including differential uptake and retention of drug controlled by membrane transporter proteins (1). The expanding use of taxanes in clinical practice has made the ability to predict primary or secondary resistance to these agents a potentially important aspect to treatment planning. Although de novo resistance is relatively rare, some malignan-cies will exhibit intrinsic resistance to taxanes, and the majority of patients with advanced disease will ultimately develop progressive disease despite continued therapy. Overexpression of EGFR, EGFRvIII, and HER2 is found in ovarian and breast carcinomas, tumors that are perhaps the best defined indications for taxane therapy (1,24). EGFR and HER2 overexpression is inversely correlated with response to chemotherapy and prognosis, and resistance is often accompanied by an increase in receptor expression (24, 40 -42). Poor response rates could result from intrinsic drug resistance mediated by these oncogenes.
We analyzed the effect of EGFRvIII and HER2 expression on cell sensitivity to paclitaxel, since these oncogenes induce constitutive signals for proliferation and transformation. Both receptors confer relative resistance to paclitaxel and confer resistance to tubulin polymerization. RT-PCR assays of RNA levels suggest that expression of EGFRvIII increases levels of class IVa and IVb 3-fold. The changes in isotype class IV are consistent with studies previously demonstrating an increase in type III or IV tubulin expression associated with paclitaxel resistance (7-9). Since class III and IV tubulins are relatively resistant to paclitaxel-induced stabilization of microtubules, RNA from subconfluent NIH3T3 cell lines expressing EGFRvIII (HC2), HER2, wtEGFR (CO12), or Ras was isolated and analyzed by RT-PCR for the presence of ␤-tubulin isotype-specific RNA using the primer sets described in Table I. PCR products were separated on agarose gels and visualized by ethidium bromide staining. Quantitation by comparison with coamplified ␤ 2 -microglobulin and 18 S RNA was used to generate ratios of PCR products from samples. Isotypes classes III and IV could not be coamplified with ␤ 2 -microglobulin, and separate reactions from samples were used for quantitation. A, ␤-tubulin isotype products. B, graphic representation of PCR data, normalized to ␤ 2 -microglobulin. , and HER2-expressing SKBR3 were analyzed by Western blot using EGFRvIII-specific antibodies, followed by stripping the blots and probing with anti-HER2 antibodies (to confirm expression of the kinasedead HER2/neu mutant) and antiphosphotyrosine antibodies. SKBR3 was included as a positive control for HER2 and negative control for EGFRvIII. B, RT-PCR quantitation of ␤-tubulin isotype IVa from cells expressing EGFRvIII (HC2) or coexpressing EGFRvIII and the kinasedead T691stop mutant (H(T691). C, bar graph representation of effects of the T691 mutant on class IVa expression. cells overexpressing these isotypes should have a selective growth advantage. The majority of published studies have demonstrated increases in IVa class rather than IVb, although there is no evidence that these isotypes differ in their sensitivity to paclitaxel (8 -10). Conditional overexpression of isotype IVb in Chinese hamster ovary cells by Blade et al. (43) did not induce paclitaxel resistance, suggesting that IVb levels do not mediate taxane resistance, at least in the cell system used in their study. In our studies, the failure to suppress isotype IVb following transfection with the T691 mutant also suggests that EGFRvIII kinase activity is not essential for tubulin isotype IVb induction. In contrast to the studies of Yu et al. (38), there was no evidence that p21 Cip1 expression played a role in paclitaxel resistance. The experimental conditions of these two studies were clearly different, involving cell lines that may have distinct protein expression and compensatory mechanisms available in response to oncogene activation, accounting for the disparate results (38).
Modulation of tubulin isotype expression by oncogenes is a novel mechanism for the induction of paclitaxel resistance. Why expression of class IV may be favored in cells expressing specific oncogenes remains to be explored. Oncogenic Ras does not influence class IV levels, suggesting a level of specificity for HER2 and EGFRvIII transformation. Transcriptional regulation of ␤-tubulin synthesis is separate from that of ␣-tubulin and is controlled by cellular tubulin levels, extracellular growth factors, and expression of transcription factors (44 -46). Studies directed at this question that rely on overexpression of specific tubulin isotypes are difficult to interpret because overexpression of isotype class IV tubulin causes compensatory decreases in synthesis of endogenous class IV, maintaining a constant cellular level (44). This is presumably related to a negative feedback control in which an amino-terminal tetrapeptide in ␤-tubulin transduces a signal that induces RNA degradation thus maintaining homeostatic levels (47)(48)(49). In Drosophila, binding sites for the transcription factors Tinman and Engrailed have been identified within isotype-specific ␤-tubulin sequences. Binding of these transcription factors is regulated in a stage-specific fashion, and expression proceeds in a tissue-specific, temporally regulated manner (45,50). In mammalian cells, exposure to cardiac stress induces transcription of specific ␤-tubulin classes within cardiac myocytes, and these tubulins appear to be intrinsically more stable, allowing hypertrophy and adaptation to stress (51). In rat brain ␤-tubulin isotype levels vary dramatically depending on the stage of neuronal development, again suggesting that expression of different isotypes has functional significance for differentiating cells (52). Few studies are available that address trends in isotype expression within specific tissues during tumor development. The stimulus for modulation of ␤-tubulin isotype expression could be related to the relatively less differentiated phenotype of tumor cells, an increased fraction of cells undergoing mitosis, or EGFR-specific induction of transcription factors that regulate ␤-tubulin stability.
Direct evidence of the contribution of isotypes III and IVa to the development of paclitaxel resistance has been hampered by the compensatory cellular mechanisms noted above, and the effects of class III and IVa overexpression have not been reported to date. More indirect evidence for the importance of isotype expression on paclitaxel resistance comes from the use of antisense oligonucleotides to inhibit production of specific isotypes. Antisense oligonucleotides specific to isotype class III ␤-tubulin are capable of blocking class III expression while reversing taxane resistance in drug-resistant lung carcinoma cells (53). Extending these studies through transient expression of EGFR family members combined with assay of isotype expression and taxane resistance will be required to confirm the general relevance of these findings. More detailed characterization of tumors from patients that overexpress oncogenic EGFR and correlation of isotype expression with clinical drug resistance will be important correlates to this work.