Microtubule-associated protein tau: A marker of paclitaxel sensitivity in breast cancer

Methods

Patients and Samples. Fine-needle aspirations of breast cancer were collected during a prospectively designed pharmacogenomic marker discovery program at the Nellie B. Connally Breast Center of University of Texas M. D. Anderson Cancer Center (10). The program includes marker discovery (n = 85 patients in single arm study) and validation (n = 220 patients in a randomized study) phases to develop a gene signature-based predictor of pCR to sequential paclitaxel and T/FAC preoperative chemotherapy for stage I-III breast cancer. The discovery phase of this program has been completed. Results on the single most predictive gene that emerged from the microarray analysis are reported here. Table 1 presents the clinical characteristics of the 82 informative patients included in the marker discovery analysis. Thirty-three of the 82 cases were also included in our previous report (8). All patients underwent breast surgery after completion of chemotherapy, and the extent of residual cancer was measured in the surgical specimen. Grossly visible residual tumor was identified, and representative sections were submitted for histopathologic study. In the absence of grossly visible residual cancer, all slices of the specimen were radiographed, and areas of radiologically and/or architecturally abnormal tissue were submitted for histopathologic study. This study was approved by the institutional review board of MDACC, and all patients signed an informed consent for voluntary participation.

For immunohistochemical validation on independent cases, a tissue microarray was used that was previously constructed from formaldehyde-fixed, paraffin-embedded tissues of pretreatment core needle biopsies left over after diagnosis. All of these patients with stage I-III breast cancer received preoperative T/FAC chemotherapy on a previous randomized clinical trial. One hundred twenty-two patients had sufficient pretreatment tissue available for tissue array analysis of tau expression. IHC and data analysis were conducted in accordance with a laboratory protocol approved by the institutional review board of the University of Texas M. D. Anderson Cancer Center.

Twelve human breast tumor cell lines (T47D, BT20, ZR75.1, MCF7, MDA-MB-231, MDA-MB-361, MDA-MB 435, MDA-453, MDA-468, BT 549, BT 474, and SKBR3) were obtained from the American Type Culture Collection. All culture media components were purchased from the M. D. Anderson Tissue Culture Core Facility (Houston).

Microarray Data Analysis. RNA was extracted from FNA samples by using the RNeasy Kit (Qiagen, Valencia, CA). The RNA yield and the cellular composition of these FNA samples were reported in ref. 11. cRNA was generated by using standard T7 amplification protocol without second-round amplification. Fragmented and biotin-labeled cRNA was hybridized to Affymetrix U133A gene chips overnight at 42°C as described in the Affymetrix technical manual.

dchip v1.3 software (http://dchip.org) was used to generate probe level intensities and quality measures, including median intensity, percentage of probe set outliers, and percentage of single probe outliers for each chip. Three chips failed the quality-control process, and subsequent analysis was performed on 82 samples. dchip software was used for normalization; this program normalizes all arrays to one standard array that represents a chip with median overall intensity. Normalized gene expression values were transformed to the log scale (base 10) for analysis. To identify differentially expressed genes between cases with pCR (n = 21) and those with residual disease (n = 61), genes were ranked by P values obtained with two-sample, unequal-variance t tests.

IHC. Two representative areas of each prechemotherapy core biopsy were selected for coring, and 5-μm sections (0.6-mm diameter) were placed in a tissue array slide. Antigen retrieval was performed with boiling in citrate buffer (pH 6.0) for 10 min in a microwave oven after deparaffinization. The slides were incubated with anti-tau monoclonal antibody that recognizes all isoforms of human tau protein, irrespective of phosphorylation status (1:50 dilution, clone T1029, United States Biological, Swampscott, MA) overnight at 4°C. Anti-mouse horseradish peroxidase-labeled secondary antibody (DAKO Envision TM+ System) and diaminobenzidine substrate were used to generate signal. Normal breast epithelium served as an internal positive control. Omission of the primary antibody served as a negative control. The specificity of the antibody was demonstrated with Western blot. Tau staining of tumor cells was scored as follows: IHC score 0, no staining; 1+, less staining than normal epithelium (Fig. 1b); 2+, similar to normal epithelium (Fig. 1c); 3+, uniform staining more intense than normal cells (Fig. 1d). Cases with 0 or 1+ staining intensity were considered tau negative, and tumors with 2+/3+ staining were considered tau positive. This dichotomization of staining results was determined by using staining intensity of normal epithelial cells as a reference and without knowledge of the clinical outcome. Slides were scored without knowledge of the clinical outcome. Correlation with complete response was assessed in univariate analysis (χ² test) and in multivariate analysis (logistic regression) including patient age, tumor size, histological type and grade, and estrogen receptor (ER), progesterone receptor, HER2, and tau status. ER and HER2 status was determined by the routine clinical pathology laboratory of University of Texas M. D. Anderson Cancer Center.

siRNA Experiments. Tau protein expression in breast cancer cell lines was assessed by Western blot using two different monoclonal anti-tau antibodies (clone T14 from Zymed and clone T1029 from United States Biological). Results were concordant with both antibodies. To transiently knock down tau expression, cells were transfected with a control siRNA (directed against lamin) or two distinct anti-tau siRNAs (5′-AATCACACCCAACGTGCAGAA-3′ and 5′-AACTGGCAGTTCTGGAGCAAA-3′) following the manufacturer's instructions (Qiagen).

Twenty-four hours after siRNA transfection, the medium was changed, and cells were treated with various concentrations of paclitaxel and epirubicin for a further 48 h. Proliferation rates were determined with the CellTiter-Glo Luminescent Cell Viability Assay (Promega). Chemosensitivity was determined from three separate experiments, each performed in triplicate.

The effect of tau suppression on cellular uptake of paclitaxel was examined by using fluorescent-conjugated paclitaxel (Oregon Green 488, Molecular Probes). Spontaneously fluorescent epirubicin was used to examine epirubicine uptake (12, 13). Forty-eight hours after siRNA transfection, 3 × 10⁵ cells were trypsinized and resuspended in 1 ml of medium containing 1 μM Oregon Green paclitaxel or 16 μM epirubicin and incubated for 20, 50, or 80 min at 37°C. Subsequently, cells were analyzed by FACS using cellquest software (BD Biosciences, Franklin Lakes, NJ). The amount of fluorescence per cell (arbitrary fluorescence units) was taken as the measure of drug uptake. Results were displayed as histograms of mean fluorescence and standard deviation of three independent experiments. The percentage of fluorescent cells and nonfluorescent cells was also determined.

Tubulin Polymerization Assays. Bovine brain tubulin polymerization assay was performed in 100-μl volumes at 37°C by using the Tubulin Polymerization Assay Kit (Cytoskeleton, Denver). Purified tau protein was purchased from Cytoskeleton (ref. no. TA01). Fluorescent BODIPY-paclitaxel (BODIPY 564/570) was purchased from Molecular Probes. OD₃₄₀ was measured every 30 s for 30 min. Results are presented as change in baseline-corrected turbidity over time. Fluorescence emission was read at excitation/emission wavelengths of 530/580 nm with a Cytofluor II fluorescence plate reader (PerSeptive Biosystems, Framingham, MA). Binding of paclitaxel to microtubules was also assessed with ³H-labeled paclitaxel (Moravek Biochemicals, Brea, CA).

Results

Gene Expression Analysis Reveals Tau mRNA as the Best Single Gene Discriminator of pCR to Preoperative Chemotherapy with Paclitaxel, 5-Fluorouracil, Doxorubicin, and Cyclophosphamide. We used gene expression profiling to discover genes that are associated with extreme chemotherapy sensitivity. The most significantly differentially expressed gene between cases with pCR and residual cancer was microtubule-associated protein tau. Tau mRNA expression measured by all four Affymetrix probe sets that were directed against this molecule was significantly lower (unequal-variance t test, P < 0.3 × 10^-5) in tumors that achieved pCR (Fig. 5, which is published as supporting information on the PNAS web site). All cases with pCR had low tau expression; however, some cases with residual cancer also had low tau expression. There was no differential expression of any other microtubule-associated proteins or tubulin subtypes in our gene expression data.

Low Tau Protein Expression Assessed by IHC Predicts Higher pCR Rate. We examined tau protein expression in 122 independent cases by using IHC. All patients received 24 weeks of preoperative paclitaxel and anthracycline-containing chemotherapy. None of these patients was included in the microarray study. Thirty-eight patients had pCR (31%). Cytoplasmic expression of tau protein was seen in all normal breast epithelium (2+ IHC score) and blood vessels but not in fibroblasts or adipocytes (Fig. 1). Normal epithelium provided a convenient internal control for scoring and represent cells that are resistant to therapeutic doses of paclitaxel. Sixty-four tumors (52%) showed less than normal expression (0/1+) and were considered tau negative. Fifty-eight tumors (48%) were tau positive (2+/3+). Forty-four percent of tau-negative tumors (28/64) had pCR, compared with 17% of tau-positive tumors (10/58) (P = 0.04). Seventy-four percent of all pCR cases were tau negative (n = 28/38; Fig. 1e). The odds ratio for pCR in tau-negative tumors was 3.7 (95% confidence interval, 1.6-8.6; P = 0.0013). Multiple logistic regression including age, tumor size, nodal status, histology, Black's modified nuclear grade, and ER, progesterone receptor, and HER2 receptor expression identified high nuclear grade (P = 0.05) and ER-negative status (P = 0.06) as independent factors associated with low tau expression. Similar multiple logistic regression including age, tumor size, nodal status, histology, Black's modified nuclear grade, and ER, progesterone receptor, and HER2 receptor expression as covariates identified nuclear grade 3 histology (P < 0.01), age <50 (P = 0.03), and tau-negative status (P = 0.04) as independent predictors of pCR (Table 2). Tau was more powerful than ER status as a single variable to identify patients who are most likely to achieve pCR.

These results confirmed the microarray data that low tau expression is associated with significantly higher rate of pCR compared with normal tau expression. However, only half of the tau protein-negative cases had pCR, which suggests that tau expression measured at the mRNA level may be a more powerful discriminator of response than tau protein expression assessed by IHC.

Down-Regulation of Tau Expression in Breast Cancer Cells Increases Sensitivity to Paclitaxel. We hypothesized that lower than normal tau expression may increase sensitivity to paclitaxel due to its effect on microtubules. We assessed tau protein expression in breast cancer cell lines with Western blot by using the same antibody that was used for IHC. Four cell lines (ZR75.1, T47D, MCF7, and MDA-MB 435) expressed tau, whereas eight other cell lines did not (Fig. 2a). ZR75.1 and MCF-7 cells were used for further in vitro studies. These cells are known to be relatively resistant to paclitaxel (14). T47D and MDA-MB 435 cells could not be successfully transfected with anti-tau siRNA.

siRNA transfection efficiency was estimated to be 30-40% (see below), and tau down-regulation lasted for at least 72 h after transfection (Fig. 2b). Decreased tau expression significantly increased the sensitivity of ZR75.1 cells to paclitaxel compared with control cells transfected with lamin siRNA or no siRNA (Fig. 2c). The IC₅₀ of paclitaxel was reduced from >10 μM to 20 nM in the tau knock-down cell pool. However, cell kill did not reach 100%, which is consistent with partial transfection efficacy and limited (48-h) exposure to paclitaxel. This finding also suggests that other mechanisms of resistance may exist in these cells. Tau down-regulation did not result in increased sensitivity to epirubicin (Fig. 2d). Identical results were obtained with MCF7 cells (data not shown). These data demonstrate that tau protein partially protects cells from the cytotoxic effects of paclitaxel. Suppression of tau expression renders cells more sensitive to paclitaxel, but not to epirubicin.

Microtubules Formed in the Presence of Tau Bind Less Paclitaxel and Show Decreased Paclitaxel-Induced Stabilization in Vitro. Pharmacological stabilization of microtubules by paclitaxel affects tau binding (15, 16). We hypothesized that the opposite may also occur; physiological stabilization of microtubules by tau may reduce paclitaxel binding to tubulin. To examine this hypothesis, we first measured the uptake of fluorescent paclitaxel in tau knock-down ZR75.1 cells and lamin siRNA-treated control cells. Intracellular paclitaxel is mostly bound to tubulin; therefore, cells with low tau expression are expected to take up more paclitaxel. Forty-four hours after siRNA transfection, cells were exposed to 1 μM Oregon Green paclitaxel for 20-80 min and then analyzed by FACS. Control cells showed unimodal distribution of fluorescence intensity (mean = 4 units), whereas the tau knock-down cell pool showed bimodal distribution indicating a fraction of highly fluorescent cells (mean = 100 units) (Fig. 3 a and b). We assume that the highly fluorescent subpopulation represents the successfully transfected cells. Twenty-seven percent (±6.3%) of cells showed >10 fluorescence units in the tau knock-down pool compared with 7.2% (±0.8%) in controls (Fig. 3c). The same FACS experiment was also conducted with spontaneously fluorescent epirubicin. The distributions were unimodal and similar in both the control and tau knocked-down cells (Fig. 3 d and e). These results suggest that cells with lowered tau protein expression accumulate more paclitaxel, but not epirubicin.

Next, we examined whether tau could reduce paclitaxel-induced tubulin polymerization in vitro. The rate of tubulin polymerization can be monitored by measuring optical absorbance of the tubulin solution at 340 nm (17, 18). When paclitaxel and tau were added simultaneously to the tubulin solution, their combined effect was partially additive (Fig. 4a). However, when tubulin was preincubated with tau before adding paclitaxel, which may represent a more physiological condition, the ability of paclitaxel to induce maximal tubulin polymerization was reduced in a dose-dependent manner (Fig. 4b). This finding may be due to (i) reduced availability of substrate because tubulin already polymerized by tau cannot be recruited by paclitaxel or (ii) tau directly competing with paclitaxel binding to tubulin.

To examine whether binding of paclitaxel to tubulin is affected by tau, we used fluorescent BODIPY-paclitaxel. When BODIPY-paclitaxel binds to microtubules, it displays enhanced fluorescence (19). BODIPY-paclitaxel (5 μM) was added to tubulin solution after 30 min of preincubation with (i) tau (15 μM), (ii) regular, nonfluorescent paclitaxel (10 and 20 μM), or (iii) control buffer (Fig. 4 c and d). Fluorescence was measured 30 min later. Preincubation with regular paclitaxel reduced the binding of BODIPY-paclitaxel to tubulin in a dose-dependent manner due to direct competition. Preincubation with tau produced a similar effect as did 10 μM regular paclitaxel (Fig. 4d). These experiments were repeated with tritium-labeled paclitaxel, and similar results were obtained (data not shown). These observations support the hypothesis that tau-stabilized microtubules bind less paclitaxel.

Discussion

In this study, we used transcriptional profiling as a screening tool to identify genes whose expression is associated with pCR to T/FAC preoperative chemotherapy for stage I-III breast cancer. Tumors that show pCR to preoperative chemotherapy represent extremely chemotherapy-sensitive cancers. The single best gene to discriminate between cases with pCR and residual disease was microtubule-associated protein tau. A subset of patients (n = 33) included in the current study was also included in a separate analysis when RNA was profiled on a cDNA microarray (8). Tau was not part of the final cDNA result-based predictor; however, it was among the top differentially expressed genes (20). Low tau mRNA expression was significantly more common among cases with pCR compared with those with residual disease. In the current study, no cases with high tau mRNA expression had pCR. This inverse correlation with pathologic response was confirmed with IHC in an independent patient cohort. These results suggest that low tau expression may be used as a marker to identify breast cancers that are particularly sensitive to paclitaxel-containing chemotherapy. We also demonstrated that reducing tau expression in breast cancer cells renders these cells more sensitive to paclitaxel but not to epirubicin in vitro. Microtubules assembled in the absence of tau were more vulnerable to paclitaxel, which could explain how low tau expression leads to increased sensitivity to this drug. However, tau alone is an imperfect marker. Approximately half of all cases with low tau expression had some residual cancer, suggesting that there are other mechanisms of resistance that can protect cells from paclitaxel even if the microtubules are hypervulnerable because of low tau expression.

The observation that tau expression modulates response to paclitaxel is unique but consistent with previously reported data. Expression of tau outside the central nervous system, including in epithelial cells, is well documented (21-23). Tau contains an imperfect estrogen response element upstream of its promoter and is an estrogen-induced protein in cultured neurons and in MCF-7 cells (24-26). Indeed, three of the four cell lines that expressed tau were ER positive (MCF7, ZR75, and T47D), and we also observed an association between low tau expression and ER-negative status in our clinical specimens. This phenomenon may partly explain why ER-negative breast cancers are more sensitive to paclitaxel chemotherapy. Furthermore, induction of tau expression with retinoic acid in neuroblastoma cells increased their resistance to paclitaxel, suggesting a protective effect (27). Tau is implicated in the pathology of Alzheimer's disease, and its microtubule binding characteristics have been studied extensively. Tau is able to bind to both the outer and inner surfaces of microtubules, and it may bind to the same inner-surface pocket as paclitaxel (28). Some investigators reported that under some conditions, tau may enhance binding of paclitaxel to microtubules (17, 19). In these reports, paclitaxel exposure either preceded tau exposure or was concomitant to it. When tau is added to paclitaxel-stabilized microtubules, tau binds to the outer surface of tubulin, rather than to the inner surface, and further enhances paclitaxel-induced polymerization (29, 30). We also observed this additive effect when the two molecules were added to tubulin concomitantly (Fig. 4a). Kinetic studies also showed that tau binds to microtubules differently depending on polymerization status (31). Microtubules assembled in the presence of paclitaxel show moderate binding affinity and rapid dissociation kinetics. In contrast, when microtubules are assembled in the presence of tau without paclitaxel, tau shows strong binding with slow dissociation. Our experiments are complementary to these observations in that we examined paclitaxel binding as a function of presence or absence of tau. Our findings suggest that microtubules assembled in the presence of tau are less susceptible to paclitaxel binding and pharmacological hyperpolymerization.

Several important questions remain to be examined. There are numerous alternative splice variants of tau, and each contains multiple phosphorylation sites that can affect interactions with tubulin. In our study, we did not address the importance of isoforms or the effect of phosphorylation. The microarray probe sets and the antibody that we used were directed against shared domains of the isoforms and were not sensitive to phosphorylation status. Whether tau confers partial protection to other microtubule-binding chemotherapy drugs, including docetaxel, epothilons, or vinca alkaloids, also remains to be investigated.

Whereas tau is a promising single gene marker of sensitivity to paclitaxel-containing chemotherapy, it is also clear that many tumors, despite low tau expression, are not fully sensitive to treatment, suggesting additional pathways of resistance. This observation is consistent with the commonly held belief that response to chemotherapy is a multifactorial process and that no single marker will be informative in all cases. Indeed, tubulin mutations, variable expression of tubulin isoforms, overexpression of multi-drug resistance transporters, or bcl-2 may all contribute to resistance to paclitaxel in tau-negative tumors (32-35). Multigene predictors that use information from several distinct molecular pathways of resistance will likely be more powerful than any single gene. However, low tau expression represents a unique molecular mechanism of hypersensitivity to paclitaxel. Inhibition of tau function could be explored as a potential therapeutic strategy to increase the anticancer activity of this drug.

Supplementary Material