Microtubule-Associated Protein-tau is a Bifunctional Predictor of Endocrine Sensitivity and Chemotherapy Resistance in Estrogen Receptor–Positive Breast Cancer

Purpose: The clinical outcome for patients with breast cancer is influenced by the metastatic competence of the cancer and its sensitivity to endocrine therapy and chemotherapy. A molecular marker may be prognostic of outcome or predictive of response to therapy, or a combination of both.

Experimental Design: We examined separately the prognostic and predictive values of tau mRNA expression in estrogen receptor (ER)–positive primary breast cancers in three patient cohorts. We used gene expression data from 209 untreated patients to assess the pure prognostic value of tau, data from 267 patients treated with adjuvant tamoxifen to assess predictive value for endocrine therapy, and data from 82 patients treated with preoperative paclitaxel followed by 5-fluorouracil, doxorubicin, and cyclophosphamide (paclitaxel/FAC) to assess predictive value for chemotherapy response. Wilcoxon rank sum test was used to compare tau expression between different outcome groups.

Results: Higher tau mRNA expression showed borderline nonsignificant association with better prognosis in the absence of systemic adjuvant therapy. Higher tau mRNA expression was significantly associated with no recurrence (at 5 and 10 years, P = 0.005 and P = 0.05, respectively) in patients treated with tamoxifen, indicating a predictive value for endocrine therapy. Tau expression was significantly lower in patients who achieved pathologic complete response to paclitaxel/FAC chemotherapy (P < 0.001).

Conclusion: This study suggests that high tau mRNA expression in ER-positive breast cancer indicates an endocrine-sensitive but chemotherapy-resistant disease. In contrast, low tau expression identifies a subset of ER-positive cancers that have poor prognosis with tamoxifen alone and may benefit from taxane-containing chemotherapy.

The long-term outcome for patients with newly diagnosed invasive breast cancer depends on the inherent biological aggressiveness of the disease and its sensitivity to adjuvant therapies. Patients with estrogen receptor (ER)–positive breast cancer who are at substantial risk for recurrence often receive both adjuvant endocrine therapy and chemotherapy. Prolonged recurrence-free survival in these patients may be due to any of the following causes: lack of micrometastatic disease at the time of diagnosis, effective eradication of micrometastases by adjuvant endocrine treatment or by adjuvant chemotherapy, or by the combination of both. When novel molecular markers of survival are proposed, it is imperative to examine if the marker identifies patients who have inherently good prognosis in the absence of any adjuvant therapy or if it predicts the efficacy of some form of systemic treatment.

We recently reported that low expression of microtubule-associated protein (MAP) tau in breast cancer was associated with higher rates of pathologic complete response to preoperative paclitaxel and 5-fluorouracil, doxorubicin, cyclophosphamide (paclitaxel/FAC) chemotherapy (1). Tau expression correlates closely with ER expression in human breast cancer. Tau is included in several previously reported ER gene signatures (2, 3). Furthermore, this gene contains an imperfect estrogen response element upstream to its promoter and it is an estrogen-induced gene in neurons and neuroblastoma cells, as well as in MCF-7 cells in vitro (4–6). This raises the possibility that tau expression is a marker of estrogen activity in breast cancer and therefore may predict benefit from antiestrogen therapy.

The goal of the current analysis was to examine separately the prognostic and predictive values of tau mRNA expression in three distinct human breast cancer gene expression data sets. The first set included 209 patients with ER-positive tumors who received no systemic adjuvant therapy, the second set included 267 patients with ER-positive tumors who received 5 years of adjuvant tamoxifen therapy but no adjuvant chemotherapy, and the third set included 82 patients with ER-positive tumors who received preoperative paclitaxel, 5-fluorouracil, doxorubicin, and cyclophosphamide chemotherapy.

Patient population. To assess the pure prognostic value of tau, we used publicly available gene expression data previously described by Wang et al. (7). This prognostic data set is available at the Gene Expression Omnibus database,⁷

accession number GSE2034, and contains gene expression information on 286 lymph node–negative patients who received no systemic adjuvant therapy. This data includes patients with both ER-negative and ER-positive tumors but, in our analysis, we included only the 209 ER-positive tumors. To examine the value of tau as predictor of survival after 5 years of adjuvant tamoxifen therapy, we used data provided by the Institut Jules Bordet (Brussels, Belgium). This included gene expression and survival data on 267 patients with ER-positive tumors treated with adjuvant tamoxifen. Results from this data set were previously reported by Sotirou et al. (8) and by Loi at al.⁸ Data are available at the Gene Expression Omnibus database (accession number pending). The third data set consisted of 82 patients with ER-positive tumors treated with preoperative chemotherapy including 12 doses of weekly (80 mg/m²) or thrice weekly (225 mg/m²) paclitaxel followed by four courses of 5-fluorouracil (500 mg/m²), doxorubicin (50 mg/m²), and cyclophosphamide (500 mg/m²; paclitaxel/FAC). We used this data set to examine the value of tau as a single gene marker of pathologic response to paclitaxel/FAC (T/FAC) chemotherapy. Response was dichotomized as pathologic complete response (defined as no residual invasive cancer in the breast or lymph nodes) or residual disease. This data set was also previously described by Hess et al., and the microarray data are available at the M. D. Anderson Cancer Center bioinformatics web site (9).⁹ In all three data sets, ER status was determined from routine pathologic assessment including immunohistochemistry or ER-ligand binding assay for the older specimens. The cutoff for ER positivity was 10 fmol/mg protein or >10% positive cells by immunohistochemistry. Clinical characteristics are reported in Table 1.

Gene expression analysis. All three gene expression data sets were obtained using Affymetrix U133A Gene Chip following standard operating procedures. Additional details on sample processing, RNA extraction, and hybridization are available in the original publications (7–9). Surgical biopsy specimens were used for microarray profiling in the prognostic and the adjuvant endocrine data sets, whereas fine needle aspiration material was used to generate the M. D. Anderson Cancer Center neoadjuvant chemotherapy data. Only the normalized probe set intensities were available for downloading and analysis from the prognostic data set. For the adjuvant endocrine and for the M. D. Anderson Cancer Center data sets, the raw intensity files (CEL) from each microarray were available. We normalized the gene expression data using dChip V1.3 software¹⁰

to a single reference array. The normalization files are available online.⁹ After normalization, the 75th percentile of the pixel level was used as the intensity level for each feature on a microarray. Multiple features representing each probe set were aggregated using the perfect match model to form a single measure of intensity for the probe set. A single Affymetrix probe set, 203929_s_at, was used as a measure of tau expression. This particular probe set was selected from the four distinct tau probe sets that are present on the U133A chip because it targets the most 3′ end of the transcript and it showed the highest mean expression in two of the three data sets. It also discriminated effectively between patients with pathologic complete response and those with residual invasive cancer after neoadjuvant chemotherapy in our previously published work (1).

Statistical analysis. The correlation between tau expression and ER status and between tau expression and pathologic complete response were assessed by two-sample Wilcoxon rank sum test. This test was used because of the nonnormal distribution of tau. The prognostic value of tau in untreated and tamoxifen-treated patients was assessed by examining the association between tau expression and the occurrence of distant metastases using the Wilcoxon test. To assess if covariates including grade, tumor size, and nodal status were associated with tau levels, we did multivariate linear regression analyses. Kaplan-Meier survival curves were compared using the log-rank test. In order to generate Kaplan-Meier curves, patients had to be assigned to various tau expression groups. However, no cutoff points have been previously established to assign low or high tau status to cases based on microarray results. Furthermore, normalized expression values differed between the data sets, particularly for the prognostic data set that used different normalization procedures. In the absence of accepted cutoff values and considering the nonnormal expression distribution of tau, we grouped patients by tertiles and quartiles. The R statistical environment was used for all statistical analyses (10).

Comparison of tau expression between ER-positive and ER-negative cancers. Tau expression was examined in relation to ER status in two distinct cohorts of patients. The first included public microarray data from 286 patients, including 209 ER-positive and 77 ER-negative cases. Gene expression data were derived from the analysis of surgical resection tissue and the ER status was assigned to each case by the original investigators based on routine pathology results (7). The second cohort included 133 cases from M. D. Anderson Cancer Center, including 82 ER-positive and 51 ER-negative tumors. Gene expression data were generated from fine needle aspiration specimens. Figure 1 illustrates the results. In both data sets, tau expression was significantly lower in ER-negative cancers compared with ER-positive cancers. The median expression was 300 normalized intensity units (range, 14-3,690) in the ER-negative and 1,511 (range, 39-8,139) in ER-positive tumors (P < 0.0001) in the Wang data set. It was 112 (range, 61-579) and 350 (range, 51-1628), respectively (P < 0.0001) in the M. D. Anderson Cancer Center data set. The differences in intensity scale are due to differences in the normalization procedures applied to the two data sets.

Next, we examined tau expression within ER-positive breast cancers only. Figure 2 illustrates that there is substantial variation in tau expression within ER-positive cancers and the expression distribution is nonnormal. This indicates that within each distinct data set, there is a minority of ER-positive cancers that show very high expression levels of tau compared with the rest of cases.

Prognostic value of tau in ER-positive, node-negative disease in the absence of systemic adjuvant therapy. We first assessed the prognostic value of tau mRNA expression in patients with node-negative, ER-positive disease who received no adjuvant systemic therapy. Of the 209 patients included in this analysis, 127 had >5 years follow-up and 34 patients had >10 years of follow-up. There was a statistically nonsignificant trend for lower tau expression in those who relapsed in the first 5 years of follow-up (median, 1,300; SD, 1,476) compared with those who did not (median, 1,558; SD, 1,498; P = 0.09; Fig. 3A). Interestingly, this trend reached significance by 10 years of follow-up (P = 0.025; Fig. 3B). We also did Kaplan-Meier survival analyses and grouped patients into tertiles and quartiles of tau expression. There was a statistically nonsignificant trend towards better survival in patients in the top two tertiles (top 66%, P = 0.09; Fig. 4A) and quartiles (top 50%, P = 0.14; Supplementary Fig. S5) compared with patients with lower levels of tau expression. These results indicate that high tau mRNA expression has a borderline association with better prognosis. However, it is not a significant prognostic marker in the absence of systemic adjuvant therapy.

Predictive value of tau in ER-positive breast cancer treated with adjuvant tamoxifen. We examined tau expression in 267 ER-positive tumors from patients who received 5 years of adjuvant tamoxifen therapy. The median follow-up for these patients was 101 months (range, 20-171). Figure 3C shows that tau expression was significantly higher (P = 0.005) in patients who did not have a distant metastatic recurrence in 5 years (n = 229; median, 258; SD, 305) compared with those who did (n = 38; median, 168; SD, 190). The same positive correlation between tau levels and better recurrence-free survival remained significant at 10 years of follow-up (P = 0.025; Fig. 3D). Multivariate linear regression analysis including grade, tumor size, and nodal status showed a significant association between tau expression and grade. Tau expression was lower in higher grade tumors (P = 0.025; Supplementary Fig. S6). Kaplan-Meier survival analysis of patients grouped into tertiles and quartiles of tau expression confirmed significantly better survival for patients in the upper two tertiles (P = 0.008; Fig. 4B). Similar results were obtained when patients were grouped by quartiles (P = 0.009; Supplementary Fig. S5). Multivariate Cox regression analysis including age, tumor size, nodal status, ER expression, and tau indicated that tau remains a significant independent predictor of outcome (Table 2). Higher MAP-tau expression was associated with a decreased risk of distant recurrence in this population of patients with tamoxifen-treated breast cancer (hazard ratio, 0.60; 95% confidence interval, 0.38-0.94; P = 0.03).

Overall, these data indicate that patients with high tau expression have good prognosis with adjuvant endocrine therapy. On the other hand, patients with the lowest one-third of tau expression values have unacceptably high rates of relapse both at 5 and at 10 years postsurgery.

Tau as predictor of response to preoperative paclitaxel/FAC chemotherapy in ER-positive breast cancer. In an earlier study including both ER-positive and ER-negative cases, we reported that low tau expression was associated with higher rate of pathologic complete response to preoperative paclitaxel/FAC chemotherapy (1). In the current study, we examined the same hypothesis focusing on ER-positive breast cancers only. A total of 82 cases were included in the current analysis and this included 35 cases from the original study. Figure 3E shows that the median tau expression was 143 (SD, 58) in cases with pathologic complete response (n = 7) compared with 429 (SD, 353) in cases with residual cancer (n = 75) after chemotherapy (P < 0.001). The pathologic complete response rate was 19% (5 of 27) in the lowest tertile of tau expression, 3% (2 of 27) in the median tertile, and 0% (0 of 28) in the highest tertile. Multivariate logistic regression analysis including age, tumor size, nodal status, ER expression, and tau indicated that tau remains a significant independent predictor of pathologic complete response (Table 2). High MAP-tau expression was associated with a decreased odds ratio for the occurrence of pathologic complete response (odds ratio, 0.14; 95% confidence interval, 0.03-0.59; P = 0.008). These results corroborate our previous observation that low MAP-tau expression was associated with paclitaxel sensitivity. In the current article, we also show that this association is particularly strong in ER-positive breast cancer.

Tau mRNA expression is highly correlated with ER status. ER-negative cancers have significantly lower tau expression than ER-positive cancers. This observation holds true in multiple independent data sets and is consistent with in vitro observations which suggested that tau is an ER-regulated molecule (6). However, there is considerable variation in tau expression within ER-positive breast cancers. We examined the association between tau levels and clinical outcome including response to chemotherapy, endocrine therapy, and distant recurrence-free survival. We provide further evidence that low tau expression is associated with extreme sensitivity to paclitaxel-containing preoperative chemotherapy. This inverse correlation is particularly strong in ER-positive cancers. This is clinically potentially important because ER-positive breast cancers are generally less sensitive to chemotherapy (11, 12). Low tau expression may identify a subset of ER-positive tumors that benefit from adjuvant (or neoadjuvant) paclitaxel-containing chemotherapy. Low tau protein expression may confer increased paclitaxel sensitivity by allowing increased binding of this drug to microtubules (1).

We also observed a positive correlation between tau expression and better distant recurrence-free survival after adjuvant tamoxifen therapy. Patients with no recurrence had higher tau levels. This is consistent with the hypothesis that tau may be an indicator of active estrogen signaling and is therefore a marker of sensitivity to antiestrogen therapy. This hypothesis is also supported by laboratory evidence. Tau is an estrogen-induced gene in several in vitro cell line models (4–6, 13, 14).

Tau expression showed only borderline significant association with better prognosis in the absence of any systemic adjuvant therapy in ER-positive breast cancer. This indicates that it is not a powerful prognostic marker but it is a bifunctional predictor of benefit from endocrine therapy and chemotherapy in ER-positive breast cancer. It could predict opposite responses to these two important treatment modalities. High tau mRNA expression in ER-positive breast cancer could indicate an endocrine-sensitive but relatively chemotherapy-resistant disease, whereas low tau expression could identify a subset of ER-positive cancers that have poor prognosis with tamoxifen alone and may benefit from adjuvant taxane-containing chemotherapy.

These data add to the growing body of evidence that indicates an inverse relationship between endocrine and chemotherapy sensitivity in ER-positive breast cancer. It was recently shown that high recurrence score measured by Oncotype DX (Genomic Health, Inc., Redwood, CA) predicted little or no benefit from adjuvant tamoxifen therapy, but at the same time also predicted substantial benefit from adjuvant CMF chemotherapy (15, 16). We also observed a similar inverse correlation between a 200-gene ER reporter score and endocrine sensitivity and chemosensitivity.¹¹

It should be noted that in the current article, we present a correlation between a molecular marker and clinical outcome and not the performance characteristics of a new test. In order to validate tau mRNA expression as a diagnostic predictive test, the measurement methodology, normalization procedure, and cutoff values to assign positive and negative status will need to be defined a priori. Based on our results, we suggest that if tau mRNA expression is measured by an Affymetrix U133A gene chip and is normalized according to our normalization procedure,⁹ then tau levels ≥183 (lowest tertile) indicate a 78% 10-year distant metastasis-free survival with adjuvant tamoxifen and predict a <4% pathologic complete response rate from paclitaxel/FAC chemotherapy. Tau mRNA levels <183 indicate a 65% 10-year distant metastasis-free survival with adjuvant tamoxifen and predict a 26% pathologic complete response rate. However, the true predictive values of these suggested Tau mRNA thresholds will need to be validated in an independent study (17). It is also important to consider that the inclusion of other genes in addition to tau in a multigene prediction score can improve its endocrine therapy and chemotherapy response predictive values (3, 9).