Molecular subtypes of breast cancer in relation to paclitaxel response and outcomes in women with metastatic disease: results from CALGB 9342

The taxanes (paclitaxel and docetaxel) are among the most active drugs for treatment of breast cancer and are an important component of treatment in the neoadjuvant, adjuvant, and metastatic settings. These drugs act, at least in part, by stabilizing microtubules and inducing G₂/M arrest, with subsequent apoptosis in malignant cells. Paclitaxel, the first taxane that was developed, has substantial antitumor activity. As a result, it has become the standard of care as first-line treatment in the metastatic setting, and it is indicated for the treatment of high-risk, early-stage breast cancer, in addition to anthracycline-based therapy [1–8]. However, little is known about the mechanisms of sensitivity and resistance to taxanes. Hence, an important therapeutic goal is to identify molecular predictors of a response to taxane therapy.

Human epidermal growth factor receptor (HER)2 (ErbB2) is over-expressed in 25% to 30% of human breast cancers and is associated with reduced disease-free survival and overall survival (OS) [9]. Data from Cancer and Leukemia Group B (CALGB) trial 8869/8541 [10] and National Surgical Adjuvant Breast and Bowel Project (NSABP) trial B11 [11] indicate an improved response to doxorubicin-containing regimens in patients whose tumors over-express HER2. Data on the association between this growth factor and taxane sensitivity are limited and conflicting. In some experimentalsystems, transfection of HER2 leads to paclitaxel resistance by blocking activation of Cdc2, which is required for paclitaxel-induced apoptosis [12, 13]. HER2 signaling also activates Akt, providing a survival signal through that pathway [14]. In contrast, HER2 activation may increase sensitivity to paclitaxel by abrogation of the G₁/S checkpoint through p27, allowing progression to G₂/M [15].

Clinical studies of HER2 tumor levels in relation to the response to taxanes have also yielded conflicting results. Seidman and collaborators [16] reported an association between HER2 over-expression and an improved response to taxanes in a retrospective analysis of patients with breast cancer who had been treated in several clinical trials. Other studies have found no association with treatment response [17, 18], resistance to paclitaxel [19], or an improved response in patients with HER2-positive tumors [20–22]. These differing results may be due to differences in patient selection, study design and sample size, and variation in the methods used to determine HER2 status.

Mutations in the tumor suppressor gene p53 are present in 18% to 25% of primary breast carcinomas [23, 24]. In general, missense mutations lead to nuclear accumulation of p53, but the mutant protein is not capable of normal p53 function. Mutations in p53 may increase sensitivity to paclitaxel by abrogating the G₁ checkpoint, allowing cells to progress to G₂/M, as suggested by in vitro models [25]. In addition, expression of MAP4 (microtubule-associated protein 4), which occurs when p53 is transcriptionally silent, stabilizes polymerized microtubules and increases sensitivity to paclitaxel [26, 27]. Alternatively, p53 mutations may lead to taxane resistance, because mutant p53 cannot upregulate expression of Bax, which participates in apoptosis [28]. Finally, specific mutations in p53 have been shown to disrupt spindle checkpoint control, potentially increasing resistance to taxane-induced damage [29].

Newer approaches to molecular classification of breast cancer have identified three distinct subclasses with both biologic and prognostic significance. These subclasses, defined by classic biomarkers, are estrogen receptor (ER) and/or progesterone receptor (PR) positive tumors, HER2-amplified tumors, and ER/PR/HER2-negative tumors. The three subtypes have been reproducibly identified by gene expression profiling in multiple breast cancer cohorts and exhibit consistent prognostic significance [30–34]. In addition, it was recently reported that African-American women are more likely to have the 'triple negative' breast cancer phenotype [35, 36]. These tumors lack expression of ER, PR, and HER2; they express basal keratins and myoepithelial markers; and they are associated with worse outcome in early-stage breast cancer [34, 37].

In CALGB 9342, women with advanced breast cancer were randomly assigned to paclitaxel, given alone as first-line or second-line chemotherapy, at one of three dose levels [8]. We conducted a biomarker companion study to determine whether HER2 status (measured by immunohistochemistry [IHC] and fluorescent in situ hybridization [FISH]), hormone receptor status, p53 status (measured by IHC and sequencing), or a combination of these markers predicted clinical outcomes in this cohort of patients with metastatic breast cancer. We also determined the frequency of molecular markers in African-American women and examined whether race influenced outcomes.

We evaluated the relationship among tumor biomarkers, treatment response, and outcomes in a subset of patients with available tumor blocks who were enrolled in a large, prospective, randomized trial of single-agent paclitaxel given as first-line or second-line therapy for metastatic breast cancer. Neither HER2 status nor p53 status significantly affected the treatment response or TTF when HER2 over-expression was scored as 3+ on the HercepTest or FISH positive. Although p53 status and ER status were not predictive of a benefit from therapy, measured by response or TTF, they were both important variables in determining OS in this cohort. Thus, it appears that p53 and ER status behaved as prognostic factors in this study, but they did not add predictive value within the context of response to paclitaxel. Although not surprising, this observation underlines the importance of separating response to therapy from the underlying biology, a distinction first made by McShane and coworkers [40]. In addition, our study provides support for the use of classification systems that include ER, PR, and HER2 status, because the outcomes according to these biomarkers in our cohort are consistent with published data [30, 34]. Furthermore, racial differences in the pattern of tumor subtypes are also seen, a finding that translates into differences in survival, even in the metastatic setting.

Other studies have examined response to taxanes based on HER2 status. Seidman and coworkers [16] reported a higher rate of response to paclitaxel in patients with HER2-positive tumors; however, this association was seen only when HER2 status was assessed with the use of the 4D5 antibody, and not when a rabbit polyclonal antibody (pAb-1) was used to determine HER2 status. The authors reported a low level of agreement between the results of the 4D5 assay and an assay with pAb-1. The monoclonal antibody CB11 and the polyclonal antibody used in the HercepTest are known to target different epitopes of HER2 and appear to vary in the specificity of their results. Therefore, in the present study we evaluated the correlation among three methods of measuring HER2 status that have been approved by the US Food and Drug Administration. We found that the results of the CB11 assay, the HercepTest, and FISH were reasonably concordant, with the highest level of agreement between the CB11 assay and FISH.

No association was seen between the rate of response to paclitaxel and HER2 status as assessed by any method. In contrast to our findings, Konecny and coworkers [21] reported a statistically significant increase in response rate to epirubicin and paclitaxel, but not to epirubicin and cyclophosphamide, in women with HER2-positive tumors, as assessed by FISH. One possible explanation for these conflicting results is that the patient populations were not entirely comparable, because all patients in the study reported by Konecny and coworkers received treatment as first-line therapy, whereas in CALGB 9342 paclitaxel was given as either first-line or second-line therapy for metastatic disease. In addition, the treatment regimens were not entirely comparable. There may be an advantage to combining an anthracycline with a taxane in women with HER2-positive tumors; if so, epirubicin plus paclitaxel may be a particularly effective regimen against HER2-positive breast cancer, whereas HER2 status may not influence the likelihood of a response to paclitaxel given alone. Furthermore, FISH techniques differed between these studies and might have influenced the findings. Finally, our sample was relatively small, although the identical response rates in women with HER2-positive tumors and those with HER2-negative tumors suggests that a substantial difference in response rates would be unlikely in a larger sample.

Our data showed that hormone receptor status is important as a prognostic factor in this study but was not predictive of clinical response or TTF after paclitaxel therapy. Conflicting data have been reported regarding the role of ER in predicting benefit from taxane therapy. In CALGB 9344, a subgroup analysis demonstrated that the addition of paclitaxel in the adjuvant setting was more beneficial in women with ER/PR-negative tumors than in those with ER/PR-positive tumors [1]. This observation is supported by a combined analysis conducted by Henderson and coworkers [41], which demonstrated a strong relationship between ER/PR negativity and a benefit from chemotherapy. However, the results of the NSABP B-28 trial [2] and the Breast Cancer International Research Group (BCIRG) 001 trial [42] do not support this finding. Our study suggests that differences in survival as a result of hormone receptor status need to be considered in analyses of trials in the metastatic setting, even among patient populations that have previously received hormonal therapy.

We further evaluated the relationship among p53 status, race, and outcomes in this cohort. We found that the presence of p53 over-expression, as detected by IHC, was not predictive of the response to paclitaxel but was associated with significantly diminished OS. There are two possible explanations for these findings. First, mutations in p53 may lead to genomic instability in vitro, with resistant clones arising more rapidly in these tumors, even though the initial response to therapy is the same as that in tumors without p53 mutations. It is also possible that other tumor features that predict a worse prognosis are associated with p53 mutations and influence survival without affecting response to therapy.

In this study p53 over-expression, as determined by IHC, was associated with decreased survival, but p53 mutations identified by sequence analysis were not. Explanations include the possibility that certain p53 mutations have little effect on the tumor phenotype, whereas the presence of p53 stabilization is a surrogate for a mutation with a functional effect. The study was not sufficiently powered to allow for a rigorous examination of the interaction between mutation type and clinical outcome. Furthermore, previously described p53 mutations were not detected in approximately one-quarter of cases in which the IHC assay for p53 was positive. We used stringent criteria for the classification of mutations; only those included in the p53 mutation database maintained at Hôpital Necker-Enfants Malades were considered true mutations. Furthermore, it is possible that some of the mutations identified by sequencing but not included in the database represent true functional mutations.

Large, population-based studies have documented racial disparities in breast cancer outcomes [43]. As demonstrated by Polite and coworkers [44], self-reported African-American women from CALGB 9342 had decreased survival as compared with their Caucasian counterparts, even with adjustment for other factors. In this biomarker substudy, we observed the association between African-American race and worse outcome. A potential explanation for these findings is the observation that these patients were twice as likely to have tumors that were negative for ER, PR and HER2, and triple-negative status was itself associated with significantly poorer OS. Because we did not collect data on post-study treatment, we cannot rule out differences in care received after the study as a cause of the observed disparity. However, our results extend the observations of Olopade [36] and Carey [35] and their groups, who reported that basal-like tumors are more common in women of African ancestry. Because the basal-like phenotype has been associated with a poor prognosis, our findings suggest that differences in the biologic features of tumors may explain the decreased survival of African-American women in our study. To our knowledge, this is the first study to demonstrate that the association between a poor prognosis and the triple-negative phenotype persists in women with metastatic breast cancer.

The study had several limitations. First, we were able to obtain adequate tumor blocks from only approximately one-third of patients enrolled in CALGB 9342. As a result, the power to detect all but substantial differences in outcomes based on biomarkers was limited. Second, we cannot rule out the possibility that the dose-response curve for paclitaxel differs according to HER2 status; unfortunately, our sample was not large enough to carry out such an analysis. Third, 90% of the biomarker assays were performed on blocks from primary tumors (and 10% on blocks from synchronous lymph-node metastases), on the assumption that HER2 status is stable over time regardless of disease progression or interim therapies. A recent abstract has questioned this assumption [45], reporting discordance in HER2 expression between primary and metastatic sites in eight (14%) of 58 patients. Finally, patterns of gene expression may ultimately be more useful than single gene markers in predicting the response to therapy [46–48].

Molecular subtyping has become commonplace in breast cancer, but the implications of molecular markers in patients with metastatic disease remains unclear. This study suggests that tumor response is not dictated by p53, HER2 or ER status, but that these tumor features impact on tumor regrowth and death in patients with metastatic breast cancer. In addition, the recent observation of increased frequency of 'triple-negative' phenotype in African-American patients was confirmed. Furthermore, this study is the first to point out that the negative prognosis of triple-negative tumors continues beyond the diagnosis of metastasis, but it does not seem to relate to tumor response, at least to paclitaxel. These observations further point out the importance of tumor biology in behavior of breast cancer in the clinical setting.