Phosphorylated p-70S6K predicts tamoxifen resistance in postmenopausal breast cancer patients randomized between adjuvant tamoxifen versus no systemic treatment

Activation of the mitogen-activated protein kinase (MAPK) and phosphatidylinositol-3-kinase (PI3K) pathways confers anti-estrogen resistance in vitro[1‐3]. Apart from activation by the canonical pathway drivers (PIK3CA mutations [4], loss of PTEN [5] or overexpression of growth factor receptors such as human epidermal growth factor receptor 2 (HER2) [6] and insulin like growth factor 1 receptor [7]), in vitro data have shown that the PI3K pathway can also be activated in response to estrogen depletion. This results in acquired hormone-resistant breast cancer cells that are sensitive to PI3K/mammalian target of rapamycine (mTOR) inhibition [8]. These preclinical data support the clinical observation that estrogen receptor alpha (ERα)-positive metastatic breast cancer patients with prior exposure to aromatase inhibitors derive substantial benefit from the addition of an mTOR inhibitor [9]. Whether or not patients who are primarily resistant to adjuvant endocrine therapy might benefit from PI3K and/or MAPK pathway inhibition remains to be defined. A biomarker of an activated PI3K and/or MAPK pathway with clinical validity to predict resistance in the adjuvant setting has not been identified [10], but could potentially be used as a companion diagnostic for non-ERα-targeted drugs, such as an mTOR inhibitor.

Several canonical pathway drivers, such as PIK3CA mutations [11, 12], loss of PTEN [12], and HER2 [13], have been studied for their validity to predict resistance. However, none of these drivers significantly predicts lack of benefit from endocrine therapy. An important issue to be considered is that the presence of these drivers in clinical samples may not necessarily result in high activation of downstream proteins [12, 14]. In vitro, PI3K pathway activation leads to phosphorylation of AKT and subsequently of mTOR and p70S6K [10]. Phosphorylation of extracellular signal-regulated kinase (ERK)1/2 is a result of MAPK pathway activation [15] and mediates activation of p70S6K [16]. However, relatively moderate activation of the PI3K pathway was seen in tumors with a PIK3CA exon 20 mutation-associated gene signature [14]. In addition, in a large series of primary breast cancer tumors, reverse-phase protein analysis did not show activation of the typical downstream proteins in the PI3K pathway in PIK3CA mutated luminal A tumors [17]. The activation status of downstream proteins rather than the presence or absence of a canonical driver therefore probably ultimately defines anti-estrogen sensitivity in breast cancer patients.

We hypothesized that activated proteins downstream in the PI3K and/or MAPK kinase pathways could potentially be used as a marker that separates patients who are likely to benefit from adjuvant tamoxifen treatment from those who are primarily resistant to this drug. The aim of our study was therefore to investigate the predictive value of different downstream activated proteins in the PI3K and/or MAPK pathways in a large series of ERα-positive postmenopausal breast cancer patients randomized between adjuvant tamoxifen versus no systemic treatment.

In this retrospective analysis of tumor samples from ERα-positive postmenopausal breast cancer patients, we have shown the clinical validity of p-p70S6K, as a marker of PI3K and/or MAPK pathway activation, to predict adjuvant tamoxifen resistance. Patients whose tumor expressed p-p70S6K did not benefit from adjuvant tamoxifen, confirming previous in vitro findings [1, 2].

We are aware of the discussion that the use of phospho-specific antibodies for IHC may be challenging, since preanalytic variables (such as fixation technique and duration) may critically affect the signal [26]. Although we did not observe an association between phospho-protein expression and either tumor size, inclusion center or relative age of the tumor sample and confirmed the phospho-specificity of the antibodies, we cannot exclude that unknown preanalytic variables might have affected the results of IHC. Nevertheless, since these unknown variables would have affected both tumor samples from control patients as well as tamoxifen-treated patients, the observed predictive value of p-p70S6K markers is probably not biased by these preanalytic variables. Unlike others [29, 30], we have used automatic immunostainings to improve robustness and reproducibility of these IHC-based tests. Nevertheless, the visual interpretation of IHC-based tests is still a subjective, time-consuming and variable process, with an inherent intra-observer and inter-observer variability [31, 32]. The technical validity of IHC-based tests may therefore be further improved by image analysis applications on digital slides [33]. We deemed a direct comparison of the observed frequencies of breast cancers scoring positive for the downstream-activated proteins analyzed with those observed in other studies [30, 34, 35] not to be appropriate due to differences in study population, antibodies and staining techniques as well as interpretation of these stainings.

Very few studies have tested the predictive validity of downstream-activated proteins in the PI3K and/or MAPK pathway in the context of a randomized clinical trial. Perez-Tenorio and colleagues observed a decreased benefit from tamoxifen only in those patients with PIK3CA mutations who also expressed AKT activation [12], but the test for interaction was not significant. In a subgroup analysis of ERα-positive and PgR-positive patients, Bostner and colleagues recently observed a reduced benefit from tamoxifen in those patients whose tumor expressed high p-mTOR [29]. These results support our data and suggest that downstream activated proteins have a superior clinical validity to predict endocrine resistance compared with the presence or absence of a pathway driver [36]. The strength of our study is that we pre-specified the cutoff value (median expression level) that was used for the primary analysis of the putative marker as a binary factor. In addition, we tested the markers as continuous linear variables and used Aikake’s information criteria to define the optimal cutoff, which coincided with the median for most putative biomarkers. p-p70S6K (as a marker of activation of either the PI3K pathway and/or the MAPK pathway) was the only significant marker according to our conservative definition of significance, although most of the other downstream activated proteins showed a similar trend. In addition, we found a significant interaction for the composed variable of either high p-ERK1/2 and/or p-mTOR expression, representing activated MAPK and PI3K pathways, respectively.

In our study, interestingly, most of the markers of PI3K and/or MAPK activation were associated with a favorable prognosis in the absence of systemic treatment. Although an association with low-grade and positive PgR status was shown, the favorable outcome was independent of these factors in multivariate analysis. The question arises of why these tumors have a favorable prognosis. Although others have studied the association of these markers with prognosis in series of patients treated with endocrine therapy [8, 30, 37], none of these studies discerned prognosis from prediction. In case a putative biomarker under investigation results in a reduced treatment efficacy, this would cancel out a potential prognostic effect if the biomarker is analyzed in series of patients who are all treated with the drug [10]. A direct comparison between our data and these studies is therefore not possible.

Our study has several limitations. First of all, for retrospective biomarker analysis we could only use the subgroup of 563 ERα-positive patients from whom sufficient tumor material was available. However, this subgroup did not differ from the total study population. Furthermore, the duration of tamoxifen therapy differs from the current duration of at least 5 years. However, we anticipate that the relative effects of the biomarkers analyzed in this study will be similar for shorter and longer duration of endocrine therapy. The patients in our study randomized to adjuvant treatment received tamoxifen only (and no aromatase inhibitors), while currently most ERα-positive postmenopausal breast cancer patients receive an aromatase inhibitor preceding or following tamoxifen treatment. Since the putative predictive biomarkers studied are considered to cause escape from hormone dependency [8], our data are thought to be applicable to current clinical practice as well. Furthermore, the patients in our trial had not received adjuvant chemotherapy and thereby the observed effects of the putative biomarkers were not biased by endocrine-resistant patients who were cured by adjuvant chemotherapy. In current clinical practice, patients who relapse despite adjuvant endocrine therapy and adjuvant chemotherapy are both endocrine therapy resistant and chemotherapy resistant. The adjuvant endocrine therapy resistance may be explained by activation of the PI3K/MAPK pathway and our data suggest that p-p70S6K may identify those patients.

It would be clinically relevant to determine whether p-p70S6K could be used as a companion diagnostic for the addition of non-ERα-targeted drugs (like PI3K/mTOR inhibitors) to endocrine therapy in the adjuvant setting. In metastatic breast cancer patients, who had been randomized between tamoxifen alone or in combination with everolimus, a trend for a better treatment efficacy was observed for patients with a tumor with high p-p70S6K [38]. Randomized clinical trials in the adjuvant setting testing the addition of everolimus to hormonal therapy are currently recruiting patients [39], and these studies are of great importance to test the treatment predictive value of p-p70S6K in this setting.

With the approval of everolimus for metastatic breast cancer patients with acquired endocrine therapy resistance, the first non-ERα-targeted drug that can overcome endocrine therapy resistance has become clinically available. Whether or not inhibitors of the PI3K/Akt/mTOR pathway may overcome endocrine therapy resistance in the adjuvant setting is currently under investigation. A biomarker of an activated PI3K and/or MAPK pathway with clinical validity to predict endocrine therapy resistance in the adjuvant setting has not been identified [10], but such a biomarker could potentially be used as a companion diagnostic for these non-ERα-targeted drugs. In our series of ERα-positive postmenopausal breast cancer patients, p-p70S6K is significantly associated with intrinsic tamoxifen resistance. Patients whose tumor expresses p-p70S6K, as a marker of downstream PI3K and/or MAPK pathway activation, have a favorable prognosis but do not benefit from adjuvant tamoxifen. It would be of great importance to validate these findings in other randomized series with endocrine therapy and to further explore this marker as a potential companion diagnostic.