The prognosis of patients treated with everolimus for advanced ER‐positive, HER2‐negative breast cancer is driven by molecular features

Abstract

Everolimus is widely used in patients with advanced ER‐positive, HER2‐negative breast cancer. We looked at alterations in the PIK3CA/AKT/mTOR pathway in a multicenter cohort as potential biomarkers of efficacy. Patients with advanced ER‐positive, HER2‐negative breast cancer treated with everolimus and endocrine therapy between 2012 and 2014 in two cancer centers were included. Targeted sequencing examined mutations in PIK3CA, ESR1, and AKT1 genes. An immunochemical analysis was conducted to evaluate expression of PTEN, INPP4B, STK11, p4EBP1, and pS6. We analyzed 71 patients (44 primary tumors; 27 metastatic tissues). Median age was 63 years [58–69]. All patients had heavily pretreated advanced disease. A mutation in the PIK3CA pathway was observed in 32 samples (PIK3CA exons 10 and 21 and AKT1 exon 4 in 15.5%, 24.0%, and 5.6% of samples), and in ESR1 in 5 samples (7.0%), respectively. Most samples showed cytoplasmic expression of the PIK3CA pathway proteins. Progression‐free survival was longer in patients with a pS6 or p4EBP1 histoscore ≥ median value (6.6 versus 3.7 months, p = 0.037), and in patients with a PTEN histoscore ≤ median value (7.1 versus 5.3 months, p = 0.02). Overall survival was longer in patients with pS6 ≥ 3rd quartile (27.6 versus 19.3 months, p = 0.038) and in patients with any mutation in the PIK3CA/AKT/mTOR pathway (27.6 versus 19.3 months, p = 0.011). The prognosis of patients treated with everolimus for advanced ER‐positive, HER2‐negative breast cancer appears primarily driven by molecular features associated with the activation of the PIK3CA/AKT/mTOR pathway.

Introduction

The pivotal phase 3 BOLERO‐2 trial demonstrated the efficacy and safety of the combination of everolimus with exemestane compared with exemestane in patients with advanced ER‐positive, HER2‐negative breast cancer and progressive disease under nonsteroidal aromatase inhibitor (AI) treatment [1]. Everolimus has since become one of the therapeutic cornerstones of advanced hormone receptor‐positive, HER2‐negative (HR+, HER2−) breast cancer [2]. Several studies have repeatedly confirmed the efficacy of everolimus in the real‐world setting, including in very advanced disease [3, 4]. Recently, the French epidemiological study and medico‐economic (ESME) program, a nation‐wide real‐world study of patients with advanced breast cancer, suggested that adding everolimus to the treatments sequences had a favorable impact on survival [5]. However, the use of everolimus in the daily practice has been limited by cumbersome adverse events, such as mucositis, weight loss, or hyperglycemia [1, 3]. Furthermore, the recent advent of CDK4/6 inhibitors as major treatment of patients with advanced HR+/HER2− breast cancer has profoundly changed the therapeutic landscape [6].

These data underline the need to better identify the subsets of patients who may actually benefit from everolimus‐based therapy. It has been suggested that activating alterations in the PIK3CA/AKT/mTOR pathway may confer sensitivity to mTOR inhibitors such as everolimus. In vitro, tumor cells carrying oncogenic PIK3CA mutations or with a PTEN 0 histoscore were sensitive to everolimus, except when KRAS or BRAF mutations were concomitantly present [7]. In the pivotal BOLERO‐2 study, progression‐free survival (PFS) was longer in patients with PIK3CA exon 10 mutation compared to patients with PIK3CA exon 21 mutation. However, no PFS difference was observed between PIK3CA wild type and PIK3CA‐mutated patients, whether captured in tumor tissue of circulating cell‐free DNA [3, 8]. More interestingly, early results from the prospective SAFIR‐TOR study have also confirmed that activation of the mTOR pathway as evaluated by phospho‐4EBP1 (p4EBP1) staining on the metastatic tissue may be associated with an enhanced efficacy of everolimus [8].

In the present report, we propose a translational analysis of our previously reported cohort which demonstrated efficacy of everolimus including in heavily pre‐treated patients, in line with the more recent ESME report [3, 5]. We comprehensively evaluated the long‐term prognostic value of the activation of the PIK3CA/AKT/mTOR pathway in primary and metastatic samples of a subset of our series.

Materials and methods

Patients and tumor samples

All patients had histologically confirmed ER‐positive, HER2‐negative advanced breast cancer and were treated with everolimus and endocrine therapy from January 2012 to April 2014 in two comprehensive cancer centers (Institut Curie, Paris, France and Institut Bergonié, Bordeaux, France). The population was previously described [3]. In brief, 123 patients were studied, 74 from Institut Curie and 49 from Institut Bergonié. All patients had a long history of advanced HR+/HER2− breast cancer, and received everolimus combined with endocrine therapy for metastatic progression. For the purpose of the present study, we gathered tumor tissue samples from 71 patients, including 44 primary tumors (62%) and 27 metastatic samples (38%). The study was approved by the French national data protection authority and our Institutional Review Board. This study followed the precepts of the Declaration of Helsinki and French laws concerning biomedical research. Per current French regulations, no written informed consent was required.

Molecular analyses

Targeted sequencing was performed with DNA extraction followed by gene mutation screening and Sanger sequencing. For DNA extraction, six tissue sections of 6‐μm thickness were obtained from formalin‐fixed paraffin–embedded (FFPE) tissues and a seventh tissue section stained with hematoxylin and eosin (H&E). The tumor‐rich areas were microdissected using a single‐use blade and the samples underwent proteinase K digestion in a rotating incubator at 56 °C for 3 days. DNA was extracted using NucleoSpin kit (Macherey‐Nalgen, Hoerdt, France) according to the supplier recommendations in two separate aliquots that were analyzed in parallel.

For gene mutation sequencing, high‐resolution melting (HRM) primers were designed to span the entire exons with product sizes under 200 base pairs. Primers were designed for PIK3CA (exons 10 and 21), AKT1 (exon 4), and ESR1 (exons 5 and 8) genes. The PCR for HRM and Sanger sequencing analysis were performed on a 384‐well plate in the presence of the fluorescent deoxyribonucleic acid (DNA) intercalating dye, LCGreen (Idaho Technology, Salt Lake City, UT, USA) in a LightCycler480 (Roche Diagnosis, Meylan, France). The reaction mixture in a 15‐μl final volume contained LCGreen, uridine diphosphate (UDP) glycosylase (Roche), and Roche master mix (Roche). The cycling and melting conditions were as follows: with an initial cycle of 10 min at 40 °C, 1 cycle of 95 °C for 10 min; 50 cycles of 95 °C for 10 s, 55–65 °C for 10 s, 72 °C for 30 s; 1 cycle of 97 °C for 1 min and a melt from 70 to 95 °C rising 0.2 °C per second. All samples were tested in duplicate. The HRM data were analyzed using GeneScan software (Roche). All samples including the wild type exons were plotted according to their melting profiles on the differential plot graph. Any difference of the horizon line based on the wild type sample was sequenced with Sanger sequencing.

For Sanger sequencing, the reaction mixture in a total of 50 μl was made using 1 μl of PCR products without first purification followed by a sequencing reaction with Big Dye Terminator v3.1 (Thermo Fisher, Courtaboeuf, France) according to the manufacturer's protocol. The sequencing products were purified with a Sephadex gel (GE Healthcare, Velizy‐Villacoublay, France) before running on a 3500 Genetic Analyser (Applied Biosystems, Foster City, CA, USA). The sequencing data were visualized using Finch TV (Geospiza, Inc., Seattle, WA, USA) with detection sensitivity of 10% mutated cells.

Immunohistochemical analyses

All analyses were centrally performed at Institut Curie. Following slide review, the most representative paraffin block was selected for immunohistochemical analysis (IHC). Three‐micrometer‐thick FFPE sections were mounted on silane‐coated slides. The immunohistochemical labeling procedure was based on the use of the BOND III automated system (Leica Biosystems, Newcastle, UK). Knowing that PTEN, INPP4B, or STK11 loss of function, and pS6 or p4EBP1 overexpression result in a deregulation of mTOR pathway and that these deregulations have been reported in many types of cancers (nine), we assessed for each case immunohistochemical staining with five different antibodies directed against inositol polyphosphate 4‐phosphatase type II (INPP4B), Thr70 phosphorylated 4EBP1 (p4EBP1), Ser235/236 phosphorylated S6 ribosomal protein (pS6), phosphatase and tensin homolog (PTEN), and serine/threonine kinase 11 (STK11). Details of antibodies, clones, and dilutions used for each single antibody are reported in supplementary material, Table S1. Visualization was achieved using diaminobenzidine (DAB), followed by counterstaining with hematoxylin (Bond Polymer Refine Detection). Positive and negative controls were included in each experiment. Pathologists who were blinded to genomic data interpreted the stains. Nuclear and cytoplasmic immunoreactivity was recorded separately. For each antibody, the stained slides were classified according to the percentage of positively stained tumor cells (0–100%) and the intensity of staining (0, absent; 1, weak; 2, moderate; and 3, strong), and a histoscore was derived, as the product of percentage and intensity (possible range: 0–300).

Statistical analyses

Results are reported with frequencies and proportions for categorical data and median and range [interquartile 25–75 range] for quantitative data. Quantitative data were compared with chi‐square test or Fisher's exact test. PFS was estimated using the Kaplan–Meier method and counted from the date of everolimus initiation to disease progression or death. Overall survival (OS) was estimated from the date of everolimus initiation to death of any cause. Univariate analyses were performed with the log‐rank test. For targeted sequencing (PIK3CA, AKT1, and ESR1), survival analyses were performed according to the presence or absence of a given mutation. For immunohistochemistry, survival analyses were performed according to the histoscore median value, or according to quartiles. Multivariate analyses were performed with Cox‐proportional hazards regression, with progression of disease or death as endpoint and using a forward variable selection procedure, based on the results of the univariate analyses. Final estimates were retained with a p value (Wald) <0.05. The statistical analyses were performed using MedCalc® Statistical Software version 20.2 (Ostend, Belgium), and with Prism 9 for Windows 64‐bit, version 9.5.1, GraphPad software, LLC (Boston, MA, USA).

Results

Patients and tumor samples

The characteristics of the population at the initiation of everolimus therapy are depicted in supplementary material, Table S2. The clinical and therapeutic features of the 71 patient subset (Figure 1) appear almost identical to those of the original 123 patient population [3], with regard either to clinical characteristics or to treatment history. Median age was 62 years [58–69]. All patients had a good performance status (Eastern Cooperative oncology Group ≤ 1) albeit with extensive prior therapies. As expected, most patients had polymetastatic disease with frequent (>50%) visceral involvement. In combination with everolimus, 68 of the 71 patients (95.8%) received an AI, and 3 received fulvestrant. Overall, clinical characteristics of the patients with primary tumor or metastatic samples did not significantly differ (supplementary material, Table S2). We only observed more frequent visceral disease in patients with metastatic samples (74.1% versus 36.4% in patients with primary tumor samples, p = 0.02), and conversely more frequent bone or distant lymph node metastases in patients whose only available sample was the primary tumor sample. The sampled metastatic sites included distant lymph nodes (31.0%), skin (24.1%), liver (20.7%), lung (10.3%), and miscellaneous sites.

Figure 1.

Flow chart of patients and samples included in this study.

Molecular analyses

Molecular analyses were performed on 71 tumors (Figure 1 and Table 1). Overall, mutations in the PIK3CA/AKT/mTOR pathway were observed in 32 patients (47.1%) including 11 patients (15.5%) with a PIK3CA exon 10 mutation, 17 patients (24.0%) with a PIK3CA exon 21 mutation, and 4 patients (5.6%) with an AKT1 exon 4 mutation. ESR1 mutations were observed in five patients (7.0%). Of note, the mutational status did not significantly differ between primary tumors and metastatic samples.

Table 1.

Molecular characterization of the PIK3CA/mTOR pathway by targeted sequencing and immunochemistry in primary tumor and metastasis samples

	Global population (71)	Primary tumor (44)	Metastasis (27)	p value*
Mutations	n (%)	n (%)	n (%)
PIK3CA pathway	32 (47.1%)	19 (43.2%)	13 (48.1%)	0.90
PIK3CA exon 21	17 (24.0%)	9 (20.5%)	8 (29.6%)	0.57
PIK3CA exon 10	11 (15.5%)	8 (18.2%)	3 (11.1%)	0.31
AKT1 exon 4	4 (5.6%)	2 (4.5%)	2 (7.4%)	0.71
ESR1	5 (7.0%)	2 (4.5%)	3 (11.1%)	0.37
Exon 5	2 (2.8%)	1 (2.3%)	1 (3.7%)	0.79
Exon 8	3 (4.2%)	1 (2.3%)	2 (7.4%)	0.36

	Global population (71)		Primary tumor (44)		Metastasis (27)		p value
IHC (histoscore)	n	Median † [IQR]	n	Median [IQR]	n	Median [IQR]
C‐p‐4EBP1 ‡	70	45.0 [5.0–80.0]	41	60.0 [8.75–120.0]	26	45.0 [0.0–70.0]	0.25
C‐p‐S6	71	10.0 [0.0–40.0]	41	5.0 [0.0–22.5]	26	20.0 [2.7–40.0]	0.03
C‐PTEN	71	100.0 [90.0–200.0]	41	120.0 [97.5–200.0]	27	100.0 [92.5–195.0]	0.40
C‐STK11	70	20.0 [10.0–60.0]	41	30.0 [10.0–80.0]	26	15.0 [5.0–30.0]	0.03
C‐INPP4B	70	0.0 [0.0–20.0]	41	5.0 [0.0–65.0]	26	0.0 [0.0–2.0]	0.01
N‐p‐4EBP1 §	25	0.0 [0.0–0.0]	24	0.0 [0.0–0.0]	0
N‐p‐S6	71	0 [0–0]	41	0 [0–0]	27	0 [0–0]	1.0
N‐PTEN	71	0 [0–0]	41	0 [0–0]	27	0 [0–0]	0.65
N‐STK11	71	0 [0–0]	41	0 [0–0]	27	0 [0–0]	0.41

^*Exploratory p values (primary versus metastatic).

^† Median [IQR 25–75].

^‡ C indicates cytoplasmic staining.

^§ N indicates nuclear staining.

Results of the cytoplasmic and nuclear staining of the PIK3CA/AKT/mTOR pathway proteins are summarized in Table 1. Nuclear staining was very limited, in contrast to cytoplasmic staining. As illustrated in Figure 2A, median cytoplasmic histoscores were 100 [interquartile range, IQR 90–200] for PTEN, 0 [IQR 0–20] for INPP‐4B, 20 [IQR 10–60] for STK11, 45 [IQR 5–80] for p4EBP1, and 10 [IQR 10–40] for pS6. Representative stained sections are shown in Figure 2B. No statistical differences between primary tumor group and metastasis tumor group were found between PTEN median histoscore (p = 0.40) or p4EBP1 median histoscore (p = 0.25), whereas INPP‐4B median histoscore (p = 0.01), STK11 median histoscore (p = 0.03), and pS6 median histoscore (p = 0.03) were different between primary tumors and metastatic tissues (Table 1). However, all differences in the cytoplasmic staining were in favor of greater activation of the PIK3CA/AKT/mTOR pathway in the metastatic tissue. Of note, low expression of PTEN or INPP4B (below median H‐score value) and mutations in the PIK3CA/AKT/mTOR pathway were associated with higher expression of p4EBP1 or pS6 with borderline significance (data not shown). There was no significant mutual exclusion of PTEN loss and mutations in the PIK3CA/AKT/mTOR pathway.

Figure 2.

Results of the immunohistochemical analysis of the PIK3CA pathway. (A) Box‐plots showing histoscore distributions of the components of the PIK3CA pathway. (B) Immunohistochemical staining of PTEN, p4EBP1, and pS6 (×200 magnification). The top images show cases lacking immunohistochemical evidence of PIK3CA/AKT/mTOR pathway activation: retained PTEN expression (top left) and absence of pS6 (top middle) or p4EBP1 (top right) expression. The bottom images show representative examples of loss of PTEN expression (bottom left), and high cytoplasmic expression of p4EBP1 (bottom middle) and pS6 (bottom right) by cancer cells.

Survival analyses

At data cutoff (January 2023), median follow‐up was 22 months. We observed no statistical differences in PFS and OS between patients with primary tumors or metastatic samples (data not shown). The median overall PFS was 6.6 months [5.3–7.9]. Details of the univariate analyses are reported in supplementary material, Table S3. The mutational status of the PIK3CA/mTOR pathway did not correlate with PFS. Conversely, activation of the pathway, as captured by cytoplasmic expression of the pathway proteins, had a prognostic impact. Specifically, low expression of INPP4B or PTEN (which negatively regulates the pathway when normally expressed), or the combined high expression of pS6 and/or p4EBP1 (flagging the activated state of the pathway) was significantly associated with a longer PFS as illustrated in Figure 3. By multivariate analysis (Table 2), the PTEN and the pS6/p4EBP1 histoscores retained a statistically significant independent prognostic value. The PFS was longer in patients with a pS6 or p4EBP1 histoscore greater than or equal to the median value (6.6 versus 3.7 months, p = 0.037) (Figure 3A), and was longer when the PTEN histoscore was inferior to the median value (7.1 versus 5.3 months, p = 0.007) (Figure 3B). The Harrell's C‐index for PFS was 0.596 (95% CI 0.527–0.665).

Figure 3.

Progression‐free survival. (A) Progression‐free survival with Kaplan–Meier curves in patients with pS6 or p4EBP1 histoscore ≥ median value (red), and with pS6 or p4EBP1 histoscore < median value (blue curve). (B) Progression‐free survival with Kaplan–Meier curves in patients with PTEN histoscore < median (red) and PTEN histoscore ≥ median (blue).

Table 2.

Progression free survival (PFS) and overall survival (OS) multivariate analyses

	Category	Hazard ratio	95% CI	p value	Median*	95% CI
PFS
PTEN histoscore	≥Median	1		0.02	5.3	4.1–6.7
	<Median	0.53	0.32–0.91		7.1	4.9–15.0
Either pS6 or p4EBP1 histoscore	≥Median	1		0.037	3.7	2.5–6.0
	<Median	0.46	0.22–0.96		6.6	5.3–7.9
OS
Visceral metastases	Yes	1		0.60	18.0	12.7–27.7
					27.6	20.7–35.2
	No	0.91	0.45–1.82
Bone‐only metastases	No	1		0.13	19.2	14.9–27.6
	Yes	0.51	0.22–1.21		34.1	20.8–50.3
PTEN histoscore	≥1st quartile	1		0.71	32.1	18.1–40.8
	<1st quartile	0.77	0.34–1.75		22.7	14.8–27.7
pS6 histoscore	<3rd quartile	1		0.038	19.3	9.3–27.6
	≥3rd quartile	0.37	0.14–0.94		27.6	20.8–34.1
Any mutation in the PIK3CA pathway	No	1		0.011	19.3	14.9–32.1
	Yes	0.40	0.19–0.82		27.6	15.4–46.9

^*Median refers to median duration of PFS or OS (months).

The median OS of the study population was 26.3 months [20.7–29.7]. Details of the univariate analysis are reported in supplementary material, Table S4. Despite a significant prognostic value by univariate analysis, clinical features such as bone‐only metastases or presence of visceral metastases were not retained in the multivariate models. Two important biological features were independently associated with a longer OS (Figure 4). First, the presence of an activating mutation in the PIK3CA/AKT/mTOR pathway (including an exon 10 or 21 PIK3CA mutation, or an exon 4 AKT1 mutation) was associated with longer OS [27.6 versus 19.3 months, hazard ration (HR) = 0.40, p = 0.011] (Figure 4A). Similarly, OS was longer in patients with pS6 in the upper quartile (27.6 versus 19.3 months, HR = 0.37, p = 0.038) (Figure 4B). The Harrell's C‐index for OS was 0.608 (95% CI 0.527–0.688).

Figure 4.

Overall survival. (A) Kaplan–Meier curves of overall survival in patients with any mutation in the PI3KCA/mTOR pathway (red) or without mutation (blue). (B) Kaplan–Meier curves of overall survival in patients with pS6 median histoscore ≥ 3rd quartile (red) or a median histoscore < 3rd quartile (blue).

Discussion

Advanced luminal breast cancer remains a noncurable disease despite recent clinically meaningful improvements with the introduction of CDK4/6 inhibitors in the therapeutic armamentarium [6]. Before the CDK4/6 inhibitor era, everolimus combined with alternative endocrine therapies had proven efficacy in the BOLERO‐2 pivotal trial [1], as well as in large multicenter real‐world studies [5, 9]. Even if its use has been hampered by several nonhematological but clinically limiting adverse events leading to the development of alternative dosing schedules [10], everolimus is one of the recommended therapeutic options in patients developing resistance to endocrine therapy combined with CDK4/6 inhibitors (https://www.esmo.org/living-guidelines/esmo-metastatic-breast-cancer-living-guideline and Reference [11]). So far, no clear predictive factor of everolimus efficacy has been identified that could help to better select patients benefiting from this therapy. It is most important to consider that endocrine therapy and CDK4/6 inhibitors may share resistance mechanisms including activation of the PIK3CA/AKT/mTOR pathway [12, 13, 14], thus providing a post hoc rationale for the use of everolimus‐based therapy in a resistant clinical setting. It has also been suggested that everolimus may reverse resistance to palbociclib [15]. Within this rapidly changing context, we examined in this report how clinical and molecular features may be associated with prognosis in a series of 71 patients with advanced luminal breast cancer treated with everolimus‐based therapy. The initial clinical cohort of 123 patients has been reported [3]. It consisted of heavily pretreated patients, who therefore could approximate the current clinical landscape where most of these patients would have been exposed to CDK4/6 inhibitors. The present subset of 71 patients was similar to the original cohort (supplementary material, Table S2).

The prognosis of patients with advanced breast cancer, as widely demonstrated in early stage breast cancer, relies on a combination of clinical and biological characteristics [16]. Probably due to lack of statistical power, we did not recapitulate the prognostic value of clinical features, such as the number of prior lines of chemotherapy or the metastatic burden, that we had initially observed [3]. However, and most interestingly, we observed that, using a supervised molecular approach, activation of the PIK3CA/AKT/mTOR pathway, as defined by key mutations or the level of expression of the pathway proteins, is tightly associated with prognosis with regard to both PFS and OS (Figures 3 and 4).

A high level of expression of pS6 or p4EBP1, the final effectors of the PIK3CA/AKT/mTOR pathway, has already been described as associated with response to everolimus, either at the preclinical [17] or at the clinical level [18]. The SAFIR‐TOR study was designed to prospectively validate the clinical utility of p4EBP1 expression as a predictive marker of everolimus efficacy [8]. This single‐arm prospective study included 150 patients with ER‐positive, HER2‐negative, metastatic breast cancer resistant to AIs. At inclusion, a metastatic site was sampled in order to perform IHC staining of p4EBP1, pS6K, pAKT1, PTEN, and LKB1, and genomic analyses targeting PIK3CA, AKT1, and ESR1. Treatment with an everolimus exemestane combination was administered until progression. In this study, patients whose tumors expressed p4EBP1 above the median level had a significantly longer PFS, but no impact of ESR1, PIK3CA, or AKT1 mutations was observed [8]. Conversely, a retrospective study based only on primary tumor tissue failed to identify any PIK3CA/AKT/mTOR pathway‐related prognostic factor, underlining the need for studies on metastatic tissue [19]. We used both primary tumor (n = 44; 62%) and metastasis tissue samples (n = 27; 38%), according to the availability. Of note, we used a larger proportion of metastatic samples than other similar studies [19, 20, 21], even if activation of the PIK3CA/AKT/mTOR pathway might be similar between primary tumor and metastatic tissue, notably with regard to PIK3CA or AKT1 mutations [16, 22].

More specifically, the impact of PIK3CA exon 10 mutations is controversial. Exon 10 mutations have already been described as a predictive biomarker of response to everolimus in the neoadjuvant setting [23] but not in advanced ER–positive, HER2‐negative breast cancer [18, 19, 20, 21]. In the BOLERO‐2 study, the PFS in the everolimus arm was maintained irrespective of PIK3CA genotypes as captured by cell‐free DNA sequencing [24]. Conversely, it is of note that an exploratory analysis suggested that patients with HER2‐positive advanced breast cancer and harboring tumors with PIK3CA mutations, PTEN loss, or an hyperactive PI3KCA/mTOR pathway could derive a PFS benefit from everolimus [25], suggesting that the predictive or prognostic value of specific mutations still has to be explored. Additionally, no impact of PTEN loss or AKT1 exon 4 mutation correlated to response to everolimus in the BOLERO‐2 study nor in the TAMRAD study [18, 19, 20, 21]. In the BOLERO‐2 study, the Y537S mutation in the ESR1 gene as detected in the cell‐free DNA may be a poor prognosis biomarker, but with no predictive value for sensitivity (or loss of sensitivity) to everolimus [26].

This study has several limitations. It is a limited retrospective series, and the molecular analyses have evaluated only the PIK3CA/AKT/mTOR pathway, thus precluding any larger biomarker analysis. Analysis of the metastatic tissue appears paramount. All these patients have been treated before the CDK4/6 inhibitor era, and new drugs precisely targeting PIKCA or AKT1 will soon enter the therapeutic landscape [27, 28], further questioning the role of everolimus in the treatment strategies. Although we conducted this study (January 2012 to April 2014) before the CDK inhibitor era, the present results may pertain to patients currently treated with CDK inhibitors because some mutations in the PIK3CA/AKT/MTOR pathway are also implicated in resistance to this therapy [29, 30, 31]. We acknowledge that another limitation of the present study is the lack of evaluation of CDK expression, which may be relevant in the present CDK inhibitor era [32].

In summary, our results suggest, in line with other studies using similar approaches, that everolimus may be preferentially beneficial in patients with advanced ER‐positive, HER2‐negative breast cancer harboring an activated state of the PIK3CA/AKT/mTOR pathway, as captured on tumor tissue by a combination of targeted sequencing and immunohistochemistry.

Author contributions statement

PC, IB and AV‐S conceived and designed the study. LD, GB, GD and AV‐S performed the pathological analyses. AS and IB performed the molecular analyses. Clinical data were acquired by all authors. HS, LD, IB, AV‐S, MD and PC analyzed and interpreted the data. The drafting of the article and its revision for important intellectual content were performed by all authors. All authors approved the final version.

Supporting information

Table S1. List of antibodies used for the immunohistochemical analyses

Table S2. Patient characteristics

Table S3. Progression‐free survival univariate analysis

Table S4. Overall survival univariate analysis

Data availability statement

Data are available from the corresponding author upon reasonable request.