Introduction

Recent studies of gene expression profiling led to the identification of breast cancer subtypes based on common molecular features [1, 2]. Triple-negative breast cancers (TNBC) do not express estrogen and progesterone receptors (ER/PgR) and HER2. They show typical morphological features and have been classified into intrinsic basal-like or claudin-low subtypes [3, 4].

TNBC represents the group of tumors with a worse prognosis. They are typically high grade, although low-grade tumors can rarely occur and these may have a better prognosis [5, 6]. Therefore, TNBC can be considered a heterogeneous entity including different subtypes [7] and several attempts have been made to establish a molecular classification of TNBC based on the results of both transcriptomic and genomic studies [8,9,10]. Recent independent studies have demonstrated that TNBC can be sub-classified either into six subtypes (basal-like I, basal-like II, mesenchymal, mesenchymal stem-like, immunomodulatory, and luminal androgen receptor) [9] or four subtypes (luminal androgen receptor, mesenchymal, basal-like immune-suppressed, and basal-like immune-activated) [10]. Following the integrative analysis of gene expression and genome-wide copy number alterations (CNAs) by the METABRIC consortium [11], TNBC were characterized by complex patterns of copy number gains and losses throughout the genome.

Genomic profiles of TNBC have recently been defined using whole exome sequencing. TNBC has a predominance of TP53 alterations that can be present in up to 80% of the cases, and a large set of mutated genes occurring with minor mutation frequencies [11,12,13]. TNBC shows mutations in PIK3CA (7%) [12], mutation or loss of phosphatase and tensin homolog (PTEN) (35%), and loss of inositol polyphosphate-4-phosphatase type II (INPP4B) (30%) [14]. Although several potentially actionable molecular alterations (e.g., PI3K/mTOR or RAS/RAF/MEK) have been found in TNBC, none have been confirmed as a ‘driver alteration.’

MALDI (Matrix-Assisted Laser Desorption/ionization)—TOF (Time of Flight) mass spectrometry is a versatile tool for high-performance DNA analysis and was selected to perform mutation analysis [15]. Pre-designed assays are commercially available for comprehensive mutation screening of solid tumors; in particular the OncoCarta Panel v1.0 used in the present study is able to screen more than 230 potentially druggable mutations across 19 common oncogenes (ABL1, AKT1, AKT2, BRAF, CDK4, EGFR, ERBB2, FGFR1, FGFR3, FLT3, JAK2, KIT, MET, HRAS, KRAS, NRAS, PDGFRA, PIK3CA, RET). These genes belong to different functional pathways including PI3K pathway, receptor tyrosine kinase, and cell cycle-metabolic group. Moreover, FGFR1 and FGFR3 codify for mitogenic signaling molecules that play a role in angiogenesis and cell migration and the mutational status of these genes may offer information about the response to antiangiogenic therapy.

The aim of this study was to explore the prognostic and predictive relationship between the presence of a selected set of somatic mutations in TNBC and the clinical outcome within the International Breast Cancer Study Group (IBCSG) Trial 22-00, to obtain a comprehensive description of the features of this important subset of tumors.

Patients and methods

IBCSG trial 22-00 population

IBCSG Trial 22-00 is a multicenter, randomized, adjuvant phase III clinical trial assessing the efficacy of low-dose metronomic CM maintenance chemotherapy (CM: cyclophosphamide 50 mg/day orally continuously for 1 year and methotrexate 2.5 mg/twice a day orally, days 1 and 2 of every week for 1 year) to no maintenance chemotherapy (No-CM), following breast cancer surgery and standard adjuvant chemotherapy [16]. The trial enrolled 1086 patients with estrogen (ER) and progesterone (PgR) receptor-negative tumors, and any HER2 and nodal status. The IBCSG Ethics Committee and ethics committees at each center approved the study, and all patients provided written informed consent. The trial is registered at clinicaltrials.gov (NCT00022516).

Assay/genotype methods

From a representative formalin-fixed and paraffin-embedded tissues (FFPE) block, 5-micron-thick histological sections were obtained for DNA extraction (QIAamp DNA FFPE Tissue Kit Qiagen, Hilden, D). Mutational analysis was performed with OncoCarta Panle v1.0 using the MassARRAY System based on MALDI-TOF mass spectrometry (Sequenom, San Diego, CA, USA), according to the manufacturer’s protocols. Genotyping by this method relies on the principle that mutant and wild-type alleles for a given point mutation produce single-allele base extension reaction products of different mass resolved by mass spectrometry. The OncoCarta Panel v1.0 includes Amplification and extension primers for the analysis of 230 mutations in 19 genes (ABL1, AKT1, AKT2, BRAF, CDK4, EGFR, ERBB2, FGFR1, FGFR3, FLT3, JAK2, KIT, MET, HRAS, KRAS, NRAS, PDGFRA, PIK3CA, RET). Assays were conducted without knowledge of clinical outcome.

Sampling design

A matched case–control sampling selected from 570 patients enrolled in IBCSG Trial 22-00 who were known to have TNBC according to the local assessment of ER, PgR, and HER2. The cases were defined as having an invasive local, regional, or distant recurrence of breast cancer. Controls without breast cancer recurrence were matched 2:1 using the following criteria: (i) nodal status (N0, 1–3 N+, 4–9 N+, 10 N+); (ii) tumor size (≤ 2 cm, > 2 cm); (iii) age (< 40, 40–55, 56–69, ≥ 70); (iv) menopausal status (premenopausal; postmenopausal); (v) year of randomization, by using a publically available macro (%gmatch; http://www.mayo.edu/research/departments-divisions/department-health-sciences-research/division-biomedical-statistics-informatics/software/locally-written-sas-macros). Of the selected 191 patients (67 cases, 124 controls), 135 had mutational assessment (49 cases and 86 controls), after excluding 56 patients with unavailable FFPE tumor blocks or insufficient material for assessment (Fig. 1). Patients with mutational assessment consented to use of their tumor tissue for research purposes, and the project was approved by the IBCSG biological protocols working group.

Fig. 1
figure 1

Summary of study patient inclusion

Full size image

Statistical analysis

Conditional logistic regression was used to assess the association of breast cancer recurrence (case) with mutation status (at least one mutation vs. wild-type (WT)); the strata variable was the matching indicator for the case and controls. Treatment assignment (CM maintenance vs. No-CM)-by-mutation status interaction was included in the model to first assess the potentially differential treatment effect (i.e., predictive effect). If there was no evidence of interaction, the model without interaction term was used to assess the prognostic mutation effect on clinical outcome. Similarly, an exploratory analysis was also performed to assess the association of recurrence with PIK3CA pathway mutation status (at least one mutation from PIK3CA, AKT1 and AKT2: yes vs. no).

Because the statistical analysis of this type of case–control study was based on a conditional logistic regression, the only patients from the assessable population contributing to the analysis were those for which there was at least one control matched to the case. The analysis included 109 patients, comprising 44 cases and 65 controls (23 cases had one matched control; 21 cases 2 matched controls; 5 cases did not have any matched controls). Thus, 65 case–control pairs were informative strata in the conditional logistic regression analysis (Fig. 1). For a matched case–control design with 49 cases and 2 controls per case, if the frequency of at least one mutation was 22% in the control group and the correlation of mutation between matched pairs was 0.2, the study had 80% power to detect an odds ratio of 3.2 (two-sided α = 0.05) [17].

Results

The characteristics of the 570-patient cohort with locally assessed TNBC are summarized in Table 1, separately for the 135 assessable patients with mutation status and the 435 patients without mutation status assessed. The characteristics were mostly consistent, although assessable patients more frequently were node-positive as compared to those not assessed (57.0% vs. 38.4%, respectively) and more frequently had tumor size > 2 cm (63.0% vs. 52.4%, respectively).

Table 1 Characteristics of patients with locally assessed TNBC for case–control sampling, according to whether mutation status was assessed
Full size table

Among the 135 assessable patients, 37 (27.4%) patients had at least one mutation, affecting 12 of the 19 targeted genes, and 98 (72.6%) patients had wild-type status. Five patients had more than one mutation (Table 2). Among the specific mutations, PIK3CA was the most common (18/135; 13.3%), two of which co-occurred either with mutation in BRAF or with mutation in both AKT2 and PDGFRA. Mutations in BRAF, KIT, and PDGFRA each occurred in 4 patients (3.0%), while mutations in AKT1 occurred in 3 patients. When looking at pathway mutations (specifically: PIK3CA + AKT1 + AKT2), 22 (16%) patients had at least one of the PIK3CA, AKT1, or AKT2 mutations. The remaining mutations were detected in a single patient (0.7%) in each of 3 genes (ABL1, FGFR3, FLT3), or detected in 2 patients (1.5%) in each of 4 genes (AKT2, EGFR, ERBB2, MET). The occurrence of each mutation is shown in Fig. 2.

Table 2 Mutation status by treatment assignment and by clinical outcome status, among 135 patients with TNBC tumors assessable for mutations
Full size table
Fig. 2
figure 2

The Oncoprint plot showing the incidence of each mutation occurred

Full size image

Using conditional logistic regression, a subset of 65 matched pairs (109 patients) formed the informative strata and contributed to the analysis of assessing the association of mutational status with breast cancer recurrence (Table 3). Patients with at least one mutation in the 19 targeted genes had increased odds of recurrence versus those with wild-type status (odds ratio (OR) 2.28; 95% CI 0.88–5.92; p = 0.09). Similar results were observed when assessing the association of at least one of PIK3CA, AKT1, and AKT2 mutation with the recurrence (ORR 2.60; 95% CI 0.62–6.41). There was no evidence of a predictive relevance of the mutation status with use or non-use of CM maintenance therapy (mutation-by-treatment interaction p = 0.73 or 0.48, respectively). For patients assigned to CM maintenance, having a targeted mutation was associated with increased odds of breast cancer recurrence (OR 1.98; 95% CI 0.57–6.92), as it was for patients assigned to No-CM (OR 2.62; 95% CI 0.76–9.07).

Table 3 Conditional logistic regression assessing association of breast cancer recurrence with targeted mutation status overall and according to treatment assignment, among the subset of 109 patients with TNBC tumors assessable for mutations that formed 65 matched case–control pairs
Full size table

Discussion

The mutational analysis performed with OncoCarta panel v1.0 in 135 patients with locally assessed TNBC within trial IBCSG 22-00 identified 27.4% patients with at least one mutation in the 19 selected genes. Among patients with a breast cancer recurrence (cases), 36.7% had at least one mutation identified. Since this is a matched case–control study, the patients selected for the analysis were guaranteed to have higher-risk features (and be enrolled earlier in the trial) than the other patients, as by definition this design enriched for patients who had a recurrence (cases) and patients with matched disease characteristics.

In similar studies, the presence of a potentially targetable mutation was found in about 36% of cases with TNBC [12], although a different method for identifying mutations was used (high-throughput RNA sequencing—RNA-seq). About 40% of breast cancers contained single rare mutations that could be potential therapeutic targets [18]. Key driver mutations are thought to provide a selective advantage to a number of cells, facilitating their clonal expansion. A key finding from recent next-generation sequencing (NGS) studies was that each tumor contains a dominant clone (> 50% of cancer cells) that has a mutational profile different from the other sub-clones [19].

In our population of patients with TNBC, PIK3CA mutation was the most frequent mutation, being identified in 13.3% of selected patients’ tumors. These results were comparable to those reported in the literature showing PIK3CA mutations in 10–25% of breast cancers [12] (http://cancer.sanger.ac.uk/cancergenome/projects/cosmic/). Other groups reported that this mutation could be present in 8.3% of breast cancers classified as basal-like [20] and that patients with TNBC had a rate of PIK3CA mutation of approximately 13–15% [21] (http://cancer.sanger.ac.uk/cancergenome/projects/cosmic/) as compared with a rate of about 30% in ER/PgR-positive breast cancers.

In addition, the frequency of mutation of the BRAF gene (3.0%) observed in the present study population was similar to other TNBC series as well as in all subtypes (http://cancer.sanger.ac.uk/cancergenome/projects/cosmic/). The frequency of mutation of KIT and PDGFRA genes (3% each) observed was higher than reported in other TNBC series (0–1.4%) (http://cancer.sanger.ac.uk/cancergenome/projects/cosmic/) though the numbers are small.

A recent trial by Zhu et al. showed that although 42.1% of TNBC expressed KIT protein, only 1 activating mutation was detected in the series; the expression of KIT in TNBC lacked correlation with activating mutations in both KIT and PDGFRA genes [22]. Nevertheless, it might be worth investigating the potential effect of dasatinib on TNBCs expressing C-kit and PDGFRA in clinical trials.

Mutation of the AKT1 gene was 2.2% in the present series, about 1% in other series and 4% in all subtypes (http://cancer.sanger.ac.uk/cancergenome/projects/cosmic/). Likewise, point mutations have been reported in multiple components of a single signaling pathway (i.e., PIK3CA, PTEN, AKT1, as molecular components of the PI3K signaling pathway) indicating the potential relevance of specific signaling pathways as a therapeutic target in breast cancer [23]. In the present series, PIK3CA pathway mutations (specifically: PIK3CA + AKT1 + AKT2) were found in 22 (16%) patients. Inhibitors of the PI3K/AKT/mTOR pathway, frequently de-regulated in TNBC, are acquiring a growing interest and several inhibitors are in preclinical development or already in early phase clinical trials [24, 25].

Mutations in the tested oncogenes were not associated with breast cancer recurrence in this TNBC subset of patients. However, our analysis suggested that the presence of at least one targeted mutation may be prognostic in patients with TNBC, having observed a numerically increased odds of recurrence (OR 2.28; 95% CI 0.88–5.92), and similar results were evident when assessing the association of at least one of PIK3CA, AKT1, and AKT2 mutation with the recurrence (ORR 2.60; 95% CI 0.62–6.41). Nevertheless, evaluation in larger datasets is warranted. Evidence from the present trial showed that mutational status either single or pathway mutation, failed to be predictive for benefit of CM maintenance, as the test of interaction was not statistically significant (p = 0.73 and p = 0.48, respectively). This may be in part attributed to the restricted sample size that represents one of the limits of the study. The prognostic role of mutation status was extensively investigated either for PIK3CA mutations [26] or for TP53 status as a prognostic or predictive marker in TNBC [23]. PIK3CA mutations role of cell-free DNA (cfDNA) in early-stage TNBC patients was recently associated with a more favorable prognosis. After a median follow-up of 54.4 months, the presence of PIK3CA mutations of cfDNA was significantly associated with relapse-free survival (p = 0.0072) and breast cancer-specific survival (p = 0.016) [27]. These discordant results—if compared with the data of the present study—may be due to the differences between PIK3CA mutation detection in plasma and in tumor tissue DNA. In fact, it has been hypothesized that this discordance might be a result of the problem of tumor heterogeneity [28].

Indeed, a recent pooled analysis confirmed that the presence of a PIK3CA mutation represents an independent negative prognostic factor [29]. However, to make clarity on this subject, more studies focusing on specific exons mutations would be useful [30].

Although TP53 mutation/expression status did not show any significant implications in terms of prognosis for patients with TNBC, the discrepancy between mutation and expression of TP53 may indicate poor prognosis in TNBC patients [31].

Another limit of the study was that TP53 as well as other suppressor genes commonly mutated in TNBC including PTEN, PIK3R1, TSC1/2, or BRCA1/2 are not included in the panel of genes evaluated, so we were unable to assess the prognostic value of these genes in our series. However, we selected a pre-validated assay (OncoCarta Panel v1.0) that is enriched in oncogenes to prioritize the investigation of mutations that might also be clinically actionable, although less frequently reported in TNBCs. Recent results from the SAFIR-01 study (NCT01414933), which performed a comparative genomic hybridization (CGH) array and Sanger sequencing of PIK3CA and AKT1 in patients with metastatic breast cancer, reported a targetable genomic alteration in 195 (46%) patients. Therapy was therefore personalized in 55 (13%) patients [32] and of those only 9% had an objective response. These results do not appear encouraging, but the study population was heavily pretreated. Very low response rates are frequently seen in patients with breast cancer previously treated with targeted agents [33]. Response rates are similar when pretreated patients receive treatment of physician choice in 3rd or more line of treatment [34]. However, the clinical relevance of many of these alterations remains to be established. In fact, it is increasingly recognized that molecular profiling of advanced disease could help elucidate the biological basis of distant recurrence and resistance to therapy [35]. Notably, the AURORA (Aiming to Understand the Molecular Aberrations in Metastatic Breast Cancer; NCT02102165) study is another trial going in this direction, recruiting breast cancer patients across Europe. Both primary and metastatic sites are analyzed for genomic alterations, and patients are offered participation in clinical trials utilizing targeted therapeutic agents.

The AURORA pilot study was recently published reporting results in 41 patients with the primary objective of investigating the feasibility in four European recruitment sites and central pathological and sequencing facilities. Results were encouraging and at least one clinically actionable mutation was identified in 73% of patients. This pilot study demonstrated that next-generation genomic techniques are ready for international molecular screening programs in routine clinical settings, although some technical challenges remain to be addressed [36].

In the present trial, the mutational profile was performed only in the primary tumor and not repeated in the recurrence, as this analysis was retrospectively done in a subset of patients participating in the IBCSG 22-00 trial and the matching tissue for recurrence was available only for a limited number of patients. Nevertheless, the identification of critical pathways involved in carcinogenesis, metastasis, and drug resistance, as well as actionable mutations, that could emerge from these trials fuels the dream of “personalized medicine,” in which the molecular landscape of an individual’s cancer will inform clinical decision making, particularly the selection of “tailored” targeted therapies.

Finally, the results of the present study have identified a consistent subgroup of patients with TNBC and PIK3CA mutations. Since there is contradictory information about the prognostic role of PIK3CA mutation in this subset of patients, it may be worth contributing to a potential pooled analysis to elucidate the role of PI3KCA mutations/alterations in TNBC.