Analysis of PIK3CA Mutations and Activation Pathways in Triple Negative Breast Cancer

Background Triple Negative Breast Cancer (TNBC) accounts for 12–24% of all breast carcinomas, and shows worse prognosis compared to other breast cancer subtypes. Molecular studies demonstrated that TNBCs are a heterogeneous group of tumors with different clinical and pathologic features, prognosis, genetic-molecular alterations and treatment responsivity. The PI3K/AKT is a major pathway involved in the regulation of cell survival and proliferation, and is the most frequently altered pathway in breast cancer, apparently with different biologic impact on specific cancer subtypes. The most common genetic abnormality is represented by PIK3CA gene activating mutations, with an overall frequency of 20–40%. The aims of our study were to investigate PIK3CA gene mutations on a large series of TNBC, to perform a wider analysis on genetic alterations involving PI3K/AKT and BRAF/RAS/MAPK pathways and to correlate the results with clinical-pathologic data. Materials and Methods PIK3CA mutation analysis was performed by using cobas® PIK3CA Mutation Test. EGFR, AKT1, BRAF, and KRAS genes were analyzed by sequencing. Immunohistochemistry was carried out to identify PTEN loss and to investigate for PI3K/AKT pathways components. Results PIK3CA mutations were detected in 23.7% of TNBC, whereas no mutations were identified in EGFR, AKT1, BRAF, and KRAS genes. Moreover, we observed PTEN loss in 11.3% of tumors. Deregulation of PI3K/AKT pathways was revealed by consistent activation of pAKT and p-p44/42 MAPK in all PIK3CA mutated TNBC. Conclusions Our data shows that PIK3CA mutations and PI3K/AKT pathway activation are common events in TNBC. A deeper investigation on specific TNBC genomic abnormalities might be helpful in order to select patients who would benefit from current targeted therapy strategies.


Introduction
Breast cancer (BC) is the most frequent malignant tumor in females and the commonest cause of cancer death among women worldwide [1]. Recently, our knowledge on BC biology has strongly been improved, with significant increase in personalized treatment options. Gene expression profiling studies have established the heterogeneous nature of BC, which might be considered as a collection of distinct "intrinsic" subtypes, based on specific genetic alterations involving different oncogenic pathways [2][3][4][5][6][7][8].
Triple Negative Breast Cancer (TNBC), which accounts for 12-24% of all breast carcinomas, is defined by the lack of expression of estrogen, progesterone receptors (ER, PgR) and

Materials and Methods
Our experimental study was approved by ASL Sassari Bioethical Committee, which also waived the need for written informed consent from the patients, according to the Italian legislation concerning the guidelines for the performance of retrospective observational studies; however, breast tissue samples were fully anonymized prior of any authors' access.
From representative formalin-fixed, paraffin-embedded (FFPE) specimens, 3μm-thick tissue sections were cut for haematoxylin and eosin stains (H & E) and immunohistochemical analysis. Additional consecutive sections were also obtained for genetic analysis. FFPE specimens from control group (non-TNBC tumors) were utilized only for PIK3CA mutational analysis.

Real-Time PCR to detect PIK3CA gene mutations
To detect PIK3CA mutations a Real-Time PCR procedure was used. TNBC and BC control group were processed for genomic DNA isolation using cobas1 DNA Sample Preparation Kit (Roche Mannheim, Germany) following the manufacturer's instructions. Briefly, deparaffinized 5 μm section, containing at least 10% of tumor cells, was used for the extraction process. The amount of genomic DNA mixture was spectrophotometrically determined (NanoDrop2000, Thermo Fisher Scientific, Waltham, MA USA) and adjusted to a fixed concentration to be added to the amplification/detection.
The PCR Real-Time cobas1 PIK3CA Mutation Test kit (Roche) uses a pool of primers divided into three different mixes for each samples and controls that define specific base-pair (bp) sequences that range from 85 to 155 bp in PIK3CA exons 1, 4, 7, 9, and 20. An additional primer pair, targeting a conserved 167 bp region in PIK3CA exon 3, provides a full process control. A derivative of Thermus species Z05-AS1 DNA-polymerase is utilized for amplification. Selective amplification of target nucleic acid from the sample is achieved in the cobas1 PIK3CA Mutation Test by the use of AmpErase (uracil-N-glycosylase) enzyme and deoxyuridine triphosphate (dUTP).
The target DNA was amplified and detected on the cobas z480 analyzer (Roche) using the RT-PCR-based amplification and detection reagents provided in the cobas1 PIK3CA Mutation Test kit (Roche). The validated instrument software of the cobas 4800 system uses a specific and tested algorithm for the interpretation of the results, through the automatic and standardized analysis of the specific kinetic of single curve, in order to guarantee the correct interpretation of the amplified curves. No operator mediated evaluation or interpretation is needed and/or possible.
Gene sequencing to detect EGFR, AKT1, BRAF and KRAS mutations Gene mutation analyses were performed on specific exons which are known to harbor the most frequent and significant mutations for each gene: exons 18, 19, 20, and 21 for EGFR, exon 2 for AKT1, exon 15 for BRAF and exons 2 and 3 for KRAS. Selected primers (Eurofins MWG, Synthesis GmbH, Munich, Germany) used for the amplification and sequencing reaction are summarized in Table 1.
Five 10 μm-thick consecutive sections from TNBC specimens were prepared, and tumors were macro-dissected with a scalpel blade under sterile conditions, using corresponding Haematoxylin & Eosin stained sections as a guide. DNA was extracted using the QIAamp DNA FFPE Tissue Kit (Qiagen, Hilden, Germany) in accordance with the manufacturer's instructions. To obtain genomic DNA 10μl of RNase A (20mg/ml, Rnase PureLink, Life Technologies, Carlsbad, CA, USA) were applied to the silica membrane to digest contaminating RNA. We assessed the quantity and the quality of nucleic acids spectrophotometrically, as described above. Gene sequencing analysis was executed as previously reported [26].

Evaluation of Immunohistochemical Staining
AR expression was interpreted as positive if at least 1% immunostained tumor nuclei were detected in the sample, according with ASCO/CAP recommendations for immunohistochemical testing of hormone receptors in BC [33]. EGFR was considered positive when 1% of neoplastic cells exhibited positivity, according to manufacturer's instructions; CK5/6 were considered positive when 5% of neoplastic cells exhibited positivity; moreover, the results were scored semi-quantitatively including intensity (0, negative; 1+, weak; 2+, moderate; 3+, strong).
p-AKT, p-p44/42 MAPK and PTEN staining were scored semiquantitatively based on staining intensity (0-3) and percentage of stained cells (0-100) using the histo-score (H-score; range 0-300) [34]. PTEN loss cut-off corresponding to the H-score 10 was used. In the same way, p-AKT and p-p44/42 MAPK with H-score 10 were defined as negative. A score ranging from 11 to 100 was considered as weakly positive (score 1), from 101 to 200 moderately positive (score 2), and from 201 to 300 strongly positive (score 3). Finally, subcellular localization of immunostaining was also assessed for each antibody in each positive tumor.
Subcellular localization of immunostaining has also been assessed for each positive tumor case.

Statistical analysis
Statistical analysis was carried out with STATAR13 (StataCorp, College Station, TX, USA). Shapiro-Wilk normality test was used to assess the parametric distribution of the quantitative variables. Median and inter-quartile range and frequencies were used to summarize quantitativeand qualitative variables, respectively. Statistical differences between individuals with and without PIK3CA mutations for quantitative and qualitative variables were evaluated performing the Mann-Whitney U test and Chi2 or Fisher's Exact Test when appropriated, respectively. Logistic regression analysis, both univariate and multivariate, was carried out to assess the association between mortality after 5 years of follow-up and the epidemiological, clinical, and molecular variables. The statistical significance was set-up at <0.05.

Clinic-pathologic features
One hundred and thirty-three primary BC were included in the study, specifically 97 TNBC and 36 non-TNBC. TNBC samples were characterized by ER, PR, and HER2 negativity, with ki67 proliferation index ranging from 6% to 90% of neoplastic cells. Patients' age ranged from 27 to 92 years (mean: 56), tumor size varied between 8 and 140 mm (mean: 30.5 mm). The clinical-pathologic data of TNBC included in this study are reported in Table 2.

Immunohistochemical analysis
Immunohistochemistry was performed to define TNBC subtypes and to analyze downstream signaling pathways.
The immunoreactivity for basal markers, EGFR and CK5/6 was reported as membranous or membranous-cytoplasmic. EGFR expression was appreciable in 69.1% of TNBC, with staining intensity ranging from 1+ to 3+, and the percentages of positive cells varying from 5% to 90%.

Mutational profiling and association analysis
No significant differences were detected between TNBC and non-TNBC PIK3CA mutational status (23.7% vs 33.3%, p = 0.262). The statistical differences between PIK3CA mutational status and standard clinical, pathological and biological features of TNBC were analyzed (Table 2). Briefly, PIK3CA mutations were significantly higher in older individuals (mean: 66 years vs 54 years; p = 0.006). However, no significant differences were obtained comparing other variables, such as tumor size, histologic type (although ductal carcinomas were prevalently represented) pT, pN, Stage, and OS. Interestingly, TNBC showing PIK3CA mutations were prevalently of lower grade (p = 0.03) with lower proliferation index (p = 0.004).
A statistically significant association was obtained between PIK3CA mutational status, AR expression (p = 0.002) and p-AKT expression (p = 0.0001). No statistical differences were observed between PIK3CA mutational status and immunohistochemical expression of EGFR, p-p44/42 MAPK, and PTEN, as summarized in Table 3.
A logistic regression analysis was carried out in order to assess the impact of established clinic-pathologic prognostic predictors on TNBC overall patient survival. Although lower Ki67 values were significantly associated with a better overall patient survival (p = 0.015), a multivariate model analysis did not confirm Ki67 value as an independent prognostic factor ( Table 4).

Discussion
Our study demonstrates that PIK3CA gene mutations and PI3K/AKT pathway activation are common events in TNBC, indicating a critical role of this pathway in TNBC pathogenesis. Several population-based studies have analyzed the association between PIK3CA mutations and different molecular subtypes of BC with variable results. PIK3CA abnormalities have commonly been related to hormone receptors and HER2 overexpression, and PIK3CA mutations are known to be highly associated with Luminal-A phenotypes [15,35,37,38]. Nevertheless, recent studies on a wide series of BC have shown that PIK3CA mutations are recognizable in TNBC, even with low frequencies [21,22,39]. Recently, Kriegsmann et al. demonstrated a higher frequency of PI3K pathway alterations, comprising mainly PIK3CA mutations (22.1%), in a large series of TNBC [23].
Our results support Kriegsmann et al. experience, demonstrating that PIK3CA mutations are recognizable in a higher percentage (23.7%) of TNBC than previously reported in the literature [17,18,[20][21][22]. Although histologic subtypes other than invasive ductal carcinoma are scarcely represented in our study, we could identify PIK3CA mutations in triple negative invasive lobular carcinoma, medullary carcinoma, and even in special variants of BC, as adenoid cystic carcinoma. These results underline the contribution of PIK3CA mutations in the biology of several histotypes of TNBC, confirming results of other authors [38,40].
To confirm the activating role of these mutations, we performed immunohistochemical analysis for PIK3CA downstream signaling pathways, such as pAKT and p-p44/42 MAPK, showing overexpression of these proteins in 100% of PIK3CA mutated TNBC. Interestingly, a strong association was identified between the presence of PIK3CA mutation and a higher score of pAKT (p = 0.0001), demonstrating the activating role of this genetic alteration in TNBC.
To perform a consistent evaluation of PI3K/AKT and BRAF/RAS/MAPK pathways, we analyzed other members of these pathways. In our experience, total loss of PTEN protein was detected in 11.3% of TNBC, all of these with basal-like phenotypes. pAKT and/or p-p44/42 MAPK were expressed in all TNBC with PTEN loss, according with the activating function of protein loss on these signaling pathways. Furthermore, PTEN loss was observed in association with both PIK3CA-mutated and PIK3CA-wild-type tumors. A very recent study by Millis et al. on a wide series of TNBC described PTEN loss in 66% of tumors, which could be explained by different immunohistochemical thresholds related to staining intensity and percentages of cells (0+ or 50%) [22].
Although PTEN loss and PIK3CA mutations have often been reported as mutually exclusive [37], it should be noted that in more recent studies these genetic events are described as concordant in BC patients [18]. We hypothesize that combined mutations might be the consequence of sequential waves of mutations during clonal cell expansion, but it is also possible that PTEN loss and PIK3CA mutation might be responsible for different pathologic events in tumor progression.
The correlation analysis between PIK3CA status and TNBC molecular subtypes showed a statistically significant association between PIK3CA mutations and Luminal Androgen Receptor (LAR) subtype. Furthermore, the highest percentage of PIK3CA mutations was obtained in TNBC basal-like variants, in accordance with a very recent study that confirmed the significant relation between PIK3CA mutations and the presence of basal markers in a series of 75 TNBC [24]. However, no significant association was appreciable between PIK3CA mutations and TNBC basal-like or non-basal-like subtypes (p = 0.258). Recently, the Cancer Genome Atlas Network has shown that PIK3CA mutations were common in Luminal and HER2-enriched tumors, whereas they were the 2 nd most common mutation in basal-like BC. Moreover, activation of PI3K/AKT pathway was the highest in this subgroup [8].
Taken together, our results demonstrate that at least 33% of TNBC have deregulated PI3K/ AKT pathways, making these pathways an attractive target for pharmacologic treatment and highlighting the importance of mutation profiling for individualized therapies. Indeed, there is a strong interest in developing rapid, reliable and sensitive methods which might be used for clinical routine detection of PIK3CA mutations in BC. In this study breast cancer samples (FFPE) were tested by Real-Time PCR. This approach is more sensitive than Sanger sequencing, having the ability to detect mutations at a level of greater than 5%, which are present only in a subgroup of cancer cells or in tumors with significant contamination. The higher frequencies of PIK3CA mutations detected in our and Kriegsmann et al. studies are related to more sensitive methodologies, widening the opportunities for TNBC patients to be appropriately selected for individualized targeted therapy strategies. Moreover, the primers used in the cobas1 PIK3CA Mutation Kit were designed to avoid false positive results due to presence of PIK3CA pseudogene sequence mismatch at the end of exon 9. [47]. We applied cobas1 PIK3CA Mutation Kit also for the analysis of a small series of non-TNBC tumors, showing 33.3% of mutations rate, namely 40% in Luminal-A-like subtypes, 25% in Luminal-B-like subtypes, and 33.3% in Her2 positive subtypes. Although the small number of tumors could not allow to reach any statistical significance, the results were similar to those in the literature, even in this small sized sample [8].
Conflicting results about the prognostic meaning of PIK3CA mutations are evident from the literature. Some authors reported better overall survival (OS) and disease free survival (DFS) in patients harboring these genetic alterations, compared to patients without mutations [18,35,36,38,48,49]. In contrast, other authors reported that PIK3CA mutations in BC patients were associated with poor clinical outcome [50][51][52].
Our data on prognostic significance of PIK3CA mutations, restricted to 65 patients due to incomplete 5-years OS information, could not demonstrate any significant association between PIK3CA mutations and OS. The association between clinical-pathological data and PIK3CA mutations on all the TNBC showed that higher grades and higher proliferation indexes were significantly associated with wild-type PIK3CA tumors.
Several drugs targeting multiple levels of the PI3K network, such as PI3K, AKT, and mTOR, have been progressively taken into consideration in clinical trials for BC [53]. However, the use of PI3K/AKT pathway inhibitors as single-agent therapies has been demonstrated minimally effective in some diseases, according to the complexity of the PI3K/AKT pathway, with activation of multiple feedback and cross-talk mechanisms that might explain this drug resistance. For this reason, PI3K pathway inhibitors are being tested in human trials in combination with HER2, MEK and ER inhibitors, supposing that the simultaneous targeting of these escape mechanisms will lead to clinically success [54][55][56][57].
Recent results of Janku et al. have shown the clinical significance of highlighting PIK3CA molecular status in BC patients, since tumors with PIK3CA mutations treated with PI3K/AKT/ mTOR inhibitors showed a response rate of 30% matched to 10% in wild-type PIK3CA tumors; the response rate in wild-type PIK3CA tumors was comparable to previous reports (4%-11%) when patients were treated on phase I trials without molecular selection [58]. Moreover, Fink et al. found that gene expression profiles for TNBC subtype definition were poor predictors of response to kinase inhibitors, whereas high sensitivity was reported when treatment options were based on specific genetic abnormalities of tumor, such as PTEN loss or PIK3CA mutations [59].
In conclusion, our study contributes to defining the complexity of TNBC category and, considering the high percentage of genetic alterations involving PI3K/AKT pathways identified in our series, supports the necessity to subclassify TNBC on the basis of their specific genomic abnormalities, in order to appropriately select patients who would more likely benefit from current targeted therapeutic strategies. Moreover, our results highlight the issue concerning the application of reliable and sensitive methodologies to detect PIK3CA mutations in routine practice.