Introduction
Female breast cancer is the most commonly diagnosed cancer with more than one million new cases every year and breast cancer is one of the leading causes of cancer-related death in women (1). The phosphatidylinositol-3-kinase (PI3K) pathway is a major signaling pathway in cells and is involved in essential cell processes such as metabolism, survival, proliferation, growth and motility (2). Dysregulation of the PI3K pathway occurs in a large variety of human cancers (3) and has been proven to be implicated in breast cancer development and progression (4). PI3K converts phosphatidylinositol-4,5-bisphosphate (PIP2) to phosphatidylinositol-3,4,5-triphosphate (PIP3). The phosphatase and tensin homolog (PTEN) antagonizes and negatively regulates PI3K by converting PIP3 to PIP2 (5).
The PI3K/AKT/mTOR pathway appears to have a major role in the response to treatment and in the development of resistance to anticancer drugs. Overactivation of the PI3K pathway downstream of human epidermal growth factor receptors (HER) can be driven by mutations of PI3K, an enzyme from the lipid kinase family involved in cell signaling. These activating mutations occur mainly on p110α, the catalytic subunit of PI3K encoded by the PIK3CA (phosphoinositide-3-kinase, catalytic, α polypeptide) gene located on chromosome 3. PIK3CA mutations are present in 25% of breast carcinomas, and the most common activating mutations occur on exons 9 and 20 according to the COSMIC database (Catalogue Of Somatic Mutations in Cancer Database, Wellcome Trust Genome Campus, Hinxton, Cambridge; accessed June 2012; http://www.sanger.ac.uk). More precisely, E542K (c.1624G→A, p.Glu542Lys), E545K (c.1633G→A, p.Glu545Lys) in exon 9 and H1047R (c.3140A→G, p.His1047Arg), H1047L (c.3140A→T, p.His1047Leu), represent more than 90% of the mutations encountered in breast carcinomas. These four mutations have been shown to have an oncogenic role in breast cancers (6–9).
Recent studies have shown that PI3K may be implicated in the resistance of breast cancers to anti-estrogen therapy agents (10,11), anti-HER2 tyrosine kinase inhibitor (lapatinib) (12) and anti-HER2 monoclonal antibody (trastuzumab) (5,11,13). Mutations of PIK3CA and loss of the PTEN protein are keys factors in the development of resistance to these drugs (14–16). Moreover, lapatinib and trastuzumab resistance can occur in HER2-amplified breast cancers bearing a PIK3CA mutation (5,14,15). HER2 overexpression in breast cancers is present in 15–25% of tumors (17,18). Approximately 75% of the breast cancers express estrogen receptors and/or progesterone receptors and a relationship between anti-estrogen resistance and activation of the PI3K pathway has recently been found (11). These new findings imply that the PI3K pathway may be an important target for novel targeted therapies.
New anti-PI3K and anti-mTOR drugs are currently under development (19,20) for breast cancer treatment. Activation of the PI3K/AKT pathway and overexpression of PI3K may play a major role in the use of new therapeutic schemes, and identification of PIK3CA mutations could be a major biomarker for predicting the response to these new therapies.
In light of these issues, there is a huge interest in developing rapid, reliable and sensitive methods that can be used for clinical routine detection of PIK3CA mutations in breast tumors. In the present study, we used a polymerase chain reaction (PCR)-high resolution melting assay (HRM) and a PCR-amplification refractory mutation system (ARMS) to analyze alcohol-formalin-acetic acid (AFA)-fixed paraffin-embedded breast tumor specimens. PCR-HRM is a cost-effective post-PCR method that enables the identification of alterations in single nucleotides, i.e., mutations through the analysis of thermal denaturation of double-stranded DNA. PCR-ARMS is a powerful mutation-specific real-time PCR-based technique combining ARMS and a bi-functional fluorescent probe/primer molecule (Scorpion).
The present study evaluated the relationship between PIK3CA exon 9 and 20 mutations and conventional clinicopathological criteria and compared the sensitivity of the two techniques by examining the correlation of the results achieved using both methods. The final goal was to validate a double technique approach, ensuring a cost-effective, rapid process yielding a high quality level of analysis, according to the recommendations of the French National Cancer Institute (INCa) for clinical routine analysis of mutations in tumors in a treatment-choosing process. In such a context, PCR-HRM could be proposed to determine the PIK3CA mutational status yielding binary results (mutated or wild-type) and PCR-ARMS-Scorpion to accurately identify the four main hotspot mutations of PIK3CA.
Patients and methods
Population
One hundred and forty-nine invasive breast carcinoma tumor specimens, from patients diagnosed between 2008 and 2009, were retrospectively included in this study. All specimens were collected as AFA-fixed paraffin-embedded tissues from our institutional Biobank. The tumor characteristics of this population were consistent with literature data regarding mean age at diagnosis, histological type (ductal and lobular), Scarff-Bloom-Richardson (SBR) grade as well as hormone receptor and HER2 status (Tables I and II). According to the Biobank procedure of our institute, all women involved in this study were informed that their tumor samples might be used for research purposes and had the opportunity to decline. No opposition was expressed.
DNA extraction
For each tumor specimen, hematoxylin and eosin slide analysis was conducted by a pathologist to ensure a minimum of 20% tumor tissue content as recommended in previous studies (21). Selected areas were macrodissected and 5 10-μm serial sections were cut from each paraffin block and collected in RNAase DNAase-free vials (SafeSeal Microcentrifuge Tubes, Sorenson Biosciences, Salt Lake City, UT, USA). Paraffin was removed by extraction with toluene (VWR BDH Prolabo, Fontenay Sous Bois, France) and centrifuged. DNA isolation was performed using the QIAamp DNA FFPE tissue kit (Qiagen, Courtaboeuf, France) protocol. The pellet was washed with ethanol, centrifuged and resuspended with 180 μl of tissue lysis buffer (ATL buffer; Qiagen) and 20 μl of proteinase K (Qiagen). The sample was then gently mixed, incubated at 56°C for 1 h and at 90°C for 1 h under agitation. DNA was extracted with MinElute Columns (Qiagen) as recommended by the manufacturer. The nucleic acids were eluted in a volume of 100 μl. Final concentration of eluates, ranging from 33.6 to 729.0 ng/ml, were suitable for PCR-HRM and PCR-ARMS analyses (Table III). DNA extracts from cell lines bearing E542K (Cal51), E545K (MCF7), H1047R (HCT116) and H1047R (SUM159PT) PI3-kinase mutations were used as positive controls. DNA extracted from the MDA231 cell line was used as a wild-type negative control. DNA quality was controlled using agarose gel electrophoresis with GAPDH (glyceraldehyde-3-phosphate dehydrogenase) as the control housekeeping gene. Only DNA with no degradation was used. DNA concentrations were determined using a Bio-photometer (Eppendorf, Hamburg, Germany).
PCR-HRM
HRM analysis was performed using the LightCycler 480® Real-Time PCR system (Roche Diagnostics, Meylan, France) and the LightCycler 480 HRM Master kit (Roche Diagnostics) in 384-well plates (Roche Diagnostics). Twenty micrograms of DNA was amplified in a final volume of 20 μl. All data and melting curves were analyzed using LightCycler SW v. 1.5.0.39 software (Roche Diagnostics). One mix was prepared for each exon. For each sample, 10 μl of Master Mix (Roche Diagnostics), 2.8 μl of MgCl2 25 mM, PCR-quality grade water and 1 μl of the primers (forward and reverse) were added. Eighteen microliters of mix was added to each well, and 2 μl of the sample was used for the analysis. One set of primers were used for each of the PIK3CA exons 9 and 20. All primers were designed as previously described (22).
PCR-HRM was divided into different phases: a phase of pre-incubation (10 min at 95°C) was followed by 45 cycles of classic PCR (10 sec at 95°C, a temperature decrease from 60 to 54°C by 0.5°C/cycle in 15 sec and, finally, 10 sec at 72°C). The PCR phase was followed by the high resolution melting phase which consisted of 1 min at 95°C, 1 min at 40°C and a temperature increase by 0.2°C/sec from 65 to 95°C. A cooling phase of 1 min at 40°C was finally performed.
PCR-ARMS
ARMS analysis was performed using the LightCycler 480 Real-Time PCR system (Roche Diagnostics) in 384-well plates. Eighty micrograms of DNA was amplified in a final volume of 20 μl. Data and fluorescence curves were analy’ed using LightCycler SW v. 1.5.0.39 software. All primers were designed as previously described (23). One mix was made for each tested mutation and all samples were proceeded as simplex. For each sample, 0.06 μl of Hot Diamond Taq polymerase (Eurogentec, Angers, France) was added together with 2 μl of reaction buffer 10X (Eurogentec), 3.2 μl of MgCl2 25 mM (Eurogentec), 0.4 μl of dNTP 10 mM (Eurogentec), PCR-quality water, 0.8 μl of ARMS primers 6.25 μM (Eurogentec) and 0.8 μl of Scorpion primers 6.25 μM (ATD Bio, Southampton, UK). Eighteen microliters of mix was added to each well, and 2 μl of the sample was used for the analysis. The mix for exon 9 mutations contained ARMS control primers for exon 15, specific ARMS primers, respectively, of E542K and E545K mutations and exon 9 and 15 specific Scorpion primers. The mix for exon 20 mutations contained ARMS control primers for exon 15, specific ARMS primers, respectively, of H1047R and H1047L mutations and exon 20 and 15 specific Scorpion primers.
Sensitivity
The sensitivity of PCR-HRM and PCR-ARMS was evaluated by mixing mutated and wild-type DNA from the cell lines at 50, 25, 10, 5, 2, 1 and 0.5% ratios.
Statistical analysis
The significance of the concordance of mutation detection using the two methods was assessed using κ statistics. κ>0.8 was considered as indicative of significance to conclude that both methods provide similar results. The χ2 test was also used to compare mutation frequencies with those obtained from the literature. χ2 and Fisher’s exact tests were used to test for differences between classes of patients and tumors based on clinical, pathological and biological characteristics. Limit of statistical significance was set at P<0.05.
Results
Mutation analysis using PCR-HRM
One hundred and eighteen specimens were analyzed using PCR-HRM (Table I). PIK3CA mutations (exons 9 and 20) were detected in 30 (27.5%) of the specimens. No correlation was found regarding patient age (≤50 or >50 years), ductal or luminal type, estrogen and progesterone status (alone or combined as hormonal receptor status), HER2 status, and for the four subtypes identified as HER2+/HR+, HER2+/HR−, HER2−/HR+ and HER2−/HR− (triple negative).
PIK3CA mutations were found to be correlated with SBR grade (P=0.050) with a lower mutation rate noted in the highest grades. Regarding the frequency of exon 9 and 20 mutations, a low rate of exon 9 mutations was observed in SBR grade III tumors (P=0.025) while no difference was observed for the exon 20 mutation rate. Exon 9 mutations were found to be more frequent in SBR grade I tumors while exon 20 mutations were predominantly observed in SBR grade II and III tumors.
Mutation analysis using PCR-ARMS
One hundred and forty-nine specimens were analyzed using PCR-ARMS (Table II). PIK3CA mutations (exons 9 and 20) were detected in 27 (18.1%) specimens. No difference was found regarding patient age (≤50 or >50 years), ductal or luminal type, estrogen and progesterone status (alone or combined as hormonal receptor status), HER2 status, and for the four subtypes (HER2+/HR+, HER2+/HR−, HER2−/HR+ and HER2−/HR−, i.e., triple negative). As with PCR-HRM, PIK3CA mutations were found to be related with SBR grade (P=0.004) with a lower mutation rate in the highest grades. Regarding the frequency of exon 9 and 20 mutations, a lower rate of exon 9 and exon 20 mutations was observed in SBR grade III tumors (P=0.009). Again, exon 9 mutations were found to be predominant in SBR grade I tumors (5/7) while exon 20 mutations were predominantly observed in SBR grade II and III tumors (13/20).
Comparative analysis of PCR-HRM and PCR-ARMS
One hundred and two breast tumor samples (Table III) were analyzed using both PCR-HRM and PCR-ARMS. PIK3CA mutations were detected (Table IV) in 28 tumors (27.5%) when PCR-HRM was used and 23 (22.5%) when PCR-ARMS was used.
Among the 28 mutated tumors (Table IV), PCR-HRM results showed that 13 (46.4%) carried a mutation on exons 9 and 15 (53.6%) on exon 20.
Among the 23 samples in which a mutation was detected using PCR-ARMS assay (Table IV), 9 (39.1%) carried a mutation in exon 9, identified as c.1624G→A in 6 cases (26.1%) and as c.1633G→A in the 3 others (13.0%). Among the remaining 14 (60.9%) specimens identified as mutated in exon 20, 10 (43.5%) carried the c.3140A→G and 4 (17.4%) the c.3140A→T mutation. No sample was found with mutations in both exons 9 and 20 with any of the assays.
Sensitivity
The analytical sensitivity of each assay was evaluated from dilution of DNA extracted from PIK3CA-mutated cell lines into DNA from PIK3CA wild-type cell lines (MDA 231) from 0.5 to 50%. PCR-HRM was able to discriminate a dilution corresponding to 5% of exon 9 mutated DNA and 10% of exon 20 mutated DNA (Fig. 1, Table VI). PCR-ARMS was able to discriminate a DNA dilution corresponding to 0.5% of c.1624G→A, 0.5% of c.1633G→A, 0.5% of c.3140A→G and 0.5% of c.3140A→T mutated DNA (Fig. 2, Table VI).
Results obtained using PCR-ARMS and PCR-HRM (Table V), were found to be statistically comparable (κ=0.845, P<0.001). Contingency table is shown in Table III.
No statistical difference in the frequencies of PIK3CA mutations was found between results achieved with any of the two assays and data from Sanger database.
Discussion
Recent studies have shown that PIK3CA mutations play a major role in resistance to trastuzumab or lapatinib (5) or to hormonal therapy (11,13) of breast carcinomas and could represent a potent response predictive marker for PI3-kinase- and mTOR-targeted therapies (24) and be used as a treatment-choosing parameter. PIK3CA mutations are mostly located within exons 9 and 20 and four hotspots (c.1624G→A, c.1633G→A, c.3140A→G and c.3140A→T) represent more than 90% of all mutations. Although the prognostic value of PIK3CA mutations in breast cancer remains controversial, in a recent evaluation of 2587 breast cancers cases from 12 independent studies, Dumont et al(24) reported a more favorable clinical outcome in patients with PIK3CA mutated tumors and that improved prognosis may pertain only to patients with a mutation in the kinase domain of p110α and to post-menopausal women with estrogen-positive cancers.
The relative prognostic value of exon 9 vs. 20 mutations also remains controversial. Barbareschi et al(25), found that exon 9 mutations have a negative prognostic value while exon 20 mutations were associated with favorable outcome while Lai et al(26) reported exon 20 mutations as associated with poor prognosis. Furthermore, Lerma et al(27) reported a decrease in survival in PIK3CA exon 20-mutated and HER2-positive patients. This study was retrospective and was based on a small number of cases (only 6 patients in the exon 20-mutated PIK3CA and HER2-positive group) and was never confirmed prospectively. Finally, in a larger series of patients, Cizkova et al(28) did not find any difference between exon 9- and exon 20-mutated tumors regarding metastasis-free survival. Using immunohistochemistry, Aleskandarany et al(29) reported that PIK3CA protein expression in invasive breast cancer was associated with poor prognosis.
The frequencies of PIK3CA mutations reported here, with PCR-HRM and PCR-ARMS, respectively, are fully consistent with COSMIC database values and previously published results reporting 18–40% of PIK3CA mutations in breast tumors (7,26,30–33).
In the present series, with a relatively limited number of cases, no significant correlation was observed between PIK3CA mutations and several clinical and histopathological criteria. Concerning the age at diagnosis, a higher PIK3CA mutation rate was noted in older patients (34) that was not observed in other series (25,26,28). The same is true for the histological type of breast cancer, when comparing ductal and lobular types. Some studies reported PIK3CA mutations to be more frequently observed in ductal than in lobular carcinomas (25,28,34) while others did not (35).
Although it has been reported that PIK3CA mutations are related to hormone receptor and HER2 status (28,34), this issue still remains controversial since no correlation between PIK3CA mutations and hormonal status has been observed (25,36,37) similarly to the results of the present study, or limited to estrogen receptor status, but not to progesterone receptor status (35). Similar findings have been found concerning the correlation between PIK3CA mutations and HER2 status (28,34) but again are controversial in other series (25,35) or observed when HER2 is determined using immunohistochemistry but not FISH (37). A higher frequency of PIK3CA mutation has also been noted in low SBR grade tumors similarly to the observation in the present series and consistently with other studies (28,34) but contrary with others (35).
All these discrepancies among the different studies could probably be explained by the relatively limited number of cases that were analyzed and the different techniques that were used for PIK3CA mutation analysis with different sensitivity limits. This also probably explains why a large range of mutation rates were reported. This point clearly justifies that consideration should be given to the validation and especially the endpoints and sensitivity limits of the assays that are routinely used.
Combining PCR-HRM and PCR-ARMS assays, several discrepancies were observed between the two techniques in 6 (5.9%) samples. All were identified as PIK3CA wild-type using PCR-ARMS and mutated using PCR-HRM; 5 samples with mutations on exon 9 and 1 sample with mutation on exon 20. Because our PCR-ARMS assay was designed to specifically, but only, identify the four main PIK3CA mutations (c.1624G→A, c.1633G→A, c.3140A→G and c.3140A→T), it is obvious that the 6 discordant cases should bear other mutations on exon 9 or 20, that were only detected using PCR-HRM which is an exon-specific method, able to detect all the somatic mutations on the entire exon 9 and 20. Exon 9 and 20 mutations, not located in the four main hotspots, were reported to represent approximately 10% of all exon 9 and 20 mutations in breast cancers (28). Our data were consistent with these values. In the present series, this rate of PIK3CA mutations, differing from the four main mutations identified by PCR-ARMS was found to be consistent with data recently reported in breast cancer (28).
Direct Sanger sequencing of these samples could confirm the hypothesis that they may carry another mutation but the lack of sensitivity of Sanger sequencing (approximately 20% mutant DNA) could be not discriminatory between a false-positive with HRM or a false-negative with sequencing (38).
The PCR-HRM and PCR-ARMS assays used here are highly sensitive, able to detect as low as 5% of mutated DNA for exon 9 (2.0 ng) and 10% for exon 20 (4.0 ng) and 0.5% (0.8 ng) of mutated DNA. The PCR-ARMS assay is more sensitive than pyrosequencing-based assay as previously described (38) and which was able to detect 5% of mutated DNA and even higher than another PCR-ARMS assay as previously reported (5 ng) (23). This high sensitivity could be explained by the high specificity of the primers used for this technique.
High sensitivity assays are required for routine analysis of mutations in clinical specimens. Recent results (39) achieved in colon cancer using KRAS mutation analysis, reported that sequencing was not sensitive enough to provide clinically relevant results yielding two many false-negative results. The authors recommended the use of more sensitive techniques with a detection limit approximating 1–2% to overcome this problem.
In conclusion, the present study highlighted the potential of PCR-HRM- and PCR-ARMS-based assays for the evaluation of the PIK3CA mutation status in breast cancers. No correlation was observed with patient age at diagnosis, histological type, hormone receptor and HER2 status. PIK3CA exon 9 and 20 mutations were found to be related to Scarff-Bloom-Richardson (SBR) grade with a lower rate of mutations and a higher frequency of exon 9 mutations in SBRI and exon 20 mutations in SBRII/III tumors. No difference was observed in the frequency of the two different mutations screened for each exon in any subcategory. Thus, we propose to use a combination of both assays, with a screening of full exon 9 and 20 using PCR-HRM and further identification of the four main mutations using PCR-ARMS. Following analysis of our data, using such a procedure would have led to only one false-negative result (<1%) which is quite satisfactory. Combining these two assays should represent a cost-effective rapid procedure that provides highly reliable results that are fully consistent with the attempts of practitioners in view of personalizing therapy for invasive breast cancers.