PIK3CA-activating mutations and chemotherapy sensitivity in stage II–III breast cancer

The study population consisted of 140 patients who participated in a pharmacogenomic predictive marker discovery study at the University of Texas M. D. Anderson Cancer Center (MDACC) [17]. During this research, patients were asked to undergo pretreatment fine needle aspiration (performed with a 23- or 25-gauge needle) of the primary breast tumor. Cells from two or three passes were collected into vials containing 1 mL of RNAlater™ solution (Ambion, Inc., Austin, TX, USA) and stored at -80°C. All patients subsequently received 6 months of preoperative chemotherapy: 63 patients (45%) received six courses of 5-fluoruracil, doxorubicin (or epirubicin), and cyclophosphamide (FAC or FEC, respectively) chemotherapy, and 77 patients (55%) received 12 weekly courses of paclitaxel followed by four courses of 5-fluoruracil, doxorubicin (or epirubicin), and cyclophosphamide (TFAC or TFEC, respectively). None of these patients received preoperative treatment with trastuzumab, lapatinib, or endocrine therapy. All patients underwent modified radical mastectomy or lumpectomy and sentinel node dissection after completion of chemotherapy. All patients with ER-positive tumors subsequently received adjuvant endocrine therapy. Each patient gave informed consent to allow molecular analysis of her tumor, and this study was approved by the institutional review board of the MDACC. Patient characteristics are summarized in Table 1.

ER expression status and progesterone receptor (PR) expression status were assessed by immunohistochemistry (IHC) (6F11; Novocastra Laboratories Ltd., Newcastle, UK) and human epidermal growth receptor 2 (HER2) status was assessed by either fluorescence in situ hybridization (FISH) or IHC as part of routine clinical care. ER positivity and PR positivity were defined as greater than 10% positive tumor cells with nuclear staining. HER2 positivity was defined as either HER2 gene amplification on FISH analysis (>2.0 CYP16/HER2 gene copy number ratio) or 3+ signal on IHC evaluation. Nuclear grade was assessed using modified Black's nuclear grading system. Pathological response was determined at the time of surgery by microscopic examination of the excised tumor and lymph nodes. Pathological complete response (pCR) was defined as no residual invasive cancer in either tumor or lymph nodes as opposed to residual disease (RD). Cases with in situ carcinoma in the absence of an invasive component were also included among the cases with pCR [18]. Cases with residual cancer (RD) represent a continuum of responses and it has long been recognized that the larger the residual cancer after preoperative chemotherapy, the worse the prognosis. We recently developed a method to quantify residual invasive cancer after preoperative chemotherapy on a continuous scale. This method combines the largest diameter of the invasive tumor, the percentage cellularity of the tumor, the number of lymph nodes involved, and the largest diameter of the nodal involvement into a residual cancer burden (RCB) score [19, 20]. The RCB score correlates with survival and also can be used to define four distinct pathological response categories: RCB-0 (same as pCR), RCB-I (near pCR), RCB-II (moderate residual cancer), and RCB-III (extensive residual cancer). These RCB categories are predictive of long-term survival; patients who achieve RCB-I pathological response have overall and disease-free survival rates similar to those of patients achieving pCR (that is, RCB-0) whereas patients with RCB-III have a very poor prognosis, particularly if they have ER-negative disease [20].

DNA was extracted from the flow-through of the RNA extraction step performed with a Qiagen RNEasy Mini Kit (#74104; Qiagen Inc., Valencia, CA, USA) using a Qiagen DNA extraction kit (#69504; Qiagen Inc.) according to the manufacturer's instructions. DNA concentration and purity were determined using a NanoDrop ND-1000 Spectrometer (NanoDrop Technologies, Wilmington, DE, USA).

Sequences for all annotated exons and adjacent intronic sequences containing the kinase domain of the PIK3CA gene were extracted from the Celera (Rockville, MD, USA) [21] or public [22] draft human genome sequences. Primers for polymerase chain reaction (PCR) amplification and sequencing were designed using the Primer3 program [23] and were synthesized by MWG (High Point, NC, USA) or Integrated DNA Technologies, Inc. (Coralville, IA, USA). PCR amplification and PIK3CA sequencing were performed using a 384-capillary automated sequencing apparatus (Spectrumedix, State College, PA, USA). Sequence traces were assembled and analyzed to identify potential genomic alterations using the Mutation Surveyor software package (SoftGenetics, LLC, State College, PA, USA). Primer sequences and conditions for PCR amplification and sequencing have been reported previously [7, 24]. Exon-specific and sequencing primers were synthesized by Invitrogen Corporation (Carlsbad, CA, USA). Purified PCR products were sequenced using a BigDye® Terminator version 3.1 Cycle Sequencing Kit (Applied Biosystems, Foster City, CA, USA) and analyzed with a 3730 ABI capillary electrophoresis system. Mutational analysis was carried out in the laboratory of author AB at the University of Torino [24].

The correlation between PIK3CA mutation status and dichotomous clinical/pathological parameters was examined by means of the chi-square test. ER, PR, and HER2 receptor expression status (positive versus negative), nuclear grade (1/2 versus 3), and lymph node status (negative versus positive) were considered as dichotomous variables. Tumor size (T0–T4) and patient ethnicity (Asian, Black, Hispanic, and Caucasian) were treated as categorical variables, and patient age was treated as a continuous variable. Pathological response was examined as both a dichotomous variable comparing pCR versus all RD and as an ordinal categorical variable (RCB-0, -I, -II, and -III). The associations between continuous variables and PIK3CA mutation status were determined using the unequal variance t test. A P value of less than 0.05 was considered significant.

The mutational status of the PIK3CA gene was assessed in all 140 tumors by direct sequencing of the gene regions encoding the helical domain (exon 9) and the catalytic domain (exon 20) of the PIK3CA gene. Tumor DNA was used from needle aspiration biopsy material that contains 75% to 90% cancer cells. One hundred seventeen tumors (83.6%) had the wild-type PIK3CA gene and 23 patients had an activating mutation in the PIK3CA gene (16.4%). Among the cases with a PIK3CA mutation, 12 had a missense mutation in exon 9 (8 E545K type, 3 E542K type and 1 Q546R) and 11 cases had a mutation in exon 20 (all but 2 were H1047R). Table 2 lists all of the detected mutations. We also examined mutations in exon 1 but no mutation was found in any of the cases.

When all of the cases were considered together, PIK3CA mutation was significantly associated with lymph node-negative status; 52% of mutant cases were node-negative compared with 25% among the wild-type cases (P = 0.012). There was also a trend for increased frequency of PIK3CA mutations in older women. The median age of patients with a PIK3CA mutation was 56 years compared with 51 years for the wild-type cases (P = 0.0535). No other clinical or pathological factor was significantly associated with PIK3CA mutation status (Table 3). In a multivariate model that included patient ethnicity, tumor grade (1/2 versus 3), tumor size, nodal stage, ER, PR, and HER2 status, patient age as well as response to chemotherapy, nodal status remained independently associated with PIK3CA mutation (P = 0.029).

We also examined the association between clinical and pathological parameters and PIK3CA mutation status in ER-negative (n = 62) and ER-positive (n = 78) tumors separately. No significant correlation was found between any clinical variable and PIK3CA mutation status among the ER-negative tumors. In contrast, among the ER-positive tumors, PIK3CA mutation status was significantly and inversely associated with nodal status. Patients with ER-positive tumors who were also positive for PIK3CA mutation had a higher incidence of node-negative disease (53% versus 22%; P = 0.025). No other clinical/pathological factor was associated with PIK3CA status in patients with ER-positive tumors (Table 4).

We examined the correlation between PIK3CA mutation status and response to chemotherapy in all cases and after stratification by ER status. When all of the cases were considered together, there was no difference in pCR rate (pCR = extreme chemotherapy sensitivity) among the PIK3CA mutant (pCR = 18%) and wild-type (pCR = 17%) cases (Table 3). In ER-positive tumors, the pCR rates were 8% and 13% (P = 0.62) in tumors with wild-type and mutant PIK3CA, respectively. In ER-negative tumors, the pCR rates were 28% and 29% (P = 1.0) for the wild-type and mutant cases, respectively (Table 4). Next, we examined pCR rates by type of chemotherapy and PIK3CA mutation status. Sixty-three patients received neoadjuvant FAC/FEC chemotherapy and the pCR rates were 6% and 8% for the wild-type and mutant cases, respectively (P = 1.0). Seventy-seven patients received neoadjuvant TFAC/TFEC chemotherapy and the pCR rates were 27% and 30% for the wild-type and mutant cases, respectively (P = 1.0) (Table 5).

It has been suggested that mutations in exon 9 may have different functional consequences than mutations in exon 20; therefore, we also tested the association between mutation type and response to chemotherapy [25]. There was no difference in pCR rates associated with mutation in either exon individually. However, in correlation analysis, nodal stage was associated with PIK3CA mutation status only for those in exon 9, with patients harboring an exon 9 mutation having an increased incidence of node-negative disease (66.7%) compared with patients with wild-type or other mutation types (24.8%; P = 0.023). Mutations in exon 20 were not significantly associated with any clinical or pathological parameter (Table 6).

RCB response category was available for 106 patients and this provided an opportunity to correlate PIK3CA mutation with graded pathological response. We compared PIK3CA mutation frequency in all four RCB categories and also in the two extreme response groups: RCB-0/I (highly chemotherapy-sensitive cases) versus RCB-III (highly chemotherapy-resistant tumors). No significant association was found between RCB response categories and PIK3CA mutation status in either analysis (P = 0.121 and 0.166, respectively) (Table 3). Even after stratification for ER status, chemotherapy regimen, and type of mutation, no significant association was found between PIK3CA mutation status and response to therapy (Tables 4, 5, 6).