Low-Frequency Mutational Heterogeneity of Invasive Ductal Carcinoma Subtypes: Information to Direct Precision Oncology

Information regarding the role of low-frequency hotspot cancer-driver mutations (CDMs) in breast carcinogenesis and therapeutic response is limited. Using the sensitive and quantitative Allele-specific Competitor Blocker PCR (ACB-PCR) approach, mutant fractions (MFs) of six CDMs (PIK3CA H1047R and E545K, KRAS G12D and G12V, HRAS G12D, and BRAF V600E) were quantified in invasive ductal carcinomas (IDCs; including ~20 samples per subtype). Measurable levels (i.e., ≥ 1 × 10−5, the lowest ACB-PCR standard employed) of the PIK3CA H1047R, PIK3CA E545K, KRAS G12D, KRAS G12V, HRAS G12D, and BRAF V600E mutations were observed in 34/81 (42%), 29/81 (36%), 51/81 (63%), 9/81 (11%), 70/81 (86%), and 48/81 (59%) of IDCs, respectively. Correlation analysis using available clinicopathological information revealed that PIK3CA H1047R and BRAF V600E MFs correlate positively with maximum tumor dimension. Analysis of IDC subtypes revealed minor mutant subpopulations of critical genes in the MAP kinase pathway (KRAS, HRAS, and BRAF) were prevalent across IDC subtypes. Few triple-negative breast cancers (TNBCs) had appreciable levels of PIK3CA mutation, suggesting that individuals with TNBC may be less responsive to inhibitors of the PI3K/AKT/mTOR pathway. These results suggest that low-frequency hotspot CDMs contribute significantly to the intertumoral and intratumoral genetic heterogeneity of IDCs, which has the potential to impact precision oncology approaches.


Introduction
Breast cancer is a heterogeneous disease, presenting with a spectrum of clinical features that impact prognosis and clinical response to treatment. Classically, breast tumors are characterized by the presence or absence of a hormone receptor [HR; estrogen receptor (ER) or progesterone receptor (PR)) and/or the overexpression of human epidermal growth factor receptor-2 (HER2)]. Breast tumors that are HR+ and/or HER2+ account for 80-85% of breast cancer cases. The remaining 15-20% of breast cancer cases are classified as triple-negative breast cancer (TNBC), which are defined by the lack of ER, PR, and HER2 overexpression. It is now standard care to evaluate new breast cancers for the presence of these markers [1-3].
(genetically and/or epigenetically). Furthermore, acquisition of subsequent CDMs, along with clonal expansion/contraction and possible recruitment of other transformed cells, culminates in the extensive intratumoral genetic heterogeneity observed in IDCs. Most CDMs quantified were at levels below the sensitivity of standard mutation detection methodologies, suggesting the potential for undetected mutant subpopulations to impact therapeutic response in the breast may be grossly underestimated.
Here, we extended our analysis of low-frequency CDMs in breast cancer subtypes, including TNBC. Specifically, we quantified six hotspot CDMs (PIK3CA H1047R, PIK3CA E545K, KRAS G12D, KRAS G12V, HRAS G12D, and BRAF V600E) in 81 IDCs (approximately 20 samples for each of the four breast cancer subtypes), using the sensitive and quantitative ACB-PCR approach. Mutant fractions (MFs) of the six CDMs were quantified in IDCs, analyzed by breast cancer subtype, and compared to levels measured in normal breast tissue [7,11].

Cancer-Driver Gene MFs in IDC Subtypes
A comparison of the distribution of measurements for PIK3CA H1047R, PIK3CA E545K, KRAS G12D, KRAS G12V, HRAS G12D, and BRAF V600E MFs in all IDCs for the four subtypes is given in Figure 4, along with MF measurements previously reported in normal breast (collected from individuals who died from causes unrelated to cancer or diseases affecting the breast [7,11]). Significant differences in the proportion of MFs above the level of accurate quantification were observed between normal breast and IDCs for PIK3CA H1047R, BRAF V600E, and HRAS G12D. Only the BRAF V600E mutation showed a significant increase compared to normal breast (Fisher's exact test, p = 0.0049). For PIK3CA H1047R, and HRAS G12D, the proportion of MFs above the level of accurate quantitation were significantly lower when compared to normal breast (p = 0.0005, and 0.0235, respectively).  [7,11] and four subtypes of IDC. Measurements in the shaded area are below the level of accurate ACB-PCR quantification (1 × 10 −5 ). * Significant compared to normal breast. † Significant difference between IDC subtypes (Fisher's exact test; p > 0.05). Few significant differences in PIK3CA, KRAS, HRAS, and BRAF MFs were observed between normal breast and IDC subtypes. Fisher's exact test was used to test for difference in the number of MFs above the limit of detection (LOD) between pairwise groups of interest. While the geometric  [7,11] and four subtypes of IDC. Measurements in the shaded area are below the level of accurate ACB-PCR quantification (1 × 10 −5 ). * Significant compared to normal breast. † Significant difference between IDC subtypes (Fisher's exact test; p > 0.05). Few significant differences in PIK3CA, KRAS, HRAS, and BRAF MFs were observed between normal breast and IDC subtypes. Fisher's exact test was used to test for difference in the number of MFs above the limit of detection (LOD) between pairwise groups of interest. While the geometric mean MF of PIK3CA H1047R for every subtype was more than 10-fold lower than that measured in normal breast (8.03 × 10 −4 [7]), only the HR−/HER2+ and HR−/HER2− (TNBC) subtypes were significantly different (p = 0.0441 and p = 0.0003, respectively). The proportions of KRAS G12D MFs above the LOD were significantly different between HR+/HER2− and normal breast (p = 0.002). There were also significant differences between the HR+/HER2− subtype and the HR+/HER2+ and HR−/HER2+ subtypes (p = 0.0099 and 0.0046, respectively). For KRAS G12V, the geometric mean MF in the HR−/HER2− and HR+/HER2− subtypes were 14-to 40-fold less than that found in normal breast (1.23 × 10 −5 [7]). However, significance was only reached in the HR−/HER2− subtype (p = 0.0177). BRAF V600E MFs were quantified and found to be greater in all four subtypes of IDCs as compared to normal breast [7]. However, only the HR+/HER2− subtype reached statistical significance (p = 0.0001) when compared to normal breast. In addition, BRAF V600E demonstrated significant difference when comparing HR+/HER2− to the HR+/HER2+ and HR−/HER2+ subtypes (p = 0.01 and 0.0015, respectively). No significant differences in PIK3CA E545K and HRAS G12D MFs were observed between normal breast ( [7] and [9], respectively) and the IDC subtypes, or among each subtype.

Cancer-Driver Gene MFs in IDCs Based on HR and HER2 Status
A comparison between cancer-driver MFs in IDCs that do or do not express HR and HER2 showed few significant differences. In HER2+ IDCs, the proportion of KRAS G12D and G12V MFs above the LOD were significantly greater compared to HER− tumors (p = 0.0012 and 0.0291, respectively). In contrast, in HER2− IDCs, a greater proportion of BRAF V600E MFs above the LOD were found compared to HER2+ IDCs (p = 0.0014). A moderate negative association in KRAS G12D and G12V, and BRAF V600E mutation was found in HR+ versus HR− IDCs, with a greater proportion of KRAS G12D MFs above the LOD in HR− IDCs, while BRAF V600E MFs above the LOD were observed in greater proportions in IDCs that are HR+, though the results were not significant.

Correlation Analyses
Exploratory correlation analyses using Spearman's correlation coefficient were performed between CDMs, and between CDMs and clinicopathological features in IDCs and IDC subtypes. Additional studies are needed to confirm the association between various relationships.

Correlation Analyses of CDMs in IDCs and IDC Subtypes
Correlation analyses revealed a few significant monotonic relationships between specific cancer-driver MFs in breast IDCs (weak to moderate in strength). Correlation analyses excluded MF measurements below our limit of accurate quantification (MFs < 1 × 10 −5 ), thus limiting the number of samples analyzed for some CDMs (i.e., KRAS G12V), and, in some cases, resulted in insufficient numbers for analysis. The results are presented in a correlation matrix (Figure S1A-E; includes p and r values for each). For all IDCs ( Figure S1A; n = 81), no significant correlation was observed between any two CDMs measured in this study. Correlation analysis within the individual subtypes revealed a positive association between PIK3CA H1047R and HRAS G12D and PIK3CA E545K mutations in the HR+/HER2+ subtype ( Figure S1B). In the HR−/HER2+ subtype, HRAS G12D mutation was inversely associated with KRAS G12D mutation ( Figure S1D). In the HR+/HER2− and HR−/HER2− (TNBC) subtypes, no significant correlations were observed between any CDMs; however, the correlation between KRAS G12D and BRAF V600E mutations in the HR−/HER2− subtype may warrant further investigation (r = 0.75, p = 0.0663; Figure S1E).

Correlation Analyses of CDMs in IDCs Based on HR and HER2 Status
Correlation analyses of PIK3CA, KRAS, HRAS, and BRAF CDMs in IDCs that do or do not express HR and HER2 showed only a few weak, but significant, monotonic relationships (Figure S1F-I). No significant associations were observed in HR+ IDCs, however, in HR− IDCs, a positive association was observed between BRAF V600E and HRAS G12D MFs ( Figure S1G). In HER2+ IDCs, negative associations included KRAS G12D and HRAS G12D MFs, as well as KRAS G12D and PIK3CA E545K MFs ( Figure S1H). In the HER2− IDCs, a positive association between BRAF V600E and KRAS G12V MFs were found ( Figure S1I).

Correlation Analyses of CDMs and Clinicopathological Features of IDCs and IDC Subtypes
Using the available clinicopathological information for all IDCs in this study, correlation analyses between age, clinical stage, nodal involvement, menopausal status, maximum dimension, and histological grade, and the PIK3CA, KRAS, HRAS, and BRAF CDM levels measured by ACB-PCR were performed (Table 4). In all IDCs, for example, correlation analyses found PIK3CA H1047R MF was correlated positively with donor age and menopausal status (pre-menopausal versus post-menopausal (r = 0.48, p ≤ 0.0001; and r = 0.74, p = 0.0001, respectively)). PIK3CA H1047R and BRAF V600E MFs were correlated positively with maximum tumor dimension (r = 0.46, p ≤ 0.01; and r = 0.47, p ≤ 0.001, respectively), while KRAS G12V MF was correlated negatively with maximum tumor dimension (r = −0.65), although this did not reach the level of statistical significance (Table 4, Figure 5; using only samples with a MF ≥ 1 × 10 −5 ).

Discussion
The identification of oncogenic mutations and the use of that information in the care and management of cancer patients has become an important area of clinical investigation, driven by the precision medicine paradigm. While the genetics of familial breast cancer is well studied, remarkably few quantitative analyses have been done with respect to somatic mutations involved in sporadic breast cancer. Women diagnosed with TNBC face a poor prognosis and less effective treatment options than women with other breast cancer subtypes [19][20][21]. Improved understanding of the molecular genetic underpinnings of TNBC, as compared with other subtypes of breast cancer, may lead to better predictions of tumor behavior and treatment response. To elucidate the roles of low-frequency CDMs in breast cancer, a sensitive and quantitative approach called ACB-PCR was employed to characterize the frequency and distribution of six CDMs between the four subtypes of breast cancer, including HR+/HER2+, HR+/HER2−, HR−/HER2+, and HR−/HER2− (TNBC). The six CDMs included PIK3CA H1047R and E545K, KRAS G12D and G12V, HRAS G12D, and BRAF V600E. These mutations were chosen because of their known roles in breast carcinogenesis (e.g., mutations in PIK3CA), likelihood to impact targeted therapies currently being explored in breast cancer (e.g., mutant PIK3CA impacts resistance to HER2 targeted therapies; and mutations in key genes of the MAP kinase pathway that impact resistance to EGFR-targeted therapies (e.g., KRAS, HRAS, and BRAF)), or their potential as a therapeutic target (e.g., PIK3CA and BRAF mutant proteins).
This study has resulted in several critical findings, including the determination that: (1) all IDCs contain measurable levels of at least one CDM, with the majority (59%) harboring three or more of the six CDMs, (2) CDMs are present in a subpopulation of cells, and therefore, would go undetected by standard DNA sequencing, (3) PIK3CA H1047R and BRAF V600E are correlated positively with maximum tumor dimension, whereas KRAS G12V is correlated negatively (non-significant), and (4) subtype specific differences in MFs exist, most notably lower PIK3CA H1047R MFs in TNBCs (as compared to normal breast [7]).
Activating somatic mutations in PIK3CA are considered early, initiating events in breast cancer, resulting in cell proliferation and resistance to apoptosis. Furthermore, transgenic mouse models were used to demonstrate that expression of mutant PIK3CA induces mammary epithelial cell transformation and mouse mammary tumors [22]. PIK3CA is the most frequently mutated gene in human breast cancer, detected in 27% of breast carcinomas in the COSMIC database [6]. Three major hotspot mutations in PIK3CA are reported in IDC tumor samples, the H1047R, E545K, and E542K mutations, with reported mutation frequencies of 10.52%, 5.41%, and 3.22%, respectively. Data obtained using the sensitive and quantitative ACB-PCR approach generated a number of interesting findings and expanded current knowledge regarding PIK3CA mutation in human breast tumors. Specifically, (1) 59% of IDCs have measurable levels (≥ 1 × 10 −5 ) of PIK3CA mutation, with 42%, 36%, and 19% having measurable levels of the PIK3CA H1047R, E545K, or both mutations, respectively, (2) PIK3CA H1047R and E545K exist as mutant subpopulations in IDCs, meaning they have levels of PIK3CA mutation that would go undetected by DNA sequencing, and (3) PIK3CA H1047R MF measurements in TNBC are significantly lower than that measured in normal breast. A greater level of PIK3CA mutation in normal breast than in breast tumors is consistent with the ideas that PIK3CA mutations are early drivers of multiclonal tumor origin, that multiple mutations are needed to overcome normal tissue homeostatic mechanisms, and that clonal selection and metapopulation dynamics drive the outgrowth of tumor cells, which may be predominantly other than PIK3CA-mutant (yet retaining minor PIK3CA mutant subpopulations), although some percentage of the time predominantly PIK3CA-mutant tumors do arise [23].
The high prevalence of mutated PIK3CA in breast cancer has created an active area of research as well as an attractive drug target. Mutation in the PIK3CA gene confers sensitivity to PI3K/AKT/mTOR pathway inhibition. Therefore, targeting components of this pathway has become a popular strategy in breast cancer therapeutics, particularly for use in TNBC [24][25][26]. Data generated from this study suggests that inhibitors of the PI3K/AKT/mTOR pathway may be a viable strategy for many breast cancer patients, considering 59% of IDCs had measurable levels of PIK3CA mutation, and 26% of IDCs had levels of PIK3CA mutation ≥ 1 × 10 −3 . For TNBC, however, very few samples had appreciable levels of PIK3CA mutation. Only four samples had a PIK3CA MF ≥ 1 × 10 −5 and only one sample had a level of mutation ≥ 1 × 10 −3 , suggesting inhibition of the PI3K/AKT/mTOR pathway in TNBCs may not be a broadly effective strategy. Nevertheless, TNBC is not one homogeneous disease [27,28], suggesting the use of sensitive mutation detection approaches could be used to identify TNBCs with actionable PIK3CA mutations, such as those found in this study. Given the reported overexpression of EGFR in TNBC [29], and the relatively low levels of therapeutic resistance-causing PIK3CA, KRAS, and BRAF mutations found in TNBC, therapies directed against the EGFR receptor may be a viable avenue for further exploration.
Previously, we showed surprisingly high levels of the PIK3CA H1047R mutation were present in normal breast tissue (median MF = 6.13 × 10 −4 ; range = 1.25 × 10 −5 − 4.59 × 10 −2 ) [7]. Interestingly, we found many tumors (~60%) in this study contained PIK3CA H1047R mutation at levels lower than the lowest PIK3CA H1047R MF measured in normal breast. We hypothesize that pre-existing PIK3CA mutant cells can drive tumor initiation and early progression but are lost or outgrown by other cancer driving clones in most advanced breast cancers. Data from the COSMIC database supports this idea, with PIK3CA mutation detected in 49% of hyperplasia, 29% of DCIS, and 24% in IDCs [6], suggesting the level of PIK3CA mutation decreases as disease progression ensues. In the future, it would be interesting to analyze spatially-disparate normal samples from individuals with IDC and compare levels of mutations between normal and tumor samples and between normal breast samples from women with and without IDC.
Although found at lower frequencies (as compared to mutations in PIK3CA), KRAS, HRAS, and BRAF mutations were highly prevalent across IDC subtypes, albeit as minor mutant subpopulations. Measurable levels of KRAS G12D and G12V, HRAS G12D and BRAF V600E were found in 63%, 11%, 84%, and 59% of IDCs, respectively. This is in stark contrast to the reported frequencies in the COSMIC database, where KRAS G12D and G12V, HRAS G12D, and BRAF V600E are reported to occur in 8/2585 (0.31%), 4/2585 (0.16%), 0/2159 (0%), and 0/1844 (0%) of IDC tumor samples [6]. The finding that minor mutant subpopulations exist in breast IDCs may have therapeutic implications, as mutations in the genes of the MAP kinase pathway are known to impact response to molecularly-targeted therapy for other cancer types [30][31][32][33][34]. Interestingly, an inverse relationship between KRAS and BRAF mutation immerged in this dataset. That is, (1) BRAF V600E and KRAS G12D as well as BRAF V600E and KRAS G12V MFs were inversely correlated in IDCs, (2) BRAF V600E was correlated positively with maximum tumor dimension, whereas KRAS G12V MF was correlated negatively (non-significant), (3) greater BRAF V600E MFs were observed in the HR+/HER2− subtype compared to the HR+/HER2+ and HR−/HER2+ subtypes, whereas significantly less KRAS G12D mutation was observed in the HR+/HER2− subtype as compared to HR+/HER2+ and HR−/HER2+ subtypes, and (4) BRAF V600E was significantly greater in HER2− samples, while KRAS (G12D and G12V) were significantly greater in HER2+ IDCs. The inverse nature of the prevalence of the BRAF and KRAS mutations may be due, in part, to redundancy in the MAP kinase pathway. While BRAF and KRAS mutations have been described as "mutually exclusive" [35], we observed 26/81 (32.1%) of IDCs contained mutations in both genes. It is currently unclear what impact these low-frequency mutant subpopulations of critical genes in the MAP kinase pathway may have on tumor biology and therapeutic response.
A growing body of evidence suggests that undetected mutant subpopulations may impact response to molecularly targeted cancer therapies [30,31,36,37]. For example, in HER2-overexpressing breast cancer, mutations in PIK3CA have been shown to confer resistance to the HER2-targeted therapy, trastuzumab [38,39]. In this study, measurable levels of PIK3CA mutation (H1047R, E545K or both) were found in 63.4% of HER2+ IDCs, as compared to 15.6% reported in the COSMIC database. The high prevalence of PIK3CA mutant subpopulations may explain, in part, the high frequency of relapse in these patients after trastuzumab therapy. Furthermore, we observed that most breast IDCs contain multiple CDMs. For example, 53.0% of IDCs contained clonal subpopulations expected to be PIK3CA and MAP kinase pathway mutant, suggesting that combinations of molecularly-targeted therapies may be required for adequate treatment of breast cancer.
The underappreciated role of MAP kinase pathway gene mutations in breast cancer pathology and response to treatment may be due, in part, to the use of insensitive mutation detection methodologies in an exceedingly heterogenous tumor. Most mutation detection techniques employed today (i.e., DNA sequencing, ARMS/Scorpions, MALDI-TOF, mutant-enriched PCR, and high-resolution melting analysis) do not have the required sensitivity to detect minor mutant subpopulations present in a heterogeneous tumor, such as breast cancer [37]. ACB-PCR is an exceptionally powerful and unique technique, shown to quantify particular cancer-related base pair substitution mutations in normal human tissues and tumors with a sensitivity of 1 × 10 −5 [7,8,11,15,30,40,41]. Only mutation detection techniques with this level of sensitivity would be able to adequately characterize the gene mutations in the MAP kinase pathway quantified in this study.
One major limitation of this study is the lack of adequate samples representing African-American women. Breast cancers in African-American women display different characteristics than in Caucasian women [42]. In African-American women, breast cancers (1) occur with an earlier onset, (2) show less favorable outcomes, and (3) present with a more aggressive tumor phenotype [43]. Many differences have been attributed to these disparities, most notably the higher prevalence of TNBC in women of African-American decent [44][45][46][47]. While differences in epidemiology and prognosis between African-American and Caucasian women with breast cancer (and most notably TNBC) have been described in the literature [44,47], few data are available regarding possible differences in the somatic mutations that may underlie these differences. Therefore, ACB-PCR analyses to detect potential differences in low frequency somatic point mutations in TNBC of women of African origin are ongoing.

DNA Isolation
All IDCs were frozen in liquid nitrogen and pulverized using mortar and pestle. The pulverized samples were further homogenized using an Omni THQ tissue homogenizer (Omni International, Kennesaw, GA, USA) and 6 mL proteinase K buffer [1 mg/mL proteinase K, 100 mM NaCl, 25 mM EDTA (pH 8), and 1% SDS] per gram of tissue. The homogenate was incubated~16 h at 37 • C, and DNA extraction completed as described previously [40].

Generation of Mutant and Wild-Type Standards and ACB-PCR
ACB-PCR is an allele-specific amplification method that can be used to selectively amplify and quantify a mutant allele in a 100,000-fold excess of wild-type allele. Each ACB-PCR assay essentially includes a reconstruction experiment. Mutant and wild-type plasmid DNAs are used as template to PCR amplify the same amplicon from the genomic sample DNAs in a high-fidelity first-round PCR, which is optimized to ensure equal number of duplications occur between the different sample types. Based on the DNA concentration of the mutant and wild-type amplicons, they are combined to generate a set of mutant fraction (MF) standards with the defined MFs of 10 −1 , 10 −2 , 10 −3 , 10 −4 , 10 −5 , and 0 (wild-type amplicon only). Equal numbers of molecules of the defined MF standards and unknowns are then analyzed by ACB-PCR. The signal generated from the wild-type-only amplicon defines the background signal of each ACB-PCR assay. The remaining MF standards are used to construct a standard curve, from which the MFs of the unknown samples can be interpolated [8]. Mutant PIK3CA H1047R (codon 1047 CGT) and E545K (codon 545 AAG), KRAS G12D (codon 12 GAT) and G12V (codon 12 GTT), HRAS G12D (codon 12 GAC), and BRAF V600E (codon 600 GAG) and wild-type (PIK3CA codon 1047 CAT and codon 545 GAG, KRAS codon 12 GGT, HRAS codon 12 GGC, and BRAF codon 600 GTG) standards were prepared by PCR amplification of cloned mutant or wild-type plasmid DNAs. For PIK3CA codon 545, KRAS codon 12, HRAS codon 12, and BRAF codon 600, primers and reaction conditions were identical to those used for amplification of the target sequences from genomic DNAs [7,11]. For PIK3CA codon 1047, extended upstream and downstream primers (up, 5 -ACATCATTTGCTCCAAACTGACCAAACTGTTCTTATTACTTA-3 ; down, 5 -ATGCATGCTGTTTAATTGTGTGGAAGATCCAATCCATTTTTGT-3') were used to generate a 308 bp product identical in size and sequence to the amplicon in the genomic DNAs. Purification, storage, and quantification of wild-type and mutant PCR products were performed as described above.
On the day of the ACB-PCR, purified PIK3CA codon 1047 and codon 545, KRAS codon 12, HRAS codon 12, or BRAF codon 12 mutant and wild-type reference DNAs (prepared by PCR amplification of cloned mutant or wild-type plasmid DNAs [7]) were mixed to generate standards with MFs of 10 −1 , 10 −2 , 10 −3 , 10 −4 , 10 −5 , and 0 (containing only the wild-type sequence), at a concentration of 5 × 10 7 copies/µL for the PIK3CA E545K, KRAS G12D and G12V, and BRAF V600E standards, and 1 × 10 8 copies/µL for the PIK3CA H1047R and HRAS G12D standards. Ten microliters of the standard DNA mixture or genomic DNA was used in the corresponding ACB-PCR reaction, for a total of 5 × 10 8 copies per reaction for PIK3CA E545K, KRAS G12D and G12V, and BRAF V600E assays, and 1 × 10 9 copies per reaction for PIK3CA H1047R and HRAS G12D assays. Duplicate reactions for each MF standard, along with a no-DNA control was included for every ACB-PCR experiment. ACB-PCR primers (including a fluorescein-labelled mutant-specific primer (MSP), a blocker-primer (BP), and an up-or downstream primer) and PCR reaction and cycling conditions were performed as previously described for HRAS G12D and BRAF V600E [7], and PIK3CA E545K [11].

Vertical Polyacrylamide Gel Electrophoresis, Image Analysis, and Data Collection
Equal volumes of ACB-PCR products, containing 1× ficoll loading buffer/dye, were loaded on non-denaturing 8% vertical polyacrylamide gels. The fluorescent ACB-PCR products were visualized using a PharosFX Molecular Imager with an external blue laser (Bio-Rad). The pixel intensities of the ACB-PCR product bands (PIK3CA H1047R, 148 bp; PIK3CA E545K, 77 bp; KRAS G12D, 103 bp; KRAS G12V, 89 bp, HRAS G12D, 95 bp; and BRAF V600E, 98 bp) were quantified using Quantity One software (Bio-Rad) and a locally averaged background correction. For PIK3CA H1047R, KRAS G12D and G12V, HRAS G12D and BRAF V600E, construction of the standard curve and interpolation of the MF of unknowns was performed as previously described [7]. For PIK3CA E545K, the construction of a standard curve utilizing all MF standards (1 × 10 −5 to 1 × 10 −1 ) did not accurately interpolate the MF of the unknown samples at both the low and high end of the standard curve based on their respective fluorescent intensities, in pixels (i.e., overestimation at the low end and underestimation at the high end of MF). To circumvent an under-or overestimation, PIK3CA E545K MF was interpolated from standard curves that gave the best quantification across an observable range of MFs. Specifically, MF standards were used to construct two log-log plots; one 1 × 10 −5 to 1 × 10 −3 for samples with pixel intensities < 1 × 10 −3 MF standard, and one 1 × 10 −3 to 1 × 10 −1 for samples with pixel intensities > 1 × 10 −3 MF standard. MFs were calculated using the power function and pixel intensities of the PIK3CA E545K ACB-PCR products.

MF Measurements in Normal Breast
Ten fresh-frozen normal breast samples (collected at autopsy from individuals who died from causes other than cancer or breast disease) were purchased from the National Disease Research Interchange (NDRI; Philadelphia, PA, USA) [7,11]. DNA isolation, first-round PCR amplification of PIK3CA codon 1047 and 545, KRAS codon 12, HRAS codon 12 and BRAF codon 600, and subsequent ACB-PCR quantification of PIK3CA H1047R and E545K, KRAS G12D and G12V, HRAS G12D, and BRAF V600E MFs were described [7,11]. The MF measurements previously reported in normal breast have been integrated into the analyses of the current study.

Statistical Analyses
The arithmetic average of three independent MF measurements was used to calculate the MF of each target of interest for all samples. The average MF measurement for each sample was log 10 -transformed. For each mutational target, the geometric mean MF was calculated as the average log 10 -transformed MFs measured in each group. Log 10 -transformed data were used for all subsequent analyses. For group comparisons, non-parametric statistical methods were employed because the log-transformed data were not normally distributed. For comparisons between normal breast and all IDCs, Fisher's exact test was employed to compare the numbers of samples above and below the levels of accurate ACB-PCR quantitation (i.e., > 1 × 10 −5 and < 1 × 10 −5 ), using GraphPad Prism 5 Software (GraphPad Software, Inc., La Jolla, CA, USA). Differences in cancer-driver MFs between IDC subtypes and normal breast tissue were assessed using Fisher's exact test with Holm's correction for multiple comparison testing. Correlation analyses were assessed using Spearman's rank correlation coefficient (unadjusted for tied data). Correlation between a quantitative variable and an ordinal variable (e.g., MF and clinical stage) was assessed with polyserial correlation. For all statistical analyses, significance was defined as p ≤ 0.05.

Conclusions
The sensitive and quantitative ACB-PCR approach was used to obtain a detailed and quantitative molecular genetic analysis of six CDMs in 81 IDCs, categorized into four IDC subtypes (including TNBC). Significant differences in levels of these CDMs were observed among the four IDC subtypes, as well as that of normal breast tissue previously published. Of significant note, the PIK3CA H1047R mutation was found to be lower in TNBCs, suggesting a lack of benefit with PI3K/AKT/mTOR inhibition in these individuals. In HER2+ breast tumors, PIK3CA mutation (H1047R, E545K, or both) was found in 63% of tumors, which may subvert the efficacy of HER2-targeted therapies and ultimately result in relapse. Furthermore, a large percentage of breast cancers were found to have measurable levels of mutations in critical genes of the MAP kinase pathway (i.e., KRAS, HRAS, and BRAF). While the MAP kinase pathway gene mutations analyzed in this study have only rarely been associated with breast cancer, their roles may be underestimated, due to the use of relatively insensitive mutation detection methods. Thus, the low-frequency mutational profile of IDC subtypes, including TNBC, revealed in this study identified points in tumor-related pathways that are the most appropriate/inappropriate for therapeutic intervention. Furthermore, the data generated provide a rationale for incorporating more sensitive and quantitative approaches (in place of standard DNA sequencing) into future clinical trials designed to address specific personalized medicine questions.