PIK3CA/AKT1 Mutations in Breast Carcinoma: a Comprehensive Review of Experimental and Clinical Studies

The phosphatidylinositol-3-kinase (PI3K) pathway is an important nexus for integration of extracellular and intracellular signals, and there are very frequent perturbations of this pathway in breast cancer, making it an attractive target for therapeutic manipulation. Hotspot mutations in PIK3CA or AKT1 are found in nearly 30% of breast cancers, especially estrogen receptor-positive and Her-2/neu-positive subgroups. This review will emphasize experimental models, clinical characterization and implications of PIK3CA/AKT1 mutations in breast cancer. In vitro studies have demonstrated that these mutations are kinase-activating and can confer cellular transforming properties in the correct context. Further, overexpression of PIK3CA H1047R in a variety of murine models results in mammary proliferation and carcinomas, and established carcinomas may become PIK3CA H1047R-independent. Data from human breast carcinomas regarding the clinicopathologic significance of PIK3CA/AKT1 mutations were contradictory at first, yet trends are beginning to emerge. PIK3CA mutation seems to impart a favorable prognosis in estrogen-receptor positive breast cancers, and mutations are seen early, in pre-invasive breast lesions. Although larger studies are needed, PIK3CA mutations may not confer selective advantage in the metastatic setting. Numerous pharmacologic compounds targeting the PI3K pathway are in development. The complexity of PI3K crosstalk with other signaling cascades, negative feedback regulation, and the myriad of other genotypic and phenotypic deviations in breast cancers argues for thorough molecular characterization of tumors in cancer trials.


Introduction
In 2004, Samuels and the Vogelstein-Velculescu group first characterized point mutations in the PIK3CA gene, encoding the 110α-catalytic subunit of phosphatidylinositol-3-kinase (PI3K) in a variety of human cancers [1]. Meta-analysis of the breast cancer literature reveals that PIK3CA mutations are present in 26% of invasive breast cancers, making this gene one of the most frequently mutated in breast cancer, with most mutations concentrated in 'hotspots' in the helical domain (exon 9) or the catalytic domain (exon 20), with a smaller number of mutations preferentially in exons 1, 4 and 7. ( Figure  1, Table 1, and references therein). More recently, Carpten et al. [36] characterized mutations in the plekstrin homology domain of AKT1 (found in 3.8% of breast cancers), an important downstream kinase. These discoveries have stimulated a large body of exciting in vitro, in vivo, and translational studies investigating the scope and implications of PIK3CA and AKT1 mutations.
Study of double-mutant cell lines have provided further insight into the disparate effects of PIK3CA helical and kinase domain mutants. Mutations in exon 9 (helical domain), disrupt the constitutive inhibitory effect of the PI3K p85 subunit, but remain dependent upon interaction with RAS [5,32,35]. In contrast, exon 20 mutations (kinase domain), become independent of RAS, yet require binding to the p85 regulatory domain [5,19,32,35]. Interestingly, recent in vitro studies have also shown that PIK3CA mutant kinases, especially exon 9 helical domain mutants, may preferentially activate PDK1 and its substrate SGK3, rather than the canonical AKT pathway [22]. These studies suggest that the different PIK3CA hotspot mutations in human tumors could have different effects, likely also influenced by cell type, microenvironment, and interplay with other aberrant tumor signaling pathways.

Activating AKT1 mutations
Mutations in AKT family members in cancer remained elusive for many years, as none were found in studies of the kinase domain. However, in 2007 Carpten et al. [36] identified transforming mutations in the plekstrin-homology domain of AKT1, present in a number of cancers, including breast cancer. The plekstrin-homology domain is a key segment involved in AKT1 localization to the cell membrane, and interaction with PIP 3 . [36][37][38]The activating mutation E17K alters the lipid binding pocket, resulting in increased membrane localization, AKT1 activation, and transformation of NIH 3T3 or Rat1 cell lines [36]. Again, expression of mutant AKT1 at physiologic levels in breast epithelial cells lines did not prove sufficient for transformation [34]. The various isoforms of AKT have been linked to different cellular phenotypes, with AKT1 associated with cell survival and growth in an environmental dependent fashion, and AKT2 linked to invasiveness. There is concern that inhibition of AKT1 could result in compensatory AKT2 overactivation, thus promoting invasion and metastasis [2,8].

PIK3CA kinase domain mutations in murine models
Three mouse models of mammary specific overexpression of PIK3CA with the activating H1047R kinase domain mutation (exon 20) have recently been published; no exon 9 transgenic models have been published to date [39][40][41]. Adams et al. [39] utilized the ROSA26 knock-in system and MMTV-Cre to drive expression of either H1047R or wild type PIK3CA in the mouse mammary gland (in a 129/CD1 background, backcrossed to FVB). Two variant lineages were studied, with differences in strength of recombination/transgene expression levels, and specificity of restriction to the mammary gland. Female mice from both H1047R mutant lines developed mammary tumors starting at 5 months of age, and showed reduced survival, while animals overexpressing wild type PIK3CA remained healthy [39]. The mammary tumors in H1047 animals included adenomyoepitheliomas and adenosquamous carcinomas, which were shown to generally express estrogen receptor and to have elevated levels of phospho-AKT(S473), PTEN and phospho-c-Jun(S73) [39]. The authors also investigated interaction of PIK3CA H1047R mutation with p53 in double transgenic animals, and found and increased rate of tumor formation, and a different spectrum of tumor histology, predominantly poorly differentiated and spindle cell, as is fairly characteristic of p53 recombinant mouse models [39,42]. As a side note, some H1047R transgenic mice without Cre recombinase developed blood vessel lesions, attributed to spontaneous activation of H1047R expression in endothelial cells [39]. This ties in with observations of Graupera et al. [43] that PIK3α is involved in angiogenesis and endothelial migration [3,43].
Meyer et al. [40] generated PIK3CA H1047 and GFP expressing animals, with expression driven by either MMTV-Cre, or WAP-Cre recombination in mammary luminal epithelial cells, in a FVB/N background. Female mice from both lineages developed proliferative mammary gland abnormalities. Although 75% of the MMTV animals died at a young age of undetermined causes (attributed to MMTV promoter leakiness causing deleterious expression elsewhere), surviving animals developed mammary carcinomas at an average of 7 months of age [40]. In the WAP strain, parous mice developed tumors more rapidly than nulliparous females, as would be expected given enhanced WAP activity and thus recombination and transgene expression with pregnancy [40]. A delayed involutional phenotype with a dramatic decrease in mammary gland apoptosis was also demonstrated after pregnancy [40]. As noted in the Toronto mouse model, the mammary carcinomas in the transgenic animals were of a variety of histologic types, predominantly adenomyoepitheliomas and adenosquamous carcinoma, and showed increased phospho-AKT levels, with about 20% of the adenomyoepitheliomas expressing estrogen receptor.
An elegant transgenic mouse model with conditional overexpression of HA-tagged PIK3CA H1047R was created by Liu et al. [41,44] In this model, PIK3CA H1047R overexpression is driven by an inducible promoter, with expression of the reverse tetracycline transactivator linked to the MMTV promoter, resulting in transgene expression in the mammary gland in the presence of tetracycline (doxycycline) in a FVB background [41,44]. Prior studies with this system demonstrated titratable levels of reporter proteins, with somewhat increased expression during pregnancy/lactation, as MMTV is hormonally dependent, and slight heterogeneity of expression in older animals [44]. Like the other models, proliferative mammary gland lesions were seen in PIK3CA H1047R expressing animals, and carcinomas developed at an average of 7 months [41]. The tumors were predominantly of solid, mixed or squamous morphology, with acinar, glandular, and papillary histologies also seen. Tumors had abundant phospho-AKT (S473), and phospho-S6RP (S235, S236) by immunohistochemical analysis. Interesting, upon withdrawing doxycycline (turning PIK3CA H1047 expression off), about one-third of tumors completely regressed, but nearly two-third of the tumors partially regressed and then resumed growth [41]. These recrudescent tumors were essentially independent of the PIK3CA transgene, and several exhibited additional genetic alterations, including Met, cdkn2a or myc, which also imparted resistance to a PIK3CA targeted compound, GDC-0941. [41] This model suggests that activating PIK3CA mutations may have an important role in tumorigenesis, but that additional 'hits' often occur, promoting PIK3CA oncogene independence and sustained tumor growth even upon inhibition of a key tumor-initiating pathway [41].
AKT1 E17K transgenic mouse mammary models have not yet been published. Prior studies with activated (myristated) AKT1 demonstrated accelerated tumorigenesis but reduced metastasis, and activated AKT2 increased metastasis in a background of activated ERBB2 or polyoma virus middle T antigen [45].
Common features of these three contemporary PIK3CA H1047R mouse models include robust development of mammary tumors upon mammary gland directed overexpression of the PIK3CA activating mutation H1047R. Further, each of the models showed a similar variety of murine tumor morphologies, and estrogen receptor expression reminiscent of human tumor data demonstrating PIK3CA mutations in a wide spectrum of breast carcinomas, as detailed below (Table  1; of note, estrogen expression is relatively rare in mouse mammary carcinoma as compared to human carcinoma) [46][47][48]. The authors have remarked on the prevalence of murine adenosquamous carcinomas in H1047R transgenic models, drawing analogy to the apparent high prevalence of PIK3CA mutations in human metaplastic breast carcinoma. However, it should be noted that mammary tumors in nongenetically engineered animals and some other genetic models have a predilection for squamous or adenomyoepitheliomatous morphology (FVB, BALB/c and C3H, and in association with chemical carcinogens; Wnt pathway transgenic) [42,48,49]. Other caveats include the supraphysiologic expression levels of the mutant transgene, and the presence of epitope tags in some of the models.
Early studies did not reveal correlation of PIK3CA mutation status with clinicopathologic parameters of carcinoma grade, stage, hormone receptor status, Her2 status, or outcome. However, with data from additional larger studies, several patterns have emerged. PIK3CA mutations are seen preferentially in estrogen receptor (ER)-positive and Her2/neu-positive tumors, and are less prevalent in 'triple-negative' carcinomas (Table 1, Figure 2) [46]. However, Hennessy et al. [46] found a notable exception, in that their cohort of 19 squamous and spindle cell metaplastic carcinomas had a nearly 50% frequency of PIK3CA mutations. Lobular carcinomas also appear to have a relatively high rate of mutation [62,63]. Few of the other morphologic 'special types of breast carcinoma' have been studied in the literature; papillary and micropapillary carcinomas appear to have a mutation profile similar to ductal carcinomas, perhaps with enrichment of AKT1 mutations in micropapillary carcinomas (our preliminary data) [61,66]. In contrast, our group and others have found that mucinous carcinomas, which are often low-grade and ER+, have a paucity of PIK3CA mutations [52,60,61,63,[67][68][69]. Further, we previous studied phyllodes tumors, and found no PIK3CA/AKT pathway mutations, while Vorkas reported one fibroadenoma with an H1047R mutation, suggesting a role in epithelial malignancy, and less of a role in fibroepithelial lesions [65,70].

PIK3CA mutation and breast cancer prognosis
Initial data was also contradictory regarding the association of PIK3CA mutations and breast cancer outcome, with studies of substantial numbers of patients demonstrating alternatively worse survival, or favorable prognosis of breast cancers harboring PIK3CA mutations. [52,63,71], Li et al. [52] reviewed a cohort of 250 breast carcinomas, and found a 35% frequency of PIK3CA mutations (exons 7,9,20). PIK3CA mutations were associated with worse survival (p=0.004); however, mutations were associated with larger tumor size (p=0.004) and ER-positivity (p=0.008) [52]. In multivariate analysis, PIK3CA mutation did not reach statistical significance in the overall group or in ER+ tumors, but was independently associated with worse survival in the Her2 negative group (p=0.016) [52]. In contrast, a Japanese study of 188 tumors found a 30.7% PIK3CA mutation frequency (exons 1,2,4,7,9,13,18,20), and found mutations to be associated with favorable prognosis in uni-and multivariate analysis (p<0.05) [63]. Mutations were also significantly associated with ERpositivity and phospho-AKT positivity [63]. Similarly, a larger study    from Memorial-Sloan Kettering demonstrated PIK3CA mutations in 32.5% of 590 breast cancers, and demonstrated that patients with PIK3CA mutated tumors have significantly improved overall (p=0.03) and cancer-specific survival (p=0.004), while mutations were associated with older age at diagnosis, hormone receptor positivity, Her2 negativity, lower grade and stage, and lymph node negativity [71]. However, the improvement in survival was independent of hormone receptor status [71]. This group also analyzed outcome based on type of PIK3CA mutation, and found that the survival benefit correlated with kinase domain (exon 20) mutation (p=0.005), but not helical domain (exon 9) mutation (p=0.54). They also noted kinase domain mutation to correlate with lymph node negativity (p=0.007), and helical domain mutation to correlate with older age at diagnosis (p=0.004) [71].

B -AKT1 Domains & Hotspot Mutations
In addition to the Memorial Sloan Kettering group, other studies have separately analyzed the association of helical domain (exon 9) and kinase domain (exon 20) mutations with breast cancer outcome, which again yielded contradictory data. [62,72] An Italian study of 163 breast carcinomas found a PIK3CA mutation rate of 28%, with 53% of mutations in exon 9, and 47% exon 20 [62]; of note, most other studies have found a greater frequency of exon 20 mutations. Patients with exon 20 mutations had significantly better overall and disease free survival, as compared to patients without PIK3CA mutations [62]. Conversely, patients with exon 9 mutations had worse 5 year overall (p=0.018) and disease free (p=0.040) survival as compared to patients without PIK3CA mutations, and certainly as compared to patients with exon 20 mutations [62]. In multivariate analysis, the presence of exon 9 mutations maintained significance in terms of prognosis, and was the strongest predictor of disease free and overall survival (p=0.001 and p=0.0003 respectively), with nodal status and progesterone receptor expression also significant [62]. In contrast, a study from Taiwan found poor prognosis associated with exon 20 mutations [72]. This study documented a 26% PIK3CA mutation frequency in 152 breast cancer cases; of the mutations, 35% were in exon 9, 70% in exon 20, 12% in exons 4, 7 [72]. Exon 20 mutations were associated with worse overall survival at 5 years in univariate analysis (p=0.005), but not with disease free survival (p=0.5877). [72] In multivariate analysis, the presence of PIK3CA exon 20 mutation was again a risk factor for significantly worse survival (p=0.0038) along with nodal status (p=0.0004). [72] More recently, PIK3CA mutations have been analyzed in the context of somewhat more homogeneous subgroups of breast cancer, notably ER+ tumors. [57,58,73] These studies have generally revealed improved prognosis with PIK3CA mutation in this subset of breast cancer, a somewhat paradoxical finding considering that these are kinase activating mutations. [57,58,73] In a study of ER+ tumors from several different neoadjuvant endocrine therapy trials, Ellis et al. [73] demonstrated a 32% PIK3CA mutation frequency, and found presence of mutation to be a favorable prognostic factor for relapse free survival in univariate (p=0.02) and multivariate analysis, despite a weak negative correlation with clinical response to neoadjuvant endocrine therapy.

PIK3CA mutation gene signature
Using PIK3CA mutation status in addition to gene expression analysis of 173 ER+/Her2-breast carcinomas, Loi et al. [57] characterized a PIK3CA mutation "gene signature," and then validated this signature in several other published datasets. In the initial dataset, 91% of PIK3CA mutations were exon 20 H1047R substitutions (overall 26% mutation rate). They found that PIK3CA mutation was indeed associated with a distinct gene expression profile, and in fact, some tumors showed this expression profile in absence of mutation, suggesting alternate mechanisms of pathway activation. Of note, the PIK3CA mutation-like gene signature appears to be distinct from the PTEN-loss expression signature. While PIK3CA mutated carcinomas were shown to have somewhat better outcome than wildtype carcinoma in the initial dataset, the PIK3CA-mutant like gene signature had still better prognostic significance [57]. The PIK3CA-gene signature was negatively correlated with signatures of proliferation, AKT activation, and PTEN loss, but positively correlated with ESR1 expression. Further protein data revealed decreased expression of mTOR, S6K and S6, with decreased phosphorylation of 4EBP1 and S6, in contrast to data from cell lines [57].
With a similar strategy, Cizkova et al. [58] characterized the transcriptome changes in PIK3CA mutated tumors. In this French cohort, 36% of cases had PIK3CA mutation, almost equally divided between exon 9 and exon 20 in the validation group of 249 tumors [58]. Differentially regulated genes amongst PIK3CA mutant and wildtype tumors were analyzed by cellular pathways, revealing impact on many signalling networks, including MAPK, Calcium, Jak-STAT, Wnt, apoptosis, with the Calcium pathway altered by downregulated genes, and Wnt pathway by upregulated genes [58]. This group further characterized a 29-gene set for discrimination of PIK3CA mutant and wild-type tumors, and verified it by RT-PCR in the 249 tumor cohort. Of note, 2 genes in this set were differentially regulated by exon 9 vs. exon 20 mutation. (TFAP2B-overexpressed in exon 20 mutated tumors and NRIP3-overexpressed in exon 9 mutated tumors). Interestingly, of the 29 discriminatory genes published by Cizkova, 6 are common to Loi's 278 gene signature (Table 3) [57,58].
In summary, current consensus associates PIK3CA mutation with favorable breast cancer outcome in ER+ tumors.

PIK3CA heterogeneity in tumors and metastasis
Several studies to date have examined PIK3CA mutational status within primary breast tumors, and between primary tumors and nodal or distant metastasis, with contrasting findings (Table 4). [74][75][76][77] Dupont-Jensen reported considerable discordance in PIK3CA mutational status between primary tumors and metastasis. [75] In a study of 104 patients with available tissue from primary and asynchronous recurrent/metastatic carcinomas, PIK3CA mutations were found in 45% of primary tumors, with a very sensitive SNaPshot PCR method. [75] Interestingly, the rate of PIK3CA mutation in metastatic tumor was 53%, but 21 patients changed genotype from wild type to mutant, and 11 from mutant to wildtype [75]. Further, positive lymph nodes (synchronous with primary breast carcinoma) were assayed in 47 patients, revealing a 34% mutation rate, with mutational discordance between primary tumor and synchronous nodal metastasis in 12/46 patients (26%), with only 4 of 12 gaining mutation in nodal metastatic tumor [75]. Further, intratumoral heterogeneity was assayed in 10 primary tumors, and discordance was found in 4 (40%), including one case that had different foci assayed as PIK3CA wildtype, H1047R mutation, and E542K mutation (Table 4) [75].
Using a sensitive PCR-mass-array mutation screen, Gonzalez-Angulo et al. [76] screened another cohort of patients with metastatic breast carcinoma and found comparable PIK3CA mutation frequencies in primary tumor and metastasis after anthracycline and/or taxane based chemotherapy (40% and 42% respectively), with a propensity for exon 20 mutations in this group. However, upon analysis of paired metastasis from the same patient, 18% of patients had discordant PIK3CA mutation status. In this study, almost equal numbers of metastases gained PIK3CA mutations as lost mutations (Table 4) [76]. This group also compared wildtype and mutant peak height as a surrogate measure of the fraction of DNA/fraction of cells harboring mutation, and noted that 5 additional cases had >50% changes in the fraction of mutated DNA. While different mixtures tumor clones with heterogeneous mutation status could explain this observation, these results might also be accounted for by differing proportions of admixed non-neoplastic tissue (desmoplastic stroma etc.); although the authors attempted to control for the latter by requiring at least 70% tumor nuclear cellularity in tested specimens [76].
The group from Memorial-Sloan Kettering employed similar sensitive methodologies (PCR and mass array analysis, but also confirmed patient identity between samples by SNP analysis), with an overall mutation rate of 38%, but found substantially higher intratumoral and primary-metastasis mutational concordance. Intratumoral heterogeneity was analyzed in a larger number of primary breast cancers, with mutational concordance in 60/63 of tumors (95% see below for DCIS) [77]. Further, mutations in primary breast tumor and synchronous nodal metastasis, as well as asynchronous distant metastasis were mostly concordant (11/12, 91% nodes; 5/5, 100% distant, Table 4) [77]. In a small sample of breast carcinomas and paired brain metastases, Da Silva et al. [74] similarly found a complete concordance for PIK3CA mutation in all 12 pairs (2 pairs mutant, 10 wildtype).
The explanation for the marked difference in findings amongst these studies is unclear, perhaps related to methodological differences, divergent populations and treatments, and small samples sizes in    ND: not done # Nodes: 8 cases had mutation in primary but not in nodal metastasis (3 exon 9, 1 exon 20); 4 had mutation in nodal metastasis but not in primary (3 exon 9, 5 exon 20). # Distant: 11 cases had mutation in primary but not in metastasis; 21 cases had mutation in metastasis but not in primary (p=0.08) *5 cases had mutation in primary but not in metastasis; 4 cases had mutation in metastasis but not primary ^Nodes: one case had exon 20 mutationin primary, but not in nodal metastasis some studies. However, a pattern is emerging, suggesting that PIK3CA mutations are neither selected for nor against in metastatic breast cancer. Further work to characterize intratumoral heterogeneity, and heterogeneity in metastatic tumors will be essential to recommending an optimal tumor testing algorithm, both for PIK3CA testing in breast carcinoma, and as applied to other types of tumors.

AKT1 mutations in breast carcinoma
Fewer studies to date have investigated AKT1 plekstrin homology domain mutations in breast cancer. Meta-analysis of 10 studies covering over 2000 samples reveals an average mutation rate of 3.8% (range 1.4-8.2%, Table 2). Interestingly, AKT1 mutations have been found almost exclusively in ER+ tumors (Table 2). Correlation with outcome has not been established, and intratumoral heterogeneity has not been investigated.

PIK3CA/AKT pathway mutations in carcinoma in-situ
Data from endometrial and colon carcinomas suggested that PIK3CA mutation may be a relatively late event in carcinogenesis. [1,6,78] Several studies have addressed the question of PIK3CA and AKT1 mutations in carcinoma in situ in the breast (Table 4), and have generally found the mutation frequency to be equivalent in ductal carcinoma in situ (DCIS) and invasive carcinoma. [51,63,64,77,79,80] Li et al. [64] reported mutations in 24% of DCIS cases (n=57), as compared to 27% of invasive carcinomas (n=108). Additionally, there is relatively good mutational concordance between DCIS and accompanying invasive carcinoma in the same specimen. Our group initially analyzed carcinoma in situ nearby invasive carcinomas which had PIK3CA or AKT1 mutations, and found 100% concordance in 13 cases, including 2 cases with AKT1 mutation in DCIS and invasive carcinoma, but we did not comprehensively examine pairs without mutated invasive carcinoma. [79] In addition to studying intratumoral heterogeneity, as discussed above, Kalinsky et al. [77] compared mutational status in 5 cases of paired DCIS and invasive carcinoma, and found complete concordance (all harboring mutations).
The largest study to date found similar PIK3CA mutation frequency among DCIS alone (30%), DCIS adjacent to invasive carcinoma (31%), DCIS paired with invasive carcinoma (DCIS=29%, IDC=31%) and invasive carcinoma alone (28%) [80]. Concordance between paired DCIS and invasive carcinoma in the group of pilot cases (which overall had a lower mutation frequency than listed above) was 93%, excluding 2 heterogeneous cases. In these two cases, multiple areas of DCIS and invasive carcinoma were tested, revealing mutational heterogeneity. In one case, 2 of 3 areas of in situ carcinoma were PIK3CA mutant, while the invasive component and one area of DCIS was wild type; in a second case, DCIS and one invasive sample was wild type, while the second invasive sample had a PIK3CA exon 20 H1047R mutation [80]. Their larger cohort demonstrated a lower DCIS-invasive mutational concordance of 66% [80]. Interestingly, discordant cases included 7 cases with mutant DCIS (6 exon 20, 1 exon 9) with accompanying PIK3CA wild type invasive carcinoma, and 8 cases with wildtype DCIS and PIK3CA mutant invasive carcinoma (all exon 20); 6 cases were concordant mutant (all exon 20), and 24 cases were concordant wildtype [80]. This data suggests that PIK3CA mutations in breast neoplasia often exist already at the stage of DCIS.
Although our group has not studied the identical spectrum of breast lesions, our data on a variety of breast lesions suggests that PIK3CA/AKT1 mutations may be very prevalent in proliferative breast epithelium. We first studied a group of papillary neoplasms, including benign papillomas (with or without usual ductal hyperplasia), papillomas with atypia or DCIS, and papillary carcinomas, using sensitive HPLC, or PCR-mass array based methods, with sequence confirmation [81]. We found a 66% frequency of PIK3CA and AKT1 mutations in benign papillomas [81]. Interestingly, there was a bias toward AKT1 mutations in papillomas without hyperplasia, and a bias toward PIK3CA exon 20 mutations in papillomas with usual hyperplasia. Although the rate of mutation was similar in papillomas with atypia or DCIS (11/18, 61% plus a RAS mutation), the mutation rate was lower in a small sample of papillary carcinomas, and seemingly closer to that of invasive breast carcinomas (3/10, 30% plus a RAS mutation) [81].
Our group has also analyzed columnar cell lesions (with and without atypia n=24), revealing a similarly high PIK3CA mutation frequency (54% using sensitive PCR mass-array screening methods). Of the columnar cell lesions with paired DCIS or invasive carcinoma available for analysis, concordance was only 5/14 cases. Of the discordant cases, most had PIK3CA mutant columnar cell lesion, with paired wildtype carcinoma. Preliminary data from a number of additional cases is proving confirmatory, suggesting that PIK3CA mutations may be beneficial in early breast proliferative lesions, and may have a different, perhaps lesser, role in invasive and /or metastatic carcinoma. It is unclear why our results differ from Li's group, although methodologic and target lesional differences are logical explanations.

Conclusion
Mutations of the PI3K110α catalytic subunit (PIK3CA) are very common in estrogen receptor positive and Her-2/neu positive breast cancers, and they provide an attractive target for therapy. Since the common hotspot mutations are kinase activating, it has been hypothesized that PIK3CA/AKT1 mutations play a role in breast carcinogenesis, yet these mutations paradoxically correlate with good prognosis in estrogen receptor positive tumors. Further, several lines of data from human and mouse suggest that these mutations may have a role early in breast epithelial cell proliferation, but may have a lesser role at later stages of carcinoma and metastasis. In patient tissue, PIK3CA mutations are found in found in proliferative lesions, and in DCIS, yet metastatic tumors gain and lose PIK3CA mutation at about equal frequency, suggesting lack of selective advantage in the metastatic process. Further, mammary overexpression of PIK3CA H1047R mutant results in proliferative lesions and carcinomas, yet in the inducible mouse model, carcinoma readily lose 'addiction' to the PIK3CA mutant as tumors accumulate other genetic changes. Indeed, data from whole genome sequencing of human breast cancers and metastases reveals the complexity of and heterogeneity of genetic changes, with 30-90 point mutations per tumor, non-withstanding indels, translocations, epigenetic and translational/post-translational changes. [91][92][93][94][95] Thus, while our current knowledge of the PI3K/ AKT signalling pathway and PIK3CA/AKT1 mutations provides insight into the pathogenetics of breast cancer, further experimental studies exploring pathway crosstalk and clinical trials of combination therapies utilizing PI3K/AKT inhibitors together with other agents will be necessary to develop effective targeted treatment strategies.