The phosphoinositide 3-kinase (PI3K) signaling pathway is crucial for cell growth, proliferation, metabolism, and survival, and is frequently deregulated in human cancer, including ~ 70% of breast tumors. PIK3CA, the gene encoding the catalytic subunit p110α of PI3K, is mutated in ~ 30% of breast cancers. However, the exact mechanism of PIK3CA-evoked breast tumorigenesis has not yet been defined. Genetically engineered mouse models are valuable for examining the initiation, development and progression of cancer. Transgenic mice harboring hotspot mutations in p110α have helped to elucidate breast cancer pathogenesis and increase our knowledge about molecular and cellular alterations in vivo. They are also useful for the development of therapeutic strategies. Here, we describe current mouse models of mutant PIK3CA in the mammary gland, and discuss differences in tumor latency and pathogenesis.
Cre driven by the mouse mammary tumor virus long terminal repeat
mouse mammary tumor virus
Cre driven by the mouse mammary tumor virus long terminal repeat promoter
Cre driven by the whey acidic protein promoter
Phosphoinositide 3-kinases (PI3Ks) belong to a family of lipid kinases involved in metabolism, growth, proliferation, and survival signaling. Class Ia PI3Ks phosphorylate the 3-hydroxyl group of phosphatidylinositol 4,5-bisphosphate, resulting in the production of the second messenger phosphatidylinositol 3,4,5-trisphosphate. This recruits and activates several signaling proteins, including AKT and PDK1, leading to the activation of their downstream effectors. PI3K action is reversed by the PTEN phosphatase [1, 2]. PI3Ks are heterodimers of regulatory (p85α, p50α, p55α, p85β, and p55γ) and catalytic (p110α, p110β, p110γ, and p110δ) subunits [2, 3].
Genomic alterations of components of the PI3K pathway are found in ~ 70% of breast cancers . The gene PIK3CA encodes the catalytic subunit p110α, and its amplification and/or mutation is associated with several kinds of human solid tumors [5-9]. Activating somatic mutations in PIK3CA are present in ~ 30% of human breast cancers at all stages [5, 8, 10-12]. In 47% of these cases, mutations occur in the kinase domain; the most frequent one is H1047R in exon 20. In 33% of these cases, mutations occur in the helical domain; the most frequent ones are E545K and E542K in exon 9 [8, 13]. These mutations lead to a constitutively active enzyme with oncogenic capacity in cell culture [14-16]. Alterations in PIK3CA are found at similar frequencies in pure ductal carcinoma in situ, ductal carcinoma in situ adjacent to invasive ductal carcinoma, and in invasive ductal carcinoma, indicating that PIK3CA mutations occur early in carcinoma development . In addition, mutant p110α has been found in distinct human breast cancer subtypes, such as estrogen receptor (ER)α-positive, progesterone receptor (PR)-positive, human epidermal growth factor 2 (HER2)/Neu-positive and triple-negative breast cancers [12, 18], but the correlation between PIK3CA mutations and pathological parameters remains controversial [5, 12, 19-25]. Also, assessment of the clinical outcome associated with these hotspot mutations showed contradictory results: some studies reported poor prognosis in breast cancer patients harboring PIK3CA exon 20 [26, 27] or exon 9 mutations , whereas others reported favorable prognosis with improved overall survival in patients with exon 20 mutations [10, 22]. Notably, PIK3CA mutations were shown to reduce the efficacy of HER2-targeted and ER-targeted therapies [28-30].
Adding to the list of mammary tumor mouse models , several groups have generated transgenic mice expressing PIK3CA H1047R in the mammary gland [32-36]. In contrast to mouse models such as Neu, Myc or polyoma middle T-antigen, which evoke tumors with a very specific phenotype, PIK3CA H1047R mutant mice have heterogeneous tumors. These new models should not only help to delineate the cellular and molecular mechanisms of action of mutant p110α in vivo, but also improve our understanding of tumor initiation, development and progression in breast cancer, provide a resource for the development of cancer therapies, and help to elucidate mechanisms of resistance to current PI3K pathway inhibitors.
Tumor formation in PIK3CA H1047R mutant mice
Transgenic mice expressing the amino acid substitution H1047R in the mammary gland were generated in order to model mutant PIK3CA breast cancer (Table 1; Fig. 1). For conditional mammary-specific expression of human PIK3CA or murine Pik3ca H1047R, two different promoters were used to drive Cre recombinase expression. First, Cre driven by the mouse mammary tumor virus (MMTV) long terminal repeat (MMTV-Cre) promoter results in mosaic expression of mutant PIK3CA/Pik3ca in differentiated mammary luminal cells and progenitor cells, and in other organs, depending on the MMTV-Cre line [35, 37-41]. Second, Cre driven by the whey acidic protein promoter (WAPi-Cre)  (Fig. 1A) results in expression of mutant PIK3CA in alveolar progenitor cells and differentiated secretory luminal cells . Tetracycline-inducible promoter systems (combined with MMTV-rtTA ; Fig. 1B) were also used to drive overexpression of H1047R, leading to a seven-fold to eight-fold change in expression of mutant PIK3CA as compared with Pik3ca . Other groups have used a knock-in system to express endogenous levels of Pik3ca H1047R under the control of the native promoter (combined with MMTV-Cre [35, 36, 39, 41]) (Fig. 1C,D).
Table 1. Summary of current mouse models of PIK3CA alterations. α-SMA, α-smooth muscle actin
Meyer et al. found mammary tumor-independent high lethality (~ 75%) in MMTV-Cre PIK3CA H1047R mice. Although the cause of death could not be identified, promoter leakiness leading to expression of PIK3CA H1047R in other tissues was suggested . Using two different MMTV-Cre lines [37-39] to induce expression of PIK3CA H1047 (Fig. 1E), Adams et al. found that some MMTV-CrelineAPik3ca H1047R and MMTV-CreNLSTPik3ca H1047R mice reached endpoint (lethargy, impaired breathing, tumor) independently of mammary tumors . These observations raise concerns about whether MMTV-Cre is the optimal promoter system with which to study PIK3CA-induced mammary cancer in mice.
Each of these systems leads to the development of heterogeneous mammary tumors. The most prominent phenotypes, adenosquamous carcinoma and adenomyoepithelioma, express ERα, as well as basal [e.g. cytokeratin (CK)5 and CK14) and luminal (e.g. CK8 and CK18) markers. The Pik3ca H1047R knock-in models of Yuan et al. and Tikoo et al. led mostly to hormone receptor-positive fibroadenomas (76.9% and 45%, respectively) or sarcomas (42.5%). Other histopathological features, such as adenocarcinoma, carcinosarcoma, and osteosarcoma, were also observed [35, 36] (Table 1). Heterogeneity, a feature of human breast cancer, was also reported in mouse models of Pten inactivation [44, 45]. In another study, however, loss of PTEN resulted only in adenomyoepithelioma  (for a review of PTEN mouse models, see ).
Nulliparous WAPi-Cre PIK3CA H1047R animals developed tumors at an average of 219 days (K.S. and M.B.A., unpublished data), and nulliparous MMTV-Cre Pik3ca knock-in H1047R animals developed tumors at an average of 484 and 492 days [35, 36]. Notably, tumor latencies in parous WAPi-Cre PIK3CA H1047R animals were, on average, 140 days , and those in MMTV-Cre Pik3ca knock-in H1047R animals were 392 and 465 days [35, 36], showing that pregnancy accelerates tumorigenesis in these models. An increase in the number of H1047R-expressing cells after pregnancy and a delay in involution (days 2 and 8), owing to a reduced number of apoptotic cells, were observed in WAPi-Cre H1047R mice .
Tumor formation was also investigated by using MMTV-driven expression of nonmutated p110α fused to a Src myristolyation sequence (Fig. 1F), which results in the recruitment of p110α to the membrane and constitutive activation of PI3K signaling (MMTV-MYR-p110α). Transgenic mice developed heterogeneous ER-positive mammary tumors, but at a frequency lower than in mice expressing mutant PIK3CA .
Synergism between PIK3CA H1047R and p53 alterations
Whole-exome capture and sequencing of mammary tumors from MMTV-Cre knock-in Pik3ca H1047R mice of various histotypes has revealed greater increases in somatic mutations in spindle cell tumors (~ 44–88) and adenocarcinoma (~ 4–61) than in fibroadenoma (~ 2–13) . Moreover, comparative genomic hybridization-array profiling showed a greater accumulation of chromosomal copy number alterations in spindle cell tumors than in adenocarcinoma and fibroadenoma . Functional validation and examination of the clinical relevance of these secondary genomic alterations are now warranted.
Some human breast tumors harbor alterations in PIK3CA in combination with mutant p53 [18, 49, 50]. p53 mutations (R245H, A135V, and I192N) were among the secondary mutations identified by Yuan et al.  in adenocarcinoma and spindle cell neoplasia. It is likely that these mutations prevented the well-established p53-dependent tumor suppression. Notably, Pik3ca mutant mouse models were used to investigate the interaction of Pik3ca H1047R and p53 [32, 36]. Heterozygosity in p53 was shown to accelerate tumor onset in MMTV-Cre Pik3ca H1047R mice . The tumor histotype in double mutants consisted mostly of ER-positive, CK14-positive and CK8-positive spindle cell tumors that express epithelial–mesenchymal transition markers, or adenosquamous carcinoma  (Table 1).
p53 has been found to be inactivated in MMTV-MYR-p110α-evoked tumors, suggesting that p53 loss is important for tumorigenesis in this model. Consistently, no difference was found in tumor latency or tumor phenotype between MMTV-MYR-p110α mice in a heterozygous p53 background and MMTV-MYR-p110α mice . Notably, MMTV-MYR-p110α mice in an inactive retinoblastoma (pRB) background (CDK4 R24C knock-in line [51, 52]) showed enhanced mammary tumorigenesis. These data suggest that PIK3CA H1047R-evoked tumor suppression mechanisms can be circumvented by inactivation of either p53 or pRB .
PIK3CA H1047R mutations and metastasis
Mouse models with an altered PI3K pathway can increase our understanding of breast cancer progression and metastatic spread. Metastases were reported in Pten heterozygous mice: one study found a metastatic tumor in the regional lymph node of one mouse, and three other mice had lung metastases. Their morphological appearance was similar to that of the primary tumor . In contrast, metastases are rarely found in PIK3CA/Pik3ca H1047R mouse models. There is only one report of isolated lung metastasis in MMTV-CreNLSTPik3ca H1047R mice . However, PIK3CA mutations occur at high frequencies in metastatic human breast cancer . Surprisingly, oncogenic PIK3CA-driven breast tumors have a longer time to recurrence after surgery , and some clinical studies reported a good prognosis [10, 22]. These observations may mean that mutant PIK3CA results in a selective advantage for breast cancer cells at the primary site, but not during metastatic progression and colonization of distant sites. The analysis of PIK3CA status in a large number of metastatic lesions, circulating tumor cells and matched primary tumors should clarify this ‘PIK3CA paradox’ . An alternative explanation is that patients with mutations in PIK3CA respond well to the current standard of care, resulting in this apparent paradox.
Constitutively active PI3K signaling, in association with further genomic alterations, induces mammary cancer in mice, suggesting a causative role for PIK3CA mutations in breast tumorigenesis. Additional gain-of-function or loss-of-function genomic alterations may also contribute to metastasis in breast tumor progression. The identification of these synergistic oncogenic pathways is of paramount importance for the elucidation of the ‘wiring diagram’ of tumor cells with PIK3CA mutations.
Therapeutic strategies and resistance in mutant PIK3CA mouse models
At least 26 PI3K pathway-targeting compounds are currently undergoing over 150 cancer-related clinical trials . PIK3CA mouse models also serve as a valuable tool for testing anticancer drugs [55, 56]. A test of the efficacy of the PI3K inhibitor GDC-0941 by Yuan et al. showed a decline in tumor growth in spindle cell tumors with Pik3ca H1047R and p53 mutations .
These mouse models have already proved useful for investigating resistance mechanisms. For example, PIK3CA H1047R-driven tumors were shown to recur after PIK3CA H1047R inactivation in a PI3K-dependent or PI3K-independent manner . Tumor survival in c-MET-elevated tumors was shown to depend on an active endogenous PI3K pathway, whereas c-MYC elevation contributed to oncogene independence and GDC-0941 resistance .
The PI3K-independent recurrence of PIK3CA H1047R-initiated mammary tumors shows how important it is to investigate associated pathways involved in tumor formation that may result in escape from treatment. Their delineation should pave the way for the development of mechanism-based combination therapies.
Current mouse models of mutant PIK3CA are contributing to a better understanding of breast cancer pathogenesis associated with alterations in the PI3K pathway. So far, characterization of PIK3CA H1047R-evoked tumors has clarified some pathophysiological aspects, but has also raised further relevant questions.
Tumor heterogeneity and the observation that PIK3CA H1047R mutants develop CK5/CK14-positive and CK8/CK18-positive mammary carcinomas [32, 34-36] suggest either a luminal and basal tumor cell of origin or the dedifferentiation of cell lineage-committed tumor cells to multipotent progenitors that then give rise to CK5/CK14-positive and CK8/CK18-positive cells. Consistent with the latter possibility, Tikoo et al. found that expression of Pik3ca H1047R results in expansion of the luminal progenitor population. Furthermore, the putative mammary stem cell-enriched basal population and the luminal progenitors of mutants show enhanced colony-forming ability and a larger colony size , but the molecular mechanisms underlying these effects have not yet been defined. The current mouse models do not definitively answer the question of what causes heterogeneity in PIK3CA H1047R-evoked tumors and which cell type gives rise to which subtype of mammary cancer. However, lineage-tracing experiments should provide further information about the cell of origin and cellular hierarchy in tumors.
There have been few studies of synergistic pathways contributing to tumor progression in vivo, especially metastatic spread, and to resistance to therapy. Such investigations should lead to an increase in our understanding of tumor relapse and therapy failure, and should help to define targets for preventing PIK3CA-driven tumor initiation and progression.
We thank members of the Bentires-Alj laboratory for fruitful discussions. Research conducted in the laboratory of M. Bentires-Alj is supported by the Novartis Research Foundation, the European Research Council (ERC starting grant 243211-PTPsBDC), the Swiss Cancer League, and the Krebsliga Beider Basel.