PI3K Positively Regulates YAP and TAZ in Mammary Tumorigenesis Through Multiple Signaling Pathways

Breast cancer is one of the most frequently diagnosed cancers in women worldwide, accounting for 25% of all cancer cases and 15% of cancer mortalities in women (1). PIK3CA, a catalytic subunit of PI3K, is one of the most frequently mutated genes identified in breast cancers (2). One of the most common and best-characterized mutations of PIK3CA identified in breast cancer is H1047R in the kinase domain (3), which causes increased catalytic activity. PIK3CB, another major catalytic subunit of PI3K, is also aberrantly activated in breast cancer cells; however, this is more commonly found to occur secondary to overexpression and/or gene amplification (3). Regardless of the mechanism, both the PIK3CA-H1047R mutation and PIK3CB overexpression result in hyperactivation of PI3K, enabling further exertion of its oncogenic functions (e.g., cell proliferation, antiapoptosis, and angiogenesis) through the PI3K–PDK1–AKT (PI3K-AKT) signaling pathway (4). In this pathway, the activation of PI3K can phosphorylate PI (3, 4) P2 into PI (3–5) P3, which then interacts with PDK1 and AKT, recruiting them to the inner leaflet of the cell membrane, where PDK1 phosphorylates and activates AKT. Activated AKT subsequently phosphorylates a variety of downstream genes to cause increased cell proliferation, diminished apoptosis, and other oncogenic functions (4, 5). Moreover, the activation of the PI3K–AKT pathway is frequently correlated with resistance to various anticancer therapies (6–8). Given this, many studies have been carried out to develop agents inhibiting PI3K-AKT for cancer treatment (3, 5, 9–12). Unfortunately, the existing drugs targeting the PI3K–AKT pathway often result in the development of drug resistance (13), through unknown mechanisms. More studies are therefore necessary to discern the molecular network of PI3K in breast cancer development and progression to design more efficient strategies for the treatment of PI3K-involved breast cancer.

YAP and its paralog TAZ, two well-known transcriptional coactivators and effectors of the Hippo pathway, are involved in tumorigenesis of breast cancer (14–17). Through interactions with transcription factors of the TEAD family, YAP and TAZ can induce transformation and epithelial–mesenchymal transition (EMT) of human immortalized mammary epithelial MCF10A cells (18, 19) as well as cellular resistance to chemotherapeutic drugs such as taxol and doxorubicin (16–18, 20). YAP/TAZ are negatively regulated by the Hippo signaling pathway. In this pathway, MST1/2 serine/threonine (S/T) kinases, the mammalian homologs of Drosophila Hippo, phosphorylate and activate LATS1/2. LATS1/2 subsequently phosphorylate S127/S89 of YAP/TAZ and prevent them from translocating to the nucleus to activate transcription of downstream genes (19, 21, 22). Although an increasing number of negative regulators of YAP and TAZ have been identified (23–31), there are few known positive regulators of YAP and TAZ. Recent studies have found that PIK3CA can positively regulate YAP in response to mitogen signals such as EGF treatment (32). Furthermore, coactivation of PIK3CA and YAP has been shown to promote liver carcinogenesis in mice and in human patients (33). However, the signaling pathway mediating the regulation of YAP by PIK3CA is not clear. PIK3CB has been found to be a major target of YAP in cardiomyocyte proliferation and survival (34), but there is currently no evidence showing PIK3CB can regulate YAP in tumorigenesis. Therefore, the role of PIK3CA/B in regulation of YAP and its paralog TAZ in mammary tumorigenesis is not yet known.

In this study, we performed a systematic gain-of-function screening of kinases involved in mammary tumorigenesis. PIK3CB was identified as a kinase of interest. We further determined that PIK3CB positively regulates YAP and TAZ to promote mammary tumorigenesis both in vitro and in vivo. Similarly, PIK3CA-H1047R was found to positively regulate YAP and TAZ in mammary tumorigenesis. This study implicates YAP/TAZ in PI3K-related breast cancer and provides a new rationale for targeting YAP/TAZ for breast cancer treatment.

HEK293, SK-BR3, MCF10A, were purchased from ATCC about 10 years ago. Passages 10–20 for HEK293, SK-BR3 and passages 105–115 for MCF10A were used in our experiment; HEK293A, HEK293A-LATS1/2-knockout (LATS-KO), HEK293A-MST1/2-KO (MST-KO) were generously provided by Dr. Kunliang Guan, passages 20–30 were used for our experiment; HCT116, HCT116-AKT-KO cells were gifts from Dr. Bert Vogelstein; MMTV-NIC cells were provided by Dr. Christopher Nicol. No authentication and Mycoplasma tests have been performed for all cell lines. HEK293, HEK293A, HEK293A-LATS1/2-knockout (LATS-KO), HEK293A-MST1/2-KO (MST-KO), and HEK293T cells were cultured in DMEM (Sigma-Aldrich) supplemented with 10 % heat-inactivated fetal bovine serum (FBS) (Sigma-Aldrich) and 1% penicillin/streptomycin (P/S). Subcultivation was performed every 3 days at the ratio of 1:10. SK-BR3 and HCT116, HCT116-AKT-KO cells were maintained in McCoy 5A modified media (Sigma-Aldrich) with 10% FBS and 1% penicillin/streptomycin added. Cells were passed every 3 days at the ratio of 1:3-1:10. MMTV-NIC and MCF10A (human immortalized mammary epithelial cells) were cultured in DMEM/Nutrient Mixture F12 Ham (Sigma-Aldrich) supplemented with 5% horse serum (HS) (Invitrogen), 20 ng/mL hEGF, 500 ng/mL hydrocortisone, 10 μg/mL insulin, 2.5 mmol/L l-glutamine, 100 ng/mL cholera toxin, and 1% P/S. Cells were maintained in a 37 °C incubator with 5% CO₂. Cells were normally subcultivated every 4 days with ratio of 1:10. All cell lines were cultured less than 2 months for experiments.

PIK3CB cDNA was amplified from a human kinase open reading frame (ORF) library (KOL, Addgene kit # 1000000014) with a FLAG tag in the forward primer and cloned into BamHI/MluI site of the WPI vector (containing hygromycin B resistant gene) or BamHI/NotI site of the pcDNA3 vector. PIK3CB-D937A was generated through site-directed mutagenesis. Primers used are as follows: PIK3CB forward primer: 5′-GAAGATCTACCATGGACTACAAAGACGATGACGACAAG ATGTGCTTCAGTTTCATAATG-3' (BglII cut site is underlined, FLAG tag is in italics); reverse primer: GTAATCATGCGGCCGCCGACGCGT TTAAGGTCCGTAGTCTTT

CCGAAC (NotI cut site is underlined, MluI cut site is in italics); PIK3CB-D937A forward primer: CAGCTCTTCCACATTGCCTTTGGACATATTCTT; reverse primer: AAGAATATGTCCAAAGGCAATGTGGAAGAGCTG (mutation site is underlined).

PIK3CA-H1047R-pBabe plasmid was a gift from Dr. Leda Raptis (Queen's University, Kingston, Ontario, Canada). PIK3CA was amplified from Myr-PIK3CA-pLNCX plasmid (gift from Dr. Raptis) using PCR and cloned into pcDNA3 at BamHI and NotI sites. PIK3CA-D933A was generated through site-directed mutagenesis and cloned into pcDNA3 vector. Myr-AKT1, Myr-AKT2, and AKT2-KD (kinase dead) plasmids were provided by Dr. Eric Asselin (University of Quebec at Trois-Rivieres, Trois-Rivieres, Quebec, Canada); PDK1-WT, PDK1-KD (K110N), PDK1-ΔPH (PH domain deletion) plasmids were generous gifts from Drs. Paolo Armando Gagliardi and Luca Primo (University of Turin, Turin, Italy) (35, 36).

The human kinase ORF library was purchased from Addgene (Kit #1000000014), which contains 558 human kinase ORFs in the pDONR-223 Gateway Entry vector in six 96-well plates with each well containing one kinase in bacteria. To further modify the library, 1 μL from each well of kinase was mixed together to make a pool of whole human kinase ORFs before being spread onto 10 × 150 mm spectinomycin-containing (50 μg/mL, Sigma) LB plates. Bacterial colonies were directly scraped off the LB plates to avoid over-representation of certain kinases during liquid culture, and then subjected to Midiprep (Qiagen) for plasmid extraction. The kinase cDNAs from the ORF library were subsequently transferred from entry vector (pDONR-223) into V5-tagged expression vector pLX304 through Gateway LR ligations (Invitrogen) according to the manufacturer's protocol. Plasmids were then electroporated into competent cells (Lucigen) and grown on 10 × 150 mm LB plates with ampicillin (100 μg/mL). Kinase overexpressing library (KOL) plasmids were collected by Midiprep. The newly established KOL was further validated by amplifications of several kinase ORFs in the library by PCR (data not shown).

Retroviral and lentiviral production, titration and infection of overexpressing constructs were as described (22). To knock down YAP and TAZ in MCF10A-WPI and MCF10A-PIK3CB-WPI cells, sgRNAs targeting YAP and TAZ were cloned in lentiCRISPRv2.0 (a gift from Dr. Feng Zhang, Addgene plasmid #52961) and were used for producing lentivirus to infect the cells. MCF10A-PIK3CA-H1047R was established through retroviral infection. Overexpression of TAZ in MCF10A-PIK3CB and MCF10A-PIK3CA-H1047R was established by infecting MCF10A-PIK3CB and MCF10A-PIK3CA-H1047R with TAZ-WPI lentivirus. For the establishment of MCF10A cells stably expressing KOL, MCF10A cells were infected with KOL lentivirus at a multiplicity of infection (MOI) of 0.3 with a 100-fold coverage (100 cells infected with virus expressing each kinase). Infected cells were selected with blasticidin (10 μg/mL) for five days.

RNA extraction and qRT-PCR were as described (22). mRNA expression levels of experimental cells were normalized to control cell lines. Results are represented as fold change.

Protein extraction and Western blotting were as described (22). Antibodies targeting YAP and CYR61 are from Santa Cruz Biotechnology, 1:1,000; antibodies against p-YAP-S127, S6K, pS6K, AKT1, AKT2, pAKT, PIK3CA, PIK3CB, LATS1, cleaved-PARP, MST1 and MST2, are from Cell Signaling Technology, 1:1,000; LATS2 antibody used is from Bethyl Laboratories (1:2,500); TAZ antibody is from BD Biosciences (1:1,000); FLAG and HA antibodies are from Sigma (1:1,000). For co-immunoprecipitation (co-IP), HEK293 cells with different transfections were grown to 85% confluency and were lysed in NP40 lysis buffer [50 mmol/L Tris-HCl (pH 8.0), 150 mmol/L NaCl, 1% NP40, 10% glycerol, dH₂O] with complete EDTA-free protease inhibitor (Roche) added. Lysates were incubated overnight with protein A or G agarose beads (Roche). Immunoprecipitates were subsequently analyzed by Western blot analysis.

For cell proliferation assay, a triplicate of 1 × 10³ MCF10A-WPI, TAZ, PIK3CA, PIK3CB, PIK3CA-TAZ, and PIK3CB-TAZ cells were seeded into each well of 24-well plates. Cell numbers were counted on days 1, 3, 4, 5, and 6 after plating. Data are shown as mean ± SD and experiments were repeated at least three times.

For anoikis cell death analysis, triplicates of 2 × 10⁵ cells were plated into ultra-low attachment flask (Corning) and culture for 60 hours. 1/10 of the floating cells were taken for Trypan blue cell viability assay. The remaining cells were lysed and used for Western blot analysis to check expression of apoptotic marker cleaved-PARP.

MCF10A cells stably overexpressing KOL (1 × 10⁵ cells/100-mm plate) or WPI control, TAZ-WPI, PIK3CB-V5, PIK3CB-V5-WPI, PIK3CB-TAZ-WPI, PIK3CA-H1047R, or PIK3CA-H1047R-TAZ-WPI (2 × 10⁴ cells/well of 6-well plate) were plated in complete media containing 0.4% agarose overtop of a 0.8% agarose layer. Medium was refreshed every three days. At around 20–30 days, colonies were visible. For screening purposes, single colonies were picked with a P-10 tip and transferred into single wells of 96-well plate for further cell growth. These cells were passed to bigger plates when they were confluent. For soft agar assays with other cell lines, colonies were stained with crystal violet (0.005% crystal violet in 20% methanol). Pictures of these stained colonies were further captured under white light using the Bio-Rad Gel Doc System (Bio-Rad) and quantified using Quantity one software. Counted colonies were averaged and summarized in data bars. Data were statistically analyzed with unpaired student t test, where “*” indicates P value is less than 0.05. The experiments were repeated at least twice.

Triplicates of either HEK293, HEK293A-related cells, HCT116-related cells (5 × 10⁵ cells/well of 12-well plate) or SK-BR3 (6 × 10⁴ cells/well of 12-well plate) were transfected with CYR61-luc (0.1 μg), super TEAD binding site (STBS)-luc (0.1 μg) alone or together with other plasmids (0.4 μg) using Polyjet (SignaGen). In addition, 10 ng of Renilla luciferase vector (pRL-TK) was cotransfected in each sample as an internal transfection control. Approximately 40 hours posttransfection, luciferase activity was measured with the Dual Luciferase Reporter Assay System (Promega) and Turner Biosystems 20/20 luminometer. Fold change values were calculated by normalizing experimental samples to controls transfected with reporter plasmids alone.

A xenograft mouse model was used to evaluate the tumor formation of a variety of transformed MCF10A cells; MCF10A/TAZ, MCF10A/PIK3CA, MCF10A/PIK3CB, MCF10A/PIK3CA+TAZ and MCF10A/PIK3CB+TAZ. A total of 4 × 10⁶ cells suspended in 200 μL of cold PBS and GFR phenol-red free Matrigel (1:1 dilution) were injected subcutaneously into the flanks of 12-week-old Rag2^−/−;Il2rg^−/− mice (n = 3/group). General health of these animals and palpable tumor volume was assessed twice a week. Six weeks postinjection, mice were sacrificed and tumors were harvested. Tumors were fixed in formaldehyde, paraffin-embedded, and sectioned for hematoxylin and eosin (H&E) staining and IHC analysis. All procedures were approved by the Queen's Animal Care Committee in accordance with Canadian Council on Animal Care guidelines.

Formalin-fixed paraffin-embedded (FFPE) tumor tissues were sectioned at the thickness of 3–4 μm and were stained by the Discovery XT Automated IHC/ISH research slide staining system (Ventana Medical Systems, Inc.). For pathology analysis, sections were stained with H&E. For IHC, antigens were retrieved with EDTA (pH = 8) solution, blocked by 1% BSA (Fraction V), and incubated with mouse monoclonal anti-TAZ antibody (1:2,500 dilution, BD Biosciences) or rabbit monoclonal anti-PIK3CB (1:100 dilution, Abcam) antibody. As a control for specificity, one slide was processed with the same IHC conditions without a primary antibody. IHC signals were developed by using biotinylated HRP-conjugated anti-mouse or anti-rabbit secondary antibody, respectively, followed by catalyzing 3,3'-diaminobenzidine (DAB) substrate-chromogen into a visible precipitate. Slides were scanned and the images were uploaded to Spectrum system for picture capture and analysis.

Statistical significance was determined using an unpaired Student t test. A P value less than or equal to 0.05 is considered as significant. “*” indicates P value < 0.05.

To screen for kinases causing transformation in mammary cells, we established a heterogeneous population of nontumorigenic MCF10A mammary cells infected with a KOL library. After antibiotic selection, surviving cells were subjected to soft agar assay to assess cellular transformation. Thirty days after culture, colonies representing transformed clones became visible. Single colonies were isolated and expanded. Protein lysates were collected from each clonal cell line of interest and the identity of the transformation-inducing kinase was determined by immunoprecipitation using a V5 antibody and mass spectrometry (MS).

Among the kinases detected by MS, PIK3CB was identified as the transformation-inducing kinase in two different clonal cell lines. PIK3CB is a catalytic subunit belonging to PI3K family class IA (3). Given that the mechanisms underlying PIK3CB-induced mammary tumorigenesis are largely unknown, we chose to further evaluate the function of this kinase in our cell lines.

To confirm the results of our screen, we first subjected cells from the two PIK3CB-overexpressing colonies (MCF10A-PIK3CB-V5-C-1/C-2) to soft agar assay alongside a control cell line. After 20 days of culture, colonies of MCF10A-PIK3CB-C-1/C-2 cells were visible (Fig. 1A and B). Thus, PIK3CB indeed enhances anchorage-independent growth of MCF10A cells.

We have previously found that overexpression of the Hippo transducers YAP and TAZ is also sufficient for MCF10A cellular transformation. Indeed, YAP/TAZ overexpression in MCF10A results in similar tumorigenic phenotypes to PIK3CB overexpression in MCF10A cells (18, 37). Thus, we next investigated whether PIK3CB might promote mammary tumorigenesis through modulation of YAP and TAZ function.

As previously described YAP and TAZ are translocated to nucleus after activation. Therefore, we first investigated whether PIK3CB overexpression might affect translocation of YAP/TAZ into the nucleus. Wild-type MCF10A (MCF10A-WT) and MCF10A-PIK3CB-C-1 cells were fixed and the subcellular locations of YAP/TAZ were examined through indirect fluorescence staining. PIK3CB dramatically increased YAP/TAZ nuclear localization (Fig. 1C; Supplementary Figs. S1A and S1B and S2A and S2B), suggesting that YAP and TAZ may be activated after PIK3CB overexpression.

To further test this hypothesis, we evaluated YAP/TAZ transcriptional coactivation using promoter reporter assays. First, a promoter reporter containing 14 TEAD DNA-binding sequences (STBS) upstream of a luciferase gene (STBS-luc) was used (38). PIK3CB was cotransfected with STBS-luc into HEK293A. Notably, PIK3CB significantly increased STBS-luc luciferase activity (Fig. 1D). We also examined the effect of PIK3CB on the promoter of CYR61—a YAP/TAZ target gene that we have previously characterized (17). Consistent with what was observed using STBS-luc, PIK3CB activates the CYR61 promoter (Fig. 1E). Similar results were obtained in a breast cancer cell line, SK-BR3 (Fig. 1F). Finally, we also determined that PIK3CA active mutant (PIK3CA-H1047R) enhances CYR61 promoter activity (Fig. 1G). Therefore, both PIK3CA and PIK3CB positive regulate YAP/TAZ.

To confirm that PIK3CA/B modulates YAP/TAZ function, we tested whether loss of PIK3CA/B reduces YAP/TAZ activity. We treated MCF7 breast cancer cells with a PI3K inhibitor, GDC0941, and assessed YAP S127 phosphorylation (YAP-pS127; LATS phosphorylation site). Interestingly, the PI3K inhibitor abolished the phosphorylation of two known PI3K downstream targets, AKT and S6K, and increased YAP-S127 phosphorylation (Fig. 2A). Likewise, genetic knockout of both PIK3CA and PIK3CB using CRISPR-Cas9 in HEK293A cells also increased phosphorylation of YAP at S127 (Fig. 2B). Furthermore, PI3K inhibitor treatment suppressed PIK3CB-induced nuclear translocation of YAP/TAZ in MCF10A-PIK3CB cells (Fig. 2C; Supplementary Figs. S1 and S1B and S2A and S2B). Thus, inhibition of PIK3CA/B regulates YAP phosphorylation status and YAP/TAZ subcellular localization, suggesting that the activation of PIK3CA/B decreases LATS-induced inhibitory phosphorylation of YAP/TAZ.

To further examine the relationship between PI3K and YAP/TAZ, we evaluated YAP/TAZ status in cells with endogenous activation of PI3K. Previous studies have shown that PI3K is activated in cells with HER2 overexpression (39). Therefore, we used cells isolated from tumors formed in an MMTV-driven HER2-overexpresion breast cancer mouse model (MMTV-NIC). In this model, overexpression of HER2 increased AKT phosphorylation/activation (Fig. 2D) and caused YAP/TAZ to be localized in the nucleus in a PI3K-dependent manner (Fig. 2E; Supplementary Figs. S1C and S2C and S2D).

PI3K kinase activity is critical for its functions in the PI3K–AKT pathway. To examine whether regulation of YAP/TAZ by PI3K depends on its kinase activity, we generated MCF10A cell lines stably overexpressing wild-type PIKCB (PIK3CB-WT) and a kinase-dead construct of PIK3CB (PIK3CB-D937A). As shown in Fig. 3A, overexpression of PIK3CB-WT, but not PIK3CB-D937A, reduced pS127-YAP and upregulated CYR61 expression (Fig. 3A and B). Similarly, PIK3CB-D937A and kinase-dead PIK3CA (PIK3CA-D933A) could not activate the CYR61 promoter in HEK293 (Fig. 3C). Therefore, PIK3CA/B regulates YAP and TAZ through their kinase activity.

As described previously, PI3K signals through PDK1 and AKT to regulate tumorigenesis (5). Therefore, we examined whether PIK3CB regulates YAP and TAZ through PDK1 and AKT. PDK1 inhibition dramatically reduced PIK3CA/PIK3CB-induced activation of the CYR61-luc construct (Fig. 4A and B). Unfortunately, we failed to establish stable HEK293 or HEK293A cell lines with PDK1 CRISPR knockout, possibly due to the essential functions of PDK1 in HEK293A/HEK293 cells. Nonetheless, we further examined how PDK1 regulates YAP/TAZ using MCF10A cell lines stably overexpressing wild-type (MCF10A-PDK1-WT), membrane binding domain deletion (MCF10A-PDK1-ΔPH) or kinase-dead PDK1 (MCF10A-PDK1-KD; Fig. 4C). Immunofluorescence staining indicates that YAP nucleus localization is dramatically increased in PDK1 and PDK1-ΔPH–overexpressing MCF10A cells but decreased in MCF10A-PDK1-KD (Fig. 4D and E), suggesting that PDK1 regulates YAP and TAZ through its kinase activity. Consistent with this, PDK1-KD showed reduced CYR61-luc reporter activity compared with the wild type and PDK1-ΔPH constructs (Supplementary Fig. S3A). Therefore, PDK1 kinase activity plays an essential role in the regulation of YAP/TAZ.

Next, we explored signaling downstream of PDK1 through AKT. AKT knockout in HCT116 colon cancer cells reduced PI3KCA/B-induced activation of the CYR61-luc reporter (Fig. 5A and B). Interestingly, AKT knockout in this cell line decreased but did not abolish activation of the reporter by PIK3CA/B suggesting either AKT knockout is incomplete in this model or that AKT is not the sole factor downstream of PI3K that regulates YAP/TAZ.

We further examined whether the upstream kinases of the Hippo pathway (MST and LATS) are essential for PIK3CA/B-induced activation of YAP/TAZ. We cotransfected PIK3CA-H1047R or PIK3CB with CYR61-luc into HEK293A-WT, HEK293A-MST knockout (MST-KO) or HEK293A-LATS knockout (LATS-KO) cell lines (Fig. 5C and D). As shown in Fig. 5E, PIK3CA/B-induced CYR61 promoter activity in HEK293A-MST-KO cells was similar to that in HEK293A-WT cells (Fig. 5E), suggesting that PI3K activates YAP/TAZ activity independent of MST. Surprisingly, PIK3CA/B-induced activation of CYR61 promoter activity was enhanced rather than abolished in LATS-KO cells (Fig. 5E), suggesting that PI3K may also activate YAP/TAZ through other proteins in the absence of LATS.

To further elucidate the signaling between PDK1, AKT, MST, and LATS, we cotransfected AKT1 and PDK1 alongside CYR61-luc into HEK293A-WT, LATS-KO, and MST-KO cells. Interestingly, AKT showed no activation of the reporter in HEK293A-WT cells and only marginal activation in MST-KO cells, but did show dramatic activation in LATS-KO cells (Fig. 5F). Indeed, cotransfection of AKT1 with YAP increased YAP levels in LATS-KO cells but not in WT cells (Fig. 5G). This suggests that LATS suppresses the function of AKT in YAP/TAZ regulation. However, PDK1 showed similar activation of the reporter in both WT and MST-KO cells (Fig. 5F), suggesting that MST does not mediate PDK1 activation of YAP/TAZ. Like AKT, PDK1 induced dramatic activation of YAP/TAZ reporter in LATS-KO cells (Fig. 5F), which is consistent with the result of PI3K-induced activation of YAP/TAZ reporter in LATS-KO cells (Fig. 5E), suggesting there exists a PI3K–PDK1–AKT–dependent signal activating YAP/TAZ, which is antagonized by LATS.

The observation that PDK1 but not AKT activated the reporter in WT cells indicates there are likely other components in addition to AKT mediating the effect of PDK1 on YAP/TAZ. As PDK1 affects subcellular localization of YAP/TAZ (Fig. 4D and E), we investigated whether there is any interaction between PDK1 and LATS (independent of AKT) in YAP/TAZ regulation. LATS and YAP were cotransfected with different PDK1 constructs and binding between these proteins was assessed by co-IP. Although differences in PDK1 constructs did not affect binding between LATS and YAP, we noted that YAP-S127 phosphorylation by LATS is decreased in PDK1-expressing cells (Supplementary Fig. S3B and S3D). Thus, PDK1 can activate YAP through LATS.

Given that AKT also seems to be involved in PI3K–PDK1–YAP/TAZ signaling (Fig. 5A and B) and activates YAP/TAZ in LATS-KO cells (Fig. 5F), we tested whether AKT and LATS interact. LATS1 or LATS2 expression was not affected by coexpression of AKT1, AKT2, or AKT2-kinase dead (KD) constructs in HEK293 (Supplementary Fig. S4A). Moreover, we did not observe any differences in LATS1 or LATS2 phosphorylation phospho-tag resolving gel (Supplementary Fig. S4B). Thus, it is unlikely that AKT directly interacts with LATS to affect the activation of YAP/TAZ.

The interactions between PI3K, PDK1, AKT, LATS, YAP/TAZ may be further complicated by feedback pathways. Indeed, it has been shown that overexpression of YAP/TAZ upregulates LATS activity, which in turn suppresses YAP/TAZ function (40). Thus, it is possible that in LATS WT cell lines, the increased YAP expression by AKT1 triggers the activation of LATS, which in turn inhibits YAP activation, such that YAP activity appears unaltered. However, in LATS-KO cells, this feedback loop is disrupted such that YAP expression levels are not suppressed by LATS, leading to activation of YAP by AKT1. To test this hypothesis, we transfected YAP5SA, a YAP construct with all five LATS phosphorylation sites mutated, alongside AKT1 into LATS-WT cells. As expected, AKT1 increased YAP5SA expression in these cells (Fig. 5H). These results are consistent with a model in which AKT1 can activate YAP by increasing YAP expression levels, and that activated YAP can trigger LATS expression, which in turn inhibits YAP function. This signaling may also explain why AKT1 dramatically increases CYR61-luc activity in LATS-KO cells but not in WT cells.

We next set out to determine whether AKT1 kinase regulates the expression of YAP through phosphorylation. We chose to use YAP5SA to test for phosphorylation by AKT to avoid confounding signals due to basal phosphorylation of YAP by LATS (Supplementary Fig. S5A). Interestingly, no YAP phosphorylation was detected with AKT1 coexpression in HEK293, suggesting the regulation of YAP by AKT1 is not due to phosphorylation (Supplementary Fig. S5A). Moreover, we did not observe any direct binding between AKT1 and YAP/YAP5SA (Supplementary Fig. S5B and S5C) or AKT1 and TAZ/TAZ4SA (Supplementary Fig. S5C and S5D) indicating that AKT1 regulates YAP/TAZ indirectly. Thus, further studies will be required to understand the precise mechanisms of AKT1-induced YAP upregulation.

To further confirm that YAP and TAZ play essential roles in mediating PIK3CB functions, we knocked out YAP and TAZ in MCF10A-PIK3CB-WPI cells to generate MCF10A-PIK3CB-WPI-sgYAP/TAZ (Supplementary Fig. S6A). Significantly, YAP/TAZ double knockdown dramatically decreased PIK3CB-induced anchorage-independent growth of MCF10A cells (Fig. 6A and B). Similar results were obtained using MCF10A-PIK3CA-H1047R cells (Supplementary Fig. S6B and S6D). Conversely, TAZ overexpression in MCF10A-PIK3CA-H1047R and MCF10A-PIK3CB cells (Supplementary Fig. S6E) increased both PIK3CA- and PIK3CB-induced anchorage-independent growth of MCF10A cells (Fig. 6C and D). Likewise, MCF10A with TAZ and PI3K coexpression significantly increased cell proliferation compared with MCF10A expressing either construct alone (Fig. 6E and F). Moreover, TAZ coexpression with PIK3CA or PIK3CB significantly decreased cell death as demonstrated by trypan blue assay and cleaved-PARP (c-PARP) expression when these cells were cultured in ultra-low attachment flask, suggesting coexpression of TAZ and PI3K decreases anoikis (Fig. 6G and H). Finally, we tested whether coexpression of TAZ and PI3K was sufficient for tumorigenesis in vivo. Notably, MCF10A-PIK3CB-TAZ cells formed tumors after 6 weeks in immunocompromised mice (Fig. 7A). These tumors had strong expression of both TAZ and PIK3CB and resembled a high-grade poorly differentiated mammary carcinoma in pathologic appearance.

Collectively, these results show that TAZ and PI3K synergistically increase malignant cell phenotypes and that YAP and TAZ are functional components mediating PI3K-induced mammary tumorigenesis.

YAP and TAZ are negatively regulated by LATS in the Hippo pathway (19, 21, 22). Mounting evidence suggests that YAP and TAZ can be also negatively regulated by other proteins, such as AMOT (41), α-catenin (24) and Cdk1 (30, 31). To date, a limited number of positive regulators of YAP and TAZ have been identified. In this study, we identified novel positive regulation of YAP and TAZ by kinase PIK3CB and its homolog PIK3CA. Moreover, we found that the activation of YAP and TAZ by PIK3CA and PIK3CB occurs through PI3K–PDK1–AKT signaling, in which AKT positively regulates the expression level of YAP to affect its activation status (Fig. 7B). This regulation was also affected by the negative feedback regulation between YAP/TAZ and LATS. In LATS-KO cells, which have no negative feedback, PI3Ks, PDK1, and AKT1 dramatically increased YAP/TAZ activity (Fig. 5E and F). Interestingly, we found that the regulation of YAP by AKT does not occur through direct phosphorylation (Supplementary Fig. S5A). No direct interaction between AKT1 and YAP/TAZ was found (Supplementary Fig. S5B-D), suggesting an indirect regulation of YAP by AKT. Therefore, more studies are required to discern the exact mechanism by which AKT regulates YAP expression levels.

We also found that PDK1 inhibition abolished PI3K-induced YAP/TAZ activation (Fig. 4A), indicating that PDK1 is the major downstream player in PI3K regulation of YAP/TAZ. However, unlike AKT1, both PI3Ks and PDK1 activated YAP/TAZ activity in HEK293A-WT cells, which have intact LATS-YAP/TAZ negative feedback regulation (Fig. 5E and F). Therefore, PDK1 may negatively regulate LATS to affect the negative feedback regulation. This notion was supported through co-IP experiments, in which PDK1 coexpression decreased LATS phosphorylation of YAP without affecting the interaction of LATS and YAP (Supplementary Fig. S3B and S3D). In previous studies, it has been shown that PKCs, downstream targets of PDK1, can regulate LATS to affect YAP activity (42, 43). Thus, we hypothesize that LATS may be negatively regulated by PDK1 indirectly and therefore involved in PI3K-YAP/TAZ regulation (Fig. 7B). One study suggests that the subcellular location of PDK1 but not its kinase activity mediates PIK3CA activation of YAP in response to mitogen treatment (32). In this study, the cytoplasm-localized PDK1 may serve as a scaffold protein to form a complex with Sav, MST, and LATS. In this complex, LATS can be activated by MST and therefore triggers the inhibition of downstream YAP and TAZ (32). Our study, however, indicates that PDK1-KD but not PDK1-ΔPH, which loses the membrane binding motif to mimic the cytoplasm-localized PDK1, decreased the activity of YAP/TAZ (Fig. 4C–E; Supplementary Fig. S3). This suggests that PDK1 kinase activity rather than subcellular location of PDK1 plays an essential role mediating the regulation of YAP/TAZ by PI3Ks in our condition of mammary tumorigenesis. While future work will be necessary to reconcile the differences in signaling between our study and others, we suspect that these observations may be due to cell line–specific or context-specific signaling.

As PI3K catalytic subunits, PIK3CA and PIK3CB share common properties and play essential roles in PI3K activation. The PI3K–AKT signaling pathway is a major pathway downstream of activated HER2, which is involved in 20% of breast cancer (44). Significantly, PIK3CA-H1047R has been found involved in HER2-targeted therapy (6). Moreover, both PIK3CA and PIK3CB are involved in breast cancer development and progression. About 30% of breast cancers have active mutations of PIK3CA, and these tend to be luminal-type breast cancers (45). Amplification of PIK3CB is identified in around 5% of breast cancer patients and is correlated with the Luminal B subtype (46). Both active mutation and amplification of PI3K isoforms result in highly active functions of the kinase. In this study, we found that PI3K can positively regulate YAP and TAZ (Figs. 1 and 2; Supplementary Fig. S1 and S2). Moreover, coactivation of TAZ can dramatically enhance the oncogenic functions of PIK3CA or PIK3CB both in vitro and in vivo (Figs. 6C–H and 7A), while knock down of YAP and TAZ dramatically decreased PI3K-induced anchorage-independent growth (Fig. 6A and B; Supplementary Fig. S6A and S6D). Surprisingly, although both PIK3CA and PIK3CB activate YAP/TAZ and function similarly in cells, co-overexpression of TAZ with PIK3CB instead of PIK3CA causes tumor formation in mice in vivo (Fig. 7A). The reason for this difference is unknown. It has previously been suggested that loss of PTEN tumor suppressor–induced tumor formation in vivo depends on PIK3CB instead of PIK3CA (47). Therefore, it is possible that PIK3B may have higher oncogenic activity than PIK3CA due to stronger activation of YAP/TAZ in vivo.

Finally, we analyzed human tumor databases and found a significant co-occurrence of PIK3CA/B and YAP/TAZ mRNA expression in breast cancer (Supplementary Table S1). It is therefore highly possible that activated PI3K positively regulates YAP and TAZ to result in tumorigenesis, either through an overexpression of PIK3CB or an active mutation of PIK3CA. Our data suggests that the resistance or relapse of cancers in patients treated with specific PI3K inhibitors may be due to the activation of YAP/TAZ through other complimentary pathways when PI3K is inhibited. Therefore, in addition to developing specific inhibitors targeting PIK3CB or PI3K signaling, our study highlights the potential for TAZ and YAP to be targeted therapeutically for the treatment of PI3K-involved breast cancers, HER2-positive breast cancers and other cancers that are resistant to PI3K inhibitor treatments. The activation status of TAZ and YAP may also be good biomarkers for clinical prognosis of these types of breast cancers.

In conclusion, our study provides the first evidence that PIK3CB functions as a novel regulator of YAP and TAZ in breast cancer development. Moreover, we clarified the signaling pathways that mediate PI3KCA/CB regulation of YAP/TAZ. Most significantly, we found TAZ enhances PI3K functions in mammary tumorigenesis both in vitro and in vivo, and knockdown of YAP/TAZ dramatically blocked PI3K-induced mammary tumorigenesis. Amplification of PIK3CB is found in luminal B subtype of breast cancer and mutations of PIK3CA (e.g., PIK3CA-H1047R) mostly exist in luminal types of breast cancer. In addition, it is well known that PIK3CA mutations (e.g., PIK3CA-H1047R) play an important role in resistances of the targeted therapy for HER2 positive breast cancer due to the well-studied mechanism that HER2 functions through PI3K–AKT pathway. Therefore, our studies suggest that YAP/TAZ are critical mediators of HER2- and PI3K-induced tumorigenesis and can be used as biomarkers and therapeutic targets for the prognosis and treatment of HER2-positive and luminal subtypes of breast cancer in the future.

No potential conflicts of interest were disclosed.

Conception and design: Y. Zhao, T. Montminy, T. Azad, X. Yang

Development of methodology: Y. Zhao, T. Montminy, E. Lightbody, Y. Hao

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y. Zhao, T. Montminy, E. Lightbody, Y. Hao, C.J.B. Nicol

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y. Zhao, T. Montminy, T. Azad, E. Lightbody, Y. Hao, S. SenGupta, C.J.B. Nicol

Writing, review, and/or revision of the manuscript: Y. Zhao, T. Montminy, E. Lightbody, S. SenGupta, X. Yang

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): T. Azad, E. Asselin

Study supervision: X. Yang

This work was supported by grants from the Canadian Institute of Health Research (CIHR) and Canadian Breast Cancer Foundation (to X. Yang), Breast Cancer Action Kingston and Kingston General Hospital Foundation (to C.J. Nicol), and Canadian Breast Cancer Foundation, Ontario Chapter #369649 (to C.J. Nicol). Y. Zhao was supported by Ontario Trillium Scholarship (OTS, Canada), CIHR/Terry Fox Foundation Training Program in Transdisciplinary Cancer Research and Chinese Government Award for Outstanding Self-financed Students Abroad (2014). T. Montminy was funded by CIHR/Terry Fox Foundation Training Program in Transdisciplinary Cancer Research and the Huang Leadership Development Scholarship. T. Azad is supported by a Vanier Canada Graduate Scholarship. E. Lightbody is supported by Queen's University Terry Fox Research Institute Transdisciplinary Training Program in Cancer Research Fellowship, and the Dr. Robert John Wilson Graduate Fellowship.

The authors would like to thank Dr. Kunliang Guan for providing the HEK293A-LATS1/2-knockout (LATS-KO) and HEK293A-MST1/2-KO (MST-KO) cell lines; Dr. Fernando D. Camargo from Harvard University for providing the STBS-luc plasmid; Dr. Jian Chen for performing the MS detection and analysis; Drs. Paolo Armando Gagliardi and Luca Primo for the PDK1 constructs (PDK1-WT, PDK1-KD, PDK1-DPH); Dr. Leda Raptis for the PIK3CA-H1047R plasmid. We also thank Ellen van Rensburg for manuscript proof-reading and Lee Boudreau for performing IHC.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.