Hippo-TAZ signaling is the master regulator of the onset of triple-negative basal-like breast cancers

Significance

A universal oncogenic driver of basal-like breast cancer (BLBC) has resisted identification. We show that continuous transcriptional coactivator with PDZ-binding motif (TAZ) activation in precancerous murine luminal cells generates luminal cancers that later become BLBCs. Subsequent TP53 alteration, a feature of invasive human BLBCs, accelerates tumor progression. Because BLBC development is inhibited by TAZ inactivation in vivo, our work provides a sound rationale for targeting Hippo-TAZ signaling as therapy for human BLBC. Our mouse model of BLBC represents a powerful tool for evaluating such drugs.

Abstract

Breast cancer is the most frequent malignancy in women worldwide. Basal-like breast cancer (BLBC) is the most aggressive form of this disease, and patients have a poor prognosis. Here, we present data suggesting that the Hippo–transcriptional coactivator with PDZ-binding motif (TAZ) pathway is a key driver of BLBC onset and progression. Deletion of Mob1a/b in mouse mammary luminal epithelium induced rapid and highly reproducible mammary tumorigenesis that was dependent on TAZ but not yes-associated protein 1 (YAP1). In situ early-stage BLBC-like malignancies developed in mutant animals by 2 wk of age, and invasive BLBC appeared by 4 wk. In a human estrogen receptor⁺ luminal breast cancer cell line, TAZ hyperactivation skewed the features of these luminal cells to the basal phenotype, consistent with the aberrant TAZ activation frequently observed in human precancerous BLBC lesions. TP53 mutation is rare in human precancerous BLBC but frequent in invasive BLBC. Addition of Trp53 deficiency to our Mob1a/b-deficient mouse model enhanced tumor grade and accelerated cancer progression. Our work justifies targeting the Hippo-TAZ pathway as a therapy for human BLBC, and our mouse model represents a powerful tool for evaluating candidate agents.

Breast cancer (BC) is the most frequent cancer in women worldwide and remains the second leading cause of death (1). Human BC is a heterogeneous disease that has been classified into various subgroups based on the expression (or not) of the estrogen receptor (ESR), progesterone receptor (PGR), and/or human epidermal growth factor receptor 2 (HER2; ERBB2) (2). Triple-negative BC (TNBC; ESR⁻PGR⁻HER2⁻) is the most complex, aggressive, and fatal subtype of BC, and no targeted therapies currently exist to treat it. TNBC patients invariably develop resistance to existing chemotherapeutic regimens, leading to ∼90% treatment failure and a very low 5-y survival rate (3). Thus, there is a pressing need to better understand TNBC onset and progression mechanisms in order to develop more effective treatments.

In humans and mice, the mammary epithelial cell (MEC) population is composed of the luminal cell and basal cell subsets. Luminal cells line the lumens of the mammary ducts and alveoli and are responsible for the production and transport of milk. Basal cells contact the basement membrane and contract to pump milk through the ducts. The majority of BCs originate from the transformation of progenitor or mature luminal cells, and the resulting tumors can exhibit histological and/or molecular features of cells undergoing either luminal- or basal-like cell differentiation (4, 5). However, the molecular mechanisms by which luminal tumors acquire basal-like characteristics and generate basal-like BCs (BLBCs) have yet to be elucidated. Importantly, BLBCs constitute the majority (71 to 78%) of human TNBCs (6).

Human BLBCs display many genetic alterations, including TP53 nonsense and frameshift somatic mutations (84%); aberrant activation of PI3K-AKT pathway genes (PIK3CA [49%]; PTEN [19%]; INPP4B [30%]; AKT3 [31%]); and activation of RAS-RAF-MEK pathway genes (KRAS [32%]; BRAF [30%]; EGFR [23%]) or other receptor tyrosine kinases (RTKs). Alterations to the BRCA (36%), MYC (31%), and RB1 (20%) genes are also observed (6–8). In mice, overexpression of single genes (e.g., Erbb2, Igf-Ir, Wnt1, Met, Notch1-IC domain, Stat5, H-Ras, or Myc) as well as inactivation of Trp53, Pten, Rb, Stat1, Brg1, or Apc all trigger the development of some form of BC-like malignancies but only after a prolonged period (9). BLBCs do arise after several months in Brca1+Trp53 (10) and Rb+Trp53 (11) double heterozygous mutant mice, but, to date, the mouse model showing the quickest onset of BLBC (within 2 wk) features a dominant stable β-catenin mutation combined with exogenous Met overexpression (12). Thus, no mouse model currently available both exhibits an endogenous single pathway alteration and quickly and faithfully replicates the generation of human BLBC.

Our previous work has revealed the importance of the Hippo–transcriptional coactivator with PDZ-binding motif (TAZ)/yes-associated protein-1 (YAP1) pathway in the development of many types of cancers (13). The Hippo signaling pathway is activated in response to changes in the extracellular environment, including the presence of excessive estrogen (14). The downstream targets of Hippo signaling are the transcriptional cofactors TAZ and YAP1. In most cell types, TAZ and YAP1 positively regulate cell proliferation via their effects on Transcriptional enhanced associated domain (TEAD) transcription factors but are negatively controlled by upstream Hippo pathway elements. These elements include the STE20-like protein kinases, the large tumor suppressor homolog (LATS) kinases, and their adaptor proteins Salvador homolog-1 and Mps one binder kinase activator-1 (MOB1) (14).

Both TAZ and YAP1 are involved in promoting the proliferation, migration, self-renewal, epithelial-mesenchymal transition (EMT), metastasis, and drug resistance of both mouse and human BC cells in vitro (15, 16). However, transgenic (Tg) mice overexpressing either TAZ or YAP1 do not spontaneously develop tumors even in extreme old age (17, 18). TAZ Tg mice did exhibit BC formation after administration of the carcinogen, 7,12-dimethylbenz[a]anthracene (DMBA) (18), although whether these tumors were luminal or basal-like in character was not reported. In humans, while there is some evidence that TAZ is activated in at least the later stages of BC development, especially that of BLBCs, it remains controversial whether the same is true for YAP1 activation (19–22). Moreover, TAZ/YAP1 expression has yet to be examined in human BLBC at its premalignant or carcinoma in situ (CIS) stages.

To resolve these issues and elucidate whether Hippo-TAZ/YAP1 activation participates in the onset and progression of BCs, we generated MOB1-deficient mice in which endogenous TAZ and YAP1 were strongly activated in normal mammary luminal cells. We also analyzed TAZ/YAP1 activation in tissue samples from patients with premalignant stage BC. We report that Hippo-TAZ signaling is a key driver of BLBC in mice and that TAZ activation is a feature of early-stage human BLBCs, making the Hippo pathway a worthy therapeutic target.

Results

Loss of Mob1a/b Triggers Early Onset of BC in Mice.

To examine the roles of TAZ and YAP1 in mature mouse mammary gland, we first sought to generate adult mutant mice that could undergo postnatal deletion of MOB1A and MOB1B, leading to hyperactivation of endogenous TAZ/YAP1 activity. To this end, we crossed Mob1a^flox/flox;Mob1b^−/− mice (23) to Rosa26-CreERT2-Tg mice (24) to produce Rosa26-CreERT2;Mob1a^flox/flox;Mob1b^−/− progeny in which Mob1a/b deletion (double knockout; DKO) could be postnatally induced by treatment with tamoxifen (TAM). We then intraperitoneally (i.p.) administered TAM to Rosa26-CreERT2;Mob1a^flox/flox;Mob1b^−/− (roMob1 DKO) mice, as well as to control Rosa26-CreERT2;Mob1a^+/+;Mob1b^+/+ mice, when the animals were 8 to 12 wk old. A test of this strategy using Rosa26-CreERT2;Mob1a^+/+;Mob1b^+/+;Rosa26-LSL-tdTomato mice revealed that by 3 wk after TAM initiation (post-TAM), this system had deleted 35, 37, and 3% of luminal, basal, and stromal cells, respectively, from among lineage-negative (CD31⁻CD45⁻TER119⁻) cells isolated from whole mammary glands (SI Appendix, Fig. S1A). The Mob1a gene was deleted frequently (50%) in proximal MECs of roMob1 DKO mice at 3 wk post-TAM, but only infrequently (10%) in distal MECs (SI Appendix, Fig. S1B). Loss of MOB1 protein in EpCAM^med/high proximal MECs isolated from roMob1 DKO mice at 3 wk post-TAM was confirmed by immunoblotting (SI Appendix, Fig. S1C).

We then sacrificed roMob1 DKO and control mice at 2.5 and 5 wk post-TAM and histologically analyzed their mammary glands. To our surprise, all of the roMob1 DKO mice had developed tumors in proximal mammary ducts by 2.5 wk post-TAM; (Fig. 1A; yellow arrowhead), with 83% of these animals showing ductal CIS (DCIS) and 17% displaying low-grade invasive ductal carcinoma (IDC). Strikingly, all roMob1 DKO mice exhibited IDC by 5 wk post-TAM. No tumors were observed in TAM-treated control (Mob1a^flox/flox;Mob1b^−/−) mice. In addition, mutant MECs showed hyperplasticity, with many mitotic cells (Fig. 1B, green arrowheads) exhibiting impaired cell polarity and containing abnormally large nuclei (Fig. 1B, red arrowhead), consistent with malignant transformation. However, all roMob1 DKO mice died by 10 wk post-TAM (Fig. 1C). In previous work, we showed that roMob1 DKO mice develop tongue cancers (25) and suffer from reduced saliva secretion (26) in response to TAM-induced MOB1 depletion, effects that may have contributed to a fatal impairment of food intake. Our roMob1 DKO mice were, thus, unsuitable for further BC analyses.

Fig. 1.

Mob1a/b deletion in adult mouse mammary epithelium results in extremely rapid carcinogenesis. (A) (Left) Representative carmine-stained stereomicroscopic views (Top) and hematoxylin and eosin (H&E)-stained microscopic views of sections (Bottom) of control and roMob1 DKO mammary epithelial layers at 2.5 wk post-TAM administration. Yellow arrowheads, tumors in proximal large ducts. Scale bars, 2 mm (Top) and 20 μm (Bottom). (Right) Percentages of the indicated lesion types in mammary epithelium of roMob1 DKO mice at the indicated weeks post-TAM (n = 10/group). (B) H&E-stained sections of control (Top) and roMob1 DKO (Bottom) mammary epithelial layers at 2.5 wk post-TAM. Mutant epithelium shows features of CIS, including hyperplasia with many mitotic cells (green arrowheads) exhibiting impaired cell polarity and abnormally large nuclei (red arrowhead). White dotted line, basement membrane. Scale bars, 20 μm. (C) Kaplan-Meier curves depicting overall survival of TAM-treated roMob1 DKO and control mice (n = 16). **P < 0.01, Log-rank test. (D) (Left) Representative carmine-stained stereomicroscopic views of mammary glands of control and prMob1 DKO mutant mice at the indicated weeks of age. Yellow or blue arrowheads indicate tumors in proximal or distal mammary ducts, respectively. (Top Right) Macroscopic view of control mammary gland at 40 wk. Black arrowhead, nipple. (Bottom Right) Macroscopic view of tumor formation in prMob1 DKO mammary gland at 40 wk. Scale bars, 2 mm. (E) (Left) Representative H&E-stained mammary tissues from control and prMob1 DKO mice (n = 12 to 17 mice/group/timepoint) sacrificed at the indicated weeks after birth. Actively growing CIS was observed in prMob1 DKO mice from 2 wk of age onward. Scale bars, 20 μm. (Right) Percentages of prMob1 DKO mice showing the indicated grades of mammary tumors at the indicated weeks after birth. (F) Low magnification (Top) and high magnification (Bottom) views of representative H&E-stained mammary duct epithelial layers from control and prMob1 DKO mice at the indicated weeks of age. As they aged, mutant mice successively developed UDH, ADH, DCIS, low-grade IDC, and high-grade IDC. Scale bars, 20 μm. (G) Representative H&E-stained sections of lung, lymph node, and liver tissues that were isolated from 60-wk-old prMob1 DKO mice and exhibited metastasis of malignant MECs. Scale bars, 20 μm.

To observe the long-term effects of MOB1 loss in the context of BC development, we generated mammary luminal cell-specific Mob1a/b DKO mutants using Progesterone receptor-Cre knock-in (PGR-Cre) mice (27). We successively mated PGR-Cre Tg mice with Mob1a^flox/flox and Mob1b^−/− mice to generate PGR-Cre;Mob1a^flox/flox;Mob1b^−/− mice (prMob1 DKO mice). This system can delete a target gene starting at E16.5 (28), the time when PGR begins to be expressed in mouse mammary glands, uterus, vagina, and ovaries (27). A test of this strategy using 6-wk-old PGR-Cre;Mob1a^+/+;Mob1b^+/+;Rosa26-LSL-YFP mice revealed that, in whole mammary glands, this system had deleted the target gene in 62% of luminal cells but failed completely to delete the target gene in basal cells (SI Appendix, Fig. S2 A and B). The Mob1a gene was deleted in 90% of proximal mammary ducts, and in 50% of distal mammary ducts, in prMob1 DKO mice at 6 wk of age (SI Appendix, Fig. S2C). Mob1a^+/+;Mob1b^+/+, Mob1^flox/flox;Mob1b^−/−, and PGR-Cre;Mob1a^+/+;Mob1b^+/+ mice at 6 wk of age were all indistinguishable from wild-type (WT) mice in gross appearance, survival, and histology (SI Appendix, Fig. S2D), allowing the use of Mob1a^flox/flox;Mob1b^−/− mice as the control in subsequent experiments.

Although all prMob1 DKO mice appeared healthy at birth and most survived for more than 40 wk, 70% of these mutants died by 60 wk of age. In total, 62% of prMob1 DKO mice developed DCIS in the proximal mammary ducts by 2 wk of age (Fig. 1D–F), and all developed low-grade IDC by 6 wk of age (Fig. 1E). Even from the exterior, proximal DCs of >5 mm in size were obvious from 10 wk of age on (Fig. 1D, Right). By 40 wk of age, high-grade IDC had developed and metastasized to the lung in 50% of mutant mice, as well as to the lymph nodes (8.3%) and liver (8.3%) (Fig. 1E and G). These carcinomas appeared in both proximal (yellow arrowhead) and distal (blue arrowhead) ducts (Fig. 1D). There were no significant differences in ductal length, or in numbers of terminal endpoints or branch points of the ducts, between control and prMob1 DKO mice of 6 or 10 wk of age (SI Appendix, Fig. S3 A and B). The most prominent cancerous lesions were observed in the proximal large ducts of prMob1 DKO mice (Fig. 1D–F). Detailed pathological analysis confirmed the successive occurrence of usual ductal hyperplasia (UDH), atypical ductal hyperplasia (ADH), DCIS, low-grade IDC, and finally high-grade IDC in prMob1 DKO mice as they aged (Fig. 1F). Unfortunately, due to the female infertility of both prMob1 DKO and roMob1 DKO mice, we were unable to examine mammary gland alterations during pregnancy and lactation. Nevertheless, both TAM-treated roMob1 DKO mice and prMob1 DKO mice showed surprisingly rapid onset and development of BCs, and so provided suitable models for further mechanistic analyses.

Luminal-Cell-Specific MOB1-Deficient Mice Develop TN BLBCs.

To classify the BCs in our prMob1 DKO mice, we performed immunohistochemistry (IHC) studies of tumor sections that were isolated from prMob1 DKO mice at 10 wk of age and stained using antibodies against KRT8, ESR, KRT14, KRT5, or p63. KRT8 and ESR are luminal cell markers, whereas KRT14, KRT5, and p63 are basal cell markers. We found that all BCs in prMob1 DKO mice were ESR⁻, KRT8⁻, KRT14⁺, KRT5⁺, and p63⁺, classifying them as BLBC (Fig. 2A). We confirmed this finding using qPCR and flow cytometric analysis of EpCAM^med/high proximal MECs isolated from control and prMob1 DKO mice at 10 wk of age. qPCR revealed that messenger RNA (mRNA) levels of the basal cell markers Krt14, Krt5, and Tp63 were increased (Fig. 2B, Left), while those of the luminal cell markers Krt8, Esr1, Pgr, Foxa1, and Her2/Erbb2 were all decreased (Fig. 2B, Middle). Increased mRNA levels of Ctgf and Ankrd1, which are downstream targets of TAZ/YAP1, were also observed and constituted a positive control (Fig. 2B, Right). Similarly, flow cytometric analysis showed that the basal cell population (CD24^medCD49f^high) was increased in prMob1 DKO mice, whereas the luminal cell population (CD24^highCD49f^med) was decreased (Fig. 2C).

Fig. 2.

Tumors in luminal-cell-specific MOB1-deficient mice are TN basal-like in phenotype. (A) Representative immunostaining to detect KRT8, ESR, KRT14, KRT5, or p63, as indicated, in sections of (Top) normal mammary epithelium from a control mouse or (Bottom) BCs resected from prMob1 DKO mice at 10 wk of age. Scale bars, 20 μm. (B) (Left and Middle) Quantitation of the fold increase or decrease in mRNA expression of the basal (Krt14, Krt5, Tp63) and luminal (Krt8, Esr1, Pgr, Foxa1, Erbb2) markers as determined by qPCR in EpCAM^med/high proximal MECs isolated from control and prMob1 DKO mice at 10 wk of age. mRNA expression data were normalized to Gapdh. Data are the mean ± SEM (n = 4/group). (Right) mRNA levels of the Hippo-TAZ/YAP1 downstream target genes Ctgf and Ankrd1, which were analyzed as positive controls. *P < 0.05 and **P < 0.01, t test. (C) (Left) Representative flow cytometric analysis to detect CD24^highCD49f^med luminal cells and CD24^medCD49f^high basal cells among EpCAM^med/high proximal MECs isolated from control or prMob1 DKO mice at 10 wk of age. (Right) Quantification of the ratio of CD24^medCD49f^high basal cells to total YFP⁺ MECs for the cells in the left panels. Data are the mean ± SEM (n = 3 mice/group). **P < 0.01, t test. (D) (Left) Representative immunostaining to detect KRT8 (red) and KRT5 (green) in tumors isolated from prMob1 DKO mice at 2, 6, and 10 wk of age. 4′,6-diamidino-2-phenylindole (DAPI) (blue), nuclei. Scale bars, 20 μm. (Right) Percentages of KRT5⁺, KRT8⁺/KRT5⁺, and KRT8⁺ cells among mammary tumor cells isolated from prMob1 DKO mice at the indicated ages.

Because BLBC originates from luminal cells (4), we used IHC to monitor changes to luminal and basal cell markers in tumors developing in prMob1 DKO mice as they aged. At 2 wk of age, prMob1 DKO mice showed increased numbers of KRT8⁺ luminal tumor cells in the proximal ducts, with approximately normal numbers of KRT5⁺ basal cells (Fig. 2D, Left). However, the level of KRT8 expression by tumor cells then gradually declined, and the number of KRT5⁺ tumor cells increased. Importantly, by 6 wk of age, as many as 37% of tumor cells were KRT8⁺KRT5⁺ in phenotype (Fig. 2D, Right). By 10 wk of age, almost all KRT8⁺ tumor cells had disappeared, leaving a few remaining KRT8⁺KRT5⁺ cells among a vast pool of KRT5⁺ BLBC cells. Thus, in the absence of MOB1, luminal cells become transformed over time, gradually losing their luminal character and gaining basal-like features until they generate a TN (ESR⁻PGR⁻HER2⁻) BLBC.

Onset of BC in MOB1-Deficient Mice Depends on TAZ Activation.

IHC analysis demonstrated that, in control mice, TAZ was activated mainly in luminal cells of large ducts but only faintly expressed in luminal cells of small ducts, both at the prepuberty and puberty virgin stages (Fig. 3A and B). TAZ expression was even weaker in the basal cells of the large ducts in control mice (Fig. 3A and C). Pregnancy and lactation increased TAZ expression in the large ducts (Fig. 3A and C). On the other hand, MOB1 was expressed in both luminal and basal cells, in both large and small ducts, and during both the prepuberty and puberty virgin periods in control mice. Furthermore, MOB1 expression was markedly reduced in pregnant control mice, with only a modest increase during the lactation period (SI Appendix, Fig. S4A and B). These results are in line with MOB1’s expected negatively regulatory effect on TAZ activation in the mammary ducts of control mice, especially in the large ducts. TAZ was expressed more strongly in the luminal cells of large ducts and tumors of prMob1 DKO and roMob1 DKO mice compared to controls (Fig. 3A and D). TAZ nuclear staining intensity in the small ducts of prMob1 DKO and roMob1 DKO mice was weaker than that in the large ducts but still stronger than that in controls (Fig. 3A and E). Because both prMob1 DKO and roMob1 DKO female mice are infertile, we could not examine TAZ activation during pregnancy or lactation in these animals.

Fig. 3.

MOB1-mediated TAZ activation controls the onset of mammary tumors in mice. (A) Representative immunostaining to detect TAZ protein in control, prMob1 DKO, and roMob1 DKO mammary epithelium at the prepuberty (1 to 3 wk), puberty virgin (8 to 12 wk), pregnancy, and lactation stages. Large and small ducts were distinguished by their localization and size as described in SI Appendix, Supplementary Methods and were analyzed separately. Scale bars, 10 μm. (B) Quantitation of the intensity of TAZ nuclear staining in luminal and basal cells in the large and small ducts of control mice at the puberty virgin stage (n = 4). (C) Quantitation of TAZ nuclear staining intensity of luminal and basal cells in the large ducts of control mice at the indicated stages (prepuberty, n = 4; puberty virgin, n = 4; pregnancy, n = 4; lactation, n = 4). (D) Quantitation of TAZ nuclear staining intensity of luminal and basal cells in the large ducts of control (n = 4), prMob1 DKO (n = 4), and roMob1 DKO (n = 4) mice at the puberty virgin stage. (E) Quantitation of TAZ nuclear staining intensity of luminal cells in the large and small ducts of control (n = 4), prMob1 DKO (n = 4), and roMob1 DKO (n = 4) mice at the puberty virgin stage. (F) Representative immunostaining to detect YAP1 protein in (Left) control large duct, (Middle) roMob1 DKO mammary tumor, and (Right) roMob1 DKO epidermis at 5 wk post-TAM. Scale bars, 10 μm. (G) Quantitation of YAP1 nuclear staining intensity in luminal (white bar) and basal (black bar) cells in the control large duct, roMob1 DKO mammary tumor, and roMob1 DKO epidermis (hatched bars) samples in F. For B–E and G, at least 20 cells were analyzed for each sample, and staining intensity was measured by ImageJ in AUs. Data are the mean ± SEM **P < 0.01, Tukey’s multiple comparison test. (H) (Top) Representative carmine-stained stereomicroscopic views of whole mounts of mammary glands from mice of the indicated genotypes at 5 wk post-TAM. Scale bars, 2 mm. (Middle) H&E-stained sections of proximal large ducts in mammary epithelium of the mice in the top row. Scale bars, 20 μm. (Bottom) Immunostaining to detect TAZ and YAP1 in proximal large ducts in mammary epithelium of the mice in the top row. Scale bars, 10 μm. (I) Quantitation of the maximum diameter of tumors estimated from serial sections of proximal ducts in mice of the indicated genotypes at 5 wk post-TAM. Data are the mean ± SEM (n = 6 mice/group). **P < 0.01, n.s., not significant, t test.

With respect to YAP1, the immunostaining revealed faint activation of YAP1 in the basal cells of control large ducts, but the activation was negligible in roMob1 DKO mammary tumors at 5 wk post-TAM. However, YAP1 was strongly activated in the nuclei of epidermal skin cells of the same mutant mice (Fig. 3 F and G). These observations suggest that the function of TAZ, but not YAP1, was affected by mammary-tissue-specific loss of MOB1.

To investigate the roles of TAZ and YAP1 in the phenotypes of MOB1-deficient MECs, we crossed roMob1 DKO, Yap1^flox/flox, and Taz^flox/flox mice to generate triple KO (TKO) mutants either lacking TAZ plus MOB1 (Rosa26-CreERT2;Mob1a^flox/flox;Mob1b^−/−;Taz^flox/flox; designated roMob1TAZ TKO) or lacking YAP1 plus MOB1 (Rosa26-CreERT2;Mob1a^flox/flox;Mob1b^−/−;Yap1^flox/flox; designated roMob1YAP1 TKO). The onset of tumors in MOB1-deficient mammary epithelium was completely prevented by additional deficiency of TAZ, but not by deficiency of YAP1 (Fig. 3H and I). This result confirms our hypothesis that the onset of mammary tumorigenesis associated with MOB1 deficiency depends largely on TAZ, but not on YAP1, and aligns with our observation that BC arises in luminal cells of MOB1-deficient large ducts, where TAZ activation is the strongest.

Alterations to Hippo-TAZ Signaling Result in a Gain of Tumorigenic Properties and a Shift toward BLBC.

We next established mammary organoid cultures using luminal cells that were purified from roMob1 DKO or control mice and cultured with or without TAM for 14 d in vitro. Both the number and size of organoids derived from roMob1 DKO luminal cells were significantly increased by TAM treatment (Fig. 4A and B). Enhanced cell proliferation in MOB1-deficient organoids was confirmed by their significantly increased Ki67⁺ positivity (Fig. 4C and D). Notably, terminal deoxynucleotidyl transferase-mediated dUTP nick end labeling (TUNEL)⁺ apoptotic cells were very rare in both control and MOB1-deficient luminal cell cultures (SI Appendix, Fig. S5A and B). Thus, MOB1 deficiency confers increases in progenitor-like activities and proliferation on luminal cells.

Fig. 4.

Tumorigenic properties of mammary cells bearing MOB1-TAZ alterations and effect of TAZ overexpression on luminal BC cells in vitro. (A) Macroscopic views of representative organoids generated by culturing luminal cells that were purified from roMob1 DKO and control mice. Cells were treated with or without TAM for 2 d, followed by the incubation in Matrigel without TAM for 12 d in vitro. Scale bars, 100 μm. (B) Quantitation of the (Top) mean number and (Bottom) mean diameter of the organoids in A. Data shown are the mean ± SEM (n = 3/group). *P < 0.05, Tukey’s multiple comparison test. (C) Representative images of Ki67 staining of the organoids in A. Nuclei were counterstained with DAPI. Scale bars, 20 μm. (D) Quantitation of the Ki67-positive cells in C. At least 1,000 cells were counted for each sample. Data are the mean ± SEM (n = 3). **P < 0.01, t test. (E) Immunostaining to detect α-tubulin (red) and γ-tubulin (green) in MCF10A-iTAZ4SA cells that were cultured with or without Dox for 7 d. DAPI (blue), nuclei. (Left) Representative normal mitotic cell (Dox−). (Right) A mitotic cell overexpressing TAZ4SA (Dox+) showing a multipolar spindle. Scale bar, 10 μm. (F) Quantitation of cells with multipolar mitotic spindles in the Dox+ and Dox− MCF10A-iTAZ4SA cultures in E. At least 50 mitotic cells/sample were evaluated. Data are the mean ± SEM (n = 3). *P < 0.05, t test. (G) Immunostaining to detect α-tubulin (red) and nuclei (DAPI, blue) in MCF10A-iTAZ4SA cells that were cultured with or without Dox for 7 d. (Left) Representative normal cell showing a normal nucleus (Dox−). (Right) A cell overexpressing TAZ4SA (Dox+) showing a micronucleus (white arrowhead). Scale bar, 10 μm. (H) Quantitation of cells with micronuclei in the Dox− (n = 3) and Dox+ (n = 5) MCF10A-iTAZ4SA cultures in G. At least 130 cells/sample were evaluated. Data are the mean ± SEM. *P < 0.05, t test. (I) Quantitation of mRNA levels of (Top Left) the basal cell marker genes KRT14, KRT5, EGFR, and KRT17; (Bottom Left) the luminal cell marker genes KRT8, ESR, FOXA1, and GATA3; and (Right) TAZ and its target genes CTGF, CYR61, and ANKRD1 in MCF7-iTAZ4SA cells that were cultured for 7 d with or without Dox (1 μg/mL). Data are the mean ± SEM (n = 3). *P < 0.05 and **P < 0.01, t test. (J) Representative immunofluorescent staining of TAZ, basal cell marker (KRT5, EGFR), and luminal cell marker (KRT8, ESR) proteins in MCF7-iTAZ4SA cells that were cultured for 7 d with or without Dox (1 μg/mL). DAPI (blue), nuclei. Scale bars, 20 μm.

To support our mammary organoid observations, we examined the TAZ-dependent tumorigenic properties of the human MEC line MCF10A-iTAZ4SA, which inducibly expresses a constitutively active mutant form of TAZ (TAZ4SA) (29) in response to doxycycline (Dox) treatment (SI Appendix, Fig. S6A). As expected, forced TAZ activation in MCF10A cells increased their proliferation (SI Appendix, Fig. S6B). The incidence of multipolar mitotic spindle formation in these cells (Fig. 4E and F) and the emergence of micronuclei (Fig. 4G and H) were also enhanced. The number of apoptotic MCF10A cells was also modestly elevated by constitutive TAZ activation (SI Appendix, Fig. S6C). These results suggest that TAZ-induced increases in cell proliferation, progenitor-like activities, and chromosomal mis-segregation may drive tumor onset in MOB1-deficient mammary glands.

Because the MOB1-deficient luminal cancers in our mutant mice had acquired a basal-like phenotype upon TAZ activation in vivo, we examined whether TAZ activation in a human luminal cancer cell line (MCF7) would induce these cells to adopt a basal-like cancer phenotype in vitro. For this purpose, we engineered MCF7 cells to overexpress iTAZ4SA (MCF7-iTAZ4SA cells). We then cultured these cells with or without Dox for 7 d and analyzed the expression of luminal and basal markers by qPCR (Fig. 4I) and immunofluorescence staining (Fig. 4J). Overexpression of activated TAZ4SA complementary DNA resulted in strong expression of TAZ at both the mRNA and nuclear protein levels and also upregulated the mRNA expression of TAZ downstream transcriptional targets. As expected, the expression of constitutively active TAZ induced by Dox treatment also increased mRNA levels of KRT14, KRT5, EGFR, and KRT17 (basal markers) while decreasing those of KRT8, ESR, FOXA1, and GATA3 (luminal markers) (Fig. 4I). Consistent with these results, the basal proteins KRT5 and EGFR were increased in MCF7-iTAZ4SA cells, but the luminal proteins KRT8 and ESR were decreased (Fig. 4J). Thus, constitutive and aberrant activation of TAZ can convert human luminal BC cells into BLBC cells in vitro.

TAZ Activation in Human Premalignant BLBC.

To our knowledge, the status of TAZ activation has not been examined in human premalignant mammary tissues, including in DCIS. We therefore sought to examine TAZ immunostaining intensity in normal human breast tissue samples as well as in patient tissue samples exhibiting UDH, ADH, DCIS, or IDC lesions. First, we evaluated the specificity of an antibody against human TAZ by immunostaining of WT, YAP1-KO, and TAZ-KO variants of the human non-small-cell lung cancer cell line H1299-Luc. This anti-TAZ antibody did not stain TAZ-KO H1299-Luc cells but generated a clear positive signal for both parental H1299-Luc and YAP1-KO H1299-Luc cells (SI Appendix, Fig. S7A). Deficiency of YAP1 and TAZ proteins in YAP1-KO and TAZ-KO variants, respectively, were confirmed by immunoblotting (SI Appendix, Fig. S7B). Thus, we deemed our anti-human TAZ antibody to be suitable for the specific detection of TAZ in human BC specimens.

Patients with basal-like type DCIS or IDC were selected on the basis of the KRT5 positivity (>50%) of their tumor cells (SI Appendix, Fig. S8A, Left). Immunostaining using anti-TAZ antibody showed that TAZ activation was upregulated from normal levels starting at the premalignant UDH stage, with a gradual increase in TAZ protein detected in both the nucleus and cytoplasm of human MECs as malignancy progressed (Fig. 5A and B). In contrast, immunostaining to detect YAP1 protein showed a limited and very gradual increase only in the cytoplasm of KRT5⁺ MECs, with no significant increase in the nucleus even at the DCIS and IDC stages of BLBC (SI Appendix, Fig. S8B). When we examined ESR⁺ human luminal BCs in the same way (SI Appendix, Fig. S8A, Right), neither nuclear TAZ (Fig. 5C and D) nor nuclear YAP1 (SI Appendix, Fig. S8C and D) was increased in DCIS or even IDC samples. To test whether TAZ expression changed during the course of tumor progression in a single patient, we examined five clinical BLBC samples that each contained cancerous lesions of different pathological stages as well as a noncancerous normal tissue area in a single slice of the specimen. Analysis of TAZ nuclear staining intensities in these samples showed that TAZ activation gradually increased with histological progression to BLBC, even within a single specimen (SI Appendix, Fig. S9A and B). Thus, the nuclear localization and activation of TAZ in premalignant mammary gland lesions is an event that leads to BLBC.

Fig. 5.

TAZ activation initiates at the precancerous stage of human BLBC. (A) (Top) Representative immunostaining to detect TAZ (with hematoxylin counterstaining) in normal proximal mammary duct (n = 14) or in lesions from KRT5⁺ BLBC patients displaying UDH (n = 17), ADH (n = 17), DCIS (n = 17), or IDC (n = 15). The scoring system to indicate degrees of nuclear and cytoplasmic TAZ staining is also shown. (Bottom) H&E-stained sections of DCIS and IDC patient samples. Scale bars, 20 μm. (B) Quantitation of the mean nuclear (Left) and cytoplasmic (Right) TAZ staining scores for normal proximal mammary duct or the indicated lesions from the KRT5⁺ BLBC patients in A. Data are the mean ± SEM (n = 15 or 17/group). *P < 0.05 and **P < 0.01, Tukey’s multiple comparison test. (C) Representative immunostaining to detect TAZ (with hematoxylin counterstaining) in sections of normal human mammary duct as well as DCIS and IDC lesions from patients with ESR⁺ luminal BCs. Scale bars, 20 μm. (D) Quantitation of mean nuclear TAZ score for normal human proximal mammary ducts (n = 10), and for DCIS (n = 10) and IDC (n = 22) lesions, from patients with KRT5⁺ BLBC or ESR⁺ luminal BC, as indicated. Data are the mean ± SEM **P < 0.01, Tukey’s multiple comparison test.

TP53 Inactivation Following Hippo-TAZ Pathway Alteration Accelerates Tumor Progression.

Because the most frequently mutated gene in human invasive BLBCs is TP53, we next used IHC to examine TP53 mutation status in the DCIS stage of human KRT5⁺ BLBC (Fig. 6A). Previous IHC work has linked strong nuclear accumulation of TP53 protein with TP53 missense mutations, and complete loss of TP53 protein with null TP53 mutations (30). We found TP53 mutations in 80% of IDC BLBC patients, consistent with a previous report (31), but in only 5.9% of DCIS BLBC patients (Fig. 6B, Left). Intriguingly, TP53 mutation was observed in 0 and 22.7% of DCIS and IDC ESR⁺ luminal BCs, respectively (Fig. 6B, Right). Thus, although TP53 mutation is frequent in BLBC IDC samples, it may not be the primary event driving the onset of BLBC.

Fig. 6.

Frequency of TP53 mutation in DCIS BLBC patients and effect of combined alterations of the Hippo-TAZ pathway plus TP53 in mice. (A) Representative immunostaining to detect TP53 (with hematoxylin counterstaining) in sections of human DCIS and IDC from (Left) KRT5⁺ BLBC and (Right) ESR⁺ luminal BC patients. Scale bars, 20 μm. (B) Quantitation of TP53 alteration frequency in human DCIS (n = 17) and IDC (n = 15) lesions from the BLBC patients in A and in DCIS (n = 10) and IDC (n = 22) lesions from the ESR⁺ luminal BC patients in A. The percentages of tumors showing nuclear accumulation of TP53 (black bars) and TP53-null tumors (white bars) are indicated. Data are the mean ± SEM. **P < 0.01, Fisher’s exact test. (C) Representative macroscopic views (Top), carmine-stained stereomicroscopic views (Middle), and H&E-stained microscopic views (Bottom) of mammary ducts in 10-wk-old prMob1 DKO and prMob1Tp53 TKO mice. Yellow arrowheads, tumors in proximal ducts. Red arrowheads, tumors in distal ducts. Scale bars are 2 mm, 2 mm, and 20 μm in macroscopic, stereoscopic, and microscopic views, respectively. (D) Quantitation of tumor diameter, percentage of distal involvement, and percentage of high- and low-grade IDC in the prMob1 DKO and prMob1Tp53 TKO mice in C. Data are the mean ± SEM (n = 10 to 11). *P < 0.05, **P < 0.01, t test (Left) or Fisher’s exact test (Middle and Right).

To determine the additional effects of TP53 inactivation on neoplasms featuring Hippo-TAZ alterations, we generated PGR-Cre;Mob1a^flox/flox;Mob1b^−/−;Trp53^flox/flox (prMob1Tp53 TKO) mice and compared mammary tumor progression in these animals with that in prMob1 DKO mice. Upon sacrifice at 10 wk of age, prMob1Tp53 TKO mice showed tumors that were larger in size than those in prMob1 DKO mice (Fig. 6C, Top and D, Left). In prMob1 DKO mice, tumors were observed only in the proximal mammary ducts, but tumors in prMob1Tp53 TKO mice infiltrated along these ducts and so also appeared in the distal mammary ducts (Fig. 6C, Middle, red arrowhead; Fig. 6D, Middle). Histologically, all tumors in prMob1 DKO mice were classified as low-grade IDC, but 54.5% of tumors in prMob1Tp53 TKO mice were considered to be high-grade IDC (Fig. 6C, Lower and 6D, Right). Thus, TP53 inactivation subsequent to Hippo-TAZ activation accelerates mammary tumor progression.

Discussion

This study has generated an animal model in which the mutation of a signal element in the Hippo pathway leads to very rapid and spontaneous BC onset. Indeed, our mutant mice show the quickest BC onset and progression induced by any single pathway alteration. Stunningly, all tumors observed in these mutant mice are BLBCs. Because the onset of these malignancies was so fast, we conclude that the Hippo-TAZ pathway is also highly likely to be a key driver of human BLBC. As was true in our mutant mice, it may be that significant activation of Hippo-TAZ signaling can initiate human BLBC in the absence of any other genetic or epigenetic alteration.

This pinpointing of altered Hippo-TAZ signaling as a key factor in human BLBC aligns with known risk factors. Obesity, high estrogen levels, and exposure to tobacco smoke have all been documented to increase BC risk; and lipid accumulation, estrogen treatment, and cigarette smoke extract have all been shown to activate TAZ/YAP1 (32–34). In addition, most of the genetic changes observed in human BLBCs—such as TP53 mutation or the aberrant activation of the WNT1, PI3K, HIF1α/ARNT, EGFR, or other RTK pathways—can activate TAZ/YAP1 (25, 35–38). These data support our conclusion that events resulting in accumulated TAZ activation might lead to the onset and progression of human BLBC.

Although our MOB1-deficient mice showed rapid BC formation, neither TAZ Tg mice (18) nor YAP1 Tg mice (17) spontaneously develop tumors at any time during their lives. A potential explanation for this discrepancy is that these TAZ Tg and YAP1 Tg animals were generated using the mouse mammary tumor virus (MMTV)-promotor system, which drives gene expression not only in MECs, but also in other tissues such as the salivary glands. It is possible that a Tg mouse strain showing strong TAZ/YAP1 activation in multiple tissues might therefore die prematurely, and that a strain with weak TAZ/YAP1 activation might survive this alteration tumor free. In fact, we have found that most of the MOB1-deficient mice we have generated using the MMTV-Cre Tg approach show early lethality due, at least in part, to decreased saliva secretion (SI Appendix, Fig. S10). Another possible reason for the lack of tumor formation in TAZ Tg and YAP1 Tg mice is the effect of TAZ/YAP1-independent LATS kinase activity. LATS can bind to and control several molecules besides TAZ/YAP1, including the TP53 E3 ligase MDM2 (39), the ESR (40), the cytokinesis regulator Aurora B (41), and the cell-cycle regulator CDC26 (42). Alteration of these TAZ/YAP1-independent downstream effectors of LATS kinases may be required for BC onset. In fact, significant downregulation of LATS mRNA expression in human breast tumors compared to matched normal tissue has been reported (43). Nevertheless, for mechanistic reasons that remain under investigation, TAZ must be important for BC development because the formation of these malignancies in our MOB1-deficient mice was substantially prevented by additional loss of TAZ.

Previous reports have demonstrated that TAZ overexpression is more often observed in patients with invasive BLBCs (70.8% positive) than in those with invasive ESR⁺/PGR⁺ (12 to 30% positive) or HER2⁺ (16 to 19.4% positive) luminal cancers (21, 22), indicating that TAZ activation is especially important in invasive BLBCs. We replicated this observation in mice, showing that TAZ is activated starting from the precancerous stage of BLBC but not ESR⁺ luminal cancers. Unlike TAZ’s involvement, however, the involvement of YAP1 activation in human BCs has been controversial. Some groups have reported the presence of greater levels of activated YAP1 protein in tumor cells from patients with invasive TNBC (59.9%) compared to invasive ESR⁺/PGR⁺ (29.7%) or HER2⁺ (10.8%) cancers (44, 45). This elevation of activated YAP1 also correlated with aggressive tumor behavior and reduced metastasis-free survival. On the other hand, several teams have shown that, compared to normal human breast tissue, YAP1 was not activated in invasive ESR⁺/PGR⁺, HER2⁺, or TNBC malignancies (20). In addition, loss of heterozygosity for the chromosomal region containing the YAP1 gene has been observed in 33 to 43% of invasive BC cases (46, 47), and the YAP1 promotor is frequently methylated in BC patients (48). We showed here that nuclear YAP1 activation was not increased at either the precancerous or invasive stage of BCs (including BLBCs), and this BLBC formation was not prevented by additional loss of YAP1 in our mutant mice. Thus, we conclude that TAZ, rather than YAP1, is the more important oncogenic effector in BLBC carcinogenesis in mice and humans. We speculate that the studies reporting YAP1 activation in BLBCs may have resulted from a lack of tight YAP1 specificity of the antibodies used in the relevant IHC experiments. Because TAZ and YAP1 are paralogous molecules, these antibodies may have, in fact, recognized activated TAZ.

A previous report demonstrated that overexpression of TAZ in freshly prepared human MECs or normal human MCF10A cells increased the appearance of the basal-like phenotype and that inhibition of TAZ in MCF10A cells led to their luminal differentiation in vitro (49). This group also showed that TAZ-deficient mice exhibited reduced numbers of normal basal cells with an accompanying decrease in mammary gland cellularity. However, when they knocked down YAP1 in MCF10A cells, they did not see a similar alteration to their lineage marker profile (49). Therefore, although it seems that TAZ (but not YAP1) can alter the fate of normal breast epithelial cells, it remains to be clarified whether TAZ activation in luminal BC cells skews their differentiation toward the BLBC type. We have shown in our study that MOB1 loss in luminal cells initiates the formation of luminal BCs that gradually change the profile of their surface markers to that of BLBCs. We also demonstrated that an ESR⁺ luminal cancer cell line induced to overexpress TAZ lost its luminal BC phenotype and acquired a BLBC phenotype. Thus, continuous activation of TAZ may skew early-onset luminal BC cells toward the BLBC fate.

ΔNp63 is the major isotype of Tp63 in mammary glands and is highly expressed specifically in normal immature stem/progenitor cells and mature myoepithelial/basal cells (50). ΔNp63 knockdown in primary human breast epithelial cells induces a decrease in basal cell markers and an increase in luminal markers (51), implying that enhanced expression of ΔNp63 in mammary tissues might drive an increase in basal markers and a decrease in luminal markers, just as we observed in our MOB1-deficient mice. It may be that TAZ hyperactivation has downstream effects that lead to excessive ΔNp63 activity, which could partly explain the basal-like phenotype of BCs in our MOB1-deficient mice.

TP53 mutations are more frequent in human invasive BLBCs (88% positive) than in either luminal BCs (26% positive) or HER2-amplified BCs (50% positive) (52). However, we showed here that TP53 mutation is rare in human BLBCs at the DCIS stage, implying that TP53 inactivation is not an initiating event in BC carcinogenesis. Using our mutant mice, we demonstrated that TP53 inactivation layered on top of Hippo-TAZ activation accelerates tumor progression. These observations suggest that TP53 mutation is a late event that causes BLBCs to acquire more aggressive features.

In conclusion, our paper has provided genetic evidence that the Hippo-TAZ pathway is a key driver of the onset and progression of BLBCs because this aberrant signaling triggers incipient luminal cell tumors to acquire a BLBC phenotype. Our results further suggest that subsequent TP53 mutation (possibly induced by the genetic instability caused by altered Hippo-TAZ signaling) confers increased aggressiveness on BLBCs. The Hippo-TAZ pathway thus becomes an appealing molecular target for novel therapies of BLBC. We are now screening various drugs capable of targeting the Hippo-TAZ pathway and employing our mutant mice as a useful tool for their evaluation. Our goal is to identify one or more such drug candidates whose administration could bring concrete benefits to BLBC patients.

Materials and Methods

Mice.

Mouse strains used in this study were PGR-Cre (27); Mob1a^flox/flox;Mob1b^−/− (23); Rosa26-LSL-YFP reporter (The Jackson Laboratory); Rosa26-LSL-tdTomato reporter (The Jackson Laboratory); Rosa26-CreERT2 (The Jackson Laboratory); Trp53^flox/flox (The Jackson Laboratory) and Taz^flox/flox (kindly provided by Dr. J. Wrana, Lunenfeld-Tanenbaum Research Institute). Yap1^flox/flox mice were generated using Yap1^flox/flox ES cells from the Knockout Mouse Project Repository. PGR-Cre and Rosa26-CreERT2 mice were of the C57BL/6 background, and Mob1a^flox/flox;Mob1b^−/− mice were back-crossed to C57BL/6 for more than six generations. Postnatal Mob1a/b double homozygous mutant mice (Rosa26-CreERT2;Mob1a^flox/flox;Mob1b^−/− mice; designated roMob1 DKO) were generated by administering a single i.p. dose of 1 mg TAM (Toronto Research Chemicals, Toronto, Canada) at 8 to 12 wk of age. Primers used for genotyping PCR are listed in SI Appendix, Table S1A. All mice were kept in specific pathogen-free facilities at Kobe University.

Mammary Organoid Culture.

Organoid culture was performed essentially as previously described (53) using mammary luminal cells that were freshly sorted from Rosa26-CreERT2;Mob1a^flox/flox;Mob1b^−/− or control mice. These cells were pretreated with 0.5 μM TAM for 2 d, mixed with 20 μL Matrigel (Corning), embedded in prewarmed (37 °C) 48-well plates, and incubated for 15 to 30 min to allow solidification. Plates were then overlaid with DMEM/F12 advanced medium supplemented with Hepes (1:100, GIBCO), L-glutamine (1:100, GIBCO), B27 (1:50, Thermo Fisher), 100 ng/mL A83-01 (Tocris Bioscience), 50 ng/mL EGF (PeproTech) and 10 μM Y-27632 (Sigma-Aldrich) and cultured without TAM for up to 12 d to generate organoids.

Clinical Samples.

Randomly selected, surgically obtained human breast DCIS (ESR/PGR-positive [n = 40] or negative [n = 60]) or IDC (ESR/PGR-positive [n = 20] or negative [n = 30]) samples that also contained noncancerous tissue were acquired from Hyogo Cancer Center, Kobe City Medical Center General Hospital, or Kyushu University Beppu Hospital. ESR/PGR-positive samples were considered to be luminal BCs. Resected tissues were fixed in formalin and stained with anti-YAP1 (Sigma-Aldrich), anti-TAZ (Sigma-Aldrich), anti-KRT5 (DAKO), anti-ESR (DAKO), or anti-p53 (DO-7, Santa Cruz) antibodies (SI Appendix, Table S2) using a standard protocol. UDH and ADH were histologically defined in regions adjacent to DCIS in the same tissue sections. Among ESR/PGR-negative samples (DCIS or IDC), KRT5-positive (50% or greater positive staining) samples were considered to be BLBCs. Scoring of YAP1 and TAZ activation was conducted either as described in Fig. 5A or by ImageJ analysis of 3,3′-diaminobenzidine (DAB) staining measured in arbitrary units (AUs). Cancer cells showing 10% or greater strong positive nuclear p53 staining, or a complete absence of nuclear and cytoplasmic p53 staining, were considered to bear TP53 alterations.

Study Approval.

Animal experiments were approved by the Kobe University Animal Experiment Committee (P60602-R8, P170603), and animal care was conducted in accordance with institutional guidelines. The full clinical study protocol was approved by the ethics committees of Kobe University Hospital (No. B210093); Hyogo Cancer Center (No. R-922); Kobe City Medical Center General Hospital (No. 21102); and Kyushu University (No. 21031-00). Written informed consent was obtained from all patients whose cancers were analyzed in this study.

Data Availability

All study data are included in the article and/or SI Appendix.