High throughput kinase inhibitor screens reveal TRB3 and MAPK-ERK/TGFβ pathways as fundamental Notch regulators in breast cancer

Abstract

Expression of the Notch ligand Jagged 1 (JAG1) and Notch activation promote poor-prognosis in breast cancer. We used high throughput screens to identify elements responsible for Notch activation in this context. Chemical kinase inhibitor and kinase-specific small interfering RNA libraries were screened in a breast cancer cell line engineered to report Notch. Pathway analyses revealed MAPK-ERK signaling to be the predominant JAG1/Notch regulator and this was supported by gene set enrichment analyses in 51 breast cancer cell lines. In accordance with the chemical screen, kinome small interfering RNA high throughput screens identified Tribbles homolog 3 (TRB3), a known regulator of MAPK-ERK, among the most significant hits. We demonstrate that TRB3 is a master regulator of Notch through the MAPK-ERK and TGFβ pathways. Complementary in vitro and in vivo studies underscore the importance of TRB3 for tumor growth. These data demonstrate a dominant role for TRB3 and MAPK-ERK/TGFβ pathways as Notch regulators in breast cancer, establishing TRB3 as a potential therapeutic target.

Breast cancer ranks second among causes of cancer death in North American women and elevated expression of the Notch ligand Jagged 1 (JAG1) is a marker of relapse and poor outcome in this malignancy (1). Consistent with this, activated Notch signaling (2–4) and up-regulation of growth-promoting Notch target genes (5–9), are observed in breast cancer. Notch is a highly conserved intercellular signaling system present in multicellular organisms (10). Mammals have four Notch proteins (NOTCH1–4) that function as receptors for five Notch ligands [Delta-like (Dll)1, -3, -4, and JAG1 and -2]. Notch ligand–receptor interaction on neighboring cells leads to presenilin-protease (γ-secretase) complex-mediated cleavage of the receptor, resulting in release of the cytoplasmic domain fragment, intracellular Notch (N^IC). N^IC enters the nucleus and modulates the expression of target genes, predominantly by converting the recombination signal sequence-binding protein J kappa (RBPJκ) transcription factor from a repressor to an activator of transcription (11).

Knowledge of the events that promote Notch activation in breast cancer is limited. Hypoxia, through up-regulation of p66^shc or IL-6 can induce NOTCH3 and ERK-dependent JAG1 expression in breast cancer cells (12, 13). JAG1 expression is regulated by TGFβ in a SMAD-dependent fashion in a mouse breast cancer bone metastasis model (14). Numb loss (2), Ras-induced γ-secretase stabilization (15) and Pin1-potentiated NOTCH1 cleavage by γ-secretase (8) promote Notch activation and have been implicated in the progression of breast cancer.

Protein kinases and pseudokinases participate extensively in cellular signaling pathways and are key regulators of cell function. Here we describe the use of large-scale chemical kinase inhibitor- and kinase-specific small interfering (si)RNA-based screens to gain insight into the mechanisms that promote Notch activation in breast cancer. We show that the pseudokinase Tribbles homolog 3 (TRB3), functions as a master regulator of JAG1-induced Notch activation and tumor growth through its control of MAPK and TGFβ signaling pathways. These findings are consistent with a previously reported association between TRB3 expression and poor overall survival in breast cancer (16) and establishes TRB3 as a potential therapeutic target in this malignancy.

Results

MAPK Regulates Notch in Breast Cancer.

To perform a functional genetic screen for regulators of Notch signaling in breast cancer, a dual-reporter cell line was created by transducing human mammary carcinoma cells (MDA MB231) that expressed Renilla luciferase under the control of the thymidine kinase (TK) promoter, with an adenovirus expressing a Notch-regulatable Hes1 promoter/firefly luciferase cassette (Fig. 1A). MDA MB231 cells were selected as a suitable cell background because they express high levels of key Notch signaling pathway constituents, and they are Notch signaling competent (5). We subjected the Notch dual-reporter line to both γ-secretase inhibitor (GSI) and siRNA-mediated knockdown of JAG1 or NOTCH1 to confirm its reliability and power to identify compounds and siRNAs that inhibit Notch (Fig. S1A).

Fig. 1.

MEK1/2 inhibitors down-regulate JAG1 expression in breast cancer cell lines. (A) Dual-reporter MDA MB231 cell line was created by transducing cells that expressed Renilla (R) luciferase under the control of the thymidine kinase (TK) promoter (MB 231 TK-R) with adenovirus containing a Notch-regulatable Hes1 promoter/firefly (FF) luciferase cassette. In a robotic high throughput screen format, cells were cultured on microtiter plates, treated with chemical kinase inhibitor or kinome siRNA libraries, and FF and R luciferase quantified. (B) Kinase inhibitor screen scatter plot: FF/R ratios are plotted on the y axis against 240 corresponding kinase inhibitors (OICR-L100) on the x axis. The dashed lines represent two standard deviations (SD) away from the mean (solid line) and the circled dots represent the MEK1/2 inhibitors in OICR-L100. (C) Western blot analysis of p-ERK1/2, ERK1/2, JAG1, N^IC, and uPA in MDA MB231 cells treated with U0126 (1 µM or 10 µM) for 24 h. (D) Expression of JAG1 in MDA MB231 or HCC1143 cells treated with U0126 (1 µM or 10 µM) for 24 h. (E) Western blot analysis of JAG1, p-ERK1/2, and ERK1/2 protein in MDA MB231 and HCC1143 cells grown in serum-free conditioned media for 24 h prior to stimulation with EGF for 4 h, followed by treatment with U0126 (10 µM) for the indicated time periods. β-Actin expression is included as a loading control. Molecular weight (M_W) markers are shown in kilodaltons.

Kinases such as c-Src, c-Abl, mitogen-activated protein kinases (MAPK), phosphotidylinositol-3-kinase (PI3K), and the epidermal growth factor (EGF) receptor are commonly activated in cancer cells and are known to contribute to tumorigenesis. To comprehensively identify cross-talk between Notch and kinase-regulated signaling pathways, a kinase inhibitor library that contains 240 kinase inhibitors was screened using our dual-reporter system to identify pathways that disproportionally influenced Hes1-firefly luciferase expression compared with TK-Renilla luciferase. Of these compounds, 26 reduced the normalized firefly/Renilla luciferase ratio greater than two standard deviations from the mean (Fig. 1B) without significantly affecting the TK-Renilla luciferase level (Fig. S1B). Control wells in the screen confirmed low variability of the firefly/Renilla luciferase ratio (Fig. S1C). The findings of the screen were validated separately (Fig. S2A) and a toxicity assay identified only one of 26 compounds as toxic under the experimental conditions (Fig. S2B). When these 26 compounds were analyzed according to major signaling pathways targeted, MAPK-ERK was found to be the predominant pathway promoting Notch activation in MDA MB231 cells. Strikingly, 6 of 26 compounds targeted MEK1/2 and this comprised 66% of all MEK1/2 inhibitors included in the library; this was statistically significant (P < 0.001). Upon repeat screen of the entire library, there was an 87% overlap in the significant hits identified and this included all 6 of the aforementioned MEK1/2 inhibitors.

To validate MAPK-ERK signaling as a regulator of Notch activation, the dual-reporter line was treated with the six MEK1/2 inhibitors identified in the screen, and Notch reporter activity was found to be reduced in a dose-dependent manner in all cases (Fig. S2C). Similarly, treatment of wild-type MDA MB231 cells with U0126 reduced phospho (p)-ERK1/2, JAG1, Notch activation as measured by N^IC levels, and expression of the Notch target, urokinase plasminogen activator (uPA) (9) (Fig. 1C).

MDA MB231 cells contain an activating mutation in the BRAF proto-oncogene so they may be uniquely dependent upon the RAF/MAPK-ERK pathway for Notch activation. Indeed, HCC1143 breast cancer cells, which do not display constitutive RAF/MAPK-ERK activation, demonstrated a marginal decrease in JAG1 expression after 24 h of U0126 treatment (Fig. 1D). However, HCC1143 cells could be rendered permissive for RAF/MAPK-ERK activation through serum starvation and EGF stimulation (Fig. 1E). Under these conditions, HCC1143 cells demonstrated a transient increase in p-ERK1/2 and exquisite MAPK-ERK–dependent JAG1 expression; this was evidenced by the ability of U0126 treatment to abrogate JAG1 expression by 24 h of treatment. These data suggest that the MAPK-ERK pathway regulates JAG1 expression in breast cancer cells.

Correlation of JAG1 and MAPK Pathway Activation in Breast Cancer Cell Lines.

To further explore the link between MAPK-ERK and JAG1/Notch we looked for an association between a gene expression signature indicative of MAP kinase activation (17, 18) and JAG1 expression in 51 breast cancer cell lines (19). We found that MAPK-ERK was activated in cell lines of the basal-like subtype, also associated significantly with JAG1 overexpression (5, 9) (Fig. 2A). To a lesser degree, the MAPK-ERK activation signature was present in human epidermal growth factor receptor 2 positive (HER2+) cell lines (Fig. S3). To determine if MAPK-ERK activation genes showed a statistically significant, concordant difference as a function of continuous JAG1 gene expression, gene set enrichment analysis (GSEA) was used (Fig. 2B). JAG1 and MAPK-ERK activation were compared in breast cancer cell lines, and in the NCI-60 cell lines originating from cancer tissues of multiple types (20). Interestingly, a statistically significant association between MAPK-ERK activation and JAG1 expression was specific to breast cancer cell lines [enrichment score (ES) = 0.60 and 0.39 and P = 0.007 and 0.035 for up-regulated and down-regulated MAPK-ERK genes, respectively], but not to diverse human cancer cell lines (NCI-60, ES = 0.25 and 0.22 and P = 0.63 and 0.57 respectively).

Fig. 2.

MAPK pathway activation in JAG1-expressing and basal-like tumor cells. (A) An ordered set of genes significantly up-regulated (yellow) or down-regulated (blue) following MAPK pathway activation in MCF7 cells, as previously defined (Materials and Methods), is comparatively visualized in a set of 51 breast tumor cell lines. Cell lines of the basal-like molecular subtype show the strongest MAPK signature pattern and JAG1 expression (red box). (B) Gene set enrichment analysis (GSEA) was used to determine if there are statistically significant, concordant differences in MAPK signature gene expression as a function of continuous JAG1 expression in 51 breast cancer cell lines (Genentech) versus a series of cancer cell lines of varied type (NCI-60). Shown are GSEA plots for a representative JAG1 probeset. ER+, estrogen receptor positive; ER-, estrogen receptor negative; HER2+, human epidermal growth factor receptor 2 positive; ES, enrichment score; NES, normalized enrichment score; p, nominal P value.

TRB3 Is a Master Regulator of MAPK and TGFβ Pathway Activation and JAG1 Expression.

As a complementary approach to identify kinases that influence Notch activation, a kinome siRNA library was screened using the dual-reporter system. The kinome siRNA library contains siRNA pools that target 720 genes encoding protein kinases. Two biological replicate screens were performed and the results of each screen were analyzed using the B-score method of high throughput screen normalization to correct for systematic and positional variability across the samples (21). The results obtained in the two screens were reproducible and identified five siRNA pools that demonstrated a reduction in the B-score of at least two SDs from the mean (correlation coefficient = 0.61; Fig. 3A). These hits were independently validated in the dual-reporter cell line (Fig. S4A), and it was confirmed that they regulated JAG1, NOTCH1, and Hes1 expression (Fig. S4B).

Fig. 3.

TRB3 is a regulator of JAG1. (A) Kinome siRNA library screen scatter plot: B-scores for two biological replicate screens are plotted on the x- and y axes. Dashed cross hairs separate hits that are 2 SD away from the mean B-scores. (B) Western blot analysis of JAG1 and N^IC in the indicated cell lines 72 h after transfection with Scr or TRB3 siRNA. (C) Western blot analysis of p-ERK1/2, ERK1/2, and JAG1 in MDA MB231 cells treated with Scr or TRB3 siRNA. (D) Western blot analysis of TRB3, p-ERK1/2, ERK1/2, and JAG1 in HepG2 cells transduced either with adenoviral control or with adenovirus expressing TRB3 [multiplicity of infection (MOI) = 5], in the absence or presence of U0126. (E) Western blot analysis of JAG1, p-ERK1/2, and ERK1/2 in MDA MB231 cells treated with or without TGFβ and either Scr or TRB3 siRNA. (F) Western blot analysis of SMAD4 and JAG1 in MDA MB231 cells treated with Scr, TRB3, or SMAD4 siRNA. β-Actin expression is included as a loading control. M_W markers are shown in kilodaltons.

Consistent with the results of the chemical kinase inhibitor screen, TRB3, a known regulator of MAPK-ERK (22), was among the top five hits. TRB3 is a member of a family of three pseudokinases (TRB1–3) and these are the mammalian orthologs of Tribbles, a protein that inhibits mitosis in early Drosophila development. Despite its lack of kinase activity, TRB3 participates as a scaffold molecule during protein complex formation (23).

The matched response of both TRB3 and JAG1 expression to all four siRNA species that comprised the TRB3 siRNA pool, reduced the likelihood that JAG1 knockdown was due to an off-target effect (Fig. S4C). To ensure that our findings were not unique to the MDA MB231 cell line, we explored additional basal-like subtype breast cancer cell lines and found that JAG1 expression and Notch activation were consistently dependent on TRB3 (Fig. 3B).

To explore the regulation of JAG1 expression and MAPK-ERK signaling by TRB3 in breast cancer, p-ERK1/2 levels were determined by Western blotting after treatment of MDA MB231 cells with TRB3 siRNA (Fig. 3C). TRB3 knockdown resulted in markedly reduced p-ERK1/2 and a corresponding reduction in JAG1 protein. To reconcile the findings of the chemical inhibitor screen we explored the ability of individual components within the MAPK-ERK pathway to regulate JAG1 expression. Although BRAF, ERK1, or ERK2 knockdown resulted in decreased JAG1 expression, combined ERK1 and ERK2 knockdown was required to overcome redundancy within the pathway and to match the effect of TRB3 knockdown (Fig. S5). Consistent with these findings, overexpression of TRB3 in HepG2 human hepatocellular carcinoma (Fig. 3D) or MCF10A mammary epithelial (Fig. S6A) cells resulted in up-regulation of p-ERK1/2 and JAG1; this effect could be reversed with U0126, demonstrating MAPK-ERK dependence. These effects were more difficult to illustrate in MDA MB231 cells, likely due to an overactivated MAPK pathway (Fig. S6B). Taken together with previous reports that TRB3 promotes activation of MAPK-ERK, and the finding that MAP-ERK regulates JAG1 (Fig. 1), these data are consistent with a model where MAPK-ERK is a mediator of TRB3-regulated JAG1 expression.

Studies in mouse models of breast cancer bone metastases have implicated TGFβ-SMAD4 signaling in the regulation of JAG1 expression (14). In nonmammary cell lines TRB3 has been shown to stabilize SMADs and potentiate SMAD-mediated transcriptional activity by triggering degradation of SMAD ubiquitination regulatory factors (SMURFs) (24, 25). Based on these findings we tested whether TRB3 regulates TGFβ/SMAD4-dependent JAG1 expression in breast cancer. TGFβ treatment of serum-starved MDA MB231 cells resulted in up-regulation of JAG1 and this effect was dependent upon TRB3 (Fig. 3E). Furthermore, TRB3 knockdown resulted in a decrease in SMAD4 and JAG1 proteins (Fig. 3F) without affecting SMAD4 mRNA levels (Fig. S7A), suggesting a posttranscriptional mechanism of SMAD4 regulation by TRB3. To confirm the importance of SMAD4 in this sequence we demonstrated that SMAD4 knockdown alone reduced JAG1 expression (Fig. 3F). To determine whether TRB3-mediated degradation of SMURFs was required for SMAD4 protein expression, we performed knockdown of TRB3 and double knockdown of TRB3 and SMURF1 or SMURF2 (Fig. S7 B and C). Knockdown of TRB3 had no effect on SMURF1 and double knockdown conditions did not rescue SMAD4 or JAG1 levels, suggesting that SMURFs are not involved in TRB3 regulation of SMAD4 and JAG1. These findings suggest that in addition to its control of MAPK-ERK, TRB3 regulates JAG1 expression through SMURF-independent TGFβ/SMAD4 signaling.

JAG1 Rescues a Proliferation Defect Imposed by TRB3 Knockdown.

Based on the effects of TRB3 on MAPK and TGFβ pathways and on JAG1 expression, we predicted that knockdown of TRB3 should have a profound effect on breast cancer cell proliferation. We examined the effect of siRNA-mediated depletion of TRB3 on the proliferation of wild-type MDA MB231 cells and on an MDA MB231 variant that was engineered to overexpress JAG1 (JAG1 MDA MB231). Compared with Scr siRNA-treated cells, TRB3 siRNA achieved a significant knockdown of JAG1 in MDA MB231 cells but not in JAG1 MDA MB231 cells (Fig. 4 A and B, Insets). Corresponding with this, TRB3 knockdown resulted in a significant reduction in MDA MB231 proliferative capacity, whereas JAG1 overexpression rescued this effect (Fig. 4). The data demonstrate a critical role for TRB3 in maintaining MDA MB231 proliferation and implicate JAG1 in mediating this effect.

Fig. 4.

JAG1 rescues the proliferation defect imposed by TRB3 knockdown. Proliferation of MDA MB231 (A) and JAG1 MDA MB231 (B) cell lines in monolayer culture after treatment with either Scr siRNA (dashed curves) or TRB3 siRNA (solid curves). Cells were counted at 24, 48, 72, and 96 h after transfection. Insets, Western blot analysis of JAG1 protein expression in MDA MB231 (A) or JAG1 MDA MB231 (B) cells treated with either Scr or TRB3 siRNA. β-Actin expression is included as a loading control. M_W markers are shown in kilodaltons. Experiments were performed in triplicate; error bars represent SD. *P < 0.05 relative to Scr control.

TRB3 Knockdown Results in a Growth Deficiency of MDA MB231 Mouse Xenografts.

To confirm the importance of TRB3 to MDA MB231 cell proliferation and tumorigenesis we performed an orthotopic MDA MB231 tumor xenograft study, where MDA MB231 cells underwent ex vivo transfection with either Scr or TRB3 siRNA (Fig. S8A) followed by injection into the mammary fat pad of severely immunodeficient nonobese diabetic (NOD/SCID/IL2Rγ[−/−]) (NSG) female mice (26). We ensured injection of equivalent numbers of viable cells in all models through the Trypan blue dye exclusion method and an anoikis assay (27) (Fig. S8B). Parallel in vitro assays demonstrated persistent knockdown of JAG1 for at least 14 d posttransfection (Fig. S8C). Tumors derived from MDA MB231 cells transfected with TRB3 siRNA grew substantially slower than those derived from control cells as evaluated by tumor size measurements (Fig. 5A) and average tumor weight at sacrifice (Fig. 5B). Taken together, these results demonstrate that TRB3 plays a crucial role in breast cancer tumorigenesis and that down-regulation of TRB3 can reduce tumor growth in vivo.

Fig. 5.

TRB3 knockdown impairs xenograft growth of MDA MB231 cells. (A) Estimated cross-sectional area of tumor xenografts derived from Scr siRNA-treated (dashed curve; six NSG mice) or TRB3 siRNA-treated (solid curve; six NSG mice) MDA MB231 cells measured over 5.5 wk. (B) Weight of xenograft tumors derived from Scr siRNA-treated or TRB3 siRNA-treated MDA MB231 cells at 5.5 wk. Error bars represent SD.*P < 0.05 relative to Scr control.

Discussion

We performed high throughput screens to identify kinases and associated signal transduction pathways that promote JAG1/Notch activation in breast cancer. A kinase inhibitor screen and follow-up GSEA identified MAPK-ERK as the predominant regulator of JAG1 expression and Notch activation in this malignancy. Accordingly TRB3, a regulator of MAPK signaling was identified as a top hit in a complementary kinome siRNA screen. Herein we show that through the control of MAPK-ERK and TGFβ pathways, TRB3 is a master regulator of JAG1 expression and is required for the growth of basal-like breast cancer.

TRB3 is a pseudokinase, a class of proteins that comprises 10% of the kinase superfamily and that lacks one or more of the conserved amino acid motifs that are essential for catalytic activity (28). Despite lacking these characteristic motifs, pseudokinases are not impotent, vestigial remnants of functional kinases but rather, are crucial regulators of diverse cellular functions. In some instances pseudokinases function as scaffolding proteins, bringing together catalytically active molecules and their substrates. TRB3 plays such a role in normal cells during conditions of hypoxic/endoplasmic reticulum stress or nutrient deprivation where it becomes up-regulated and promotes survival by blunting potentially deleterious stress signals (29, 30). Indeed, TRB3 is up-regulated in cancers as a response to hypoxia (16, 29) and is associated with poor outcome (16, 31). However, an apparent discordance between the influence of TRB3 mRNA (16) and protein (32) on outcome in breast cancer remains to be explained. We hypothesize that in response to cellular stress, TRB3 protein influences outcome by promoting activation of key cancer signaling pathways such as MAPK-ERK, TGFβ, and JAG1/Notch.

JAG1/Notch signaling is emerging as a mediator of breast cancer progression and metastasis and is associated with the basal-like subtype (6, 33–35). Approximately 20% of breast cancer patients have basal-like disease, and despite an initial response to systemic cytotoxic chemotherapy, their disease follows an aggressive clinical course with early recurrence (36). Therefore, JAG1/Notch signaling and regulators of this pathway are attractive therapeutic targets in this breast cancer subtype.

How TRB3 influences tumor cell biology may be inferred from knowledge about the pathways that it regulates. Tumor-initiating cells (TICs) represent a small population of cells within some tumors that possess the unique ability to self-renew and to produce derivatives that maintain the tumor. The TRB3 target pathways Notch, MAPK-ERK, and TGFβ (37–39) have all been implicated in TIC maintenance, suggesting that through the control of these pathways TRB3 may regulate tumor initiation. The metastatic potential of epithelial tumors likely depends on a process known as epithelial-to-mesenchymal transition (EMT), where epithelial cells acquire a migratory mesenchymal phenotype (40). The Notch, TGFβ, and MAPK-ERK pathways interact with each other and have a synergistic effect on the production of factors that promote EMT and metastasis (41–43). Metastases also depend upon the seeding of metastatic sites by tumor cells, a process that Notch and TGFβ pathways have been shown to facilitate (14); TGFβ released from sites of bone metastases stimulates JAG1 expression in tumor cells, enabling paracrine Notch activation in osteoblasts and preosteoclasts and bone invasion. Collectively, these findings predict that TRB3 can potentiate tumor initiation and the metastatic capacity of breast cancer cells through its regulation of MAPK-ERK– and TGFβ-mediated JAG1/Notch activation. It remains to be shown whether activation of these pathways and processes account for the reduced survival associated with tumors with elevated TRB3.

Interestingly, TRB3 knockout mice are phenotypically identical to their wild-type littermates (44), suggesting that under normal physiological conditions this gene is either unnecessary or is functionally redundant with the other TRB paralogs. Thus, whereas TRB3 may have minimal effects on normal cells, it has a major impact on the survival and fitness of stressed cells including hypoxic tumor cells. This establishes TRB3 as an ideal therapeutic target whose elimination may render a low side-effect profile while being deleterious to tumor progression and metastasis.

Materials and Methods

Cell Culture and Reagents.

Cell lines were purchased from American Type Culture Collection. The MB231 TK-R cell line was generated by transfecting MDA MB231 cells with the pRL-TK vector (TK promoter-driven Renilla luciferase; Promega) together with pBABE-puro plasmid with selection of stable transfectants in 1µg/mL puromycin. JAG1-pcDNA 3.1 and pcDNA 3.1 were transfected into MDA MB231 cells to generate control and JAG1 overexpressing cell lines. MEK1/2 inhibitor U0126 was purchased from Cell Signaling Technology. GSI (N-[N-(3,5-difluorophenacetyl-L-alanyl)]-S-phenylglycine t-butyl ester; Calbiochem) was used at 50 μM. Human epidermal growth factor (EGF) (Austral Biologicals) was used at 50 ng/mL and TGFβ1 (Perpotech) was used at 1 ng/mL. Transfections of siRNA (50 nmol/L; Table S1) were performed using Lipofectamine RNAiMAX (Invitrogen).

High Throughput Screens for Notch Signaling Regulators.

High throughput screens were performed at the Simple Modular Assay Robotic Technology (SMART) facility at the Samuel Lunenfeld Research Institute. For the kinase inhibitor screen MB231 TK-R cells were plated at a density of 6,500 cells per well in 384-well plates together with HES1 promoter/firefly (Hes1-FF) adenovirus at a multiplicity of infection (MOI) of 100 and cultured for 18 h. Transduced cells were treated with aliquots from a collection of 240 kinase inhibitors (OICR-L100, Medicinal Chemistry Platform at the Ontario Institute for Cancer Research) at a final concentration of 5 µM in 0.1% DMSO for 48 h. After incubation, Notch activation was assessed by luciferase assay. The siRNA screen was performed using the Human siGENOME siRNA Library Protein Kinases SMARTpool (Dharmacon) targeting 720 genes encoding all known human protein kinases and consisting of a smart pool of 4 siRNAs per gene. Dharmacon nontargeting siRNA was used as a negative control and JAG1 siRNA as a positive control. MB231 TK-R cells were infected with Hes1-FF adenovirus at MOI 100 and then plated at a density of 16,500 cells per well in 96-well plates and cultured for 18 h before transfection with the siRNA library and controls at 34 nm concentration using (Lipofectamine RNAiMAX; Invitogen). Forty-eight hours posttransfection, cells were assayed for luciferase activity.

Gene Expression Analysis of MAP Kinase Activation.

A gene expression signature indicative of MAP kinase activation (17, 18) was mapped to the Affymetrix U133 2.0 plus platform yielding 387 of a possible 395 genes and to the Affymetrix 133A platform yielding 341 genes. For each gene, a single probeset was chosen based upon the highest level of variability between samples. To determine if MAPK kinase activation genes show statistically significant, concordant differences as a function of continuous JAG1 gene expression, the GSEA method was used with the dataset collapsed to gene symbols, 1,000 permutations and phenotype permutation type, and Pearson metric for ranking genes as previously described (45, 46). JAG1 and MAPK activation was compared in two sets of breast cancer cell lines (19, 47), and in the National Cancer Institute (NCI)-60 cell lines originating from cancer tissues of multiple types (20) prepared as previously described. Methodology of hierarchical clustering and classification of molecular subtypes of breast cancer cell lines was previously reported (9).

Western Blotting and Antibodies.

Cells were lysed in RIPA buffer (25 mM Tris, pH 7.6, 150 mM NaCl, 1% NP40, 1% DOC, 0.1% SDS) and protein was resolved by SDS/PAGE and blotted on PVDF membranes (Bio-Rad). Anti-JAG1 (sc-8303), anti-TRB3 (sc-271572), anti-Smad4 (sc-7966), anti-SMURF1 (sc-25510), and horseradish peroxidase (HRP) (sc-1615 HRP) were purchased from Santa Cruz Biotechnology. Anticleaved NOTCH1-Val-1744 (no. 2421), anti-ERK1/2 (no. 4695), and anti-p-ERK1/2 (no. 9101S) were purchased from Cell Signaling Technology and anti-uPA was purchased from Chemicon.

Proliferation Assays.

Sixty thousand cells were transfected in 24-well tissue culture plates in triplicate. At each time point cells were trypsinized and counted in a Vi-Cell XR Coulter counter.

Quantitative Real-Time (QRT) PCR.

Total RNA was extracted using RNeasy plus (Qiagen) following the manufacturer’s instructions. Reverse transcription of total RNA was carried out using iScript cDNA synthesis kit (Bio-Rad). PCR was performed with Power Sybr Green (Applied Biosystems). See Table S2 for primer sequences.

Adenovirus Construction, Amplification, and Purification.

Adenoviruses were constructed using the AdEasy system (Stratagene). Notch reporter was generated by inserting the human Hes1 promoter (−106 to −4) upstream of the firefly luciferase gene in pGL3-Basic vector (Promega) using traditional cloning methods. Hes1-FF was than cloned into the pShuttle vector. ADTRB3 was generated by cloning the TRB3 cDNA (Origene) into the pShuttle CMV vector. AD-293 packaging cells were used for recombinant adenovirus generation and amplification. For adenovirus purification and viral titer determination, the Adeno-X purification and qPCR titration kits (Clontech) were used, respectively.

Mouse Xenografts.

MDA MB231 cells were transfected with Scr or TRB3 siRNA followed by injection into the mammary fat pad of six female NSG mice to generate xenografts. Tumors were measured at constant intervals and growth curves plotted. Excised tumors were weighed.

Data Analysis.

The screens were analyzed using two different high throughput screen normalization methods: The first method was based on the FF/R ratios that were calculated for each test compound and siRNA. The second method determined a “B-score” where FF/R ratios were normalized and corrected for systematic and positional variability within the samples (21). Each screen was performed in duplicate. Inhibitors or siRNAs that reproducibly decreased the FF/R ratio or the B-score by more than 2 SD from the mean were considered as hits. Luciferase activity was assessed using the Dual-Glo luciferase reporter assay system (Promega).