Downregulation of Notch Pathway by a γ-Secretase Inhibitor Attenuates AKT/Mammalian Target of Rapamycin Signaling and Glucose Uptake in an ERBB2 Transgenic Breast Cancer Model

The Notch pathway is a conserved cell signaling pathway involved in differentiation, cell fate, and tissue homeostasis (1). Activation through binding with cell-surface ligand families Delta and Jagged produces a series of cleavages in the Notch receptor (2, 3). The final cleavage by the protease complex γ-secretase releases the Notch intracellular domain (NICD), which then translocates to the nucleus (4). NICD forms a complex with the transcriptional repressor CSL (5), which is subsequently transformed from repressor to activator, transcribing a number of downstream target genes, including members of the HES and HEY families as well as many others (1, 6).

Notch signaling is deregulated in a variety of cancers including T-cell acute lymphoblastic leukemia (T-ALL), non–small cell lung cancer, and breast cancer. In T-ALL, Notch signaling is activated through NOTCH1 receptor mutations found in >50% of patients (7) with translocations found in ∼1% of cases (reviewed in ref. 8). NOTCH3 activation has been observed in lung cancer cell lines and human tumors in which γ-secretase inhibitor (GSI) has been shown to have antitumor effects (9). Coexpression of NOTCH1 and its ligand JAG1 in breast cancer correlates with poor survival (10), and Notch pathway activation is seen in patients through decreased expression of Numb (11), a negative regulator of Notch signaling (12). Notch signaling events can be blocked by intervening at various sites, including inhibition of the final cleavage by γ-secretase inhibitors (GSI), which prevent the formation of NICD and are currently being evaluated in cancer patients.

Recent data indicate that GSI is effective in inhibiting proliferation and in induction of apoptosis. GSI effects are much higher in ERBB2-positive cell lines where ERBB2 is amplified and/or overexpressed (ZR-75-1 and MDA-MB-453) compared with ERBB2-negative cells (MCF-7 and MDA-MB-231) that lack ERBB2 amplification and where ERBB2 expression is low (13–15). A detailed mechanism of GSI effects and an evaluation of pharmacodynamic markers have not been explored in breast cancer xenografts. In addition, no studies have been reported in transgenic models of breast cancer where tumor development and microenvironment more closely reflect the human disease.

One transgenic mouse model is the ERBB2-driven mammary carcinogenesis model (neuT; ref. 16). Amplification or overexpression of the ERBB2 oncogene, a member of the receptor tyrosine kinase family, is amplified in more than one fourth of human breast cancers and is associated with poor prognosis (17). A variety of pathways or transcription factors (18) may contribute to tumor development and maintenance in ERBB2/neu–driven cancers, including the phosphoinositide 3-kinase (PI3K)/AKT/mammalian target of rapamycin (mTOR) axis (reviewed in ref. 19). ERBB2 overexpression is known to activate the PI3K/AKT/mTOR axis in human breast cancer cell lines and primary tumors (20). The AKT/mTOR pathway is activated in neuT tumors, which are sensitive to mTOR inhibitors (21), as inactivation of Pten accelerates the formation of primary and metastatic breast tumor in mouse mammary tumor virus (MMTV)-ERBB2 transgenic mice (22).

ERBB2 has been shown to influence Notch signaling, yet details of this response remain unclear. ERBB2 was reported to activate NOTCH1 in breast cancer cells (23), although inactivation of ERBB2 signaling with trastuzumab led to Notch pathway activation (24). Similar to ERBB2, Notch signaling is known to impinge on the PI3K/AKT/mTOR axis. The role of the Notch pathway in PI3K/AKT/mTOR signaling was discovered in anoikis and epithelial-mesenchymal transition in human papilloma virus–derived cancers (24). Ectopic expression of NICD increased AKT activation in Jurkat cells (25), and Notch signaling via PDK1 in pre-T cells promoted thymocyte growth as well as an mTOR-dependent proliferation of T-cell progenitors (26). Notch also induced activation of AKT in the breast cancer cell line MCF10A, and inhibition of Notch decreased AKT signaling (27). NOTCH1 activated PI3K/AKT/mTOR signaling by direct repression of PTEN via the Notch target gene HES1 (28). GSI inhibition of the PI3K/AKT/mTOR signaling resulted in decreased cell size in T-ALL cell lines (29), and PTEN mutation induced resistance to GSI (28). These studies indicate that Notch is an important regulator of the PI3K/AKT/mTOR signaling network.

Because ERBB2 is known to influence Notch and function partially through PI3K/AKT/mTOR signaling, we sought to determine if Notch signaling was active in the neuT model and how GSI would affect AKT signaling, tumor growth, and initiation. Here we report that Notch signaling is active in the neuT model and impinges on the PI3K/AKT/mTOR axis. GSI inhibits tumor growth and increases survival of neuT mice but has little effect on tumor initiation. Our data also suggest that the effects of GSI on the PI3K/AKT/mTOR pathway can be measured by noninvasive imaging such as positron emission tomography (PET) and may be useful for studying GSI response in a subset of ERBB2-positive breast cancer patients.

BALB/c-neuT-transgenic female mice were used in these studies (30). Mice with one or two palpable tumors were enrolled in the efficacy experiments. Mice received either 0.5% methylcellulose (vehicle) or GSI (MRK-003) at specified concentrations administered p.o. at different dosing schedules including daily, 3 d on/4 d off, or 1 d on/10 d off. Efficacy was measured by tumor number (initiation of palpable tumors) and/or tumor volume (by caliper), which were monitored twice a week throughout the study. Body weights and other clinical signs were assessed twice per week to determine tolerability. Mice were euthanized 6 h after the final dose. Tumors were weighed and analyzed by biochemical and immunohistochemical techniques. Blood was collected to determine GSI concentration. All procedures were done according to protocols approved by the Institutional Animal Care and Use Committee of Merck Research Laboratories.

Cells procured from the American Type Culture Collection were grown in complete medium in a 37°C incubator and treated with either DMSO or GSI at 1 μmol/L. Six hours after the GSI treatment, cells were harvested and protein lysate was prepared as described (31).

Formalin-fixed paraffin-embedded mouse breast cancer tissues or tissue microarrays (TMA) from human tumors were sectioned at 5-μm thickness and analyzed for Ki67 (1:200, clone SP6, Neomarkers), phospho-S6RP (1:75; Cell Signaling Technology), Glut1 (1:250; Abcam), ERBB2 (Ventana), and Hes1 (1:200; MBL). Automated staining was done using the ChromoMap Kit on the Discovery XT (Ventana Medical Systems) under standard conditions. TMA sections of each tumor were stained with antibody against ERBB2, HES-1, GLUT1, or pS6RP. Three consecutive TMA sections were stained with each antibody at the same time. A sample scored positive if more than 50% of the tumor retained stain in two of three consecutive sections. The intensity of staining was scored manually (high, 3; medium, 2; low, 1; no staining, 0) by two independent investigators (Supplementary Table S2).

Tumors were collected and protein lysates were made as described (32). Briefly, 30 μg of lysate were immunoblotted with antibodies (1:500 dilution) against S6RP, phospho-S6RP, ERBB2, phospho-ERBB2, Glut1, Pten, Akt, and phospho-Akt (S473) (Cell Signaling Technology). β-Actin or tubulin (Abcam) was used as loading control.

neuT mice were enrolled (n = 10 per group) into the imaging study when the total tumor volume of each mouse was between 150 and 300 mm³ as determined by micro-computed tomography (microCT) image analysis. Baseline 2-[¹⁸F]fluoro-2-deoxy-d-glucose ([¹⁸F]FDG), microPET, and microCT imaging scans were done before initiation of therapy and then weekly for 3 wk. After enrollment, mice were gavaged either with vehicle or with GSI at 300 mg/kg once a week. [¹⁸F]FDG was synthesized using standard methods (33), and 100 μCi were injected i.v. under isoflurane anesthesia. Sixty minutes after the radiotracer injection, mice were given 2 mL of a 1:9 dilution of a microCT contrast agent (Omnipaque-300, GE Healthcare). The contrast agent aided in the identification and segmentation of individual tumors (Supplementary Fig. S2). MicroPET images were collected using a Concorde Focus 220 microPET (Siemens Medical Solutions) using the 5-min static acquisition protocol. Immediately following microPET image acquisition, microCT images were acquired on a GE eXplore Locus Ultra microCT system (GE Healthcare) using the 16-s anatomic protocol. After completion of microPET and microCT image acquisition, the image data were reconstructed to give the respective image data.

Two tumors per mouse that could be identified longitudinally in all four scans by both microPET and microCT were selected for analysis. For microPET image analysis, tumor regions of interest were manually drawn using ASPIRO (Concorde Microsystems) and uptake of [¹⁸F]FDG was reported in normalized activity units. Tumor volumes were determined from microCT images using the Amira 4.1.1 software (Mercury Computer Systems, Inc.). Fold increases of tumor volumes from baseline scans were calculated.

One-way ANOVA followed by Dunnett's multiple comparison test was conducted to evaluate the effect of GSI on tumor (number and volume) and body weight. Student's t test was applied for the difference between control and treated for Ki67, tumor weight, pharmacokinetic, and survival analysis.

Female neuT-transgenic mice were monitored weekly until a minimum of one but not more than two palpable tumors were identified for enrollment. Following enrollment, neuT mice were given either vehicle or GSI under several different dosing regimens. Body weight and the number of palpated tumors were determined twice per week for 3 weeks. Tumor growth inhibition required an intermittent dosing schedule, as we showed previously in T-ALL models (31, 34; Supplementary Fig. S1). Antitumor activity was achieved using 150 mg/kg 3 days a week, a single dose of 300 mg/kg p.o. once a week, or a dose of 300 to 500 mg/kg every 10 days (Fig. 1A and B; Supplementary Fig. S1A, C, and D). Other doses tested were either not tolerated or not efficacious (Supplementary Fig. S1; Fig. 2A, and data not shown).

In addition to the significant effect on tumor number, antitumor effects were also observed when tumor volume was quantified using caliper or microCT imaging (Fig. 1B; Supplementary Fig. S1A). The volumetric data correlated with tumor weight, reduced 27% and 52%, respectively, in mice treated with 300 and 500 mg/kg GSI once every 10 days compared with vehicle (Supplementary Fig. S1D). To determine if antitumor effects translated to increased survival, tumor-bearing mice were treated for 3 weeks and monitored biweekly thereafter without treatment. When tumor volume reached 10% of the body weight, the mouse was removed from study. The survival analysis suggests that the GSI antitumor effects translate into a survival benefit when compared with the vehicle (Fig. 1C). These data indicate that GSI may provide clinical benefit by decreasing the tumor burden and increasing the survival of ERBB2-driven breast cancers.

The Notch signaling pathway may be an important regulator of tumor-initiating cells or breast cancer stem cells (35). To determine if Notch might play a role in onset of tumor formation in this neuT model, we enrolled 30 neuT mice at the age of 4 weeks where no microscopic tumor was evident (data not shown). Mice were divided into three groups and treated either with vehicle or with 150 or 300 mg/kg GSI weekly for 18 weeks. GSI did not delay the appearance of the first palpable tumor when treated at doses known to provide antitumor effects and inhibit Notch signaling in these tumors (Fig. 2B and data not shown). These results suggest that Notch signaling affects the maintenance of established tumors (microscopic and macroscopic), yet provides little support in tumor initiation in this model. Because the detection of tumor initiating cells is a rapidly advancing field, more detailed studies are warranted.

Following confirmation of the GSI antitumor effects, we wanted to determine the correlation between response and Notch pathway inhibition in this model. We began by determining the plasma concentration of GSI-treated animals. Plasma levels peaked ∼5 hours after a single 150 mg/kg p.o. administration of GSI and returned toward baseline levels by 48 hours (Fig. 3A and data not shown). Similar levels and dose-dependent response were observed in mice treated with GSI over the course of 5 weeks (150 mg/kg 3 days on 4 days off and 300 mg/kg once weekly; Fig. 3B). Pharmacodynamic response was evaluated by quantitative PCR and immunohistochemistry after 3 weeks of treatment. In tumors, quantitative PCR provided evidence that the plasma levels achieved were capable of modulating Notch target genes (c-Myc, Hes1, Dtx1, Hey1, Hes7, Hey2, and Hey). A 10-fold decrease in Dtx1, Hes7, HeyL, and Myc was associated with doses that provided antitumor effects, whereas noneffective doses provided a <10-fold decrease in these target genes (Fig. 3C). Although less dramatic effects were observed in Hes1 mRNA levels relative to other target genes monitored, effective doses decreased Hes1 protein levels as measured by immunohistochemistry (Fig. 3D). These events corresponded with a decrease in Ki67 (Fig. 3D, bottom), indicating that GSI-induced inhibition of Notch signaling results in decreased proliferation and antitumor activity in the neuT model.

ERBB2 and Notch signaling are known to impinge on the PI3K/AKT pathway. We evaluated the status and response of three PI3K pathway markers (phospho-Akt, phospho-S6RP, and Pten) by immunoblot and immunohistochemistry in neuT tumors treated with GSI. Phospho-AKT and phospho-S6RP show that the AKT pathway is activated in these tumors relative to normal mammary glands (Fig. 4A). AKT/mTOR pathway activation is also associated with Glut family members in many solid tumors, and we therefore wanted to determine whether Glut proteins are expressed in neuT tumors. Our data suggest that Glut1 is overexpressed in the epithelial cells of neuT tumors (Fig. 4B). Based on these findings, we initiated studies to observe whether Notch inhibition has any effect on PI3K/AKT/mTOR pathway activity and Glut1 expression in this model. GSI decreased phospho-Akt as measured by both immunoblot and immunohistochemistry (Fig. 4C and data not shown). To understand the cause of decreased level of phospho-Akt, we have analyzed the level of total Pten protein. Data show that GSI has no effect on the total protein level of Pten in neuT tumors (Fig. 4C). The effects of GSI on ERBB2 were evaluated by the levels of total and phospho-ERBB2 in neuT tumors treated with either GSI or vehicle. Both total and phospho-ERBB2 remained unchanged after GSI treatment, whereas a decreased level of phospho-ERBB2 was observed in tumors treated with lapatinib (Fig. 4C and data not shown). Similarly, data indicate a downregulation of phospho-S6RP by treatment with GSI in neuT mice (Fig. 4D, top). Together, these results show that the PI3K pathway is activated in these tumors and was downregulated by GSI. To determine the relationship between inhibition of the AKT/mTOR pathway by GSI and the Glut1 levels, we performed immunohistochemistry analysis of Glut1. Data showed a significant decrease of Glut1 by GSI (Fig. 4D, bottom).

Following observation of GSI-mediated downregulation of Glut1 in neuT tumors, we wanted to determine whether these tumors display increased glucose uptake as compared with the surrounding normal tissues and modulation by GSI. We administered [¹⁸F]FDG and analyzed the uptake of the radiotracer in vivo by microPET imaging. Baseline images showed measurable levels of [¹⁸F]FDG uptake in neuT tumors as compared with other tissues (Fig. 5A). The radiotracer was also observed in mouse tissues (heart and brown fat) known to express high levels of Glut1 as well as in the urinary bladder because the radiotracer is cleared via the kidneys. To determine if GSI affects glucose uptake and whether [¹⁸F]FDG-PET imaging could serve as a functional readout, we conducted microPET imaging analysis on neuT mice for 3 weeks with an efficacious regimen of 300 mg/kg GSI once per week (Fig. 1). GSI treatment decreased [¹⁸F]FDG uptake at all treatment points measured starting from day 7 to day 21 (Fig. 5A and B). These data show the active uptake of glucose by neuT tumors and that GSI can decrease the [¹⁸F]FDG radiotracer uptake as monitored by noninvasive microPET imaging.

Our data show a role for γ-secretase in tumor growth progression and the connection of the Notch and PI3K pathways. Next, we wanted to determine whether these results were restricted to mouse tumors or were reflected in human breast cancer. We treated ERBB2-positive (MDA-MB-453) human breast cancer cells with GSI and observed that HES1 and GLUT1 protein expression was downregulated when cells were treated with 1 μmol/L GSI as compared with DMSO (Fig. 6A). No difference in ERBB2 expression was observed in these cells, indicating that the GSI effects are downstream of ERBB2 and not a direct effect on the receptor (Fig. 6A). Next, we wanted to determine the levels of ERBB2, HES1, pS6RP, and GLUT1 in human breast cancer. Seventy-six human breast tumor sections from primary (n = 29) and metastatic (n = 47) patients were stained with antibodies against ERBB2, HES1, pS6RP, or GLUT1 (Fig. 6B). We observed that 51% (39 of 76) of the tumors are ERBB2 positive, 58% (44 of 76) are HES1 positive, 75% (57 of 76) are pS6RP positive, and 71% (54 of 76) are GLUT1 positive (Fig. 6C; Supplementary Table S1). No signal from phospho-Akt (S473 and T308) was observed and may be a result of phospho-epitope stability. We observed a correlation of ERBB2 protein level with pS6RP, GLUT1, and HES1 proteins in a subset (27%) of human breast tumor specimens. The high level of ERBB2 protein may be due to the amplification and/or overexpression of ERBB2 in these tumors. We assessed tumor grade by pathologic scoring and no correlation between tumor grade and increased staining of these markers was observed (Supplementary Table S2). However, a trend toward positive staining of these markers was observed in metastatic tumors (Supplementary Tables S1 and S2). Expression of all four proteins was observed in 27% (21 of 76) of the tumors, showing that this pathway (ERBB2, Notch, and AKT/mTOR) is active in approximately one quarter of human breast cancers. Together, these data suggest that GSIs may provide therapeutic benefit to this population of breast cancer patients and their response may be monitored by noninvasive functional imaging techniques.

The neuT transgenic mouse is important for understanding ERBB2-driven breast cancers because it aptly mimics some form of human disease. It provides advantages over human breast cancer xenograft models by its nature as a primary autochthonous tumor and development in the presence of an intact immune system. However, like other models, it has limitations, one being the overexpression of ERBB2 driven by MMTV promoter (not by amplification), which begins at puberty. These factors may influence complex developmental signaling pathways such as Notch. Here we describe the effect of Notch pathway modulation in the neuT breast cancer model using a γ-secretase inhibitor.

We observed an active Notch signaling pathway as evidenced by the ability of GSI to induce a dose-dependent decrease in the expression of several Notch target genes, including Hes1 and Myc. The Notch target genes Hes1 and Myc are important regulators of the GSI antitumor effect. Alteration of MYC is associated with a very short-term prognosis in human breast cancer (36, 37). Myc has been shown to induce breast tumor formation in MMTV mouse models, and turning off MYC expression resulted in tumor regression (38, 39). In Notch-dependent T-ALL cells, overexpression of MYC can block the effect of GSI (32). In neuT tumors, downregulation of Myc associated with antitumor activity on administration of GSI suggests that Notch is driving Myc and may be partly responsible for the antitumor effects.

Our results also support a role of AKT signaling in the GSI-dependent response of neuT tumors because a decrease in phospho-Akt and phospho-S6RP was observed. The Notch target gene HES1 has been shown to regulate cell growth by two mechanisms, by repressing the cell cycle regulator p27^kip1 (40) and through direct regulation of the AKT/mTOR signaling pathway by repressing PTEN expression (28). Therefore, regulation of this pathway in neuT tumor may be due to the regulation of PTEN activity by HES1. Here we show that inhibition of Notch signaling in neuT mammary tumors is associated with decreased cell proliferation as measured by Ki67.

The AKT/mTOR signaling pathway is known to influence cell growth through anabolic processes that control glucose metabolism, ATP production, and glucose uptake through HIF1 transcriptional control of the glucose transporter GLUT1 (41, 42). Elevated levels of GLUT1 provide increased uptake of the substrate [¹⁸F]FDG, which serves as a clinically relevant PET imaging agent for measuring the response of metabolically active tumors. Increased Glut1 expression and increased [¹⁸F]FDG uptake in neuT breast tumors reflect their elevated metabolic activity relative to normal breast tissue that can be modulated by GSI. Although many factors may contribute to the GSI antitumor effects, we introduce data showing that a reduction of Glut1 may also be a contributing factor. Changes in tumor volume as measured by caliper, microCT, and tumor weight were similar to the effects observed by [¹⁸F]FDG-PET and suggest that this latter measure may provide an early readout of clinical response. These data suggest that [¹⁸F]FDG-PET may be useful in monitoring GSI response in other ERBB2-positive cancers, including non–small-cell lung, pancreatic, endometrial, bladder, and gastric cancers.

We also evaluated the effects of GSI on tumor initiation using the neuT model by treating mice at 4 weeks of age, before the appearance of microscopic tumor. Results show that, under these parameters, GSI was ineffective in preventing the occurrence of palpable tumors. These functional data and the fact that GSI had no effect on ERBB2 levels in the human cell lines tested suggest that the transgene in the neuT model was not directly affected by MRK-003. Because GSI does not affect ERBB2, a combination of GSI with ERBB2 inhibitors may prove beneficial and is further supported by observations that ERBB2 can influence the activity of Notch (43) and inhibition of ERBB2 via trastuzumab can activate Notch signaling (23). This information will be important in considering GSI as a monotherapy or in combination with trastuzumab or lapatinib in metastatic breast cancers, which stain for the four markers (ERBB2, HES1, pS6RP, and GLUT1).

Due to its complex nature, it is important to understand how gain- or loss-of-function events influence GSI response in other cancer types. Activation of the PI3K/AKT/mTOR pathway by mutations of PIK3CA, PTEN, LKB1, and TSC2 or by amplification of PIK3CA, AKT1, AKT2, p70S6K, and IKBKE is common in a majority of solid tumors, including breast (44), and PIK3CA mutation is more common in hormone receptor–positive and ERBB2-positive breast cancers (45). Likewise, FBW7 mutations have also been shown to increase Notch signaling; however, they also lead to diminished Notch-dependent regulation of MYC, conferring GSI resistance to cells harboring both NOTCH1 and FBW7 mutations (46). It will be important to understand how these events influence Notch signaling if clinical benefits from GSI treatment are to be fully realized.

GSIs provide antitumor effects through several mechanisms, including direct effects on Notch-dependent tumor cells as shown here, in addition to indirect effects via tumor angiogenesis (47), immune modulation (48), and tumor-initiating cells (49). In addition, γ-secretase processes a number of substrates (50) besides Notch; therefore, the antitumor effects of GSI shown here may also be influenced by these events. In summary, we show that the Notch/AKT/mTOR signaling pathway is active in the neuT model as well as in human breast cancer (Supplementary Fig. S3). Inhibition of Notch signaling in ERBB2-positive preclinical models results in modulation of the Notch/AKT/mTOR signaling pathway and attenuates glucose uptake associated with antitumor effect. These data support the use of GSI for treatment of ERBB2/NOTCH/pS6RP/GLUT1–positive breast cancer patients whose response may be monitored by noninvasive [¹⁸F]FDG-PET imaging.

All authors except T. Yeatman and D. Coppola are employees of and hold ownership interests in Merck & Co., Inc. T. Yeatman and D. Coppola received a research grant from Merck & Co., Inc.

We thank Nirah Shomer, Nacy E. Kohl, and Carol Rohl of Merck Research Laboratories-Boston for technical support, and Maria K. Miyawa of Moffitt Cancer Center for assistance in tumor grading.

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.