Targeting Specific Regions of the Notch3 Ligand-Binding Domain Induces Apoptosis And Inhibits Tumor Growth In Lung Cancer

Abstract

Like many signaling pathways in development, the Notch pathway also plays a role in cancers when it is dysregulated. Studies involving deletions of several epidermal growth factor (EGF)-like repeats in the extracellular domains of Drosophila Notch and mammalian Notch1 have identified potential ligand binding sites. However, the ligand binding domain in Notch3 is not known. Using a screening library of 155 peptides representing the entire 34 EGF-like repeats in Notch3, we discovered two distinct ligand binding regions involving the 7–10 and 21–22 repeats. They are also distinct from the putative ligand binding region of Notch1. The peptides from these regions induced apoptosis and reduced expression of the Notch3-dependent gene Hey1. They also bind directly to the ligand Jagged1, suggesting that the mechanism of activity involves the disruption of the interaction between Notch3 and Jagged1. Notch3 recombinant Fc-fusion proteins based on peptide sequences induced apoptosis and suppressed tumor growth in xenografts. The findings from this study provide a promising rationale and mechanistic basis for Notch3 receptor-targeted therapeutic development in the treatment of patients with lung cancer.

INTRODUCTION

Notch3 is a type I transmembrane receptor belonging to a family of proteins essential for cellular differentiation and embryonic development. In mammals, there are 4 Notch receptors (Notch1-4) and two families of ligands, Jagged (Jagged1, 2) and Delta-like (Dll1, -3, -4). Binding of the ligand to the extracellular domain of the Notch receptor triggers two successive, proteolytic cleavages and untethers the Notch intracellular domain (ICD) from the cytoplasmic membrane. The Notch ICD is then translocated to the nucleus, binds to the transcription factor CSL, and induces expression of target genes. These genes include the Hairy-enhancer of Split (HES) and hairy and Enhancer-of-split related with YRPW motif (Hey) families.

Activation of the Notch pathway depends on the interaction between the ligand and the receptor’s extracellular domain (ECD) with subsequent release of the activated ICD. Notch3 is a large protein containing 2321 amino acids with a predicted molecular mass of 243.66 kD. The Notch3 ECD, a region containing the ligand recognition site, is estimated to be 210 kD. Identifying the part within the large ECD important for receptor-ligand interaction will help to better understand the biology of Notch3 signaling and therapeutic design. Using deletion mutants and point mutations of Drosophila Notch and mammalian Notch1, the identified ligand binding site appears to involve EGF-like repeats 11–12 (1, 2). However, given the functional diversity and the variation in tissue distribution among the different Notch family members, we hypothesized that the targetable ligand recognition sites on Notch3 receptor differ from that of other family members.

Notch3 is overexpressed in about 40% of resected non-small cell lung cancers (NSCLC), and its suppression results in loss of the malignant phenotype both in vitro and in vivo (3, 4). In both development and cancer, Notch has been shown to crosstalk with oncogenic pathways such as the EGFR/ras/MAPK (mitogen-activated protein kinase) pathway (3, 5–7). Thus, targeting this pathway represents a rational strategy in the treatment of patients with lung cancer. One approach currently being explored in clinical trials is blocking the essential proteolytic processing of Notch receptors with γ-secretase inhibitors. The efficacy of this class of compounds needs exploring, but the relative lack of target specificity suggests that new, more specific, strategies targeting this pathway should be pursued.

In this study, we identify the domains within Notch3 ECD important for ligand recognition and binding. Using a high-throughput system and a Notch3 peptide library, we discovered two previously unknown regions, EGF-like repeats 7–10 and 21–22, important for Notch3 activation. In addition, we demonstrated that interfering peptides and recombinant proteins mimicking these regions can abrogate Notch3 activation, induce apoptosis, and inhibit tumor growth in vivo. The findings of present study not only give novel insights into Notch3 signaling, but also establish a foundation in which targeted therapy can be developed.

MATERIALS AND METHODS

Peptide Library

The peptide library consisted of 155, synthetic peptides. Their sequences were 5–15 amino acids in length and spanned nearly the entire Notch3 ECD. Each peptide represented a unique extracellular site on the ECD, with peptide 1 representing the amino terminus and peptide 155 representing the carboxyl terminus of the last EGF-like repeat. They were synthesized by SynPep (San Francisco, CA) and diluted in deionized H₂O to bring the concentration to approximately 10 mg/ml of peptide in 1X PBS. The peptides were biotinylated using E-Z Link Biotin BMCC (Thermo Fisher Scientific, Inc., Rockford, IL) in PBS at a molar ratio of approximately 1–2 moles of biotin per mole of synthetic peptide for imunofluorescence staining and pull-down assays.

Cell Culture and inhibitor

The Notch3-expressing lung cancer cell line HCC2429 was established as previously described (8). HEK293T cells and HeLa cells were obtained from the American Type Culture Collection (ATCC) and maintained in DMEM with 10% FCS. MRK-003 was provided by Merck, Inc. & Co. (Whitestation, NJ), and its formulation was described previously (9).

Apoptosis Screen of Peptide Library

Both HCC2429 and HeLa cells were seeded onto 384-well plates at 3,000 cells in 50 µl per well. 25 µl of Annexin V Alexa Fluor 680 (Invitrogen, Inc., Carlsbad, CA), diluted 1:2,200 in RPMI and 10 µl of peptide (diluted to 0.1 mg/ml in RPMI), were added. After an overnight incubation, the treated cells were analyzed with a FMAT 8100 HTS System fluorescent plate reader (Applied Biosystems, Foster City, CA). Each peptide was assayed in quadruplicate.

Notch3 Deletion Mutants

With the Notch3 ECD as template, inverse PCR was used to generate deletions of Notch3 EGF-like repeats 7–10 and EGF 21–22. The ECD containing the deletions were confirmed by DNA sequencing, cloned into pcDNA4, and re-ligated to Notch3 intracellular fragment, to generate full-length Notch3 deletion mutants.

In vitro Pull-down Assay

HEK293T cells were transfected with HA-tagged Jagged1, (provided by Dr. Artavanis-Tsakonas) using Lipofectamine 2000. The cells were lysed in NP-40 Buffer (10 mM Tris-HCl pH 7.5, 150 mM NaCl, 1% NP40 plus 50mM protease inhibitors). One microgram of biotin-labeled peptides and Streptavidin-conjugated magnetic beads (Promega, Madison, WI, USA) was used to pull down HA-tagged Jagged1. The resulting proteins were resolved on SDS-PAGE and detected with an anti-HA antibody. For the Fc-fusion protein binding assay, 5 µg of Fc-fusion protein and protein A agarose beads (Sigma-Aldrich, Inc, St. Louis, MO) were used.

Immunofluorescent Staining Assay

HCC2429 and HEK293T cells were plated on glass chamber slides. After 24 hrs, the µg/cells were rinsed twice in PBS and fixed in 4% paraformaldelyde and treated with 1 ml biotin-labeled peptides and 0.5 µg/ml AlexaFluor 488 labeled streptavidin (Invitrogen, Inc., Carlsbad, CA). TO-PRO3 (Invitrogen, Inc., Carlsbad, CA) was used for nuclear staining. The cells were then examined under confocal fluorescence microscopy.

Antibodies

Notch3 and HA-targeted Jagged1 were detected, using a rabbitNotch3 antibody (Orbigen, Inc., San Diego, CA) and a anti-HA monoclonal antibody (HA-7) (Sigma-Aldrich, St. Louis, MO), respectively, at 1:1000 dilution. The goat, anti-human IgG-HRP antibody (sc-2453) (Santa Cruz Biotechnology, Inc., Santa Cruz, CA) and the mouse anti β-tubulin monoclonal antibody (AA2) (Millipore, Billerica, MA) at 1:5000 dilution were used to detected hFc-fusion protein and β-tubulin, respectively.

Fc-fusion protein expression

The peptide DNA sequences were cloned into the N-terminal of pFUSE-hIgG1-Fc2 and pFUSE-mIgG1-Fc2 vectors (Invivogen, San Diego, CA). These vectors produce secreted fusion protein in mammalian cells. The plasmids were then transiently expressed in HEK293E, and the proteins were purified from culture medium with a protein A/G column (GE Healthcare Life Sciences). The eluted Fc-fusion proteins were buffer-equilibrated with PBS buffer using a HiTrap desalting column (GE Healthcare Life Sciences).

Real-time PCR

Total RNA was extracted from HCC2429 or HeLa cells 24 hrs after peptides treatment or transfection with deletion mutants, using the Qiagen RNase Mini kit. RNA was reverse transcribed with the Superscript II First-Strand Synthesis kit (Invitrogen, Inc., Carlsbad, CA) and quantitated using the iQ5 Multicolor Real-Time PCR detection system (Bio-RAD, Hercules, CA) and QuantiText SYBR green RT-PCR kit (Qiagen, Valencia, CA). Annealing temperature for PCR was 58° C with the following primers: for GAPDH, 5' TGCACCACCAACTGCTTAGC 3' for sense and 5' GGCATGGACTGTGGTCATGAG 3' for antisense; for Hey1, 5 ’ AGATGACCGTGGATCACCTG 3’ for sense and 5’ TGTTGAGAGCGAAACCAGTC 3’ for antisense; and for Hes1: 5’ AGAAGGCGGACATTCTGGA 3’ for sense and 5’ GAGTGCGCACCTCGGTATTA 3’ for antisense. The threshold cycle value (Ct) was determined with iCycler Optical system interface software. Mean Ct of Hey1 or Hes1 was calculated from triplicate measurements and normalized with the mean Ct of the gene GAPDH as internal control.

Apoptosis Assay

HCC2429 cells were treated with peptides or Fc-fusion proteins for 24 hours and maintained in serum-free RPMI medium. Percent apoptosis was determined using the Annexin V-FITC Apoptosis Detection kit (Calbiochem, La Jolla, CA) and a Beckman Coulter FACS Calibur Flow Cytometer (Beckman Coulter, Inc., Fullerton, CA).

In vivo tumorigenicity

1 × 10⁶ HCC2429 cells suspended in 50% Matrigel (BD Bioscience, San Jose, CA) were injected subcutaneously into hind limbs of thymic 4- to 6-week-old female nude mice (nu+/nu+). When the tumors were palpable, the mice were treated with Fc-control or with a single loading dose of recombinant protein at 15 mg/kg followed by dose of 10 mg/kg every 3 days. The tumor size was measured every 3 days with a caliper. Tumor Volume (TV) was calculated with the formula: TV = (Length) × (Width)² / 2.

Statistical analyses

The size of implanted tumors at different time intervals after treatment was compared with that of control treated with mouse Fc. Unless specifically stated, statistical inference in comparative experiments both in vivo and in vitro was obtained using Wilcoxon rank-sum test. For all statistical comparisons, the differences were considered significant at P < 0.05.

RESULTS

Peptide Library Screening Identifies Potential Ligand Binding Sites

Notch receptors differ in the number of tandem EGF-like repeats in the ECD. Notch3 contains 34 EGF-like repeats, whereas Notch1 possesses 36. In contrast to Notch1, the ligand binding site for Notch3 is not as well characterized. Therefore, to identify Notch3 binding sites, we created a peptide library consisting of 155 short peptides sequences, 5–15 amino acids in length, spanning the entire 34 EGF-like repeats within the Notch3 ECD. Because inhibition of Notch3 induces apoptosis in tumor cells, the conjugated carrier peptides were then screened for the ability to induce apoptosis. Out of the 155 peptides, we identified 15 peptides with reproducible apoptosis-promoting activities in both HCC2429 and HeLa cells (Table 1). The effect on apoptosis by these peptides was dose-dependent (Figures 1A, B). Interestingly, the locations of these peptides mapped to two discrete regions, EGF-like repeats 7–10 and 21–22 (Figure 1C). The amino acid sequences from Notch3 repeats 7–10 are most similar to Notch1 EGF-like repeats 8–11 with 79% percent identity (data not shown). When we compared the Notch3 sequence with that from putative Notch1 ligand binding sites, the EGF-like repeats 11–13, only 40% identity was observed, suggesting that the Notch3 binding domain differs from that of Notch1. An example of the alignment between Notch3 EGF-like repeat 7 and Notch 1 EGF-like repeat 11 is shown in Figure 1D.

Table 1.

List of peptides, their Location, and effect on apoptosis in HCC2429

Peptide ID	Peptides Sequences	Effects On Apoptosis	EGF-like Repeats
N16	CFNTLGGHS	+	7
N17	CVCVNGWTGES	++	7
N22	CVNTQGSFL	+	10
N52	CTCHGGYTGPS	+	21
N65	CREAAAQIGVRLEQL	+	21
N70	CIDLVARYL	+	29
N102	CATAV	++	8
N103	CFHGAT	++	8
N105	CVSNP	+	9
N17	CTFGV	+	16
N130	CDQDIND	+	21
N132	CLNGGS	++	22

Figure 1.

(A) Result from representative experiment performed on the FMAT™ 8100 HTS fluorescent plate reader and assayed with annexin V-Alexa 680 showing that Notch3 peptides N132, N105, N103, and N102 induced apoptosis. Fifteen of the total 155 peptides induced apoptosis in HCC2429 cells . Each peptide was assayed in quadruplicate (upper panel). A dose response was noted in which signal intensity was correlated with peptide concentration (lower panel). (B) The peptides with pro-apoptotic activity mapped to two distinct regions, EGF-like repeats 7–10 and 21–22, within the extracellular domain. Z axis is fluorescence intensity, as a measure of apoptosis. Axes X and Y show the location of individual well on the plate. (C) Alignment of Notch3 EGF-like repeat 7 and Notch1 EGF-like repeat 11 demonstrating 40% identity (blue background). Conserved substitution (*) was observed in five residues within these repeats. (D) Cartoon and surface representations of an EGF-like domain reconstructed using PyMOL molecular visualization software. Location of N17 (upper panel) and N132 (lower panel) sequences based on Class II EGF domain consensus sequence are colored in blue, illustrating the putative surface involved in ligand binding.

The highly conserved class II EGF-like repeat is observed in all Notch receptors, and its secondary structure contains a core with β-pleated sheet, three disulfide bonds and a series of loops (10). At this time, only the structure of the Class II EGF-like domain from human Notch1 is known. Because Notch3 also contains Class II EGF-like repeats, we mapped the Notch3 peptides with pro-apoptotic promoting activity to the Class II consensus sequence (PDB ID: 2VJ3). Interestingly, the peptide sequences mapped to the loop regions of the EGF-like repeats, suggesting that the loop region is responsible for ligand interaction. Molecular visualizations of the relative positions of N17 and N132 peptides within an EGF-like repeat are shown in Figure 1E (blue).

Notch3 peptides bind to Jagged1 and inhibit Hey1 transcription

To determine whether the apoptosis induced by the peptides is Notch-dependent, we examined the ability of the peptides to bind to Jagged1 and to alter transcription of Notch-dependent genes. Using biotin-labeled Notch3 peptides, we found that the labeled peptides bind to Jagged1-expressing HCC2429 but not HEK293T, which does not express endogenous Jagged1 (Figure 2A). Of the fifteen peptides identified, six were found to both induce apoptosis and bind to Jagged1. This interaction was subsequently confirmed with the in vitro pull-down studies (Figure 2B). The peptides also inhibited Hey1 transcription with varying potency (Figure 2C). Interestingly, Hes1 transcription was not altered (data not shown). These findings were consistent with our earlier observation that Notch3 preferentially regulates Hey1 and not Hes1 in our lung cancer models (4). The ability of the peptides to induce apoptosis confirmed the findings from the fluorescent screening assay (Figure 2D). Of the 15 peptides identified by FMAT 8100 HTS screen, 6 induced apoptosis and bound to Jagged1.

Figure 2.

Notch3 peptides bound to Jagged1 and inhibited transcription of Notch3-dependent gene Hey1. (A) Fluorescent-labeled Notch3 peptides N16, N17, N102, N103, N132 (green) bind to HCC2429 cells expressing Jagged1, but not to HEK 293T cells that do not express endogenous Jagged1. (B) Notch3 peptides inhibited signaling through binding to Jagged1. Immunoprecipitation experiment demonstrating that Jagged1 binds to Notch3 peptides but not control peptide. No: no input; C: control peptide. (C) Treatment of lung cancer cell line HCC2429 with Notch3 peptides reduced transcription of Notch3-dependent gene Hey1 determined by real-time RT-PCR. Note N17 peptide exhibited both the highest apoptotic activity and greatest reduction in Hey1 transcription. (D) All Notch3 peptides can induce apoptosis in Notch3-expressing HCC2429 cancer cells. Cells treated with a γ-secretase inhibitor MRK003 were used as positive control.

Notch3 Fc-fusion proteins bind to Jagged1 and inhibit Notch3 activation

A major limitation to using peptides for in vivo applications is their short biological half-life in the blood stream. To overcome this limitation, we utilized Notch3 Fc-fusion proteins, in which recombinant protein is fused to the Fc domain of human gamma immunoglobulin (IgG). Fc-N17, Fc-N16,17 and Fc-N132 Fc-fusion proteins reduced activated Notch3 at differing degree. Fc-N132 had a greater effect than either Fc-N16 or Fc-N130, suggesting that the inhibiting activity may be related to the sequences themselves and not the length. A similar effect was observed when conditioned media containing secreted Fc-fusion protein were used (Figure 3A). Interestingly, while not all Fc-fusion constructs affected Notch3 activation, they all retained the ability to bind to Jagged1 (Figure 3B).

Figure 3.

Notch3 Fc-fusion proteins bind to Jagged1 and inhibit Notch3 activation. (A) Transfection of Notch3 Fc-fusion expression plasmids Fc-N16, Fc-N16,17 and Fc-N132 into HCC2429 down-regulated expression of Notch3 ICD. Conditioned media from transfected HEK293T also reduced activated Notch3 in HCC2429. Similar to the previous transfection experiment, Fc-N16, Fc-N16,17 and Fc-N132 can reduce Notch3 ICD level but not Fc-N16 or Fc-N130. (B) Consistent with the peptide data, the immunoprecipitation experiment shows Fc-fusion proteins bound to Jagged1, but not Fc control.

Notch3 Fc-fusion proteins induce apoptosis and inhibit tumor growth in vivo

Treatment with purified Fc-N16,N17 and Fc-132 proteins resulted in inhibition of Notch3 activation to levels resemble those obtained with MRK003 (Figure 4A). This observation confirmed our early peptide data (Figure 2D). To determine the effect of Notch3 Fc-fusion proteins in vivo, we utilized a HCC2429 human lung cancer xenograft model. We observed a statistically significant reduction of tumor volume with Fc-N16, N17 and Fc-N132 treatment compared with Fc control after 12 days of treatment. After 16 days, the average tumor volumes with Fc-N16, N17 (0.256 cm³) and Fc-N132 (0.256 cm³) treatment showed a two-fold reduction compared with Fc control (0.612 cm³) (Figure 4B, C).

Figure 4.

Notch3 Fc-fusion proteins induced apoptosis and inhibited tumor growth in vivo. (A) Purified Fc-fusion proteins Fc-N16,N17 and Fc-N132 inhibited Notch3 activation compared to control. (B) When HCC2429 xenografts were treated with Fc-fusion protein Fc-N16,17 and Fc-N132, tumor growth was significantly reduced. (C) Tumors resected from mice treated with Fc-N132, Fc-N16,N17, Fc control and PBS. The tumors from Fc-fusion protein-treated animals were significantly smaller than those treated with Fc control and PBS.

Deletion of putative ligand-binding sites abrogated Notch3 activation in vitro

To determine whether EGF-like repeats 7–10 and 21–22 are necessary for signaling, we created constructs N3Δ7-10, N3Δ21-22 and N3Δ7-10, Δ21-22, similar to the full-length receptor but lacking EGF-like domains 7-10, 21-22 or both (Figure 5A). Similar strategies were used in Drosophila to better understand domain functions of Drosophila Notch (11). Unlike native full-length receptor, both N3Δ7-10 and N3Δ21-22 constructs were unable to activate Notch3 cleavage in the presence of Jagged1 (Figure 5B).

Figure 5.

Deletions of EGF-like repeats 7–10 and 21–22 reduces Notch3 activity in vitro. (A) Diagrams of full-length receptors, and mutants with deletions of EGF-like repeats 7–10, 21–22 or both. (B) Co-transfection of Jagged1 and full-length Notch3 into HeLa cells resulted in induction of activated Notch3 (ICD). In contrast, activated Notch3 was absent in cells transfected with Jagged1 and mutants N3Δ7-10 and N3Δ21-22. Full-length Notch3 induced transcription of Hes1 (C) and Hey1 (D) in HeLa cells in the presence of Jagged1. Deletions of EGF-like repeats 7–10, 21–22 or both resulted in decreased level of Hes1 transcription compared to full-length Notch3.

As in many biological systems, modulation of Notch3-dependent genes is context-dependent. In contrast to early finding that Notch3 regulates only Hey1 in lung cancer cell lines, in HeLa cells transcription of both Hes1 and Hey1 was reduced when Notch3 with deletions of EGF-like domains 7–10, 21–22 or both were used (Figure 5C, D), supporting the hypothesis that these regions are important for Notch activation (4).

DISCUSSION

Information regarding the binding site for Notch receptors mostly gleaned from studies using Notch deletion mutants in Drosophila. Of the 36 EGF-like repeats in Drosophila Notch, repeats 11–12 are sufficient and necessary for interaction with both Delta and Serrate (12). Similar observations have been made for mammalian Notch1, in which the loss of calcium binding EGF-like repeats 11, 12 and 13 have been shown to abrogate receptor function (1).

However, there are differences in structure, tissue distribution, and activation of downstream target genes among the Notch receptors (13). Unlike Notch1, Notch3 contains 34 instead of 36 EGF-like repeats. Notch3 also differs from Notch1 and Notch2 by its lack of a transactivation domain (TAD). Structural and functional differences, therefore, can implicate different ligand recognition sites among the four mammalian Notch receptors. In this study, using a Notch3 peptide library we discovered two regions within the Notch3 ECD important for ligand binding. Unlike EGF-like repeats 11–13 in Notch1, our findings suggest that the binding site on Notch3 involves EGF-like repeats 7–10, with the strongest functional activity in EGF-like repeats 7–8. This observation differs from that of Joutel et al., in which the mutation C428S located on EGF-like repeat 11 abrogated ligand binding (14). It is possible that this mutation results in a conformational change in the receptor that prevents ligand binding without being within the ligand-interacting surface of the receptor.

The present data demonstrates two potential ligand binding sites, EGF-like repeats 7–10 and 21–22, within the ECD of Notch3. Either site appeared sufficient for receptor activation. Since similar studies have not been carried out for Notch1, it is not known whether Notch3 is the only mammalian Notch receptor with two functional domains. In Drosophila, deletion of Notch EGF-like repeats 24–26, a part of the genetically-defined Abruptex region, results in reduced signaling by the ligand Serrate but not the ligand Delta (15). This observation suggests that all mammalian Notch receptors may possess two ligand domains and that the second binding site is important for regulating ligand specificity. Further studies are needed to test this hypothesis.

The expression of Notch3 in adult mammals is limited to the vascular system. Embryonic deletion of Notch3 in mice results in vascular smooth vessel defects, suggesting that targeting this pathway in cancer will result in anti-angiogenic effects (10). Given the role of Notch signaling in maintaining stem cells, it is possible that inhibiting Notch3 could also have gut toxicity as observed with γ-secretase inhibitors (11). Unlike Notch1 in T-cell acute lymphoblastic leukemia, oncogenic mutation has not been associated with Notch3 (16, 17). By contrast, the dysregulation of the Notch3 pathway in cancer has mostly been associated with overexpression and gene amplification (3, 18). Thus, interfering with ligand-receptor interaction using peptides and recombinant protein constitutes promising strategies targeting this pathway.

While the effects of the peptides and recombinant proteins may not be specific to Notch3, since other Notch receptors utilize similar ligands, the strategies used in the present study can potentially be developed for clinical use. Furthermore, the identified functional sites within the receptor could also serve as targets for therapeutic antibodies or chemical peptidomimetic screening and production. Our study, therefore, provides not only insights into the mechanism of Notch3 signaling in lung cancer, but also uncovers specific receptor regions whose targeting results in anti-tumor activity and may serve as the basis for designing specific anticancer therapeutics targeting this pathway.