Abstract
Like many signaling pathways in development, the Notch receptor pathway plays an important role in cancer pathobiology when it is dysregulated. Potential ligand-binding sites within the epidermal growth factor (EGF)–like repeats of Notch1 have been identified, but the ligand-binding domains in Notch3, which is implicated in lung cancer, are not known. In screening a library of 155 peptides representing all 34 EGF-like repeats in Notch3, we discovered two distinct ligand-binding regions involving the 7–10 and 21–22 repeats that are distinct from the putative ligand-binding domain of Notch1. In cell-based assays, peptides from these regions induced apoptosis and reduced expression of the Notch3-dependent gene Hey1. They also bound directly to the Notch ligand Jagged1, suggesting that their mechanism of action involves disrupting interactions between Notch3 and Jagged1. Recombinant Fc fusion peptides engineered for in vivo testing showed that the Notch3 peptides defined could trigger apoptosis and suppress tumor growth in tumor xenograft assays. These findings rationalize a mechanistic approach to lung cancer treatment based on Notch3 receptor–targeted therapeutic development. Cancer Res; 70(2); 632–8
- Notch3
- Peptides
- 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 four Notch receptors (Notch1–Notch4) and two families of ligands, Jagged (Jagged1 and Jagged2) and Delta-like (Dll1, Dll3, and Dll4). Binding of the ligand to the extracellular domain (ECD) 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 of the ECD between the ligand and the receptor with subsequent release of the activated ICD. Notch3 is a large protein containing 2,321 amino acids with a predicted molecular mass of 243.66 kDa. The Notch3 ECD, a region containing the ligand recognition site, is estimated to be 210 kDa. 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 seems to involve epidermal growth factor (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 those of other family members.
Notch3 is overexpressed in ∼40% of resected non–small cell lung cancers, 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 cross-talk with oncogenic pathways such as the EGF receptor/ras/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 showed that interfering peptides and recombinant proteins mimicking these regions can abrogate Notch3 activation, induce apoptosis, and inhibit tumor growth in vivo. The findings of the present study not only give novel insights into Notch3 signaling but also establish a foundation on which targeted therapy can be developed.
Materials and Methods
Peptide library
The peptide library consisted of 155 synthetic peptides. Their sequences were 5 to 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 NH2 terminus and peptide 155 representing the COOH terminus of the last EGF-like repeat. They were synthesized by SynPep and diluted in deionized H2O to bring the concentration to ∼10 mg/mL of peptide in 1× PBS. The peptides were biotinylated using E-Z Link Biotin BMCC (Thermo Fisher Scientific, Inc.) in PBS at a molar ratio of approximately 1 to 2 moles of biotin per mole of synthetic peptide for immunofluorescence staining and pull-down assays.
Cell culture and inhibitor
The Notch3-expressing lung cancer cell line HCC2429 was established as previously described (8). HEK293T and HeLa cells were obtained from the American Type Culture Collection and maintained in DMEM with 10% FCS. MRK003 was provided by Merck, Inc. & Co., 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 in 50 μL per well. Twenty-five microliters of Annexin V–Alexa Fluor 680 (Invitrogen, Inc.), diluted 1:2,200 in RPMI 1640 and 10 μL of peptide (diluted to 0.1 mg/mL in RPMI 1640), were added. After an overnight incubation, the treated cells were analyzed with a FMAT 8100 HTS System fluorescent plate reader (Applied Biosystems). 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 was confirmed by DNA sequencing, cloned into pcDNA4, and religated to Notch3 intracellular fragment to generate full-length Notch3 deletion mutants.
In vitro pull-down assay
HEK293T cells were transfected with hemagglutinin (HA)–tagged Jagged1 (provided by Dr. Artavanis-Tsakonas, Harvard Medical School, Boston, MA) using Lipofectamine 2000. The cells were lysed in NP40 buffer [10 mmol/L Tris-HCl (pH 7.5), 150 mmol/L NaCl, 1% NP40 plus 50 mmol/L protease inhibitors]. One microgram of biotin-labeled peptides and streptavidin-conjugated magnetic beads (Promega) 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.) were used.
Immunofluorescent staining assay
HCC2429 and HEK293T cells were plated on glass chamber slides. After 24 h, the cells were rinsed twice in PBS and fixed in 4% paraformaldehyde and treated with 1 mL of biotin-labeled peptides and 0.5 μg/mL of Alexa Fluor 488–labeled streptavidin (Invitrogen). TO-PRO3 (Invitrogen) was used for nuclear staining. The cells were then examined under confocal fluorescence microscopy.
Antibodies
Notch3 and HA-targeted Jagged1 were detected using a rabbit Notch3 antibody (Orbigen, Inc.) and an anti-HA monoclonal antibody (HA-7; Sigma-Aldrich), respectively, at 1:1,000 dilution. The goat anti-human IgG-horseradish peroxidase antibody (Santa Cruz Biotechnology, Inc.) and the mouse anti–β-tubulin monoclonal antibody (AA2; Millipore) at 1:5,000 dilution were used to detect human Fc fusion protein and β-tubulin, respectively.
Fc fusion protein expression
The peptide DNA sequences were cloned into the NH2 terminus of pFUSE-hIgG1-Fc2 and pFUSE-mIgG1-Fc2 vectors (Invivogen). 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 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 h after peptide 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) and quantitated using the iQ5 Multicolor Real-Time PCR detection system (Bio-Rad) and QuantiTect SYBR Green reverse transcription-PCR (RT-PCR) kit (Qiagen). Annealing temperature for PCR was 58°C with the following primers: glyceraldehyde-3-phosphate dehydrogenase (GAPDH), 5′-TGCACCACCAACTGCTTAGC-3′ (sense) and 5′-GGCATGGACTGTGGTCATGAG-3′ (antisense); Hey1, 5′-AGATGACCGTGGATCACCTG-3′ (sense) and 5′-TGTTGAGAGCGAAACCAGTC-3′ (antisense); and Hes1, 5′-AGAAGGCGGACATTCTGGA-3′ (sense) and 5′-GAGTGCGCACCTCGGTATTA-3′ (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 h and maintained in serum-free RPMI 1640. Percent apoptosis was determined using the Annexin V-FITC Apoptosis Detection kit (Calbiochem) and a FACSCalibur Flow Cytometer (Beckman Coulter, Inc.).
In vivo tumorigenicity
HCC2429 cells (1 × 106) suspended in 50% Matrigel (BD Biosciences) were injected s.c. into hind limbs of athymic 4- to 6-wk-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 d. The tumor size was measured every 3 d with a caliper. Tumor volume was calculated with the formula: volume = (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 tests. 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 to 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. 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 (Fig. 1A and B). Interestingly, the locations of these peptides mapped to two discrete regions, EGF-like repeats 7–10 and 21–22 (Fig. 1C). The amino acid sequences from Notch3 repeats 7–10 are most similar to Notch1 EGF-like repeats 8–11 with 79% 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 Fig. 1D.
List of peptides, their location, and effect on apoptosis in HCC2429
A, result from a representative experiment performed on the FMAT 8100 HTS fluorescent plate reader and assayed with Annexin V–Alexa Fluor 680 showing that Notch3 peptides N132, N105, N103, and N102 induced apoptosis. Fifteen of the total 155 peptides induced apoptosis in HCC2429 cells. Top, each peptide was assayed in quadruplicate; bottom, a dose response was noted in which signal intensity was correlated with peptide concentration. B, the peptides with proapoptotic activity mapped to two distinct regions, EGF-like repeats 7–10 and 21–22, within the ECD. Z axis is fluorescence intensity, as a measure of apoptosis. X and Y axes show the location of individual wells on the plate. C, blue background, alignment of Notch3 EGF-like repeat 7 and Notch1 EGF-like repeat 11 showing 40% identity. 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 (top) and N132 (bottom) sequences based on class II EGF domain consensus sequence is 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 a 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 proapoptotic-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 regions are responsible for ligand interaction. Molecular visualizations of the relative positions of N17 and N132 peptides within an EGF-like repeat are shown in Fig. 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 (Fig. 2A). Of the 15 peptides identified, 6 were found to both induce apoptosis and bind to Jagged1. This interaction was subsequently confirmed with the in vitro pull-down studies (Fig. 2B). The peptides also inhibited Hey1 transcription with varying potency (Fig. 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 (Fig. 2D). Of the 15 peptides identified by FMAT 8100 HTS screen, six induced apoptosis and bound to Jagged1.
Notch3 peptides bound to Jagged1 and inhibited transcription of Notch3-dependent gene Hey1. A, fluorescent-labeled Notch3 peptides N16, N17, N102, N103, and N132 (green) bind to HCC2429 cells expressing Jagged1 but not to HEK293T cells that do not express endogenous Jagged1. B, Notch3 peptides inhibited signaling through binding to Jagged1. Immunoprecipitation experiment showing that Jagged1 binds to Notch3 peptides but not to control peptide (C). No, no input. 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 that the 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. *, P < 0.05.
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 bloodstream. To overcome this limitation, we used Notch3 Fc fusion proteins, in which recombinant protein is fused to the Fc domain of human IgG. Fc-N17, Fc-N16,N17, and Fc-N132 Fc fusion proteins reduced activated Notch3 to differing degrees. 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 (Fig. 3A). Interestingly, although not all Fc fusion constructs affected Notch3 activation, they all retained the ability to bind to Jagged1 (Fig. 3B).
Notch3 Fc fusion proteins bind to Jagged1 and inhibit Notch3 activation. A, transfection of Notch3 Fc fusion expression plasmids Fc-N16, Fc-N16,N17, and Fc-N132 into HCC2429 downregulated 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,N17, 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 that Fc fusion proteins bound to Jagged1 but not to Fc control.
Notch3 Fc fusion proteins induce apoptosis and inhibit tumor growth in vivo
Treatment with purified Fc-N16,N17 and Fc-N132 proteins resulted in inhibition of Notch3 activation to levels resemble those obtained with MRK003 (Fig. 4A). This observation confirmed our early peptide data (Fig. 2D). To determine the effect of Notch3 Fc fusion proteins in vivo, we used 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 2-fold reduction compared with Fc control (0.612 cm3; Fig. 4B and C).
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 with control. B, when HCC2429 xenografts were treated with Fc fusion proteins Fc-N16,N17 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. *, P < 0.05.
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 (Fig. 5A). Similar strategies were used in Drosophila to better understand domain functions of Drosophila Notch (11). Unlike the native full-length receptor, both N3Δ7-10 and N3Δ21-22 constructs were unable to activate Notch3 cleavage in the presence of Jagged1 (Fig. 5B).
Deletions of EGF-like repeats 7–10 and 21–22 reduce 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, cotransfection 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 with full-length Notch3. *, P < 0.05.
As in many biological systems, modulation of Notch3-dependent genes is context dependent. In contrast to early findings 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 was used (Fig. 5C and D), supporting the hypothesis that these regions are important for Notch activation (4).
Discussion
Information about 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 has 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. 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 and colleagues (14), in which the mutation C428S located on EGF-like repeat 11 abrogated ligand binding. 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 show two potential ligand-binding sites, EGF-like repeats 7–10 and 21–22, within the ECD of Notch3. Either site seemed sufficient for receptor activation. Because 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 antiangiogenic 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.
Although the effects of the peptides and recombinant proteins may not be specific to Notch3, because other Notch receptors use 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 antitumor activity and may serve as the basis for designing specific anticancer therapeutics targeting this pathway.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Acknowledgments
We thank Jing Hao (Vanderbilt Protein Expression User's Club) and Kimberly Cook (Vanderbilt Monoclonal Antibody Core) for their assistance in Fc fusion protein production and purification, respectively.
Grant Support: National Cancer Institute grant 1R01 CA115707 and National Cancer Institute Lung Specialized Program of Research Excellence grant P50 CA90949.
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.
- Received September 3, 2009.
- Revision received November 2, 2009.
- Accepted November 2, 2009.
- ©2010 American Association for Cancer Research.
References
Volume 70, Issue 2
