From Fly Wings to Targeted Cancer Therapies: A Centennial for Notch Signaling

Abstract

Since Notch phenotypes in Drosophila melanogaster were identified 100 years, Notch signaling has been extensively characterized as a regulator of cell fate decisions in a variety of organisms and tissues. However, in the past 20 years, accumulating evidence has linked alterations in the Notch pathway to tumorigenesis. In this Perspective, we discuss the pro-tumorigenic and tumor suppressive functions of Notch signaling and dissect the molecular mechanisms that underlie these functions in hematopoietic cancers and solid tumors. Finally, we link these mechanisms and observations to possible therapeutic strategies targeting the Notch pathway in human cancers.

This year will be the centennial of the discovery of a signaling pathway that has fascinated developmental, molecular, and cancer biologists around the world. Mutant Notch phenotypes in the fly wing were characterized by John S. Dexter 100 years ago (Dexter, 1914) and, rapidly after, Thomas Hunt Morgan identified the mutant alleles (Morgan, 1917). Almost seven decades later, after the molecular biology revolution, Spyros Artavanis-Tsakonas and Michael Young cloned the Notch receptor and attributed the wing-notching phenotype to gene haplo-insufficiency (Kidd et al., 1986; Wharton et al., 1985). These studies brought a revolution in a large number of fields including developmental and stem cell biology, neuroscience, and – related to this Perspective – cancer biology (Fortini et al., 1993). Indeed, in the early nineties, gain-of-function mutations of the pathway were identified in cancer (Ellisen et al., 1991; Gallahan and Callahan, 1997; Gallahan et al., 1987; Jhappan et al., 1992). A deluge of reports followed, cementing the role of Notch signaling as oncogenic but also tumor suppressive, depending on the context. In this Perspective, we attempt to provide a detailed characterization of Notch functions in both solid and hematopoietic cancers and discuss the molecular mechanisms explaining such functions as well as approaches to target Notch signaling in human cancers.

A brief description of the Notch signaling pathway

There are four Notch receptors (named Notch1–4) in mammals. Notch1 and Notch2 each have 36 EGF-like repeats, while Notch3 and Notch4 have 34 and 29 repeats, respectively, which affect their affinity for corresponding ligands (Haines and Irvine, 2003; Okajima and Irvine, 2002; Rebay et al., 1991) (Figure 1). Notch receptors are single pass type I transmembrane molecules coded by a single precursor that becomes a non-covalently linked heterodimer consisting of an N-terminal extracellular (NEC) fragment and a C-terminal transmembrane-intracellular subunit (NTM) as a result of cleavage by a furin-like protease in the trans-Golgi network (Blaumueller et al., 1997) (Figures 1 and 2). The Notch pathway is normally activated upon interactions with ligands such as Delta-like and Jagged, which are also transmembrane proteins containing EGF-like repeats. In mammals, there are three Delta-like ligands (Dll1, Dll3, and Dll4) and two Jagged ligands (Jag1 and Jag2). The Notch pathway gets activated in a strictly controlled fashion: ADAM10/17 metalloproteases cause an S2 cleavage in the receptor, followed by a third cleavage (S3 cleavage) mediated by the presenilin–γ-secretase complex, composed of presenilin 1 (PSEN1), PSEN2, nicastrin (NCSTN), presenilin enhancer 2 (PEN2), and anterior pharynx-defective 1 (APH1) (Shah et al., 2005). This series of events releases the intracellular portion of the Notch receptor (termed ICN) that then translocates into the nucleus to mediate target gene activation (De Strooper et al., 1999; Schroeter et al., 1998). Notch-ICN is a transcriptional activator (Bray, 2006) consisting of ankyrin repeats, a RAM (RBP-Jκ associated molecule) domain, a transactivation domain (TAD), a nuclear localization signal (NLS), and a PEST domain regulating protein stability (Figures 1 and 2). Notch ligands are also cleaved by γ-secretase and ADAM metalloprotease complexes, thus providing an additional level of regulation of the pathway (LaVoie and Selkoe, 2003; Six et al., 2003). Despite the overall similarities between the receptors, the differences in the ligand-binding extracellular domains and the transactivation intracellular domains lead to distinct ligand affinities and capacity to activate downstream transcription.

Figure 1. Protein structure and mutations of a typical Notch receptor.

The structure of the NOTCH1 receptor and genetic alterations of the protein in representative types of cancer are depicted. ADAM metalloproteases and the γ-secretase complex cleave the receptor and free the ICN domain. Major mutations are clustered according to their effects on protein activity. Both gain-and loss-of-function mutations are shown. The majority of the T-ALL mutations are clustered in the heterodimerization (HD) and PEST domains controlling processing of the receptors by proteases and the stability of the protein correspondingly. Different characteristic cases of hematopoietic disorders (affecting NOTCH2 as well) are shown. In CLL tumors there is an apparent mutational hotspot at the PEST domain of NOTCH1. In the case of SqCC mutations, they are mainly clustered in the EGF repeat region potentially affecting interaction with the ligands. T-ALL: T-cell acute lymphoblastic leukemia, SMZL: splenic marginal zone lymphoma, CLL: chronic lymphocytic leukemia, DLBCL: diffuse large B cell lymphoma, SqCC: squamous cell carcinoma. Percentages are approximations based on current literature.

Figure 2. Overview of the Notch signaling pathway.

A visual description of the signaling cascade is shown for the signal-receiving cell (i.e. the cell expressing the Notch receptor). Pathway inhibitors used include antibodies against NOTCH receptors and DLL ligands, γ-secretase complex inhibitors (GSI), and small peptides inhibiting formation of the transcriptional complex. Antibody-based treatments are shown in purple, GSI compounds in pink, peptide-based drugs in red. Potential epigenetic inhibitors (in green) can include BRD inhibitors like JQ1. HDAC: histone deacetylase, ICN1: intracellular part of NOTCH1, LSD1: lysine specific demethylase 1, SMRT: Silencing-Mediator for Retinoid/Thyroid hormone receptors, GSK3β: glycogen synthase kinase 3 beta, DNMAML1: dominant negative MAML1.

In the nucleus, Notch binds to initially inactive CBF1-Su(H)–LAG1 (CSL) (aka RBP-Jκ) complexes and mediates their conversion to a transcriptional activator followed by the recruitment of the co-activator protein mastermind-like 1 (MAML1) (Figure 2) (Nam et al., 2006; Wilson and Kovall, 2006; Wu et al., 2000). The ankyrin repeats seem to play an important role for MAML1 recruitment. The list of target genes regulated by Notch is very much dependent on cell type and can include genes whose products are involved in fundamental aspects of cell biology, such as cell cycle regulation (Joshi et al., 2009; Lewis et al., 2007), cellular differentiation, and metabolism (Palomero et al., 2006). Common targets of the pathway include the HES and HEY (Iso et al., 2001a; Iso et al., 2001b; Jarriault et al., 1995) families of transcription repressors as well as MYC transcription factor (Palomero et al., 2006; Sharma et al., 2006; Weng et al., 2006). The binding and function of Notch on DNA appears to be a rapid and dynamic process controlled by the kinase CDK8 and the ubiquitin ligase Fbxw7 leading to Notch phosphorylation, ubiquitination, and its subsequent proteasomal degradation (Fryer et al., 2004; Mukherjee et al., 2005; O'Neil et al., 2007; Thompson et al., 2007), which shuts off the pathway (Figure 2).

Various tools have been developed to study the transcriptional activity of the pathway, such as ChIP-seq and ChIP-chip to map Notch1 binding on the genome (Castel et al., 2013; Ntziachristos et al., 2012; Palomero et al., 2006; Wang et al., 2011a) and mouse models that allow efficient tracing of receptor cleavage/activity in many different tissues (Hansson et al., 2006; Liu et al., 2011; Mizutani et al., 2007; Souilhol et al., 2006). Recently, the group of Artavanis-Tsakonas (Fre et al., 2011; Sale et al., 2013) and our laboratory (Oh et al., 2013) have traced Notch pathway activity in vivo by using reporter systems for Notch receptors expression and Hes1 activity by coupling them to fluorescent proteins (Figure 3). As there are several unanswered questions regarding Notch ligand expression, even under physiological conditions, an exciting next step could involve development of fluorescent tools to probe ligand expression together with pathway activation in real-time within a living organism.

Figure 3. In vivo mapping of Notch pathway activity using a Hes1^GFP reporter.

(A) Targeting strategy for the generation of transgenic animals expressing Emerald GFP (emGFP) from the endogenous Hes1 locus. (B) Immuno-fluorescence staining for thymus of the Hes1^GFP mice. DAPI stains DNA (nucleus), VE-cadherin is a vascular endothelial marker and K14 is a marker of thymic medullary cells. (C) Increased levels of Notch pathway help differentiation of thymic T cell progenitors through the DN2/3 CD4⁻8⁻ differentiation stage and the pathway activity is decreased immediately at the DP stage. (D) Activity of the Notch pathway in the mouse bone marrow is detected at the HSC level and is decreased as cells differentiate. Subsequently it is reactivated at the level of a megakaryocytic-erythrocytic progenitor (MEP). HSC: hematopoietic stem cells, MPP: multipotent progenitors, CMP: common myeloid progenitors, CLP: common lymphoid progenitors, MEP: megakaryocyte-erythrocyte progenitor, GMP: granulocyte-macrophage progenitor, DN: double negative (CD4⁻CD8⁻), DP: double positive (CD4⁺CD8⁺), SP: single positive.

NOTCH signaling pathway in cancer

The Notch pathway is genetically altered in a large number of hematopoietic and solid tumors (Figure 1). Intriguingly, these alterations can lead to either activation or repression of the pathway depending on the context and the activation status of other potentially oncogenic pathways (Table 1 and Figure 4). Interestingly, it appears that there are multiple and distinct modes of aberrant regulation of the pathway and its targets in cancer. They include activating and inactivating mutations, receptor/ligand over-expression, epigenetic regulation, and effects of post-translational modifications, most notably receptor and ligand fucosylation (especially O-fucosylation) (Haines and Irvine, 2003; Lei et al., 2003; Okajima et al., 2003) and ubiquitination (Fryer et al., 2004; Thompson et al., 2007). We initially discuss T cell acute leukemia, a disease in which Notch has a well-characterized oncogenic role. Subsequently we present several other cases of hematopoietic and solid tumors where Notch has tumor suppressive or oncogenic roles, along with its potential mechanisms of action and partners.

Table 1.

Oncogenic and tumor suppressive roles of Notch signaling in human cancers ^*

Tumor Type	Oncogene or Tumor Suppressor	Mutations (%) or noteworthy observations *	References *
T-Cell Acute Lymphoblastic Leukemia (T-ALL)	Oncogene	50–60% NOTCH1, 30% FBXW7 Role in cancer initiation and maintenance	(Malyukova et al., 2007; Weng et al., 2004)
Chronic Lymphocytic Leukemia (CLL)	Oncogene	5–12% NOTCH1 Role in cancer initiation and survival	(Fabbri et al., 2011; Puente et al., 2011)
Melanoma	Oncogenic	~50% NOTCH1 overexpression in human samples Possible role in metastasis	(Balint et al., 2005; Bedogni et al., 2008)
Cholangiocarcinoma (CCC)	Oncogenic	35% FBXW7 Notch1 promotes tumor initiation and maintenance	(Akhoondi et al., 2007; Zender et al., 2013)
Colorectal cancer	Oncogenic	8–9% FBXW7 Crosstalk with Wnt and Hippo signaling	(Miyaki et al., 2009) (Akhoondi et al., 2007)
Lung adenocarcinoma	Oncogenic	10% NOTCH1 Role in initiation and maintenance (Notch1), and metastasis (Jagged2) Specific role for Notch3 in tumor propagation	(Licciulli et al., 2013; Westhoff et al., 2009; Zheng et al., 2013)
Glioblastoma	Oncogenic	Role in tumor propagation and radioresistance	(Chu et al., 2013) (Wang et al., 2010)
Renal Cell Carcinoma	Oncogenic	Role in progression and maintenance	(Sjolund et al., 2008)
Ovarian cancer	Oncogenic	Role in maintenance and therapy response	(2011; Cancer Genome Atlas Research, 2011; McAuliffe et al., 2012)
Prostate	Oncogenic	Activation of the pathway associated with tumor progression, metastasis, and recurrence	(Marignol et al., 2013; Santagata et al., 2004)
Breast cancer	Mostly Oncogenic	NOTCH1 and NOTCH4 fusions Potential NOTCH2 dominant-negative truncated mutant Other alterations activating Notch signaling But hyperactive Notch signaling may inhibit cancer growth	(Fu et al., 2010; Imatani and Callahan, 2000; Jhappan et al., 1992)
Pancreatic Ductal Adenocarcinoma (PDAC)	Mostly Oncogenic	Notch2 loss inhibits progression and maintenance Overexpression of ligands –Jagged2 (90%), Dll4 (50%) But Notch1 loss may promote tumor initiation	(Hanlon et al., 2010; Mazur et al., 2010; Mullendore et al., 2009)
Cervical cancer	Mostly Oncogenic	Pathway activation in human tumors, but dose-dependent effects Possible role in tumor-propagating cells	(Bajaj et al., 2011; Maliekal et al., 2008; Zagouras et al., 1995)
Head and neck squamous cell carcinomas (HNSCC)	Mostly oncogenic	Possible bimodal pattern of Notch pathway alterations with a small subset of tumors with inactivating NOTCH1 mutations but a larger group with pathway activation	(Sun et al., 2013)
Hepatocellular carcinoma (HCC)	Oncogenic and Tumor Suppressive	Context-dependent effects that may be related to various molecular subtypes	(Qi et al., 2003; Villanueva et al., 2012)
Medulloblastoma	Oncogenic and tumor suppressive	Opposite roles for Notch1 and Notch2	(Fan et al., 2004)
B-Cell Acute Lymphoblastic Leukemia (B-ALL)	Tumor Suppressive	No mutations Role in maintenance (activation induces growth arrest and death)	(Zweidler-McKay et al., 2005)
Acute Myeloid Leukemia (AML)	Tumor Suppressive	Notch1 and 2 expressed but the pathway is not active Role in cancer initiation and maintenance	(Kannan et al., 2013; Lobry et al., 2013)
Small Cell Lung Carcinoma (SCLC)	Tumor Suppressive	No mutations Inhibits tumor maintenance (possible similar role in other neuroendocrine tumor types)	(Sriuranpong et al., 2001)
Lung Squamous Cell Carcinoma (SqCC)	Tumor Suppressor	5–12.5% NOTCH1, NOTCH2	(2012; Cancer Genome Atlas Research, 2012; Wang et al., 2011b)
Cutaneous Squamous Cell Carcinoma (SqCC)	Tumor Suppressor	60–75% NOTCH1, NOTCH2	(Wang et al., 2011b)
Chronic myelo-monocytic leukemia (CMML)	Tumor Suppressor	12% various pathway genes (NCSTN, APH1, MAML1, NOTCH2) Role in cancer initiation	(Klinakis et al., 2011)

^*Cancers indicated in this table have been selected for historical reasons (first examples of mutations in the Notch pathway), because they affect large populations of cancer patients, or because of the particular insight of some studies to the role of Notch signaling in cancer. Among the selected tumor types, selected observations and references are shown, see text for additional references and details. In particular, data from large cancer genomes efforts indicate that many alterations in the extended Notch pathway exist in human tumors, most of these alterations are awaiting additional analyses and functional validation.

Figure 4. Simplified scheme of Notch interactions with other signaling pathways in cancer.

The TGFβ, PI3K, NFκB, and WNT pathways are some of the most important pathways interacting with NOTCH. Notably, Jagged 1 is activated by TGFβ pathway and in turn activates NOTCH receptors in neighboring cells. Phosphorylation of NOTCH from the WNT-induced GSK3β leads to ubiquitination through FBXW7 and final degradation. Also, a classical NOTCH target, HES1, represses PTEN, a competitor of another pathway with oncogenic roles, PI3K, which in turn activates NFκB, a pathway important for leukemia progression. Important parameters of the interactions, such as regulation of NFκB pathway by NOTCH through HES1 action, or the interaction of NOTCH with the WNT member DVL (Dishevelled) protein that inhibits both WNT and NOTCH pathways are not shown in this figure. ICJ1: intracellular part of JAG1, ECJ1: extracellular part of JAG1.

T-Cell Acute Lymphoblastic Leukemia (T-ALL)

NOTCH1 is a master transcription factor that controls innate and adaptive immunity and plays an important role in directing hematopoietic development towards T cells (Aifantis et al., 2008; Li and von Boehmer, 2011; Radtke et al., 2013). The very first finding of Notch pathway alterations in cancer comes from the work of Ellisen and colleagues, which was confirmed subsequently by other groups, that revealed a rearrangement between the intracellular part of NOTCH1 (ICN1) and the T cell receptor beta (TRB) locus, leading to high level expression of truncated, constitutively-active NOTCH1 in leukemia (Ellisen et al., 1991). In vitro studies (Capobianco et al., 1997), as well as animal modeling (Girard et al., 1996; Pear et al., 1996) then revealed that ICN1 is a strong oncogenic allele. Most importantly, ten years ago, the Aster and Look laboratories reported the first activating NOTCH1 mutations in human T-ALL, occurring in approximately 50% of all cases (Weng et al., 2004).

The majority of these mutations encompass single amino acid substitutions, insertions and deletions located in exons 26 and 27 of the genetic locus, which encode the N-terminal and C-terminal components of the hetero-dimerization domain respectively. These mutations lead to lower protection of S2 cleavage of Notch, resulting in either ligand independent activation or hypersensitivity of the pathway to ligands. Another rare group of mutations, the Juxtamembrane Expansion Mutants (JME) also augments NOTCH1 activation at the cell membrane (Sulis et al., 2008) (Figure 1). Finally, PEST domain mutants encompass another category of NOTCH1 mutations in 20–25% of T-ALLs. PEST domain alterations lead to truncation or loss of the domain due to frame-shift or nonsense nucleotide substitutions, which impair proteasomal degradation mediated by the ubiquitin ligase FBXW7 and lead to higher ICN1 cellular concentrations (Weng et al., 2004). The importance of ICN1 degradation in physiology becomes more evident by the fact that 15 % of T-ALL cases harbor mutations or deletions in FBXW7 (Asnafi et al., 2009; O'Neil et al., 2007; Thompson et al., 2007). These changes are localized in three arginine residues critical for its interaction with ICN1. Mutations in the PEST domain and FBXW7 do not occur concurrently which implies that they play the same role to increase stability of ICN1 (Asnafi et al., 2009; O'Neil et al., 2007).

The fact that FBXW7 mutations directly affect cells with leukemia-initiating (LIC) properties though the stabilization and overexpression of MYC, another well-characterized substrate of this ubiquitin ligase, further demonstrates that NOTCH and MYC actions are intertwined in cancer cells (King et al., 2013). Interestingly, NOTCH1 mutations in T-ALL were shown to have a favorable prognosis and better outcome post-treatment in a number of studies including the ALL-Berlin-Frankfurt-Munster 2000 study (Breit et al., 2006), a study by the Japan Association of Childhood Leukemia Study that examined NOTCH1 and FBXW7 mutational status in T-ALL and T-cell lymphoblastic lymphoma patients (Park et al., 2009), and the Lymphoblastic Acute Leukemia in Adults (LALA)-94 and the GRAALL-2003 trials (Asnafi et al., 2009). Finally, another report on 134 pediatric patients from the EORTC-CLG 58881 and 58951 protocols concluded that NOTCH1 and FBXW7 mutations associate with improved early chemotherapeutic response and lower minimal residual disease (MRD) levels (Clappier et al., 2010). It remains to be seen whether Notch pathway inhibition will be used successfully to target T-ALL, especially in relapsed disease that is refractory to conventional chemotherapy-based treatments.

Chronic lymphocytic leukemia (CLL)

CLL is the most common leukemia in adults. Recently, it was demonstrated using next generation sequencing-based approaches that 10–12% of CLL cases exhibit activating mutations of NOTCH1, underlining the significance of such mutations as a prognostic marker. The vast majority of these mutations are in the PEST domain, leading to truncated protein variants with a longer half-life (Figure 1). Interestingly, there seems to be a mutational hotspot in this disease, with P2515Rfs being the most prevalent mutation (Fabbri et al., 2011; Puente et al., 2011; Rossi et al., 2012a). Mutations of NOTCH1 are mutually exclusive with TP53 abnormalities and survival outcomes are poor in both cases (Rossi et al., 2012a; Wickremasinghe et al., 2011). NOTCH1 and SF3B1 (a splicing factor) mutations were associated with decreased overall survival, and both retained independent prognostic significance for survival outcomes (Oscier et al., 2013). Mutational activation of NOTCH1 was observed at significantly higher frequency during disease progression towards the high-risk Richter transformation (30%) and chemo-refractory CLL (20%) (Fabbri et al., 2011). This later study and a very recent large-scale clinical analysis of CLL patients (Weissmann et al., 2013) confirmed that NOTCH1 mutations are an adverse prognostic parameter in this disease.

Lymphoma

Non-Hodgkin lymphoma (NHL) is a heterotypic mix of diseases, the most prevalent amongst them consist of Burkitt lymphoma, Follicular lymphoma (FL, the most indolent amongst NHL cases) (Pasqualucci et al., 2014), (Roulland et al., 2011), and diffuse large B cell lymphoma (DLBCL). Burkitt lymphoma, mainly characterized by the upregulation of MYC due to its translocation to the immunoglobulin locus, display recurrent gain-of-function NOTCH1 mutations in 8–9% of patients (Love et al., 2012). FL and DLBCL are malignancies of B cell origin and together comprise 60% of new NHL diagnoses in North America. FL and the germinal center B-cell (GCB) DLBCL subtype are derived from germinal center B cells, whereas the more aggressive activated B-cell (ABC) DLBCL subtype is most likely derived from cells that have exited the germinal center. NOTCH2 is mutated in ~8% of DLBCL cases (Lee et al., 2009). These are mainly gain-of-function mutations affecting the PEST domain (and thus the stability of the protein) as well as copy number alterations (Morin et al., 2011). Interestingly, NOTCH2 is required for B-cell development in the spleen marginal zone environment and has been implicated in splenic marginal zone lymphoma (SMZL) (Kiel et al., 2012; Rossi et al., 2012b), as 20% of SMZL cases exhibit gain-of-function NOTCH2 mutations accompanied by mutations of NOTCH1, SPEN and DTX1 (Rossi et al., 2012b). It was suggested that these genetic changes are associated with adverse prognosis (Kiel et al., 2012). Finally, Jundt and colleagues have characterized an activating role for NOTCH1 in classic Hodgkin lymphoma (Schwarzer et al., 2012; Schwarzer and Jundt, 2011). These authors suggested that NOTCH1 is activated through the upregulation of its ligands within the tumor niche and suppresses genes important for B cell identity, such as E12/E47 and the early B cell factor (EBF) (Jundt et al., 2008). Additional studies are required to better define Notch receptor and ligand expression, targeted signaling pathways in the distinct subtypes of lymphoma.

Acute myeloid leukemia and myelo-monocytic neoplasms

Several years ago, emerging evidence indicated that the Notch pathway could have tumor suppressive roles in various types of tumors, in stark contrast to its oncogenic role in the aforementioned hematopoietic malignancies (Nicolas et al., 2003; Rangarajan et al., 2001). In contrast to the tumorigenic role of NOTCH1 in T-ALL, our laboratory and others have recently characterized a tumor suppressive role of the Notch pathway in myeloid malignancies. We have shown that deletion of nicastrin (Ncstn), an essential component of the γ-secretase complex, leads to the induction of chronic myelomonocytic leukemia (CMML) (Klinakis et al., 2011), a disease characterized by increased extramedullary hematopoiesis, monocytosis, myeloproliferation, and frequent progression to acute myeloid leukemia (AML). This is a Notch-mediated effect, as compound deletion of Notch1/2 in vivo led to similar effects. This was further attested when analysis of the conditional model for the deletion of FX (the homologue of human GDP-L-fucose synthase) or O-fucosyltransferase 1 (Pofut1) showed myeloid hyperplasia (Yao et al., 2011), underlining the importance of Notch receptor fucosylation for ligand binding and pathway activation. Ablation of MAML1 can lead to similar phenotypes (Chen et al., 2008). Mechanistically, the tumor suppressor role of NOTCH in this disease is mediated by direct repression of the PU.1 and CEBPα promoters by HES1. Subsequent screening of primary CMML samples for Notch pathway mutations showed that NCSTN, Mastermind-like 1 (MAML1), APH1A, and NOTCH2 are mutated and genetically inactivated in about 12% of CMML patients. These mutations are unique to CMML and not found in other myeloproliferative disorders such as polycythemia vera (PV) and myelofibrosis (MF). Notch inactivating mutations co-occurred with other described myeloid mutations in genes such as TET2, FLT3, and ASXL1 (Klinakis et al., 2011). Based on these findings we were able to show that combination of Notch pathway and TET2 inactivation leads to acute myeloid leukemia (AML). AML cells specifically express NOTCH2 on their surface but show no signs of pathway activity. Interestingly, re-activation of the Notch pathway in established AML leads to complete disease remission (Kannan et al., 2013; Lobry et al., 2013). This observation provides a rationale for the use of specific NOTCH2 activating antibodies or specific agonists as a viable therapeutic strategy in this type of leukemia. Mechanistically, there might be several ways to suppress Notch pathway activity in AML. Initially, AML cells might reside in microenvironments that lack Notch ligands. Another putative mechanism is epigenetic silencing, achieved by DNA and histone methylation of target gene promoters/transcriptional start sites. In agreement with this possibility, we found that Notch target genes are characterized by H3K27me3 marks (Lobry et al., 2013) and mice carrying the R132H mutation of isocitrate dehydrogenase 1 (IDH1) (Figueroa et al., 2010; Gross et al., 2010; Xu et al., 2011) develop a myeloproliferative disease characterized by marked DNA hyper-methylation of Notch pathway genes such as Lfng, Maml3, and Hes5 (Sasaki et al., 2012).

Acute B cell leukemia (B-ALL)

Interestingly, Notch signaling also appears to act as a tumor suppressor in B cell ALL (B-ALL) (Zweidler-McKay et al., 2005). In agreement with the AML findings, Notch pathway re-activation leads to growth inhibition and induces apoptosis in human B-ALL cells. In a recent follow-up publication, it was shown that several Notch pathway targets in B-ALL are suppressed by DNA cytosine hyper-methylation on their promoters followed by histone H3K27 and H3K9 trimethylation (Kuang et al., 2013). The parallel between AML and B-ALL is intriguing and can potentially be explained by a recent Notch activity mapping effort (Oh et al., 2013) that demonstrated activity of the pathway in T cell progenitors and pre-erythrocytes and a lack of pathway activation in the B cell and myelo-monocytic lineages. These findings provide support for a key role for NOTCH as a developmental regulator that can determine the fate of progenitors in the hematopoietic system. In this model, NOTCH action needs the addition of other oncogenic stimuli to transform cells. In agreement with this idea, we found that Notch pathway inactivation can lead to increased frequency of granulocyte-monocyte progenitors (GMP), cells that can initiate diseases like CMML and AML upon further alterations (Klinakis et al., 2011).

Notch signaling in solid tumors

A number of recent reviews have thoroughly summarized our knowledge on Notch signaling in solid tumors (Nowell and Radtke, 2013; Ranganathan et al., 2011; South et al., 2012) (see also Table 1). Recent genomic data, including resources from The Cancer Genome Atlas (TCGA), underscore the prevalence and the complexity of Notch pathway alterations in human cancers, although they do not currently provide detailed functional interpretations for these alterations. Our goal in the section below is not to provide an exhaustive list of the solid tumor types in which Notch signaling is altered and the possible consequences of these alterations. Rather we aim to highlight some key observations in a few prominent tumor types and draw some points of discussion from these studies, including the plethora of partners used by Notch (Figure 4) and the distinct roles of the four Notch receptors.

Breast cancer

Breast cancer is a very prevalent form of cancer in which the Notch pathway may act as a tumor suppressor or an oncogene depending on the subtype. One of the first indications that Notch signaling may play a role in solid tumors actually came from experiments with mouse mammary tumor viruses (MMTV). Integration of the MMTV genome next to the “Int-3” locus resulted in an activating mutation of Notch4, leading to the constitutive activation of the receptor and breast cancer development (Gallahan and Callahan, 1997; Jhappan et al., 1992; Robbins et al., 1992). Since this seminal discovery, a number of studies have confirmed that activation of Notch signaling plays an oncogenic role in breast cancer (Colaluca et al., 2008; Pece et al., 2004; Robinson et al., 2011; Xu et al., 2012). In breast cancer cells, Notch signaling can be activated by functional interactions with other signaling pathways, including the Ras and the Wnt pathways (Ayyanan et al., 2006; Fitzgerald et al., 2000; Izrailit et al., 2013; Meurette et al., 2009; Weijzen et al., 2002). Recent observations indicate that Notch4 may play a more specific role compared to other Notch receptors in breast cancer stem cells (Harrison et al., 2010). In contrast, a recent study indicates that hyperactivation of NOTCH3 may actually be detrimental to breast cancer cells by inducing senescence (Cui et al., 2013). Interestingly, mammary epithelial cells respond differently to different levels of activation of the Notch pathway (Mazzone et al., 2010). Thus, while accumulating evidence indicates that Notch is pro-tumorigenic in breast cancer, in certain contexts, specific (high) levels of activation may be tumor suppressive; alternatively, different Notch receptors may have unique signaling outputs in mammary epithelial cells or in different subtypes of breast cancer. Once more, the notion of a “differentiation switch” could explain the many faces of Notch signaling in this type of tumor.

Lung cancer

Lung adenocarcinoma (LAC) is a major subtype of lung cancer. Initial observations suggested that Notch signaling promotes the expansion of LAC cells in culture (Dang et al., 2003; Eliasz et al., 2010; Haruki et al., 2005). More recent in vivo studies demonstrate that Notch signaling is a key promoter of LAC development and maintenance (Allen et al., 2011; Licciulli et al., 2013; Maraver et al., 2012) and that NOTCH3 plays a unique role in the self-renewal of LAC tumor-propagating cells (Zheng et al., 2013). Expression of JAG2 at the surface of lung adenocarcinoma cells leads to homotypic interactions with Notch receptors and promotes the metastatic potential of these LAC stem cells (Yang et al., 2011). Thus, while mutations and other alterations may not be frequent in LAC (Westhoff et al., 2009), Notch pathway activity correlates significantly with worse survival in lung cancer patients (Hassan et al., 2013; Zheng et al., 2013) and activation of Notch may be important for the sustained growth of LAC. Targeting NOTCH3 and/or JAG2 may benefit a very large number of lung cancer patients worldwide.

Squamous cell lung carcinoma (SqCC) is the second major type of non-small cell lung cancer. In stark contrast to LAC, Notch signaling is thought to be a tumor suppressor of SqCC development, as evidenced by the identification of loss-of-function mutations in human tumors (Wang et al., 2011b). These mutations mainly cluster in the EGF-like repeat region of NOTCH1 and thus have the potential to disrupt ligand binding or to produce truncated receptors (Figure 1). While functional validation for these observations is still missing owing to the current lack of appropriate mouse models, numerous observations indicate that inactivation of Notch signaling promotes the development of squamous cell carcinoma in other tissues, including in cutaneous and head-and-neck tumors (Agrawal et al., 2012; Pickering et al., 2013; Proweller et al., 2006; Rothenberg and Ellisen, 2012; Wang et al., 2011b). These observations suggest that loss of Notch pathway activity may be critical for the growth of tumor cells with squamous differentiation characteristics.

Small cell lung carcinoma (SCLC) is a neuroendocrine subtype of lung cancer, representing a smaller fraction of lung cancer cases (~12–15%) but with the highest mortality rate. Genomic studies have failed to identify recurrent mutations in the Notch pathway in SCLC (Peifer et al., 2012; Rudin et al., 2012). However, early observations indicated that hyperactivation of Notch signaling blocks the cell cycle of SCLC cells (Sriuranpong et al., 2001; Sriuranpong et al., 2002). A tumor suppressive role for Notch in SCLC is supported by evidence that Notch may play a similar role in other neuroendocrine tumors such as medullary thyroid carcinoma (Cook et al., 2010). Thus far, however, no functional evidence has been obtained in vivo that activation of Notch may block SCLC development or maintenance, and it is still possible that subpopulations of cells in SCLC tumors may display some Notch activity and contribute to SCLC growth (Kluk et al., 2013; Salcido et al., 2010).

Thus, three different subtypes of lung cancer display strikingly different roles for Notch signaling in cancer development, from an active oncogenic role with rare genetic alterations in LAC to tumor suppressor with inactivating mutations in SqCC and then possibly tumor suppressive with no sign of mutations in SCLC. It is possible that these differences are related to the role of Notch in cell fate decisions during lung embryonic development.

Liver cancer

Genome sequencing analyses did not reveal recurrent mutations in Notch pathway genes in hepatocellular carcinoma (HCC), a leading cause of cancer-related deaths worldwide (Fujimoto et al., 2012; Guichard et al., 2012). Nevertheless, Notch signaling has been of interest to liver cancer biologists because of the prominent role of Notch signaling in liver development, including mutations in NOTCH2 or JAG1 in patients with Alagille syndrome (syndromic bile duct paucity) (McDaniell et al., 2006; Oda et al., 1997). Haploinsufficient mutations in a specific ligand and a specific receptor in the Notch pathway would suggest that Notch signaling could play very context-dependent and level-dependent roles in liver tumors. Initial observations suggested that low levels of Notch correlates with high activity of the Wnt pathway, a major oncogenic pathway in HCC (Wang et al., 2009). Also, high levels of active Notch1 may inhibit the expansion of HCC cells (Qi et al., 2003), and deletion of Notch1 in the liver of mice results in hyperproliferative hepatocytes, suggesting a tumor suppressive role for Notch in HCC (Croquelois et al., 2005). Similarly, Notch signaling has a tumor suppressive effect in HCC initiated by inactivation of the RB pathway (Viatour et al., 2011). However, other reports have more recently provided evidence that Notch signaling is active and oncogenic in HCC (Dill et al., 2013; Tschaharganeh et al., 2013; Villanueva et al., 2012), and possibly important for the development of tumors following hepatitis B virus infection (Jeliazkova et al., 2013). These observations suggest that the role of Notch signaling in HCC may be different in the distinct molecular subgroups of this cancer type and underscore the need to further explore the molecular contexts associated with tumor suppressive or oncogenic roles of Notch in the liver.

In contrast to the complex roles of Notch signaling in HCC, accumulating evidence supports a pro-tumorigenic role for Notch signaling in cholangiocarcinoma (CCC). Mutations of the Notch repressor FBXW7 are found in a subset of human tumors (Akhoondi et al., 2007). Similar to the disruption of bile ducts in Alagille patients, activation of Notch2 in liver progenitors and adult hepatocytes promotes biliary tubulogenesis (Jeliazkova et al., 2013). Finally, constitutive activation of NOTCH1 is sufficient to initiate CCC development in mice (Zender et al., 2013).

It is likely that the sometimes contradictory consequences of Notch activation in liver cells are due to a combination of the strength of the downstream signal, the timing of the activation, the cell type in which this activation occurs, and the receptor involved (Ortica et al., 2013). There seems to be a consensus that higher Notch levels in liver progenitors favors bile duct differentiation versus hepatocytic differentiation. Possibly activation of Notch (e.g. NOTCH2) in these progenitors promotes CCC while suppressing HCC (Guest et al., 2013). It is also possible that Notch switches from a suppressive role in the early stages of HCC development to a more oncogenic role. Although it was proposed that Notch signaling plays a role in liver cancer invasion and metastasis (Lim et al., 2011; Zhou et al., 2013), more work is required to further support this notion.

Colorectal cancer

The intestinal epithelium possesses an unprecedented self-renewal rate that appears to be linked to a high susceptibility to malignant transformation. Notch signaling has been known for many years now to be involved in both the control of homeostatic self-renewal in stem cell populations and the development of colorectal cancer (CRC) (Fre et al., 2005; Radtke and Clevers, 2005; van Es et al., 2005). While mutations in NOTCH genes are rare, Notch signaling is overexpressed or constitutively activated in CRC in part because of mutations in regulators of Notch signaling, including in FBXW7 (although FBXW7 clearly controls other cellular pathways beyond Notch) (Akhoondi et al., 2007; Babaei-Jadidi et al., 2011; Camps et al., 2013; Miyaki et al., 2009; Sancho et al., 2010; Zhu et al., 2013). In addition, Notch activation has been linked to activation of Wnt signaling and Hippo/YAP signaling in CRC cells, although the various levels of crosstalk between these pathways are still not fully understood (Camargo et al., 2007; Fre et al., 2009; Kim et al., 2012; Kwon et al., 2011; Peignon et al., 2011; Rodilla et al., 2009; Tschaharganeh et al., 2013). In particular, Jagged1, expressed on tumor cells themselves or produced from endothelial cells, is thought to be a key ligand for Notch activation in CRC cells (Lu et al., 2013; Rodilla et al., 2009; Tschaharganeh et al., 2013). Another Notch ligand, DLL4, plays a non-cell autonomous role in CRC development in large part by controlling the development of blood vessels necessary for tumor growth (Fischer et al., 2011; Ridgway et al., 2006). Expression of miR-34a in CRC stem cells may help control Notch output and generate a bimodal Notch response (Bu et al., 2013). Finally, Notch signaling may play a crucial role not only in the early stages of CRC development by controlling the fate of stem cells and cancer stem cells but also at the later stages of tumor invasion and metastasis (Sonoshita et al., 2011).

Pancreatic cancer

The major and most lethal type of pancreatic cancer is pancreatic ductal adenocarcinoma (PDAC). An early study detected evidence of Notch pathway activation in PDAC and showed that Notch lies downstream of TGFβ during ductal metaplasia, an early stage of PDAC development (Miyamoto et al., 2003). Mouse genetics studies have demonstrated that activation of Notch signaling cooperates with oncogenic K-Ras to promote both initiation and dysplastic progression from acinar cells by inducing their rapid reprogramming to a duct-like phenotype (De La et al., 2008). Indeed, pharmacological inhibition of Notch signaling slows the progression of the disease in mutant mice and prevents the expansion of some human PDAC cell lines (Cook et al., 2012; Mizuma et al., 2012; Plentz et al., 2009), possibly in part because of an inhibition of PDAC stem cells (Bailey et al., 2013). Genetic inactivation of Notch2, but not Notch1 (Avila et al., 2012; Mazur et al., 2010), inhibits PDAC development initiated by oncogenic K-Ras. In fact, loss of Notch1 function may even promote PDAC development, although the basis of this observation remains unknown (Hanlon et al., 2010).

Melanoma

The Notch pathway has been found to be active in melanoma (Asnaghi et al., 2012). NOTCH1 appears to promote disease progression (Rangarajan et al., 2001; Zhang et al., 2012) and growth of melanocytes under hypoxic conditions (Bedogni et al., 2008). There are no documented gain-of-function mutations affecting the pathway in this disease (Hodis et al., 2012), suggesting that the pathway might be affected through transcriptional and epigenetic control possibly through contrasting actions of BRN2, a possible activator, and MITF, which acts as a repressor of the Notch pathway (Thurber et al., 2011). Whatever the mechanism of activation, recent pre-clinical studies have reported that γ-secretase inhibitors (GSI) can reduce the tumor initiating potential and suggested that GSI combination with chemotherapy could be a useful new therapeutic approach in melanoma (Huynh et al., 2011).

In conclusion, Notch signaling plays distinct roles in different types of tumors, both solid and liquid (hematopoietic). The presented list is by no means exhaustive but provides us with an extensive overview of Notch signaling in cancer. Most of these studies are relatively recent emphasizing the increased interest in the study of Notch signaling in cancer during the last decade. Mechanistically, more work is required to pinpoint specific molecular pathways and gene targets in each tumor type, but emerging technologies and most notably DNA and RNA next generation sequencing-based approaches will continue to help us further dissect the role of this pathway in tumor initiation and progression.

A perspective on two decades of Notch-centered cancer research: Remaining intriguing questions

This brief overview of some of the most common and/or lethal human cancers, both hematopoietic and solid, highlight several key aspects of Notch signaling in cancer development that hold true in other tumors in which Notch signaling is also altered, including myeloma, prostate, ovarian, skin, and brain cancers. Obviously, there are several outstanding questions that have to be addressed to not only help us better understand pathway function in cancer but also enable more efficient therapeutic targeting (see below). An initial question is whether Notch pathway mutations are tumor-initiating or tumor-propagating. Most likely, both types of mutations can be described, depending on the tumor type. We discussed an intriguing example in myeloid neoplasms where Notch signaling loss of activity seems to expand the frequency of leukemia-initiating cells (LIC) but requires secondary mutational events to lead to full-blown disease (Klinakis et al., 2011; Lobry et al., 2013). A similar scenario might play out in BALL, as it was shown that Notch activity directs lymphocyte progenitors exclusively to the T cell lineage, at the expense of B cell differentiation (Pui et al., 1999; Radtke et al., 1999). On the other hand, it is intriguing to ask whether NOTCH1 activating mutations in T-ALL occur to simply define lineage, by locking cells in a specific differentiation status (T cell in this case), or to truly transform the cells. Further studies that can genetically sequence both leukemia and normal stem cell/progenitor populations, preferably at the single cell level, might address such questions.

One particularly interesting aspect of Notch signaling in cancer progression that has been emerging in the last few years is its potential impact on metastasis, which may be linked to the role of Notch in cancer stem cells (see (Giancotti, 2013) for a recent discussion). Early studies had identified JAG1 expression as a marker of metastatic prostate cancer (Santagata et al., 2004) and found a role for Notch in the control of epithelial-mesenchymal transitions (EMT) (Timmerman et al., 2004). Indeed, JAG1 expression on tumor cells may help promote the spread of breast cancer cells to the bone microenvironment by activating Notch signaling in bone cells (Sethi et al., 2011). Activation of Notch during EMT and metastasis may be under the control of the miR-200 microRNA (Brabletz et al., 2011; Yang et al., 2011). This pro-metastatic function of Notch signaling may be promoted by its crosstalk with the machinery responding to hypoxic environments (Sahlgren et al., 2008; Yeung et al., 2011). Furthermore, an increasing number of studies connect Notch signaling to molecules and pathways involved in tumor invasion and metastatic growth, including Tenascin C (Oskarsson et al., 2011) and regulators of polarity (McCaffrey et al., 2012) in breast cancer.

As discussed above, Notch signaling in tumor cells has been involved in various aspects of angiogenesis in multiple studies, especially via the DLL4 and JAG1 ligands (Benedito et al., 2009; Li and Harris, 2005; Phng and Gerhardt, 2009; Zeng et al., 2005). In particular, the Notch ligand DLL4 is upregulated in the angiogenic vasculature in response to VEGF and blockade of DLL4 was shown to lead to markedly increased non-productive tumor vascularity, which inhibits tumor growth (Noguera-Troise et al., 2006; Ridgway et al., 2006). While these observations seemed promising clinically (Hoey et al., 2009), long-term blockade of DLL4 leads to the development of vascular neoplasms (Yan et al., 2010), potentially limiting the therapeutic potential of DLL4-blocking strategies. Thus, activation of Notch signaling may contribute to tumor spread via multiple mechanisms, including by maintaining the self-renewal of cancer stem cells, by directly contributing to the cellular processes involved in tumor invasion (e.g. EMT and response to hypoxia) (Wang et al., 2011c), by controlling neo-vascularization, as well as by playing a key role in the metastatic niche. Such issues could be more important in solid tumors than leukemia, however, it is intriguing to define Notch-ligand expressing niches in different types of hematopoietic tumors and test whether ligand expression is important for leukemia cell homing to different tissues and response to drug treatments. For example, targeting the expression or function of a specific ligand could affect NOTCH1-expressing T-ALL homing and metastasis. As most cancer patients die from metastatic disease, it will be important in the near future to continue to investigate the molecular and cellular basis of tumor spread in connection with Notch signaling.

Therapeutic targeting of the Notch pathway in tumors

As proteolytic cleavage of NOTCH receptors by the presenilin/γ-secretase complex is a prerequisite for the activation of signaling (in the absence of downstream activating mutations), small molecule GSI efficiently blocks NOTCH1 activity in T-ALL cells and has been proposed as a molecular targeted therapy for the treatment of this disease (Aster and Blacklow, 2012; Palomero and Ferrando, 2009). However, animal studies have shown that systemic inhibition of NOTCH signaling results in “on-target” gastrointestinal toxicity because of the accumulation of secretory goblet cells in the intestine due to alterations in the differentiation of intestinal stem cells following Notch inactivation. Phase 1 clinical trials further confirmed these treatment side effects. As a result, inhibition of the pathway using GSI alone may not be the most viable therapeutic choice in the future. An alternative to the use of single GSI treatment is the combinatorial use of glucocorticoids and GSI, where glucocorticoids ameliorate the GSI-induced gut toxicity by inducing the expression of Cyclin D2, protecting the animals from developing intestinal goblet cell metaplasia (Real et al., 2009).

Notch signaling targeting is, however, not restricted to the usage of GSI. Alpha-secretase inhibitors (ASI) against the ADAM10/17 metalloproteases that mediate receptor S2 cleavage are available (Zhou et al., 2006) and are currently being tested (Purow, 2012). Furthermore, using phage display technology, pharmaceutical companies have generated highly specialized antibodies against NOTCH1 and NOTCH2 that act mainly through stabilization of the negative regulatory region (NRR) of the receptors, and the protection from proteolytic cleavage, thus inhibiting the production of ICN1/2 (Wu et al., 2010). These antibodies lead to lower levels of gastrointestinal toxicity and other side effects emanating from pan-Notch pathway inhibition achieved by GSI. Selective blocking of NOTCH1 inhibits tumor growth in pre-clinical models through at least two mechanisms: inhibition of cancer cell growth and deregulation of angiogenesis. Soluble extracellular fractions of Notch receptors and ligands can also act as decoys and inhibit the pathway in a dominant-negative manner. A Notch1 decoy decreased tumor cell viability in xenograft models (Funahashi et al., 2008). However, under different conditions, a DLL1 decoy can play either an activating or inhibitory role (Hicks et al., 2002). Thus, a better understanding of the dynamics by which decoys work is needed before they can be considered as a viable therapeutic strategy.

Other types of experimental inhibitors entail synthetic peptides that mimic MAML1 but lack its active domains. Despite the use of these peptides to serve basic research purposes, their use for therapeutic purposes is still limited. Moellering et al. generated a synthetic, cell-permeable, alpha-helical peptide (SAHM1) blocking MAML1 recruitment and NOTCH-mediated transcription as it binds with high affinity to the interface on the NOTCH-CSL transactivation complex (Moellering et al., 2009). Treatment of human T-ALL cell lines and a mouse model of NOTCH1-driven T-ALL with SAHM1 resulted in strong, NOTCH-specific inhibition of cell proliferation, and leukemia progression (Moellering et al., 2009).

Another intriguing idea for the treatment of tumors that are induced by NOTCH and depend on pathway activity is to not target the Notch pathway itself but focus on its signaling targets. Several such efforts are currently underway. Briefly, we have recently demonstrated in vivo T-ALL remission when we target: a) the NOTCH1-induced IKK kinase complex with a pivotal role in controlling the NF-kB pathway which-in turn-is strongly related to NOTCH in leukemia (Figure 4) (Dan et al., 2008; Espinosa et al., 2010; Vilimas et al., 2007), b) the CyclinD:CDK4/6 kinase complex, hyperactivated in this type of acute leukemia (Sawai et al., 2012), and c) the bromodomain-containing protein BRD4 (King et al., 2013). BRD proteins can be transcriptional co-activators and share common binding patterns with T-ALL oncogenes NOTCH1 and MYC in promoters and enhancers of key genes for the induction and progression of the disease. Bradner and colleagues recently modified a thienodiazepine molecule so that it inhibits binding of BRD to the acetylated residues of histone H4 (Filippakopoulos et al., 2010). We were able to show that such drugs can target both NOTCH1- and MYC-regulated transcription in T-ALL, leading to complete disease remission in vivo. Such, “epigenetically”-targeted therapies might be particularly attractive considering the ability of Notch to alter locus accessibility and initiate transcription. We have recently connected NOTCH1 binding to loss of H3K27me3 on target promoters and demonstrated an antagonism between NOTCH1 binding and polycomb complex 2 (PRC2) recruitment and activity (Ntziachristos et al., 2012). Based on these findings, H3K27me3 demethylation inhibitors might be an attractive therapy option in NOTCH1-induced T-ALL (or CLL). Finally, recent evidence suggests that it may be possible to inhibit Notch signaling by interfering with its trafficking in cancer cell secretory pathways (Ilagan and Kopan, 2013; Kramer et al., 2013).

While a number of “anti-Notch” strategies are emerging, it may be as important to specifically activate Notch in tumors where activation of the Notch pathway is tumor suppressive. As discussed above for AML, in the case where tumor cells express a Notch receptor (NOTCH2) but do not show signs of pathway activation, providing a ligand for these receptors, or treating with activating antibodies may be sufficient to activate the pathway in certain contexts and inhibit tumor growth. Furthermore, in cases where Notch receptors are not expressed (e.g. they are transcriptionally silenced), or if some of their key target genes are silenced, approaches to de-repress the expression of these genes may be useful to slow cancer growth (Stockhausen et al., 2005).

Future directions in the understanding and treatment of Notch-induced tumors

The NOTCH pathway has been the intense focus of cancer researchers for the last two decades. Unfortunately, there are still no FDA-approved, Notch-targeted therapies. Retrospectively, this is not surprising, as we now know that the pathway plays key roles in several tissues, including adult differentiating and regenerating tissues, explaining the potential side-effects of general inhibitors of the Notch pathway such as GSI. The critical question is whether one can successfully target the Notch pathway to significantly inhibit cancer growth. Another major conundrum comes from possible distinct roles for Notch at several stages of the tumorigenic process, an idea that was not been thoroughly examined. Furthermore, it is likely that inhibition of Notch signaling in tumors initiated by Notch activating mutations will have a therapeutic effect, as tumors are often addicted to early initiating events. However, in tumors where alterations in Notch pathway members occur late during tumor evolution, tumors may rapidly invent ways around the targeting of Notch.

We would suggest that specificity should be the key for future attempts to target Notch activity in cancer cells: one should have a complete map of both Notch receptor and ligand expression in different cancers and their microenvironments to be able to use antibodies or other small molecules that specifically inhibit only the relevant molecules. Targeted (Notch-focused) sequencing of tumors is also important to provide a clear idea of the type of mutation and its potential impact on pathway activity. Importantly, a large number of tumors containing Notch activating mutations, like the ICN1 translocation, cannot be treated with GSI. In contrast, receptor-specific antibody agonists could be of significance clinical value for tumors in which Notch signaling has a tumor suppressive function. Myeloid neoplasms are a cancer subtype that could benefit from targeted pathway activation as we have shown that in such tumors, the pathway is inactive but the NOTCH2 receptor is expressed on the surface of the cells and can be activated by ligand binding, leading to cell death. Strategies to activate Notch in some cancer are worth testing, first in pre-clinical models and hopefully in the near future in patients. Moreover, both Notch agonists and antagonists could also be used in combination with current treatments, including chemotherapy and more recent targeted therapies. A recent intriguing example of such treatments is the combination of anti-DLL4 antibodies with either chemotherapy or Avastin or VEGF traps to target tumor angiogenesis (Lobov et al., 2011; Noguera-Troise et al., 2006). Another one is the combination of Notch receptor inhibition (using GSI or antibodies) and glucocorticoids for the treatment of T-ALL (Real et al., 2009).

As with other signaling pathways involved in cancer, identification and targeting of Notch-interacting partners and targets could be pivotal for the development of anti-tumor therapy protocols. Some attempts have been made to identify such genes/proteins using whole proteome (mass spectrometry) (Yatim et al., 2012) and genome/transcriptome (RNA-seq, gene array, ChIP-seq for NOTCH1 and HES1) (Ntziachristos et al., 2012; Wang et al., 2011a) approaches. These studies suggest that, apart from a small fraction of “universal” targets, including members of the HES and HEY families, Notch pathway activity controls the expression of a large number of tissue and cell-type specific gene targets. Indeed, we have previously shown that NOTCH2-HES1 signaling can regulate the expression of CEBPA and PU1, two key regulators of myeloid differentiation, but these genes are not affected by Notch pathway regulation in T-ALL. Thus, potential targeting of the function of tissue-specific Notch pathway targets could offer more targeted therapies with fewer side effects. In a similar fashion, it will be intriguing to define the biochemical composition of the nuclear Notch complex in different tissues to see whether there is specificity that can guide small molecule inhibition efforts.

Overall, it is fair to say that it took the scientific community almost a century to reach to the point that the basic molecular tenets of Notch signaling are well understood. Similarly, although we know for the last two decades that Notch signaling is involved in cancer, only recently we developed the means (small molecules, antibodies) to effectively target pathway activation in this disease. There is still a significant need for further research efforts that will better define the pathway and propose drugs or drug combinations that can affect Notch signaling specifically in cancer, avoiding harmful side-effects and improving both survival and quality of life for patients with Notch-induced tumors.