Notch signaling: control of cell communication and cell fate

The ability to form biological patterns is key to the orderly and reproducible development of all multicellular life. Pattern formation is made possible by molecular mechanisms of cell-cell signaling, which permit cells to influence each other's fate and behavior. One of the most important mechanisms of cell signaling is mediated by Notch, a transmembrane receptor that coordinates a signaling system known as the Notch pathway. Notch was identified genetically almost 100 years ago by a mutant fly with `notches' in its wings (Morgan, 1917),which indicated its requirement in wing outgrowth. Notch has since been found to be crucial for patterning in a great number of other developmental settings throughout the animal kingdom, from worms to humans.

This primer describes the key molecular features of Notch signaling and some representative biological processes that it controls by first introducing the core players and mechanism of Notch signal transduction; Notch is an unusual protein in that it functions both at the cell surface to receive extracellular signals and in the nucleus to regulate gene expression. In addition, some of the many biological settings where Notch signaling regulates cell fates and pattern formation are discussed, together with molecular strategies that influence Notch signaling in specific developmental locations. Finally, how the knowledge of Notch function and biology from model organism studies has lent insight into human diseases caused by aberrant Notch signaling is also considered.

Key components of Notch signaling were originally recognized genetically through mutant animals whose phenotypes resembled those of Notchmutants. In flies, Notch was the founding member of a collection of`neurogenic' mutants (see Box 1), so named because they produce a remarkable excess of neurons at the expense of epidermis(Poulson, 1945; Lehmann et al., 1983). Nematodes have two homologs of Notch (LIN-12 and GLP-1), which were identified by mutations that affect cell lineages and germ-line proliferation. The lin-12/glp-1 double mutant displays an aggregate phenotype that constitutes the full loss of Notch activity in the worm; this phenotype is characteristic of a small class of `LIN and GLP' or `LAG' mutants in worms(Lambie and Kimble, 1991).

These mutants laid the foundation for genetic, molecular and biochemical studies that established the core Notch signaling apparatus. At its heart lies a Delta-type ligand, a Notch-type receptor and a transcription factor of the CBF1/Su(H)/LAG1 (CSL) family (Fig. 1). All metazoan organisms studied to date contain one or more orthologs of each of these proteins, and these are summarized in Table 1. Delta- and Notch-related proteins are all single-pass transmembrane proteins that contain extracellular arrays of epidermal growth factor (EGF) repeats; specific EGF repeats mediate direct contact between ligand and receptor(Rebay et al., 1991). CSL proteins are sequence-specific DNA-binding proteins(Henkel et al., 1994) that function downstream of Notch. Because almost all locations of Notch signaling involve this ligand-receptor-transcription factor trio, they are generally considered as the `core' components of Notch signaling.

Some signaling cascades are truly cascades, and involve a complicated sequence of proteins that pass a message from the outside of the cell into the nucleus. At the opposite end of the spectrum lies Notch signaling, which operates by a remarkably direct mechanism. The route towards understanding how Notch works, however, has not been so direct.

For many different types of signal-activated cell-surface receptors,removal of the extracellular domain creates a mutant receptor that is permanently in the active mode. This is the case for Notch: an artificial,truncated Notch protein consisting of only its intracellular domain(Notch^intra) has strong constitutive activity in flies and worms(Lieber et al., 1993; Struhl et al., 1993). Interestingly, these engineered Notch^intra proteins localized to the nuclei of transgenic animals, which indicated that the transmembrane receptor Notch might have a nuclear function. Consistent with this model, a direct protein-protein interaction has been observed between Notch^intra and the CSL transcription factor(Fortini and Artavanis-Tsakonas,1994). However, a competing model based upon tissue culture data proposed that the purpose of Notch-CSL binding was to hold CSL in the cytoplasm until receptor activation, at which point CSL would be released and travel to the nucleus (Fortini and Artavanis-Tsakonas, 1994).

Both models were challenged by the locations of the natural proteins in tissues engaged in Notch signaling: endogenous nuclear Notch is essentially never seen, while CSL appears constitutively nuclear. Eventually, the evidence came together to support strongly the Notch nuclear translocation model. The key findings were: (1) that Notch is proteolyzed in response to its interaction with ligand, which releases a soluble intracellular fragment (a natural Notch^intra molecule; see Box 2)(Kopan et al., 1996; Schroeter et al., 1998; Struhl and Adachi, 1998); (2)that Notch^intra is a transcriptional co-activator(Jarriault et al., 1995; Hsieh et al., 1996); and (3)that exceedingly small, histochemically invisible, amounts of Notch^intra suffice to activate target genes(Schroeter et al., 1998; Struhl and Adachi, 1998). The current `canonical' view of Notch signaling is that ligand-induced activation of Notch triggers the cleavage and liberation of a small amount of Notch<sup>intra</sup>, which then translocates to the nucleus and serves as a CSL transcriptional co-activator (Fig. 1, but see Box 3 for some examples of `non-canonical' Notch signaling).

If CSL proteins reside in the nucleus, do they do anything when Notch is at the cell surface? CSL function was initially perplexing; vertebrate CSL proteins were first characterized as transcriptional repressors(Dou et al., 1994), but genetic tests in flies showed that CSL activated target genes during Notch signaling (Bailey and Posakony,1995; Lecourtois and Schweisguth, 1995). How can the same protein be both a repressor and an activator?

Insight into this puzzle came from a virus. The EBNA2 protein from Epstein-Barr Virus (EBV) is a transcriptional co-activator that binds to and hijacks CSL in infected B cells. Interestingly, EBNA2 converts CSL from a default repressor into an activator of transcription(Hsieh and Hayward, 1995; Waltzer et al., 1995). Notch^intra was later found to use the same strategy(Jarriault et al., 1995; Hsieh et al., 1996). The basis of this switch involves distinct CSL co-repressor and co-activator complexes(Fig. 1, `nucleus'). In the absence of Notch signaling, CSL associates with transcriptional co-repressors that actively keep target gene expression switched off(Kao et al., 1998). Following Notch activation, the CSL co-repressor complex is replaced by a co-activator complex coordinated by Notch^intra. Active repression of targets in the absence of Notch signaling allows cells to tightly control signaling outputs (see Box 4 for unexpected consequences of this), and this dual mode of regulation is now understood to be a common feature of most of the major signaling cascades (reviewed by Barolo and Posakony, 2002).

Although engagement of the Notch^intra/CSL activator complex is normally necessary for a target gene to respond to Notch signaling, it is not always sufficient. Indeed, no Notch target gene is expressed in every location where Notch itself is activated. This specificity is due to co-regulation of Notch target genes by other transcription factors and/or signaling pathways(reviewed by Bray and Furriols,2001). For example, cone cell-specific expression of pax2(sv - FlyBase) during Drosophila eye development is achieved through coordinated regulation by at least three inputs - Notch signaling, EGF receptor signaling and Lozenge (Flores et al., 2000). Combinatorial regulation allows for a transcriptional response that is appropriate for each different developmental setting (see also Box 4 for consequences of complex gene regulation).

Notch mutant fly embryos are so strongly affected that Donald Poulson, who pioneered the use of mutants to study fly development, was compelled to write that `All in all, a kind of hopeless monster is produced which can not develop beyond the embryonic stage'(Poulson, 1945). We now know that Notch is likely to be involved in the development of most tissues in species throughout the animal kingdom, with myriad effects on cell fate specification, proliferation and cell death(Table 2). The following examples are a brief glimpse into the repertoire of Notch, and illustrate some different types of Notch-regulated patterning events.

Contingency plans are part and parcel of development. Often, more cells than necessary have the opportunity to become a specialized cell type, a scheme that allows for backups. An important role for Notch signaling is to prevent these `extras' from actually taking on such specialized fates(Fig. 2A,B).

The classic example of this is the Drosophila neural-epidermal choice. Special groups of cells known as proneural clusters have neural potential because of their expression of basic helix-loop-helix (bHLH)transcriptional activators, also known as proneural proteins. Notch signaling restricts neural differentiation by repressing the expression of proneural genes (Parks et al., 1997). The failure of Notch signaling causes all proneural cluster cells to have high levels of proneural proteins and become neural, manifest as the `neurogenic'phenotype (Fig. 2C,D). Constitutive Notch signaling has the opposite effect and suppresses neural differentiation.

Many proteins that mediate neural repression in Drosophila are encoded by the Enhancer of split Complex [E(spl)-C]. This genomic region contains multiple genes that are directly turned on by Notch signaling via CSL-binding sites in their promoters(Bailey and Posakony, 1995; Lecourtois and Schweisguth,1995), including several genes that encode related bHLH repressor proteins. Forced expression of E(spl)bHLH repressors is sufficient to inhibit neural development (Nakao and Campos-Ortega, 1996), possibly by directly inhibiting the expression of bHLH proneural activators(Heitzler et al., 1996).

The same E(spl)bHLH repressors are deployed in other settings where Notch signaling restricts the number of progenitor cells during Drosophiladevelopment, including those of the visceral and somatic musculature, midgut and intestine, heart and a host of other internal organs(Hartenstein et al., 1992). Notch signaling in vertebrates also represses neurogenesis and myogenesis(Fig. 2E,F and Table 2) via homologous Hairy/E(spl)-related bHLH repressors known as HES (mammal), HER (fish) or ESR(frog) proteins (Chitnis et al.,1995) (reviewed by Artavanis-Tsakonas et al.,1999). Activation of bHLH repressors by Notch signaling is therefore a general strategy for preventing equipotent cells from all acquiring the same fate, a role sometimes referred to as `inhibitory Notch signaling'.

A second general role of Notch is to promote the development of a given cell type or body region, often by inducing the expression of positively acting regulatory molecules. In many of these cases, Notch signaling creates a new cell type as a result of cell-cell interactions at the boundary between distinct cell populations (Fig. 3A,B). This is sometimes referred to as `inductive' Notch signaling, and contrasts with the `inhibitory' role of Notch where Notch signaling represses a given cell fate among equipotent cells.

A well-understood example of this occurs during Drosophila wing development. Notch signaling between the dorsal and ventral compartments of the future wing specifies the wing margin, a line of cells that organizes the outgrowth of the wing. A loss of Notch signaling eliminates the wing margin and wing tissue (visible as the eponymous wing `notching'), while ectopic Notch signaling results in extra wing tissue(Fig. 3C-E). In this setting, a key output of Notch activation is to directly turn on vestigial in the presumptive wing margin (Kim et al.,1996). Vestigial is a transcriptional co-activator that is essential for wing development, and ectopic Vestigial will direct the formation of wing-like outgrowths in inappropriate locations.

Notch signaling also works at boundaries during vertebrate somitogenesis. The segmented vertebrate body plan is founded upon regularly spaced blocks of mesoderm known as somites. Somites split off progressively from the presomitic mesoderm, a process driven by periodic oscillations in gene expression that are called the segmentation `clock'. Notch signaling appears to be central to the segmentation clock, as the expression of many Notch pathway components oscillates within the presomitic mesoderm, and mutation of members of the Notch pathway causes defects in clock oscillation and segmentation(Fig. 3F,G)(Conlon et al., 1995; Palmeirim et al., 1997)(reviewed by Bessho and Kageyama,2003). Oscillating gene expression involves an auto-repressive activity of the Notch-activated bHLH repressors Hes1 and Hes7, which turn their own expression off. As these repressor proteins are short lived, their rapid degradation permits a new cycle of Hes1/7 transcription to begin (reviewed by Bessho and Kageyama,2003).

It remains to be fully understood, though, how the segmentation clock physically leads to somitogenesis, and whether or not Notch/CSL or Hes proteins directly regulate any genes that mediate somite partitioning. More generally, there are many `inductive' Notch-regulated developmental processes,in which Notch signaling promotes rather than represses a cell type or behavior, for which the relevant target genes remain to be identified. For example, it is not well understood how Notch promotes germline proliferation in C. elegans (Berry et al.,1997) (Fig. 3H,I)or specifies cell fates such as mammalian astrocytes, fly glia and early worm blastomeres (Table 2). Although bHLH repressors are expressed in some `Notch-inductive' settings, most of the well-characterized examples also involve the direct activation of genes that encode positively acting transcriptional regulators or proteins specific to terminally differentiated cells. Whether this is generally true or not will be revealed by a more detailed understanding of genes directly regulated by Notch^intra/CSL during normal development.

As we have now seen, the major biological role of Notch signaling is to control the developmental fates of cells and to make cells different from one another. Therefore, cells become distinguished from one another according to whether they predominantly send or receive Notch signals. In some cases, this is easily explained by the exclusive distribution of ligand and receptor. For example, during C. elegans gonadal development, the somatic distal tip cell signals to adjacent germline cells to induce their division, which generates the pool of germline cells. The directionality of signaling is due to the fact that the distal tip cell expresses only the ligand LAG-2, while the germline cells express only the Notch receptor GLP-1(Henderson et al., 1994). Notch ligands often show spatially patterned expression, so this is a general focal point for controlling Notch signaling. However, there are many situations where all of the cells involved in Notch signaling express both ligand and receptor. How is signaling made directional in these cases?

Notch signaling sometimes shows a remarkable ability to amplify small differences in the signaling capacities of different cells. One tactic is for signaling to regulate receptor and/or ligand transcription. In this way, the degree to which a cell activates Notch signaling has a dynamic effect on how well it sends and/or responds to Notch signals in return.

A clear example of this occurs during the anchor cell/ventral uterine cell(AC/VU) decision in C. elegans gonadal development. Although the destiny of almost all nematode cells is predetermined by lineage, AC and VU acquire their identity through a Notch-mediated discussion between two cells,referred to as Z1.ppp and Z4.aaa. Initially, LIN-12 (Notch) and LAG-2 (ligand)are expressed by both cells, and they engage in `mutual' bi-directional Notch signaling. However, random fluctuation in expression of LIN-12 and LAG-2 is amplified by positive feedback; activation of LIN-12 promotes its own expression and inhibits the expression of LAG-2 by the cell. In the end, one cell expresses only the ligand and differentiates as AC, while the other cell expresses only receptor and turns into VU(Wilkinson et al., 1994; Christensen et al., 1996). A directional signaling situation like this is sometimes called `lateral'signaling.

Similar `mutual' to `lateral' Notch signaling is seen during the selection of Drosophila neural precursors from groups of equipotent proneural cluster cells discussed earlier (Fig. 2A), a process known to be very sensitive to the levels of ligand and receptor in each individual cell(Heitzler and Simpson, 1991). A proposed mechanism for Notch-mediated regulation of ligand incorporated the finding that the expression of Delta can be activated by proneural proteins, which are, in turn, inhibited by Notch-activated bHLH repressors(Heitzler et al., 1996). However, this regulatory chain does not fully explain how signaling becomes directional, as increased levels of proneural proteins in neural precursors and decreased levels of proneural proteins in the inhibited cells are not well correlated with changes in Delta or Notch levels during normal development(Parks et al., 1997). Thus,post-transcriptional mechanisms to amplify differences in Delta-Notch signaling must also exist, some of which are discussed in the next section.

A second way to influence Notch signaling among cells that co-express ligand and receptor involves post-translational modifications. Two molecules that target the Delta ligand are Neuralized and Mindbomb. Although they are not related to one another structurally, both are RING-type E3 ubiquitin ligases that directly ubiquitinate Delta and cause it to be internalized from the plasma membrane. Recent studies show that the effect of Neuralized and Mindbomb is actually to increase the ability of Delta to activate Notch signaling in neighboring cells (Itoh et al., 2003) (reviewed by Le Borgne and Schweisguth, 2003). One way in which this might work is if the extracellular domain of Notch is co-internalized with Delta into the Delta-expressing cell, which might facilitate the `S2' cleavage of Notch in the Notch expressing cell (Fig. 1). The expression of these ubiquitin ligases is initiated in a subset of Notch-dependent processes, including neurogenesis, and may be responsible for making Notch signaling directional during the neural/epidermal fate decision.

Notch activity is regulated by sequential glycosylation of particular EGF repeats (reviewed by Schweisguth,2004). Notch modification by the glycosyltransferase Fringe has complex effects. Fringe inhibits the ability of Notch to be activated by Serrate- and Jagged-type ligands, but stimulates the response of Notch to Delta-type ligands (Panin et al.,1997). In vitro data suggest that Fringe enhances Notch-Delta binding and decreases Notch-Serrate binding. Fringe is often deployed in a spatially restricted pattern to position or create directional Notch signaling during cell fate-inductive events. Some examples of this occur during development of the fly wing margin, at the dorsoventral boundary of the fly eye, at the dorsoventral limb borders of vertebrate limbs and in developing vertebrate somites (reviewed by Irvine,1999).

A special type of directional inhibitory Notch signaling occurs during many instances of asymmetric cell divisions(Fig. 4A). In this situation,although both sister cells are capable of sending and receiving Notch signals,directionality is imposed by the asymmetric segregation of factors that influence Notch signaling. Such factors localize to a crescent centered at one end of the mitotic spindle in the mother cell(Fig. 4B).

A crucial Notch-inhibitory factor is Numb(Rhyu et al., 1994), a phosphotyrosine binding (PTB) domain adaptor protein that binds directly to Notch and prevents the Notch pathway from becoming activated in that cell. This is well illustrated during Drosophila sensory lineages, in which asymmetric cell division is the consequence of Numb being inherited by only one daughter cell, and Notch signaling being suppressed in that daughter cell(reviewed by Posakony, 1994). For example, one cell in the adult bristle lineage divides to give rise to one socket cell and one shaft cell (Fig. 4A,C). Notch signaling is activated only in the socket cell and prevents it from becoming a shaft cell; the presumptive shaft cell is immune to Notch signaling because it inherits Numb(Fig. 4A,B). If both cells have activated Notch signaling - either through loss of Numb or constitutively active Notch - they both turn into socket cells(Fig. 4D). Conversely, if neither cell has activated Notch signaling, a double-shaft structure results(Fig. 4E). Therefore, both Numb and Notch signaling are needed to ensure that the daughter cells of this asymmetric cell division adopt different fates.

Given the profound and widespread roles of Notch signaling across a range of tissues, it is perhaps no surprise that deviant Notch signaling underlies some human diseases. Consistent with the roles of Notch signaling in development and neural behavior, relevant heriditary conditions include Alagille syndrome (where mutations in the Notch ligand JAG1 affect the development of many organs, including that of the liver, skeleton, heart and eye) (Li et al., 1997; Oda et al., 1997), certain forms of spondylocostal dysostosis (where mutations in the ligand DLL3 result in rib fusions and trunk dwarfism)(Bulman et al., 2000) and cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy (CADASIL) (where mutations in the extracellular EGF repeats of NOTCH3 predispose individuals to dementia, migraines and strokes) (Joutel et al.,1996). In addition, a mild decrease in Notch signaling in mice causes specific defects in spatial learning and memory, which might make Notch a more general culprit in cognitive deficits(Costa et al., 2003).

Aberrant Notch signaling is also intimately involved in several human cancers. This was first demonstrated in a recurrent t(7;9) translocation that is associated with certain pre-T cell acute lymphoblastic leukemias (T-ALL)(Ellisen et al., 1991). This chromosomal rearrangement results in constitutive Notch activity in T cells. Conversely, cells that are mutant for Notch1 form skin and corneal tumors in mice, indicating that Notch also suppresses tumorigenesis(Nicolas et al., 2003). Many other human and murine cancers, including certain neuroblastomas, and mammary,skin, cervical and prostate cancers, are correlated with alterations in expression of Notch proteins and/or ligands. Although causal relationships in many cases await better characterization, these observations suggest broad roles for Notch dysfunction in cellular transformation.

The CSL transcription factor is a particular victim of viruses that redirect CSL function towards their own ends. Multiple proteins from Epstein-Barr virus (EBV), Kaposi's sarcoma-associated herpesvirus (KSHV), and adenovirus type 5 bind CSL directly and modulate CSL-dependent transcription of both viral and host target genes in a Notch-independent fashion (reviewed by Allenspach et al., 2002). A majority of humans are latently infected with EBV and adenovirus, but, as is the case for KSHV, initial infection is usually asymptomatic. A serious problem arises in immunocompromised individuals, though, where opportunistic behavior of EBV and KSHV causes oncogenic transformation and malignancy.

The molecular elucidation of these diseases is now the basis of diagnostic tools. However, it is hoped that knowledge of the mechanism of Notch signaling will be relevant for therapeutic design. This hope remains to be realized, but one can imagine that conditions that arise from Notch pathway gain-of-function might be alleviated by small molecule antagonists of genetic inhibitors of Notch processing/Notch^intra-co-activator function, while conditions that arise from Notch pathway loss-of-function might be treated with targeted delivery of soluble Notch ligands or other strategies that either locally activate Notch activity or suppress CSL co-repressor activity. A variety of relevant mouse disease models are now available to help test these strategies. Of course, given the very general functions of the Notch pathway, minimization of collateral or non-specific effects will be necessary to make such therapies clinically useful.

A clear theme of developmental and evolutionary biology is that nature is frugal: useful proteins and signaling pathways are reused in diverse settings. Notch signaling is an example of a particularly successful signaling cascade that is used repeatedly throughout the development of all metazoan organisms. Many Notch-mediated affairs are highly analogous; for example, Notch signaling mediates the inhibition of many different cell fates through the same bHLH repressor proteins. However, the redeployment of this pathway during evolution has resulted in diverse outcomes to Notch activation in different situations. Depending on the setting, Notch can inhibit, delay or induce differentiation,and can variously promote apoptosis, cell division or a static state(Table 2). Things become more complicated when one considers how signaling pathways interact with one another. For example, Notch signaling and EGF receptor signaling can work in parallel or in series to regulate a given process, and can either cooperate or antagonize each other during transcriptional regulation of a given target. Notch signaling is also influenced by a great number of other factors in specific settings, only some of which have been addressed in this primer(reviewed by Greenwald, 1998; Schweisguth, 2004). All of these considerations mean that although one can be guided by previous studies,it is imperative to evaluate each new example of Notch signaling carefully.

Given the pervasive use of the Notch pathway in animal development, how was this signaling cascade originally assembled during evolution? Although only metazoans are known to use Notch signaling, certain `prospective' components of the Notch pathway are found in other types of eukaryotes. A striking example is the definitive existence of CSL homologs in some fungi. If these species never had a Notch pathway, they may shed light on ancestral functions of CSL transcription factors. Will they be found to be transcriptional activators or repressors, or both? And are there any species in which both Notch and CSL are present, but are not associated in a common signaling process? It will be a great future challenge to understand when and how functional connections arose between components of a presumptive Notch pathway during evolution.

Elise Lamar, Jose F. de Celis, Tim Schedl, Clarissa Henry, Fabrice Roegiers and Scott Barolo graciously contributed images.