It has been proposed that EGF-like ligands are bivalent; in case of NRG-1, an N-terminally located high affinity site with narrow specificity first binds a direct receptor (ErbB-3 or ErbB-4), and then a second C-terminal site recruits a co-receptor with lower affinity but broader specificity, ErbB-2 usually being the preferred one (Tzahar et al., 1997). Such a mechanism may account for the diversity of receptor dimers that are observed with a single ligand, as well as for the activation of the orphan receptor ErbB-2 in response to different EGF-like growth factors (Karunagaran et al., 1996; Graus-Porta et al., 1997). Evidence for the existence of all ten possible homo- and heterodimers of ErbB proteins has been reported, including the ErbB-2 homodimer that is stabilized by oncogenic mutation or overexpression (Riese et al., 1995; Tzahar et al., 1996). This network of inter-receptor interactions displays a strict hierarchy rather than a random pattern (Tzahar et al., 1996). In fact, ErbB-2 is the preferred heterodimerization partner for all other ErbB family members. A driving force for the preferential binding to ErbB-2 could be that heterodimers containing ErbB-2 have a very low rate of ligand dissociation compared to other receptor pairs; this property of ErbB-2 can significantly prolong cell stimulation by every ErbB ligand (Graus-Porta et al., 1995; Karunagaran et al., 1996). In the absence of ErbB-2, NRG-1 induces formation of other heterodimers like ErbB-1/ErbB-3 and ErbB-1/ErbB-4 heterodimers, which explains the inhibition in trans of EGF binding to ErbB-1 when NRG-1 is present (Karunagaran et al., 1995). The existence of several ligands, together with their distinct ability to stabilize preferential homo- and heterodimeric receptor pairs, points to the existence of a hierarchical network of ligand-stimulated receptor dimerization events within the ErbB family (Pinkas-Kramarski et al., 1996).

Autophosphorylation of tyrosine residues may also be influenced by the heterodimer combination (Olayioye et al., 1998); in this way, each receptor has the ability to interact with distinct sets of intracellular signaling proteins thus increasing the functional versatility of the ErbB family. There are four potential mechanisms which may account for ligand-induced differential phosphorylation of the receptors (reviewed in Sweeney and Carraway, 2000). One ligand may induce receptor dimerization and phosphorylation of a particular subset of tyrosine residues. Binding to a different ligand could influence site usage for phosphorylation by promoting the association of the dimeric receptor complex with other cellular proteins like kinases, phosphatases or even cell surface molecules, which may mediate phosphorylation or dephosphorylation of specific sites. Alternatively, this second ligand could stimulate the assembly of oligomeric receptor complexes, or induce a different conformation of the receptor pair. Taken together, combinatorial dimerization and ligand-induced diversification of signaling appear to confer ErbB receptors the potential to give rise to a broad range of cellular responses; moreover, because each receptor is unique in terms of catalytic activity, cellular routing and transmodulation, the resulting network allows fine tuning and stringent control of cellular functions.

As for other receptor tyrosine kinases, ligand-induced autophosphorylation of ErbB receptors on specific tyrosine residues creates docking sites for cytoplasmic signaling proteins containing Src-homology2 (SH2) or phosphotyrosine-binding (PTB) domains (reviewed in Pawson, 1995). The binding specificity of these proteins is determined by the amino acid sequences adjacent to the phosphorylated tyrosines; amino acids located N-terminally determine the binding of specific PTB domains, and amino acids that are C-terminally located select SH2 domains. All ErbB family members, including the C. elegans and D. melanogaster homologs Let23 and DER, couple via Shc and/or Grb-2 to the mitogen-activated protein kinase (MAPK) pathway (Pinkas-Kramarski et al., 1996). However, certain intracellular proteins are preferential substrates of specific ErbB receptors. ErbB-3 is a preferred activator of p85 subunit of phosphatidylinositol-3-kinase (PI-3-K) due to the multiple specific binding motifs present in the ErbB-3 C-terminal tail, which are virtually absent in case of ErbB-2 (Fedi et al., 1994; Prigent et al., 1994). Similarly, the negative regulator c-Cbl and phospholipase Cγ1 (PLCγ1) couple to both ErbB-1 and ErbB-2 but not to ErbB-3 or ErbB-4 (Fazioli et al., 1991; Fedi et al., 1994; Cohen et al., 1996; Levkowitz et al., 1998; Klapper et al., 2000; Levkowitz et al., 2000). Olayioye et al. (2000) offers an excellent review of the present knowledge about the intracellular mediators of ErbB signals. As this work aimed to find novel substrates for the ErbB-2 receptor, its known downstream effectors will be described in detail.

ErbB-2 contains 5 putative autophosphorylation sites in its C-terminal tail, termed here Y1 (the most N-terminal tyrosine residue) to Y5 (the most C-terminal one; Hazan et al., 1989; Segatto et al., 1990; Akiyama et al., 1991). It has been described that Shc binds to Y4 through its PTB domain (Segatto et al., 1993; Ricci et al., 1995; Dankort et al., 1997). Grb-2 directly binds to Y2 via its SH2 domain, and indirectly via Shc (Ricci et al., 1995; Dankort et al., 1997). Chk binds to Y5 (Zrihan-Licht et al., 1998). Grb-7, c-Src, Ras-GTPase activating protein (Ras-GAP) and the abovementioned PLCγ1 also interact with ErbB-2, though the binding sites are unclear (Fazioli et al., 1991; Jallal et al., 1992; Stein et al., 1994; Muthuswamy and Muller, 1995b).

The functional role of the ErbB-2 autophosphorylation sites in receptor-mediated transformation has been assessed by mutational analysis of the rodent constitutively active ErbB-2, termed Neu (Dankort et al., 1997). Absence of all major autophosphorylation sites of Neu dramatically decreases transforming activity upon overexpression in fibroblasts. The C-terminal tyrosine residues Y2 to Y5 can independently mediate transformation, since they fully restore transforming activity when individually added back to the inactive receptor. In contrast, the first tyrosine residue Y1 may not be involved in receptor-mediated transformation, as the resulting add-back mutant lacks transforming potential; moreover, Y1 may represent a negative regulatory site, since it supresses transforming activity when restored in combination with any other single tyrosine residue. Recent studies show that the functionally-redundant add-back mutants containing tyrosines Y2, Y4 and Y5 activate Ras to induce transformation, whereas the add-back mutant containing tyrosine Y3 operates independently of Ras to activate MAPK (Dankort et al., 2001).

The importance of ErbB signaling in development was revealed by genetic studies in mice. Null mutations of individual ErbB receptor loci are embryonic lethal. Loss of erbB-1 leads to embryonic or perinatal lethality depending on the genetic background of the host (Miettinen et al., 1995; Sibilia et al., 1995; Threadgill et al., 1995; Sibilia et al., 1998); the mice display abnormalities in multiple organs including brain, skin, lung and gastrointestinal tract. Mice that show spontaneous or targeted mutation of TGFa, one of the various ErbB-1 ligands, show only part of the phenotype observed in erbB-1 null mice, like impaired development of the eyes and hair follicles (Luetteke et al., 1993; Mann et al., 1993); this partial overlap suggests that each ErbB-1 ligand may play a distinct developmental role. The information gained by targeted mutation of erbB-2, erbB-3 and erbB-4 receptors and their ligand NRG-1 clearly demonstrates that distinct receptor heterodimers are essential for different developmental events. ErbB-2 -/- mice die at midgestation (E10.5) due to malformation of myocardial trabeculae in the heart ventricle (Lee et al., 1995), a phenotype that is shared by the NRG-1 (Meyer et al., 1995) and the erbB-4 null mice (Gassmann et al., 1995). These results are consistent with the view that NRG-1, which is expressed in endothelial cells of the endocardium (Meyer and Birchmeier, 1994), is required for activation of myocardial ErbB-2/ErbB-4 heterodimers to promote trabecular formation in the developing heart. In contrast to ErbB-2 and ErbB-4, ErbB-3 is not expressed in myocardium but in mesenchyme of the pre-valvular endocardial cushions. Accordingly, erbB-3 null mice die at day E13.5 displaying normal heart trabeculation but defective valve formation (Erickson et al., 1997; Riethmacher et al., 1997).

In addition to cardiac abnormalities, erbB-3 -/- mice show a generalized neural crest defect that affects both central and peripheral nervous structures. ErbB-3 mutant mice fail to form cranial sensory ganglia due to impaired migration of neurons from the hindbrain neural crest (Erickson et al., 1997; Riethmacher et al., 1997). This phenotype is also observed in mice lacking erbB-2 or NRG-1 (Lee et al., 95; Meyer et al., 1995); in contrast, erbB-4 deficient mice do not exhibit deficient cellularity of cranial ganglia but rather the innervation of these ganglia is disrupted, thus suggesting a unique role for ErbB-4 (Gassmann et al., 1995). In the peripheral nervous system, erbB-3 -/- mice completely lack Schwann cell precursors and Schwann cells that normally accompany nerves (Erickson et al., 1997; Riethmacher et al., 1997). In addition, degenerative motor and sensory neurons are found in the dorsal root ganglia (Erickson et al., 1997; Riethmacher et al., 1997), and migration of sympathogenic neural crest cells is also impaired in erbB-3 mutants (Britsch et al., 1998); similar defects have been observed in NRG-1 and erbB-2 deficient mice (Meyer et al., 1995; Kramer et al., 1996; Britsch et al., 1998). It is evident from these studies that ErbB-2/ErbB-3 heterodimers transmit NRG-1 signals for neural crest cells to migrate. Recently, the early mortality of erbB-2 null mice has been rescued by myocardial expression of an erbB-2 transgene (Morris et al., 1999; Woldeyesus et al., 1999); erbB-2 rescued mice show striking similarities with the erbB-3 null mice, thus confirming the role of ErbB-2/ErbB-3 complexes in the development of the peripheral nervous system. Data on genetic ablation of the other NRG genes are largely missing. Mice carrying a targeted mutation of the NRG-3 gene, however, do not show an overt phenotype, but may deserve a more detailed analysis (T. Müller and C. Birchmeier, unpublished data). So far, the differential embryonic expression of the various neuregulin proteins and their distinct biological properties in vitro suggest different physiological functions (Carraway et al., 1997; Chang et al., 1997; Crovello et al., 1998). Together, these observations show the critical function of ErbB receptors and their ligands during embryogenesis. Moreover, the above genetic studies define distinct developmental roles for certain receptor combinations and are therefore strong support for the occurrence of signal diversification in vivo.

There is considerable evidence that ErbB signaling has important roles in both normal and neoplastic growth of the mammary gland. All four ErbB receptors are found in mammary tissue (Schroeder et al., 1998). In prepubescent mammary gland, ErbB-1 and ErbB-2 are widely expressed in epithelium, stroma and mesenchymal fat, with ErbB-1 levels being particularly high in stromal cells. In the mature gland, ErbB-3 and ErbB-4 are also detected; just at this stage, ErbB-1 and ErbB-2 are differentially located: ErbB-1 is present in stroma and adipose compartments, while ErbB-2 is prominent in epithelium. During pregnancy, the four ErbB receptors are coordinately expressed in mammary epithelium; ErbB-1 and ErbB-2 are found at high levels in the alveolar epithelium throughout pregnancy, whereas ErbB-3 and ErbB-4 levels increase preferentially in the ductal epithelium later in pregnancy. During lactation, receptor levels are low. ErbB-1 and ErbB-2 expression markedly increases during involution, while ErbB-3 expression declines and ErbB-4 expression is not detectable.

Despite expression of all ErbB receptors during puberty, only ErbB-1 and ErbB-2 are tyrosine-phosphorylated while ductal growth occurs (Sebastian et al., 1998). ErbB-3 and ErbB-4 seem to be present in a non-phosphorylated, inactive state at this stage, suggesting that none of them is relevant for ductal morphogenesis. Endogenous phosphorylation of all four receptors is detected during late pregnancy and lactation, with increasing levels of tyrosine-phosphorylated ErbB-1, ErbB-2 and ErbB-4 later at lactation. Information of the phosphorylation state of the ErbB receptors during involution is missing.

Like the receptors, the six EGFR ligands and NRG-1 are differentially expressed in mammary tissue (Schroeder et al., 1998). TGFα, betacellulin and heparin-binding EGF (HB-EGF) transcripts are found in prepubescent mammary gland and through mid-pregnancy, drop markedly during late pregnancy and lactation and again increase at involution. Amphiregulin and epiregulin transcripts appear in mature virgin glands and early in pregnancy, respectively; transcript levels of both factors decline later in pregnancy, remain low throughout lactation and involution, and reappear as the gland again resembles the mature virgin state. EGF transcripts are present at low levels in the virgin gland, and they dramatically increase during late pregnancy and lactation and return to low levels during involution. The NRG-1α isoform is found in mammary mesenchyme and shows a strongly regulated pattern of expression (Yang et al., 1995). NRG-1 is present at low levels in the virgin gland. At mid-pregnancy, NRG-1 exhibits a sudden concentration peak and then rapidly decreases to basal levels, which are constant later in pregnancy, throughout lactation and involution.

Several lines of evidence support a role of ErbB signaling in the development of the mammary gland. In early studies using mice, pellets containing EGF-like factors were surgically implanted into the mammary glands to allow the slow local release of these factors. Implants of EGF, TGFα and NRG-1α stimulate ductal side-branching and lobulo-alveolar morphogenesis in virgin glands (Vonderhaar et al., 1987; Jones et al., 1996). EGF- or TGFα-induced alveoli lack secretory activity; in contrast, alveoli that derive from NRG-1α treatment are differentiated into secretory structures, which accumulate secreted milk proteins in their luminal compartment. These observations indicate that all factors can promote formation of alveoli but only NRG-1α stimulates their terminal differentiation. In other experiments, EGF-like growth factors were directly injected into the mammary glands to study their ability to induce tyrosine phosphorylation of ErbB receptors. Treatment of prepubescent glands with EGF stimulates tyrosine phosphorylation of stromal ErbB-1 (the EGF receptor) and ErbB-2 (Schroeder et al., 1998; Sebastian et al., 1998); as above mentioned, endogenous phosphorylation of ErbB-1 and ErbB-2 is observed in mammary tissue during puberty (Sebastian et al., 1998). Therefore, it is likely that locally-produced EGF induces at puberty the formation of active ErbB-1/ErbB-2 heterodimers, which may be essential for mammary ductal growth. Exogenous EGF induces phosphorylation of ErbB-1 and ErbB-2 at pregnancy, despite all ErbB receptors being present; in contrast, administration of NRG-1β results in trans-phosphorylation of all ErbB receptors, clearly indicating that neuregulin can induce the formation of combinatorial receptor complexes at pregnancy. Together with the pregnancy-restricted expression of neuregulin in mammary mesenchyme (Yang et al., 1995), these findings strongly suggest that NRG-1 may play a major role in the mammary gland during pregnancy to promote alveolar morphogenesis via trans-activation of ErbB heterodimeric receptors.

Convincing evidence of the physiological role of ErbB signals in mammary development has been supplied by genetic studies in mice; moreover, the phenotypes of knockout or transgenic mice support the differential roles of the ErbB receptors and their ligands that could be expected from the abovementioned stimulation studies, and from expression and activation patterns in vivo. Mammary glands from mice carrying a targeted mutation of the amphiregulin gene show impaired ductal growth, whereas TGFa/EGF double null mice show normal mammary development at this stage (Luetteke et al., 1999). Amphiregulin null mammary glands are competent for lobuloalveolar differentiation; however, additional loss of TGFa and/or EGF severely compromises lactogenesis. Coherent with these findings, expression of a TGFa transgene in mammary tissue induces precocious alveolar development in virgin females, alveolar hyperplasia during pregnancy and delayed involution (Matsui et al., 1990; Sandgren et al., 1995). Together, these results suggest distinct functions of the various ErbB-1 ligands in the mammary gland: amphiregulin may be critical for ductal growth during puberty, while TGFα or EGF may be involved in alveolar differentiation during pregnancy. ErbB-1 is activated in virgin tissue; mammary expression of a transgene encoding a dominant-negative ErbB-1 receptor (the EGF receptor) inhibits ductal branching in the glands of virgin mice thus showing a role for ErbB-1 signaling in pubertal mammary development (Xie et al., 1997); similarly, female waved-2 mice, which carry a spontaneous inactivating mutation of the erbB-1 gene, display impaired glandular development (Fowler et al., 1995). Transplantation and tissue recombination experiments further support a role for ErbB-1 in ductal morphogenesis. Mammary gland grafts from neonatal erbB-1 -/- mice fail to undergo ductal growth (Sebastian et al., 1998); however, they develop lobulo-alveolar structures when stimulated by prolactin, indicating that the EGF receptor is essential for ductal branching but not for alveolar morphogenesis (Wiesen et al., 1999). Moreover, tissue recombinants revealed that wild-type fat pad supports outgrowth of erbB-1 -/- epithelium whereas the -/- fat pad does not, thus clearly showing the relevant role of stromal ErbB-1 in mammary ductal growth.

A role for NRG-1 in mammary gland development during pregnancy is substantiated by its restricted expression and pan-activating effect on ErbB receptors at this stage (see above). Indeed, mammary glands from mice that lack NRG-1a fail to undergo lobulo-alveolar morphogenesis at pregnancy (Li et al., submitted); NRG-1α is the isoform that is normally expressed in the glands. In contrast, NRG-1b null mice die during embryogenesis; the NRG-1β isoform accounts for the cardiac and neural crest phenotypes that were described in section 2.4 of this Introduction (C. Birchmeier, unpublished data). These results definitely demonstrate that NRG-1α is the naturally occurring isoform in mammary tissue, where it functions during pregnancy as a critical growth factor that promotes alveolar morphogenesis and secretory differentiation. Further insights into the mechanisms of neuregulin-induced morphogenesis have been gained from genetic studies with ErbB receptors. Expression of dominant-negative forms of erbB-2 and erbB-4 in the mammary gland of transgenic mice revealed a physiological role for both receptors in lobulo-alveolar development at late pregnancy and lactation (Jones et al., 1999; Jones and Stern, 1999). Moreover, there is genetic and biochemical evidence that Stat5a mediates morphogenic signals of ErbB-4 in the mammary gland during lactation (Liu et al., 1997; Jones et al., 1999). According to this data, it can be speculated that NRG-1α induces alveolar morphogenesis via signaling pathways that involve trans-activation of ErbB-2 and ErbB-4 and transcriptional regulation by Stat5a.

As already mentioned in Section 1.3 of this Introduction, the physiological processes that prepare the mammary gland for lactogenesis can be mimicked in vitro using organ culture and organotypic assays. By these means, our group has extensively contributed to understand the function of neuregulin in the mammary gland. The pioneering work of Yang et al. (1995) has shown that NRG promotes lobulo-alveolar differentiation of mammary gland explants in organ culture. The same morphogenic effect of neuregulin has been observed by Niemann et al. (1998) in organotypic cell culture. Treatment with neuregulin induces EpH4 mammary epithelial cells to form alveoli-like structures when cultured on a Matrigel reconstituted matrix; moreover, alveolar cells functionally differentiate and secrete milk components into a luminal compartment, thus reproducing in vitro the physiological responses of mammary epithelium to neuregulin. In addition, these studies show that activation of exogenous ErbB-2 receptor tyrosine kinase is sufficient for EpH4 cells to undergo alveolar morphogenesis in the Matrigel system. The present work contributes to understand the intracellular mechanisms underlying the morphogenic events of neuregulin-stimulated EpH4 cells in organotypic culture, for it identifies Vav2 as an essential mediator of ErbB-2 specific signals leading to alveolar morphogenesis.