Modulation of Erbb2 signaling during development: a threshold level of Erbb2 signaling is required for development

The Erbb2 receptor tyrosine kinase is a member of the epidermal growth factor receptor (Egfr) family, which also includes the Egfr/Erbb1, Erbb3, and Erbb4. Although there is considerable homology amongst the Egfr family members, each Egfr family member can interact with a distinct subset of ligands (Olayioye et al.,2000; Pinkas-Kramarski et al.,1996a; Yarden and Sliwkowski,2001). For example, Egfr ligands such as Egf or Tgfβ bind directly to the Egfr whereas the heregulins are specific ligands for ErbB3 or ErbB4 (Carraway et al., 1994; Peles et al., 1993). In contrast to these Egfr family members, a direct soluble ligand for Erbb2 has yet to be described. However, there have been several reports suggesting that a member of the mucin family known as Muc4 may act as a membrane-bound ligand for Erbb2 (Carraway et al.,1999).

Despite the lack of a direct soluble ligand for Erbb2, the tyrosine kinase activity of Erbb2 can be stimulated by other Egfr ligands through the formation of specific heterodimers with other members of the Egfr family(Goldman et al., 1990; Karunagaran et al., 1996; Pinkas-Kramarski et al.,1996a; Pinkas-Kramarski et al., 1996b; Tzahar et al.,1996). For example, stimulation of cells with Egf can result in transphosphorylation of Erbb2 through the formation of Egfr-Erbb2 heterodimers(King et al., 1988; Stern and Kamps, 1988). Similarly, stimulation of cells with heregulin can result in the transphosphorylation of Erbb2 through the formation of specific heterodimers of Erbb2 and Erbb4 or heterodimers of Erbb2 and Erbb3(Goldman et al., 1990; Karunagaran et al., 1996; Pinkas-Kramarski et al.,1998a; Pinkas-Kramarski et al., 1996a; Pinkas-Kramarski et al., 1998b; Pinkas-Kramarski et al.,1996b; Sliwkowski et al.,1994; Stern and Kamps,1988; Tzahar et al.,1997; Tzahar et al.,1996). Transphosphorylation of Erbb3 is absolutely dependent on its capacity to form specific heterodimers with other members of the Egfr family since it lacks intrinsic tyrosine kinase activity(Guy et al., 1994). Taken together, these observations argue that the formation of specific heterodimers play an important biological role in the function of the Egfr family.

Members of the Egfr family share structural similarities including an extracellular ligand-binding domain, a single-pass transmembrane domain, a highly conserved tyrosine kinase domain, and a regulatory carboxyl-terminal tail containing several tyrosine autophosphorylation sites(Hynes and Stern, 1994). Ligand-mediated activation results in strong mitogenic signals from the receptors, leading to cellular growth and differentiation. Not surprisingly,members of the Egfr family are collectively involved in both development and disease. Whereas amplification and aberrant overexpression of Erbb2 has been implicated in various cancers, most notably in breast cancer(Simpson et al., 1995; Slamon et al., 1987; Slamon et al., 1989), the loss of Erbb2, Erbb3 or Erbb4 expression in knock-out mice has deleterious effects on the developing embryo (Britsch et al., 1998; Gassmann et al.,1995; Lee et al.,1995; Riethmacher et al.,1997). For example, Erbb2- and Erbb3-deficient animals share similar hypoplastic development of the sympathetic nervous system (Britsch et al., 1998),and Erbb2- and Erbb4-deficient animals exhibit defective formation of cardiac ventricular trabecules(Gassmann et al., 1995; Lee et al., 1995).

Upon receptor activation, specific tyrosine residues in the terminal tail of the receptor are autophosphorylated(Akiyama et al., 1991; Hazan et al., 1990), and then serve as potential binding sites for intracellular signaling proteins harboring phosphotyrosine binding (PTB)(Kavanaugh et al., 1995) or Src homology 2 (SH2) (Pawson,1995) domains. In a series of studies by Dankort et al.(Dankort et al., 1997), the five major tyrosine autophosphorylation sites (Y1028, Y1144, Y1201, Y1227,Y1253) were evaluated systematically for their roles in constitutively activated Erbb2-mediated transformation of fibroblast cells. Although simultaneous ablation of all five sites in the tyrosine phosphorylation-deficient (NYPD) mutant drastically impaired transformation, independent tyrosine-tophenylalanine mutations at four of five sites only modestly reduced the transforming ability. Substitution at the one remaining site (Y1028) resulted in a consistent increase in transformation. Conversely, restoration of individual tyrosine residues to the NYPD mutant at one of four sites (`add-back' mutants) was not only able to fully restore the transforming ability, but was also able to transform with a modest increase over the fully functional receptor. The Y1028add-back mutant had the opposite effect and completely ablated the residual transforming abilities of the NYPD mutant. These observations suggested that while four of the five tyrosine sites positively mediate Erbb2 signaling, Y1028 negatively modulates Erbb2 activity.

We describe here the introduction of wild-type and phosphotyrosine mutant Erbb2 cDNAs into the endogenous mouse Erbb2 locus to examine their physiological roles, in vivo. Although mice derived from the different knock-in alleles were viable, examination of the levels of Erbb2 expression revealed that the knock-in strains expressed only 10% of the expected Erbb2 protein. A further reduction of Erbb2 protein, achieved by intercrossing the Erbb2 knock-in mutants with Erbb2 null mice, resulted in perinatal lethality. Thus, we established a minimum threshold level of Erbb2 required for viability. In contrast, expression of the Y1028F mutation in knock-in animals genetically rescued this perinatal lethality. Biochemical analyses revealed that these animals expressed higher levels of Erbb2 above the threshold. We further identify that Y1028 mediates the negative regulation of Erbb2 signaling by influencing the turnover rate of the receptor.

cDNAs encoding the non-activated rat Erbb2 (rErbb2) receptor harboring single tyrosine phosphorylation mutations were generated from the corresponding C-terminal domain of the oncogenic rErbb2 mutants previously described by Dankort et al.(Dankort et al., 1997). The cDNAs were cloned into the targeting vector containing flanking 3′ and 5′ arms of homology as described previously(Chan et al., 2002). These plasmids were electroporated into R1 ES cells and G418 (Geneticin; GIBCO)-resistant colonies were picked and subsequently screened by Southern Blot analysis for correctly targeted mutants. Mutant mice were generated from the positive ES cell clones and were maintained in an SV129/Balb/c background. All animal work was performed with approval and under the strict guidelines of the Canadian Council on Animal Care (CCAC).

Embryos were lysed in modified TNE lysis buffer (50 mM Tris-HCl pH 7.6, 150 mM NaCl, 1% NP-40, 10 mM NaF, 2 mM EDTA) supplemented with 10 μg/ml leupeptin, 10 μg/ml aprotinin and 1 mM sodium orthovanadate. For immunoblotting, the membranes were incubated with an anti-Erbb2 monoclonal(Ab-3, Oncogene Research) or anti-Grb2 polyclonal (1:2500; C-23, Santa Cruz)antibody overnight at 4°C. Subsequently, the membranes were treated with anti-mouseHRP or anti-rabbitHRP secondary antibodies for 1 hour at room temperature, washed three times in TBS and visualized by enhanced chemiluminescence (Amersham), as specified by the manufacturer.

For whole-mount in situ hybridization, embryos were fixed in 4%paraformaldehyde/0.2% glutaraldehyde, dehydrated through a graded series of methanol/PBT (PBS+0.1% Tween-20) baths and stored in 100% MeOH at–20°C until needed. Whole-mount in situ hybridization was carried out as previously described (Wilkinson and Nieto, 1993).

E12.5 embryos were fixed in fresh 4% paraformaldehyde, washed several times with PBS and then post-fixed and bleached in Dent's fixative (methanol:DMSO,4:1)+5% H₂O₂. After extensive washes, the embryos were incubated in blocking solution (5% goat serum+1% DMSO in TBS) overnight at room temperature. Incubation in anti-NF150 (1:1000, Chemicon) diluted in TBS+1% DMSO was performed for 2 days at room temperature. After 5×1 hour washes in TBS-T (TBS+1% Tween20), the embryos were incubated overnight in goat anti-rabbitHRP (1:1000 in TBS-T). DAB+ kit (DAKO, catalogue no. K3468) was used to visualize the neurofilament staining. Whole-mount immunohistochemistry was performed on diaphragm dissected from E18.5 embryos and fixed in 4% PFA. They were first treated in 0.1 M glycine in PBS, washed several times, then incubated in PBT (PBS+0.5% Triton-X-100) for 5 minutes. Diaphragms were incubated overnight with anti-NF150 diluted 1:1000 in PBT+5% goat serum. After extensive washes in PBT, they were incubated overnight with goat anti-rabbit-Alexa488 (Molecular Probes) diluted 1:1000 in PBT andα-bungarotoxin-Alexa594 (Molecular Probes). Diaphragms were visualized using the Zeiss LSM 510Meta confocal microscope.

Rat-1 cells stably expressing the oncogenic Erbb2 tyrosine phosphorylation site mutants were seeded (13×10⁶ cells) onto 60-mm plates and pulse labeled using 0.1 mCi/ml of ³⁵S-EXPRESS Protein Labeling mix(NEG772, NEN Life Science Products). The cells were then washed and chased with unlabeled medium for the indicated times. Protein lysates from the labeled cells were immunoprecipitated using an anti-Erbb2 antibody (Ab4,Oncogene Science) and resolved on an 8% SDS-PAGE. The gel was dried and exposed to a PhosphoImager screen (Molecular Dynamics) and scanned using the Typhoon8600 scanner (Amersham). Image analyses were performed using ImageQuant software (Molecular Dynamics).

To assess the relative contribution of the different Erbb2 tyrosine autophosphorylation sites, we employed a targeted knock-in strategy involving replacement of the first coding exon of the mouse Erbb2 gene(m-Erbb2) with the rat Erbb2 cDNA (r-Erbb2)(Fig. 1A). Thus, expression of mouse Erbb2 is disrupted and replaced by expression of the rat Erbb2 cDNA under the control of the endogenous promoter. For genotyping purposes, note that the `wt' superscript (e.g. Erbb2<sup>wt</sup>) will be used herein to designate the endogenous mouse Erbb2 genotype whereas the `Erbb2' superscript (e.g. Erbb2^Erbb2) will be used to designate the knock-in rat Erbb2 cDNA allele. Following germline transmission of the targeted knock-in allele, transgenic animals homozygous for the Erbb2 knock-in cDNA (Erbb2^Erbb2/Erbb2) were viable, appeared healthy, and reproduced at the expected Mendelian frequencies(Table 1A).

To examine the functional roles of the tyrosine autophosphorylation sites in vivo, mutant rat Erbb2 cDNA harboring single tyrosine-to-phenylalanine substitutions(Fig. 1C) were also expressed in mice using an identical targeting strategy to that described above but substituting the cDNA cassette (Fig. 1B). Specifically, we generated knock-in mice expressing the Erbb2-Y1028F mutant, as well as the Erbb2-Y1144F mutant(loss of the direct binding site for Grb2) and the Erbb2-Y1227Fmutant (loss of the direct binding site for Shc). Homozygous Erbb2^{Y1028F/Y1028F} knock-in animals expressing the Y1028F mutation were generated from heterozygous matings at the expected Mendelian frequencies (Table 1A) and appeared normal and healthy. Interestingly, homozygous Erbb2^{Y1144F/Y1144F} and Erbb2^{Y1227F/Y1227F} knock-in mice were also normal, healthy and fertile (Table 1A).

The use of a rat Erbb2 cDNA instead of a mouse cDNA was necessary to distinguish expression of the knock-in allele from the endogenous gene and allow for the validation of the knock-in strategy. The mouse and rat Erbb2 share a greater than 93% sequence similarity at both the nucleic acid and amino acid levels. Importantly, the tyrosine autophosphorylation sites and the sequences surrounding these sites in the carboxyl-terminus are conserved. To confirm that rescue of the embryonic lethality associated with disruption and inactivation of the mouse Erbb2 gene (Morris et al.,1999) was attributed solely to expression of the knock-in rat Erbb2 cDNA, RT-PCR was performed on total RNA isolated from wild-type and knock-in animals. Analysis of the RT-PCR product suggested that homozygous animals indeed only expressed the rat Erbb2 cDNA (data not shown).

Since specific expression of the knock-in allele was validated, we next performed immunoblot analyses on protein lysates prepared from E12.5 wild-type, heterozygous, and homozygous mutant embryos(Fig. 1E) to determine whether the knock-in Erbb2 allele expressed wild-type levels of Erbb2 protein. Surprisingly, the levels of Erbb2 expressed in the Erbb2^Erbb2/Erbb2 embryos(Fig. 1E, lanes 4-5) were dramatically reduced relative to the expression of endogenous Erbb2 in wild-type littermates (Fig. 1E,lanes 1 and 2). However, in spite of the considerably reduced levels of Erbb2 protein, homozygous Erbb2^Erbb2/Erbb2 knock-in animals did not display any obvious phenotype and appeared generally healthy.

Although mice bearing the knock-in rat Erbb2 cDNA appeared phenotypically normal, we have previously demonstrated that they indeed exhibited subtle defects. In particular, our lab showed that rat Erbb2 cDNA knock-in animals possessed only 10% of the number of muscle spindles found in wild-type animals(Andrechek et al., 2002). Consequently, we investigated the effects of further reducing the expression of Erbb2 by interbreeding heterozygous knock-in(Erbb2^wt/Erbb2) animals with heterozygous Erbb2knock-out (Erbb2^wt/ko) animals. This strategy allowed us to express a single Erbb2 knock-in allele in an Erbb2-deficient background to generate hemizygous Erbb2^Erbb2/ko animals. Significantly, no Erbb2^Erbb2/ko animals were found at weaning age(three-weeks old) when the animals were genotyped(Table 1A).

To determine the precise reason that the Erbb2^Erbb2/koanimals were dying, we interbred homozygous Erbb2^Erbb2/Erbb2 knock-in animals with heterozygous Erbb2^wt/ko animals and assessed whether we could detect viable hemizygous Erbb2^Erbb2/ko animals at birth. Although the expected number of animals were present at birth, 18 of the 38 newborn pups were either stillborn or started dying immediately after birth(Table 1B). They could not breathe independently, despite the ability to open their mouths, and they became cyanotic and died within a few minutes. Indeed, all of the dead pups were genotyped to be Erbb2^Erbb2/ko animals. Subsequent postmortem histological analyses of the lungs confirmed that the hemizygous Erbb2^Erbb2/ko pups were unable to inflate and expand their lungs, despite being vascularized and structurally intact (data not shown). These observations suggest that a critical minimal threshold level of Erbb2 protein is required to maintain viability.

Previous in vitro analyses of the tyrosine phosphorylation mutants identified tyrosine 1028 as a negative regulator of Erbb2-induced transformation via an undetermined mechanism. Accordingly, we next asked whether we could genetically rescue the perinatal lethality observed with the hemizygous Erbb2^Erbb2/ko animals, by removing the putative negative regulatory tyrosine residue in Erbb2. This was accomplished by similarly crossing the heterozygous Erbb2^wt/Y1028Fknock-in animals with Erbb2^wt/ko animals to generate mice expressing a single Y1028F cDNA knock-in allele in an Erbb2-deficient background. Interestingly, in contrast to the hemizygous Erbb2^Erbb2/ko pups, no perinatal lethality was observed with the hemizygous Erbb2^Y1028F/ko animals(Table 1A). In fact, all of the progeny survived and the adult Erbb2^Y1028F/ko animals appeared normal and healthy and were fertile. To preclude any possibility that the phenotypic differences observed in the Erbb2-Y1028F knock-in animals versus the control Erbb2 knock-in animals were not by chance and that it is specifically due to this particular point mutation, mice expressing either the Y1144F or the Y1227F phosphotyrosine mutant Erbb2 receptor in an Erbb2-deficient background were also generated. As indicated in Table 1B, none of the hemizygous Erbb2^Y1144F/ko or Erbb2^Y1227F/ko animals survived, and similar to the Erbb2^wt/ko animals, they died shortly after birth struggling unsuccessfully to breathe.

To further elucidate the nature of the defect resulting in the inability of the animals to inflate their lungs, the neuromuscular junctions in the diaphragm muscles of E18.5 embryos were examined by whole-mount immunostaining with a neurofilament-150 antibody and with α-bungarotoxin. In the control animals, Erbb2^wt/wt and Erbb2^Erbb2/Erbb2, the central region of the diaphragm muscles was innervated by the relatively large main phrenic nerve with smaller intramuscular branches (Fig. 2A,B). Accordingly, innervation of the diaphragm from the genetically rescued embryos, Erbb2^Y1028F/ko was similar to that in the control animals (Fig. 2C). In contrast, mutant diaphragms from Erbb2^Erbb2/ko, Erbb2^Y1144F/ko,Erbb2^Y1227F/ko embryos were very poorly innervated, which probably resulted in the perinatal respiratory failure. As shown in Fig. 2D-F, the innervations were thin, disorganized, and discontinuous or incomplete and they lacked an apparent main nerve trunk from which smaller presynaptic branches originate. However, labeling of postsynaptic acetylcholine receptor (AChR) clusters withα-bungarotoxin did not reveal any significant differences in the number,density or shape of AChR clusters (Fig. 2G-L,G′-L′).

Additional evidence for the genetic rescue of the hypomorphic Erbb2 knock-in allele by expression of the Y1028F mutation was observed in the developing sympathetic nervous system, where Erbb2 plays a critical role (Britsch et al.,1998). For comparison, Fig. 3A,C show the normal development of the primary sympathetic chain ganglia in heterozygous embryos at E12.5, using Phox2a as a marker. In Erbb2^Erbb2/ko embryos, there was only partial development of the thoracic sympathetic ganglia and the defects became more severe in a rostral-caudal gradient (Fig. 3B). In comparison, the sympathetic ganglia of Erbb2^Y1028F/ko embryos were similar in size and length to those of heterozygous littermates (Fig. 3D). Take together, these results strongly suggest that Erbb2-Y1028 specifically mediates a negative regulatory effect on the receptor and this plays an important function in vivo to modulate the activity of the receptor throughout development.

Since we hypothesized earlier that a minimal threshold level of Erbb2 is required for normal development, we explored whether this Y1028Fmutation resulted in increased levels of Erbb2 such that the hemizygous animals survived. To determine the molecular basis for the hypermorphic Erbb2^Y1028F allele, immunoblot analyses were performed on protein lysates derived from E12.5 embryos. Expression of the Y1028F receptor mutant in the Erbb2^Y1028F knock-in animals consistently resulted in significantly higher levels of Erbb2(Fig. 4A, lane 2 versus lane 5,lanes 3-4 versus lanes 6-7; Fig. 4B, lane 3 versus lane 4). In the hemizygous embryos, there is a critical difference in Erbb2 levels in the Erbb2^Y1028F/koembryos (Fig. 4A, lanes 6-7)versus the Erbb2^Erbb2/ko embryos(Fig. 4A, lanes 3-4), which probably determined the difference as to whether the animals survived or died shortly after birth. Note that the Erbb2 levels in Erbb2^Y1028F/ko embryos(Fig. 4A, lanes 6-7) are comparable to the levels in Erbb2^Erbb2/Erbb2 embryos(Fig. 4A, lane 2), which are essentially healthy animals.

To confirm that the observed increase in Erbb2 protein is specific to expression of the Y1028F allele, we also examined the levels of Erbb2 in homozygous E12.5 embryos expressing either the Erbb2-Y1144F mutation or the Erbb2-Y1227Fmutation. The results revealed that only embryos with the Y1028Fallele expressed elevated levels of Erbb2 when compared to the levels observed in knock-in embryos expressing the Erbb2 cDNA, the Y1144FcDNA, or the Y1227F cDNA alleles(Fig. 4B, compare lane 4 versus lanes 3,6,7). In fact, quantitative immunoblotting using ¹²⁵I-conjugated secondary antibodies and subsequent ImageQuant(Molecular Dynamics) analyses revealed that the levels of Erbb2 in Erbb2^{Y1028F/Y1028F} embryos were at least 3- to 8-fold higher than in Erbb2^Erbb2/Erbb2,Erbb2^{Y1144F/Y1144F} and Erbb2^{Y1227F/Y1227F}embryos (Fig. 4C). Thus, the quantitative difference in Erbb2 levels alone probably determined the difference between animals surviving or dying at birth.

In order to assess whether the variation in Erbb2 protein levels observed in the Erbb2 or Y1028F knock-in alleles were occurring at a transcriptional or post-transcriptional level, Erbb2 transcripts in E12.5 embryos were detected using an RNase protection assay(Fig. 4D). The results showed that samples harboring either the Erbb2 or the Y1028Fknock-in allele expressed identical levels of Erbb2 transcripts(Fig. 4D, lanes 4-5 versus lanes 6-8). Note that either knock-in alleles express significantly lower levels of Erbb2 transcripts compared to wild-type animals. This phenomenon is probably a result of using a cDNA knock-in strategy and explains the overall lower Erbb2 expression levels in the knock-in animals. Regardless of this, our observations argue that specifically expressing the Y1028F knock-in allele results in increased Erbb2 protein levels without affecting transcriptional activity, when compared to the Erbb2 knock-in allele.

Since the function and activity of the ErbB family of receptors are closely related, it is plausible that other family members may be affected by the low Erbb2 levels or may be elevated to compensate for the loss Erbb2 activity. In this regard, the expression level of other ErbB receptor family members was measured to determine if there were corresponding differences in expression. Interestingly, Egfr levels were modestly affected depending on the level of Erbb2 in the knock-in animals, which may augment the defects observed. No significant differences in ErbB3 or ErbB4 levels were noted in the wild-type and knock-in animals that would suggest any kind of compensatory role for these receptors.

Since many receptor tyrosine kinases are downregulated by endocytosis and subsequently targeted for degradation(Katzmann et al., 2002), we examined whether Y1028 affected Erbb2 receptor turnover rate(Fig. 5A,B). Rat-1 cell lines stably expressing oncogenic Erbb2 (V664E mutation) or its mutant tyrosine phosphorylation site derivatives were established. Specifically, we compared the turnover rate of the oncogenic Erbb2 receptor versus the Y1028Fmutant receptor. Conversely, we also assessed the effects of restoring Y1028 to the Y1144 add-back mutant to generate the Y1028/Y1144 double add-back mutant Erbb2 receptor. Add-back mutants are derived from an Erbb2 receptor stripped of the five major tyrosine autophosphorylation sites by tyrosine-to-phenylalanine mutations and then individual mutant sites are `added-back' or reverted to tyrosine residues.

Pulse-chase analyses using Rat-1-derived stable cell lines revealed that the Y1028F mutation significantly stabilized the Erbb2 receptor compared to the wild-type receptor (Fig. 5A). Alternatively, when Y1028 was restored to the Y1144add-back mutant, the receptor turnover rate of Erbb2-Y1028/Y1144 increased(Fig. 5B). Endocytosis and turnover of the Erbb2 receptor at the plasma membrane, as determined by monitoring biotin-labeled surface receptor (data not shown), was consistent with the ³⁵S-labeled pulse-chase data. The subcellular localization of Erbb2 or the Y1028F mutant receptor was subsequently determined in transiently transfected Rat-1 cells. Consistent with the retention of the Y1028F receptor at the cell surface, there was a significantly higher proportion of Y1028F receptor localized at the cell surface as determined by immunofluorescent detection when compared to the wild-type Erbb2 receptor(Fig. 5C).

Based on the above results, we next examined whether the E3-ubiquitin ligase, c-Cbl, was responsible for mediating the negative regulatory effect of Y1028 on Erbb2 activity since c-Cbl has been shown to mediate the downregulation of the Egfr and other receptor tyrosine kinases(Thien and Langdon, 2001). Our data suggest that c-Cbl is able to associate with Erbb2 through other specific phosphotyrosine residues, and that Y1028 is not responsible for the recruitment of c-Cbl to Erbb2, nor is it required for the ubiquitylation of Erbb2 (Fig. 5D). Taken together, these observations suggest that Y1028 modulates Erbb2 protein levels in a c-Cbl- and ubiquitylation-independent manner.

Although knock-in animals expressing a single Erbb2 receptor tyrosine phosphorylation site mutation were grossly normal, it was conceivable that there may be subtle defects as a result of ablating a particular Erbb2-initiated signaling pathway. Thus, whole embryos (E12.5) were immunostained with anti-neurofilament (Fig. 6) to examine the appearance of sensory cutaneous nerves, which were abnormally defasciculated in the cardiac rescued Erbb2 mutants and in Erbb3-deficient mutants(Woldeyesus et al., 1999). Interestingly, in comparison with Erbb2^wt/wt embryos(Fig. 6A), there was a striking defect in the morphology of the developing cutaneous nerves in the thoracic body wall of Erbb2^{Y1227F/Y1227F} embryos(Fig. 6D) that was not observed in Erbb2^{Y1028F/Y1028F} or Erbb2^{Y1144F/Y1144F} knock-in embryos(Fig. 6B,C). The trajectories of the sensory nerves in Erbb2^{Y1227F/Y1227F} mutants appeared intact but were highly disorganized and severely defasciculated,similar to the defects in the cardiac-rescued Erbb2 mutants(Woldeyesus et al., 1999). These results suggest that the signal transduction pathways initiated specifically by Erbb2-Y1227 are absolutely required for the proper development of the cutaneous sensory nerves in the thoracic body wall.

Using a series of Erbb2 cDNA knock-in animals encoding Erbb2 receptors with different tyrosine autophosphorylation site mutations, we have identified a minimum threshold level of Erbb2 (above ∼5% of wild-type levels) that is required to sustain viability and physiological function. We have also presented evidence suggesting that the individual Erbb2 tyrosine phosphorylation sites have specific and unique roles in development. Inadvertently, the cDNA knock-in allele turned out to be a hypomorph expressing only ∼10% of the level of Erbb2 expressed in wild-type littermates. Despite the low levels of Erbb2, homozygous Erbb2knock-in animals were viable and overtly normal and healthy without any compensatory effect from other ErbB receptor family members. We cannot however exclude the possibility that there may be more subtle defects that are undetectable or have not been identified. Indeed, we have recently demonstrated that similar hypomorphic knock-in animals displayed no phenotypic differences compared to their wild-type siblings despite having tenfold fewer spindle cells (Andrechek et al.,2002).

Nonetheless, it is intriguing that the animals express significantly higher levels of Erbb2 but only require about 10% for development. Since Erbb2 acts primarily as a co-receptor and its activation is dependent on the ligand-mediated activation of its ErbB counterparts, the high levels of Erbb2 may suggest its excess availability to engage in a dimer signaling unit. Thus,the threshold that we have identified may also reflect the threshold activity of the other ErbB receptors. When the Erbb2 knock-in animals were inter-crossed with Erbb2-deficient animals to generate hemizygous knock-in animals (Erbb2^Erbb2/ko), this additional twofold reduction in Erbb2 levels fell below the threshold level required and resulted in perinatal lethality due to acute respiratory distress. The inability of newborn Erbb2^Erbb2/ko pups to inflate their lungs is remarkably similar in phenotype to the cardiac-specific rescue of Erbb2-deficient mutants (Morris et al., 1999; Woldeyesus et al., 1999).

Evaluation of the diaphragm muscles of E18.5 embryos revealed an obvious defect in the phrenic nerve of mutant embryos. In the cardiac-rescued Erbb2 mutants, Woldeyesus et al.(Woldeyesus et al., 1999)reported that the phrenic nerve was thin and poorly fasciculated at E14.5 but only remnants were detected at later stages. Observations by Morris et al.(Morris et al., 1999) showed that diaphragms of E18.5 mutant embryos were completely devoid of innervations, and AChR clusters were more dispersed and rounded in shape. Interestingly, the diaphragm muscles of our Erbb2^Erbb2/koembryos at E18.5 (Fig. 2) were innervated, although a main nerve trunk was absent and any nerves that were present were very thin, highly disorganized and fragmented. We also did not observe any differences in the density or shape of AChR receptor clusters compared to those of wild-type embryos. These results support the notion of progressive threshold sensitivities to the level of Erbb2 signaling in the developing peripheral nervous system. Although the low levels of Erbb2 prevented or impeded the complete degeneration of motor neurons in the diaphragm muscle and maintained apparently normal clustering of AChR at the neuromuscular junctions, this was not sufficient for the diaphragm to respond and function. Also note that the low levels of Erbb2 were sufficient to bypass the cardiac trabeculation defects in Erbb2 null embryos, suggesting that the cardiac system is not as sensitive as the peripheral nervous system to this threshold level of Erbb2.

Conversely, we were able to genetically rescue the perinatal lethality in hemizygous Erbb2^Erbb2/ko embryos by similarly introducing the Y1028F knock-in allele into an Erbb2-deficient background to generate Erbb2^Y1028F2/ko animals. Tyrosine 1028 is a negative regulatory signal that affects the stability of the receptor (discussed below). All Erbb2 knock-in mice bearing the Y1028F allele expressed higher levels of Erbb2 protein than in the Erbb2 knock-in mice. Thus, this genetic manipulation ablating a negative regulatory phosphotyrosine site in the carboxyl-terminus increased the level of Erbb2 above the minimal threshold required for the phrenic nerve in the diaphragm to develop normally. Erbb2^Y1028F2/koanimals survived and were phenotypically healthy. In contrast, similar experiments performed with the Y1144F or the Y1227F knock-in alleles also resulted in perinatal lethality with similar defects in the innervation of the diaphragm muscle. These two mutations, of course, did not affect the level of Erbb2 expressed in the knock-in animals.

Although previous in vitro experiments concluded that Y1028 negatively regulated oncogenic Erbb2-mediated transformation, we were also interested in determining whether Y1028 would also have this suppressive effect in vivo in the context of a non-oncogenic Erbb2 receptor. As described above, the results of the in vivo experiments with the Y1028F knock-in animals are consistent with a negative regulatory effect of Y1028 on Erbb2 activity. Moreover, here we have also identified the biochemical basis for this effect,which was previously unknown. Phosphotyrosine 1028 modulates Erbb2 protein levels by promoting the downregulation and turnover of the receptor in a c-Cbl- and ubiquitin-independent manner. Conversely, loss of Y1028 resulted in an increase in Erbb2 protein stability. Thus, the stabilization of Erbb2 levels above the minimum threshold in the hemizygous Erbb2^Y1028F2/ko animals allowed these particular strains to develop normally.

Sequence alignment of Erbb2 with the Egfr shows that Erbb2-Y1028 and its surrounding sequences share a high degree of similarity to the region surrounding the Egfr-Y992 (Fig. 1D). Despite the identity and function of Egfr-Y992 suggested in 1989 (Chen et al., 1989), the mechanism of its function has not been clearly elucidated. Carboxyl-terminal truncation of the Egfr has suggested that an 18 amino acid region surrounding Y992 conforms to an `internalization' domain and is required for Egf-dependent receptor internalization. However, a subsequent study showed that point mutation of Y992 did not affect Egfr internalization(Sorkin et al., 1992). In another study, the Y992F mutation actually increased the rate of Egfr internalization and it was suggested that the increase in negative charge associated with phosphorylation of Y992 would reduce the rate of ligand-induced endocytosis (Holbrook et al., 1999). The nature of these discrepancies may be due to artefactual differences in cell types, the level of ectopic Egfr expression in the cells, or the nature of the mutation used in their analyses. Our results showing the effect of Y1028, in a physiologically relevant context, to promote the turnover rate of the Erbb2 receptor from the cell surface should lead to clarification of these conflicting reports concerning the role of Y992 in the Egfr.

Homozygous knock-in animals expressing Erbb2 phosphotyrosine mutations where the Grb2 site (Y1144) or the Shc-binding site (Y1227) were ablated had little impact on the gross development of mice, even at tenfold less expression. As discussed above, previous studies in our lab revealed a significant reduction in the number of muscle spindle cells in Erbb2^Erbb2/Erbb2 animals without any obvious phenotypic consequences (Andrechek et al.,2002). Therefore, it is conceivable that other specific cell or tissue types may be more sensitive to the altered Erbb2 signaling. Indeed, the cutaneous sensory nerves in the thoracic body wall of Erbb2^Y1227F/1227F mutants were thin and defasciculated,whereas Erbb2^{Y1144F/Y1144F} embryos were similar to wild type. Since Erbb2-Y1227 binds to the Shc adapter protein, these results suggest that the Shc signaling pathway may play an important and unique role downstream of Erbb2 in the development of these sensory nerves. Furthermore,it is interesting to note that Shc can recruit phosphatidylinositol 3-kinase(PI3K) (Gu et al., 2000). PI3K is a major signaling pathway downstream of ErbB3, which also display a strikingly similar abnormal sensory nerve phenotype in Erbb3-deficient animals(Woldeyesus et al., 1999). Thus, it is conceivable that the Erbb2 and ErbB3 signaling pathways converge downstream at the level of PI3K, such that PI3K plays a key role in the development of sensory nerves. In general, this raises the idea that signaling pathways downstream of the individual receptor phosphotyrosine sites can function differentially and independently in the development of specific tissues.

Unlike with Erbb2, where loss of the Grb2 binding site (Y1144) did not result in any discernable phenotype, uncoupling of Grb2 from the Met receptor in a knock-in model resulted in severe defects in muscle development(Maina et al., 1996). Taken together, these observations suggest that the remaining Erbb2 autophosphorylation sites are able to functionally substitute for the inability of these Erbb2 mutants to recruit Grb2. Alternatively, it is possible that the Erbb2 heterodimerization partners such as Egfr, ErbB3 or ErbB4 can compensate for the lack of Grb2-binding sites on Erbb2. Indeed, it has been demonstrated that Egfr or ErbB3 can independently bind Shc and Grb2(Batzer et al., 1994; Carraway and Cantley, 1994; Okutani et al., 1994; Prigent and Gullick, 1994). It is also possible that there may be other more subtle developmental defects in different tissues of the Erbb2^Y1144F knock-in mutants. In this regard, transgenic mice expressing constitutively activated Erbb2 mutants coupled to either Grb2 or Shc specifically in the mammary epithelium develop morphologically distinct mammary tumors that possess inherently different metastatic properties(Dankort et al., 2001).

In summary, using a series of unique Erbb2 knock-in animals and genetic manipulation, we have established that a minimum threshold level of Erbb2 receptor expression is required during proper development. This critical threshold is only 5-10% of normal Erbb2 levels in wild-type embryos. We have also identified an important in vivo function of the intrinsic negative regulatory site in Erbb2 (Y1028) to modulate the stability/turnover rate of the receptor throughout development. In addition, our data suggest that individual Erbb2 signaling pathways are not redundant, but rather play unique roles in specific tissues throughout development.

We thank M. Hibbs, A. Dunn, M. Rudnicki, and C. Goridis for generously providing plasmids. Oligonucleotide synthesis and DNA sequencing were performed by D. Gooden and B. Allore at the MOBIX Central Facility (McMaster University). Thanks to Monica Graham (McMaster University) for general technical assistance and to Dongmei Zuo (McGill University) for confocal imaging. This work was supported by a CIHR grant (MOP-10594). W.J.M. is a recipient of the Canadian Research Chair in Molecular Oncology. R.C. was supported by a predoctoral studentship from the U.S. Army DOD Breast Cancer Research Program (DAMD no. 17-00-1-0333).