Journal of Mammary Gland Biology and Neoplasia, Vol. 2, No. 2, 1997

ErbB Receptors and EGF-like Ligands: Cell LineageDetermination and Oncogenesis Through CombinatorialSignaling

Ronit Pinkas-Kramarski,1 Iris Alroy,1 and Yosef Yarden1'2

The ErbB/HER family of transmembrane receptor tyrosine kinases includes four membersthat bind more than two dozens ligands sharing an epidermal growth factor- (EGF)3 like motif.This family plays a pivotal role in cell lineage determination in a variety of tissues, includingmesenchyme-epithelial inductive processes and the interactions between neurons and muscle,glia and Schwann cells. Certain ligands and receptors of the family, especially the ErbB-2receptor tyrosine kinase, contribute to a relatively virulent phenotype of some human tumors;most notable are carcinomas of secretory epithelia. This large variety of biological signals isgenerated through a combinatorial network of signal transduction in which different ErbBligands are apparently capable of stabilizing discrete homo- and heterodimeric receptor com-plexes, each coupled to a specific set of cytoplasmic signaling proteins. Because each receptoris unique in terms of catalytic activity, cellular routing and transmodulation, the resultingnetwork allows not only an enormous potential for signal diversification but also fine tuningand stringent control of cellular functions. ErbB-2 emerges as a master coordinator of thenetwork, prolonging and amplifying signaling by decelerating the dissociation rates of itsheterologous ligands. Thus, the tumorigenic action of ErbB-2 may be attributed to its abilityto act as a shared signaling subunit, rather than by functioning as a bone fide receptor.

KEY WORDS: Growth factor; receptor; tyrosine kinase; signal transduction; cancer.

INTRODUCTION

Since the original discovery of epidermal growthfactor (EGF) in the mouse submaxillary gland and thesubsequent resolution of the three disulfide loops of themolecule, lessons learned with EGF and its signalingpathway have paved the way to understanding otherfamilies of growth factors. Thus, the realization that

1 Department of Molecular Cell Biology, The Weizmann Instituteof Science, Rehovot 76100, Israel.

2 To whom correspondence should be addressed. e-mail:liyarden@weiztnann.weizmann.ac.il

3 Abbreviations: amphiregulin (AR), cysteine-rich domain (CRD),ductal carcinoma in situ (DCIS), epidermal growth factor (EGF),heparin binding EGF (HB-EGF), Neu differentiation factor (NDF),neuregulin (NRG), receptor tyrosine kinase (RTK), transforminggrowth factor a (TGFa).

the membrane receptor of EGF carries an intrinsictyrosine kinase activity has been confirmed by molecu-lar cloning and later shown to be shared by all othergrowth factor receptors. Similarly, the necessity ofEGF-receptor dimerization for transmembrane signal-ing appears to be true for most, if not all, other growthfactor receptors. Signaling by EGF receptor attractedattention not only because of its basic science aspects,but also due to the many links to cancer. Thus, ahomlog of EGF, termed transforming growth factor a(TGFa) was isolated from the medium of oncogene-transformed cells and a transforming avian retroviruswas found to encode a truncated form of the EGFreceptor. A decade later, the number of EGF-likeligands is close to 30 and three homologs of the EGFreceptor joined to form the ErbB family (also calledtype I receptor tyrosine kinases), with EGF receptor

971083-3021/97/0400-0097$12.50/0 O 1997 Plenum Publishing Corporation

98 Pinkas-Kramarski, Alroy, and Yarden

assigned the name ErbB-1. Moreover, many additionallinks to cancer have been found, especially with ErbB-2, a receptor whose overexpression is correlated withpoor prognosis of several types of adenocarcinomas(1, 2). As with other landmarks of the family thatgeneralize to other RTKs, multiplicity of ligands thatbind to a group of structurally related receptors isshared by most other growth factor receptors (3).

ErbB LIGANDS: EGF AND NEUREGULINFAMILIES

The characteristic structure of the EGF-like motifis widely involved in protein-protein interactions,including cell adhesion processes. This domain isdefined by a set of six cysteine residues typicallyspaced over a sequence of 35-50 amino acids, thatform three disulfide bonds. Strict length conservationof the disulfide-held loops characterize those EGFunits that belong to RTK ligands. Several ligands, otherthan EGF and TGFa, are known to bind to the extracel-lular domain of the EGF receptor (Fig. 1) and theyinclude amphiregulin (AR), betacellulin (BTC), hepa-rin-binding EGF (HB-EGF) epiregulin and three viralproteins. Both EGF and TGFa are highly expressedin early stages of embryonic development and also inthe nervous system, but at early preimplantation stagesonly TGFa seems to be functional. Amphiregulin andHB-EGF are secreted heparin-binding growth factorsof 43 and 48 amino acids, respectively. Epiregulinwas recently purified from a mouse fibroblast-derivedtumor cell line (4) and found to inhibit growth ofseveral epithelial tumor cell lines while stimulatingother types of cells. Interestingly, functional duality ofErbB ligands as both growth inhibitory and stimulatoryhas been reported in several cases, and may dependon ligand concentration, serum factors and the reper-toire and relative expression level of the various ErbBproteins expressed in a given cell type. Betacellulinactivity was initially identified in a mouse insulinomacell line. Although it binds and activates ErbB-1, itcan also interact with ErbB-4 (5).

The search for ligands that activate ErbB-2 ledto the isolation of the Neu differentiation factor (NDF)from the growth medium of ras-transformed rat fibro-blasts and heregulin or gp30 from the growth mediumof breast cancer cells (6). It has been later found thatNDF/heregulin activates ErbB-2 only indirectly bybinding to both ErbB-3 and ErbB-4. Two related neu-ronal factors, namely the glial growth factor (GGF)

(7) and the acetylcholine receptor inducing activity(ARIA) (8) were isolated on the basis of a Schwann cellmitogenic activity and transcriptional up-regulation ofthe acetylcholine receptor, respectively. More recently,another related factor was isolated and named sensoryand motor neuron-derived factor (SMDF). Here wewill use the name neuregulins (NRGs) to refer to the15 isoforms of the factor (Fig. 1), which arise byalternative mRNA splicing from a single gene onhuman chromosome 8 (9). Like most of the ErbB-1ligands, the precursors of NRGs are transmembraneproteins that undergo rapid processing and secretion(10). The mosaic structure of the proNRG moleculescontains several structural motifs other than the EGF-like domain, which is sufficient for receptor activation.Two variants of this domain exist, either a or B. Thelatter is shared by all of the neural isoforms and exhibitsa somewhat higher receptor affinity. Multiple tran-scripts of NRGs are detectable in various tissues andorgans. Nevertheless, the major sources appear to beeither stroma cells that underlie epithelial sheets orneural cells.

Despite the apparent strict receptor specificity ofthe various ligands, several lines of evidence indicatethat certain ligands may have broader specificity. Thus,betacellulin can bind and activate ErbB-1 as well asErbB-4 (5). Likewise, HB-EGF interacts weakly withErbB-4 (11). Domain swapping between EGF andNDF produced biregulin, an EGF molecule that carriesthe five N-terminal amino acids of NDF and activatesboth ErbB-1 and the NRG receptors (12). Differencesin biological potency of EGF and TGFa are well docu-mented, but the molecular basis is not understood. Itmay be relevant that various ErbB-1 ligands differ intheir ability to stimulate the four ErbB family members(11), implying broader receptor specificity than pre-viously anticipated. The solution structure of the mini-mal domain necessary for receptor recognition hasbeen solved for EGF, TGFa and NRGa (13, 14). Inter-estingly, only small differences were noted when thestructures were aligned. The NRG ligands have a three-residue insertion between Cys3 and Cys4 in the EGF-like domain resulting in a longer B-loop (the regionbetween cysteines 2 and 4). This insertion is the mostsignificant structural difference between NRGa andEGF or TGFa. Extensive mutagenesis and domainswapping analysis of EGF and TGFa established theimportance of the B-loop and the C-terminal tail forbinding to ErbB-1. Interestingly, the a and fj isoformsof NDF differ in this latter region, which confers dis-tinct receptor recruitment abilities (15), implying that

100 Pinkas-Kramarski, Alroy, and Yarden

EGF and TGFa may differ in specificity to heterodim-eric combinations of ErbB-1.

THE ErbB QUARTET

The four members of the ErbB family of RTKsshare a common molecular structure that includes alarge glycosylated extracellular domain with two cys-teine-rich domains (CRDs), a single transmembranestretch, and a long cytoplasmic tail that contains thetyrosine kinase and the autophosphorylation sites (Fig.2). The ligand binding cleft is confined to the inter-CRD domain (domain III) and the N-terminal tail, atleast in the case of ErbB-1. Domain III is relativelyconserved, despite remarkable differences in ligand

Fig. 2. Schematic representation of the domain structure of ErbBproteins and their homology. The following extracellular domainsare shown: an N-terminal signal peptide (SP), and two cysteine-rich domains (CRDs, domains II and IV). The transmembrane (TM)and the cytoplasmic domains of the tyrosine kinase (TK) and thecarboxy-terminal tail (CT) are indicated. Amino acid identitiesbetween each of the ErbB proteins and ErbB-1 are indicated forindividual domains. Note that ErbB-3 is the least homologousmember.

specificities. Likewise, the catalytic tyrosine kinasedomain is relatively conserved throughout the family,but the flanking juxtamembrane and C-tail regions,which undergo auto- or trans-phosphorylation, displaysequence variation. In spite of an overall conservationof primary structure (Fig. 2), each of the four receptorshas unique functional landmarks. Most important isthe potency and substrate specificity of the tyrosinekinase activity. ErbB-3, unlike the others, is practicallydevoid of kinase activity although it binds ATP (16).By contrast, ErbB-2 exhibits a constitutively high basalkinase activity. The amino acid sequences that flankeach receptor autophosphorylation site are alsoremarkably different. Since these regions determine theidentity of Src homology 2 (SH2) containing signalingproteins that become associated with each receptor,this diversity probably specifies signaling specificity(17). Apparently, the cellular routing of each receptor,after ligand binding, is another functional difference;although ErbB-1 undergoes rapid internalization andtargets EGF for lysosomal degradation, the other ErbBmolecules are slowly internalized (18, 19) and mayrecycle back to the cell surface without significantdegradation of the endocytosed ligand. Lastly, signifi-cantly different patterns of expression characterize thefour receptors. ErbB-2 is the most widely expressedmember, whereas ErbB-3 is found on many typesof epithelial cells and ErbB-4 expression is generallyconfined to nerve, muscle and glia cells. ErbB-1 isexpressed by liver parenchymal cells, fibroblasts andepithelial cells of all kinds. Taken together, the fourErbB receptors display distinct rather than redundantcharacteristics, consistent with specific biologicalactivities.

PHYSIOLOGY OF ErbB SIGNALING

Numerous studies have documented the potentmitogenic action of most ErbB ligands on various typesof cultured cells. However, responses may differdepending on cell types, ligand concentration andligand identity. For example, NRGa was found to bea survival factor for cultured MK keratinocytes,whereas in the same cellular system NRG|3 induced astrong mitogenic signal, probably due to the higherbinding affinity and prolonged signaling by this iso-form (20). NRG isoforms were found to induce pheno-typic differentiation of cultured mammary cells derivedfrom both malignant or normal tissues. The differenti-ated phenotype involves growth arrest, synthesis and

ErbB Receptors and EGF-Like Ligands 101

secretion of milk components and enhanced expressionof several cell adhesion molecules, including the inter-cellular adhesion molecule 1 and CD44 isoforms. Inorgan cultures of mammary glands, NRG stimulatedlobulo-alveolar budding and the production of milkproteins, in accordance with its expression in the mes-enchyme during lobulo-alveolar development at preg-nancy (21). Another example of the action of amesenchyme-derived NRG on epithelial cells is pro-vided by excisional wound repair experiments per-formed in a rabbit ear model (22). In this system,NRGa2, but not other isoforms, induced increased re-epithelialization and epidermal thickening, that wasassociated with expression of epidermal differentiationmarkers without significant epidermal proliferation.

Similar to the mesenchymal-epithelial system,NRGs act in the nervous system as paracrine mediatorsthat instruct cell fate determination. NRGs promoteproliferation of mature Schwann cells, especially whenadministered at high concentration and in the presenceof forskolin. At lower concentrations, the factorincreases motility of mature Schwann cells and pre-vents their apoptosis in vivo (23). Maturation and sur-vival of embryonic brain astrocytes, but not other celltypes, were also induced by NRG (24). NRGs aremost likely the long-sought neuron-derived maturationfactor for embryonic Schwann and glia cell precursors(25, 26) and their action is apparently independent ofcell proliferation. Similarly, nerve cell-made NRGspromote slow myogenic differentiation in vitro (27).The well documented effects of NRG/ARIA on theneuromuscular synapse are also consistent with cellu-lar differentiation rather than proliferation. The factorselectively elevates the e subunit of the nicotinic ace-tylcholine receptor, a component that is induced bysynaptic input, along with the voltage-gated sodiumchannel. NRG is concentrated at the synapse site andits synthesis coincides with the phase of synapse for-mation in embryogenesis, but its source may not berestricted to the presynaptic axon, because transcriptswere localized also to the adjacent muscle cell (28).Lastly, the cell type specificity of NRGs in the nervoussystem appears to be even broader, because the factorenhances the development of oligodendrocytes frombipotential (O2A) glial progenitor cells and it indirectlyaffects nerve cells by up-regulating neurotrophin 3, asurvival factor of sympathetic neuroblasts, in non-neu-ronal cells near embryonic sympathetic ganglia (29).

Consistent with the multiple biological roles ofNRGs, inactivation of the corresponding single murinegene led to embryonic lethality (30). Likewise, knock-

out (KO) of the direct NRG receptor, ErbB-4, as wellas the co-receptor, ErbB-2, had a similar effect (31,32). Strikingly, in all three mutant embryos the finger-like extensions of the ventricular myocardium (trabec-ulae) were absent, but elsewhere the myocardiumappeared normal. By contrast, neural abnormalitiesvaried among the mutants, but in all cases they wereconfined to the rhombencephalon, an embryonic brainstructure that develops into the brainstem. In bothNRG"X~ and ErbB-2~'~ embryos the neural crest-derived portions of cranial sensory ganglia were miss-ing. In addition, certain Schwann cell precursors weremissing in NRG~'~ mutants. Unexpectedly, in ErbB-4~'~ mutants the cranial sensory ganglia looked nor-mal, but their spacing and innervation to and from thecentral nervous system were disrupted. Remarkably,axons that connect cranial ganglia V (trigeminal) andVII/VIII (facial and acoustic) with the hindbrainthrough rhombomers 2 and 4, respectively, were mis-targeted to rhombomers 4 and 2, respectively. Conse-quently, it was proposed that ErbB-4, which isexpressed at this stage of development (E9.5) in theintervening rhombomer 3, mediates a repulsive signalfor innervating axons. Whether this unique activity ofErbB-4 is mediated by a ligand distinct from NRG, orit simply reflects nonredundancy with ErbB-3, may beresolved by the production of ErbB-3 ~'~ animals.

Gene targeting of the TGFa gene, as well as thephenotype of the waved-1 mutant mouse, carrying adefective TGFa gene, revealed dramatic derangementof hair follicles, waviness of the whiskers and fur,and eye abnormalities, including open eyelids at birth.Another recessive mouse mutation that affects coattexture, waved-2, is due to a point mutation within thekinase domain of ErbB-1. Inactivation of the erbB-1gene yielded additional information that may relate tothe in vivo action of other ErbB-1 ligands (33-35).Mutant viability displayed strain-specificity rangingfrom peri-implantation death, due to inner-cell massdegeneration, to in-utero death at mid-gestation. How-ever, on a specific genetic background mutant micesurvived for up to three weeks. Most of the observedphenotypic defects involved aberrant proliferation anddifferentiation of specific epithelia: the spongiotropho-blast layer of the placenta, the corneal epithelium, epi-dermis and hair follicles, gastrointestinal tract, and thelung epithelium. Thus, only part of the traits of ErbB-\~'~ mutants may be attributed to defective TGFasignaling. Nevertheless, considering the multiplicity ofErbB-1 ligands, the overall phenotype is surprisinglymild. This observation may be explained by partial

102 Pinkas-Kramarski, Alroy, and Yarden

compensation of ErbB-1 deficiency by other ErbBmembers.

ErbB AND CANCER

Genetic alterations in growth factors and theirreceptors and the components of their signal transduc-tion pathways are important in the multi-step accumu-lation of cellular defects that lead to carcinogenesis.The ErbB family emerges as a major target for tumor-associated alterations that occur in nonhematopoieticcancers, probably because it provides a major prolifer-ative engine for epithelial and neuronal cells. Extensivediscussion of the relevant literature is out of the scopeof the present review, but may be found in some recentreviews (36, 37).

ErbB ligands. Because most ErbB ligands aresynthesized as transmembrane precursors, their actionmay not be limited to the soluble form. Interaction ofmembrane-bound ligands with surface receptors hasbeen termed juxtacrine. It has not yet been directlydemonstrated in neoplastic tissues, but association ofseveral ErbB ligands with the plasma membrane hasbeen reported and in the case of NRGs, displayeddependency on isoform identity (10). AR, NRG, andHB-EGF contain a heparin-binding motif that allowsretention by extracellular matrix components. Anothermode of stimulation may be intracellular (intracrine)as implied by the presence of a nuclear localizationsignal in AR and translocation of NRG to the nucleus(38). Nevertheless, paracrine stimulation of cancercells by stroma-derived ligands and autocrine activityof tumor cell-derived peptides, appear to be the majormechanisms of cellular activation.

The causative transforming effect of an EGF fam-ily ligand is exemplified by the role of TGFa in coloncancer (37). Transgenic mice bearing the TGFa geneunder control of an exogenous promoter expressedthis growth factor in the colon and developed hyper-plasia of the colonic mucosa. Transfection of welldifferentiated colon cancer cells with the TGFa geneincreased their anchorage-independent growth in vitroand markedly enhanced tumorigeniciry in vivo. Con-sistent with a functional role in transformation, anti-bodies to ErbB-1 and neutralizing antibodies to TGFareduced the rate of growth of colon cancer cells thatexpress the peptide, as did antisense oligonucleotidesand antisense TGFa RNA. TGFa appears to havea prognostic significance in several gastrointestinalmalignancies: for example 35% of esophageal tumors

displayed co-expression of the growth factor and itsreceptor, and expression correlated with shorterpatient survival rates. Poorer prognosis of TGFa-expressing tumors, especially those that co-expressErbB-1, has also been reported in the case of lungand ovarian cancers. Likewise, co-expression of eitherEGF or TGFa, together with ErbB-1, was reported in38% of pancreatic tumors and correlated with tumorsize and decreased patient survival. Because over 90%of pancreatic carcinomas exhibit Ki-ras gene muta-tions and transcription of both TGFa and NRG genesis up-regulated by ras, the prognostic significance ofTGFa in pancreatic, as well as in other cancers, maybe indirectly related to ras activation.

Although less significant than TGFa expression,enhanced synthesis of EGF has been reported in severaltumor types, including lung and ovarian cancers. BothEGF and TGFa are elevated in the urine of gliomapatients. Extensive analysis of NRG expression in can-cer has not yet been reported. Likewise, data on ARare still limited. Nevertheless, some lines of evidencesuggest that AR may act as an autocrine factor in breastcancer (37). The peptide is generally expressed athigher level in invasive breast carcinomas than in duc-tal carcinomas in situ (DCIS) or in normal, nonin-volved mammary epithelium, and its expression, likethat of TGFa, is enhanced by steroid hormones, acti-vated ras and EGF.

ErbB-1. In addition to the clinically-relevantcoexpression of ErbB-1 with its ligands, deletionmutants and overexpression of the erbB-1 gene havebeen observed in several types of human cancer. ErbB-1 overexpression is frequently observed in nonsmallcell lung carcinoma (NSCLC), particularly of the squa-mous variety. Expression levels were correlated withhigh metastatic rate, poor differentiation and shortpatient survival time (39). Cancers of other organs,such as renal carcinoma, ovarian and breast cancer,display elevated levels of ErbB-1 but the correlationwith clinical outcome is inconclusive. On the otherhand, brain malignancies present several types of clini-cally relevant alterations of the erbB-1 gene. First,overexpression, often due to amplification of the erbB-1 gene, is the most common genetic alteration in glialtumors, and it correlates with poor prognosis (40). TheerbB-1 gene is rearranged in a significant fraction ofgliomas in which it is amplified, resulting in in-framedeletions of portions of the extracellular domain ofErbB-1. The most common mutation (type III) deletesamino acids 6-273, without affecting EGF binding.Although this alteration occurs in 17% of glioblastoma

ErbB Receptors and EGF-Like Ligands 103

multiforme tumors, at present its clinical significanceis uncertain.

ErbB-2. Evidence derived from experimentalsystems indicates that ErbB-2 is a unique member ofthe ErbB family in that it transforms fibroblasts evenin the absence of a ligand or an activating mutation.Transformation is observed only with very highexpression driven by strong promoters. A carcinogen-induced mutation that replaces a valine for glutamicacid in the transmembrane domain of rat ErbB-2/Neuis transforming even at low expression levels, probablybecause its kinase is constitutively active. Consistentwith an oncogenic role, monoclonal antibodies toeither rat or human ErbB-2 can inhibit tumorigenicgrowth of ErbB-2-transformed cells probably due toaccelerated endocytosis. Recent clinical trials suggestthat such serotherapy utilizing ErbB-2-specific anti-bodies may be clinically useful (41). Along with atransformed phenotype, experimental overexpressionof the erbB-2 gene also led to cellular resistance tothe cytotoxic effect of tumor necrosis factor and con-ferred an increased metastatic potential (42). Althoughno mutation analogous to the rat neu oncogene hasbeen detected in human tumors, overexpression of theerbB-2 gene is a common genetic event that appearsto confer a selective advantage to several types ofcarcinomas. Overexpression, with or without geneamplification, has been detected in 15-30% of primarybreast tumors. It is associated with inflammatory ductalcarcinomas, rather than noninflammatory tumors, butit is most abundant in DCIS of the comedo type (43).Why most DCIS tumors, whose prognosis is relativelygood, overexpress ErbB-2 is currently an open ques-tion, but it implies that overexpression occurs at anearly stage of breast cancer development. Neverthe-less, DCIS tumors that overexpress ErbB-2 are charac-terized by a relatively high invasive potential. ThatErbB-2 overexpression enhances the metastatic poten-tial of breast tumors is supported by the observationthat the incidence of overexpression is significantlyincreased in patients with 3 or more axillary lymphnodes positive for metastases. Moreover, ErbB-2 over-expression is predictive of poor prognosis particularlyin node-positive patients (2, 44).

Clinical studies revealed another important hall-mark of ErbB-2-overexpressing tumors, namely lowerresponsiveness to adjuvant therapy regimens. Lymphnode-positive patients whose tumors overexpressErbB-2 benefit from high dose chemotherapy, whereaspatients with no overexpression are indifferent to ahigher dose regimens (45). Similar to breast cancer,

overexpression of ErbB-2 was observed in 25-32% ofovarian carcinomas. Again, gene amplificationaccounts for only a portion of the overexpressingtumors, implying that ErbB-2 overexpression, no mat-ter by which mechanism, confers a selective advantageto the tumor. Regardless of whether or not gene ampli-fication was involved, overexpression was correlatedin several studies with poor prognosis and shorter over-all patient survival. Analyses of other types of tumorssuggest that the clinical importance of ErbB-2 mayextend to most carcinomas (46). Approximately 30%of lung adenocarcinomas display overexpression ofErbB-2 associated with advanced stage carcinomasand shorter survival. Similarly, advanced stage gastriccarcinomas overexpress ErbB-2 more than early stagetumors of the stomach; expression correlates withpoorer survival rates. A strong correlation betweenErbB-2 overexpression and poor survival was alsoobserved in cervical carcinomas.

NRG receptors. Because the two NRG receptors,ErbB-3 and ErbB-4, joined the family relativelyrecently, only few reports on their involvement in can-cer are currently available. Transfection experimentsindicate that neither receptor can transform 3T3 fibro-blasts in the presence of NRG, but coexpression ofNRG receptors with either ErbB 1 or ErbB-2 produceda transformed phenotype that is ligand-dependent (47-49). Overexpression of ErbB-3 has been observed athigh frequency in several types of human carcinomas,including breast, colonic, gastric and pancreatictumors, but so far no significant correlation withpatient clinical outcome has been reported. For exam-ple, in 212 node-negative breast tumors, ErbB-3 pro-tein expression was found to be significantly associatedonly with ErbB-2 expression, and no correlation withrelapse-free survival was observed even after a medianfollow-up of 63 months (50).

THE ErbB SIGNALING NETWORK

In vitro analyses of signal transduction by ErbBproteins may provide a functional basis for understand-ing the biological and clinical aspects of the family.Especially important are questions that relate towhether or not ErbB-2 has a ligand of its own, whatis the significance of the defective ErbB-3 kinase, whythere are multiple ligands that share receptor specific-ity, and why ErbB-2 is more oncogenic than othermembers of its family. In many respects signaling byErbB proteins is analogous to the mechanism by which

104 Pinkas-Kramarski, Alroy, and Yarden

other tyrosine kinases transmit hormonal signals intothe cytoplasm and to the nucleus (3). Growth factorbinding to the monomeric binding site induces rapidreceptor dimerization, allowing the two receptors tophosphorylate each other on tyrosine residues. Thesesites function as docking terminals for a large numberof SH2-containing proteins that either couple to othersignaling proteins or by themselves carry an enzymaticactivity. Because binding of SH2 motifs is determinedby the exact amino acid sequences that flank the recep-tor's phosphotyrosine residues and each ErbB proteinis characterized by a unique set of tyrosine phosphory-lation sites, different sets of signaling proteins areexpected to become associated with each receptor. Pro-teins like SHC and GRB-2 apparently associate withall ErbB proteins. On the other hand, two cytoplasmicsubstrates display receptor selectivity. These are thephosphatidylinositol 3'-kinase, whose activationappears to require recruitment of ErbB-3 (51), and aputative negative regulator of the ErbB pathway, theproduct of the c-cbl proto-oncogene. This protein isphosphorylated upon ErbB-1 activation, but not byany other ErbB protein (52). Substrate selectivity isapparently a major determinant of signal diversifica-tion that is shared by all receptor tyrosine kinases.However, the ErbB family appears to utilize an addi-tional diversity generator, namely: ligand-dependentformation of receptor heterodimers, that enables com-binatorial recruitment of various sets of signalingmolecules.

The first heterodimers to be detected were thoseof ErbB-1 with ErbB-2, and their existence explainednot only the efficient tyrosine phosphorylation ofErbB-2 upon cell stimulation with EGF but also asynergy between ErbB-1 and ErbB-2 in cell transfor-mation (53). Similarly, heterodimers between NRGreceptors and ErbB-2 enabled isolation of NRG on thebasis of its ability to increase ErbB-2 phosphorylation,although this ligand does not bind with high affinityto ErbB-2. Evidence for the existence of all 10 differenthomo- and heterodimers of ErbB proteins has beenreported (54,55), including an ErbB-2 homodimer thatmay be stabilized by an oncogenic mutation, bivalentantibodies or overexpression at the plasma membrane.This network of inter-receptor interactions, however,displays a strict hierarchy rather than a random pattern(Fig. 3) (55). ErbB-2 emerges as the preferred partnerof the other three receptors, once they are occupiedby ligands. Importantly, the interactions are directionaland competitive. If given a choice, ErbB-2 prefers tointeract with the kinase-defective NRG receptor, ErbB-

Fig. 3. Inter-receptor interactions in the ErbB family. A. The car-toon depicts a model of ErbB-2/ErbB-3 heterodimer that is boundby one ligand (NDF). The defective tyrosine kinase domain ofErbB-3 is represented by a cross and phosphorylation sites aresymbolized by encircled P letters. Note that upon ligand bindingto ErbB-3 and heterodimer formation, the catalytic activity of ErbB-2 autophosphorylates its own sites, as well as sites on ErbB-3. B.Presumed hierarchy of receptor interactions. The relative strength ofinteractions is represented by the width of each arrow. Arrowheadsindicate the directional nature of each ligand-driven interaction.Note that major interactions occur between ErbB-3 and ErbB-2(unidirectional) and ErbB-1 and ErbB-4 (bi-directional).

ErbB Receptors and EGF-Like Ligands 105 

3. The ErbB-2/ErbB-3 heterodimer appears to be themost mitogenic receptor combination, probably due tothe many docking sites on ErbB-3, which becomephosphorylated only in trans, by ErbB-2 (47). Thismechanism, that is illustrated in Fig. 3B, provides dualcontrol of the potent mitogenic action of ErbB-3: bothNRG binding and heterodimerization with ErbB-2must occur. In the absence of ErbB-2 other interactionsoccur, the most notable are formation of ErbB-1/ErbB-4 and ErbB-l/ErbB-3 heterodimers, that are selectivelyformed by NRG. In addition to the relative strength oftrans-phosphorylation events, this hierarchy explainswhy NRG inhibits in trans binding of EGF to ErbB-1 (56). In fact, the thermodynamic driving force forthe whole network appears to be higher ligand bindingaffinity to heterodimers, as compared with homodi-mers and receptor monomers. Especially important areErbB-2-containing heterodimers because they displayhighest affinity for both EGF and NRG (57, 58). Thereason for this high affinity ligand binding is thathomo- and especially heterodimers of ErbB-2 are char-acterized by a very low rate of ligand dissociation (k^)compared with the rate of release from other receptorcombinations. From a signal transduction point ofview, this property of ErbB-2 makes it a very potentreceptor because it can significantly prolong cell stimu-lation by all ErbB ligands (59, 60). In addition, dueto a potent ability of ErbB-2 to activate phosphatidyl-inositol turnover by recruitment of phospholipase Cy

as well as to spark the linear cascade of kinases leadingto stimulation of Erk kinases, heterodimers that containErbB-2 are extremely active.

Definite resolution of the biological activity ofindividual receptor pairs requires co-expression in acellular environment that lacks endogenous expressionof all four members of the family. Several importantlessons were learned from experiments with ErbB-expressing Baf/3 and 32D myeloid cells, that arestrictly dependent on interleukin-3 for survival (19,54). A graded order of relative strength of mitogenicsignals was observed, in which heterodimers weremore potent than homodimers and ErbB-2-containingheterodimers displaying maximal activity. Asexpected, ErbB-3 homodimers displayed no activity,but co-expression of either ErbB-2 or ErbB-1 reconsti-tuted a proliferative signal. Interestingly, homodimersof ErbB-4 displayed minimal mitogenic activity, evenin the presence of NRG and ErbB-3, but co-expressionof either ErbB-1 or ErbB-2 reconstituted mitogenesisin response to NRG. In cell transformation assays,NRG only caused proliferation of ErbB-4-expressing

cells, but in the presence of ErbB-1 or ErbB-2 it causedcell transformation (49). Thus, whereas EGF can trans-mit mitogenic signals also through ErbB-1 alone, amitogenic effect of NRG appears to depend on thepresence of more than one type of ErbB molecules.

An additional layer of control that augments sig-nal diversification is related to the multiplicity of ErbBligands. Comparison of the biological activities of thea and (3 isoforms of NRGs revealed that the (3 isoformis usually more potent, probably because of its higherreceptor affinity (20). However, a qualitative differ-ence was observed when myeloid cells expressing twocombinations of ErbB-3 were assayed: although bothisoforms were mitogenic to cells expressing an ErbB-2/ErbB-3 combination, only the (3 isoform was mito-genic in cells bearing ErbB-3 and ErbB-1 (15). Theseobservations imply that different ligands of ErbB pro-teins may have distinct abilities to recruit heterodimer-izing partners to the direct receptor. In the case of themultiple ErbB-1 ligands, which include seven mamma-lian molecules and three viral proteins, this model maybe even more important. For example, different ErbB-1 ligands have a different potential to stimulate trans-phosphorylation of ErbB proteins (11) and TGFa isa superior agonist compared to EGF in a series ofbiological assays.

PERSPECTIVES AND OPEN QUESTIONS

It is clear by now that the ErbB family providesan enormous diversification potential for inter- andintracellular signaling. Presumably, this diversity playsan instructive role that is needed for specifying themany cell lineages of the mammalian nervous andepithelial systems. Consistent with this possibility,lower organisms, like Drosophila and C. eleganswhose cellular heterogeneity is more limited, expressonly one ErbB protein. Pivotal to the combinatorialsignaling by the ErbB family is ErbB-2. This receptoremerges as a pan-ErbB signaling subunit, whose func-tion is analogous to nonligand binding subunits oflymphokine receptors (e.g., gp130). As such, ErbB-2may not function as a direct receptor for any growthfactor. Consequently, overexpression of ErbB-2 in cer-tain tumor cells may confer a growth advantagebecause it favors formation of the more mitogenicheterodimers, especially the superior ErbB-2/ErbB-3combination. The structural basis of these inter-recep-tor interactions remains unknown. While the existenceof a dimerization site on each receptor is possible,

106 Pinkas-Kramarski, Alroy, and Yarden

bivalency of the relatively small EGF-like ligands is anobvious alternative. Regardless of the exact molecularmechanism, it must explain why the interactions arelimited to the ErbB family. Other subfamilies of RTKsmay display signaling networks, but these also appearto be limited to intra-family interactions. How exten-sive are inter-receptor interactions in other subfamiliesof growth factor receptors, and to what extent they arerelated to the ErbB network remain to be seen.

REFERENCES

1. D. J. Slamon, G. M. Clark, S. G. Wong, W. J. Levin, A. Ullrich,and W. L. McGuire (1987). Human breast cancer: correlationof relapse and survival with amplification of the HER-2//neuoncogene. Science 235:177-182.

2. D. J. Slamon, W. Godolphin, L. A. Jones, J. A. Holt, S. G.Wong, D. E. Keith, W. J. Levin, S. G. Stuart, J. Udove, A.Ullrich, and M. F. Press (1989). Studies of the HER-2/neuproto-oncogene in human breast and ovarian cancer. Science244:707-712.

3. P. van der Geer, T. Hunter, and R. A. Lindberg (1994). Receptorprotein-tyrosine kinases and their signal transduction pathways.Ann. Rev. Cell Biol. 10:251-337.

4. H. Toyoda, T. Komursaki, D. Uchida, Y. Takayama, T. Isobe, T.Okuyama, and K. Hanada (1995). Epiregulin, a novel epidermalgrowth factor with mitogenic activity for rat primary hepato-cytes. J. Biol. Chem. 270:7495-7500.

5. D. J. Riese, Y. Bermingham, T. M. van Raaij, S. Buckley, G.D. Plowman, and D. F. Stern (1996). Betacellulin activatesthe epidermal growth factor receptor and erbB-4, and inducescellular response patterns distinct from those stimulated byepidermal growth factor or neuregulin-beta. Oncogene12:345-353.

6. E. Peles and Y. Yarden (1993). Neu and its ligands: from anoncogene to neural factors. BioEssays 15:815-824.

7. M. A. Marchionni, A. D. J. Goodearl, M. S. Chen, O. Berming-ham-McDonogh, C. Kirk, M. Hendricks, F. Denehy, D. Misumi,J. Sudhalter, K. Kobayashi, D. Wroblewski, C. Lynch, M.Baldassare, I. Hiles, J. B. Davis, J. J. Hsuan, N. F. Totty, M.Otsu, R. N. McBury, M. D. Waterfield, P. Stroobant, and D.Gwynne (1993). Glial growth factors are alternatively splicederbB-2 ligands expressed in the nervous system. Nature362:312-318.

8. D. L. Falls, K. M. Rosen, G. Corfas, W. S. Lane, and G. D.Fischbach (1993). ARIA, a protein that stimulates acetylcholinereceptor synthesis, is a member of the neu ligand family.Cell. 72:801-815.

9. A. Orr-Urtreger, L. Trakhtenbrot, R. Ben-Levi, D. Wen, G.Rechavi, P. Lonai, and Y. Yarden (1993). Neural expressionand chromosomal mapping of Neu Differentiation factor Cto8pl2-p21. Proc. Natl. Acad. Sci. U.S.A. 90:1867-1871.

10. D. Wen, S. V. Suggs, D. Karunagaran, N. Liu, R. L. Cupples,Y. Luo, A. M. Jansen, N. Ben-Baruch, D. B. Trollinger, V. L.Jacobson, T. Meng, H. S. Lu, S. Hu, D. Chang, D. Yanigahara,R. A. Koski, and Y. Yarden (1994). Structural and functionalaspects of the multiplicity of Neu differentiation factors. Mol.Cell Biol. 14:1909-1919.

11. R. R. Beerli and N. E. Hynes (1996). Epidermal growth factor-related peptides activate distinct subsets of ErbB receptors anddiffer in their biological activities. J. Biol. Chem.271:6071-6076.

12. E. G. Barbacci, B. C. Guarino, J. G. Stroh, D. H. Singleton,K. J. Rosnack, J. D. Moyer, and G. C. Andrews (1995). Thestructural basis for the specificity of epidermal growth factorand heregulin binding. J. Biol. Chem. 270:9585-9589.

13. N. E. Jacobsen, N. Abadi, M. X. Sliwkowski, D. Reilly, N. J.Skelton, and W. J. Fairbrother (1996). High-resolution solutionstructure of the EGF-like domain of heregulin-alpha. Biochem-istry 35:3402-3417.

14. K. Nagata, D. Kohda, H. Hatanaka, S. Ichikawa, S. Matsuda,T. Yamamoto, A. Suzuki, and F. Inagaki (1994). Solution struc-ture of the epidermal growth factor-like domain of heregulin-alpha, a ligand for pl80erbB-4. EMBO J. 13:3517-3523.

15. R. Pinkas-Kramarski, M. Shelly, S. Glathe, B. J. Ratzkin, andY. Yarden (1996). Neu differentiation factor/neuregulin iso-forms activate distinct receptor combinations. J. Biol. Chem.271:19029-19032.

16. P. M. Guy, J. V. Platko, L. C. Cantley, R. A. Cerione, and K.L. Carraway (1994). Insect cell-expressed p180ErbB3 pos-sesses an impaired tyrosine kinase activity. Proc. Natl. Acad.Sci. U.S.A. 91:8132-8136.

17. K. L. Carraway and L. C. Cantley (1994). A neu acquaintancefor ErbB3 and ErbB4: A role for receptor heterodimerizationin growth signaling. Cell. 78:5-8.

18. J. Baulida, M. H. Kraus, M. Alimandi, P. P. Di Fiore, and G.Carpenter (1996). All ErbB receptors other than the epidermalgrowth factor receptor are endocytosis impaired. J. Biol.Chem. 271:5251-5257.

19. R. Pinkas-Kramarski, L. Soussan, H. Waterman, G. Levkowitz,I. Alroy, L. Klapper, S. Lavi, R. Seger, B. Ratzkin, M. Sela,and Y. Yarden (1996). Diversification of Neu differentiationfactor and epidermal growth factor signaling by combinatorialreceptor interactions. EMBO J. 15:2452-2467.

20. M. Marikovsky, S. Lavi, R. Pinkas-Kramarski, D. Karunagaran,N. Liu, D. Wen, and Y. Yarden (1995). ErbB-3 mediates differ-ential mitogenic effects of NDF/heregulin isoforms on mousekeratinocytes. Oncogene 10:1403-1411.

21. Y. Yang, E. Spitzer, D. Meyer, M. Sachs, C. Niemann, G.Hartmann, K. M. Weidner, C. Birchmeier, and W. Birchmeier(1995). Sequential requirement of hepatocyte growth factorand neuregulin in the morphogenesis and differentiation of themammary gland. J. Cell. Biol. 131:215-226.

22. D. M. Danilenko, B. D. Ring, J. Z. Lu, J. E. Tarpley, D. Chang,N. Liu, D. Wen, and G. F. Pierce (1995). Neu differentiationfactor upregulates epidermal migration and integrin expressionin excisional wounds. J. Clin. Invest. 95:842-851.

23. J. T. Trachtenberg and W. J. Thompson (1996). Schwann cellapoptosis at developing neuromuscular junctions is regulatedby glial growth factor. Nature 379:174-177.

24. R. Pinkas-Kramarski, R. Eilam, O. Spiegler, S. Lavi, N. Liu,D. Chang, D. Wen, M. Schwartz, and Y. Yarden (1994). Brainneurons and glial cells express Neu differentiation factor/here-gulin: a survival factor for astrocytes. Proc. Natl. Acad. Sci.U.S.A. 91:9387-9391.

25. N. M. Shah, M. A. Marchionni, I. Isaacs, P. Stroobant, andD. J. Anderson (1994). Glial growth factor restricts mammalianneural crest stem cells to a glial fate. Cell. 77:349-360.

26. Z. Dong, A. Brennan, N. Liu, Y. Yarden, G. Lefkowitz, R.Mirsky, and K. R. lessen (1995). Neu differentiation factor isa neuron-glia signal and regulates survival, proliferation, andmaturation of rat Schwann cell precursors. Neuron 15:585-596.

27. J. R. Florini, D. S. Samuel, D. Z. Ewton, C. Kirk, and R. M.Sklar (1996). Stimulation of myogenic differentiation by aneuregulin, glial growth factor 2. Are neuregulins the long-sought muscle trophic factors secreted by nerves? J. Biol.Chem. 271:12699-12702.

28. L. M. Moscoso, G. C. Chu, M. Gautam, P. G. Noakes, J. P.Merlie, and J. R. Sanes (1995). Synapse-associated expressionof an acetylcholine receptor-inducing protein, ARIA/heregulin,

ErbB Receptors and EGF-Like Ligands 107

and its putative receptors, ErbB2, and ErbB3, in developingmammalian muscle. Develop. Biol. 172:158-169.

29. J. M. Verdi, A. K. Groves, I. Farinas, K. Jones, M. A. Marchi-onni, L. F. Reichardt, and D. J. Anderson (1996). A reciprocalcell-cell interaction mediated by NT-3 and neuregulins controlsthe early survival and development of sympathetic neuroblasts.Neuron. 16:515-527.

30. D. Meyer and C. Birchmeier (1995). Multiple essential func-tions of neuregulin in development [see comments]. Nature378:386-390.

31. M. Gassmann, F. Casagranda, D. Orioli, H. Simon, C. Lai,R. Klein, and G. Lemke (1995). Aberrant neural and cardiacdevelopment in mice lacking the ErbB4 neuregulin receptor.Nature 378:390-394.

32. K.-F. Lee, H. Simon, H. Chen, B. Bates, M.-C. Hung, and C.Hauser (1995). Requirement for neuregulin receptor erbB2 inneural and cardiac development. Nature 378:394-398.

33. M. Sibilia and E. F. Wagner (1995). Strain-dependent epithelialdefects in mice lacking the EGF receptor. Science 269:234-238.

34. P. J. Miettinen, J. E. Berger, J. Meneses, Y. Phung, R. A.Pedersen, Z. Werb, and R. Derynck (1995). Epithelial immatu-rity and multiorgan failure in mice lacking epidermal growthfactor receptor. Nature 376:337-341.

35. D. W. Threadgill, A. A. Dlugosz, L. A. Hansen, T. Tennenbaum,U. Lichti, D. Yee, C. LaMantia, T. Mourton, K. Herrup, andR. C. Harris (1995). Targeted disruption of mouse EGF recep-tor: effect of genetic background on mutant phenotype. Sci-ence 269:230-234.

36. N. E. Hynes and D. F. Stern (1994). The biology of erbB-2/neu/HER-2 and its role in cancer. Biochem. Biophys. Acta.1198:165-184.

37. D. S. Salomon, R. Brandt, F. Ciardiello, and N. Normanno(1995). Epidermal growth factor-related peptides and theirreceptors in human malignancies. Crit. Rev. Oncol. Hematol.19:183-232.

38. W. Li, J. W. Park, A. Nuijens, M. X. Sliwkowski, and G. A.Keller (1996). Heregulin is rapidly translocated to the nucleusand its transport is correlated with c-myc induction in breastcancer cells. Oncogene. 12:2473-2477.

39. K. Pavelic, Z. Benjac, J. Pavelic, and S. Spaventi (1993).Evidence for a role of EGF receptor in the progression ofhuman lung carcinoma. Anticancer Res. 13:1133-1138.

40. A. J. Wong, J. M. Ruppert, S. H. Bigner, C. H. Grzeschik, P.A. Humphrey, D. S. Bigner, and B. Vogelstein (1992). Structuralalterations of the epidermal growth factor receptor gene inhuman gliomas. Proc. Natl. Acad. Sci U.S.A. 89:2965-2969.

41. J. Baselga, D. Tripathy, J. Mendelsohn, S. Baughman, C. C.Benz, L. Dantis, N. T. Sklarin, A. D. Seidman, C. A. Hudis,J. Moore, P. P. Rosen, T. Twaddell, I C. Henderson, and L.Norton (1996). Phase II study of weekly intravenous recombi-nant humanized anti-p185 HER2 monoclonal antibody inpatients with HER2/neu-overexpressing metastatic breast can-cer. J. Clin. Oncol. 14:737-744.

42. D. Yu, S. Wang, K. M. Dulski, C. Tsai, G. L. Nicolson, and M.Hung (1994). c-erbB-2/neu overexpression enhances metastaticpotential of human lung cancer cells by induction of metas-tatsis-associated properties. Cancer Res. 54:3260-3266.

43. M. van de Vijver, J. L. Peterse, W. J. Mooi, P. Wisman, J.Lomans, O. Dalesio, and R. Nusse (1988). Neu-protein overex-pression in breast cancer. Association with comedo-type ductalcarcinoma in situ and limited prognostic value in stage II breastcancer. N. Engl. J. Med. 319:1239-1245.

44. A. K. Tandon, G. M. Clark, G. C. Charmless, A. Ullrich,and W. L. McGuire (1989). HER-2/neu oncogene protein andprognosis in breast cancer. J. Clin. Oncol. 7:1120-1128.

45. H. B. Muss, A. D. Thorr, D. A. Berry, T. Kute, E. T. Liu, F.Koerner, C. T. Cirrincione, D. R. Rudman, W. C. Wood, M.Barcos, and I. C. Henderson (1994). c-erbB-2 expression andresponse to adjuvant therapy in women with node-positiveearly breast cancer. New Eng. J. Med. 330:1260-1266.

46. I. Stancovski, M. Sela, and Y. Yarden (1994). Molecular andclinical aspects of the Neu/ErbB-2 receptor tyrosine kinase.Cancer Treat Res. 71:161-191.

47. C. Wallasch, F. U. Weiss, G. Niederfellner, B. Jallal, W. Issing,and A. Ullrich (1995). Heregulin-dependent regulation ofHER2/neu oncogenic signaling by heterodimerization withHER3. EMBO J. 14:4267-275.

48. M. Alimandi, A. Romano, M. C. Curia, R. Muraro, P. Fedi, S. A.Aaronson, P. P. Di Fiore, and M. H. Kraus (1995). Cooperativesignaling of ErbB-3 and ErbB-2 in neoplastic transformationof human mammary carcinoma cells. Oncogene 10:1813-1821.

49. K. Zhang, J. Sun, N. Liu, D. Wen, D. Chang, A. Thomason,and S. K. Yoshinaga (1996). Transformation of NIH 3T3 cellsby HER3 or HER4 receptors requires the presence of HER1or HER2. J. Biol. Chem. 271:3884-3890.

50. G. Gasparini, W. J. Gullick, S. Maluta, P. D. Palma, O. Caffo, E.Leonardi, P. Boracchi, O. F. P, N R. Lemoine, and P. Bevilacqua(1994). c-erbB-3 and c-erbB-2 protein expression in node-negative breast carcinoma—an immunocytochemical study.Eur. J. Cancer. 30a:V2.

51. S. P. Soltoff, K. L. Carraway, S. A. Prigent, W. G. Gullick,and L. C. Cantley (1994). ErbB3 is involved in activation ofphosphatidylinositol 3-kinase by epidermal growth factor. Mol.Cell Biol. 14:3550-3558.

52. G. Levkowitz, L. N. Klapper, E. Tzahar, A. Freywald, M. Sela,and Y. Yarden (1996). Coupling of the c-Cbl protooncogeneproduct to ErbB- 1/EGF-receptor but not to other ErbB proteins.Oncogene 12:1117-1125.

53. Y. Kokai, J. N. Myers, T. Wada, V. I. Brown, C. M. LeVea, J.G. Davis, K. Dobashi, and M. I. Greene (1989). Synergisticinteraction of p185c-neu and the EGF receptor leads to transfor-mation of rodent fibroblasts. Cell 58:287-292.

54. D. J. Riese, T. M. van Raaij, G. D. Plowman, G. C. Andrews,and D. F. Stern (1995). The cellular response to neuregulinsis governed by complex interactions of the ErbB receptor fam-ily. Mol. Cell Biol. 15:5770-5776.

55. E. Tzahar, H. Waterman, X. Chen, G. Levkowitz, D. Karuna-garan, S. Lavi, B. J. Ratzkin, and Y. Yarden (1996). A hierarchi-cal network of inter-receptor interactions determines signaltransduction by NDF/neuregulin and EGF. Mol. Cell Biol.16:5276-5287.

56. D. Karunagaran, E. Tzahar, N. Liu, D. Wen, and Y. Yarden(1995). Neu differentiation factor inhibits EGF binding: amodel for trans-regulation within the ErbB family of receptortyrosine kinases. J. Biol. Chem. 270:9982-9990.

57. T. Wada, X. Qian, and M. I. Greene (1990). Intermolecularassociation of the p185neu protein and EGF receptor modulatesEGF receptor function. Cell. 61:1339-1347.

58. M. X. Sliwkowski, G. Schaefer, R. W. Akita, J. A. Lofgren,V. D. Fitzpatrick, A. Nuijens, B. M. Fendly, R. A. Cerione, R.L. Vandlen, and K. L. R. Carraway (1994). Coexpression oferbB2 and erbB3 proteins reconstitutes a high affinity receptorfor heregulin. J. Biol. Chem. 269:14661-14665.

59. D. Graus-Porta, R. R. Beerli, and N. E. Hynes (1995). Single-chain antibody-mediated intracellular retention of ErbB-2impairs Neu differentiation factor and epidermal growth factorsignaling. Mol. Cell Biol. 15:1182-1191.

60. D. Karunagaran, E. Tzahar, R. R. Beerli, X. Chen, D. Graus-Porta, B. J. Ratzkin, R. Seger, N. E. Hynes, and Y. Yarden(1996). ErbB-2 is a common auxiliary subunit of NDF and EGFreceptors: implications for breast cancer. EMBO J. 15:254-264.