The N-terminal Domains of Neuregulin 1 Confer Signal Attenuation*

Degradation of activated ERBB receptors is an important mechanism for signal attenuation. However, compared with epidermal growth factor (EGF) receptor, the ERBB2/ERBB3 signaling pair is considered to be attenuation-deficient. The ERBB2/ERBB3 ligands of the neuregulin family rely on an EGF-like domain for signaling and are generated from larger membrane-bound precursors. In contrast to EGF, which is processed to yield a 6-kDa peptide ligand, mature neuregulins retain a variety of segments N-terminal to the EGF-like domain. Here we evaluate the role of the N-terminal domain of neuregulin 1 in signaling and turnover of ERBB2/ERBB3. Our data suggest that whereas the EGF-like domain of neuregulin 1 is required and sufficient for the formation of active receptor heterodimers, the presence of the N-terminal Ig-like domain is required for efficient signal attenuation. This manifests itself for both ERBB2 and ERBB3 but is more pronounced and coupled directly to degradation for ERBB3. When stimulated with only the EGF-like domain, ERBB3 shows degradation rates comparable with constitutive turnover, but stimulation with full-length neuregulin 1 resulted in receptor degradation at rates that are comparable with activated EGF receptor. Most of the enhancement in down-regulation was maintained after replacing the Ig-like domain with a thioredoxin protein of comparable size but different amino acid composition, suggesting that the physical presence but not specific properties of the Ig-like domain are needed. This sequence-independent effect of the N-terminal domain correlates with an enhanced ability of full-size neuregulin 1 to disrupt higher order oligomers of the ERBB3 extracellular domains in vitro.

The ERBB family of receptor tyrosine kinases is involved in a broad spectrum of growth control and cell differentiation events. Members of this receptor family in humans include the epidermal growth factor receptor (EGFR, 2 ERBB1), ERBB2 (HER2/Neu), ERBB3 (HER3), and ERBB4 (HER4) (1). Overexpression of ERBB receptors is associated with a number of solid tumor malignancies and often results in constitutive activation and uncontrolled growth regulation (2)(3)(4)(5)(6). Controlled signaling by ERBB receptors is initiated by binding of a ligand to preformed or ligand-induced dimers, and ERBB receptors can form various combinations of homodimers and heterodimers. The heterodimer composed of ERBB2 and ERBB3 is the most potent pair in terms of stimulating cell proliferation (7). ERBB2 is unique in that no ligand has been identified that binds directly to its extracellular domains (ECDs). However, ERBB2 exhibits the strongest cytoplasmic kinase activity of the ERBB receptors (8,9). In contrast, ERBB3 has a catalytically deficient kinase domain, but its ECDs bind many isoforms of the neuregulin (heregulin) family of ligands with high affinity (10). In addition, ERBB3 has a high propensity to self-associate, and oligomers of ERBB3 are destabilized by the binding of neuregulin, thereby favoring the formation of signaling-competent heterodimers with ERBB2 (11,12).
The neuregulin (NRG) family of ligands is encoded by four genes: NRG1, NRG2, NRG3, and NRG4 (13). All NRG ligands share an EGF-like domain, which is both necessary and sufficient for binding to and activating ERBB receptors (14,15). The EGF-like domain of NRG1 ligands has been shown to be structurally highly homologous to EGF (16). NRG1 and NRG2 ligands bind to both ERBB3 and ERBB4, whereas NRG3 and NRG4 only bind to and activate ERBB4 (17,18). Whereas little is known about NRG2, NRG3, and NRG4, NRG1 has been extensively studied. More than 15 NRG1 isoforms, which result from alternative splicing of a single gene, have been identified (19). These isoforms can be divided into three types (I, II, or III), based on their N-terminal segments (Fig. 1A). NRG1 ligands of type I (heregulin; Neu differentiation factor; acetylcholine receptor-inducing activity (ARIA)) contain an Iglike domain and a glycosylation-rich segment (15, 20 -22). Type II isoforms (glial growth factor) also contain an Ig-like domain but lack the glycosylation-rich segment (23). Type III isoforms (sensory and motor neuron-derived factor) lack both the Ig-like domain and glycosylation-rich segment but contain a cysteine-rich domain of a size comparable with the Ig-like domains of type I and II NRG1s (24). Variations in the C-terminal portion of the EGF-like domain of NRG1 differentiate subtypes (␣, ␤1, ␤2, ␤3) and convey preferential binding to either ERBB3 or ERBB4. All data presented here use recombinant and nonglycosylated NRG1-␤1 with or without N-terminal domains. This subtype is known to bind preferentially to ERBB3. NRG1 isoforms are either generated from short transcripts leading to directly secreted ligands or are synthesized as transmembrane precursor proteins (14,25). The membrane-bound precursors undergo cleavage between the EGF-like domain and the transmembrane domain. The result is a soluble NRG1 ligand containing both the N-terminal segments and the EGFlike domain, equivalent to NRG1 ligands obtained by direct secretion. However, direct activation of cells through cell-cell contacts between receptor-expressing cells and cells expressing membrane-bound NRG1 has also been demonstrated (26). For comparison, EGF is also synthesized as a 190-kDa transmembrane precursor protein that, in contrast to NRG1, undergoes cleavage both N-terminal and C-terminal to the EGF domain, producing a small peptide ligand (ϳ6 kDa) for EGFR (27)(28)(29) (Fig. 1B). An exception for EGFR ligands is heparin-binding EGF-like growth factor. Similar to NRGs, heparin-binding EGF-like growth factor retains a 45-amino acid N-terminal domain with heparan sulfate binding affinity (30). To better distinguish the different NRG1 ligands used in this paper, and unless a general reference to the NRG1 ligand is intended, we will refer to NRG1 containing the Ig-like domain as Ig-NRG1 and refer to the EGF-like domain of NRG1 as NRG1 176 , denoting the N terminus of this artificial peptide ligand (Fig. 1C).
The EGF-like domain of NRG-1 has been reported to be sufficient for the basic activation of ERBB2/ERBB3 heterodimers. Furthermore, the similarity of NRG1 176 to EGF in terms of size and structure underscores the structural and functional similarities between their target receptors, EGFR, ERBB3, and ERBB4. As a result, most studies involving NRG1 have been carried out using NRG1 176 or comparable peptide ligands. However, the N-terminal segments of NRG1 are consistently retained in all isoforms in vivo, with the exception of a small fraction of NRG1 type III, which undergoes an additional cleavage event, leaving an N-terminal portion of reduced size. This suggests that the retention of the N-terminal segments of NRG1 may reflect a functional conservation despite wide variability of these N-terminal domains on the primary sequence level. One example of a functional benefit conferred by the N-terminal Ig-like domain has been reported for NRG1-␤1 stimulation of acetylcholine receptor transcription in myotubes (31). In this case, the ability of the Ig-like domain to bind heparan sulfates facilitates the enrichment of ligand on the cell surface, resulting in an enhanced growth stimulation response at low ligand concentrations.
Following initial activation events, ligand-induced internalization and degradation of receptor-ligand complexes serves as an important mechanism for attenuating receptor-mediated signaling alongside nondegradative processes, such as dephosphorylation by phosphatases. In response to receptor activation, EGFR and EGF are rapidly internalized through the endocytic pathway and are sent to lysosomes for degradation (32,33). However, ERBB2, ERBB3, and ERBB4 have been reported FIGURE 1. Schematic diagram of NRG1 and EGF processing. A, NRG1 isoforms differ with respect to their N-terminal segments, represented by boxes 1-3. The EGF-like domain is marked E, and the transmembrane region is marked T. The classification of segments 1, 2, and 3 is outlined in the table for NRG1 of types I, II, and III. The annotated segments are of variable size and can represent structural domains or short, functionally annotated sequence segments. NC, not classified; Ig-like, immunoglobulin-like domain; GR, glycosylation-rich; CRD, cysteine-rich domain; TM, secondary transmembrane region. The observed range in molecular weights for the different isoforms is listed to the right of each diagram. The indicated secondary cleavage of type III NRG1 occurs for only a fraction of the ligand, following insertion of the secondary transmembrane region in segment 2. B, EGF is also synthesized as a large transmembrane precursor but undergoes successive cleavage events at both its N and C terminus, producing a 6-kDa soluble ligand (adapted from Ref. 29). C, schematic diagram of NRG1 176 , Ig-NRG1, and Trx-NRG1 drawn to scale within the indicated domains and linker regions.
to display impaired internalization compared with EGFR and hence to show reduced signal attenuation. A study comparing the EGF-driven endocytosis of EGFR with chimeras containing the extracellular domain of EGFR fused to the cytoplasmic domains of either ERBB2, ERBB3, or ERBB4 showed that the chimeras internalized 125 I-labeled EGF at a much slower rate than EGFR (34). Using the EGF-like domain of NRG1 as a ligand, the rate of NRG1 endocytosis in cells expressing various combinations of ERBB receptors was reported to be much slower than the internalization rate of EGF in cells expressing EGFR (7,35).
In this study, we evaluate the role of the N-terminal segment of NRG1 with respect to receptor activation, heterodimerization, and down-regulation of ERBB3 and ERBB2. We show that whereas the initial dimerization and phosphorylation events are comparable for Ig-NRG1 and NRG1 176 , the presence of the Ig-like domain is important in accelerating the removal of activated ERBB3 and ERBB2, primarily through enhanced degradation in the case of ERBB3, thus conferring enhanced signal attenuation to the ERBB2/ERBB3 system. The enhanced downregulation observed for Ig-NRG1 was also observed with an unrelated N-terminal fusion protein of similar size, indicating that the physical presence but not the amino acid composition of the Ig-like domain is responsible for its effect on receptor down-regulation. In an in vitro assay, the presence of domains N-terminal to the EGF-like domain of NRG1 correlates with enhanced destabilization of ERBB3 oligomers. We propose that the N-terminal domains of NRG1 ligands, maintained in different variations, are functionally conserved and are required to improve signal attenuation in the otherwise attenuation-deficient ERBB2/ERBB3 system.
Co-immunoprecipitations-MCF7 cells were seeded at 1,600,000 cells/100-mm dish and then serum-starved in RPMI 1640 for 16 h, 24 h after seeding. Cells were then stimulated with 10 nM NRG1 176 or Ig-NRG1 for 30 min at 4°C. Following simulation, cells were washed with PBS, lysed in mild lysis buffer, and further homogenized by five passages through a 26-gauge needle. Following incubation at 37°C for 5 min to ensure the disruption of Triton-resistant lipid fractions and mild sonication, insoluble material was removed by centrifugation for 10 min at 14,000 ϫ g. Immunoprecipitation was carried out with 200 g of lysate with 30 l of a slurry of the anti-ERBB3 (C-17)-agarose resin. The immunoprecipitated material was washed three times with 10 ml of cold mild lysis buffer and eluted from the resin with SDS-PAGE sample buffer. Following SDS-PAGE, immunoblot analysis was performed with anti-ERBB2 and anti-ERBB3 as the primary antibodies and goat antirabbit IgG-HRP as the secondary antibody.
Cell Surface Biotinylation-MCF7 cells were prepared as described under "Tyrosine Phosphorylation Assays." Following a 3-h serum starvation or without starvation, cells were pretreated with 10 nM NRG1 176 or Ig-NRG1 or left untreated in RPMI 1640 with or without 10% FBS for 30 min at 4°C. After two washes with PBS containing 0.1 mM CaCl 2 and 1 mM MgCl 2 (PBS-CM), the cells were labeled with 1 mg/ml Sulfo-N-hydroxy-succinimido-LC-Biotin in either PBS-CM or RPMI 1640 ϩ 10% FBS containing 10 nM ligand two times for 20 min at 4°C. The labeling reaction was stopped with 100 mM glycine in PBS, pH 8.5, plus 5 mM EDTA for 5 min at 4°C, and the cells were washed two times with RPMI 1640 and then incubated in RPMI 1640 with or without 10% FBS with 10 nM ligand or medium alone for 0 -9 h at 37°C. Cells were lysed in SDS lysis buffer and boiled for 15 min. Approximately 100 g of each lysate was diluted 1:10 with mild lysis buffer and incubated with 100 l of a slurry of immobilized streptavidin-agarose resin for 1 h at room temperature. Proteins bound to the resin were washed three times with 10 ml of PBS containing 0.1% SDS and eluted with SDS-PAGE sample buffer. SDS-PAGE and immunoblot analysis were performed with anti-ERBB2 or anti-ERBB3 as a primary antibody and goat anti-rabbit IgG-HRP as the secondary antibody.
ERBB3 Down-regulation Assay-MCF7 cells were prepared as described for the tyrosine phosphorylation assays and then stimulated with 10 nM NRG1 176 , Ig-NRG1, or Trx-NRG1 in RPMI 1640 or incubated with RPMI 1640 alone for 0 -9 h at 37°C. Cells were washed in PBS and lysed in SDS lysis buffer. Lysates were directly evaluated for ERBB3 levels by immunoblot analysis with anti-ERBB3 and goat anti-rabbit IgG-HRP.
Prolonged Ligand Exposure-MCF7 cells were seeded as described for co-immunoprecipitations and then starved with RPMI 1640 with 2% FBS for 24 h. Following starvation, cells were incubated with either NRG1 176 , Ig-NRG1, or Trx-NRG1 at 10 nM or were left untreated in RPMI 1640 with 2% FBS for 24 h at 37°C. Cells were lysed in SDS lysis buffer and boiled for 15 min, and total protein concentrations were determined using a BCA protein assay kit. SDS-PAGE was carried out on 30 g of each lysate. The lysates were then evaluated for tyrosine phosphorylation and ERBB3 levels by immunoblot analysis with anti-phosphotyrosine-HRP or anti-ERBB3 followed by goat anti-rabbit IgG-HRP, respectively. Each experiment was done in duplicate. Autoradiograms were quantified using densitometry analysis.
Growth Stimulation Assay-MCF7 cells were serum-starved in RPMI 1640 for 24 h before the experiment. Following starvation, cells were seeded at 4000 cells/well in a 96-well plate in RPMI with 2% FBS alone or containing 10 nM NRG1 176 or Ig-NRG1 and incubated at 37°C for 3 days. After 3 days, cell proliferation was measured using an XTT (({2,3-bis(2-methoxy-4-nitro-5-sulfophenyl)-5-[(phenylamino)carbonyl]-2H-tetrazolium hydroxide})) assay. In this assay, the cells were incubated with 0.3 mg/ml XTT and 5 g/ml phenazine methosulfate in RPMI for 4 h at 37°C. The reaction was stopped using 3.3% SDS, and the cells were incubated at 37°C for 2 h. Cell growth was calculated using absorbance readings at 490 nm. Each experiment was done six times.
ERBB3-ECD Oligomer Disruption Assay by Chemical Crosslinking-V5 and His 6 epitope-tagged ERBB3-ECD was diluted with PBS or with PBS containing a 5-fold molar excess of Ig-NRG1, NRG1 176 , or bovine serum albumin control at a final ECD concentration of 100 nM. After a 30-min incubation at room temperature, a 100-fold molar excess of bis(sulfosuccinimidyl)suberate was added to each construct, and samples were incubated for an additional 2 h at room temperature. The crosslinking reaction was quenched with Tris base (1 M) and spermidine (100 mM). Samples were run on 4 -15% gradient gels, transferred to polyvinylidene difluoride membranes, and probed with anti-V5-HRP or anti-His 5 -HRP antibodies.
ERBB3-ECD Oligomer Disruption Assay by ERBB3-ECD Pull-down-ERBB3-H565F-ECDs carrying a C-terminal S-tag were preincubated in PBS for 15 min at room temperature at a final concentration of 0.50 M. A total of 1.25 ϫ 10 Ϫ11 mol of ECD were subsequently immobilized on the S-protein-agarose resin that was blocked with 1 mg/ml bovine serum albumin (equivalent of 50 l of initial S-protein resin suspension) by incubation for 30 min at room temperature. The samples were centrifuged for 2 min at 1000 ϫ g, and unbound S-tag ECDs were removed by two washes with S-tag wash buffer (20 mM Tris, 150 mM NaCl, 0.1% Tween 20, pH 7.5). After immobilization of the ECDs, free S-tag binding sites on the resin were saturated with excess S-tagged S-peptide, and the resin was washed twice with PBS to remove excess S-peptide and ECDs that were potentially displaced. Following washes, the immobilized ECDs were incubated for 15 min with PBS or the different indicated ligands at a final concentration of 1 M. The eluted samples and resin after elution were probed with S-protein conjugated with HRP.
Surface Plasmon Resonance-Binding experiments were performed on a BIAcore 3000 (BIAcore AB) at 25°C as described previously (12) but with immobilized receptor instead of ligand. This precludes the analysis of small ligands but facilitates the direct comparison of different ligands of sufficient size. To ensure that ERBB3-ECDs immobilized as spatially separated monomers and to protect the ligand binding site from damage by chemical modification (known to occur with amine-modifying reagents), ERBB3-ECD (36 nM) was immobilized in the presence of 75 nM Trx-NRG1. Coupling was carried out in HEPES (25 mM, pH 6.0) on a BIAcore CM5 chip functionalized with N-ethyl-NЈ-(dimethylaminopropyl)carbodiimide N-hydroxy-succinimide (EDC/NHS) to introduce N-hydroxysuccinimide esters. The reference flow cell was coated with Trx-NRG1 only. Injections of ligands were carried out at multiple concentrations and flow rates to rule out mass transfer effects, and the presented K D values were obtained by direct fitting of dissociation on and off rates to individual surface plasmon resonance curves.
Homology Model of ERBB3 Dimer-An approximate homology model of ligand-bound ERBB3 was generated using Swiss-Model (36). This model is a composite based on the dimer of domains I-III of EGFR with bound ligand (37,38) and the structure and relative orientation of domain IV in the extended but ligand-free form of ERBB2 (39), since domain IV is absent in the ligand-bound EGFR crystal structure. The design of this EGFR/ERBB2 chimeric structure is aided by the high degree of structural homology for the individual domains of ERBB receptors. The sequence of ERBB3-ECD was aligned to the sequence of the EGFR/ERBB2 chimera and was generated using Swiss-Model. The EGF-like domain of neuregulin was aligned and modeled to transforming growth factor-␣ by the same method. In order to provide a graphic representation of the location and relative size of the Ig-like domain, the Ig-like domain with the predicted closest match to the Ig-like domain of NRG1 was identified using the UCLA/Department of Energy fold server (40,41). The sequence with the highest score (Protein Data Bank code 1qp1) was used as a structural and sequence template for a model of the Ig-like domain. However, due to large variations in the length of loop regions between Ig-like domains, the generated "structure" can only serve as model for the relative size of the Ig-like domain and is unlikely to have much predictive value beyond this application.

The N-terminal Ig-like Domain of NRG1 Does Not Influence
Heterodimerization and Tyrosine Phosphorylation-The EGFlike domain of NRG1 is considered to be sufficient for activation of ERBB2/ERBB3 tyrosine phosphorylation. We confirmed that both Ig-NRG1 and NRG1 176 performed in a comparable manner in the MCF7 test system used for our analysis. Stimulation of MCF7 cells with increasing concentrations of Ig-NRG1 or NRG1 176 for 30 min at 37°C resulted in comparable total tyrosine phosphorylation of ERBB receptors (Fig. 2, A and B, top (done at 4°C)). Whereas Ig-NRG1 resulted in a slightly earlier onset in tyrosine phosphorylation, both ligands achieved maximal phosphorylation at concentrations of 7-10 nM. When analyzed for the receptor-specific phosphorylation of either ERBB2 or ERBB3, both NRG1 ligands stimulated tyrosine phosphorylation of the individual receptors at comparable levels ( Fig. 2B, center). A comparison of the ability of both ligands to stimulate ERBB2/ERBB3 heterodimer formation in MCF7 cells showed comparable complex formation for both ligands, although coimmunoprecipitation of ERBB2 seemed marginally stronger with NRG1 176 (Fig. 2B, bottom), whereas matching amounts of ERBB3 were precipitated in mild lysis conditions in the presence of either ligand. Therefore, receptor complex formation and the initial tyrosine phosphorylation response induced by Ig-NRG1 and NRG1 176 are comparable in MCF7 cells, consistent with reports that the EGF-like domain of NRG1 is sufficient for activation (15).
Ig-NRG1 Enhances the Down-regulation of Activated ERBB2 and ERBB3-Next, we evaluated the lifetime of the tyrosine phosphorylation signal generated by stimulation with either NRG1 176 or Ig-NRG1. The lifetime of the total tyrosine phosphorylation signal was reduced when stimulated with Ig-NRG1 compared with NRG1 176 (Fig. 3A), and the plateau reached with NRG1 176 over the course of 4 -5 h was ϳ10% higher compared with stimulation with Ig-NRG1. Overall, the half-life of tyrosine phosphorylation following stimulation with Ig-NRG1 was ϳ1 h, whereas stimulation with NRG1 176 resulted in a halflife of ϳ2 h. In order to evaluate the specific contribution of ERBB2 and ERBB3, we compared receptor-specific tyrosine phosphorylation for both ligands after 3 h, the time point that reproducibly showed the largest ligand-specific differences (Fig. 3B). Both ERBB2 and ERBB3 tyrosine phosphorylation is diminished more after 3 h of stimulation with Ig-NRG1. However, with respect to receptor phosphorylation, the difference was more pronounced for ERBB2. However, when the same lysates were analyzed for receptor degradation, again, Ig-NRG1 proved to be more potent than NRG1 176 , but its impact was more pronounced for ERBB3 (Fig. 3C).
For Ig-NRG, the reduced lifetime of tyrosine-phosphorylated ERBB3 correlated well with the overall steady-state levels of ERBB3 after ligand stimulation. In order to evaluate if the shorter half-life of the tyrosine phosphorylation signal and steady-state receptor levels of ERBB3 is reflected in a shorter half-life of the cell surface receptor, we measured the half-life of ERBB3 that resided on the cell surface at the time of stimulation. To this end, cell surface receptors were labeled with biotin, and the lifetime of biotinylated ERBB3 was monitored in the presence and absence of Ig-NRG1 or NRG1 176 . The half-life of biotinylated ERBB3 in the absence of ligand was ϳ4 h (Fig. 4A,  Ctrl), representing the rate of constitutive turnover of cell surface ERBB3. In the presence of NRG1 176 , the half-life of ERBB3 was very similar to that of unstimulated ERBB3 (ϳ3.5-4 h). These results are consistent with previous reports that the halflife of ERBB3 changes very little upon ligand stimulation (34). However, in the presence of Ig-NRG1, the ERBB3 receptor was degraded at rates faster than constitutive turnover, which is consistent with the observed shorter half-life of tyrosine phosphorylation after Ig-NRG1 stimulation. The down-regulation of ERBB3 in the presence of Ig-NRG1 appears to be biphasic, with a rapid drop between 1 and 3 h, followed by a slow turnover of the residual receptors. Since the analysis of the half-life of tyrosine phosphorylation revealed that all but a minute amount of residual signal had been removed by 4 h, the receptors showing slower turnover are likely to represent ERBB3 receptors that were not stimulated and hence show only constitutive turnover. This incomplete activation is likely to be a reflection of the fact that MCF7 cells express a modest excess of ERBB3 over ERBB2 (42). Therefore, the biphasic nature of ERBB3 degradation upon stimulation with Ig-NRG1 may reflect the rapid downregulation of activated ERBB3 (t1 ⁄ 2 ϭ 1-2 h), followed by the slower constitutive degradation of nonactivated ERBB3 (t1 ⁄ 2 ϭ 4 h). A comparison of the turnover of cell surface ERBB2 and ERBB3 at 3 h (Fig. 4B) is in good agreement with the changes in steady-state levels (Fig. 3C), showing increased down-regulation for both receptors with Ig-NRG1 but only a relatively small increase in the rate of turnover for cell surface ERBB2. Treatment with Ig-NRG1 results in a slightly earlier onset in tyrosine phosphorylation, but both ligands achieve comparable maximal tyrosine phosphorylation at concentrations of 7-10 nM. B, tyrosine phosphorylation of ERBB2 or ERBB3 and receptor heterodimerization at 10 nM ligand. MCF7 cells were untreated (Ctrl ) or treated with 10 nM NRG1 176 or Ig-NRG1 for 30 min at 4°C. Immunoblots were carried out with the antibodies listed to the left after immunoprecipitations (IP) with antibodies indicated on the right (where applicable). For the phosphorylation analysis of ERBB2 and ERBB3 (center), cells were lysed in SDS-lysis buffer, and the immunoprecipitation was done after 1:10 dilution of the SDS to avoid coimmunoprecipitation of receptors. Coimmunoprecipitations were carried out in mild lysis buffer. All experiments were carried out in duplicate.

The Enhanced Attenuation for Ig-NRG1 Is Not Specific to the Ig-like Domain-For
Ig-NRG1, part of the enhanced attenuation and signaling potency, in terms of onset of full activation ( Fig. 2A), is likely to be due to the heparan sulfate binding properties of its Ig-like domain. However, not all NRGs contain heparan sulfate binding domains, but all NRGs retain an N-terminal domain of some sort. We therefore evaluated if the simple presence of an N-terminal domain contributes to the enhancement of signal attenuation. Specifically, we asked if an N-terminal thioredoxin fusion of NRG1 (Trx-NRG1) could substitute for the Ig-like domain in the enhancement of receptor down-regulation. This comparison is facilitated by the fact that the N-terminal thioredoxin fusion protein is very different in composition but similar in size to Ig-NRG1 (Fig. 1C). Trx-NRG1 was initially constructed to aid in the expression of NRG1 176 in Escherichia coli but has been shown in several cases to be a suitable substitute for NRG1 in stimulation assays and to show comparable binding to ERBB3 and ERBB2/ERBB3 heterodimers (17,43). In order to ensure that the properties of Ig-NRG1 in our test system were not due to differences in affinity, we measured the affinity of Trx-NRG1 and Ig-NRG1 by plasmon surface resonance and compared the binding constants with published values ( Table 1). The binding constant for the commercially obtained NRG1 176 was not directly accessible by this method due to the small size of the ligand; however, binding constants for both NRG1 176 and Trx-NRG1 are available in the literature for comparison. The binding constant obtained for Trx-NRG1 in our analysis matches that of previously published values and is comparable with that of Ig-NRG1 as well as the published values for NRG1 176 . Therefore, any FIGURE 3. Ig-NRG1 stimulation results in faster decrease of activated ERBB2 and ERBB3 levels. MCF7 cells were incubated in the presence of 10 nM NRG1 176 or Ig-NRG1 for the indicated times. A, total receptor tyrosine phosphorylation, shown as ␣-pY in the Western blot (IB) and graphic representation of scanned results, decreases faster after stimulation with Ig-NRG1 (f) compared with NRG1 176 (E). B, the difference in the decrease of receptor-specific tyrosine phosphorylation is more pronounced for ERBB2 than ERBB3. Shown is anti-receptor Western blot after 30-and 180-min ligand stimulation followed by ␣-pY immunoprecipitation of denatured receptors and quantitative presentation of the average of data obtained in triplicate. C, receptor down-regulation by Ig-NRG1 is more pronounced for ERBB3. Cell lysates were probed before (0 min) or after (180 min) stimulation with the indicated ligands. Total tyrosine phosphorylation levels (at 30 and 180 min of stimulation) are shown for comparison. Bar diagrams illustrate the percentage of total tyrosine phosphorylation remaining at 180 min relative to 30 min (pY; white) and the percentage of ERBB2 (gray) and ERBB3 (black) remaining at 180 min relative to unstimulated levels. Experiments were carried out in duplicate, and graph representations represent the average of both data sets. differences in the biological activity between Ig-NRG1 and NRG1 176 or possibly Trx-NRG1 are not due to significant differences in its affinity for ERBB3.
In order to evaluate the impact of replacing the Ig-like domain with thioredoxin, we measured the decrease in cellular steadystate levels of ERBB3 after stimulation of MCF7 cells with the different ligands (Fig. 5A). ERBB3 levels decreased rapidly and at a rate that matched the degradation of cell surface ERBB3 following treatment with Ig-NRG1 (Fig. 4A). Likewise, the drop in steadystate levels after treatment with NRG1 176 was comparable with the corresponding rate observed for biotinylated cell surface ERBB3 as well as the rate of unchallenged turnover of the receptor on the cell surface. The drop in steady-state levels after Trx-NRG1 stimulation was intermediate but closer to that observed for Ig-NRG1, suggesting that the physical presence of the Ig-like domain, but not its specific composition, is sufficient to convey much of the enhanced downregulation. The failure of a stoichiometric mixture of Trx and NRG1 176 to elicit the same enhancement in the drop of tyrosine phosphorylation levels demonstrates that the physical linkage of both functionally unrelated proteins is causal for the enhanced signal attenuation (Fig. 5B). A comparison of phosphorylated extracellular signal-regulated kinase 1/2 levels (Fig. 5B) did not show significant differences on the time scale at which differences at receptor degradation and phosphorylation were observed.
Since all signal attenuation assays of either receptor degradation (Figs. 4B and 5B) or removal of tyrosine phosphorylation signal (Fig. 3B) had consistently indicated a difference in the plateau level after 4 -5 h, we asked if this difference would be perpetuated during prolonged stimulation. When MCF7 cells were treated with the different NRG1 ligands for 24 h, the resulting steady-state levels of ERBB3 were lower than in unstimulated cells in all cases (Fig. 5C), and ERBB3 levels after prolonged stimulation with Ig-NRG1 were indeed 10 -15% lower than the steady-state levels for ERBB3 observed in NRG1 176 -treated cells. This difference is relatively small but consistent with the magnitude of the difference in plateau levels observed previously (Figs. 3, 4B, and 5B). Treatment with Trx-NRG1 also resulted on average in lower ERBB3 levels, but the associated error makes these results less informative. The latter may be due to the relative instability of the artificial fusion protein in serum after prolonged incubation, resulting in varying ratios of Trx-NRG1 and NRG1 176 -like cleavage products (data not shown).
As a further readout of prolonged stimulation with NRG1 176 compared with Ig-NRG, we evaluated the ability of both ligands to stimulate the proliferation of MCF7 cells (Fig. 5D). This assay involved the growth of MCF7 cells in the presence of ligand for 3 days. In this assay, Ig-NRG1 showed stronger growth stimulation than NRG1 176 . As was the case for the determination of ERBB3 levels, data obtained with Trx-NRG1 showed variations that were large relative to the observed differences for NRG1 176 and Ig-NRG (data not shown), probably due to the instability of the fusion protein over time.
Enhanced Down-regulation after Stimulation with Ig-NRG1 Correlates with an Increased Ability to Disrupt ERBB3-ECD Interactions-Our previous studies on the interactions of ERBB receptors (11,12) demonstrated that the ECDs of ERBB3 are exceptionally prone to forming higher order oligomers through

The N-terminal Domains of NRG1 Confer Signal Attenuation
the further association of receptor dimers. Furthermore, the binding of ligand destabilizes these higher order oligomers but not ERBB3 dimers, which are intrinsically unstable (44) and are neither destabilized nor stabilized by NRG1 (12). Ligand-independent oligomerization of ERBB3 and the disruption by NRG1 has been observed for purified ERBB3-ECDs in solution and full-length ERBB3 on the surface of insect cells (11) as well as ERBB3/ERBB2 chimera in Chinese hamster ovary cells (12). Our current working model for the mode of disruption of higher order oligomers by NRG1 is based on steric interference by the ligand with larger oligomeric assemblies and is consistent with the structure-based model, which places the ligand on the exterior of receptor dimers (Fig. 6). We had previously observed efficient disruption of higher order oligomers of ERBB3-ECDs using both Trx-NRG1 and NRG1 176 , although higher concentrations of NRG1 176 were required for efficient disruption (12). A comparison of NRG1 176 with Ig-NRG1 in an ERBB3-ECD cross-linking assay (Fig. 7A) shows that Ig-NRG1 is also more potent than NRG1 176 to disrupt oligomeric interactions of ERBB3-ECDs in this assay. The overall formation of higher order cross-linked species is very inefficient on statistical grounds, but the observed covalent cross-linking of oligomers is reduced efficiently by Ig-NRG1 but not at intermediate levels of NRG1 176 (500 nM ligand and 100 nM ECD). Lower ECD concentrations or higher NRG1 176 concentrations are needed to achieve disruption by NRG1 176 (data not shown).

FIGURE 5. The physical presence but not the biochemical composition of the Ig-like domain is important to mediate enhanced down-regulation.
A, autoradiogram and scan results of steady-state ERBB3 levels in MCF7 cells following treatment with NRG1 176 (E), Ig-NRG1 (f), or Trx-NRG1 (F) for 0 -9 h. Treatment with the unrelated but size-comparable Trx-NRG1 fusion protein can recover much of the enhanced down-regulation, conveyed by Ig-NRG1. B, the enhanced decrease in tyrosine phosphorylation seen after Ig-NRG1 stimulation occurs also for Trx-NRG1 but not for a stoichiometric mixture of Trx and NRG1 176 . Within the same time frame, neither ligand displays significant differences in levels of phosphorylated extracellular signal-regulated kinase 1/2. C, compared with NRG1 176 and unstimulated control cells (Ctrl), extended (24-h) exposure to Ig-NRG1 or Trx-NRG1 results in a small but detectable decrease in the steady-state levels of ERBB3. Data represent the average of three independent measurements. D, sustained stimulation of MCF7 cells with Ig-NRG1 results in enhanced growth stimulation, compared with NRG1 176 (data are averages of two sets of triplicates, and the p value (Student's t test) for the difference between growth stimulation by both ligands is p Ͻ 0.0001 (n ϭ 6)). IB, immunoblot.
To evaluate the ability of Ig-NRG1 to disrupt ERBB3-ECD oligomerization in an alternative assay that is not dependent on cross-linking efficiency and better suited for quantitative comparisons, we measured the release of indirectly immobilized ECDs from S-protein-agarose resin. In this previously described assay (12), oligomeric S-tagged ECDs are immobilized on a resin that is only sparsely populated with S-protein. Under those conditions, a large portion of ERBB3-ECDs are immobilized indirectly (i.e. through their secondary interaction with ECDs that are bound directly to the resin via the high affinity interaction of their S-tag). Provided that all free S-protein moieties are subsequently blocked with an excess of S-tagged blocking protein (S-tagged glutathione S-transferase in our case), the addition of ligand results in the release of all indirectly bound ECDs, thereby providing a readout for oligomerization and its ligandmediated disruption.
A sample data set, comparing the ability of Ig-NRG1 and NRG1 176 to elute indirectly bound ERBB3-ECDs from the resin, is shown in Fig. 7B, which shows the amount of ERBB3-ECD eluted through ligand challenge (top) versus the amount of ECD remaining on the resin (bottom). For quantitation purposes, and to correct for variations in load between experiments, the elution efficiency is expressed as the fraction of eluted ECD relative to the total ECD (eluted plus residual bound). This fraction is shown in the graph in Fig. 7B as an average of three experiments. This analysis shows that NRG1 176 -mediated elution is marginally higher than unchallenged dissociation in PBS (Ctrl) but significantly lower than the elution obtained with Ig-NRG1, consistent with a model of enhanced disruption of higher order oligomers by Ig-NRG1.

DISCUSSION
In this paper, we examine the contribution of the N-terminal domain of NRG1 to ERBB receptor down-regulation. We show that the previously described deficiency of down-regulation for the ERBB2/ERBB3 system (7,34) is, especially with respect to ERBB3, in large part correlated with the absence of the N-terminal domains in artificial peptide ligands that represent only the EGF-like domain of NRG. On the one hand, our analysis confirms that the EGF-like domain of NRG1 is both necessary and sufficient for binding and activation of target ERBB3-ERBB2 heterodimers, and the presence of the additional N-terminal Ig-like domain of NRG1 does not significantly alter potency in terms of the initial tyrosine phosphorylation response. Furthermore, Ig-NRG1 and NRG1 176 stabilize heterodimers of ERBB2 and ERBB3 to a comparable extent, although slightly more ERBB3 was coimmunoprecipitated with ERBB2 following stimulation with the EGF-like domain alone (NRG1 176 ) (Fig. 2). It is not clear from our study to what extent this difference may reflect the stability of heterodimers in a cellular context or the stability and stoichiometry of higher order complexes in the presence of the artificial peptide ligand.
However, beyond these similarities in the immediate outcome of stimulation, we observed significant differences in signal attenuation in terms of both receptor degradation and loss of tyrosine phosphorylation. Both processes were significantly accelerated after stimulation with the full-length ligand, especially for ERBB3. In fact, we found the half-life of Ig-NRG1stimulated ERBB3 as well as the associated tyrosine phosphorylation of ERBB2 and ERBB3 to be on the order of 1-2 h, which is comparable with the rate of 1.5 h reported for the turnover of ligand-stimulated EGFR (34). However, in the absence of the N-terminal Ig-like domain, the rate of degradation of ERBB3 following stimulation with NRG1 176 is very similar to the ligand-independent constitutive turnover rate. Interestingly, this enhanced signal attenuation and receptor down-regulation after stimulation with Ig-NRG1 results not only in marginally lower steady-state levels of receptors during prolonged stimulation but also a modest increase in growth stimulation. This observation is in contrast to the assumption that slower turnover will necessarily lead to enhanced growth stimulation but is similar to the observation that a short intermittent activation of the mitogen-activated protein kinase pathway via EGF stimulation of EGFR is growth-stimulatory in PC12 cells, whereas prolonged activation through neural growth factor stimulation results in differentiation (45). A comparison of mitogenactivated protein kinase activation by Ig-NRG on the time scale at which marked differences existed in receptor signal attenuation did not show differences between Ig-NRG1 and NRG1 176 in the activation of extracellular signal-regulated kinase 1/2 (Fig. 5B). However, differences may exist on a shorter time scale, and a further analysis may also require insight into the impact of the N-terminal domains of NRG1 on routing and subcellular localization of activated recep- Bound NRG1 176 as well as the Ig-like domain do not interfere directly with the dimerization interface but would interfere with the secondary interaction of receptor dimers with other receptor complexes. The stable ERBB3/ERBB2 heterodimer is expected to assume a similar conformation. Direct structural information for an ERBB dimer is only available for domains I-III of EGFR with bound transforming growth factor-␣, providing the basis for the model of ERBB3-ECD domains I-III and the EGF-like domain of NRG1. Domain IV and its relative orientation to domain III was modeled based on the structure of ERBB2-ECD in the extended conformation. The distance of 100 Å between the N termini of EGF in an EGFR dimer or the N termini of the EGF-like domain of NRG1 in this model is indicated in the figure.
tors. It is conceivable that differences in subcellular localization of activated receptors supersede the lower steady-state levels after Ig-NRG1 stimulation in terms of favoring a growth-stimulatory response.
A comparison of the differences in receptor degradation compared with the loss of tyrosine phosphorylation for ERBB2 and ERBB3 also shows a very pronounced difference in the preferred route of signal attenuation for both receptors. In the case of ERBB2, dephosphorylation without receptor degradation appears to be significantly more pronounced. This is consistent with previous reports that show a very slow degradation in ERBB2 receptor levels after stimulation of MDAMB435 cells with full-length NRG1, whereas ERBB2 tyrosine phosphorylation dropped rapidly (46). In contrast, ERBB3 receptor degradation correlates well in our studies with the loss of tyrosine phosphorylation when Ig-NRG1 was used as the ligand, but not when cells were stimulated with NRG1 176 (Fig. 3, B and C). These apparent differences between ERBB2 and ERBB3 in the preferred mode of signal attenuation are also reflected in very basic differences in adapter protein recruitment. Whereas activated ERBB3 features several specific and high affinity Src homology 2 and phosphotyrosine-binding protein (PTB) sites, ERBB2 features primarily low affinity, low specificity sites (47).
The differences in down-regulation may therefore reflect a differential control of alternate subsets of adapter proteins, recruited by both activated receptors.
How might the presence of the Ig-like domain of NRG1 confer increased signal attenuation capabilities? Specific properties of the Ig-like domains, such as the binding of heparan sulfates, are implicated in promoting an enhanced signal response at low ligand concentrations. Indeed, we observed a slight increase in the potency of Ig-NRG1 over NRG1 176 at low ligand concentrations (Fig. 2), as was reported previously (31). However, with respect to the impact on receptor down-regulation, no specific biochemical property of the Ig-like domain appears to be needed to confer at least most of the enhancement in downregulation, and the Ig-like domain can be replaced by a heterologous domain of similar size but completely different sequence and structure (Fig. 5).
We did, however, observe a correlation between the enhanced ability of Ig-NRG1 compared with NRG1 176 to disrupt higher order oligomers of ERBB3 and the enhanced down-regulation of ERBB3. Our previous studies demonstrated that this disruption does not directly affect the dimer interface but rather the formation of higher order oligomers (12). This is consistent with the observation that the initial tyrosine phosphorylation response was not affected by the N-terminal domains of NRG1 and is illustrated in Fig. 6, which shows a homology model, outlining the relative orientation of two ERBB receptors in a receptor dimer bound to NRG1 176 or Ig-NRG1. From this model, it is apparent that the binding of NRG1 would not directly interfere with the dimer interface, which is intrinsically unstable for ERBB3 dimers and stable for ERBB2/ERBB3 heterodimers. However, the binding of NRG1 would sterically interfere with the formation of larger order oligomers from dimers, and the presence of an Ig-like or similar domain would greatly enhance this interference. How this interference with larger order complexes would change the properties of the ligand-bound receptors on the cell surface in a way that alters down-regulation requires further investigation. In fact, the complete disruption of ERBB3-ECD interactions by Ig-NRG1 seen in vitro, for fulllength ERBB3 receptors in insect cells (11), and for recombinant ERBB3/ERBB2 chimeras in Chinese hamster ovary cells (12) is much less pronounced for epitope-tagged ERBB3 in a mammalian cell context, such as MCF7 (data not shown). This would argue that a rearrangement of ECD interactions rather than a complete physical disruption of ERBB receptor complexes may be enforced by the presence of the Ig-like domain of Ig-NRG1.
Recent FRET studies on EGFR may provide some insight into the observed disruption of oligomers and its consequences. Whitson et al. (48) demonstrated that FRET between EGFR bound and N-terminally labeled EGF molecules does not occur in reconstituted and signaling competent dimers of the receptor. This is consistent with the distance of 100 Å between the N termini of both EGF ligands and the Foerster distance of 50 Å in the experimental setup used in this assay. However, FRET was observed in a cellular context. This argued in favor of additional receptor contacts that occurred only in the context of additional components of the cellular machinery and might involve FIGURE 7. Ig-NRG1 has an increased ability to destabilize oligomers of the ERBB3-ECD. A, the formation of cross-linked higher order products is suppressed more efficiently with Ig-NRG1. Cross-linking of 100 nM ECD was carried out in the presence of a 5-fold molar excess of ligand (or bovine serum albumin) in PBS as indicated above the lanes. The position of monomeric ECD, ECD dimers, and a small amount of higher order ECD complex, resulting from multiple cross-linking events, is indicated to the right. The receptor was detected by immunoblotting for its C-terminal His tag. B, Ig-NRG1 is more effective in a quantitative assay in eluting S-tagged but indirectly bound ECDs from S-protein-agarose resin. ECDs were eluted with a 5-fold molar excess of either NRG1 176 or Ig-NRG1 as indicated in a sample data set. The autoradiogram shows the indirectly bound and eluted fraction and the fraction remaining on the resin (directly bound plus indirectly bound that failed to elute). Bar diagrams show the fraction of eluted ERBB3-ECD from experiments in triplicate, calculated as the percentage of eluted protein relative to total (eluted plus remaining) for each experiment. The p value (Student's t test) for the difference in disruption between both ligands is p ϭ 0.0016 (n ϭ 3). a "head to head" interface, with respect to the ligand binding site, that was previously observed as a crystallographic dimer interface (49). A similar model of a ligand-induced transition between dimers and tetramers was recently proposed based on fluorescence lifetime studies of EGFR activation (50). In the case of ERBB3 interactions, the Ig-like domain of NRG1 would fall into this putative head-to-head interface between dimers and might, through modulation of this secondary interaction, alter receptor internalization and down-regulation.