Structural Insights into the Down-regulation of Overexpressed p185her2/neu Protein of Transformed Cells by the Antibody chA21*

p185her2/neu belongs to the ErbB receptor tyrosine kinase family, which has been associated with human breast, ovarian, and lung cancers. Targeted therapies employing ectodomain-specific p185her2/neu monoclonal antibodies (mAbs) have demonstrated clinical efficacy for breast cancer. Our previous studies have shown that p185her2/neu mAbs are able to disable the kinase activity of homomeric and heteromeric kinase complexes and induce the conversion of the malignant to normal phenotype. We previously developed a chimeric antibody chA21 that specifically inhibits the growth of p185her2/neu-overexpressing cancer cells in vitro and in vivo. Herein, we report the crystal structure of the single-chain Fv of chA21 in complex with an N-terminal fragment of p185her2/neu, which reveals that chA21 binds a region opposite to the dimerization interface, indicating that chA21 does not directly disrupt the dimerization. In contrast, the bivalent chA21 leads to internalization and down-regulation of p185her2/neu. We propose a structure-based model in which chA21 cross-links two p185her2/neu molecules on separate homo- or heterodimers to form a large oligomer in the cell membrane. This model reveals a mechanism for mAbs to drive the receptors into the internalization/degradation path from the inactive hypophosphorylated tetramers formed dynamically by active dimers during a “physiologic process.”

p185 her2/neu is one of the four receptor tyrosine kinases of the ErbB family. We initially found that both homodimerization and heterodimerization with other ErbB receptors will induce transphosphorylation of the intracellular domains and result in the downstream signaling for cell proliferation and transformation. Moreover our studies have established that heterodimerization leads to increased signaling and transforming activity (1,2). Significant overexpression of p185 her2/neu results in abnormalities in cell signaling and can cause cell transformation. Early studies from several laboratories found that her2/neu gene was amplified and overexpressed in 20 -30% of breast and ovarian cancers. Breast cancers that have p185 her2/neu overexpressed have a more aggressive course associated with higher relapse rates (3). p185 her2/neu represents the first oncoprotein target amenable for drug intervention and immunotherapy in which disabling the kinase reverses the malignant properties of the transformed cell and renders the tumor sensitive to chemotherapy and radiation therapy (4 -8).
The p185 her2/neu protein possesses a similar architecture to the other three ErbB members of this family. These kinases are type 1 transmembrane proteins and comprise an extracellular domain (ECD) 4 with four subdomains (I/L1, II/S1, III/L2, IV/S2), a single transmembrane helix, an intracellular tyrosine kinase domain, and a C-terminal tail (9). Recent crystallographic studies revealed that the subdomains II and IV contribute to dimerization events of the ErbB receptors (10,11). Monoclonal antibodies that bind the ectodomains of these ErbB proteins have many consequences that can be associated with different epitope regions of the ectodomains. Binding to subdomains II and IV can limit dimerization of p185 her2/neu . The mechanical disabling of the formation of dimeric complex is the therapeutic purpose of certain anti-p185 her2/neu monoclonal antibodies (mAbs), such as the humanized mAb Pertuzumab. Pertuzumab binds near the center of subdomain II and appears able to directly interrupt the dimerization of p185 her2/neu (12). The crystal structure of the Trastuzumab Fab fragment in complex with the p185 her2/neu ECD revealed that Trastuzumab binds to the juxtamembrane subdomain IV of ECD (13). This interaction may also influence dimerization of the p185 her2/neu transmembrane region as well as blocking the proteolytic cleavage of p185 her2/neu ECD, a mechanism that may be relevant to altering p185 her2/neu functionality in vivo (13). p185 her2/neu is expressed at about 40,000 copies/cell in normal tissues, whereas in cancer cells, it can reach 10 6 copies/cell. Down-regulation of p185 her2/neu has been shown for several inhibitory antibodies as a mechanism to dampen p185 her2/neumediated transformation. Although some studies indicated that 4D5 could down-regulate cell surface p185 her2/neu receptor, it was reported that by binding to its epitope in the receptor, Trastuzumab prevented p185 her2/neu ectodomain cleavage mediated by matrix metalloprotease (14). The cleavage produces a shed ectodomain as well as a kinase-active membranebound p95 fragment, which can be detected from advanced breast tumors that are insensitive to Trastuzumab treatment (15). However, degradation of both p185 her2/neu and p95 can be observed in the presence of HSP90 inhibitor (16), suggesting an important role for HSP90 in the prevention of p185 her2/neu internalization and degradation.
Synergistic down-regulation of p185 her2/neu levels on her2/ neu transformed cells have been observed when combinations of antibodies recognizing different epitopes on p185 her2/neu ECD are used (17). A mechanism has been proposed in which the combination of antibodies cross-links the receptors, thus forming larger antibody-receptor lattices to enhance internalization (17,18). A physiologic pathway has been defined in which antibody disables ErbB receptor dimers, leading to the formation of inactivated tetramers and then internalization (19).
Trastuzumab-resistant tumors have been identified after patients achieved an initial response to Trastuzumab-based regimens. Development of new antibody drugs or combinations of ectodomain binding monoclonal antibodies is a reasonable approach to limit the emergence of resistance or even to treat the single antibody-resistant tumor (8,20). The anti-p185 her2/neu antibody chA21 mediates specific inhibitory effects on p185 her2/neu -overexpressed cancer cells (21,22), as well as human breast and ovarian cancer xenograft (23,24). Our previous data revealed that chA21 is a potent down-regulator of p185 her2/neu (22).
Here, we report the x-ray crystal structure of the chA21 single-chain Fv (scFv) in complex with the first 192 residues (designated as EPI) of the N terminus of p185 her2/neu . The complex structure, scFV-EPI, presents the first detailed atomic level description of a monoclonal antibody binding to this region of the p185 her2/neu ectodomain. The study documents that chA21 recognizes an epitope located on the opposite surface of the putative dimerization interface of the subdomain II of p185 her2/neu ECD. Our experiments indicate that interactions with this epitope facilitate chA21 cross-linking of receptors and leads to down-regulation of overexpressed ErbB receptors on the cancer cell surface.

EXPERIMENTAL PROCEDURES
Materials-Trastuzumab and Pertuzumab were obtained from Genentech. Preparation of chA21 and its scFv was described previously (25). The p185 her2/neu antibody Ab-17 (NeoMarkers) contains two monoclonal antibodies, e2-4001 and 3B5, that were raised against the cytoplasmic domain and the C terminus of the receptor, respectively. The polyclonal antibody Ab-1 (NeoMarkers) was generated using an immunogen similar to what was used to obtain 3B5. Polyclonal anti-EGFR antibody (sc-03) and polyclonal anti-ErbB3 antibody (sc-285) were from Santa Cruz Biotechnology. The GAPDH antibody was obtained from KangChen Bio-tech. Horseradish peroxidase-conjugated goat anti-mouse/rabbit IgG and FITCconjugated goat anti-rabbit IgG were purchased from Pierce.
Crystallography-Crystallization of scFv-EPI complex was described previously (25). The complex structure was solved by the molecular replacement method using the program PHASER (26). The structures of the free chA21 scFv (Protein Data Bank (PDB) code 2GJJ) and residues 1-192 of p185 her2/neu ECD (PDB code 2A91) were used as the searching models.  and COOT (28) were employed for refinement and iterative adjustments of the model. TLS (29) refinement was performed using Refmac5. The final model was validated with PROCHECK (30) for stereochemical restraints. All details for structure refinement and model quality are described in Table  1. The atomic coordinates and structure factors have been deposited in the PDB under the code 3H3B.
Cell Culture-The human mammary cancer cell line SK-BR-3, which endogenously expresses epidermal growth factor receptor (EGFR), p185 her2/neu , and ErbB3, was a generous gift from Prof. Jun Wang (University of Science and Technology of China, Hefei, China). The cells were incubated at 37°C, 5% CO 2 in RPMI 1640 supplemented with 10% FBS, 10 units/ml penicillin, 10 g/ml streptomycin (all from Invitrogen).

RESULTS
Overall Structure of scFv-EPI Complex-The structure of scFv-EPI complex was determined to 2.45 Å resolution with R/R free factors ϭ 20.1%/25.5% (Table 1). There are two EPI (Chains A and B) and two scFv (Chains C and D) molecules in the asymmetric unit of the complex crystal (supplemental Fig.  S1). The overall electron density for these four molecules was quite good, and only a few residues on the N and C terminus of each molecule, the linkers (residues 113-134) between the light and heavy chains of both scFv chains, and residues Lys-178, Gly-179, and Ser-180 in Chain B were not observed. Ramachandran plot analysis revealed that only Ala-57 residues in both scFvs fell into disallowed regions. Ala-57 locates at the tip of the complementarity-determining region (CDR) L2 and has been found to participate in forming classic ␥-turn structures seen in immunoglobulins (31). The details of refinement statistics and model quality of the final model are summarized in Table 1.
Among the four molecules in the asymmetric unit, Chains A and C form one antigen-antibody complex, whereas Chains B and D form another. These two complexes are disposed in the asymmetric unit through a non-crystallographic two-fold rotation symmetry axis (supplemental Fig. S1), and their structures are almost identical (root mean square deviation (r.m.s.d.) of 0.705 Å for 405 comparable C␣ atoms)) (supplemental Fig. S2). Therefore, only the complex structure comprising Chains A and C, which has better electron density, is used in the following structural analysis and discussion.
In the complex, the scFv folds into two ␤-barrel domains formed by the V H and V L fragments, respectively (Fig. 1A). In comparison with the structure of the free chA21 scFv that was determined previously (22), both their overall structure (r.m.s.d. of 0.527 Å for 218 comparable C␣ atoms) and the conformations of the six CDRs are almost identical (Fig. 1B). These data indicate that antibody-antigen interactions do not cause obviously structural changes in the chA21 scFv.
EPI consists of the p185 her2/neu ECD subdomain I (residues 1-172) and the first C2 module (residues 173-192) of subdomain II. Subdomain I adopts a ␤-helical structure, and the N terminus of subdomain II folds closely to it (Fig. 1A). The overall structure of EPI in this complex is very similar to that of the corresponding region of the previously described p185 her2/neu ECD (12,13,32).
One distinction exists at surface-exposed residues 99 -112, which is a p185 her2/neu -specific loop not found in other ErbB receptors (supplemental Fig. S3). In p185 her2/neu structures reported previously (12,13,32) this loop could not be fully traced due to its high flexibility, whereas it is quite ordered in our structural model because of its extensive interactions with the scFv (Fig. 1C).
Overview of the Interface of the scFv-EPI Complex-The complex structure indicates that the CDRs of light chain (L) and heavy chain (H) of chA21 scFv form a pocket able to interact with a large area of the C-terminal part of EPI ( Fig. 2A). The antibody-antigen interface buries 953 Å 2 of the scFv surface and 886 Å 2 of EPI, which represent about 8.5% of the total scFv accessible surface area (about 11,251 Å 2 ) and 9.7% of the EPI accessible surface area (9168 Å 2 ), respectively.
The CDRs L1, L2, L3, H1, H2, and H3 contribute 220 (23.2%), 3.3 (0.3%), 93 (9.8%), 71 (7.5%), 263 (27.7%), and 303 Å 2 (32.0%) of the interface, respectively. These values indicate that the heavy chain of chA21 contributes more interface area for EPI binding than the light chain. This is a frequently observed pat-FIGURE 1. Structure of chA21 scFv in complex with p185 her2/neu EPI. A, the stereo drawing of the scFv-EPI complex. In EPI, the subdomain I and the first C2 module of subdomain II are colored green and yellow, respectively. The light chain (V L ) and the heavy chain (V H ) of chA21 scFv are colored blue and orange, respectively. The CDRs of chA21 scFv are labeled. B, the structure superposition of chA21 scFv in the free form (PDB code 2GJJ, magenta) and in the complex (colored the same as panel A). Their overall structures are almost identical. C, the structure superposition of EPI in the free form (PDB code 2A91, red; PDB code 1S78, cyan; PDB code 1N8Z, gray) and in the complex (colored the same as panel A). Their overall structures are very similar, but the p185 her2/neu -specific loop is disordered before the binding of chA21 scFv. The figures were prepared with the program PyMOL (48). tern in many antibody-antigen complexes, which we have described previously (33,34). Particularly, the L2 loop of the light chain does not protrude from the scFv surface and only forms three van der Waals contacts with EPI (supplemental Table S1). A limited if not minimal role of L2 in antibody-antigen interaction is frequently observed, and we as well as others have noted this feature previously (33)(34)(35).
Analysis of the complex structure shows that 28 residues of p185 her2/neu participate in the antibody-antigen interaction. These residues belong to three discontinuous loops, which are labeled as loop I, II, and III from the N terminus to the C terminus, respectively ( Fig. 2A). Loop I (residues 100 -105) is the N-terminal part of the p185 her2/neu -specific loop. It locates nearby the CDRs L3 and H2 of chA21 and contributes 22.9% of the total interface area of the EPI molecule. Loop II (residues 135-144) is behind loop I and III and contributes only about 7.9% of the total interface. The loop III (residues 163-187) contributes 69.3% area of the total interface, which seems to be the major binding region for chA21 scFv.
Interactions between chA21 scFv and p185 her2/neu EPI-In total, the interactions between chA21 scFv and EPI consist of 17 hydrogen bonds and 172 van der Waals contacts (Ͻ4 Å) (supplemental Tables S1 and S2). Neither salt bridges nor bridging water molecules were found. All of the 17 hydrogen bonds are contributed by L1, L3, H2, and H3 of scFv (supplemental Table  S2). As for p185 her2/neu , the loops I, II, and III participate in forming 4, 3, and 10 hydrogen bonds, respectively (supplemental Table S2).
Loop III could be separated into an N-terminal part (residues 163-175) and a C-terminal part (residues 185-187), which are linked by a loop behind these two parts. The N-terminal part of loop III lies on the interface, and the main chain atoms of its residues Arg-166 (III), Cys-170 (III), and Pro-172 (III) contribute six hydrogen bonds altogether (Fig. 2, E and F). The C-terminal part of loop III locates nearby CDR L1, where Glu-185 (III) forms two hydrogen bonds with Tyr-31 (L1) and Asn-33 (L1) of chA21 scFv (Fig. 2F). Additionally, the imidazole ring of His-171 (III) of EPI was sandwiched between the side chains of Tyr-31 (L1) and Tyr-239 (H3) of scFv. The Pro-172 (III) was deeply embedded to the pocket formed by Asn-34 (L1), Lys-36 (L1), Tyr-236 (H3), and Glu-237 (H3) (Fig. 2B). These observations indicate the essential roles of His-171 (III) and Pro-172  Fig. 1. The other parts of chA21 scFv are colored light gray. EPI is colored the same as in Fig. 1, but only the three loops containing the epitope are drawn. B, the three loops containing the epitope of EPI lie on the chA21 surface. The important residues on EPI are shown as sticks and labeled with black characters. The key residues on chA21 are colored the same as in Fig. 1 and labeled with white characters, and the other parts of chA21 are colored light gray. C-F, the detailed hydrogenbonding interaction network between chA21 and EPI. The hydrogen bonds are drawn as gray dashes. The residues are labeled, and characters in the parentheses indicate which CDR (chA21 scFv) or loop (EPI) the residues belong to. In panels B-F, the carbon atoms and the drawing are colored the same as in Fig. 1. The oxygen, nitrogen, and sulfur atoms are colored red, blue, and orange, respectively.
(III) in the interaction affinity and specificity of antibodyantigen interactions and agree with our former mutagenesis results, in which mutations H171A and P172A reduced the affinity ϳ200and 700-fold as compared with the wild-type p185 her2/neu (22). chA21 Binds to the Opposite Surface of the p185 her2/neu Dimerization Interface-The complex structure indicates that the chA21 epitope is located at the C-terminal part of subdomain I and the N terminus of subdomain II of p185 her2/neu ECD. This is an epitope distinct from those described for Trastuzumab and Pertuzumab, which are located at subdomains IV and II, respectively. By recognizing this epitope, chA21 binds to the region of p185 her2/neu located on the opposite side of its dimerization interface (Fig. 3A).
To understand the role of chA21 in dimerization of p185 her2/neu and other ErbB members, we developed a model for a complex comprising a p185 her2/neu ECD homodimer and chA21 scFv. The EGFR homodimer structure (PDB code 1IVO) (11) was used as the template to model the p185 her2/neu homodimer, and the two EGFRs were replaced with two p185 her2/neu proteins (PDB code 1N8Z) (13) by superimposing residues 232-292 of p185 her2/neu to residues 226 -286 of EGFR. The scFv-EPI complex structure model was next superimposed to one p185 her2/neu member of the homodimer.
In this modeled complex, the chA21 scFv does not cause steric conflict that is able to disrupt the homodimer directly (Fig. 3B). The heterodimers of p185 her2/neu with EGFR and ErbB3 were also modeled as described (36), and binding of chA21 would not cause steric conflicts in these heterodimers either (data no shown). These structural observations suggest that chA21 uses a different strategy to inhibit functional tyrosine kinase complexes of p185 her2/neu -overexpressing cancer cells.
Down-regulation of p185 her2/neu by chA21 Requires Bivalency and Is Dose-dependent-Previously, we showed in FACS analysis that p185 her2/neu was internalized from the cell surface by the original murine antibody A21 (22). We investigated receptor down-regulation ability of chA21 in SK-BR-3 cells and compared chA21 with other antibodies. p185 her2/neu was clearly down-regulated after 24 h of treatment with chA21 and was hardly detectable after 48 h (Fig. 4A). Before the treatment reached 24 h, we noted the accumulation of a fragment of about 130 kDa, which appears to be an intermediate product formed during p185 her2/neu degradation (Fig. 4A). The overall level of p185 her2/neu was not significantly affected by Trastuzumab or Pertuzumab (Fig. 4B). Interestingly, the combination of chA21  Fig. 1) binds to the back and top of the p185 her2/neu ECD domain, which is distinct to Trastuzumab (light cyan) and Pertuzumab (brown). B, chA21 scFv binds to the modeled p185 her2/neu homodimer without steric conflict. , and Trastuzumab (100 nM) ϩ Pertuzumab (100 nM), respectively, for 24 h. Western blotting was performed using the anti-p185 her2/neu antibody Ab-1. Cont, control. C, SK-BR-3 cells were treated for 24 h with chA21 in the concentration range of 10 -1000 nM. The levels of p185 her2/neu were detected by Western blotting using Ab-1. D, SK-BR-3 cells were treated with chA21 (100 nM) for the indicated time intervals. The levels of EGFR and ErbB3 were detected by Western blotting using the anti-EGFR antibody (sc-03) and anti-ErbB3 antibody (sc-285), respectively. E, SK-BR-3 cells were preincubated with or without heregulin (HRG) (20 ng/ml) for 1 h at 4°C. Then medium was replaced by fresh RPMI 1640 medium with or without chA21 (100 nM) for 24 h of incubation at 37°C. The levels of ErbB3 were detected by Western blotting using the anti-ErbB3 antibody (sc-285).
with Trastuzumab or Pertuzumab displayed synergistic downregulation effects, whereas the combination of Trastuzumab and Pertuzumab had apparent but less significant effects (Fig.  4B). We have shown that the activity of anti-p185 monoclonal antibodies to reduce the malignant phenotype required bivalent antibodies (4), and we noted that the growth inhibitory activity of chA21 was dependent on its bivalency (22). Here, we also show that the monovalent scFv of chA21 was unable to induce p185 her2/neu down-regulation (Fig. 4B).
It is unlikely that the disappearance of p185 her2/neu after chA21 treatment is due to blocked protein synthesis or cell death. p185 her2/neu is a relatively stable protein with a long halflife. In SK-BR-3 cells, it has been reported that the p185 her2/neu protein level was only reduced by about 20% in 16 h after the protein synthesis was blocked by cycloheximide (37). In our proliferation assay, we did not observe significant cell death after chA21 treatment (data not shown). In addition, comparable GAPDH protein levels (the loading control) in all the samples rule out the possibility of dramatic loss of cells after treatment (Fig. 4).
Another unique feature of chA21-mediated p185 her2/neu down-regulation is its bell-shaped dose-dependent curve. As shown in Fig. 4C, chA21 demonstrated the highest p185 her2/neu down-regulation activity at 100 nM. Down-regulation of p185 her2/neu was less obvious when chA21 concentration was reduced (10 nM) or increased (1000 nM).
Down-regulation of EGFR and ErbB3 in SK-BR-3 Cells by chA21-p185 her2/neu forms heterodimers with other members of the ErbB family (1-3), and we also examined the effect of chA21 on EGFR and ErbB3 levels. Western blotting revealed that EGFR and ErbB3 were also down-regulated by chA21 treatment (Fig. 4D), and the time course of down-regulation of these two receptors was similar to that of p185 her2/neu , with obvious degradation observed after 12 h of treatment. Pretreatment with heregulin, which leads to the formation of p185 her2/neu -ErbB3 heterodimers, further enhanced the downregulation of ErbB3 (Fig. 4E).

DISCUSSION
The elucidation of targeted therapy has changed cancer treatment (4,7). We have tried to define the principles by which ectodomain targeting of ErbB receptors can disable kinase complexes and reverse the malignant phenotype of human tumors in vitro and in vivo.
There are shared biological effects of most p185 her2/neu ectodomain binding monoclonal antibodies. However, some epitopes to which inhibitory antibodies bind may help explain their anti-tumor activities (38 -40). Epitopes for some anti-p185 her2/neu antibody drugs are located at ECD subdomains II and IV. Both Trastuzumab and Pertuzumab recognize what may represent dimerization interfaces of p185 her2/neu .
In this work, the crystal structure of scFv-EPI complex reveals that chA21 recognizes a novel epitope at the top of p185 her2/neu ECD. This unique epitope locates to the opposite side of the dimerization interface. chA21 exploits a related but partially distinct mechanism to mediate p185 her2/neu downmodulation. We propose a cross-linking model as an explanation of the molecular mechanism of action of chA21, which also extends our previous observation of receptor inactivated tetrameric forms induced by antibodies (19).
In cancer cells that overexpress p185 her2/neu and other ErbB receptors, we first reported that p185 her2/neu forms homodimers or heterodimers on the cell surface (2,41). By binding to its epitope in p185 her2/neu , the bivalent chA21 molecule can cross-link two p185 her2/neu receptors in different homodimers to form a large complex (Fig. 5). This cross-linked complex would be efficiently internalized and degraded. Such a model explains perfectly the receptor down-regulation data we observed for chA21.
In a preliminary characterization of such chA21-induced oligomeric complexes, we have analyzed chA21 and p185 her2/neu ectodomain-human Fc fusion protein (ECD-Fc) on non-reducing SDS-PAGE. In the presence of cross-linking reagent bis(sulfosuccinimidyl)suberate, several high molecular weight species can be observed in the sample containing chA21 and ECD-Fc (supplemental Fig. S4). Our current data have outlined some biochemical features regarding the process of forming chA21-induced complexes.
Bivalency-chA21 requires bivalency to function as a crosslinker. The monovalent scFv form of chA21 cannot simultaneously capture two p185 her2/neu molecules. As a result, although the scFv recognizes the same epitope as chA21, it fails to cross-link p185 her2/neu molecules to form a large complex. Detectable down-regulation of p185 her2/neu cannot be observed for scFv (Fig. 4B). We also observed this bivalencydependent feature for the original mAb that down-regulates oncogenic p185 her2/neu (4).
Stoichiometry-A certain stoichiometric ratio between the antibody and the receptor is required. In our experiment, we observed the highest down-regulation when cells were treated with 100 nM chA21. With lower concentration of chA21 (e.g. 10 nM), the large receptor-antibody complex could not efficiently form due to the lack of enough cross-linkers. When the chA21 concentration is too high (e.g. 1000 nM), chA21 levels may lead to self-competition, resulting in binding to single species of p185 her2/neu . A cross-linked complex would not efficiently form in this case (supplemental Fig. S5). This explains the bell- shaped dose-dependent curve (Fig. 4C). To confirm this stoichiometry aspect of chA21-mediated down-regulation, we also studied the effect of chA21 on MCF-7, another cell line with much less expression of p185 her2/neu . As predicted, p185 her2/neu down-regulation was not observed at 100 nM, the concentration optimized for SK-BR-3 (supplemental Fig. S6A). Consistently, treating the SK-OV-3 cells, which also express high levels of p185 her2/neu , with 100 nM chA21 also caused significant downregulation (supplemental Fig. S6B).
Synergistic Effect with Other Antibodies-Our studies have shown that inactive tetrameric receptors can be readily induced by Trastuzumab (19). We expect that in the presence of chA21, these oligomeric structures can be further cross-linked for degradation (Fig. 4B). This would lead to the synergistic effect of the combination of treatment of chA21 with Trastuzumab or Pertuzumab.
Effect on Heterodimers of ErbB Receptors-ErbB receptors other than EGFR are considered to be resistant to ligand-induced downregulation (42,43), and p185 her2/neu can also impose an inhibitory effect on EGFR internalization (44,45). Because the pathogenesis and prognosis of cancers are correlated with abnormal signaling events shared by the activation of members of EGFR family, it seems wise and necessary to disrupt multiple ErbB receptor functions at the same time. However, antibodies such as Trastuzumab and Pertuzumab cannot universally prevent the formation of heterodimers as the dimer interfaces they bind to are only important for certain types of heterodimers (36).
Although EGFR and ErbB3 are not directly recognized by chA21, they can form heterodimers with p185 her2/neu . In SK-BR-3 cells, it has been estimated by a mass spectrometry study that about 10 -20% of p185 her2/neu proteins are associated with EGFR. 5 These heteromers can be cross-linked by chA21 (Fig.  5B), leading to our observation of EGFR and ErbB3 degradation after chA21 treatment (Fig. 4D).
The unique capability of chA21 to down-regulate multiple ErbB receptors may provide a therapeutic opportunity. Acquired resistance to monomeric antibodies has been reported as a result of the up-regulation of active heterodimers of ErbB receptors after anti-EGFR or anti-p185 her2/neu antibody treatment (46,47). It will be of interest to investigate whether this feature of chA21 to target heterodimers of ErbB receptors leads to less acquired resistance in human therapy.
The crystal structure of scFv-EPI complex was determined to 2.45 Å resolution and revealed the detailed antibody-antigen interactions necessary to further improve chA21 for better affinity and specificity. The complex also provides insight into the structural basis for understanding the p185 her2/neu down-regulating activity of chA21. Our experiments indicate that chA21, a bivalent antibody, binds to an epitope of p185 her2/neu that promotes cross-linking to form larger complexes that are internalized and degraded. These studies add to our efforts to delineate the principles needed to rationally target tumor oncoproteins.