Structural Basis for Negative Cooperativity in Growth Factor Binding to an EGF Receptor  Diego Alvarado, Daryl E. Klein, Mark A. Lemmon  Cell  Volume 142, Issue 4, Pages 568-579 (August 2010) DOI: 10.1016/j.cell.2010.07.015 Copyright © 2010 Elsevier Inc. Terms and Conditions

Cell 2010 142, 568-579DOI: (10.1016/j.cell.2010.07.015) Copyright © 2010 Elsevier Inc. Terms and Conditions

Figure 1 An Asymmetric Ligand-Induced s-dEGFRΔV Dimer (A) The (SpitzEGF)2⋅(s-dEGFRΔV)2 dimer is asymmetric. Domains I, III, and IV are blue, yellow, and red, respectively. Domain II is green in the left-hand molecule (IIL) and dark gray in the right-hand molecule (IIR). Bound ligand is magenta. The domain II dimerization arm is labeled. An asterisk marks the amino-terminal part of domain II where asymmetry is most evident. (B) Structure (PDB ID code 1IVO) of the symmetric EGF-induced dimer of the human EGFR extracellular region (s-hEGFR) lacking domain IV (Ogiso et al., 2002), colored as in (A). (C) Overlay of the left (green) and right (red) molecules from the s-dEGFRΔV dimer, using domain I as reference. A double-headed curved arrow illustrates “wedging” apart of domains I and III in the green molecule compared with the red molecule, breaking direct domain I/III interactions detailed in Figure S1, and altering the domain II conformation so that the dimerization arm is substantially reoriented. (D) Overlay of the right-hand molecule from the asymmetric s-dEGFRΔV dimer (red) on unligated s-dEGFRΔV (cyan) from PDB ID code 3I2T (Alvarado et al., 2009), using domain I as reference. (E) Overlay of the two receptor molecules in the human (EGF)2⋅(s-hEGFRΔIV)2 dimer. See Table S1 for crystallographic statistics. Cell 2010 142, 568-579DOI: (10.1016/j.cell.2010.07.015) Copyright © 2010 Elsevier Inc. Terms and Conditions

Figure 2 The Ligand-Binding Sites in the (SpitzEGF)2⋅(s-dEGFRΔV)2 Dimer Are Inequivalent (A) Overlay of the two ligands (gray) in the (SpitzEGF)2⋅(s-dEGFRΔV)2 dimer, illustrating differences in their binding sites (see also Figure S2). The green structure corresponds to the left-hand molecule in Figure 1A, and the red structure to the right-hand molecule. Green arrows denote the ∼3–5 Å shift of the green domain I toward the top left of the figure and the ∼7 Å translation of the N-terminal helix described in the text. A, B, and C loops of the bound ligand are labeled. The upper inset details s-dEGFRΔV side chains that interact with SpitzEGF, highlighting significant changes. The lower inset gives a similar view of domain III interactions, which are only modestly changed. Residues underlined (E400, S401, H433, and E460) are mentioned in the text. (B) Analogous overlay of the two bound ligands in the human (EGF)2⋅(s-hEGFRΔIV)2 dimer from Figure 1B (Ogiso et al., 2002), illustrating similarity of the two binding sites. Most side chains that contact bound ligand overlay very well in this superimposition. Cell 2010 142, 568-579DOI: (10.1016/j.cell.2010.07.015) Copyright © 2010 Elsevier Inc. Terms and Conditions

Figure 3 SpitzEGF Binding to s-dEGFR Yields Curved Scatchard Plots (A) Experimental data for binding of fluorescently labeled SpitzEGF to biotinylated s-dEGFR are well fit by the Hill equation with a Hill coefficient (nH) of 0.31 (red curve), but not a simple hyperbola (black). The inset shows saturation at >6 μM SpitzEGF. Data are representative of over six independent experiments. (B) Scatchard transformation of binding data shown in (A). The characteristic concave-up curvature is fit well by the Hill equation (nH = 0.31)—suggesting negative cooperativity. (C) Data for fluorescent SpitzEGF binding to a dimerization-defective s-dEGFR variant (s-dEGFRdim-arm) are well fit by a simple hyperbolic binding curve (black) or by the Hill equation with Hill coefficient of 1.02 (dashed red curve) , suggesting no cooperativity. Data are representative of over six independent experiments. (D) Scatchard transformation of data shown in (C) yields a straight line, arguing that s-dEGFR dimerization is required for negative cooperativity. Cell 2010 142, 568-579DOI: (10.1016/j.cell.2010.07.015) Copyright © 2010 Elsevier Inc. Terms and Conditions

Figure 4 Half-of-the-Sites Reactivity in s-dEGFRΔV (A) Crystal structure of an s-dEGFRΔV dimer bound to SpitzEGFΔC. Ligand is bound to the left-hand molecule in which domains I and III are “wedged” apart, but not the right-hand receptor molecule, which structurally resembles unligated s-dEGFRΔV. SpitzEGFΔC lacks six amino acids from its C terminus, and binds s-dEGFRΔV with apparent KD = 4.37 ± 0.26 μM, 12-fold weaker than the value of 368 ± 23 nM measured for SpitzEGF (Figure S3). (B) Electron density is shown from a 2Fo − Fc map (blue) contoured at 1.0σ, calculated with model phases from the receptor molecules alone. In the region corresponding to the left-hand binding site in (A), clear density for bound ligand is seen. Cα traces for domains I and III are shown in blue and yellow, respectively, in the density and the small part of domain IV seen is colored red. (C) By contrast, the 2Fo − Fc map suggests no density for bound ligand in the region corresponding to the right-hand binding site in (A). This binding site appears to be vacant in crystals of a SpitzEGFΔC⋅(s-dEGFRΔV)2 dimer. Cell 2010 142, 568-579DOI: (10.1016/j.cell.2010.07.015) Copyright © 2010 Elsevier Inc. Terms and Conditions

Figure 5 Ligand Binding Promotes an Extensive Asymmetric Dimerization Interface (A) Crystallographic dimer of unligated s-dEGFRΔV reported previously (Alvarado et al., 2009), shown surface rendered with individual domains colored as in Figure 1A. (B) Close-up of the unligated s-dEGFRΔV dimer in the domain II region. Disulfide-bonded modules m2–m8 are labeled, as are selected residues that interact across the ligated dimer interface in (D). An arrow marks the location between modules m4 and m5 of the ligand-induced kink (of ∼12°) that allows the amino-terminal region of the left-hand domain II to “collapse” into its right-hand counterpart in (D). (C) Surface-rendered asymmetric (SpitzEGF)2⋅(s-dEGFRΔV)2 dimer, with individual domains and ligand colored as in (A). (D) Domain II region close-up of the (SpitzEGF)2⋅(s-dEGFRΔV)2 dimer. Disulfide-bonded modules m2, m3, and m4 from the left-hand molecule (green) have “collapsed” onto their counterparts in the right-hand molecule (gray), burying 1160 Å2 in an intimate domain II interface. Dimerization arm-mediated contacts are largely unaltered. Cell 2010 142, 568-579DOI: (10.1016/j.cell.2010.07.015) Copyright © 2010 Elsevier Inc. Terms and Conditions

Figure 6 Model for Negatively Cooperative Ligand Binding to s-dEGFR (A–C) Structures and cartoons describe a model for negatively cooperative ligand binding to s-dEGFRΔV. Domains (and ligand) are colored as in Figure 1. (A) Binding of a single ligand either to “preformed” s-dEGFRΔV dimers (which have two identical binding sites) or to s-dEGFRΔV monomers yields the singly ligated dimer shown in (B). (B) Singly ligated s-dEGFRΔV dimers are asymmetric. Binding of SpitzEGFΔC to the left-hand molecule wedges apart domains I and III (blue and yellow), and thus “bends” domain II (green) such that it collapses against its counterpart (gray) in the neighboring right-hand molecule, as in Figure 5D. (C) A second SpitzEGF binds to the singly ligated dimer, and occupies the binding site in the right-hand molecule with no change in s-dEGFR conformation. The intimate dimer interface in (B) restrains domain II in the right-hand molecule, so that domains I and III cannot readily be wedged apart. Thus, the binding event that occurs in going from (B) to (C) involves a compromised set of ligand/receptor interactions as described in Figure 2A, reducing binding affinity (and retaining asymmetry in the doubly ligated dimer). (D) The dimer of human s-hEGFRΔIV formed upon EGF binding is symmetric (Ogiso et al., 2002), with both ligands bound in the same manner (Figure 2B). A symmetric dimer of this sort would form following ligand binding to the dimer in (B) if ligand/receptor contacts were maximized at the expense of contacts in the dimerization interface. Cell 2010 142, 568-579DOI: (10.1016/j.cell.2010.07.015) Copyright © 2010 Elsevier Inc. Terms and Conditions

Figure S1 Changes in Direct Domain I/III Contacts upon Ligand Binding to s-dEGFRΔV, Related to Figure 1 Details of domain I/domain III interactions are shown with the same orientation (using domain III as reference) for unligated s-dEGFRΔV (A), and (for the (SpitzEGF)2⋅(s-dEGFRΔV)2 complex shown in Figure 1A) the right-hand (B) and left-hand (C) molecules. (A) In unligated s-dEGFRΔV (cyan highlights), direct interactions between domains I (blue) and III (yellow) contribute to stabilization of the extended structure and restrain the domain II configuration. An α helix close to the domain I amino terminus makes several direct contacts with a loop (gray) that protrudes from the domain III surface and contains residues 396–405. The side chains of R21 and N22 in domain I form predicted hydrogen bonds with the S401 side chain and the E400 carbonyl oxygen from domain III. R26, F403 and M402 also participate in the domain I/domain III interface. (B) Direct domain I/domain III interactions are slightly remodeled when SpitzEGF (magenta) binds s-dEGFRΔV without wedging apart domains I and III (presumed low-affinity binding to the right-hand molecule in Figure 1A—shown here with red highlights). The s-dEGFRΔV conformation is changed very little from that seen without ligand (A). The relative positions of domains I and III are unaltered, but a remodeled set of domain I/III contacts forms. The side chain and backbone carbonyl oxygen of E400 in the domain III loop protrusion form predicted hydrogen bonds with the side chains of S8 and R26, respectively, in domain I. The R21 and N22 side chains from domain I interact with N100 of the bound SpitzEGF molecule, rather than with domain III as seen in (A), and M402 in domain III now packs against I98 in the bound SpitzEGF molecule. (C) When SpitzEGF binding wedges domains I and III apart, as in the left-hand molecule in Figure 1A (with green highlights in this panel), domain I is significantly relocated with respect to domain III, and a distinct set of domain I/domain III contacts forms. Domain I is shifted to the left in (C) by 7–8 Å compared with (A) and (B). R26 from the N-terminal domain I helix now forms a predicted salt bridge with the E400 side chain from domain III, and the I4 and F403 side chains (in domains I and III, respectively) are now in direct contact (but are >8Å apart in the absence of ligand). The N22 side chain in domain I forms a predicted hydrogen bond with E97 in the bound ligand, and M402 of s-dEGFR contacts the SpitzEGF I98 side chain. Cell 2010 142, 568-579DOI: (10.1016/j.cell.2010.07.015) Copyright © 2010 Elsevier Inc. Terms and Conditions

Figure S2 SpitzEGF Is Positioned Differently on the Domain I and Domain III Surfaces in the Two (SpitzEGF)2⋅(s-dEGFRΔV)2-Binding Sites, Related to Figure 2 (A) Surface representation of domain I of s-dEGFRΔV (gray) with the relative positions of SpitzEGF on its surface in the two binding sites shown in Figure 1A. The green SpitzEGF molecule represents that bound to the left-hand site in Figure 1A, in which domains I and III are wedged apart. The red SpitzEGF molecule represents that bound to the right-hand binding site in Figure 1A, in which domains I and III are not wedged apart. Key residues on the ligand that are involved in receptor binding are shown in stick representation. In this panel, the N- and C-terminal portions of the ligand (amino acids 49–69 and 102–105), which do not contact domain I, have been omitted for clarity. The location of the SpitzEGF B-loop (residues 70–81) on the domain I surface is clearly different in the two binding sites, and the green ligand molecule appears to pack more intimately against domain I in this region—consistent with the increased buried surface area in the left-hand site in Figure 1A. The C-terminal portion of SpitzEGF (residues 81–105) is also quite differently located on the domain I surface in the two binding sites, representing a major difference in ligand environment in the two sites. For example, E97 and N100 contact quite different regions of domain I. (B) Surface representation of domain III of s-dEGFRΔV (gray) showing the relative positions of SpitzEGF as it binds domain III in the left-hand (green) and right-hand (red) sites in Figure 1A. Alterations in the mode of domain III interaction in the two binding sites are concentrated largely in the A-loop (residues 54–69) and C-loop (residues 84–93) as shown, which contact a large loop at the back of domain III in the orientation shown here. By contrast, the SpitzEGF C terminus (C-tail) occupies almost exactly the same location on the domain III surface in the two binding sites. Side-chains from Y95, K96, Y102, and K105 make very similar predicted hydrogen bonds with the receptor in both cases, and the I98 and P104 side chains are involved in almost identical hydrophobic interactions. As discussed in the main text, the C-tails of ErbB/EGFR ligands are primary determinants of ligand-binding affinity and specificity, and this region of SpitzEGF appears to be the primary ‘anchor’ for ligand binding to both sites in Figure 1A. Cell 2010 142, 568-579DOI: (10.1016/j.cell.2010.07.015) Copyright © 2010 Elsevier Inc. Terms and Conditions

Figure S3 Comparison of s-dEGFRΔV-Binding Affinities for SpitzEGF and SpitzEGFΔC, Related to Figure 4 Surface Plasmon Resonance (SPR) curves are shown for binding of different concentrations of s-dEGFRΔV to immobilized SpitzEGF (black) and SpitzEGFΔC (red). Owing to the design of the assay (with immobilized ligand and soluble receptor extracellular region), the data can only be fit to a single class of binding sites at these s-dEGFRΔV concentrations. Immobilized SpitzEGF binds s-dEGFRΔV with an apparent KD in this assay of 368 ± 23 nM. SpitzEGFΔC, which lacks the most C-terminal 6 amino acids (100–105) binds approximately 12-fold more weakly, with an apparent KD of 4.37 ± 0.26 μM. The reported average and standard deviation values were generated from at least three independent measurements. Cell 2010 142, 568-579DOI: (10.1016/j.cell.2010.07.015) Copyright © 2010 Elsevier Inc. Terms and Conditions

Figure S4 The Singly Ligated s-dEGFRΔV Dimer as a Model for Human ErbB Receptor Heterodimers, Related to Figure 6 The asymmetric SpitzEGFΔC⋅(s-dEGFRΔV)2 dimer in Figure 4A and Figure 6B suggests a structural model for heterodimers that are thought to form between the extracellular regions of ligand-bound hEGFR (or human ErbB3) and unligated human ErbB2 (Burgess et al., 2003; Citri and Yarden, 2006). A structural model for a putative asymmetric human ErbB receptor heterodimer is shown. This model was generated by overlaying EGF-bound s-hEGFRΔIV (Ogiso et al., 2002) onto the left-hand molecule of the asymmetric s-dEGFRΔV dimer and unligated sErbB2 (Cho et al., 2003) onto the right-hand molecule—using domain I as the reference point for superimposition in each case. The N-terminal part of domain II in ligand-bound s-hEGFR appears to have ‘collapsed’ onto the corresponding region of unligated ErbB2 in this heterodimer model—as seen in asymmetric s-dEGFR dimers—and this could play an important role in stabilizing heterodimers through formation of an extended asymmetric interface. This hypothetical s-hEGFR/sErbB2 heterodimer structure may also be a reasonable model for heterodimers formed between neuregulin-bound ErbB3 and unligated ErbB2. Cell 2010 142, 568-579DOI: (10.1016/j.cell.2010.07.015) Copyright © 2010 Elsevier Inc. Terms and Conditions