Insulin Mimetic Peptide Disrupts the Primary Binding Site of the Insulin Receptor*

Sets of synthetic peptides that interact with the insulin receptor ectodomain have been discovered by phage display and reported in the literature. These peptides were grouped into three classes termed Site 1, Site 2, and Site 3 based on their mutual competition of binding to the receptor. Further refinement has yielded, in particular, a 36-residue Site 2-Site 1 fusion peptide, S519, that binds the insulin receptor with subnanomolar affinity and exhibits agonist activity in both lipogenesis and glucose uptake assays. Here, we report three-dimensional crystallographic detail of the interaction of the C-terminal, 16-residue Site 1 component (S519C16) of S519 with the first leucine-rich repeat domain (L1) of the insulin receptor. Our structure shows that S519C16 binds to the same site on the L1 surface as that occupied by a critical component of the primary binding site, namely the helical C-terminal segment of the insulin receptor α-chain (termed αCT). In particular, the two phenylalanine residues within the FYXWF motif of S519C16 are seen to engage the insulin receptor L1 domain surface in a fashion almost identical to the respective αCT residues Phe701 and Phe705. The structure provides a platform for the further development of peptidic and/or small molecule agents directed toward the insulin receptor and/or the type 1 insulin-like growth factor receptor.

In 2002, sets of synthetic peptides were identified that could compete for insulin binding to the human insulin receptor extracellular region and that had affinities in the high nanomolar to low micromolar range (1). Based on competition studies, the putative epitopes of these peptides were grouped into three non-overlapping sites, termed Sites 1, 2, and 3. 3 Some Site 1 peptides were able to activate the receptor tyrosine kinase and act as agonists in an insulin-dependent fat cell assay, whereas Site 2 and Site 3 peptides were found to act as antagonists both in phosphorylation and fat cell assays. The highest affinity Site 1 peptides contained the motif FYXWF, whereas Site 2 peptides were characterized by either a short or long disulfide loop (1). Optimized versions of the Site 1 and Site 2 peptides were subsequently described (2), in particular those that were linked in tandem in various ways. Several of these linked peptides were found to function as either potent antagonists (e.g. peptide S661, a Site 1-Site 2 combination) or as potent agonists (e.g. peptide S597, a Site 2-Site 1 combination), depending on the order in which the individual Site 1 and Site 2 peptides were linked (2,3).
In earlier studies (4,5), we provided isothermal titration calorimetry (ITC) 4 data that suggested how Site 1 peptides might bind to the insulin receptor. The insulin receptor (IR) itself is a disulfide-linked (␣␤) 2 homodimer; each ␣␤ monomer comprises, from its N terminus, two homologous leucine-rich repeat domains (L1 and L2) separated by a cysteine-rich region (CR) comprising eight disulfide-linked modules (6). These domains are followed by three fibronectin type III domains (FnIII-1, FnIII-2, and FnIII-3), one of which, FnIII-2, contains a 120-residue insert domain that includes the ␣␤ cleavage site (6). C-terminal to FnIII-3 lie the transmembrane helix, intracellular juxtamembrane region, tyrosine kinase domain, and C-terminal tail. The extracellular domains of the homodimer adopt a folded over, ⌳-shaped conformation (6), each "leg" of which consists of the L1-CR-L2 module of one receptor monomer packed against the linearly arranged FnIII-1, FnIII-2, and  FnIII-3 domains of the alternate receptor monomer (Fig. 1A). The primary insulin binding site comprises the central ␤-sheet (L1-␤ 2 ) of the L1 of one IR monomer and the C-terminal segment of the ␣-chain (termed ␣CT) of the alternate monomer (5,7). Our ITC experiments (4,5) showed that the Site 1 component of the optimized Site 2-Site 1 peptide S519 (SLEEE-WAQVECEVYGRGCPS-GSLDESFYDWFERQLG where the hyphen denotes the Site 2/Site 1 junction (2)) bound to an isolated three-domain L1-CR-L2 construct of IR (IR485) (8) with a K d of 11 nM. Entropic considerations led us to conclude further that in so doing the Site 1 peptide was capable of displacing bound exogenous ␣CT peptide from the L1 surface. Such binding would thus disrupt the primary insulin binding site. We noted further that the Site 1 component of S519 has sequence similarity to the insulin receptor ␣CT segment (Fig. 1B). However, to date, no three-dimensional structural data exist to support these conclusions. We describe here the crystal structure of the C-terminal 16-residue Site 1 component of S519 ("S519C16") in complex with the N-terminal L1-CR fragment IR310.T of human IR bound in turn to the Fv domain of the monoclonal antibody 83-7 (7,9). To increase the crystallization propensity of the ternary complex, the IR310.T⅐83-7 Fv complex was subjected to mild endoglycosidase H (EH) treatment prior to the addition of the peptide. Our crystal structure reveals that S519C16 binds to the central ␤-sheet of the L1 domain in a fashion closely similar to that of the ␣CT peptide in the aporeceptor (5), demonstrating that agonistic properties of S519 arise from a disruption of the primary insulin binding site.

Results
Sample Characterization-SDS-PAGE analysis of the EH-treated IR310.T⅐83-7 Fv revealed a small drop in molecular mass (ϳ3 kDa) compared with that of the untreated complex, slightly less than the drop in molecular mass (ϳ5 kDa) upon similar EH treatment of IR310.T in the absence of attached Fv ( Fig. 2A). These data suggested that attachment of the Fv module resulted in a slight reduction in the accessibility of the endoglycosidase to the IR310.T N-linked glycan. The dissociation constant, K d , of the S519C16 peptide for the EH-treated IR310.T⅐Fv 83-7 complex was determined by ITC to be 2.5 Ϯ 0.6 nM ( Fig. 2B; see "Experimental Procedures"), which is similar to that reported (K d ϭ 2.6 Ϯ 0.7 nM) for S519C16 upon ITC against the IR485 construct (4), indicating that its relative affinity was not adversely affected by either removal of the L2 domain, partial glycosylation, or Fv 83-7 complexation. Errors are S.E. in both instances.
Crystallographic Phasing and Refinement-The successful initial phasing of the diffraction data by molecular replacement (see "Experimental Procedures") revealed that the IR310.T⅐83-7 Fv complex was assembled in a fashion closely similar to that seen in the previous structures of which this complex is a subset, in particular that of the intact IR ectodomain in complex with two copies each of 83-7 Fab and 83-14 Fab (10) (Fig. 3). (2mF obs Ϫ DF calc ) difference electron density maps calculated after refinement of the molecular replacement solution revealed a helix-like tube of difference density lying over the hydrophobic trough of the central ␤-sheet of domain L1 of IR310.T. Protuberances compatible with side chains at the resolution of the data set extended from the feature in a pattern suggestive of an ␣-helix. Given the high affinity of the peptide for the IR310.T⅐Fv 83-7 complex, we interpreted this density to be associated with the S519C16 peptide (and indeed there was no other candidate for it from either the buffer or the remaining protein within the crystal). An initial polyalanine ␣-helix was placed objectively into the density feature using the "place-helix-here" tool within Coot (11). Various trials were performed with different origin points for the placed helix; these resulted in helices of very similar respective length (ϳ21 residues) and disposition and with identical N-to-C directions. The extra length in all instances arose from the polyalanine ␣-helix C terminus extending artifactually into unrelated density associated with a neighboring crystallographic monomer. The seven C-terminal residues of the template ␣-helix were therefore deleted. Assignment of the sequence register was guided by the qualitative nature of the side chain-related density features as well as by the amphipathic nature of a helix generated by the S519C16 sequence and its implied engagement with the overwhelmingly hydrophobic trough on the L1-␤ 2 surface. In particular, we noted that the engagement was mediated predominantly by a 5-residue motif of which the first two and last two were presumed hydrophobic by virtue of their positioning within the L1-␤ 2 trough. Within the S519C16 sequence (Fig. 1B), the only such motif is the Site 1-defining FYDWF quintuplet; the 14-mer template thus corresponding to the 14 central residues of S519C16, with the N-and C-terminal glycines then being left unmodeled. The alternative registers resulting from a shift of FIGURE 1. Structures of the IR ectodomain and the insulin mimetic peptides. A, the folded over, ⌳-shaped structure of the IR ectodomain. One monomer is in ribbon representation with domains labeled; the other is in molecular surface, apart from its insert domain (ID), which is in ribbon. B, the sequence similarity between IR ␣CT (residues 694 -710) and S519C16 (residues [21][22][23][24][25][26][27][28][29][30][31][32][33][34][35][36] proposed to underlie their competition for binding the three-domain IR L1-CR-L2 construct IR485 (4,8).
Ϯ1 residue resulted in implausible interactions with the L1 domain. Crystallographic refinement of the structure inclusive of the peptide proceeded according to standard iterative building/refinement protocols. A number of residues within the IR310.T and 83-7 Fv moieties were left unmodeled due to poor or absent electron density (see "Experimental Procedures"). These residues all lay remote from the S519C16 binding site (Fig. 3) and corresponded to either (i) loop regions within IR310.T known to be of relatively high mobility (i.e. B-factor) in other structures of which IR310.T is a subset (e.g. Protein Data Bank code 2HR7 (8)) or (ii) polypeptide termini that arise as artificial truncations with respect to parent protein (i.e. human IR or 83-7 mAb). The final (2mF obs Ϫ DF calc ) difference electron density for the peptide is presented in Fig. 4, and final refinement statistics are presented in Table 1.  The overlay is based on the L1-CR modules. The polypeptide termini of the IR310.T⅐83-7 Fv complex that abut unmodeled residues of the complex are indicated by spheres (red for C termini and blue for N termini); these are remote from the S519C16 moiety. Also absent from the model are the N-and C-terminal glycine residues of S519C16. Detail of the S519C16/IR L1-␤ 2 Interface-The manner in which S519C16 engages the IR L1-␤ 2 surface is shown in Fig.  5A. S519C16 forms an amphipathic ␣-helix, its hydrophobic surface engaging the same trough on the L1-␤ 2 surface as that engaged by the ␣CT peptide within the apo form of the receptor (Protein Data Bank codes 4XST and 4ZXB) (5, 10, 12) ( Table 2). The periphery of this trough is formed by the side chains of IR residues Gln 34 , Tyr 60 , Phe 88 , Phe 89 , Tyr 91 , Arg 118 , Glu 120 , Phe 96 , Phe 64 , Leu 37 , and Arg 14 (in clockwise order), and its floor is formed by the side chains of residues Leu 36 , Leu 62 , and Leu 94 . The side chain of S519C16 Phe 27 5 lies close to IR Leu 94 at one end of the trough, that of S519C16 Phe 31 lies close to IR Leu 62 near the middle of the trough, and that of S519C16 Trp 30 lies close to IR Leu 36 at the opposite end of the trough. The side chains of S519C16 Ser 26 , Tyr 28 , Gln 34 , and Leu 35 engage the periphery of the L1-␤ 2 trough. None of the S519C16 side chains appear to be in van der Waals contact with the floor of the trough as far as can be discerned at the current resolution. Notable further interactions include the parallel stacking of the side chain of S519C16 Phe 27 against the side-chain guanidinium group of IR Arg 118 and the perpendicular stacking of the side chain of S519C16 Phe 31 against the side chains of IR Phe 64 and Phe 96 . Although the interaction of peptide residues C-terminal to Phe 31 with L1-␤ 2 is not extensive, we note that extension of the peptide beyond the FYXWF motif to at least Gln 34 is required for maximum affinity to IR (13); these additional residues may assist in stabilizing the helical conformation of the peptide. Details of the peptide/IR interactions are presented in Table 2.
Comparison with the IR ␣CT/IR L1-␤ 2 Interface-Two structures exist that reveal the disposition of IR ␣CT on the surface of IR L1-␤ 2 in the absence of bound ligand. The first is that of the intact IR ectodomain (Protein Data Bank code 4ZXB), and the second is that of the so-called insulin microreceptor (IR; a complex of IR310.T plus exogenous IR ␣CT peptide; Protein Data Bank code 4XST). Direct superposition of the structure presented here with that of Protein Data Bank code 4ZXB via their L1 domains results in a root mean square difference of 1.1 Å across the C␣ atoms of residues 24 -35 in the S519C16 helix and their respective counterparts in the IR ␣CT helix (i.e. the C␣ atoms of IR residues 698 -709). The equivalent root mean square difference of S519C16 and the ␣CT component of a similarly overlaid Protein Data Bank code 4XST is 2.2 Å. A more detailed comparison of the respective helical axes of S519C16 and that of IR ␣CT is given in Table 3 and illustrated in Fig. 5B. Of the set of L1-␤ 2 residues involved in engaging ␣CT in the aporeceptor structure, almost all play a role here in engaging S519C16 with these residues displaying conserved side chain conformation across the two structures. Furthermore, the manner in which S519C16 residues Phe 27 , Phe 31 , and Leu 35 engage the L1-␤ 2 surface is similar to that in which IR ␣CT residues Phe 701 , Phe 705 , and Leu 709 engage the L1-␤ 2 surface (Fig. 5C).
The total molecular surface area buried by S519C16 as it engages IR L1-␤ 2 is 959 Å 2 compared with 930 Å 2 for the equivalent interaction between IR ␣CT and apo-IR L1-␤ 2 . The shape complementarity, S c (14), of the interface between S519C16 and IR L1-␤ 2 is 0.80 compared with a shape complementarity, S c , of 0.81 for the interface between IR apo-␣CT and IR L1-␤ 2 in Protein Data Bank code 4XST.

Discussion
The interaction of S519C16 with the L1 surface of the insulin receptor aligns with our earlier suggestion that it may function by competition for binding with the native IR ␣CT segment (4) rather than by mimicking the interaction of insulin itself. Competitive displacement of the ␣CT segment from the L1 surface will dramatically reduce the affinity of the receptor for insulin as L1 itself has no measurable affinity for insulin (4). ␣CT displacement from the L1-␤ 2 surface may destabilize the receptor as a whole given that the ␣CT segments are coupled to each other via inter-␣-chain disulfide bond(s) within the Cys 682 -Cys 683 -Cys 685 motif and form a bridge across the pair of L1 domains (10,15).
We note that crystallization of an S519C16⅐IR 310.T complex proved refractory prior to our use of recombinantly derived 83-7 Fv adjunct; the most likely reason for such failure is the relatively high level of surface glycosylation of IR310.T. The adjunct provides additional, non-glycosylated polypeptide surface to the IR310.T complex and in so doing likely increases its propensity for crystallization. As such, it acts as a "crystallization chaperone" (16). In addition, attachment of the Fv was found to maintain solubility of IR310.T after endoglycosidase treatment and thus increase protein purification yield. 83-7 Fv has the advantage over the (hybridoma-derived) 83-7 Fab used in our prior crystallographic studies in that it lacks the (conformationally flexible) hinge region between the Fab variable and constant domains; such hinge regions also have the propensity to decrease crystallizability. Use of Fv domains as crystallization adjuncts is well established in membrane protein crystallography (for example, see Ref. 17), but we are unaware of any attempt to use them within the context of receptor tyrosine kinase crystallization. The manner in which the Site 2 component of S519 binds to the receptor remains enigmatic. However, the now determined location of S519C16 on the L1-␤ 2 surface places constraints on the location of the S519 Site 2 component in the context of the intact receptor. We have undertaken some tentative molecular modeling of the Site 2 component to gain insight into what role their characterizing disulfide loop might play in receptor binding (see "Experimental Procedures"). As may be anticipated, in the three classes of models that emerged from the modeling, the disulfide loop motif is directed toward the 2-fold axis of the receptor (Fig. 6) and not toward the putative second binding site of insulin (i.e. the FnIII-1/FnIII-2 junction) (7). We hence speculate that Site 2 peptides might function by disulfide exchange with either of the two inter-␣-chain IR disulfide bonds (Cys 524 or within the Cys 682 -Cys 683 -Cys 685 triplet) as these are nearby, exposed, and highly mobile. Such exchange is, of course, conjectural. We note, however, that disulfide  Two residues are deemed to be in contact if any pair of their non-hydrogen atoms are within 3.8 Å of each other. Calculation was performed using the programs CONTACT and SC within the CCP4 suite (38  The values reported in each cell of the table are the angle and distance of closest approach between the respective axes of the ␣CT or S519C16 helices after structural overlay of the L1 domains. Angles and distances were computed with the "axes/ planes/centroid" tool within Chimera (39).

Insulin Mimetic Peptide S519
exchange of insulin with its receptor has been reported (18,19), implying a labile disulfide bond in proximity to the insulin binding site. Finally, we note that the agonistic interaction of the S519 helix with the IR L1-␤ 2 surface is reminiscent of that of two other regulatory helices relevant in therapeutic contexts. The first is the interaction of the helix of the MDM2 oncoprotein with the transactivation domain of the tumor suppressor protein p53 (20), and the second is the interaction of individual helices from the proapoptotic Bcl-2 family proteins with a groove on the surface of their prosurvival relatives (21). In the first case, the interaction of the MDM2 helix with a groove on the surface of p53 is mediated primarily by the side chains of a tryptophan residue and a phenylalanine residue in the same (i, i ϩ 4) helical spacing as is observed here for the S519C16 engagement with IR L1-␤ 2 . In the second case, the helix/groove interaction is more extensive and mediated by a set of smaller hydrophobic residues. In both cases, attempts have been made to develop mimetics (either small molecule or stabilized peptide) of the cognate helix with reasonable success (22,23). These examples raise the question whether the S519C16 interaction with IR can be mimicked by non-peptide druglike molecules. Such compounds may find application as agonists of IR in the context of diabetes or antagonists of both IR and type 1 insulin-like growth factor receptor in the context of cancer.

83-7 Variable Heavy (VH) and Variable Light (VL) Domain
Cloning and Expression-Codon-and expression-optimized DNA corresponding to murine monoclonal antibody 83-7 VH chain residues 1-118 (6,9,10) followed by the sequence SLVPRGSSSEQKLISEEDLN (thrombin cleavage site ϩ c-myc tag) was synthesized and then cloned into the vector pcDNA3.1 by DNA2.0. Similarly, DNA encoding the 83-7 VL chain residues 1-112 (6,9,10) followed by the sequence SSDYKD (FLAG tag) was synthesized and then cloned into the vector pJ201 (DNA2.0). Both genes were then individually transferred into the BamHI/XbaI sites (in-frame with a secretion signal) of the plasmid pNCM02 (Takara Bio, Japan) for independent trans-formation into Brevibacillus choshinensis cells (Takara Bio, Japan). Isolated colonies of the transformed B. choshinensis cells were then screened by Western blotting (antibodies 9E10 and M2, respectively) for overexpression of the expected domain. The highest expressing colonies were then stored as glycerol stocks. For 1-liter scale-up, each glycerol stock was used to inoculate 2 ml of 2SY broth containing 10 g/ml neomycin sulfate (Sigma-Aldrich) (2SYnm) followed by incubation overnight at 30°C at 120 rpm. 0.2 ml of these respective cultures was then used to inoculate a further 20 ml of 2SYnm broth, and once sufficiently grown, 5 ml of this inoculum was used to inoculate a further 500 ml of 2SYnm broth in Tunair TM flasks (Sigma-Aldrich). Cultures were incubated for 72-96 h with 1-ml samples taken at 24-h intervals to monitor production via SDS-PAGE and Western blotting. Optical density was monitored at 660 nm. Samples were centrifuged at 13,000 rpm for 5 min to pellet bacteria and recover the supernatant containing the secreted product.
Assembly and Purification of 83-7 Fv-700 ml of conditioned medium containing the c-myc-tagged 83-7 VH domain was combined with 850 ml of conditioned medium containing FLAG-tagged 83-7 VL domain and incubated for 30 min at room temperature followed by addition of 3 M Tris-HCl, pH 8.5, at a ratio of 5 ml/liter of combined media. This process was estimated to give a slight excess of VL monomers in the VL/VH mixture. The pH-adjusted combined media was then run through a 9E10 Mini-Leak low affinity column (Kem En Tec, Denmark) (24), and the desired 83-7 Fv was eluted with c-myc peptide (decameric form) prepared in Tris-buffered saline plus azide (24.8 mM Tris-HCl, pH 8.0, 137 mM NaCl, 2.7 mM KCl plus 0.02% NaN 3 ; TBSA). Fractions were combined with one tablet of cOmplete protease inhibitor mixture (Roche Applied Science). Superdex 200 10/300 size exclusion chromatography (GE Healthcare) fractions examined by SDS-PAGE showed the presence of two bands of molecular mass 14 and 16 kDa, respectively, indicating a correctly formed Fv eluting at 22 kDa. The c-myc tag was then removed from the 83-7 VH domain of the 83-7 Fv as follows. The 83-7 Fv was diluted to ϳ4 mg/ml in TBSA and then combined with 10 mM CaCl 2 and 0.5 unit of human thrombin/mg of 83-7 Fv (Roche Applied Science). The sample was incubated at 37°C for 4 h, and the reaction was stopped by the addition of 1 mM phenylmethylsulfonyl fluoride and incubation on ice. The sample was then repurified by means of a Superdex S75 column (GE Healthcare). Size exclusion chromatography and SDS-PAGE estimates of molecular weight agreed with the expected masses. IR binding activity of 83-7 Fv was confirmed by size exclusion chromatography with SDS-PAGE with IR310.T and 83-7 Fv co-eluting as a single peak (data not shown).
Partial Deglycosylation of IR310.T⅐83-7 Fv complex-IR310.T (produced as described previously (7) and prepared in TBSA) was combined with a 1.25-fold molar ratio of the c-myc tagremoved 83-7 Fv and then incubated with EH (New England Biosciences catalogue number P0702) in G5 buffer (according to the manufacturer's protocol) for ϳ30 h at 37°C at a ratio of 1 mg of IR310.T/10,000 units of EH. Sample pH was immediately thereafter adjusted by addition of 1 ⁄ 20 volume of 3 M Tris-HCl, pH 8.5. The EH-treated sample was then purified by size exclusion chromatography using a Superdex S200 column (GE Healthcare) equilibrated with TBSA buffer to separate the desired form from aggregated complex and to remove EH and excess 83-7 Fv. The final complex was concentrated to 29 mg/ml in TBSA buffer. The extent of deglycosylation was assessed by SDS-PAGE ( Fig. 2A).
Isothermal Titration Calorimetry-S519C16 peptide (synthesized by Genscript and prepared at a concentration of 60 M in TBSA) was titrated against the IR310.T⅐83-7 Fv complex (prepared at a concentration of 6 M in TBSA) using a MicroCal iTC200 instrument (Malvern Instruments, UK). Experiments were conducted at 25°C in "highest quality" mode of the instrument. The volume of the sample placed in the cell was 0.30 ml, and the titrant was injected in 2.52-l volumes over 5.04 s at 3-min intervals with the total number of injections being 16 (the first injection was 1.0 l and judged unreliable). The sample contents were stirred at a speed of 750 rpm over the duration of the titration. S519C16 peptide was first injected into a solution of TBSA alone, and the heat of dilution was found to be a constant, which was then subtracted from the subsequent titration data. All data were analyzed using the MicroCal Analysis software incorporated within Origin 7 using a single site interaction model based on minimization of 2 . Measurements were conducted in triplicate, and the resultant ITC-derived thermodynamic parameters were averaged. A sample titration curve and its integration are shown in Fig. 2B. ITC experiments, wherein 83-7 Fv was titrated against IR485 using protocols similar to those above, yielded a K d of 9.4 Ϯ 0.3 nM, which is comparable with the K d of ϳ40 nM for papain-released 83-7 Fab (6) upon titration against IR485 (data not shown).
Crystallization of the IR310.T⅐83-7 Fv⅐S519C16 Complex-15 l of the above deglycosylated IR310.T⅐83-7 Fv complex was combined with 130 l of S519C16 peptide prepared in 10 mM HEPES-NaOH, pH 7.5, to yield a final complex-to-peptide molar ratio of 1:3. The resultant mixture was then subjected to a 576-condition robotic sparse matrix sitting drop screen (Commonwealth Scientific and Industrial Research Organisation Collaborative Crystallization Centre, Parkville, Australia) to identify hit crystallization conditions. A crystal of ϳ120 m in size was obtained in a 1.75 M (NH 4 ) 2 HPO 4 condition. The crystal was transferred to a solution containing 1.75 M (NH 4 ) 2 HPO 4 ϩ 30% glycerol immediately prior to loop mounting and cryocooling in liquid nitrogen. Diffraction data were collected at 100 K at the MX2 beamline of the Australian Synchrotron (25) and then processed using XDS (build 2014118) (26). Data collection statistics are presented in Table 1. Attempts to improve the initial crystallization conditions via manual hanging drop vapor diffusion protocols were unsuccessful. Structure Determination and Crystallographic Refinement-Initial phases were obtained by molecular replacement using Phaser (27) with an IR310.T⅐83-7 Fv entity used as a search object (excised as a subset of Protein Data Bank code 4OGA (28), which contains as a subset the IR310.T⅐83-7 Fab complex). Phases were then improved by crystallographic refinement using autoBUSTER (version 2.10.2) (29). Difference electron density was apparent on the surface of the L1-␤ 2 sheet of IR310.T in a location approximating that of the receptor ␣CT peptide (5,12) and was interpreted as the S519C16 peptide in an ␣-helical conformation. The direction of the peptide was deduced using the place-helix-here tool within Coot (11), allowing ready model building of the central 14 residues of the S519C16 peptide in a helical conformation into B-factor-sharpened (2mF obs Ϫ DF calc ) difference electron density (5,12). The in-principle Ϯ1-residue register ambiguity was resolved by assessing both the shape of the density side chains and the hydrophobic complementarity to the L1-␤ 2 surface. Carbohydrate residues were identified at five of the six anticipated N-linked glycosylation sites (30) on IR310.T, and these were then rebuilt independently of their conformation in Protein Data Bank code 4OGA. Refinement was carried out using PHENIX (version 1.10pre-2104-1692) (31) and finally autoBUSTER (version 2.10.2) (29) iterated with model building within Coot (version 0.8.1). Reference model restraints were included for IR310.T (to both the A and B chains of Protein Data Bank code 2HR7) and 83-7 Fv (variable heavy chain to Protein Data Bank code 1FNS and variable light chain to Protein Data Bank code 3MBX). Electron density associated with IR310.T residues 1-4, 151-153, 160 -167, 174 -177, 266 -275, and 304 onward, S519C16 residues 21 and 36, 83-7 variable heavy chain residues 117 onward, and 83-7 variable light chain residues 113 onward was poor and did not allow these residues to be convincingly modeled; therefore, they were omitted from the final model. Final refinement statistics are presented in Table 1.
Molecular Modeling-Comparative modeling was performed using MODELLER (version 9.15) (32). First, a model of the S519C16 peptide bound to the IR ectodomain was created by superposing the structure determined here onto the recently improved structure of the ectodomain (10) with the ␣CT peptide (residues 686 -719) of one monomer being excluded to allow for accommodation of S519C16. Models of full-length S519 were then created separately by imposing secondary structural constraints on the S519 peptide (as predicted by PSIPRED (version 3.3) (33), namely that S519 residues Leu 2 -Val 13 and Leu 25 -Leu 35 are ␣-helical in conformation), and linking the pair of cysteine residues (S519 Cys 11 and Cys 18 ) to form a disulfide bond. Finally, models of full-length S519 bound to the IR ectodomain were created by superimposing the C-terminal residues of the S519 models onto those of the S519C16 peptide in complex with the IR ectodomain. Models with overlap between S519 and the IR were rejected, reducing the 5,000 initial models to 901. The 901 models were clustered using the MMTSB cluster.pl utility (34). The final models were subjected to molecular dynamics minimization with the YASARA (version 15.7.12) program. Hydrogen atoms were added to fill missing valencies, water molecules were included, and a short simulated annealing protocol was applied using the YAMBER3 force field.