Aromatic Anchor at an Invariant Hormone-Receptor Interface

Background: Invariant insulin residue PheB24 (a site of diabetes-associated mutation) contacts the insulin receptor. Results: Hormonal function requires hydrophobicity rather than aromaticity at this site. Conclusion: The B24 side chain provides a nonpolar anchor at the receptor interface. Significance: Nonstandard aliphatic modification of residue B24 may enhance therapeutic properties of insulin analogs. Crystallographic studies of insulin bound to fragments of the insulin receptor have recently defined the topography of the primary hormone-receptor interface. Here, we have investigated the role of PheB24, an invariant aromatic anchor at this interface and site of a human mutation causing diabetes mellitus. An extensive set of B24 substitutions has been constructed and tested for effects on receptor binding. Although aromaticity has long been considered a key requirement at this position, MetB24 was found to confer essentially native affinity and bioactivity. Molecular modeling suggests that this linear side chain can serve as an alternative hydrophobic anchor at the hormone-receptor interface. These findings motivated further substitution of PheB24 by cyclohexanylalanine (Cha), which contains a nonplanar aliphatic ring. Contrary to expectations, [ChaB24]insulin likewise exhibited high activity. Furthermore, its resistance to fibrillation and the rapid rate of hexamer disassembly, properties of potential therapeutic advantage, were enhanced. The crystal structure of the ChaB24 analog, determined as an R6 zinc-stabilized hexamer at a resolution of 1.5 Å, closely resembles that of wild-type insulin. The nonplanar aliphatic ring exhibits two chair conformations with partial occupancies, each recapitulating the role of PheB24 at the dimer interface. Together, these studies have defined structural requirements of an anchor residue within the B24-binding pocket of the insulin receptor; similar molecular principles are likely to pertain to insulin-related growth factors. Our results highlight in particular the utility of nonaromatic side chains as probes of the B24 pocket and suggest that the nonstandard Cha side chain may have therapeutic utility.

with these structural elements (designated the "micro-receptor"; IR) has recently been determined ( Fig. 1C) (5). In the IR complex, the classical position of the C-terminal segment of the insulin B chain (brown in Fig. 1C) (2) would intersect ␣CT (magenta) and so must be displaced (5,6). In this study we have investigated a key feature of this complex, stabilization of the bound conformation of insulin by an invariant aromatic residue (Phe B24 ) docked within a nonpolar pocket of the IR (6).
Vertebrate insulin sequences are remarkable for an invariant phenylalanine (Phe) residue at position B24 (Phe B24 ) and for the broad conservation of phenylalanine and tyrosine at B25 and B26 (Phe B25 and Tyr B26 ) ( Fig. 2A) (7). A homologous triplet of aromatic residues is conserved among vertebrate insulin-related growth factors (IGF-I and IGF-II; Phe-Tyr-Phe) (8,9). The contributions of these residues to the structure, stability, selfassembly, and activity of insulin have attracted long standing interest (10 -12). The crystal structure of the zinc insulin hexamer, determined in 1969 by Hodgkin and co-workers (2,13), revealed that residues B24 -B26 participate in an aromatic-rich dimer interface (middle panel of Fig. 1A); this interface occurs three times in the hexamer (lower panel of Fig. 1A). The C-terminal deletion of this segment markedly impairs IR binding (14). The importance of these side chains was further highlighted by the discovery of clinical mutations associated with diabetes mellitus (DM) (2); effects of the substitution of Phe B24 by Ser (clinical variant insulin Los Angeles) attracted detailed characterization (15,16). Despite their intense study in the 45 years since the publication of the Hodgkin structure (13), how these aromatic side chains bind to the IR has, until recently, remained the subject of speculation. Photo-cross-linking studies suggested that in the holoreceptor complex Phe B24 and Tyr B26 primarily contact L1, whereas Phe B25 primarily contacts ␣CT (17,18).
The co-crystal structure of an insulin-IR complex, recently refined at 3.5 Å resolution, has revealed that on receptor binding the B24 -B27 segment is displaced from its position in the free hormone (Fig. 2B) (6). Such displacement, associated with reorganization of ␣CT on the surface of L1 (5), is critical to high affinity binding of the hormone. At this reorganized interface, the side chain of Phe B24 plays a unique role as an "anchor" Conserved aromatic residues Phe B25 and Tyr B26 are shown as black sticks, and Phe B24 is shown in red. The Zn 2ϩ ion is depicted in blue. B, ⌳-shaped IR ectodomain homodimer. One protomer is shown as a ribbon (labeled), and the other as a molecular surface. Domains are as follows: L1, first leucine-rich repeat domain; CR, cysteine-rich domain; L2, second leucine-rich repeat domain; FnIII-1-3, respective first, second, and third fibronectin type III domains; and ␣CT, ␣-chain C-terminal segment. C, overlay illustrating insulin in its classical free conformation bound to site 1 of the microreceptor (L1-CR ϩ ␣CT(704 -719); designated IR) (5). L1 and part of CR are shown in cyan, and ␣CT in magenta. Phe B24 , Phe B25 , and Tyr B26 are as in A. The B chain is shown in dark gray (B6 -B19); the position of the brown tube (residues B20 -B30) would lead to a steric clash between B26 -B30 and ␣CT. The figure was in part modified from Menting et al. (6), with permission of the authors. Coordinates were obtained from PDB entries 4INS, 2DTG, and 3W11. within a nonpolar pocket defined by the IR L1 surface, ␣CT, and central insulin B chain ␣-helix (Fig. 2C). The immediate environment of Phe B24 in the IR is delimited by L1 side chains Asn 15 , Leu 37 , and Phe 39 by the ␣CT side chain Phe 714 and by the insulin side chains Leu B15 , Tyr B16 , and Cys B19 (stereo model in Fig. 2D; for clarity B16 is omitted). Although past studies of selected insulin analogs suggested that aromaticity is required for such anchoring (12), the structure of the B24-related binding pocket in the IR would seem compatible with aliphatic side chains of appropriate size and shape. To resolve this apparent discrepancy, we undertook systematic mutagenesis of the B24 position in human insulin. To facilitate trypsin-mediated semi-synthesis (19), our collection of analogs was prepared with ornithine (Orn) rather than lysine at position B29. 7 Intriguingly, we observed that a nonaromatic side chain (Met B24 ) confers essentially native receptor-binding affinity and substantial biological activity in a rat model of DM. Molecular modeling of a simulated interface between [Met B24 ]insulin and the IR suggested that this linear side chain might adopt a compact conformation within the confines of the B24-related pocket, intrinsically nonplanar due to the tetrahedral geometry of the side-chain carbon atoms and related sp 3 configuration of the ␦-sulfur atom. This model motivated design and characterization of a nonstandard insulin analog in which Phe B24 was substituted by cyclohexanylalanine (Cha), an amino acid containing a cyclic and nonplanar aliphatic side chain. The Cha B24 insulin analog likewise exhibited high activity, and its crystal structure (determined at a resolution of 1.5 Å as an R 6 zinc insulin hexamer) closely resembles that of wild-type (WT) insulin, including within the hexamer's three aromatic-rich B24-related dimer interfaces. The crystallographic analysis was enriched by the presence of an entire hexamer within the asymmetric unit (20). To our knowledge, the crystal structure of [Cha B24 ]insulin represents the first high resolution view of a cyclic aliphatic side chain in a globular protein (21).
In the crystal structure, the six Cha B24 side chains each exhibited a pair of overlapping chair conformations (oriented up or down) with variable occupancies. Although Cha B24 itself is unable to participate in weakly polar interactions (unlike Phe B24 ) (22), neighboring aromatic-aromatic interactions among Tyr B16 , Phe B25 , Tyr B26 , and their dimer-related partners are preserved in the structure. Molecular modeling of the variant IR complex likewise suggested that the side chain of Cha B24 would readily be accommodated within the aromaticrich confines of the B24-related pocket. Nonetheless, the distinctive structure and flexibility of Cha's chair conformations may confer novel biophysical properties. In the context of a mealtime insulin analog in current clinical use ([Lys B28 ,Pro B29 ]insulin; named KP-insulin (23)), we observed that Cha B24 enhanced the rate of disassembly of the analog hexamer and delayed onset of protein fibrillation, properties of potential therapeutic advantage in the treatment of DM (24).
Together, our results have defined the structural requirements of an invariant anchor residue at a hormone-receptor interface (6). Conservation of Phe B24 among vertebrate insulins and IGFs (2) may be enjoined by biological constraints (such as efficiency of folding and avoidance of toxic misfolding) unrelated to structure-activity relationships. Despite the broad conservation of Phe B24 and its cognate IR binding pocket, nonstandard modifications at this site may be of translational interest.

EXPERIMENTAL PROCEDURES
Preparation of Insulin Analogs-Analogs were prepared by trypsin-catalyzed semi-synthesis using an insulin fragment, des-octapeptide (B23-B30)-insulin, and modified octapeptides as described (19). The fragment was made by tryptic cleavage of human insulin and purified by reverse phase HPLC. Modified octapeptides were prepared by solid phase synthetic methods (25). Trypsin-mediated formation of a peptide bond between Arg B22 and a synthetic octapeptide (in the series GXFYTPOT or GXFYTKPT, where X represents a standard or nonstandard amino acid and O designates ornithine) was favored by reaction in a mixed aqueous-organic solvent system containing 1,4-butanediol and dimethylacetamide as described (26); the resulting 51-residue insulin analogs were purified by preparative reverse phase C4 HPLC (Higgins Analytical Inc., Proto 300 C4 10 M, 250 ϫ 20 mm), and their purity was assessed by analytical reverse phase C4 HPLC (Higgins Analytical Inc., Proto 300 C4 5 M, 250 ϫ 4.6 mm). Predicted molecular masses were in each case verified using an Applied Biosystems 4700 proteomics analyzer MALDI-TOF.
X-ray Crystallography-Crystals of [Cha B24 ,Orn B29 ]insulin were obtained via hanging-drop vapor diffusion at 25°C. 1-l drops containing the protein at 10 mg/ml in 0.02 N HCl were mixed with a 1-l drop of reservoir buffer containing 0.1 M sodium citrate, 0.08% zinc acetate, 2% phenol. Drops were suspended over 1 ml of reservoir buffer. A single crystal was transferred to a solution containing 20% (v/v) glycerol in the mother liquor for flash freezing. Diffraction data were collected at ϳ100 K at a wavelength of 0.9537 Å at the Argonne National Laboratory, Structural Biology Center (beamline 19-ID; Argonne, IL).
Data were processed using XDS, version January 10, 2014 (27), and the resolution limit was set according to the CC 1/2 criterion (28) at the p ϭ 0.001 level of significance. The structure was then determined by molecular replacement using PDB entry 1ZNJ as an initial search model and refined using PHENIX version 1.9.1692 (29). Hydrogen atoms were included in the final stages of refinement, which employed anisotropic B-factors for all non-hydrogen atoms. Within each insulin monomer, the segment Gly B23 -Cha B24 -Phe B25 was modeled as having two alternate conformations to allow for gauche and anti-gauche chair conformations (defined with respect to the C ␤ -C ␥ bond). Occupancies of these paired segments were then refined independently across the protomers. The final model included residues A1-A21 and B1-B28 for each of the six insulin molecules in the asymmetric unit, 207 water molecules, two Zn 2ϩ cations, two Cl Ϫ anions, and six phenol molecules. Full data processing and refinement, the statistics are given in Table 1.
Circular Dichroism Spectroscopy-Far-ultraviolet (UV) spectra were obtained from 200 to 250 nm on an AVIV spectropolarimeter equipped with an automated syringe-driven titration unit. Insulin or analogs were made 50 M in a buffer containing 10 mM potassium phosphate (pH 7.4) and 50 mM KCl. Helixsensitive wavelength 222 nm was used as a probe of protein denaturation by guanidine hydrochloride. Thermodynamic parameters were obtained by application of a two-state model as described (30). In brief, data were fit by nonlinear least squares to a two-state-model as shown in Equation 1, where x is the concentration of guanidine hydrochloride, and A and B represent respective estimates of the baseline ellipticities of the protein in its native and unfolded states as extrapolated to a guanidine concentration of 0. Baseline values were approximated via pre-and post-transition lines represented by  (30,31). Spectra of DKP-insulin and [Cha B24 -DKP]insulin were each acquired at protein concentrations of 0.35 mM. Homonuclear two-dimensional nuclear Overhauser effect (NOE) spectroscopy (NOESY) and total correlation spectroscopy (TOCSY) were obtained with respective mixing times of 200 ms and 55 ms. Presumptive assignments were obtained by analogy to the assigned spectrum of [L-Ala B24 -DKP]insulin, an analog that exhibits a native-like fold but without the aromatic ring current of Phe B24 (30,31). 1 H NMR chemical shifts were calibrated to parts/million (ppm) relative to trimethylsilyl propionate as an internal standard, assumed to be 0.0 ppm.
Receptor Binding Assays-Affinities of WT insulin or semisynthetic analogs for the IR (isoform B) were measured by a competitive-displacement assay (32). In brief, microtiter strip plates (Nunc Maxisorb) were incubated at 4°C overnight with a stock solution containing 40 g/ml anti-FLAG immunoglobulin G. Lysates of 293 PEAK cells transfected with cDNAs encoding the IR with a C-terminal FLAG tag were purified using wheat germ agglutinin chromatography as described previously (32). Partially purified receptors were then immobilized in the coated plates. Competitive binding assays employed human [3-125 I-Tyr A14 ]insulin as a tracer.
Receptor-based Screening of Insulin Analogs-Analogs were initially tested for their ability to displace pre-bound [ 125 I-Tyr A14 ]insulin from the IR at a single concentration of analog (0.75 nM), calibrated based on baseline displacement of 90% of receptor-bound [ 125 I-Tyr A14 ]insulin by control analog [Orn B29 ]insulin. The nondisplaced fraction of [ 125 I-Tyr A14 ]insulin (less than 20, 21-40, 41-70, and 71-100%) by a given B24 analog permitted its qualitative assignment to a corresponding class of relative affinities (designated high (I), intermediate (II), low (III), or very low (IV)).
Rodent Assays-Male Sprague-Dawley rats (mean body mass ϳ300 g) were rendered diabetic by streptozotocin (33). To test the in vivo potency of insulin analogs relative to [Orn B29 ]insulin, the analogs were made 10 g per 100 l in a formulation buffer (16 mg/ml glycerin, 1.6 mg/ml meta-cresol, 0.65 mg/ml phenol, and 3.8 mg/ml sodium phosphate (pH 7.4)) and injected intravenously into tail veins. The normalized dose was 10 g of insulin analog per 300-g rat with the actual dose (and so injection volume) being adjusted to each rat's body weight. Resulting changes in blood glucose concentration were monitored using a clinical glucometer (Hypoguard Advance Micro-Draw meter). WT insulin or analogs were each re-purified by HPLC, dried to powder, dissolved in diluent at the same maximum protein concentration, and re-quantitated by analytical C4 reverse phase HPLC to ensure uniformity of formulations; dilutions were made using the above buffer. Rats were injected at time t ϭ 0. Blood was obtained from the clipped tip of the tail at time t ϭ 0 and every 10 min for the 1st h, every 20 min for the 2nd h, every 30 min for the 3rd h, and every hour thereafter to a final time of 5 h. The efficacy of WT insulin or analog to reduce the blood glucose concentration was calculated using the following: (a) the change in blood glucose concentration over the first hour, and (b) the integrated area between the glucose time dependence and a horizontal line at the starting blood glucose concentration (area over the curve; AOC). Statistical significance was assessed using a Student's t test.
Assessment of Fibril Formation-To estimate lag times prior to onset of fibril formation, zinc-free WT insulin or analogs were made 60 M in phosphate-buffered saline (pH 7.4) containing 0.01% sodium azide as an antimicrobial agent. The solutions were gently rocked at 37°C in glass vials containing a liquid-air interface. Aliquots, taken at regular intervals, were frozen to enable later analysis of thioflavin T fluorescence. For a given tube, the assay was terminated on the 2nd day of the appearance of cloudiness in the solution (34).
Cobalt-coordinated Insulin Hexamers-Visual absorption spectroscopy was used to probe the formation and disassembly of phenol-stabilized R 6 Co 2ϩ -substituted insulin hexamers. WT insulin or analogs were made 0.6 mM in a buffer containing 50 mM Tris-HCl (pH 7.4), 50 mM phenol, 0.2 mM CoCl 2 , and 1 mM NaSCN (35). Samples were incubated overnight at room temperature prior to the studies to ensure that a conformational equilibrium was reached. Spectra (450 -700 nm) were obtained to monitor tetrahedral Co 2ϩ coordination with its signature peak absorption band at 574 nm (35). To determine the rate of Co 2ϩ release from the hexamers, metal ion sequestration was initiated at 25°C by addition of an aliquot of EDTA (50 mM at pH 7.4) to a final concentration of 2 mM. Attenuation of the 574-nm absorption band was monitored on a time scale of seconds to hours. Kinetic data were consistent with a monoexponential decay (36).
Molecular Dynamics Simulations-Molecular dynamics (MD) simulations were performed using GROMACS (version 4.5.5) (37), which employs the CHARMM all-atom additive force field (38,39). Proteins were solvated in a cubic box of TIP3P water molecules. Ionizable residues and protein termini were assumed to be in their charged state. Sodium and chloride ions were added to neutralize the system at a final ionic strength of 0.10 M. Protein and solvent (including ions) were coupled separately to a thermal bath at 300 K employing velocity rescaling (40) applied with a coupling time of 1.0 ps. Pressure was maintained at 1 bar by coupling to a Berendsen barostat (41) with a coupling constant of 5.0 ps and compressibility of 4.5 ϫ 10 Ϫ5 bar. The time step was 2 fs. Simulations were performed with a single nonbonded cutoff of 10 Å and neighbor list update frequency of 10 steps (20 fs). The particle-mesh Ewald method was used to account for long range electrostatics (42), applying a grid width of 1.2 Å with fourth-order spline interpolation. Bond lengths were constrained using the LINCS algorithm (43). All simulations consisted of an initial minimization of water molecules, followed by 100 ps of MD with the protein restrained to permit equilibration of the solvent. Simulations of the WT, [Met B24 ]-, and [Cha B24 ]insulin monomers used as initial model coordinates the structure 2KJJ obtained from the Protein Data Bank (PDB). Calculations were continued for 120 ns from the geometries obtained after initial positionally restrained MD (below) at a temperature of 300 K.
Molecular Modeling of IR Complexes-Comparative modeling of the Met B24 -and Cha B24 variant IR complexes was performed using MODELLER (44). The simulations employed as molecular templates the structure of the WT IR-insulin complex (PDB entry 4OGA; Ref 6). The models included residues disordered in the crystal structure, viz. Cys 159 -Asn 168 and Lys 265 -Gln 276 of IR and Asp B28 -Thr B30 of insulin. An initial 50 models were created for each complex, and the structure with the lowest energy was selected for MD simulations. The IR L1-CR fragment contains N-linked glycosylation sites at residues Asn 16 , Asn 25 , Asn 111 , Asn 215 , and Asn 255 (45); N-acetylglucosamine carbohydrate was thus attached at each of these sites. Following positionally restrained MD, all restraints on the protein were removed, and MD continued for a further 50 ns. The surface area and volumes of protein cavities and crevices were estimated using programs VoroProt and PROVAT (46,47), respectively. The surface areas of cavities were calculated based on rolling a sphere of radius 1.4 Å, whereas cavity volumes represent the difference between the sizes of cavities in the WT structure and rigid-body models wherein the residue of interest was substituted by Gly.

RESULTS
Receptor Binding Screen-Eighteen insulin analogs containing substitutions at B24 were prepared at small scale in the context of [Orn B29 ]insulin (Table 2), an active template chosen to facilitate trypsin-mediated semi-synthesis (6). To avoid disulfide interchange reactions, we did not prepare a Cys B24 analog. To avoid introducing a tryptic site, Lys B24 and Arg B24 analogs were also not prepared. A coarse receptor binding assay was designed to discriminate among insulin analogs based on displacement of pre-bound [ 125 I-Tyr A14 ]insulin at an analog concentration of 0.75 nM (Fig. 3A). At this concentration, [Orn B29 ]insulin was found to displace Ն80% of the bound [ 125 I-Tyr A14 ]insulin; a maximum of 20% of bound [ 125 I-Tyr A14 ]insulin was thus employed to define a high affinity group of analogs (class I in Fig. 3B). Intermediate, low, and very low affinity groups were similarly defined based on fractions of bound [ 125 I-Tyr A14 ]insulin in the presence of the 0.75 nM analog of 21-40, 41-70, and 71-100%, respectively (classes II-IV in Fig. 3B). This preliminary classification facilitated design of definitive binding assays.
The high affinity group (class I) included Phe (the WT residue), an aliphatic residue (Met), and Gly; the anomalous activity of [Gly B24 ]insulin has been widely discussed in past studies (asterisk in Fig. 3B) (10,48,49). Class II contained Ile and Leu. Class III contained Val as well as the aromatic residues Tyr and Trp; the low affinity of [Tyr B24 ,Orn B29 ]insulin is in accordance with prior studies (11,12). Class IV (very low affinity) contained His (in accordance with Ref. 12), charged (Asp and Glu), constrained (Pro), polar (Asn, Gln, Ser, and Thr), and Ala (in accordance with Ref. 50). Following this screen, all analogs in classes I-III and representative analogs in class IV (Ala, Asn, Asp, Ser, and Thr) were prepared at larger scale to enable measurement of hormone-receptor dissociation constants (K d ) and, in four cases (Gly, Ile, Leu, and Met), measurement of thermodynamic stabilities (⌬G u ).
Results of selected complete competitive displacement assays of receptor binding are given in Table 2; representative data are shown in Fig. 3C. The three analogs with highest affinities were Gly B24 (K d 0.05 Ϯ 0.01 nM), Phe B24 (parent analog; K d 0.05 Ϯ 0.01 nM), and Met B24 (K d 0.06 Ϯ 0.01 nM) as compared with WT (K d 0.05 Ϯ 0.01 nM). To test whether the high affinity of the Met B24 analog depends on side-chain hydrophobicity, we prepared an Orn B24 analog (a basic isostere of Met); only negligible receptor binding was observed (K d Ͼ50 nM; Table 2). The affinity of the Leu B24 variant was reduced by 2-fold (K d 0.10 Ϯ 0.02 nM) in accordance with past studies (19,51); the affinity of the Ile B24 variant was reduced by 3-fold (K d 0.15 Ϯ 0.02 nM). Such small differences are consistent with natural variation among mammalian insulins (species variants (52)). Aside from the special case of Gly B24 (which may reflect a shift in binding mode (6,12,53)), 8 these findings suggest that the hormone-receptor interface selects for nonpolar side chains at B24 of suitable shape and size, irrespective of aromaticity. Avoidance of packing defects within the B24-related pocket presumably underlies the affinity order Phe, Met Ͼ Ile, Leu Ͼ Val Ͼ Ala.
Residues in class IV are likely to have imposed a combination of cavity-related and polarity penalties. Because of its clinical association with DM (15,16), detailed binding studies were undertaken of [Ser B24 ,Orn B29 ]insulin in relation to [Ala B24 ,Orn B29 ]insulin and other small polar or charged side chains in class IV (Thr, Asn, or Asp; Table 2). In accordance with previous studies, the affinity of the Ser B24 analog (15,54) was low (ϳ1% relative to [Orn B29 ]insulin) but 3-fold greater than that of the Ala B24 analog (50). Also preferred relative to Ala B24 were Thr, Asn, or Asp, which conferred affinities similar to that of [Ser B24 ,Orn B29 ]insulin. Unlike Ala, each of these side chains would be able to participate in hydrogen bonding. 8 The high activity of [Gly B24 ]insulin has long been noted as an anomaly (10,49,66).

TABLE 2 Receptor-binding affinities of insulin analogs
Analogs were prepared in a template in which Lys B29 was substituted by Orn. Assays employed the B isoform of the lectin-purified and detergent-solubilized IR.

B24 residue
This represents ͓Orn B29 ͔insulin; its dissociation constant was indistinguishable from that of WT insulin. b VL means very low affinity (class IV in coarse screen). c Complete binding assays were performed on these representative members of class IV to obtain a quantitative calibration of the assay. d ND means not determined. e Orn B24 provided a model of a basic side chain as an isostere of Met; its K d value was estimated to be 73 Ϯ 17 nM. Lys and Arg were not included in the initial screen as these substitutions would have complicated semi-synthesis (see "Experimental Procedures").
Folding and Stability-Overall ␣-helicity of the high affinity B24 insulin analogs (Met, Gly, Ile, and Leu) was monitored by far-UV CD (Fig. 4, A and C). Although the spectra indicated that a predominance of ␣-helix is maintained, attenuation of mean residue ellipticity at 222 nm and more negative values at 208 nm suggested overall ␣-helical destabilization. Such features were previously observed in the CD spectrum of KP-insulin in relation to WT insulin (49) and may reflect decreased dimerization at this protein concentration (50 M) (55). In accordance with this interpretation, previous studies have shown that the spectrum of an Ala B24 analog of engineered monomer DKP-insulin 9 was similar to that of DKP-insulin itself (30).
Helix-related ellipticity at 222 nm was exploited as a probe of fractional protein unfolding as a function of increasing concentrations of guanidine hydrochloride (Fig. 4, B, and D). Estimates of free energies of unfolding (⌬G u ), based on a two-state model (30), demonstrated that the aliphatic substitutions at B24 were in each case associated with lower thermodynamic stability ( Fig. 8A and Table 3). The B24 substitutions compatible with high affinity (Gly, Ile, Leu, and Met) exhibited decrements (⌬⌬G u ) in the range 0.7-1.9 kcal⅐mol Ϫ1 (including uncertainties) relative to the baseline stability of [Orn B29 ]insulin (⌬G u 3.5 Ϯ 0.2 kcal/mol). The m values obtained in fitting the twostate model (i.e. the slope in a plot of ⌬G u versus denaturation concentration; column 5 in Table 3) were also reduced. Because this slope correlates with extent of exposure of nonpolar surfaces on unfolding (30), this trend suggests that the nonaromatic side chains at B24 are associated with less efficient burial of such surfaces in the respective native states.
Nonstandard Mutagenesis-The B24-related pocket, as visualized in the crystal structure of the WT IR-complex (6), has a calculated volume of ϳ200 Å 3 in the absence of Phe B24 . Of this potential space, the side chain of Phe B24 occupies only 117 Å 3 (packing efficiency 58%); residual packing defects are predicted above, below, and to the sides of the aromatic ring (see "Discussion"). Although such quantitative analysis is limited by the low resolution of the structure (3.5 Å), modeling predicted that the planar aromatic ring of Phe B24 could be replaced by a nonplanar aliphatic ring without steric clash. We therefore prepared nonstandard analogs [Cha B24 ,Orn B29 ]insulin (for comparison with our original set of B24 variants) and [Cha B24 -KP]insulin (for assessment of its potential pharmacologic properties (56)); [Cha B24 -DKP]insulin was also prepared to facilitate NMR studies (30, 31). These analogs exhibited a relative affinity for the insulin receptor (ϳ50 -60%; class II) similar to that of the Leu B24 analog ( Fig. 3D and Table 2). To test whether Phe and Cha are generally interchangeable at protein-protein interfaces, we also prepared a Cha B25 analog (in the context of KP-insulin); at this site the nonstandard substitution impaired receptor binding by ϳ50-fold relative to KP-insulin (Table 4) in accordance with the low activity of DM-associated clinical variant [Leu B25 ]insulin (19,51). The substantial receptor-binding affinity of the Cha B24 analogs is therefore an intrinsic feature of the B24-related pocket.   (30,31). The absence of B24-related aromatic TOCSY and NOESY cross-peaks in the variant spectra were readily apparent (Fig. 5). Nevertheless resolved were the inter-residue NOEs involving the aromatic protons of Tyr B26 (with presumptive assignments of B26-Val B12 and B26-Leu B15 ) and of Tyr A19 (A19-Ile A2 and A19-Leu B15 ). These qualitative NMR features indicate that the Cha B24 analog maintains a native-like conformation in solution, including with respect to the overall orientation of the C-terminal segment of the B chain relative to the ␣-helical core (30,31 (20). Indeed, a classical monoclinic lattice (Table 1) was observed in which one R 6 hexamer (with protomers designated as having chains A and B, C and D, E and F, G and H, I and J, and K and L, respectively) defined the asymmetric unit. 10 A ribbon model of the zinc insulin analog hexamer is shown in Fig. 6A; in      ). B, superposition of a representative analog protomer (gray) and WT protomer (white). The side chains of Cha B24 (red) and Phe B24 (white) are shown as sticks. C, stereo view of aromatic residues (gray) and Cha B24 (red) at dimer interface. D, expanded view of the side chains near B24 in the analog structure (gray) and WT structure (white). Sulfur atoms of cysteine A20 -B19 are shown as yellow spheres. Coordinates of the WT R 6 hexamer were obtained from PDB entry 1ZNJ. DECEMBER 12, 2014 • VOLUME 289 • NUMBER 50 JOURNAL OF BIOLOGICAL CHEMISTRY 34717 axial zinc ions (overlying blue spheres at center of Fig. 6A) by the side chains of His B10 (three per R 3 trimer; light gray side chains) was essentially identical to that observed in WT R 6 hexamers (20). The fourth coordination site contained a presumed chloride anion. A representative (2F obs Ϫ F calc ) electron density map of one metal ion-binding site (Fig. 7A) is shown in relation to a superposition of the analog and WT coordination sites (dark and light gray in Fig. 7B; stereo stick models). Zn-N distances are 1.98, 1.98, and 2.00 Å in the zinc-binding site shown and 1.96, 2.01, 2.04 Å in the other trimer.

Anchor Residue at the Insulin Receptor Interface
Superposition of a representative analog protomer and WT protomer is shown in Fig. 6B (dark and light gray ribbons, respectively). Average pairwise differences between individual protomers of [Cha B24 ,Orn B29 ]insulin and the WT R-state protomers (main-chain r.m.s.d. 0.65 ϭ Å and side-chain r.m.s.d. ϭ 2.03 Å) are only marginally different from the structural variation within a collection of WT protomers (mainchain r.m.s.d. ϭ 0.54 Å and side-chain r.m.s.d. ϭ 1.87 Å). Similarly, the variant dimer interface closely resembles that of WT insulin; a representative structure is shown in Fig. 6C. Side chains in close proximity to B24 are shown in Fig. 6D. Alignment of a representative analog dimer and WT dimer yielded main-chain r.m.s.d. ϭ 0.29 Å and side-chain r.m.s.d. ϭ 2.50 Å, the latter being inflated by the varying rotameric disposition of the side chain of residue B25 (and its 2-fold related partner) within and across both structures.
The analog hexamer contained six bound molecules of phenol, located at an interface between dimers as in the WT R 6 hexamer (20). The six independent phenol-binding sites are essentially identical. The (2F obs Ϫ F calc ) electron density map volume associated with one such phenol is shown in Fig. 7C in relation to a superposition of variant and WT structures (dark and light gray; stereo stick models; Fig. 7D). In each case, a characteristic pair of hydrogen bonds from the phenolic -OH group was formed to the main-chain carbonyl oxygen (acceptor) and amide group (donor) of Cys A6 and Cys A11 , respectively.
The positions of Phe B24 (light gray side chain) and Cha B24 (red) closely overlap, with the side chain of Cha B24 (red in Fig.  7F) packing among the cluster of aromatic residues Tyr B16 , Phe B25 , Tyr B26 , and their dimer-related partners (black) (Fig.  6C). The six independent side chains of Cha B24 each exhibited the expected chair conformations but with varying relative occupancies of two overlapping conformations (with respective C atoms oriented toward or away from Leu B15 of the same FIGURE 7. (2F obs ؊ F calc ) electron density maps. A, on-axis view of a representative zinc ion-binding site in R 6 analog hexamer. B, superposition (stereo view) of zinc-binding sites in the analog hexamer (gray) and WT R 6 hexamer (white). Zn 2ϩ ion is shown as a black sphere. Not shown: Zn 2ϩ -coordinating Cl Ϫ anion. C, electron density of a representative bound phenolic ligand. Its para-OH group forms hydrogen bonds with the carbonyl oxygen and amide nitrogen of Cys A6 and Cys A11 , respectively (cysteine A6 -A11). The phenolic ring makes van der Waals contacts within the dimer-dimer interface of the hexamer; contacts include His B5 of another dimer. D, stereo view corresponding to C. E, electron density of Cha B24 and surrounding residues. F, stereo view corresponding to E (stick representation). The two chair conformations of Cha B24 and Cha B24Ј at this dimer interface are highlighted in red. Coordinates of the WT R 6 hexamer were obtained from PDB entry 1ZNJ. chain; defined as anti-gauche or gauche with respect to the B24 C ␤ -C ␥ bond). Occupancy refinement suggested that in four of the protomers the cyclohexanyl ring has a predominant antigauche conformation (anti-gauche occupancies 0.73 (chain B), 0.70 (chain D), 0.68 (chain F), and 0.82 (chain L), respectively); in the other two protomers the occupancy was split somewhat more evenly (anti-gauche occupancies 0.64 (chain H) and 0.54 (chain J), respectively). Dimer-related pairs of Cha B24 side chains thus exhibited adjoining anti-gauche occupancies of (0.73, 0.86), (0.68, 0.64), and (0.54, 0.82). Although these occupancies are consistent in each case with the appearance of the associated volumes of the final (2F obs Ϫ F calc ) electron density map (Fig. 7E), the diffraction data do not distinguish between dynamic disorder within a hexamer from static disorder among hexamers in the crystal lattice. Observation of nonequivalent occupancies across a given dimer interface nonetheless suggests that both anti-gauche/anti-gauche and anti-gauche/ gauche packing schemes are allowed.
The environment of Cha B24 is similar but not identical to that of Phe B24 at the dimer interface. A (2F obs Ϫ F calc ) electron density map of one such interface is shown in Fig. 7E in relation to a superposition of analog and WT structures (Fig. 7F). Although the four canonical inter-chain hydrogen bonds observed in the B24 -B28/B28Ј-B24Ј anti-parallel ␤-sheet (2) are indistinguishable from those observed in a collection of WT R 6 structures (20), small adjustments occurred in the side-chain conformations of Tyr B26 Ј (near Cha B24 ) and Tyr B26 (near Cha B24 Ј), where primed residues represent the dimer-related partners. Although the positions of the aromatic rings overlap, on average their respective 1 and 2 angles differ slightly (by 2°a nd 5°with respect to WT hexamer 1ZNJ). These adjustments are presumably related to the nonplanarity of Cha B24 and Cha B24 Ј and to changes in weakly polar interactions (22). The similarity of the variant and WT structures suggests that the essential features of the dimer interface do not depend on the asymmetric distribution of partial charges in the aromatic ring of Phe B24 and associated pattern of aromatic-aromatic interactions.
Kinetic Analysis of Hexamer Disassembly-To gain further insight into native self-assembly, we next probed the rate of disassembly of [Cha B24 ,Orn B29 ]insulin and [Cha B24 -KP]insulin hexamers in relation to their respective parent templates, WT insulin, and [Met B24 ,Orn B29 ]insulin. These studies employed Co 2ϩ as an optical probe of the tetrahedral zinc-binding site in the R 6 insulin hexamer (35).
Visible absorption spectra of [Orn B29 ]insulin and its derivatives are shown in Fig. 8C. The spectra of the variant complexes exhibited d-d absorption bands similar in shape to those of [Orn B29 ]insulin hexamers (black line) but reduced in magnitude. The respective absorption bands of the Met B24 and Cha B24 analogs were attenuated by 37 Ϯ 2 and 7 Ϯ 1% (blue and green lines in Fig. 8C). Such attenuation suggests either incomplete hexamer assembly at the protein concentration employed (0.6 mM) or a partial shift in the conformational equilibrium to subpopulations of T 3  The conformational equilibrium of insulin is characterized by rates of hexamer disassembly (with loss of bound metal ions) and reassembly (with restoration of metal ion binding). To probe the lifetimes of the tetrahedral Co 2ϩ -binding sites, we added an excess of EDTA to these protein solutions. The ensuing attenuation of the d-d absorption bands provided a kinetic probe for dissociation of the hexamer with release of the bound  Table 5. DECEMBER 12, 2014 • VOLUME 289 • NUMBER 50 metal ion leading to its sequestration as an octahedral EDTA complex. In this assay, originally developed at the Lilly Research Laboratories (36), the time-course of d-d attenuation is essentially mono-exponential. Lifetimes are given in Table 5. In the context of [Orn B29 ]insulin substitution of Phe B24 by Met led to a 10-fold reduction in lifetime, whereas substitution by Cha led to 3-fold reduction (Fig. 8D). The Cha B24 substitution had a similar effect in the context of KP-insulin (Fig. 8F). Accelerated disassembly of cobalt insulin hexamers is of potential pharmacologic relevance as the more rapid disassembly of KP-insulin (4-fold relative to WT) has been shown to correlate with its more rapid absorption from a subcutaneous depot in patients (23).

Anchor Residue at the Insulin Receptor Interface
Insulin Fibrillation-Whereas native state stabilities (summarized for the present analogs in Fig. 8A) are associated with relative rates of chemical degradation of clinical insulin formulations (58), the predominant mechanism of degradation above room temperature is through fibrillation (amyloid formation) (59). Accordingly, the susceptibility of the Cha B24 analogs to fibrillation was probed on gentle agitation at 37°C in relation to WT insulin, KP-insulin, [Orn B29 ]insulin, and [Met B24 ,Orn B29 ]insulin (Table 3); the lag times were defined on the basis of a 2-fold enhancement of thioflavin T fluorescence (34). Because of the stochastic nature of a nucleation-propagation reaction (60), results are shown for each individual vial (Fig. 8B).
Whereas substitution of Lys B29 by Orn was observed to prolong the lag time, this effect was mitigated by the destabilizing Met B24 substitution. By contrast, lag times of [Orn B29 ]insulin and [Cha B24 ,Orn B29 ]insulin were indistinguishable (p ϭ 0.89). Furthermore, in the context of KP-insulin, Cha B24 appeared to provide a protective effect. Statistical analysis of the two groups (n ϭ 6 and n ϭ 5, respectively) indicated that the observed difference in mean lag times (3.3 and 13 days, respectively) was significant (p Ͻ 0.05). The unperturbed lag time of [Cha B24 ,Orn B29 ]insulin (relative to [Orn B29 ]insulin) and the prolonged lag time of [Cha B24 -KP]insulin (relative to KP-insulin) were surprising given their reduced thermodynamic stabilities (above). Indeed, an [Ala B24 ]insulin analog, which is less stable than wild type insulin, exhibits accelerated fibrillation (30). Relative prolongation of the lag time of [Cha B24 -KP]insulin seems remarkable as protective substitutions in insulin are uncommon in surveys of diverse insulin analogs (61).
Biological Activity in Diabetic Rats-The potency and duration of action of [Cha B24 ,Orn B29 ]insulin were tested in a rat model of DM (33) in relation to control studies of the parent [Orn B29 ]insulin, class I analog [Met B24 ,Orn B29 ]insulin, and class IV analog [Pro B24 ,Orn B29 ]insulin. The analogs were administered by i.v. bolus injection at a submaximal dose (10 g insulin analog per 300 g rat). The activity of [Orn B29 ]insulin was similar to that described previously in rat studies of KP-insulin (62), whereas the negligible activity of the Pro B24 analog was similar to that seen on injection of a buffer alone (the gradual linear fall in mean blood glucose concentration is likely to reflect progressive fasting). Because of variation in the initial blood glucose level between rats, data are shown both in relation to the actual blood glucose concentrations at time t ϭ 0 (Fig. 9A) and after normalization with respect to their initial values (defined as 1.0; Fig. 9B). The pattern of fall and recovery of the blood glucose concentrations observed on injection of [Cha B24 ,Orn B29 ]insulin and [Orn B29 ]insulin (black and green lines in Fig. 9A) were indistinguishable. A trend toward a more rapid recovery phase was observed on injection of [Met B24 ,Orn B29 ]insulin (blue line), but the large errors precluded definitive assessment.
Initial rates of fall of the blood glucose concentration (calculated between 0 and 60 min) were similar among the three active analogs; any differences were not statistically significant (Fig. 9C). Although the Met B24 analog appeared to be less longlived when mean values were plotted, quantitative analysis of total AOC demonstrated that any differences did not reach statistical significance (Fig. 9D). 11

DISCUSSION
In a co-crystal structure of insulin bound to a domain-minimized model of the IR (6), insulin segment B24 -B27 was displaced from the ␣-helical core of the hormone to insert between the tandem ␣CT-L1 element of the IR ␣ subunit (4). The side chain of Phe B24 inserts within a conserved pocket lined by L1 residues (Asn 15 , Leu 37 , and Phe 39 ), ␣CT residue Phe 714 , and insulin residues Leu B15 , Tyr B16 and Cys B19 . The B24-binding cavity (with surface area 444 Å 2 as defined by a 1.4-Å spherical probe) is predominantly nonpolar (ϳ80%). Whereas in the free insulin monomer one edge of the B24 aromatic ring is exposed to solvent, in the IR complex this side chain is buried. Of the conserved aromatic residues in the B chain C-terminal segment (Phe B24 , Phe B25 , and Tyr B26 ), only Phe B24 inserts within a classical binding pocket to anchor the displaced segment; the B25and B26 side chains are by contrast in contact with shallow depressions in otherwise exposed surfaces (6).
Structure-Activity Relationships-The long standing view that an aromatic side chain is required at B24 (10), recently reinforced by the studies of Brzozowski and co-workers (12), is broadly consistent with possible weakly polar interactions at the receptor 11 Initial rates of fall of the blood glucose concentration were not statistically significant (Fig. 9C). Although [Met B24 ,Orn B29 ]insulin appeared to be less potent than [Orn B29 ]insulin when mean values were plotted (Fig. 9, A and  B), quantitative analysis of total AOC demonstrated that any differences were not of statistical significance (Fig. 9D). The p value was 0.43. We cannot exclude that in a larger study (i.e. with a greater number of animals per group) statistically significant differences might be observed. If the small apparent differences in mean AOC in the present data should represent a true difference in potency, a power calculation suggests that a trial of Ͼ24 rats per group would be required to demonstrate statistical significance. In standard pharmaceutical formulations, any such differences in intrinsic potency would be compensated by redefinition of the formulation strength in relation to international units (IU) per mg of protein such that equivalent biological activity per ml is obtained. a Kinetic features of Co 2ϩ release from protein hexamers were monitored by following the attenuation in absorbance at 574 nm after addition of excess EDTA; data were fitted to a mono-exponential decay.
interface, including potential -(Phe 39 and Phe 714 ), aryl-sulfur (cystine B19-A20), and aryl-amino (Asn 15 ) interactions. Because such structural details were not well defined at the resolution of the refined co-crystal structure (3.5 Å), we undertook systematic mutagenesis of the B24 position. All possible amino acid substitutions were introduced with the exceptions of Cys (to avoid a free thiol group), Lys, and Arg; the latter were singly represented by Orn, a basic analog resistant to tryptic cleavage. In this collection of analogs, Lys B29 was substituted by Orn. The highest affinities were conferred by the native Phe B24 , Met B24 , and Gly B24 (the latter being in accordance with past results (10, 49) as discussed separately below). Relative to these analogs, our results (Table 2) exhibited a trend in favor of aliphatic substitutions with substantial affinity requiring a sidechain volume greater than that of Ala (affinity Ͻ1% relative to WT insulin). Whereas alternative aromatic side chains (His, Tyr, and Trp) conferred low affinity, nonaromatic side chains at B24 exhibited the relationship Phe (WT) ϭ ϳ Met Ͼ Leu Ͼ Ile Ͼ Val Ͼ Ͼ Ala. The affinity of [Leu B24 ,Orn B29 ]insulin was only 2-fold lower than that of WT insulin, whereas the affinity of [Ile B24 ,Orn B29 ]insulin was reduced by 3-fold. Charged or polar side chains impaired hormone binding. This trend in structure-activity relationships suggests that the B24-related pocket of the IR selects for nonpolar side chains whose detailed size and shape are complementary to the borders of the pocket. The marked discrimination between Phe B24 and Tyr B24 suggests that at least one portion of this border is rigid in the hormonereceptor complex. Understanding how the para-OH group of Tyr B24 impairs binding will require a higher resolution structure of the IR complex. We imagine that the smaller size and increased polarity of His (relative to Phe) accounts for its essential inactivity (relative affinity Ͻ1%; see also Ref. 12), whereas the larger bicyclic structure of Trp may make unfavorable contacts with the cavity walls. [Met B24 ,Orn B24 ]Insulin also displayed native or near-native in vivo potency in male Lewis rats (Fig. 9).
With the exception of Gly B24 , the above structure-activity relationships are in general accordance with occupancy of a delimited nonpolar cavity (63). When expressed in terms of differences in free energies of association (RT ln(K a /KЈ a )), limited discrimination exists between Phe, Met, Leu, and Ile (Ͻ0.5 kcal/mol at room temperature). Why might Met be preferred even to this small extent? The side chains of Met, Ile, and Leu are similar in volume (64) but differ in shape and repertoire of conformations. We speculate that whereas branched side chains (Leu and Ile) are restricted in potential conformations, the linear side chain of Met might be more adaptable (65) in conforming to the dimensions of the cavity. Molecular modeling (MD simulation followed by energy minimization) of Met B24 at the IR interface supported the plausibility of this model (Fig. 10, C and D); essential features of the WT model (subjected to the same MD-energy minimization protocol) were retained (Fig. 10, A and B). The favorable bound conformation of Met B24 presumably compensates for its greater loss of conformational entropy relative to Leu or Ile.
We imagine that increasing the size of an L-amino acid side chain at B24 would enhance the occupied percentage of the B24-related cavity and so would reduce the size of any potential cavity penalty. Similarly, increasing the hydrophobicity of the side chain would reduce the polarity penalty. These expectations are in accordance with the activities of Val B24 and Thr B24 analogs. Although these side chains are of similar shape and size, a preference is shown for Val versus Thr (respective affinities 8 and 2.5% relative to WT insulin). This further indicates that the B24-binding pocket displays a strong preference for the binding of nonpolar side chains relative to polar side chains.
Anomalous Activity of [Gly B24 ]Insulin-Whereas the low affinity of [Ala B24 ]insulin may be attributed to a cavity penalty (due to the predicted packing defect at the variant interface (63)), the high activity of the Gly B24 analog has long posed a seeming paradox (10,11,49,66). How may an unanchored hormone-receptor interface be tolerated? This enigma is deepened by the enhanced affinity of [D-Ala B24 ]insulin and other D-analogs at B24 (10,12,30). Although destabilization of the B20 -B30 segment (Gly B24 ) (6) or its frank detachment (D-Ala B24 ) (30) might reduce the cost of induced fit (and so enhance affinity), an alternative model has envisaged a one-residue shift in register between the C-terminal B chain ␤-strand and the site 1 surface. 12 In this model, Phe B25 would occupy the B24-binding pocket as an equivalent anchor; Tyr B26 would occupy the B25 binding pocket (which would be expected to be well tolerated (53)), and Thr B27 would occupy the exposed B26-related surface observed in the IR complex.
The register shift model may rationalize why small polar or charged side chains at position B24 (Ser, Thr, Asn, or Asp) confer a higher affinity (ϳ1%) than does the nonpolar side chain of L-Ala (Ͻ 0.5%). Although the differences are small, an intriguing idea is that higher observed affinities of [Ser B24 ]insulin and related analogs reflect the coexistence of two binding modes, one resembling WT insulin in which the B24-related cavity is incompletely filled and the other resembling the register-shifted model envisaged for [Gly B24 ]insulin. The latter scheme may be favored (relative to Ala B24 analog) due to possible hydrogen bonding by the side chains of Ser, Thr, Asn, or Asp to the main chain as in turns. We imagine that such compensation is incomplete, i.e. the putative GERGX element (where X indicates the B24 substitution) Ser, Thr, Asn, or Asp is less compatible, relative to Gly B24 or D-Ala B24 , with the novel pentaloop conformation. These possibilities may in the future be tested through photo-cross-linking studies of insulin analogs containing one or another of the B24 substitutions together with para-azido-Phe at position B25 or B26 (17,18). A shift in register would be associated with a shift in pattern of domain-specific photo-cross-linking in respective holoreceptor complexes between L1 and ␣CT. Coexistence of native-like and register-shifted binding modes would in turn be implied by dual photo-cross-linking at each site to both L1 and ␣CT.
Evolutionary Constraints-To probe why neither Met nor Gly was selected by nature as an alternative to Phe B24 among vertebrate insulin sequences (7), we evaluated three other aspects of structure. (i) Thermodynamic stability: in the zincfree monomer Met B24 and Gly B24 are each associated with reduced free energies of unfolding (⌬⌬G u ϭ 2.2 Ϯ 0.1 and 2.2 Ϯ 0.2 kcal⅐mol Ϫ1 , respectively). Such destabilization is likely to extend to the variant proinsulins and could be associated with impaired efficiency of disulfide pairing in pancreatic ␤-cells (48,67). (ii) Susceptibility to fibrillation: just as a zinc-free Gly B24 variant exhibited a decreased lag time (in the context of KPinsulin (6)) relative to its parent monomer, analogous studies of [Met B24 ,Orn B29 ]insulin demonstrated marked foreshortening of the lag time relative to [Orn B29 ]insulin. (iii) Hexamer assembly: [Gly B24 ]insulin has previously been shown to impair R 6 hexamer assembly (48). In this study, we observed analogous attenuation of the R 6 -specific Co 2ϩ -visible absorption band of the Met B24 analog with more rapid disassembly of the variant R 6 hexamer relative to WT insulin or [Orn B29 ]insulin. Although insulin is stored in the secretory granules of pancre-atic ␤-cells as micro-crystals of zinc insulin hexamers (68,69), the structure of these hexamers is not known (i.e. whether R 6 , T 3 R f 3 , T 6 , or in a novel allosteric state). It is possible that the perturbed spectroscopic and kinetic features of the [Met B24 ,Orn B29 ]insulin hexamer might be associated with perturbed packaging and storage in secretory granules.
Our findings highlight the multiple biological constraints that may have governed the evolution and divergence of vertebrate insulins. By analogy to the endoreticular stress previously described in studies of ␤-cell lines expressing [Ser B24 ]proinsulin (48,67), we speculate that Met and Gly have been excluded at position B24 due to structural requirements of biosynthesis and storage (70). The general conservation of insulin sequences among vertebrates, which is more stringent than that typically observed in globular domains (2), may reflect the multiple roles played by specific side chains in the course of a complex "conformational life cycle" from nascent folding to receptor binding. Of particular interest, we envisage that a combination of inefficient or unstable disulfide pairing, perturbed hexamer assembly, and heightened susceptibility to fibrillation might be associated with a risk of toxic protein deposition as an amyloidogenic disease (70). Proof of principle is provided by the selective deposition of insulin fibrils in the islet of Langerhans as observed in the South American rodent Octodon degus (71), whose insulin contains a divergent B chain sequence (72).
Insulin-like Growth Factor System-We anticipate that the homologous B domains of IGF-I and IGF-II undergo hinge-like detachment on binding to site 1 of the type 1 IGF receptor (IGF-1R) similar to that observed on binding of insulin to the IR model (6). Indeed, insulin, IGF-I, and IGF-II are each capable of binding to IR (isoforms A and B) and IGF-1R (1). In structures of free IGFs (73)(74)(75)(76), corresponding B22-B26 segments exhibit almost identical dispositions (with respect to the ␣-helical core) as in insulin; the structure of L1 and sequence of ␣CT are also conserved between IR and IGF-1R (5). We thus envision that Phe 23 in IGF-I and Phe 26 in IGF-II (homologs of Phe B24 in insulin) function as corresponding anchor residues within homologous nonpolar pockets. This proposal is supported by a detailed correspondence of structural features between these homologous signaling systems. The B20-B23 ␤-turn (sequence GERG in insulin and maintained in the IR complex) is conserved among IGFs both in sequence (GDRG) and structure (75,76), suggesting that the homologous turn is also maintained as IGFs engage IGF-1R. Furthermore, the aromatic triplet in insulin (Phe B24 -Phe B25 -Tyr B26 ) is conserved in IGFs (as Phe-Tyr-Phe) as are the cognate binding residues in IR and IGF-1R (Fig. 11A). An apparent violation is at L1 residue Phe 39 in the IR; this nonpolar side chain in the Phe B24 binding pocket of the IR is substituted by Ser in IGF-1R. However, this exception reflects a two-residue insertion in the L1 domain of IGF-1R that repositions the Ser outside of the predicted binding pocket (Fig. 11B). 13 The detached model of the IGF-IGF-1R interface is broadly consistent with the structure of the IR complex. On replacement of Phe B25 by Tyr in IGFs, for example, the para carbon (C ) would be directed away from ␣CT and L1, and so Tyr could readily be accommodated. The Phe in IGFs corresponding to Tyr B26 likewise poses no conflict as its contact residues (L1 residues Arg 14 and Asp 12 ) are identical in IGF-1R (Fig. 11A). In accordance with these expectations, substitution of Phe B25 by Tyr is well tolerated in insulin (77), whereas substitution of Tyr B26 by Phe results in only a small reduction in binding to the insulin receptor (78). Despite such similarities, a salient difference distinguishes the IGFs, their respective C domains linking B and A domains. Modeling suggests that, due to their length and flexibility, these linking peptides would not constrain segmental detachment of the respective B domain ␤-strands from the ␣-helical cores of these single-chain growth factors. Furthermore, the bound position of the B24 -B27 segment (running anti-parallel to the central ␤-sheet (L1-␤ 2 ) of L1) suggests that the IGF C domains would be directed toward the CR domain of the receptor (whether IR or IGF-1R). We speculate that the shorter C domain of IGF-II (relative to IGF-I) may explain its higher affinity for the A isoform of IR (1).
Clinical Pharmacology and Nonstandard Mutagenesis-A pioneering application of protein engineering in pharmacology was provided by the design of rapid-acting insulin analogs (79). The essential goal was to accelerate the disassembly of insulin hexamers in the subcutaneous depot, thereby facilitating the absorption of insulin into the bloodstream (80). This was accomplished by interchange of residues Pro B28 and Lys B29 (in Humalog (23); KP-insulin in Table 3) and alternatively by substitution of Pro B28 by Asp (in Novolog (79)). These modifications lie at the edge of the dimer-related anti-parallel ␤-sheet (residues B24 -B26 and B26Ј-B24Ј) in the hexamer (2). Although the Cha B24 substitution receptor binding led to a 2-3-fold attenuation of receptor-binding affinity in vitro (in both the Orn B29 and KP frameworks; Tables 2 and 4), full potency was retained in a rat model of DM. Molecular modeling suggested that this cyclic aliphatic side chain could fill the B24related pocket without steric clash (Fig. 10, E and F).
Although our studies of Cha B24 analogs were motivated by the unexpected high affinity of [Met B24 ,Orn B29 ]insulin, it is possible that this analog might confer clinical benefits. Intriguingly, substitution of Phe B24 by Cha led to accelerated disassembly of [Orn B29 ]insulin and to a further acceleration in the disassembly of KP-insulin. Although these assays employed cobalt insulin hexamers, we expect that these trends would extend to the zinc insulin hexamers as conventionally employed in insulin formations (81). Such ultra-rapid dissociation of insulin hexamers, if pertinent to the subcutaneous depot, is of current interest in relation to the safety and efficacy of "smart" insulin pumps in which a computer-based algorithm couples the output of a continuous monitor to insulin infusion rate (82). Such closed-loop systems promise to enhance glycemic control in type 1 DM. The utility of the Cha B24 derivative of KP-insulin in continuous pump-based infusion may be further enhanced by its augmented resistance to fibrillation relative to KP-insulin or WT insulin.
We speculate that the enhanced rate of disassembly of Cha B24 cobalt hexamers is related to the flexibility of the aliphatic ring as visualized in the present high resolution crystal structure. Indeed, the electron density of the Cha B24 side chain indicated the presence of at least two conformational substrates, ascribed to gauche and anti-gauche chair conformations. Their interconversion (with respect to the C ␤ -C ␥ bond), which seems to occur independently at the three dimer interfaces, may facilitate access to transition states for hexamer dissociation. Although mechanisms of WT hexamer disassembly are not well understood, the nature of such transition states may in principle be probed by MD simulations at elevated temperature and at prolonged time scales (83).
Concluding Remarks-Phe B24 , invariant among vertebrate insulins and IGFs (7), plays a key role in the three-dimensional structure of the hormone and its self-assembly (2). A recent co-crystal structure of insulin bound to a fragment of the insulin receptor has further demonstrated that the aromatic ring of Phe B24 inserts into a nonpolar pocket at the interface between the L1 ␤-helix of the receptor ␣ subunit, ␣CT residue Phe 714 , and the central B chain ␣-helix (6). Such side-chain insertion appears to anchor the displaced B24 -B27 segment in a groove between L1 and ␣CT. The B24-related pocket in the IR complex is lined by aromatic (Tyr B16 , Phe 39 , and Phe 714 ), aliphatic (Leu B15 and Leu 37 ), and polar (Asn 15 ) residues. Like Phe B24 itself, the lining residues are broadly conserved among vertebrate insulins, IGFs, and their respective receptors (5).
This study exploited systematic mutagenesis to define structure-activity relationships at position B24. Surprisingly, we have found that aromaticity (long thought to be central (10,12)) is not required to achieve high affinity binding in vitro and native potency in a rat model of DM. Although the lining of the B24-related pocket in principle offers an opportunity for weakly polar interactions (which may be provided by aromatic-aromatic and aromatic-carboxyamide contacts (84)), the high affinity and biological activity of analogs containing Met B24 and Cha B24 provide evidence that any such directional interactions are replaceable by an aliphatic anchor. It is possible that the similar affinities of these analogs masks complex and compensating differences in underlying thermodynamic driving forces. The general resemblance of the B24-related pocket in the IR complex (6) to a "druggable" binding pocket (as seen in a variety of pharmaceutical targets) and its tolerance to modified "ligands" (i.e. the present B24 modifications) suggest that nonstandard insulin analogs or IR agonists may be of therapeutic utility in the future.
Why has Phe B24 been conserved within the vertebrate insulin family, an evolutionary history extending for 500 million years? Such invariance is particularly striking given the following: (i) the essentially native activities of Gly B24 and Met B24 analogs; (ii) the divergence generally observed among hydrophobic cores of globular proteins over this time scale (85); and (iii) the corresponding plasticity of core packing in model systems as revealed by random combinatorial mutagenesis (86). In light of the clinical association between Ser B24 (16) and a monogenic form of DM (15) linked to endoreticular stress in pancreatic ␤-cells (67), we speculate that Phe B24 confers a uniquely favorable combination of foldability, avoidance of misfolding, native assembly, stability, receptor binding and hormonal signaling. We thus envisage that, at this and other sites invariant among vertebrate insulins and IGFs, exploration of sequence space has become frozen at the narrow intersection of multiple biophysical and biological constraints.