Mutations in human insulin cause an autosomal-dominant syndrome of diabetes and fasting hyperinsulinemia. We demonstrate by residue-specific photo cross-linking that diabetes-associated mutations occur at receptor-binding sites. The studies use para-azido-phenylalanine, introduced at five sites by total protein synthesis. Because two such sites (ValA3 and PheB24) are largely buried in crystal structures of the free hormone, their participation in receptor binding is likely to require a conformational change to expose a hidden functional surface. Our results demonstrate that this surface spans both chains of the insulin molecule and includes sites of rare human mutations that cause diabetes.
The insulinopathies describe a monogenic form of adult-onset diabetes due to mutations in the insulin gene (1,2). Patients respond normally to exogenous insulin but exhibit fasting mutant hyperinsulinemia due to delayed receptor-mediated clearance of the variant hormone (2). Inheritance is autosomal dominant with variable penetrance. The presence of one wild-type allele permits normal growth and development; homozygous or hemizygous mutations have not been observed and are presumably incompatible with life. Diabetes-associated mutations may either impair binding of the variant insulin to its receptor or perturb subcellular trafficking and processing of the variant proinsulin in the pancreatic β-cell (2). Mutations that impair binding have been identified at three invariant sites: ValA3 → Leu, PheB24 → Ser, and PheB25 → Leu. By analogy to the nomenclature describing abnormal hemoglobins, these are designated insulins Wakayama, Los Angeles, and Chicago, respectively (1). We demonstrate here that these mutations occur at contact sites between insulin and the α-subunit of the insulin receptor.
The structure of insulin is well characterized by crystallography (3) and nuclear magnetic resonance spectroscopy (4,5) (Fig. 1A). Residues A3, B24, and B25 exhibit distinct environments. Whereas PheB25 projects from the surface, ValA3 and PheB24 are engaged in long-range interactions (Fig. 2). ValA3 contacts TyrB26 and ProB28 at an interface between the NH2-terminal A-chain α-helix and COOH-terminal B-chain β-strand (Fig. 2B). PheB24 packs against ValB12, LeuB15, TyrB16, and CysB19 to stabilize the supersecondary structure of the B-chain. In dimers and hexamers, PheB24 and PheB25 also participate in an intermolecular β-sheet, an essential element of insulin’s storage form in the β-cell (3). Whereas considerable evidence indicates that the exposed side chain of PheB25 contacts the insulin receptor (6, including previous photo cross-linking studies [7,8]), the roles of ValA3 and PheB24 have long been the subject of speculation (3,9–13).
To test whether residues A3, B24, and B25 contact the insulin receptor, we have synthesized insulin analogs containing a photo-activatable derivative of phenylalanine (Phe), para-azido-Phe (Pap) (8,14). Pap was chosen based on its rigidity and small size (relative to other photoactivable moieties), thus limiting the distance range for cross-linking. Modified A- and B-chains were prepared by solid-phase synthesis using the photostable precursor para-amino-Phe. To enable efficient detection of cross-linked peptides, the α-amino group of the B-chain was biotinylated (8). The nonstandard side chain was introduced into an engineered insulin monomer (DKP-insulin, which contains three B-chain substitutions: HisB10→Asp, ProB28→Lys, and LysB29→Pro), chosen as a template for its efficiency of synthesis, enhanced receptor binding, and absence of confounding self-association (4). A3, B24, and B25 para-amino-Phe analogs exhibit respective receptor-binding affinities of 2.0 ± 0.2, 59 ± 2, and 147 ± 3% relative to native insulin (Kd 0.48 nmol/l); the affinity of the biotin adduct of DKP-insulin is 132 ± 5% (assays performed in triplicate). Corresponding analogs were prepared at terminal positions A1 and A21 (ordinarily conserved as Gly and Asn, respectively) at the periphery of insulin’s putative receptor-binding surface (3,15); their relative receptor-binding affinities are 46 ± 5 and 79 ± 22%. AsnA21 projects from the surface near GlyB23 and PheB25 (Fig. 2A). Conversion of para-amino-Phe to Pap in the intact hormones was verified by mass spectrometry.
A3, B24, and B25 Pap analogs each exhibit rapid and efficient cross-linking to the ectodomain of the receptor on ultraviolet irradiation (red asterisks in Figs. 1B and C). Efficiency (defined as the probability of photo cross-linking of the Pap derivative once bound to the receptor) is highest for PapA3. Similar results are obtained with the lectin-purified holoreceptor. No covalent complex is observed in the absence of irradiation or in control studies of the para-amino-PheB25 precursor. Photo cross-linking is successively diminished by the addition of native insulin or by higher concentrations of IGF-1 (Fig. 1D). In contrast to the cross-linking at sites of clinical mutation, photo cross-linking of Pap derivatives at A1 and A21 is markedly less efficient (green arrows in lanes 6 and 14 in Fig. 1C; relative to A3, complexes are reduced by 19- and 4-fold, respectively). As a first step in identifying sites of cross-linking in the insulin receptor, we used partial proteolysis with trypsin and chymotrypsin to characterize fragments of the receptor α-subunit covalently bound to insulin. Analysis of such fragments demonstrates that PapB24 contacts the NH2-terminal L1 β-helix domain, the major hormone-binding region of the receptor (13). By contrast, PapB25 contacts the COOH-terminal region of the α-subunit in accordance with the pioneering study of Kurose et al. (8). Edman sequencing demonstrated that PapB25 cross-links to tryptic peptide 704-718 in the insert-domain tail. (The COOH-terminal residue of the α-subunit is 731.) In our hands, limited chymotryptic digestion of the covalent hormone-receptor complex, followed by reduction with dithiothreitol (DTT), yields a 34-kDa adduct that, on further digestion, yields a 20-kDa adduct. Following enzymatic deglycosylation, the apparent mass of this fragment is 14 kDa and thus contains about 120 amino acids. The results of Kurose et al. imply that the latter fragment contains the COOH-terminal portion of the α-subunit derived from the second fibronectin-homology domain (FnIII1) and insert domain (ID) (13).
PapA3, which is not predicted to contact the receptor in a current model based on electron-microscopic image reconstruction (16), cross-links the COOH-terminal to the L1 and cysteine-rich domains. To localize this site more precisely, a second PapA3 derivative was prepared in which the biotin tag was attached at the NH2-terminus of the A-chain (rather than the NH2-terminus of the B-chain; see research design and methods). This design facilitates mapping following DTT reduction as above. Limited chymotryptic digestion thus demonstrates that PapA3 cross-links to the same COOH-terminal 34-kDa and 20-kDa adducts as PapB25, i.e., within the FnIII1-ID–derived tail. We suggest that the A3 binding site (like that of PapB25 ) resides within the ID-derived portion, since the FnIII1 moiety may be deleted in active fragments of the α-subunit (13). Because the L1 domain and the COOH-terminal domain are distant in the sequence of the α-subunit, the present results suggest that these and other sites of photo cross-linking are nearby in the three-dimensional structure of the hormone-receptor complex (13,16). It is not known whether PapA3 and PapB25 cross-link to the same α-subunit or to dimer-related α-subunits within the α2β2 heterotetramer. Sites of weak cross-linking by PapA1 and PapA21 derivatives were not characterized.
Photo cross-linking of PapA3 and PapB24 derivatives is of structural interest. Because ValA3 and PheB24 are largely buried in crystal structures of insulin (3), it has been unclear whether these residues contact the receptor or serve as structural supports. A possible role for PheB24 in redirecting the main chain of insulin on receptor binding has been proposed based on the unexpectedly high activities of d-amino acid substitutions (9). We and others have hypothesized that detachment or reorganization of the COOH-terminal region of the B-chain near B24 exposes the side chains of PheB24 and ValA3 and thus enables them to contact the receptor (11–13,17). The present results support (but do not establish) this hypothesis. Although we cannot exclude that Pap-mediated contacts are probe dependent (i.e., not ordinarily made by PheB24 or ValA3), a direct interaction would rationalize the exquisite sensitivity of binding at each site to subtle modifications (such as TyrB24, AlaA3, ThrA3, and LeuA3) (9,10). A direct interaction is consistent with the structure and function of a truncated insulin analog lacking the COOH-terminal five residues of the B-chain (residues B26–B30) (magenta in Fig. 1A). In the crystal structure of this analog, ValA3 is exposed in an otherwise native-like conformation (18). When the new COOH-terminus is amidated, this analog is fully active (19). Conversely, tethering the COOH-terminal segment of the B-chain to the A-chain yields a native-like single-chain analog with essentially no biological activity (17). Furthermore, a conformational change in the B-chain would rationalize the low activity of a “chiral” analog in which the internal side chain of IleA2 (also shielded by TyrB26 and ProB28) is substituted by allo-isoleucine (10). This modification does not perturb the structure or stability of insulin but would alter its “hidden” functional surface (20).
In the decades since the crystal structure of insulin was elucidated in 1969 by D. Hodgkin et al. (3), the residues required for its function have been extensively investigated by mutagenesis and chemical modification (3,13). Interpretation of these results is incomplete, however, as such approaches do not generally distinguish between side chains that contact the receptor and those required to stabilize insulin’s active conformation. By exploiting site-specific photo cross-linking, the present studies strongly suggest that sites of clinical mutations (1,2) are in direct contact with the insulin receptor. A molecular understanding of such contacts, likely to emerge from a crystal structure of the hormone-receptor complex, may enable design of nonpeptide insulin agonists for the treatment of diabetes.
RESEARCH DESIGN AND METHODS
Insulin analogs were synthesized and purified as described (4,8). The relative receptor-binding affinities of para-amino-Phe analogs were determined by competitive displacement of 125I-insulin from a human placental membrane preparation, as previously described (10). Conversion of such analogs to Pap derivatives was effected as described (4). Lectin-purification of the insulin receptor overexpressed in cell line P3-A was performed by the procedure of Yoshimasa et al. (21). Photo cross-linking reactions were performed at high concentrations of hormone and receptor (ca. 200 nmol/l) to enable essentially complete binding of the PapA3 analog. Short-wave ultraviolet light (254 nm) generated from a Mineralight Lamp (Model UVG-54; UVP, Upland, CA) was used with an optimum exposure time of 20 s and a distance of 1 cm from the light source. Identification of photo cross-linked receptor domains utilized prior characterization of chymotryptic and tryptic sites (22,23). Western blots used Neutravidin (Pierce, IL) and a polyclonal anti-receptor antiserum that recognizes the NH2-terminal region of the α subunit (N-20; Santa Cruz Biotech, Santa Cruz, CA). For such mapping studies, analogs contained an amino-caproyl-biotin tag either at the α-amino group of residue B1 or, in the case of the second PapA3 derivative, the ε-amino group of d-lysine introduced in place of glycine at position A1. Domain-mapping studies are supported by SDS-PAGE analysis of cross-linked fragments following enzymatic deglycosylation as described (24).
We thank G.D. Smith and C. Yip for kindly providing the receptor ectodomain and D.F. Steiner for a mammalian cell line overexpressing the human insulin receptor. S.H.N. was supported in part by the Diabetes Research & Training Center of the University of Chicago. This work was supported in part by grants from the National Institutes of Health to P.G.K. (DK56673) and M.A.W. (DK40949).