Contribution of Residue B5 to the Folding and Function of Insulin and IGF-I

Abstract

Proinsulin exhibits a single structure, whereas insulin-like growth factors refold as two disulfide isomers in equilibrium. Native insulin-related growth factor (IGF)-I has canonical cystines (A6—A11, A7–B7, and A20—B19) maintained by IGF-binding proteins; IGF-swap has alternative pairing (A7–A11, A6—B7, and A20—B19) and impaired activity. Studies of mini-domain models suggest that residue B5 (His in insulin and Thr in IGFs) governs the ambiguity or uniqueness of disulfide pairing. Residue B5, a site of mutation in proinsulin causing neonatal diabetes, is thus of broad biophysical interest. Here, we characterize reciprocal B5 substitutions in the two proteins. In insulin, His^B5 → Thr markedly destabilizes the hormone (ΔΔG_u 2.0 ± 0.2 kcal/mol), impairs chain combination, and blocks cellular secretion of proinsulin. The reciprocal IGF-I substitution Thr^B5 → His (residue 4) specifies a unique structure with native ¹H NMR signature. Chemical shifts and nuclear Overhauser effects are similar to those of native IGF-I. Whereas wild-type IGF-I undergoes thiol-catalyzed disulfide exchange to yield IGF-swap, His^B5-IGF-I retains canonical pairing. Chemical denaturation studies indicate that His^B5 does not significantly enhance thermodynamic stability (ΔΔG_u 0.2 ± 0.2 kcal/mol), implying that the substitution favors canonical pairing by destabilizing competing folds. Whereas the activity of Thr^B5-insulin is decreased 5-fold, His^B5-IGF-I exhibits 2-fold increased affinity for the IGF receptor and augmented post-receptor signaling. We propose that conservation of Thr^B5 in IGF-I, rescued from structural ambiguity by IGF-binding proteins, reflects fine-tuning of signal transduction. In contrast, the conservation of His^B5 in insulin highlights its critical role in insulin biosynthesis.

Introduction

The vertebrate insulin-related superfamily consists of insulin and insulin-related growth factors (IGF-I⁵ and IGF-II) (1, 2), relaxin (3–5), and relaxin-related factors (6–9). Insulin and IGFs function as ligands for receptor tyrosine kinases (the insulin receptor and type 1 IGF receptor; IR and IGF-1R) (10), whereas relaxin and related factors bind to G-protein-coupled receptors (11). Interest in the evolution and folding properties of insulin-related polypeptides has recently been invigorated by the discovery of mutations in the human insulin gene associated with permanent neonatal-onset diabetes mellitus (12). These dominant mutations impair the foldability of variant and (in trans) wild-type proinsulin, leading to β-cell dysfunction, endoplasmic reticular (ER) stress, and impaired β-cell viability (13). One such mutation occurs at position B5 (12, 14). Insulin contains a conserved His at B5, whereas IGF-I contains a conserved Thr (Table 1). Here, we investigate reciprocal substitutions in these proteins, Thr^B5 in insulin and His^B5 in IGF-I, as probes of competing evolutionary constraints among otherwise homologous sequences.⁶

TABLE 1.

Sequences of human insulin and related factors

Conservation among vertebrate species is illustrated in supplemental Table S1. Residue at position B5 (IGF position 4) is shown in green; conserved sites in the insulin superfamily are highlighted in red. Phe^B24 (boldface) is proposed to function as a structural switch on binding of insulin and IGFs to their cognate receptors. Residues A1–A3 in insulin and IGFs (underlined) participate in receptor binding but are not conserved among relaxin-related factors, which bind to an unrelated class of receptors.

IGFs are single-chain polypeptides containing A- and B-domains, an intervening connecting (C)-domain, and a C-terminal D-domain (Fig. 1A) (1, 2); insulin (like relaxin and related factors) contains two chains (designated A and B; Fig. 1B) as a consequence of proteolytic processing in the trans-Golgi network (3–5, 15). Crystal structures of IGF-I and insulin exhibit similar α-helical domains (Fig. 1, A and B) (1–9). Protein folding is linked to specific disulfide pairing. The canonical cystines in insulin are A6—A11, A7–B7, and A20—B19, and the corresponding cystines in IGF-I are at polypeptide positions 47–52, 6–48, and 18–61. Whereas the six cysteines and selected core residues are broadly conserved throughout the vertebrate insulin-related superfamily (Table 1; see also supplemental Table S1), other residues are restricted to particular proteins. An example is residue B5; its restricted conservation suggests distinct IGF- and insulin-specific contributions to structure or function.

FIGURE 1.

Structures of IGF-I and insulin. A, ribbon model of IGF-I (PDB accession code 1PMX). The A-domain is shown in red, B-domain in blue, and C- and D-domains in gray. The three disulfide bridges are shown in gold (canonical numbering A7–B7, A6–A11, and A20–B19; IGF residues 6–48, 47–52, and 61–18); balls represent sulfur atoms. Side chain of Thr⁴ (canonical position B5) is also shown (green). B, ribbon model of human insulin (T-state protomer; PDB accession code 4INS). The coloring scheme is as in A. His^B5 is shown in green. C, position of Thr⁴ (canonical position B5) adjoining cystine A6–A11 (IGF residues 47–52) in inter-domain crevice of IGF-I. The β-OH group and γ-CH3 of Thr⁴ (Thr^B5) are shown as green and black spheres; neighboring A-chain carbonyl oxygens are shown as red spheres. D, position of His^B5 in inter-chain crevice of insulin. Imidazolic NH is shown as green sphere, and neighboring A-chain carbonyl oxygens in red spheres. Whereas His^B5 consistently engages in hydrogen bonding to the A-chain in crystal structures of insulin (T-state protomers), Thr⁴ in IGF-I engages in analogous hydrogen in only a detergent-stabilized crystal structure and not in multiple other structures (see supplemental Table S2).

B5 side chains occupy similar but not identical structural environments in IGF-I and insulin. The side chains project in each case into an inter-domain crevice adjoining cystine A6—A11 and buttressing the solvent-exposed A7–B7 disulfide bridge (Fig. 1, C and D) (15–17). The respective N-terminal arms of IGF-I and insulin (residues 1–3 and B1–B4, respectively) exhibit different orientations (Fig. 1, A and B, arrows). An imidazolic NH of His^B5 in insulin (Fig. 1D, green ball) donates one or bifurcating hydrogen bonds to A-chain carbonyl oxygens (red balls; see also supplemental Table S2), proposed to facilitate the folding of proinsulin (18, 19). An Ala^B5 substitution in insulin impairs chain combination, whereas binding to the insulin receptor (IR) is decreased by less than 2-fold (18). Although the side chain of Thr^B5 in IGF-I is not well ordered in solution (20, 21) and variably positioned in crystal structures (16, 17), its β-OH group (Fig. 1C, green ball) can be oriented to donate an analogous hydrogen bond to an A-domain carbonyl oxygen (red) (see supplemental Table S2). His^B5-IGF-I exhibits native interactions with IGF-binding proteins (IGFBP) with 2-fold enhanced binding to the IGF-1R (22).

IGF sequences anomalously encode two alternative structures in equilibrium (Fig. 2A) (23–27). Native IGF-I retains canonical disulfide pairing, whereas IGF-swap exhibits the alternative pairing A7–A11, A6–B7, and A20–B19 (IGF-I positions 46–52, 6–47, and 18–61). The non-native isomer is at least 20-fold less active (26). ¹H NMR studies have demonstrated that IGF-swap possesses a well organized three-dimensional structure with salient differences from that of native IGF-I (26, 28). The alternative structure is as stable (or more stable) than native IGF-I as probed by thiol-catalyzed disulfide exchange (26) and chemical denaturation (29). Proinsulin in contrast refolds to form a unique ground state (30); disulfide isomers exist only as metastable kinetic traps (29, 31, 32). Striking similarities are nonetheless observed between the solution structures of IGF-swap and insulin-swap (Fig. 2B) (28). Such structural similarity suggests that one or a few sites of sequence difference might account for the contrasting refolding properties of proinsulin and IGF-I. The defective folding properties of IGF-I are circumvented in vivo by selective binding of native IGF-I to IGFBPs (33, 34).

FIGURE 2.

Structures of disulfide isomers. A and B, corresponding ribbon models of native insulin (A) and insulin-swap (B). A-chain α-helices are shown in red: (native insulin) A1–A8 and A12–A18 and (insulin-swap) A12–A18. The A-chains are otherwise shown in gray. In each structure the central B-chain α-helix (residues B9–B19) is shown in blue; the B-chain is otherwise gray. His^B5 in each case is shown in green (PDB accession numbers 4INS and 1XGL). Disulfide pairing is as labeled (boxes). C, stereo pair showing crystal structure of mini-IGF-I-swap (PDB accession code 1TGR). The crystal structure exhibits additional α-helical segments relative to the solution structure of intact IGF-swap (purple): residues B4–B9 and A1–A8. Canonical A- and B-domain α-helices are shown in red and blue, respectively.

Studies of inactive mini-IGF-I models (lacking C- and D- domains) and corresponding mini-proinsulin analogues suggest that this distinction is mediated by differences between B-domains (35–37). Construction of chimeric models has highlighted the proposed role of residue B5; substitution of Thr⁴ in mini-IGF-I by His led to unique canonical pairing (19). Interpretation of these studies has remained uncertain, however, as the crystal structure of mini-IGF-I-swap (38) (Fig. 2C) was found to exhibit marked and unexpected differences from the structures of either IGF-swap (28) or insulin-swap (Fig. 2B) (32).⁷ These structural differences imply that the conformational equilibrium between mini-IGF-I and mini-IGF-I-swap may not be representative of the intact proteins.⁸ It is therefore of interest to investigate the effects of His^B5 in native IGF-I and the effects of Thr^B5 in insulin. To this end, we have introduced reciprocal B5 substitutions into insulin and IGF-I. Our studies yield asymmetrical outcomes. In insulin Thr^B5 profoundly destabilizes the native state but does not promote isomer formation. In IGF-I His^B5 enables unique disulfide pairing (native IGF-I) by selectively destabilizing the competing fold (IGF-swap).

Why is a folding-defective residue excluded in insulin but conserved in IGF-I? On the one hand, we demonstrate that Thr^B5 blocks the secretion of proinsulin in a human cell line and so presumably would cause β-cell dysfunction in vivo (39). On the other hand, IGF-IGFBP co-evolution has circumvented the role of residue B5 in folding, allowing sequence variation associated with fine-tuning of IGF-1R signaling in growth, development, and cellular homeostasis. Taken together, our results suggest that the absence or presence of folding partners can profoundly alter the fitness landscape of a protein sequence.

MATERIALS AND METHODS

Synthesis of Thr^B5-Insulin

The protocol for solid-phase synthesis is as described previously (40). The wild-type A-chain was obtained by oxidative sulfitolysis of human insulin as described previously (18, 41). Thr^B5 B-chain was prepared by automated Fmoc (N-(9-fluorenyl)methoxycarbonyl)-based synthesis and derivatized by S-sulfonation. Insulin chain combination was achieved as described previously (41). Purification of Thr^B5-insulin was accomplished by cation-exchange chromatography (CM52, 24 × 1.2 cm, Whatman) using acetate buffer (pH 3.3) and a NaCl gradient. Final purification was by reverse-phase high-performance liquid chromatography (rp-HPLC) using a C18 column with a 0.1% trifluoroacetic acid/acetonitrile solvent system. Combination of 40 mg of A-chain and 20 mg of the variant B-chain resulted in a final yield of 1.1 mg. The corresponding yield of wild-type chain combination under these conditions is ∼4.0 mg. The purified analogue was observed to be a single component on analytical rp-HPLC (C18 column, 25 × 0.46 cm) using two independent elution gradients, acetonitrile in 0.1% trifluoroacetic acid and methanol in 0.1% trifluoroacetic acid. Because insulin isomers exhibit very low affinity for IR, native disulfide pairing was implied by the higher residual receptor-binding affinity of the analogue (Fig. 3C; see “Results”). The yield of chain combination was reduced by 4-fold relative to wild-type chain combination; no non- native isomers were encountered in the course of chain combination. The matrix-assisted laser desorption-ionization (MALDI) mass spectrum of the product was in accord with its predicted value (5770.4 Da).

FIGURE 3.

CD and receptor-binding studies of Thr^B5-insulin. A, far-ultraviolet CD spectrum of Thr^B5-insulin (dashed line) in relation to spectrum of wild-type insulin (black line). Spectrum of analogue, acquired at 4 °C and pH 7.4, exhibits attenuated α-helical features. B, CD-detected guanidine titrations of Thr^B5-insulin (dashed line) in relation to titration of wild-type insulin (black line). Inferred values of ΔG_u are provided in Table 2. C, receptor-binding studies employing IR (isoform B). Competitive displacement assay in which pre-bound ¹²⁵I-labeled insulin is displaced by successive concentrations of unlabeled Thr^B5-insulin (▾) or wild-type insulin (▴) is shown. Ratio of bound-to-total tracer counts are plotted as a function of the logarithm (base 10) of the ligand concentration (nanomolar). D, analogous receptor-binding studies employing IGF-1R. Competitive displacement assay in which pre-bound ¹²⁵I-labeled IGF-I is displaced by successive concentrations of unlabeled Thr^B5-insulin (▾) or wild-type insulin (▴).

Synthesis of His^B5-IGF-I

Total chemical synthesis of His^B5-IGF-I was based on native chemical ligation of three unprotected peptide segments as described for native IGF-I (see supplemental Fig. S1) (42). His^B5-IGF-1[1–17]^αthioester-(Arg)₄-Ala-OH was prepared by in situ neutralization Boc chemistry stepwise solid-phase peptide synthesis (43) on HSCH₂CH₂CO-(Arg)₄-Ala-OCH₂-Pam-resin (44) (electrospray ionization-mass spectrometry: observed, 2581.2 ± 0.3 Da; calculated, 2581.9 Da). Ligation of IGF-I[Thz¹⁸-47]^αthioester and IGF-I[Cys⁴⁸-70] was complete in 14 h. IGF-I[Thz¹⁸-70] was quantitatively converted to IGF-I[Cys¹⁸-70] in 3 h by the addition of 0.2 m methoxylamine·HCl to the crude ligation mixture at pH 4. After solid-phase extraction and lyophilization, ligation of the His^B5-IGF-I[1–17]^αthioester and crude IGF-1[Cys¹⁸-70] was performed. This ligation made use of high concentration (200 mm) of (4-carboxylmethyl)thiophenol, an improved aryl thiol catalyst for high yield ligation at a hindered Val-Cys site (45). The His^B5-IGF-I[1–17]^αthioester-(Arg)₄-Ala-OH contained a solubilizing “Arg tag” in the C-terminal region (44). Reaction of IGF-I[1–17]^αthioester and IGF-I[Cys¹⁸-70] (2 mm each) was ∼70–80% complete in 18 h, yielding the final full-length reduced product. HPLC purification provided pure reduced full-length IGF-I(SH)₆ (see supplemental Fig. S2).

Cellular Expression and Folding of Proinsulin

Plasmids encoding human proinsulin or B5 variants in mammalian cell culture were constructed to enable analysis of protein folding and secretion as described previously (18, 46). A control for an uncleaved preproinsulin was provided by a cleavage-site mutation associated with neonatal diabetes mellitus (Asp^S24 in the signal sequence (12)). In brief, HEK293T cells (human) were cultured in high glucose Dulbecco's modified Eagle's medium containing 10% fetal bovine serum and 0.1% penicillin/streptomycin at 37 °C with 5% CO₂. For metabolic labeling, cells were plated into 6-well plates 1 day before transfection. Plasmid DNA (2 μg) was transfected into each well using Lipofectamine (Invitrogen). At 40 h post-transfection, cells were preincubated in methionine/cysteine-deficient medium with or without 10 μg/ml tunicamycin (TM) as indicated for 30 min, metabolically labeled in the same medium containing ³⁵S- labeled Met and Cys for 30 or 60 min, washed once with complete medium, and chased in complete medium with or without 10 μg/ml TM at different time points as indicated (35). After chase, media were collected, and cells were lysed in 100 mm NaCl, 1% Triton X-100, 0.2% sodium deoxycholate, 0.1% SDS, 10 mm EDTA, and 25 mm Tris-HCl (pH 7.4) with protease inhibitors. Lysates were immunoprecipitated with guinea pig anti-insulin antiserum (Linco Diagnostic) and analyzed by Tris/Tricine-urea-SDS-PAGE under nonreducing conditions or reducing conditions as indicated (18, 47). For treatment with peptide N-glycosidase F (PNGase F), the anti-insulin immunoprecipitates were incubated with PNGase F at 37 °C for 1 h prior to gel analysis.

Refolding of IGF-I and His^B5-IGF-I

Redox-coupled folding was effected in the following conditions: ∼0.5 mg/ml IGF-I(SH)₆ or His^B5-IGF-I, 20 mm Tris-HCl (pH 7.7), 8 mm cysteine, 1 mm cystine·HCl, and 0.5 m guanidine hydrochloride. Folding, as monitored by liquid chromatography-mass spectrometry, was almost complete in <1 min (to yield native IGF-I and IGF-swap) for wild-type sequence, whereas folding of the variant sequence (to yield a single product) was approximately half-complete at <1 min; folding was thus extended for 5 h. Whereas refolding of the native polypeptide yielded two predominant products, refolding of the variant yielded a single dominant product (see “Results”). Following oxidation, the mass of the protein decreased by 6.6 ± 0.7 Da in each case, indicating the formation of three disulfide bonds. Principal components of the folding reaction were purified by preparative C18 rp-HPLC, enabling native IGF-I, IGF-swap, and His^B5-IGF-I to be obtained. Respective overall yields of native IGF-I and His^B5-IGF-I were 6.7 and 8.3% from peptide segment IGF-I[Cys⁴⁸-70]. Disulfide pairing in the variant protein was established by NMR fingerprinting (see “Results”).

Synthesis of His^B5 Two-chain Hybrid

The wild-type insulin A-chain S-sulfonate was combined with a 29-residue B-chain derived from the B-domain of IGF-I in which Thr⁴ was substituted by His (canonical position B5). The IGF-I-derivated peptide was synthesized and purified as in the synthesis of human insulin analogues (41). Combination of 30 mg of A-chain and 15 mg of the variant B-chain resulted in a final yield of 0.9 mg. The corresponding yield of wild-type chain combination under these conditions is ∼3.0 mg. The MALDI-mass spectrum of the product was in accord with its predicted value (5562.8 Da).

Receptor Binding Assays

Dissociation constants for binding of insulin, IGF-I, and respective B5 analogues to IR were determined in competitive radioligand binding assays with ¹²⁵I-Tyr^A14 human insulin. The assay employed the B isoform of IR (IR-B). Experimental details are provided as supplemental material. In brief, assays were performed with the isolated IR-B with a C-terminal FLAG tag using a microtiter plate antibody capture technique as described previously (48). Microtiter strip plates (Nunc Maxisorb) were incubated overnight at 4 °C with FLAG M2 IgG (100 μl/well of 40 μg/ml in phosphate-buffered saline). In all assays, the percentage of tracer bound in the absence of competing ligand was less than 15% to avoid ligand-depletion artifacts. Dissociation constants of analogues were obtained by nonlinear regression analysis of binding data by the method of Wang (49); this employs an exact mathematical expression to describe the competitive binding of two different ligands to a receptor. A corresponding microtiter plate antibody assay using IGF-1R was employed to assess binding of insulin, IGF-I, and respective B5 analogues to this homologous receptor using ¹²⁵I-Tyr31 IGF-I as tracer (50). Results are summarized in Table 2. Control studies of cellular extracts in the absence of prior transfection of the epitope-tagged receptor constructs (either IR or IGF-1R) demonstrated that background binding to endogenous cellular proteins was negligible.

TABLE 2.

Activity and stability of protein analogues

Activity is defined by relative receptor-binding affinities for immobilized epitope-tagged IGF-1R or IR (isoform B). Values are in each case normalized to 100%. The absolute affinities of native IGF-I for IGF-1R and of insulin for IR are similar under the conditions employed (∼0.04 nm). Stability is inferred from guanidine denaturation studies based on a two-state model extrapolated to zero denaturant concentration. Uncertainties are ±0.1 kcal/mol. ND means not determined.

Protein	IGF-1R (Kd)	%	IR (Kd)	%	ΔGu
	nm		nm		kcal/mol
IGF-I	0.04 ± 0.005	100	13.9 ± 2.2	0.4 ± 0.1	2.9a
IGF-swap	2.8 ± 0.4	1.4 ± 0.19	ND	ND	3.7a
HisB5-IGF-I	0.022 ± 0.002	182 ± 16	5.7 ± 1.0	1 ± 0.2	3.1a
Insulin	8.7 ± 1.3	0.50 ± 0.08	0.06 ± 0.01	100	4.4b
ThrB5-insulin	53.8 ± 5	0.07 ± 0.001	0.33 ± 0.04	18.0 ± 2.2	2.4b

^a Studies were conducted in phosphate-buffered saline at 25 °C. ΔG_u errors are all ±0.1.

^b Studies conducted in 50 mm potassium phosphate (pH 7.4) at 4 °C. The stability of insulin in phosphate-buffered saline at 25 °C is 3.6 ± 0.1 kcal/mol. ΔG_u errors are all ±0.1.

Cellular Studies of IGF-I Signaling

Mouse anti-Akt antibody was from Cell Signaling Technologies (Beverly, MA); rabbit anti-phospho-Akt (Ser(P)⁴⁹³) was from Invitrogen; mouse Cy5-labeled IgG and rabbit Cy3-labeled IgG secondary antibodies were from GE Healthcare. An IGF-1R-deficient mouse embryo fibroblast cell line expressing the human IGF-1R was provided by Prof. R. Baserga (see Ref. 51). Akt phosphorylation as a parameter for activation of the IGF-IR signaling pathway was assayed essentially as described previously (52). In brief, cells were treated with 0.1–10 nm ligand for 5 min and lysed in SDS sample buffer without dithiothreitol or bromphenol blue and boiled immediately to inhibit protease and phosphatase action. Protein concentration was determined with a detergent-compatible protein assay kit (Bio-Rad). Dithiothreitol (100 mm) and bromphenol blue (0.1%) were then added. Whole-cell lysates (20 mg) were subjected to reducing SDS-PAGE on 10% Criterion gels (Bio-Rad), and resolved proteins were transferred to Immobilon FL polyvinylidene difluoride membranes (Millipore Corp., Billerica, MA). Blots were probed simultaneously with the Akt and phospho-Akt antibodies (1:2500 and 1:1000 dilutions, respectively), washed three times with Tris-buffered saline/Tween 20 (TBS-T), and then probed with a combination of the Cy-3- and Cy-5-labeled secondary antisera (1:5000 dilution each). Membranes were washed again three times with TBS-T buffer and dried, and fluorescent images were captured and quantified using a Fluorchem Q imaging system (Alpha Innotech, San Leandro, CA). Paired Student's t tests were used for all statistical analyses. Significance was accepted at p < 0.05.

Disulfide Reassortment

β-Mercaptoethanol-catalyzed disulfide exchange was assessed as described previously (26). In brief, purified IGF-I or His^B5-IGF-I was dissolved in 50 mm sodium phosphate (pH 7.5 at 20 °C) containing 1 mm EDTA at a protein concentration of 100 μg/ml (13 μm). The solutions were incubated at 20 °C after initiating the disulfide exchange reaction. To catalyze exchange, a freshly made stock solution of 0.02% (v/v) β-mercaptoethanol was added to a final concentration of 0.001% (v/v) (0.17 mm). Aliquots were removed after overnight incubation, quenched with 10-fold excess 0.1% trifluoroacetic acid, and analyzed by rp-HPLC. Samples were applied through an autosampler onto a Vydac RP-C4 column (214TP54, 250 × 0.46 cm; Grace Vydac, Hesperia, CA) at a flow rate of 1 ml/min. The proteins were eluted with an acetonitrile gradient using a solvent system consisting of 0.1% aqueous trifluoroacetic acid (solvent A) and acetonitrile containing 0.1% trifluoroacetic acid (solvent B). Proteins were also eluted with an alternative gradient system consisting of 0.1% trifluoroacetic acid and methanol. Protein elution was monitored at 215 nm using a dual-λ 2487 absorbance detector. Data acquisition and processing utilized the Waters Breeze HPLC software. A parallel sample set was also analyzed at 37 °C.

Circular Dichroism

Far-ultraviolet CD spectra were acquired using an Aviv spectropolarimeter equipped with thermistor control. Samples were made 30–50 μm in 140 mm NaCl and 10 mm sodium phosphate at pH 7.4 (phosphate-buffered saline) and observed at 4 or 25 °C in quartz cuvettes with a path length of 1 mm. Samples were diluted to 5 μm for protein denaturation studies. Guanidine-induced denaturation data at 4 or 25 °C were fitted by nonlinear least squares to a two-state model as described previously (53).

NMR Studies of IGF-I and His^B5-IGF-I

One- and two-dimensional ¹H NMR spectra of IGF-1, IGF-swap, and His^B5-IGF-I were obtained at 700 MHz in 10% deuterated acetic acid at 40 °C as described previously (21, 28, 54). These conditions have been employed previously in NMR studies of IGF-I and its analogues (55). Two-dimensional nuclear Overhauser enhancement spectroscopy (mixing time of 200 ms) and total correlation spectroscopy (mixing time of 55 ms) spectra were obtained in 10% deuterated acetic acid and 90% D₂O. Resonance assignment of the ¹H NMR spectrum of native IGF-I was independently obtained based on ¹⁵N-based heteronuclear NMR studies (56); results were consistent with prior studies (21, 54). Presumptive assignment of the spectrum of His^B5-IGF-I was obtained by analogy.

RESULTS

Our results are presented in two parts. We first describe the synthesis of Thr^B5-insulin and the effects of the substitution on structure, stability, activity, and cellular biosynthesis. Investigation of the reciprocal IGF-I analogue is then described in relation to native IGF-I and IGF-swap. His^B5-IGF-I exclusively exhibits native disulfide pairing.

Studies of an Insulin Analogue

Characterization of Thr^B5-Insulin

Insulin chain combination recapitulates the native folding of proinsulin (57); the combination of IGF-I-derived A- and B-domain peptides likewise yields disulfide isomers corresponding to native IGF-I and IGF-swap (58). Combination of the Thr^B5 insulin B-chain with the wild-type A-chain yields a single predominant product with an efficiency 4-fold lower than that of wild-type chain combination. Analysis of the crude chain combination mixture by rp-HPLC revealed several side products (cyclic A- and B-chains and B-chain dimer) but none with the mass of the insulin analogue. This purified product exhibited the predicted mass (5770.4 Da) and was found to be a single component following rp-HPLC using two independent elution co-solvents in either 0.1% trifluoroacetic acid/acetonitrile or 0.1% trifluoroacetic acid/methanol. These HPLC systems readily resolve insulin disulfide isomers (29).

The Thr^B5 substitution perturbs the structure of insulin as probed by CD spectroscopy (Fig. 3A) and causes a marked reduction in thermodynamic stability as probed by guanidine denaturation at 4 °C (Fig. 3B). Application of a two-state model provided estimates of respective free energies of unfolding (ΔG_u) of 4.4 ± 0.1 kcal/mol (wild-type insulin) and 2.4 ± 0.1 kcal/mol (Thr^B5-insulin). The extent of destabilization (ΔΔG_u 2.0 ± 0.2 kcal/mol) is more severe than ordinarily observed in studies of mutant insulins but similar to that previously observed in studies of an Ala^B5 analogue (18). Relative to Ala^B5, introduction of a longer side chain and potential hydrogen bond donor and acceptor (the β-OH group of Thr) into the B5-related inter-chain crevice does not augment stability. These analogues each exhibit attenuated helix-associated CD features, presumably representing transmitted perturbations in structure or dynamics.

The affinity of Thr^B5-insulin for IR (isoform B) is reduced by ∼5-fold relative to human insulin (Fig. 3C). Wild-type and variant hormone receptor dissociation constants in vitro are 0.06 ± 0.01 and 0.33 ± 0.04 nm, respectively (Table 2). The cross-binding of Thr^B5-insulin to IGF-1R is reduced by ∼6-fold relative to that of wild-type insulin (respective dissociation constants were 53.8 ± 5 and 8.7 ± 1.3 nm; Fig. 3D). Within statistical uncertainties, the extent of reduction in affinities to IR and IGF-1R is congruent. The activity of Thr^B5-insulin is at least 2-fold lower than that of an Ala^B5 analogue (18) but 4-fold higher than that of non-native disulfide isomers of wild-type sequence (18).

Effects on Proinsulin Biosynthesis

Insulin undergoes a complex pathway of biosynthesis in pancreatic β-cells (Fig. 4A). Folding and trafficking of variant proinsulins were evaluated in live HEK293T cells transfected with human preproinsulin cDNA bearing or lacking mutations. Following transfection, cells were pulse-labeled with ³⁵S-amino acids for 1 h and then chased for 1 h to examine expression, disulfide isomer formation, and secretion of newly synthesized proinsulin (Fig. 4, B and C). Denaturing PAGE (Tris/Tricine-urea-SDS-PAGE) in the absence of reducing agent permitted examination of distinct proinsulin disulfide isomers as formed in the ER. The wild-type construct (Fig. 4B, lanes 3 and 4) gave rise to robust expression, primarily of a fast-migrating band; previous studies have established that this is the native species (46, 59). This is efficiently secreted from transfected cells (“C” in Fig. 4) to medium (“M”). Two minor species are also present as slower migrating isomers with mispaired disulfide bonds; these exhibit lower secretion efficiency (Fig. 4B, brackets). Substitution of His^B5 by Met or Glu markedly impairs expression (Fig. 4B, lanes 5 and 7) and increases the fraction that forms mispaired disulfide isomers in the ER (Fig. 4B, lanes 5 and 7); moreover, secretion is blocked (lanes 6 and 8). Arg^B5-proinsulin exhibits efficiencies of folding and secretion similar to those of wild-type proinsulin (Fig. 4, lanes 11 and 12).

FIGURE 4.

Cellular folding and secretion of proinsulin and B5 variants. A, cellular pathway of insulin biosynthesis. Nascent proinsulin folds as a monomer in the ER (left) wherein the zinc ion concentration is low; in post-Golgi granules proinsulin is processed by cleavage of the connecting peptide to yield mature insulin, and zinc-stabilized hexamers begin to assemble. Zinc-insulin crystals are observed in secretory granules. On secretion into the portal circulation (right), hexamers dissociated to yield bioactive insulin monomers. B, pulse-chase studies in HEK293T cells as analyzed under nonreduced conditions (46). Cells were pulse-labeled with ³⁵S-labeled amino acids for 1 h and chased 1 h. Chase media (M, even-numbered lanes) and cell lysates (C, odd-numbered lanes) were immunoprecipitated with anti-insulin antiserum and analyzed by nonreducing Tris/Tricine-urea-SDS-PAGE. Whereas wild-type human proinsulin (lanes 3 and 4) and Arg^B5-proinsulin (lanes 11 and 12) exhibited efficient folding in the ER and secretion to the medium, Met^B5 and Glu^B5 variants exhibited impaired folding associated with non-native disulfide isomers (lanes 5 and 7) and a block to secretion (lanes 6 and 8). Thr^B5-proinsulin forms a larger molecular weight intracellular aggregate (lane 9; asterisk) also with absent secretion (lane 10). An empty vector control (pcDNA3) is provided in lanes 1 and 2. C, control studies of unstable protein variants in transfected HEK293T cells suggest that cellular foldability is not under thermodynamic control. Folding and secretion of a molten two-disulfide analogue (containing paired substitutions Cys^A6 → Ser and Cys^A11 → Ser; lanes 17 and 18) and partially folded analogue (Ile^A2 → Gly; lanes 19 and 20) are similar to wild type (lanes 15 and 16); an empty vector control is provided in lanes 13 and 14.

Strikingly, folding of Thr^B5-proinsulin leads to a product of higher molecular mass (Fig. 4, lane 9, asterisk) that is not secreted (lane 10). The marked shift in electrophoretic mobility might be the result of three possible molecular perturbations as follows: (i) aberrant aggregation of the mutant proinsulin leading to inter-molecular disulfide bridges; (ii) failure of cleavage of the signal peptide leading to expression of a mutant preproinsulin; or (iii) anomalous post-translational modification giving rise to increased molecular mass and decreased mobility. The latter mechanism was suggested by the fortuitous creation of a potential N-linked glycosylation site (Asn^B3–Gln^A4–Thr^B5) (60) following substitution of His^B5. To distinguish between these possibilities, SDS-PAGE analysis was performed under nonreducing or reducing conditions (Fig. 5A, left- and right-hand panels. respectively). On reduction, proinsulin and proinsulin analogues behave as random coil polypeptides, and so their electrophoretic mobilities on denaturing SDS-PAGE reflect molecular mass. Dithiothreitol reduction of wild-type insulin led to the collapse of discrete disulfide isomers (Fig. 5A, lane 2) to a single band of reduced mobility (lane 7). Dithiothreitol reduction of the Thr^B5-associated set of bands (Fig. 5A, lane 3) likewise led to a single band (lane 8) whose mobility is proportionately reduced relative to the reduced wild-type polypeptide (lane 7). The relative mobility of Thr^B5-proinsulin-associated polypeptide (Fig. 5A, lane 8) is slower than either that of wild-type proinsulin (lane 7) or that of Asp^S24-preproinsulin (lane 12); the latter is a clinical variant, recently discovered in a patient with neonatal-onset diabetes mellitus (12), that we and others predicted to be defective in signal-peptide processing. Failure to match the mobility of either proinsulin or preproinsulin indicates that anomalous mobility of biosynthetic Thr^B5-proinsulin is not due to either aberrant inter-molecular disulfide bridges or failure of signal-peptide cleavage.

FIGURE 5.

Aberrant glycosylation and trafficking of Thr^B5-proinsulin. A, transfected 293T cells were preincubated with (lanes 6 and 11) or without 10 μg/ml tunicamycin (TM) for 30 min. Cells were then pulse-labeled with ³⁵S-labeled amino acids in the same media for an additional 30 min. The newly synthesized proinsulin or proinsulin variants were immunoprecipitated with anti-insulin antiserum. Immunoprecipitates were treated with (lanes 4 and 9) or without PNGase F at 37 °C for 1 h before being analyzed under both reducing and nonreducing conditions. As a control for impaired cleavage of the signal peptide, a mutation (Asp^S24 in the preproprotein; lane 12) showed uncleaved mutant preproinsulin as a band of reduced electrophoretic mobility. B, transfected 293T cells were pulse-labeled with ³⁵S-labeled amino acids in the presence (lanes 8 and 11) or absence of 10 μg/ml TM for 30 min as A. The labeled cells were either immediately lysed (lanes 1, 2, 5, 8, and 10) or chased with (lanes 9, 10, 12, and 13) or without 10 μg/ml TM for 2 h. Cell lysates (C) and chase media (M) were analyzed under reducing conditions. EV, empty vector negative control; WT, wild type.

The expressed intracellular proteins were further characterized for possible N-linked glycosylation in the ER either by digestion of the extracted proteins with PNGase F (which removes N-linked carbohydrates) or by synthesis in cells treated with TM (which inhibits the addition of N-linked glycans (61)). Following digestion with PNGase F, the gel mobility of Thr^B5-proinsulin (Fig. 5A, lanes 4 and 9) reverted to that seen for wild-type proinsulin (lanes 2 and 7). This finding indicates that the anomalous mobility of Thr^B5-proinsulin is due to a novel N-linked glycosylation. Further support for this conclusion was obtained in cells treated with TM in which an additional set of Thr^B5-associated bands (Fig. 5A, lanes 6 and 11) was observed similar to those of wild-type proinsulin (lanes 2 and 7).

To address whether the impaired secretion of Thr^B5-proinsulin was strictly a result of aberrant glycosylation, we performed pulse-chase experiments in the absence and presence of TM. Intracellular (C) and secreted fractions of proinsulin (M) were analyzed by SDS-PAGE under reducing conditions (Fig. 5B). The key result was obtained on comparison of secreted proteins after a chase period of 2 h. Whereas wild-type proinsulin is efficiently secreted regardless of the presence of TM (Fig. 5B, lanes 4 and 10), the secretion of Thr^B5-proinsulin is blocked even when produced as an unglycosylated species (Fig. 5B, lane 13).

Because Thr^B5 markedly impairs the stability of insulin in vitro (see above), we further compared the intracellular expression of Thr^B5-proinsulin to that of two partially folded mutant proinsulins as follows: a two-disulfide analogue containing pairwise substitution of Cys^A6 and Cys^A11 by Ser (and so lacking cystine A6–A11; Fig. 4C, lanes 17 and 18), and an analogue containing core substitution Ile^A2 → Gly (lanes 19 and 20). Biophysical studies of these analogues have previously been described in the context of an engineered monomer (DKP-insulin) (62). The two-disulfide analogue is a molten globule of marginal stability lacking the A1–A8 α-helix; Gly^A2-insulin also exhibits segmental unfolding of this α-helix (63–66). Denaturation studies indicate that the stabilities of these analogues are reduced to an extent similar to or greater than that of Thr^B5-insulin,⁹ and yet the corresponding mutant proinsulins are efficiently folded and secreted by HEK293T cells. These results suggest that, even in the absence of aberrant glycosylation, Thr^B5-proinsulin exhibits an altered conformation that fails a quality control checkpoint in the biosynthetic pathway and so undergoes proteosomal degradation (67). Possible molecular mechanisms include increased exposure of nonpolar surfaces (ordinarily hidden in the B5-related crevice) and/or non-native aggregation mediated by such surfaces. These quality control criteria may not be reflected in the in vitro thermodynamic impairment of the monomer in guanidine solutions.

Studies of an IGF-I Analogue

Characterization of His^B5-IGF-I

The two competing ground states of wild-type IGF-I are shown in schematic form in Fig. 6A, panel a. Redox-coupled refolding was conveniently monitored by rp-HPLC. In accord with prior studies (24, 26), IGF-swap elutes just before native IGF-I, whereas intermediates containing one or two disulfide bridges elute after native IGF-I (Fig. 6A, panels b and c). The early eluting isomer was not observed on refolding of His^B5-IGF-I (Fig. 6B, chromatogram panels d–f). In addition, folding of the variant was slower than that of native IGF-I as probed by appearance of the native elution peak and disappearance of later eluting intermediates. The predominant product was purified and verified as a mono-component by analytical rp-HPLC (see supplemental Fig. S2). The molecular mass of this fraction indicated formation of three disulfide bridges. Physical evidence that the eluted His^B5-IGF-I fraction contains a single pairing scheme (i.e. is not a co-eluted mixture of isomers) and that this pairing is canonical (i.e. is like native IGF-I) was provided by NMR fingerprinting (below).

FIGURE 6.

Refolding of native and variant IGF-I. A, wild-type IGF-I folds to two competing ground states in a bifurcating energy landscape (panel a), yielding native IGF-I (N) and IGF-swap. MG indicates molten globule states. Redox-coupled folding was monitored by rp-HPLC at successive times: panel b, <1 min, and panel c, 2 h. B, refolding of His^B5-IGF-I yields a single product at higher yield but is slower than that of wild-type IGF-I: panel d, <1 min; panel e, 5 h, and panel f, 8 h. Arrow in panel d indicates accumulation of two-disulfide intermediate.

Although we cannot exclude that His^B5-IGF-I-swap was formed at a low level but with an anomalously delayed elution time, the instability or inaccessibility of the putative His^B5 isomer was demonstrated by studies of disulfide exchange. Starting with either purified wild-type IGF-I or IGF-swap, an equilibrium distribution of the two forms can readily be catalyzed within 60 min at room temperature by a sub-stoichiometric concentration of β-mercaptoethanol (26). This protocol by contrast leads to no change in rp-HPLC elution profile (using 0.1% trifluoroacetic acid/acetonitrile co-solvent) of folded His^B5-IGF-I, which elutes as a single peak (data not shown). After a 24-h incubation, the equilibrium elution of IGF-I was at 22.7 min, whereas IGF-swap eluted at 20.9 min. Under the same conditions, His^B5-IGF-I eluted as a single peak at 21.2 min. To address the possibility that kinetic barriers preclude disulfide exchange at room temperature, the protocol was repeated for 12 h at 37 °C with the same outcome. In addition, the single peak status of His^B5-IGF-I was maintained in the presence of β-mercaptoethanol as monitored by another co-solvent elution gradient (methanol/trifluoroacetic acid), making unlikely the fortuitous co-elution of disulfide isomers. In this co-solvent system the His^B5-IGF-I also eluted as a single peak at 28.5 min.

¹H NMR studies of His^B5-IGF-I were conducted under conditions previously well characterized in studies of wild-type IGF-I and IGF-swap (10% deuterated acetic acid and pH 2.0 at 40 °C). Under these conditions, the proteins are monomeric, exhibit well resolved ¹H NMR resonances, and retain insulin-like folds (28). The spectrum of His^B5-IGF-I closely resembles that of native IGF-I (Fig. 7, A and B) and is unrelated to the spectrum of IGF-swap (supplemental Fig. S3 and S4). Aliphatic spin systems (Fig. 7, C and D) and aromatic spin systems (Fig. 7, E and F) exhibit similar patterns of chemical shifts. Published assignments of the wild-type spectrum immediately suggested corresponding assignments in the spectrum of the variant.

FIGURE 7.

¹H NMR analysis of His^B5-IGF-I. A and B, aliphatic region of one-dimensional ¹H NMR spectra of wild-type IGF-I (A) and His^B5-IGF-I (B). Brackets indicate methyl resonances of Leu-14 (canonical position B15); arrow indicates δ-methyl resonance of Ile-43 (position A2). C and D, corresponding total correlation spectroscopy spectra of wild-type IGF-I (C) and His^B5-IGF-I (D). Selected spin systems are as indicated. E and F, aromatic total correlation spectroscopy spectra of wild-type IGF-I (E) and His^B5-IGF-I (F). Assignments are as indicated (IGF numbering scheme). Residues Phe-16, Phe-23, Tyr-25, Phe-26, Tyr-32, and Tyr-60 correspond to respective canonical positions B17, B24–B26, C2, and A19. Spectra were acquired in 10% deuterioacetic acid at 40 °C at 700 MHz.

Small shifts are observed in the chemical shifts of aliphatic side chains in the hydrophobic core (IGF-I residues L14 and I43 in Fig. 7D; canonical positions B15 and A2, respectively); these may be due to either adjustments in core structure or the introduced aromatic ring current of His^B5. Evidence for structural similarity between the core structures of wild-type IGF-I and His^B5-IGF-I was provided by comparison of nuclear Overhauser enhancement spectra (Fig. 8). These include contacts between α-helices 1 and 3 (Tyr-60/Leu-14, canonical positions A19 and B15), between α-helices 1 and 2 (Tyr-60/Ile-43, positions A19 and A2), and the α/β B-domain super-secondary structure (Phe-23/Leu-14, positions B24 and B15). These framework residues are conserved among both vertebrate insulin and IGF sequences (supplemental Table S1). No nuclear Overhauser enhancements are observed between the ring of His-4 (His^B5) and methyl resonances (Fig. 8B, dashed line) in accord with the structure of IGF-I.¹⁰ The crowded nature of this spectra region precludes definitive assignment.

FIGURE 8.

Two-dimensional NMR nuclear Overhauser enhancement spectra of native and variant IGF-I. A, wild-type IGF-I, and B, His^B5-IGF-I. Assignments of selected cross-peaks are as labeled. Phe-23 and Tyr-60 (vertical axis) correspond to canonical positions B24 and A19; Leu-14 and Ile-43 correspond to canonical positions B15 and A2. Dashed line in B indicates resonance position of His-4-C₄H (canonical position B5). Spectra were acquired in 10% deuterioacetic acid at 40 °C at 700 MHz.

The far-ultraviolet CD spectrum of His^B5-IGF-I (Fig. 9A, filled circles) is also similar to that of native IGF-I (solid line) and distinct from that of IGF-swap (dotted line). The latter exhibits attenuated α-helical features consistent with the absence of α-helix 2 (IGF residues 42–49; canonical positions A1–A8) in the solution structure of IGF-swap (28). CD-detected guanidine denaturation studies indicate that native IGF-I and His^B5-IGF-I exhibit similar stabilities at 25 °C (Fig. 9B, solid line and circles). Application of a two-state model implies respective free energies 2.9 ± 0.1 kcal/mol (wild type) and 3.1 ± 0.1 kcal/mol (variant). Stabilization by the B5 substitution is at the limit of significance (ΔΔG_u 0.2 ± 0.2 kcal/mol). The corresponding denaturation transition of IGF-swap is more sharply sigmoidal (Fig. 9B, dashed line), leading to a higher estimate of ΔG_u (3.7 ± 0.1 kcal/mol). Direct comparison to native IGF-I to estimate ΔΔG_u (which would differ in sign from that obtained by studies of disulfide exchange) may be confounded by the difference in respective denatured states in the two guanidine titrations.

FIGURE 9.

CD and receptor binding studies of His^B5-IGF-I. A, far-UV CD spectra of native IGF-I (solid line), IGF-swap (dashed line), and His^B5-IGF-I (●). B, CD-detected guanidine titrations of native IGF-I (solid line), IGF-swap (dashed line), and His^B5-IGF-I (●). Ellipticity was detected at 222 nm. C, IGF-1R receptor binding assay showing competitive displacement of ¹²⁵I-labeled IGF-I tracer as a function of unlabeled ligand concentration: ■, native IGF-I, and ●, His^B5-IGF-I. D, corresponding IR receptor binding assay (isoform B) showing competitive displacement of ¹²⁵I-labeled insulin tracer as a function of unlabeled ligand concentration: ▴, human insulin; ■, native IGF-I; and ●, His^B5-IGF-I.

Biological Activity of IGF-I Analogue

Functional studies of native IGF-I and His^B5-IGF-I demonstrated that the substitution augments IGF-IR binding (Fig. 9C) in accord with past studies (22). Respective ligand-receptor dissociation constants are 0.040 ± 0.005 nm (wild-type) and 0.022 ± 0.002 nm (variant). Studies of IR demonstrated that the B5 substitution also enhances cross-binding of the IGF-I analogue by 2-fold (Fig. 9D). Respective ligand-IR dissociation constants are 13.9 ± 2.2 nm (wild type) and 5.7 ± 1 nm (variant). Thus, as in reciprocal case in insulin, the B5 substitution does not alter receptor binding specificity.

The biological activity of His-4-IGF-I was evaluated in a mammalian cell line expressing IGF-1R (51). Activation of Akt kinase by Ser and Thr phosphorylation is a critical step in the phosphoinositol 3-kinase pathway downstream of the receptor. We therefore employed measurements of the relative extent of Akt phosphorylation in response to stimulation by wild-type IGF-I or His-4-IGF-I as a surrogate for assessment of IGF-1R signaling. Relative to the basal level of Akt phosphorylation (normalized to 1.00 ± 0.33 in arbitrary units), addition of 1 nm IGF-I triggered a 2.26(±0.27)-fold increase in phosphorylation, whereas addition of 1 nm His-4-IGF-I increased Akt phosphorylation by 3.45(±0.31)-fold. Similarly, addition of the ligands at a concentration of 10 nm led to respective fold increases of 7.69(±0.46) (wild-type) and 8.46(±0.38) (analogue). The increased affinity of His-4-IGF-I is thus associated with augmented post-receptor signaling.

Informational Content of B-chain Sequences

Di Marchi et al. (58) demonstrated that isolated A- and B-chain peptides derived from IGF-I undergo chain combination to yield native and isomeric two-chain products analogous to those observed on refolding of IGF-I. These observations first demonstrated that information is missing in the IGF A- or B-domains. Combination of the IGF-I-derived B-chain (29 residues) and insulin A-chain likewise yield two isomeric products as in single-chain chimeric models (68). To test whether residue B5 is sufficient in this model to restore the fidelity of chain combination, we synthesized the His^B5-IGF-I B-chain and attempted chain combination with the insulin A-chain. Yield was reduced 3-fold relative to wild-type insulin chain combination. Unlike reciprocal chain combination of Thr^B5-insulin, which exhibited low yield but no isomers (above), the hybrid reaction gave rise to a major product with three minor isomeric contaminants (each resolved on rp-HPLC with longer elution times than that of the major product and relative peak intensity <10%). These studies suggest that the differences in folding information between the isolated B-chains of IGF-I and insulin are not confined to residue B5 alone.

DISCUSSION

The insulin-related superfamily provides a model for the evolution and divergence of a structural motif (15, 69–71). The A- and B-domains encode an α-helical fold that is shared by vertebrate insulins, growth factors, and relaxin-related factors. These polypeptide sequences contain five classes of residues as follows: (a) conserved in the superfamily; (b) conserved among insulin, IGF-I, and IGF-II but not only relaxin-related factors; (c) conserved among IGF-I and IGF-II but not more broadly; (d) restricted to one protein type; and (e) not conserved. Invariant in the superfamily is a pattern of six cysteines (supplemental Table S1). Gly^A1–Ile^A2–Val^A3 provides a conserved element within vertebrate insulins and IGFs required to bind a cognate family of receptor tyrosine kinases (IR and IGF-1R); this element is not characteristic of relaxin-related factors, which bind to an unrelated class of receptors (G-protein-coupled receptors) (72). The evolutionary history of the insulin superfamily extends to invertebrate insulin-like proteins exhibiting further sequence variation, including the pattern of disulfide bridges (71, 73).

In this study, we have investigated reciprocal substitutions in human insulin and IGF-I at position B5. This residue is conserved as His in insulin and as Thr in the IGFs; relaxin-related factors contain a basic residue (Lys or Arg). We discuss our results in relation to general principles of protein folding and biological constraints governing the evolution of the insulin-IGF family. The importance of residue B5 is underscored by the recent discovery of a mutation at this site in human proinsulin causing permanent neonatal-onset diabetes mellitus (12).

Principles of Protein Folding

Protein folding represents a major unsolved problem. Anfinsen and co-workers (74) demonstrated that the native state of a globular protein is encoded by its sequence. Although this view has undergone recent modification (embracing, for example, the cellular roles of chaperones and of “folding enzymes” such as peptidylprolyl isomerase and protein-disulfide isomerase (75)), it is nevertheless believed that the protein structures are thermodynamically (rather than kinetically) determined as the ground state of a multidimensional energy landscape (76). Fundamental to protein biophysics, these ideas have broad biological implication as folding defects underlie diverse human diseases (77, 78). IGF-I with its two alternative ground states provides an example of a polypeptide sequence that is missing information.

Oxidative protein folding pertains to a major class of secreted and membrane proteins whose folding is coupled to disulfide pairing in the endoplasmic reticulum (79, 80). In vitro, the relative reactivities of thiol groups have long provided site-specific kinetic probes, enabling reaction intermediates to be trapped (81). As demonstrated by Creighton and co-workers (82, 83), the time course of formation and disappearance of free cysteines and specific pairing arrangements provides a chemical map of the oxidative folding pathway. An experimental paradigm has been provided by bovine pancreatic trypsin inhibitor (BPTI) (84–86). The folding of BPTI is remarkable for intermediates that are well populated at neutral pH, exhibit native structures, and face large kinetic barriers to reach the ground state (84–86). The preferred final step is formation of an external disulfide bridge between solvent-exposed loops.

Like BPTI, IGF-I and proinsulin are small globular proteins containing three disulfide bridges. Profound differences from BPTI are observed. IGF-I and proinsulin are significantly less stable than BPTI (87, 88); kinetic barriers among populated two-disulfide species are also lower (63, 88). Furthermore, the conformational search is intertwined with formation of disulfide bridges: analogues of IGF-I and insulin exhibit stepwise stabilization of native structural elements with successive disulfide pairing (24, 29, 89). Oxidative folding of IGF-I and proinsulin thus occurs on successive energy landscapes whose topography is constrained by disulfide pairing (13, 88). Evidence for nonrandom initial folding trajectories has been provided in studies of IGF-I, mini-IGF-I, and mini-proinsulin by the trapping of a unique one-disulfide intermediate containing cystine A20–B19 (55, 90). Although several two-disulfide species have been characterized, a slow subsequent step is formation of the solvent-exposed A7–B7 bridge. Diabetes-associated mutations in proinsulin cluster in residues neighboring either cystine A20–B19 or A7–B7 (12, 14).

IGF-I refolding bifurcates to yield two products (24, 26). Designated native IGF-I and IGF-swap (26), the two products share the A20–B19 disulfide bridge. The isomers exhibit near-equal thermodynamic stabilities. Proinsulin and single-chain insulin analogues by contrast refold to form a unique ground state; alternative pairing schemes are metastable and, if present, rearrange to the native state. That the number of products is precisely two demonstrates that a nonrandom folding pathway is encoded but must be “ambiguous” following formation of cystine A20–B19. In studies of mini-protein models Feng and co-workers (37) obtained evidence that the respective B-domains of proinsulin and IGF-I, in particular His^B5 or Thr^B5, are responsible for determining the relative stability of the swapped isomer. The present studies of His^B5-IGF-I supports this conclusion. The B5 substitution in IGF-I selectively destabilizes the non-native isomer. The accumulation of a two-disulfide intermediate in the refolding of His^B5-IGF-I presumably represents a non-native species containing cystines A20–B19 and A7–A11 (IGF-I residues 18–61 and 48–52), ordinarily the precursor of IGF-swap but here an off-pathway intermediate. We speculate that the delayed overall time course of folding reflects the formation and rearrangement of this species.

The yield of insulin chain combination is impaired by the reciprocal substitution His^B5 → Thr. Such impairment stands in contrast to the general robustness of chain combination to substitutions, especially in the C-terminal B-chain segment and N-terminal A-chain segment (65). Sites neighboring Cys^B7 define an Achilles' heel; impaired chain combination has also been observed following substitution of Leu^B6, Ile^B11, and Val^B12 (41, 91), which like B5 are sites of neonatal diabetes-associated mutations (12, 14). It is possible that substitutions at these sites, like Thr^B5, markedly perturb the stability of insulin. Synthetic and cell biological studies nonetheless suggest that thermodynamic criteria are insufficient to account for differences in foldability. Pairwise substitution of Cys^A7 and Cys^B7 by Ser was observed to rescue chain combination otherwise blocked by substitution Gly^B8 → Ala (92). Because in the absence of cystine, A7–B7 insulin exhibits a partial fold of lower stability than that of Ala^B8-insulin (18), such rescue suggests that Ala^B8 imposes a kinetic block to pairing of Cys^A7 and Cys^B7 (92). Pairwise substitution of Cys^A6 and Cys^A11 by Ala (or Ser) likewise leads to a partial fold of marginal stability. Not only is chain combination preserved, but also in transfected mammalian cells a corresponding variant proinsulin exhibits efficient folding and secretion (63).

How might His^B5 contribute to chain combination and the biosynthesis of proinsulin? A clue is provided by the presence of a basic side chain (Arg or Lys) at this site in relaxin-related factors and the occasional occurrence of Arg among nonmammalian insulins (93). Whereas chain combination fails in the case of Met^B5, near wild-type efficiencies are obtained for Arg^B5 (93). We imagine two possible molecular roles for His^B5 or Arg^B5 in the mechanism of chain combination. First, because crystal structures of insulin and Arg^B5-insulin exhibit analogous hydrogen bonds between a B5 side-chain NH donor (NδH in His^B5 and NϵH in Arg^B5) to A-chain carbonyl oxygens, native-like interactions might stabilize an on-pathway folding intermediate whose local structure facilitates pairing of Cys^A7 and Cys^B7. Second, the positive charge of Arg^B5 and potential positive charge of His^B5 may stabilize a thiolate anion at A7 or B7 in the course of disulfide bonding. The latter mechanism would be consistent with the pK_a of His^B5, which is shifted from 6.0 to 7.0 in native insulin (93, 94). The neonatal diabetes-associated mutation at B5 is Asp (12), whose negative charge would be expected to destabilize a thiolate reaction intermediate.

Our studies of the cellular folding of proinsulin analogues on transfection in human fibroblast-derived HEK293T cells demonstrate that His^B5 or Arg^B5 support folding and secretion to an equivalent extent. Met^B5 and Glu^B5 (like Asp^B5, Gln^B5, and Phe^B5 in a previous study (18)) impair folding in the ER, leading to increased formation of disulfide isomers and a block to secretion. Expression of Thr^B5-proinsulin leads to fortuitous modification by an N-linked glycan whose secretion is blocked. Experiments employing TM to impair N-linked glycosylation nonetheless demonstrated that secretion of Thr^B5-proinsulin, folded in the absence of such modification, is itself markedly impaired. Despite this block in the secretory pathway, disulfide pairing of Thr^B5-proinsulin (in the presence of absence of glycosylation) is more efficient than that of Ala^B5-, Met^B5-, Glu^B5-, or Asp^B5-proinsulin (present study and see Ref. 18) as indicated by native SDS-PAGE in the absence of reduction (i.e. the ratio of less compact non-native disulfide isomers to the most rapidly migrating band).

We speculate that the β-OH group of Thr^B5 (unlike the above folding-defective side chains) can in principle donate a hydrogen bond to a putative thiolate intermediate and, once pairing is complete, to the A-domain within the B5-associated crevice but that these interactions are weaker (or more transient) than those of secretion-competent side chains His^B5 and Arg^B5. This model is supported by the decreased thermodynamic stability of Thr^B5-insulin. We imagine that conformational fluctuations in Thr^B5-proinsulin expose nonpolar surfaces, leading to recognition of a folding defect by quality control checkpoints in the ER and Golgi apparatus. The blocked secretion of Thr^B5-proinsulin highlights in the breach the puzzling conservation of Thr at this position in IGF-I and IGF-II.

Co-evolution of a Regulatory Network

Why is a secretion-defective residue at B5 invariant among IGF-I and IGF-II? Such a conversation poses a seeming paradox, especially in light of the genetic susceptibility of proinsulin to misfolding-associated mutations. Specification of native IGF-I disulfide pairing in vivo is attributed to specific IGFBPs, present in equimolar proportions. Such partner proteins bind native IGF-I but not IGF-swap, thus favoring formation of the native isomer by Le Chatelier's Principle (95). We propose that the availability of IGFBPs as ubiquitous heterodimeric folding partners (33) has enabled IGF-I to diverge from insulin at B5. The extent of divergence is not so marked as to preclude disulfide pairing as in vitro reduced IGF-I successfully refolds to form only 2 of the 15 possible disulfide isomers (native IGF-I and IGF-swap). Discrimination against the 13 other possible combinatoric pairing schemes demonstrates that almost all folding information (but not quite all) is retained in the polypeptide sequence of IGF-I. This subtle defect is repaired on substitution of Thr^B5 by His. In this context it is noteworthy that in cellular assays of proinsulin folding, we found that the defect in disulfide pairing associated with Thr^B5 is less profound (despite the subsequent block to secretion) than that of clinical mutation of Asp^B5 or the other substitutions tested in cell culture. We speculate that IGFBPs enable IGF-I to overcome the block to secretion observed in Thr^B5-proinsulin; formation of a heterodimer enables quality control checkpoints to be satisfied.

Conservation of Thr^B5 in IGF-I and IGF-II has presumably been enjoined by a selective advantage. His^B5-IGF-I exhibits a 2-fold increase in binding to both IGF-1R and IR. Thus, threonine subtly attenuates IGF signaling at the expense of uniqueness of folding of the isolated polypeptide. We therefore imagine that IGFBPs (in addition to sequestering circulating IGF-I and their direct signaling activities) have implicitly expanded the space of foldable sequences available to IGFs. Such co-evolution may in general enable the fine-tuning of receptor affinities of hormones and growth factors under physiological selection to coordinate the regulation of development and metabolism. We envisage that variation at B5 in IGFs enabled fine-tuning of receptor affinities without perturbations in receptor selectivity or binding to IGFBPs. Analogous variation in proinsulin was disallowed by impaired secretion and, in the case of a broader set of substitutions at B5, by proteotoxicity due misfolding of the nascent polypeptide in the β-cell. Whereas mutation of His^B5 by Asp in the human insulin gene leads to neonatal diabetes mellitus because of β-cell death (12), we speculate that an analogous Thr^B5 substitution might lead to decreased insulin secretion (from the remaining wild-type insulin allele) but would otherwise be well tolerated.

Why could fine-tuning of IGF-I activity, even by a factor of 2, be of biological importance? Factors of 2 can seem negligible in biochemical studies, and yet in vivo such variation may have profound consequences, especially in regulatory control systems. An instructive example in pharmacology was provided by a once-promising insulin analogue under development 20 years ago (96). Just as this study has investigated reciprocal insulin and IGF-I substitutions at position B5, a candidate rapid-acting insulin formulation (97) was proposed in which His^B10 was substituted by Asp^B10 (98), whose side-chain carboxylate recapitulates a feature of IGF-I (Glu^B10; position 9 in the polypeptide sequence of IGF-I). The Asp^B10 substitution was observed to enhance the stability of insulin (99) and accelerate its absorption after subcutaneous injection (97), both desirable pharmacological properties. This substitution also enhanced binding to the IR and IGF-1R by 2-fold due to increased residence times (98, 100). Although seemingly minor or even advantageous, this 2-fold alteration in the receptor-binding properties of Asp^B10-insulin was associated with an increased incidence of mammary tumors in Sprague-Dawley rats and enhanced mitogenicity in human breast cancer cell lines (96). Human trials were therefore discontinued. By analogy, we suggest that the intrinsic activity of free IGF-I may likewise be under strict biological selection. Fine-tuning of IGF-I by co-evolution of its sequence and IGFBPs is likely to affect both growth rates and base-line risk of tumorigenesis in an organism.

The fine-tuning of IGF-I may extend to the extent of its instability to disulfide exchange. In a pioneering study, Nillson and co-workers (27) demonstrated that under the redox conditions of the Golgi apparatus, secretory granules, and bloodstream, isolated IGF-I undergoes in vitro disulfide rearrangement to form an equilibrium distribution of native and swapped species. Because, to our knowledge, the biological implications of these findings have not been explored, it would be of future interest to extend this study to molecular analysis of the state of IGF-I during cellular biosynthesis. Co-expression or RNA interference knockdown of IGFBPs might be found to modulate the fidelity of disulfide pairing. Nillson and co-workers (27) further speculated that isomerization of free IGF-I in the bloodstream and tissues might provide a mechanism to down-regulate its growth-promoting activity (27). The proposed mechanism thus exploits the instability of free IGF-I and the low activity of IGF-swap as a biological defense against excessive mitogenic signaling. To our knowledge, this intriguing hypothesis has not been tested.

Extension to Invertebrate Insulin-like Sequences

The insulin superfamily includes invertebrate insulin-like polypeptides. Although these sequences are more divergent than those of the vertebrate insulin-IGF family, selected NMR structures exhibit insulin-like folds as exemplified by bombyxin (extracted from the silkworm Bombyx mori) (70) and INS-6 (found in the nematode Caenorhabditis elegans) (71). These proteins contain Thr^B5 and Arg^B5, respectively. Of the 36 insulin-like genes identified in C. elegans by Ruvkun and co-workers (73), 26 encode a basic residue at B5 and 2 encode Thr. The genome of Drosophila melanogaster encodes 7 insulin-like polypeptides of which 4 contain a basic residue at B5; the remaining sequences contain Thr, Ser, or Met. The Thr^B5-containing protein (designated DILP-2) exhibits respective sequence identities of 35 and 26% with respect to human insulin and IGF-I. Conserved residues include Gly^A1–Ile^A2–Val^A3, part of a key recognition α-helix (101) shared by insulin and IGF-I but not critical for folding (65). Interestingly, DILP-2 has been demonstrated to bind to a partner protein in D. melanogaster (102, 103). This binding protein, designated IMP-2, is a member of the immunoglobulin superfamily homologous to human IGFBP7. We speculate that, like vertebrate IGFs, folding of DILP-2 may be coupled to co-expression of this or another binding protein. Should the folding properties of human insulin and IGFs generalize to invertebrate insulin-like polypeptides, we would speculate that, on the one hand, autonomous folding requires Arg, Lys, or His at B5 and that, on the other hand, Thr^B5 (or other divergent residue at B5) might indicate the presence of a corresponding binding protein.

Concluding Remarks

A major challenge is posed by the problem of how members of the insulin superfamily fold and pass quality control checkpoints en route to secretion. This problem has both biophysical and cell biological dimensions. Its subtlety is highlighted by the implicit role of His^B5 in a variant IGF-I to destabilize a competing but unseen fold. Despite decades of investigation, rules governing cellular foldability remain elusive. Indeed, Arvan and co-workers (46) have found that substitutions in proinsulin that are well tolerated in vitro can be associated with disulfide mispairing in cell culture. Furthermore, mini-proinsulin analogues, although highly efficient in in vitro refolding assays (104), can quantitatively misfold in mammalian cell culture to yield a metastable disulfide isomer (presumably analogous to IGF-swap). These observations strongly suggest that cellular folding of proinsulin is under kinetic control, whereas thermodynamic control of IGF folding is imposed, at least at the final steps, by IGFBPs.

This study has demonstrated the value of reciprocal substitutions in insulin and IGF-I as probes of structure and function. These homologous systems provide a model in which kinetic and thermodynamic determinants of protein folding may be dissected. In addition to their biophysical interest, future studies may define a relationship between misfolding and ER stress in the pathogenesis of β-cell exhaustion in type II diabetes mellitus.