Implications for the active form of human insulin based on the structural convergence of highly active hormone analogues

Abstract

Insulin is a key protein hormone that regulates blood glucose levels and, thus, has widespread impact on lipid and protein metabolism. Insulin action is manifested through binding of its monomeric form to the Insulin Receptor (IR). At present, however, our knowledge about the structural behavior of insulin is based upon inactive, multimeric, and storage-like states. The active monomeric structure, when in complex with the receptor, must be different as the residues crucial for the interactions are buried within the multimeric forms. Although the exact nature of the insulin’s induced-fit is unknown, there is strong evidence that the C-terminal part of the B-chain is a dynamic element in insulin activation and receptor binding. Here, we present the design and analysis of highly active (200–500%) insulin analogues that are truncated at residue 26 of the B-chain (B²⁶). They show a structural convergence in the form of a new β-turn at B²⁴-B²⁶. We propose that the key element in insulin’s transition, from an inactive to an active state, may be the formation of the β-turn at B²⁴-B²⁶ associated with a trans to cis isomerisation at the B²⁵-B²⁶ peptide bond. Here, this turn is achieved with N-methylated L-amino acids adjacent to the trans to cis switch at the B²⁵-B²⁶ peptide bond or by the insertion of certain D-amino acids at B²⁶. The resultant conformational changes unmask previously buried amino acids that are implicated in IR binding and provide structural details for new approaches in rational design of ligands effective in combating diabetes.

The peptide hormone insulin regulates blood glucose levels with a widespread impact on lipid and protein metabolism. It is a molecule of major therapeutic importance in the treatment of diabetes. The mature form of insulin is formed by two chains “A” and “B” with a B chain running from Phe^B1-Thr^B30 and an A chain Gly^A1-Asn^A21, stabilized by two inter and one intra chain disulphide bonds. Insulin’s metabolic actions are expressed through binding as a monomer to the insulin receptor (IR). The structure of insulin, which has been known for four decades (1), has not provided insight into the mode of receptor binding and hormone activation. This is because detailed three-dimensional knowledge of insulin’s complex structural behavior is limited to its inactive storage (hexameric, dimeric) states (2–4). The NMR structures of the monomeric form of insulin facilitated by mutations (5), applications of organic co-solvents (6) or truncation of the B-chain (7) merely confirm the conformations known from the inactive forms but also indicate intrinsic mobility of the N- and C termini of the B-chain. It has also been found that the N terminus of insulin can exist in so called T (extended) or R (helical) conformations, however, their role for insulin activation is still ambiguous (3, 4).

It is widely acknowledged that insulin must therefore undergo induced-fit structural changes upon binding to the IR because residues crucial for receptor interactions are hidden in the known structures of the native hormone (8–11). Although the exact nature of the changes that expose these residues to the receptor surface is unknown, there is strong evidence that the C-terminal part of the B-chain is a dynamic element in the transition that leads to IR binding (10, 12, 13). The structure of the insulin-IR complex is not known but extensive insulin sequence and mutagenesis studies have identified the likely functional sites responsible for key interactions with the receptor (4, 14–18). The main IR-binding site one consists of Leu^B11, Val^B12, Leu^B15, Gly^B23, Phe^B24, Phe^B25, and Tyr^B26 on chain B and Gly^A1, Ile^A2, Val^A3, Gln^A5, Tyr^A19, and Asn^A21 on chain A (19, 20). The shielding of the A-chain residues in site one by the ∼B²⁵-B³⁰ residues of B-chain C terminus in native insulin structures led to a widely accepted induced-fit hypothesis for insulin-IR binding, requiring this part of the B chain to be displaced (3, 4, 10).

As the lack of structural detail of the active form of insulin bound to the receptor restrains the progress in combating diabetes, we have tried to characterise the structural signatures of the active conformation of this hormone employing partial chemical synthesis/modification of highly-active (i.e., 200–500%) insulins, IR-binding studies and x-ray crystallography. Two types of “semisynthetic” insulins were prepared: one in which the nitrogen atom in the Phe^B25-Xaa^B26 peptide bond was methylated (NMe^B26), and others in which the chirality at the B²⁶ Cα was changed to the D-enantiomer. The nine insulin crystal structures reported here provide persuasive structural evidence concerning the nature of the induced-fit changes upon receptor binding.

Results

Highly active B²⁶-Shortened/N-Methylated Insulins.

Chemical changes were systematically introduced in the insulin molecule to achieve high-activity. Semisynthesis (21) of several unique and some previously reported (22, 23) insulin analogues with an N-methylated (NMe) Phe^B25-Xaa^B26 peptide bond or with a D-enantiomer at position B²⁶ yielded insulins with various binding activities (Table 1, Figs. S1–S3). These modifications indicate a clear pattern that correlates with high binding activity of the hormone: (i) shortening of its B-chain by deletion of residues B27-B30, (ii) the use of a carboxyamide (CONH₂) C terminus at position B²⁶, and (iii) a concurrent incorporation of NMe^B26 (22), D-amino acid (23) or D/L-Pro at position B²⁶. In contrast, an increase of the bulkiness of the B²⁶ amino acid in shortened analogues ([NMePhe^B26]-DTI-NH₂ (22), [NMeTyr^B26]-DTI-NH₂) or similar modifications in the full-length analogues [[NMeTyr^B26]-insulin and [NMeAla^B26]-insulin (22)] resulted in low or moderate (21–72%) binding activities (Table 1).

Table 1.

Comparison of relative receptor binding affinity of some insulin analogues

B-chain position:	24	25	26	27	28	29	30	Affinity, %
Human insulin:	F	F	Y	T	P	K	T	100
[NMeAlaB26]-DTI-NH2†	F	F	Me A - CONH2	—	—	—	—	465 (22)
[D - ProB26]-DTI-NH2†	F	F	D - P - CONH2	—	—	—	—	359‡
[NMeHisB26]-DTI-NH2†	F	F	Me H - CONH2	—	—	—	—	214 (22)
[NMeAlaB26]-insulin†	F	F	Me A	T	P	K	T	21 (22)
[NMeTyrB26]-insulin†§	F	F	Me Y	T	P	K	T	21‡
[ProB26]-DTI†§	F	F	P - COOH	—	—	—	—	81‡
[ProB26]-DTI-NH2	F	F	P - CONH2	—	—	—	—	359‡
[NMePheB26]-DTI-NH2	F	F	Me F - CONH2	—	—	—	—	36 (22)
[NMeTyrB26]-DTI-NH2	F	F	Me Y - CONH2	—	—	—	—	72‡
[D - AlaB26]-DTI-NH2	F	F	D - A - CONH2	—	—	—	—	∼400(1252)¶

Relative receptor binding affinity is defined as (IC₅₀ of human insulin/IC₅₀ of analogue) × 100.

^†Crystal structures presented in this report,

^‡data presented in this study (the values of IC₅₀s are given in legends for Figs. S1–S3). Changes in B-chain sequence, high activities, and nonstandard chemical groups in bold;

^§analogue for which both monomer and hexamer crystal structures have been determined, DTI: des-tetrapeptide; [NMeXaaBⁿ]: methylation of the peptide bond N-atom preceding Xaa Bⁿ amino acid; -NH₂: C-terminal carboxyamide.

^¶Only one analogue: [D-Ala^B26]-DTI-NH₂, 1,252% from ref. 23, was reevaluated here as it showed IR affinity within the range of other analogues semisynthesised and presented in this work.

B-Chain of the Highly Active Insulins Possess a Unique β-Turn.

The apparent correlation between chemical composition and binding activity within the present insulin analogues prompted their crystallographic characterization (Tables S1–S4) in order to identify the underpinning structural features and to get further insight into the structural origins of the active conformation of insulin.

The most active (465%) insulin presented in this study: N^B26-methylated and B²⁶-truncated analogue [NMeAla^B26]-DTI-NH₂ (22), was crystallized as a monomer with a unique conformation at B²⁴-B²⁶. Its most striking feature is a type II β-turn at Phe^B24-NMeAla^B26 referred to here as the B26 turn (Fig. 1, Fig. S4A). This turn is associated with a cis conformation of the Phe^B25-NMeAla^B26 peptide bond. The B²⁴ CO group forms a 2.8 Å hydrogen bond with NH₂ of the terminal Ala^B26 carboxyamide, in which the NH₂ group mimics the peptide NH of the i + 3 (B²⁷) amino acid.

Fig. 1.

Structural features of the highly-active insulin analogues. (A) An overlay of the general fold of crystal structures of the highly-active and other insulins reported here: Magenta: [NMeAla^B26]-DTI-NH₂, Coral: [NMeHis^B26]-DTI-NH₂, green: [D-Pro^B26]-DTI-NH₂, Blue: [NMeAla^B26]-insulin, Yellow: [NMeTyr^B26]-insulin (monomer structure), Red: wild type human insulin (32); the location of the B26 turn is marked by the Star in Magenta. (B) A close-up view of the B26 turn in the most representative analogues. N, O atoms in Blue and Red, resp. [except for normal human insulin (32) that is in White], C atoms color code: Green — [NMeAla^B26]-DTI-NH₂, Yellow — [NMeAla^B26]-insulin, Pink — [NMeTyr^B26]-insulin, W — water molecule, hydrogen bonds (Dashed Lines) lengths in Å. *Me (Red) — NMe^B26, Blue * — indicates the three-dimensional convergence of N atoms from the B²⁶CONH₂ and B²⁶NH peptide groups. (C) Superposition (on B²³-B²⁵ Cα) of B26 turns in [NMeAla^B26]-DTI-NH₂ and [D-Pro^B26]-DTI-NH₂ showing a conformational convergence of peptide methylation (NMe^B26) and D-Pro^B26 chirality.

The most important consequence of the B26 turn is the departure of the B-chain B²²-B³⁰ β-strand from its typical hexamer/dimer like conformation (Fig. 1). The truncated B-chain C-terminal segment now points away from IR site one making the latter accessible for a direct interaction with the IR. Another structural consequence of the B26 turn is a rotation of the Phe^B25 side chain that now overhangs the A-chain part of site one (∼7.7 Å from nearest Ile^A2) (Fig. 2, 4). In contrast, Phe^B24 remains in place. These B-chain changes are complemented by large movements of the A-chain, most notably the N-terminal helix is displaced by ∼3 Å and rotated by ∼40°, whereas Tyr^A19 shifts 2.9 Å away to form a hydrogen bond with Gln^A5 extending the non-polar receptor binding surface (Fig. 2). These “high-activity motifs” of the insulin structure were confirmed by the crystal structure of an another highly-active (214%) truncated analogue ([NMeHis^B26]-DTI-NH₂) (22, Fig. S5). These both highly-active insulins possess B26 turns with identical geometries (e.g., the corresponding Cβ atoms of His^B26 and Ala^B26 are only ca. 0.3 Å from each other) and hydrogen bond networks.

Fig. 2.

Main structural changes occurring in insulin during a transition from ground-state to a proposed activated/active form based on the structures of some highly-active analogues; Yellow — human insulin (32); Green — [NMeTyr^B26]-insulin (isomorphous with the highly-active truncated [NMeAla^B26]-DTI-NH₂ but with the almost full B-chain structural definition), Magenta Star — Phe^B24 side-chain structural pivot, Red Star — Phe^B25 main chain structural pivot, Black Star — Ile^A2 Cα main chain pivot, Green Star — Tyr^A19 shift and formation of a hydrogen bond (Dashed Lines) with Gln^A5, Blue Star — rotation of the A1 helix (in degrees), Cyan Star — translational move of the A1 helix (in Å).

Fig. 4.

Molecular shape of a putative active form of insulin based on the electrostatic surface of the [NMeTyr^B26]-insulin analogue as the representative of B26 turn-containing insulins. (A) A “side” view showing the formation of the hydrophobic IR-binding hand-like insulin shape composed of site one residues (Red Numbering), and lined by Leu^B11, Leu^B15; labels for some residues of site two are in Magenta. (B) As (A) but with B²⁵-B³⁰ chain of the wild type human insulin modeled in (32) (Yellow Surface, Blue-Italic Numbering) showing the obstruction of site one prior to the formation of the B26 turn. (C-D) “Top” views of insulin surfaces equivalent to the (A and B) representations after 90° rotation.

D-Pro^B26 Mimics β-turn Resulting in High Activity.

To explain the structural role of D-amino acid chirality at position B²⁶ in achieving highly active (359%) insulin analogue (Table 1), the crystal structure of the truncated [D-Pro^B26]-DTI-NH₂ analogue was determined. The change of the chirality at the B²⁶ Cα center in this analogue caused by the introduction of D-Pro generates a local structure similar to that seen in the cis Phe^B25-Xaa^B26 N-methylated peptide bond high-activity insulin analogues. The absence of the cis-peptide bond in [D-Pro^B26]-DTI-NH₂ is compensated by the D-Pro in attaining a conformation similar to the B26 turn (Fig. 1C, Fig. S5). Although the i + 3 β-turn hydrogen bond is not formed here (COⁱ-NHⁱ⁺³ distance of ∼4.0 Å), the D-Pro residue positions the B-chain C-terminal CONH₂ moiety in a manner similar to that, observed in insulins containing a B26 turn; the NH₂-terminal group is only ∼1.1 Å from its hydrogen-bonded equivalents in two other highly-active analogues. Hence, the conformations of the B-chain C-termini are associated with a β-turn in all three highly-active insulins and are correlated with their similar high affinity receptor binding.

As a very similar high activity (∼360%) was also observed with the truncated L-Pro containing analogue with the carboxyamide group at the B-chain C terminus ([L-Pro^B26]-DTI-NH₂, Table 1), the shortened L-Pro analogues were also investigated by x-ray crystallography. However, only the monomeric and hexameric structures of C terminus-amide lacking truncated L-Pro homologue ([L-Pro^B26]-DTI, 81% affinity) could be crystallized and determined. They show that the B²⁰-B²⁶ segments are severely disordered in both crystals forms.

Full-length Insulin Analogues Can Adopt B26 Turn.

The capacity of full-length insulin to adopt the B26 turn was investigated further in the crystal structures of monomers of the full-length analogues: [NMeTyr^B26]-insulin and [NMeAla^B26]-insulin (22). These variants also possess the B26 turn and all other general structural features characteristic for the most active truncated insulin molecules including the exposure of site one (Fig. 1A and B, Fig. 2, Fig. S4B). The B26 β-turn is stabilized in these analogues by a typical i + 3 β-turn hydrogen bond between Phe^B24 and Thr^B27 (CO^B24-NH^B27 distances of ∼2.9–3.3 Å). As expected, the Thr^B27 peptide bond NH groups of the full-length analogues are spatially equivalent to CONH₂ terminus NH₂ group of the truncated insulins. The Cαⁱ-Cαⁱ⁺³ distances of ∼6 Å are well within the ∼7 Å limit typical for a type II β-turn. The section of the C-terminal segment that is well defined, Thr^B27-Pro^B28 (Lys^B29-Thr^B30 is disordered) follows the direction enforced by the B26 turn (identical to its conformation found in truncated insulins), and departs further from site one (Fig. 1A and B, Fig. 2) that is, thus, fully revealed. The B26 β-turn is further stabilized here by additional hydrogen bonds between Phe^B24 CO with the side-chain OH of Thr^B27 and water molecules. As the crystallization experiments were carried out in parallel in the monomer- and hexamer-like conditions (pH 3.0 and pH 7.5–8.2 respectively) (Fig. S6), the hexamer crystals of the full-length [NMeTyr^B26]-insulin have also been obtained. Despite the presence of NMe group at position B²⁶, this insulin analogue is still capable of forming a typical hexamer, in which the dimer interface adapts to the NMe-induced disruption as described before (24). Interestingly, crystallizations of the full-length [NMeAla^B26]-insulin, even under typical hexamer conditions, always yield the monomer crystal form. It has to be noted that some crystal structures of analogues described here are unique crystallographic descriptions of the full-length monomeric form of insulin.

New Pseudo β-Turn Conformation of B-Chain N Terminus.

All monomers of the highly-active insulin analogues reported here also show a structural convergence of their N termini that adopts a unique conformation not previously seen. Here, the main chain of B-chain N terminus follows the direction typical for the T-state till the Cα atom of the Phe^B5 at which it deviates from the T-fold forming, instead, a pseudo type II β-turn (Fig. 3). The turn is stabilized by the i + 3 hydrogen bond (3.0 Å) between main chain NH of the His^B5 (i + 3) and the side chain of the Asn^B3 that mimics the i- main chain CO group. This turn is stabilized further by an additional side chain Asn^B3- side chain (Nδ2) His^B5 hydrogen bond (3.0 Å). The Nδ1 atom of the His^B5 is also close to the main chain CO of the Thr^A8 (∼3.3 Å) reflecting the importance of B⁵ side chain in this unique formation. Only the main chain hydrogen bond: NH B⁶-CO A⁶ (2.8 Å) remains here from the network of B-chain—A-chain interactions that contribute to the stabilization of a typical T-state B chain N terminus.

Fig. 3.

Unique conformation of B-chain N terminus in the highly-active insulin analogues. One representative pseudo β-turn from the [NMeAla^B26]-DTI-NH₂ analogue is given in Green (only for the B²-B⁵ residues all atoms are given; Cαs trace for the B⁶-B¹⁹ region, hydrogen bonds in Dash Lines, distances in Å). Cα trace of the B¹-B¹⁹ region typical for the T-state form of insulin (1 mso.pdb) in Yellow; the same region representative for the R-state of the hormone — derived from the hexamer form of the [L-Pro^B26]-DTI) — in Magenta.

Discussion

Structural Signatures of the Activated/Active Form of Insulin.

The conservation and three-dimensional convergence of specific structural features that lead to the exposure of buried, biologically active residues in chemically modified, highly-active insulin analogues discussed here suggest key structural traits that may be present in insulin’s active conformation on the receptor or during its activation process. Although the evidence for the conformational motifs of the activated/active insulin conformation is indirect, the convergence of the structural signatures of highly-active hormone analogues observed here is to our mind quite persuasive.

It can be envisaged that the most striking element of the activated/active form of insulin is the formation of B26 turn associated with trans-to-cis isomerisation of the Phe^B25-Tyr^B26 peptide bond. The Phe^B24(CO)-Thr^B27(NH) hydrogen bond—crucial for the stabilization of B26 turn—is probably also critical to the productive receptor binding in wild-type insulin because the inhibition of its formation by selective reduction of the B²⁴ CO group to ψ(CH₂-NH) in the native hormone molecule abolishes its activity (25). The B26 turn conformation at B²⁶ is found in all three truncated, highly-active insulins described here. Although the structure of the highly-active [L-Pro^B26]-DTI-NH₂ insulin is not known the loss of potency in carboxyamide-lacking [L-Pro^B26]-DTI correlates well with a loss of B26 turn. Furthermore, the NMR structures of another very potent [D-Ala^B26]-DTI-NH₂ insulin (23) (reevaluated here) indicated mobility of the B²⁶ region. It is to be expected that there is conformational freedom in this structural element and that the formation of B26 turn can be readily achieved.

Therefore, we propose that these structural changes (formation of B26 turn and trans-cis isomerisation of the Phe^B25-Tyr^B26 peptide bond) follow the detachment of the Tyr^B26-Thr^B30 segment from the insulin molecule during, or on, binding to the receptor. These observations are consistent with the induced-fit theory of insulin-IR binding and with trans/cis peptide bond isomerisation analyses (26–29). The protein environment supplied by IR is important in the unmasking and stabilisation of the active surfaces of insulin. We suggest that the analogues with higher affinity described here have an active surface more representative of the conformation of insulin bound to the receptor than do the structures of the storage forms of the native hormone. The scale and nature of the structural rearrangements leading to highly active analogues underlines a “chaperone-like” role of the IR in insulin activation (30), especially if trans-cis peptide isomerisation is one of the key steps.

The restructuring of the B-chain and the concerted changes to the N-terminal helix of the A-chain reveal a large hydrophobic area made up of Gly^A1, Ile^A2, Val^A3, and Tyr^A19 priming them for direct interactions with the IR (Fig. 2, 4). The importance of Ile^A2 to receptor binding is indicated by its complete exposure on the IR-binding surface and by the striking spatial invariance of its Cα atom in these transformations. The structural conservation of the A² Cα suggests its role as the A-chain main chain pivot in insulin activation. The shift of Tyr^A19 into hydrogen bond contact with Gln^A5 further stabilises the A-chain IR-binding surface with the edge of the Tyr^A19 ring transforming the Gly^A1-Ile^A2-Val^A3 hydrophobic surface into a three-rim (Tyr^A19, Gly^A1-Ile^A2, and Val^A3) evenly spaced “epitope” (Fig. 4).

The key structural roles of Phe^B24 and Phe^B25 in insulin activity, well known from several studies (4, 15), are firmly confirmed by structures of the highly-active insulins described here. The aromatic ring of Phe^B24 is spatially invariant in native hormone and all other insulin analogues. This is in contrast to the ∼90° rotation of Phe^B25 side chain that subsequently overhangs (by ∼7 Å) the A-chain part of site one (Tyr^A19-Gly^A1-Ile^A2-Val^A3) “epitope” in the highly active insulins. However, in spite of its considerable side-chain movement the position of the Phe^B25 Cα atom is well conserved in different forms of insulin, confirming further the role of Phe^B24 in stereo-specific detachment of the B-chain β-strand (10). Therefore, it may be suggested that the Phe^B24 is a B-chain side-chain pivot during insulin activation (Fig. 2). Phe^B24 is structurally invariant side chain that, being anchored into the hydrophobic cavity lined up by Leu^B15, Val^B12, Tyr^B16 (its side chain collapses on this cavity in the activation process sealing it off) and Cys^B19, provides structural stability during rearrangement of the C-terminal segment of the B-chain. Phe^B25 complements the role of Phe^B24 in this process acting as the B-chain main chain pivot with the B²⁵ Cα as the first B-main chain turning point in formation of the B26 turn and, subsequently, initiation of IR binding (Fig. S7).

In addition to the structural convergence of the B²⁴-B²⁶ region in the highly active insulin analogues the consistent similarity of the unique B³-B⁵ pseudo β-turn of B-chain N termini observed there should be also noted. This turn locates the N terminus conformation in-between the T- and R-forms. However, the physiological relevance of this unique state of the hormone (e.g., as a snapshot of the T-R transition or as its more stable high-activity related form) is unclear and requires further studies.

Lower Activity Insulin Analogues Fit the B26 Turn-Related Activation Model.

Interestingly, some of the B²⁶-truncated analogues: [NMeTyr^B26]-DTI-NH₂ and [NMePhe^B26]-DTI-NH₂ (22), show lower activities, 72% and 36% respectively (Table 1). At present, there is no straightforward structural explanation for this behavior. In the native, full-length insulin the replacement of Tyr^B26 to Ala (without any N-methylation) results in a lowering of its activity to ∼36% (14). However, the presence of a more bulky aromatic side chains (Phe^B26/Tyr^B26) in the B²⁶-truncated (and N-methylated) analogues may result, for example, in a detrimental, entropic effect leading to the destabilization of the short B26 turn. It is interesting that these analogues have not crystallized, suggesting an increased flexibility that has also been observed by NMR for a related truncated insulin analogue (31).

The crystal structures of the full-length analogues result in intriguing observations. Firstly, only the monomer structures of [NMeAla^B26]-insulin have been obtained under both “monomer” and “hexamer” crystallization conditions whilst both monomer and hexamer crystals of [NMeTyr^B26]-insulin were grown in the relevant monomer and hexamer crystallization environments. This indicates that the monomer conformation is preferentially stabilized by Ala in the B²⁶ position, consistent with the high potency of the B²⁶ truncated insulin. The reduced potencies of [NMeTyr^B26]-insulin and [NMeAla^B26]-insulin analogues (∼21%, Table 1) confirm the trend that the full-length analogues are generally less active than their truncated homologues (4) and suggest that this may result from several disadvantageous steric and entropic factors originating in the B²⁷-B³⁰ region. For example, (i) the B²⁷-B³⁰ segment may require a significant contribution from the IR to be induced to a fully active conformer of insulin, (ii) the structural constrain of the B26 turn may not impose the optimal conformation for the B²⁷-B³⁰ residues, and (iii) the B26 turn may not represent the final and the most optimum active insulin conformation (Fig. S7).

Molecular Shape of Active Insulin.

The structure-function relationships of the highly-active insulin analogues described here not only give an insight into detailed definition of the putative local structural signatures of an active form of insulin, but shed also light on the possible molecular shape of the hormone’s active conformer. In the process of insulin activation the Phe^B25 and the B26 turn form a hydrophobic knob that together with the A-chain “epitope” defines a unique, highly hydrophobic IR-binding cleft, lined by Leu^B15, Leu^B11, and Val^B12 side chains [Fig. 4; it should be mentioned that the normal insulin B²⁵-B³⁰ substructure modeled in Fig. 4B is actually in more intimate contacts with other A- and B-chain residues in the ground-state insulin molecule due to a different relative orientation of these chains in highly-active insulin analogues and its ground form (Fig. 2)]. The shape of this cleft is reminiscent of a macromolecular hand that ensures the grip on the receptor by a hydrophobic thumb (Phe^B25 and B26 turn), a hydrophobic palm (Leu^B15/Leu^B11/Val^B12), and hydrophobic fingers (Tyr^A19, Gly^A1-Ile^A2, and Val^A3).

The molecular features of several highly-active, and related, insulin analogues discussed here provide unique structural elements for the conformational landscape of the hormone. They give considerable insight into the activation of human insulin and should help in development of unique types of hormone analogues for the treatment of diabetes mellitus.

Material and Methods

Semisynthesis of Insulin Analogues and Binding Assays.

In general, solid-phase synthesis of peptides, enzymatic semisynthesis of analogues, purification of analogues, preparation of rat adipose membranes, and binding assays were performed as described previously (21, 22) (SI Text).

Crystallization, Structure Determination, and Refinement.

Crystallization conditions (Table S1) were found initially by an in-house developed screen based on the available insulin crystallization protocols and they were refined individually by the hanging drop method. X-ray data (Tables S2–S4) were processed by HKL2000 (33) and all computational crystallography was performed by the CCP4 suite of programs (34). Structures were determined by molecular replacement with differently truncated 1mso.pdb-derived monomeric insulins as the model (32). Crystallization details, x-ray data, refinement statistics, and model building protocols are provided in SI Text.