Se‐Glargine: Chemical Synthesis of a Basal Insulin Analogue Stabilized by an Internal Diselenide Bridge

Insulin has long provided a model for studies of protein folding and stability, enabling enhanced treatment of diabetes mellitus via analogue design. We describe the chemical synthesis of a basal insulin analogue stabilized by substitution of an internal cystine (A6–A11) by a diselenide bridge. The studies focused on insulin glargine (formulated as Lantus® and Toujeo®; Sanofi). Prepared at pH 4 in the presence of zinc ions, glargine exhibits a shifted isoelectric point due to a basic B chain extension (ArgB31−ArgB32). Subcutaneous injection leads to pH‐dependent precipitation of a long‐lived depot. Pairwise substitution of CysA6 and CysA11 by selenocysteine was effected by solid‐phase peptide synthesis; the modified A chain also contained substitution of AsnA21 by Gly, circumventing acid‐catalyzed deamidation. Although chain combination of native glargine yielded negligible product, in accordance with previous synthetic studies, the pairwise selenocysteine substitution partially rescued this reaction: substantial product was obtained through repeated combination, yielding a stabilized insulin analogue. This strategy thus exploited both (a) the unique redox properties of selenocysteine in protein folding and (b) favorable packing of an internal diselenide bridge in the native state, once achieved. Such rational optimization of protein folding and stability may be generalizable to diverse disulfide‐stabilized proteins of therapeutic interest.


Introduction
Insulin provides a classical model for foundational studies of protein biophysics with potential therapeutic application (Figure 1). [1]The total chemical synthesis of insulin 60 years ago defined a landmark in this field. [2]A critical step was provided by insulin chain combination, enabling native disulfide pairing: such fidelity demonstrated that the folding information in proinsulin resides within the isolated A and B chains.Despite the broad utilization of chain combination to investigate structure-activity relationships and to prepare therapeutic analogs, [3] a particular and surprising challenge has been posed by the chemical synthesis of insulin glargine: classical chain combination fails. [4]Given the clinical importance of glargine to global health [5] (magnified by a growing pandemic of Type 2 diabetes mellitus; T2D), [6] complex protocols have been developed to circumvent technical problems associated with its shifted isoelectric point (due to a di-Arg C-terminal extension of the B chain) and limited solubility of the variant A chain (due to substitution of Asn A21 by the more hydrophobic Gly) (Figure 1B).Tactics have included use of solubility tags, [7] isoacyl dipeptides and pairwise cysteine deprotection to direct specific disulfide pairing. [8,9]Here, we investigate the utility of a simpler strategy to rescue the yield of chain combination: substitution of Cys A6 and Cys A11 by selenocysteine (Sec), also in the case of a more challenging analogue such as glargine.In the context of wildtype human insulin, an A6-A11 diselenide bridge is compatible with native function (as probed in a cell-based model) [10] and structure (as probed by high-resolution NMR spectroscopy [11] ).The overarching goal of this study was to test whether this strategy might generalize to a clinical analogueinsulin glargine-otherwise refractory to conventional chain combination.If so, Se-insulin analogues could provide a universal platform to enhance the yield and stability of all current clinical formulations, enhancing global access to insulin replacement therapy.
Selenocysteine (Sec, U), a near isostere of Cys and the 21 st proteinogenic amino acid, [12] has provided an elegant tool in peptide chemistry [13] to enhance the efficiency of protein folding. [14]Because the selenol group of Sec has a lower pK a (near 5.2) [15] and lower reduction potential (E 0 = À 388 mV) [16] relative to the thiol group of Cys, pairwise Cys-to-Sec are pertinent to the nascent folding of proteins (including insulin) [17] the present study has investigated whether such redox properties might rescue native combination, at least in part, between the modified A and B chains of insulin glargine.We also sought to test whether a specific diselenide bridge at positions A6-A11 might stabilize the folded state, once obtained.The latter possibility was suggested by studies of wild-type human insulin, [17] which motivated the hypothesis that the selenium atom's larger atomic size and longer bond lengths (relative to sulfur) [18] might mitigate cryptic packing defects in the hydrophobic core. [19]ermodynamic stabilization of insulin glargine could be of translational interest as a way to extend its shelf life at or above room temperature.This issue reflects the pharmaceutical formulation of insulin glargine in an acidic solution wherein insulin cannot form protective zinc-stabilized hexamers. [20]trikingly, our findings demonstrate that whereas control chain-combination of glargine's A and B chains resulted in negligible product (in accord with prior studies), [4,21] the Secbased route enabled purification of an active insulin glargine analogue (designated "Se-glargine").Although yield was lower than that of wildtype insulin chain combination, sufficient product could be obtained through repeated combination reactions to enable assessment of function and stability.6a,22] Although prepared herein by chemical synthesis and chain combination, [17] advances in methods for recombinant Sec incorporation suggest that our strategy will be compatible with large-scale manufacture and generalizable to diverse proteins of therapeutic or industrial interest. [23]

Results and Discussion
We previously described the preparation of human Se-insulin, in which the internal intrachain disulfide bond A6-A11 was replaced with diselenide via pairwise substitution of Cys A6 and Cys A11 by Sec. [17]To this end, a modified 21-residue peptide, designated A chain[C6U, C11U], was prepared in high yield by solid-phase peptide synthesis (SPPS); the wild-type B chain was prepared by sulfitolysis of human insulin (of either synthetic or biosynthetic origin).The purified chains were then mixed in the combination buffer to yield, after only 8 h (foreshortened relative to wildtype chain combination; 24-38 h), the desired Se-insulin in 34 % isolated yield. [17]The redox properties of Sec A6 and Sec A11 thus resulted in a > fourfold acceleration of oxidative protein folding.Further, whereas structure and activity were almost identical to wild-type (WT) insulin, the diselenide substitution enhanced stability (ΔG u ) by ca. 1 kcal/mol, delaying both protease degradation and reductive unfolding. [17]ncouraged by these results, we sought to test whether an analogous A6-A11 diselenide substitution could likewise stabilize a clinical insulin analogue without affecting its function.Insulin glargine was chosen based on the medical importance of products Lantus® and Toujeo® (Sanofi) as basal (long-acting) formulations. [24]his analogue contains two types of modifications.The first is addition of two basic residues at the C-terminus of the B chain (Arg B31 À Arg B32 ), which shifts the isoelectric point to neutral pH, thereby making the protein molecule more soluble at pH 4 (as in an acidic pharmaceutical formulation) and yet insoluble at physiological pH (as on injection into a subcutaneous depot).The second modification is replacement of Asn A21 by Gly, which enhances chemical stability [24] by circumventing acid-catalyzed deamination of the native side chain. [20,25]e therefore undertook the chemical synthesis of the two component peptide chains of Se-glargine: chain A [C6U, C11U, N21G] variant and a chain B variant containing Arg B31 À Arg B32 by SPPS; the latter B chain analogue was also obtained from Lantus® SoloStar® pen by sulfitolysis (see below).Fluorenylmethyloxycarbonyl (Fmoc)-based SPPS was used to synthesize A chain [C6U, C11U, N21G], wherein Sec residues were inserted manually at positions 6 and 11.After cleavage and deprotection by standard conditions, the peptide was isolated in 9 % yield (Figure S1 in the SI).The variant B chain was also synthesized and purified in 20 % yield (Figure S2), then sulfonated (Figure S3).Oxidative sulfitolysis of a commercial formulation of Lantus® SoloStar® pen was also employed to obtain sulfonated glargine A-and B chains (Figure S4).
With the A chain [C6U, C11U, N21G] and sulfonated B chain in hand, we initiated the combination reaction in 0.1 M glycine buffer at pH 10.6 (standard conditions).First, A chain [C6U, C11U, N21G] was dissolved (at a final concentration of 0.6 mM), and the sulfonated B chain was added (final concentration of 0.5 mM).At this pH, precipitation of the B chain was observed, presumably due to the Arg B31 À Arg B32 extension, which shifts the isoelectric point of the peptide.The pH was adjusted to 11.2 (by addition of 0.5 M NaOH), which enabled the B chain to be completely dissolved, and then DTT was added at a stoichiometric concentration relative to the sulfonate groups present in the B chain (1 mM).The reaction was left at 4 °C with exposure to air to permit oxidation. [26]Aliquots were taken at successive time points, quenched with 0.1 % trifluoroacetic acid (TFA) in water and left in the freezer until injected into the reversedphase HPLC.Under these conditions, Se-glargine was observed after 5 h in small amount, and so the reaction was left overnight; as expected, aggregates of the variant B chain were also observable.After 24 h, Se-glargine was isolated (ca.20 % by HPLC integration; with 7 % isolated yield, Figure S5).The sulfonated B chain from Lantus® SoloStar® pen (Sanofi) obtained by oxidative sulfitolysis, was also tested in the combination reaction.Surprisingly, the yield of the combination reaction was significantly improved (Figure 2), and Se-glargine started to form after only 3 h; the reaction was completed after 18 h.The Se-glargine product was the major peak under these conditions; A chain [C6U, C11U, N21G] was still present at the end of the reaction (which had been introduced in slight excess), whereas negligible amounts of the B chain were observed.Under these conditions, Se-glargine was isolated in 20 % yield (ca. 60 % by HPLC integration; Figure 2A).For comparison, the same conditions were applied using sulfonated A and B chains isolated from a Lantus® SoloStar® pen by sulfitolysis.Glargine insulin was observed after 24 h at very low yield; after 48 h, the yield was still low (> fivefold lower yield than that of Se-glargine; ca.11 % by HPLC integration) whereas aggregation of the chains was prominent (Figure 2B).
To evaluate the biological activity of Se-glargine in relation to recombinant control analogues, we employed a cell-based assay probing hormone-dependent autophosphorylation of the insulin receptor (IR) in human liver-derived cell line HepG2. [11,27]This assay exploited immunofluorescent detection of the total IR and the phosphorylated IR (pIR) to determine the pIR/IR ratio on exposure of the cells to successive concentrations of an insulin analogue (Figure 3A,B).The dose-dependence of IR autophosphorylation, as triggered by addition of synthetic Se-glargine, was indistinguish-able (Figure 3C) from that of control analogues repurified by HPLC from commercial sources: recombinant insulin glargine (Lantus® [Sanofi]) or recombinant insulin lispro (Humalog®, a rapid-acting analogue formulation [Eli Lilly and Co.]).
In accordance with our previous study of Se-insulin (at neutral pH), Se-glargine (in a zinc-free acidic solution at pH 4.0 and 25 °C) exhibited marked resistance to protein unfolding (relative to recombinant glargine or wildtype human insulin) on addition of a chemical denaturant (guanidine hydrochloride) as indicated by quantitative thermodynamic modelling [11] (shifted red curve in Figure 3D).Such analysis suggested that the free energy of unfolding (ΔG u ) of native glargine (2.8(� 0.1) kcal/ mol) was augmented by 1.0(� 0.2) kcal/mol (ΔΔG u ) on insertion of the A6-A11 diselenide bridge.Under these conditions, the inferred thermodynamic stability of wildtype human insulin (3.8(� 0.1) kcal/mol) is similar to that of Se-glargine (3.8(� 0.1) kcal/mol), despite distinct titration midpoints (C m = 5.0(� 0.1) M vs. 6.0(�0.1) M; SI Table S1), due to enhanced cooperativity of unfolding (m values in Table S1).
Because insulin disulfide isomers exhibit markedly decreased activity and stability, [28] the findings in Figure 3 also imply that Se-glargine chain combination gives native pairing (A6-A11, A7-B7 and A20-B19). [17,28]The A6-A11 diselenide bridge in Se-glargine thus "rescues" the impaired stability of native glargine.Fitting parameters of a two-state model are provided in the Supporting information.

Pairwise substitution of Cysteine by Selenocysteine
Selenocysteine (Sec, U), the 21 st encoded amino acid, [12a] is a near isostere of cysteine (Cys, C) that confers unique chemical properties, such as its low redox potential (E 0 = À 386 mV) and pK a (~5.2) [15] relative to Cys.Therefore, at physiological pH, the Sec side chain is predominantly deprotonated (selenolate), whereas most Cys side chains remain protonated.These characteristics have been exploited in the field of protein folding in vitro. [14,29]Marked differences in redox properties occur even at extremely basic pH wherein both Cys and Sec are deprotonated.These differences motivated the present investigation to test whether a critical pairwise Sec substitution in insulin glargine might mitigate its anomalous impairment of classical chain combination at pH 10.6.
In addition to its clinical importance, insulin glargine provides a model for general issues in protein biophysics and peptide chemistry.Correct protein folding is critical for protein function and ordinarily occurs spontaneously under suitable in vivo conditions.The associated in vitro folding process can nonetheless be challenging, especially in the case of Cys-rich proteins such as insulin.Such challenges reflect formation of non-native disulfide bonds and formation of trapped intermediates-both of which impairing the yield of properly folded protein molecules.Remarkably, and yet in accordance with foundational chemical principles, Cys-to-Sec substitutions have been demonstrated to enhance the oxidative folding of Cys-rich proteins in vitro without perturbing their respective native conformations or biological activities. [14,30]n our previous paper we observed that substitution of a single disulfide pair (at positions 6 and 11 in A chain) with Sec markedly enhanced chain-combination rate and yield; the purified product maintained nativelike biological structure and activity. [17]Further, Se-insulin demonstrated enhanced stability, which was attributed to the enlarged atomic radii of the Se atoms (in Sec) relative to S (in Cys): such subtle enlargement presumably afforded more efficient packing of the diselenide bond in the hydrophobic core without incurring steric clash.Such enhanced packing efficiency in the native state would not be related to redox chemistry and would in principle depend on the detailed structural environment of a given disulfide or diselenide bridge.

Challenges in chain combination of glargine
The Arg B31 À Arg B32 extension in the glargine B chain shifts the isoelectric point of the chain and its pH-dependent solubility, making specific combination with the A chain more challenging than analogous reactions employing the wildtype B chain or its many conventional analogues. [24]In fact, the glargine B chain is soluble in slightly more basic conditions relative to regular B chain (pH 11.2 vs 10.6), and this small change in pH during the reaction can cause insolubility of the B chain and its aggregations.Whereas synthetic glargine B chain under standard combination conditions yields essentially no product, increasing the pH to 11.2 provided the Se-glargine in 7 % isolated yield.Synthetic Se-glargine was isolated by HPLC (Figure S6 in the SI) and characterized by HR-MS (Figure S7 in the SI).
The use of glargine B chain isolated by an oxidative sulfitolysis reaction (as applied to commercially available Lantus® SoloStar® pen; Sanofi) further enhanced the rate of the combination reaction and its yield (by threefold to 20 %); only negligible amounts of the di-Arg-extended B chain were observed at the end of the reaction.These improvements were presumably due, at least in part, to the presence of polysorbate 20, a surfactant present in the Lantus® SoloStar® pen.Known to enhance stability, [31] polysorbate 20 may have remained present in low concentration after purification of the glargine B chain.We suggest that this surfactant functioned to reduce formation of glargine B chain aggregates that compete with chain combination, thereby improving the yield of Seglargine formation.In a future study [32] we will report that native protein structure is maintained (as probed by highresolution NMR spectroscopy) with nonlocal damping of conformational fluctuations (as probed by amide-proton exchange in D 2 O solution). [11]e envision that A6-A11 diselenide bridge provides a dual advantage in insulin chemistry: kinetic and thermodynamic.The kinetic advantage derives from accelerated initial pairing of Sec A6 and Sec A11 in chain combination, which enables productive pairing to compete more effectively with off-pathway reactions.Kinetic control of classical chain combination rationalizes its successful application to the chemical synthesis even of analogues of impaired product stability. [26]In contrast, the augmented thermodynamic stability of Se-glargine (relative to its parent analogue) accrues to the native state, once reached.We ascribe such stabilization to enhanced efficiency of core packing, a subtle mechanism unrelated to the redox chemistry of selenocysteine. [17]The ability of a diselenide bridge to enhance core-packing efficiency would reflect the details of side-chain packing in the native state, which are likely to vary from protein to protein.

Conclusions
The present study has exploited the striking finding that the intrachain diselenide bond A6-A11 provides an efficient tool to enhance the rate and yield of insulin chain combination.Here, we have demonstrated that even in the case of a more challenging insulin analogue, such as glargine, this "bridge substitution" can overcome kinetic barriers to chain combination.Indeed, Se-glargine was obtained in sufficient yield to permit its characterization, whereas control reactions employing the unmodified glargine A chain exhibited negligible yield.Synthetic Se-glargine retained native-like activity.
Strikingly, the A6-A11 diselenide bridge was observed to mitigate the reduced thermodynamic stability of native glargine, attenuated relative to wildtype insulin under acidic formulation.This bridge substitution provides a subtle modification of a protein molecule as the atomic radius of selenium (in Sec) is slightly larger than that of sulfur (in Cys). [32]This improvement in stability may have clinical implications: studies of Se-glargine's shelf life and rate of physical degradation (in a formulation similar to those of Lantus® or Toujeo®; Sanofi) are in progress and appear promising.We anticipate that substitution of cystine A6-A11 by a diselenide bridge can provide a general approach to stabilize all pharmacokinetic classes of insulin products (e.g., from ultra-fast and rapid-acting to basal and ultra-basal).Such application may in turn improve the distribution, storage and use of insulin, particularly in the developing world.Insulin stability and access are of overarching importance in global health, particularly in the face of an emerging pandemic of obesity-associated T2D. [6]Further, beyond insulin chemistry, Sec-based optimization of diverse therapeutic and industrial proteins-amenable to non-standard recombinant manufacturing [23] -holds great promise as an emerging biotechnology.

Supporting Information
Methods in detail pertaining to the synthesis, purification and characterization of Se-glargine, including assays of hormonal activity and thermodynamic stability; Table S1 providing twostate modeling of chemical denaturation assays.

Notes and references
[ §] Unlike an internal diselenide bridge at position A6-A11, an external diselenide bridge at A7-B7 does not stabilize the insulin molecule as probed by chemical-denaturation studies. [33] recent [34] report by Arai et al. suggests that the diselenide bridge at A7-B7 can lead to aggregation of the bovine Seinsulin, which explains its observed extended activity. Because the thermodynamic stability of a protein reflects the difference between the free energies of the bridged native state and bridged denatured state (as distinct from the reduced and unfolded state), any intrinsic physico-chemical differences in the redox properties between Cys and Sec do not in themselves contribute to net free-energy differences (ΔΔG u ).By contrast, differences in steric properties between a disulfide bridge and diselenide bridge can in principle introduce favorable or unfavorable changes in core packing efficiency.[33] [#] Any potential risk of selenium toxicity that might be posed by Se-insulin analogues for diabetes treatment is mitigated by the low molar dose of selenium atoms relative to the toxic threshold (900 μg/day) and maximum daily dose (400 μg).[41] Considering a 50-kg adolescent with T1D, an average of 0.75 unit/kg insulin injection per day, spread over three meals, would typically be used; this is equivalent to 0.4 mg of insulin, containing only ~11 μg of selenium per injection, much below the Recommended Daily Allowance (RDA, 55 μg/day) and the toxic limit (glargine is given once daily).Even in T2D with extreme insulin resistance (with daily insulin doses as high as 200 units), a similar calculation indicates that the cumulative selenium exposure would be markedly below the toxic threshold.We note in passing that selenium is an essential micronutrient, which is found in at least 25 human selenoproteins (in the form of selenocysteine). [12]

Figure 1 .
Figure 1.Three-dimensional structure and schematic representation of glargine.(A) 3D structure of glargine (the residues Thr B30 , Arg B31 , Arg B32 are not shown due to limited electron density), with three disulfide bonds A6-A11, A7-B7 and A20-B19 indicated as sticks (PDB : 4Iyd).(B) Schematic representation of glargine with the changes shown (in green).For Se-insulin and Se-glargine, the Cys residues at A6 and A11 (in orange) were replaced by Sec.

Figure 2 .
Figure 2. Analytical RP-HPLC of the recombination reaction assay for Seglargine and glargine.(A) For Se-glargine, chain A[C6U, C11U, N21G] analogue with sulfonated B chain from Lantus® Solostar® pen in a 0.1 M glycine buffer (pH~11.2) at 4 °C, with the addition of DTT in quantity stoichiometric to the concentration of sulfonate groups in the reaction.Conditions: A chain [C6U, C11U, N21G] = 0.6 mM, [sulfonated B chain] -= 0.5 mM, [DTT] = 1 mM.The reaction was completed after 18 h, with only minor aggregation of the B chain observed at the end of the reaction, and with the Se-glargine as the major peak.(B) For glargine preparation: sulfonated A and B chains from Lantus® Solostar® pen in a 0.1 M glycine buffer (pH~11.2) at 4 °C, with the addition of DTT in quantity stoichiometric to the concentration of sulfonate groups initiated the reaction.Conditions: [sulfonated A chain] = 0.6 mM, [sulfonated B chain] = 0.5 mM, [DTT] -= 3.4 mM.A peak corresponds to the mass of glargine was observed after 24 h in very small amount, and after 48 h the yield of the reaction was still low.Aggregations of the glargine A chain (a broad peak Rt = 8.2-9.2); and B chain (a broad peak Rt = 11-12.2) were observed at the end of the reaction.

Figure 3 .
Figure 3. Biological activity and stability of Se-glargine.(A, B) Design of an in-cell Western blot assay: activities of insulin analogues were evaluated in human cell line HepG2 in parallelizable 96-well plates.For detailed experiment design, see Supporting Information (section "Plate-based Fluorescence Immunoblotting Assay").Treatment of the cells with successive doses of a given insulin analogue (A) enables fluorescent detection of hormonedependent receptor autophosphorylation (B).Hormone-induced IR autophosphorylation was detected via optical readouts (schematic in the upper panel: pseudo-green colour signals) as normalized by DRAQ5 700 nm emission (lower panel; pseudo-red colour signals); the latter provided a control for cell number (as described in the Supporting information).(C) Fold-change in pIR/IR autophosphorylation (vertical axis) is shown as a function of insulin analogue dose in the range 50 pM-1 μM (horizonal axis; log scale): ( * ), insulin lispro, ( * ), insulin glargine and ( * ), Se-glargine.Vertical error bars represent standard errors; assays were performed in biological triplicate.(D) CD-monitored chemical denaturation studies.Attenuation of helix-sensitive ellipticity at 222 nm is shown as a function of guanidine-HCl concentration.Symbols are the same as in panel C except human insulin (^).Fitting of these data to a two-state model (native state versus unfolded state) is described in the Supporting information.