Supramolecular approaches for insulin stabilization without prolonged duration of action

Abstract

Aggregation represents a significant challenge for the long-term formulation stability of insulin therapeutics. The supramolecular PEGylation of insulin with conjugates of cucurbit[7]uril and polyethylene glycol (CB[7]‒PEG) has been shown to stabilize insulin formulations by reducing aggregation propensity. Yet prolonged in vivo duration of action, arising from sustained complex formation in the subcutaneous depot, limits the application scope for meal-time insulin uses and could increase hypoglycemic risk several hours after a meal. Supramolecular affinity of CB[7] in binding the B1-Phe residue on insulin is central to supramolecular PEGylation using this approach. Accordingly, here we synthesized N-terminal acid-modified insulin analogs to reduce CB[7] interaction affinity at physiological pH and reduce the duration of action by decreasing the subcutaneous depot effect of the formulation. These insulin analogs show weak to no interaction with CB[7]‒PEG at physiological pH but demonstrate high formulation stability at reduced pH. Accordingly, N-terminal modified analogs have in vitro and in vivo bioactivity comparable to native insulin. Furthermore, in a rat model of diabetes, the acid-modified insulin formulated with CB[7]‒PEG offers a reduced duration of action compared to native insulin formulated with CB[7]‒PEG. This work extends the application of supramolecular PEGylation of insulin to achieve enhanced stability while reducing the risks arising from a subcutaneous depot effect prolonging in vivo duration of action.

Graphical abstract

The supramolecular PEGylation of N-terminal acid-modified insulin derivatives produces a stable formulation with a shorter duration of in vivo action, reducing the risk of hypoglycemia arising from subcutaneous depot.

1. Introduction

Insulin has been the primary therapy for blood glucose control in people with diabetes since its discovery in 1921¹. However, native insulin instability has also been a major obstacle to its long-term use and storage, especially in tropical areas². Prandial and basal insulin analogs have been developed over the years and shown to achieve better blood glucose control^3,4. Despite being stabilized through hexamer formation, these commercial formulations still have a high propensity to form aggregated fibrils at ambient temperatures, requiring continuous refrigerated storage and transportation5, 6, 7, 8. Several strategies were developed to improve insulin stability through insulin modifications, including designing insulin with four disulfide bonds, single-chain insulins, and replacing one or more of the three disulfide bonds with non-disulfide bonds9, 10, 11, 12, 13.

Supramolecular or non-covalent PEGylation is a new approach to enhance the shelf-life stability of insulin formulations¹⁴. Non-covalent PEGylation offers advantages over covalent PEGylation in that it does not permanently modify the protein or reduce its in vivo potency, while also having limited impact on insulin pharmacokinetics¹⁵. Insulin interacts via its B1-Phe in binding to the cucurbit[7]uril (CB[7]) macrocycle with K_d = 0.5 μmol/L¹⁶. This supramolecular complex was thus used to develop a route for supramolecular PEGylation of insulin using a CB[7]‒PEG conjugate to prepare an insulin formulation free from aggregation during continuous agitation for >100 days at 37 °C (Fig. 1A)¹⁴. The enhanced in vitro insulin stability, however, led to prolonged in vivo duration of action due to a depot effect in the high local concentrations of the subcutaneous (SC) space. This effect was accentuated when the affinity between CB[7] and insulin was further increased with non-native guest chemistries replacing the CB[7]–Phe interaction¹⁷. Though a stable formulation is desirable for use in insulin pumps, the accompanying protracted in vivo action may limit its application in SC meal-time uses (Fig. 1A). The concentration of insulin in a standard U100 pharmaceutical preparation is ∼0.6 mmol/L and this concentration decreases by about 10-fold after SC injection. Since the K_d of the CB[7]–insulin complex is ∼0.5 μmol/L, most of the polymeric excipient remains bound to insulin following SC administration, leading to a depot effect extending the in vivo duration of action. We hypothesized that the introduction of pH-tunable affinity interactions between CB[7] and insulin, specifically targeting affinity in vivo that is more closely aligned with the SC insulin concentration (K_d ∼ 0.06–0.6 mmol/L), would enable insulin stabilization in formulation while also offering rapid complex dissociation in the SC space, thereby reducing the in vivo duration of action (Fig. 1B).

Figure 1.

Non-covalent PEGylation of both native insulin and N-terminal modified analogs with CB[7]‒PEG results in stabilized formulations. (A) The high affinity between CB[7]‒PEG and insulin promotes complex formation in the subcutaneous (SC) depot and prolongs insulin in vivo action. (B) N-terminal modification with a benzoic acid group results in a stable formulation when coupled with CB[7]‒PEG while offering a shorter duration of in vivo action. As the acid group has weak to no affinity for CB[7] at physiological pH, this enables complex dissociation once introduced into the SC space, enabling improved use of the formulation for mealtime SC insulin.

Previously, we developed a reductive alkylation strategy to selectively functionalize N-terminal amines in the presence of internal lysine residues and further showed that this strategy could be used to modulate peptide–CB[7] interaction in a pH-dependent manner through N-terminal substitution with benzyl carboxylate substituents^18,19. In the present work, we employed this approach to design an insulin variant with lower CB[7] binding affinity than the native protein. Specifically, the B1-Phe was mutated to Ala and further modified with benzyl carboxylates to generate N-terminal acid-modified insulin derivatives. Equilibrium binding studies have demonstrated that the interaction between N-terminal B1-Phe of insulin and CB[7] offers higher binding affinity than CB[7] interactions with internal residues^16,20. Thus, modifying the B1-Phe offers an opportunity to tune the interaction affinity between CB[7]‒PEG and insulin. Importantly, the charge state of the benzyl carboxylate is expected to afford CB[7] binding dependent on the pH level relative to pK_a of the acid group; this design leverages the known aversion to CB[7] binding of negatively charged guests²¹. Attaching benzyl carboxylate groups to the B1 N-terminus of insulin indeed resulted in a pH-dependent interaction with the CB[7]‒PEG, with weak to no affinity at neutral pH where the carboxylate is deprotonated (Fig. 1B). At formulation concentrations, these modified insulins showed pH-dependent aggregation stability when formulated with equimolar CB[7]‒PEG. However, the weakened interaction between the acid substituents and CB[7]‒PEG in the SC space expedited insulin uptake from the depot and reduced its duration of action compared with native insulin formulations.

2. Results and discussion

2.1. Synthesis of N-terminal acid-modified insulin analogs

The synthetic path to N-terminal acid-modified insulins, Ins-1 and Ins-2 is described in Scheme 1. First, the full-length A and B chains (with B1 Phe➔Ala mutation) were synthesized by solid-phase peptide synthesis (SPPS) (Scheme 1). The protected A-chain (A⁶, A⁷, A¹¹-Acm) A1 with A⁸‒A⁹ isoacyl dipeptide was synthesized on a Rink amide resin and treated with a cocktail of TFA/TIPS/H₂O to afford the Acm-protected product A2 at 40% yield. Then, the protected B-chain (B⁷-Acm) B1 was synthesized on a 2-chlorotrityl chloride (2-CTC) resin (Scheme 1). B1 was cleaved from resin by treating with an acid cocktail (TFA/TIPS/H₂O) containing 15 eq. of 2,2′-dithiodipyridine (DTDP) to provide B2 at 42% yield. A chain A2 and B chain B2 were combined in 6 mol/L urea added to a 0.2 mol/L ammonium bicarbonate (NH₄HCO₃) pH 7.5 buffer for 30 min to provide the two-chain AB1 with a disulfide bond between A²⁰ and B¹⁹ at 68% yield. Iodine oxidation of AB1 performed in 33% acetic acid/water afforded the intermediate insulin product B1-Ala insulin at 25% yield. B1-Ala insulin can be further modified using N-terminal reductive alkylation as previously reported¹⁹. Briefly, the insulin product was treated with 2 eq. of either 4-formyl benzoic acid or 4-formyl phenylacetic acid in pH 6.1 citric acid buffer in the presence of excess sodium cyanoborohydride (NaBH₃CN) to afford the desired N-terminal acid-modified insulins, Ins-1 and Ins-2, at yields of 45%‒50% after purification. The final insulin products were subjected to 1,4-dithiothreitol (DTT) cleavage and trypsin digestion to confirm selective N-terminal B-chain modification (Supporting Information). Alternatively, Ins-1 and Ins-2 could be obtained via a different synthetic route, whereby the acid substituents were attached at the N-terminus of the B-chain during the early synthesis phase (Supporting Information Fig. S1).

Scheme 1.

Synthesis of N-terminal acid-modified insulin analogs Ins-1 and Ins-2 with benzoic acid motifs installed for pH-dependent CB[7] recognition.

2.2. N-terminal acid-modified insulins have native-like potency in vitro and in vivo

Insulin is known to be tolerant of N-terminal B-chain substitutions with limited impact on bioactivity^22,23. To confirm retention of function for analogs here, a phospho-AKT (Ser 473) insulin signaling stimulation assay was used to evaluate the bioactivity of Ins-1 and Ins-2. The acid-modified analogs displayed similar potency (EC₅₀ = 4.00 nmol/L for Ins-1 and 5.88 nmol/L for Ins-2) as native insulin (EC₅₀ = 3.38 nmol/L) (Fig. 2a). These results suggest that N-terminal substitution does not affect engagement with the insulin receptor. The activity confirmed in cell-based assay results were further supported by an insulin tolerance test (ITT) in healthy rats. Briefly, equal amounts of native insulin, Ins-1, and Ins-2 (1 IU/kg) were subcutaneously injected in 8–10 weeks old male rats and blood glucose levels were serially measured. As shown in Fig. 2b, the two insulin analogs have glucose-reducing effects comparable to native insulin, suggesting similar in vivo potency.

Figure 2.

Binding studies, activity assays, pH-dependent aggregation assays, and in vivo activity assays of native insulin and N-terminal acid-modified analogs Ins-1 and Ins-2. (a) Cellular activities of insulin and analogs as assessed by phospho-AKT (pAKT) in NIH-3T3 cells overexpressing human IR isoform B (IR-B). Data represent the average of 4 independent measurements and error bar represents standard deviations (SD). (b) Insulin tolerance test in rats (n = 4/group) determined by the lowering of blood glucose following subcutaneous injection of 1 U/kg of native insulin (Nat Ins), Ins-1 or Ins-2 analogs. Data represent mean ± SD. (c) CB[7] binding affinity measured by isothermal titration calorimetry (ITC) for insulins in 10 mmol/L sodium phosphate buffer at 27 °C. Mean values from at least three ITC experiments and SD are given in parentheses. ∗K_d > 10⁻³ indicate affinity below ITC detection limit and is treated as 1 mmol/L when calculated for fold changes. (d) Aggregation assay of insulins at 40 °C and 3 mg/mL at pH 3.5 (in 20 mmol/L citrate/10 mmol/L phosphate buffer) and pH 7.4 (in 10 mmol/L PBS) with and without addition of 1 eq. of CB[7]‒PEG_20k. Experiments done in 3–5 replicates and at 100 μL per well with continuous shaking and transmittance measured at 540 nm.

2.3. pH-dependent CB[7] binding of N-terminal acid-modified insulin analogs

Isothermal titration calorimetry (ITC) was used to measure the affinity of CB[7] for the Ins-1 and Ins-2 analogs. CB[7] is known to bind native insulin with low micromolar affinity via inclusion of its B1-Phe. In solution, the B1-Phe is solvent-accessible and unfolds from the rest of the molecule to interact with CB[7]. We observed that CB[7] binds native insulin with similar affinity (K_d ∼10⁻⁷ mol/L) at both acidic and neutral pH (Fig. 2c and Supporting Information Fig. S2); these values agree with reported literature for binding at physiological pH (0.66 μmol/L)^16,24. The specific chemistry of N-terminal guest substituents determines the affinity of binding between peptides and CB[7], with cationic guests having much stronger interactions with CB[7] than neutral or anionic ones^18,25. At both acidic and neutral pH, the N-terminal amine of Phe is protonated and thus binds similarly to CB[7]. However, the insulin analogs Ins-1 and Ins-2 instead bind CB[7] in a pH-dependent manner. At pH 7.4, both Ins-1 and Ins-2 showed no measurable interaction with CB[7] (Fig. 2c and Supporting Information Fig. S3). Yet, at pH 3.5 Ins-1 and Ins-2 bound CB[7] with affinities of ∼0.88 and 0.12 μmol/L, respectively. These values corresponded to an enhancement in binding affinity of >12 and 82-fold for Ins-1 and Ins-2, respectively, compared to their binding at pH 7.4. We previously demonstrated the pH-dependent interaction of N-terminal acid-modified 6-mer peptides with CB[7]¹⁸, and the present study therefore extends the observation of this effect to N-terminal modified therapeutic peptides such as insulin. At neutral pH, the acid groups on Ins-1 and Ins-2 are anionic and therefore subjected to electrostatic repulsion from the carbonyl-fringed CB[7] portal, reducing the affinity of interaction. However, at pH 3.5 these acid groups are protonated, alleviating this electrostatic repulsion and enhancing the affinity of interaction with CB[7]. As previously observed, the added carbon spacer of Ins-2 enables this guest motif to be more deeply buried within the cavity of CB[7], underlying its stronger interaction in binding CB[7] at pH 3.5 than that measured for Ins-1.

Subsequently, the interaction between the various insulin analogs and CB[7] or CB[7]‒PEG_20k was explored; the conjugation of CB[7] to 20 kDa mPEG was synthesized as previously reported to yield CB[7]‒PEG_20k as a formulation additive for study here¹⁷. A competition assay with fluorescent acridine orange (AO), a guest for CB[7] that affords fluorescent enhancement in its bound state, was used to study the interaction between CB[7] or CB[7]‒PEG_20k and the various insulins analogs at pH 7.4 (Supporting Information Fig. S4), as previously reported¹⁴. As expected, insulin bound both CB[7] and CB[7]‒PEG_20k with similar affinity at neutral pH, with K_d values estimated to be 2.2 and 2.4 μmol/L, respectively. Meanwhile, Ins-1 and Ins-2 analogs showed minimal to no binding interaction with either CB[7] or CB[7]‒PEG_20k at pH 7.4. Together, these binding assays demonstrate that the acid-modified insulin analogs have minimal interaction with CB[7]‒PEG_20k at physiological pH, supporting their potential to curtail the SC depot effect and reduce duration of in vivo action compared to native insulin formulated with CB[7]‒PEG_20k.

2.4. pH-dependent aggregation stability of insulin analogs with PEGylated CB[7]

We further evaluated the aggregation time of PEGylated insulin formulations at an equimolar ratio of insulin to CB[7]‒PEG_20k under both acidic and neutral pH. The kinetic profiling of insulin fibrillation is usually performed at 37 °C; however, acid-modified insulin analogs showed prolonged stability at pH 7.4 (34 h for Ins-1 and 54 h for Ins-2) and even longer stability at pH 3.5 (58 h for Ins-1 and 88 h for Ins-2) when tested at 3 mg/mL concentration (Supporting Information Fig. S5). Therefore, the assay temperature was increased to 40 °C to accelerate aggregation and more closely evaluate the effect of pH over a 100-h assay. At pH 7.4, Ins-1 and Ins-2 were stable for 14 and 15 h, respectively when tested alone and 12 and 13 h, respectively when formulated with 1 eq. of CB[7]‒PEG_20k (Fig. 2d and Supporting Information Fig. S6). At pH 3.5, Ins-1 an Ins-2 were stable for 15 and 20 h, respectively, when tested alone but stability increased to 24 and 30 h, respectively, when formulated with 1 eq. of CB[7]‒PEG_20k. By comparison, at these two different pH values native insulin had similar stability when formulated alone at pH 7.4 (6.4 h) as at pH 3.5 (7.2 h). When formulated with 1 eq. of CB[7]‒PEG_20k, the stability of native insulin was extended for ∼11–12 h at both pH (Fig. 2d and Fig. S6). Thus, strong interaction affinity between insulin (B1-Phe) and CB[7]‒PEG_20k is needed to maintain PEG shielding around the insulin and enhance stability by preventing intermolecular aggregation. At both acidic and neutral pH, these strong interactions remain unchanged for native insulin, as observed in Fig. 2c, which explains the extended stability irrespective of pH. There was no measurable interaction between Ins-1 or Ins-2 and CB[7]‒PEG_20k at pH 7.4 (Fig. 2c); this is suspected to arise from deprotonation of the acid groups leading to electrostatic repulsion with the electronegative carbonyl portal of CB[7]. Accordingly, no extension in the aggregation stability of acid-modified insulins was observed at pH 7.4. However, at pH 3.5 the acid groups of both Ins-1 and Ins-2 are protonated to enable stronger interaction with the CB[7]‒PEG_20k, thereby enhancing stability by delaying aggregation for ∼10 h in each case. These results confirm the hypothesis that strong N-terminal interaction between insulin and CB[7]‒PEG_20k at a 1:1 M ratio is necessary for stabilization. As Ins-1 and Ins-2 have limited interaction with CB[7]‒PEG_20k at physiological pH compared to native insulin, this finding suggests that these formulations could benefit from a shorter in vivo duration of action following SC administration.

2.5. In vivo pharmacodynamic modulation of insulin formulations

In this study, 3 eq. of the CB[7]‒PEG_20k excipient successfully stabilized all insulin analogs when tested at 3 mg/mL concentration and 40 °C in 10 mmol/L phosphate buffered saline (PBS) and at physiological pH (Fig. 3a). This stable formulation was therefore used for in vivo studies. The in vivo pharmacokinetics of these stable acid-modified insulins formulations were thus evaluated in diabetic rats in comparison to native insulin formulations. Male rats aged 8–10 weeks were fasted for 6–8 h prior to the experiment and injected SC at 1 IU/kg with native insulin, Ins-1, or Ins-2, in each case administered alone or as a formulation with 3 eq. of CB[7]‒PEG_20k. The insulin formulations were prepared at 10 IU/mL (∼0.06 mmol/L) in PBS at pH 7.4 and the blood glucose level was serially monitored for 6 h. We hypothesized that the limited binding affinity between Ins-1 or Ins-2 and the CB[7]‒PEG_20k excipient at physiological pH would lead to faster dissociation of the complex and more rapid absorption of insulin from the SC space, thus reducing the duration of in vivo action. Such a result could have impact in reducing the risk of hypoglycemic events that may arise from a sustained depot effect. Native insulin, Ins-1, and Ins-2 all showed similar glucose-reducing effects when administered alone (Fig. 3b). In accordance with previous reports, injection of native insulin formulated with CB[7]‒PEG_20k resulted in prolonged in vivo insulin activity, as evidenced by reduced blood glucose levels between 2 and 6 h following administration (Fig. 3b). Though administration of Ins-1 and Ins-2 formulated with CB[7]‒PEG_20k resulted in reduced blood glucose levels compared to the free analogs, the effect was less pronounced than for native insulin formulated with CB[7]‒PEG_20k (Fig. 3b), supporting reduced SC depot retention for these analogs. Moreover, error analysis demonstrates no significant difference between the acid-modified analogs in their free versus CB[7]‒PEG_20k formulated forms, and both cases resulted in a rapid rise in blood glucose within ∼2 h of administration.

Figure 3.

Aggregation assay and in vivo modulation of insulin and N-terminal analogs alone or formulated with CB[7]‒PEG_20k. (a) 100-h aggregation assay of insulins with and without 3 eq. of CB[7]‒PEG_20k at 40 °C, 3 mg/mL at physiological pH (PBS, pH 7.4). Tests done in 3–5 replicates at 100 μL per well with continuous shaking and transmittance measured at 540 nm. (b) In vivo assessment of blood glucose levels in diabetic rats (n = 4–5/group) following subcutaneous administration of insulins (1 IU/kg) injected alone or with 3 eq. of CB[7]‒PEG_20k. Insulins with and without CB[7]‒PEG_20k were prepared at 0.06 mmol/L (10 IU/mL) concentrations and blood glucose was serially monitored for 6 h. (c) Statistical significance (P < 0.05) between the insulins and polymer formulations in lowering blood glucose level at 2.5- and 3.5-h time points were assessed using Student (unpaired) t-test. (d) Blood glucose lowering effect of each formulation calculated based on area under curve (AUC) from 2 to 4 h and 2–6 h and statistical significance determined using Student (unpaired) t-test in GraphPad Prism.

Statistical analyses were performed to determine whether the observed differences in the glucose-lowering effect of free versus CB[7]‒PEG_20k formulated insulins were statistically significant. Direct comparison between the free and PEGylated formulations was assessed using a Student (unpaired) t-test in GraphPad Prism version 9. The glucose-lowering effect between each free insulin and their corresponding PEGylated formulations were compared at each time point. Statistical significance (P < 0.05) was observed when comparing free native insulin to its formulation with CB[7]‒PEG_20k beginning from the 2-h time point onward. No significant difference (P > 0.05) was observed for either of the acid-modified analogs compared to their PEGylated formulations over the duration of the 6-h experiment. A display of the analysis for blood glucose-lowering effect at 2.5- and 3.5-h time points is shown in Fig. 3c. For native insulin, the difference between the two formulations increases over time beginning from the 2-h mark. Moreover, significance was also observed between the PEGylated formulations (native insulin vs acid-modified insulins) at both 2.5 and 3.5 h time points (Fig. 3c) indicating the acid-modified insulin formulations are subject to less SC depot residency. The areas under the curve (AUC) from 2 to 4 h (AUC_{2–4 h}) and 2–6 h (AUC_{2–6 h}) were also calculated to compare the duration of glucose-lowering effect of the different treatments. As a result, differences in blood glucose-lowering duration were statistically significant between native insulin and its PEGylated formulation for AUC_{2–4 h} and AUC_{2–6 h} (P = 0.0003 and P = 0.0001, respectively) (Fig. 3d). On the other hand, no significant difference (P > 0.05) was observed between the acid-modified analogs and their corresponding CB[7]‒PEG_20k formulations at both AUC_{2–4 h} and AUC_{2–6 h}. Likewise, statistical significance (P < 0.05) was noted when comparing AUC between PEGylated native insulin formulation and PEGylated formulations of Ins-1 and Ins-2 (Fig. 3d). These results support reduced SC residency for administration of acid-modified insulins formulated with CB[7]‒PEG_20k leading to a shorter duration of in vivo action.

3. Conclusions

We designed and synthesized two insulin analogs using our previously reported reductive alkylation strategy. The attachment of acid substituents to the N-terminal position on insulin resulted in pH-dependent binding to a supramolecular excipient, CB[7]‒PEG_20k, conferring corresponding pH-dependent formulation stability. These modifications do not have deleterious impact on in vitro and in vivo insulin potency. The N-terminal acidic modifications afforded negligible interactions with CB[7]‒PEG_20k at physiological pH, which should enable rapid complex dissociation and insulin uptake when administered in the SC space. Indeed, administration of stabilized insulin formulations in a rat model of insulin-deficient diabetes demonstrated the N-terminal modified analogs to have a shorter duration of in vivo action, whereas comparable native insulin formulation demonstrated prolonged duration of action arising from more stable complexation with CB[7]‒PEG_20k. N-terminal modified insulin analogs with pH-tunable supramolecular excipient interactions could thus be used to enable enhanced formulation stability without impacting the precision of blood glucose control through excessive SC depot retention.

4. Experimental

4.1. General information

4.1.1. Chemical and protein materials

Native insulin was purchased from Life Technologies. 4-Formylbenzoic acid was purchased form Accela ChemBio Inc and 2-(4-formylphenyl)acetic acid from aablocks LLC. Peptides were synthesized via Fmoc solid-phase peptide synthesis (SPPS). Streptozotocin (STZ), acridine orange (A.O), cucurbit[7]uril (CB[7]), methyl viologen, N,N-diisopropylethylamine (DIEA), triisopropylsilane, l-ascorbic acid, acetic acid (AcOH), iodine, piperidine, methanol (MeOH), urea and dichloromethane (DCM) were purchased from Sigma–Aldrich. Fmoc protected amino acids and 1-[bis(dimethylamino)methylene]-1H-1,2,3-triazolo [4,5-b]pyridinium 3-oxid hexafluorophosphate (HATU) were purchased from Chem-Impex Int'l. Inc. Boc-Ser [FmocThr (tBu)] was purchased from AAPPTec. 2-Chlorotrityl chloride (2-CTC) and Rink amide resins were purchased from ChemPep. Dimethylformamide (DMF), trifluoroacetic acid (TFA), acetonitrile (ACN) and ethyl ether were purchased from Fisher Scientific and used as supplied.

4.1.2. HPLC and LC/MS

All crude peptides were purified with a water/acetonitrile gradient in 0.1% TFA on an Agilent 1260 HPLC system. Fractions collected from HPLC were analyzed by LC/MS on a XBridge C18 5-μm (50 mm × 2.1 mm) column at 0.4 mL/min with a water/acetonitrile gradient in 0.1% formic acid on an Agilent 6120 Quadrupole LC/MS system. Fractions containing targeted product (based on LC/MS) were combined and lyophilized. Detailed LC/MS data of each target product are included in the Supporting Information.

4.1.3. General RP-HPLC conditions

Method A: Individual chains were purified by a Preparative C18 (2) column (Luna®, 5 μm, 250 mm × 21.2 mm) with a linear gradient from 25% aqueous ACN (0.1% TFA) to 45% aqueous ACN (0.1% TFA) over 40 min at a flow rate of 5 mL/min for A chain and from 30% aqueous ACN (0.1% TFA) to 55% aqueous ACN (0.1% TFA) over 45 min at a flow rate of 5 mL/min for B chains.

Method B: All AB dimer peptides were purified using a Preparative C18 (2) column (Luna®, 5 μm, 250 mm × 21.2 mm) with a linear gradient from 30% aqueous ACN (0.1% TFA) to 55% aqueous ACN (0.1% TFA) over 45 min at a flow rate of 5 mL/min.

Method C: All final products were purified by a Preparative C18 (2) column (Luna®, 5 μm, 250 mm × 21.2 mm) with a linear gradient from 25% aqueous ACN (0.1% TFA) to 50% aqueous ACN (0.1% TFA) over 45 min at a flow rate of 5 mL/min.

4.2. Peptide synthesis

Peptides were synthesized via Fmoc SPPS on a peptide synthesizer (SyroI; Biotage, Inc). Automated peptide synthesis was carried out in a 10 mL reactor vial with the following protocols (for 0.1 mmol scale). For Fmoc deprotection: (i) 4.5 mL of 20% piperidine in DMF; (ii) mix 2 × 3 min (new solvent delivered for each mixing cycle). For amino acid coupling: (i) 1.25 mL of 0.4 mol/L Fmoc-protected amino acid in DMF; (ii) 1.25 mL of 0.4 mol/L HATU; (iii) 1.0 mL of 1.0 mol/L DIPEA in DMF; and (iv) mix for 10 min at 70 °C (for cysteine coupling: mix for 15 min at 50 °C). For DMF washing (performed between deprotection and coupling steps): (i) 4.5 mL of DMF; (ii) mix 45 s. Upon completion of the peptide chain, resins were washed with DCM and dried (using vacuum) for 30 min.

4.2.1. Procedure for A-chain

A-chain was synthesized on Rink amide resin using a peptide synthesizer with a standard Fmoc/HATU/DIEA method. The resulting resin-bound A chain (0.1 mmol scale) was treated with 6.0 mL TFA solution containing 2.5% TIS, 2.5% H₂O at rt, with gentle shaking for 2 h. The resin was filtered off and the filtrate was precipitated by cold ether (40 mL). The precipitate was collected by centrifugation then washed with cold ether (40 mL × 3), and vacuum dried and purified by preparative C18 column. 103 mg of A2 (from 0.1 mmol starting resin) was obtained after lyophilization, with a yield of 40%.

4.2.2. Procedure for B chain

B-chain (B1) synthesis was conducted on 2-CTC resin using a standard Fmoc/HATU/DIEA method. The first amino acid was synthesized manually. Cleavage was conducted by treating the resin (0.1 mmol scale) with 6.0 mL TFA solution that contained 2.5% TIS, 2.5% H₂O and 15 eq. of DTDP at rt, with shaking for 2.5 h. The resin was filtered off; and the filtrate was precipitated with cold ether (45 mL). The precipitate was collected by centrifugation, and then washed with cold ether (45 mL × 3). Crude B chain was dissolved in 0.05% TFA containing aqueous acetonitrile (ACN/H₂O: 50/50 v/v, 40 mL), purified on a preparative C18 column and lyophilized to afford 150 mg of B2 (from 0.10 mmol starting resin) with a yield of 42%.

4.2.3. Reductive alkylation in solution phase

B1-Ala insulin (10 mg, 1.74 μmol) was dissolved in 1 mL citric acid buffer (pH 6.1), then NaBH₃CN (5 eq., 8.72 μmol) and 0.5 mol/L of 4-formylbenzoic acid/DMSO solution (2 eq., 7 μL) or 2-(4-formyphenyl)acetic acid/DMSO solution (2 eq., 7 μL) were added into the reaction mixture and stirred at room temperature for 20 h. The reaction solution was diluted with distilled deionized water (5 mL), purified and lyophilized to give 6 mg of B5 and B6 in 59% yield each. N-terminal B-chain modification of the products was confirmed by running a DTT test and a trypsin digestion.

Peptide B2 (40 mg, 11.31 μmol) was dissolved in 5 mL citric acid buffer (pH 6.1), then NaBH₃CN (5 eq., 70 μmol) and 0.5 mol/L of 4-formylbenzoic acid/DMSO solution (2 eq., 57 μL) or 2-(4-formyphenyl)acetic acid/DMSO solution (2 eq., 46 μL) were added into the reaction mixture and stirred at room temperature for 5 h. The reaction solution was diluted with distilled deionized water (20 mL), purified and lyophilized to give 25 mg of B5 and B6 in 60% yield each.

4.2.4. Reductive alkylation on solid support (resin)

The synthesis was adapted from previously described procedure²⁶. Briefly, the linear Acm-protected B-chain resin B1 (0.1 mmol scale) was washed with DMF (3 × 8 mL) and incubated with 10 eq. of 4-formylbenzoic acid or 2-(4-formylphenyl)acetic acid in DMF for 1 h. The resin was drained to filter off the aldehyde, then a freshly prepared 75/25 DCM/MeOH (10 mL) solution was added to the resin followed by addition of excess NaBH₃CN (20–30 eq.). Effervescence occurred after addition of the reducing agent and the reaction was observed for 30 min. Few resins beads were subjected to trinitrobenzotoluylsulfonic acid (TNBS) test (negative) to monitor reaction completion. After which, the resin was drained and washed with MeOH (3 × 8 mL), then DMF (3 × 8 mL) and DCM (3 × 8 mL) and dried. The peptide was cleaved from the resin by treatment with 6.0 mL TFA solution containing 2.5% TIS, 2.5% H₂O and 15 eq. of DTDP at rt, with shaking for 2.5 h. The resin was filtered off and the filtrate was precipitated with cold ether (45 mL). The precipitate was collected by centrifugation, and then washed with cold ether (45 mL × 3). Crude B chain was purified on a preparative C18 column and lyophilized to afford 150 mg (40% yield) of B5 and 155 mg (40% yield) of B6.

4.2.5. Preparation of B1-Ala insulin, Ins-1 and Ins-2 analogs by two-step method

A chain A2 (31 mg) and B chain B2 (40 mg) or B5 (40 mg) or B6 (40 mg) were mixed in 6 mol/L urea, 0.2 mol/L NH₄HCO₃ buffer (pH 7.5, 7 mL). The mixture was left at rt for 15 min. The resulting solution was purified by preparative C18 column with a linear gradient from 30% aqueous ACN (0.1% TFA) to 55% aqueous ACN (0.1% TFA) over 45 min at a flow rate of 5 mL/min. 45 mg of dimer AB1 and 48 mg dimer AB2 and 50 mg of dimer AB3 were obtained in 68%, 70% and 71% yields respectively.

The lyophilized powder each of AB1, AB2, AB3 dimer (30 mg) was dissolved in 33% aqueous acetic acid (6.0 mL) and treated with a freshly prepared solution of iodine (8 eq.) in MeOH (10 mg/mL). The resulting solution was gently agitated at rt for 20 min before the addition of 1 mol/L ascorbic acid until the iodine color (purple) disappeared. The crude was purified by preparative column with a linear gradient from 25% aqueous ACN (0.1% TFA) to 50% aqueous ACN (0.1% TFA) over 45 min at a flow rate of 5 mL/min. Ins-1, Ins-2 and B1-Ala insulin analogs were obtained in 20%–25% yield from their corresponding dimers and in 10%–12% yield over 2 steps based on A2.

4.3. Procedure for binding studies

4.3.1. Acridine orange (AO) competitive binding assay

6 μmol/L of either unmodified CB[7] or the CB[7]‒20KPEG conjugate and 8 μmol/L AO were combined with native insulin, Ins-1 and Ins-2 analogs at different concentrations in H₂O. The highest insulin concentration was measured to be 91 μmol/L and a 1:3 dilution downward was done to obtain ten other diluents. Also, a blank concentration with just the dye and polymer was measured as reference. Samples were incubated for few minutes and fluorescent spectra were collected on a SpectraMax ID5 plate reader (Molecular Devices, Sunnyvale, CA, USA), exciting at 485 nm and collecting the resulting fluorescent spectra from 510 nm. The decay in the peak of AO fluorescent signal was fitted to a one-site competitive binding model (GraphPad Prism, version 9.0), using the CB[7]·AO equilibrium constant reported previously (K_eq = 2 × 10⁵ L/mol)²⁷, to determine binding constants of unmodified CB[7] and CB[7]PEG to insulin and analogs.

4.3.2. ITC titration

Titration experiments were carried out at 27 °C in 10 mmol/L sodium phosphate buffer either at neutral (pH 7.4) or acidic pH (pH 3.5) on a Micro ITC-200 calorimeter from Malvern Panalytical. The concentration of CB[7], accounting for acids and waters of crystallization, was determined by titration with methyl viologen. Sample concentrations were determined by UV–visible spectroscopy in 10 mmol/L sodium phosphate, using molar absorptivities of 6335 L/(mol·cm) at 280 nm for native insulin and Ins-2, 7345 L/(mol·cm) at 280 nm for Ins-1 and 20,400 L/(mol·cm) at 257 nm for methyl viologen. During the titration, insulin and analogs were in the sample cell at a concentration of 0.025–0.035 mmol/L, and CB[7] in the injection syringe at a concentration of 0.275–0.400 mmol/L. The titration schedule consisted of 20 consecutive injections of 2 μL with 200 s interval between injections. Heats of dilution were subtracted from each data set. Detailed ITC data can be found in the Supporting Information. The data were analyzed using Origin software and fitted to a “one-set-of-sites” binding model supplied with the software. This is a standard binary equilibrium model that assumes all sites are independent and non-interacting. Dissociation constants (K_d) were determined from an average of at least 3 titrations.

4.4. Procedure for aggregation assay

4.4.1. Aggregation assay at neutral pH (pH 7.4)

5 mg of insulin and analogs were weighed on an analytical balance and dissolved in 200 μL of NaHCO₃ (5 mg/mL). Then 200 μL of 300 mmol/L NaCl were added, followed by addition of 300 μL PBS (pH 7.4). The pH of the solution was adjusted to 7.4 and the exact insulin concentration determined by measuring absorbance at 280 nm (using ε = 6335 L/(mol·cm) for insulin and Ins-2 and 7345 L/(mol·cm) for Ins-1) with a nanodrop. The final insulin concentration was adjusted to 3 mg/mL with PBS. Insulin and analogs were either measured alone or formulated with 1 eq. and 3eq. of CB[7]‒20KPEG. Samples were plated at 100 μL per well (n = 3–5/group) in a clear 96-well plate (Thermo Scientific Nunc) and sealed with optically clear and thermally stable seal. The plate was immediately placed into an SpectraMax ID5 plate reader (Molecular Devices, Sunnyvale, CA, USA) at 40 °C. Absorbance readings at 540 nm were collected every 10 min with 480 s shaking between reads for 100 h, and absorbance values were subsequently converted to transmittance. Aggregation time is defined as the time at which a 10% reduction in transmittance is observed.

4.4.2. Aggregation assay in acidic pH (pH 3.5)

5 mg of insulin and analogs were weighed on an analytical balance and dissolved in 200 μL of NaHCO₃ (5 mg/mL) followed by addition of 500 μL of buffer (pH 3.5). The pH of the solution was adjusted to 3.5 and the exact insulin concentration determined by measuring absorbance at 280 nm (using ε = 6335 L/(mol·cm) for insulin and Ins-2 and 7345 L/(mol·cm) for Ins-1) with a nanodrop. The final insulin concentration was adjusted to 3 mg/mL with buffer. Insulin and analogs were either measured alone or formulated with 1eq and 3eq. of CB[7]‒20KPEG. Samples were plated at 100 μL per well (n = 3–5/group) in a clear 96-well plate (Thermo Scientific Nunc) and sealed with optically clear and thermally stable seal. The plate was immediately placed into an SpectraMax ID5 plate reader (Molecular Devices) at 40 °C. Absorbance readings at 540 nm were collected every 10 min with 480 s shaking between reads for 100 h, and absorbance values were subsequently converted to transmittance.

4.5. Procedure for activity assay

4.5.1. Phospho-AKT (Ser473) cell-based assay

pAKT Ser473 levels were measured in a mouse fibroblast cell line, NIH 3T3, overexpressing human insulin receptor isoform B (IR-B). The cell line was cultured in DMEM (Sigma–Aldrich) with 10% fetal bovine serum (Gibco), 100 U/mL penicillin streptomycin (Thermo Fisher Scientific) and 2 mg/mL puromycin (Thermo Fisher Scientific). For each assay, 40,000 cells per well and 100 μL per well, were plated in 96-well plates with culture media containing 1% FBS. 20 h later, 50 μL of acid-modified insulin or native insulin was pipetted into each well after the removal of the original media. After a 30-min treatment, the insulin solution was removed and the HTRF pAKT Ser473 kit (Cisbio, MA, USA) was used to measure the intracellular level of pAKT Ser473. Briefly, the cells were first treated with cell lysis buffer (50 μL per well) for 1 h under mild shaking. 16 μL of cell lysate was then added to 4 μL of detecting reagent in a white 384-well plate. After 4 h incubation, the plate was read in a SpectraMax ID5 plate reader (Molecular Devices) and the data processed according to the manufacturer's protocol.

4.5.2. Insulin tolerance test (ITT)

ITT was performed on normal chow fed, 8–10 weeks old rats (n = 4/group). On the day of experiment, rats were fasted for 6–8 h in the morning, with the food removed and new bedding in the cages. The food was withdrawn for the entire experimental duration. The body weights and basal blood glucose concentrations (using a handheld Bayer Contour Next glucose monitor (Bayer) were measured. Following the basal measurements, rats received subcutaneous injections of 1 IU/kg body weight of either Nat Ins or Ins-1 or Ins-2 analogs dissolved in PBS. The blood glucose concentrations were measured up to 4 h after injection after injection.

4.6. Procedure for diabetes induction in rats

Male Sprague–Dawley rats (Charles River) were used for experiments. Animal studies were performed in accordance with the guidelines for the care and use of laboratory animals; all protocols were approved by the Stanford Institutional Animal Care and Use Committee (protocol #33909). The procedure used for diabetes induction using streptozotocin (STZ) was adapted from previous protocols²⁸. Briefly, male Sprague–Dawley rats 180–230 g (8–10 weeks) were weighed and fasted in the morning 6–8 h prior to treatment with STZ. STZ was protected from light and diluted to 10 mg/mL in 50 mmol/L sodium citrate buffer (pH 4.5) immediately before injection. The STZ solution was injected intraperitoneally at 65 mg/kg into each rat. Rats were provided with water containing 10% sucrose for 24 h after injection with STZ and were given subcutaneous saline injections daily to prevent dehydration. Rat blood glucose levels were tested for hyperglycemia daily after the STZ treatment via blood collection from a tail vein using a handheld Bayer Contour Next glucose monitor (Bayer). Diabetes was defined as having three consecutive blood glucose measurements >400 mg/dL in non-fasted rats.

4.7. Insulin pharmacodynamic test in rats

Diabetic rats were fasted for 6–8 h in the morning, with the food removed and new bedding in the cages. Rats were bled at the beginning of the study and their fasting blood glucose level measured. Rats were divided into 4–5 groups (n = 4–5/groups) and injected subcutaneously with Nat Ins, Ins-1 or Ins-2 analogs dosed at 1 IU/kg either alone or formulated with 3 molar equivalents of CB[7]-20KPEG conjugate. In all cases, insulin and analogs, with or without CB[7]-20KPEG were prepared at a concentration of 10 IU/mL (∼0.06 mmol/L) in PBS (pH 7.4). Blood glucose readings were collected every 10 min for the first hour and then every 30 min for the next 5 h using a handheld Bayer Contour Next glucose monitor (Bayer).

4.8. Statistical analysis

All data are reported as mean ± standard error mean (SEM) unless specified. All results are expressed as the mean ± standard deviation (SD). Comparisons between two formulation groups were conducted using a two-tailed Student's t-test (unpaired) in GraphPad Prism 9. Statistical significance was considered at P < 0.05.

Acknowledgments

This work is supported by NIDDK (DK120430, DK121336, USA) to Danny Hung-Chieh Chou and JDRF (5-CDA-2020-947-A-N, USA) to Matthew J. Webber. The authors thank the Stanford Veterinary Service Centre staff for their technical assistance.

Author contributions

Rolande Meudom, Conceptualization, Methodology, Investigation, Writing—original draft; Yanxian Zhang, Investigation, review & editing; Michael A. VandenBerg, Investigation; Lei Zou, Investigation; Yi Zhang, investigation; Matthew J Webber, Resources, Writing—review & editing. Danny Hung-Chieh Chou, Writing—review & editing, Supervision, Funding acquisition.

Conflicts of interest

The authors declare no competing interest.

Appendix A. Supplementary data

The following is the Supplementary data to this article.