Monomeric Insulin Analog and Stabilizing Excipient Enabled Ultrafast Insulin Formulation

ABSTRACT

The speed of insulin therapy remains fundamentally constrained by the self‐association of insulin into hexamers. Here, a materials‐based strategy is introduced to stabilize HALQ, a monomeric insulin analog, using a non‐interacting inulin‐derived excipient (BN‐Inu). BN‐Inu markedly mitigates aggregation and maintains HALQ stability for 96 h under stress and for at least 30 days at room temperature. In a porcine model of diabetes, monomeric HALQ exhibits significantly accelerated absorption and a shorter duration of action than the ultrarapid insulin aspart Fiasp. This “fast‐on, fast‐off” profile is consistent with faster clearance from the subcutaneous depot, more closely synchronizes with endogenous prandial insulin physiology. Addition of clinically used absorption enhancers further accelerates its pharmacokinetic profile, producing a faster time‐to‐peak and reduced exposure relative to ultrarapid insulin lispro Lyumjev in this animal model. Furthermore, translation to human physiology was evaluated through pharmacokinetic modeling, which predicts that HALQ could reduce time‐to‐peak from 60 to 39 min and shorten duration of action from 143 to 84 min in humans. These simulations suggest the potential utility of achieving a step‐change in the speed of insulin therapy. These findings demonstrate that monomer‐stabilizing excipients enable next‐generation ultrafast insulin formulations with the potential to improve glycemic control in diabetes.

Monomer‐stabilizing excipients are introduced to overcome the fundamental speed limit of insulin therapy caused by hexamer self‐association. By utilizing a non‐interacting inulin‐derived excipient (BN‐Inu), the monomeric insulin analog HALQ is stabilized, enabling a “fast‐on, fast‐off” pharmacokinetic profile. This materials‐based strategy realizes the ultrafast monomeric insulin formulation, offering a step‐change in physiological synchronization for potentially improved glycemic control in diabetes.

1. Introduction

Insulin therapy, the primary treatment for insulin‐requiring diabetes, relies heavily on precise injection management that is difficult to achieve consistently, resulting in frequent deviations from correct glycemic control. Ideally, exogenous insulin requires fast‐onset and fast‐offset kinetics to mimic endogenous insulin secretion, which responds rapidly to glucose spikes during meals. Rapid absorption and clearance mimic endogenous postprandial insulin secretion is clinically documented to improve glycemic control and reduce the risk of late postprandial hypoglycemia [1, 2]. However, even current ultrarapid‐acting insulins (e.g., Lyumjev, Fiasp) exhibit delayed onsets of action (20–30 min) and relatively long effect last for 5–7 h [3] (Figure 1A). This kinetic mismatch often results in a period of uncontrolled hyperglycemia during mealtime excursions, followed by an increased risk of postprandial hypoglycemia. Consequently, insulin therapy creates a considerable burden for people with diabetes requiring precise matching with meals, and the possibility of both hypoglycemia and hyperglycemia when there is a mismatch. A recent clinical data review underscores the severity of this issue, showing that only 28% of kids and 44% of adults met all glucose targets (>70% TIR, <4% time <70 mg/dL) despite using the most aggressive glucose target (110 mg/dL) with a modern automated insulin dosing system [4].

FIGURE 1.

Monomeric insulin analog HALQ design with potential of translating into ultrafast‐acting insulin therapeutics. (A) Conceptual overview comparing endogenous insulin dynamics with the pharmacokinetic profiles of current subcutaneous insulin formulations, highlighting the gap that limits tight glycemic control. (B) Mechanism of insulin dissociation and absorption from the subcutaneous depot. Slow hexamer‐to‐monomer conversion underlies delayed onset and prolonged action; a fully monomeric analog is expected to enable ultrarapid absorption. (C) Primary sequences of HALQ, native human insulin, and the parent venom‐derived insulin. (D) HALQ exhibits minimal self‐association relative to native insulin, consistent with its behavior as a true monomeric analog. The stable monomeric state of HALQ in solution was confirmed by sedimentation velocity analytical ultracentrifugation, as shown in Figure S1.

These pharmacokinetic limitations are thought to arise from the subcutaneous depot effect, in which insulin must dissociate from its associated states after injection before it can be absorbed into the bloodstream circulation (Figure 1B). To mitigate this issue, fast‐acting insulin analogs such as insulin aspart and insulin lispro were developed to reduce self‐association through amino acid substitutions. For example, native human insulin has an association constant of 1.4 × 10⁻⁵ m for dimer formation [5, 6], whereas insulin lispro exhibits a weaker association constant in the 10⁻⁴ m range [7], leading to a faster onset of action compared with native insulin. However, due to stability concerns, all these insulin analogs are still formulated as hexamers in vials and subsequently form dimers within the subcutaneous depot. As a result, the rate‐limiting step for subcutaneous insulin absorption remains the dissociation into monomeric insulin, contributing to their prolonged duration of action. Thus, fully monomeric insulin formulations are expected to bypass the subcutaneous depot effect, enabling accelerated onset and offset kinetics and potentially yielding clinical improvements in postprandial glucose control. Yet, the development of bioactive monomeric insulin has been hindered by a fundamental challenge that insulin dimerization is mediated by aromatic residues at positions B24‐B26, which are also essential for insulin receptor binding [8]. Consequently, straightforward mutations that disrupt dimerization typically lead to dramatically reduced bioactivity. Moreover, even if a monomeric analog is successfully engineered, such analogs tend to exhibit heightened propensities for aggregation, presenting additional formulation challenges [9].

The discovery of insulin‐like molecules from cone snail venoms [10] inspired the recent design of monomeric human insulin analogs, by incorporating structural features of the venom insulin (Con‐Ins K1) from C. kinoshitai [11] (Figure 1C). In particular, the analog Vh‐Ins‐HALQ contains a four‐amino acid (HALQ) C‐terminal elongation on the A‐chain, which enables an alternative mechanism of receptor activation with full potency in the absence of the human B‐chain C‐terminus. Because this analog lacks the B‐chain aromatic residues that mediate insulin dimerization, it represents a promising therapeutic candidate for a new generation of ultrafast‐acting prandial insulins (Figure 1D). However, a significant gap remains between engineering monomeric insulin analogs and translating them into viable monomeric insulin therapeutics due to their inherent low stability. Monomeric insulin is more prone to unfolding and misfolding, rapidly aggregating into amyloid species. Consequently, current insulin formulations rely on zinc or phenolic preservatives to stabilize insulin in hexameric or multi‐hexameric states, thereby increasing the energetic barrier for unfolding and preventing aggregation [12].

Alternative stabilization strategies have been developed to preserve insulin in formulation and maintain it in a monomeric state, but each presents significant challenges. Covalent [13], non‐covalent [14], and pH‐dependent reversible non‐covalent [15] PEGylation strategies have been explored to stabilize insulin monomers, but all result in prolonged pharmacological performance, which is undesirable for the development of fast‐acting insulins. Nonionic surfactants such as poloxamer 171 have been used in the Insuman U400 formulation (Sanofi‐Aventis) to enhance insulin stability, yet the insulin remains in a hexameric state in the presence of zinc. Similarly, polysorbate 20 is used in Apidra (Sanofi Aventis) to stabilize insulin glulisine, but the formulation still contains compact hexamers even in the absence of zinc [16]. Recent efforts using amphiphilic polyacrylamide‐based copolymers have shown effectiveness in stabilizing zinc‐free insulin lispro by mitigating interfacial aggregation [17]. However, polyacrylamide production raises safety concerns, including the need for stringent procedures to minimize toxic residual acrylamide, as well as uncertainties regarding degradation products that may contribute to toxicity [18]. Given these limitations, a fundamentally new approach is required to stabilize the monomeric insulin analog HALQ while preserving its optimal pharmacological performance.

In this study, we examined the key factors that compromise insulin stability and translated these insights into the rational design of excipients that maximize stabilization. Protein aggregation typically occurs when the molecule undergoes partial unfolding and exposes hydrophobic or aggregation‐prone regions, promoting strong protein–protein interactions that drive aggregate formation [19]. It is well established that surface hydration is critical for maintaining a protein's native folded structure, whereas hydrophilic–hydrophobic interfaces (e.g., the air–water interface) can promote protein adsorption, expose hydrophobic patches, and trigger aggregation cascades [20]. To address these two major instability factors, excipients should be designed to (1) preserve or enhance water‐protein interactions while reducing protein–protein contacts, and (2) prevent protein–surface interactions. Sugars are known to strengthen water‐protein interactions [21, 22, 23], whereas nonionic surfactants effectively suppress interface‐driven aggregation [24, 25]. Building on this understanding, a sugar‐based nonionic surfactant was designed with the intent to stabilize HALQ via these dual pathways, offering a strategic approach to optimize its stability. Here, we show that appropriately modifying the “Generally Recognized as Safe” (GRAS) polysaccharide inulin yields an excipient that stabilizes the monomeric insulin analog HALQ without direct interaction. Notably, the resulting HALQ formulation exhibited a shorter duration of action than commercial fast‐acting insulin formulations in a pig model.

2. Results and Discussion

2.1. Monomeric Insulin Analog HALQ and the Instability Challenge

Structural studies have demonstrated that the A‐chain C‐terminus can compensate for the loss of receptor contacts typically contributed by the insulin B‐chain C‐terminal octapeptide [11]. A representative example is the human–venom insulin hybrid analog HALQ, which lacks the B‐chain C‐terminal residues B22‐B28 of human insulin (hIns) that mediate dimerization, yet contains a C‐terminal extension on the A‐chain (HALQ, A21‐A24) and displays bioactivity comparable to hIns (Figure 2A). However, evaluating the in vivo activity of such promising monomeric insulin analogs is challenging due to their inherent stability limitations. After 7 days of stress‐aging (37°C, continuous shaking at 150 rpm), HALQ exhibited a dramatic loss of bioactivity compared with hIns, as evidenced by an approximately twofold increase in the EC₅₀ for pAkt relative to hIns (Figure 2A). Under stress‐aging conditions, hIns aggregates after approximately 4 h, whereas HALQ aggregates much more rapidly, with onset occurring at around 1 h (Figure 2B). In addition to reduced physical stability, monomeric HALQ is also more susceptible to chemical degradation than hIns (Figure 2C,D). The pronounced instability of monomeric HALQ relative to hIns arises from its bypassing of the pre‐aggregation pathway, which involves transitions between association states, where hexamer dissociation is the rate‐limiting step, followed by conformational rearrangements [26]. Skipping these steps leads directly to the formation of an expanded monomer that initiates the downstream aggregation cascade (Figure 3A). Therefore, mitigating unfolding and aggregation depends on maintaining the conformational stability of monomeric HALQ in solution, minimizing protein–protein interactions, and preventing exposure to hydrophobic interfaces.

FIGURE 2.

Monomeric HALQ shows reduced stability relative to native human insulin. (A) Cellular activity of fresh and aged HALQ compared with native human insulin, measured by pAKT signaling in NIH‐3T3 cells expressing human IR‐B. Data points (n = 3, mean ± s.d.) were fitted using a three‐parameter logistic model to determine the EC₅₀. The solid line represents the best‐fit curve, and the calculated EC₅₀ is listed in the table with a 95% confidence interval (CI). (B) Representative time‐dependent absorbance traces at 540 nm demonstrating accelerated aggregation of HALQ during stressed aging (continuous shaking, 37°C) (n = 3, mean ± s.d.). (C,D) Chemical degradation profiles of native insulin (C) and HALQ (D) after 3 days of stressed aging, shown by LC chromatograms (280 nm) and corresponding MS spectra.

FIGURE 3.

Design and optimization of the inulin‐based stabilizing excipient. (A) Conceptual illustration of the reduced stability of monomeric insulin compared with hexameric native insulin, highlighting aggregation driven by protein unfolding and aggregation (left) and by adsorption and assembly at the air–liquid interface (right). (B) Chemical structure of the modified inulin polymer eb‐Inu. Yellow schematically highlights the hydrophobic functional groups introduced in this study to obtain the amphiphilic inulin derivative eb‐Inu. (C) Representative absorbance at 540 nm during stressed aging (37°C, continuous shaking), demonstrating improved HALQ stability with 1 wt.% eb‐Inu (n = 3, mean ± s.d.). (D) Blood‐glucose profiles in diabetic rats showing delayed pharmacodynamic response of eb‐Inu–stabilized HALQ compared with unformulated HALQ (n = 6, mean ± s.d.). Statistical significance was determined by a restricted maximum likelihood repeated measures mixed model, with post‐hoc Bonferroni test, ^* p <0.05, ^** p <0.01. Statistical comparisons not shown in the figure were non‐significant (p > 0.05). (E) ¹H NMR spectra of HALQ with increasing eb‐Inu concentrations showing direct interaction, evidenced by proton chemical‐shift changes. (F) ¹H NMR spectra of HALQ in the presence or absence of BN‐Inu show no proton chemical‐shift changes, indicating loss of direct interaction upon butyl‐nitrile side‐chain modification of inulin.

2.2. Monomeric Insulin Analog HALQ‐Stabilizing Excipient Design

Given the challenges around the stability of HALQ, excipients capable of targeting multiple aggregation pathways are expected to provide the greatest enhancement of monomeric insulin HALQ stability in solutions. Sugars such as disaccharide trehalose are anticipated to stabilize proteins by being preferentially or weakly excluded from the protein surface, thereby forcing the protein to remain preferentially hydrated, reducing protein–protein interactions and aggregation [27, 28]. Such molecules are widely used in freeze‐dried formulations and cryopreservation [29]. Hydrophobic modification of a sugar backbone can further impart surface activity, possibly providing an additional mechanism to prevent protein assembly at interfaces. Based on this rationale, inulin was selected as the sugar scaffold for surface activity modification for several reasons. Inulin is a naturally occurring polysaccharide abundant in plants such as chicory root, is classified as safe as a dietary fiber ingredient [30]. It passes through the digestive tract undigested without raising blood glucose levels and has been associated with beneficial effects on lipid metabolism, weight regulation, and glycemic control [31]. Its structure comprises a terminal glucosyl moiety and repeating fructosyl units, and it presents multiple saccharide sites that can contribute to protein stabilization. Inulin is a small, inert, electrostatically neutral polysaccharide with a molecular weight range of 500–3600 Da. It is freely filtered by the glomeruli and neither reabsorbed nor secreted by the renal tubules, allowing for efficient renal clearance and making it suitable for chronic exposure during daily insulin injections without significant safety concerns [32]. However, like many polysaccharides, inulin has low water solubility due to strong intermolecular hydrogen bonding between polymer chains, which is energetically favored over interactions with water [33]. Side chain modification of inulin can improve solubility by purposely disrupting the rigid intermolecular hydrogen bonding, thereby increasing solubility by improving the exposure of remaining hydroxyl groups to water [34]. Moreover, such modifications can enhance surface activity, enabling the excipient to displace protein from the air–water interface and prevent interface‐driven aggregation.

Accordingly, inulin‐based stabilizing excipients were designed via hydrophobic side‐chain modification. Direct substitution of hydroxyl groups with alkyl chains is a common strategy for introducing hydrophobicity and generating amphiphilic molecules with surface activity [35]. Substitution of inulin hydroxyl groups with a branched six‐carbon alkyl group (2‐ethylbutyl) (eb‐Inu, Figure 3B and Figure S2) yields a highly amphiphilic derivative capable of stabilizing HALQ against aggregation (Figure 3C). However, in vivo evaluation of eb‐Inu‐stabilized HALQ in diabetic rats revealed a delayed pharmacodynamic response compared with unformulated HALQ (Figure 3D). We hypothesized that this attenuated pharmacodynamic response resulted from undesirable interactions between HALQ and the stabilizing excipient eb‐Inu.

¹H NMR analysis revealed chemical shift in the HALQ spectra within the 6.5–7.5 ppm region upon addition of increasing concentrations of eb‐Inu (Figure 3E). This spectral region corresponds to aromatic protons, suggesting weak hydrogen‐bonding or hydrophobic interactions between eb‐Inu and exposed aromatic residues in HALQ. Notably, this interaction was abolished when the C‐terminal histidine of HALQ was mutated to asparagine (NALQ, Figure S3), confirming that the interaction originates from the imidazole ring of the C‐terminal histidine, likely involving contacts with terminal ─CH groups of eb‐Inu.

These findings highlight the importance of carefully selecting side‐chain moieties for excipient design, as direct interactions between the hydrophobic substituents of stabilizing excipients and HALQ can adversely impact in vivo pharmacological performance. To mitigate these undesired interactions, we introduced a butyl nitrile moiety terminating in a nitrile group. The butyl nitrile group confers moderate hydrophobicity and interacts less favorably with histidine side chains due to weaker non‐specific hydrophobic interactions compared with non‐polar alkyl groups [36]. Consistent with this design rationale, ¹H NMR spectra showed that butyl nitrile–modified inulin (BN‐Inu) eliminated detectable proton chemical shift perturbations, confirming minimized interaction with HALQ (Figure 3F). In addition, nitrile groups are metabolically stable and non‐toxic, typically remaining unmodified during clearance from the human body [37].

2.3. HALQ Formulation Stability

Like eb‐Inu, substitution of inulin hydroxyl groups with a butyl nitrile moiety (BN‐Inu) also effectively inhibits HALQ aggregation, and BN‐Inu variants with increasing degrees of substitution (DS) exhibit progressively enhanced stabilization of HALQ (Figure S4). A DS of 0.2 was sufficient to fully suppress HALQ aggregation and was therefore selected for formulation and subsequent characterization (Figure 4A). Beyond that, the inulin‐derived excipient was designed based on general protein aggregation mechanisms, with the aim of providing broad stabilization effects. Consistent with this rationale, preliminary aggregation studies suggest that BN‐Inu may also improve the stability of native insulin as well as rapid‐acting insulin analogs, including aspart and lispro (Figure S5), supporting its potential applicability beyond HALQ. In vivo evaluation of BN‐Inu‐stabilized HALQ in rats with diabetes demonstrated a pharmacodynamic response comparable to that of unformulated HALQ (Figure 4B), representing a significant improvement over eb‐Inu–stabilized HALQ.

FIGURE 4.

BN‐Inu stabilizes monomeric HALQ during stressed aging and ambient storage. (A) BN‐Inu (1 wt.%) suppresses HALQ aggregation during 96 h of continuous stressed aging, shown by representative time‐dependent absorbance traces at 540 nm (n = 3, mean ± s.d.). (B) Blood‐glucose profiles in diabetic rats showing comparable pharmacodynamic response of BN‐Inu–stabilized HALQ compared with unformulated HALQ (n = 6, mean ± s.d.). Statistical significance was determined by a restricted maximum likelihood repeated measures mixed model, which showed no significant effect (p > 0.05). (C) Surface activity of BN‐Inu, shown by concentration‐dependent reductions in surface tension in PBS. Surface tension was measured using the pendant‐drop method and calculated by fitting the drop contour to the Young–Laplace equation (n = 4, mean ± s.d.). (D) BN‐Inu displaces HALQ from the air–liquid interface, as indicated by decreased surface tension of a 100 U/mL HALQ formulation containing 1 wt.% BN‐Inu (n = 4, mean ± s.d.). (E) BN‐Inu reduces attractive intermolecular interactions between HALQ, reflected by an increase in the second virial coefficient (B₂₂, n = 6, mean ± s.d.). B₂₂ was determined by static light scattering, with the slope of the scattering response in the low‐concentration region corresponding to B₂₂ (shown in Figure S6). (F, G) Long‐term stability of HALQ with and without BN‐Inu, assessed by LC chromatograms (280 nm) and MS spectra after 30 days of ambient storage. Formulation with 1 wt.% BN‐Inu preserves HALQ integrity. (H) pAKT signaling in NIH‐3T3 cells expressing human IR‐B demonstrates preserved biological activity of HALQ after 30 days of aging when formulated with BN‐Inu, comparable to fresh HALQ and native insulin. Data points (n = 3, mean ± s.d.) were fitted using a three‐parameter logistic model to determine the EC₅₀. The solid line represents the best‐fit curve, and the calculated EC₅₀ is listed in the table with a 95% confidence interval (CI).

Consistent with our design, BN‐Inu exhibits surface activity, as evidenced by a concentration‐dependent reduction in the surface tension of PBS solution. Surface tension reflects the cohesive intermolecular forces within a liquid [38]. The addition of <0.01 wt.% BN‐Inu reduced the surface tension of PBS from 65.7 to 53.1 mN/m, suggesting that BN‐Inu molecules preferentially diffuse to the air–liquid interface, insert between water molecules, and decrease intermolecular attraction (Figure 4C). As BN‐Inu concentration increased, surface tension reached a plateau near 0.1 wt.%, a characteristic signature of surfactant behavior associated with the critical micelle concentration [39], at which the interface becomes saturated, and additional BN‐Inu molecules assemble into micelles in the bulk solution. Beyond buffer alone, BN‐Inu similarly reduced the surface tension of HALQ solutions. HALQ at 100 U/mL in PBS lowered the surface tension from 67.5 ± 1.0 to 45.8 ± 1.7 mN/m, consistent with HALQ adsorption at the interface. The addition of 1 wt.% BN‐Inu further decreased the surface tension to 29.2 ± 3.9 mN/m (Figure 4D), a value comparable to 1 wt.% BN‐Inu in PBS alone (Figure 4C), suggesting that BN‐Inu could effectively outcompete or displace HALQ from the interface. In addition to its surface activity, BN‐Inu was evaluated for its capacity to alter protein–protein attractive interactions within HALQ, as quantified by the second virial coefficient (B₂₂). B₂₂ is a measure of protein interactions in dilute solutions [40]. Negative values reflect net attraction and increased aggregation propensity, whereas positive values indicate repulsion [41, 42]. Increasing BN‐Inu concentrations shifted B₂₂ toward more positive values (Figure 4E), corresponding to stronger repulsive interactions and enhanced HALQ colloidal stability. As BN‐Inu is a neutral polymer and lacks specific binding to HALQ, the increased repulsion may primarily be attributed to entropic mechanisms such as excluded volume effects and preferential hydration [43, 44, 45]. Collectively, while these macroscopic biophysical data do not provide direct atomistic tracking of the interface, they strongly support a proposed mechanism wherein the excipient reduces protein self‐association in solution and attenuates interfacial adsorption, offering a cohesive explanation for the improved monomeric insulin stability.

The availability of BN‐Inu provides an efficient strategy for formulating HALQ with improved long‐term stability. In contrast to the extensive chemical degradation observed for HALQ alone (100 U/mL, ∼3.15 mg/mL) in PBS after 30 days of storage at room temperature (Figure 4F), co‐formulation with 1 wt.% BN‐Inu effectively preserved HALQ molecular integrity under identical conditions (Figure 4G). The resulting stability at ambient temperature highlights the potential of this formulation to reduce cold‐chain dependence, lower logistical costs, and expand access to insulin therapeutics.

To further evaluate the association state of HALQ in the BN‐Inu formulation, we assessed self‐association using sedimentation velocity analytical ultracentrifugation (SV‐AUC, Figure S1). As a control, we characterized the association state of the fast‐acting insulin analog Fiasp (insulin aspart). Continuous c(s) analysis showed that at 100 U/mL, insulin aspart is predominantly hexameric in Fiasp, exhibiting a sedimentation coefficient close to 3s, with a small population of lower‐association states (dimers and monomers) [46]. In contrast, HALQ in PBS is almost entirely monomeric, with a predominant peak around 1s. The sedimentation coefficient distribution of HALQ formulated with BN‐Inu remained within the monomeric range but resolved into two closely spaced peaks at approximately 0.8s and 1.2s. These values are also characteristic of monomeric insulin species [46] and suggest subtle conformational heterogeneity within the formulation. The slower‐sediment species at 0.8s likely correspond to more highly hydrated monomers [47], whereas the peak at 1.2s may represent a more compact state [48].

Although HALQ remains monomeric in the BN‐Inu formulation, it exhibits markedly enhanced stability. Under 96 h of continuous stress‐aging conditions, increasing concentrations of BN‐Inu progressively delayed HALQ aggregation and reduced the overall magnitude of aggregate formation. At 1 wt.% BN‐Inu, HALQ aggregation was completely suppressed (Figure 4A). This stability exceeds that of the fast‐acting insulin analogs Fiasp (insulin aspart) and Lyumjev (insulin lispro), both of which exhibited aggregation within 3 h under the same stress conditions (Figure S7). Owing to this exceptional formulation stability, HALQ retained its full bioactivity even after 30 days of storage at ambient temperature (Figure 4H). While the present study demonstrates the feasibility and functional performance of the excipient within the proposed system, comprehensive pharmaceutical characterization, such as batch‐to‐batch consistency, long‐term stability, and formulation‐relevant properties, remains to be systematically investigated in future work.

2.4. Biocompatibility of HALQ Formulation

As a new chemical entity, albeit derived from the GRAS polymer inulin with a mild butyl nitrile modification, the initial, sub‐chronic safety profile of BN‐Inu was assessed in the context of the HALQ formulation. Commercial human insulin Humulin R U‐100 served as a control. Three groups of STZ‐induced mice with diabetes (equal numbers of males and females, n = 6 per group) received a single daily subcutaneous dose of 1U/kg Humulin R U‐100, HALQ in PBS, or HALQ formulated with BN‐Inu for 7 consecutive weeks. Blood samples were collected in week 6 for complete blood cell counts and in week 7 for blood serum chemistry and immunological analyses (Figure 5A). Whole blood cell count was used to identify early hematological indicators of toxicity (Figure 5B). Blood serum chemistry panel measurements, including enzymes and proteins like aspartate transaminase (AST), alanine transaminase (ALT), alkaline phosphatase (ALP), and albumin (ALB), were examined as markers of liver function. Blood urea nitrogen (BUN) and creatinine (Creat) were measured to assess kidney function (Figure 5C). Cytokines, including IFNα, TNFα, IL‐6, IL‐10, IL‐12, and IL‐17, were quantified to evaluate innate and regulatory immune responses and to screen for potential acute immunotoxicity or inflammatory reactions (Figure 5D). Across all hematological, biochemical, and immunological parameters, no significant differences were detected among the three treatment groups based on two‐way ANOVA with multiple comparisons. These results suggest that both HALQ and HALQ formulated with BN‐Inu are well tolerated within the scope of this 7‐week repeated exposure model.

FIGURE 5.

BN‐Inu–formulated HALQ is biocompatible and well‐tolerated in vivo. (A) Schematic of the in vivo biocompatibility study comparing Humulin R, HALQ, and HALQ formulated with BN‐Inu in mice with diabetes. (B) Complete blood counts show no treatment‐related abnormalities across groups (WBC, white blood cell; RBC, red blood cell; PLT, platelet; Neu, neutrophil; Lym, lymphocyte; Mon, monocyte; Eos, eosinophil). (C) Serum chemistry panels indicate no differences in liver or kidney function (AST, aspartate transaminase; ALT, alanine transaminase; ALP, alkaline phosphatase; BUN, blood urea nitrogen; Creat, creatinine; ALB, albumin). (D) Cytokine measurements show no evidence of immunotoxicity or inflammatory responses in HALQ or BN‐Inu–formulated HALQ relative to Humulin R (n = 6, mean ± s.d.). Statistical analysis was performed using one‐way ANOVA. Statistical comparisons not shown in the figure were non‐significant (p > 0.05).

2.5. Pharmacokinetics and Pharmacodynamics of Monomeric HALQ Formulation

Because HALQ is a novel insulin analog, specific antibody pairs and an appropriate immunoassay were required to enable its detection and quantification for pharmacokinetic (PK) analysis. An anti‐HALQ antibody was generated by immunizing mice with HALQ‐KLH conjugate, followed by hybridoma production using fused spleen cells with SP2/0 myeloma cells (Figure S8A). A matched antibody pair, anti‐HALQ clone 34B7D4 and anti‐human insulin clone AE9D6, was identified for sensitive detection of HALQ in serum using a sandwich ELISA, without cross‐activity with native human insulin or swine insulin (Figure S8B). A representative standard curve confirmed the suitability of the ELISA immunoassay for quantifying HALQ concentrations in serum samples (Figure S8C).

The pharmacodynamics (PD) and PK of HALQ formulated with BN‐Inu were compared with those of the fast‐acting insulin analog Fiasp (insulin aspart) in a randomized crossover study in pigs with diabetes. Due to individual differences in insulin sensitivity, fasted diabetic pigs received 0.1–0.2 U/kg of HALQ (100 U/mL with 1 wt.% BN‐Inu) or Fiasp to lower blood glucose (BG) to approximately 100 mg/dl. Normalized BG profiles showed similar glucose‐lowering kinetics between the two treatments, whereas BG recovery was noticeably faster in the HALQ group (Figure 6A). Normalized serum profiles were used to compare kinetic differences between the two treatment groups (Figure 6B). To quantify PD differences, parameters were calculated, including time to reach 50% of the maximum BG‐lowering effect during BG decline (Figure 6C), time to reach maximum BG lowering effect (i.e., lowest BG, normalized as −100%) (Figure 6D), time to reach 50% of maximum effect during BG recovery (Figure 6E), and duration of BG levels below 50% of maximum lowering (duration of action, Figure 6F). In these animals, HALQ produced significantly faster BG recovery (Figure 6D) and a shorter duration of action (Figure 6F) compared with Fiasp. Serum concentration of HALQ and insulin aspart was measured over time to assess PK. Overall exposure, assessed by the area under the curve during the first 100 min (AUC₁₀₀), did not differ between groups (Figure S9A–C). Fiasp, formulated with niacinamide, is designed to exhibit accelerated absorption due to promotion of monomer formation and local vasodilation that increases blood flow [49]. Accordingly, the onset of action (time to 50% of normalized peak concentration, and time to peak concentration) was similar between HALQ and Fiasp (Figure 6G,H). However, as the vasodilatory effect of niacinamide is transient [49], and the monomer fraction in Fiasp is substantially lower than in HALQ, as confirmed by SV‐AUC (Figure S1), the potentially superior absorption of monomeric HALQ appears more evident in its faster clearance from circulation (Figure 6I). Consequently, within this porcine model, the duration of action, defined as the time during which serum insulin remains above 50% of peak concentration, was 1.5‐fold shorter for HALQ (43 ± 9 min) compared with Fiasp (62 ± 15 min) (Figure 6J). This rapid clearance is a key pharmacological objective in prandial therapy [1, 2], as it more closely mimics the endogenous postprandial insulin spike and potentially reduces the risk of late‐phase hypoglycemia.

FIGURE 6.

Pharmacodynamics and pharmacokinetics of BN‐Inu–stabilized HALQ in diabetic pigs. Fasted pigs with diabetes received subcutaneous injections of BN‐Inu stabilized HALQ or Fiasp on separate study days, followed by serial blood sampling for glucose and insulin measurements. (A) Blood glucose profiles show a similar onset of action across treatments but faster glucose recovery with HALQ in comparison with Fiasp. (B) Normalized serum insulin concentrations indicate more rapid pharmacokinetics of HALQ, including faster post‐peak depletion relative to Fiasp. (C–F) Quantification of pharmacodynamic parameters: (C) time to 50% glucose reduction; (D) time to maximal glucose reduction; (E) time to 50% glucose recovery; and (F) duration of action, defined as time maintaining ≥ 50% of maximal glucose reduction. (G–J) Quantification of pharmacokinetic parameters: (G) time to 50% of peak insulin concentration; (H) time to peak concentration; (I) time to 50% decline from peak; and (J) duration of action, defined as time maintaining ≥ 50% of peak insulin concentration. Data are all presented as mean ± s.d., n = 5. Statistical significance was determined by a restricted maximum likelihood repeated measures mixed model with post‐hoc Bonferroni test, ^* p <0.05. Statistical comparisons not shown in the figure were non‐significant (p > 0.05).

2.6. Pharmacokinetics and Pharmacodynamics of HALQ Formulated With Absorption Enhancers

To better understand the potential of monomeric HALQ formulations to compete with commercial fast‐acting insulins, we co‐formulated HALQ with absorption enhancers used in the ultrarapid Lyumjev and compared the resulting PD and PK profiles. Both Lyumjev (insulin lispro) and Fiasp (insulin aspart) are classified as very fast‐acting insulins, while Lyumjev demonstrates even faster absorption, with a reported time to maximum serum insulin concentration of 57 min compared with 63 min for Fiasp in patients with diabetes. Lyumjev also exhibits faster elimination, with a median half‐life of 44 min vs. an apparent half‐life of approximately 1.1 h for Fiasp [23, 24]. The enhanced absorption of Lyumjev arises from two excipients: treprostinil, which increases local blood flow through vasodilation, and citrate, which enhances local vascular permeability and thereby accelerates insulin uptake [50]. By incorporating these same absorption enhancers at the concentration used in Lyumjev, we generated a HALQ formulation (HALQ 100 U/mL, 1 wt.% BN‐Inu, 0.000106 wt.% treprostinil sodium, 0.441 wt.% sodium citrate dihydrate) that was expected to exhibit accelerated pharmacokinetics relative to the BN‐Inu formulation alone. It is worthwhile emphasizing that the objective of faster kinetics of subcutaneous insulin is not unbounded, which is constrained by physiological considerations due to inherent physiological barriers, including diffusion from the subcutaneous depot into the capillary circulation and subsequent systemic distribution. The practical goal is not to achieve infinitely fast kinetics, but rather to minimize the mismatch between insulin action and the temporal profile of endogenous prandial insulin secretion.

Treprostinil and citrate, the small molecule absorption enhancers, did not affect HALQ stability nor interfere with the ability of BN‐Inu to prevent HALQ aggregation (Figure S7B). Consistent PD profiles were observed between HALQ formulated with BN‐Inu alone and the HALQ formulation containing treprostinil and citrate following subcutaneous injection in rats with diabetes (Figure 7A). This is not surprising, as unlike humans, rats possess only a single layer of subcutaneous fat and have more permeable skin, reducing the subcutaneous depot effect and limiting the ability to distinguish absorption‐enhancing mechanisms. Therefore, pigs with diabetes whose subcutaneous tissue architecture more closely resembles that of humans were used to differentiate pharmacological behaviors across formulations. Compared with Lyumjev, the HALQ formulation produced a similar BG depletion effect in this cohort, whereas HALQ formulated with BN‐Inu alone displayed slightly slower BG reduction, particularly between 40–110 min post‐administration (Figure 7B). Such differences likely reflect the additional absorption enhancement contributed by treprostinil and citrate, whose effects appear to outweigh the inherent absorption‐rate advantage of monomeric HALQ relative to the dimeric/hexameric insulin lispro species present in Lyumjev under these experimental conditions. It is also important to note that the high variability in HALQ PD responses in this pig cohort, compared with the previous HALQ vs. Fiasp study, is attributable to physiological heterogeneity among individual pigs, which strongly influences glucose dynamics.

FIGURE 7.

Pharmacodynamics and pharmacokinetics of BN‐Inu–stabilized HALQ formulated with absorption enhancers from Lyumjev. Fasted rats or pigs with diabetes received subcutaneous injections of BN‐Inu–stabilized HALQ with or without Lyumjev absorption enhancers, or Lyumjev itself, followed by serial blood sampling. (A) Blood glucose profiles show that BN‐Inu stabilized HALQ with or without absorption enhancers yields comparable pharmacodynamics in rats. (B) Blood glucose profiles show that adding absorption enhancers yields pharmacodynamics comparable to Lyumjev, whereas HALQ without enhancers displays a slower onset of action. Asterisks denote statistically significant differences between HALQ+BN‐Inu and Lyumjev. (C) Normalized serum insulin concentrations demonstrate accelerated HALQ pharmacokinetics when formulated with absorption enhancers. (D) Normalized pharmacokinetic profile of insulin lispro in Lyumjev after subcutaneous injection. (E–H) Quantification of pharmacokinetic parameters: (E) time to 50% of peak insulin concentration; (F) time to peak concentration; (G) time to 50% decline from peak; and (H) duration of action, defined as time maintaining ≥50% of peak insulin concentration. Data are all presented as mean ± s.d., n = 5. Statistical significance was determined by restricted maximum likelihood repeated measures mixed model with post‐hoc Bonferroni test, ^* p <0.05, ^** p <0.01, ^*** p <0.001. Statistical comparisons not shown in the figure were non‐significant (p > 0.05).

Serum concentration of HALQ and insulin aspart was measured over time to assess PK (Figure S9D–F). The Normalized serum concentration profiles revealed that HALQ exhibited a similar time to reach 50% of its maximal concentration with or without absorption enhancers, whereas the HALQ formulation containing treprostinil and citrate displayed a significantly faster elimination rate beyond 15 min post‐administration (Figure 7C), consistent with increased local blood flow. Although absorption enhancers are expected to further accelerate HALQ uptake, detecting such differences in pigs is challenging because BN‐Inu–formulated HALQ already exhibited extremely rapid absorption kinetics in this model. The inherently fast kinetics in pigs [51] also potentially mask BG‐lowering differences between Lyumjev and the enhanced HALQ formulation, explaining the lack of clear PD separation observed in Figure 7B. Notably, within this pig model with diabetes, the monomeric HALQ formulation appeared to mitigate the subcutaneous depot effect compared with Lyumjev in pigs with diabetes (Figure 7D), as reflected by its shorter time to reach 50% of maximum serum concentration (Figure 7E) and peak concentration (Figure 7F), faster elimination (shorter time to 50% of peak during decline, Figure 7G), and reduced duration of action (Figure 7H).

2.7. Pharmacokinetics Modeling and Prediction in Humans

As physiological parameters governing insulin absorption differ substantially between pigs and humans [52], the PK profiles observed in pigs cannot be directly translated to human performance. Accordingly, we used the pig data to predict human pharmacokinetics of the monomeric HALQ formulation through a simplified subcutaneous insulin absorption model [53]. This physiological compartmental model uses first‐order kinetics to describe three key kinetic parameters governing insulin absorption from the subcutaneous injection site to the interstitum area (k₁ ), subsequently into blood circulation (k₂ ), and finally to elimination (k₃ ) (Figure 8A). The post‐injection insulin concentration at the depot (I_inj ) was normalized to 1. Among these parameters, the rate constant k₁ reflects both formulation properties and species‐dependent physiology, whereas k₂ and k₃ are determined solely by species characteristics. Importantly, the ratio of k₁ values between formulations is assumed by this model framework to be species‐independent, enabling predictive cross‐species translation. To simplify the model, insulin lispro is expected to rapidly dissociate into dimers upon subcutaneous injection [54], and both the dimeric and monomeric species are represented together in the [I _inj ] compartment. Although dimers absorb more slowly than monomers, both have higher absorption rates than hexamers, and the dissociation of the dimer can be negligible for simplicity [53]. Using this framework, we first fitted the pig PK data to estimate the three rate constants for both Lyumjev and the HALQ formulation in pigs. We then incorporated known clinical PK data for Lyumjev in humans to derive human‐specific k₂ and k₃ values. Finally, leveraging the species‐independent ratio of k₁ between formulations, we projected the corresponding predicted rate constants for the HALQ formulation in humans.

FIGURE 8.

Pharmacokinetic modeling of subcutaneous insulin to predict human performance. (A) Schematic of the physiological compartmental model employing first‐order kinetics to simulate subcutaneous insulin absorption. (B) Model fits normalized pharmacokinetic profiles of HALQ formulation (BN‐Inu stabilized HALQ with absorption enhancers) and Lyumjev in diabetic pigs (data from Figure 7c,d). (C) Predicted human pharmacokinetics for HALQ formulation and Lyumjev based on model‐derived parameters. (D–G) Quantification of predicted pharmacokinetic parameters: (D) time to 50% of peak insulin concentration; (E) time to peak; (F) time to 50% decline from peak; and (G) duration of action, defined as time maintaining ≥ 50% of peak insulin concentration.

The fits of the pig PK data for both Lyumjev and the HALQ formulation are shown in Figure 8B, and the corresponding parameter estimates are summarized in Table S1. To construct the human PK model for the HALQ formulation, human kinetic parameters for Lyumjev were first obtained by fitting clinical PK datasets from the literature. Treprostinil and citrate act locally at the injection site by increasing blood flow and vascular permeability. Although these excipients accelerate insulin absorption from the subcutaneous space, they exert negligible influence on systemic plasma flow, where insulin distribution and elimination occur. Therefore, the human elimination constant k₃ was fixed at the literature‐reported, patient‐independent population value of 0.16 min⁻¹ [55]. Human PK data for Lyumjev were compiled from several clinical studies [23, 56, 57, 58], and the fitted human PK curve with fixed k₃ is shown in Figure S10A. With the human k₂ and k₃ values established from Lyumjev, the k₁ value for the HALQ formulation can be translated from the pig model using the species‐independent ratio of k₁ between formulations. This enabled the simulated prediction of the human PK profile for the HALQ formulation. The predicted human HALQ PK curve suggested a faster onset and more rapid elimination relative to Lyumjev (Figure 8C). Specifically, the model predicts that the HALQ formulation would exhibit a shorter time to reach 50% of peak concentrations of 13.2 min, compared to 15 min for Lyumjev (Figure 8D). Moreover, the model predicts that HALQ could achieve a time‐to‐peak concentration of 38.8 min (Figure 8E), an elimination time to 50% of peak decline of 97.2 min (Figure 8F), and a duration of action (half‐life) of 84 min (Figure 8G). These simulations represent an overall PK rate projected to be more than 1.5‐fold faster than Lyumjev.

3. Conclusion

The clinical demand for ultrarapid insulin has driven extensive efforts to develop formulations with accelerated onset and optimized pharmacological profiles. While insulin analogs such as aspart and lispro enabled the design of faster‐acting insulin formulations, the improvements achieved to date have been modest, particularly regarding duration of action. Fiasp and Lyumjev represent optimized ultrarapid insulin formulations, both demonstrating faster onset compared with earlier fast‐acting insulin Novolog and Humalog through formulation strategies. However, Fiasp shows little difference in duration compared with Novolog, and Lyumjev exhibits a similar duration to Humalog [59]. Such incremental improvements are anticipated to be largely attributed to the fact that currently available fast‐acting insulins are formulated as hexamers in vials, and the rate‐limiting step for absorption into the bloodstream is the dissociation of hexamers into monomers at the subcutaneous depot [17]. The development of stable, fully monomeric insulin formulations is expected to bypass the subcutaneous depot effect, enabling true ultrarapid insulin action with substantially enhanced pharmacological performance.

In this study, building on our previous discovery of a fully active, monomeric humanized cone snail venom insulin (HALQ), we addressed the inherent stability challenges of monomeric insulin and successfully developed a stable, ultrafast‐acting formulation that does not require cold storage. A stabilizing excipient, BN‐Inu, was designed based on the natural polymer inulin to formulate HALQ without directly interacting with the insulin, thereby avoiding interference with its activity or pharmacological performance. The efficacy of the inulin‐based stabilizing excipient is evident in the exceptional stability of the HALQ formulation, which exhibited no aggregation over 96 h under stress‐aging conditions. This performance contrasts sharply with commercial fast‐acting insulins, Fiasp and Lyumjev, which aggregate within 3 h despite being formulated as hexamers and theoretically more stable than monomeric formulations [60]. Furthermore, HALQ formulated with BN‐Inu maintained long‐term stability for 30 days at room temperature, as confirmed by comprehensive evaluations of both chemical and physical degradation. Based on collective biophysical observations, we propose a mechanistic model wherein the polysaccharide backbone of BN‐Inu reduces protein–protein attractive interactions, while its surface activity could displace protein from the air–liquid interface, thereby mitigating aggregation at this interface.

The advantage of monomeric HALQ is most pronounced in its duration of action, outperforming commercial ultrarapid insulin Fiasp and Lyumjev in the evaluated pig models of diabetes. After subcutaneous injection, insulin absorption is largely governed by dissociation into absorbable monomers at the subcutaneous depot, as well as local vascular permeability and blood flow, which transport insulin into the systemic circulation. Absorption enhancers, such as vasodilators and permeation agents, can accelerate this process. However, the subcutaneous depot effect persists, resulting in sustained release and prolonged insulin exposure. For example, Fiasp employs niacinamide to promote insulin dissociation and cutaneous vasodilation. Our results showed that although Fiasp achieves a comparable short time to onset, its duration of action remains approximately 1.5‐fold longer than that of monomeric HALQ in these animals. Lyumjev represents another approach, combining the potent vasodilator treprostinil with the vascular permeation enhancer citrate to accelerate insulin lispro absorption. In our pig studies, formulating monomeric HALQ with the same absorption enhancers further accelerated absorption and shortened the duration of action. The combination of inherently rapid monomeric absorption and formulation‐mediated acceleration significantly reduced both the time to peak concentration and overall duration of action relative to Lyumjev under these experimental conditions. Human PK modeling predicts that HALQ formulation could reduce time‐to‐peak by 21 min and duration of action by 59 min compared with Lyumjev, suggesting an unprecedented ultrafast‐acting activity among commercial insulin products. Such improvements hold the potential to exert a meaningful clinical impact on glucose regulation, similarly to trends suggested in prior clinical trials investigating the benefits of fast‐acting insulin pharmacologic [3, 61].

In summary, we developed a highly effective stabilizing excipient projected to be capable of preventing both interfacial and bulk insulin aggregation, without modifying the protein, thereby preserving monomeric HALQ for long‐term storage at room temperature. This stabilizing excipient overcomes the extreme stability challenges of insulin monomers, potentially enabling the translation of promising monomeric HALQ into reliable injection formulations. Based on a natural polymer backbone, the stabilizing excipient also demonstrates a favorable biocompatibility profile in the preliminary short‐term assessment on mice with diabetes. Pharmacological evaluation in pig models confirmed the faster absorption of monomeric HALQ compared with commercial ultrarapid insulins, particularly reflected in its shorter duration of action. Human PK predictions suggest a 1.5‐fold reduction in duration of action relative to existing ultrarapid formulations such as Lyumjev.

While the results across multiple animal models are promising, this study has several limitations. First, large animal models inherently involve smaller cohorts. Although we utilized a randomized crossover design to maximize statistical power, we acknowledge that in specific instances (e.g., Figure 6H), high variance precluded reaching formal statistical significance. Furthermore, while the observed ‘fast‐on, fast‐off’ profile aligns with established pharmacological goals for prandial insulin, the direct translation of accelerated clearance to long‐term clinical outcomes remains to be fully validated. Consequently, these specific observations should be considered preliminary, and future large‐scale studies are required to confirm the full extent of these pharmacokinetic trends and therapeutic advantages. STZ‐induced porcine and rodent models are the gold standard for pre‐clinical diabetes research, but they do not fully recapitulate the complexities of human metabolic physiology and subcutaneous absorption kinetics. While our predictive modeling suggests that monomeric HALQ achieves a shorter duration of action compared to ultrarapid Lyumjev, the PK profile in humans remains to be established in clinical trials. To further validate the potential clinical impact of the HALQ formulation, we compared our predictive PK data with established clinical benchmarks, rapid‐acting Humalog, and ultrarapid Fiasp (Figure S10B). Ultrarapid Fiasp and Lyumjev offer a modestly faster onset of action but an offset of action similar to traditional rapid‐acting Humalog. In contrast, our models predict that the monomeric HALQ formulation could provide both a moderately faster onset and a substantially accelerated offset of action compared to both Lyumjev and Fiasp. Finally, while our initial biocompatibility assays indicate a favorable safety profile for the stabilizing excipient BN‐Inu, further long‐term toxicology studies and regulatory evaluation will be required to transition this monomeric HALQ formulation from a preclinical candidate to a clinically available therapeutic. The availability of the ultrafast‐acting monomeric HALQ formulation could ultimately provide substantial clinical benefits for patients with diabetes, offering a pathway toward tighter glycemic control and improving overall outcomes of insulin therapy.

Moving forward, as the inulin‐derived excipient was designed based on general protein aggregation mechanisms, broader exploration of its stabilization effects and formulation optimization across diverse protein systems will be of interest for future development. While our current study demonstrates excellent thermal stability of monomeric insulin, expanding this formulation for use in continuous subcutaneous insulin infusion pumps represents a key next step. Accordingly, further formulation optimization under constant agitation and shear stress to mimic the mechanical environment of wearable pump reservoirs, particularly for fully automated, closed‐loop insulin delivery without bolus announcement, will require rigorous assessment of the formulation's compatibility with the plastics, tubing, and fluidic compartments to ensure minimal adsorption.

4. Experimental Section

4.1. Materials

All Fmoc amino acids, reagents, and solvents were used without purification. Fmoc amino acids were purchased from ChemPep and PurePep. 2‐chlorotrityl chloride (2‐CTC) resin (100–200 mesh, 0.4–1.0 mmol/g) from ChemPep. 1‐[bis(dimethylamino)methylene]‐1H‐1,2,3‐triazolo[4,5‐b]pyridinium 3‐oxid hexafluorophosphate (HATU, 99%) was purchased from Oakwood Chemical. N,N‐dimethylformamide (DMF, ≥99.8%), dichloromethane (DCM, >99.9%), acetonitrile (MeCN, >99.9%), methanol (MeOH, >99.9%), diethyl ether (Et₂O, >99%), N,N‐diisopropylethylamine (DIPEA, >99.5%), trifluoroacetic acid (TFA, >99.5%), 4‐methylpiperidine, triisopropylsilane (TIPS, >98%), and Dulbecco's phosphate buffer saline (PBS) were purchased from Fisher Scientific. Inulin from chicory (≤0.05% free glucose), 4‐bromobutyronitrile (>97%), and streptozotocin (≥75% α‐anomer basis, ≥98% HPLC) were purchased from Sigma–Aldrich. 1‐bromo‐2‐ethylbutane (>97.0%) was purchased from TCI. Anti‐Insulin [HB125 (mAb1, AE9D6, Ab 125), rabbit IgG] was purchased from Absolute Antibody. Recombinant human insulin and goat anti‐rabbit IgG Fc secondary antibody (HRP) were purchased from Thermo Fisher Scientific. Anti‐HALQ (34B7D4, 99%) was produced and purified by GenScript.

4.2. Synthesis of Monomeric Insulin Analog HALQ

Monomeric human insulin analog HALQ was synthesized based on the previously reported method [11]. Briefly, peptides were synthesized via Fmoc chemistry on a peptide synthesizer (Syro I MultiSynTech GmbH) in a 10 mL reactor vial with a 0.1 mmol total loading capacity of the resin. The first C‐terminal amino acid of the carboxylic acid C‐terminus was coupled manually to 2‐CTC resin: Fmoc‐amino acid (0.1 mmol) and DIPEA (87.1 µL, 0.5 mmol) were dissolved in a solution of DMF and DCM (1:1, 2.5 mL). This solution was added to 2‐CTC resin (250 mg), which was pre‐washed with DMF and DCM. After the reaction mixture rotated for 2 h at room temperature, the resin was washed with DMF and DCM and then capped with a solution of DCM, MeOH, and DIPEA (17:2:1, 5.0 mL) for 10 s four times. The resin was finally washed with DCM and DMF. Fmoc deprotection: (i) 4.5 mL of 20% piperidine in DMF; and (ii) mix 2 × 3 min (new solvent delivered for each mixing cycle). Amino acid coupling: (i) 1.25 mL of 0.4 m Fmoc‐protected amino acid in DMF; (ii) 1.225 mL of 0.4 m HATU; (iii) 1.0 mL of 1.0 m DIPEA in DMF; and (iv) mix for 10 min at 70°C (for cysteine and histidine coupling: mix 10 min at 50°C; for arginine coupling: mix 10 min at 50°C). DMF washing (performed between deprotection and coupling steps): (i) 4.5 mL of DMF; and (ii) mix for 45 s. Upon completion of synthesis, the resin was washed with DCM and dried (using vacuum) for 30 min. Peptide was cleaved from resin for 2 h with cleavage cocktail (TFA: H₂O: TIPS, 38:1:1). Peptide was precipitated with ethyl ether at 4°C, followed by HPLC purification and lyophilization.

4.3. Synthesis of Inulin‐Based Stabilizing Excipients

Inulin from chicory (≤0.05% free glucose) was functionalized with either 2‐ethylbutyl or butyl nitrile. Taking 2‐ethylbutyl modified inulin as an example, inulin (1 g, 5.5 mmol of fructose) was added to 30 mL N, N‐dimethylacetamide (DMAc, anhydrous 99.8%, Thermo Scientific) with continuous stirring at 50°C for 30 min to get complete solubilization. After being cooled to room temperature, 4 eq sodium hydride (NaH, 57%–63% oil dispersion, Thermo Scientific) was added slowly in several portions to get a homogeneous suspension, continued with 30 min stirring at 50°C. The reaction mixture was cooled to room temperature again, followed by the addition of 5 eq of 1‐bromo‐2‐ethylbutane (>97.0%, TCI America), with continuous stirring for 24 h at room temperature, to achieve a degree of substitution (DS) of 0.3. The reaction was stopped by the addition of cold ethanol (EtOH, absolute 200 proof, Fisher BioReagents), precipitated by the addition of ether (99%, HPLC grade, stab. with ethanol, Thermo Scientific), and separated by slow centrifugation (3000 rpm for 5 min). The DS can be controlled by the amount of NaH and alkyl bromide, as well as the reaction time. The batch‐to‐batch consistency of BN‐Inu for HALQ formulation stabilization was confirmed by randomly collected 3 BN‐Inu samples from 3 different batches to validate the consistent efficiency of stabilizing HALQ (Figure S11). The Mn, Mw, and dispersity of the inulin/BN‐Inu were determined by matrix‐assisted laser desorption/ionization mass spectrometry (MALDI‐MS). MALDI‐MS was performed using AB SCIEX 5800 TOF/TOF System. Sample matrix 2,5‐dihydroxybenzoic acid (DHB) was dissolved in water to give a final concentration of 10 mg/mL. Inulin/BN‐Inu was dissolved in water to prepare a 1 mg/mL solution. The samples were added to the matrix solution to give 1:1 v/v. 1 µL of sample/matrix mixture was placed on a target plate and air dried. MALDI‐MS spectra were acquired in positive ion mode, and the molecular weight distribution was calculated based on the mass spectra in the m/z range of 1000–4000.

4.4. NMR Characterization

¹H one‐dimensional NMR spectra were recorded at a HALQ or NALQ concentration of 3.15 mg/mL, eb‐Inu or BN‐Inu at a concentration of 1 wt.%, 3.15 mg/mL HALQ with 0.2 wt.% (∼ 1: 0.5 molar ratio) and 2 wt.% (∼1: 5 molar ratio) concentration of eb‐Inu, 3.15 mg/mL HALQ with 1 wt.% BN‐Inu, 3.15 mg/mL NALQ with 1wt.% eb‐Inu in deuterium oxide (D₂O, > 98.0%, Thermo Scientific). A Varian VNMRS 400 MHz NMR instrument was used to acquire the data.

BN‐Inu was characterized by ¹H NMR, and the DS was calculated based on the ¹H NMR peaks. Inulin consists of a chain‐terminating glucosyl moiety and a repetitive fructosyl moiety. NMR peaks from the H_glu proton of glucosyl moiety located at 5.4 ppm and the terminal methyl proton at <1.0 ppm. Meanwhile, the peaks of 7 protons from each fructose unit, 6 protons from the glucosyl moiety, and the protons from the butylnitrile substitution groups (2 protons) were all in the region of 3.3–4.3 ppm. The integrals of all other peaks in the spectrum were then normalized with respect to the methyl peak at <1.0 ppm. Because there is only a single glucosyl, the contribution from the glucosyl unit per inulin polymer group is 6 × (integral of H_glu). Herein, the estimation of DS value was calculated by the following equation, independent of the molecular weight (Mn) of inulin.

$$\begin{array}{cc}& DegreeofSubstitution\left(DS\right)\hfill \\ & =\frac{7}{\left[Integralof3.3-4.3ppm\right]-\left[6\times \left(Integralof{H}_{Glu}\right)\right]-2}\hfill \end{array}$$

4.5. High Performance Liquid Chromatography (HPLC) and LC‐Mass Spectrometry (LC‐MS)

Peptides were purified with water/ACN gradient in 0.1% TFA on an Agilent 1260 HPLC. Fractions 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/ACN gradient in 0.1% formic acid on an Agilent 6120 Quadrupole LC/MS system. Fractions containing products were collected and lyophilized. 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 20% aqueous ACN (0.1% TFA) to 50% aqueous ACN (0.1% TFA) over 40 min at a flow rate of 5 mL/min for A chains and from 30% aqueous ACN (0.1% TFA) to 60% aqueous ACN (0.1% TFA) over 40 min at a flow rate of 5 mL/min for B chains. Method B: Folded peptides were purified using a Phenomenex semi‐preparative C18 column (5 µm, 250 mm × 10 mm) with a linear gradient from 20% aqueous ACN (0.1% TFA) to 50% aqueous CAN (0.1% TFA) over 35 min at a flow rate of 3 mL/min.

4.6. Sedimentation Velocity Analytical Ultracentrifugation

SV‐AUC was performed using a Beckman Coulter XL‐I centrifuge. HALQ was prepared in phosphate buffer at a concentration of 100 U/mL (3.15 mg/mL). HALQ in formulation was prepared in phosphate buffer at the same concentration (3.15 mg/mL) with 1 wt.% BN‐Inu. Lyophilized HALQ were reconstituted in sterile diluent and dialyzed overnight against 100 sample volumes of sterile diluent. Concentrations of HALQ in sterile diluent were determined using a Direct Detect spectrometer (EMD Millipore) in AM1 quantification mode. Fiasp (U‐100, insulin aspart injection) and sterile diluent (3.48 mg/mL arginine, 0.53 mg/mL disodium phosphate dihydrate, 3.3 mg/mL glycerol, 1.72 mg/mL m‐cresol, 20.8 mg/mL niacinamide, 1. 5 mg/mL phenol, and 19.6 µg/mL zinc, pH 7.4) were obtained from Novo Nordisk. All AUC samples were centrifuged at 50 000 rpm. Data collection using the absorbance optics was conducted with sample volumes of 408 µL in 12‐mm path‐length epon resin centerpieces with quartz windows. Data for insulin in sterile diluent were collected using the interference optics, 12‐mm meniscus‐matching epon centerpieces, and sapphire windows. Dialysate (350 µL) was loaded into the reference chamber, and 330 µL of insulin was loaded into the sample chamber. An initial centrifugation at 3000 rpm was conducted to match the menisci, after which samples were removed and mixed by gentle agitation before data collection. Data were collected for ∼15 h (absorbance optics) or until the maximum number of scans had been recorded (interference optics, 999 scans with 30‐s interval, ∼8.5 h). Buffer density, viscosity, and partial specific volume of insulins were calculated using SEDNTERP 3. Data analysis was performed using the continuous c(s) model in sedfit software (v16.1c) with maximum entropy regularization and an F ratio of 0.68 (1 s.d.). Linear regression was performed with alternating use of the Marquardt–Levenberg and simplex algorithms with meniscus position, frictional ratio, and time‐invariant noise floated until the model root mean squared deviation reached a minimum value. Interference data were fit with radius‐invariant noise as an additional parameter. SV‐AUC figures were generated with the GUSSI57 software.

4.7. Second Virial Coefficient B₂₂

The second virial coefficient B₂₂ was calculated based on light scattering measurements conducted by Prometheus Panta (NanoTemper). HALQ in PBS buffer with BN‐Inu were prepared in a series of six dilutions in the concentration range of 0.625–20 mg/mL to capture concentration‐dependent effects. HALQ and BN‐Inu stock solutions were prepared after filtering through a 0.22 µm syringe filter to remove particulates and avoid scattering artifacts. The Size Analysis mode was selected to acquire static light scattering intensity (SLS) for each concentration. The scattering response was plotted, and a built‐in non‐linear Debye plot regression of the low‐concentration region was analyzed in PantaControl, where the slope corresponds to B₂₂. The non‐linearized version of the Debye plot generated by the software is shown below. I_A  is intensity of analyte corrected by the buffer signal. K_PaCo is constant in each experiment that accounts for the solvent refractive index and laser wavelength. c is sample concentration.

$$I_A=I_{sample}-I_{solvent}=\frac{K_{PaCo}·c}{\frac{1}{MW}+2B_{22}c}$$

4.8. Surface Tension

Surface tension was measured using the pendant‐drop method. Samples were filtered and equilibrated to the measurement temperature before analysis. A clean syringe fitted with a stainless‐steel needle was filled with the sample, and a pendant drop was formed by slowly dispensing liquid until a stable drop shape was obtained. The drop profile was recorded with a contact angle goniometer (Rame‐Hart 290), and surface tension was calculated by fitting the drop contour to the Young–Laplace equation using the built‐in analysis software. For each sample, at least three independent drops were measured. Temperature, needle size, and equilibration time were kept constant across all measurements.

4.9. In Vitro Stability Characterization

Aggregation of insulin was analyzed by absorbance‐aggregation assay. Formulation samples were plated at 100 µL per well in a clear 96‐well (n = 3) and placed into a plate reader (SpectraMax ID5) after being sealed with an optically clear and thermally stable adhesive plate seal (Excel Scientific). The plate was incubated with continuous shaking (480s rpm medium speed shaking between reads) at 37°C in the plate reader, and the absorbance was measured every 10 min at the wavelength of 540 nm. The aggregation of insulin leads to light scattering, which results in an increase in the measured absorbance. Chemical degradation of insulin and formulations after different aging times was evaluated by the mass spectra acquired by the liquid chromatography–mass spectrometry (LC–MS). Aliquots of aged insulin solution samples were analyzed on an Agilent LC/MS system (1260 Infinity LC Systems with Agilent InfinityLab LC/MSD iQ system) on an XBridge C18 5‐µm (50 mm × 2.1 mm) column at a flow rate of 0.4 mL/min with a water/acetonitrile gradient in 0.1% formic acid. The mass spectra were analyzed using an OpenLab CDS ChemStation.

4.10. Cellular AKT Activation Assay

pAKT Ser 473 levels were measured in NIH‐3T3 cells overexpressing IR‐B (a gift from A. Morrione, Thomas Jefferson University). Cells were cultured in DMEM (Sigma–Aldrich) with 10% fetal bovine serum (FBS; Gibco), 100 U/mL penicillin–streptomycin (Thermo Fisher Scientific), and 2 mg/mL puromycin (Thermo Fisher Scientific). For each assay, 40 000 cells in 100 µL per well were plated in a 96‐well plate with culture medium containing 1% FBS. Twenty hours later, 50 µL of insulin in serial dilution was pipetted into each well after removal of the original medium. After 30 min, the insulin solution was removed, and the homogeneous time‐resolved fluorescence pAKT Ser 473 kit (Cisbio) was used to measure pSer 473. Cells were treated with lysis buffer (50 µL per well) for 1 h under mild shaking. Lysate (16 µL) was added to 4 µL of detecting reagent in a white 384‐well plate. After 4 h, the plate was read in a SpectraMax iD5 microplate reader.

4.11. In Vivo Biocompatibility Measurements

Male and female C57BL/6 mice with diabetes were treated with (i) Humulin R U‐100 (n = 6), (ii) HALQ (n = 6), or (iii) 1wt.% BN‐Inu stabilized HALQ for 7 consecutive weeks. Formulations were administered subcutaneously at a dose of 1 U/kg. Whole blood was collected for cell count on week 6. Blood serum was collected for chemistry and cytokine profiling. Automated hematology was performed on the Sysmex XN‐1000 V hematology analyzer system. Blood smears were made for all CBC samples and reviewed by a clinical laboratory scientist. Manual differentials were performed as indicated by species and automated analysis. Chemistry analysis was performed on the Siemens Dimension EXL200/LOCI analyzer. A clinical laboratory scientist performed all testing, including dilutions and repeat tests as indicated, and reviewed all data. Cytokine profiling was performed using a bead‐based immunoassay platform (Luminex ProcartaPlex), with appropriate standards (commercial mix of mouse serum) and controls (AssayChex beads) included for accuracy.

4.12. Streptozotocin‐Induced Diabetes in Rats and Pigs

Animal studies were performed in accordance with the Guidelines for the Care and Use of Laboratory Animals and the Animal Welfare Act Regulations; all animal procedures were approved by the Stanford Institutional Animal Care and Use Committee (protocol number: 33909 and 33908). Streptozotocin (STZ) is an antibiotic that produces pancreatic islet β‐cell destruction and is widely used experimentally to produce a model of type 1 diabetes [62]. Sprague Dawley rats (Charles River) 160–230 g (8–10 weeks) were weighed and fasted 6–8 h before treatment with Streptozotocin (STZ). STZ was diluted to 10 mg/mL in the sodium citrate buffer and immediately injected intraperitoneally at a dosage of 65 mg/kg. Rats were later provided with water containing 10% sucrose for 24 h after STZ administration. Rat blood glucose levels were then tested daily via tail vein blood collection. Type 1‐like diabetes can be confirmed as having 3 consecutive blood glucose measurements > 400 mg/dL in non‐fasted rats. Type 1‐like diabetes was induced in Yorkshire pigs (25–30 kg, Pork Power) by infusing STZ intravenously at a dose of 125 mg/kg, followed by 24 h of monitoring. Food and 10% dextrose solutions were given as needed to prevent hypoglycemia. Diabetes was defined as fasting blood glucose greater than 300 mg/dl.

4.13. Insulin Tolerance Test in STZ‐Induced T1D Rats

Insulin tolerance tests were performed in diabetic Sprague–Dawley rats following a 4‐6‐h fast. HALQ was reconstituted in a 10 mg/mL sodium bicarbonate buffer and diluted in formulation diluent. Following baseline blood glucose measurements, diabetic rats were injected with 1 U/kg HALQ subcutaneously. Tail vein samples were obtained to assess blood glucose levels every 10 min for the first 60 min, every 20 min for the next 60 min, and then every 30 min up to 5 h.

4.14. In Vivo Pharmacokinetics and Pharmacodynamics Study in Diabetic Pigs

Diabetic pigs were fasted for 4 to 6 h and injected subcutaneously with a 0.1–0.2 U/kg dose of insulin formulations. Doses were determined on the basis of individual pig insulin sensitivity values with a target of a decrease in blood glucose of about 200 mg/dl. Individual pigs received the same dose for each treatment group. Pigs received each formulation once on separate days, and the order of the treatment groups was randomized. Before injection, baseline blood was sampled from an intravenous catheter line and measured using a handheld glucose monitor (Bayer Contour Next). After injection, blood was sampled from the intravenous catheter line every 5 min for the first 60 min, every 10 min for another 60 min, and then every 30 min up to 4 h. Blood was collected in serum tubes (Sarstedt) for analysis with ELISA. Serum aspart and lispro concentrations were quantified using an Iso‐insulin ELISA kit (Mercodia). All results were expressed as a mean ± SD.

4.15. HALQ ELISA Assay

Wells in MaxiSorp treated Nunc‐Immuno modules (Thermo Scientific 468667) were coated with capture antibody anti‐HALQ [34B7D4] by incubating with 100 µL of 1 µg/mL antibody in 0.2 m sodium carbonate/bicarbonate buffer (pH 9.4) at 4°C overnight. After washing with wash buffer (PBS containing 0.05% v/v Tween ‐20), the plate was then blocked with 300 µL of Pierce Protein‐Free Blocking buffer (Fisher Scientific 37572) for 1 h at room temperature and washed three times before storage. HALQ standards in serial concentrations were prepared in PBS pH 7.4. Serum samples and HALQ standards were all mixed with eBioscience ELISA/ELISPOT diluent in 1:3 dilution in an equal volume and loaded onto separate wells in duplicate. The wells were sealed and incubated on a rotary shaker at room temperature for 1 h, then washed and incubated with detection antibody, 100 µL 0.5 µg/mL anti‐insulin [AE9D6] (Rabbit IgG) in blocking buffer, at room temperature for 1 h. After washing the wells, 100 µL of goat anti‐Rabbit‐HRP (Thermo Fisher Scientific A16116) diluted in blocking buffer 1:10 000 was added to each well and incubated at room temperature on a rotary shaker for 1 h. The wells were washed, and 100 µL 1‐Step Ultra TMB ELISA Substrate (Thermo Scientific 34029) was added to each well and incubated for 15 min at room temperature. The reaction was stopped by adding 100 µL of 2 m sulfuric acid, and the absorbance at 450 nm was immediately measured on a SpectraMax iD5 microplate reader.

4.16. Pharmacokinetics Modeling and Prediction

The pharmacokinetic model was adopted from the literature [53]. Kinetic rate constants for pigs were fit for the normalized pig pharmacokinetic curves by minimizing the sum of squared errors (SSEs) between the generated, normalized insulin serum concentrations derived from the model and the experimental normalized pig serum insulin concentrations for HALQ and insulin lispro at the experimental time points from 0 to 120 min. The SciPy solve_ivp function was used to solve differential equations, and SciPy's minimize function (L‐BFGS‐B method) in Python for parameter optimization. Kinetic rate constants for human were fit for the normalized clinical pharmacokinetic curves of Lyumjev by minimizing the sum of squared errors (SSEs) between the generated serum concentrations derived from the model and the reported normalized serum lispro concentrations. The k₂ and k₃ are species‐dependent, and k₁ is both species and formulation‐dependent. Value for human k₃ was set as 0.16 min⁻¹ as reported in the literature. k₁ for HALQ was calculated based on the relationship: k_{1, HALQ, Pig} / k_{1, Lyumjev, Pig} = k_{1, HALQ, human} / k_{1, Lyumjev, human}

4.17. Statistical Analysis

All quantitative data were presented as means ± SD as indicated in the figure legends. Comparisons between two groups were conducted using two‐tailed unpaired Student's t tests, whereas comparisons among three or more groups were performed using one‐way analysis of variance (ANOVA) with post hoc multiple comparisons as appropriate. The restricted maximum likelihood repeated measures mixed model was used for pharmacodynamics and pharmacokinetics comparison between formulations to ensure the variance due to animals was properly estimated. This approach accounted for the inherent correlation within individual animals by treating the subject identity as a random effect, thereby ensuring accurate estimation of variance components. All statistical analyses were performed using GraphPad Prism. Statistical significance was considered at p <0.05.

Funding

This work was supported by the Breakthrough T1D Foundation 1‐SRA‐2024‐1606‐M‐N & 2‐SRA‐ 2025‐1698‐M‐B (DC), National Institutes of Health NIDDK DK120430 (DC), the National Institutes of Health NIDDK DK127268 (CPH), the Breakthrough T1D Foundation Postdoctoral Fellowship 3‐PDF‐2023‐1329‐A‐N (YZ), and the National Science Foundation Graduate Research Fellowship DGE 2146755 (MJA).

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting File: adma73767‐sup‐0001‐SuppMat.docx.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.