Abstract

The risk of inducing hypoglycaemia (low blood glucose) constitutes the main challenge associated with insulin therapy for diabetes1,2. Insulin doses must be adjusted to ensure that blood glucose values are within the normal range, but matching insulin doses to fluctuating glucose levels is difficult because even a slightly higher insulin dose than needed can lead to a hypoglycaemic incidence, which can be anything from uncomfortable to life-threatening. It has therefore been a long-standing goal to engineer a glucose-sensitive insulin that can reversibly auto-adjust its bioactivity according to ambient glucose levels to ultimately achieve better glycaemic control while lowering the risk of hypoglycaemia3. Here we report the design and properties of NNC2215, an insulin conjugate with bioactivity that is reversibly responsive to a glucose range relevant for diabetes, as demonstrated in vitro and in vivo. NNC2215 was engineered by conjugating a glucose-binding macrocycle4 and a glucoside to insulin, thereby introducing a switch that can open and close in response to glucose and thereby equilibrate insulin between active and less-active conformations. The insulin receptor affinity for NNC2215 increased 3.2-fold when the glucose concentration was increased from 3 to 20 mM. In animal studies, the glucose-sensitive bioactivity of NNC2215 was demonstrated to lead to protection against hypoglycaemia while partially covering glucose excursions.

Similar content being viewed by others

Week-long normoglycaemia in diabetic mice and minipigs via a subcutaneous dose of a glucose-responsive insulin complex

Article 06 December 2023

Molecular engineering of safe and efficacious oral basal insulin

Article Open access27 July 2020

Oral nanotherapeutic formulation of insulin with reduced episodes of hypoglycaemia

Article Open access02 January 2024

Main

Using insulin to control diabetes comes with the risk of introducing hypoglycaemia, namely blood glucose values below 3.9 mM. This is because blood glucose fluctuations are difficult to predict owing to many factors, such as the character and timing of meals, exercise, infections and changing individual insulin sensitivity. People with diabetes must therefore adjust their daily doses of insulin (both basal and meal insulin) to account for these factors. However, to avoid events of low blood glucose, which can be dangerous especially during the night, many opt for conservative insulin doses. Compromising insulin doses due to the fear of hypoglycaemia subsequently results in suboptimal glucose control, thereby increasing the risk of complications arising from long-term hyperglycaemia. To facilitate improved glycaemic control without the risk of hypoglycaemia, the idea of engineering an insulin that can modify its bioactivity in response to varying blood glucose levels has been pursued since the 1970s3. Despite many publications and patents, to date, no mechanism has proven to solve the issue to the extent that it can be applied to treat diabetes. Most papers in the field describe polymer systems that can release insulin from subcutaneous (s.c.) depots in response to glucose fluctuations, but such systems are limited by delayed glucose diffusion to the subcutis, as well as a delay in the released insulin entering the blood circulation. Moreover, such systems release insulin irreversibly, meaning that, once the insulin is released from the depot, it is no longer glucose sensitive. A better approach seems to be equipping insulin itself with glucose-responsive properties, so it can reversibly respond to glucose. Notably, glucose values vary over a narrow range (from approximately 2 to 20–30 mM in people with diabetes), so a rather steep change in insulin bioactivity must be attained for the glucose-sensitive insulin to have an impact. To achieve such sensitivity to glucose, a chemical group able to bind to glucose with maximal sensitivity in this glucose range will be required. One system was based on oligofucose/mannose insulin conjugates that can be cleared from the circulation in an equilibrium between glucose-sensitive binding to the mannose receptor versus insulin binding to the insulin receptor, but this did not merit pursuing beyond phase I clinical trials. The glucose response was found to be shallow, and high clearance at the mannose receptor led to a very low in vivo potency, implicating the eventual need for prohibitively high insulin doses.

The concept of introducing a glucose-sensitive switch into the insulin molecule has been pursued over many years. A switch involves dual conjugation of a glucose-binding motif plus a binding partner onto insulin such that, at low glucose, the switch will induce a closed less-active state, equilibrating towards an open more-active state with higher glucose concentrations. The glucose-binding motif must therefore have an affinity for both glucose and the binding partner within the narrow glucose range that occurs in people with diabetes (approximately 2 to 20–30 mM). Furthermore, the two components of the switch must be attached to insulin in a manner that ensures that, in the closed state, there is a lower insulin bioactivity by altering the insulin conformation and/or blocking the receptor binding surfaces of insulin. This switch idea has been pursued by using boronates as glucose binders, but the glucose sensitivity of such designs has so far been too limited for pharmacological use. The best previous example of a carbohydrate-sensitive switch working with insulin showed sensitivity to fructose at high concentrations (50 mM), but the compound was insensitive to glucose. A recently identified macrocycle offers another option for a glucose-binding element. The macrocycle was designed to provide a glucose-binding cavity that secures a relevant affinity for glucose as well as selectivity over other carbohydrates and potentially interfering small molecules. Here we describe the molecular design of NNC2215, an insulin with a glucose switch by incorporating the macrocycle at B29Lys and introducing an O1-glucoside through a short linker at B1Phe . This combination of glucose binder, glucoside, linker and conjugation sites was found to impart glucose-sensitive bioactivity to NNC2215, which demonstrated a 12.5-fold increase in insulin receptor binding affinity when glucose was raised from 0 to 20 mM and a 3.2-fold increase when raised from 3 to 20 mM. Furthermore, NNC2215 was shown to be glucose sensitive in vivo, to attenuate hypoglycaemia in pigs and to reduce the glucose excursions during glucose tolerance tests (GTTs) in diabetic rats.

Fig. 1: Functional principle and 3D model of NNC2215.

a, NNC2215 is an insulin molecule with a glucose-sensitive switch. At increasing glucose concentrations, the switch equilibrates towards an open state and the insulin receptor affinity of NNC2215 is high, thereby contributing to preventing hyperglycaemia. When glucose levels decrease, the switch equilibrates towards a closed state, interfering with the ability of NNC2215 to bind to the insulin receptor, thereby contributing to preventing hypoglycaemia. Insulin backbone, macrocycle, glucoside and glucose models were prepared using BIOVIA Discovery Studio (Dassault Systèmes). b, 3D models of NNC2215 in the open and closed forms. The insulin backbone is shown as ribbons and the switch elements (glucoside and macrocycle) are shown as stick representations. Insulin receptor chains A and C from PDB 6PXV are shown as white and grey surface representations. The open form of NNC2215 (yellow) has free glucose (orange; top left corner) bound to the macrocycle at B29. The B1–glucoside of NNC2215 in the open form is shown in orange (on the right). The closed form of NNC2215 (cyan) has the glucoside bound in the macrocycle and shows a clash between the C-terminal part of the insulin B-chain, including the switch and the C-terminal domain of the insulin receptor (α-CT, purple).

Chemistry

The macrocycle (Fig. 2a) was conjugated to desB30 human insulin at the B29Lys Nε amino group through triazole formation17 between a macrocycle derivative carrying an azido propyl linker from the macrocycle roof and an alkyne linker attached to B29Lys (by conjugation at pH > 10 to obtain the B29 product). Conjugation through the macrocycle roof circumvents the need to orthogonally address one of three carboxylic acids, as would be needed for conjugation to either of the COOH groups on the pillars of the macrocycle. Dendrimers that were used for securing good aqueous solubility of the originally reported macrocycle were found to be unnecessary when working with insulin–macrocycle conjugates. Besides the macrocycle at B29, an O1-glucoside with a short linker was attached to the B-chain N-terminal amino group (PheB1) by using the corresponding bromo trifluoromethyl sulfate phenyl ester at pH 7.5 (ref. 18). The glucoside was used as its O-peracetyl protected building block, and the acetyl groups were removed from the insulin conjugate by gentle saponification. A control compound, NNC2215a, with only the macrocycle at B29 was prepared similarly to NNC2215 by omitting the glucoside step. The conjugation sites on insulin were documented by liquid chromatography coupled with mass spectrometry (LC–MS) analysis of a sample that was treated with trypsin followed by tris(2-carboxyethyl)phosphine (TCEP). The trypsin treatment cleaved NNC2215 after B22Arg to release the B23–B29 fragment, and TCEP cleaved the disulfides of NNC2215 to give separate A and B chains. LC–MS analysis of the resulting analytical mixture showed the macrocycle attached to the B23–B29 fragment, and the glucoside attached to the B1–B22 fragment, along with free A-chain (chemistry details and LC–MS spectra are provided in the Supplementary information and Supplementary Data).

Fig. 2: Chemical structure and glucose-binding properties of NNC2215.

a, NNC2215 with dual conjugation of the macrocycle at B29 and the glucoside at B1. The control compound NNC2215a has only the B29–macrocycle. Prepared using BIOVIA Draw (Dassault Systèmes). b, ITC measurement of the affinity of glucose to the free macrocycle. A Kd of 98 μM was obtained by fitting the data to a 1:1 binding model. c, Analysis of the binding of glucose towards NNC2215 (Kd = 2.1 mM) and NNC2215a (Kd = 0.5 mM) using native MS (the raw native mass spectra are shown in the Supplementary Data). R2 values from nonlinear regression are 0.9989 for NNC2215 and 0.9994 for NNC2215a. Data are mean ± s.d. n = 3 technical replicates. The s.d. error bars are shorter than the size of the symbols. Individual values are shown as black squares/circles.

Glucose binding of macrocycle and NNC2215

Using isothermal titration calorimetry (ITC), the free macrocycle of NNC2215 was shown to bind to glucose with a dissociation constant (Kd) of 98 µM (Fig. 2b). Native MS analysis was used to study the binding of NNC2215 towards glucose. As expected, the presence of the glucoside in the dual conjugate changed the glucose affinity of NNC2215 relative to the free macrocycle, such that NNC2215 was found to bind to glucose with a Kd of 2.1 mM, which, in the context of diabetes, is considered to be severe hypoglycaemia. The control compound, NNC2215a with macrocycle only, was found by native MS to bind to glucose with a Kd of 0.5 mM. It is not surprising that conjugation of the macrocycle to insulin will moderately change its glucose affinity (relative to the free macrocycle). Similarly, attaching albumin-binding moieties to insulin or other proteins also generally changes their affinity relative to the free ligands, probably due to weak interactions of the conjugated moieties with the proximal protein and its side chains. Overall, the glucose affinity of NNC2215 and the control compound NNC2215a as measured by native MS was approximately 20-fold and 5-fold weaker compared with the free macrocycle, thereby demonstrating that suitable switch dynamics had been achieved. The glucose binding and concurrent opening of the switch in NNC2215 and NNC2215a in response to 0 to 20 mM glucose can be followed by the native MS binding plot in Fig. 2c. The steepest part of the binding curve is consistent with what would be considered to be hypoglycemia, below 4 mM glucose. Although the native MS data are obtained in the gas phase (MS vacuum), such data often reflect the interactions of molecules corresponding to the aqueous solutions from which the complexes are sampled19.

3D model of NNC2215–insulin receptor

Structural models were built of the insulin receptor binding to NNC2215 with the switch in either the open or closed state by superimposing models of NNC2215 on the insulin–insulin receptor complex Protein Data Bank (PDB) 6PXV (refs. 20,21,22,23,24) to exemplify its functioning. As illustrated in Fig. 1b, when the switch is closed, a clash occurs between the C-terminal part of the insulin B-chain and the C-terminal domain of the insulin receptor, termed α-CT. The α-CT domain is known to be a crucial part for insulin binding25. We believe that this steric hindrance is the driving force for the observed lower receptor affinity of NNC2215 at low glucose concentrations, that is, with the switch populating mainly the closed conformation. By contrast, the model of the open state is compatible with a fully active insulin conformation able to bind to the receptor.

In vitro biology

To study the glucose-sensitive interaction of NNC2215 with the insulin receptor in vitro, insulin-receptor-binding studies were conducted in the absence or presence of varying glucose concentrations whereby the binding of human insulin receptor A (hIR-A) to NNC2215 was compared to that of human insulin and insulin degludec (an acylated, long-acting basal insulin)26 (Fig. 3a–c). Unique to NNC2215, it can be seen that the binding curves are affected by the increasing glucose concentrations (Fig. 3a). Relative to the hIR-A affinity for human insulin, the affinity for NNC2215 increases from 0.75% to 2.9%, 4.3%, 6.5% and 9.2% over the glucose concentration range of 0, 3, 5, 10 and 20 mM, whereas insulin degludec has a constant affinity relative to human insulin (Table 1 and Fig. 3d). The increase in hIR-A affinity for NNC2215 is 12.5-fold from 0 to 20 mM glucose, while it is 3.2-fold from 3 to 20 mM glucose, a concentration range that can be observed in people with diabetes (Table 1). Notably, this increase in hIR-A affinity was determined in the presence of 1.5% albumin, which is the maximum concentration that can be tolerated in the assay (compared to approximately 4% albumin in the human circulation). In the absence of albumin, the increase in hIR-A affinity was reduced to 6.8-fold (from 0 to 20 mM glucose), suggesting that albumin binding contributes to some extent to the glucose sensitivity of NNC2215 (Extended Data Fig. 1a). Furthermore, the observation that, in the presence of 1.5% albumin, there is a small, 2.4-fold increase in hIR-A affinity (from 0 to 20 mM glucose) for NNC2215a, the control compound with only a macrocycle, also indicates a weak albumin-binding effect (Extended Data Fig. 1b).

Fig. 3: Glucose-dependent hIR-A affinity of NNC2215.

a–c, Representative displacement curves of 125I-insulin from hIR-A for NNC2215 (a), human insulin (b) and insulin degludec (c) in the presence of 0 to 20 mM D-glucose. Data are mean ± s.d. n = 3 technical replicates. For some datapoints, the s.d. error bars are shorter than the size of the symbols. d, hIR-A affinity of NNC2215 and insulin degludec relative to human insulin over increasing glucose concentrations. Data are mean ± s.d. n = 3 independent replicates. For some datapoints, the s.d. error bars are shorter than the size of the symbols. e, Representative curves of NNC2215 and insulin degludec dose-dependent conversion of 3H-D-glucose into lipid in isolated rat adipocytes at low (3 mM) and high (20 mM) L-glucose concentrations. Data are mean. n = 2 technical replicates. CPM, counts per minute.

Table 1 Glucose sensitivity of NNC2215 activity with respect to hIR-A binding and downstream metabolic response

The specificity towards the insulin receptor compared to the insulin-like growth factor 1 receptor (IGF-1R) is very important to avoid any increased mitogenicity. The IGF-1R binding was measured in the presence and absence of 20 mM D-glucose (Extended Data Fig. 1c). Relative to human insulin, at both 0 and 20 mM glucose, NNC2215 had approximately 10% IGF-1R affinity compared with its insulin receptor affinity (Extended Data Fig. 1a,d). Thus, as compared to human insulin, NNC2215 has a higher specificity towards the insulin receptor versus the IGF-1R.

The ability of NNC2215 to activate the insulin signalling pathway was examined in Chinese hamster ovary cells expressing the cloned human insulin receptor (CHO-hIR). Full dose–response curves were obtained for NNC2215 stimulating tyrosine phosphorylation of the human insulin receptor with a potency of 57.1% (95% confidence interval (CI) = 41.7–78.1) compared with human insulin. Downstream signalling through AKT and ERK activation had the same balance relative to human insulin with potencies of 68.4% (95% CI = 55.7–84.0) and 48.0% (95% CI = 34.4–66.9), respectively (Extended Data Fig. 1e–g).

The glucose sensitivity of NNC2215 was reflected in a metabolic end point measured ex vivo in rat adipocytes27. The induction of glucose uptake and incorporation into lipids in rat adipocytes, that is, lipogenesis, was measured in an assay in which we took advantage of the enantiomer of the physiologically active D-glucose, namely L-glucose. The rationale for using L-glucose is that cells do not catabolize or take up L-glucose in appreciable amounts, at least not through saturable transport, except in the case of some Gram-negative bacteria and plants under certain conditions28. In control experiments, L-glucose did not compete with the uptake of D-glucose into adipocytes, but L-glucose binds to the achiral macrocycle of NNC2215 with the same affinity as D-glucose29. In the hIR-A-binding assay, L-glucose is not as potent as D-glucose in activating NNC2215. The increase in the binding affinity of NNC2215 observed between 0 and 20 mM is 7.6-fold for L-glucose compared with 12.5-fold for D-glucose (Table 1). The metabolic response measured with L-glucose will therefore be underestimated compared with the effect of D-glucose. The lipogenesis measurement showed a 2.2-fold difference in the dose leading to half-maximal effect determined for NNC2215-induced 3H-D-glucose conversion into lipid in rat adipocytes at low (3 mM) versus high (20 mM) L-glucose (Fig. 3e). This demonstrates the glucose sensitivity of NNC2215 ex vivo with respect to insulin-induced metabolic response. With insulin degludec, lipogenesis showed no response to L-glucose.

In vivo pharmacology

To investigate the glucose-concentration-sensitive activation and deactivation of NNC2215 in vivo, we developed three different protocols. In the simplest protocol, rats were dosed intravenously with NNC2215 followed by L-glucose, thereby triggering NNC2215, resulting in the lowering of D-glucose in an L-glucose-dose-dependent manner. Furthermore, the deactivation of NNC2215 at low glucose was investigated in pigs by an acute drop in plasma glucose in comparison to the glucose drop induced by a non-glucose-sensitive insulin (insulin degludec). Finally, the activation of NNC2215 during meal-like glucose fluctuations was studied during a glucose challenge in diabetic rats. Insulin degludec was used as a control in the intravenous (i.v.) rat study and as a comparator in the pig study due to similar pharmacokinetic properties after i.v. administration (Extended Data Table 1a), whereas human insulin was used as comparator in the rat glucose challenge study. The in vivo half-life of NNC2215 was determined to be 1.2 h by i.v. dosing to rats and 1.3 h by i.v. dosing to pigs (Extended Data Table 1a).

L-Glucose study in rats

L-glucose was used to trigger NNC2215 in rats, without triggering endogenous insulin release as dosing of D-glucose would otherwise do in non-diabetic rats. Figure 4a shows how i.v. administration of the same dose of NNC2215 followed by L-glucose (4 dose levels including vehicle) dose-dependently triggers insulin action of NNC2215 by lowering of D-glucose. As expected, NNC2215 is dose-dependently cleared from the plasma after the triggering by L-glucose, as evidenced by the lowering of NNC2215 plasma concentrations (Fig. 4b). The initial peak in D-glucose after L-glucose or vehicle injection (Fig. 4a) is a short-lived response related to the dosing (handling) of the animals. This response is not associated with NNC2215 as the same pattern is also seen in a similar L-glucose protocol testing the non-glucose-sensitive insulin degludec (Extended Data Fig.