The CDS and SEC data show the molecular conformation and self-association states of the study analogues before and after injection. As elaborated below, the properties of insulin degludec emerged in these studies as particularly appealing for further clinical development.
Firstly, the SEC analysis reveals the extent to which the insulin analogues were self-associated in environments simulating the pharmaceutical formulation and the subcutaneous environment. In a pharmaceutical formulation containing phenol and cresol, and when using a phenol-containing elution buffer to avoid dissociation of phenol from the insulin complex, all the formulated insulin analogues eluted primarily as a single species with a molecular weight corresponding to the size of an insulin dihexamer (Table ). All the studied analogues displayed this property apart from insulin detemir, showing that the glutamic acid linker must be present for dihexamers to form. Removal of phenol from the elution buffer during SEC resembles the dilution of phenol after subcutaneous injection because phenol is a small molecule that diffuses readily in the tissue and across membranes, in contrast to the insulin zinc complex. The removal of phenol during SEC was associated with an increase in the apparent molecular mass at which some of the analogues eluted. This effect was particularly marked for insulin degludec and octadecandioyl-γ-L-Glu desB30 human insulin, both of which displayed multihexamer formation with an apparent MW >5 MDa (Table ). Multihexamer formation was impaired, however, in analogues featuring minor changes of the glutamic spacer molecule compared to degludec. Furthermore, the terminal carboxylic acid appeared to be critical for multihexamer formation since substitution with a hydroxymethyl group (16-hydroxyhexadecandioyl-γ-L-Glu) or a methyl group (hexadecanoyl-γ-L-Glu) significantly reduced multihexamer formation. In addition, multihexamer formation was inhibited by changing attachment of the glutamic acid spacer molecule to hexadecandioyl-α-L-Glu, or substituting the spacer molecule with its D-form (hexadecandioyl-γ-D-Glu). It therefore appears that multihexamer formation is dependent on the presence of a diacid of at least 16 carbon atoms connected to LysB29 of insulin via a γ-L-glutamic acid spacer molecule.
When the zinc content was reduced in the insulin formulation applied to the SEC column and elution was conducted with phenol-free buffer, insulin degludec eluted with a smaller apparent molecular mass indicating that the multihexamer complex is dependent on the zinc concentration. This was confirmed in the zinc dose–response SEC analysis, where, in the absence of zinc, insulin degludec eluted almost exclusively as a monomer.
Octadecandioyl-γ-L-Glu desB30 human insulin also eluted at smaller sizes when applied in a formulation with only two zinc per six insulin molecules, but the apparent molecular mass of the largest elution peak was still in excess of 1.3 MDa, showing that the multihexamer chains of this analogue are much slower to disassemble, hence the absorption kinetics of this analogue may to be too long to be clinically applicable.
When light scattering was used to monitor SEC employing the six zinc formulation, an almost quantitative formation of multihexamers was obtained but insulin degludec displayed a broader MW distribution of higher MW than octadecandioyl-γ-L-Glu desB30 human insulin. These differences might be caused by differences in the kinetics of formation of the two multihexamers.
The CDS analysis showed that analogues acylated by ligands composed of a spacer molecule and a fatty acid readily adopt the T3
hexamer conformation in a pharmaceutical formulation, with the SEC data suggesting that this species will assemble into dihexamers. Human insulin and insulin detemir also adopt the T3
conformation, but in contrast they eventually adopt R6
conformation at increasing concentrations of phenol (Fig. ; 14). In the case of insulin degludec, the T3
dihexamer appears to be highly stable since in further experiments (analytical ultra centrifugation and CDS, data not shown) this analogue could only be made to obtain R6
conformation as an insulin-zinc hexamer when formulated at neutral pH together with resorcinol and imidazol, which bind more tightly to the insulin’s ‘phenol binding pockets’ and are stronger inducers of the R state than phenol (22
The CDS and SEC observations suggest that T3R3 dihexamers are held together by the binding of two insulin hexamer T3 surfaces, and it can be speculated that this is the result of a single fatty acid contact between these surfaces. In the case of insulin degludec and octadecandioyl-γ-L-Glu desB30 human insulin, the change in conformation to T6 enables side chain contacts to be made at each T3 pole of a hexamer so that the large molecular mass species eluted in phenol-free SEC likely represent soluble hexamer chains, with each chain potentially linking hundreds of hexamer units.
The exact nature of the interaction between hexamers in the T3
dihexamer and the T6
multihexamer is not known. However, a mechanism that would be plausible and compatible with the experimental results would involve coordination of the terminal carboxylic acid residue of the fatty di-acid chain to a zinc atom in the neighbouring hexamer. In the multihexamer state, hexamers would therefore be joined together in a structure that could be likened, speculatively, to ‘pearls on a string’, with the fatty acid chains analogous to the ‘string’ and the zinc ions within the hexamers being the ‘pearls’. For the T3
hexamer, this mechanism would only allow formation of di-hexamers because the T3
hexamer is only open at one pole hence a fatty acid coordination would only be feasible by the two hexamers coming together via their open poles. When the conformation changes to T6
following dissociation of phenol, the hexamer opens at both poles, and the fatty di-acid could access zinc atoms in neighbouring hexamers to both sides, allowing formation of multihexamer chains. This hypothesis is supported by the light scattering data showing that ρ
) for insulin degludec; this value is in good agreement with values obtained for elongated and relatively stiff molecules (21
This putative mechanism explains the zinc-dependency of multihexamer formation, and can also explain why the fatty acid side chains need to be of a sufficient length and stereochemical orientation (favouring L-Glu over D-Glu and gamma-Glu over alpha-Glu for the linker), because a certain three dimensional orientation would likely be needed for the fatty acid side chain to contact the neighbouring zinc. Finally, this mechanism would explain why a di-carboxylic acid can induce multihexamer formation, whereas a mono-carboxylic acid cannot, since the terminal carboxylic acid would be needed for the fatty acid to act as a ligand with a zinc ion in the insulin hexamer.
Importantly, although ligand-mediated multihexamer formation leads to high molecular weight insulin analogue-zinc complexes, all the studied insulin analogues displayed full solubility as zinc phenol formulations and as multihexamers in physiological saline. The soluble state of the insulin multihexamer is of particular interest from the point of view of developing formulations for clinical use, since previous studies have highlighted the advantages of a soluble depot over a depot composed of precipitate particles with respect to reproducible absorption kinetics after subcutaneous injection (5
It is important to note that the formation of complicated structures like the insulin degludec multihexamer is likely to be sensitive to the environment during the time of formation. Thus, a previous study employing analytical ultra centrifugation reported a molecular weight of 44MDa (24
). The present investigation can only suggest that the multihexamers are large structures, likely comprising several hundred hexamer units. The exact size of multihexamer chains formed in vivo
at the site of subcutaneous injection remains elusive. In future studies, further analysis of larger aggregates using other types of methodologies should be performed (25
). Large-sized protein-like structures could potentially evoke immunological responses, and this issue is relevant, for example, for antibody-based therapeutics for which neutralising antibodies can impact efficacy over time (26
). Aggregates of endogenous proteins, particularly when denatured, may also induce an immunological response (27
). Therefore, the phase 3 clinical studies of insulin degludec, in which have been included close to 10,000 patients, have collected data on antibody formation and injection site reactions. Available data are reassuring, showing that, for all participants, the level of insulin degludec-specific antibodies was close to, or below, the limit of detection at screening and remained at the same level after 16 weeks of treatment (28
). Furthermore, dermatological injection site reactions occurred only rarely in these studies with no evidence of an increased incidence vs.
The CDS analyses in zinc-free formulations show that the test analogues, including insulin degludec, are monomers at concentrations higher than 1 mM, which means that once multihexamers dissociate due to zinc diffusion, the liberated dimers will immediately dissociate into monomers. This is likely to be the state in which insulin degludec is absorbed into the circulation, since monomers of insulin pass faster through the tissue and across membranes than higher molecular weight complexes.
A potentially important consequence of the property of forming very stable T3
dihexamers in a pharmaceutical formulation (and multihexamers at the site of injection) is that other insulins (e.g. rapid-acting insulins) can be co-formulated without the risk of inter-exchange of monomers to form hybrid hexamers either in the cartridge or injection depot. This could enable the development of combination products with discrete preservation of the PK profiles of the component insulins. Indeed, a co-formulation of insulin degludec plus insulin aspart is currently undergoing clinical investigation (30
In summary, insulin degludec and octadecandioyl-γ-L-Glu desB30 human insulin stood out in our analyses as compounds able to form multihexamer chains following subcutaneous injection, and insulin degludec further stood out as suitable for clinical development due to a favourable PK/PD profile with a plasma t½ of 25 h in humans in subsequent studies. It can be concluded that insulin degludec has a unique mechanism of protraction; in the presence of phenol and zinc, (as in a pharmaceutical formulation) it forms a soluble and stable dihexamer, but after injection, as phenol diffuses away, this re-organises to form multi-hexamer chains that will have a long residence time at the injection depot. With the gradual diffusion of zinc, however, these chains are expected to gradually disassemble to release monomers from the terminal ends of multihexamers since the zinc ions of the terminal ends are exposed due to their T3
conformation (Fig. ). Further studies are required to investigate the kinetics of the multihexamer dissociation process in vivo
as the rate-limiting step in the absorption of insulin degludec from the subcutaneous depot into the circulation. In contrast to other basal insulins, the release of monomers would become the rate-limiting step in absorption rather than, e.g. depot blood flow, and this might limit variability in PD response. Furthermore, the affinity for albumin is likely to buffer any variability that does occur in absorption rate through the mechanisms described previously for insulin detemir by Kurtzhals (31
). Indeed, the glucose-lowering variability profile of degludec has been shown to be superior to that of insulin glargine (32
Fig. 5 Schematic representation of the hypothesis for the mode of retarded absorption of insulin degludec: Insulin degludec is injected subcutaneously as a zinc phenol formulation containing insulin degludec dihexamer in the T3R3 conformation. Rapid loss of (more ...)
The pharmacological consequence of this protraction mechanism in terms of the time-action profile are shown by PK and PD data (24
). These demonstrate that in clinical use, insulin degludec has a duration of blood-glucose lowering action that extends beyond 42 h (33
), and hence insulin degludec reaches steady-state with daily dosing to produce a flat and stable PK/PD profile (24
). Clinical results obtained so far show that the ultra-long and flat action profile of insulin degludec can offer greater flexibility of dosing and a reduced risk of nocturnal hypoglycaemia compared to other basal insulins (28