Logo: to the web site of Uppsala University

Transient permeation enhancers (PEs) have been widely used to improve the oral absorption of macromolecules. During pharmaceutical development, the correct selection of the macromolecule, PE, and the combination needs to be made to maximize oral bioavailability and ensure successful clinical development. Various in vitro and in vivo methods have been investigated to optimize this selection. In vitro methods are generally preferred by the pharmaceutical industry to reduce the use of animals according to the "replacement, reduction, and refinement" principle commonly termed "3Rs," and in vitro methods typically have a higher throughput. This paper compares two in vitro methods that are commonly used within the pharmaceutical industry, being Caco-2 and an Ussing chamber, to two in vivo models, being in situ intestinal instillation to rats and in vivo administration via an endoscope to pigs. All studies use solution formulation of sodium caprate, which has been widely used as a PE, and two macromolecules, being FITC-dextran 4000 Da and MEDI7219, a GLP-1 receptor agonist peptide. The paper shares our experiences of using these models and the challenges with the in vitro models in mimicking the processes occurring in vivo. The paper highlights the need to consider these differences when translating data generated using these in vitro models for evaluating macromolecules, PE, and combinations thereof for enabling oral delivery.

Oral delivery of therapeutic peptides is limited by degradation by digestive proteases and poor gastrointestinal permeability. We have investigated how physicochemical properties, including degree of lipidation and degree of amino acid sequence modification, along with formulation with a permeation enhancer (PE), influence the enzymatic stability and intestinal absorption of glucagon-like peptide-1(GLP-1) receptor agonists. We compared four peptides: J211 (non-lipidated; modified), J229 (mono-lipidated; modified), MEDI7219 (bis-lipidated; modified), and semaglutide (mono-lipidated control; least modified). J211, J229 and MEDI7219 have similar amino acid modifications in the peptide sequence to reduce the number of labile proteolytic sites. An in vitro head-to-head comparison between MEDI7219 and semaglutide showed that MEDI7219 was more proteolytically stable (% remaining after 90 min) than semaglutide, which was degraded completely within 10 min. Notably, co-formulation with sodium caprate (C10) improved semaglutide stability, and at least doubled its half-life. Results from in vivo studies in rats following intraduodenal bolus administration, showed that in the absence of C10, the absorption of all the peptides was minimal, with cumulative fractions absorbed below 1 % for all four compounds. Co-formulation with C10 increased the bioavailability of the modified peptides by 35-40-fold, with J211, J229, and MEDI7219 reaching 7.5 %, 4 %, and 17.3 % respectively. Semaglutide's bioavailability improved by similar to 200-fold, however bioavailability did not exceed 2 %. These results demonstrate that C10 enhances peptide absorption primarily by increasing intestinal permeability but also likely by improving enzymatic stability of a labile peptide like semaglutide. Furthermore, when comparing the three modified peptides, the degree of lipidation positively correlated with increased intestinal absorption in both the presence and absence of C10.

Oral peptide delivery increasingly relies on permeation enhancers (PEs), but how enhancer self-assembly couples to peptide self-assembly and co-assembly remains poorly understood. Here, we investigated the interactions between glucagon-like peptide-1 (GLP-1) analogues and the permeation enhancer sodium caprate (C10), with emphasis on how peptide lipidation and enhancer self-assembly control peptide availability for absorption. A non-lipidated analogue (J211), two mono-lipidated analogues (J229 and semaglutide), and a bis-lipidated analogue (MEDI7219) were characterized by 1H NMR spectroscopy, diffusion NMR, small-angle X-ray scattering (SAXS), circular dichroism (CD) and all-atom molecular dynamics (MD) simulations. In deuterated buffer, the peptides formed distinct oligomeric states with the size ranking semaglutide > J211 > J229 ≈ MEDI7219, while retaining predominantly α-helical secondary structure. C10 underwent a marked colloidal transition with the formulation conditions, forming non-spherical micelles at pD 8.5 and vesicular aggregates at pD 7.4 in fasted-state simulated intestinal fluid (FaSSIF-V2). Under alkaline micellar conditions, C10 associated with all peptides and remodeled the larger oligomers, J211 and semaglutide, by loosening peptide-peptide packing without major changes in global dimensions or helicity. Under near-neutral biorelevant conditions, peptide-C10 co-assembly became strongly peptide dependent. MEDI7219 showed the weakest association with C10 aggregates and retained the largest unbound fraction, whereas semaglutide and J229 had the most bound fraction. This ranking correlated to previously observed intestinal absorption ranking in vivo. These results show that peptide absorption is impacted by the coupled equilibria between peptide self-assembly, enhancer colloidal state, and peptide bound versus unbound fractions.

Lipidation is widely used to improve peptide pharmacokinetics; however, its influence on peptide–membrane interactions relevant to oral delivery requires further investigation. Here, we investigated how lipidation pattern, amino acid sequence, and peptide oligomeric state influence the interactions of four GLP-1 analogues with a model membrane composed of 1-palmitoyl-2-oleoyl-glycero-3-phosphocholine (POPC). The analogues comprised non-lipidated J211; mono-lipidated, acid-terminated J229 and semaglutide; and bis-lipidated, methyl-terminated MEDI7219. Quartz crystal microbalance with dissipation monitoring (QCM-D) was combined with molecular dynamics (MD) simulations to characterise peptide–membrane interactions.

QCM-D revealed distinct analogue-dependent interfacial responses, indicating that lipidation alone did not determine peptide–bilayer interactions. The non-lipidated analogue J211 produced the largest transient decrease in frequency and the greatest increase in dissipation, consistent with formation of a hydrated, mechanically soft, and largely reversible membrane-associated state. Interestingly, the mono-lipidated J229 produced a weaker and reversible response. In contrast, semaglutide and bis-lipidated MEDI7219 induced only small changes in frequency and dissipation, consistent with weak interactions but the membrane perturbation was retained under the studied conditions. In the presence of 2.5 mM sodium caprate (C10), peptide-specific responses were generally attenuated and shifted towards a C10-dominated membrane-perturbation profile, although analogue-dependent differences still remained.

MD simulations showed analogue-dependent membrane engagement, where interaction depth and interaction stability were influenced by lipidation and peptide sequence. MEDI7219 was found to insert itself deepest into the membrane leaflet with stable interactions between the peptide and the lipid bilayer. The lipid tail of semaglutide remained persistently inserted although not as deep as MEDI7219, whereas (the similarly monolipidated) J229 engaged more transiently and reversibly. The non-lipidated J211 only associated with the bilayer surface through aromatic and charged residues. Overall, the lipidated peptides perturbed the local membrane structure the most, providing a molecular-level context for the distinct experimental QCM-D responses.

Together, these findings demonstrate that peptide–membrane interactions are governed not by lipidation alone, but by the combined effects of lipidation pattern, backbone sequence, peptide association state, and formulation conditions. This work provides mechanistic insight relevant to the developability of orally delivered peptides.