Solution behaviour of Human Serum Albumin and GLP-1variants

Abstract

The formation of macromolecular complexes and the solution behaviour of proteins are results of protein-protein interactions (PPI) and the nature of the solution components. PPI’s can be divided into two broad categories of attractive and repulsive forces. From a formulation perspective, proper interaction is critical for the long term stability of a pharmaceutical. Protein complex formation is important for extended half-life in vivo and is essential to cellular communication such as the induction of the insulin response. This thesis focuses on human serum albumin (HSA) as a central player and the ability of small-angle X-ray scattering (SAXS) to study proteins under diverse solution conditions. HSA is utilized in many ways in the pharmaceutical industry such as the formulations of other proteins and in particular for half-life extension of peptides, where the long half-life of up to 21 days of HSA is deployed. The thesis is divided into four parts: 1) Self-interaction of HSA in pharmaceutical relevant systems; 2) The liraglutide oligomer and interaction with HSA; 3) Conjugation of GLP-1 to HSA and last but not least 4) A benchmark study of the implementation of SAXS in RosettaDock. HSA is used as a stabilizer in formulation where it is known to decrease adsorption to the air/liquid and liquid/container interfaces, thereby reducing the aggregation propensity of the target protein. However, with high concentration liquid formulations coming into focus limited information is available on the behaviour of HSA in these crowded environments. We have applied SAXS in order to shed light on the self-interaction of HSA in pharmaceutical relevant buffer systems on protein concentrations up to 150 mg/ml. HSA stabilized by octanoate is observed to interact by purely repulsive forces in the investigated concentration ranges while defatted HSA show attractive forces at low protein concentration. The tonicity corresponding to isotonicity, is observed to coincide with the maximum screening observed in all systems while this is not observed as a function of ionic strength, making tonicity an important parameter for protein interaction in the investigated systems. Trehalose is seen to provide a screening effect of added NaCl leading us to think that trehalose somehow determines the range of interaction of the proteins in solution. Finally it is proposed that the stabilizing effect of HSA could be mediated by a repulsive network of HSA molecules screening the interaction of other proteins, hereby decreasing the aggregation propensity. Glucagon-like peptide 1 (GLP-1) is an incretin hormone used in the treatment of diabetes mellitus type 2. Due to its very limited half-life of approximately 3 minutes, Novo Nordisk A/S developed an acylated analogue, liraglutide, which due to multimerization and probable interaction with HSA in-vivo, makes it applicable for once daily administration. The oligomerization and the interaction with HSA were investigated by SAXS and static light scattering (SLS). The oligomeric state of liraglutide was approximately heptameric but due to uncertainties, the experimental studies were complimented by a series of molecular dynamic (MD) simulations. The results from MD where compared to the experimental SAXS data and the combined effort show that the liraglutide heptamer was the most probable multimer. The studies of liraglutide with HSA both in a defatted form and a stabilized form, were capable of confirming the presence of a hetero complex though at very low concentration. This could be a result of very transient interaction or the consequence of non-optimal buffer conditions. Another approach to half-life extension is conjugation of molecules to HSA. In this part of the thesis, novel GLP-1-albumin conjugates developed by Albumedix A/S where examined by a combined approach of pharmacokinetic studies and solution structure determination with SAXS. GLP-1 was conjugated to Cys34 of recombinant HSA (rHSA) and two rHSA variants with lower (NB) and higher binding (HB) affinity to the neonatal Fc receptor (FcRn). Binding kinetics showed that the conjugation had limited effect on the binding properties of the conjugates to FcRn compared to the respective rHSA variants. Increased in-vivo half-life of the conjugates was observed in NMRI WT mice compared to GLP-1, with the HB-variant displaying ~300 times longer half-life, while the potency of GLP-1-albumin was decreased by 3-4 orders of magnitude compared to GLP-1. The solution structure of the rHSA variants and the conjugates indicate a flexible nature of the conjugate, with the GLP-1 pointing away from the surface of rHSA. The low resolution structure from SAXS combined with high resolution structural information from X-ray crystallography, explains the pharmacokinetics results, as limited interference is seen between FcRn and the conjugate, while albumin steric hindrance explains the decreased potency of the conjugates. Structure determination of macromolecular complexes can be challenging by traditional approaches such as X-ray crystallography and NMR. An alternative approach is the use of experimental knowledge in combination with in-silico modelling to gain knowledge about macromolecular complexes. The final part of the thesis regards the benchmarking of the protein docking tool, RosettaDock, in combination with SAXS. The RosettaDock protocol is a two-step process involving a low resolution rigid body protocol, set up to imitate the initial protein-protein encounter, followed by a high resolution protocol where the position of the proteins and amino acid side chains are optimized. We applied SAXS as constrain (SAXSconstrain) after the low resolution step to filter out complexes with overall shapes which would not match the SAXS data. Since it is a rigid body approach larger conformational changes in any part of the complex provides a limit to the method either from interface non-complementarity or a resulting shape which deviates from the SAXS data. 38 structures from Benchmark 4.0 of different difficulty levels were investigated using this approach. In general our result indicate that implementation of SAXSconstrain reduces the sampling space and increases the probability of finding a near-native structure. In a wider perspective, the strength of RosettaDockSAXS lies in the combination of low-resolution structural information from SAXS experiments and the protein-protein interaction energies obtained from RosettaDock which provides means to gain insight to higher resolution information about the interface between two protein partners. This allows for the prediction of unknown three-dimensional atomic structures of protein-protein complexes.