Deep learning approaches for conformational flexibility and switching properties in protein design

Following the hugely successful application of deep learning methods to protein structure prediction, an increasing number of design methods seek to leverage generative models to design proteins with improved functionality over native proteins or novel structure and function. The inherent flexibility of proteins, from side-chain motion to larger conformational reshuffling, poses a challenge to design methods, where the ideal approach must consider both the spatial and temporal evolution of proteins in the context of their functional capacity. In this review, we highlight existing methods for protein design before discussing how methods at the forefront of deep learning-based design accommodate flexibility and where the field could evolve in the future.

Explicit solvent effects on protein physics - modeling lattice proteins in aqueous environments

Protein folding stands as one of the major interdisciplinary challenges of the last ten years, involving biology, chemistry, medicine and physics. In this PhD thesis, we investigate the solvent effects and the resulting hydrophobic effect on proteins. In particular, we focus on a simple lattice model of proteins (the HPW model), where the solvent is (semi)explicitly taken into account (by means of the model of Muller, Lee and Graziano (MLG)), investigating thermodynamical quantities of the modeled proteins and comparing the results with those of real proteins. We approach the protein design problem using the HPW model and we investigate some statistical and thermodynamic properties of the designed sequences. In particular we show that the HPW model is able to capture, within a single framework, both warm and cold denaturation of proteins. We further adapt the MLG model model to include chaotrope cosolvents in the solution. We show that the alteration of the network of hydrogen bonds between water molecules due to the presence of chaotropic agents induces destabilization of proteins. Moreover, we find that the presence of chaotropes originates an effective interaction between the chaotropes and the protein. In addition, we use a common method to extract effective amino-acid interactions on real proteins: a perceptron algorithm. The method applied to proteins designed with the HPW model reveals that the solvent effects can not be reproduced by effective residue interactions. Indeed, we show that it is not possible to determine a set of global interaction values able to stabilize all proteins at once.

EPFL2004

A general method for the covalent labeling of fusion proteins with small molecules in vivo

Thomas Gronemeyer, Kai Johnsson, Antje Keppler, Horst Pick, Horst Vogel

Characterizing the movement, interactions, and chem. microenvironment of a protein inside the living cell is crucial to a detailed understanding of its function. Most strategies aimed at realizing this objective are based on genetically fusing the protein of interest to a reporter protein that monitors changes in the environment of the coupled protein. Examples include fusions with fluorescent proteins, the yeast two-hybrid system, and split ubiquitin. However, these techniques have various limitations, and considerable effort is being devoted to specific labeling of proteins in vivo with small synthetic mols. capable of probing and modulating their function. These approaches are currently based on the noncovalent binding of a small mol. to a protein, the formation of stable complexes between biarsenical compds. and peptides contg. cysteines, or the use of biotin acceptor domains. Here we describe a general method for the covalent labeling of fusion proteins in vivo that complements existing methods for noncovalent labeling of proteins and that may open up new ways of studying proteins in living cells.

2003

Transient Gene Expression in HEK-293E Cells for Recombinant Protein Production

Sophie Nallet

The growing demand of biopharmaceutical products is boosting the market for therapeutic recombinant proteins (r-proteins). More than half of the 140 r-proteins that have gained approval for human therapeutic use are manufactured in mammalian cells that make possible complex post-translational modifications, like glycosylation. In the industry, r-proteins are manufactured by stable gene expression (SGE) following Good Manufacturing Practice (GMP) standards. Transient gene expression (TGE) in mammalian cells is an alternative method for the expression of r-proteins whose main advantages are rapidity and flexibility. TGE involves the expression of the protein of interest from plasmids, which are transfected into the cells with a non-viral DNA transfecting agent, usually polyethylenimine (PEI). DNA is transcribed from the plasmids without the latter having been integrated into the host genome. Despite the recent improvements in r-protein titers by TGE and the scale up of this method to the 100-L scale, TGE remains so far a tool for protein production for research purposes. Therefore, to determine if TGE can be used for manufacturing therapeutic r-protein following GMP standards, different features related to reproducibility and product quality from TGE of a recombinant anti-rhesus D immunoglobulin G in HEK-293E cells using linear PEI were characterized. To test the effects of the size distribution and chemical structure of PEI on TGE, different commercially available PEIs were characterized and their impact on transfection and r-protein expression were assessed. The results showed that both the molecular weight distribution and the chemical structure of the PEI had an effect on TGE. However, the small variations in size distribution and structure between the batches of PEI from the same supplier did not affect significantly the transfection and r-protein production by TGE. Then, the fate of PEI and plasmid DNA in cell culture over time and their removal from the final r-protein during purification were determined and measured. Most of the PEI and pDNA remained in the cell culture medium after transfection. However, both PEI and plasmid DNA, which are additional components in TGE compared to processes based on recombinant cell lines, could be reduced to a concentration below the limit of detection of the assays after Protein A affinity purification of the antibody. To test the reproducibility of a TGE process, 10 batches of transfected cells were run in 500-mL orbitally shaken bottles. The cell specific productivity of the antibody was 20.2 ± 2.6 pg.cell-1.day-1 on average for the 10 batches. The purity and size consistency of the recombinant antibody produced in those 10 batches was demonstrated by gel electrophoresis and size exclusion chromatography. Then, the glycosylation profile of the antibody produced by TGE in HEK-293E cells was determined by mass spectrometry and was reproducible over the 10 batches. Moreover, it was similar to the glycosylation profile of the antibody produced by 3 stable HEK-293E cell lines. Finally, as a proof of concept, TGE was applied for the development of a vaccine against the respiratory syncytial virus. The fusion protein of the respiratory syncytial virus was produced by TGE and used as an antigen for the development of a recombinant vaccine. TGE enabled the rapid evaluation of the candidate vaccine in animal trials. Overall, the results of this study suggest that TGE is a reliable and reproducible method for manufacturing r-proteins of a consistent quality, provided that some particular features, such as reagents quality and process parameters, are well defined and controlled. Since TGE makes possible the rapid production of virtually any protein, even toxic, it could be used to provide GMP approved biopharmaceuticals for clinical trials in a short time, reducing development costs of biological therapeutic products.

EPFL2010

Transient Gene Expression in HEK-293E Cells for Recombinant Protein Production

Sophie Nallet

EPFL2010