LIFE SCIENCES

Project

Chiral amines, a group of small chemicals, are central building blocks to a variety of fine chemical products. These include agrochemicals and pharmaceuticals such as Sitagliptin, a potent drug used to treat type II diabetes. Accordingly, biotech and pharmaceutical companies are highly interested in the efficient and sustainable production of these compounds. A group of enzymes already in use to fill this need are Transaminases (TAs). As biocatalysts, they can catalyze the amination of small chemicals at room temperature and atmospheric pressure using the automatically regenerated co-factor pyridoxal-5-phosphate (PLP). In contrast, a chemical synthesis of the desired products would generally require high temperatures and pressures coupled with toxic heavy metal catalysts. One major challenge in using TAs industrially is that the introduction of new substrates requires significant experimental enzyme engineering efforts to find a balance between TA stability and activity towards the desired substrates.

In this project, Prof. Dr. Gohlke and Steffen Docter investigated the thermal unfolding behavior of two sets of TA variants of fold type I and IV families of PLP-dependant enzymes by simulating rigid cluster decompositions using Constraint Network Analysis (CNA). A deeper understanding of this process will help researchers to identify specific mutations to optimize both enzymes for stability, instead of relying on the screening of large mutant libraries experimentally.

The team generated structural models for 50 TA variants of each fold type family using AlphaFold2 and performed Constraint Network Analysis on large ensembles of network topologies extracted from extensive molecular dynamics simulations. Using the GAUSSIAN09 software, partial charges of the covalently bound PLP were calculated. Solvated simulation boxes for all 100 TA variants with covalent PLP were created using the AmberTools22 software package, leading to system sizes of 100,000-130,000 atoms. The team used the GPU particle mesh Ewald implementation of the AMBER22 molecular dynamics simulation suite to perform a combined total of 400 ms of simulation time over the two sets of TA variants, which was made possible using the highly efficient A100 GPUs on the JUWELS Booster partition.

The ensembles of network topologies were analyzed with CNA and, interestingly, revealed different thermal unfolding behaviors of the two enzyme families, as shown for two representative TA variants in Figure 1. The (R)-selective fold type IV TA variants follow a two-step unfolding process. Here, the first major loss of rigid contacts upon heating occurs between residues of the homodimer interface, followed by individual unfolding of the monomeric constituents. Contrary to that, the simulated unfolding of the (S)-selective fold type I TA variants starts at the surface of the homodimer and evolves towards the enzyme core, including the dimer interface as a rigid cluster. This information is helpful in informing enzyme engineering campaigns aiming for improved TA stability to focus mutations on areas with early unfolding events over more rigid residue clusters.

In further work, building upon the findings above, the team investigated conformational changes for different solvent pH values based on extensive constant pH molecular dynamics simulations and designed several potentially stabilizing variants for both enzyme families. They applied different design approaches, focusing on the introduction of disulfide bridges in the dimer interface and the identification and stabilization of specific structural weak spot residues, and developed integrated enzyme design workflows incorporating state-of-the-art deep learning models such as ProteinMPNN and SaProt.