Discovery, Design, and Structural Characterization of Alkane-Producing Enzymes across the Ferritin-like Superfamily.

To complement established rational and evolutionary protein design approaches, significant efforts are being made to utilize computational modeling and the diversity of naturally occurring protein sequences. Here, we combine structural biology, genomic mining, and computational modeling to identify structural features critical to aldehyde deformylating oxygenases (ADOs), an enzyme family that has significant implications in synthetic biology and chemoenzymatic synthesis. Through these efforts, we discovered latent ADO-like function across the ferritin-like superfamily in various species of Bacteria and Archaea. We created a machine learning model that uses protein structural features to discriminate ADO-like activity. Computational enzyme design tools were then utilized to introduce ADO-like activity into the small subunit of Escherichia coli class I ribonucleotide reductase. The integrated approach of genomic mining, structural biology, molecular modeling, and machine learning has the potential to be utilized for rapid discovery and modulation of functions across enzyme families.

A new benchmark illustrates that integration of geometric constraints inferred from enzyme reaction chemistry can increase enzyme active site modeling accuracy.

Enzymes play a critical role in a wide array of industrial, medical, and research applications and with the recent explosion of genomic sequencing, we now have sequences for millions of enzymes for which there is no known structure. In order to utilize modern computational design tools for constructing inhibitors or engineering novel catalysts, the ability to accurately model enzymes is critical. A popular approach for modeling enzymes are comparative modeling techniques which can often accurately predict the global structural features. However, achieving atomic accuracy of an active site remains a challenge and is an issue when trying to utilize the molecular details for designing inhibitors or enhanced catalysts. Here we explore integrating knowledge about the required geometric orientation of conserved catalytic residues into the comparative modeling process in order to improve modeling accuracy. In order to investigate the utility of adding this information, we first carefully construct a benchmark set of reference structures to use. Consistent with previous findings, our benchmark demonstrates that the geometry between catalytic residues across an enzyme family is conserved and does not tend to deviate by more than 0.5Å. We then find that by integrating these geometric constraints during modeling, we can double the number of atomic level accuracy models (<1Å RMSD to the crystal structure ligand) within our benchmarking dataset, even for targets with templates as low as 20-30% sequence identity. Catalytic residues within an enzyme family are highly conserved and can often be readily identified through comparative sequence analysis to a known structure within the enzyme family. Therefore utilizing this readily available information has the potential to significantly improve drug design and enzyme engineering efforts for which there is no known structure for the enzyme of interest.

A Benchmark for Homomeric Enzyme Active Site Structure Prediction Highlights the Importance of Accurate Modeling of Protein Symmetry.

Accurate prediction and modeling of an enzyme's active site are critical for engineering efforts as well as providing insight into an enzyme's naturally occurring function. Previous efforts demonstrated that the integration of constraints enforcing strict geometric orientations between catalytic residues significantly improved the modeling accuracy for the active sites of monomeric enzymes. In this study, a similar approach was explored to evaluate the effect on the active sites of homomeric enzymes. A benchmark of 17 homomeric enzymes with known structures and a bound ligand relevant to the established chemistry were identified from the protein data bank. The enzymes identified span multiple classes as well as symmetries. Unlike what was observed for the monomeric enzymes, upon the application of catalytic geometric constraints, there was no significant improvement observed in modeling accuracy for either the active site of the protein structure or the accuracy of the subsequently docked ligand. Upon further analysis, it is apparent that the symmetric interface being modeled is inaccurate and prevented the active sites from being modeled at atomic-level accuracy. This is consistent with the challenge others have identified in being able to predict de novo protein symmetry. To further improve the accuracy of active site modeling for homomeric proteins, new methodologies to accurately model the symmetric interfaces of these complexes are needed.

The Zinc Linchpin Motif in the DNA Repair Glycosylase MUTYH: Identifying the Zn2+ Ligands and Roles in Damage Recognition and Repair.

The DNA base excision repair (BER) glycosylase MUTYH prevents DNA mutations by catalyzing adenine (A) excision from inappropriately formed 8-oxoguanine (8-oxoG):A mismatches. The importance of this mutation suppression activity in tumor suppressor genes is underscored by the association of inherited variants of MUTYH with colorectal polyposis in a hereditary colorectal cancer syndrome known as MUTYH-associated polyposis, or MAP. Many of the MAP variants encompass amino acid changes that occur at positions surrounding the two-metal cofactor-binding sites of MUTYH. One of these cofactors, found in nearly all MUTYH orthologs, is a [4Fe-4S]2+ cluster coordinated by four Cys residues located in the N-terminal catalytic domain. We recently uncovered a second functionally relevant metal cofactor site present only in higher eukaryotic MUTYH orthologs: a Zn2+ ion coordinated by three Cys residues located within the extended interdomain connector (IDC) region of MUTYH that connects the N-terminal adenine excision and C-terminal 8-oxoG recognition domains. In this work, we identified a candidate for the fourth Zn2+ coordinating ligand using a combination of bioinformatics and computational modeling. In addition, using in vitro enzyme activity assays, fluorescence polarization DNA binding assays, circular dichroism spectroscopy, and cell-based rifampicin resistance assays, the functional impact of reduced Zn2+ chelation was evaluated. Taken together, these results illustrate the critical role that the "Zn2+ linchpin motif" plays in MUTYH repair activity by providing for proper engagement of the functional domains on the 8-oxoG:A mismatch required for base excision catalysis. The functional importance of the Zn2+ linchpin also suggests that adjacent MAP variants or exposure to environmental chemicals may compromise Zn2+ coordination, and ability of MUTYH to prevent disease.

Computational Introduction of Catalytic Activity into Proteins.

Recently, there have been several successful cases of introducing catalytic activity into proteins. One method that has been used successfully to achieve this is the theozyme placement and enzyme design algorithms implemented in Rosetta Molecular Modeling Suite. Here, we illustrate how to use this software to recapitulate the placement of catalytic residues and ligand into a protein using a theozyme, protein scaffold, and catalytic constraints as input.

Integrative genomic mining for enzyme function to enable engineering of a non-natural biosynthetic pathway.

The ability to biosynthetically produce chemicals beyond what is commonly found in Nature requires the discovery of novel enzyme function. Here we utilize two approaches to discover enzymes that enable specific production of longer-chain (C5-C8) alcohols from sugar. The first approach combines bioinformatics and molecular modelling to mine sequence databases, resulting in a diverse panel of enzymes capable of catalysing the targeted reaction. The median catalytic efficiency of the computationally selected enzymes is 75-fold greater than a panel of naively selected homologues. This integrative genomic mining approach establishes a unique avenue for enzyme function discovery in the rapidly expanding sequence databases. The second approach uses computational enzyme design to reprogramme specificity. Both approaches result in enzymes with >100-fold increase in specificity for the targeted reaction. When enzymes from either approach are integrated in vivo, longer-chain alcohol production increases over 10-fold and represents >95% of the total alcohol products.