Ribonucleotide reductases: divergent evolution of an ancient enzyme

Ribonucleotide reductases (RNRs) are uniquely responsible for converting nucleotides to deoxynucleotides in all dividing cells. The three known classes of RNRs operate through a free radical mechanism but differ in the way in which the protein radical is generated. Class I enzymes depend on oxygen for radical generation, class II uses adenosylcobalamin, and the anaerobic class III requires S-adenosylmethionine and an iron–sulfur cluster. Despite their metabolic prominence, the evolutionary origin and relationships between these enzymes remain elusive. This gap in RNR knowledge can, to a major extent, be attributed to the fact that different RNR classes exhibit greatly diverged polypeptide chains, rendering homology assessments inconclusive. Evolutionary studies of RNRs conducted until now have focused on comparison of the amino acid sequence of the proteins, without considering how they fold into space. The present study is an attempt to understand the evolutionary history of RNRs taking into account their three-dimensional structure. We first infer the structural alignment by superposing the equivalent stretches of the three-dimensional structures of representatives of each family. We then use the structural alignment to guide the alignment of all publicly available RNR sequences. Our results support the hypothesis that the three RNR classes diverged from a common ancestor currently represented by the anaerobic class III. Also, lateral transfer appears to have played a significant role in the evolution of this protein family.     

Uniform binding and negative catalysis at the origin of enzymes

Enzymes are well known for their catalytic abilities, some even reaching “catalytic perfection” in the sense that the reaction they catalyze has reached the physical bound of the diffusion rate. However, our growing understanding of enzyme superfamilies has revealed that only some share a catalytic chemistry while others share a substrate‐handle binding motif, for example, for a particular phosphate group. This suggests that some families emerged through a “substrate‐handle‐binding‐first” mechanism (“binding‐first” for brevity) instead of “chemistry‐first” and we are, therefore, left to wonder what the role of non‐catalytic binders might have been during enzyme evolution. In the last of their eight seminal, back‐to‐back articles from 1976, John Albery and Jeremy Knowles addressed the question of enzyme evolution by arguing that the simplest mode of enzyme evolution is what they defined as “uniform binding” (parallel stabilization of all enzyme‐bound states to the same degree). Indeed, we show that a uniform‐binding proto‐catalyst can accelerate a reaction, but only when catalysis is already present, that is, when the transition state is already stabilized to some degree. Thus, we sought an alternative explanation for the cases where substrate‐handle‐binding preceded any involvement of a catalyst. We find that evolutionary starting points that exhibit negative catalysis can redirect the reaction''s course to a preferred product without need for rate acceleration or product release; that is, if they do not stabilize, or even destabilize, the transition state corresponding to an undesired product. Such a mechanism might explain the emergence of “binding‐first” enzyme families like the aldolase superfamily.     

Ribonucleotide reductase modularity: Atypical duplication of the ATP-cone domain in Pseudomonas aeruginosa

The opportunistic pathogen Pseudomonas aeruginosa, which causes serious nosocomial infections, is a gamma-proteobacterium that can live in many different environments. Interestingly P. aeruginosa encodes three ribonucleotide reductases (RNRs) that all differ from other well known RNRs. The RNR enzymes are central for de novo synthesis of deoxyribonucleotides and essential to all living cells. The RNR of this study (class Ia) is a complex of the NrdA protein harboring the active site and the allosteric sites and the NrdB protein harboring a tyrosyl radical necessary to initiate catalysis. P. aeruginosa NrdA contains an atypical duplication of the N-terminal ATP-cone, an allosteric domain that can bind either ATP or dATP and regulates the overall enzyme activity. Here we characterized the wild type NrdA and two truncated NrdA variants with precise N-terminal deletions. The N-terminal ATP-cone (ATP-c1) is allosterically functional, whereas the internal ATP-cone lacks allosteric activity. The P. aeruginosa NrdB is also atypical with an unusually short lived tyrosyl radical, which is efficiently regenerated in presence of oxygen as the iron ions remain tightly bound to the protein. The P. aeruginosa wild type NrdA and NrdB proteins form an extraordinarily tight complex with a suggested alpha4beta4 composition. An alpha2beta2 composition is suggested for the complex of truncated NrdA (lacking ATP-c1) and wild type NrdB. Duplication or triplication of the ATP-cone is found in some other bacterial class Ia RNRs. We suggest that protein modularity built on the common catalytic core of all RNRs plays an important role in class diversification within the RNR family.     

Benchmarking AlphaFold‐enabled molecular docking predictions for antibiotic discovery

Efficient identification of drug mechanisms of action remains a challenge. Computational docking approaches have been widely used to predict drug binding targets; yet, such approaches depend on existing protein structures, and accurate structural predictions have only recently become available from AlphaFold2. Here, we combine AlphaFold2 with molecular docking simulations to predict protein‐ligand interactions between 296 proteins spanning Escherichia coli''s essential proteome, and 218 active antibacterial compounds and 100 inactive compounds, respectively, pointing to widespread compound and protein promiscuity. We benchmark model performance by measuring enzymatic activity for 12 essential proteins treated with each antibacterial compound. We confirm extensive promiscuity, but find that the average area under the receiver operating characteristic curve (auROC) is 0.48, indicating weak model performance. We demonstrate that rescoring of docking poses using machine learning‐based approaches improves model performance, resulting in average auROCs as large as 0.63, and that ensembles of rescoring functions improve prediction accuracy and the ratio of true‐positive rate to false‐positive rate. This work indicates that advances in modeling protein‐ligand interactions, particularly using machine learning‐based approaches, are needed to better harness AlphaFold2 for drug discovery.     

Completing the family of human Eps15 homology domains: Solution structure of the internal Eps15 homology domain of γ‐synergin

Eps15 homology (EH) domains are universal interaction domains to establish networks of protein–protein interactions in the cell. These networks mainly coordinate cellular functions including endocytosis, actin remodeling, and other intracellular signaling pathways. They are well characterized in structural terms, except for the internal EH domain from human γ‐synergin (EHγ). Here, we complete the family of EH domain structures by determining the solution structure of the EHγ domain. The structural ensemble follows the canonical EH domain fold and the identified binding site is similar to other known EH domains. But EHγ differs significantly in the N‐ and C‐terminal regions. The N‐terminal α‐helix is shortened compared to known homologues, while the C‐terminal one is fully formed. A significant proportion of the remaining N‐ and C‐terminal regions are well structured, a feature not seen in other EH domains. Single mutations in both the N‐terminal and the C‐terminal structured extensions lead to the loss of the distinct three‐dimensional fold and turn EHγ into a molten globule like state. Therefore, we propose that the structural extensions in EHγ function as a clamp and are undoubtedly required to maintain its tertiary fold.     

Benchmarking AlphaFold for protein complex modeling reveals accuracy determinants

High‐resolution experimental structural determination of protein–protein interactions has led to valuable mechanistic insights, yet due to the massive number of interactions and experimental limitations there is a need for computational methods that can accurately model their structures. Here we explore the use of the recently developed deep learning method, AlphaFold, to predict structures of protein complexes from sequence. With a benchmark of 152 diverse heterodimeric protein complexes, multiple implementations and parameters of AlphaFold were tested for accuracy. Remarkably, many cases (43%) had near‐native models (medium or high critical assessment of predicted interactions accuracy) generated as top‐ranked predictions by AlphaFold, greatly surpassing the performance of unbound protein–protein docking (9% success rate for near‐native top‐ranked models), however AlphaFold modeling of antibody–antigen complexes within our set was unsuccessful. We identified sequence and structural features associated with lack of AlphaFold success, and we also investigated the impact of multiple sequence alignment input. Benchmarking of a multimer‐optimized version of AlphaFold (AlphaFold‐Multimer) with a set of recently released antibody–antigen structures confirmed a low rate of success for antibody–antigen complexes (11% success), and we found that T cell receptor–antigen complexes are likewise not accurately modeled by that algorithm, showing that adaptive immune recognition poses a challenge for the current AlphaFold algorithm and model. Overall, our study demonstrates that end‐to‐end deep learning can accurately model many transient protein complexes, and highlights areas of improvement for future developments to reliably model any protein–protein interaction of interest.     

The periodic table of ribonucleotide reductases

In most organisms, transition metal ions are necessary cofactors of ribonucleotide reductase (RNR), the enzyme responsible for biosynthesis of the 2′-deoxynucleotide building blocks of DNA. The metal ion generates an oxidant for an active site cysteine (Cys), yielding a thiyl radical that is necessary for initiation of catalysis in all RNRs. Class I enzymes, widespread in eukaryotes and aerobic microbes, share a common requirement for dioxygen in assembly of the active Cys oxidant and a unique quaternary structure, in which the metallo- or radical-cofactor is found in a separate subunit, β, from the catalytic α subunit. The first class I RNRs, the class Ia enzymes, discovered and characterized more than 30 years ago, were found to use a diiron(III)-tyrosyl-radical Cys oxidant. Although class Ia RNRs have historically served as the model for understanding enzyme mechanism and function, more recently, remarkably diverse bioinorganic and radical cofactors have been discovered in class I RNRs from pathogenic microbes. These enzymes use alternative transition metal ions, such as manganese, or posttranslationally installed tyrosyl radicals for initiation of ribonucleotide reduction. Here we summarize the recent progress in discovery and characterization of novel class I RNR radical-initiating cofactors, their mechanisms of assembly, and how they might function in the context of the active class I holoenzyme complex.     

Inhibition of chlamydial class Ic ribonucleotide reductase by C‐terminal peptides from protein R2

Chlamydia trachomatis ribonucleotide reductase (RNR) is a class Ic RNR. It has two homodimeric subunits: proteins R1 and R2. Class Ic protein R2 in its most active form has a manganese–iron metal cofactor, which functions in catalysis like the tyrosyl radical in classical class Ia and Ib RNRs. Oligopeptides with the same sequence as the C‐terminus of C. trachomatis protein R2 inhibit the catalytic activity of C. trachomatis RNR, showing that the class Ic enzyme shares a similar highly specific inhibition mechanism with the previously studied radical‐containing class Ia and Ib RNRs. The results indicate that the catalytic mechanism of this class of RNRs with a manganese–iron cofactor is similar to that of the tyrosyl‐radical‐containing RNRs, involving reversible long‐range radical transfer between proteins R1 and R2. The competitive binding of the inhibitory R2‐derived oligopeptide blocks the transfer pathway. We have constructed three‐dimensional structure models of C. trachomatis protein R1, based on homologous R1 crystal structures, and used them to discuss possible binding modes of the peptide to protein R1. Typical half maximal inhibitory concentration values for C. trachomatis RNR are about 200 µ m for a 20‐mer peptide, indicating a less efficient inhibition compared with those for an equally long peptide in the Escherichia coli class Ia RNR. A possible explanation is that the C. trachomatis R1/R2 complex has other important interactions, in addition to the binding mediated by the R1 interaction with the C‐terminus of protein R2. Copyright © 2011 European Peptide Society and John Wiley & Sons, Ltd.     

ScanMoment: a web server for combinatorial analysis of basic residues in nucleic acid binding sites

ScanMoment is a webserver designed to identify the presence of the basic faced α‐helix (BFAH) motif in the nucleic acid binding sites of proteins. The program calculates the ’Basic Moment‘, a parameter that quantitizes the distribution of basic residues on the surface of an α‐helix. A sliding window is used to generate a plot displaying regions of the protein sequence that possesses a high Basic Moment and hus likely to possess a BFAH motif. The user may vary the periodicity from that of an alpha‐helix (100°), to those of other secondary structures such as beta sheets and 3₁₀ helices. The program can also plot the periodicity of basic residues in a protein sequence using a Fourier transformation. The procedure has been used to characterize the presence of BFAHs in the N‐terminal extensions of the eukaryotic aminoacyl‐tRNA synthetases and to indicate the presence of a BFAH in the tRNA binding site of alanyl‐tRNA synthetase.     

Membranome 3.0: Database of single‐pass membrane proteins with AlphaFold models

The Membranome database provides comprehensive structural information on single‐pass (i.e., bitopic) membrane proteins from six evolutionarily distant organisms, including protein–protein interactions, complexes, mutations, experimental structures, and models of transmembrane α‐helical dimers. We present a new version of this database, Membranome 3.0, which was significantly updated by revising the set of 5,758 bitopic proteins and incorporating models generated by AlphaFold 2 in the database. The AlphaFold models were parsed into structural domains located at the different membrane sides, modified to exclude low‐confidence unstructured terminal regions and signal sequences, validated through comparison with available experimental structures, and positioned with respect to membrane boundaries. Membranome 3.0 was re‐developed to facilitate visualization and comparative analysis of multiple 3D structures of proteins that belong to a specified family, complex, biological pathway, or membrane type. New tools for advanced search and analysis of proteins, their interactions, complexes, and mutations were included. The database is freely accessible at https://membranome.org.     

The crystal structure of class II ribonucleotide reductase reveals how an allosterically regulated monomer mimics a dimer

Ribonucleotide reductases (RNRs) catalyze the conversion of ribonucleotides to deoxyribonucleotides, an essential step in DNA biosynthesis and repair. Here we present the crystal structure of class II (coenzyme B12-dependent) ribonucleoside triphosphate reductase (RTPR) from Lactobacillus leichmannii in the apo enzyme form and in complex with the B12 analog adeninylpentylcobalamin at 1.75 and 2.0 A resolution, respectively. This monomeric, allosterically regulated class II RNR retains all the key structural features associated with the catalytic and regulatory machinery of oligomeric RNRs. Surprisingly, the dimer interface responsible for effector binding in class I RNR is preserved through a single 130-residue insertion in the class II structure. Thus, L. leichmannii RNR is a paradigm for the simplest structural entity capable of ribonucleotide reduction, a reaction linking the RNA and DNA worlds.     

Structural determinants and distribution of phosphate specificity in ribonucleotide reductases

Ribonucleotide reductases (RNRs) catalyze the reduction of ribonucleotides to the corresponding deoxyribonucleotides, the building blocks of DNA. RNRs are specific for either ribonucleoside diphosphates or triphosphates as substrates. As far as is known, oxygen-dependent class I RNRs (NrdAB) all reduce ribonucleoside diphosphates, and oxygen-sensitive class III RNRs (NrdD) are all ribonucleoside triphosphate reducers, whereas the adenosylcobalamin-dependent class II (NrdJ) contains both ribonucleoside diphosphate and triphosphate reducers. However, it is unknown how this specificity is conveyed by the active site of the enzymes and how this feature developed in RNR evolution. By structural comparison of the active sites in different RNRs, we identified the apical loop of the phosphate-binding site as a potential structural determinant of substrate specificity. Grafting two residues from this loop from a diphosphate- to a triphosphate-specific RNR caused a change in preference from ribonucleoside triphosphate to diphosphate substrates in a class II model enzyme, confirming them as the structural determinants of phosphate specificity. The investigation of the phylogenetic distribution of this motif in class II RNRs yielded a likely monophyletic clade with the diphosphate-defining motif. This indicates a single evolutionary-split event early in NrdJ evolution in which diphosphate specificity developed from the earlier triphosphate specificity. For those interesting cases where organisms contain more than one nrdJ gene, we observed a preference for encoding enzymes with diverse phosphate specificities, suggesting that this varying phosphate specificity confers a selective advantage.     

NMR hawk‐eyed view of AlphaFold2 structures

The prediction of the three‐dimensional (3D) structure of proteins from the amino acid sequence made a stunning breakthrough reaching atomic accuracy. Using the neural network‐based method AlphaFold2, 3D structures of almost the entire human proteome have been predicted and made available (https://www.alphafold.ebi.ac.uk). To gain insight into how well AlphaFold2 structures represent the conformation of proteins in solution, I here compare the AlphaFold2 structures of selected small proteins with their 3D structures that were determined by nuclear magnetic resonance (NMR) spectroscopy. Proteins were selected for which the 3D solution structures were determined on the basis of a very large number of distance restraints and residual dipolar couplings and are thus some of the best‐resolved solution structures of proteins to date. The quality of the backbone conformation of the AlphaFold2 structures is assessed by fitting a large set of experimental residual dipolar couplings (RDCs). The analysis shows that experimental RDCs fit extremely well to the AlphaFold2 structures predicted for GB3, DinI, and ubiquitin. In the case of GB3, the accuracy of the AlphaFold2 structure even surpasses that of a 1.1 Å crystal structure. Fitting of experimental RDCs furthermore allows identification of AlphaFold2 structures that are best representative of the protein''s conformation in solution as seen for the EF hands of the N‐terminal domain of Ca²⁺‐ligated calmodulin. Taken together, the analysis shows that structures predicted by AlphaFold2 can be highly representative of the solution conformation of proteins. The combination of AlphaFold2 structures with RDCs promises to be a powerful approach to study structural changes in proteins.     

Starting a new chapter on class Ia ribonucleotide reductases

Ribonucleotide reductases (RNRs) use radical-based chemistry to convert ribonucleotides into deoxyribonucleotides, an essential step in DNA biosynthesis and repair. There are multiple RNR classes, the best studied of which is the class Ia RNR that is found in Escherichia coli, eukaryotes including humans, and many pathogenic and nonpathogenic prokaryotes. This review covers recent advances in our understanding of class Ia RNRs, including a recent reporting of a structure of the active state of the E. coli enzyme and the impacts that the structure has had on spurring research into the mechanism of long-range radical transfer. Additionally, the review considers other recent structural and biochemical research on class Ia RNRs and the potential of that work for the development of anticancer and antibiotic therapeutics.     

Spectroscopic and theoretical approaches for studying radical reactions in class I ribonucleotide reductase

Ribonucleotide reductases (RNRs) catalyze the production of deoxyribonucleotides, which are essential for DNA synthesis and repair in all organisms. The three currently known classes of RNRs are postulated to utilize a similar mechanism for ribonucleotide reduction via a transient thiyl radical, but they differ in the way this radical is generated. Class I RNR, found in all eukaryotic organisms and in some eubacteria and viruses, employs a diferric iron center and a stable tyrosyl radical in a second protein subunit, R2, to drive thiyl radical generation near the substrate binding site in subunit R1. From extensive experimental and theoretical research during the last decades, a general mechanistic model for class I RNR has emerged, showing three major mechanistic steps: generation of the tyrosyl radical by the diiron center in subunit R2, radical transfer to generate the proposed thiyl radical near the substrate bound in subunit R1, and finally catalytic reduction of the bound ribonucleotide. Amino acid- or substrate-derived radicals are involved in all three major reactions. This article summarizes the present mechanistic picture of class I RNR and highlights experimental and theoretical approaches that have contributed to our current understanding of this important class of radical enzymes.