Non-host class II ribonucleotide reductase in Thermus viruses: sequence adaptation and host interaction

Bioinformatics

All NrdJm amino acid sequences from NCBI’s RefSeq database were downloaded from RNRdb (http://rnrdb.pfitmap.org) and clustered with USEARCH (Edgar, 2010) at 90% sequence identity to reduce redundancy. Non full-length sequences and sequences of dubious quality were manually removed, before aligning all 364 sequences with ProbCons (Do et al., 2005). Trustworthy alignment positions were selected with the BMGE algorithm (Criscuolo & Gribaldo, 2010) using the BLOSUM30 substitution matrix, ending up with 363 well-aligned positions forming 350 distinct alignment patterns. RAxML version 8.2.4 (Stamatakis, 2014) was used to estimate a phylogeny, using the PROTGAMMAAUTO model, rapid bootstrapping with the autoMRE bootstopping followed by a full maximum likelihood tree search. The phylogeny is available at Figshare (https://doi.org/10.17045/sthlmuni.7117430.v1).

To search for sequences from metagenomes, we downloaded all TARA Ocean ORFs (https://doi.org/10.6084/m9.figshare.4902917, (Delmont et al., 2018)), all ORFs from the Human Microbiome Project (2017-01-09; (Human Microbiome Project C, 2012a; Human Microbiome Project C, 2012b)) the majority of archaeal and bacterial metagenome assembled genomes (MAGs) and single amplified genomes (SAGs) from IMG/MER (4910 MAGs, 2230 SAGs) plus 53 aquatic and soil metagenomes, in particular those with project names containing “virus”, “phage”, “therm” or “hot“ (SI) (Markowitz et al., 2008). Together, we downloaded a total of 250,881,638 ORFs. We used hmm profiles designed for each clan in the phylogeny to search the sequences. We found 181 sequences with a best match to the profile designed from the TV clan. These were aligned to the original alignment using Clustal Omega in profile mode (Sievers et al., 2011) and phylogenetically placed in the full phylogeny with RAxML (Stamatakis, 2014) (https://doi.org/10.17045/sthlmuni.7642343.v1).

A likelihood ratio test of significant overrepresentation of non-synonymous to silent (dN/dS) was estimated with codonml from PAML (Stamatakis, 2014; Yang, 1997; Yang, 2007) by running the program with a fixed and free dN/dS respectively, and the branch leading to the two Thermus virus sequences (marked with “#1” in Fig. 1B) designated as the “foreground” branch in the free dN/dS run. This analysis was performed on the subtree in Fig. 1B, using the same alignment as for the full protein sequence tree, reverse translated into the correct nucleotide sequence for each protein sequence. A χ² p-value of the likelihood ratio test was calculated with one degree of freedom. We could not find correct gene sequences for seven taxa (RefSeq accession numbers: WP_102410887, WP_033167051, WP_054955013, WP_065068364, WP_088370373, WP_087372021, WP_093315575), so they were left out of the analysis (https://doi.org/10.17045/sthlmuni.7642463.v1).

The similarity of codon usage as euclidean distance between the codon frequencies was calculated as described elsewhere (Popa, Landan & Dagan, 2017).

The model of the TV P74-26 nrdJm (TVNrdJm) was constructed with SwissModel (Waterhouse et al., 2018) using the Lactobacillus leichmannii monomeric NrdJ as template (PDB: 1L1L) (Sintchak et al., 2002) and default parameters.

Recombinant gene expression and purification

Recombinant expression of the TV P74-26 nrdJm (UniprotKB: A7XXH5) was performed with a synthetic gene in a pET28b(+) expression vector and E. coli BL21(DE3) expression strain. 800 mL of LB-medium, containing 30 µg mL⁻¹ kanamycin, were inoculated from a preculture to a cell density of OD₆₀₀ = 0.1 and incubated at 37 °C and 130 RPM. At an optical density of 1.0 expression was induced by addition of IPTG to a final concentration of 0.1 mmol L⁻¹. Expression was performed at 37 °C and 130 RPM for 4 h. After cell lysis with lysozyme and sonication, the enzyme was purified by Ni-NTA affinity chromatography followed by size exclusion chromatography.

Thermus thermophilus thioredoxin gene trx1 (UniprotKB: Q72HU9) and the thioredoxin reductase tr gene (UniprotKB: Q72HD8) were cloned from genomic DNA in a pET52b(+) expression vector with C-terminal His-Tag. Expression was performed in a E. coli BL21(DE3) expression strain. For all genes, 800 mL of LB-medium, containing 100 µg mL⁻¹ ampicillin, were inoculated from a preculture to a cell density of OD₆₀₀ = 0.1 and incubated at 37 °C and 130 RPM. At an optical density of 1.0 expression was induced by addition of IPTG to a final concentration of 0.1 µmol L⁻¹. Expression was performed at 130 RPM for 16 h at 37 °C for Trx1 and at 20 °C for TR and. After cell lysis with lysozyme and sonication, the enzyme was purified by Ni-NTA affinity chromatography followed by desalting. Details of the purification procedure are given in the supporting information. An SDS-PAGE showing the purified NrdJm, Trx1 and TR is attached in the supporting information (Fig. S1).

Analytical size exclusion chromatography was performed on a Superdex™ 200 10/300 GL column. 500 µg of TV P74-26 NrdJm were loaded on the column and eluted with SEC buffer (50 mmol L⁻¹ tris–HCl, 300 mmol L⁻¹ NaCl, 1 mmol L⁻¹DTT, pH = 8.0) with a flow rate of 0.2 ml min⁻¹.

Ribonucleotide reductase activity assays

If not stated differently, RNR activity was measured for the conversion of the substrate GTP to the corresponding product dGTP in the presence of a dTTP effector. The standard setup of the assay contains 50 mmol L⁻¹ Tris (pH = 8.0), 50 mmol L⁻¹ MgCl₂, 50 mmol L⁻¹ Dithiothreitol (DTT), 1 mmol L⁻¹ dTTP, 1 mmol L⁻¹ GTP, 10 µmol L⁻¹ AdoCbl and 0.1–0.4 µmol L⁻¹ NrdJm. In assays containing redoxin DTT was omitted and 2.5 µmol L⁻¹ Trx1, 1 µmol L⁻¹ TR and 2 mmol L⁻¹ NADPH were added. The assays were performed in 50 µL scale in three independent experiments for each condition. The assay was started by addition of the RNR and stopped after 5 to 15 min by addition of 50 µL Methanol and incubation at 50 °C for 5 min. After centrifugation and addition of 200 µL dH₂O, the samples were analyzed via UHPLC.

UHPLC analysis was performed on a Knauer Platinblue^® UHPLC. 5 µL of sample were injected on a Luna Omega 1.6 um Polar C18^® column from Phenomenex. GTP/dGTP and UTP/dUTP were eluted with a linear MeOH gradient from 26–30% over 15 min in a 50 mmol L⁻¹ KP_i-buffer (pH = 7.0) containing 0.25% (v/v) tetrabutylammonium hydroxide. CTP/dCTP were eluted with a linear MeOH gradient over 15 min from 27–30% the same buffer. ATP/dATP were eluted with a linear MeOH gradient over 15 min from 10–30% in the buffer as described before. A complete table of retention times is given in the supporting information (Table S17).

Gene identification and phylogenetic analysis

Screening the RNRdb (http://rnrdb.pfitmap.org) for viral RNRs with no similar host RNRs, we found three NrdJm enzymes, i.e., monomeric class II RNR enzymes, from viruses infecting Thermus species. The viruses belong to the Siphoviridae family and were found in hot biotopes. Thermus virus P74-26 and P23-45 were isolated at Uzon caldera in Kamchatka (Minakhin et al., 2008; Yu, Slater & Ackermann, 2006). The similar Thermus virus g20c was isolated at Geyzer Valley, Kamchatka (Xu et al., 2017). The NrdJm protein from Thermus viruses P74-26 (TVNrdJm, txid466052) and P23-45 (txid466051) have 96% identical amino acid sequences. Since the sequence from TV g20c contains a likely intein, we excluded this sequence from further study. The phylogenetically closest well characterized homologue, from which a crystal structure is available, is the NrdJm from Lactobacillus leichmannii (Booker & Stubbe, 1993; Sintchak et al., 2002). This is currently the only well studied NrdJm enzyme and is only distantly related to TVNrdJm (Fig. 1A). Within the phylum Deinococcus-Thermus no NrdJm enzymes could be found in current databases. The most closely related enzyme found in Thermus spp. are dimeric NrdJ enzymes with no significant sequence similarity to the TV enzymes.

We investigated the evolutionary descent of the three viral nrdJm genes by phylogenetic analysis of their respective amino acid sequences by estimating a maximum likelihood phylogenetic tree of the full diversity of NrdJm sequences from NCBI’s RefSeq database (Fig. 1A). The two viral sequences grouped inside a well-supported (82% bootstrap support) clan consisting exclusively of Firmicutes sequences except for the viral sequences (Fig. 1B). Searching for metagenomic sequences similar to those in the TVNrdJm clan, only resulted in sequences placed in the Firmicute part of the clan (https://doi.org/10.17045/sthlmuni.7642343.v1). The identity between the viral sequences and the Firmicutes sequences was 35–42% at the amino acid level. At nucleotide sequence level the similarity was hardly detectable using the BLAST algorithm. The similarity to genes from Firmicutes was shared with five other genes from the virus, whereas eleven genes were most similar to genes from Proteobacteria and eight from Deinococcus-Thermus (Table S8).

Since the NrdJm protein sequences are most closely related to Firmicutes sequences, we wanted to see if the same held true for the genetic signature of sequences. Therefore, we compared codon usage and GC-content of the TV mrdJm genes to Thermus and Firmicutes species. In codon usage, measured in euclidean distance, TV nrdJm genes showed high similarities to the genomes of the Thermus species and lower similarities to Firmicutes (Fig. 1C). Since the similarity between Firmicutes and the Thermus species on average is even lower, the TV nrdJm codon usage is intermediary between Thermus spp. and Firmicutes. The GC content of the viral genomes as well as the nrdJm genes was 57–58% (Fig. 1B), relatively close to the GC-content of the Thermus hosts at 67–69%. For the Firmicutes harboring the most similar nrdJms, the GC-content was, with the exception of three taxa around 60%, between 27 and 54% (Fig. 1B). Thermus spp. are thermophilic organisms with optimal growth temperatures between 60 and 70 °C. Most of the relevant Firmicutes are mesophiles, but some are at least moderate thermophiles despite their low GC content. Herbinix hemicellulosilytica, for instance, has an optimal growth temperature of 55 °C with a GC content of only 36% (Koeck et al., 2015) (Fig. 1B).

To investigate if positive selection had played a role in the evolution of the Thermus viruses since their divergence from the common ancestor with the Firmicute sequences, we performed a test of the ratio of non-synonymous to silent nucleotide substitutions (dN/dS) on this branch (marked with “#1” in Fig. 1B) with codonml from PAML (Yang, 1997; Yang, 2007). We found a dN/dS ratio of 17.5, significantly greater than 1 (p-value: 0.036) and hence suggesting positive selection for amino acid substitutions. A Bayes Empirical Bayes analysis discovered 40 amino acid positions with a ≥95% posterior probability of being under positive selection (Table S10). To gain some insight into the possible role of the 40 amino acids, we modeled the TVNrdJm on the Lactobacillus leichmannii monomeric NrdJm structure (Sintchak et al., 2002) (Model in the SI) and mapped the positions of the significant non conservative mutations on to the model. In total, 18 of these positions are located around the specificity site or on the dimer mimicking domain. Ten positions are located in the B₁₂ binding domain and another five cluster in a patch on the surface of the protein (Fig. 1D).

Recombinant expression of nrdJm genes and oligomeric state

The gene coding for TVNrdJm was synthesized commercially with codon optimization for recombinant expression in E. coli. High levels of soluble expression were obtained after only 4 h of expression at 37 °C. After purification with IMAC and SEC we obtained 40 mg of pure enzyme per liter of culture. The enzyme eluted after 16.8 mL on the size exclusion column, corresponding to a molecular weight of 65 kDa. With a calculated molecular weight of 73 kDa, this corresponds to a monomeric state of the enzyme in solution (Fig. S2).

The viral nrdJm gene encodes a functional B₁₂-dependent ribonucleotide reductase with high temperature optimum

The purified TVNrdJm was able to reduce GTP to the corresponding deoxyribonucleotide but not GDP, defining it as a nucleoside triphosphate reductase. Equimolar concentrations of cofactor B₁₂ were sufficient to support the reaction, while higher concentrations reduced the activity to 40% at 0.1 mmol L⁻¹ (Fig. 2A). Without the presence of cofactor B₁₂, no reaction was observed.

As the enzyme was isolated from a virus with a thermophilic host, reaction temperatures were monitored over a temperature range from 25 °C to 90 °C. Enzyme activity peaks between 60 °C and 70 °C with complete inactivation at higher temperatures (Fig. 2B). The pH profile of the enzyme was investigated between pH values 5 and 9. The pH optimum was found to be between 7 and 8 with about 50% residual activity between 6.5 and 9 (Fig. 2C).

The allosteric regulation of TVNrdJm was investigated by a survey of the reduction of the four ribonucleotides ATP, CTP, GTP and UTP with the potential allosteric effectors dATP, dCTP, dGTP and dTTP. Without allosteric effector, no activity could be observed for any of the four substrates (Table 1). Significant activity was only present for the substrate/effector pairs ATP/dGTP, CTP/dATP, GTP/dTTP.

Terminal reduction of NrdJm with artificial reducing agents and Thermus Thioredoxins

Activity of TVNrdJm was tested with the artificial reductants DTT and TCEP. While DTT is able to reduce the active site cysteine pair directly, TCEP can only work via the C-terminal cysteine pair (Domkin & Chabes, 2014; Loderer et al., 2017). The NrdJm exhibited 33% higher activity with TCEP compared to DTT at their respective maxima (Fig. 3A). For both reductants, the maximal activity was reached already at 2 mmol L⁻¹. Higher concentrations lead to a reduction of the activity to about 50% for both reductants at 100 mmol L⁻¹.

For the RNR catalyzed reaction in vivo, reducing equivalents are required in form of gluta- or thioredoxins. Since the TVs do not encode redoxins, we tested TVNrdJm with the thioredoxin/thioredoxin reductase system from the strain Thermus thermophilus. With the redoxin system, the viral NrdJm was active with a comparable reaction velocity to the artificial reductants (Fig. 3B). The control reaction without thioredoxin also displayed some activity. This may have been caused by residual DTT in the reaction mixture coming from the enzyme stocks. The experiment also showed that the complete redox reaction chain from NADPH over thioredoxin reductase, thioredoxin and RNR to the final electron recipient GTP is functional in vitro.