Structure of a tRNA-specific deaminase with compromised deamination activity

Nucleotide 34 in tRNA is extensively modified to ensure translational fidelity and efficacy in cells. The deamination of adenosine at this site catalyzed by the enzyme TadA gives rise to inosine (I), which serves as a typical example of the wobble hypothesis due to its diverse basepairing capability. However, recent studies have shown that tRNA ArgACG in Mycoplasma capricolum contains unmodified adenosine, in order to decode the CGG codon. The structural basis behind the poorly performing enzyme M. capricolum TadA (named McTadA) is largely unclear. Here we present the structures of the WT and a mutant form of McTadA determined at high resolutions. Through structural comparison between McTadA and other active TadA enzymes as well as modeling efforts, we found that McTadA presents multiple structural conflicts with RNA substrates and thus offered support to previous studies from a structural perspective. These clashes would potentially lead to reduced substrate binding affinity of McTadA, consistent with our in-vitro deamination activity and binding assays. To rescue the deamination activity of McTadA, we carried out two rounds of protein engineering through structure-guided design. The unsuccessful attempts of activity restoration could be attributed to altered dimer interface and stereo hindrance from the non-catalytic subunit of McTadA, which could be the inevitable outcome of the natural evolution. Our study provides structural insight into an alternative decoding and evolutionary strategy by a compromised TadA enzyme at a molecular level.


Introduction
step [10]. Although employing identical chemistry for catalysis, there are clear differences between the Tad enzymes from eukarya and bacteria, in terms of the deamination process. The bacterial enzyme named TadA usually functions a homodimer. In contrast, the eukaryotic Tad consists of the heterodimeric Tad2/Tad3 complex [12,13]. Tad2 is the catalytic subunit and Tad3 serves as the regulatory subunit because it lacks a conserved glutamate in the conserved H(C)XE motif, which is replaced by a valine residue. Accordingly, the modification subjects (tRNA species) in the two kingdoms are quite different as well. Sequence alignment indicates that eukaryotic Tads share only limited homology with that of bacteria.
Following the structure determination of numerous bacterial TadA enzymes such as Escherichia coli TadA (EcTadA, PDB 1Z3A), Aquifex aeolicus TadA (AaTadA, PDB 1WWR) [11,14]  Using the bioinformatics approach, Yokobori et al. found that more than 30 species from Mycoplasmas and other Mollicutes were missing the tRNA Arg -encoding gene that specifically pairs with the arginine codon CGG [15]. Additionally, 25 such species completely lack the tadA gene, suggesting non-classical decoding strategies in these species. In several species including Mycoplasma capricolum, only tRNA Arg ACG , tRNA Arg TCT and a copy of tadA gene were found to be present in the genome, and the former was presumably to decode the CGN quartet codons. Because inosine (resulted from A34 deamination) is considered as a poor basepairing partner of guanosine due to the unfavorable purine-purine base-pairing pattern (except for the A-I pairing) in Crick's theory, it raises the question as to how the CGG codon in these species was decoded. Further investigation of the M. capricolum tRNA Arg ACG sequence by reverse transcriptase-PCR on extracted tRNA Arg ACG indicated that some of these tRNA molecules contain un-deaminated A34, as adenosine at the wobble position were still found in 5 out of the 81 clones, while it was zero for B. subtilis [15], implying reduced deamination efficiency of the TadA enzyme in the former species. Moreover, sequence alignment and structural modeling suggested that this observation might be the result of the reduction in deamination efficiency due to selective mutations in TadA during evolution, which interferes with the binding and modification of the tRNA substrate. Consequently, the unmodified tRNA Arg ACG is responsible for reading the CGG arginine codon [15]. However, their hypothesis was not confirmed due to a lack of structural information on M. capricolum TadA (McTadA).
In the current study, we report the crystal structures of McTadA and its mutants.  with a wavelength of 0.979 Å, which was subsequently processed with the program HKL3000 [16]. All data were collected at 100 K with a nitrogen cryo stream. The WT crystals belonged to the P4 3 2 1 2 space group with 2.40-Å resolution and the asymmetric unit was predicted to contain one monomer. Molecular replacement (MR) was first performed with Phenix using the AfTadA (PDB 2A8N) as the search model [17]. After a plausible solution was obtained, the model was manually rebuilt by COOT according to the electron density map [18,19]. The rebuilt model was fed to the phenix.refine and multiple cycles of refinement were conducted, followed by model rebuilding [20,21]. The R free /R work factors were 21.9-and 19.8 % respectively and the final model was validated by Molprobity [22]. The structure of the TM mutant was solved by MR using the WT structure as the model. The R free /R work factors for the mutant were 25.1-and 19.8 % respectively. The structural presentation of McTadA was prepared by PyMOL (http://www.pymol.org/). All data collection and refinement statistics are presented in Table 2.

In-vitro transcription of RNA
The MctRNA Arg was prepared by in-vitro transcription [23]. EcEndoV [24][25][26][27] were added to this mixture, and the reaction was incubated at 37 ℃ for another 50 min. The reaction was stopped by adding 2 × Urea loading buffer. The product was analyzed by 18 % urea-PAGE gel electrophoresis followed by ethidium bromide staining and UV-light exposure.

Circular dichroism (CD) analysis
The WT and mutant proteins were first diluted to a concentration of 0.3-0.5 mg/mL in

The in-vitro deamination activity of McTadA
To test the deamination activity toward the ACG codon-containing RNA substrates, we first expressed the full-length recombinant McTadA. The protein was prone to precipitation upon isolation, and had to be stabilized by 10-15% glycerol. The The density from Ser115 to Asp118 was rather poor, and was considered a disorder and not modeled. Otherwise, residues from Asp2 to Leu148 were completely resolved.
In accordance to the ICP-MS result, there is a zinc ion bound to the catalytic center.
The zinc ion is fixed by four ligands，comprising His54, Cys84, Cys87, and a water molecule. The four ligands form a tetrahedral geometry, with the water molecule held in position by the general base Glu56, whose OE2 atom is only 2.9 Å away (Fig. 2B).
Additionally, Phe108 from the β4-β5 loop forms possible hydrophobic interactions with Phe143, Leu139, Ile136, all of which are located on the same side of the final helix (Fig. 2C). These interactions suggest that there might be crosstalks between these two regions. Crystallographic symmetry operations could generate the enzyme dimer, i.e. the functional catalytic unit, with the 2-fold axis coinciding with the dimer interface ( Fig. 2A).  5A). The K146A mutation considerably reduced the size of the side chain, and now poses little hindrance to C35 as well. Lastly, the substitution at the Ser105-Ile110 region produced local disorder between Lys107-Tyr111 (new numbering according to the TM mutant) and pulled the fragment away from U33. Additionally, the unresolvable density suggested flexibility of this region, which may help to relieve the rigidity and structural clashes (Fig. 5B). From our model, the current the structure of the mutant would not interfere with the binding of RNA at these two sites.

Comparison to other TadAs and generation of a McTadA-RNA complex model
With these desired engineering purposes presumably accomplished, we next performed activity assays as described above. However, we still could not detect the deamination activity for the McTadA-TM. Circular dichroism studies showed that the TM mutant had quite similar profiles to that of WT, suggesting little structural perturbation by the mutations (Fig. S1). Thorough structural analyses revealed that the which is considerably smaller than that of EcTadA (2870 Å 2 based on PDB 1Z3A) [14], AaTadA (2800 Å 2 based on PDB 1WWR) [11], or Agrobacterium fabrum tRNA adenosine deaminase TadA (AfTadA, 2620 Å 2 based on PDB 2A8N) [17]. When RNA substrates bind, they would make contacts with both enzyme subunits. While the anticodon loop bases (especially A34) primarily interact with the catalytic subunit, other bases also require support from the non-catalytic subunit. A larger interfacial area suggests a weaker association between the enzyme and RNA substrates. This is probably caused by the translation of α4 (Ala55-Leu67), which moves away from the dimer interface (Fig. 6A). Additionally, close examination of the McTadA dimer further revealed that the R45Y46 fragment from the catalytic subunit and Ser69 from the non-catalytic subunit would block bases C35 and G37 respectively (Fig. 6B, C).  (Fig. 7). This result supported our theory that structural conflicts would arise upon the binding of tRNA and the inert activity was due to reduced binding affinity (k D ) rather than catalysis (kcat). Additionally, EMSA was also carried out using  S2).

Discussion
The In accordance with this theory, base 34, or the "so-called" wobble base is usually extensively modified to adapt to the different basepairing needs.
One frequently occurring modification is that the wobble A34 in some precursor tRNAs is enzymatically converted to inosine by specific tRNA:A34 deaminases during tRNA maturation. The deaminated inosine is capable of reading the C, U and A-ending codons, but the basepairing with a G-ending codon is not allowed [9]. Thus Upon structural comparison to that of EcTadA or SaTadA, we introduced site-specific mutations to the enzyme and tried to restore its activity. However, the In this work, Sibter and co-workers identified a yeast mitochondria tRNA Arg belonging to the four CGN arginine codon family that contained unmodified A34.
They tried to explain the decoding problem of the CGN codons with the classical 'two out of three' hypothesis, as the other two base pairs are of G-C combination.

Author contribution
Wei Xie conceived and designed the research; Huijuan Liu, Saibin Wu and Dewei Ran performed the research; Wei Xie analyzed data and wrote the paper. All authors reviewed the manuscript. Table 1. Sequences of the primers and tRNA used in this study.
The restriction sites of the cloning primers were underlined in the amplifying primers while the mutated bases were in italic in the Quikchange primers.

Primers name
Sequence (  Values in parentheses are for the highest-resolution shell. b R merge =Σ |(I -< I > )|/σ(I), where I is the observed intensity. c R work = Σ hkl ||Fo| -|Fc||/ Σ hkl |Fo|, calculated from working data set. d R free is calculated from 5.0 % of data randomly chosen and not included in refinement.