Archaeal NSUN6 catalyzes m5C72 modification on a wide-range of specific tRNAs

Abstract Human NOL1/NOP2/Sun RNA methyltransferase family member 6 (hNSun6) generates 5-methylcytosine (m5C) at C72 of four specific tRNAs, and its homologs are present only in higher eukaryotes and hyperthermophilic archaea. Archaeal NSun6 homologs possess conserved catalytic residues, but have distinct differences in their RNA recognition motifs from eukaryotic NSun6s. Until now, the biochemical properties and functions of archaeal NSun6 homologs were unknown. In archaeon Pyrococcus horikoshii OT3, the gene encoding the NSun6 homolog is PH1991. We demonstrated that the PH1991 protein could catalyze m5C72 formation on some specific PhtRNAs in vitro and was thus named as PhNSun6. Remarkably, PhNSun6 has a much wider range of tRNA substrates than hNSun6, which was attributed to its tRNA substrate specificity. The mechanism was further elucidated using biochemical and crystallographic experiments. Structurally, the binding pocket for nucleotide 73 in PhNSun6 is specific to accommodate U73 or G73-containing PhtRNAs. Furthermore, PhNSun6 lacks the eukaryotic NSun6-specific Lys-rich loop, resulting in the non-recognition of D-stem region by PhNSun6. Functionally, the m5C72 modification could slightly promote the thermal stability of PhtRNAs, but did not affect the amino acid accepting activity of PhtRNAs.


INTRODUCTION
As linker molecules between ribosomes and growing polypeptide chains, transfer RNAs (tRNAs) play a central role in protein synthesis. Recently, further functions of tR-NAs have been identified in addition to translation (1)(2)(3).
5-methylcytosine (m 5 C) is a ubiquitous modification in DNA and different cellular RNAs, which is formed by site-specific methyltransferases (MTases). The NSun family, which contains an S-adenosyl-methionine (SAM)dependent NOL1/NOP2/Sun (NSun) domain, is the main RNA:m 5 C MTase and comprises seven family members in eukaryotes. Among them, NSun2, NSun3 and NSun6 are primarily responsible for m 5 C modifications on tRNAs (21)(22)(23). Genetic mutations and aberrant expression of several NSun family members are linked to human diseases and disorders, including intellectual disability (24,25), mitochondrial deficiency (26), Williams-Beuren syndrome (27), male infertility (28), and cancer development (29,30). In archaea, m 5 C modifications are widespread in cellular RNAs, especially in tRNAs. Positions C34, C38, C48, C49 and C50 of tRNAs are the known m 5 C modification sites in archaea (31,32). m 5 C72 is a newly identified modification in the acceptor stem region of human cytoplasmic tRNA Thr and tRNA Cys , established by human NSun6 (hNSun6) (23). Our previous study showed that well-folded and sequence-specific tRNA substrates are strictly selected by hNSun6 (33). The struc-ture of hNSun6 in complex with tRNA revealed that the MTase domain works together with pseudouridine synthase and archaeosine transglycosylase (PUA) domain to recognize regions in the acceptor and D-stem of tRNAs (34). Homologs of hNSun6 are present only in higher eukaryotes and hyperthermophilic archaea. In contrast to the wellstudied identified genes that encode eukaryotic NSun6s, the corresponding genes in archaea are unknown.
In the genome of Pyrococcus horikoshii OT3 (P. horikoshii), five genes encode putative RNA m 5 C MTases: PH1537, PH1374, PH1078, PH1991, and PH0851. Among them, PH1991 shares the highest sequence similarity (23.9% identity) with hNSun6, suggesting that PH1991 is the hN-Sun6 homolog. Interestingly, compared with hNSun6, the amino acid sequence of PH1991 is shorter and the predicted RNA recognition residues, especially the predicted PUA domain (residues 86-164) and RNA recognition motifs (RRMs, residues 34-79 and 173-185), are extremely divergent from eukaryotic NSun6s (Supplementary Figure  S1). Moreover, PH1991 lacks the eukaryotic NSun6-specific conserved Lys-rich loop in the PUA domain, which is essential for D-stem recognition by eukaryotic NSun6s. Until now, the properties and functions of PH1991 were uncharacterized. Thus, the present study aimed to determine whether PH1991 retains the m 5 C MTase catalytic activity; what are the RNA substrates of PH1991, considering the differences in the predicted RNA binding residues; and what is the biological function of PH1991.
In this study, we identified that PH1991 is a tRNA:m 5 C72 MTase NSun6 (PhNSun6). In contrast to hNSun6, which could only methylate four tRNAs with U73, PhNSun6 could catalyze m 5 C72 formation on eleven PhtRNAs harboring either U73 or G73. We further characterized the tRNA recognition elements for PhNSun6, which were quite different from those of hNSun6. To better understand the recognition mechanism of PhNSun6, we solved the crystal structures of PhNSun6 in the apo form and in complex with SAM, sinefungin (an analog of SAM, SFG), and S-adenosyl-homocysteine (demethylated SAM, SAH). Furthermore, we found that m 5 C72-modified PhtRNAs show slightly higher thermal stability than their unmodified counterparts, but this does not affect the amino acid accepting activity of PhtRNAs. Combined with previous work, we proposed an evolutionary model for eukaryotic and archaeal NSun6 and its tRNA recognition mechanism.

Preparation of tRNA
The PhtRNAs genes were inserted into pTrc99b to construct pTrc99b-T7-PhtRNAs. Mutants of the PhtRNA Thr (CGU) were obtained using the KODplus mutagenesis kit. All tRNAs were produced using in vitro T7 RNA polymerase transcription, as described previously (38). The transcribed tRNAs were separated by urea-denaturing 12% polyacrylamide gel electrophoresis (PAGE), eluted with 0.5 M NaAc (pH 5.2), precipitated with three volumes of ethanol at −20 • C and dissolved in 5 mM MgCl 2 . The tRNAs were annealed at 85 • C for 10 min and cooled naturally to room temperature for correct folding. The tRNA concentration was determined using UV absorbance at 260 nm, and the molar absorption coefficient was calculated according to the sequence of each tRNA (39).

Protein expression and purification
The gene encoding PH1991 (PhNSun6) was chemically synthesized by BioSune and inserted into the pET28a vector using NdeI and NotI. Escherichia coli Rosetta (DE3) was transformed with the pET28a-PhNSun6 plasmid. Protein overproduction was induced by adding 0.2 mM IPTG into cells. After culturing for 12 h at 18 • C, the cells were collected by centrifugation. PhNSun6 was purified by affinity Nucleic Acids Research, 2019, Vol. 47, No. 4 2043 chromatography on Ni-NTA Superflow resin according to the manufacturer's protocol. The wet cells (∼3 g) were suspended in 15 ml of 10 mM imidazole in buffer A [20 mM Tris-HCl (pH 7.5), 500 mM NaCl, 5 mM MgCl 2 , 5 mM DTT and 10% (v/v) glycerol] and sonicated on ice. Then, the sonicated crude extracts were heated at 65 • C for 10 min. The crude extracts were centrifuged at 40 000 × g for 40 min to remove the debris and insoluble fractions. The supernatant was mixed gently with 0.5 ml Ni-NTA resin for 30 min at 4 • C and then washed with 50 ml of 20 mM imidazole in buffer A to remove nonspecific binding proteins. Binding proteins were eluted in 8 ml of 250 mM imidazole in buffer A. The eluted protein was concentrated and purified by gel filtration on a Superdex™ 200 column.
HNSun6 was expressed and purified according to our previous method (33,34). The active P. horikoshii SerRS and mouse ThrRS were purified as described previously (40,41). The protein concentrations were determined using UV absorbance at 280 nm, and the molar absorption coefficient was calculated according to the sequence of each protein (42).

Protein crystallization, structure determination, and refinement
Crystallization was performed at 16 • C by the hanging drop vapor diffusion method. For crystallization, PhNSun6 was concentrated to ∼13.3 mg/ml. Protein solution (1 l) was mixed with an equal volume of the reservoir solution, consisting of 0.2 M magnesium acetate tetrahydrate, 0.1 M sodium cacodylate trihydrate pH 6.5 and 20% (w/v) polyethylene glycol 8000. PhNSun6 and 1 mM SAM, or SFG, or SAH, were co-crystallized under the same conditions. The crystals were frozen in liquid nitrogen after transferring for a few seconds in the mother liquor, which contained 15% (v/v) glycerol as a cryoprotectant.
All crystal diffraction data sets were collected at the Shanghai Synchrotron Radiation Facility beamlines (SSRF, Shanghai, China) BL-19U1. The diffraction data were processed using the HKL3000 program package (43). Further data analysis was performed with the CCP4 suite (44). The structure of PhNSun6 in the apo form was initially solved by molecular replacement using PHASER (45) with hN-Sun6 (PDB ID: 5WWQ) structure as starting model. The model was further improved by manual adjustments using COOT (46). The structures of PhNSun6-SAM, PhNSun6-SFG, and PhNSun6-SAH were solved by molecular replacement using PHASER. All models were refined using REFINE program in the PHENIX suite (47). The quality of the final model was evaluated using MOLPROBITY (http://molprobity.biochem.duke.edu/). Figures were drawn using PyMOL (http://www.pymol.org/). A structure-based multiple amino acid sequence alignment of NSun6s from model organisms was generated using ESPript (48).

tRNA methyl transfer assay
To measure the methyl transfer activity of PhNSun6 and its mutants, 5 M of tRNAs were used as substrates. The reactions were performed at 65 • C under the same conditions in a 25-l reaction mixture containing 200 M 3 H-SAM, 50 mM Tris-HCl (pH 7.5), 200 mM NaCl, 10 mM MgCl 2 , 100 g/ml BSA, and 5 mM DTT. Reactions were initiated by the addition of 100 nM PhNSun6. Aliquots (5 l) of the reaction mixtures were removed at time intervals between 2 and 8 min, quenched on Whatman glass-fiber filter discs, and soaked in 5% trichloroacetic acid (TCA). After washing, the amount of radioactive [ 3 H]-methyl-tRNA on each disc was measured in a Beckman Las6500 scintillation counting apparatus. The steady-state kinetics were measured under the same conditions with 50-300 nM PhNSun6 and 0.1-40 M tRNA; the reaction time was 2 or 4 min. For mutants that exhibited extremely low activity, higher concentrations of enzyme and tRNA substrates were used. The data were fitted to Lineweaver-Burk plots, and the K m , k cat and k cat /K m values for each tRNA were calculated from the curve fitting.
To obtain m 5 C72 PhtRNAs, the reactions were performed at 65 • C in a mixture comprising 2 M PhNSun6, 200 M SAM, 50 mM Tris-HCl (pH 7.5), 200 mM NaCl, 10 mM MgCl 2 , 100 g/ml BSA, and PhtRNAs. The wildtype (WT) PhtRNAs were prepared in the same mixture by adding the PhNSun6 restored buffer. The reaction was performed for 1 h and stopped using phenol/chloroform extraction. The PhtRNAs were precipitated using a 3-fold volume of ethanol.

Aminoacylation assay
The time course curve for aminoacylation by SerRS for PhtRNA Ser was determined in a 25-l mixture containing 60 mM Tris-HCl (pH 7.5), 10 mM MgCl 2 , 5 mM DTT, 2.5 mM ATP, 100 g/ml BSA, 40 M [ 3 H] serine, and 200 nM enzyme, with 5 M of PhtRNA Ser at 60 • C. Aliquots (5 l) of the reaction mixtures were removed at time intervals between 1 and 4 min, quenched on Whatman glass-fiber filter discs, and soaked in 5% TCA. The time course curve for aminoacylation by ThrRS for PhtRNA Thr was determined at 37 • C, as described previously (41).

Mass spectrometry analysis of m 5 C modification
One microgram of tRNA was hydrolyzed using benzonase, phosphodiesterase I, and bacterial alkaline phosphatase overnight at 37 • C in a 20-l solution including 20 mM Tris-HCl (pH 8.0), 2 mM MgCl 2 , and 20 mM NaCl. The solution was then diluted with H 2 O 100 times and 10 l was applied to Ultra-Performance Liquid Chromatography-Mass Spectrometry (UPLC-MS). The nucleosides were separated using UPLC on a C18 column (Agilent Zorbax Eclipse Plus C18, 2.1 × 50 mm, 1.8-Micron) and then detected using a triple-quadruple mass spectrometer (Agilent 6400 QQQ) in the positive ion multiple reaction-monitoring (MRM) mode. Mass transition from m/z 258 to 126 (m 5 C) was monitored and recorded.

Melting temperature (T m ) assay of PhtRNAs
Specific tRNAs were dissolved in T m buffer [10 mM sodium phosphate monobasic/sodium phosphate dibasic (pH 7.0), 100 M EDTA and 10 mM NaCl]. The initial absorbance of the tRNA at 260 nm was controlled between 0.2-0.3. A melting temperature curve was measured at 260 nm using a heating rate of 1 • C/min from 25 to 95 • C via an Agilent Cary 100 spectrophotometer. The T m value was determined from the first derivative of the melting curve.
Subsequently, we purified PH1991 and subjected it to SDS-PAGE analysis ( Figure 1A). To assay the MTase activity of PH1991 for PhtRNA Thr (CGU), we tested the methyl transfer activity at 37, 55 and 65 • C. The highest methyl transfer activity of PH1991 was at 65 • C (Supplementary Figure S2). We found that PH1991 was able to catalyze the methylation of three PhtRNA Thr isoacceptors and PhtRNA Cys (GCA) at 65 • C, similar to hNSun6 at 37 • C ( Figure 1B). The PhtRNA Thr (CGU) methylated by PH1991 were analyzed using UPLC-MS and the modification was confirmed to be m 5 C ( Figure 1C), as it was in the methylated PhtRNA Thr (GGU), PhtRNA Thr (UGU), and PhtRNA Cys (GCA) (Supplementary Figure S3). To confirm that C72 was the target site, three mutants, PhtRNA Thr (CGU)-C72A, -C72G, and -C72U, were constructed. None of these mutants could be methylated by PhNSun6 ( Figure 1D). We measured the steady state kinetic parameters of PhNSun6 in the presence of four PhtRNA substrates (Supplementary Table S1). The K m values for the four PhtRNAs were approximately 0.5 M; however, the k cat values showed differences. The k cat values of PhNSun6 for three PhtRNA Thr isoacceptors were approximately 4.30 min −1 , which was higher than that for PhtRNA Cys (3.04 min −1 ). Thus, we identified that PH1991 is the P. horikoshii tRNA:m 5 C72 MTase NSun6 and named it PhNSun6.

PhNSun6 shows a different recognition mechanism for the acceptor stem of tRNAs compared with that of hNSun6
Our previous study on hNSun6 showed that the CCA terminus, the target site C72, the discriminator base U73, and base pairs 2:71 and 3:70 in the acceptor stem are the tRNA recognition elements (33). We then determined whether these recognition elements in the acceptor stem were conserved in tRNA recognition by PhNSun6. We assayed the methylation of a PhtRNA Thr (CGU) mutant with a deleted CCA terminus and several mutations in U73 and C2:G71 ( Figure 2A).
PhNSun6 could not methylate the PhtRNA Thr (CGU) mutant lacking the CCA terminus ( Figure 2B), implying that the common CCA terminus of tRNA was still essential for the methylation reaction catalyzed by PhNSun6.
According to the tRNA database (31,49), the second base pair in the acceptor stem is C2:G71 or G2:C71 in most PhtRNAs, and in few PhtRNAs is G2:U71; the third base pair exists as C3:G70, G3:C70, A3:U70 and G3:U70. In the four above-mentioned PhtRNAs, the second and third base pairs are C2:G71 and C3:G70, respectively. To determine whether PhNSun6 recognized the second base pair, two mutants, PhtRNA Thr (CGU)-C2G:G71C and -C2G:G71U, were constructed and their methylation by PhNSun6 was assessed (Figure 2A). The results showed that PhNSun6 could not methylate these two mutants ( Figure 2D; Supplementary Table S2), indicating clearly that C2:G71 is a determinant base pair of tRNA recognition by PhNSun6.
Collectively, we identified four elements in the acceptor stems of tRNAs that are essential for substrate recognition by PhNSun6: The target site C72, the CCA terminus, U or G73, and the second base pair C2:G71.

The D-stem is not recognized by PhNSun6
Besides elements in the acceptor stem, hNSun6 also interacts with the D-stem of tRNAs (33,34). Base pairs 11:24 and 12:23 of the D-stem are recognized via the eukaryotic NSun6-specific Lys-rich loop (residues 157 KCKKGAK 163 in hNSun6, Supplementary Figure S1). However, archaeal NSun6 homologs lack the corresponding eukaryotic NSun6-specific Lys-rich loop, indicating that archaeal NSun6 homologs presumably possess a distinct recognition mechanism in D-stem of tRNA substrates. Therefore, we constructed a series PhtRNA Thr (CGU) mutants by replacing C11:G24 or C12:G23 with A:U, U:A, G:C, G:U and U:G, respectively ( Figure 3A). Interestingly, these mutants could all be methylated by PhNSun6 ( Figure 3B, C), and the K m and k cat values of PhNSun6 for these mutants were similar to that for PhtRNA Thr (CGU) (Supplementary Table S2).
These results suggested that C11:G24 and C12:G23 base pairs in the D-stem are not involved in RNA recognition by PhNSun6.

PhNSun6 has a wider specificity of tRNA substrates than hN-Sun6
Based on the above results, PhNSun6 requires fewer identity elements for tRNA substrates than hNSun6. To examine whether PhNSun6 has other tRNA substrates besides the above four PhtRNAs (three PhtRNA Thr isoacceptors and PhtRNA Cys (GCA)), we determined the methylation activity of PhNSun6 for all the PhtRNAs that harbor the recognition elements of PhNSun6. These PhtRNA Arg (GCG). In addition, PhtRNA Phe (GAA) with A73 and G2:C71, which does not match the recognition requirements for PhNSun6, was chosen as a negative control ( Figure 4A). Intriguingly, PhNSun6 could catalyze methylation of the above seven tRNAs but not the negative control PhtRNA Phe (GAA) ( Figure 4B; Supplementary  Table S3).
Thus, the results showed that PhNSun6 could efficiently catalyze m 5 C72 formation on eleven PhtRNAs, which suggested that PhNSun6 has a wider range of tRNA substrates than hNSun6.
Cofactors bind at the Rossmann-fold catalytic core of PhNSun6. When comparing the structures of PhNSun6 in the apo form and in the form bound with cofactors, the overall structures do not show distinct changes, except that helix ␣9 moves a slightly toward the cofactors (Supplementary Figure S5). Residues Met208, Ala209, Pro279, Phe308, Asp260, Asp233, Arg238, Ala211, Asp277, Lys215, Tyr304, Gly213 and Gly214 form the binding pocket for the cofactor SAH, as shown in Figure 5B. These residues are mostly conserved or semi-conserved in eukaryotic and archaeal NSun6s, except for the archaeal-specific residue Tyr304, which recognizes N6 and N7 of the adenine ring of SAH. In the PhNSun6-SAM and PhNSun6-SFG complexes, similar binding models of the cofactors to that of SAH were observed, but are not discussed here.
Although PhNSun6 contains a different RNA recognition motif compared with that in hNSun6, the main structure of PhNSun6 adopts a similar topology to hNSun6 and the two active Cys residues of PhNSun6 are conserved. The two conserved Cys residues point from opposite site sides of the cofactor, which is the same as that in hNSun6 ( Figure  5C). Some structural differences were observed between PhNSun6 and hNSun6. The first difference is that hNSun6 contains two unique insertions (hNSun6 specific insertion 1 and 2) in the surface of MTase catalytic core (Supplementary Figure S6). Second, for the PUA domain, most   important difference concerns the change in the orientations of PhNSun6 helix ␣4 and hNSun6 ␣5 ( Figure 6F). Third, PhNSun6 lacks the eukaryotic specific Lys-rich loop ( Figure 6A). Taken together, the structural differences between PhNSun6 and hNSun6 might contribute to the distinct tRNA substrate recognition mechanisms.

Binding model of PhNSun6 with tRNA
We were not able to solve the structure of the PhNSun6 in complex with its tRNA substrate. Thus, we generated a tRNA binding model of PhNSun6 by superimposing the structure of PhNSun6/SAH onto that of the hNSun6/SFG/tRNA complex (PDB ID: 5WWR) (34). In this model, the PUA domain of PhNSun6 has no interactions with the D-stem region of tRNA, because of the lack of the eukaryotic NSun6-specific Lys-rich loop ( Figure 6A), which is consistent with D-stem of tRNA not being recognized by PhNSun6.
The CCA terminus of tRNA is essential for PhNSun6 recognition. According to the PhNSun6/tRNA model, all the three nucleotides of the CCA terminus interact extensively with residues from the PUA domain and linkers (Figure 6B, C). The binding mode between the CCA terminus and PhNSun6 is very similar to that observed from the hNSun6/tRNA complex. However, as mentioned above, the PUA domain, especially those residues on the protein surface, are quite divergent between PhNSun6 and hN-Sun6. To test this superposition model, we mutated putative CCA terminus-interacting residues (Asn98, Glu174 and Tyr112) of PhNSun6 to Ala. All the three mutants showed reduced methyltransferase activities compared with WT protein. The K m values of PhNSun6-N98A and -E174A for PhtRNA Thr (CGU) were ∼2-fold higher than that of PhNSun6, while their k cat values remain similar ( Figure  6D; Supplementary Table S4); the K m value of PhNSun6-Y112A for PhtRNA Thr (CGU) was 5.3-fold higher than that of PhNSun6 while its k cat value decreased ( Figure 6D; Supplementary Table S4). These results suggest that all the three residues of PhNSun6 contribute to tRNA binding. Thus, we conjectured that the PUA domain has retained the CCA binding capability during evolution.
The hypothetical binding pocket of PhNSun6 for the base at the 73 rd nucleotide of the PhtRNA substrate comprises Ser190, Gln107, and Tyr41 ( Figure 6E), which are absolutely conserved in archaeal NSun6 homologs. Intriguingly, the archaeal-specific hypothetical binding pocket for N73 seems propitious to accommodate either U or G, mainly thanks to the orientation of helix ␣4, which enlarges the size of the pocket in PhNSun6. In contrast, the corresponding helix ␣5 in hNSun6 limits the binding pocket to U73 only ( Figure 6F). In the PhNSun6/tRNA model, the O4 of U73 or O6 of G73 could form a hydrogen bond with the main chain amide group of Tyr41. However, when U73/G73 is substituted by C73 or A73, the N4 of C73 or N6 of A73 could not form this hydrogen bond, thus tRNAs with C73 or A73 will be excluded by PhNSun6. The archaeal-specific residues in the hypothetical binding pocket for N73 might explain why PhNSun6 could accommodate U or G, but not C or A, at site 73 of tRNA substrates. To test our hypothesis, we made mutations on Ser190 and Gln107 of PhNSun6. The OH group from the side chain of Ser190 forms a hydrogen bond to the phosphate group of N73. When Ser190 was mutated to Ala, PhNSun6-S190A had little methylation activity compared to PhNSun6 ( Figure 6G). On the other hand, in the hypothetical binding model, side chain of Gln107 is close to the uracil ring of U73. To modify the size and the charge of the binding pocket for the base ring of N73, Gln107 was replaced by Trp with large side chain or by Arg with positive charge respectively. Both PhNSun6-Q107W and -Q107R mutants lost methylation activity completely ( Figure 6G). These results suggest that residues in the proposed N73 binding pocket are important for methylation of PhNSun6.

The m 5 C72 modification of tRNA does not affect the aminoacylation activity of aminoacyl-tRNA synthetase (aaRS)
Considering that the m 5 C72 modification is close to the CCA terminus, we investigated whether the m 5 C72 modification affected the amino acid accepting activity of tRNA by the corresponding aaRS. We tested seven (PhtRNA Ser (CGA), -(GGA), -(UGA) and -(GCU), PhtRNA Thr (CGU), -(GGU) and -(UGU)) of the eleven PhtRNAs. The aminoacylation activity of aaRS for the m 5 C72-modified PhtRNAs showed no obvious difference compared that for their unmodified counterparts (Supplementary Figure S7). These results suggest that the m 5 C72 modification does not affect the amino acid accepting activity of the tested PhtRNAs.

The m 5 C72 modification slightly increases the thermal stability of PhtRNAs
Archaea usually exist in extreme environments, such as hot springs and salt lakes. P. horikoshii is a hyperthermophilic anaerobic archaeon, growing optimally at 98 • C (50). Critically, according to statistical analyses, archaeal NSun6 homologs only exist in hyperthermophilic archaea.
In hyperthermophilic organisms, tRNAs have to tolerate high temperature, resulting in the development of certain strategies to protect tRNAs from degradation and denaturation, and to increase the stability of tRNA tertiary structure, such as high GC-content and tRNA modifications (16,18,51). Notably, loss of some tRNA modifications in hyperthermophilic tRNAs affects their thermal stability.
Nucleic Acids Research, 2019, Vol. 47, No. 4 2051  Each of melting temperature was calculated from the average of three independent measurements and standard deviation values were also shown.
For example, in Thermus thermophilus, lack of the m 7 G46 modification causes decreased T m values of class I tRNAs and degradation of tRNA Phe and tRNA Ile , resulting in reduced protein synthesis at high temperatures (52).
To determine whether there is a correlation between the m 5 C72 modification and the thermal stability of tRNAs, we measured the T m of the eleven m 5 C72-modified PhtRNAs and their unmodified counterparts. The data were shown in Table 2. Except for PhtRNA Arg (GCG), the T m values for the other ten PhtRNAs with the m 5 C72 modifications all increased compared with those for their unmodified counterparts. However, the delta T m values are between ∼1 and 2 • C in most cases. Only PhtRNA Asp (GUC) showed a 3 • C difference between methylated and unmethylated tRNAs. The results suggested that m 5 C72 contributes thermal stability towards tRNA, although the effect from this single modification is small.

Archaeal or eukaryotic NSun6 distinguish their cognate tRNA substrates
Based on this work and our previous work on eukaryotic NSun6 (33,34), a remarkable aspect of NSun6 is the difference in substrate specificity between archaeal and eukaryotic homologs (23,33). This raises an interesting question of whether archaeal or eukaryotic NSun6 could discriminate each other's tRNA substrates.
First, we verified whether the tRNA substrates of hN-Sun6 could be methylated by PhNSun6. Considering the stability of hctRNAs, we tested the methyl transferase activity of PhNSun6 to hctRNA Thr (AGU), hctRNA Thr (UGU), and hctRNA Cys (GCA) at 65 or 55 • C. The results showed that the tRNA substrates of hNSun6 could not be methylated by PhNSun6 at 65 • C ( Figure 7A) or 55 • C ( Figure 7B). Considering that C2:G71 is the determinant element for PhNSun6 recognition, we speculated that the tRNA substrates of hNSun6 were not distinguished by PhNSun6 because these native tRNAs harbor a G2:C71 base pair.
In conclusion, PhNSun6 or hNSun6 could only methylate its own cognate tRNA substrates, and the two enzymes could not recognize each other's cognate tRNA substrates ( Figure 7D).

tRNA recognition mechanism by archaeal NSun6s
In this study, we identified that PH1991 is the P. horikoshii tRNA:m 5 C72 MTase NSun6, PhNSun6. The tRNA substrates recognized by PhNSun6 were based on (i) the 3 CCA terminus, (ii) the target site C72, (iii) U73 or G73, and (iv) the discriminatory base pair C2:G71 at the acceptor stem. The CCA terminus and the target site C72 are common recognition elements for PhNSun6 and hNSun6. However, they exist in almost all tRNAs, which are not enough to distinguish substrate tRNAs of NSun6. U73/G73 and C2:G71 function as discriminatory elements to identify PhNSun6 substrate tRNAs from nonsubstrate tRNAs.
In the superimposition model of PhNSun6/tRNA, the binding pocket of site 73 is capable of accommodating both U and G, which was in line with our methylation experiment in which eleven PhtRNAs with U73 or G73 could be substrates of PhNSun6. In the model, the base moiety of G71 was not in contact with PhNSun6. However, based on our results, C2:G71 functions as a discriminatory base pair that can select substrate tRNAs from other PhtRNAs. The exact recognition mechanism of PhNSun6 for C2:G71 requires further mechanistic and structural studies.
Based on the PhNSun6/tRNA superimposition model, PhNSun6 does not interact with the D-stem region of the tRNA, because of the lack of the eukaryotic NSun6-specific Lys-rich loop, which is consistent with biochemical data on the effect of the D-stem on activity of PhNSun6. Thus, our biochemical and crystallographic data suggest that PhNSun6 ignores the sequential specificity of the D-stem of PhtRNA substrates. Considering that archaeal NSun6 homologs share high identity with each other, we speculated that the tRNA substrate specificity of PhNSun6 is conserved in other archaeal NSun6 homologs.

A proposed evolutionary model for NSun6
Based on the results of this study, together with previous findings on hNSun6, we proposed an evolutionary model for NSun6 (Figure 8). Originally, the RRM and PUA domain were recruited and inserted into an ancient m 5 C catalytic core to bind RNA substrates, resulting in the formation of NSun6's last universal common ancestor (LUCA). Specifically, for the recognition by NSun6's LUCA of tRNA substrates, the PUA domain recognized the common CCA terminus; the RRM recognized site 73; and the m 5 C catalytic core recognized C72 and base pairs near site 72. Later in evolution, NSun6 divided into two branches: Archaeal NSun6 and eukaryotic NSun6. The divergence of the RRM and PUA domain during evolution, which increased the precision and accuracy of RNA substrate discrimination, archaeal and eukaryotic NSun6 exhibit different discrimination mechanisms for its cognate tRNA substrates. For RNA substrate discrimination, eukaryotic NSun6 recognizes the acceptor stem and D-stem, while archaeal NSun6 only interacts with the acceptor stem. The evolutionary difference between archaeal and eukaryotic NSun6 has led to diverse cognate tRNA recognition mechanisms.

Biological functions of m 5 C72 modification
The presence of m 5 C modifications in RNA has been observed across the three domains of life. However, the tRNA:m 5 C modification is present only in archaea and eukaryotes (31,32,53). The m 5 C modification could increase hydrophobicity and reinforce stacking interactions, probably leading to tRNA stabilization (54). In Saccharomyces cerevisiae, the m 5 C40 modification of tRNA Phe stabilizes the tRNA anticodon stem (55); lack of m 5 C48 and m 5 C49, together with m 7 G46, cause the rapid decay of certain tR-NAs (56).
In archaea, NSun6s are present only in hyperthermophilic species (57). In terms of tRNA canonical functions, m 5 C72 modification does not affect the amino acid accepting activity of tRNAs, but slightly increases the thermal stability of PhtRNAs. Acceptor stem of tRNA is rarely modified, and m 5 C72 is one of the few modifications in this region. It remains possible that m 5 C72 has other biological functions in vivo, which awaits further studies on both archaeal and eukaryotic NSun6s.

DATA AVAILABILITY
Protein Data Bank: atomic coordinates and structure factors for apo PhNSun6 have been deposited with accession code 5ZVD; for the PhNSun6-SAM, PhNSun6-SFG, and PhNSun6-SAH complex under accession code 5ZVG, 5ZVH and 5ZVE, respectively.
Nucleic Acids Research, 2019, Vol. 47, No. 4 2053 Figure 8. A proposed evolutionary model for NSun6. Originally, ancient m 5 C catalytic core recruited and inserted RRM motif and PUA domain, resulting in NSun6's LUCA. Later in evolution, NSun6 divided into two branches: archaeal NSun6 and eukaryotic NSun6. Due to the divergent RRM motif and PUA domain during evolution, archaeal and eukaryotic NSun6 exhibit different discrimination mechanisms for the cognate tRNA substrates. For RNA substrates recognition, eukaryotic NSun6 recognizes the acceptor stem and D-stem region, while archaeal NSun6 only interacts with the acceptor stem region. The tRNA elements that are recognized by archaeal or eukaryotic NSun6 are summarized here and highlighted in sticks.