PylSn and the Homologous N-terminal Domain of Pyrrolysyl-tRNA Synthetase Bind the tRNA That Is Essential for the Genetic Encoding of Pyrrolysine*

Background: Pyrrolysyl-tRNA synthetase attaches pyrrolysine to tRNAPyl. It is encoded by pylS in Archaea but by pylSn and pylSc in Bacteria. Results: PylSn binds tRNAPyl specifically in an electrophoretic mobility shift assay. Conclusion: PylSn and the PylS N terminus participate in binding the tRNA that is key to the genetic encoding of pyrrolysine. Significance: PylSn and the homologous PylS N terminus previously had no known function. Pyrrolysine is represented by an amber codon in genes encoding proteins such as the methylamine methyltransferases present in some Archaea and Bacteria. Pyrrolysyl-tRNA synthetase (PylRS) attaches pyrrolysine to the amber-suppressing tRNAPyl. Archaeal PylRS, encoded by pylS, has a catalytic C-terminal domain but an N-terminal region of unknown function and structure. In Bacteria, homologs of the N- and C-terminal regions of archaeal PylRS are respectively encoded by pylSn and pylSc. We show here that wild type PylS from Methanosarcina barkeri and PylSn from Desulfitobacterium hafniense bind tRNAPyl in EMSA with apparent Kd values of 0.12 and 0.13 μm, respectively. Truncation of the N-terminal region of PylS eliminated detectable tRNAPyl binding as measured by EMSA, but not catalytic activity. A chimeric protein with PylSn fused to the N terminus of truncated PylS regained EMSA-detectable tRNAPyl binding. PylSn did not bind other D. hafniense tRNAs, nor did the competition by the Escherichia coli tRNA pool interfere with tRNAPyl binding. Further indicating the specificity of PylSn interaction with tRNAPyl, substitutions of conserved residues in tRNAPyl in the variable loop, D stem, and T stem and loop had significant impact in binding, whereas those having base changes in the acceptor stem or anticodon stem and loop still retained the ability to complex with PylSn. PylSn and the N terminus of PylS comprise the protein superfamily TIGR03129. The members of this family are not similar to any known RNA-binding protein, but our results suggest their common function involves specific binding of tRNAPyl.

Pyrrolysine has been called "the 22 nd amino acid" (1), because it is the most recently discovered genetically encoded protein residue found in a naturally occurring organism (2). Pyrrolysine was first described in the structure of the methyltransferase initiating monomethylamine metabolism in Methanosarcina barkeri (3), and subsequently residues with the mass of pyrrolysine were demonstrated in the dimethylamine and trimethylamine methyltransferases (4). For each of these proteins, the position of pyrrolysine corresponds in the encoding gene to an amber codon (TAG/UAG) (5,6). Pyrrolysine is inserted into the growing monomethylamine methyltransferase protein with little detectable UAG termination product in Methanosarcina spp. Even foreign genes with in-frame amber codons are expressed in the methanogen, albeit with less efficient UAG translation (7,8).
Underlying UAG translation in Methanosarcinaceae is the pyl gene cluster (see Fig. 1) (9,10). The first gene in the pyl gene cluster is pylT, encoding the tRNA Pyl that possesses the CUA anticodon required to translate amber codons as pyrrolysine. Deletion of pylT renders Methanosarcina acetivorans unable to produce methylamine methyltransferases but with no other apparent defect (11). The secondary structure of tRNA Pyl is unusual, having a small D loop and variable loop and an elongated anticodon stem, as well as lacking the base typically found between the acceptor and D stems (9). Nonetheless, tRNA Pyl assumes the expected L-shaped tertiary structure but with a more compact core relative to most tRNA species (12,13).
Following pylT in the archaeal pyl gene cluster is the pylS gene that encodes a pyrrolysyl-tRNA synthetase able to ligate tRNA Pyl to chemically synthesized pyrrolysine (14) or pyrrolysine analogs (15). Transformation of Escherichia coli with the pylS and pylT genes allows in vivo suppression of UAG codons when supplemented with pyrrolysine (14) or its derivatives (16). Pyrrolysine is made independently of tRNA Pyl by the products of the downstream pylB, pylC, and pylD genes (10,17). Each molecule of pyrrolysine is made from two molecules of lysine (18) with (3R)-3-methyl-D-ornithine an intermediate (18 -20). Further transformation of E. coli with pylB, pylC, and pylD allows E. coli to decode UAG as endogenously produced pyrrolysine (17). The ability of pylS and pylT to encode a functional orthogonal pair within E. coli has led to their application for insertion of modifiable lysine derivatives into recombinant proteins (10,21).
The structure of the Methanosarcina mazei PylS with an N-terminal truncation of 185 amino acids (⌬185PylS) revealed the catalytic site of PylS (27,28). The deletion was necessary to eliminate the highly basic yet hydrophobic N-terminal portion of the protein, whose presence interfered with protein stability and crystallization (29). The structure of D. hafniense PylSc is similar to that of M. mazei ⌬185PylS, although with an apparently tighter binding pocket for pyrrolysine (13,30). PylSc binds tRNA Pyl primarily through interactions with residues of the acceptor and D stem with no direct contact to residues of the T arm, anticodon arm, or variable loop (13).
Loss of the N-terminal portion of archaeal PylS is detrimental to in vivo activity. Truncation or mutation in the first 106 residues from M. barkeri PylS led to the loss of in vivo aminoacyla-tion of tRNA Pyl , as well as UAG codon suppression in a recombinant system (31). However, these experiments did not delineate whether protein structure or activity was compromised by the N-terminal truncation. Recombinant D. hafniense PylSc was reported to charge cognate tRNA Pyl with a pyrrolysine analog in vitro, but pylSc has been reported to produce undetectable or very low level UAG translation in vivo (31). The biochemical function of the N terminus of PylS or PylSn has never been demonstrated.
Here we show using EMSA that PylS can bind with specificity to tRNA Pyl , an activity that is lost when the N-terminal portion of M. barkeri PylS is deleted. Furthermore, we show that the D. hafniense PylSn is also capable of binding tRNA Pyl and that PylSn recognizes tRNA Pyl specifically. This binding is susceptible to mutations in the variable loop and T-arm of tRNA Pyl , areas that do not interact with PylSc. Thus, PylSn and the homologous region of PylS are tRNA Pyl -binding proteins, although neither has detectable homology with known families of RNA-binding proteins.

EXPERIMENTAL PROCEDURES
Expression of Genes and Purification of Products-The D. hafniense PylSc and M. barkeri PylS proteins were purified following recombinant expression of their genes as described previously (14,30). The D. hafniense pylSn gene was amplified by PCR from genomic DNA using Ex Taq (Takara Mirus Bio, Madison, WI) and primers PylSnF and PylSnR (the sequence of these and other primers are listed in supplemental Table S1). The amplified pylSn gene was subsequently ligated into PCR2.1-Topo (Invitrogen) and then inserted into the NcoI and NotI sites of expression vector pET22b(ϩ) (Novagen, Madison, WI) so as to produce sequence encoding PylSn with a C-terminal hexahistidine tag under the control of an inducible T7 promoter. The plasmid was then transformed into E. coli BL21 (DE3) (Stratagene, Santa Clara, CA). The desired sequence of this and the following gene constructs were confirmed by sequencing of both sense and antisense strands using a 3730 DNA sequence analyzer (Applied Biosystems, Carlsbad, CA) at the Ohio State University Plant-Microbe Genomics Facility.
The cells were grown at 37°C in LB broth (10 g/liter tryptone, 5 g/liter yeast extract, and 10 g/liter sodium chloride) containing 100 mg/liter ampicillin. When the culture had grown to early log phase (A 600 ϭ 0.4 -0.6), 1 mM isopropyl ␤-D-1-thiogalactopyranoside was added to induce production of PylSn. The cells were harvested by centrifugation 3.5 h later and then suspended in 500 mM NaCl, 10 mM imidazole, and 20 mM sodium phosphate buffer, pH 7.4, prior to lysis using a French pressure cell. Following centrifugation at 27,000 ϫ g, the supernatant was applied to a 1-ml bed volume nickel-activated HiTrap HP column (GE Healthcare). Bound protein was then eluted with a 40-ml linear gradient of 10 -500 mM imidazole in 500 mM NaCl and 20 mM sodium phosphate, pH 7.4. PylSn eluted at 250 mM imidazole, and the collected peak was 95% pure judged by densitometry of the preparation following separation by SDS-acrylamide gel electrophoresis and subsequent Coomassie staining. PylSn was stored in elution buffer supplemented with 20% glycerol at Ϫ20°C. N-terminally truncated M. barkeri MS PylS (⌬92PylS) was prepared using genomic DNA as template for PCR amplification with primers 92SF and 92SR that resulted in amplification of the gene such that it would produce PylS lacking N-terminal residues 1-92. The amplified truncated pylS gene was inserted into pET22b (ϩ) at the NdeI and XhoI sites so as to produce ⌬92PylS with a hexahistidine C terminus. The ⌬92PylS protein was purified by nickel affinity column as described above and eluted at ϳ250 mM imidazole.
The PylSn⌬120S protein in which PylSn replaces the N terminus of M. barkeri PylS was made as follows. The M. barkeri PylS gene was amplified from genomic DNA using FusCF and FusCR and cloned into pET15b using the NcoI and BamH1 sites. This resulted in a truncated pylS gene whose product would lack the first 120 N-terminal residues but have a C-terminal hexahistidine tag. The D. hafniense pylSn gene was then amplified from genomic DNA using primers FusNF and FusNR and then ligated into the pET15b clone bearing the inserted truncated PylS gene using the NcoI and EcoRV sites. The resultant plasmid was transformed into E. coli BL21 and used to produce the purified fusion protein designated PylSn⌬120S protein using the methods outlined above. The PylSn⌬120S eluted from the nickel affinity column again at ϳ250 mM histidine.
Recombinant D. hafniense tRNA Pyl was produced in E. coli cells bearing the pETDuet-1 plasmid in which the genomic D. hafniense pylT gene was inserted into the XbaI site lying directly behind the T7 promoter. The pylT gene was amplified from genomic DNA using primers PylTF and PylTR. The tRNA Pyl was then isolated with the E. coli RNA pool as described below.
E. coli RNA pool isolation-E. coli BL21 (DE3) was grown in LB broth at 37°C to an A 600 of ϳ0.5 and then centrifuged at 5000 ϫ g. The cell pellet was suspended in 10 mM EDTA in 0.3 M sodium acetate, pH 4.5, before extraction with phenol:chloroform (5:1) equilibrated with the same buffer (Invitrogen). Following centrifugation at 18,000 ϫ g for 15 min, the aqueous layer was extracted a second time. The RNA was then precipitated by addition of a three-fold volume of 100% ethanol, resuspended in 0.3 M sodium acetate, pH 4.5, and precipitated again. The final pellet was air-dried and dissolved in TE buffer (10 mM Tris-Cl, 1 mM EDTA, pH 7.0). The isolated cellular RNA pool contained 30% tRNA when examined by fluorescent intensity of agarose electrophoretic gels stained with ethidium bromide. The deacylated tRNA pool was prepared by adding equal volume of deacylation buffer (100 mM NaCl in Tris-HCl, pH 9.5) and then heating at 70°C for 30 min. The deacylated tRNA pool was exchanged into TE buffer using Microspin G-25 columns (GE Healthcare).
In Vitro Transcription of tRNA-The tRNAs used in this work were prepared following modifications of published procedures (32,33). An oligonucleotide corresponding to the 5Ј end of the tRNA gene was made such that it also contained an upstream sequence with a T7 promoter. A second oligonucleotide corresponding to the 3Ј end of the tRNA gene was made such that it overlapped the first oligonucleotide for 10 -12 nucleotides centered on the sequence encoding the anticodon loop. The annealed oligonucleotides were extended using Sequenase (Affymetrix, Santa Clara, CA) to make a template for run-off transcription employing the Megashortscript T7 kit (Invitrogen) according to the manufacturer's recommendations and having 0.5 M template incubated for a reaction time of 4 h. The resultant tRNA transcript was purified on a denaturing 14% polyacrylamide gel and recovered by elution in TE buffer for 4 h at 22°C, followed by ethanol precipitation. The recovered tRNA transcript pellet was dissolved in TE buffer and stored at Ϫ20°C. To produce 32 P-tRNA, [ 32 P]ATP (20 Ci; final specific radioactivity of 133 Ci/nmol; PerkinElmer Life Sciences) was added to the transcription reaction.
Assay for Amino Acid Activation and tRNA Charging-Aminoacylation of tRNA Pyl was detected using acid-urea acrylamide gel electrophoresis as described previously (14). The reactions were conducted at 37°C in 50 mM KCl, 10 mM MgCl 2 , 5 mM ATP, and 5 mM dithiothreitol in 10 mM Hepes buffer, pH 7.2. Additionally, assays contained as substrate 50 M pyrrolysine (kind gift of Michael Chan) (34) and recombinant tRNA Pyl present in the E. coli RNA pool (1 mg/ml total RNA). Enzyme was added as indicated in the text. All of the reactions were initiated with enzyme and ended after 30 min by the addition of 7 M urea in 0.3 M sodium acetate buffer, pH 5.0. The reaction mixture was then loaded on a 14% polyacrylamide gel containing 7 M urea in 0.3 M sodium acetate, pH 5.0, and ran at 50 V in the same acetate buffer for 16 h at 4°C. The tRNA was then electroblotted onto a Hybond N ϩ membrane (GE Healthcare) and detected by hybridization with a 5Ј-labeled 32 P-tRNA Pyl probe employing a Storm phosphorimager (GE Healthcare).
Amino acid activation was assayed using ATP-pyrophosphate exchange, essentially as performed previously (14,35). All assays were performed at 37°C in duplicate and were repeated with at least two independent enzyme preparations. Assays contained 20 mM Hepes (pH 7.2), 10 mM MgCl 2 , 25 mM KCl, 4 mM DTT, 5 mM ATP, 2 mM [ 32 P]PPi (5-15 dpms/pmol; PerkinElmer Life Sciences). In standard assays, 50 M pyrrolysine and 0.5-1 M enzymes were added. Identical reactions without enzyme were set up as negative controls. To determine kinetic parameters of PylSc, pyrrolysine concentrations were varied over the range of 0.1-4 K m , and the data were analyzed by nonlinear regression using Prism software (GraphPad, San Diego, CA).
Electrophoretic Mobility Shift Assay-Binding of 32 P-labeled tRNA by various proteins was assayed by EMSA. Protein-tRNA complexes were formed by incubation of 17-50 nM 32 P-tRNA with various amount of protein on ice for 15 min in 30 l of binding buffer (50 mM KCl, 10 mM MgCl 2 , and 5 mM DTT in 10 mM Hepes, pH 7.2). As indicated, in experiments having high amounts of PylSn, 33 g/ml heparin was also added to the reaction buffer to suppress protein aggregation. The tRNA-protein mixture was then loaded onto a 10% polyacrylamide gel containing 10% glycerol, 1 mM EDTA in 45 mM Tris borate buffer, pH 8.0, which was electrophoresed in the same buffer at 175 V for 2 h at 4°C. The gel was then exposed to a phosphorus screen for 32 P-tRNA and 32 P-tRNA-protein complex detection using a Storm phosphorimager. The density of individual bands was quantified by ImageQuant (GE Healthcare). The apparent K d values of tRNA-protein complexes were obtained by nonlinear regression of the data using Prism software. For assays with variant tRNA Pyl for which PylSn had poor affinity, the apparent K d was estimated by applying the data to following equation: where is the fractional saturation of tRNA-protein complex seen at the upper end of the PylSn concentrations tested (36).

RESULTS
The N-terminal Region of PylS Influences Complex Formation with tRNA Pyl -We first tested whether binding of PylS with its cognate tRNA could be detected by electrophoretic mobility shift assay, because this has not previously been documented. Increasing concentrations of M. barkeri MS PylS were incubated with in vitro transcribed M. barkeri 32 P-tRNA Pyl (20 nM) prior to electrophoresis in a nondenaturing 10% acrylamide gel. Migration of the tRNA Pyl band was retarded in a manner allowing fit to a binding curve having an apparent K d for the PylS-tRNA Pyl complex of 0.12 Ϯ 0.01 M (Fig. 2A).
N-terminally truncated PylS was unable to form a stable complex with tRNA Pyl detectable by EMSA. We made three constructs of the truncated pylS genes with deletions that would correspond to the first 92, 119, or 145 residues of the protein. Of these, only the first construct, ⌬92PylS, produced a stable form of the protein that could be successfully purified. However, ⌬92PylS could not form a stable complex in EMSA experiments performed under the same conditions as with intact PylS. Even with 10 M ⌬92PylS, which was the highest concentration tested, no shift of the free tRNA Pyl band was apparent (Fig. 2B). The lack of activity in these assays was not due to an improperly folded protein, as evidenced by the retention of pyrrolysine-dependent PP i -ATP exchange activity, which ⌬92PylS catalyzed with a k cat of 14.4 Ϯ 0.6 min Ϫ1 , relative to 6 min Ϫ1 observed with wild type PylS preparations. ⌬92PylS also aminoacylated tRNA Pyl in a manner indistinguishable from wild type PylS when assayed using acid-urea electrophoresis (data not shown). These results suggested that although the N terminus is not required for activity in these assays, the presence of the N terminus sufficiently stabilized interaction with tRNA Pyl to form a complex detectable by EMSA.
Bacterial PylSn Binds tRNA Pyl -The co-crystalization of PylSc and tRNA Pyl , as well as modeling studies, revealed how the C terminus of PylS interacts with tRNA Pyl (13,27). However, the above experiments indicate that N-terminal domain might recognize tRNA Pyl as well. Unfortunately, our attempts to generate a stable protein with only the N-terminal region of PylS did not succeed. Therefore, we turned to PylSn, the bacterial homolog of the N terminus of archaeal PylS domain, which is encoded as a separate gene independent of the gene encoding the catalytic PylSc domain in some Bacteria. D. hafniense PylSn aligns with residues 1-95 of M. barkeri PylS (supplemental Fig.  S1).
Recombinant D. hafniense PylSn bearing a C-terminal hexahistidine tag was relatively stable if kept at low protein concentration at high ionic strength in stock solutions supplemented with sufficient glycerol to keep the solution from freezing at Ϫ20°C. Even under these conditions, stored protein aggregated and precipitated. Therefore, fresh preparations were consistently used in studies of PylSn binding to tRNA. We noted that aggregated PylSn nonspecifically bound nucleic acid in high molecular weight complexes do not enter the gel during EMSA. However, such artifacts were minimized with fresh preparations stored as indicated.
The addition of freshly prepared PylSn to assay mixtures with in vitro transcribed D. hafniense tRNA Pyl resulted in a pronounced shift in position of tRNA Pyl following migration in PAGE gels (Fig. 3A), indicating that PylSn forms a distinct complex with tRNA Pyl . No other band was visible in the gel other than those for tRNA Pyl and PylSn-tRNA Pyl complex, and significant aggregation of protein-tRNA complex was not observed at the tRNA and protein concentrations tested. Increasing amounts of PylSn were added to 17 nM tRNA Pyl to determine the apparent K d for the formation of the PylSn-tRNA Pyl complex. The data best fit a nonlinear regression curve corresponding to a K d of 0.13 Ϯ 0.01 M.
Similar to the results obtained with ⌬92PylS, bacterial PylSc did not form an EMSA-detectable complex when incubated with D. hafniense tRNA Pyl (Fig. 3B), nor did the addition of PylSc cause a change in the migration of the PylSn-tRNA Pyl complex. This was not due to an inactive enzyme, D. hafniense PylSc carried out an PP i -ATP exchange reaction dependent on pyrrolysine (k cat ϭ 19.5 Ϯ 1.2 min Ϫ1 , K m ϭ 44 Ϯ 4 M) and charged D. hafniense tRNA Pyl recombinantly produced in E. coli with pyrrolysine in a manner indistinguishable from wild PylS when assayed using acid-urea electrophoresis.
PylSn could rescue the ability of N-terminally truncated M. barkeri PylS to bind tRNA Pyl as detectable by EMSA. Six different chimeric proteins were prepared in which variable lengths of the N-terminal region of PylS were replaced by PylSn. Only one of these, the fusion protein PylSn⌬120S, in which PylSn replaced the first 120 residues of PylS, could be successfully purified. PylSn⌬120S formed an EMSA-detectable complex when incubated with D. hafniense tRNA Pyl , and the apparent K d for the formation of the PylSn⌬120S-tRNA Pyl complex was measured as 0.35 Ϯ 0.03 M (data not shown). Unlike wild type, the fusion protein was not able to detectably aminoacylate tRNA Pyl , possibly because of altered spacing of the N-and C-terminal tRNA Pyl -binding domains.
PylSn Binding of tRNA Is Selective for tRNA Pyl -The above data indicated that PylSn binds tRNA Pyl but did not reveal whether binding was specific for tRNA Pyl or could be generalized to other tRNAs. Therefore, we evaluated the ability of D. hafniense PylSn to bind various in vitro transcribed tRNA species from D. hafniense including tRNA GAG (Leu) (Fig. 4), tRNA UCC (Gly), tRNA UUU (Lys), and tRNA UGC (Ala) (data not shown). We also tested Bos taurus mitochondrial tRNA UGA (Ser) (Fig. 4), which shares some of the unique features of tRNA Pyl including a three-base variable loop and elongated anticodon stem (37). When assayed under conditions where PylSn readily caused a gel shift of free tRNA Pyl , PylSn did not cause a shift analogous to the PylSn-tRNA Pyl complex with these tRNA species.
The specificity of PylSn binding to D. hafniense tRNA Pyl was further tested by competition assays in which increasing concentration of the crude deacylated tRNA pool extracted from E. coli was incubated with 17 nM of 32 P-tRNA Pyl transcript prior to the addition of 1 M of PylSn. No decrease in the band representing the tRNA Pyl -PylSn complex could be detected because the estimated tRNA concentration increased from 0 to 1.7 M; nor did the free tRNA Pyl concentration appear to increase (Fig. 5). These data suggest that PylSn can form a stable complex with tRNA Pyl even in the presence of a 100-fold molar excess of other tRNA species.
Certain Base Substitutions in tRNA Pyl Compromise Binding by PylSn-As a further test of the specificity of PylSn for tRNA-Pyl , we produced in vitro transcribed variants of D. hafniense tRNA Pyl having base substitutions at the positions indicated in Figs. 6 and 7. The positions changed were chosen based on their high conservation in known examples of tRNA Pyl because these bases might serve as determinants for PylSn binding or disrupt tertiary structure important for PylSn recognition. The variant tRNA Pyl were then compared in the ability to form complexes with wild type D. hafniense tRNA Pyl in EMSA. Representative EMSA gels are shown in Fig. 8. We screened mutations under conditions that would detect base changes yielding strong bind-ing defects, i.e. the variant tRNA Pyl -PylSn complex must have a K d greater than 2.5 M to be distinguishable from wild type.
No bases are universally conserved in the anticodon stem of tRNA Pyl , although several base pairs are found in the majority of examples, such as G29:C41 and A31:U39 (10). However, substitution of either with an alternative base pair did not result in a severe defect in PylSn binding under our assay conditions (Figs. 6 and 8). Furthermore, outright deletion of either base pair did not affect binding of PylSn in our assays relative to wild type, indicating the elongated anticodon stem of tRNA Pyl is not essential for PylSn binding. Six of the seven residues in the   5 and 6). Both panels are from the same polyacrylamide gel. Mutations of these residues are deleterious to the in vivo ability of tRNA Pyl to act as an amber suppressor in E. coli, especially U33 and A37, the residues that flank the anticodon (38). However, none of the four residues flanking the anticodon are essential for binding of tRNA Pyl by PylSn (Fig. 6). The CUA anticodon itself was changed to AGC, and this construct is still readily bound by PylSn. Similarly, a change to the acceptor stem did not detectably change PylSn binding, as shown by substitution of C71 and C72, along with the discriminator base G73, with the sequence UUA.
In contrast, changes to the D stem strongly affected PylSn binding. Three of the four base pairs in the D stem are universally conserved in tRNA Pyl and the remaining pair, A11:U24, is found in 70% of bacterial tRNA Pyl sequences. Mutation of the two base pairs distal to the D loop (Fig. 8) led to weak binding (K d ϭ ϳ30 M), whereas independent mutation of the two base pairs proximal to the D loop (Fig. 7) led to similarly weak (U12: A23 to C12:G23) or no detectable binding (C13:G22 to U13: A22). The assay conditions were set up such that a shift of 10% of the tRNA would have indicated an apparent K d of 200 M.
One of the most distinctive features of tRNA Pyl is the very small variable loop composed of only three residues (Fig. 7). Mutations in any of these three residues significantly decreased the affinity of PylSn for tRNA Pyl (Figs. 7 and 8). The least disruptive mutation was C45A, which nonetheless had only weak binding with an estimated K d of 10 M. In contrast, tRNA Pyl having A46G or G47U substitution did not detectably interact with PylSn in EMSA. Insertion of an additional base into the variable loop also led to no detectable binding. The detrimental effect of substitution of the latter two residues may reflect their important role in stabilizing the L-shaped tertiary structure by interaction with the D stem helix (13), indicating the importance of the tRNA Pyl structure for binding to PylSn.
Mutations within the T stem and T loop also strongly affected binding between PylSn and tRNA Pyl . Two highly conserved residue pairs are present in the T-stem, namely G52:C62 and G53:C61. Mutation of the first or second GC pair reduced affinity of PylSn for the tRNA with apparent K d values of ϳ10 and 40 M, respectively (Fig. 7). Mutations in the T-loop also reduced PylSn affinity for the tRNA. For example, no interaction of tRNA Pyl with PylSn was detectable (K d Ͼ 200 M) following substitution of A58 with C. This effect may be due to the need for A58 interaction with U54 in the properly folded tRNA (13); however, alteration of U54 to G led to a more modest decrease in affinity, with a K d of ϳ20 -30 M, perhaps indicating that A58 might be involved in direct PylSn binding and recognition. Mutation of A58 was also reported to have similar detrimental effects on archaeal PylRS function in vivo (38).

DISCUSSION
PylSn has an unknown structure and has had no demonstrated function. Here we show for the first time that the bacterial PylSn forms a complex with tRNA Pyl , but not noncognate   (13). The base changes that strongly affected PylSn binding to tRNA, but still showed detectable binding in EMSA, are underlined. Base changes that eliminated the tRNA ability to form an EMSA-detectable complex with PylSn are boxed.
tRNAs or tRNA Pyl with certain bases modified. Loss of the N terminus of archaeal PylS leads to a loss of activity in EMSA, indicating that this homolog of PylSn has the same capacity to form a stable complex with tRNA Pyl . D. hafniense PylSn used as a search query for a conserved domain BLAST search at the National Center for Biotechnology Information revealed only a single hit against one domain superfamily, TIGR03912, which includes the discrete pylSn gene products of Bacteria as well as the N termini of archaeal PylS proteins. A further search using either the BLAST or PSI-BLAST programs did not reveal significant homology to a known RNA-binding protein. PylSn and the N terminus of PylS thus appear to represent a novel tRNA binding domain.
A recent survey of the genomes hosted at NCBI revealed those from 12 species of Bacteria and Archaea that possessed the pyrrolysine biosynthetic genes pylBCD (10). All these genomes encode a TIGR03912 domain as the N-terminal portion of pylS or as a separate pylSn gene. Those with pylSn also possess the pylSc gene. This implies the tRNA Pyl binding ability of the TIGR03912 family plays a role in the genetic encoding of tRNA Pyl sufficient to lead to the retention of the independent gene or linked domain along with the other pyl genes. Given the current data, the simplest hypothesis is that the binding of tRNA Pyl by PylSn, or the N terminus of PylS, plays a role in facilitating the charging of tRNA Pyl by the catalytic domain, possibly by increasing its affinity for tRNA Pyl . The published K d of PylSc for D. hafniense tRNA Pyl is 6.9 M. This is remarkably high in comparison with the K d values measured for archaeal PylS and tRNA Pyl , which are typically 10-fold lower (31). The PylSn-tRNA Pyl complex has a K d in this same low range. Notably, loss of the ability of PylS to form an electrophoretically detectable complex with tRNA Pyl occurs upon deletion of the N-terminal domain, and PylSc was not capable of forming a complex with tRNA Pyl detectable by electromobility shift, suggesting higher stability of the PylSn and tRNA Pyl complex.
The flexible linker connecting the catalytic and N-terminal domains of PylS provides a direct connection between the high affinity tRNA Pyl -binding domain and the catalytic domain, whose complex with tRNA Pyl is less stable. However, for PylSn to function similarly to increase the overall affinity of bacterial pyrrolysyl-tRNA synthetase for tRNA Pyl , it would presumably require the formation of a ternary complex of the two proteins and the tRNA. Such a complex may or may not include direct interaction of the two proteins, because PylSn could act to change the conformation of tRNA Pyl , so as to increase the binding affinity of PylSc.
Substitutions of highly conserved residues in tRNA Pyl revealed portions of the tRNA whose sequence are critical for binding of tRNA Pyl by PylSn. Substitution of conserved residue pairs in the D stem, residues of which directly interact with PylSc (13), strongly influence the binding of PylSn to tRNA Pyl (Figs. 6 and 8). PylSc binds the D stem from the minor groove side, leaving it possible that the decreased PylSn affinity for these variants reflects PylSn interaction with the opposite side of the D stem. However, both A46 and G47 of the variable loop respectively interact with U12:A23 and C13:G22 of the D stem (13), and it is likely these mutations disrupt the tertiary structure of tRNA Pyl . On the other hand, the G10:C25 and A11:U24 pairs do not directly participate in this interaction, and changes of these base pairs still strongly affected PylSn affinity. Interestingly, base substitutions within the T stem and loop also strongly influenced tRNA Pyl binding to PylSn, indicating that parts of the tRNA Pyl core that have no reported contact with PylSc are important for PylSn recognition. The requirement for A58 is especially interesting; given that this residue does not directly pair or stack with any other residue in the core, it may represent a determinant directly recognized by PylSn.
The cloverleaf structure of tRNA Pyl is distinguished by the exceptionally small variable loop, as well as an elongated anticodon stem. Only the sequence of one of these elements appears crucial for PylSn binding. Substitution of any base within the variable loop, or insertion of additional base into this region, either eliminated or compromised interaction between tRNA Pyl In contrast, the elongated anticodon stem and anticodon loop are not a strict requirement for PylSn binding; deletion of a base pair in the stem at most moderately interfered with PylSn binding. Although the effects these changes of tRNA Pyl have on PylSn binding may reflect effects on structure or direct base interactions, the sensitivity of PylSn binding to changes in tRNA Pyl D and T stems and T loop further illustrate the specificity of PylSn for tRNA Pyl .
Although the retention of the pylSn gene (or the corresponding part of archaeal pylS) in presently sequenced genomes suggests a long term advantage to retention of this tRNA Pyl -bind- ing domain, the physiological advantage is not yet apparent. In vitro, PylSc acts as an aminoacyl-tRNA synthetase, when presented with pyrrolysine, or pyrrolysine analogs, and tRNA Pyl (this work and Ref. 39). Similarly, we show here that an N-terminal truncation of archaeal PylS retains the ability to activate pyrrolysine and charge tRNA Pyl . However, the current data from recombinant systems in which PylS or PylSc and tRNA Pyl are employed in vivo to suppress amber codons in reporter genes yield conflicting information. PylS was reported to lose the ability to function in vivo when pylS suffers deletion of the N-terminal domain, and pylSc did not detectably function in amber suppression leading to expression of a reporter gene (31). A very sensitive genetic assay revealed weak UAG translation with pylSc, which was not enhanced by co-expression with pylSn (13). A more recent paper indicated that PylSc by itself is capable of supporting fairly high levels of amber codon translation in E. coli (40). Our own in vivo experiments agree with the latter (data not shown) and suggest that PylSc does not require PylSn to function in a recombinant system. This suggests subtle yet important effects of PylSn on PylSc function, or of the N-terminal region of PylS on catalytic domain, have led to retention of this tRNA Pyl -binding protein/ domain in genomes. The tighter and more stable binding displayed by PylSn for tRNA Pyl relative to PylSc may influence the kinetics of tRNA Pyl acylation so that lower expression of pylT allows an adequate pool of pyrrolysyl-tRNA Pyl to fuel UAG translation. Alternatively, orthogonality of PylSc and tRNA Pyl has yet to be formally proven, and it is possible the specificity of tRNA Pyl binding by PylSn and its homologs may increase the fidelity of pyrrolysine incorporation into protein. Even a small error rate in which pyrrolysylation of a noncognate tRNA occurs could result in a deleterious physiological effect. In any case, it is clear that the specific binding of tRNA Pyl by the N terminus of PylS, or by PylSn, is of sufficient merit to render retention of their encoding genes a positive trait in those organisms that decode UAG as pyrrolysine.