Human Lysyl-tRNA Synthetase Accepts Nucleotide 73 Variants and Rescues Escherichia coli Double-defective Mutant*

The nucleotide 73 (N73) “discriminator” base in the acceptor stem is a key element for efficient and specific aminoacylation of tRNAs and of microhelix substrates derived from tRNA acceptor stems. This nucleotide was possibly one of the first to be used for differentiating among groups of early RNA substrates by tRNA synthetases. In contrast to many other synthetases, we report here that the class II human lysyl-tRNA synthetase is relatively insensitive to the nature of N73. We cloned, sequenced, and expressed the enzyme, which is a close homologue of the class II yeast aspartyl-tRNA synthetase whose co-crystal structure (with tRNAAsp) is known. The latter enzyme has a strong requirement for G73, which interacts with 4 of the 14 residues within the “motif 2” loop of the enzyme. Even though eukaryotic lysine tRNAs also encode G73, the motif 2 loop sequence of lysyl-tRNA synthetase differs at multiple positions from that of the aspartate enzyme. Indeed, the recombinant human lysine enzyme shows little preference for G, and even charges human tRNA transcripts encoding the A73 found in E. coli lysine tRNAs. Moreover, while the lysine enzyme is the only one in E. coli to be encoded by two separate genes, a double mutant that disables both genes is complemented by a cDNA expressing the human protein. Thus, the sequence of the loop of motif 2 of human lysyl-tRNA synthetase specifies a structural variation that accommodates nucleotide degeneracy at position 73. This sequence might be used as a starting point for obtaining highly specific interactions with any given N73 by simple amino acid replacements.

The canonical tRNA structure contains 76 nucleotides in a cloverleaf secondary structure that terminates at the 3Ј-end in the N 73 CCA 76 OH single-stranded tetranucleotide acceptor stem (where N indicates any nucleotide), and also contains the TC, dihydrouridine, and anticodon stem-loop domains. In three dimensions, the four helical stems of the cloverleaf are organized into a two-domain L-shaped structure that places the anticodon and amino acid attachment site in separate domains where they are about 76 Å apart (1). The N 73 CCA 76 OH tetranucleotide is at the 3Ј-end of the 12-base pair minihelix domain, which is formed by coaxial stacking of the acceptor stem with the TC stem-loop (2).
For at least half of the tRNA synthetases, the minihelix, and a 7-base pair microhelix comprising the acceptor stem alone, are substrates for specific aminoacylation (3). Nucleotides required for aminoacylation of these substrates generally include N 73 and two or three base pairs in the helix. The constellation of acceptor helix atoms needed for aminoacylation of RNA oligonucleotide substrates constitutes an operational RNA code for amino acids that may have been part of the earliest aminoacylation systems (4,5).
The strong dependence of aminoacylation on the nature of N 73 has also been seen with the full tRNAs. The N 73 "discriminator" nucleotide was recognized early as a possible site for synthetases to discriminate between groups of tRNAs, according to their amino acid specificities (6). N 73 -dependent aminoacylation is found for synthetases that do and do not make contact with the anticodon trinucleotide in the second domain of the tRNA structure. This dependence can be so strong that, even for synthetases for which interactions with the anticodon trinucleotide contribute significantly to aminoacylation efficiency, an incorrect N 73 nucleotide can obscure positive interactions with the anticodon (7).
To understand the molecular basis for recognition of the discriminator base, an especially useful paradigm is provided by the crystal structure of yeast aspartyl-tRNA synthetase complexed with tRNA Asp (8,9). This protein is one of ten class II tRNA synthetases that are characterized by three highly degenerate sequence motifs in the class-defining active site domain (10). Motifs 1, 2, and 3 make up a helix-loop-strand, strand-loop-strand, and strand-helix, respectively. In the cocrystal, four residues (Glu, Asn, Ser, and Thr) in the 14-amino acid loop of motif 2 provide interactions with G 73 of tRNA Asp . These include a backbone contact by Asn and a hydrogen bond between the carboxyl of Glu to the exocyclic N-2 of G, and between the OH groups of Ser and Thr with N-1 and O-6 of the base. These interactions are consistent with the high sensitivity of aminoacylation of minihelices or of tRNA Asp transcripts to the nature of N 73 (11). High sensitivity of aminoacylation to the nature of N 73 is also seen with other class II enzymes (12)(13)(14)(15)(16)(17)(18)(19)(20)(21).
The crystal structure of the class II Escherichia coli lysyl-tRNA synthetase (LysRS) 1 has also been solved to 2.8-Å reso-lution (22). While within class II much diversity in structure and sequence is seen, the structure of the lysine enzyme is a close homologue to that of aspartyl-tRNA synthetase, thus establishing that the two enzymes are more related to each other than to most other class II enzymes. While a co-crystal with tRNA Lys in the region of the acceptor stem has not been solved, the close similarity of the lysine enzyme to aspartyl-tRNA synthetase suggests that tRNA interactions are likely to be similar. In particular, while the sizes and sequences of the loop of motif 2 are highly variable among class II synthetases, aspartyl and lysyl-tRNA synthetases align well throughout this region.
The importance of A 73 in the E. coli lysine system was suggested from in vitro (23) and in vivo (24) studies. While G 73 is conserved among prokaryotic and eukaryotic aspartate tRNAs, the conserved A 73 of prokaryotic lysine tRNAs is replaced by a G that is conserved among cytoplasmic eukaryotic lysine tRNAs (25). In addition, despite the close similarity of the aspartyl-and lysyl-tRNA synthetase structure, the sequence of the loop of motif 2 in Saccharomyces cerevisiae lysyl-tRNA synthetase differs at eight positions from that of the aspartate enzyme, including three of the four positions that contact G 73 of tRNA Asp (8,26). This observation suggests that a significantly different motif 2 sequence in eukaryotic lysyl-tRNA synthetases could recognize G 73 or, alternatively, that the eukaryotic lysine enzyme, unlike others, does not have a high specificity for the discriminator base.
Because of our interest in the development of aminoacylation systems throughout evolution, and in gaining an understanding of species barriers to aminoacylation, we focused on human LysRS as an example from the highest eukaryotes. The sequence of the loop of motif 2 of the human enzyme provided another opportunity for assessment of the relationship (if any) between the sequence of that loop and the nucleotide specificity at position 73. In addition, we wanted to determine whether the A 73 of the prokaryotic lysine tRNAs would block aminoacylation of E. coli tRNA Lys by the human protein. A discriminator-basedependent species barrier to aminoacylation had been seen earlier with the human and E. coli glycyl-tRNA synthetases (15,27).

EXPERIMENTAL PROCEDURES
Isolation of Human cDNA for Lysyl-tRNA Synthetase from a cDNA Library-To obtain a probe for library screening, cross-species PCR was performed as described (27)(28)(29). Two conserved regions in LysRS (indicated in Fig. 1) were selected from four sequences that were available, and three degenerate primers were synthesized for these regions. Using cDNA synthesized from the mRNA of a human fetal fibroblast cell line as a template, cross-species PCRs were performed for 35 cycles of 30 s at 94°C, 30 s at 55°C, and 90 s at 72°C. A 0.25-kbp DNA fragment that was amplified was then cloned and used as a probe for screening of a cDNA DR2 phage library constructed from human fetal brain (5Ј-Stretch cDNA library, CLONTECH). Ten positive clones were obtained from about 2 ϫ 10 6 plaques.
DNA Sequencing and 5Ј-and 3Ј-Rapid Amplification of cDNA Ends (RACE) PCR-DNA was prepared from each phage lysate. The longest insert to be observed from some of the clones was 1.4 kbp. The insert was digested with BamHI and XbaI and was subcloned into the BamHI-XbaI sites of pBluescript KS(ϩ) to construct plasmid pM102. Several subclones were constructed from pM102 using standard methods, and their sequences were determined. Comparison of the deduced amino acid sequence with those of LysRSs from other organisms showed sequences coding for the human lysine enzyme. However, the 1.4-kbp insert lacked 5Ј-coding and 3Ј-noncoding regions of the human coding sequence. Therefore, both ends of the cDNA were extended by the modified PCR method (30). Fragments of 0.5 and 0.16 kbp, respectively, were obtained from the 5Ј-and 3Ј-RACE PCR amplification (CLON-TECH), resulting in a total length of the combined sequences of 1997 nucleotides.
Amino Acid Sequence Analysis-Multiple sequence alignments were performed using the PILEUP program (Genetics Computer Group, Madison, WI) that makes alignments based on the method of Needleman and Wunsch (31).
Construction of Expression Vectors-The cDNA insert of pM102 and the 0.5-kbp fragment of the 5Ј-RACE PCR product were assembled to construct plasmid pM109 that contains a full open reading frame of LysRS. The gene fragment that encodes codons 1-597 or 66 -597 was cloned behind the T7 promoter of plasmid vector pET-3a (32) to give plasmid pM116 and pM131, respectively. Plasmid pM116 produces full-length human lysyl-tRNA synthetase (hLysRS), and plasmid pM131 expresses a protein that is missing 65 residues at the N terminus (hLysRS-⌬N65). Codons 66 -597 of hLysRS were also cloned under the lac promoter of plasmid pTZ19R (which in addition harbors the gene for ampicillin resistance) (33) to construct plasmid pM130.
To construct oligohistidine-tagged hLysRS, a 1.8-kbp fragment from pM116 was cloned into pKS583, 2 which is a derivative of pET19b (32). The resultant plasmid pM368 produces a fusion protein and containing N-terminal MRGSHHHHHHSSGWVD sequence appended to fulllength hLysRS.
In Vivo Complementation Test-To test cross-species complementation, a lysS and lysU double-defective strain, PAL⌬S⌬UTR/pMAK705 lysU ϩ (constructed by Chen et al.;Ref. 34), was used. The growth of this strain is maintained by a wild type copy of lysU on a temperaturesensitive plasmid pMAK705 (Cm r ) (34), so that the strain can grow only at the lower temperature (30°C). Strain PAL⌬S⌬UTR/pMAK705 lysU ϩ was transformed to ampicillin resistance at 30°C with pM130 (which produces the 66 -597 fragment of hLysRS; see above) or with pTZ19R (33) (as a control) and then scored for colony formation at 42°C on LB plates containing ampicillin. A similar set of experiments was carried out with an E. coli strain, pM131/PAL⌬S⌬UTR (DE3), which was constructed by lysogenization with phage DE3 that produces T7 RNA polymerase (32). Plasmid pM131 (see above) was used for the DE3lysogenized tester strain. Plasmid pKS506, which contains E. coli lysU under a tetracycline promoter in pBR322 (35), 2 and pBR322 were used as controls.
Following lysozyme treatment, cells were sonicated gently and spun at 14,000 ϫ g for 30 min to pellet cell debris. The supernatant was partially purified by a 30 -60% ammonium sulfate cut and dialyzed overnight against buffer A (50 mM K 2 HPO 4 (pH 7.5), 2 mM ␤-ME). The sample was further purified on a HiLoad (26/10) Q-Sepharose Fast Flow FPLC anion exchange chromatography (Pharmacia Biotech Inc.). The bound protein was eluted by application of a linear gradient of NaCl from 0 to 300 mM (in buffer A). Peak fractions were pooled and concentrated using Centricon 30 (Amicon) concentrators. The sample was then injected onto a HiLoad (16/60) Superdex 200 prep grade sizing column (Pharmacia) preequilibrated in 50 mM HEPES (pH 7.5), 100 mM NaCl, 2 mM ␤-ME. The sample was eluted with 120 ml of the same buffer. The protein was exchanged into 100 mM HEPES (pH 7.5), 6 mM dithiothreitol, and glycerol was added to 50% concentration by volume. Samples were stored at Ϫ80°C. The above purification yielded approximately 3 mg of hLysRS-⌬N65 that was judged to be Ͼ95% pure by SDS-polyacrylamide gel electrophoresis. (Purified hLysRS-⌬N65 migrates with an apparent molecular mass of 61 kDa.) The histidine-tagged full-length hLysRS was purified using a nickelnitrilotriacetic acid resin (QIAGEN) according to the manufacturer's instructions. Enzyme concentrations were estimated by the Bradford method using the Bio-Rad protein assay kit and bovine serum albumin as the standard (36).
Transfer RNA Preparation and Mutagenesis-To prepare unmodified human tRNA Lys,3 , a DNA insert containing a T7 RNA polymerase promoter and the gene encoding the tRNA was made by ligating six chemically synthesized oligodeoxynucleotides together as described previously (37). The insert was ligated into EcoRI/BamHI-digested plasmid pVAL119 (38), which was a gift of Dr. Jack Horowitz (Iowa State University, Ames, IA) to generate plasmid pLYSF119. FokI-digested pLYSF119 was used to prepare tRNA Lys,3 by in vitro transcription according to established procedures (12). T7 RNA polymerase was purified according to Grodberg and Dunn (39) from E. coli strain BL21/ pAR 1219, a gift of F. William Studier (Brookhaven National Laboratory, Upton, NY). RNA transcripts were purified on 16% polyacrylamide gels and stored in 10 mM Tris-HCl/1 mM EDTA (pH 8.0) at Ϫ20°C. Prior to aminoacylation assays, tRNAs were renatured in the presence of MgCl 2 as described previously (40). N 73 variants of tRNA Lys,3 were made by overlap extension PCR mutagenesis and were verified by sequencing. A plasmid containing the gene encoding E. coli tRNA Lys in front of a T7 RNA polymerase promoter was obtained as a gift from Dr. Margaret Saks (California Institute of Technology, Pasadena, CA). Digestion with BstNI allowed preparation of the unmodified transcript as described above. Fully modified E. coli tRNA Lys was purchased from Sigma. To determine tRNA concentrations, an ⑀ 260 of 60.4 ϫ 10 4 M Ϫ1 was used.
Aminoacylation Assays-Aminoacylation assays with purified enzymes were conducted at 30°C in a reaction mixture containing 50 mM HEPES, pH 7.

Isolation of cDNA for a Human
Lysyl-tRNA Synthetase-The cDNA for hLysRS was isolated by an alignment-guided crossspecies PCR approach (27)(28)(29). We first designed cross-species PCR primers for LysRS from an alignment of four sequences from Campylobacter jejuni (42), S. cerevisiae (26), and two isozymes from E. coli (43)(44)(45) that were available when we started the experiment. We chose two conserved regions that overlap with the class-defining motif 1 and motif 2 (10) and synthesized degenerate primers KY-138, -139, and -140. PCR amplifications with KY-138/139 and KY-139/140 on a human cDNA gave 240-base pair DNA fragments whose sequences were identical and showed enough similarity to those of LysRSs from other organisms to conclude that they were derived from the human enzyme. The full-length cDNA was subsequently isolated from a human fetal brain poly(A) ϩ cDNA library, and 5Ј-and 3Ј-RACE PCR was performed. The resultant sequence has 1997 nucleotides in total without the poly(dA) track. This length was in agreement with the apparent size of the transcript (2.1 kbp) based on Northern blot experiments (data not shown).
Coding Sequence for a Human Lysyl-tRNA Synthetase and Multiple Sequence Alignment-The DNA sequence obtained contains one long open reading frame with 597 codons that starts at nucleotide 15 and extends to nucleotide 1808. The first ATG codon and the flanking sequences (AagATGG) are consistent with the Kozak's optimal translation initiation sequence (ACCATGG) (46). The predicted translation product shows high sequence similarity to all of the reported LysRSs. Fig. 1   mitochondria). Also given are structural elements that are based on the crystal structure of the E. coli LysRS (LysU) (22). Of the 597 amino acids in the human LysRS, 58 amino acids (10%) are strictly conserved across the 21 sequences. These conserved residues are shown by black boxes in Fig. 1. There is no large insert in the alignment. However, there is an Nterminal extension of approximately 60 amino acids in the five eukaryotic enzymes. This N-terminal extension of human LysRS is dispensable for aminoacylation activity (see below).
The x-ray crystal analysis (22) showed that the E. coli LysRS comprises two main domains: an N-terminal anticodon recognition domain and a C-terminal catalytic domain. The N-terminal domain forms a ␤-barrel structure with five ␤-strands that are surrounded by four ␣-helices (H1-H4). The C-terminal catalytic domain forms an eight-stranded ␤-sheet (␤ B1 -␤ B8 ) flanked by four helices: H7, -8, -9, and H16. The class defining motifs 1, 2, and 3 are all located within this core structure. Between helix H9 and ␤-strand ␤ B5 , a long polypeptide is inserted into the structure. A similar insertion is also observed in yeast aspartyl-tRNA synthetase (9). One of the ␤-strands of this insertion (␤ B4 ) forms the core of the ␤-sheet in the catalytic domain (22). Thus, the primary structure of E. coli LysRS (LysU) is divided into four sub-regions based on its tertiary structure. These are the N-terminal anticodon binding domain, the first half of the catalytic domain that includes motifs 1 and 2, the insertion domain, and the second half of the catalytic domain that contains motif 3.
Our multiple sequence alignment shows that the catalytic core is highly conserved in this enzyme. About 18% of the residues in the core region are identical across the 21 sequences. By comparison, a similar calculation with 17 class I isoleucyl-tRNA synthetase sequences gave 10% of the residues as being strictly conserved in the catalytic core region. 2 (The isoleucine enzyme is representative of a group of the class I enzymes (28).) The N-terminal anticodon binding domain shows moderate conservation (8%). In contrast, the insertion that splits the catalytic domain of LysRS shows very low sequence conservation and no residue is conserved across all sequences. By comparison, for the CP1 region that splits the catalytic domain of the class I isoleucyl-tRNA synthetase (47), the sequences are relatively well conserved as they are in the catalytic core region (28). This difference in sequence conservation of the two insertions could reflect distinct roles of these subdomains in the respective enzymes (48).
Further analyses of sequences for other tRNA synthetases suggest that the catalytic domain of LysRS is one of the most conserved domains in synthetases. However, the recently published genome sequence of Methanococcus jannaschii, which belongs to the Euryarchaeota line of Archaea, does not contain an open reading frame that shows similarity to any LysRS sequence (49). Thus, in this organism, an entirely different type of LysRS may be present.
The N-terminal Extension Is Dispensable for Enzyme Activity-To demonstrate that the cloned cDNA codes for an active LysRS, the encoded polypeptide was expressed in E. coli and its aminoacylation activity was tested. Plasmids pM116 and pM131 express codons 1-597 and 66 -597, respectively, of human LysRS from a T7 promoter in the pET-3a vector (32). Cell extracts prepared from BL21(DE3)/pM116 and BL21(DE3)/ pM131 showed aminoacylation activity in the presence of crude bovine tRNAs, whereas an extract from BL21(DE3) having the vector pET-3a alone did not show activity under these conditions (Fig. 2). These data show that the encoded polypeptide has LysRS activity and that codons 1-65 (that are unique to the eukaryotic enzymes) are dispensable for the activity. A comparison of the kinetic parameters for aminoacylation by the purified 1-597 and 66 -597 enzymes was carried out, and the results are described below. This result (with the extract from BL21(DE3)) also shows that E. coli LysRS does not charge the mammalian substrate. Experiments using purified E. coli LysRS (lysU gene product) confirmed the latter conclusion (data not shown).
Cross-species Complementation-To determine if a human LysRS can recognize E. coli tRNA Lys as a substrate in vivo, we performed an in vivo cross-species complementation test using an E. coli lysS and lysU double mutant null strain. E. coli and other enteric bacteria generally have two genes for LysRS (50). In E. coli, these genes are lysS and lysU (51). The lysS gene is expressed constitutively, and the expression of lysU is induced under certain conditions such as a temperature shift. In strain PAL⌬S⌬UTR/pMAK705 lysU ϩ (34), both lysS and lysU were disrupted by gene targeting and cell growth was maintained by a wild type copy of lysU on plasmid pMAK705. Because replication of this plasmid is temperature-sensitive (52), the growth of PAL⌬S⌬UTR/pMAK705 lysU ϩ is also temperature-sensitive.
PAL⌬S⌬UTR/pMAK705 lysU ϩ was transformed at 30°C with plasmid pM130 (which expresses hLysRS-⌬N65 from a lac promoter), and transformants were tested for their growth at 42°C. Whereas the tester strain having a vector alone (pTZ19R) could not grow at 42°C, cells carrying pM130 form colonies at 42°C. The cells that grew at 42°C showed sensitivity to chloramphenicol, indicating that they had lost the maintenance plasmid pMAK705 lysU ϩ . Experiments with PAL⌬S⌬UTR(DE3)/pMAK705 and pM131, in which the expression of the human gene (hLysRS-⌬N65) was driven by T7 RNA polymerase, showed the same results (Fig. 3). The elimination of the lysU gene was further confirmed by PCR using lysU-specific primers (data not shown).  (23,24). Mutation of A 73 to G, C, or T led to decrease of aminoacylation activity in vitro (23). Interestingly, the three cytoplasmic human lysine tRNAs have a G instead of an A at position 73 (Fig. 4). Because human tRNA Lys has G 73 and because human LysRS can replace the E. coli enzymes in vivo, we imagined that human LysRS, in contrast to the E. coli enzyme, does not have a high specificity for N 73 .
To investigate this issue further, we purified the human enzyme and performed in vitro aminoacylation assays with N 73 variants of tRNA Lys,3 transcripts (Fig. 4). Human LysRS was purified from the lysS lysU double mutant strain whose growth was maintained by hLysRS-⌬N65. By using such a "humanized" E. coli strain, we avoided possible contamination by the E. coli enzyme during the purification. Table I summarizes the in vitro aminoacylation assay results using variants of tRNA Lys,3 transcripts. In addition to the wild type transcript that has G 73 , the A 73 , C 73 , and U 73 human tRNA Lys,3 variants were all efficiently aminoacylated by hLysRS-⌬N65. This result confirmed that N 73 of tRNA Lys does not serve as an important recognition element for human LysRS. The hLysRS-⌬N65 also charged both natural and in vitro transcribed E. coli tRNA Lys with k cat /K m within a factor of 2 of the wild type human transcript. This cross-species aminoacylation is in contrast to results obtained with E. coli LysRS, which charge unmodified tRNA with a 140-fold lower activity than natural E. coli tRNA Lys (23).
We also purified the full-length form of hLysRS using a His tag attached to its N terminus and tested the contribution of the N-terminal extension to the kinetic parameters for aminoacylation of a wild type and human tRNA Lys,3 transcript. The k cat and K m of the His-tagged full-length form of hLysRS were 1.5 Ϯ 1.2 s Ϫ1 and 15 Ϯ 12 M, respectively. These data show that truncation at the N terminus causes relatively small changes in k cat and K m , so that k cat /K m is reduced by 1.3-fold in the truncated variant (Table I). DISCUSSION We previously reported that cross-acylation was not observed between human glycyl-tRNA synthetase and E. coli tRNAs, or between E. coli glycyl-tRNA synthetase and bovine tRNAs (27). This species-specific aminoacylation in the glycine system was proposed to correlate with a nucleotide difference at N 73 (27). In E. coli, the C 2 :G 71 base pair and U 73 of tRNA Gly are required for recognition by E. coli glycyl-tRNA synthetase (13,14), and all prokaryotic tRNA Gly have "U" at position 73. In contrast, all reported eukaryotic cytoplasmic and archaebacterial tRNA Gly have "A" at this position (25). Subsequent experiments using minihelices demonstrated that the difference at the discriminator base position is indeed relevant for speciesspecific recognition in the glycine system (15). Thus, in the glycine case, the discriminator base is a key element for synthetase recognition both in E. coli and in human, and this difference in a single nucleotide results in a barrier to crossspecies recognition.
The primary structures of human and E. coli glycyl-tRNA synthetases are so diverged that we could not conclude (from their primary structures) that both genes originated from a common ancestor (27,53). If the N 73 discriminator served as an important element for tRNA recognition in an early stage of evolution (5), then the nucleotide changes in the N 73 of glycine tRNAs could have contributed to the existence of two distinct glycine enzymes or to a significant divergence in the primary structures of enzymes derived from a single ancestral enzyme.
Because the structure of aspartyl-tRNA synthetase is closely related to that of lysyl-tRNA synthetase (22), the basis of N 73 recognition by aspartyl-tRNA synthetase might be relevant to how LysRS interacts with the same nucleotide. In the aspartate enzyme, specific residues in the loop of motif 2 make critical contacts with G 73 (8). Given that yeast aspartyl-tRNA synthetase is specific for G 73 , E. coli LysRS for A 73 , and human LysRS has no specificity for the nucleotide at position 73, a comparison of the sequences of the loops of motif 2 in these three enzymes could be instructive (Fig. 5). Consistent with the close relationship between these enzymes, the loop of motif 2 is the same length (Ϯ1) for the three enzymes. (Across all class II enzymes, the loop of motif 2 ranges from about 10 to 20 amino acids in length.) At several positions where contacts occur between G 73 and residues in the loop of motif 2 of aspartyl-tRNA synthetase, the sequences of the two lysine enzymes have changes. Given the success of switching anticodon recognition by a single amino acid swap in a designed peptide that was transplanted into either of two synthetases (54), the comparisons in Fig. 5 suggest the possibility of changes in the sequence of the loop of motif 2 of human LysRS that could switch the nonspecific N 73 behavior of the human enzyme into specific recognition of either an A or a G. These sorts of experiments test the depth of our understanding of those coadaptations between synthetases and tRNAs that occur in evolution to preserve the genetic code. We showed here that human LysRS is relatively insensitive to base modifications for its aminoacylation of E. coli tRNA Lys , in contrast to the E. coli enzyme that showed sensitivity to base modifications (23). Unmodified transcripts of human tRNAs were also efficiently aminoacylated by the purified human enzyme in vitro (Table I). Interestingly, both prokaryotic and eukaryotic tRNA Lys are extensively modified (55) and, in human cells, the modifications in the anticodon loop of tRNA Lys, 3 have been proposed to play a role in its selection as a tRNA primer in human immunodeficiency virus type-1 (56). It would be interesting to determine the effect of base modifications on aminoacylation by human LysRS using purified native mammalian tRNA Lys,3 .  Table I were made based on these structures. E. coli has one lysine tRNA, whereas human cells have three tRNA Lys isoacceptors. The N 73 position of all three human lysine tRNAs is G (indicated by a gray box). Human LysRS has an extra N-terminal extension of approximately 60 amino acids that is absent from prokaryotic enzymes. Similar extensions were observed in S. cerevisiae (cytoplasm) (26), C. longicaudatus (Chinese hamster) (57), C. elegans (GenBank U41105), and L. esculentum (tomato) (Gen-Bank X94451) LysRSs. Similarities are observed among these five sequences (Fig. 1), suggesting that the sequences have been diverged from a common ancestor. The extensions are relatively rich in charged amino acids, and the possible role of the extension could be in the formation of a multi-enzyme complex that is unique to higher eukaryotes (58). Removal of the extensions from the yeast or hamster LysRSs did not abolish their aminoacylation activity (57,59). We showed that the extension of the human enzyme is also dispensable for its in vitro aminoacylation activity and for in vivo cross-species complementation. Thus, the selective pressure to retain this extension throughout eukaryote evolution must arise from a function other than aminoacylation.
The alignment of the N-terminal extensions (Fig. 1) suggests that the human lysine enzyme studied here is of cytoplasmic origin. For example, the sequence of the extension of the human enzyme has similarities to that of yeast cytoplasmic LysRS but not to that of the yeast mitochondrial enzyme.
This report is the first of a human tRNA synthetase rescuing a bacterial strain that is defective in a specific tRNA synthetase. Recent attempts to rescue the null strain of the class I E. coli isoleucyl-tRNA synthetase with its human counterpart, or to rescue the E. coli null strain of the class II glycyl-or alanyl-tRNA synthetases with their respective human homologues, have failed. 2 The failure of the human glycine enzyme to rescue the bacterial null strain was unsurprising, because of the strict species-specific aminoacylation observed in vitro (15). As shown here, the successful cross-species complementation in the case of LysRS results, at least in part, from its unusual tolerance for the difference in N 73 between E. coli and human lysine tRNAs.