Aminoacyl-tRNA recognition by the leucyl/phenylalanyl-tRNA-protein transferase.

We employ mutant and mischarged aminoacyl-tRNAs to characterize aminoacyl-tRNA recognition by the leucyl/phenylalanyl-tRNA-protein transferase (L/Ftransferase). Wild type Met-tRNAMetm (CAU anticodon) and mischarged Met-tRNAVal-1 (CAU anticodon) are substrates for the L/F-transferase during the NH2-terminal aminoacylation of alpha-casein, whereas Val-tRNAVal-1 (UAC), Val-tRNAMetm (UAC), and Arg-tRNAMetm (CCG, A20) are not. Mutations in the anticodon and extra arm of tRNALeu-1 do not measurably effect its ability to serve as a substrate for the L/F-transferase, and the dissociation constants of the complexes between L/F-transferase and either wild type Leu-tRNALeu-4 (UAA) or mutant Leu-tRNALeu-4 (CUA) are each 0.4 +/- 0.2 microM. The dissociation constants for the complexes between the L/F-transferase and uncharged tRNA, leucine methyl ester, and puromycin are all 10-1,000-fold greater than that of the Leu-tRNA.L/F-transferase complex. Dissociation of the Leu-tRNA.L/F-transferase complex is slow, relative to the rate calculated assuming that association is diffusion controlled. Finally, deoxyoligonucleotide.aminoacyl-tRNA hybrids (dO.AA-tRNAs) are employed to characterize the determinants of the Leu-tRNALeu-4 acceptor stem recognized by the L/F-transferase. A dO.AA-tRNA completely lacking acceptor stem base pairs remains a substrate for the L/F-transferase, whereas a dO.AA-tRNA containing a 2-base pair single-stranded region, at its 3' terminus, does not.

The L/F-transferase 1 catalyzes the transfer of Leu, Phe, and Met from aminoacyl-tRNAs to the amino termini of acceptor proteins (1,2). All known acceptor proteins or peptides contain the basic amino-terminal residues Arg or Lys and are predominantly unstructured at their amino termini (3). The use of aminoacyl-tRNA as the aminoacyl donor is analogous to the peptidyl transferase reaction catalyzed on the large subunit of the ribosome and makes the L/F-transferase unique among enzymes that catalyze non-ribosomal peptide bond synthesis in Escherichia coli cells. Previously characterized enzymes activate the amino acid as an enzyme-bound thioester (4) or an aminoacyl-phosphate (5) prior to peptide bond formation. Interestingly, many species of bacteria use a reaction similar to that catalyzed by the L/F-transferase in peptidoglycan biosyn-thesis. For example, the cell wall of Staphylococcus aureus contains a (Gly) 5 cross-link between adjacent (L)-Ala-(D)-Glu-(L)-Lys-(D)-Ala peptides. This cross-link is synthesized using Gly-tRNA as the amino acid donor (6). However, the E. coli cell wall lacks this cross-link and no role of the L/F-transferase in cell wall biosynthesis has been demonstrated to date.
Although the precise cellular role of the L/F-transferase remains an open question, recent data demonstrate a role in protein degradation (7,8). For example, a ␤-galactosidase variant, engineered to expose an amino-terminal Arg (Arg-␤-galactosidase) in E. coli cells, is modified to form Leu-Arg-␤galactosidase by the L/F-transferase and subsequently degraded by the Clp A/P (Ti) protease (9,10). This result implicates the L/F-transferase as a component of the N-end rule pathway of protein degradation (11). Amino-terminal aminoacylation via the L/F-transferase is a simple yet powerful signal for protein turnover, as E. coli aat::minitet mutant cells that lack L/F-transferase activity degrade Arg-␤-galactosidase 100 to 1,000 times more slowly than wild type cells. Finally, the aat genes of both E. coli and Provadencia stuartii, that encode L/F-transferase homologs, lie at the end of three gene operons whose other members are homologs of the P-glycoproteins responsible for multi-drug resistance in mammalian cells (8,12). These multi-drug resistance homologs are essential for exit from the stationary phase (13) and for the proper expression of cytochrome d (14) in E. coli cells. This chromosomal position remains an additional clue to the cellular function of the L/F-transferase.
The L/F-transferase is a monomeric enzyme of 234 residues that lacks a detectable RNA component or other organic cofactors (15,16). The aat gene encoding the L/F-transferase has been cloned and used to overexpress, purify, and characterize an affinity tagged L/F-transferase (16). The specific activity of the recombinant L/F-transferase is comparable to that of the previously purified wild type enzyme (15). Circular dichroism analysis demonstrates that the L/F-transferase is ϳ50% ␣-helical and lacks detectable ␤-sheet structure. The absence of ␤-sheet structure strongly suggests that the L/F-transferase recognizes tRNA using a domain other than the ribonucleoprotein RNA-binding domain, found throughout the prokaryotic and eukaryotic kingdoms (17). Using the recombinant enzyme, we demonstrated previously that the modified nucleotides found in natural tRNAs are not essential for recognition by the L/F-transferase (16). This result opens the way for the use of T7 RNA polymerase-derived mutant tRNAs to dissect the essential determinants of aminoacyl-tRNAs recognized by the L/F-transferase.

MATERIALS AND METHODS
tRNA and Aminoacyl-tRNA Synthesis and Purification-Unmodified tRNAs were transcribed using T7 RNA polymerase using standard procedures with minor modifications (18). Briefly, plasmid-borne tRNA genes were digested with BstNI restriction endonuclease (New England Biolabs) and the enzyme removed by phenol/chloroform extraction. Transcription reactions include: 40 mM Tris (pH 8.0), 20 mM MgCl 2 , 25 mM NaCl, 2.0 mM spermidine, 5 mM dithiothreitol, 16 mM each nucleotide triphosphate, 40 mM GMP, 80 units of RNasin (Promega), and purified T7 RNA polymerase. Transcription reactions were incubated 8 -10 h at 37°C. Heterogeneous tRNA transcripts were separated from unincorporated nucleotides and pyrophosphate using a NAP-5 column (Pharmacia) followed by precipitation with 2 volumes of ethanol in the presence of 0.2 M NaCl. Full-length tRNA transcripts were isolated from incorrectly sized transcripts by electrophoresis on 10% acrylamide gels containing 50% urea. Full-length tRNA transcripts were identified by their UV shadow, eluted from the gel slice by soaking overnight at 37°C, and concentrated by ethanol precipitation. Leu-tRNAs were generated using partially purified E. coli Leu-tRNA synthetase (LeuRS) as described previously (16). Charging reactions were carried out in 30 mM Hepes (pH 7.4), 15 mM MgCl 2 , 25 mM KCl, 2.0 mM dithiothreitol, 2.0 mM ATP, 20 M Leu. Aminoacyl-tRNAs were separated from the LeuRS by phenol/chloroform extraction using phenol equilibrated at pH 5.0. Aminoacyl-tRNAs were concentrated by precipitation with 2.5 volumes of ethanol in 0.2 M NaCl.
L/F-transferase Activity Assays-Qualitative reactions to characterize the ability of mischarged and mutant aminoacyl-tRNAs to serve as substrates for the L/F-transferase employ a mixture of aminoacyl-tRNA synthetases (Sigma) and purified L/F-transferase enzyme (16). All reactions contain 50 mM Tris (pH 8.0), 0.1 M KCl, 10 mM MgCl 2 , 10 mM 2-mercaptoethanol, 12 M ␣-casein, 10 mM ATP, ϳ500 units of E. coli aminoacyl-tRNA synthetases (Sigma), and purified L/F-transferase. Reactions were incubated for 5 min at 37°C. Reactions are assayed for 14 C-amino acid incorporated into protein (hot and cold trichloroacetic acid stable) and 14 C-amino acid incorporated aminoacyl-tRNA (cold trichloroacetic acid stable) as described previously (16).
Substrate Binding and Competitive Dissociation -Dissociation constants were determined using the ribonuclease A protection assay that is commonly employed to characterize aminoacyl-tRNA⅐elongation factor-Tu (AA-tRNA⅐EF-Tu) complex formation (19). Binding reactions are carried out in 50 mM Hepes buffer (pH 7.2) and contain 1-2 M [ 14 C]Leu-tRNA Leu-1 (CAG) or the amber suppressor [ 14 C]Leu-tRNA Leu-1 (UCA) and variable amounts of purified L/F-transferase. Binding reactions are incubated for 15 min at 25 or 0°C (on ice). The amount of aminoacyl-tRNA in complex with the L/F-transferase is determined by the addition of ribonuclease A to a final concentration of 0.25 mg/ml, followed by incubation on ice or at 25°C for 15 s. Ribonuclease A digestion is stopped by the addition of equal volumes of 20% trichloroacetic acid, and 10-l aliquots are spotted on Whatmann 3MM filters. Filters are subsequently washed and the amount of radioactivity that is insoluble in trichloroacetic acid determined as described previously (16). In these experiments each value required to determine K d is measured directly. Deoxyoligonucleotide⅐Aminoacyl-tRNA Hybrid Formation-Hybridization of deoxynucleotides with aminoacyl-tRNAs was performed by mixing Leu-tRNA Leu-1 (4 -6 M) with oligonucleotide (40 -50 M) and incubation for 30 s at 75-80°C, followed by slow cooling to 20°C. These hybridizations are carried out in 10 mM sodium acetate (pH 5.0) to minimize spontaneous hydrolysis of aminoacyl-tRNAs. Hybridization reactions are subsequently precipitated with ethanol and stored at Ϫ75°C. Immediately prior to use, hybrids are resuspended in H 2 0. One-half of the resuspended hybrid is used as substrate in the L/Ftransferase reaction, while the remainder is assayed for the fraction of aminoacyl-tRNA in the hybrid form by nondenaturing gel electrophoresis in TAE buffer (40 mM Tris acetate, 10 mM EDTA). Deoxyoligonucleotides used in hybrid formation were as follows: LP, 5Ј-ACCCGGAGC-GGGACTTCAAC-3Ј; LCA, 5Ј-TCCCGGAGCGGGACTTGAAC-3Ј; L2, 5Ј-GTACCCGGAGCGGGACTTGAAC-3Ј; L30, 5Ј-CTACCGATTCCA-CC ATCCGG-3Ј.

RESULTS
Mischarged Aminoacyl-tRNAs in the L/F-transferase Reaction-The abilities of selected aminoacyl-tRNAs and aminoacyl-tRNA mutants to function as substrates of the L/F-transferase during the NH 2 -terminal aminoacylation of ␣-casein are shown in Table I (upper section). Wild type tRNA Metm (CAU) and the anticodon mutant tRNA Val-1 (CAU) are each aminoacylated with methionine to form Met-tRNA Metm (CAU) and Met-tRNA Val-1 (CAU), respectively. These two Met-tRNAs serve as substrates for the L/F-transferase in the NH 2 -terminal aminoacylation of ␣-casein (Met-tRNA Val-1 (CAU) contains the nucleotide sequence of tRNA Val-1 but with a CAU anticodon; the anticodon of wild type methionine tRNAs). Conversely, Val is not transferred from either its wild type cognate Val-tRNA-Val-1 (UAC) or from the anticodon mutant Val-tRNA Metm (UAC), and Arg is not transferred from Arg-tRNA Metm (CCG, A20). These results suggest that either the aminoacyl moiety of an aminoacyl-tRNA is important for recognition by the L/Ftransferase, or that the L/F-transferase recognizes directly the anticodons of its aminoacyl-tRNA substrates (this second model is not supported by additional experiments, described below). Each of the amino acids transferred by the L/F-transferase are hydrophobic and contain an unbranched ␤-carbon. The lack of transfer of either Val (hydrophobic) or Arg (unbranched ␤-carbon) from tRNAs that are nearly identical with tRNA Metm , suggests that both properties of the amino acid moiety are essential for recognition by the L/F-transferase.
The experiments described above rely on the recognition specificities of specific aminoacyl-tRNA synthetases. For example, the Met-and Val-tRNA synthetases from E. coli identify their cognate tRNAs predominantly via direct recognition of the nucleotides within the tRNA anticodon (20). Other nucleotides in the tRNA structure may be mutated with minor effect. Correspondingly a tRNA Metm with the Val anticodon (UAC) is efficiently charged with Val but not Met, while tRNA Val-1 with the Met anticodon (CAU) is efficiently charged with Met but not Val. In addition to the Arg anticodon (CCG), a second mutation, U20 3 A20, is required to convert tRNA Metm into an efficient substrate for the Arg-tRNA synthetase (21).
Effects of Mutations in the Anticodon and Extra Arm of tRNA Leu-1 on the L/F-transferase Reaction-The specificities of the L/F-transferase for specific aminoacyl-tRNA mutants (Table I) left open the possibility that the L/F-transferase might recognize the anticodon of its cognate aminoacyl-tRNAs. This mode of recognition has been demonstrated for several aminoacyl synthetases in addition to those mentioned above (20). Also, the base of the central nucleotide of the anticodon of all L/F-transferase-substrate aminoacyl-tRNAs is an adenine. Central adenines in the tRNA anticodons are required to recognize Leu codons (CUN and UUR), Phe codons (UUY), and the Met codon (AUG). This common adenine provides for a plausible mechanism for the specific recognition of these three classes of tRNA by a single enzyme. However, this mode of recognition is strongly argued against by the data in Table I (lower section). A tRNA Leu-1 anticodon mutant (UCA anticodon), a second mutant containing an G37 3 U37 substitution in the anticodon loop (immediately 3Ј to the anticodon), and a third mutant tRNA Leu-1 lacking four nucleotides of the extra arm region (⌬2 lacks the C47⅐G47g and C47a⅐G47f base pairs (22)), are all active as L/F-transferase substrates. These data demonstrate that the anticodon is not an essential determinant for the recognition of aminoacyl-tRNAs by the L/F-transferase. Nonetheless, experiments described below reveal that the nucleic acid makes a contribution to the overall affinity of the AA-tRNA⅐L/F complex. It should be noted that the above series of experiments takes advantage of the fact that the majority of the nucleotides recognized by the Leu-tRNA synthetase are external to the anticodon of tRNA Leu , allowing each mutant tRNA Leu-1 to be charged efficiently (22).
The L/F-transferase⅐Leu-tRNA Dissociation Constant Is in the Micromolar Range-In the previous series of experiments, anticodon mutants in tRNA Leu-1 did not affect the ability of this tRNA to act as an L/F-transferase substrate (Table I). To separate possible effects of anticodon mutations on AA-tRNA substrate binding versus effects on catalysis, and to confirm the generality of our conclusion that the L/F-transferase does not bind the tRNA anticodon, the dissociation constants for an independent pair of Leu-tRNA Leu-4 substrates were determined. The dissociation constants for the complexes between the L/F-transferase and wild type Leu-tRNA Leu-4 (UAA) and mutant Leu-tRNA Leu-4 (CUA) were determined using a ribonuclease A protection assay. This assay has been employed previously to characterize aminoacyl-tRNA⅐EF-Tu complex formation and relies on the observation that nearly complete digestion of an aminoacyl-tRNA is required to render the amino acid moiety soluble in 10% trichloroacetic acid (19). Therefore complex formation is monitored directly by determining the fraction of [ 14 C]Leu that is not rendered acid soluble by brief exposure to high concentrations of ribonuclease A. The data in Fig. 1 is presented as the fraction of the [ 14 C]Leu-tRNA precipitated by 10% trichloroacetic acid relative to the fraction precipitated under saturating levels of L/F-transferase (plateau region). Shown are the data for the binding of the L/Ftransferase to wild type Leu-tRNA Leu-4 (UAA) and the anticodon mutant Leu-tRNA Leu-4 (CUA) at 0°C (on ice). Also shown are the data for the formation of the L/F-trans ferase⅐Leu-tRNA Leu-4 (UAA) complex at 25°C. In agreement with the activity data described above, mutations in the anticodon region of a Leu-tRNA do not affect binding to the L/Ftransferase. The calculated value of the dissociation constant in each case is 0.4 Ϯ 0.2 M. This value is similar to the K m measured previously for the L/F-transferase using wild type tRNA Leu-4 as a substrate (0.31 M, (23)). Conversely, this K d value is significantly lower that the K d value of AA-tRNA⅐EF-Tu complexes (2-10 nM) (19). This lower affinity of the L/F-transferase for aminoacyl-tRNAs is consistent with our earlier findings that the L/F-transferase can be overexpressed to high cellular levels without causing significant lethality (16).
Competition for L/F-transferase Binding by Leucine Methyl Ester, tRNA, and Puromycin-To dissect further the regions of Leu-tRNA recognized by the L/F-transferase we carried out binding reactions in the presence of variable amounts of competitor molecules. The simplest competitor tested was the methyl ester of leucine (Leu-O-Me). Competition experiments employing the Leu-O-Me molecule allow estimation of the importance of the leucyl side chain and free amino group of Leu-tRNA for recognition by L/F-transferase. NH 2 terminally acetylated Phe-tRNA Phe is not a substrate for the L/F-transferase (24). The use of the ester removes the possible confusing contribution to the binding reaction of the charged acid group of free Leu, as it seems likely that the L/F-transferase uses discrimination against this charge to avoid binding free leucine in the cell. Leu-O-Me demonstrates a measurable ability to compete with Leu-tRNA for L/F-transferase binding (Fig. 2). Based on the competition data, the K d of the Leu-O-Me⅐L/F complex is Ͼ1.0 mM, a value significantly higher than that of the L-tRNA⅐L/F complex (see Fig. 1). This suggests that the nucleic acid moieties of aminoacyl-tRNAs contribute to their overall affinity for the L/F-transferase. Unfortunately, a quantitative determination of the ability of uncharged tRNA to bind the L/F-transferase and inhibit the formation of the AA-tRNA⅐L/F complex is not possible using the ribonuclease A protection assay. The concentrations of tRNA Leu-4 required to compete significantly with Leu-tRNA Leu-4 for limiting L/F-transferase overwhelm the ability of ribonuclease A to degrade completely the unbound Leu-tRNAs. However, an 8-fold excess of uncharged tRNA Leu-4 over Leu-tRNA Leu-4 shows no detectable effect on L/F-transferase⅐Leu-tRNA Leu-4 complex formation (data not shown). This result demonstrates, at least qualita-TABLE I Certain unmodified and mischarged AA-tRNAs are substrates for the L/F-transferase Top, the ability of a Met-tRNA to serve as a substrate for the L/F-transferase strongly depends on the methionyl moiety. Qualitative comparisons of the effects of mutations in tRNA Metm and tRNA Val-1 on their ability to serve as substrates for the L/F-transferase were made using coupled reactions containing purified L/F-transferase and a mixture of total E. coli aminoacyl-tRNA synthetases. Cold trichloroacetic acid is the amount of [ 14 C]valine or [ 35 S]methionine (in counts) that is precipitated by 10% trichloroacetic acid after the reaction and is indicative of radiolabeled amino acid in the form of AA-tRNA or incorporated into protein. Hot trichloroacetic acid is the amount of radiolabeled amino acid that remains precipitable in 10% trichloroacetic acid after boiling in 5% trichloroacetic acid and is indicative of radiolabeled amino acid incorporated into protein. R is the ratio: Hot trichloroacetic acid/Cold trichloroacetic acid (in percent). tRNA Metm (CAU) ϭ 5Ј-GGCUACGUAGCUCAGUUGGuUAGAGCA-CAUCACUCAUAAUGAUGGGGUCACAGGUUCGAAUCCCGUCGUAGCCACCA-3Ј is the E. coli wild type elongator tRNA Met (anticodon underlined), tRNA Metm (UAC) ϭ tRNA Metm (CAU) with a CAU3 UAC anticodon mutation, tRNA Metm (CCG, A20) ϭ tRNA Metm (CAU) with a CAU3 CCG anticodon mutation and a T3 A mutation at nucleotide 20 (lower case). tRNA Val-1 (UAC) ϭ 5Ј-GGGUGAUUAGCUCAGCUGGGAGAGCAC-CUCCCUUACAAGGAGGGGGUCGGCGGUUCGAUCCCGUCAUCACCCACCA-3Ј is one of the three wild type tRNA Val s of E. coli, tRNA Val-1 (CAU) ϭ tRNA Val-1 (UAC) with a UAC3 CAU anticodon mutation.
Lower, mutations in neither the anticodon nor the extra arm abolish the ability of Leu-tRNA Leu-1 mutants to function as L/F-transferase substrates. Qualitative comparisons of the effects of mutations in tRNA Leu-1 on its activity as a substrate for the L/F-transferase were made using coupled reactions containing purified L/F-transferase and purified Leu-tRNA synthetase. tRNA Leu-1 ϭ 5Ј-GCGAAGGUGGCGGAAUUGGUA-GACGCGCUAGCUUCAGGUGUUAGUGUccUUACggACGUGGGGGUUCAAGUCCCCCCCCUCGC-3Ј, tRNA Leu-1 ⌬2 ϭ tRNA Leu-1 but lacking 4 nucleotides from the extra arm region (shown in lower case), tRNA Leu-1 U37 ϭ tRNA Leu-1 with a G3 U mutation at nucleotide 37 (immediately 3Ј to the anticodon), tRNA Leu-1 (UGA) ϭ tRNA Leu-1 with a CAG3 UGA mutation in the anticodon.

AA-tRNA⅐L/F-Transferase Interactions
tively, that the K d of the tRNA⅐L/F complex is significantly greater than the K d of the L-tRNA⅐L/F complex. Puromycin (3Ј-[amino-p-methoxy-hydrocinnamamido]-3Ј-deoxy-N,N-dimethyladenosine) is a widely used inhibitor of ribosomal protein synthesis that acts via its ability to mimic the 3Ј terminus of an aminoacyl-tRNA and cause polypeptide chain termination. Puromycin also inhibits aminoacylation of ␣-casein by the L/F-transferase (25). As seen in Fig. 2, puromycin, which mimics both the aminoacyl and tRNA moieties of an aminoacyl-tRNA, is a significantly better inhibitor of L-tRNA⅐L/F complex formation than is Leu-O-Me. This result supports the claim that the tRNA moiety of aminoacyl-tRNAs make a contribution to the overall affinity of the L-tRNA⅐L/F complex and further suggests that a significant fraction of this contribution comes from the 3Ј-terminal adenine of aminoacyl-tRNAs. The observed ability of puromycin to inhibit Leu-tRNA Leu-4 ⅐L/F-transferase complex formation is similar to published inhibition data. For example, 0.4 mM puromycin had a 75% inhibitory effect on an assay used to detect ribosomeindependent incorporation of methionine into ␣-casein (25). The responsible enzyme was identified later as the L/F-transferase (26).
Dissociation of the L/F-transferase⅐Leu-tRNA  Complex-The ribonuclease A assay described above also provides a qualitative method to estimate the dissociation rates of aminoacyl-tRNA⅐protein complexes. For example, during prolonged incubation of the L-tRNA⅐L/F complex with excess ribonuclease A, aminoacyl-tRNAs that dissociate from the enzyme become accessible to the nuclease and may be degraded prior to rebinding. Although, the possibility of rebinding makes this assay qualitative, it has been exploited to estimate the k off of the aminoacyl-tRNA⅐EF-Tu complex (27). Data reflecting the rate of release of [ 14 C]Leu, as a trichloroacetic acid-soluble form, from the Leu-tRNA Leu-1 ⅐L/F complex is graphed in Fig. 3  (left panel). The amount of Leu-tRNA Leu-1 protected is stable with time for each L/F-transferase/Leu-tRNA Leu-1 ratio tested.
The point reflecting ϳ60% protection was followed for a longer time and a significant fraction of the complex is stable for Ͼ10 min (Fig. 3, right panel). The rate of disappearance of the complex is ϳ2 orders of magnitude slower than the rate of the disappearance of the aminoacyl-tRNA⅐EF-Tu complex (k off ϳ0.2 s Ϫ1 ) (27). This slower rate of dissociation of the L-tRNA⅐L/F complex is striking in light of the overall tighter binding of the aminoacyl-tRNA⅐EF-Tu complexes (K d values of EF-Tu for various aminoacyl-tRNAs are in the 2-10 nM range (19)). Assigning this extremely slow dissociation rate as equivalent to the k Ϫ1 rate of the L/F-transferase-aminoacyl-tRNA-␣-casein ternary complex is not consistent with overall rate of catalysis of the L/F-transferase enzyme (discussed below).
Activity of Deoxyoligonucleotide⅐Aminoacyl-tRNA Hybrids in the L/F-transferase Reaction-The ability of puromycin to act as an efficient inhibitor of L-tRNA⅐L/F complex formation suggests that the tRNA acceptor stem region is recognized by the L/F-transferase. In order to determine the importance of the different structural elements contained within the tRNA acceptor stem, the structure of this region was modulated by constructing deoxyoligonucleotide⅐aminoacyl-tRNA hybrids (dOligo⅐AA-tRNAs). This approach was necessitated by the inability of the Leu-, Phe-, or Met-tRNA synthetases to charge efficiently a tRNA minihelix or microhelix (28). Because the hybrids are not expected to be aminoacylated efficiently by the LeuRS enzyme, the oligonucleotides were hybridized with previously aminoacylated tRNA Leu-4 . The dO⅐AA-tRNAs are schematized in Fig. 4A and the completeness of hybrid formation, for each oligonucleotide, is revealed by the non-denaturing gel in Fig. 4b. Hybridization of the LP deoxyoligonucleotide recreates an anticodon helix that is structurally analogous to that seen in native aminoacyl-tRNAs, while use of the LCA oligonucleotide creates an acceptor helix with a terminal non-Watson-Crick T:U base pair. Oligonucleotide L2 generates an acceptor helix that exposes only a 2-base pair single stranded 3Ј-overhang, while oligonucleotide L30 creates a completely single stranded 3Ј terminus by hybridizing with the 5Ј region of the tRNA. The data in Table II demonstrates that the dOligo⅐L-tRNAs constructed from deoxyoligonucleotides LP and LCA remain substrates for the L/F-transferase. Therefore, no 2Ј-OH groups from the 5Ј-strand are required for nucleic acid recognition by the L/F-transferase. In addition, the dO⅐L-tRNA constructed from L30 serves as an L/F-transferase substrate while the dO⅐L-tRNA constructed from L2 does not. Overall, these experiments reveal that no double stranded structure is required for recognition by the L/F-transferase, but rather that Ͼ2 single stranded nucleotides must be exposed at the 3Ј terminus of the aminoacyl-tRNA. DISCUSSION The L/F-transferase utilizes a highly degenerate family of macromolecules as substrates during its catalysis of peptide bond formation. For example, acceptor proteins and peptides bearing either of the basic residues Arg or Lys at their NH 2 terminus are efficient amino acid acceptors (3). In addition, Leu, Phe, Met, and the amino acid analog p-fluoro-Phe are all transferred from their cognate tRNAs, and certain mutant tRNAs, to acceptor proteins (24). The molecular bases for the degenerate recognition of aminoacyl-tRNAs by the L/F-transferase is becoming clear. Both the aminoacyl and nucleic acid moieties of an aminoacyl-tRNA are utilized by the enzyme during this discrimination. Previous authors have demonstrated that Phe-tRNA Val-1 is a substrate but Val-tRNA Val-1 and N-acetyl-Phe-tRNA Phe are not (24). We have generalized and extended these findings by demonstrating that Met-tRNA recognition is also strongly dependent on the aminoacyl group. The substrate specificity of the L/F-transferase for Leu-, Phe-, and Met-tRNAs, but not Ile-or Val-tRNAs suggests that an unbranched ␤-carbon is recognized by the enzyme. However, the inability of the enzyme to accept Arg-tRNA Metm (CCG, A20) as a substrate suggest that overall side chain hydrophobicity is also recognized.
Neither mutations in the anticodon region nor in the extra arm of tRNA Leu1 abolish recognition of this nucleic acid by the L/F-transferase. Direct recognition of the anticodon is observed for several aminoacyl-tRNA synthetases (20), however, the size of the L/F-transferase argued against its use of this mechanism. A globular protein of 234 amino acids would have an approximate diameter of 40 Å, while the distance from the anticodon loop to the aminoacyl group of an aminoacyl-tRNA is ϳ75 Å (29). Overall, the small size of the L/F-transferase and the lower K d of the puromycin⅐L/F complex, relative to Leu-O-Me⅐L/F complex, suggests that the nucleic acid residues recognized by the enzyme are very close to the 3Ј terminus of the tRNA. We demonstrated that no intramolecular base pairs involving the nucleotides from the 5Ј terminus of a tRNA are essential for recognition by the L/F-transferase. This is informative in light of the previous demonstration that the 3Ј-pen-tanucleotide of Phe-tRNA Phe is not a substrate for the L/Ftransferase (24). We conclude that nucleotides from the 3Ј terminus of an aminoacyl-tRNA 5Ј to the single-stranded 3Јterminal rA-rC-rC-rA are recognized by the L/F-transferase (not necessarily in base paired form).
The existing data reflecting the ability of Phe-, Leu-, and Met-tRNAs to serve as L/F-transferase substrates are schematized in Fig. 5. The acceptor stem and anticodon of each tRNA is shown along with published data from L/F-transferase assays. The reported K m values for the different Leu isoacceptors vary over a nearly 10-fold range (Fig. 5) (23). A strong dependence on the nucleic acid moiety is also seen when the published data for the initiator and elongator Met-tRNA Met s are compared. These changes to the nucleic acid moiety result in a   FIG. 3. Dissociation of the AA-tRNA⅐L/F complex is slow for each substrate/enzyme ratio tested. Qualitative estimation of the half-life of the L-tRNA⅐L/F complex is made using prolonged digestion of the complex with ribonuclease A. At each Leu-tRNA Leu-4 (UAA)/L/Ftransferase ratio, used to construct the dissociation curve of Fig. 2, the complex is stable for Ͼ45 s (left panel). For the binding reaction resulting in ϳ60% protection of tRNA Leu-4 (UUA) a significant fraction of the complex is stable for Ͼ10 min (right panel).

FIG. 4.
A, schematic of the dO⅐AA-tRNA hybrids. dO⅐L-tRNAs are constructed by annealing each deoxyoligonucleotide with previously aminoacylated tRNA Leu-4 (UUA). The nucleotide sequence of tRNA Leu-4 (UUA) is given in the legend to Fig. 1, the sequences of the deoxyoligonucleotides are: LP, 5Ј-ACCCGGAGCGGGACTTCAAC-3Ј; LCA, 5Ј-TC-CCGGAGCGGGACTTGAAC-3Ј (the nucleotide involved in the T⅐U base pair is underlined); L2, 5Ј-GTACCCGGAGCGGGACTTGAAC-3Ј; L30, 5Ј-CTACCGATTCCACCATCCGG-3Ј. B, a large fraction of each Leu-tRNA is converted to dO⅐Leu-tRNA hybrid form. Nondenaturing gel electrophoresis was used to demonstrate that the majority of the tRNA Leu-4 (UUA) is converted to the dO⅐L-tRNA form in the annealing reactions. dO⅐Leu-tRNAs are expected to be less compact than fully folded tRNAs and therefore to have lower mobility during electrophoresis. 30-fold difference in the kinetic assay employed by the Soffer laboratory (26). No strictly conserved nucleotides are observed in the acceptor stem regions of the L/F-transferase substrate tRNAs depicted in Fig. 5. This is also true when the complete sequences of these tRNAs are compared; there are no conserved nucleotides that are shared only by L/F-transferase substrate tRNAs. However, there is a correlation between the occurrence of A:T base pairs near the base of the acceptor stem and overall tRNA activity in L/F-transferase assays. All tRNAs with purely G:C acceptor stems, which are expected to be most stable to denaturation, are poor L/F-transferase substrates (Fig. 5). This correlation, between the acceptor stem stability and activity, indicates a tRNA deformation or unwinding during the reaction with the L/F-transferase. The inability of initiator tRNA Metf to act as a substrate might implicate any of a number of tRNA regions as recognized by the L/F-transferase and we originally suspected that the acceptor helix terminal non-canonical C:A base pair was involved (30). However, the activity of the LCA and L30-derived dO⅐L-tRNAs in the L/F-transferase reaction demonstrates that a canonical Watson-Crick base pair is not required at this, or any, position. In fact, a direct comparison of the K m values of Phe-tRNA Phe and Leu-tRNA  suggest that the non-canonical terminus is favored (if differential effects of the aminoacyl groups may be ignored). Overall, the ability of enzymes to deform structured RNA substrates during their reaction cycle is important and experiments are planned to construct chimeric tRNAs to test this model.
The extreme stability of the L-tRNA⅐L/F complex is intriguing as a similar kinetic stability is also displayed by the AA-tRNA⅐EF-Tu complex at low temperatures (27). However, for the binding reaction described in Equation I, equating the dissociation rate of the L-tRNA⅐L/F complex k off , with the k Ϫ1 value for analogous step in peptide bond synthesis (Equation II), results in calculated reaction rates that are significantly lower than those observed experimentally (16).
AA-tRNA ϩ ␣-casein ϩ L/F l | : During prolonged incubation in the presence of ribonuclease A, the observed t 1 ⁄2 for dissociation of the L-tRNA⅐L/F complex is ϳ10 min (see Fig. 3, right panel). If this process is equated to the dissolution of the ternary complex, the corresponding k Ϫ1 value is 1.1 ϫ 10 Ϫ3 s Ϫ1 . Using the measured value: K d ϭ 0.4 ϫ 10 Ϫ6 M the calculated k 1 value for this process is then 2,750 s Ϫ1 M Ϫ1 , a value much lower than that expected for a diffusion controlled reaction (31). For the standard L/F-transferase reaction, as described by Soffer and colleagues (2), a 75-l reaction containing 1.8 M L-tRNA and 4.0 M ␣-casein are used. This concentration of ␣-casein is ϳfour times greater than the K m value for this substrate (1.0 M (3)), allowing the reaction to be approximated as pseudo-first order. Under these conditions in a reaction containing 1.0 g of L/F-transferase, formation of the ternary complex would occur at a calculated initial rate of: d[ternary complex]/dt ϭ k on [L/F][AA-tRNA] ϭ 0.23 ϫ 10 Ϫ8 M s Ϫ1 . However, in actual L/F-transferase assays, performed under these conditions, the rate of product formation is 5.3 ϫ 10 Ϫ8 M s Ϫ1 (16). Therefore, the observed rate of product formation is significantly greater than the calculated maximum. This paradox is further exaggerated when the near equality of our measured K d (0.4 M, Fig. 1) and the previously reported value for K m (0.32 M for tRNA Leu-4 (23)) is considered. The near equality of K d and K m for Equation 2 results from the condition k Ϫ1 Ͼ Ͼ k 2 . Under these conditions, if the true k Ϫ1 ϭ k off ϭ 1.1 ϫ 10 Ϫ3 s Ϫ1 , product formation would be immeasurably slow; a result not observed.
The physical basis for the unusual binding kinetics of the AA-tRNA⅐L/F complex remain to be determined. Our working model is that the AA-tRNA⅐L/F complex is able to isomerize into non-productive conformations that are unable to proceed to product but remain inaccessible to ribonuclease. It should be noted that in this scenario, the K d we report here is the value that is relevant to discussions of the fraction of cellular aminoacyl-tRNAs complexed with L/F-transferase. Two lines of reasoning have led us to adopt our model. First, although one form of the AA-tRNA⅐EF-Tu complex is very long-lived (the low

dO⅐L-tRNAs hybrids as substrates for the L/F-transferase
The ability of the different dO⅐L-tRNAs to function as L/F-transferase substrates is monitored as described for other substrates (see methods). R 1 is the fraction of [ 14 C]leucine, present in the reaction, that is incorporated into ␣-casein from each species of dO⅐L-tRNA. R 2 for each dO⅐L-tRNA is the ratio R 1 (dO⅐L-tRNA)/R 1 (none).  5. Schematic of the acceptor stems of L/F-transferase substrate tRNAs and current data concerning their interaction with the enzyme. Data for the tRNA Leu isoacceptors is from Rao and Kaji (23), while the data for the tRNA Met isoacceptors is from Scarpulla et al. (26). temperature complex), this stable AA-tRNA⅐EF-Tu complex displays a 1:1 stoichiometry. Conversely, the physiologically relevant complex at higher temperature displays a 1:2 stoichiometry, and is much less long-lived (27). Second, non-productive binding modes are most easily envisioned for an enzyme with degenerate substrate binding specificities, and aminoacyl-tRNA recognition by the L/F-transferase is unusually degenerate. For example, if the L/F-transferase's substrate binding site recognizes the charged polyphosphates of single-stranded regions of the tRNA in conjunction with a hydrophobic side chain, these interactions might be provided by multiple conformations of the AA-tRNA⅐L/F complex.