Portal de Periódicos da CAPES

Crystal structures of leucyl/phenylalanyl-tRNA-protein transferase and its complex with an aminoacyl-tRNA analog

2006; Springer Nature; Volume: 25; Issue: 24 Linguagem: Inglês

10.1038/sj.emboj.7601433

1460-2075

Kyoko Suto, Yoshihiro Shimizu, Kazunori Watanabe, Takuya Ueda, Shuya Fukai, Osamu Nureki, Kozo Tomita,

RNA Research and Splicing

Article16 November 2006free access Crystal structures of leucyl/phenylalanyl-tRNA-protein transferase and its complex with an aminoacyl-tRNA analog Kyoko Suto Kyoko Suto Institute for Biological Resources and Functions, National Institute of Advanced Industrial Science and Technology (AIST), Tsukuba, Ibaraki, Japan Search for more papers by this author Yoshihiro Shimizu Yoshihiro Shimizu Department of Medical Genome Sciences, Graduate School of Frontier Sciences, University of Tokyo, Kashiwa, Chiba, Japan Search for more papers by this author Kazunori Watanabe Kazunori Watanabe Institute for Biological Resources and Functions, National Institute of Advanced Industrial Science and Technology (AIST), Tsukuba, Ibaraki, Japan Search for more papers by this author Takuya Ueda Takuya Ueda Department of Medical Genome Sciences, Graduate School of Frontier Sciences, University of Tokyo, Kashiwa, Chiba, Japan Search for more papers by this author Shuya Fukai Shuya Fukai Department of Biological Information, Graduate School of Bioscience and Technology, Tokyo Institute of Technology, Nagatsuda-cho, Midori-ku, Yokohama, Kanagawa, Japan Search for more papers by this author Osamu Nureki Osamu Nureki Department of Biological Information, Graduate School of Bioscience and Technology, Tokyo Institute of Technology, Nagatsuda-cho, Midori-ku, Yokohama, Kanagawa, Japan Search for more papers by this author Kozo Tomita Corresponding Author Kozo Tomita Institute for Biological Resources and Functions, National Institute of Advanced Industrial Science and Technology (AIST), Tsukuba, Ibaraki, Japan Search for more papers by this author Kyoko Suto Kyoko Suto Institute for Biological Resources and Functions, National Institute of Advanced Industrial Science and Technology (AIST), Tsukuba, Ibaraki, Japan Search for more papers by this author Yoshihiro Shimizu Yoshihiro Shimizu Department of Medical Genome Sciences, Graduate School of Frontier Sciences, University of Tokyo, Kashiwa, Chiba, Japan Search for more papers by this author Kazunori Watanabe Kazunori Watanabe Institute for Biological Resources and Functions, National Institute of Advanced Industrial Science and Technology (AIST), Tsukuba, Ibaraki, Japan Search for more papers by this author Takuya Ueda Takuya Ueda Department of Medical Genome Sciences, Graduate School of Frontier Sciences, University of Tokyo, Kashiwa, Chiba, Japan Search for more papers by this author Shuya Fukai Shuya Fukai Department of Biological Information, Graduate School of Bioscience and Technology, Tokyo Institute of Technology, Nagatsuda-cho, Midori-ku, Yokohama, Kanagawa, Japan Search for more papers by this author Osamu Nureki Osamu Nureki Department of Biological Information, Graduate School of Bioscience and Technology, Tokyo Institute of Technology, Nagatsuda-cho, Midori-ku, Yokohama, Kanagawa, Japan Search for more papers by this author Kozo Tomita Corresponding Author Kozo Tomita Institute for Biological Resources and Functions, National Institute of Advanced Industrial Science and Technology (AIST), Tsukuba, Ibaraki, Japan Search for more papers by this author Author Information Kyoko Suto1,‡, Yoshihiro Shimizu2,‡, Kazunori Watanabe1, Takuya Ueda2, Shuya Fukai3, Osamu Nureki3 and Kozo Tomita 1 1Institute for Biological Resources and Functions, National Institute of Advanced Industrial Science and Technology (AIST), Tsukuba, Ibaraki, Japan 2Department of Medical Genome Sciences, Graduate School of Frontier Sciences, University of Tokyo, Kashiwa, Chiba, Japan 3Department of Biological Information, Graduate School of Bioscience and Technology, Tokyo Institute of Technology, Nagatsuda-cho, Midori-ku, Yokohama, Kanagawa, Japan ‡These authors contributed equally to this work *Corresponding author. Institute for Biological Resources and Functions, National Institute of Advanced Industrial Science and Technology (AIST), 1-1-1, Higashi, Tsukuba, Ibaraki 305-8566, Japan. Tel.: +81 29 861 6085; Fax: +81 29 861 6095; E-mail: [email protected] The EMBO Journal (2006)25:5942-5950https://doi.org/10.1038/sj.emboj.7601433 PDFDownload PDF of article text and main figures. ToolsAdd to favoritesDownload CitationsTrack CitationsPermissions ShareFacebookTwitterLinked InMendeleyWechatReddit Figures & Info Eubacterial leucyl/phenylalanyl-tRNA protein transferase (L/F-transferase), encoded by the aat gene, conjugates leucine or phenylalanine to the N-terminal Arg or Lys residue of proteins, using Leu-tRNALeu or Phe-tRNAPhe as a substrate. The resulting N-terminal Leu or Phe acts as a degradation signal for the ClpS-ClpAP-mediated N-end rule protein degradation pathway. Here, we present the crystal structures of Escherichia coli L/F-transferase and its complex with an aminoacyl-tRNA analog, puromycin. The C-terminal domain of L/F-transferase consists of the GCN5-related N-acetyltransferase fold, commonly observed in the acetyltransferase superfamily. The p-methoxybenzyl group of puromycin, corresponding to the side chain of Leu or Phe of Leu-tRNALeu or Phe-tRNAPhe, is accommodated in a highly hydrophobic pocket, with a shape and size suitable for hydrophobic amino-acid residues lacking a branched β-carbon, such as leucine and phenylalanine. Structure-based mutagenesis of L/F-transferase revealed its substrate specificity. Furthermore, we present a model of the L/F-transferase complex with tRNA and substrate proteins bearing an N-terminal Arg or Lys. Introduction Regulated degradation of intracellular proteins is essential for protein quality control in both eubacteria and eukaryotes. The N-end rule pathway, one of the most prevalent proteolytic pathways, functions in the in vivo half-life control of selected proteins, by destroying them according to the N-terminal residue (Varshavsky, 1996, 2003). In eukaryotes, the N-end rule pathway is part of the ubiquitin (Ub) system (Hu et al, 2005; Nandi et al, 2006), which controls peptide import (Turner et al, 2000; Du et al, 2002), chromosomal segregation fidelity (Rao et al, 2001), apoptosis regulation (Ditzel et al, 2003; Varshavsky, 2003), and nitric oxide detection (Hu et al, 2005). In eukaryotes, the destabilizing N-terminal residues are hierarchical. The primary destabilizing residues are categorized into two types: the type-1 destabilizing residues are basic (Arg, Lys, or His), whereas the type-2 residues are bulky and hydrophobic (Phe, Leu, Trp, Tyr, or Ile). Multiple E3 Ub ligases (N-recognins) recognize these N-terminal residues, and conjugate a poly-Ub chain to the target protein. The resulting ubiquitylated proteins are thereafter progressively degraded by the 26S proteasome in an ATP-dependent manner. The secondary destabilizing N-terminal residues, Asp and Glu, are recognized by Arg-tRNA-protein transferase (R-transferase), which conjugates an arginine, the primary destabilizing residue, to them. The N-terminal Asn and Gln are tertiary destabilizing residues, in that they are deaminated to yield the secondary destabilizing N-termini of Asp and Glu, respectively. Recently, N-terminal Cys and oxidized Cys residues have also been reported to behave as tertiary and secondary destabilizing residues, respectively, in mammals (Kwon et al, 2002; Hu et al, 2005). Thus, the functions of the N-end rule pathway have been well elucidated in eukaryotes (Hu et al, 2005), whereas they still remain unknown in eubacteria. Eubacteria have an analogous, but Ub-independent, N-end rule pathway (Tobias et al, 1991; Shrader et al, 1993). As in eukaryotes, the destabilizing residues in bacteria are also hierarchical. In Escherichia coli, the primary destabilizing N-terminal residues are Leu, Phe, Trp, and Tyr. Proteins with these N-terminal residues are degraded by ClpAP, a proteasome-like protease consisting of the AAA+ chaperone ClpA and ClpP peptidase (ClpAP) (Tobias et al, 1991). Recently, the ClpAP-specific adaptor, ClpS, was identified as an N-recognin, an essential component of the pathway (Erbse et al, 2006). ClpS shares secondary structural features with UBR1, the E3 Ub ligase of the eukaryotic N-end rule pathway (Lupas and Koretke, 2003), and binds to a primary destabilizing residue of a substrate to deliver it to the ClpAP protease for degradation (Erbse et al, 2006). In E. coli, the secondary destabilizing N-terminal residues are Arg and Lys. These N-terminal residues are recognized by leucyl/phenylalanyl-tRNA-protein transferase (L/F-transferase), which conjugates leucine or phenylalanine to the N-terminal Arg or Lys, followed by ClpS recognition. However, the physiological functions of the eubacterial N-end rule pathway remain to be elucidated. There are several similarities between the eukaryotic and eubacterial N-end rule pathways, such as the hierarchic structure of the destabilizing N-terminal residues, the requirement of the specific N-recognin (E3 Ub ligase in eukaryotes and ClpS in eubacteria), and the involvement of ATP-dependent proteases (the 26S proteasome complex in eukaryotes and the proteasome-like ClpAP in eubacteria). Furthermore, in both systems, an aminoacyl-tRNA-protein transferase (R-transferase in eukaryotes and L/F-transferase in eubacteria) conjugates a primary destabilizing residue to the substrate's secondary destabilizing N-terminal residue. These similarities between the N-end rule pathways in the two kingdoms imply that these pathways and their components share a common origin. However, in contrast to the sequence similarity between the eukaryotic E3 Ub ligase and the eubacterial ClpS (Lupas and Koretke, 2003), no significant sequence similarity exists between the eukaryotic R-transferase and the eubacterial L/F-transferase. E. coli L/F-transferase was identified about four decades ago (Kaji et al, 1965a, 1965b). Biochemical studies revealed that E. coli L/F-transferase catalyzes the transfer of Leu and Phe, and Met and Trp less efficiently, using the cognate aminoacyl-tRNAs (Kaji et al, 1965a, 1965b; Horinishi et al, 1975; Abramochkin and Shrader, 1996) to the N-terminal Arg and Lys of acceptor proteins (Soffer, 1973). It was also reported that the anticodon of the aminoacyl-tRNA is not a determinant for the enzyme recognition, but that the single-stranded acceptor region of the aminoacyl-tRNA is required for the L/F-transferase activity (Abramochkin and Shrader, 1996). However, the detailed molecular basis for the recognition of the aminoacyl moiety of an aminoacyl-tRNA by L/F-transferase has remained obscure. Here, we have determined the crystal structures of E. coli L/F-transferase and its complex with puromycin, an analog of the aminoacyl-tRNA. The structure of this complex, together with extensive biochemical mutational studies, revealed the mechanism by which L/F-transferase specifically recognizes the aminoacyl-moiety of aminoacyl-tRNAs. Based on the current structure, we also present a model of the L/F-transferase complex with tRNA and substrate proteins bearing an N-terminal Arg or Lys residue. Results and discussion Overall architecture of E. coli L/F-transferase and structural similarity with the FemABX enzyme family The E. coli L/F-transferase was overexpressed in E. coli and crystallized. The apo structure was initially solved by multi-wavelength anomalous dispersion (MAD), using the selenomethionine-labeled protein, and the native structure was refined to an R factor of 27.5% (Rfree of 22.2%) using reflections up to 2.4 Å resolution. Subsequently, the data sets were collected from crystals soaked in puromycin. The structure of the complex with puromycin was refined to an R factor of 27.5% (Rfree of 22.5%) up to 2.8 Å resolution (Supplementary Table 1). E. coli L/F-transferase forms a compact structure and consists of two domains: an NH2-terminal domain (amino-acid residues 3–62) and a COOH-terminal domain (amino-acid residues 63–232) (Figure 1A). The NH2-terminal domain is composed of four β-strands (β1–4) and one α-helix (α1), whereas the COOH-terminal domain is composed of eight β-strands (β5–12) surrounded by six α-helices (α2–7) (Figure 1B). Puromycin, an aminoacyl-tRNA analog, binds to the cleft formed by β9, α5, and β10 in the COOH-terminal domain and by β3 in the NH2-terminal domain (see below) (Figures 1A and 2). Figure 1.Overall architecture of E. coli L/F-transferase. (A) Stereo view of the E. coli L/F-transferase structure. The NH2-terminal domain (residues 2–62) and the COOH-terminal domain (residues 63–232) are colored blue and green, respectively. The puromycin bound to the hydrophobic pocket is colored yellow. (B) Topology diagram of L/F-transferase. The rimmed elements in the COOH-terminal domain (α3–α5) and (β5–β12) are common to the GNAT superfamily fold. The α-helices and β-strands in the COOH-terminal domains are colored red and yellow, respectively. (C) Comparison of the structures of E. coli L/F-transferase (left), W. viridescens FemX (wvFemX; middle, PDB accession number 1P4N; Biarrotte-Sorin et al, 2004) and S. aureus FemA (saFemA; PDB accession number 1LRZ; Benson et al, 2002). The COOH-terminal domain of L/F-transferase is topologically similar to the domain 2′s of wvFemX and saFemA. The conserved α-helices and β-strands in L/F-transferase, wvFemX and saFemA, are colored red and yellow, respectively. Download figure Download PowerPoint Figure 2.Sequence alignment of the L/F-transferases from E. coli (Aat_E. coli), Xylella fastidosa (Aat_X-fast; accession number ZP_00683190.1), Pseudomonas aeruginosa (Aat_P. aeru; accession number ZP_00204849.1), Synechocystis (Aat_Synech; accession number NP_440931.1), and Mesorhizobium loti (Aat_M. loti; accession number NP_102051.1). The sequence of R-transferase from Plasmodium falciparum (ATEL_P. fa; accession number NP_473045.1). The secondary structure elements of E. coli L/F-transferase are indicated above the alignment. The α-helices and β-strands in the NH2-terminal domain are colored blue, and those of the COOH-terminal domain are colored red and yellow, respectively, as shown in Figure 1B. Download figure Download PowerPoint The COOH-terminal domain of L/F-transferase is topologically similar to domain 2 of Weissella viridescens FemX (wvFemX; Biarrotte-Sorin et al, 2004) and that of Staphylococcus aureus FemA (saFemA; Benson et al, 2002) (Figure 1C, middle and right, respectively). Both wvFemX and saFemA belong to the FemABX family (Hegde and Shrader, 2001), containing a GCN5-related N-acetyltransferase (GNAT) fold (Sterner and Berger, 2000) (Figure 1B), and catalyze the transfer of an amino acid to a precursor of peptidoglycan, using an aminoacyl-tRNA as a substrate. WvFemX transfers an alanine to the ε-amino group of a Lys side chain using Ala-tRNAAla as a substrate, whereas saFemA transfers a glycine to the α-amino group of a Gly main chain, utilizing Gly-tRNAGly, to form an interpeptide bridge of the peptidoglycan. Therefore, the topological similarity between the COOH-terminal domain of L/F-transferase and the domain 2′s of wvFemX and saFemA may reflect the similarity of their chemical reactions, where an amino acid is transferred from an aminoacyl-tRNA to the amino group of a protein (or peptide). It is interesting to note that, although the topological and structural similarity is readily apparent between L/F-transferase and the FemABX families (Figure 1C), there is no significant amino-acid similarity between these enzymes, suggesting that they might have arisen from a common ancestor, but have divergently evolved. In contrast to the topological similarity between the COOH-terminal domain of L/F-transferase and the domain 2′s of wvFemx and saFemA, no significant topological similarity can be identified between the NH2-terminal domain of L/F-transferase and the domain 1's of wvFemX and saFemA (Figure 1C). Recognition of an aminoacyl-tRNA analog, puromycin, by L/F-transferase The chemical structure of puromycin is similar to that of the 3′-terminus of an aminoacyl-tRNA; the carboxyl group of p-methoxyphenylalanine is linked to the 3′-amino group of 3′-amino-6-N,N-dimethyladenosine by an amide bond (Figure 3A). The p-methoxybenzyl group and the 6-N,N-dimethyladenosine correspond to the side chain of an amino acid and the adenosine at the CCA end of an aminoacyl-tRNA, respectively. Figure 3.Recognition of the puromycin by E. coli L/F-transferase. (A) Chemical structure of puromycin (left) and that of the 3′-ends of Leu-tRNALeu and Phe-tRNAPhe (middle and right, respectively). The amino-acid moiety and the base moiety are colored pink and blue, respectively. (B) ∣Fo−Fc∣ omit map of puromycin (contour level 3.0σ). (C) Recognition of the p-methoxybenzyl group and the puromycin base by the hydrophobic pocket, as shown by a surface model. (D) Ribbon model of (C). The hydrophobic amino acid involved in the recognition of the p-methoxybenzyl group and the base moiety of puromycin are colored green and blue, respectively. (E) The C-shaped edge of the hydrophobic pocket is composed of continuous amino-acid residues (Gly155-Glu156-Ser157-Met158; colored yellow and highlighted). The α-, β- and γ-carbons of puromycin are also shown. Download figure Download PowerPoint Puromycin reportedly inhibits the activity of L/F-transferase by preventing the aminoacyl-tRNA from binding to the enzyme, suggesting that puromycin binds to the same position in L/F-transferase as the 3′-end of an aminoacyl-tRNA (Horinishi et al, 1975; Abramochkin and Shrader, 1996). To elucidate the molecular basis of the recognition of the aminoacyl moiety of aminoacyl-tRNAs by L/F-transferase, apo L/F-transferase crystals were soaked in a solution containing puromycin, and the crystal structure was determined (Supplementary Table 1). The electron density corresponding to the puromycin was clearly visible in the complex structure (Figure 3B), with no significant structural change observed between the apo and complex structures. The puromycin binds to a cleft formed by β9, α5, and β10 in the COOH-terminal domain and β3 in the NH2-terminal domain (Figure 1A). The p-methoxybenzyl group of puromycin is docked within a deep pocket at the bottom of the cleft (Figure 3C and D). The pocket is composed of the side chains of several hydrophobic residues (Met144, Phe153, Leu170, Phe173, Ile185), and thus the inner surface is quite hydrophobic. The recognition of the p-methoxybenzyl group of puromycin by L/F-transferase is achieved through a hydrophobic interaction with these amino-acid residues. The edge of the pocket is formed by the main chains of continuous residues (Gly155, Glu156, Ser157) and the side chain of Met158 (Figure 3E), adopting a C-shaped structure. The β-carbon of the p-methoxyphenylalanine group of puromycin is cramped by the C-shaped edge of the pocket: the side chain of Met158 and the peptide bond between Gly155 and Glu156 sandwich the β-carbon of puromycin (Figure 3E). The distance between the Cγ of puromycin and the Cε of Met158 is 3.8 Å, and the distances between the Cβ of puromycin and the Cα of Ser157 and the main chain N of Glu156 are 3.5 and 4.0 Å, respectively (Figure 3E). The 6-N,N-dimethyladenine group of puromycin is stabilized mainly by a π–π stacking interaction with Trp49, which is further stabilized by a stacking interaction with Trp111 (Figure 3D). The hydrophobic amino-acid residues, Trp59, Phe47, and Val189, interact with the base moiety of puromycin through hydrophobic interactions, and these interactions are not specific to the 6-N,N-dimethyladenine of puromycin. The 2′-hydroxyl group of the ribose moiety of puromycin hydrogen bonds with the main-chain carbonyl group of Glu156 (Figure 3D). Recognition mechanism of the amino-acid moiety of the aminoacyl-tRNA by L/F-transferase The chemical structure of puromycin resembles that of the 3′-termini of aminoacyl-tRNAs, as described (Figure 3A). The reported inhibitory effect by puromycin on L/F-transferase activity might reflect the similar hydrophobic properties of the p-methoxybenzyl group of puromycin to the side chains of phenylalanine and leucine (Figure 3A). Moreover, the 6-N,N-dimethyladenosine corresponds to the 3′-terminal adenosine at position 76 in the aminoacyl-tRNAs. To explore the detailed molecular mechanism of the recognition of the 3′-end of Leu-tRNALeu or Phe-tRNAPhe by L/F-transferase, a series of site-directed mutations were introduced into the L/F-transferase residues that are proximal to puromycin, based on the present complex structure, and the activities of leucyl and phenylalanyl transfer to α-casein bearing an NH2-terminal Arg, from the respective aminoacyl-tRNAs, were analyzed (Figure 4A and B). Figure 4.In vitro activity assays for various mutants of L/F-transferase. (A) The activity of [14C]Leu incorporation into α-casein from [14C]Leu-tRNALeu by mutant L/F-transferases (see Materials and methods). The upper gel shows the [14C]Leu-labeled α-casein. The lower graph shows the quantification of the relative intensity of the labeled product of the upper gel, where the incorporation of [14C]Leu into α-casein by wild-type L/F-transferase was taken as 1.0. The bars on the graph indicate the s.d. of more than three independent experiments. (B) The activity of [14C]Phe incorporation into α-casein from [14C]Phe-tRNAPhe by mutant L/F-transferases, as in (A). Download figure Download PowerPoint Mutations of hydrophobic amino-acid residues (Met144, Phe153, Leu170, Phe173, Ile185) that accommodate the p-methoxybenzyl group of puromycin (Figure 3C and D) considerably reduced the L/F-transferase activity (Figure 4A and B). Therefore, it is reasonable to assume that the hydrophobic leucyl and phenylalanyl moieties of Leu-tRNALeu and Phe-tRNAPhe, respectively, are recognized by this highly hydrophobic pocket of L/F-transferase. Actually, superposition of the α and β carbons of Phe and Leu onto those of puromycin revealed that the side chains of Phe and Leu can enter the hydrophobic pocket without any steric clashes (Figure 5A and B). The corresponding amino-acid residues of L/F-transferases from other eubacteria are well conserved as hydrophobic amino-acid residues (Figure 2). These results explain why L/F-transferase recognizes aminoacyl-tRNAs attached to hydrophobic amino acids, and excludes those coupled to hydrophilic or charged amino acids. Figure 5.Recognition of the aminoacyl-moiety by L/F-transferase. Recognition of the side chains of Leu (A) and Phe (B) by L/F-transferase. The α- and β-carbons of Leu and Phe were superimposed onto those of puromycin (colored orange). Discrimination of Ile (C) and Val (D) by L/F-transferase. The α- and β-carbons of Ile and Val were superimposed as in (A) and (B). The steric hindrance between the branched methyl groups at the β-carbons of Ile and Val. The C-shaped edge of the hydrophobic pocket is colored yellow. The chemical structures of each amino acid are depicted in parentheses. Download figure Download PowerPoint The β-carbon of the aminoacyl moiety of puromycin is sandwiched between the side chain of Met158 and the peptide-bond plane of Gly155–Glu156 (Figure 3E). Both the side chains of the leucyl and phenylalanyl moieties of the respective aminoacyl-tRNAs lack branched β-carbons (Figures 3A, 5A, and B). When the Cα and Cβ of Ile and Val, possessing branched β-carbons, are superimposed onto those of puromycin (Figure 5C and D), the distances between the branched Cγ2 of Ile and the carbonyl carbon of Glu156, and between the branched Cγ2 of Val and the Cα of Ser157 are 2.6 and 2.5 Å, respectively. These close distances between the branched methyl groups of Ile and Val and the C-shaped edge would cause steric hindrance, thus precluding aminoacyl-tRNAs charged with β-branched amino acids, such as Ile and Val. These results explain the specificity of L/F-transferase for amino acids bearing an unbranched β-carbon (Abramochkin and Shrader, 1996). For the other hydrophobic amino-acid residues, such as Ala, Pro, Trp, and Met, the size of the Ala and Pro side chains is not large enough to fit within the hydrophobic pocket (Supplementary Figure 1A and B). On the other hand, L/F-transferase reportedly transfers Met and Trp to proteins bearing an N-terminal Arg or Lys, although the activities are less efficient as compared with those of Leu and Phe (Kaji et al, 1965a; Abramochkin and Shrader, 1996). Our in vitro assays using Met-tRNAMet and Trp-tRNATrp as substrates also showed that both Met and Trp could be transferred to the NH2-terminal Arg of the α-casein fragment (Supplementary Figure 2D and E). However, in the presence of Phe-tRNAPhe, Leu-tRNALeu, Met-tRNAMet, and Trp-tRNATrp, Phe and Leu were predominantly transferred to the α-casein fragment (Supplementary Figure 2F). The superposition of the Cα and Cβ of Trp onto those of puromycin (Supplementary Figure 1C) revealed that the size of the Trp side chain (indole group) is too large to be properly accommodated in the hydrophobic pocket, and thus the indole group might become snagged on the C-shaped edge of the pocket. On the other hand, the size of the Met side chain is sufficiently large enough to be accommodated into the hydrophobic pocket properly (Supplementary Figure 1D). However, the contact area of the Met side chain with the hydrophobic pocket is smaller than that of Leu and Phe. These findings explain the previous and present data showing that Trp and Met are incorporated into the protein by L/F-transferase less efficiently than Leu or Phe in vitro, and that Leu and Phe are incorporated predominantly in vivo. Taken together, the hydrophobicity and size of the pocket and the confined C-shaped structure of the edge of the pocket could collaboratively discriminate Leu and Phe predominantly (and Trp and Met less efficiently), from other amino-acid residues to be transferred to acceptor proteins bearing Arg or Lys at the N-terminus. It is notable that the continuous amino-acid residues composing the C-shaped edge (Gly155–Met158) of the hydrophobic pocket are well conserved among the eubacterial L/F-transferases (Figure 2). The 6-N,N-dimethyladenosine of puromycin corresponds to the 3′-terminal adenosine residue (A76) of tRNA, as described above (Figure 3A). The 6-N,N-dimethyladenosine of puromycin interacts with several amino-acid residues (Figure 3D). The mutations of Trp49 and Trp111 drastically reduced the L/F-transferase activity (Figure 4), suggesting that the stacking interaction between the base moiety of A76 and Trp49 and the further stabilization by Trp111 are crucial for recognition of the 3′-terminal nucleotide of the aminoacyl-tRNAs. This result is consistent with the previous finding that the activity of a recombinant L/F-transferase, lacking the NH2-terminal 78 amino acids, is significantly reduced (Ichetovkin et al, 1997). Moreover, the mutation of Val189, which interacts with the base moiety through a hydrophobic interaction, reduced the activity, as expected (Figure 4). The present structural and biochemical studies of L/F-transferase (Figures 3 and 4) strongly suggest that the 3′-terminus of the aminoacyl-tRNA is recognized by the combination of the hydrophobic aminoacyl moiety recognizing pocket and the 3′-nucleotide recognition site, although the 3′-nucleotide binding site is not specific for the adenosine. Docking model of L/F-transferase and aminoacyl-tRNA The electrostatic potential surface of L/F-transferase reveals the highly biased distribution of charged residues (Figure 6A, lower panel). Especially, the α2 helix in the COOH-terminal domain contains a cluster of positively charged residues (Arg76, Arg80, Lys83, and Arg84). These amino-acid residues protrude toward the solvent and are conserved among the eubacterial L/F-transferases (Figure 2). Moreover, next to the charged region, a cleft suitable for the accommodation of the 3′-region of a tRNA-acceptor helix is formed by helices α5 and α6, extending toward the puromycin-binding pocket (Figure 6A, lower and right panels). The amino-acid residues in the NH2-terminal half of α5 (Asn164, Ser166, and Lys167) and Asn191 and His193 in α6 are well conserved among the eubacterial L/F-transferases (Figure 2). Mutations of the positively charged residues in α2 (Arg76, Arg80, Lys83, and Arg84) and α6 (His193) remarkably reduced the L/F-transferase activity (Figure 4). A comparison of the surface electrostatic potential of L/F-transferase with that of wvFemX and saFemA revealed similar distributions of positively charged amino-acid residues, in the regions corresponding to the α2 helix of L/F-transferase, and the distances between the positively charged region and the substrate-binding pocket were almost the same as those observed in L/F-transferase (Figure 6B). Figure 6.Model of aminoacyl-tRNA binding to L/F-transferase. (A) Two views of a ribbon diagram of the docking model of L/F-transferase and tRNA. The tRNA backbone is shown as a green line (upper two panels). Two views of the L/F-transferase-tRNA complex model, showing the surface colored according to its calculated electrostatic potential (lower panel; blue, positively charged +8KT; red, negatively charged –8KT). The electrostatic surface model was calculated by the program APBS (Baker et al, 2001). The upper view displays the top side of the complex of front views. (B) Electrostatic potentials of wvFemX (upper) and saFemA (lower). The marked regions on the diagrams show the positively charged regions corresponding to α2 of L/F-transferase and predicted as RNA binding regions. (C) The simultaneous tRNA binding to L/F-transferase (colored green) and EF-Tu (colored pink). The tRNA phosphate backbones are shown as a green line and a pink line for the L/F-transferase-tRNA complex and the EE-Tu-tRNA complex, respectively. The EF-Tu and Cys-tRNACys complex structure was