tRNA as an active chemical scaffold for diverse chemical transformations

Abstract

During protein synthesis, tRNA serves as the intermediary between cognate amino acids and their corresponding RNA trinucleotide codons. Aminoacyl-tRNA is also a biosynthetic precursor and amino acid donor for other macromolecules. AA-tRNAs allow transformations of acidic amino acids into their amide-containing counterparts, and seryl-tRNA^Ser donates serine for antibiotic synthesis. Aminoacyl-tRNA is also used to cross-link peptidoglycan, to lysinylate the lipid bilayer, and to allow proteolytic turnover via the N-end rule. These alternative functions may signal the use of RNA in early evolution as both a biological scaffold and a catalyst to achieve a wide variety of chemical transformations.

Transfer RNA (tRNA) enjoys the distinction of being discovered as a biochemical entity (technically as sRNA) at virtually the same time as its existence was predicted as an essential component in protein synthesis. While tRNA would later receive attention as the first nucleic acid molecule to have its atomic structure determined by X-ray crystallography [1], thereby confirming many of the Watson and Crick base pairing postulates, Crick’s “adaptor hypothesis” [2] remains a singular scientific achievement, because it originated almost entirely from theoretical considerations associated with the genetic coding problem. In the adaptor hypothesis, Crick made the bold conceptual leap that no obvious physical and chemical basis exists for associating particular combinations of the four ribonucleotides with each of the twenty amino acids, so hydrogen bonding and base pairing would have to remain the chemical basis by which genetic information would ultimately be expressed. Accordingly, a set of adaptors was proposed corresponding to the 20 canonical standard amino acids, each possessing a distinct end to accommodate the cognate amino acid, and a second end to interact with the templating RNA. These, in turn, would dictate the existence of a set of aminoacylating enzymes, each one specific for one of the twenty amino acids and their cognate adaptors. In conjunction with a non-overlapping three base degenerate code, this model provided a remarkably elegant and essentially correct resolution to the coding problem.

Despite the durability of the adaptor hypothesis as a workable answer to the coding problem, its principal assumptions were soon challenged by new observations in the literature. Crick never entertained the possibility that cells might be able to survive with less than a full set of adaptor-charging enzymes, and yet still produce proteins with all 20 amino acids. Similarly, the possibility that tRNA adaptors might have roles outside of protein synthesis was not envisioned. By the late 60’s, however, Wilcox & Nirenberg would report that some bacterial species lack a GlnRS ortholog, with Gln-tRNA^Gln being produced by a transamidation reaction [3]. Unanticipated roles for tRNA also emerged in the synthesis of the peptide crosslinker of peptidoglycan [4] and in lipid modification [5], representing some of the first forays of tRNA outside of protein synthesis.

Yet while these observations were being reported, theoretical work during this period addressed the importance of self-replicating nucleic acids in the origin of life, and the potential significance of small molecules linked to RNA [6]. The possibility that RNA might be able to catalyze aminoacylation was also noted (and validated much later [7]). Others would later propose that modern day enzymatic co-factors, particularly nicotinamide and flavin co-factors, are vestiges of the RNA world [8]. Support for the importance of pre-biotic chemistries involving RNA-based catalysis would later come with the discovery that RNase P [9] and the ribosome [10,11] both have RNA-based catalytic engines, and with the development of in vitro selection protocols that allowed catalytic RNAs to be isolated from large complexity nucleic acid pools [12]. With the current explosion of new regulatory roles for small RNAs [13], it is perhaps a timely occasion to consider some of the functions of transfer RNA that are distinct from its canonical role in protein synthesis. Here, we review the “non-protein synthesis” roles of tRNA, particularly those in which it serves as a biosynthetic scaffold for other biochemical functions (Figure 1). Interestingly, a number of these secondary functions are concerned with the plasma membrane and the cell wall, structures that separate the contents of the cell from its external environment. In other functions, the aminoacylated tRNA serves as a substrate for further biosynthetic transformations, generating final compounds whose functions are outside the realm of protein synthesis.

Figure 1. Secondary roles of tRNA in cellular metabolism.

The canonical roles of tRNA in protein synthesis are represented by “Chemistry of Aminoacylation and Editing” and “Ribosomal Protein Synthesis,” while other functions discussed in this review are arrranged in a circular fashion around a central tRNA molecule. Depending upon its role in a particular process, the tRNA molecule shown in the circle should be considered as either free tRNA (as shown) or an aminoacyl-tRNA. The details of these processes are discussed in the text.

tRNA as an active participant in reaction chemistry

The concept of tRNA as an active chemical scaffold implies that specific chemical reactions involving the aminoacyl moiety are facilitated as a direct consequence of the presence of the tRNA. Recent work on a number of reactions, including peptidyl transfer, aminoacyl transfer, and deacylation, underscores a potential catalytic role for tRNA. As representative examples, we focus here on reactions that are localized to one of the two vicinal hydroxyls of nucleotide A76, which constitutes the 3′ terminus [14]. Each hydroxyl group is able to engage in H-bonds as either a donor or acceptor, making these groups particularly valuable for bridging interactions between the tRNA, the enzyme, and the solvent. Relative to a free hydroxyl group, the pK_a’s of these vicinal groups are depressed by at least three units (12.5 vs. 16 units for ethanol), implying a higher degree of reactivity than for a single hydroxyl group.

A particularly significant example of the direct participation of the vicinal hydroxyls in reaction chemistry is found in the peptidyl transferase reaction [15]. One component of the ribosome’s enhancement of the rate of peptidyl transfer is the deprotonation of the -NH₃ group on the A-site tRNA, which serves as the nucleophile. The 3′OH leaving group on the P-site tRNA must also be protonated. There is an absence of protein-based catalytic groups in the ribosomal 50S subunit structure where the peptidyl transferase center is located, motivating the search for RNA-based functional groups able to contribute to catalysis [11]. While early attention initially focused on A2451 of the 23S large subunit RNA [11,16], experiments featuring mutant ribosomes subsequently showed that this nucleotide does not make an essential contribution to catalysis [17,18]. Analysis of ribosome crystal structures indicates that one of the only ionizable groups in the neighborhood of the NH₂ nucleophile is the 2′OH of the P-site tRNA, highlighting it as a possible general base [19]. Support for this model was provided by tRNA substitution experiments, which showed that there is an ~10⁶-fold decrease in the rate of peptidyl transfer when the 2′OH is replaced by a deoxy or fluoro [20]. While the catalytic power of the ribosome likely has a number of different origins, including a significant entropic component originating from fixing the geometry of the reactants, as well as excluding solvent water [21,22], a proton shuttle mechanism (Figure 2A) involving the P-site 2′OH is arguably an important catalytic element. In such a mechanism, the concerted nature of the proton transfer means that the 2′OH need not assume a full charge during the reaction, arguing that a major pK_a perturbation is not required [14].

Figure 2. Role of tRNA in facilitating key steps in protein synthesis.

A., Substrate-assisted catalysis of peptide bond formation in the ribosome, adapted from Weinger at al (ref 20). The -amino group of the A-site tRNA (blue) serves as the nucleophile in the attack on the carbonyl carbon of the P-site tRNA (magenta). The specific function of the 2′OH as a general base in the reaction is still under investigation. B., Substrate assisted catalysis of aminoacyl transfer by threonyl-tRNA synthetase, adapted from Minajigi & Francklyn (ref. 23). The threonyl adenylate is depicted in blue, and the A76 group of incoming tRNA^Thr in magenta. C., Proposed role for the adjacent hydroxyl of A76 in stabilizing the geometry of the attacking water molecule in the deacylation of Thr-tRNA^Phe by phenylalanyl-tRNA synthetase, adapted from Ling et al. (ref 26). The decreases in aminoacyl transfer by threonyl synthetase and post transfer editing by phenylalanyl-tRNA synthetase associated with removal of the adjacent hydroxyl are of the same order (750 vs 300-fold, respectively), but several orders magnitude less than the effect of loss of the adjacent hydroxyl on peptidyltransferase (~10⁶).

Recent studies have provided evidence that similar mechanisms may be involved in the aminoacyl transfer and deacylation reactions catalyzed by the aminoacyl-tRNA synthetases (ARSs). These enzymes catalyze a two step reaction in which amino acid and ATP are first condensed to form an enzyme linked adenylate, followed by a subsequent step in which the amino acid is transferred to the 3′ end of the cognate tRNA. In ThrRS, a systematic study of the conserved and likely catalytic residues in the active site showed that only one (His309) is likely to be directly involved in aminoacyl transfer [23]. Notably, His309 is hydrogen bonded to the 2′OH of tRNA^Thr [24]. Substitution studies on ATP and tRNA substrates showed that replacement of the non-bridging oxygen of the adenylate with sulfur had only minimal effect on the rate of transfer, whereas substitutions of the 2′OH with deoxy or fluoro significantly decreased (100–700-fold) the rate of aminoacyl transfer [23]. Notably, there was a synergistic effect of combining the 2′OH substitution with a His309A substitution. Accordingly, a “proton transfer” model (Figure 2B) for the activation of tRNA^Thr was proposed where His309 assists in the deprotonation of the 2′OH, which in turn assists in the activation of the 3′OH [23]. An analogous proton shuttle has been proposed for liver alcohol dehydrogenase, where the substrate alcoholic protein is transferred first to a conserved serine, then to one of the vicinal hydroxyl groups of the co-factor NADH ribose, and then finally to a conserved histidine [25]. These observations draw an explicit functional parallel for the first time between a tRNA and a nicotinamide-based cofactor, supporting the White hypothesis [8] that contemporary cofactors are remnants of early RNA-based metabolism.

A catalytic role for the vicinal hydroxyl of the A76 of tRNA^Phe in the ARS-catalyzed deacylation of mischarged-tRNA similar to that of ThrRS was proposed in the PheRS system [26]. A survey of the PheRS editing active site by scanning mutagenesis showed that none of the active side chains contributes significantly to deacylation. Loss of the 3′OH, however, dramatically slowed deacylation, leading to the model that the vicinal hydroxyl assists in catalysis by helping to activate a catalytic water molecule for attack on the aminoacyl ester (Figure 2C). The ablation of the adjacent A76 hydroxyl in those tRNAs that participate in editing in other systems is also been accompanied by significant losses in activity [27,28]. These results suggest that substrate assisted catalysis, particularly that involving adjacent hydroxyl groups of A76, may be a general feature of diverse tRNA reactions. One hallmark of such reactions may be the relative insensitivity of the reaction to amino acids substitutions in active site residues adjacent to the scissile bond.

AA-tRNA as a precursor in synthesis of other tRNAs

In many bacterial genera and in all Archaea, several canonical ARS are missing. The most frequently missing ARS is GlnRS, but asparaginyl- (AsnRS) and, in several Archaeal genera, cysteinyl-tRNA synthetase (CysRS) can also be absent [29]. To compensate for the missing ARSs, these species employ indirect pathways in which one amino acid serves as precursor, i.e. glutamate is converted to glutamine, aspartate is converted to asparagine, and serine is converted to cysteine [29]. In all of these pathways, an initial AA-tRNA^AA intermediate is produced by a mischarging reaction that joins the precursor amino acid to the cognate tRNA with the appropriate anticodon for the eventual product. The intermediate is enzymatically converted into the final amino acid, thereby creating a cognate aminoacyl tRNA that can participate in translation. In addition to the three canonical amino acids listed above, selenocysteine is also incorporated into proteins via an indirect pathway [30]. All indirect pathways share several common features, including a “non-discriminating” ARS that is able to aminoacylate the tRNA associated with the indirect pathway as well as its standard cognate tRNA; chemical activation of one or more small molecule substrates; and the coupling of aminoacyl-tRNA binding to other enzymatic activities distal to the immediate site of tRNA binding, such as the production of ammonia by the glutaminase site in the amidotransferases (described below).

Among the indirect pathways, the best-characterized are those associated with formation of Gln-tRNA^Gln and Asn-tRNA^Asn. Two different families of amidotransferases are responsible for catalyzing the reaction, the heterotrimeric GatCAB [31] and heterodimeric GatDE [32]. In the bacteria, GatCAB can function as both a glutamine amidotransferase (Gln-AdT) and an asparagine amidotransferase (Asn-AdT) [29]. In Archaea, GatDE is always a Gln-AdT and GatCAB is an Asp-AdT. Following a heterologous aminoacylation by a non-discriminating ARS, the basic amidotransferase reaction is identical in both families [29]. In the first step, the terminal carboxylate of the aminoacyl-tRNA (Asp-tRNA^Asn or Glu-tRNA^Gln) is phosphorylated, activating it for the subsequent transfer step. Next, a glutaminase activity cleaves glutamine to liberate ammonia, which is then transported to the active site where the aminoacyl moiety is bound. (This activity can be complemented by free NH₃+ in the reaction.) Finally, the amidotransferase amidates the activated intermediate to form Gln-tRNA^Gln or Asn-tRNA^Asn. Thus, the reaction formally couples a tRNA-dependent kinase activity, localized to GatB or GatE, to a glutaminase activity, localized to GatA or GatD.

To ensure a 1:1 stoichiometry between the equivalents of ammonia generated and equivalents of aminoacylated tRNA formed, the activities of the two catalytic sites must be tightly coordinated. The X-ray structures of GatCAB [33] and GatDE [34] suggest the existence of “tunnels” in both complexes that connect the glutaminase and amidase active sites. In the dimeric histidyl- and threonyl-tRNA synthetases, there is evidence that the presence of tRNA provides allosteric coupling between the amino acid activation reactions carried out in both active site of the dimer, equalizing the rates of reactions which are unequal in the absence of tRNA [35,36]. In both Gln-AdT and Asn-AdT, one hypothesis is that the presence of tRNA similarly serves to couple the activity of the two active sites, thereby preventing the wasteful liberation of free ammonia from glutamine under conditions where Glu-tRNA^Gln and Asp-tRNA^Asn are not available to receive it. This remains to be tested experimentally. Similarly, all indirect pathways generate a mis-charged aminoacyl-tRNA species that would be detrimental to growth, were it allowed to enter protein synthesis. This is likely prevented in two ways. First, elongation factors (and the ribosome’s A-site) have been shown to discriminate against mis-charged tRNA at the level of binding [37,38]. Secondly, the formation of a direct complex between the non-discriminating ARS and a Gln-AdT or Asn-AdT may provide an opportunity for “substrate channeling” between the synthetase and the transamidase, such that the aminoacyl-tRNA intermediate formed by the ARS is prevented from entering the general pool of charged tRNAs for protein synthesis. Evidence for a “transamidosome” is still in the formative stages [39,40]. The transamidosome protects the aspartyl-tRNA^Asn from hydrolysis, by some 1.4-2 fold relative to free tRNA. Dissociation of the final tRNA appears to be the rate-limiting step. One complication associated with complex formation is that GatCAB is required to associate with both Class I and Class II ARS: tRNA complexes, which recognize different faces of their cognate tRNAs. (GatDE is only obliged to form a complex with Class I GluRS, so there is not this issue.)

The entrance of both selenocysteine and cysteine into protein synthesis requires indirect pathways for aminoacyl-tRNA formation, and these share several interesting similarities [29]. A critical difference between the two is that cysteine is coded directly in the genetic code, while selenocysteine employs a complex mechanism involving recoding of the UGA stop codon. (Details of this latter process are outside the scope of this review.) Selenocysteine is formed only by the indirect pathway, whereas cysteine can be formed through both a direct route, requiring the canonical cysteinyl-tRNA synthetase (CysRS), or an indirect route, which requires the transformation of phosphoseryl-tRNA^Cys into Cys-tRNA^Cys [41]. The existence of the indirect pathway was first established in those methanogenic Archaea in which the canonical CysRS is absent [42]. Subsequently, additional genera were identified in which the direct pathway is dispensable [43].

The indirect pathway for cysteinyl-tRNA^Cys formation consists of series of linked reactions [41]. First, tRNA^Cys is aminoacylated with phosphoserine by a dedicated ARS (SepRS) to form phosphoserine-tRNA^Cys. This intermediate is subsequently converted to Cys-tRNA^Cys by the action of SepCysS, a pyroxidal phosphate dependent enzyme [41]. The form in which cysteine is donated is not yet known, but one possible scenario, suggested by the relationship between SepCysS and the cysteine desulfurases CsdB, NifS, and IscS, is that a persulfide intermediate is involved [44] [45]. As described above for the indirect glutamine and asparaginyl-tRNA pathways, the entrance of phosphoseryl-tRNA^Cys into protein synthesis must be blocked. Poor affinity of EF-Tu is clearly one factor [46], and there is some evidence that supports the existence of a complex between SepRS and SepCysS [47]. SepRS is a tetramer, while SepCysS is a dimer [45,48,49]. A significant question that remains is how they interact, and the mechanism by which the aminoacyl-tRNA is handed off from SepRS to SepCysS [45].

The mechanisms for the formation of selenocysteine-tRNA^Sec differ between the prokaryotes and eukaryotes. In bacteria, tRNA^Sec is aminoacylated by SerRS, which essentially behaves like a “non-discriminating” ARS in this context, able to aminoacylate both tRNA^Ser and tRNA^Sec [30]. Selenium is introduced into the system in the form of selenophosphate, which is synthesized by SelD. The key enzyme in the pathway, selenocysteine synthase, employs Ser-tRNA^Sec and selenophosphate as substrates, producing the Sec-tRNA^Sec that functions directly in protein synthesis, albeit by a special recoding pathway that involves a dedicated elongation factor (SelB) and a special structural context around the UGA codon where selenocysteine is incorporated.

The archaeal/eukaryotic pathway for incorporation of selenocysteine resembles the indirect Cys-pathway with respect to the phosphoseryl-tRNA intermediate. A notable parallel between the indirect cysteine and selenocysteine pathways is that the key enzymes that transform the aminoacyl-tRNA intermediates are PLP-dependent enzymes [50,51]. In the eukaryotic pathway, tRNA^Sec is aminoacylated with serine, and then phosphorylated by the O-phosphoseryl-tRNA kinase (PSTK) [52]. The resulting Sep-tRNA^Sec is converted to Sec-tRNA^Sec by the action of SepSecS, which uses selenophosphate [51,53]. Crystal structures of SepSecS have been solved from M. maripaludis [54], mouse [55], and human [56]. All three orthologs are homotetrameric, with their active sites localized at the dimer interface. In the proposed mechanisms for both reactions, the Schiff base between the PLP and conserved lysine is attacked by the -NH₃ group of the amino acid, forming an aldimine. Upon loss of phosphate, this species is converted into an aminoacryl-tRNA intermediate, which then undergoes attack by either sulfur (in the indirect cysteine pathway) or selenophosphate (in the selenocysteine pathway). While the cysteine and selenocysteine mechanisms share similarities with respect to the structure of the intermediates, sulfur and selenium are added across the double bond of the aminoacryl-tRNA intermediate by different chemistries. Sulfur addition may occur via a persulfated SepCysS intermediate, while selenium addition in SepSecS may simply require selenophosphate. In contrast to the aminoacylation and peptidyl transferase reactions, tRNA serves as an essentially passive scaffold in both reactions, and a role for the adjacent 2′OH has not been invoked in either case.

A general principle of indirect pathways for aminoacyl-tRNA synthesis is that the interactions among the enzymatic components of the pathway serve to limit the incidental transfer of aminoacyl-tRNA intermediates into protein synthesis before the required chemical transformation is completed. One potentially general strategy is the formation of biologically/physiologically stable complexes between the non-discriminating ARS and the next enzyme in the transformation pathway. In the case of the AspRS/AdT transamidosome, the tRNA plays an essential role in bringing the components together [39]. After formation of the final cognate product, the AspRS/AdT transamidosome has to dissociate in order for the product to be released. Provided that such complexes actually form, the scaffolding model requires that the CCA end translocate from the aminoacylation active site to the amidotransferase active site. This translocation event may be subject to coordination with the rate of another chemical process (e.g. ammonia formation). Significantly, modeling of the AspRS/AdT transamidosome indicates that the appropriate geometry for these relevant partners is available. Translocation is thus reminiscent of amino acid editing in the class I and class II ARSs, which is generally dependent on structurally distinct aminoacylation and editing sites.

Aminoacyl-tRNAs are donors of amino acids for the interpeptide bridge synthesis of bacterial peptidoglycan

Peptidoglycan is a macromolecular, multilayer complex structure that maintains bacterial cell shape and prevents damage from internal osmotic pressure. The peptidoglycan polymer is composed of alternating N-acetyl glucosamine (NAG) and N-acetylmuramic acid (NAM) units, joined together by β-1→4 linkage. A pentapeptide stem extends from the carboxyl group of NAM, and the stems of different chains are linked together via a peptide bridge. In Staphylococcus aureus, the bridge consists of five glycines, and is initiated at the -amino group of Lys₃ on one peptide stem and terminates at D-Ala₄ on the other peptide stem. The composition and sequence of the “stem” and “bridge” peptides varies between different bacterial species. Peptidoglycan synthesis involves a cytoplasmic as well as membrane phase, consisting of multiple sets of distinct enzymes, substrates, metal ions and coenzymes.

In Staphylococcus aureus, the synthesis of the peptide stem involves the sequential action of the Mur family enzymes. These enzymes use free amino acids as substrates, linking them sequentially at each of the successive carboxyl moieties of NAM [57]. By contrast, AA-tRNAs are used as amino acid donors for synthesis of the peptide interbridge. The involvement of an RNA component was proposed nearly five decades ago, when the RNase-sensitive nature of peptidoglycan synthesis was first observed [58]. Peptide interbridge synthesis is insensitive to antibiotics that typically block protein synthesis, suggesting that bridge synthesis is ribosome-independent [4]. The use of charged tRNAs in cell wall synthesis was the first non-canonical function attributed to AA-tRNAs synthesized by ARSs. The first AA-tRNA to be identified was Gly-tRNA, which could be demonstrated to participate in the addition of the first glycine in the bridge peptide [4]. This process was subsequently shown to be dependent on glycyl-tRNA synthetase, tRNA and ATP [59]. The generality of this approach and its extension to other AA-tRNAs, including Thr-tRNA, Gly-tRNA and Ser-tRNA, was demonstrated several years later [60,61]. Interestingly, at least one of the glycine tRNAs in S. aureus is restricted to peptidolglycan cross bridge synthesis, showing no activity in canonical protein synthesis [62]. The molecular basis of this restriction has not been resolved, and poses an interesting problem with respect to compartmentalization of cellular components.

The sequential synthesis of the pentapeptide bridge is catalyzed by the Fem XAB families of enzymes, which have been studied by a combination of genetic, biochemical and structural approaches. Fem is an abbreviation based on “factors essential for methicillin” resistance, and includes FemX (now fmhB), Fem A, and Fem B. All enzymes in the class are non-ribosomal peptidyl transferases, and are members of the GCN5-related N-acetyltransferase (GNAT) superfamily [63]. FemX catalyzes the addition of the first glycine at the -amino group of lysine, while Fem A adds the second and third glycines. The sequential addition of the last two glycine residues is catalyzed by Fem B [64], which adds these residues by a polarity (C-terminal to N-terminal) that is opposite to that of ribosome-mediated protein synthesis. Glycyl-tRNA is the source of glycine for each of the five glycines [65]. The specificity and essential nature of these enzymes was initially suggested by genetic analysis, which demonstrated that fem mutants showed reduced glycine in their cell walls, reduced cell wall turnover, and greater sensitivity to antibiotics. Although Fem A and Fem B are 39% identical, Fem B cannot substitute for Fem A [66]. Therefore, these enzymes show position-dependent activity during pentapeptide bridge synthesis, even when they are expressed in a heterologous system, and exhibit a form of functional “cross-talk” [67].

The crystal structure of the FemX enzyme from Lactobacillus viridescens (or Weisella viridescens) showed two long channels running across the surface for the binding of the two substrates, UDP-MurNAc-pentapeptide and AA-tRNA [68]. The conformation of the former substrate is in a bent conformation, while the orientation of the latter is currently unknown. Interestingly, structural comparisons with the tRNA binding to seryl-tRNA synthetase revealed that the helical arms of the FemX and FemA proteins contain flexible arms that can participate in tRNA binding [69] [68]. FemX exhibits an ordered sequential mechanism where UDP-MurNAc-pentapeptide binding leads that of Ala-tRNA^Ala, followed by ordered product release with deacylated tRNA^Ala [70]. The specificity of the Fem enzymes with respect to amino acid and tRNA substrates has been studied in detail. FemX_Wv exhibits a preference for Ala-tRNA^Ala over Ser-tRNA^Ser or Gly-tRNA^Gly, but shows no requirement for the post-transcriptional modifications present in mature tRNA^Ala and tRNA^Ser molecules. With respect to different tRNA identity elements, C71, and C72 are essential for FemX_Wv but the G3-U70 pair that serves as a crucial identity element for alanyl-tRNA synthetase is not required [71]. Non-cognate tRNAs are excluded on the basis of both the aminoacyl moiety and the sequence of the tRNA. L-alanine is preferred 110-fold over D-alanine, suggesting relatively weak specificity towards different stereoisomers. The exclusion of serine is due to steric hindrance at the FemX_Wv active site rather than nucleotide sequence of tRNA^Ser. The exclusion of glycyl-tRNA^Gly is mediated by the presence of the C2-G71 base pair of tRNA^Gly, which acts as anti-determinant, rather than the glycyl moiety [72].

The FemXAB enzymes all function as AA-tRNA dependent non-ribosomal peptidyl transferases. The current model for the mechanism of the transferase reaction catalyzed by these enzymes suggests that, in contrast to ribosomal peptidyl transferase, all principal catalytic roles are carried out by enzyme functional groups rather than by those of tRNA [70]. This mechanism imposes a requirement for a general base to de-protonate the -NH₃ group of the lysine nucleophile, and a general acid to protonate the tRNA^Ala leaving group. The pH rate profile of the LvFemX enzyme is bell-shaped, and suggests that groups with pK_a’s 5.5 and 9.3 are important for catalysis. Structural information, as well as comparative sequence analysis, suggests that a conserved Asp is the general base, while a conserved Tyr or Lys serves as general acid. The Fem family enzymes are thus attractive targets for antimicrobial therapy, since these enzymes they possess a unique mechanism and are universally absent from eukaryotes [73].

The role of AA-tRNA in the degradation pathway of proteins

Prokaryotic and eukaryotic cells possess highly specific systems to ensure that protein turnover is carefully regulated. One major route is the N-end rule, where the identity of the N-terminal residue controls the half-life of the targeted protein in vivo [74]. In this pathway, the amino terminus of the protein that undergoes turnover is extended at its N-terminus by a single residue. The resulting modified protein is then processed by cellular proteolytic machinery, which varies between species. The principal mechanism governing protein turnover in eukaryotic cells is polyubiquitination, which involves conjugation of protein targets with ubiquitin, converting them into substrates for the 26S proteosome. Bacterial cells as a general rule do not regulate turnover through ubiquitination. Rather, unstable proteins are readily bound to the AAA+ ClpA complex, which delivers them to the ClpP peptidase with which ClpA is directly associated [74]. Here, we focus on the tRNA-dependent addition of amino acids that promote turnover. These are catalyzed by the L/F and R- transferases, which are found in the prokaryotes and eukaryotes, respectively. The involvement of tRNA in these processes was originally suggested by the RNase sensitivity of these reactions [75] [76] [74]. The following discussion focuses on the L/F transferase, which performs tRNA dependent proteolytic modification in eubacteria.

A monomeric enzyme of 234 residues, the L/F transferase catalyzes the transfer of Leu/Phe to the NH₂-terminal arginine or lysine of the target protein [77]. It possesses two domains, an N-terminal domain that is idiosyncratic to the family, and a C-terminal domain that exhibits structural homology to the FemXAB family (Figure 3) [78]. The L/F transferase exhibits high specificity for the aminoacyl moiety, but much less for the nucleic acid. Specificity for Leu and Phe is nearly absolute, but Met and Trp can be tolerated at substantially reduced activity. No activity is seen with Val, Ile, or Arg. While there is very little specificity for the cognate tRNA, and oligoDNA/tRNA annealed hybrids are accepted as substrates, presentation of the aminoacyl group via a single stranded RNA region is required [79]. In contrast, the requirement for an arginine at the NH₂ terminus of the target peptide is absolute.

Figure 3. A common fold for aminoacyl transfer in two different systems.

A. Structure of FemX from Weisella viridecens, in complex with UDP-MurNAc-pentapeptide, as reported by Biarotte-Sorin et al (ref. 68). The enzyme is depicted in ribbon form in magenta, with the ligand rendered in stick form. B., Structure of the leucyl/phenylalanyl-tRNA protein transferase from Escherichia coli, as reported by Watanabe et al., (ref 80). The adenosine moiety of the rA-Phe substrate analog is rendered in stick form.

The first complex of the E. coli L/F transferase to provide significant insights into substrate recognition was that with puromycin, a chemical analog of the 3′ end of aminoacylated tRNA in which a p-methoxyl phenylalanine moiety is linked to an amino 6-N,N dimethyl adenosine group [78]. These groups are chemically similar to the Phe and 3′-adenosyl group of Phe-tRNA^Phe. Puromycin occupies a cleft in between the two domains of the L/F transferase, with the p-methoxy Phe binding to a hydrophobic pocket rich in Phe and Met residues, and the dimethyl adenosine group being recognized largely through - stacking with conserved Trp residues. Inspection of the binding pocket reveals that while Phe and Leu are readily accommodated, an apparent steric clash exists with amino acid substrates that are branched at the beta carbon. Thus, the 3′ terminus of the aminoacylated tRNA is recognized by the combination of a largely hydrophobic pocket that interacts with the amino acid, and more polar interactions with the 3′ acceptor end [79]. Notably, there are few obvious structural changes in the enzyme that accompany binding. More detailed insight into the reaction was provided by a complex between the L/F transferase, riboadenosyl-Phe (rA-Phe), and a short peptide (RYLGYL) derived from casein which functions as an active acceptor [80]. The former, a catalytically active substrate analog of a 3′-aminoacyl terminus, accurately mimics conformational features of the authentic RNA substrate, including the ester bond and the C3′ endo ribose confirmation, and binds essentially in the same pocket as puromycin. When rA-Phe is soaked into a crystal of the L/F transferase with the acceptor peptide, density for the adenosyl moiety is not observed. However, new density for Phe appears at a location proximal to the N-terminal arginine. The terminal YL of the acceptor peptide is not observed in the complex. Only N-terminal Phe and Arg residues are specifically recognized, but rest of the acceptor peptide recognition is sequence-independent.

Comparison of the rA-Phe and acceptor peptide complexes permits a mechanism for the peptide bond formation reaction to be proposed. Gln188, which is positioned to initially extract a proton as a general base from the -NH₃⁺ nucleophile then donate a proton to the tRNA leaving group, appears to be critical. This mechanism is analogous to a reverse acylation step catalyzed by chymotrypsin [81]. Like the Fem XAB family, L/F transferase catalyzes a reaction that is analogous to peptidyl transferase, but which relies on enzyme functional groups rather than the adjacent hydroxyl of A76 for proton transfer [80]. Interestingly, however, phenylalanylated-tRNA^Phe with a 3′ deoxy group is unable to transfer its aminoacyl moiety to the donor peptide. Despite the presence of a similar protein fold, FemWv and L/F-transferase are characterized by many functional differences, particularly with regard to potential catalytic residues. L/F enzyme has specificity for hydrophobic amino acids, and does not require the acceptor stem [72].

Modification of bacterial membrane lipids by aminoacyl-tRNAs

Aminoacyl-phosphatidyl glycerol synthases (aa-PGs) attach different amino acids to the polar heads of phosphatidylglycerol (PGs), effectively neutralizing the negative charge of the membrane [82]. This mechanism decreases membrane permeability and confers resistance to antibacterial peptides (defensins), and cationic antibiotics (aminoglycosides etc). The presence of RNase inhibited the formation of AA-PGs, suggesting the mechanism is RNA dependent. Through reconstitutions of the minimal system in-vitro, AA-tRNA has been confirmed as the amino acid donor.

The aa-PGs that have been investigated in most detail are MprF (multiple peptide resistance factor) 1 and MprF2, which are responsible for the synthesis of Ala-PG and Lys-PG, respectively, in Clostridium perfringens [83,84]. These proteins show different distributions in different organisms, with MprF1 activity being found in Pseudomonas aeruginosa [85] while MprF2 is found in Staphylococcus aureus [86]. In both cases, the amino acid donors are AA-tRNAs. Heterologous expression of S. aureus MprF leads to synthesis of Lys-PG in E. coli, which otherwise does not synthesize it. Based on the ability of MprF1 to efficiently recognize tRNA^Ala, tRNA^Pro, and a minihelix^Ala, the specificity of MprF1 arises from the direct recognition of the aminoacyl moiety of Ala-tRNA. Similarly, the tRNA^Lys species from Borrelia burgdoferi and humans, which share less than 50% sequence identity, are also well recognized by MprF2. Variation in the discriminator base of tRNA did not affect the activity of these enzymes. Thus, the nucleotide sequence of tRNA, the anticodon, and the shape of the tRNA may not be required for the activity of MprF proteins [84]. The activity of MprF raises the issue of how AA-tRNAs are partitioned between their canonical function of protein synthesis, and lipid modification and cell wall/crosslinking functions. A determination of the K_d’s of Lys-tRNA for EF-Tu and MprF suggests that the two proteins have similar affinities for tRNA under physiological conditions [84]. An interesting hypothesis that has not been experimentally tested is that the different MprF-like proteins sample different AA-tRNAs under different environmental contexts.

The role of aminoacyl-tRNA as an intermediate in antibiotic synthesis

For most of the pathways described above that feature aminoacyl-tRNA as a donor, the product either feeds directly into protein synthesis (indirect pathways for aminoacyl synthesis), into the synthesis of peptides by non-ribosomal means (peptidoglycan cross-linking), or into modifications where the amino acid moiety remains largely intact (lipid modification). Pathways also exist where the donated amino acid moiety undergoes transformation into a compound significantly different from the parent compound. One such example is the synthesis in Streptomyces viridifaciens of valanimycin (Val), an azoxy-based antibiotic that includes the skeletons of the amino acids valine and serine in its structure [87]. Production of the antibiotic is dependent on a gene cluster composed of some 14 different genes [88]. Three of the genes, VlmD, VlmH, and VlmR participate in the conversion of valine into the intermediate isobutylhydroxylamine. The latter compound attacks Ser-tRNA^Ser in a reaction catalyzed by VlmA, forming the intermediate O-(L-seryl)-isobutylhydroxylamine [89,90]. In a series of steps not yet deduced, this intermediate is converted into the final product valinamycin [91]. Notably, the Ser-tRNA^Ser used in this reaction is produced by a dedicated seryl-tRNA synthetase, VlmL [89]. It is unclear why a special tRNA synthetase is needed, as the reaction does not involve a special tRNA^Ser isoacceptor. In view of the essential role of VlmL, and its inability to be substituted by the canonical SerRS, complex formation between VlmL and VlmA may be required in order for the VlmA-catalyzed reaction to proceed. Of note, VlmA is weakly related to MprF (~23% identical, 46% similar) [91]. Both share a common sequence motif (the DUF470 element), which likely contains the MprF active site. Thus, it is possible that VlmA recapitulates the active site domain of MprF, and that the lysyl-phosphatidyl glycerol transferase activity shares some degree of similarity with VlmA’s activity, which transfers L-serine to isobutylhydrolxylamine. Such a similarity would imply a common evolutionary lineage.

It would be surprising if valanimycin were the only example of a “natural product” derived from amino acids, and this is in fact not the case. Cyclodipeptides have recently emerged as another class of secondary metabolites derived from aminoacyl-tRNA. Examples of these compounds include cyclo(L-Phe-L-Pro) and cyclo (L-Phe-trans-4-OH-L-Pro). A number apparently have antibacterial and antifungal activity. Prior to the recent work, the synthesis of cyclic dipeptides was thought to occur exclusively by the action of non-ribosomal peptide synthetases (NRPSs). These enzymes are modular in nature, and a separate adenylation domain selects the cognate amino acid and joins it to ATP. Aminoacyl – tRNAs are not part of this process. In a recently discovered pathway, synthesis of the dipeptide albonoursin (precursor = cycle L-Phe-L-leu) was reported to be catalyzed by AlbC, a 239-residue protein that bears no sequence similarity to other proteins whose structure is known [92]. Notably, AlbC was shown to require amino acid in the form of aminoacyl-tRNA in order to catalyze albonoursin synthesis.

These initial studies featured the addition of exogenous tRNAs to AlbC-containing extracts; the involvement of tRNA was later confirmed by a direct assay in which AlbC was incubated with purified [¹⁴C]-labeled AA-tRNAs. Bioinformatics searches indicated that potential AlbC orthologs exist in other bacteria genera including the Firmicutes (Bacillus, Staphylococcus) Actinobacteria (Corynebacterium, Mycobacterium) and -Proteobacteria. The degree of sequence conservation, however, is relatively low, with 19–27% identity and 37–45% similarity. AlbC orthologs can be identified by the consensus sequences Hx(lVI][LVI]G[LVI]S and Y[LVI]xxExP. Notably, these are not found in any other proteins in the database. When expressed in E. coli, these proteins also produced compounds that could be identified as cyclic dipeptides, and the reactions employed AA-tRNA as substrate. For a number of these dipeptide synthases, the dipeptide is an intermediate and not a final product. For example, the cFL peptide formed by AlbC is subsequently oxidized by a cyclodipeptide oxidase to generate the authentic albonursin. In other systems, the localization of cytochrome P450 enzymes near the cyclodipeptide synthase (CDPS) enzymes suggests that redox-mediated modification of the cyclodipeptide intermediate may a general feature of the modification of these compounds. As yet, the specificity determinants of AlbC and its homologs with respect to tRNA are unknown. In view of their utilization of AA-tRNAs as substrates, these biosynthetic pathways would be expected to compete with normal protein synthesis. Accordingly, it is not unexpected that synthesis of their products would be maximal during log phase, which is seen for albonoursin in S. noursei.

The role of AA-tRNA in the synthesis of tetrapyrroles

The biosynthesis of hemes, chlorophylls, and bilins in plants, algae, and some bacteria is regulated by the availability of the first metabolic intermediate, δ-aminolevulinic acid (ALA). Synthesis of ALA begins with the formation of Glu-tRNA^Glu by glutamyl-tRNA synthetase (GluRS), reduction of glutamate to glutamate-semialdehyde (GSA) by Glu-tRNA reductase (GluTR), and then finally the transamidation of GSA to ALA by GSA aminomutase [93]. The conversion of glutamate to ALA was decreased in the presence of ribonuclease, which suggested the involvement of an RNA component [94]; this was later confirmed to be tRNA^Glu [95]. The specificity for tRNA^Glu is absolute, as tRNAs specific for other amino acids are not active in the reaction [96].

Subsequent work raised the possibility of separate tRNA^Glu species for translation and heme synthesis [97] [98]. Two GluRSs are present in Acidithiobacillus ferrooxidans, the first a discriminating GluRS-1 that glutamylates both tRNA^Glu isoacceptors and tRNA^Gln, and the second a non-discriminating GluRS that preferentially glutamylates tRNA^Gln. GluRS-1 appears to be the enzyme most closely associated with heme synthesis. When cells were grown under high heme requirements (Fe²⁺), the level of GluRS-1 was increased. However, under the conditions of excess heme synthesis (Fe²⁺/ALA), GluRS-1 activity was drastically lowered without change in the protein level. Thus, increased intracellular heme concentrations corresponded to the reduced GluRS-1 activity, suggesting its regulation by heme. GluRS-1 was also found to be dispensable for protein synthesis, suggesting that GluRS-2 is sufficient to generate Glu-tRNA [93].

Substrate requirements and the mechanism of glutamate reduction by GluTR have also been examined. Notably, the activity of GluTR on in vitro transcripts is comparable to its activity on tRNA produced in vivo, suggesting that the post-transcriptional modifications typically found in Glu-tRNA are not obligatory. GluTR also possesses a conserved cysteine (Cys50) whose chemical modification or substitution via mutagenesis significantly reduces GluTR activity. It has been hypothesized that Cys50 attacks the α-carbon of Glu-tRNA glutamate, forming a thioester intermediate (acyl-enzyme), and releasing tRNA^Glu. Reduction of the thioester intermediate to GSA is subsequently accomplished using NADPH as cofactor, facilitated by an active site histidine. Alternatively, in the absence of NADPH, an esterase activity can release free glutamate from the acyl-enzyme [99,100]. While sequence variations in the acceptor arm and anticodon stem/loop of tRNA^Glu apparently have little effect on the activity of GluTR, variants harboring mutations in the tertiary core (D-stem/loop, variable loop, and T-stem/loop) exhibit significant differences in their recognition by GluTR [101]. Thus, Glu-tRNA presents a different set of identity determinants for the two different enzymes, GluRS-1 and GluTR.

Summary and Perspectives

The list of reactions in a variety of cell types where AA-tRNA serves as the amino acid donor for functions other than protein synthesis continues to grow. As reviewed above, these include reactions where the amino acid is used essentially in unmodified form, as in the case of peptidoglycan cross-linking, lipid modification, and N-end rule conjugation. There are also examples where the attached amino acid undergoes significant chemical transformation, as seen in the indirect pathways for aminoacyl-tRNA formation, for synthesis of antibiotics, cyclic dipeptides, and tetrapyrrole synthesis. In these non-proteogenic pathways, the data currently available suggest that the tRNA principally serves as a molecular support, facilitating the transfer of the amino acid from one active site (i.e. the aminoacylation active site) to the other, where additional chemistry can occur. In contrast to the aminoacylation and peptidyl transferase reaction, where the adjacent hydroxyl groups on A76 play an active role in chemistry, direct participation of tRNA functional groups in the chemistry of these non-proteogenic reactions has not yet been reported. One might speculate that this reflects the fact that the aminoacylation and peptidyl transferase reactions evolved contemporaneously in the context of the RNA world, whereas the other reactions evolved in a context where proteinaceous enzymes predominated. A critical problem that all non-proteogenic reactions that employ AA-tRNA must address is the diversion of aminoacylated intermediates into protein synthesis. Current work argues that the selectivity by elongation factors and the formation of complexes with enzymes that follow in the reaction sequence provide at least two different mechanisms to achieve sequestration away from ribosomes. Whether additional tRNAs remain to be discovered that participate only in non-proteogenic functions (like the Gly-tRNA from S. aureus) is an interesting question that further whole genome sequencing projects should enlighten. A final question concerns the extension of the scaffolding principle to RNAs distinct from tRNA. A recent chemical screen for biological small molecule-RNA conjugates indicates in E. coli and Streptomyces venezuelae led to the detection of a number of such compounds, including 3′-dephospho-CoA and other thioester derivatives of RNA [102]. The finding of such compounds, and the depth of undiscovered biology they imply, suggests that further functions of tRNA outside of protein synthesis may yet be discovered.