The structure of an E. coli tRNAfMet A1-U72 variant shows an unusual conformation of the A1-U72 base pair.

Translation initiation in eukaryotes and archaea involves a methionylated initiator tRNA delivered to the ribosome in a ternary complex with e/aIF2 and GTP. Eukaryotic and archaeal initiator tRNAs contain a highly conserved A1-U72 base pair at the top of the acceptor stem. The importance of this base pair to discriminate initiator tRNAs from elongator tRNAs has been established previously using genetics and biochemistry. However, no structural data illustrating how the A1-U72 base pair participates in the accurate selection of the initiator tRNAs by the translation initiation systems are available. Here, we describe the crystal structure of a mutant E. coli initiator tRNAfMetA1-U72, aminoacylated with methionine, in which the C1:A72 mismatch at the end of the tRNA acceptor stem has been changed to an A1-U72 base pair. Sequence alignments show that the mutant E. coli tRNA is a good mimic of archaeal initiator tRNAs. The crystal structure, determined at 2.8 Å resolution, shows that the A1-U72 pair adopts an unusual arrangement. A1 is in a syn conformation and forms a single H-bond interaction with U72 This interaction requires protonation of the N1 atom of A1 Moreover, the 5' phosphoryl group folds back into the major groove of the acceptor stem and interacts with the N7 atom of G2 A possible role of this unusual geometry of the A1-U72 pair in the recognition of the initiator tRNA by its partners during eukaryotic and archaeal translation initiation is discussed.

INTRODUCTION

The initiator tRNAiMet (called tRNAfMet in bacteria and organelles) plays a central role in decoding mRNA into proteins. Its recruitment by the translation initiation machinery allows setting of the reading frame of the mRNA by decoding the correct initiation codon. Selection of the initiator tRNA by the translation initiation machinery and its rejection by the elongation system involve some nucleotidic sequence features that distinguish it from elongator tRNAs, in particular elongator Met-tRNAs (Marck and Grosjean 2002; Kolitz and Lorsch 2010). Among them, whatever the domain of life, tRNAiMet harbors a weak base pair or unpaired bases at position 1–72 whereas most elongator tRNAs possess a strong G–C base pair at this position. In bacteria, a conserved mismatch at position 1:72 (most frequently C1:A72) is involved in the recognition of Met-tRNAfMet by methionyl-tRNAfMet transformylase, the enzyme responsible for the formylation of the esterified methionine. However, biochemical studies have shown that a weak base pair was tolerated by the enzyme (Guillon et al. 1992). Indeed, the specific interaction of methionyl-tRNAfMet transformylase with a methionylated initiator tRNA is based on the ability of the enzyme to open the 1–72 pair (Schmitt et al. 1998). Moreover, the absence of strong base-pairing at position 1–72 of the initiator tRNA was shown to be important for the rejection of the initiator tRNA by the elongation factor EF-Tu (Lee et al. 1991; Guillon et al. 1992; Rudinger et al. 1996) as well as for the protection of formyl-Met-tRNAfMet against hydrolysis by peptidyl-tRNA hydrolase (Schulman and Pelka 1975; Dutka et al. 1993).

In eukaryotes and in archaea, Met-tRNAiMet binds to the small ribosomal subunit as a complex with the heterotrimeric factor e/aIF2 in its GTP-bound form. When pairing between AUG codon and CAU anticodon is correct, e/aIF2 is released from the ribosome in a GDP-bound state. Therefore, the selection of the start codon is achieved through the control of the nucleotide state of e/aIF2. Eukaryotic and archaeal initiator tRNAs almost systematically carry an A1–U72 base pair (one exception in Schizosaccharomyces pombe tRNAiMet carrying a Ψ1–A72 base pair ([Keith et al. 1993; Marck and Grosjean 2002]). This base pair has been shown crucial for the initiation process in various studies (von Pawel-Rammingen et al. 1992; Farruggio et al. 1996; Kapp and Lorsch 2004; Yatime et al. 2004; Kapp et al. 2006). By analogy with the interaction of methionyl-tRNAfMet transformylase with Met-tRNAfMet in bacteria, it was proposed that the opening of the 1–72 pair was important to ensure correct binding of the tRNAiMet to e/aIF2 (Yatime et al. 2004). However, by studying Met-tRNAiMet affinity for S. cerevisiae eIF2–GTP or eIF2–GDP, it was proposed that the identity and/or geometry of the 1–72 base pair was important for positioning the methionine moiety of the initiator tRNA in its recognition pocket and that an easily disrupted base pair may not be sufficient to ensure full binding specificity (Kapp and Lorsch 2004). Finally, beyond its function in e/aIF2 recognition, the A1–U72 base pair was also shown to influence the binding of the ternary complex to the 40S subunit and later steps in translation initiation (Kapp et al. 2006). The available structure of the archaeal ternary complex (TC), aIF2:GTP:tRNA, revealed the positioning of the initiator tRNA with respect to aIF2. However, the resolution (5 Å) was too low to reveal the details of atomic interactions involving the A1–U72 pair (Schmitt et al. 2012).

Structural studies of e/aIF2:tRNA, and more generally of eukaryotic and archaeal translation initiation complexes, would benefit from the availability of large amounts of pure initiator tRNA carrying an A1–U72 base pair. Toward this aim, we are using an E. coli tRNAfMetA1–U72 variant as a good mimic of archaeal initiator tRNAs. To obtain pure fractions of such a tRNA, we designed an E. coli strain in which the four genes encoding initiator tRNA, metZWV and metY are inactivated and complemented by a plasmid-borne gene coding for the E. coli tRNAfMetA1–U72 variant as the sole source of initiator tRNA. We purified from this new strain large amounts of tRNAfMetA1–U72 devoid of wild-type tRNAfMet. This mutant tRNA, aminoacylated with methionine, was crystallized and the structure was solved and refined to 2.8 Å resolution. Interestingly, we find that A1 adopts an unexpected syn conformation and forms a single H-bond interaction with U72. This interaction requires protonation of the N1 atom of A1. Moreover, the 5′ phosphoryl group folds back into the major groove of the acceptor stem and interacts with the N7 atom of G2. A possible role of the peculiar geometry of the A1–U72 base pair in the selection of the initiator tRNA by eukaryotic and archaeal translation initiation systems is discussed.

RESULTS

E. coli tRNAfMetA1–U72 resembles archaeal tRNAiMet

Sequences of archaeal tRNAiMet were extracted from the tRNAscan-SE database (Lowe and Eddy 1997). A sequence logo (Schneider and Stephens 1990; Crooks et al. 2004) calculated from a multiple alignment of the 170 extracted DNA sequences shows a high level of conservation among archaeal initiator tRNA sequences (Fig. 1). For instance, S. solfataricus tRNAiMet deviates from the consensus by only 13 positions. Interestingly, the sequence of a mutated E. coli Met-tRNAfMet carrying A1–U72 instead of C1:A72 appears to be close to the archaeal consensus, with 13 deviations only (83% identity). In particular, the sequences of the acceptor stems are identical. Therefore, the E. coli tRNAfMetA1–U72 variant appears to be a good mimic of an archaeal initiator tRNAiMet.

FIGURE 1.

Comparison of archaeal initiator tRNAs with the E. coli tRNAfMet A1–U72 variant. (A) Sequences of archaeal tRNAiMet were extracted from the tRNAscan-SE database (Lowe and Eddy 1997). A sequence logo calculated from a multiple alignment of the 170 DNA sequences obtained is shown. The DNA sequence of the S. solfataricus tRNAiMet is shown above the logo and that of E. coli Met-tRNAfMetA1–U72 variant is shown below. Nucleotidic bases differing from the consensus sequence are in red. (B) Cloverleaf representations of initiator Met-tRNAsMet. (Left) S. solfataricus Met-tRNAiMet, (right) E. coli Met-tRNAfMetA1–U72 variant. Note that wild-type tRNAfMet carries C1:A72. Nucleotides differing from the archaeal initiator tRNA consensus are in red. Post-transcriptional modifications of S. solfataricus Met-tRNAiMet are not known.

Construction of a strain for tRNAfMetA1–U72 expression

A strain using pBSTNAV-tRNAfMetA1–U72 as its sole source of initiator tRNA was constructed in three steps: knockdown of metZWV, transformation by pBSTNAV-tRNAfMetA1–U72 and knockdown of metY. First, the metZWV genes were deleted in E. coli MG1655 strain (Guyer et al. 1981). For this purpose, we have used the modified λRed/FLP procedure (Datsenko and Wanner 2000; Chen et al. 2015). The whole knockout procedure involved λRed recombinase-mediated replacement of the metZWV region by a kan marker flanked by two FRT sequences, one of which is mutated (Senecoff et al. 1988). The kan marker was then eliminated through FLP recombinase mediated reaction. The advantage of this procedure is that the mutant FRT sequence remaining on the chromosome at the knocked-out locus cannot react with any other FRT during subsequent modifications (Chen et al. 2015). Because of this and of the removal of the kan marker, the whole procedure can be repeated for knocking-out other genes. Recombination events were verified using PCR (Fig. 2A). The resulting strain PalΔmetZWV could grow because intiator tRNA was produced from the intact metY locus, as previously observed (Kenri et al. 1991; Kapoor et al. 2011; Samhita et al. 2012). However, it displayed a slow growth phenotype, yielding small colonies, in agreement with previous studies of a ΔmetZWV::kan strain (Fig. 2B; Samhita et al. 2014). In contrast, when we deleted metY using the same procedure, normal growth of the obtained PalΔmetY strain was observed.

FIGURE 2.

Inactivation of metZWV leads to a slow growth phenotype. (A) Verification by PCR of gene inactivation. The metY and metZ loci of each strain were probed by PCR analysis using whole cells as template DNA sources. Primer couples for probing the metY locus (VerY, lanes 1,2,3) or the metZ locus (VerZ, lanes 4,5,6). Shown are the analyses of PalΔmetZWVΔmetY:kanR-pBSTNAV2-tRNAfMetA1–U72 (lanes 1,4), PalΔmetZWV (lanes 2,5), and PalΔmetY (lanes 3,6). Expected sizes (after knockdown) are 1411 base pairs (bp) (lane 1), 406 bp (lane 2), 398 bp (lane 3), 364 bp (lanes 4,5), and 591 bp (lane 6). Lane M was loaded with a molecular weight marker (MWD-1, Nippongenetics), relevant band sizes of which are indicated on the left. (B) Strain PalΔmetZWV transformed with PBSTNAV2 displays a slow growth phenotype (small colonies, 1). The phenotype can be reversed upon expression of either tRNAfMet (2), tRNAfMetA1–U72 (3), or S. solfataricus tRNAiMet (4).

In order to knock out the metY gene in PalΔmetZWV, cells were transformed with the λRed-producing plasmid pKD46 (Datsenko and Wanner 2000) and then with pBSTNAV-tRNAfMetA1–U72. Introduction of the latter plasmid restored standard colony size, suggesting that the overproduced tRNAfMetA1–U72 was functional in translation initiation (Fig. 2B). Interestingly, standard colony sizes were also restored with pBSTNAV-S. solfataricus-tRNAiMetA1–U72 (Fig. 2B). Accordingly, the metY locus was successfully knocked-out through its λRed-mediated replacement by the kan gene. The obtained E. coli strain, PalΔmetZWVΔmetY::kan:pBSTNAV-tRNAfMetA1–U72, has tRNAfMetA1–U72 as its sole source of initiator tRNA.

Overall structure of tRNAfMetA1–U72 and crystal packing analysis

The strain described above was then used to overexpress and purify Met-tRNAfMetA1–U72 for use in crystallization trials. Crystals belonging to space group C2221 were obtained using 20% PEG 6000 and 2 M NaCl as precipitants. The structure was solved by molecular replacement and refined to 2.8 Å resolution. All bases are visible except C75 and the methionylated A76.

An analysis using DSSR (Lu et al. 2015) identifies all the secondary structure elements characteristic of the classical cloverleaf secondary structure as well as usual tertiary interactions that stabilize the L-shaped tertiary fold of the molecule. A global superimposition of the present structure to that of wild-type E. coli tRNAfMet (PDB entry 3CW5) (Barraud et al. 2008) shows modification in the angle made by the two branches of the tRNA, i.e., the anticodon helix (from P56 to P35) and the acceptor helix (from P56 to P73) as defined (Giege 2008). The angle varies from 71° for E. coli tRNAfMet to 81° for tRNAfMetA1–U72. This difference reflects tRNA plasticity. For instance, the angle of the two tRNA branches varies from 110° to 65° in free and AspRS-bound tRNAAsp (Ruff et al. 1991). Crystals containing tRNAfMetA1–U72 have been obtained in space group C2221, whereas previous crystals of wild-type E. coli tRNAfMet had been obtained in space group P6422. In the present crystals, three regions of the tRNA are involved in packing interactions (Fig. 3). One contact point involves the anticodon loops of two tRNA molecules. Within this region, anticodon–anticodon base-pairing between the C34A35U36A38 nucleotides of one molecule and the corresponding nucleotides in a symmetry-related molecule is observed (Fig. 3A). Two Watson–Crick A–U pairs and two noncanonical C–A hydrogen bonds are holding this interaction. Notably, the same packing of anticodon loops was observed in P6422 crystals of wild-type E. coli tRNAfMet (Barraud et al. 2008).

FIGURE 3.

Crystal packing interactions. The central view shows arrangement of the tRNA molecules within the C2221 crystals with the three contact points boxed and detailed in the close-up views. The reference molecule of the asymmetric unit is in gray and the neighboring molecules are in orange. In the close-up views, one molecule is shown with bases in cyan and the neighboring molecule is in orange. Intramolecular interactions are drawn with red dashed lines and intermolecular interactions are drawn with black dashed lines. Note that the close-up views are not drawn in the same orientations as the central view. (A) Crystal packing at the anticodon loop. (B) Crystal packing at the elbow. (C) Crystal packing at the top of the acceptor stem. The left view shows A73 of a tRNA molecule making an A-minor interaction with the C51–G63 pair of a neighboring molecule. Distances are given in Å. The right view shows interaction of the acceptor stem of a tRNA molecule with the T-stem of a crystallographic neighbor. rT stands for ribothymidine.

A second contact point, specific to C2221 crystals, involves the universal G19–C56 base pair located at the elbow of the tRNA. The G19–C56 base pair presents a flat hydrophobic surface stacked on the corresponding base pair of a neighboring tRNA molecule (Fig. 3B). This surface at the elbow is exploited for tRNA recognition by many partners including the ribosome (Korostelev et al. 2006; Zhang and Ferre-D'Amare 2013, 2016).

Finally, a third contact point, also specific to C2221 crystals, exists between the top of the acceptor stem of one molecule and the minor groove of the T-stem of a neighbor. In particular, A73 makes an A-minor type I interaction (Nissen et al. 2001) with the G51–C63 receptor base pair (Fig. 3C) as observed in a variant of group I self-splicing intron (Battle and Doudna 2002). A second interaction involves the NH2 group of A1 and the 2′OH group of a neighboring G64. Also, one phosphate oxygen of C74 is bound to the NH2 group of G52.

Conformation of the anticodon stem–loop

The structure of the anticodon stem and loop is identical to that described for E. coli tRNAfMet (Barraud et al. 2008). A37 protrudes outside of the anticodon loop and A38 is directly stacked onto U36. An unusual Cm32–A38 base pair is formed thereby extending the anticodon stem and restricting the anticodon loop to 5 nt. The Cm32–A38 base pair is stabilized by a polar interaction between the N1 atom of A38 and the O2 atom of Cm32 and by stacking onto G31–C39. Finally, A37 is positioned within the major groove of the anticodon stem and forms a base triple with G29–C41 (Fig. 4), as previously observed (Barraud et al. 2008). Superimposition of tRNAfMet and tRNAfMetA1–U72 structures shows that bending of the anticodon loops slightly differ. However, despite this movement, the network of interactions in the anticodon loop is conserved in the two tRNA structures.

FIGURE 4.

Conformation of the anticodon loop of the E. coli tRNAfMetA1–U72 variant. (A) Global view of the anticodon loop of the E. coli tRNAfMetA1−U72. The figure shows A37 outside of the anticodon loop forming an A-minor interaction with the G29–C41 base pair. A38 is stacked between U36 and C39 and forms an unusual base pair with Cm32. (B) A37 (G29–C41) base triple. A simulated annealing mFo–DFc omit map contoured at 3.5 σ is shown for A37. Distances are given in Å. They are consistent with hydrogen bonding within standard error on coordinates (Table 1).

TABLE 1.

Data collection and refinement statistics for tRNAfMetA1–U72 structure determination

An unusual A1–U72 pair

Characteristic of eukaryotic and archaeal initiator tRNAs, the A1–U72 pair at the top of the acceptor stem, shows a peculiar conformation (Fig. 5). Watson–Crick (WC) interactions between A1 and U72 are unexpectedly not observed. When compared to a standard G1–C72 WC base pair, as in tRNAPhe (PDB entry 1EHZ [Shi and Moore 2000]), a rotation of the A1 base around bond P–O3′ is observed. Moreover, instead of adopting the usual anti conformation, the adenine ring rotates around the glycosidic bond and adopts a syn conformation (χ = −14.9°) (Fig. 5A,B). The 5′ phosphoryl group folds back into the major groove of the acceptor stem and interacts with the N7 atom of G2 (Fig. 5A,C). On the other hand, base U72 remains stacked onto C71 as in a canonical tRNAPhe acceptor stem. As a consequence, the opening parameter (Lu et al. 2015) of the A1–U72 base pair becomes unusually high (153.42°). The NH2 group of A1 points toward the minor groove of the acceptor helix. An interaction between the N1 of A1 and the O2 of U72 (d = 3.0 Å) is observed which requires protonation of the N1 atom of A1. In addition, A73 is stacked onto A1 instead of being stacked on the first base pair of the acceptor stem as in tRNAPhe (Fig. 5D). Finally, a water molecule bound to N7 of A73 located just above U72 could make an O–π or an H–π interaction with the uridine ring (Sarkhel et al. 2003). It is interesting to note that such water–nucleobase interactions would make up for the absence of base stacking on U72 (Fig. 5D).

FIGURE 5.

Conformation of the top of the acceptor stem of the E. coli tRNAfMetA1–U72 variant. (A) Views in two orientations of the 2.8 Å resolution 2mFo–DFc map of E. coli tRNAfMetA1–U72 contoured at 2.5 SDs in the region of the A1–U72 base pair. (B) Stereo view of the A1–U72 base pair. Simulated annealing mFo–DFc (green) and 2mFo–DFc (blue) omit maps contoured at 5 and 2.5 σ, respectively, are shown. (C) Stereo view showing interaction of the 5′PO4 group with N7 of G2. Represented maps are the same as in panel B. The view shows that the 5′PO4 lies in the G2–C71 plane. (D) The views in two different orientations compare the geometry of a standard G1–C72 base pair as observed in tRNAPhe (yellow sticks, PDB entry 1EHZ [Shi and Moore 2000]) with that of the A1–U72 base pair of E. coli tRNAfMetA1–U72 (cyan sticks). Comparison results from the superimposition of the acceptor stems of the two tRNAs. The figures were drawn with Pymol (Schrodinger 2010).

Molecular dynamics simulation of a protonated A1

In order to further study a possible relevance of the protonated state of A1, molecular simulation techniques were used. Protonated N1 of adenine can be observed either in a neutral tautomer or in a positively charged form of the base. In a first step, we have evaluated a possible relevance of tautomeric forms of adenine using quantum chemistry methods (Warshel and Naray-Szabo 1997). The calculated relative energies of the adenine tautomers in solvent were 12.0 and 10.0 kcal/mol for the N6 proton being cis and trans relative to N1, respectively. These energies are high and do not depend on the medium pH. Therefore, neutral tautomers are not likely to be relevant. In a second step, we analyzed the stability of the positively charged form of adenine. The experimentally determined pKa of N1 of adenine is 4.2 (Kampf et al. 2002). Therefore, the relative energy of the positively charged N1 protonated form of adenine is only 3.1 kcal/mol higher than the energy of the neutral form at a physiological pH of 6.5 suggesting that this form may exist in RNA complexes stabilized by favorable electrostatic interactions with negative phosphates. Therefore, in a second step, we used molecular dynamics simulations (Sham et al. 1997; Aleksandrov et al. 2007) to compute protonation free energies of adenine A1 of tRNAfMetA1–U72 in solution and in the crystal. Molecular dynamics was used to average values over the multiple conformations observed during the simulations and to model protonation states not observed in the crystal. The structure of tRNAfMetA1–U72 with the A1–U72 pair observed in the crystal is stable during molecular dynamics simulation with A1 in the N1 protonated form (Fig. 6A). This suggests that this conformation can exist in solution. In MD simulations with A1 in the neutral form the A1–U72 base pair adopts a reverse Watson–Crick conformation (Fig. 6B). We calculated protonation free energies using these MD simulations. For tRNAfMetA1–U72 in solution, the protonation free energy for A1 (2.2 kcal/mol) is lowered by 0.9 kcal/mol as compared to what is computed for free adenine. This corresponds to a pKa value of 4.9 for the N1 of A1 (see relationships between ΔΔG and pKa in Materials and Methods). If the crystal environment is taken into account, the protonation of A1 of tRNAfMetA1–U72 is favored by an additional 0.9 kcal/mol, leading to a protonation free energy of 1.3 kcal/mol that corresponds to a pKa of 5.6 for the N1 atom of A1. Notably, further increase of the pKa value may occur through interaction of an anion in the vicinity of A1 that would not have been detected at the resolution reached in the crystal structure.

FIGURE 6.

Final structures of the A1–U72 base pair after 5 nsec of MD simulations of tRNAfMetA1–U72 in the crystal environment with A1 in the N1 protonated form (A) and with A1 in the neutral form (B).

DISCUSSION

The present work describes the construction of an E. coli strain able to grow in a genetic context where a variant of the initiator tRNA is used as the sole source of initiator tRNA. Using such a strain we observed that E. coli growth is supported by overexpression of an initiator tRNA carrying an A1–U72 base pair instead of the natural C1:A72. This result is consistent with previous work showing that the absence of strong base-pairing at position 1–72 is sufficient to ensure efficient formylation of Met-tRNAfMet, an important step for initiation of translation (Guillon et al. 1993; Mangroo and Rajbhandary 1995). Because productive interaction with the formylase requires opening of the 1–72 bp, this shows the propensity of the A1–U72 base pair to reach an open conformation. Accordingly, we observed that E. coli growth is supported by overexpression of the archaeal initiator tRNAiMet from S. solfataricus, a tRNA having a sequence very close to that of an E. coli tRNAfMetA1–U72 variant (Fig. 1B). Hence, the λRed/FLP procedure applied on PalΔmetZWV in the presence of an overproducing initiator tRNA plasmid gives a useful tool to test for initiator tRNA complementation and to allow for overexpression and efficient purification of effective initiator tRNA variants in an E. coli context. This method represents a useful alternative to the P1 transduction strategy (Samhita et al. 2012) in particular in cases where it is desirable to limit the number of antibiotic resistance genes.

Using pure fractions of the E. coli tRNAfMetA1–U72 variant, we obtained crystals and solved the 3D structure of the tRNA. The overall structure presents all the tertiary interactions responsible for the L-shaped folding of the tRNA as previously described (Barraud et al. 2008). Two regions exhibit specific structural features: the anticodon loop and the top of the acceptor stem. Within the anticodon loop, A37 is extruded and A38 makes a peculiar base pair with Cm32. Notably, the unusual conformation of the anticodon region was already observed in wild-type tRNAfMet crystals obtained in a different space group (Barraud et al. 2008). In this view, it is notable that yeast initiator tRNA also shows an extruded base 37 (Basavappa and Sigler 1991). However, in this case, the presence of an N6-threonylcarbamoyladenosine modification in A37 might explain the absence of a base triple interaction between t6A37 and the G29–C41 base pair. Together, these structures reflect the propensity of the anticodon loop of initiator tRNAs to adopt a “37-unstacked” conformation. This observation reinforces the idea that the switch between a “37-unstacked” conformation observed in free-initiator tRNAs (present work; Basavappa and Sigler 1991; Barraud et al. 2008) as compared to the “37-stacked” conformation observed in tRNAfMet bound to the P site (Selmer et al. 2006) might be important for initiator tRNA selection during the translation initiation process.

At the top of the acceptor stem, tRNAfMetA1–U72 shows an unusual conformation of the A1–U72 pair. Bases A1 and U72 are systematically encountered in archaeal initiator tRNAs and only one exception is known in eukaryotic initiator tRNAs (Marck and Grosjean 2002). Recent biochemical results have suggested that the identity of the A1–U72 base pair was important to ensure the correct selection of the initiator tRNA during translation initiation and that a weak base pair at position 1:72 was not sufficient (for review, see Kolitz and Lorsch 2010). In eukaryotes and in archaea, the initiator tRNA is delivered to the ribosome as a ternary complex, e/aIF2:GTP:tRNA. In the TC, it was suggested that the geometry of the base pair at 1–72 was important for the correct positioning of the methionine moiety in its binding pocket (Kapp and Lorsch 2004). Furthermore, it was also proposed that the A1–U72 base pair plays a role in the binding of the TC within the translation initiation complex and directly affects translation efficiency (Kapp et al. 2006).

Notably, interaction of the N1 atom of A1 with the O2 atom of U72 (d = 3.0 Å) assumes that N1 of A1 is protonated. In the present study, crystals have been obtained at pH 6.3, a pH where the N1 of adenosine is normally not protonated (Saenger 1984). In order to evaluate the relevance of the peculiar conformation of the A1–U72 pair, we performed a computational study of the protonation state of A1 using molecular simulation techniques. As expected (Fonseca Guerra et al. 2006), the results of the calculations make it unlikely that the occurrence of adenine tautomers play a significant role in the observed conformation. However, formation of a positively charged A1 through protonation of the N1 atom has a much lower energetic cost. Several conclusions may be drawn from the calculations. First, the energetic penalty for protonation of N1 of A1 is not very high (2.2 kcal/mol in solution or 1.3 kcal/mol in the crystal). Second, crystal packing does not seem to have a predominant influence on the decrease of the free energy necessary for protonation because it contributes 0.9 kcal/mol only. Finally, the structure of tRNAfMetA1–U72 observed here is stable during molecular dynamics simulations with protonated A1. Therefore, the unusual conformation observed here must exist at least transiently in solution. Accordingly, the crystals are able to nucleate and grow. Overall, these calculations strongly suggest that interaction of the A1–U72 base pair with a proteic partner could easily stabilize a structure of the tRNAfMet with the peculiar A1–U72 arrangement observed in the crystal including protonation of N1 of A1. pKa value shifting toward neutrality explained by electrostatic stabilization has been observed, for instance, in specific bases of ribozymes (Ravindranathan et al. 2000; Wilcox et al. 2011; Liberman et al. 2012) or in tRNA3Lys anticodon stem–loop (Durant and Davis 1999).

One obvious candidate for the stabilization of the unusual A1–U72 base pair is e/aIF2-GTP within the TC. In this view, the peculiar A1–U72 conformation might explain its role in the positioning of the methionine moiety of the initiator tRNA in its recognition pocket within the γ subunit of eIF2-GTP (Kapp and Lorsch 2004). Notably, the crystal structure of nonaminoacylated S. cerevisiae initiator tRNA, obtained in the P6422 space group, shows a standard WC A1–U72 base pair (Basavappa and Sigler 1991). In the present study, the use of a methionylated tRNA may have influenced the conformation of the A1–U72 base pair. Notably, attempts to obtain crystals of nonaminoacylated tRNAfMetA1–U72 in the same conditions were unsuccessful. However, a possible influence of the methionyl moiety remains unlikely because the C75 and met-A76 bases are disordered in the crystal. Nevertheless, the fact that the A1–U72 base pair can be easily opened suggests that the closing base pair of the initiator tRNA may switch between different conformations depending on its partners at the various stages of its functional cycle. Unfortunately, the available 3D structure of archaeal TC aIF2:GDPNP:Met-tRNAfMet was obtained at low resolution (5Å) thus preventing observation of the fine geometrical requirements for optimal tRNA binding (Schmitt et al. 2012). Therefore, further structural studies of the TC are required to reveal the conformation of the A1–U72 base pair of the initiator tRNA bound to e/aIF2. Finally, recent Cryo-EM studies of archaeal translation initiation complexes (Coureux et al. 2016) show transient deformation of the TC structure at some stages of the translation initiation process. Resolutions of the cryo-EM structures were not sufficient to observe the status of the A1–U72 base pair. It is, however, conceivable that adjustments of the A1–U72 base pair geometry participate in the control of the dynamics of the initiation complex during start codon recognition. To conclude, the present work gives a new example of an RNA structure with rare ionic forms of bases that could be important for RNA function (Wilcox et al. 2011; Liberman et al. 2012; Rozov et al. 2016).

MATERIALS AND METHODS

Inactivation of genes in the E. coli chromosome

Gene deletions were performed using the λRed/FLP procedure (Datsenko and Wanner 2000) as modified (Chen et al. 2015). E. coli K12 MG1655 strain (F− λ− ilvG− rfb-50 rph-1) was transformed with the pKD46 plasmid, carrying a thermosensitive replicon and the λ Red operon under the control of an arabinose inducible promoter. Cells were plated on LB plates supplemented with ampicillin (50 µg/mL) and grown at 30°C. The kan gene flanked by FRT (Flp recombinase Recognition Target) sequences carried by pEAW507 was PCR-amplified using primers made of approximately 50 bases immediately adjacent to the region to be deleted fused to 19 bases bordering the FRT-kan-FRT cassette on the plasmid. Sequences of the oligonucleotides used for the deletions of metZWV and metY were as follows :

5′AAAAAAAGGTTGCATGAAAACGCGAGCGGAGTATAGTGCGCATCCACGGACGTAGAGGATCCCGGAAGT3′ (metZWV-forward)

5′TTTATGGGAGACCAGAAACAAAAAAACACCCGTTAGGGTGTTCAATAATTCCCTTTCGTCTTCAAGAAT3′ (metZWV-reverse)

5′GCAGTATTTGCATTTTTTACCCAAAACGAGTAGAATTTGCC ACGTTTCAGGCGTAGAGGATCCCGGAAGT3′ (metY-forward)

5′CGAAGGCCGAAGTCTTCACAGTATATTTGAAAAAGGACTCTAAGGGAAAGCCCTTTCGTCTTCAAGAAT3′ (metY-reverse)

The PCR product was digested by DpnI in order to degrade the plasmid template, and transformed by electroporation into MG1655-pKD46 cells grown at 30°C in LB medium containing 1 mM L-arabinose. After electroporation, 1 mL of LB medium containing 1 mM L-arabinose was added and cells were incubated 2 h at 30°C plus 4 h at 37°C, and finally plated at 37°C on LB plates containing kanamycin (25 µg/mL). Clones were then tested by PCR for correct recombination of the cassette at the desired locus and for loss of the pKD46 plasmid. PCR primers were as follows: 5′GGCGCAACGAAGATAACAAACCGCCGC3′ (ver-metZWV-forward), 5′GTATTTTGCGATGGGGCGACGTGGCGA3′ (ver-metZWV-reverse), 5′AGTATTCAACAAATGAAAGTGAACTGG3′ (ver-metY-forward), and 5′TAACAAAAAACCCCGATAAATCGGGGC3′ (ver-metY-reverse).

For elimination of the kan gene, cells were first transformed with pLH29 (Huang et al. 1997), a plasmid expressing the FLP recombinase. One colony was grown overnight at 37°C in 20 mL of LB containing 25 µg/mL chloramphenicol and 1 mM IPTG, and then left 2 d at room temperature. Clones were then screened for loss of resistance to kanamycin, and removal of the kan gene was then verified by PCR using the primers described above (Fig. 2A). Finally, plasmid pLH29 was eliminated by performing four successive liquid cultures to saturation in LB at 37°C. Individual clones were finally tested for loss of pLH29 by assaying sensitivity to chloramphenicol. This procedure yielded PalΔmetZWV and PalΔmetY strains.

For inactivation of metY in PalΔmetZWV carrying pBSTNAV-tRNAfMetA1–U72 derivative, the same procedure was used except that pKD46 was selected with gentamicin instead of ampicillin. In this case, the kan gene at the metY locus was not removed after metY deletion. This procedure yielded PalΔmetZWVΔmetY::kan strains carrying a pBSTNAV-tRNAfMetA1–U72.

tRNA purification

The E. coli tRNAfMetA1–U72 variant was overexpressed in PalΔmetZWVΔmetY::kan carrying pBSTNAV-tRNAfMetA1–U72 and purified as previously described (Mechulam et al. 2007). Briefly, a whole tRNA extract was prepared according to the Zubay procedure (Zubay 1962). The overexpressed tRNAfMet was then separated from other tRNAs by an anion exchange step on a Q-Hiload column (16 × 100 mm) equilibrated in 20 mM Tris–HCl, pH 7.6, 200 mM NaCl, 8 mM MgCl2, 0.1 mM EDTA. tRNAs were eluted using a 350 mM to 550 mM NaCl gradient in the same buffer. About 10 mg of tRNAfMetA1–U72 was obtained from 1 L of E. coli culture. Methionine acceptance of the tRNAfMetA1–U72 preparation ranged between 1450 and 1600 pmol/A260 unit. Before crystallization, the tRNA was methionylated and purified as described in Mechulam et al. (2007). The aminoacylated tRNA was resuspended in water (final concentration 400 µM) and diluted to 50 µM in a buffer containing 10 mM MOPS, pH 6.7, 5 mM MgCl2, 200 mM NaCl, 10 mM 2-mercaptoethanol before crystallization trials.

Crystallization, data collection, structure solution, and refinement

Crystals belonging to space group C2221 were obtained using the sitting drop method with a 1:1 mixture of a 50 µM solution of Met-tRNAfMetA1–U72 and 20% PEG 6000, 2 M NaCl as precipitants. Final pH of the 1:1 mixture of tRNA and precipitant was 6.3. Interestingly, the C2221 space group was never observed for crystals of wild-type E. coli initiator tRNAfMet. According to the CC(1/2) values (Table 1) the high-resolution limit of the data set was 2.8 Å (Karplus and Diederichs 2012; Diederichs and Karplus 2013). The structure was solved by molecular replacement using PHASER and the coordinates of E. coli tRNAfMet (PDB entry 3CW5 [Barraud et al. 2008]) from which the terminal CCA bases have been removed. Initial map showed different orientations of the anticodon region and of the 3′ extremity of the acceptor helix as compared to the molecular replacement-starting model. Cycles of manual reconstructions, performed with the program COOT (Emsley et al. 2010), and of refinements steps with PHENIX (Adams et al. 2010), led to a model refined to 2.8 Å resolution (Rwork = 0.196, Rfree = 0.225). Because of crystal packing interactions, the tRNA molecules are more tightly bound together along the acceptor helices. This is reflected by lower B-values along the acceptor helix comparing with those observed for atoms belonging to the anticodon helix. According to this anisotropy, the TLS procedure (Adams et al. 2010) was used in the last steps of refinement. All nucleotides are visible except A76. Moreover, positioning of C75 is only tentative. Refinement parameters in Table 1 were calculated with PHENIX.

Relative stability of adenine tautomers in solvent

The structures of adenine tautomers were energetically optimized at the density functional theory (DFT) level with the ORCA program version 3.0 (Neese 2012). The TPSS (Tao et al. 2003) functional and the Ahlrich's def2-TZVP AO basis set (Weigend and Ahlrichs 2005) were used. Single-point energy calculations at the higher level of theory, second-order Møller–Plesset perturbation (MP2), using the same basis set were performed on the minimized structures. The implicit solvent calculations were performed within the COSMO conductor-like screening model (Eckert and Klamt 2002) with a dielectric constant of ε = 80.0.

Molecular dynamics simulations

The simulations included tRNA nucleotides within a 24 Å sphere, centered on the A1 nucleotide of tRNA. tRNA atoms between 20 and 24 Å from the sphere's center were harmonically restrained to their experimentally determined positions. A 75 Å cubic box of water was overlaid, and waters overlapping the RNA atoms were removed. Periodic boundary conditions were assumed, i.e., the entire 75 Å box was replicated periodically in all directions. All long-range electrostatic interactions were computed efficiently by the particle mesh Ewald method (Darden 2001) and the appropriate number of potassium counterions were included to render the system electrically neutral. MD simulations were performed at constant room temperature and pressure, after 200 ps of thermalization. The CHARMM27 force field was used for the tRNA (Foloppe and MacKerell 2000) and the TIP3 model for water (Jorgensen et al. 1983). For the protonated form of adenine, we used a force field specifically developed in this work. Calculations were done with the NAMD program (Phillips et al. 2005). Ten nanoseconds of molecular dynamics were performed at constant room temperature and pressure.

To model residue A1 of tRNA in the crystalline environment, the symmetric units were constructed according to the crystallographic symmetry operators. Nucleotides within a 24 Å sphere around nucleotide A1 were retained for simulations. The rest of the system preparation was identical to the setup used for the simulations of tRNA in solvent.

Poisson–Boltzmann linear response approximation (PBLRA)

For the protonation free energy of the A1 adenine of tRNA in solution and in crystal, we used a simpler Poisson–Boltzmann linear response approximation, or PB/LRA (Aleksandrov et al. 2007). The protonation of the A1 adenine was modeled by changing selected atomic charges. The corresponding free energy change was computed both in tRNA and for the adenine molecule alone in solution, using conformations taken from the last 2 ns of tRNA MD simulations. The free energy was approximated by the continuum electrostatic free energy, where the tRNA atoms are explicitly included but the solvent is replaced by a dielectric continuum. The electrostatic potentials were computed for each MD conformation by numerically solving the Poisson–Boltzmann equation of continuum electrostatics (Aleksandrov et al. 2007), where the tRNA was treated as a single dielectric medium with a dielectric constant of 2; solvent was treated as another medium with a dielectric constant of 80, the experimental value for bulk water. The boundary between the two dielectric media was defined as the tRNA molecular surface, computed with a 1.6 Å radius probe sphere. For the potential calculation, the system was discretized using a cubic grid with a 68 Å edge and a spacing of 0.4 Å. The Poisson–Boltzmann equation was solved with Coulombic boundary conditions, using the CHARMM program (Brooks et al. 2009). We used an ionic strength corresponding to a 0.15 M concentration of monovalent ions.

The proton binding constant of A1 is related to ΔΔG by the relationship: where pKatRNA and pKaadenine are the pKa values of the titrating group in the tRNA and in the reference molecule, adenine, respectively.

DATA DEPOSITION

The Protein Data Bank accession number for the structure reported in this article is 5L4O.