The human mitochondrial tRNA^Met: Structure/function relationship of a unique modification in the decoding of unconventional codons

Abstract

Human mitochondrial mRNAs utilize the universal AUG and the unconventional isoleucine AUA codons for methionine. In contrast to translation in the cytoplasm, human mitochondria use one tRNA, hmtRNA^Met_CAU, to read AUG and AUA codons at both the peptidyl- (P-), and aminoacyl-(A-) sites of the ribosome. The hmtRNA^Met_CAU has a unique post-transcriptional modification, 5-formylcytidine, at the wobble position 34 (f⁵C₃₄), and a cytidine substituting for the invariant uridine at position 33 of the canonical “U-turn” in tRNAs. The structure of the tRNA's anticodon stem and loop domain (hmtASL^Met_CAU), determined by NMR restrained molecular modeling, revealed how the f⁵C₃₄ modification facilitates the decoding of AUA at the P- and A-sites. The f⁵C₃₄ defined a reduced conformational space for the nucleoside, in what appears to have restricted the conformational dynamics of the anticodon bases of the modified hmtASL^Met_CAU. The hmtASL^Met_CAU exhibited a “C-turn” conformation that has some characteristics of the U-turn motif. Codon binding studies with both E. coli and bovine mitochondrial ribosomes revealed that the f⁵C₃₄ facilitates AUA binding in the A-site and suggested that the modification favorably alters the ASL's binding kinetics. Mitochondrial translation by many organisms including humans sometimes initiates with the universal isoleucine codons AUU and AUC. The f⁵C₃₄ enabled P-site codon binding to these normally isoleucine codons. Thus, the physicochemical properties of this one modification, f⁵C₃₄, expand codon recognition from the traditional AUG to the non-traditional, synonymous codons AUU and AUC as well as AUA, in the reassignment of universal codons in the mitochondria.

Introduction

Mitochondria are membrane-bound organelles with their own genetic information. Through oxidative phosphorylation, the organelle generates the majority of the eukaryotic cell's energy resource, ATP. Most of the approximately 2000 proteins that compose the mitochondrion are encoded by nuclear DNA and are imported. However, the mitochondrial genome encodes thirteen critical components of the electron transfer chain and the ATP synthase. These proteins are the only products of the organelle's protein synthesizing machinery. Other components required for translation, including the ribosomal proteins and the auxiliary initiation, elongation and termination factors, are the products of nuclear genes. The mammalian mitochondrial DNA also encodes 22 tRNAs (mtRNA), one for each of 18 amino acids and two each for serine and leucine.¹ Cytoplasmic tRNAs are not imported into the mammalian mitochondria,² perhaps with the one exception of cytoplasmic glutamine tRNA imported into the isolated mitochondria of cells from individuals with myoclonic epilepsy with ragged-red fiber (MERRF).³

This minimal set of 22 mtRNAs must respond to all of the amino acid codons. In order to facilitate the reading of the amino acid codes in mammalian mitochondria, an unmodified uridine in the “wobble” position 34 of the mtRNA recognizes all four nucleotides at the third position of the synonymous codons, a guanine in position 34 recognizes only pyrimidines, and a modified uridine discriminates between purines and pyrimidines.⁴ Thus, there are some exceptions to the reading of the universal genetic code. Sequence analysis of mitochondrial genomes from various organisms has revealed that there are many codon reassignments, a unique feature of the non-plant mitochondrial translational systems.^4,5 For example, in mammalian mitochondria the universal stop codon UGA codes for tryptophan, and the two universal arginine codons AGA and AGG either terminate translation, or promote frameshifting in the human mitochondrial ribosome.⁶ Of particular interest is the universal isoleucine codon AUA that is used to specify methionine for initiation and elongation of translation.^1,7-10

In the cytoplasm, two tRNA^Met_CAU decode the single methionine codon AUG.¹ One is an initiator tRNA (tRNA_i^Met_CAU for eukaryotes and tRNA_f^Met_CAU for prokaryotes) that decodes AUG in the ribosome's peptidyl- or P-site at the initiation of translation. The other tRNA^Met_CAU is responsible for recognizing the AUG codons internal to the mRNA (tRNA_m^Met_CAU). In contrast, the single mitochondrial tRNA^Met_CAU plays a unique role since it supports both the initiation and elongation of protein synthesis by responding to the universal methionine codon AUG in the A- and P-sites.¹⁰ The tRNA^Met_CAU anticodon is the Watson-Crick complement for the methionine codon AUG. This single mtRNA^Met_CAU also reads the codon AUA which is present in mitochondrial mRNAs as 20% of the initiator codons, and 80% of internal methionine codons. Francis Crick's Wobble Hypothesis explained how a tRNA would be able to read more than one codon, but also predicted an unfavorable pairing of the anticodon CAU to the codon AUA due to the C●A mismatch at the wobble position.¹¹ Since the AUA codon occurs much more frequently than the AUG codon in mammalian mitochondrial genomes,¹ it has been assumed that some special mechanism must exist in the mitochondrial translation system to allow the AUA codon to be translated as methionine.

The unique post-transcriptional modification at the human mtRNA^Met_CAU (hmtRNA^Met_CAU) wobble position 34, 5-formylcytidine (f⁵C₃₄) (Fig. 1c),^12,13 may play a crucial role in the recognition of both methionine codons, the non-universal AUA and the universal AUG.^14,15 The hmtRNA^Met_CAU shares an important similarity to elongator tRNA^Met_CAU in that both have a wobble position-34 modification; in Escherichia coli the modification is N⁴-acetylcytidine, while in humans it is 2′-O-methylcytidine.¹⁶ The bacterial and eukaryotic cytoplasmic initiator tRNA^Met_CAU are unmodified at the wobble position. The f⁵C modification has been found in the mitochondrial tRNA^Met_CAU of humans, bovines, and rodents and is considered universal in mammalian mtRNA^Met_CAU.^10,13,17 The modification also is found in the mtRNA^Met_CAU of nematodes, squids, frogs, chickens, and fruit flies.^4,13,18,19 Human mtRNA^Met_CAU is a hybrid of the mammalian cytoplasmic initiator and elongator tRNA^Met_CAU species. The hmtRNA^Met_CAU has the three consecutive G-C pairs at the bottom of the anticodon stem characteristic of initiator tRNAs (Fig. 1). However, this tRNA has a number of unusual features. The dihydrouridine (D)-loop of the human mitochondrial tRNA^Met_CAU is smaller than that of the canonical cytoplasmic tRNAs and lacks the G residues at 18 and 19 that normally facilitate interactions with the T-loop. The minor loop is small and does not contain the common residue G47. The T-stem has two adjacent pyrimidine:pyrimidine pairs (U-U and U-Ψ). The T-loop lacks the TΨC sequence and contains only 6 nucleotides instead of the normal seven. There is also a mismatch in the acceptor stem.

Figure 1.

Primary sequence and secondary structure of the human mtRNA^Met_CAU and its anticodon stem and loop domain (hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄). Primary sequence and secondary structure of the human (a) mtRNA^Met_CAU and the (b) hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄. (c) Chemical structure of the 5-formylcytydine modification, f⁵C₃₄. Numbering of the nucleobase atoms is shown. (d) Nucleoside sequence and secondary structure of the anticodon stem and loop domains of the yeast tRNA^Met_i (left) and E.coli tRNA^Met_m (right).

Our studies focus on the folding, structure and codon binding properties of the hmtRNA^Met_CAU,^15,20,21 in particular that of tRNA's anticodon stem and loop domain, hmtASL^Met_CAU.¹⁵ The heptadecamer hmtASL^Met_CAU has a five base-paired stem consisting of the three contiguous G●C/C●G base pairs (residues 29–31/39–41) that are proximal to the loop and contribute thermodynamic stability to the RNA through stacking (Fig. 1) - a hallmark of initiator tRNA^Met that increases its affinity for the P-site.²² As with the mammalian initiator tRNA^Met, the hmtASL^Met_CAU has a cytidine at position-33. The 33^rd residue of tRNAs is usually occupied by the invariant U₃₃, responsible for the “U-turn” in the backbone that presents the anticodon for codon binding. Other than in the hmtASL^Met_CAU, C₃₃ has been found only in the cytoplasmic initiator tRNA^Met_CAU of insects, plants, and vertebrates.¹⁶ In the yeast initiator tRNA_i^Met_CAU, the loop is closed by a non-canonical C₃₂●A₃₈ base-pair, whereas in the hmtRNA^Met_CAU the loop has only the potential of a non-canonical C₃₂●C₃₈ base pair. Position 37, 3′-adjacent to the anticodons of many of the cytoplasmic tRNA_i,f^Met_CAU and tRNA_m^Met_CAU, have the t⁶A modification, or a variant,¹⁶ whereas the hmtRNA^Met_CAU has no modification at position 37. Previously, we had reported the procedures for chemical synthesis of the f⁵C mononucleoside and for its successful, site-selected incorporation into the hmtASL^Met_CAU.¹⁵ In the present paper, we use this chemistry to explore in detail the impact of the modification on the structure and dynamics of the ASL, and the biological ramifications of the formyl modification in the hmtASL^Met_CAU recognition of non-cognate codons. In our initial report the 5-formyl group appeared to be significant to the binding of the unconventional methionine codon AUA.¹⁵ The introduction of the f⁵C modification lowered the melting temperature and ellipticity, suggesting that this modification can enhance the dynamic of the loop¹⁵. Thus, the unique modification such as the f⁵C can represent a powerful tool in the stabilization of the codon-anticodon interaction. Nevertheless, how the mtRNA^Met interacts with the mRNA into the ribosome still remained unclear. Further, ribosome binding experiments using native tRNA^Met_CAU having the normal modifications and an unmodified transcript of tRNA^Met_CAU indicate that f⁵C is important for decoding in the A-site of the bovine mitochondrial ribosome.¹⁴ We have sought to understand the role of the 5-formyl group in the efficiency of binding of the hmtRNA^Met_CAU to the mitochondrial codons AUA or AUG at both the ribosomal P- or A-sites. Here, we report the first high resolution three dimensional structures of the modified and unmodified human mitochondrial anticodon domain of tRNA^Met_CAU. The conformational dynamic of the hmtASL^Met_CAU appears to be influenced by the f⁵C₃₄ modification in the direction of enhanced order in the anticodon loop. In addition, we demonstrate that the wobble position f⁵C₃₄ of hmtRNA^Met_CAU is particularly important to the binding of the hmtASL^Met_CAU to AUA at the A-site. Interestingly, some organisms use the isoleucine codons AUU and AUC in initiation of translation at the P-site.^9,18,23-25 The structure and dynamics of the anticodon may be critical to recognition of the unconventional AUA codon at the P- and A-sites of the ribosome, and in expanding its P-site recognition further to AUU and AUC.

Results

In order to determine the effect of f⁵C₃₄ on codon recognition and its physicochemical contribution to the conformation and dynamics of the hmtASL^Met_CAU, we have solved the solution structures of the hmtASL^Met_CAU-Ψ₂₇ and the hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ by NMR and restrained molecular modeling. In addition, we have analyzed the codon binding of the ASLs and the full length tRNA with E. coli and bovine mitochondrial ribosomes. The heptadecamer hmtASL^Met_CAU constructs were chemically synthesized with pseudouridine at the 5′-terminal position 27 (Ψ₂₇), and with or without f⁵C₃₄. The ASLs are designated hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ and hmtASL^Met_CAU-Ψ₂₇, respectively (Fig. 1).

Contribution of f⁵C₃₄ to the structure of the hmtRNA^Met_CAU anticodon: NMR resonance assignments

The structural contributions of f⁵C₃₄ to the hmtASL^Met_CAU were determined by NMR-restrained molecular modeling. The non-exchangeable resonances were assigned using well established techniques.²⁶ The 1D spectrum recorded exhibited three sharp peaks in the imino region which were assigned to G₂₉, G₃₀ and G₃₁ imino proton resonances, involved in canonical G●C base pairing (Fig. 2a). One other resonance, broader than the previous ones, at 11.0 ppm was assigned to Ψ₂₇H3 (Fig. 2a) and two NOE connectivities were observed between Ψ₂₇H3 and the amino group of A₄₃. The NMR data clearly demonstrated that the stem residues (C₂₈ – G₃₁ and C₃₉ – G₄₂) adopt a classical A-form RNA structure. The sequential pathway (aromatic to H1′ of the 3′ residue) could be followed using the 400ms NOESY spectrum from Ψ₂₇ to C₃₃ and from A₃₇ to G₄₃ (Fig. 3). A break appeared between C₃₃ and f⁵C₃₄/C₃₄, however, a weak sequential NOE was observed at 400 ms between f⁵C₃₄/C₃₄ H1′ and A₃₅H8. The two ASLs exhibited non-sequential NOEs between the A₃₅H8 and C₃₃H1′/H2′ protons. No NOE connectivity was observed between A₃₅H8 and U₃₆H1′, nor between U₃₆H6 and A₃₇H1′, even with a long mixing time (400ms). On the 5′ side of the loop, C₃₂ was stacked with G₃₁ and C₃₃, as shown by the inter-aromatic NOEs between the bases (C₃₂H5/G₃₁H8, C₃₂H6/G₃₁H8, C₃₂H6/C₃₃H6 and C₃₂H6/C₃₃H5). Moreover, the chemical shifts of C₃₈ and the C₃₂H5/H6 protons were characteristic of protons experiencing a 5′ purine ring current,²⁶ confirming the stacking of G₃₁/C₃₂, and of A₃₇/C₃₈. The A₃₇H2 resonance was identified based on the natural abundance ¹H-¹³C HSQC spectrum and a strong NOE connectivity was observed between the A₃₇H2 and C₃₈H1′. Non-sequential NOEs were observed between A₃₅H2 and A₃₇H1′, as well as between A₃₅H2 and A₃₇H8 (Fig. 3). The chemical shifts of U₃₆H5 and H6 were not in agreement with the chemical shift of a uridine experiencing a purine ring current, and thus were not supportive of the stacking of A₃₅ and U₃₆. Moreover, weak NOEs were observed between U₃₆H6 and the A₃₅H3′/H2′ protons, as well as very weak NOEs between U₃₆H5 and A₃₅H3′, and H2′. Weak inter-aromatic NOEs were observed between A₃₇H8 and U₃₆H6 and between U₃₆H6 and A₃₅H8. Nevertheless, the intensities of these NOE connectivities were less than the NOE between A₃₅H2 and A₃₇H1′ (Fig. 3). A considerable number of inter-residue NOEs were observed for the hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ and the hmtASL^Met_CAU-Ψ₂₇. The inter-nucleoside NOEs of the anticodon loop residues (C₃₂-C₃₈) were particularly important to establishing the contribution of the f⁵C₃₄. A significant number of these were attributed to the modified nucleoside, and to C₃₃ and A₃₅ (Fig. 4).

Figure 2.

¹H and ³¹P NMR spectra (1D) of the hmtASL^Met_CAU. (a) ¹H NMR spectrum of the hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄. The H₂O-exchangeable proton resonances, as well as the formyl-proton (H7) resonance of f⁵C₃₄ are shown. (b) ³¹P-NMR spectra (1D) of the hmtASL^Met_CAU-Ψ₂₇ (top) and the hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ (bottom). The ³¹P resonances for f⁵C₃₄, C₃₄, A₃₇ and U₃₆ are denoted in red.

Figure 3.

The hmtASL^Met_CAU NOSEY spectra. NOESY spectra of hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ (a) and hmtASL^Met_CAU-Ψ₂₇ (b). NOSEY spectra of the hmtASL^Met_CAU in D₂O were recorded at 25° C with a mixing time of 400 ms. The H1′ and H5 to aromatic portion of the spectra are shown. The sequential H1′-aromatic connectivities are indicated, as well as the unusual NOEs connectivities between A₃₅H2 – A₃₇H1′ (blue) and A₃₅H8 – C₃₃H1′ (red).

Figure 4.

Compilation of inter-nucleoside NOE connectivities. The inter-nucleoside NOEs observed in the loop residues 31 to 39 of the hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ were both sequential and non-sequential. The sequential (i to i+1) inter-nucleosides NOEs are shown in blue and the nonsequential NOEs are shown in red. The hmtASL^Met_CAU-Ψ₂₇ exhibited comparable NOEs.

Sugar conformations and backbone geometry

All of the loop residues, with one exception, exhibited small to very small ³J_H1′-H2′ coupling constants corresponding to the C3′-endo sugar pucker conformation. The COSY spectra indicated that U₃₆ adopted a C2′-endo sugar pucker conformation. The ³¹P phosphorous spectra of both ASLs showed that all the resonances are grouped together within ± 0.5 ppm except for three resonances belonging to the residues f⁵C₃₄/C₃₄, U₃₆ and A₃₇ (Fig. 2b,c). The phosphorous resonance of C₃₄ was downfield by 0.9 ppm from the ³¹P main cluster, whereas the f⁵C₃₄ ³¹P resonance was downfield by 1.1ppm (Fig. 2b,c). The ³¹P resonance of A₃₇ was slightly downfield (∼0.3 ppm), whereas that of U₃₆ was upfield (∼0.3 ppm) from the main cluster of peaks. All of the other resonances remained in the main cluster, indicating that the α and ζ angles were typical of an A-form RNA conformation (Fig. 2b,c).

Structures of the hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ and the hmtASL^Met_CAU-Ψ₂₇

The structures of the hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ and hmtASL^Met_CAU-Ψ₂₇ were derived from the NMR data using a simulated annealing protocol.²⁶ Ten of the lowest energy structures for the hmtASL^Met_CAU Ψ₂₇;f⁵C₃₄ and for the hmtASL^Met_CAU-Ψ₂₇ were superimposed (Fig. 5). The stems of both ASLs adopt a classical A-form RNA conformation. The ten lowest energy structures for the hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ and for the hmtASL^Met_CAU-Ψ₂₇ had overall heavy atom RMSDs relative to the mean structure of 1.71 ± 0.58 Å and 1.51 ± 0.46 Å, respectively (Table 1). The C₃₂●C₃₈, C1′ – C1′ distance was on average 10.19 ± 0.54 Å and 10.39 ± 0.35 Å for the hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ and for the hmtASL^Met_CAU-Ψ₂₇, respectively. These values are comparable to the classical A-RNA C1′-C1′ distance (∼10.6 Å). The heavy atom RMSD value observed for the loop residues (C₃₂ to C₃₈) relative to the mean structure were 0.38 ± 0.20 Å and 0.51 ± 0.17 Å, respectively.

Figure 5.

Structures of the hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ and hmtASL^Met_CAU-Ψ₂₇. The ten lowest energy structures for the hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ (a) and those for the hmtASL^Met_CAU-Ψ₂₇ (b) are superimposed. To illustrate differences in the degree to which the bases of a family of structures converge, the ribose and phosphodiester bond details are not shown.

Table 1. Structural statistics for hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ and hmtASL ^Met_CAU-Ψ₂₇.

	hmtASL MetCAU-Ψ27;f5C34	hmtASL MetCAU-Ψ27
Number of experimental restraints
Distance restraints	266	275
Intranucleotides	157	161
Internucleotides	109	114
Exchangeable	13	13
Restraints by residue	15.6	16.1
Hydrogen bonding distances restraints	24	24
Dihedral restraints	82	82
Negative NOEs	26	25
Heavy atoms RMSD from mean structure (Å)
All residues (Y27 - A43)	1.51 ± 0.58	1.71 ± 0.58
Loop residues (C32 – C38)	0.51 ± 0.17	0.38 ± 0.20

The two loops (C₃₂ to C₃₈) exhibited the same overall conformation (Fig. 6). The stem was continued from the canonical base pair G₃₁●C₃₉ at the bottom of the stem through the non-canonical C₃₂●C₃₈ mismatch pair (Fig. 6c). Taken together with the stacking of G₃₁/C₃₂ and C₃₉/C₃₈, the C₃₂●C₃₈ non-canonical base pair extended the stem. In the loop, the position of C₃₂ was stabilized by a hydrogen bond between the C₃₂O2 and the C₃₈N4 with acceptor-donor distances of 2.98 ± 0.15 Å and 3.30 ± 0.20 Å for the hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ and for the hmtASL^Met_CAU-Ψ₂₇, respectively. Moreover, the structures were stabilized by a probable hydrogen bond between C₃₂N4 and A₃₇O2P with acceptor donor distances of 3.12 ± 0.47 Å and 3.49 ± 0.48 Å for the hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ and for the hmtASL^Met_CAU-Ψ₂₇, respectively.

Figure 6.

Conformational differences and similarities in the anticodon loops of the f⁵C₃₄-modified and unmodified hmtASL^Met_CAU-Ψ₂₇. Stereo view of the average structures of the anticodon loops (residues C₃₂ to A₃₈) for (a) hmtASL^Met_CAU-Ψ₂₇ and (b) hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄. (c) The anticodon loops contain a C₃₂●C₃₈ intra-loop base pair in both the hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ and the hmtASL^Met_CAU-Ψ₂₇. (d) Stereo view of the two average structures superimposed (green, hmtASL^Met-Ψ₂₇;f⁵C₃₄; and purple, the hmtASL^Met_CAU-Ψ₂₇).

In both ASLs, the unusual C₃₃ residue was stacked with C₃₂. An abrupt turn in the backbone was observed between the residues C₃₃ and f⁵C₃₄/C₃₄ (Fig. 6a and b) The turn was characterized by a trans α dihedral angle (173 ± 20° for hmtASL^Met_CAU-Ψ₂₇;f⁵C and -160.5 ± 8.6° for hmtASL^Met_CAU-Ψ₂₇). The f⁵C₃₅/C₃₄ and A₃₅ residues were stacked nicely, consistent with the observation of n/n+2 NOEs between the two residues in the NMR spectra. Both nucleosides adopted the C3′-endo sugar conformation. The average distance between the C₃₃2′OH and the A₃₅N7 was 2.88 ± 0.08 Å and 2.90 ± 0.07 Å for the hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ and for the hmtASL^Met_CAU-Ψ₂₇, respectively, allowing the formation of a direct hydrogen bond. In both structures, the U₃₆ adopted the C2′-endo sugar pucker conformation (δ dihedral angles of 137.9 ± 4.6° for hmtASL^Met_CAU-Ψ₂₇;f⁵C and 143.1 ± 1.6° for hmtASL^Met_CAU-Ψ₂₇), which is consistent with the NMR data. Thus, the base resides outside the loop and in direct contact with the solvent. The intra-residue NOEs of f⁵C₃₄ include those between the formyl proton and H6 and H1′. These fix the position of the formyl proton relative to the nucleoside conformation. In addition, there are weak NOEs between the formyl proton and A₃₅H8 and the formyl proton and C₃₃H3′. Importantly, the conformational space occupied by the f⁵C₃₄ in the ten structures of hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ was considerably less than that occupied by C₃₄ within the ten structures of hmtASL^Met_CAU-Ψ₂₇. The limited space that confined the motional dynamics of the modification may be indicative of the modification's contributions to the reading of unconventional codons.

The anticodon wobble position modification, 5-formylcytidine, enhances AUA A-site codon recognition by the hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄

Mitochondrial tRNA^Met must recognize both AUG and AUA in the P-site for initiation and in the A-site during polypeptide chain elongation. To study the effect of the f⁵C₃₄ residue on the codon binding characteristics of the human mitochondrial tRNA^Met, we determined the equilibrium binding properties of the hmtASL^Met_CAU-Ψ₂₇, with and without the f⁵C₃₄ to the AUG and AUA codons at the P- and A-sites by titrating mRNA-programmed, E. coli 70S ribosomes with increasing amounts of the hmtASL^Met_CAU (Fig. 7; Table 2). As would be expected,²⁷ the two ASLs bound the P-site codons to a degree (3-4 pmoles bound) exceeding by several fold their binding of the A-site codons (0.5-1.0 pmoles). The 5-formyl modification did not affect the equilibrium binding of hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ to AUG at the P- site (Table 2). Neither the dissociation constant, nor the extent of binding to AUA, at the P-site appear to be affected greatly by the f⁵C₃₄ modification (Table 2; Fig. 7a). In binding AUA at the P-site, the hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ and the hmtASL^Met_CAU-Ψ₂₇ exhibited steady-state dissociation constants within experimental error of each other, 7.9 ± 5.5 μM and 11.0 ± 4.7 μM, respectively. Binding to AUA in the P-site was weak resulting in larger errors in the determination of the K_ds. The most striking observation in this data is that the extent of either hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ or hmtASL^Met_CAU-Ψ₂₇ binding at the P-site was substantially lower than observed with the AUG codon. The differences in saturation level between binding to AUA and binding to AUG probably arises from differences in kinetics. First, there appears to be an intrinsically lower ability of the hmtASL^Met_CAU to capture the formed complex when AUA is present. This is most likely a consequence of the difficulty of forming the f⁵C₃₄-A base pair and the weakness of this pairing. We suspect that on the E. coli ribosome two pathways to codon binding occur, as has been reviewed previously.²⁸ In one, the ASL binds before the mRNA, but leaves rapidly due to unstable interactions. In the other, the mRNA binds first via the Shine-Delgarno sequence, but also dissociates rapidly in the absence of codon:anticodon interactions. Thus, stable binding of the ASL is only observed when both hmtASL^Met_CAU and mRNA are present at the same time leading to codon:anticodon interactions. With AUA, the kinetics of the codon:anticodon interaction and/or the stability of the interaction is weaker than that with AUG and the complex cannot form to the same extent. Therefore, the major difference in saturation is that attributed to differences in codon:anticodon interactions. This observation suggests that the use of AUA as the initiation codon may reduce the efficiency of initiation complex formation for specific mRNAs in mammalian mitochondria.

Figure 7.

AUG and AUA codon binding properties of the hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ and hmtASL^Met_CAU-Ψ₂₇ in the P- and A-sites of the ribosome. (a) AUG and AUA codon binding of the hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ and hmtASL^Met_CAU-Ψ₂₇ in the P-site of the ribosome. The E. coli 70S ribosome (5 pmoles), programmed with AUG in the P-site, was titrated with hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ (▲) and with hmtASL^Met_CAU-Ψ₂₇ (■), or programmed with AUA and titrated with hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ () and with hmtASL^Met_CAU-Ψ₂₇ (). The maximum hmtASL^Met_CUA bound was between 3-4 pmoles. (b) AUG and AUA codon binding properties of the hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ and hmtASL^Met_CAU-Ψ₂₇ in the A-site of the ribosome. The E. coli 70S ribosome (5 pmoles), programmed with AUG in the A-site, was titrated with hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ (▲) and with hmtASL^Met_CAU-Ψ₂₇ (■), or programmed with AUA and titrated with hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ () and with hmtASL^Met_CAU-Ψ₂₇ (). The maximum hmtASL^Met_CUA bound was between 0.5-1.0 pmoles. In contrast to equilibrium binding studies, the time-dependent binding of the hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ (▲) and hmtASL^Met_CAU-Ψ₂₇ (■) to AUG and AUA were conducted at the single concentration of ASL (0.5 μM), far too low to saturate the ribosomes. The hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ and hmtASL^Met_CAU-Ψ₂₇ were bound to (c) AUG in the P-site of E. coli 70S ribosome, and to (d) AUA in the P-site). The hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ and hmtASL^Met_CAU-Ψ₂₇ were bound to (e) AUG in the A-site of E. coli 70S ribosome, and to (f) AUA in the A-site.

Table 2. Dissociation constants (K_d) of Unmodified and f⁵C₃₄-Modified hmtASL^Met_CAU-Ψ₂₇ for the Mitochondrial methionine Codons at the E. coli Ribosomal P or A-sites.

Dissociation Constants (Kd, μM)
hmtASLMetCAU-Ψ27	P-site		A-site
	AUG	AUA	AUG	AUA
unmodified	1.4 ± 0.29	11.0 ± 4.7	0.9 ± 0.1	4.5 ± 2.1
f5C34	1.3 ± 0.22	7.9 ± 5.5	2.4 ± 0.4	1.6 ± 0.4

The f⁵C₃₄-modified hmtASL^Met_CAU-Ψ₂₇ bound AUG in the A-site with a higher affinity than that of the hmtASL^Met_CAU-Ψ₂₇ as indicated by the lower dissociation constant (Table 2; Fig. 7b). At the A-site, the hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ bound AUA with a K_d of 1.6 ± 0.4 μM (Fig. 7b; Table 2). However, the hmtASL^Met_CAU-Ψ₂₇ exhibited a a higher K_d of 4.5 ± 2.1 μM. Thus, there appears to be a modification-dependent difference in the hmtASL binding of AUA at the A-site. Again, the largest difference observed in the AUG versus AUA-dependent binding at the A-site is reflected in a lower extent of binding suggesting that the off-rate for codon-anticodon interactions could be more rapid with the AUA codon.

Kinetics of the f⁵C₃₄-modified and unmodified hmtASL^Met_CAU-Ψ₂₇ binding AUG and AUA at the P- and A-site

The unmodified CAU anticodon of tRNA^Met is a canonical match for the codon AUG. Therefore, we did not expect the 5-formyl modification of hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ to have a significant effect on the binding of the ASL to AUG. As expected, steady-state binding assays of hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ or the hmtASL^Met_CAU-Ψ₂₇ to AUG at the P-site resulted in similar K_ds. However, there were differences in the levels of ASL concentration at which saturation was attained for the binding of AUG at the P-site, and also for AUA at the A-site (Fig. 7a,b). Thus, we explored the possibility that the two ASLs had significant differences in the kinetics of their binding. A significant difference in the kinetics of binding became apparent at a concentration of ASL (0.5 μM) far too low to saturate the ribosomes (Fig. 7c-d). The amount of hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ bound to AUG in the P-site was five-fold greater than that of the hmtASL^Met_CAU-Ψ₂₇ after only 30 seconds (Fig. 7). While the hmtASL^Met_CAU-Ψ₂₇ maintained a steady amount of weak binding over time, the amount of hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ bound to AUG continued to increase, reaching a ten-fold greater amount bound than that of the hmtASL^Met_CAU-Ψ₂₇ at the final time point of 30 min. These results were repeated for the binding of the two ASLs to AUG at the A-site, and AUA at both the P- and A-sites (Fig. 7). Though we were deterred in determining the rate constants for lack of sufficient modified hmtASL^Met_CAU, the time courses clearly indicate a significant difference in rates of association that are dependent on the presence of f⁵C₃₄. The binding appears to be biphasic. We suspect that the 70S preparation has some portion of 30S subunits present. The subunits would form complexes rapidly accounting for an initial rapid phase in the kinetics. The slower phase may represent binding to 70S particles which probably takes longer either because access to the P-site is more restricted or because there is a dissociation step required in providing 30S subunits prior to the mRNA binding step. The equilibrium binding constants of the hmtASL^Met_CAU-Ψ₂₇ to the P-site codons AUG and AUA are not affected by the presence of f⁵C₃₄ (Table 2), though there seems to be an effect on the extent of binding (Fig. 7). The similarities of the K_ds taken together with the results of the time courses at low ASL concentration indicate that the modification affects both on- and off- rates. In comparison, the equilibrium binding of the hmtASL^Met_CAU-Ψ₂₇ to the A-site codons AUG and AUA is enhanced by the presence of the f⁵C₃₄, and the respective time courses reflect this result. This suggests that in the absence of f⁵C, the off-rate may predominate.

Codon binding of hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ and the hmtASL^Met_CAU-Ψ₂₇ on the bovine 55S mitochondrial ribosome

Thus far, the codon binding experiments we have described for hmtASL^Met_CAU were conducted with the model E. coli 70S ribosomes. However, previous studies of translation in vitro using bovine mitochondrial 55S ribosomes succeeded in demonstrating low levels of incorporation of methionine into protein using the native bovine mtRNA^Met_CAU, unmodified transcripts and mitochondrial ribosomes programmed with poly(AUA).¹⁴ Therefore, we thought it appropriate to attempt the binding of hmtASL^Met_CAU to programmed, purified bovine mitochondrial 55S ribosomes. The bovine mitochondrial 55S ribosomes were programmed with mRNA, and then either hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ or the hmtASL^Met_CAU-Ψ₂₇ was added. Mitochondrial mRNAs tend to lack the 5′-UTR. In order to study the binding of hmtASL^Met_CAU to the initiation site (P-site), we used a 30-mer oligonucleotide mRNA with either AUG or AUA at the 5′ terminal end followed by the valine codon GUA as the second codon. The binding of the ASLs to AUG in the P-site was detectable, although the level of codon binding was considerably lower than that observed on the E. coli ribosomes. The 5-formyl-group of the f⁵C₃₄-modified hmtASL^Met_CAU enhanced the binding of AUG by approximately 40% (Fig. 8a, first two bars). Though the P-site binding of the hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ to AUA was detectable, the binding of the hmtASL^Met_CAU-Ψ₂₇ was not (Fig. 8a, second two bars). The presence of f⁵C₃₄ appears to have enhanced P-site binding. To observe A-site binding, the 55S ribosomes were programmed with a 30-mer mRNA construct consisting of a 5′-leader of twelve uridines, followed by GUA at the P-site and AUG or AUA at the A-site, and ending with another stretch of twelve U residues. The presence of the 5-formyl modification on the ASL enhanced A-site binding to AUG sufficiently for it to be observed (Fig. 8a). However, neither the binding of hmtASL^Met_CAU-Ψ₂₇ to AUG, nor the binding of either ASL to AUA was observed in the A-site.

Figure 8.

Codon binding by hmtASL^Met_CAU, the native hmtRNA^Met_CAU and the unmodified transcript on the bovine mitochondrial ribosome. (a) AUG and AUA codon binding by the hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ (f⁵C₃₄) and hmtASL^Met_CAU-Ψ₂₇ (unm) on the 55S bovine mitochondrial ribosome. The modification f⁵C₃₄ enhanced P-site binding of the AUG and AUA codons, and A-site binding to the AUG codon. However, A-site binding of the AUA codon by either ASL was not detectable. The binding (pmoles) of hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ and hmtASL^Met_CAU-Ψ₂₇ are shown under conditions in which the ASLs are saturating. (b) The IF2-facilitated binding of bovine mtRNA^Met_CAU and its transcript to AUG and AUA in the P-site of the bovine mitochondrial ribosome. The bovine mitochondrial ribosome programmed with AUG was titrated with either mtRNA^Met_CAU (○) or its transcript (). Also, the AUA-programmed ribosome was titrated with either mtRNA^Met_CAU (◊) or its transcript ().

In our mitochondrial 55S ribosome binding assays, only small amounts of the hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ bound AUG at the A-site and AUA at the P-site. The low level of detectable binding to the P-site, and the inability to detect the binding of either hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ or the unmodified hmtASL^Met_CAU to AUA at the A-site may have been due to the very low levels of binding to mitochondrial ribosomes, in general, experienced by us (approximately 0.8 pmol of hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ bound to AUG at the P-site) and others^10,14 in comparison to E. coli 70S ribosomes (4 pmol of hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ bound to AUG at the P-site¹⁰). Also, the codon AUA is not as efficient as AUG in the decoding of methionine in mitochondrial translation. Translation in vitro with a poly(AUG) mRNA resulted in five-fold greater amounts of Met-containing polypeptides being produced than from a poly(AUA) mRNA.¹⁴

The lack of full tRNAs, aided by initiation and elongation factors, may also have contribute to the low levels of binding. Therefore, we conducted initiation factor, IF2_mt-facilitated, binding of the native bovine mtRNA^Met_CAU and the full length, but unmodified, transcript to codons at the P-site of the bovine mitochondrial 55S ribosome. Native bovine mtRNA^Met_CAU was isolated and purified,²⁹ and the unmodified mtRNA^Met_CAU obtained through transcription in vitro.³⁰ The native tRNA and transcript were aminoacylated and formylated. The IF-2_mt-facilitated binding of the native bovine and unmodified transcript of fmet-mtRNA^Met_CAU to AUG at the P-site were essentially identical at very low levels of input tRNA (Fig. 8b). The affinity of the two tRNAs for AUA at the P-site was considerably lower. However, the absence of modifications on the transcript made no apparent difference in the binding. Thus, the binding of full length native bovine fMet-mtRNA^Met_CAU and the unmodified transcript to AUG at the P-site of the mitochondrial ribosome was consistent with the results obtained in the binding of the hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ and hmtASL^Met_CAU-Ψ₂₇. Though the fMet-mtRNA^Met_CAU and the unmodified transcript comparably bound AUA at the P-site, the extent was diminished to the degree as to not be able to make a reasonable comparison to the binding of the respective mtASLs. The mitochondria uses AUA to initiate translation (AUA at the P-site) only about twenty percent of time, but approximately eighty percent of mitochondrial methionine A-site codons are AUA (twenty percent are AUG).^21,31 Thus, the probability of AUA being at the mitochondrial P-site is significantly lower than that of AUG.

Ability of the 5-formyl-modification to expand codon recognition within the codon box AUN

Mitochondria not only initiate translation from AUG and AUA, but also have been found to initiate from the two mitochondrial isoleucine codons AUU and AUC.^9,24,32 AUU and AUC are rare initiation codons and may be used to regulate gene expression. The ability of this single and modest modification, f⁵C₃₄, to enable the one hmtRNA^Met_CAU to initiate translation with AUA, as well as with AUG, begs the question as to whether f⁵C₃₄ expands the tRNA's ability to decode the other synonymous codons, AUU and AUG. We found that the hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ bound to AUU and to AUC with a K_d similar to that for AUA (Fig. 9). Furthermore, as it did in binding to AUA, the modified hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ bound to AUU and AUC with a dissociation constant that was two- to three-fold less than that of the hmtASL^Met_CAU-Ψ₂₇ (Fig. 9). The hmtASL^Met_CAU-Ψ₂₇ was bound to AUU, AUC, and AUA with K_ds comparable to, and within error of, that of the negative-control codon for valine GUA (Fig. 9).

Figure 9.

The f⁵C₃₄-modified hmtASL^Met_CAU-Ψ₂₇ binds the near-cognate codons AUU and AUC. The modification f⁵C₃₄ enabled hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ to bind the near-cognate codons AUU and AUC, as well as AUA, in the P-site of the E. coli ribosome. The relative binding of non-cognate codons by hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ (f⁵C₃₄) and hmtASL^Met_CAU-Ψ₂₇ (unm) are shown under conditions in which the ribosomes are saturated with ASLs, and the results internally normalized (percent). The binding of the valine codon GUA by hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ (f⁵C₃₄) is an indication of the degree of non-specific binding in the assay.

Discussion

Determining the structural contributions of f⁵C₃₄ to the function of mtRNA^Met_CAU presented a unique challenge because this single RNA has the dual role of participating in initiation and elongation, as well as reading both AUA and AUG. The hmtRNA^Met_CAU has unique decoding properties because the f⁵C₃₄ extends codon recognition often to AUA, and to a small extent to AUU and AUC. The hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ and hmtASL^Met_CAU-Ψ₂₇ exhibited the same overall conformation expected of a hairpin RNA with a stem adopting an A-form RNA structure extended by a non-canonical base pair between C₃₂●C₃₈. The 5′ side of the loop was characterized by the stacking of C₃₂ with the unique C₃₃, and the 3′ side by the stacking of A₃₅, A₃₇ and C₃₈. A sharp turn was observed in the backbone and was characterized by a trans α dihedral angle between the residues C₃₃ and f⁵C₃₄/C₃₄. This observation is in accordance with the downfield ³¹P resonance belonging to f⁵C₃₄/C₃₄ and represents one of the characteristics of the U-turn motif.^33,34 Moreover, the “classical” U-turn motif is stabilized by a network of hydrogen bonds between the U₃₃2′OH and the N7 of the purine at residue 35 (most often an adenosine), and by a hydrogen bond between the U₃₃H3 exchangeable proton and the phosphate group of residue 36. These hydrogen bonds have been shown to be essential in the formation of the U-turn and to the ribosomal binding properties of the tRNAs.³⁵ Because C₃₃ substitutes for the invariant U₃₃, some of the hydrogen bonds that stabilize the canonical U-turn conformation adopted by cytoplasmic tRNA^Met_CAU and other tRNAs should not be possible for the hmtRNA^Met_CAU.^33,35 The last hydrogen bond is not possible in the case of a cytosine in position 33, unless the amino group of the cytosine plays the role of hydrogen bond donor. The three dimensional structures of the hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ and the hmtASL^Met_CAU-Ψ₂₇ showed that the C₃₃ amino group is directed toward the minor groove and is not involved in direct hydrogen bonding. Nevertheless, distances between the C₃₃N4 and the O2P of residue C₃₂ are on average 4.07 ± 0.72 Å and 3.93 ± 0.55 Å for hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ and hmtASL^Met_CAU-Ψ₂₇. If these distances preclude the formation of a direct hydrogen bond, the formation of some indirect hydrogen bonds involving water molecules is highly possible. Water mediated hydrogen bonds can be of importance in the stabilization of the loop conformation and the formation of the observed ersatz U-turn, or “C-turn”. The projection of the U₃₆ residue from the structure and into the solvent is consistent with the NMR data. This is not very surprising when we consider that the uridine is unpaired and between the two stacked purines, A₃₅ and A₃₇. Unpaired uridines have a tendency to bulge from the structure.³⁴ Modifications in position 37 provide stability to the anticodon-codon interaction,³⁶ and are important in driving the conformation of the anticodon loop toward that of the well-known U-turn important to the binding of the codon on the ribosome.^37-39 In the anticodon domain structure of the S. pombe initiator tRNA^Met_CAU, with a uridine in position 33 and the t⁶A₃₇ modified residue, U₃₆ is also bulged out from the structure,³⁸ whereas in the structure of the unmodified yeast initiator tRNA^Met, U₃₆ adopted the C2′-endo sugar conformation and is loosely stacked with A₃₅.⁴¹ Direct exposure of U₃₆ to the solvent makes it available for tertiary interactions, in particular with the mitochondrial methionyl-tRNA^Met synthetase (mtMRS). mtMRS recognition of the cytoplasmic tRNA^Met species includes the anticodon,^42,43 and especially residue C₃₄ as a strong identity element.⁴⁴ Moreover, the residue U36 has been shown to be essential for the tRNA^Met – MetRS interaction by being specifically recognized by the MetRS C-terminal extended loop.⁴⁵

The f⁵C₃₄ modified residue is one of the most conformationally rigid residues with a C3′-endo sugar conformation that could help in pre-structuring the anticodon for codon interaction.^39;46 Yet, UV-monitored studies of thermal stability and circular dichroism studies of stacking interactions of the hmtASL^Met_CAU-Ψ₂₇ with and without f⁵C₃₄ indicated that the introduction of f⁵C₃₄ may enhance the ASL's motional dynamics.¹⁵ While the biological function of f⁵C₃₄ is evident from our work and that of others, the physicochemical contribution of the f⁵C₃₄ distinguishing hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ from hmtASL^Met_CAU-Ψ₂₇ appears ambiguous. However, this apparent contradiction of thermal instability and restricted conformational dynamics is found in other ASLs when the introduction of a modified nucleoside at positions 34 and/or 37 increases the width of the anticodon loop, as does f⁵C₃₄.⁴⁷ As with other wobble position-34 modifications,^37,48,49 the f⁵C₃₄ also takes a defined chemical and physical space within the anticodon loop that possibly restricts the conformational dynamics of the other loop nucleosides, in particular A₃₅ and A₃₇, in comparison to that of hmtASL^Met_CAU-Ψ₂₇ (Fig. 5). In the absence of stable isotope labeling, the restricted dynamics are deduced from a comparison of the family of calculated structures for hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ with that of hmtASL^Met_CAU-Ψ₂₇ (Fig. 5) We suggest that the mechanism by which the formyl oxygen of f⁵C₃₄ may restrict conformational dynamics is similar to that of the C5-oxygen of 5-oxyacetic acid uridine (cmo⁵U₃₄). Cmo⁵U₃₄, found in one of the isoaccepting species for each of the E. coli tRNAs for alanine, proline, threonine, serine and valine, enables the tRNA's reading of all four synonymous codons by pre-structuring of the anticodon, and fixing the geometry of the base in the A-site of the ribosome.^48;49

Position-37 is not modified in the mtRNA^Met_CAU. Modifications of the purine at position-37 provide stability to the tRNA/mRNA interactions, and have been shown to be important in driving the conformation of the ASL toward the well-known U-turn structure that accommodates the binding to cognate codons on the ribosome.^39,47 The presence of a cytosine in position-33 is quite unusual.¹⁶ Of the 21 sequences of tRNAs having a cytosine in position-33, 12 are cytoplasmic initiator tRNAi^Met_CAU and three are mitochondrial mtRNA^Met_CAU.¹⁶ Surprisingly, the presence of a cytosine residue in position-33 is correlated with the absence of modifications in position-37 in mitochondria, except for that of plants. The structures of the two cytoplasmic tRNA^Met_CAU species have been studied. Crystal structures have been solved for the entire eukaryotic yeast and E. coli tRNA^Met_CAU species free,^50-53 and in complex with the methionyl-tRNA synthetase (metRS).⁵⁴ In addition, the anticodon stem and loops of the E. coli tRNA_f^Met_CAU and unmodified tRNA_m^Met_CAU were solved by NMR,⁴¹ as well as that of the S. pombe tRNA^Met_CAU ASL containing the t⁶A₃₇ modification.³⁸ The anticodon stem and loop domains of yeast, E. coli, human and plant initiator tRNAs have the same primary sequence (residues 29-41) except for position-33 which is a uridine in yeast and E. coli initiator tRNAs, and a cytosine in human and plant initiator tRNAs. This singular difference appears to be responsible for the specific, H₂O-mediated hydrolysis of the C₃₄-A₃₅ phosphodiester bond in human and plant initiator tRNAs, and not in yeast and E. coli initiator tRNAs.⁵⁵ The substitution of U₃₃ with C₃₃ in elongating tRNAs has been shown to decrease codon binding on the ribosome.^35,56 Substitution of U₃₃ for C₃₃ in hmtASL^Met_CAU does not affect P-site binding to either AUG or AUA (not shown). Nevertheless, the S1 nuclease cleavage patterns for the anticodon stem and loop domains of the prokaryotic and eukaryotic initiator tRNAs remain the same whether a cytosine or a uridine is in position-33.⁵⁷

The A-site anticodon/mRNA interaction is proof-read and stabilized by the 16S rRNA residues G₅₃₀, A₁₄₉₂ and A₁₄₉₃.⁵⁸ Crystal structures of the ribosome show that U₃₃ does not make contact with the ribosome, and that the U₃₃N3 is involved in the “classical” U-turn hydrogen bond.^18,36 In the NMR structures of the hmtASL^Met_CAU, C₃₃ lacks that hydrogen bond. Once on the ribosome, the amino group of C₃₃ is available to form tertiary interactions with the rRNA that could stabilize the anticodon/codon binding. Interestingly, the crystal structure mentioned above reveals that the anticodon P-site tRNA^Phe wobble base (G₃₄) is stabilized by stacking interactions with the 16S rRNA base C₁₄₀₀ and by packing between the ribose of G₃₄ and the base G₉₆₆.⁵⁹ If the same interactions are observed in the case of the hmtRNA^Met_CAU, the enhanced stacking properties of the f⁵C₃₄ modified residue could account for stabilization of the hmtASL^Met/mRNA interaction. It appears that the C₃₃ may change the conformational dynamic of the ASL loop, and thus could be of importance in stabilizing the f⁵C●A base pair, as well as that with AUG.

A pre-structured anticodon is the result of modification-induced, restricted dynamics, and is advantageous for codon binding.³⁹ The three hydrogen bonds between an f⁵C₃₄ and a guanosine are no different than that of a tRNA's C₃₄●G base pair with the mRNA codon. However, codon recognition and binding of the hmtRNA^Met_CAU to AUA in the P- or A-sites is problematic because the unconventional f⁵C₃₄●A interaction would be expected to be quite weak.⁶⁰ Yet, eighty percent of all methionine codons internal to mitochondrial mRNAs are AUA. We have found that that the f⁵C₃₄ facilitates the binding of the hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ at the A-site, and also may effect the kinetics of codon recognition at both the P- and A-sites. In addition, humans, giant pandas and fruit flies have been found to use AUU for initiation with methionine, honey bees use AUC, and mice use both AUU and AUC to initiate translation.^24,25 The hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ read AUU and AUC at the P-site to an extent similar to its binding to AUA, whereas the hmtASL^Met_CAU-Ψ₂₇ was not able to do so. Thus, the f⁵C₃₄ appears to have expanded the ability of the tRNA to read all four synonymous codons within the AUN codon box.

Molecular dynamics simulations of the base pairing of the 5′-monophosphate of f⁵C₃₄ with that of A under aqueous and neutral conditions⁶¹ indicate the potential for novel interactions, including a bridging H₂O molecule that stabilizes the interaction (Fig. 10). The f⁵C₃₄ intranucleotide hydrogen bond that we observed between the f⁵C₃₄ amino and formyl groups in the NMR-derived structure could increase the stacking interaction between f⁵C₃₄ and A₃₅ and possibly A₃₇, and thus compensate for the lack of hydrogen bonds in an anticodon f⁵C●A codon interaction.¹⁴ A hydrogen bond between the N4-amino group and the formyl-group oxygen at the position C5 of the f⁵C₃₄ modification may have altered the electronic properties of the modified residue, and thereby affected the base pairing properties of the f⁵C₃₄, enabling a f⁵C₃₄●A mismatch to form on the ribosome, as it has been proposed.¹⁴ An additional hydrogen-bond could be formed between f⁵C₃₄ of the anticodon and A, the third base of the codon, if the adenosine were protonated. This is possible since protonation of an adenosine has been shown to be effective in certain RNA constructs at a pH near 7.⁶² Nevertheless, an f⁵C●A⁺ interaction is still weaker than an f⁵C●G or C●G interaction. Further studies need to be conducted to reveal the geometry, and stability that the f⁵C₃₄ modification brings to the interactions of the hmtRNA^Met with unconventional codons of mRNA in the A- and P-sites of the ribosome.

Figure 10.

Molecular dynamics simulation of the f⁵C₃₄●A base pair. The f⁵C₃₄●A base pair must occur on the mitochondrial ribosome when hmtRNA^Met_CAU reads the AUA codon in either the A-or P-sites. The base pairing of the 5′-monophosphate of f⁵C₃₄ with that of A under aqueous and neutral conditions was simulated with molecular dynamics,⁶¹ and the lowest energy structure is presented here. The resulting model indicated a possible bridging H₂O molecule that stabilizes the interaction.

Materials and Methods

Oligonucleotide Chemical Synthesis and tRNA Preparation

RNAs were chemically synthesized (Thermo Fisher Scientific, Lafayette, CO) using “ACE” chemistry.⁶³ The protected 5-formylcytidine modified nucleotide was chemically synthesized and incorporated at position-34.¹⁵ The RNA molecules were deprotected and dialyzed extensively against 20 mM phosphate buffer (10 mM KH₂PO₄; 10 mM Na₂HPO₄; pH 6.8; 0.05 mM EDTA) using a membrane with a 3500 Da cutoff. For observation of the exchangeable protons by NMR, the samples were resuspended in 300 μl of 90% H₂O/10% D₂O. For experiments involving the non-exchangeable protons, the samples were exchanged twice with 99.9% D₂O. The NMR sample concentration was between 1.0 and 1.5 mM. Full length, native bovine mtRNA^Met_CAU and its unmodified transcript were prepared as previously described.³⁰

NMR Spectroscopy

NMR spectra were collected on a DMX Bruker 500 MHz instrument equipped with a triple-resonance ¹H, ¹³C, and ¹⁵N probe and three-axis pulsed field gradient capabilities, and on a Varian INOVA 600 MHz instrument equipped with a cryoprobe. NMR spectra were collected with the RNA samples under solution and temperature conditions required and widely used for determination of nucleic acid structures.^49,62,64 All NMR data were processed using NMRPIPE.⁶⁵ Spectra were displayed and analyzed using SPARKY software.⁶⁶ NOESY spectra in D₂O were recorded at 10, 25 and 30° C with mixing times of 50, 150 and 400 ms. TOCSY, COSY and DQF-COSY experiments were acquired at 25 °C. Natural abundance ¹H-¹³C HSQC and ¹H-³¹P HETCOR experiments were recorded in D₂O at 25 °C. For the study of exchangeable protons, 1D 1H and 2D NOESY spectra were recorded at different temperatures with various mixing time using the Watergate water suppression sequence. In confirmation of the unimolecular character of the ASLs after purification and dialysis, 1D 1H and 1D-³¹P spectra were recorded immediately. The ³¹P NMR spectrum exhibited a highly downfield resonance (characteristic of the canonical U-turn conformation) that we attributed to be P₃₄. This property was confirmed with the ³¹P resonance assignments using the ³¹P-1H HECTOR NMR spectrum. Other assignments confirmed that we had a C-turn (a pseudo-like U-turn). For both RNAs (with or without the f⁵C₃₄ modification), a dipolar connectivity between the protons A₃₅-H8/C₃₃-H1′ was observed within the NOESY spectra also characteristic of the canonical U-turn (Fig. 3). The formation of a C-turn is incompatible with the RNA adopting a duplex conformation. A duplex conformation would probably result in non-canonical, mismatch base pairs between C₃₂-C₃₈, C₃₃-A₃₇ and others. In a duplex conformation, observation of the A₃₅-H8/C₃₃-H1′ dipolar connectivity would require the flipping out of the residue f⁵C₃₄. This was absolutely not the case as we observed NOE connectivities between the residues f⁵C₃₄ and A₃₅ (C₃₄/f⁵C₃₄-H6/A₃₅-H8 and between f⁵C₃₄ H7–A₃₅ H8) demonstrating that those two residues are stacked.

NMR restraints generation and structure calculation

NMR data collected at 25° and 30° C were used for the structure calculations. The NOE cross-peaks were integrated using the peak fitting Gaussian function and volume integration in SPARKY software. Using the intensity of the cross peak, and as reference the H5-H6 covalent distance for uridine and cytosine residues, the cross-peaks were qualitatively classified as strong (3.0 1.2 0.0), medium strong (4.0 2.0 0.0), medium (5.0 3.0 0.0), weak (5.0 2.0 0.0) or very weak (6.0 3.0 0.0). No inter-nucleotide restraints were used from the H5′ and H5″ protons. For restraints involving imino protons, (3.0 1.2 0.0) and (5.0 3.0 0.0) intervals were used. A total number of 266 and 275 distance restraints were employed for generating the structures of the hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ and the hmtASL^Met_CAU-Ψ₂₇ (Table 2). Consistent with the NMR data, the four stem G●C Watson-Crick base pairs were subjected to six hydrogen-bond restraints (1.8-2.5 Å between the donated hydrogen and the acceptor, and 2.7 - 3.5 Å between the two heavy atoms). No hydrogen bond restraints were used between Ψ₂₇ and A₄₃, despite the observation of an NOE between the Ψ₂₇H3 imino proton and the A₄₂NH2 protons. Dihedral constraints on the ribose ring and backbone were derived from semi-quantitative measurement of the ³J_H-H and ³J_H-P couplings. Sugar pucker conformations were derived from the observation of the COSY and DQF-COSY spectra. Due to the large U₃₆H1′-H2′ coupling observed, this residue was constrained to the C2′-endo (δ: 160 ± 20°), whereas for the other residues only weak (A₄₃) or very weak H1′-H2′ connectivity was observed on the COSY spectra and these residues were therefore, left unrestrained (A₄₃) or constrained to the C3′-endo conformation (δ: 85 ± 20°). Independent confirmation of the U₃₆ sugar pucker conformation was derived from the ¹H-¹³C HSQC spectra, and the C3′ and C4′ resonances downfield from the main cluster of sugar resonances. The γ torsion angle was derived from the ³J_H4′-H5′ and ³J_H4′-H5″ peaks in the COSY and DQF-COSY. For residues that lacked H4′-H5′ and H4′-H5″ peaks (J < 5Hz), γ was constrained to the g⁺ conformation (54 ± 20°). For the other residues, γ was left unrestrained. β dihedral angles were derived from the ³J_P-H5′ and ³J_P-H5″ couplings observed in the ¹H-³¹P HETCOR NMR spectrum. When the ³J_P-H5′ and ³J_P-H5″ were clearly absent, β was constrained to the trans conformation (178 ± 20°). When ³J_P-H5′ and ³J_P-H5″ could be observed, β was left unrestrained. All H3′-P were clearly identified on the ¹H-³¹P HETCOR (³J > 5Hz), so the ε dihedral angle were loosely constraint to avoid the steric forbidden g⁺ conformation (-120 ± 120°). Dihedral angle restraints for the α and ζ were derived from the observation of the ³¹P chemical shifts. All of the ³¹P resonances, except for ³¹P of the f⁵C₃₄ and C₃₄ which were found strongly downfield and A₃₇ to a lesser extent, were in a sharp range of chemical shift. The U₃₆ ³¹P chemical shift was upfield to the majority of the ³¹P resonances. Thus, α and ζ were constraints to avoid the trans conformation resulting in strongly downfield 31P resonances (0 ± 120°). F₅C₃₄/C₃₄, A₃₇ and U₃₆ α and ζ dihedral angles were left unrestrained. Each of the intra-aromatic to H1′ NOE connectivities of the different ASL residues had an intensity significantly less than that for the H5-H6 NOE connectivity showing that all the bases adopted an anti glycosidic torsional angle. Thus, the χ glycosidic dihedral angle was constrained to the anti conformation for all residues. NMR restrained molecular modeling of the hmtASL^Met_CAU-Ψ₂₇;f⁵C and hmtASL^Met_CAU-Ψ₂₇ structures was achieved using CNS 1.2.^67,68 One hundred structures were calculated using a standard NMR restrained annealing protocol.⁶⁹ The 10 structures having the lowest total energy, without dihedral or distances restraint violations, were chosen for further analysis. The structure helical parameters were analyzed using X3DNA software.⁷⁰ The output structures were visualized with MOLMOL and PYMOL.⁷¹

Codon Binding

In order to observe binding at the P-site, ribosomes were programmed with a 27-mer oligonucleotide mRNA containing a Shine-Dalgarno sequence at the 5′-leader, an AUG or AUA at the P-site, and the valine codon GUA at the A-site. The binding of mRNA was assessed using radiolabeled mRNAs to ensure that the different mRNAs were bound by the ribosome similarly. The binding of the different mRNA sequences to the ribosomes revealed no significant differences (data not shown). Since ASLs bind P-site codons significantly better (∼100-fold) than A-site codons²⁷ and both the hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ and hmtASL^Met_CAU-Ψ₂₇ bound the valine codon GUA poorly in the A- or P-sites, similar to that of nonspecific ribosome binding (data not shown), the A-site was left empty. Near saturation, the amount of hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ and hmtASL^Met_CAU-Ψ₂₇ bound to AUG in the P-site exceeded their binding of AUG in the A-site by many fold (2-4 pmoles bound compared to 0.5-1.0 pmoles). In order to study the binding of the hmtASLs to the A-site codons AUG or AUA, the E. coli 70S ribosome was programmed with a 27-nucleotide mRNA with a 5′-leader with containing the Shine-Dalgarno sequence, the Val codon GUA at the P-site, and either AUG or AUA at the A-site. The P-site was saturated with unmodified ASL^Val_UAC and A-site binding by either hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ or the hmtASL^Met_CAU-Ψ₂₇ was tested. Control experiments saturating the GUA-programmed A-site with ASL^Val_UAC after P-site binding revealed similar results as those in the absence of ASL^Val_UAC (data not shown). The 70S ribosomal subunits were prepared from E. coli MRE600.⁷² Insignificant amounts of the ASLs that had been 5′-endlabeled with ³²P using [γ-³²P] ATP (MP Biomedicals)¹⁵, and were radiochemically detectable (2,000-5,000 CPM), were mixed in a fixed ratio with unlabeled ASLs in a range of concentrations (0-5 μM).⁴⁹ The filter binding assay (30 min; 25 °C) was performed with activated ribosomes (0.25 μM; 5 pmoles) that had been programmed with mRNA (2.5 μM) and used the ribosomal binding buffer (50 mM HEPES, pH 7.0; 30 mM KCl; 70 mM NH₄Cl; 1 mM DTT; 100 μM EDTA; 20 mM MgCl₂).⁴⁹ Data collected from a phosphoimager were measured for radioactive intensity using ImageQuant (Amersham). Nonspecific binding was determined by the binding of ASLs to ribosomes without mRNA and subtracted from the experimental data. In general, the final data is a result of at least three separate experiments, each done with samples in triplicate, ie. minimally nine determinations for each binding with the result in the absence of mRNA subtracted. Data were analyzed using single-site, nonlinear regression (Prism, GraphPad Software, Inc., San Diego, CA).⁴⁹ Dissociation constants (K_ds) were extracted from the binding curves.

Ribosomal Kinetic Assays

The kinetic assays were conducted in a similar fashion as the ribosomal binding assays described above with the exception that only a single low concentration of ASL was used (0.5 μM) and binding was monitored on a time course. Aliquots of ribosomal reactions were filtered after: 0.0, 0.25, 0.5, 1, 5, 10, 15, 20, 30 minutes. The nitrocellulose was washed, dried, and processed as described.

Bovine Mitochondrial 55S Binding Assays

Bovine mitochondrial, 55S, ribosomes were isolated and purified as previously described.¹⁰ Bovine 55S mitochondria were programmed with a chemically synthesized 30-mer oligonucleotide. Mammalian mitochondrial mRNAs are leaderless and translation initiation begins at the first nucleoside triplet on the mRNA. Therefore, to study binding at the P-site, we designed the mRNAs (see the following sequences) with an AUG or AUA (in bold) as the first three nucleosides, followed by the valine codon GUA (underlined) at the A-site, and a stretch of poly-Uridines:

1. 5′-AUGGUAUUUUUUUUUUUUUUUUUUUUUUUU-3′

2. 5′-AUAGUAUUUUUUUUUUUUUUUUUUUUUUUU-3′.

The mRNAs of mammalian mitochondria do not consist of a ribosomal binding sequence such as the Shine-Dalgarno sequence found in bacterial mRNAs. However, an mRNA consisting of stretches of poly(U) with the mitochondrial methionine codon AUG has been shown to be successful in translating methionine using an mitochondrial translational system in vitro.^10,14 Therefore, this poly(U)-methionine codon mRNA can be incorporated into the mitochondrial 55S ribosome. To study the binding of the hmtASL^Met_CAU-Ψ₂₇;f⁵C₃₄ to the A-site, we incorporated the valine codon GUA (intended for the P-site; underlined in the following mRNA sequences), followed by AUG or AUA (in bold) into a string of 24 Uridines:

1. 5′-UUUUUUUUUUUUGUAAUGUUUUUUUUUUUU-3′

2. 5′-UUUUUUUUUUUUGUAAUAUUUUUUUUUUUU-3′.

Full length, native bovine mtRNA^Met_CAU and its transcript were aminoacylated and formylated.¹⁴ The IF2-facilitated binding of the full length tRNAs to AUG and AUA in the P-site was conducted as previously described.^73,74

Abbreviations

f⁵C₃₄

5-formylcytidine at position 34

ASL

anticodon stem and loop

NOE

nuclear Overhauser effect

NOESY

nuclear Overhauser spectroscopy

COSY

correlation spectroscopy

TOCSY

total correlation spectroscopy

HSQC

heteronuclear single quantum correlation

RNA

ribonucleic acids

tRNA

transfer ribonucleic acids

mRNA

messenger RNA

ppm

part per million

RMSD

root mean square deviation

NMR

nuclear magnetic resonance