Amber suppression in Escherichia coli by unusual mitochondria-like transfer RNAs

Abstract

The “cloverleaf” base-pairing pattern was established as the structural paradigm of active tRNA species some 30 years ago. Nevertheless, this pattern does not accommodate the folding of certain mitochondrial tRNAs. For these recalcitrant tRNAs, we have proposed structures having from 5 to 10 base pairs in the anticodon stem rather than the canonical 6. The absence of these types of tRNAs in cytoplasmic translation systems, however, raises the possibility that they may not be bona fide alternate folding patterns for active tRNA molecules. For this reason, we have designed new tRNA genes based on our model of unusual mitochondrial tRNAs, having 7, 8, 9, and 10 base pairs in the anticodon stem with other modifications to the D-stem and connector regions. We show here that these synthetic genes produce tRNAs that actively suppress amber codons in vivo.

The vast majority of tRNAs encoded in the genomes of all cell types, chloroplasts, mitochondria (mt), and viruses fold into the standard cloverleaf secondary structural pattern. Some mt tRNA sequences are unusual, however, in that they do not fit this pattern. Structures for these tRNAs have been proposed based on optimal base pairing patterns and the hypothesis that the distance between the anticodon and acceptor stem must be identical in all tRNAs, so that they can extend themselves in the same way between the messenger RNA and the site of aminoacyl transfer on the ribosome (1, 2). These patterns involve a double zipper principle, where the 5–10 base pairs in the anticodon stem vary inversely with the number of base pairs in the D-stem and the length of the connector regions (Fig. 1). These changes are compensatory and result in three-dimensional structures for tRNAs, which are virtually superimposable on the normal three-dimensional “L” structure in spite of their unusual secondary structure. Other unusual features of mt translation including abbreviated ribosomal RNAs (3–5) and nonuniversal genetic codes (6–8) etc. raise the issue that these tRNAs may be simply another anomaly of mt rather than representatives of alternate folding patterns, because they are not found in cytoplasmic translation. We show here, to the contrary, that tRNAs incorporating these unusual features actively suppress amber mutations in Escherichia coli, and covariation analysis of their sequences supports the patterns that we have assigned to the unusual mt tRNAs.

Figure 1.

The base-pairing pattern of mitochondrial tRNAs. This representation inspired by the three-dimensional L structure of the tRNA shows schematically how changes in the number of base pairs in the anticodon stem and those of the D-stem and the number of nucleotides in the connector regions can be compensatory (1–5). Each black or white circle represents a nucleotide, and pairing is indicated by a bar between positions. Alternate structures are produced by attaching the connector regions to the two nucleotides indicated in the boxes from 5 to 10. These numbers also represent the number of base pairs in the structure resulting from the attachment. AN refers to the anticodon loop and adjacent stem; D, to the D-loop and -stem; and T, to the T-loop and -stem.

MATERIALS AND METHODS

Strains.

Three E. coli strains have been used: Top10 (F⁻ mrcA Δ(mrr-hdsRMS-mcrBC) Φ80lacZΔM15 ΔlacX74 deoR recA1 araD139 Δ(ara-leu)7697 galU galK rpsL endA1 nupG) from Invitrogen; XAC-1 (F′ lacI₃₇₃lacZ_{μ118 am} proB⁺/F⁻ Δ(lac-proB)_XIII nalA rif argE_am ara) (9); and XAC/A16 (ΔlacproB nalA rif argE_am/F′ lacIq amber-Z fusion proB) containing the pDa3am plasmid (10).

Construction of the Combinatorial Library.

The template oligonucleotide coding for the combinatorial tRNA library and two flanking primers with restriction enzyme sites for EcoRI and PstI (sequences on request) were synthesized by General Synthesis and Diagnostics (Toronto). These were then PCR-amplified for 5 min at 95°C, followed by 30 cycles of: 30 s at 94°C, 30 s at 42°C, and 30 s at 72°C, with Vent DNA polymerase (2 units)/100 pmol of each primer/20 mM of dNTPs/100 ng of the template. All enzymes were from New England Biolabs. The double-stranded DNA obtained was then digested with PstI and EcoRI, purified on a Sephadex G-50 column, and cloned into the pGFIB-I plasmid, using a 3:1 insert-to-plasmid ratio and T4 DNA ligase. The plasmid had been predigested with the same enzymes and dephosphorylated with calf intestine alkaline phosphatase by using a 3:1 insert:plasmid ratio and T4 DNA ligase. Electroporation in the TOP10 strain yielded over 500,000 colonies (four times the sequence complexity of the library). Plasmid DNA from 12 randomly selected clones was isolated, and the sequences of the encoded tRNA genes confirmed the randomized nature of the expected positions. Plasmid DNA from this library was prepared by extracting the DNA from all 5 × 10⁵ colonies by using the alkaline lysis protocol (11). This preparation was used to transform cells of the XAC-1 strain. Cells were plated on LB medium with 50 μg/ml ampicillin and 20 μg/ml 5-bromo-4-chloro-3-indolyl β-d-galactopyranoside (X-Gal) and left to grow overnight at 37°C. Blue colonies developed during a 24-h incubation at 4°C. Plasmids from the blue colonies were isolated and used to retransform XAC-1 to ensure that the phenotype was dependent on the presence of the plasmid. All other protocols unless otherwise mentioned were carried out according to Sambrook et al. (11).

tRNA Levels and Aminoacyl-tRNA Levels.

The tRNA and aminoacyl-tRNA levels were measured by using an adaptation of previous protocols (12, 13). Cells from a fresh transformation of XAC-1 with isolated plasmids were grown in 10-ml cultures by using 2× YT broth containing 50 μg/ml ampicillin. During log phase, cells were harvested by centrifugation at 4°C. All subsequent steps were carried out at 4°C. The pellet was resuspended in 0.5 ml of 0.3 M sodium acetate (pH 5.2) and 1 mM EDTA and extracted with an equal volume of phenol equilibrated with 0.3 M sodium acetate (pH 5.2). After vortexing for 1 min and 10 min on ice, the mixture was centrifuged to isolate the aqueous layer, which was subjected to precipitation on dry ice with 2 vol of ethanol. The pellet from centrifugation was washed with a 0.5-ml vol of 70% ethanol/10 mM sodium acetate (pH 5.2) and resuspended in 40 μl of 10 mM sodium acetate (pH 5.2)/1 mM EDTA. The concentration of the RNA isolations is determined according to the absorbance of a 1:1000 dilution. Samples of 2–4 μg of each RNA preparation were distributed in two tubes, one of which was maintained at 4°C, whereas 1.5 μl of 0.5 M Tris (pH 9) was added to the other and incubated at 37°C for 25 min. To each sample, 1.5 μl of loading buffer containing 0.1 M sodium acetate (pH 5.2), 8 M urea, 0.05% bromophenol blue, and 0.05% xylene cyanol was added to the samples, and they were loaded on a 6.5% polyacrylamide gel containing 8 M urea, 0.1 M sodium acetate (pH 5.2) and run overnight at 300 V at 4°C. After electrophoresis, RNA was transferred by electroblotting the portion of the gel around the xylene cyanol on a nylon membrane (Hybond-N, Amersham, Buckinghampshire, UK) and hybridized with a probe corresponding to the anticodon stem-loop of the tRNA library and with a probe for 5S RNA (positions 34–53 in the E. coli 5S sequence).

Aminoacylation levels were obtained from the quantity of aminoacyl-tRNA and uncharged tRNA in lanes that had not been treated with Tris by scanning autoradiograms of the gels and evaluating the intensity with the Scion Image PC program (Beta 1 release, 1997). tRNA levels were calculated by scanning the autoradiograms and normalizing to the quantity of 5S RNA in the same lane.

Identification of the Charged Amino Acid.

The identification of the amino acid charged on the tRNAs from clone T7 and T59 was accomplished by their introduction into strain XAC/A16 carrying the plasmid pDa3am, which contains an amber codon at the third position of the dihydrofolate reductase (DHFR) gene. DHFR was isolated on a methotrexate resin (Sigma) according to the protocol of McClain and Foss (10) and Normanly et al. (14) and subsequently eluted with folic acid. Separation of the folic acid from DHFR was done with a Bio-Rad Econopac Q column with a gradient from 0 to 1 M KCl. The fractions with DHFR were pooled and desalted on Centricon-30 (Amicon). DHFR was microsequenced in the laboratory of C. Lazure (Institut de recherche clinique de Montréal) after SDS/15% PAGE and transfer to a PVDF membrane (Amersham).

RESULTS

Experimental Design.

Although the activity of these unusual structures in protein synthesis was the main purpose of this work, we also wished to evaluate the structural inferences that have been made to rationalize how the unusual tRNAs could be fitted to the standard three-dimensional structure (1, 2, 15). Thus, we adopted the strategy to incorporate the unusual features into the background of a completely different, nonmitochondrial tRNA as a stringent test of our knowledge. Because the conversion of a normal tRNA molecule to the mt type requires modification of the D-stem and connector regions, a tRNA must be selected whose identity determinants are elsewhere so that amino acid charging ability and specificity would not be jeopardized. The E. coli tRNA_UGC^Ala gene was chosen for modification because its main identity element, the G3-U70 base pair, is located in the acceptor stem (16–18). Fig. 2 shows the modifications of the E. coli tRNA_UGC^Ala gene to incorporate the following mt characteristics. 1) The G26⋅A44 base pair was changed to G26⋅C44 in the designed gene. Because the geometry of this interaction may be a factor in arresting base pairing in the anticodon stem, rendering it a Watson–Crick interaction could favor extension of the stem (19). 2) G10 was deleted to eliminate the 10⋅25 base pair and to favor the formation of a 25⋅45 pair. Position 25 was represented in the designed gene as either an A, a common nucleotide at the last position of the anticodon stem, or a C to encourage propagation of the stem. 3) T8 was deleted from the gene sequence to shorten the connector region between the acceptor stem and the D-stem. This change plus deletion of G10 guarantees that tRNAs from this gene could not have a standard cloverleaf secondary structure (1, 15). 4) Positions 13, 14, 15, 21, and 22 of the D-domain were randomized to permit all combinations of nucleotides at these positions, which could be sensitive to both secondary and tertiary interactions. 5) The unpredictability of interactions at strategic positions of the variable loop prompted the additional randomization of positions 46, 47, and 48. 6) Finally, the TGC anticodon of tRNA^Ala gene was converted to CTA to permit reading of the UAG amber codon. Judicious placement of this codon in a reporter gene allows simple evaluation of tRNA in vivo activity by the well known amber suppression test. Moreover, amber mutations in both the argE and the lacI/lacZ genes of the XAC-1 strain of E. coli prompted its use in this test: suppression would render the strain prototrophic for arginine and blue in the presence of X-Gal (9).

Figure 2.

Modifications to the tRNA^Ala gene sequence. The base pair G26⋅A44 was converted to the Watson–Crick pair G26⋅C44. T8 and G10 were deleted. Nucleotide 25 was encoded by a mixture of A and C. Positions 13–15 and 21–22 of the D-domain as well as positions 46–48 of the variable loop were randomized, and the anticodon TGC was replaced by CTA. Ts are used here, because the gene sequence is given.

Clones Selection and Characterization.

The tRNA gene library was prepared by PCR amplification of synthetic DNA encoding the tRNA gene with eight randomized (4 and 5 above) and one binary position (2 above) producing a library having a sequence complexity of 1.3 × 10⁵. The amplification product was cloned into the pGFIB-I plasmid (9) and then used to transform the XAC-1 strain (see Materials and Methods). Plasmids from blue colonies were isolated and utilized to retransform XAC-1 to ensure that the suppressor phenotype was not a result of a host mutation. Fig. 3A shows the plate growth of several isolated clones in the presence and absence of X-Gal and in the absence of an arginine growth supplement. Generally, darker colored colonies also had better growth characteristics in the medium lacking arginine, suggesting a certain correlation between the suppression levels of the two amber codons.

Figure 3.

Properties of active tRNA clones. (A) Colony growth of different tRNA clones in XAC-1. Newly transformed colonies were replated on M9 minimal medium with ampicillin (50 μg/ml), without arginine, and with or without X-Gal (20 μg/ml). Ala represents the plasmid containing the tRNA_su+^Ala gene; Phe, the tRNA^Phe suppressor gene; and “−,” the plasmid with no insert. The top image is the clone index; the middle shows colony growth in the presence of X-Gal with no added arginine; and the bottom shows growth in the absence of X-Gal and arginine. (B) Aminoacylation levels of various tRNA clones. Total RNA was isolated from different clones. One sample of each was treated with Tris (pH 9) and then analyzed with a second, untreated sample on a 6.5% polyacrylamide gel at pH 5.2 (see Materials and Methods). Lanes containing samples which were treated with Tris are indicated by “+,” and those which were untreated are indicated by “−.” Northern blot hybridization was carried out with a probe representing the anticodon part of the tRNA library and with a 5S probe. The percentage of aminoacylation was calculated by determining the intensity of the bands representing the aminoacylated and nonaminoacylated species. The levels of expression were determined in the same way and normalized to the level of the 5S RNA in each lane.

Thirty-three plasmids demonstrated reproducible suppression of the amber mutations, and their characterization included the determination of the β-galactosidase activity of cellular extracts and the sequence of the plasmid-borne tRNA genes (Table 1). Suppression varied between 0.3 and 40% of the tRNA_su+^Ala control. Certain tRNA genes contained deletions (T59 and T39) or insertions of nucleotides (T28, T42, etc.) not present in the original gene.

Table 1.

β-Galactosidase activities and gene sequences of suppressor tRNAs

β-Galactosidase activity was determined using the method of Miller (21) in the presence of 50 μg/ml ampicillin and the use of chloroform and SDS. The activity is presented as a percentage of the activity of the control tRNA_su+^Ala. The same plasmid without a tRNA insert gave 0.075% activity. tRNA genes were sequenced using the Sanger method by the Organellar Genome Megasequencing Project Laboratory of the Université de Montréal under the direction of G. Burger. Sequences were grouped depending on the predicted number of base pairs in the anticodon stem. Type 7-1 and 7-2 tRNAs differ in the number of nucleotides (1 or 2) in the Connector 1. ∗ indicates the presence of a bulged nucleotide in the anticodon stem (15). Nucleotides in the randomized positions are in bold; and those unexpected from the original design are underlined. Noncanonical base pairs are in italics. These sequence data have been submitted to the GenBank database under accession numbers AF003201–AF003233. AC refers to the aminoacceptor stem; D, to the D-stem; AN, to the anticodon stem; and T, to the T stem. 

tRNAs encoded in these plasmids were further characterized by the determination of their in vivo aminoacylation level and their level of expression. Fig. 3B is a representative autoradiogram of a polyacrylamide gel run at acid pH, demonstrating the amino acid charging level of the unusual tRNAs and the control tRNA_su+^Ala. Analysis of these data showed that aminoacylation of all tRNAs was between 70 and 90 ± 5% (Table 2). Although generally the charging level of tRNA_su+^Ala was greater than the variant tRNAs, these differences were within the experimental error. The steady state level for all tRNAs was also calculated by using the amount of 5S RNA as an internal standard. The levels of the variant tRNAs were found to be consistently lower than that of the control tRNA (Table 2). The specific activity of these tRNAs (that is the level of activity divided by quantity of aminoacyl-tRNA) is as much as three times the value of the control. This value must be taken with some caution, however, because suppression is saturable due to the negative effect of excessive read-through of stop codons. On the other hand, the gel mobility of the charged and noncharged tRNAs is consistent with the predicted length, indicating that nuclease processing of the precursor tRNA was not disrupted by the unusual structure of the tRNAs.

Table 2.

Levels of expression and aminoacylation

Clone	Type*	% of activity†	% of aminoacylation‡	% of expression§
Ala	6	100	91  ± 4	100
46	7-1	10  ± 0.7	86  ± 10	5  ± 3
49	7-1	6  ± 0.3	83  ± 13	13  ± 12
43	7-2	23  ± 0.7	96  ± 6	6  ± 2
25	8	4  ± 0.6	86  ± 14	9  ± 7
7	8	11  ± 3.8	66  ± 24	12  ± 3
39	9	7  ± 0.1	90  ± 10	15  ± 8
37	10	2  ± 0.2	84  ± 14	22  ± 15

^*Structural type defined in Fig. 4. 

^†Obtained by averaging β-galactosidase activity from three independent cellular extracts (21). 

^‡Determined by scanning the first lane of each clone in the autoradiogram of acid polyacrylamide gels (Fig. 3B). These values are the average of three independent experiments. 

^§Obtained from the total tRNA (tRNA + aminoacylated tRNA) in each of the two lanes of the clones. The quantity of tRNA in each band was first normalized to the quantity of 5S RNA in the same lane (internal standard) and expressed as a percentage of the tRNA_su+^Ala control (Fig. 3B). These values were obtained from three experiments. 

The identity of the amino acid inserted at the site of the amber codon was determined by using the pDa3am plasmid, which contains an amber mutation at the third codon of the encoded DHFR gene. The XAC/A16 strain containing the pDa3am plasmid was transformed with the plasmid bearing either the T7 or the T59 tRNA clone. These tRNA clones were chosen because they were associated with high β-galactosidase activity and represented vastly different structural groups (see Table 1). DHFR was isolated on a methotrexate resin, and microsequencing of the protein from both transformants showed the presence of alanine at the position of the nonsense codon.

Sequence Analysis.

In all, 22 of the 33 tRNA gene sequences had 7 base pairs in the anticodon stem and one or two connector nucleotides (Type 7-1 and 7-2, Fig. 4). For the remaining 11 tRNAs, the optimal base pairing produces structures with 8, 9, and 10 base pairs in the anticodon stem (Fig. 4). All tRNAs fold into patterns that have been proposed previously for mt tRNAs (1, 2). None of the 33 tRNA sequences could be folded into the common cloverleaf pattern.

Figure 4.

Secondary structure of isolated tRNAs. The optimal secondary structure patterns are shown for the type 7-1, 7-2, 8, 9, and 10 tRNAs, which were isolated in these experiments. Types 7-1 and 7-2 are shown with the covariation data from the D-stem of these tRNAs. None of these sequences can be folded into the canonical cloverleaf structure.

The top of Fig. 4 focuses on the analysis of nucleotide covariation in the positions of the D-stem that were randomized in the original library. For Type 7-1 tRNAs, the third and fourth pairs of the D-stem, positions 13–22 and 14–21, are Watson–Crick pairs in 12 and 10 cases of 13, respectively. In all nine Type 7-2 tRNAs, the three D-stem base pairs are Watson–Crick base pairs. Because an insufficient number of unusual mt tRNA sequences is available in the sequence database to establish the base pairing pattern we have proposed, our data help to validate the base pairing pattern of at least the Type 7 tRNAs. Moreover, the fact that randomized positions in the synthetic library produced highly base paired D-stems testifies to the importance of this stem in active tRNA species.

DISCUSSION

Because mt may be more tolerant to slower and/or less precise protein synthesis, we had anticipated that the mt-like tRNAs may not be active in cytoplasmic translation. Our suggestion that eukaryotes and archaea are protected from the alternate, mt-type, folding pattern of certain tRNAs by dimethylation of a particular guanosine (19) also raises the spector that these tRNAs would be detrimental. In contrast, our data show that these tRNAs are efficient suppressors (Table 2), although it is true that high speed and/or high precision protein synthesis in response to a normal codon has not been demonstrated. Only the state level of these tRNAs seems deficient, possibly reflecting instability or poor processing. On the other hand, efficient suppressors might only exist at low levels, because otherwise they would interfere with normal termination of translation. Also, the acid gel experiment shown in Fig. 3B demonstrates that aminoacylation and maturation of the 5′- and 3′-termini are normal.

The fact that these mt-like tRNAs function well as suppressor tRNAs in the cytoplasmic protein synthesis begs the question of why they are not normally found in cytoplasmic systems. Of course, our data address only the effect of the presence of these tRNAs during a few generations; the true evolutionary issue deals with the long term effect on the survival of the cell. It might be possible, however, that the conformational space of the unusual tRNAs was never explored during the evolution of the cytoplasmic protein synthesis system. We think not because: (i) the origin of mt is rooted in the α-purple bacteria, and the unusual tRNAs must be considered as derived from normal structures rather than remnants of primordial molecules; (ii) the simple mechanism for the production of the unusual tRNAs has been proposed, which involves only slippage of base pairing in the D-stem area (1); (iii) the distribution of these tRNAs among distantly related mt clearly demonstrates several independent origins for this class of tRNAs (15). Therefore, this structure must have originated several times during the evolution of cytoplasmic translation, but then it must have been rejected systematically.

Our ability to prepare active tRNAs having unusual folding motifs in light of previous attempts (refs. 22 and 23 and V. Bourdeau et al., unpublished results) could be because of our gene design strategy and the use of randomized positions in the unusual gene sequence such that many sequence variants could be screened simultaneously. In any case, these experiments demonstrate that the classic cloverleaf pattern of tRNA is not a necessary condition for tRNA activity even in cytoplasmic protein synthesis and that the self-compensating structures that we have proposed for the unusual mt tRNAs more faithfully represent the structure/function paradigm of active tRNA species.