Mutations which alter the elbow region of tRNA2Gly reduce T4 gene 60 translational bypassing efficiency

Alan J. Herr, John F. Atkins, Raymond F. Gesteland

Author Affiliations

  1. Alan J. Herr1,
  2. John F. Atkins1 and
  3. Raymond F. Gesteland*,1
  1. 1 Department of Human Genetics, 15N 2030 E RM 6160, University of Utah, Salt Lake City, UT, 84112‐5330, USA
  1. *Corresponding author. E-mail: rayg{at}


Translating ribosomes bypass a 50 nucleotide coding gap in bacteriophage T4 gene 60 mRNA between codons 46 and 47 in order to synthesize the full‐length protein. Bypassing of the coding gap requires peptidyl‐tRNA2Gly detachment from a GGA codon (codon 46) followed by re‐pairing at a matching GGA codon just before codon 47. Using negative selection, based on the sacB gene from Bacillus subtilis, Escherichia coli mutants were isolated which reduce bypassing efficiency. All of the mutations are in the gene for tRNA2Gly. Most of the mutations disrupt the hydrogen bonding interactions between the D‐ and T‐loops (G18●ψ55 and G19●C56) which stabilize the elbow region in nearly all tRNAs. The lone mutation not in the elbow region destabilizes the anticodon stem at position 40. Previously described Salmonella typhimurium mutants of tRNA2Gly, which reduce the stability of the T‐loop, were also tested and found to decrease bypassing efficiency. Each tRNA2Gly mutant is functional in translation (tRNA2Gly is essential), but has a decoding efficiency 10‐ to 20‐fold lower than wild‐type. This suggests that rigidity of the elbow region and the anticodon stem is critical for both codon–anticodon stability and bypassing.


Unerring protein synthesis depends on correct tRNA–mRNA pairing. Cognate codon–anticodon interactions are stabilized by the ribosome in both the A‐site and the P‐site in order to minimize translation errors. While stabilization in the A‐site ensures incorporation of correct amino acids in the polypeptide chain, stabilization in the P‐site helps the ribosome maintain the correct reading frame between translocation and the next decoding step. Yet some genes require a temporary destabilization of peptidyl‐tRNA–mRNA pairing in the ribosome for proper expression. These genes carry cryptic signals in their mRNAs which facilitate peptidyl‐tRNA detachment from a specific codon and re‐pairing at an alternate identical or synonymous codon (Atkins and Gesteland, 1995; Farabaugh, 1996). Since in most cases the alternate codon overlaps the first codon, the result is a programmed change in reading frame [examples include the +1 frameshift found in RF2 (Craigen and Caskey, 1986; Weiss et al., 1988) or −1 frameshifts found in many retroviruses (Brierley, 1995; Gesteland and Atkins, 1996)]. But in some cases, called translational bypassing, the peptidyl‐tRNA re‐pairs at a non‐overlapping codon. The existence of prokaryote examples (Weiss et al., 1987; Huang et al., 1988; O'Connor et al., 1989; Kane et al., 1992; Manch‐Citron and London, 1994; Gallant and Lindsley, 1998), along with a recent report showing bypassing in the expression of rabbit β‐globin (Chittum et al., 1998), indicates that this phenomenon may be common to most organisms.

The best characterized example of translational bypassing is in gene 60 of bacteriophage T4 (Weiss et al., 1990), encoding a topoisomerase subunit. In this case, the peptidyl‐tRNA (tRNA2Gly) detaches from a GGA codon (codon 46) and re‐pairs at a second GGA codon 47 nucleotides downstream. To stimulate bypassing of this non‐coding information (or coding gap), gene 60 mRNA relies on several cis‐acting signals (Figure 1). These signals include matching GGA codons (referred to as take‐off and landing sites) flanking an optimally sized coding gap, a stop codon immediately downstream of the take‐off site, a stem–loop structure in the 5′ end of the coding gap, and a stretch of basic and hydrophobic amino acids translated from codons preceding the coding gap.

Figure 1.

Efficient translational bypassing of the gene 60 coding gap requires matching take‐off and landing site GGA codons, a stop codon, a stem–loop structure, and a nascent peptide signal. In the current model, peptidyl‐tRNA2Gly dissociates from the take‐off site GGA, scans the mRNA as it moves through the decoding center, and re‐pairs with the landing site GGA.

Modulation of peptidyl‐tRNA2Gly pairing with the crucial GGA codon is at the center of the proposed model for bypassing: peptidyl‐tRNA2Gly releases from the take‐off site GGA, the mRNA moves through the ribosome, and peptidyl‐tRNA2Gly re‐pairs at the landing site GGA. Peptidyl‐tRNA2Gly appears to locate the landing site by scanning the coding gap for a cognate codon as the mRNA moves through the ribosome (F.M.Adamski, M.B.Moore, R.F.Gesteland and J.F.Atkins, in preparation). Recent estimates show that 50% of ribosomes bypass the coding gap (productive bypassing) (Maldonado and Herr, 1998). The other 50% either terminate at the UAG codon just after the take‐off site, resume translation in an unproductive reading frame, or abort landing by releasing the mRNA or the peptidyl‐tRNA (unproductive bypassing).

How the combination of signals stimulates bypassing remains a mystery. One possible function for the signals is to switch the ribosome into a scanning mode which allows the mRNA to move through the decoding center. Another possibility is that the signals target components of the translational apparatus which stabilize peptidyl‐tRNA pairing. Previous genetic studies identified mutations in the C‐terminal domain of ribosomal protein L9 which suppress mutations in the stem–loop and nascent peptide, suggesting that perhaps L9 is the target of these signals (Herbst et al., 1994; Adamski et al., 1996). Here, we report a novel set of mutations in tRNA2Gly which reduce the bypassing efficiency of wild‐type gene 60.


We developed a genetic approach to identify features of the translational apparatus required for bypassing. The strategy was to isolate mutants of Escherichia coli which bypass the coding gap less efficiently using a selection based on the sucrose sensitivity conferred by the sacB gene from Bacillus subtilis (Gay et al., 1985).

Negative selection based on the sucrose sensitivity conferred by sacB has been used in Gram‐negative bacteria to select for allelic exchange (Ried and Collmer, 1987; Blomfield et al., 1991), to select for plasmids containing inserts (Pierce et al., 1992), to capture insertional sequences (Gay et al., 1985), and to create counterselectable transposons for use in mapping of chromosomal mutations (Lawes and Maloy, 1995). The product of sacB, levansucrase, releases glucose from sucrose by a transfructosylation reaction in which the byproducts are levans—large branched fructosyl polymers (Steinmetz et al., 1985). When expressed in E.coli, levansucrase is exported to the periplasmic space and appears to cause membrane fragility (Gay et al., 1983). We found that levansucrase requires a signal peptide at its N‐terminus in order to confer sucrose sensitivity (data not shown), consistent with disruption of membrane integrity by levans.

For a selection in which efficient gene 60 expression was lethal, we created a gene 60–sacB fusion with the E.coli phoA signal peptide coding sequence fused in‐frame and upstream of gene 60. This cassette was placed under control of an isopropyl‐β‐d‐galactopyranoside (IPTG)‐inducible tac promoter on a pBR322‐based plasmid. Selection for surviving mutants was carried out in a strain which had a constitutively expressed gene 60–lacZ fusion on the chromosome conferring blue color to colonies grown on plates with X‐gal. Using this combination of gene fusions, we could first select for spontaneous mutants with reduced sacB expression and then screen on the basis of color intensity for those with lower bypassing efficiency.

We optimized selection conditions by varying sucrose and IPTG concentrations in low salt, tryptone‐casamino acid broth plates until a 3‐ to 5‐fold drop in bypassing efficiency could be detected. Using these selection conditions we screened for light blue or white colonies that survived after 24–48 h of growth. The frequency of sucrose‐resistant cells was 1×106. Of these, one in 100 gave rise to light blue or white colonies. After confirming the color phenotype, 13 independent mutants were transformed with plasmids that allowed quantification of bypassing efficiency. Five mutants displayed a 3‐ to 5‐fold reduction in bypassing efficiency (data not shown) and were designated byp mutants.

All five mutants mapped to glyT, the sole gene encoding tRNA2Gly. Sequencing PCR products amplified from each mutant provided confirmation. Four mutations are located at the universally conserved positions G18, G19 and C56 (Figure 2). One mutation is in the anticodon stem at C40. Thirteen additional independent mutants were isolated: eight were at C56 and the rest were at G18. No mutations were identified in any gene besides glyT.

Figure 2.

Identification of byp mutants. Secondary structure prediction of WT tRNA2Gly showing positions of byp mutants.

These results are intriguing because in nearly all tRNAs, G18, G19, ψ55 and C56 stabilize the elbow region by forming base pairs between the dihydrouracil (D) and ribothymidine (T) loops (Dirheimer et al., 1995). G19 forms a Watson–Crick pair with C56, and G18 pairs with ψ55 through a single hydrogen bond. Importantly, the mutant versions of tRNA2Gly that reduce bypassing efficiency must retain the ability to function in translation because the GGA decoding ability of tRNA2Gly is required for growth of E.coli (Murgola and Pagel, 1980).

We chose the byp mutants G18A (G18 to A), G19U, C40G and C56A for further characterization. Each was transduced into a fresh background (CSH142; Table I) and bypassing efficiency and growth rates were determined. To measure bypassing efficiency, transductants were transformed with a glutathione S‐transferase (GST)–gene 60 fusion plasmid which allows detection of proteins due to both productive and unproductive bypassing (Maldonado and Herr, 1998). Expression of these proteins was assayed by pulse–chase labeling of total protein with [35S]methionine followed by separation on an SDS–polyacylamide gel (Figure 3A). All mutants bypass the coding gap at only 20–30% of wild type (WT) confirming that the byp phenotype is due to the mutant glyT alleles (Figure 3B). The growth curves and doubling times (28 min) of G18A and G19U are virtually indistinguishable from WT (data not shown). C40G and C56A also grow at nearly the same rate as WT (28–30 min) but appear to enter stationary phase at a lower density.

Figure 3.

Characterization of byp mutants. (A) Pulse–chase analysis of isogenic WT and byp mutant strains expressing a GST–gene 60 fusion. The lanes labeled ‘−’ and ‘+’ are negative (without the plasmid) and positive (expressing only GST) controls, respectively. The GST–gene 60 fusion and products due to productive (45.1 kDa) or unproductive bypassing (31.7 kDa) are represented schematically to the right of the autoradiograph. (B) The average bypassing efficiency determined from three separate experiments.

View this table:
Table 1. Bacterial strains

All of the mutants described here are predicted to disrupt base‐pairing interactions within tRNA2Gly. To confirm that the loss of base pairing was responsible for the mutant phenotypes, compensatory mutations which restore base pairing were made in three of the mutants (Figure 4). Over‐expression of tRNA2Gly single mutants [G19U (pLG19U), C56A(pLC56A) or C40G(pLC40G)] does not overcome the defect in bypassing in strains carrying a chromosomal byp mutation. In contrast, double mutants carrying the compensatory changes correct the defect in bypassing. G19U/C56A(pLcom2) restores bypassing efficiency to ∼80% of WT in both G19U and C56A backgrounds. (The lack of complete restoration may reflect either a residual defect in bypassing or competition with the chromosomal byp mutant.) G30C/C40G(pLcom1) restores bypassing efficiency to ∼100% of WT in a C40G background. These results show that the structural integrity of the tRNA is critical for bypassing.

Figure 4.

byp mutants carrying compensatory mutations function normally in bypassing. Plasmids and the over‐expressed tRNA2Gly variant: pLglyT, WT; pLG19U, G19U; pLC56A, C56A; pLcom2 G19U/C56A double mutant; pLC40G, C40G; pLcom1, G30C/C40G double mutant. Top panels show pulse–chase analysis of the strains carrying both tRNA over‐expression plasmids and a GST–gene 60 fusion (see Figure 3). Bottom panel shows the average bypassing efficiency determined from three separate experiments.

There are mutants of Salmonella typhimurium in glyT that promote −1 frameshifting at the sequence G GGA (Riyasaty and Atkins, 1968; O'Mahony et al., 1989a; Pagel et al., 1992). These recessive alleles, called sufS for suppressor of frameshift mutation, include the mutations U34C, 60+U (an extra U between U59 and U60), C61U, C62A and G1A (Figure 5A). We tested these mutations for altered bypassing efficiency (Figure 5B). U34C and G1A reduce bypassing efficiency to 70% of WT while 60+U, C61U and C62A reduce bypassing efficiency to nearly the same extent as byp mutants (30–40% of WT) (Figure 5C). Those mutants with T‐stem or T‐loop defects (60+U, C61U, C62A) may be analogous to the byp mutants which disrupt the elbow region. Previous work showed 60+U and C61U variants are present at low abundance and aminoacylated poorly (O'Mahony et al., 1989b). This suggests that aminoacyl‐tRNA2Gly levels may influence bypassing efficiency.

Figure 5.

sufS mutants and bypassing efficiency. (A) Secondary structure prediction of WT tRNA2Gly showing positions of sufS mutants. (B) Pulse–chase analysis of S.typhimurium trpE91 WT and sufS variants expressing a GST–gene 60 fusion. (C) The average bypassing efficiency determined from three separate experiments.

To determine whether aminoacyl‐tRNA2Gly levels are correlated with the decrease in bypassing, we measured tRNA abundance by Northern analysis of tRNA preparations purified and separated under conditions which preserve the aminoacyl‐tRNA bond and separate aminoacyl‐tRNA from deacyl‐tRNA. A sample subjected to base‐treatment (which hydrolyzes the aminoacyl bond) is analyzed alongside each untreated sample to provide a size marker for the faster migrating deacyl‐tRNA. Our analysis of tRNA2Gly levels in WT, G18A, G19U, C56A, 60+U and C61U preparations reveal two species of aminoacyl‐tRNA2Gly with different electrophoretic mobilities (Figure 6A). C40G and U34C preparations, by contrast, show only the faster migrating form of aminoacyl‐tRNA2Gly. The most likely explanation for the two species is the presence or absence of the modified nucleotide U* at position 34. This is consistent with U34C showing a single species and with a partial characterization of U* determined during the sequencing of tRNA2Gly (Roberts and Carbon, 1975). This characterization showed U* was similar to 2‐thio‐5(methylaminomethyl) uridylic acid and that U* had a lower electrophoretic mobility than uridine. Furthermore, the authors noted in some preparations of tRNA, only 70% of tRNA2Gly carries U*, which is consistent with the ratios of the two products we see in most samples. The observation that C40G preparations lack U* suggests the stability of the anticodon stem may be important for modification.

Figure 6.

Aminoacyl‐tRNA2Gly levels in byp and sufS mutants. (A) Northern blot analysis of total tRNA purified and separated under acidic conditions. Strains: for complete genotypes see Table I. Abbreviations: aa‐tRNA*, modified aminoacyl‐tRNA2Gly; aa‐tRNA, unmodified aminoacyl‐tRNA2Gly; tRNA*, modified deacyl‐tRNA2Gly; tRNA, unmodified deacyl‐tRNA2Gly. Deacylation treatment: ‘−’, untreated sample; ‘+’ treated with Tris pH 9.0 prior to analysis. An oligonucleotide complementary to nucleotides 27–54 of tRNA2Gly served as the probe. (B) The average tRNA levels in each strain (determined from untreated samples) from three separate experiments normalized to the appropriate WT control. (C) The average estimated percentage of tRNA2Gly which is aminoacylated.

Determining aminoacyl‐tRNA levels is complicated by the U* modification because deacylated modified tRNA2Gly co‐migrates with unmodified aminoacyl‐tRNA2Gly. Therefore, to determine aminoacyl‐tRNA levels, we first measured total tRNA levels and then estimated the percentage which was aminoacylated.

Total tRNA levels were determined by quantifying the amount of hybridization to tRNA2Gly in each sample (modified and unmodified), correcting for loading differences using the amount of hybridization to 5S rRNA, and then normalizing the amount of each mutant tRNA to the wild‐type control (Figure 6B). From this analysis, C40G, C56A and 60+U are present at <20% of WT levels. G18A, G19U and C61U are present at 30–50% of WT levels. U34C (which was assayed in both S.typhimurium and E.coli) showed a substantial amount of variability, but in some assays was as abundant as WT.

The percentage of aminoacyl‐tRNA2Gly in WT, G18A, G19U, C56A, 60+U and C61U preparations was estimated by comparing amounts of modified aminoacyl‐tRNA (Figure 6A, ‘−’ lanes) to base‐treated modified deacyl‐tRNA (Figure 6A, ‘+’ lanes). The unmodified variants (C40G and U34C) were analyzed in a similar fashion. The results (Figure 6C) suggest the percentage of tRNA2Gly which is aminoacylated is similar for both mutants and WT, and indicates that total tRNA levels are a reasonable estimate for aminoacyl‐tRNA levels. The variability in aminoacyl‐tRNA levels alone does not seem to account for the decrease in bypassing efficiency: C56A is 3‐fold less abundant than G19U, yet both have the same decrease in bypassing efficiency.

We investigated whether the decoding ability of the mutants is correlated with the decrease in bypassing efficiency. Decoding efficiency of tRNA2Gly variants with a 3′AUC anticodon was measured using an amber suppression assay. Expression vectors were constructed in which transcription of amber suppressor tRNA2Gly variants was controlled by a temperature‐sensitive λ repressor (Figure 7A). Amber suppression was monitored using a protA‐cat‐lacZ construct interrupted by a UAG codon and normalized to a control with a sense codon replacing the UAG. After 1 h of growth at 40°C to ensure steady‐state levels of suppressor tRNA, lacZ expression was induced. At the end of induction, total tRNA was purified and separated using conditions which preserve the aminoacyl bond. The samples were then subjected to Northern analysis using a probe (Figure 7B) and hybridization conditions under which only amber suppressor tRNA2Gly was detected (Figure 7C).

Figure 7.

Decoding efficiency of tRNA2Gly mutants. (A) Expression construct. (B) Secondary structure prediction of tRNA2Gly amber suppressor [tRNA2Gly (SuUAG)] variants. The bold line represents the oligonucleotide probe used to detect tRNA2Gly (SuUAG) variants. (C) Northern analysis of total tRNA purified and separated under acidic conditions from WT strains over‐expressing tRNA2Gly (SuUAG) variants or WT tRNA2Gly. Deacylation treatment: see Figure 6. (D) Decoding efficiency determination. The area of the each circle corresponds to the normalized activity or concentration of the corresponding tRNA. Estimates represent the average of three separate experiments.

Northern analysis of both base‐treated and untreated C56A reveals a slower migrating product which may reflect a processing defect (Figure 7C). This is consistent with a loss of recognition by RNaseP which makes specific contacts with the T‐stem and loop of pre‐tRNA (Svard and Kirsebom, 1993; Gaur et al., 1996). Alternatively, it may reflect a polyadenylated degradation intermediate (Li et al., 1998). The reduced mutant tRNA levels seen both here and in Figure 7 may indicate an additional processing defect: complete degradation during 3′ processing by 3′→5′ exonucleases (Li and Deutscher, 1996).

Decoding efficiency was calculated by correcting the normalized amber suppression levels by the normalized aminoacyl‐tRNA2Gly (SuUAG) levels (Figure 7D). The results show that all tRNA2Gly variants that reduce bypassing are 10‐ to 20‐fold down in decoding efficiency.

The observation that these tRNA2Gly mutants are reduced in both decoding and bypassing efficiency can be explained by two models. In the first model, the drop in tRNA2Gly decoding efficiency starves ribosomes at GGA codons. As a result, ribosomes stalled just before the take‐off site GGA are diverted from bypassing by competing events such as tRNA1Gly near‐cognate decoding. [tRNA1Gly near‐cognate decoding has been invoked to account for the −1 frameshift activity of 60+U and C61U (O'Connor, 1998; Qian et al., 1998). Ribosomes which shift −1 at the take‐off site would terminate at a UAG within the coding gap. tRNA3Gly (anticodon = 3′CCG) has been shown to read GGA in vitro under conditions of no competition, but competes poorly with tRNA2Gly or tRNA1Gly for GGA codons (Samuelsson et al., 1983).] In the second model, despite the decrease in decoding efficiency, tRNA2Gly still decodes the take‐off site GGA and a reduction in stability in the P‐site leads to a decrease in landing efficiency.

To distinguish between these two models, we first tested whether the reduction in bypassing was due to tRNA1Gly near‐cognate decoding. tRNA1Gly (encoded by glyU) has a 3′CCC anticodon and decodes primarily GGG codons. A glyU(SuUAG) strain, which expresses an amber suppressor variant of tRNA1Gly (anticodon 3′CUA), was transduced to tetracycline resistance with P1 lysates grown on byp mutants carrying a Tn10 linked to glyT (argE86::Tn10). Mutant tRNA2Gly must have sufficient activity to decode all GGA and GGG codons for the double mutants to be viable. Transductional crosses with C40G and C56A mutant showed that these alleles do not meet this requirement (data not shown). No double mutants were obtained even though argE86::Tn10 is 50% linked to glyT. On the other hand, G18A‐glyU(SuUAG) and G19U‐glyU(SuUAG) double mutants were readily obtained. Bypassing efficiency in these strains remains 3‐ to 5‐fold less than WT, showing that tRNA1Gly is not necessary for the byp phenotype of these mutants (Figure 8).

Figure 8.

tRNA1Gly and the byp mutant phenotype. Strains and relevant genotypes are listed to the left (see also Table I). Pulse–chase analysis of the strains expressing a GST–gene 60 is shown to the right.

Additional evidence discounts the starvation‐dependent model. This model predicts that over‐expression of byp mutants may partially suppress the bypassing defects associated with a chromosomal byp mutant. The results from Figure 4 show no indication that over‐expression of G19U, C56A or C40G increases bypassing in a byp mutant strain. The starvation‐dependent model also predicts that over‐expression of byp mutants will have no effect on bypassing in a WT cell. We tested the second prediction by monitoring bypassing in WT cells over‐expressing different tRNA2Gly variants (data not shown). Over‐expression of C40G, G18A or G19U reduces bypassing by 20–30%, arguing against the starvation‐dependent model. Instead, these results support a model in which tRNA2Gly mutants directly interfere with the bypassing mechanism in the P‐site.


The decoding properties of peptidyl‐tRNA2Gly are critical for each stage of gene 60 bypassing. Destabilization of peptidyl‐tRNA2Gly pairing with the take‐off site GGA is required for bypassing initiation. Peptidyl‐tRNA2Gly probing of the coding gap for a cognate interaction is required during scanning, and peptidyl‐tRNA2Gly pairing with the landing site GGA is necessary for resumption of decoding in the correct downstream reading frame. The glyT mutants described here may disrupt bypassing during either take‐off, scanning or landing. The observation that each mutant pairs inefficiently with its cognate codon in the A‐site suggests that the defect occurs after take‐off.

In one model for reduced bypassing, increased dissociation of mutant peptidyl‐tRNA2Gly from cognate codons diminishes recognition of the landing site GGA during scanning. Ribosomes which fail to form a stable interaction with the landing site either resume translation at random non‐cognate sites or dissociate from the mRNA completely. In a second model, weakened tRNA–ribosome contacts in the P site during scanning lead to high rates of peptidyl‐tRNA dissociation from the ribosome. Non‐cognate landing and mRNA or peptidyl‐tRNA2Gly dissociation from the ribosome may all occur during WT bypassing—only 50% of ribosomes poised at the take‐off site bypass productively. Experiments to determine the fate of ribosomes which fail to bypass in WT and mutant glyT strains are underway.

Mutations in the anticodon stem and elbow region of tRNA2Gly show nearly identical decreases in bypassing efficiency and decoding ability. This poses the question of how these two regions of tRNA act to stabilize codon–anticodon pairing. To address this question, we examined the predicted structural consequences of the mutants and then looked at the role structure may play in stabilizing codon–anticodon interactions.

Structural consequences of mutations

Three of the seven tRNA2Gly mutants that display a byp phenotype are predicted to increase the flexibility of the elbow region by altering the highly conserved base pairs between the D‐ and T‐loops (Dirheimer et al., 1995). G19U and C56A disrupt the G19●C56 Watson–Crick base pair at the corner of the elbow which locks the loops together. Rare, naturally occurring examples of the disrupted pairs (U19●C56 or G19●A56) exist (Sprinzl et al., 1998) but the properties of these tRNA have not been investigated. Mutations at G18 interfere with the other hydrogen bonding contact between the loops (N2 of G18 to O4 of Ψ55) and may reduce the stacking energy gained by G18 intercalation between G57 and m1A58.

Three of the remaining mutants are predicted to increase the flexibility of the elbow region by altering the T‐loop. Two of these may affect a stabilizing hydrogen bond between phosphate 60 (P60) and N4 of C61. The extra U in 60+U may distort the backbone and move P60 out of hydrogen bonding range with N4. This type of disruption has been reported for the addition of an A to the T‐loop of tRNAAsp (Romby et al., 1987). Interestingly, addition of the A also destabilizes the interaction between the D‐ and T‐loops as shown by an increase of C56 to chemical modification. C61U disrupts the P60–N4 hydrogen bond by virtue of eliminating N4. Romby et al. (1987) noted, in a C61U variant of tRNAAsp, that P60 became highly accessible to chemical modification suggesting the other hydrogen bond it makes (to 02′ of A58) may also be weakened. The third mutation (C62A) creates a G–A mismatch in the T‐stem which must distort or destabilize G53●C61, perhaps also affecting the P60–N4 hydrogen bond. Thus, all of the byp mutations in the T‐stem–loop may increase the T‐loop flexibility and potentially alter its interaction with the D‐loop.

The lone byp mutation in the anticodon stem (C40G) creates a G–G mismatch in a region defined genetically as the extended anticodon (Yarus, 1982). The work of Yarus and colleagues indicates that for each possible 3′ nucleotide in the anticodon (which pairs with the first base of the codon) there is an optimal sequence for certain nucleotides in the anticodon loop and for the first two base pairs of the anticodon stem (31●39 and 30●40) (Yarus et al., 1986a,b). Changing these positions or base pairs from the optimum reduces decoding efficiency. Structurally, the G–G pair in tRNA2Gly may destabilize the stem and alter its stacking interactions with bases in the anticodon loop.

tRNA structure and stability of codon–anticodon pairing

The preceding discussion suggests that each of the tRNA2Gly mutants reduces the structural rigidity of the tRNA. The immediate effects of this increased flexibility may be different for mutants in the elbow region versus mutants in the anticodon stem, but ultimately the loss of stability results in diminished codon–anticodon pairing in both the A‐ and P‐sites of the ribosome. The coincident loss of tRNA structure and decoding ability suggests the mutations may interfere with a dynamic interaction between the tRNA and ribosome which stabilizes cognate pairing.

The contribution that ribosomes make to codon–anticodon stability is substantial. Estimates indicate that cognate interactions in the A‐site of the ribosome are stabilized between 10‐ and 100‐fold (at 20–25°C) over both non‐cognate interactions in the A‐site and cognate interactions in solution (Grosjean et al., 1976; Karim and Thompson, 1986; Pape et al., 1998). Chemical protection studies with A‐site tRNA or anticodon loop analogues suggest that the majority of the contacts are between the anticodon stem–loop and the decoding center of 16S rRNA (Green and Noller, 1997). These same studies indicate P‐site tRNA also makes extensive contacts with 16S rRNA suggesting ribosome‐mediated stabilization continues in the P‐site. Contacts between 23S rRNA and tRNA appear to be almost exclusively with the CCA end (Green and Noller, 1997). Thus, with respect to rRNA contacts, the main ribosome–tRNA interactions are at the extremities of the tRNA.

A conformational adjustment by tRNA may trigger the ribosome to stabilize cognate codon–anticodon pairing. Evidence for a conformational change in tRNA during the decoding step comes from several sources. One source is genetic studies which examined the consequences (with respect to decoding) of altering the D‐anticodon domain of tRNA. Mutations affecting the 12‐23‐9 triple base pair allow normally forbidden third position wobble pairs during the decoding step (Hirsh, 1971; Smith and Yarus, 1989). Similarly, mutations at the 27–43 pair (Schultz and Yarus, 1994) allow first position wobble pairs. Together, these results suggest the conformation of the D‐anticodon domain of tRNA influences dissociation rates from the A‐site (Yarus and Smith, 1995).

Additional support for a conformational change in tRNA upon codon–anticodon pairing comes from fluorescence studies which showed that ternary complex binding to a cognate codon in the A‐site induces a conformational change in the D‐loop (Rodnina et al., 1994). A striking parallel observation comes from comparison of the crystal structure of tRNAPhe (of a monomer) with the crystal structure of tRNAAsp [of a dimer of monomers bound by their nearly self‐complementary anticodon loops (GUC)] (Moras et al., 1980; Westhof et al., 1985). The G19●C56 base pair does not form in tRNAAsp in stark contrast to tRNAPhe. This lack of pairing correlates with increased flexibility in the elbow region and increased rigidity in the anticodon stem. The crystal structure, however, may not tell the whole story. The fluorescence studies suggest that the conformation of the D‐loop is at least partially restored following release of the tRNA into the A‐site (Rodnina et al., 1994). In agreement with this conclusion, the results presented here show that stability of the elbow contributes to codon–anticodon stability.

Given this view of the interplay between the ribosome and tRNA, it is not hard to imagine how mutants of the elbow region and anticodon stem destabilize codon–anticodon pairing. The C40G mutation may directly interfere with 30S subunit contacts to the anticodon stem region. Alternatively, it may disrupt a conformational change in the tRNA which improves tRNA–ribosome contacts. The flexibility found in mutants of the elbow region may destabilize tRNA–30S subunit contacts by altering the D‐anticodon domain or by altering inter‐subunit contacts. A final possibility, applicable to both types of mutants, takes into account the low but significant rates of WT peptidyl‐tRNA dissociation and re‐pairing. Evidence for this comes from the ‘stop‐hopping’ observed during translation of the sequence CUU UAG CUA. Peptidyl‐tRNALeu detaches from the CUU, bypasses the stop codon, and re‐pairs at the CUA at an efficiency of 1% (Weiss et al., 1987). The predicted increased flexibility found in mutants of the elbow region and anticodon stem of tRNA2Gly may lower the efficiency of re‐pairing (whether at ordinary GGA codons or at the landing site GGA) by causing misalignment of the D‐anticodon domain with respect to the P‐site codon.

The mutations in tRNA2Gly described here add to a collection of mutants which affect the stability of peptidyl‐tRNA pairing with mRNA. Not surprisingly, tRNA mutations which alter the anticodon loop have significant effects, that is, addition of an extra nucleotide to the anticodon of tRNA1Val allows this tRNA to detach and re‐pair at matching valine codons at high frequency (O'Connor et al., 1989). Mutations which influence peptidyl‐tRNA decoding have also been identified in nucleotides comprising the decoding domain of 16S rRNA (530 loop, helix 34 and 1400–1500 region) (O'Connor et al., 1997). Finally, mutations in ribosomal protein L9 enhance both gene 60 bypassing and stop‐hopping (Herbst et al., 1994). The continued exploitation of genetic systems in which peptidyl‐tRNA detachment and re‐pairing occurs will undoubtedly yield additional insights into how the translational apparatus maintains peptidyl‐tRNA–mRNA pairing.

The importance of the L structure of tRNA is implicit in its extraordinary conservation in all organisms. The three dimensional structure of the functional sites of the ribosome maintains a selection for tRNA molecules and translation factors that conform to certain structural specifications. The tRNA2Gly mutants described here, especially the mutants of the elbow region, indicate that one functional role of the rigidity of the L structure is to ensure stable codon–anticodon interactions.

Materials and methods

Plasmid constructions

All plasmid constructions were confirmed by DNA sequencing on automated sequencing machines (ABI‐100).

Sucrose selection plasmid. The phoA‐gene 60‐sacB plasmid (GSB228×3) was constructed in three steps. (i) SacB was amplified by PCR using a sense primer with an embedded ApaI site (GCTTGGGCCCATCTTCTAAA GCTAACATCAAAAAGTTTGCAAAACAAGC) and an anti‐sense primer with an embedded SalI site (GCGCGCGTCGACTTATTTGTTAACTGTTAATTGTCCTTGTTC) from a sacB expression vector (unpublished data) in which the HindIII site [starting at nt 1798 (DDBJ/EMBL/GenBank accession No. X02730)] and the KpnI site (starting at 1521) had been mutated to CAGCTT and GGTATC, respectively. This PCR fragment was cloned into the ApaI and SalI sites of a gene 60–lacZ expression vector (GLZ30; Maldonado and Herr, 1998) in which the lacZ fusion is controlled by an IPTG‐inducible tac promoter. (ii) The ribosome binding site (RBS) and first 43 codons of E.coli phoA was amplified and cloned into the XbaI and HindIII sites using a sense primer with an embedded XbaI site (CATGTCTAGATTAATTGGAGAAAATAAAATGAAACAAAGCA CTATTGCACTGG) and an anti‐sense primer with an embedded HindIII site (CTAGAAGCTTACCGCCGGGTGCAGTAATATCGCCCTG). And (iii), the first 48 codons of gene 60 was amplified using a sense primer with an embedded HindIII site (GATCAAGCTTATGAAATTTGTAAAAATTGATTCTTC) and an anti‐sense primer with an embedded ApaI site (GGTTAGGGCCCCCTAATCCATCGTGATCTGCGTCTG) and cloned between phoA and sacB. A control with a 3‐fold reduction in bypassing [amplified from (BX5b) (Weiss et al., 1990)] was also made (GSB228×5b).

tRNA over‐expression plasmids. The XbaI–SalI fragment of pACYC184 (Chang and Cohen, 1978) was replaced with a sequence which eliminates the XbaI site and introduces the bacteriophage λ 6S terminator and a KpnI site upstream of SalI. The 3.4 kb KpnI–SalI fragment from GLZ30 was subsequently cloned into this vector to create GLZ30/ACYC. The tac promoter, flanked by KpnI and XbaI sites, was then replaced with the pL region of bacteriophage λ amplified by PCR using primers with embedded XbaI and KpnI sites (GATCTCTAGAGCTGATGTGCTCAGTATCACCGCCAGTGG and GATCGGTACCTTGCCT CACGATCGCCCCCAAAACACATAACC, respectively) to create pLEX1. All tRNAGly genes were cloned into pLEX1 using downstream HindIII and SalI sites. glyT PCR products were amplified from E.coli or S.typhimurium strains using a sense primer with an embedded HindIII site located upstream of glyT in tyrU (GATCAAGCTTCTTCGAATCCTT CCCCCACCACC) and an anti‐sense primer with an embedded SalI site located downstream in thrT (GATCGTCGACGACCTCACCCTTACCAAGGGTGCG). pLcom1 and pLcom2 (Figure 4) were made using a two‐step PCR strategy and internal primers which allowed the appropriate compensatory mutation to be made. E.coli glyU was amplified using primers with embedded HindIII and SalI sites (GATCAAGCTTAAAATGCTGATGGCGAATAATCGCAGTC and GATCGGGCCCTTGTTAAAAAGCGTAGATAA CCACTTATTTAAG, respectively).

To make pSuGlyT plasmids (Figure 7), a temperature‐sensitive bacteriophage λ cI allele (cI857) was amplified using a primer with an overhang carrying an embedded SalI site, the lacIq promoter, and a synthetic RBS (GATCGTCGACACCATCGAATGGTGCAAAACCTTTCGCGGTATGGCATGATAGCGCCCGGAAGAGAGTCAATTCAGGAGAATCGTATGAGCACAAAAAAGAAACCATTAACACAA‐GAGCAGC) and a primer with an embedded EagI site (GATCCGGCCGTGCCGATCAGCCAAACGTCTCTTCAGGCCACTG). This was cloned into pLEX1 using SalI and EagI sites to create pLEX2. Amber suppressor derivatives of glyT alleles were amplified using a two‐step PCR strategy and cloned into the XbaI and SalI sites of pLEX2. The outside primers were the same as detailed above except the primer in tyrU carried an embedded XbaI site. The internal primers changed the anticodon of glyT to 3′AUC.

ProtA‐cat‐lacZ expression plasmids for the amber suppression assays were constructed by amplifying the RBS and coding sequence of protA‐cat from 4p112 (Weiss et al., 1987) by PCR and subcloning it into the XbaI and HindIII sites of GLZ30 creating GLZ69. An oligonucleotide insert which separated the protA‐cat and lacZ open reading frames by a UAG codon was then cloned between HindIII and ApaI sites creating GLZ96. A test construct with a CAG sense codon replacing the UAG codon was also made: GLZ95.

Genetic experiments

All genetic manipulations were carried out essentially as described (Miller, 1972, 1992). The genotypes of strains used are listed in Table I.

The E.coli strain for the selection (which carries a gene 60–lacZ fusion on the chromosome) was created in three steps. First, the promoter of the plasmid GLZ16 (Maldonado and Herr, 1998) was replaced with the constitutive σ70 E.coli consensus promoter (Watson et al., 1983) to create GLZ60. Secondly, this plasmid was recombined with λRS45, as described by Simons et al. (1987), to create λGLZ60. Finally, a Su1675Iq λGLZ60 lysogen was made and designated AH108.

Selection conditions where a 3‐fold decrease in bypassing efficiency could be detected were determined by monitoring the survival of 103 Su1675Iq [GSB228×5b] cells against a background of 108 AH108[GSB228×3] cells on TCB plates (10 g tryptone, 5 g casamino acids, 17 g Bacto‐agar, 0.2% glucose, 40 μg/ml X‐gal, 1× A‐salts per liter) with variable concentrations of IPTG and sucrose. Plates with 10–20 μm IPTG and 2.5% sucrose proved to be optimal.

Identification of mutants. Aliquots (0.1 ml) of 8 h 2 ml Luria–Bertani (LB) cultures were plated on selection plates and incubated for 24–48 h. Light blue or white colonies were picked and streaked on LB X‐gal plates to confirm the color phenotype. To test if bypassing was reduced, putative mutants were transformed with compatible plasmids which allowed bypassing to be quantified.

Mutants displaying a significant reduction in bypassing efficiency were mapped and sequenced in seven steps. (i) AH108byp[GSB228×3] strains were cured of the selection plasmid using courmermycin (Danilevskaya and Gragerov, 1980). (ii) AH108byp F phenocopies were mated with BW5660 to introduce recA+ linked to srbC300::Tn10. (iii) The resulting kanr tetr UVr exconjugants were transduced to sorbital utilization and tet sensitivity. (iv) Matings of F phenocopies of the resulting transductants to tetr Hfr mapping strains (Singer et al., 1989), were used to localize the mutants to the region between 89 and 6 min (clockwise) on the E.coli chromosome. (v) Transductional crosses with strains having Tn10's near genes involved in translation in the localized region (Singer et al., 1989) were used to map the mutations to the tufB operon. (The mutants were 50% linked to ArgE86::Tn10 and 70% linked to thi‐39::Tn10.) (vi) Complementation performed using previously described tufB expression plasmids suggested that the mutations were in glyT (Van Delft et al., 1987). (vii) PCR products were amplified and sequenced using primers upstream of glyT in tyrU and downstream in thrT. For further characterization, the byp mutants were moved into CSH142 using P1 transduction and a linked Tn10 (argE86::Tn10). The resulting transductants were then transduced to arginine prototrophy to remove the Tn10.

Note on the sacB selection. One limitation of the sacB selection system is that mutants which reduce both levansucrase production and growth rate may not survive. Mutants appear to survive only if they increase the ratio of growth rate/levan synthesis beyond a certain threshold. This may explain why 60+U, C61U and C62A were not isolated even though these reduce bypassing to levels comparable to byp mutants.

tRNA1Gly and the byp mutant phentype. AH122 and byp mutant derivatives were constructed in three steps. (i) A P1 lysate grown on BGS334 was used to transduce a CSH142 λGLZ60 lysogen to tetr and lysA. (ii) This strain was transduced to lysine prototrophy and glyU(SuUAG) with a P1 lysate grown on BGS347. Presence of glyU(SuUAG) was confirmed by transducing BGS334 to lysine prototropy and testing for suppression of trpA(UAG211). (iii) This strain was transduced to tetr using P1 lysates grown on byp mutants carrying argE86::Tn10. A CSH142 λGLZ60 lysogen was transduced as a control in these experiments creating AH123.

Bypassing efficiency assays

Bypassing efficiency measurements were performed essentially as described (Maldonado and Herr, 1998) using [35S]methionine pulse–chase analysis. Cells carrying a GST–gene 60 expression plasmid were grown to mid‐log phase in MOPS–glucose (Neidhardt et al., 1974) containing all amino acids (150 μg/ml each) except methionine. Expression was induced with IPTG for 8 min, after which total protein was labeled by the addition of 7.5 μCi [35S]methionine. After a 1 min labeling period, the radioactivity was chased by addition of an excess of cold methionine for 3 min. Following pulse–chase labeling, total protein in each sample was separated on a 12% Tris–glycine SDS polyacylamide and visualized with a Molecular Dynamics PhosphorImager. Efficiency estimates represent the amount of bypassing product divided by the total protein synthesized from the GST–gene 60 fusion (bypassing product + termination product). All efficiency estimates were determined by averaging at least three independent assays and were corrected for differences in methionine content between the bypassing product and termination product.

Aminoacyl‐tRNA purification and detection

Whole aminoacyl‐tRNA was purified as described (Varshney et al., 1991; Bourdeau et al., 1998), separated on an acidic denaturing 6% polyacrylamide gel, blotted to a Zeta‐probe membrane (Bio‐Rad) and probed using 32P‐end labeled oligonucleotide probes complementary to nucleotides 27–54 of tRNA2Gly or tRNA2Gly (SuUAG). 5S rRNA was simultaneously probed using a 32P‐end labeled oligonucleotide complementary to nucleotides 30–60. Loading differences between the samples were corrected for by normalizing tRNA2Gly levels to 5S rRNA levels. Mutant tRNA levels were then normalized relative to WT levels. Three separate preparations of tRNA were assayed for each estimate.

Decoding efficiency estimates

O/N 37°C cultures of AH108 strains carrying tRNA2Gly (SuUAG) over‐expression plasmids and GLZ96 (UAG containing test construct) or GLZ95 (control construct; see above) were diluted 1:50 and grown for 1 h at 40°C. LacZ expression was then induced with 2 mM IPTG for an additional 90 min of growth before β‐galactosidase activity was assayed (Miller, 1972; Maldonado and Herr, 1998) and whole tRNA was purified. β‐galactosidase levels from test constructs were corrected by levels from control constructs to give an amber suppression efficiency for each tRNA2Gly (SuUAG) variant. Decoding efficiency estimates were made by correcting the amber suppression efficiency by the levels of aminoacyl‐tRNA2Gly (SuUAG) in each sample. Estimates represent the average of three independent measurements.


We thank Steve Alam, Mike Howard and Doug Bucklin for critical reading of the manuscript. We thank Fran Pagel and Manny Murgola for bacterial strains BGS334 and BGS347. We also thank Alan Collmer for providing a sacB expression vector. Finally, we are grateful to Michael O'Connor for both bacterial strains and helpful discussions. This work was supported by a grants (to J.F.A.) from NIH (GM48152), and (to R.F.G.) from DOE (DE‐FG03‐99ER62732/A000). A.J.H. is supported by an NIH training grant (#5T32GM07464‐24).