Reassigning Cysteine in the Genetic Code of Escherichia coli

Corresponding author: Philippe Marlière, Unité de Biochimie Cellulaire, Institut Pasteur, 28 rue du Docteur-Roux, 75015 Paris, France., marliere{at}pasteur.fr (E-mail).

Communicating editor: S. YOKOYAMA

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

We investigated directed deviations from the universal genetic code. Mutant tRNAs that incorporate cysteine at positions corresponding to the isoleucine AUU, AUC, and AUA and methionine AUG codons were introduced in Escherichia coli K12. Missense mutations at the cysteine catalytic site of thymidylate synthase were systematically crossed with synthetic suppressor tRNA^Cys genes coexpressed from compatible plasmids. Strains harboring complementary codon/anticodon associations could be stably propagated as thymidine prototrophs. A plasmid-encoded tRNA^Cys reading the codon AUA persisted for more than 500 generations in a strain requiring its suppressor activity for thymidylate biosynthesis, but was eliminated from a strain not requiring it. Cysteine miscoding at the codon AUA was also enforced in the active site of amidase, an enzyme found in Helicobacter pylori and not present in wild-type E. coli. Propagating the amidase missense mutation in E. coli with an aliphatic amide as nitrogen source required the overproduction of Cys-tRNA synthetase together with the complementary suppressor tRNA^Cys. The toxicity of cysteine miscoding was low in all our strains. The small size and amphiphilic character of this amino acid may render it acceptable as a replacement at most protein positions and thus apt to overcome the steric and polar constraints that limit evolution of the genetic code.

THE genetic code, the overall molecular rule that governs the formation of polypeptides from instructions read from polynucleotides, stands as the main product of evolution (WOESE 1967 ) and as an almost invariant feature of living cells (OSAWA et al. 1992 ). If descendants of natural species could be progressively remodeled in the laboratory so as to adopt different genetic codes, protein evolution could be redirected and artificial sources of biodiversity thereby established. An alternative genetic code could specify a smaller or a larger set of amino acids, a set substituted with noncanonical monomers, or a set of canonical amino acids among which codons would be redistributed [reviewed in RICH 1962 , WEBER and MILLER 1981 , WONG 1988 , and LEMEIGNAN et al. 1993 ]. The natural deviations of the genetic code that have been characterized thus far are mostly the reassignment of termination codons to amino acids and, less frequently, codon reassignments among the 20 canonical amino acids (FOX 1987 ).

Two models of the mechanism by which sense codons are switched to noncognate amino acids have been proposed. In the disuse model, advocated by JUKES and OSAWA 1993 , a drift toward A+T- or G+C-rich genomes brings a codon for an amino acid to extinction, and then, following loss of the cognate tRNA, that codon reappears to be captured by a mutant tRNA carrying amino acid ß. In the ambiguity model, advocated by NINIO 1990 , ANDERSSON and KURLAND 1991 , and SCHULTZ and YARUS 1996 , a codon in use specifying amino acid when read by a tRNA carrying amino acid ß generates protein variants with diversified functional abilities, until eventually the codon is reassigned to ß. The disuse model may be tested experimentally by introducing synthetic tRNA genes into species devoid of certain codons, such as Mycoplasma capricolum (OBA et al. 1991 ), Micrococcus luteus (KANO et al. 1993 ), and possibly Fusobacterium nucleatum (B. LEMEIGNAN and P. MARLIÈRE, unpublished results), to decode a selectable gene bearing that codon. Here we report how we tested the ambiguity model by perpetuating an alternative codon reading that provides a selective advantage to the host cell.

We used cysteine as the invader amino acid for capturing codons for several reasons. It is small and amphiphilic and fits in the volume occupied by all other amino acids except glycine, alanine and serine (CREIGHTON 1993 ). Therefore, it should be acceptable as a replacement at most positions in the sequence of proteins. Its thiol group facilitates radioisotopic labeling and derivatization with specific reagents (CREIGHTON 1993 ). Thus, cysteine misincorporation into proteins normally devoid of cysteine has been used as a tool for measuring missense translation (EDELMANN and GALLANT 1977 ). Finally, no deviation of the genetic code involving the reassignment of a sense codon from or to cysteine is known in nature (FOX 1987 ). Experimental creation of such a reassignment may thus generate nonstandard translation processes. In Escherichia coli K12, cysteine is incorporated in response to the two codons UGU and UGC through the action of tRNA^Cys and Cys-tRNA synthetase each encoded by a single gene, cysT and cysS, respectively.

We found that cysteine miscoding at isoleucine codons AUU, AUC, AUA and at methionine codon AUG was deleterious but well tolerated by E. coli. A single mutation of a catalytic cysteine into isoleucine in the active site of thymidylate synthase or of a foreign aliphatic amidase allowed selection for a miscoding Cys-tRNA reading the rare isoleucine codon AUA. Either selection was sufficient for stable propagation of the miscoding Cys-tRNA.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Strain propagation:
The E. coli K12 strains and the plasmids used in this study are listed in Table 1. Bacteria were routinely grown in mineral standard (MS) medium containing 2 g/liter glucose with or without 0.3 mM thymidine or in LB rich medium as described previously (RICHAUD et al. 1993 ). Growth media were solidified with 15 g/liter agar (Difco) for the preparation of plates.

Growth kinetics:
Growth rates were determined by measuring the optical density at 600 nm of at least five successive samples taken from cultures in balanced growth in MS glucose medium at constant temperature. Three independent experiments were performed in parallel, from which the mean value for the growth rate was computed.

Serial cultures:
The thyA strain ß5301 harboring the plasmids pTS0 and pCT1 and the thyA strain ß5249 harboring the plasmids pTS1 and pCT1 were serially subcultured in ¹/₃ ml mineral medium supplemented with 2 g/liter glucose at 37° by daily 1/100 dilutions for 80 days. Every 10 days, the cell density of stationary phase cultures was measured immediately prior to subculturing by plating dilutions on mineral glucose plates with and without 50 mg/liter carbenicillin or 25 mg/liter chloramphenicol. Colonies were counted after incubation at 37° for 24 hr. Plasmid DNA was prepared from serial cultures (QIAGEN) and used to transform the thyA strain ß1308 following a standard procedure (SAMBROOK et al. 1989 ). Transformants were selected on LB plates containing 0.3 mM thymidine and carbenicillin at 50 mg/liter and tested for thymidine auxotrophy by growth on MS glucose plates.

tRNA constructs:
The genes of miscoding Cys-tRNAs were constructed according to the method of NORMANLY et al. 1986 , by annealing 5'-phosphorylated oligonucleotides: 1: 5'-pATAACCGCTTTGTTAACGCGCCG; 2: 5'-pGTGGAGGCGCGTTCCGGAGT-CGAACC; 3: 5'-pAATTCGGCGCGTTAACAAAGCGGTTATGTAGCGG; 4: 5'-pGACTAGACGGATTTAGAATCCGCTAC; 5: 5'-pATTCTAAATCCGTCTAGTCCGGTTCG; 6: 5'-pACTCCGGAACGCGCCTCCACTGCA; 4a: 5'-pGACTAGACGGATTATAAATCCGCTAC; 5a: 5'-pATTTATAATCCGTCTAGTCCGGTTCG; 4b: 5'-pGACTAGACGGATTATTAATCCGCTAC; 5b: 5'-pATTAATAATCCGTCTAGTCCGGTTCG; 4c: 5'-pGACTAGACGGATTATGAATCCGCTAC; 5c: 5'-pATTCATAATCCGTCTAGTCCGGTTCG; 4d: 5'-pGACTAGACGGATTATCAATCCGCTAC; and 5d: 5'-pATTGATAATCCGTCTAGTCCGGTTCG. Mutated bases are underlined. tRNA^Cys_CUA was constructed from oligonucleotides 1–6; tRNA^Cys_UAU from oligonucleotides 1–3, 4a, 5a, and 6; tRNA^Cys_AAU, from oligonucleotides 1–3, 4b, 5b, and 6; tRNA^Cys_CAU, from oligonucleotides 1–3, 4c, 5c, and 6; and tRNA^Cys_GAU, from oligonucleotides 1–3, 4d, 5d, and 6. The sequences assembled had 5' EcoRI- and 3' PstI-ligatable ends. A fivefold molar excess of the assembled DNA was then ligated into pTZ18 digested with EcoRI and PstI (Boehringer Mannheim) in a volume of 10 µl for 14 hr at 7° with 1 unit of T4 DNA ligase (Amersham) in a buffer containing ATP provided by the vendor. Six microliters of the mix was used to transform GT869 cells (Table 1) following a standard procedure (SAMBROOK et al. 1989 ). Plasmid DNA from six transformants selected on LB plates containing 50 µg/ml ampicillin was analyzed by restriction endonuclease digestion. The insert in one plasmid (pCT11–pCT15) from each construction that showed the appropriate restriction-fragment pattern for the cloned tRNA gene was sequenced with Taq-DNA polymerase (Amersham), according to the supplier's instructions to confirm the cloned sequence. Plasmids pCT11–pCT14 were used to transform ß5153 cells to give strains ß5372–ß5375 (Table 1) and to transform ß5236 cells to give strains ß5379–ß5382 (Table 1).

thyA constructs:
Mutant alleles of the E. coli thymidylate synthase gene were constructed by site-directed mutagenesis of plasmid pTS0 (Table 1), performed according to the method of KUNKEL et al. 1987 . The following mutagenic oligonucleotides phosphorylated at the 5' end were used: 7: 5'-pTGGATAAAATGGCGCTGGCACCGATGCATGCATTC-TTCCAGTTCTATGT (allele 146AUG); 8: 5'-pTGGATAAAATGGCGCTGGCACCGATACATGCATTCTTCCAGTTCTATGT(allele 146AUA); 9: 5'-pTGGATAAAATGGCGCTGGCACCGATTCATGCATTCTTCCAGTTCTATGT (allele 146AUU); and 10: 5'-pTGGATAAAATGGCGCTGGCACCGATCCATGCATTCTTCCAGTTCTATGT (allele 146AUC). Mutated bases are underlined. Mutants were screened by using the ligation product to transform the thyA strain ß1308 to carbenicillin resistance. Transformants were tested for thymidine auxotrophy on MS glucose plates. For each construction, one carbenicillin-resistant thymidine-auxotrophic transformant was analyzed by DNA sequencing of the plasmid insert as described in the previous section. Double transformants were constructed by transforming cells expressing each of the four thyA missense alleles from a ColE1 plasmid mutated at codon 146 (plasmids pTS1–pTS4) with pSU18 or a pSU18 derivative carrying one of the four cysT alleles for missense tRNA^Cys (pCT1–pCT4). Twenty strains (ß5245–ß5264) were thereby obtained as shown in Table 1.

Amide utilization and amiE constructs:
E. coli strains producing amidase from Helicobacter pylori were constructed by transformation of MG1655 with pILL417, which carries the amiE gene. Growth on aliphatic amides was tested on an ammonia-free mineral medium (AFM): a buffer solution, pH 7.3, containing 50 mM dipotassium phosphate, 4 mM citric acid, 1 mM magnesium sulfate, 3 mM ferric chloride, 1 mM manganese chloride, and 1 mM calcium chloride. To this medium was added 2 g/liter glucose and 20 mM of either acetamide, propionamide, or butyramide. Directed mutagenesis of the H. pylori amiE gene was carried out by a method based on an accurate polymerase chain reaction (ANSALDI et al. 1996 ) using pILL417 (Table 3). Primers 11: 5'-GGGCTTGAAAGTTTCTTTGATCATTATAGATGATGGGAACTACCCTG, 12: 5'-GGGCTTGAAAGTTTCTTTGATCATTTAGGATGATGGGAACTACCCTG, and 13: 5'-CAAAGAAACTTTCAAGCCCTTAGGCCCATCAA were synthesized (mutated bases are underlined) and used to construct the amiE allele 165AUA (primers 11 and 13) and the amiE allele 165UAG (primers 12 and 13). Both primer pairs overlap for 19 nucleotides. The polymerase chain-reaction protocol involved a denaturation of 120 sec at 95° followed by 15 cycles of 30 sec at 95°, 30 sec at 58°, and 240 sec at 72° followed by an extension step of 400 sec at 72°. The reaction was carried out using 2 units of Vent DNA polymerase (New England BioLabs) and 50 ng of plasmid template in 50 µl reaction mixture. The concentrations of primers, deoxynucleotides, and buffer were adjusted according to the supplier's instructions. Treatment with DpnI (Boehringer Mannheim) and transformation of MG1655 cells were performed according to ANSALDI et al. 1996 . Transformants were tested for growth on AFM glucose plates supplemented with 20 mM acetamide. pAM1 (allele 165AUA) in strain ß5412 (Table 1) and pAM2 (allele 165UAG) in strain ß5413 were sequenced as described. ß5412 cells were transformed with pCT1 and the resulting strain, ß5417, was grown on aliphatic amides to test for restoration of amidase activity by translational suppression. pAM3 carrying the gene coding for the tRNA^Cys_UAU and the allele amiE:165AUA in tandem was constructed by transferring the PstI-HindIII fragment from pAM1 to pCT11. pAM3 was introduced into the strain ß5236, harboring pCS, and the resulting strain ß5420 tested for amidase activity.

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Cysteine miscoding in DNA precursor biosynthesis:
We first studied cysteine miscoding using a biosynthetic reaction which, although generating only small amounts of product, is essential in E. coli. The enzyme catalyzing such a reaction would presumably be required only in small amounts, and therefore low-level suppression of a missense mutation of the gene for this enzyme borne by a high-copy number plasmid should permit growth. Thymidylate synthase, which catalyzes the conversion of dUMP to dTMP in DNA biosynthesis, fulfilled these requirements. Fewer than 100 molecules of active thymidylate synthase per E. coli cell are sufficient to sustain proliferation. The active center of the enzyme involves a cysteine residue at position 146 that was shown by systematic codon substitution to be indispensible for activity (DEV et al. 1988 ; MICHAELS et al. 1990 ). Strains of E. coli whose thymidylate synthase gene thyA is inactive require the nucleoside thymidine or the base thymine for growth in mineral glucose media (see MATERIALS AND METHODS).

The cysteine codon UGC 146 in thyA on the 200-copy number ColE1 plasmid pTS0 was replaced with each codon of the AUN family, i.e., the three isoleucine codons AUU, AUC, and AUA and the methionine codon AUG, by site-directed mutagenesis. Strain ß1308, lacking thyA, was transformed with each of the four mutated plasmids pTS1–pTS4. The transformants could grow only if they were supplied with exogenous thymidine. The four strains were then transformed with a compatible 20-copy number P15A plasmid expressing a synthetic gene for a missense tRNA^Cys with one of the four anticodons AAU, GAU, UAU, and CAU, respectively pCT1–pCT4. The growth of the 16 resulting strains, bearing all possible codon/anticodon crossings of the AUN series, was assessed in mineral glucose medium at 30° in the absence of thymidine (Table 2).

The four miscoding Cys-tRNAs suppressed thyA alleles with complementary codons, allowing growth and thereby demonstrating their ability to act as substrates of Cys-tRNA synthetase and to incorporate cysteine during translation elongation. Three noncomplementary combinations also grew: the suppressor Cys-tRNA_GAU with the codon AUU, Cys-tRNA_AAU with AUA, and Cys-tRNA_CAU with AUA (Table 2). As judged from growth rates, there were differences in the relative suppression efficiencies of missense Cys-tRNAs.

The miscoding Cys-tRNA_GAU suppressed thymidine auxotrophy so as to allow thyA strains ß5256 and ß5260 to grow nearly as fast in the absence of thymidine as in its presence (Table 2). This efficient suppression may indicate that tRNA^Cys_GAU is a better substrate for Cys-tRNA synthetase than the other mutated tRNAs. Note that tRNA^Cys_GAU , like the bona fide tRNA^Cys_GCA , has a G at position 34. Charging assays demonstrated that mutants of E. coli tRNA^Cys substituted at position G34 were cysteinylated by the cognate synthetase 2000-fold less well than the wild-type substrate, whereas a mutant retaining G34, tRNA^Cys_GAA was almost as good a substrate as the bona fide tRNA^Cys (KOMATSOULIS and ABELSON 1993 ).

The weakest suppression was observed for the miscoding Cys-tRNA with A34 in our in vivo study (Table 2). No mature tRNA from E. coli bears an A in the anticodon position 34. The synthetic variant of E. coli tRNA^Phe with A34 has a disabled phenotype in vivo (GAVINI and PULAKAT 1992 ). The low efficiency of AUU suppression obtained with the complementary suppressor Cys-tRNA_AAU is in line with these observations. It remains to be established whether A at position 34 prevents efficient charging of the tRNA molecule by its cognate synthetase or interferes during translation elongation.

Toxicity of cysteine miscoding:
We attempted to amplify the charging of missense tRNA^Cys in order to assess the toxicity of cysteine miscoding. The ratio of tRNA and aminoacyl-tRNA synthetase concentrations is known to affect misincorporation rates in nonsense suppression (SWANSON et al. 1988 ). An E. coli strain overexpressing the cysS gene for Cys-tRNA synthetase from a 20-copy number P15A plasmid (pSU38) was transformed with compatible 200-copy number ColE1 plasmids (pTZ18) carrying the four synthetic cysT alleles encoding missense tRNA^Cys. Misincorporation of amino acids through erroneous codon reading is known to cause heat sensitivity in microorganisms (HALL and GALLANT 1972 ). The growth characteristics of E. coli double transformants were thus compared at 30°, 37°, and 42°. The growth rates of each strain overproducing a miscoding tRNA^Cys but not Cys-tRNA synthetase were indistinguishable from those of the host strain MG1655 bearing the vectors, except for the strain carrying the allele cysT:GAU (Figure 1). As for double transformants also expressing the cysS gene, the growth rates at 37° of the strains bearing the alleles cysT:CAU and cysT:GAU and, at 42°, of all four strains carrying a cysT miscoding allele were lower than the growth rate of the control strain ß5371 overexpressing cysS but no cysT allele (Figure 1).

View larger version (39K): In this window In a new window Download PPT slide   Figure 1. Toxicity of cysteine miscoding. Growth rates, expressed in h-1 units, of wild-type E. coli MG1655 overexpressing alleles of the cysT gene with mutated anticodons at various temperatures in mineral glucose medium are indicated by hatched bars. Solid bars represent the corresponding values obtained with the same strains overexpressing the cysS gene for Cys-tRNA synthetase from a compatible plasmid.

The generation time of the strain overexpressing the cysT:GAU allele was 26% longer at 42° than that of the control. The strain overexpressing both cysS and cysT:GAU grew 20, 67, and 260% more slowly at 30°, 37°, and 42°, respectively, than the control strain bearing no tRNA^Cys gene on the vector (Figure 1). Thus, this allele has the largest effect on cell growth, in agreement with the observation (above) that Cys-tRNA_GAU efficiently reads the two codons AUU and AUC. About 95% of isoleucine is specified by these two codons in E. coli (BLATTNER et al. 1997 ), and therefore it is not surprising that their misreading as cysteine hinders growth. Although the cysT:UAU allele encodes an efficient suppressor tRNA^Cys (Table 1), it had the smallest impact on the cell growth rate even with concomitant overproduction of Cys-tRNA synthetase (Figure 1). There are about 3500 AUA codons in the E. coli genome, making it one of the rarest codons (BLATTNER et al. 1997 ). Presumably, this provides less opportunity to cause damage by misreading.

The cysT:CAU allele efficiently reads the methionine codon AUG and, less well, the minor isoleucine codon AUA (Table 1). Its toxicity was similar to that of cysT:GAU, with 52 and 95% lower generation times at 37° and 42°, respectively, only when cysS was overexpressed (Figure 1). Again, this could be due to the high abundance of methionine AUG codons, which amount to about 35000 in the E. coli genome (BLATTNER et al. 1997 ).

Genetic stability of cysteine miscoding:
We next tested how our model miscoding system could be maintained over subsequent generations. ß5301 and ß5249, thyA-deleted strains expressing the AUA-decoding tRNA^Cys from a P15A plasmid with an associated chloramphenicol-resistance marker, were propagated in parallel liquid cultures. Both strains also harbored a ColE1 plasmid carrying a thyA allele and an associated carbenicillin-resistance marker. ß5249 expressed thyA:AUA146 and ß5301 expressed wild-type thyA. The growth rates of the two strains were identical (data not shown).

They were serially subcultured in mineral medium with a limiting glucose supply at 37° by daily 1/100 reinoculation. Over 80 days, totalling about 530 generations, the viable cell titer at the end of each subculture remained constant for both strains, around 1.8 x 10⁹ cells/ml. The viable cells of both lines remained resistant to carbenicillin. The chloramphenicol-resistant cell count remained stable in ß5249 cultures, but declined about 100-fold in ß5301 cultures after 530 generations (Figure 2).

View larger version (9K): In this window In a new window Download PPT slide   Figure 2. Persistence of miscoding Cys-tRNAUAU. A P15A vector carrying the synthetic cysT:UAU allele and a chloramphenicol resistance marker was propagated in two strains also harboring either an active thyA allele with cysteine 146 UGC codon (+, ß5301) or an inactive allele with isoleucine 146 AUA codon (, ß5249) on a ColE1 plasmid with a carbenicillin resistance marker. Serial subcultures were reinoculated at 1/100. Stationary subcultures were periodically sampled and cells counted by dilution and plating on plates with carbenicillin, chloramphenicol, or without antibiotic. The titer of carbenicillin-resistant cells remained nearly constant at ~1.8 x 109 for the active and inactive thyA alleles.

Retransformation of strain ß1308 with plasmids extracted from subcultures of ß5249 conferring carbenicillin resistance did not result in thymidine prototrophs. Therefore, codon 146 had not reverted from isoleucine AUA into cysteine UGU or UGC. DNA analysis of clones isolated from subcultures of strain ß5301 indicated that the pCT1 plasmid expressing suppressor Cys-tRNA and chloramphenicol resistance had been lost. A single missense mutation thus seems sufficient to perpetuate an ambiguity in the genetic code.

Cysteine miscoding in an acquired catabolic function:
Finally, we devised a suppression scheme for enforcing cysteine miscoding at a high rate. Assimilation of carbon or nitrogen involves abundant flows through metabolic pathways and high activity of the corresponding enzymes. Therefore, any mutation in one of these enzymes abolishing its activity would require being suppressed at a high rate to restore growth of the cell. Our attention focused on hydrolases, which release ammonia from amides and nitriles. One class of these enzymes (BORK and KOONIN 1994 ) displays a conserved catalytic cysteine residue in its active site, which has been shown to be indispensible for the activity of a nitrilase (KOBAYASHI et al. 1992 ). A gene specifying an enzyme of this class, amiE, was recently identified in H. pylori, a bacterium of the human stomach microflora (SKOULOUBRIS et al. 1997 ). Wild-type E. coli has no amidase activity but grows in mineral glucose medium with aliphatic amides as the nitrogen source when carrying amiE. Acetamide, propionamide, butyramide, and acetoacetamide, but neither formamide nor acetonitrile, were utilized by E. coli transformed with the amiE gene (data not shown). With amides as the only nitrogen source, the heterologous amidase goes through one catalytic cycle to generate every nitrogen atom incorporated into the biomass.

An isoleucine AUA missense mutation was introduced into codon 165 of amiE, which in the wild type specifies the cysteine residue conserved among amidase-nitrilase homologs (BORK and KOONIN 1994 ). This mutation completely abolished growth on aliphatic amides as the only nitrogen source (Table 3, strain ß5412). The production of a complementary suppressor tRNA^Cys from a compatible plasmid did not restore utilization of amides, though marginal growth could be reproducibly detected in liquid cultures with acetamide (Table 3, strain ß5417). By comparison, amide catabolism by an amiE:165UAG amber mutant allele could be restored by coexpression of an amber suppressor allele of cysT (Table 3, strain ß5418).

Efficient rescue of the allele amiE:165AUA was obtained by overexpressing the cysS gene in addition to the corresponding missense cysT allele, as judged from growth with acetamide as sole nitrogen source (Table 3, strain ß5420). This suggests that tRNA^Cys_UAU is a poor substrate for Cys-tRNA synthetase. Genetic constructs overexpressing jointly amiE:165AUA, cysT:UAU, and cysS grew slowly on propionamide or butyramide (Table 3, strain ß5420). Selection for growth on either of these substrates may thus enhance cysteine miscoding in E. coli.

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

We have constructed and characterized stable bacterial strains in which the ambiguous reading of a codon can be perpetuated. Among the 16 possible codon/anticodon combinations tested by crossing thyA alleles with artificial tRNA genes, 6 were expected to lead to suppression by incorporation of cysteine, namely the 4 complementary combinations and the combinations with U:G and G:U pairs at the wobble position. The 4 strictly complementary codon/anticodon combinations indeed resulted in suppression, notwithstanding relative miscoding efficiencies (Table 2). Pairing of third codon base U with first anticodon base G34 also led to efficient suppression, in accordance with the rule that codons ending with either pyrimidine specify the same amino acid. By contrast, the reciprocal pairing of third codon base G with first anticodon base U34 did not lead to observable suppression (Table 2). Because suppression tests provide only an indirect measurement of codon reading, we could not investigate further why Cys-tRNA_UAU was accepted for AUA but not for AUG. Possibly, the AUG/UAU codon/anticodon interaction is not sufficiently stable to trigger productive elongation, or the Cys-tRNA anticodon UAU may be outcompeted by the Met-tRNA anticodon N4-acetyl-CAU for binding to codon AUG. Suppression assays indicated that codon AUA was efficiently read by anticodons AAU and CAU in addition to the complementary UAU (Table 2). The tRNA^Ile decoding the rare codon AUA is produced in only small amounts in E. coli (MURAMATSU et al. 1988 ) and it may thus not be sufficiently abundant to outcompete even weak pairing with suppressor tRNAs. In addition, the C34 in some of the suppressor tRNA^Cys_CAU molecules may possibly be changed to lysidine (L), enabling the modified anticodon LAU to pair with AUA, as in the minor tRNA^Ile_LAU (MURAMATSU et al. 1988 ). The anticodon arm sequences encompassing the anticodon of the minor tRNA^Ile and suppressor tRNA^Cys indeed share a long sequence GA(U/C)UC*AUAAUC that might suffice for substrate modification, at the site indicated by an asterisk, by the lysidine-forming enzyme. It is clear from our data that decoding processes of normal translation can be explored in vivo by systematic crossings of artificial tRNAs with missense codons for restoring the function of a selectable protein.

Steric and polar factors are known to be important determinants of the phenotypes of missense mutations (GLASS et al. 1982 ; MARKIEWICZ et al. 1994 ). In no case was viability abolished by misincorportion of cysteine in place of isoleucine and methionine (Figure 1). This contrasts strikingly with the lethality of phenylalanine miscoding at leucine CUC codons due to overexpression of a mutant pheV allele from a plasmid (PAGES et al. 1991 ). A likely explanation is that the aromatic side-chain of phenylalanine exceeds that of leucine, whereas cysteine fits within the volumes occupied by isoleucine and methionine. The capture of the leucine codon CUG by serine in the yeast genus Candida, the most radical known deviation from the standard genetic code (OHAMA et al. 1993 ), also respects steric constraints. The miscoding tRNA^Ser_CAG from Candida albicans is highly toxic in Saccharomyces cerevisiae but nevertheless tolerated (SANTOS et al. 1996 ). Similarly, a chromosomal allele of the serU gene, supH, an anticodon mutation generating the suppressor Ser-tRNA_CAA able to read the leucine codon UUG, confers heat sensitivity on E. coli (THORBJARNARDOTTIR et al. 1985 ). Although serine fits within the space occupied by leucine, this miscoding substitutes a hydrophilic for a hydrophobic moiety. Cysteine is amphiphilic and thus presumably less liable to disrupt the interior of protein molecules when replacing the hydrophobic isoleucine and methionine.

Cysteine was satisfactory as a codon captor despite the rarity of cysteine codons in genomes. This rarity may at first sight suggest that this reactive amino acid would not be suitable. For instance, exported proteins could be vulnerable to the generation, through miscoding, of thiol side-chains that cross-link when exposed to oxygen. Also, cytoplasmic proteins sprinkled with miscoded cysteine could react with alkylating metabolites such as S-adenosylmethionine. However, random substitution of isoleucine and of methionine with cysteine did not appear to be more toxic than substitution of these two apolar amino acids with structurally similar and chemically inert analogs such as O-methylthreonine and norleucine, respectively (HORTIN and BOIME 1983 ). The choice of isoleucine as codon donor was less critical because apparently no amino acid with an aliphatic side-chain has been identified as a catalytic residue in any enzyme. Altogether, the steric and polar considerations that guided the design of our system are validated by the healthy phenotypes of our genetic constructs.

The innocuity of cysteine misincorporation at AUA suggests that it may be possible to totally reassign this rare isoleucine codon in E. coli by natural selection under controlled conditions. Three evolutionary paths appear accessible to our system, as cysteine miscoding can be maintained under selection over numerous generations: (i) the mutated AUA codon may revert and the suppressor tRNA^Cys then be lost; (ii) the decoding of the mutated AUA into cysteine could acquire additional signals so as to render it context dependent while maintaining systematic isoleucine decoding at other AUA codons, as in the case of selenocysteine coding by unique UGA stop codons (BOCK et al. 1991 ); (iii) the 3500 AUA codons of the E. coli genome could progressively tolerate decoding into cysteine by fixation of adjustment mutations until the minor tRNA^Ile loses function. It will be of interest to investigate which of these three paths is favored by natural selection and whether the amiE or the thyA missense alleles are, individually or combined, able to cause an evolutionary bifurcation by codon reassignment.

	ACKNOWLEDGMENTS

The project reported in this article originated from the thesis work of Béatrice Lemeignan. By our efforts, we have sought to pay tribute to her insight, intelligence and fortitude. We are grateful to Agnès Ullmann and Michel Goldberg for their encouragement and support. We also thank Hilde de Reuse for sharing information and material, Valérie de Crécy, Simon Wain-Hobson and Didier Mazel for constructive comments, and Emmett Johnson and Steven Quentzel for linguistic help. The writing of the manuscript greatly benefited from the advice of Alex Edelman and an anonymous referee. V.D. was supported by fellowships from the Ministère des Affaires Etrangères, the Deutsche Forschungsgemeinschaft, and the Fondation Pasteur-Weizmann.

Manuscript received May 28, 1998; Accepted for publication July 9, 1998.

	LITERATURE CITED

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

ANDERSSON, G. E. and C. G. KURLAND, 1991  An extreme codon preference strategy: codon reassignment. Mol. Biol. Evol. 8:530-544[Abstract].

ANSALDI, M., M. LEPELLETIER, and V. MÉJEAN, 1996  Site-specific mutagenesis by using an accurate recombinant polymerase chain reaction method. Anal. Biochem. 234:110-111[Medline].

BARTOLOMÉ, B., Y. JUBETE, E. MARTINEZ, and F. DE LA CRUZ, 1991  Construction and properties of a family of pACYC184-derived cloning vectors compatible with pBR322 and its derivatives. Gene 102:75-78[Medline].

BLATTNER, F. R., G. PLUNKETT, III, C. A. BLOCH, N. T. PERNA, and V. BURLAND et al., 1997  The complete genome sequence of Escherichia coli K-12. Science 277:1453-1474[Abstract/Free Full Text].

BÖCK, A., K. FORCHHAMMER, J. HEIDER, W. LEINFELDER, and G. SAWERS et al., 1991  Selenocysteine: the 21st amino acid. Mol. Microbiol. 5:515-520[Medline].

BORK, P. and E. V. KOONIN, 1994  A new family of carbon-nitrogen hydrolases. Protein Sci. 3:1344-1346[Abstract].

CREIGHTON, T. E., 1993 Proteins, Structures and Molecular Properties. Freeman and Company, New York.

DEV, I. K., B. B. YATES, J. LEONG, and W. S. DALLAS, 1988  Functional role of cysteine-146 in Escherichia coli thymidylate synthase. Proc. Natl. Acad. Sci. USA 85:1472-1476[Abstract/Free Full Text].

EDELMANN, P. and J. GALLANT, 1977  Mistranslation in E. coli. Cell 10:131-137[Medline].

FOX, T. D., 1987  Natural variation in the genetic code. Annu. Rev. Genet. 21:67-91[Medline].

GAVINI, N. and L. PULAKAT, 1992  The tRNA species for redundant genetic codons NNU and NNC. A thought on the absence of phenylalanine tRNA with AAA anticodon in Escherichia coli. J. Biol. Chem. 266:2240-2243.

GLASS, R. E., V. NENE, and M. G. HUNTER, 1982  Informational suppression as a tool for the investigation of gene structure and function. Biochem. J. 203:1-13[Medline].

HALL, B. and J. GALLANT, 1972  Defective translation of RC^-cells. Nature New Biol. 237:131-135[Medline].

HORTIN, G. and I. BOIME, 1983  Applications of amino acids analogs for studying co- and posttranslational modifications of proteins. Methods Enzymol. 96:777-784[Medline].

JUKES, T. H. and S. OSAWA, 1993  Evolutionary changes in the genetic code. Comp. Biochem. Physiol. 106B:489-494.

KANO, A., T. OHAMA, R. ABE, and S. OSAWA, 1993  Unassigned or nonsense codons in Micrococcus luteus. J. Mol. Biol. 230:51-56[Medline].

KOBAYASHI, M., N. YANAKA, T. NAGASAWA, and H. YAMADA, 1992  Primary structure of an aliphatic nitrile-degrading enzyme, aliphatic nitrilase, from Rhodococcus rhodochrous K22 and expression of its gene and identification of its active site residue. Biochemistry 31:9000-9007[Medline].

KOMATSOULIS, G. A. and J. ABELSON, 1993  Recognition of tRNA(Cys) by Escherichia coli cysteinyl-tRNA synthetase. Biochemistry 32:7435-7444[Medline].

KUNKEL, T. A., J. D. ROBERTS, and R. A. ZAKOUR, 1987  Rapid and efficient site-specific mutagenesis without phenotypic selection. Methods Enzymol. 154:367-382[Medline].

LEMEIGNAN, B., P. SONIGO, and P. MARLIERE, 1993  Phenotypic suppression by incorporation of an alien amino acid. J. Mol. Biol. 231:161-166[Medline].

MARKIEWICZ, P., L. G. KLEINA, C. CRUZ, S. EHRET, and J. H. MILLER, 1994  Genetic studies of the lac repressor. XIV. Analysis of 4000 altered Escherichia coli lac repressors reveals essential and non-essential residues, as well as "spacers" which do not require a specific sequence. J. Mol. Biol. 240:421-433[Medline].

MICHAELS, M. L., C. W. KIM, D. A. MATTHEWS, and J. H. MILLER, 1990  Escherichia coli thymidylate synthase: amino acid substitutions by suppression of amber nonsense mutations. Proc. Natl. Acad. Sci. USA 87:3957-3961[Abstract/Free Full Text].

MURAMATSU, T., K. NISHIKAWA, F. NEMOTO, Y. KUCHINO, and S. NISHIMURA et al., 1988  Codon and amino-acid specificities of a transfer RNA are both converted by a single post-transcriptional modification. Nature 336:179-181[Medline].

NINIO, J., 1990  The revised genetic code. Origins Life Evol. Biosphere 20:167-171[Medline].

NORMANLY, J., R. C. OGDEN, S. J. HORVATH, and J. ABELSON, 1986  Changing the identity of a transfer RNA. Nature 321:213-219[Medline].

OBA, T., Y. ANDACHI, A. MUTO, and S. OSAWA, 1991  CGG: an unassigned or nonsense codon in Mycoplasma capricolum. Proc. Natl. Acad. Sci. USA 88:921-925[Abstract/Free Full Text].

OHAMA, T., T. SUSUKI, M. MORI, S. OSAWA, and T. UEDA et al., 1993  Non-universal decoding of the leucine codon CUG in several Candida species. Nucleic Acids Res. 21:4039-4045[Abstract/Free Full Text].

OSAWA, S., T. H. JUKES, K. WATANABE, and A. MUTO, 1992  Recent evidence for evolution of the genetic code. Microbiol. Rev. 5:229-264.

PAGES, D., K. HIJAZI, E. J. MURGOLA, J. FINELLI, and R. H. BUCKINGHAM, 1991  Suppression of a double missense mutation by a mutant tRNA^Phe in Escherichia coli. Nucleic Acids Res. 19:867-869[Abstract/Free Full Text].

RICH, A., 1962 On the problems of evolution and biochemical information transfer, pp.103–126 in Horizons in Biochemistry. Academic Press, New York.

RICHAUD, C., D. MENGIN-LECREULX, S. POCHET, E. J. JOHNSON, and G. N. COHEN et al., 1993  Directed evolution of biosynthetic pathways. J. Biol. Chem. 268:26827-26835[Abstract/Free Full Text].

SAMBROOK, J., E. F. FRITSCH and T. MANIATIS, 1989 Molecular cloning: A Laboratory Manual. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY.

SANTOS, M. A., V. M. PERREAU, and M. F. TUITE, 1996  Transfer RNA structural change is a key element in the reassignment of the CUG codon in Candida albicans. EMBO J. 15:5060-5068[Medline].

SCHULTZ, D. W. and M. YARUS, 1996  On malleability in the genetic code. J. Mol. Evol. 42:597-601[Medline].

SKOULOUBRIS, S., A. LABIGNE, and H. DE REUSE, 1997  Identification and characterization of an aliphatic amidase in Heliobacter pylori. Mol. Microbiol. 25:989-998[Medline].

SWANSON, R., P. HOBEN, M. SUMNER-SMITH, H. UEMURA, and L. WATSON et al., 1988  Accuracy of in vivo aminoacylation requires proper balance of tRNA and aminoacyl-tRNA synthetase. Science 242:1548-1551[Abstract/Free Full Text].

THORBJARNARDOTTIR, S., H. UEMURA, T. DINGERMANN, T. RAFNAR, and S. THORSTEINSDOTTIR et al., 1985  Escherichia coli supH suppressor: temperature-sensitive missense suppression caused by an anticodon change in tRNA^Ser₂ J. Bacteriol. 161:207-211[Abstract/Free Full Text].

WEBER, A. L. and S. L. MILLER, 1981  Reasons for the occurrence of the twenty coded protein amino acids. J. Mol. Evol. 17:273-284[Medline].

WOESE, C. R., 1967 The Genetic Code. The Molecular Basis for Genetic Expression. Harper and Row, New York.

WONG, J. T. F., 1988  Evolution of the genetic code. Microbiol. Sci. 5:174-181[Medline].

This article has been cited by other articles:

				R. A. Hughes and A. D. Ellington Mistakes in translation don't translate into termination PNAS, February 1, 2005; 102(5): 1273 - 1274. [Full Text] [PDF]

				I. Ahel, C. Stathopoulos, A. Ambrogelly, A. Sauerwald, H. Toogood, T. Hartsch, and D. Soll Cysteine Activation Is an Inherent in Vitro Property of Prolyl-tRNA Synthetases J. Biol. Chem., September 13, 2002; 277(38): 34743 - 34748. [Abstract] [Full Text] [PDF]

				J. A. Silverman and P. B. Harbury Rapid Mapping of Protein Structure, Interactions, and Ligand Binding by Misincorporation Proton-Alkyl Exchange J. Biol. Chem., August 16, 2002; 277(34): 30968 - 30975. [Abstract] [Full Text] [PDF]

				V. Döring, H. D. Mootz, L. A. Nangle, T. L. Hendrickson, V. de Crécy-Lagard, P. Schimmel, and P. Marlière Enlarging the Amino Acid Set of Escherichia coli by Infiltration of the Valine Coding Pathway Science, April 20, 2001; 292(5516): 501 - 504. [Abstract] [Full Text]