Case for the as a triplet of triplets

Fabienne F. V. Chevancea and Kelly T. Hughesa,1

aDepartment of Biology, University of Utah, Salt Lake City, UT 84112

Edited by John R. Roth, University of California, Davis, CA, and approved March 21, 2017 (received for review September 5, 2016) The efficiency of codon in vivo is controlled by many Watson–Crick pairing, plus energy of stacking between bases of factors, including codon context. At a site early in the Salmonella the mRNA and tRNA. flgM , the effects on translation of replacing codons Thr6 and Based on the work presented here, we propose a model to Pro8 of flgM with synonymous alternates produced a 600-fold explain the effects of codon context on mRNA translation effi- range in FlgM activity. Synonymous changes at Thr6 and Leu9 ciency. In the model, bases of the anticodon loop of the tRNA, resulted in a twofold range in FlgM activity. The level of FlgM ac- which are often heavily modified, and mRNA bases flanking the tivity produced by any codon arrangement was directly propor- codon contribute to the stacking energy of codon–anticodon tional to the degree of in vivo stalling at synonymous recognition (shaded gray in Fig. 1). These interactions combine codons. Synonymous codon suppressors that corrected the effect to determine the efficiency at which a given codon is translated. of a translation-defective synonymous flgM allele were restricted The tRNA modifications vary throughout the three kingdoms to two codons flanking the translation-defective codon. The various of life (3) and could affect codon–anticodon pairing. The dif- codon arrangements had no apparent effects on flgM mRNA stabil- ferences in tRNA modifications could account for differences in ity or predicted mRNA secondary structures. Our data suggest that efficient mRNA translation is determined by a triplet-of-triplet ge- synonymous codon biases and for the effects of codon context netic code. That is, the efficiency of translating a particular codon is (the ability to translate specific triplet bases relative to specific influenced by the nature of the immediately adjacent flanking co- neighboring codons) on translation among different species. dons. A model explains these codon-context effects by suggesting Here, using in vivo genetic systems of Salmonella, we demon- that codon recognition by -bound aminoacyl-tRNA strate that the translation of a specific codon depends on the ′ ′

is initiated by hydrogen bond interactions between the first two nature of the codons flanking both the 5 and 3 sides of the GENETICS nucleotides of the codon and anticodon and then is stabilized by translated codon, thus generating higher-order genetic codes for base-stacking energy over three successive codons. proteins that can include codon pairs and codon triplets. The effect of the flanking codons on the translation of a translation | genetic code | context effects on translation | tRNA | specific codon varies from insignificant to profound. It has been synonymous codon effects known for decades that highly expressed use highly biased codon pairs, which can vary from one species to the next. The he DNA sequence is transcribed to form mRNA, which then speed of translation depends heavily on flanking codons (4). Tis translated into protein by . The genetic code consists of 64 triplet RNA codons that specify the 20 amino acids Results and sites of translation termination (stop codons). The decoding The Effects of Synonymous Codon Changes Adjacent to Ser7 on FlgM process involves interaction between a specific codon in the Activity. The assay system described below involves the FlgM protein mRNA and the three-base anticodon of the cognate aminoacyl- of Salmonella, which coordinates flagellar gene expression and fla- tRNA (bases 36, 35, and 34 within the anticodon loop) (1). The gellum assembly. FlgM binds and inhibits a flagellum-specific tran- code is redundant, in that many amino acids are specified by two scription factor, σ28, before completion of the flagellar motor (the or more triplet sequences. In addition, codons specifying a par- hook basal body). When the flagellar motor is completed, un- ticular are recognized by single or multiple tRNAs. associated FlgM is conducted out of the cell through the completed Most tRNA species recognize codons that differ in the third or flagellar motor, thereby freeing internal σ28 and stimulating the “wobble” position. Modifications of tRNA bases also contribute transcription of flagellar genes whose products are needed following to tRNA recognition of single or multiple anticodons as the tRNA enters the ribosome at the A site (1). Codons are read in Significance the 5′–3′ direction of the mRNA by sequential interactions with tRNA aligned 3′–5′ vis-à-vis each codon. Stacking-related wob- The genetic code for life is a triplet base code. It is known that ble codon recognition is mediated by extensive modification of adjacent codons can influence translation of a given codon and position 34 of the tRNA codon. that codon pair biases occur throughout nature. We show that Other base modifications stabilize the interaction of the tRNA mRNA translation at a given codon can be affected by the two with ribosomal A site, i.e., methylation of a G at position previous codons. Data presented here support a model in 37 prevents tRNA slippage, which would result in a shift in the which the evolutionary selection pressure on a single codon is translation (1). Fig. S1 illustrates the initial over five successive codons, including synonymous codons. translation steps starting with codon recognition and proceeding This work provides a foundation for the interpretation of how to peptidyl transfer. GTP-bound elongation factor Tu (EF-Tu) – single DNA base changes might affect translation over multiple brings aminoacyl-tRNA to the A site of the ribosome. Watson codons and should be considered in the characterization of the – Crick pairing at the first two bases of the codon initiates EF-Tu effects of DNA base changes on human disease. tRNA docking followed by triplet stabilization at the third (wobble) position. GTP hydrolysis by EF-Tu is a slow step in the Author contributions: F.F.V.C. and K.T.H. designed research, performed research, and process providing a test of the correctness of triplet recognition. wrote the paper. If an incorrect match is sensed, the tRNA is rejected, and the The authors declare no conflict of interest. process begins again. A correct match leads to the acceptance This article is a PNAS Direct Submission. of the incoming tRNA followed by peptidyl transfer. Correct 1To whom correspondence should be addressed. Email: [email protected]. – codon anticodon recognition results from stabilization of the EF- This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. Tu aminoacyl-tRNA at the A site, which is determined by the 1073/pnas.1614896114/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1614896114 PNAS Early Edition | 1of6 Downloaded by guest on September 28, 2021 + σ28 . GTP FlgM production in a wild-type serT background by scoring - . dependent transcription of luminescence transcribed from a Occasionally . * Modified σ28 Peptide -dependent promoter (PmotA-lux) (Fig. 2). Reduction of FlgM “Dangling” Often . . 28 * Base 37 } Modified . . σ . . . causes an increase in -dependent transcription. Always . 2xWC * * * The effects of synonymous changes at Thr6 and Pro8 on FlgM base pairs * 34 39 activity are presented in Fig. 2A. The 16 context backgrounds Wobble } Often . 32 . Watson-Crick Base Modified 38 * * Base Pairing * *33 * * exhibited a >600-fold range in FlgM activity. We conclude that 37 * * 36 35 34 codon context has a profound effect on flgM mRNA translation 5' 3' through the UCA Ser7 codon. Base . Stacking . Remarkably, various pairs of synonymous changes at positions } . Energy . Thr6 ACN and Pro8 CCN showed synergistic context effects on translation of the central Ser7 UCA codon. The single change of the Pro8 codon in the wild-type sequence from CCU to CCG (with the Thr6 codon unchanged) resulted in about a 20-fold Peptide . . GTP . . reduction in FlgM activity (FlgM activity was 0.053 of wild type).

. . * * * * * * * *

5' 3' ...... Base . . Stacking Energy } 3BaseShift Fig. 1. A model showing the effect of Watson–Crick base pairing and the A influence of base stacking with bases in the anticodon loop and adjacent bases in the mRNA sequence on codon–anticodon stabilization in the ac- commodation step in translation and showing how translation can be af- fected by codons adjacent to the codon being translated (context).

motor completion. These proteins include the external flagellar fil- ament and chemosensory proteins (5, 6). Our work on the effect of synonymous mutations in codon translation began with the isolation of a tRNA mutant in which translation of flgM mRNA was reduced. The mutant resulted B from a G–A base change at position 10 within the D-stem of the essential serT tRNA [tRNA(cmo5UGA)(Ser)] (7). This mutation resulted in a defect in the translation of the UCA codon for amino acid 7 of FlgM (8). There are dozens of UCA codons in the more than 50 genes required for flagellum production and chemotaxis. Therefore we were surprised that this serT mutant was perfectly motile. Why did the mutant apparently translate the many UCA codons of all C the flagellar genes normally but fail to translate the UCA codon efficiently in flgM? It seemed possible that the defect in flgM translation might be caused by context effects of codons flanking the UCA codon for amino acid 7 (Ser7) of FlgM. The Ser7 UCA codon in flgM is flanked at the 5′ side by a codon, ACC (Thr6), and at the 3′ side by the CCU (Pro8) codon. We wanted to determine whether codons adjacent to the UCA codon for Ser7 affected its translation. To avoid effects of amino acid substitutions on FlgM activity, we focused on synonymous codon changes. Synonymous changes at Fig. 2. The effects of silent mutations on FlgM inhibitory activity of the σ28- Thr6 and Pro8 would not affect the amino acid sequence of – FlgM protein. Any effects of adjacent, synonymous substitutions dependent transcription of a PmotA lux reporter construct. (A) FlgM in- hibitory activity is presented as 1/luminescence levels measured for each would be limited to effects on mRNA secondary structure and strain (SI Materials and Methods) divided by the levels measured for the mRNA stability or translation. If synonymous changes at wild-type flgM strains with ACC at position 6 and CCU at position 8. The Thr6 and Pro8 affected the ability of the ribosome to translate boxes depicting inhibitory levels are shaded red for increased FlgM activity the Ser7 UCA triplet because of codon-context effects, then the and blue for decreased activity to highlight the severe translation defect of levels of FlgM protein could increase or decrease. Slight changes the Thr6 ACC–Pro8 CCG context. (B) FlgM inhibitory activity was determined in the level of this protein are detected easily because of the for isogenic strains that contain the six synonymous Leu9 positions placed in effect of FlgM on the regulation of flagellar gene expression. context with the severe Pro8 CCG substitution (SI Materials and Methods) and compared with the wild type. (C) FlgM inhibitory activity was de- Four triplets encode Thr (ACN), and four others encode Pro termined for 24 isogenic strains that contain each of four threonine codons (CCN). Thus, there are 16 possible silent change combinations at (ACN) at amino acid position 6 in combination with each of six co- codons 6 and 8 of flgM that do not change the translated FlgM amino dons (UUG/A and CUN) at amino acid position 9 of flgM and was compared acid sequence. We assayed the effects of these 16 combinations on with the wild-type activity.

2of6 | www.pnas.org/cgi/doi/10.1073/pnas.1614896114 Chevance and Hughes Downloaded by guest on September 28, 2021 The single change in the wild-type sequence of the Thr6 codon A 134562 from ACC to ACU resulted in an 11-fold increase in FlgM activity. Pro8 = CCU (wild-type) If the synonymous changes at Thr6 ACN and Pro8 CCN affected P -flgM - -lacZ FlgM + LacZ flgM translation independently, then the effect of combining codon flgM RBS Pro8 = CCU changes at these positions should be additive. However, the P -flgM - lacZ FlgM-LacZ combination of CCU to CCG at Pro8 and ACC to ACU at flgM Thr6 produced a synergistic effect on flgM mRNA translation that Pro8 = CCG P -flgM - lacZ FlgM-LacZ + LacZ resulted in a 34-fold increase in FlgM activity. The synonymous flgM Pro8 = UGA (stop) codon changes had no apparent effects on flgM mRNA secondary P -flgM - lacZ LacZ structure or stability (SI Results, Table S1, Dataset S1,andFig. S2). flgM Taken together, our results demonstrate that translation of the Ser7 UCA codon is dependent on both the 5′ and 3′ flanking B codons and that the synergistic effects on translation of the FlgM-LacZ central codon depend on context effects extending over three LacZ codons. Because these synonymous changes had no apparent effect on flgM mRNA secondary structure or stability, our results suggests that the ribosome senses at least three consecutive Fig. 3. The effects of codon context on flgM and flgM–lacZ translation. triplet codons during the process of translation. (A) A Western blot using anti-LacZ antibody shows the effect of reduced flgM translation in a flgM–lacZ construct with the CCG codon at position 8 of flgM The Reduced Translation at Pro8 CCG of flgM Is Context Dependent. compared with the wild-type CCU codon at position 8. Lane 1 shows LacZ The severely reduced FlgM activity caused by the synergistic effects levels in a wild-type strain with a flgM–lacZ operon fusion (flgM5222::MudJ) of the Pro8 CCG codon and the Thr6 ACC codon were not seen where flgM and lacZ are translated separately. Lanes 2 through 6 show levels with the other Thr6 synonymous codons. When the Thr6 ACC of LacZ proteins in strains carrying a flgM–lacZ gene fusion (flgM7512:: codon was replaced by ACU, ACA, or ACG in combination with MudK). Lanes 2 and 3 show the full-length FlgM–LacZ fusion using wild-type flgM coding sequences (Pro8 = CCU). The strains used in lanes 3 and 5 carry a CCG at Pro8, we observed FlgM activities increased by 34-, 11-, 28 σ H14D substitution that results in increased PflgM transcription and facil- and 6.7-fold above wild type, respectively (Fig. 2A). If the defect in itates visualization of the differences in expression of the lower LacZ band

translation of flgM mRNA with the Pro8 CCG substitution was (24). The effect of the CCU-to-CCG change at Pro8 on FlgM–LacZ expression GENETICS caused by context, we wondered if translation of Pro8 CCG could is shown in lanes 4 and 5. The strain used for lane 6 carries a UAG be sensitive to synonymous changes on the 3′ side of Pro8 at Leu9. at the Pro8 position in flgM.(B) Effects of mutant SerT tRNA on FlgM–lacZ If synonymous changes at Leu9 did not affect Pro8 CCG trans- expression. Lane 1 is the same as in lane 1 in A. Lanes 2 and 3 are the FlgM– lation, then FlgM activities would be expected to remain low. All lacZ translational fusion with Pro8 CCU and Pro8 CCG, respectively. Lanes six Leu9 codons were constructed adjacent to Pro8 CCG with wild- 4 through 7 are samples from strains that carry the σ28 H14D substitution type codons at Thr6 (ACC) and Ser7 (UCA), and the cells were resulting in increased PflgM transcription that are used only in this gel to assessed for FlgM activity. An 11-fold range in FlgM activity levels optimize the visualization of the LacZ protein band (24). Lanes 4 and 5 are Pro8 CCU and Pro8 CCG, respectively, in a serT mutant background. Lanes was observed for the six Leu9 codons when adjacent to Pro8 CCG 6 and 7 are Pro8 CCU and Pro8 CCG, respectively, in a serT+ background. (Thr6 ACC), suggesting that translation at CCG could be context dependent on both 5′ and 3′ codons (Fig. 2B). lane 4). However, in a strain allowing increased transcription from The Severe Translation Defect at Thr6 ACC-Ser7 UCA-Pro8 CCG of FlgM the flgM promoter, the CCU–CCG codon change at Pro8 resulted Results in Ribosome Pausing in Vivo. In the tests described above, it in a substantial increase of the lower band corresponding to LacZ was assumed that the 20-fold reduction in the in vivo FlgM ac- aloneaswellasintheFlgM–LacZ fusion band (Fig. 3A,lane5). tivity for the CCU-to-CCG synonymous codon change at Pro8 The presence of the lower-MW LacZ band resulting from the was caused by reduced translation of flgM mRNA. To test this CCU–CCG synonymous change at Pro8 is consistent with a pause assumption, we set up an assay for ribosome pausing to see if the in translation at Pro8 that resulted in translation reinitiation reduction in flgM translation correlated with a pause in ribosome downstream within the flgM–lacZ coding sequence to produce the translation of Pro8 CCG in vivo. shorter LacZ band observed. To confirm this observation, a UAG The pausing assay uses a flgM–lacZ gene fusion that allows lacZ stop codon was introduced at the Pro8 position in the flgM–lacZ to be translated independently of flgM if translation is stalled at gene fusion construct. As expected, the Pro8 CCU-to-UAG (stop Pro8. The flgM and lacZ genes were placed in a single operon – where they were transcribed into a single mRNA but could be codon) change produced no full-length FlgM LacZ protein; only translated independently. Cells with this structure were pheno- the lower-MW LacZ band was observed, indicating translation of + + lacZ independent of flgM translation in the flgM–lacZ fusion typically FlgM ,LacZ , and the LacZ protein was easily identified – by Western blot using anti-LacZ antibody, as shown in Fig. 3A, construct (Fig. 3A, lane 6). This result also indicated that the flgM lane 1. The flgM and lacZ genes were fused so that the 5′ end of lacZ mRNA was stable in the absence of full flgM translation. the flgM coding sequence contained a ribosome-binding site These data are consistent with the stalling of translation at codon (RBS), and the lacZ part of the fusion started with a GTG start Pro8 CCG, which resulted in a substantial reinitiation of trans- codon. With no stalling, translation would produce a single fusion lation downstream and producing the lower-MW product. protein including both FlgM and LacZ sequences. If ribosomes We also tested the effect of the serT tRNA mutant that pre- stalled upstream in flgM at Pro8 CCG codon, this stalling would viously was shown to be defective in translating the Ser7 UCA expose the downstream RBS and allow lacZ to be translated in- codon of flgM on flgM–lacZ translation of the Pro8 CCU and dependently of flgM. When full-length flgM was fused to lacZ,the Pro8 CCG constructs (Fig. 3B). The control of a flgM–lacZ larger FlgM–LacZ protein fusion product was observed (Fig. 3A, transcriptional fusion in which flgM and lacZ are translated in- lane 2). Increased transcription of the flgM promoter allowed us to dependently (Fig. 3A, lane 1) is repeated in Fig. 3B, lane 1. As detect a small amount of LacZ in addition to the FlgM–LacZ shown in Fig. 3B, reduction in the level of FlgM–LacZ fusion was fusion (Fig. 3A, lane 3). The lower-molecular-weight (MW) LacZ observed by the Pro8 CCU–CCG change (Fig. 3B, lanes 2 and 3). band could result from translation restart at lacZ or from protein In a strain background with increased transcription from the flgM degradation. Introduction of the CCU–CCG codon change at promoter, the lower-MW LacZ band accumulates in the pres- Pro8 resulted in decreased FlgM–LacZ fusion product (Fig. 3A, ence of Pro8 CCG change (Fig. 3B, lanes 4 and 5). In the serT

Chevance and Hughes PNAS Early Edition | 3of6 Downloaded by guest on September 28, 2021 mutant we observe a decrease in FlgM–LacZ levels but not in the Thr6: Synonymous Changes that lower-MW LacZ band (Fig. 3B, lanes 6 and 7). These results ACC->ACA (6) ACC->ACG (1) Suppress the Pro8 translation defect demonstrate that the serT mutant is significantly defective in ACC->ACT (3) flgM–lacZ translation relative to translation of lacZ. Ser7: TCA->TCG (6) The Effects of Synonymous Codon Changes at Thr6 and Leu9 on FlgM TCA->TCC (6) Activity. We looked for synergistic effects of synonymous changes TCA->TCT (2) at codons Thr6 and Leu9, which are separated by two triplet Leu9 codons. In this case, there are four synonymous codons for Thr TTG->TTA (3) FlgM (ACN) and six codons for Leu (UUA/G and UCN), including 24 possible silent change combinations that do not change the MSIDRTSPLKPVSTVQTRETSDTPVQKTRQEKTSAATSASVTLSDAQAKLMQPGVSDINMERVEALKTAIRNGELKMDTGKIADSLIREAQSYLQSK* Pro8 Val25 CCG CCG->CCC (2) translated FlgM amino acid sequence. We assayed these Ser2 Pro24 ACG->AGT CCG->CCC Thr23 24 combinations for FlgM regulatory activity; the results are ACG->ACA Asp4 Thr6: Ser21 Synonymous Changes with shown in Fig. 2C. Here the range in FlgM regulatory activity GAC->GAT ACC-> (2) AGC->AGT Thr20 ACC->ACA No (or slight) phenotypic varied from a 10% reduction in FlgM activity to a 22-fold in- ACC->ACG Val12 crease in FlgM activity over wild type. For the most part, the GTT->GTG (2) Gln16 change GTT->GTC (1) CAG->CAA contributions of synonymous changes at Thr6 and Leu9 were Val15 additive. We did not observe any dramatic synergistic effects as GTC->GTG we did with synonymous combinations of Thr6 and Pro8 codons Fig. 4. Suppression of the flgM Pro8 CCG allele by doped oligonucleotide on Ser7 reading. However, we did observe small synergistic ef- synonymous codon mutagenesis of codons 2–25. In a strain unable to as- − fects resulting in an approximately twofold synergistic effect on semble the flagellar motor (a hook basal body-mutant strain, HBB ), FlgM is 28 not secreted, resulting in the inhibition of σ -dependent PfliC–lac expression FlgM activity for the Thr6 ACG substitution in combination with − the Leu9 CUU, CUC, CUA, and CUG substitutions (Fig. 2C). and a Lac phenotype on MacConkey lactose indicator plates. Introduction of the flgM Pro8 CCG results in reduced FlgM levels, increased σ28-dependent The remaining changes in FlgM activity caused by the synony- + PfliC–lac expression, and a Lac phenotype on MacConkey lactose indicator mous Thr6 and Leu9 combinations could be accounted for by plates. Doped oligonucleotide mutagenesis that introduced synonymous additive changes of the single synonymous substitutions at changes in amino acids 2–25 (excluding Pro8) was performed on the flgM Pro8 − Thr6 and Leu9. This result supports the conclusion that if the CCG, HBB mutant strain. More than 10,000 recombinant strains were screened ribosome does sense more than three consecutive triplet codons for increased FlgM activity that resulted in a Lac− phenotype. Mutants resulting during the process of translation, the effect is small. in a Lac+ phenotype are shown in red and occurred at amino acid codons Thr6, + Ser7, and Leu9. Mutants with a slight Lac phenotype are in blue and occurred − Isolation of Synonymous Suppressor Mutations of Pro8 CCG. The above at codons Ser2 and Val12. Ten independent colonies with the parental Lac data suggest that the translation of a specific codon can be influenced phenotype were sequenced and are shown in black. The numbers in paren- significantly by the adjacent two codons. Neither mRNA secondary thesis are the number of independent isolates that were sequenced. structure nor mRNA stability appeared to be a factor. However, it is very difficult to rule completely out the effects of mRNA secondary substitutions, which also suppressed the CCG translation defect structure and mRNA stability on translation.Inanattempttorule (Fig. 2B), were not included in this experiment. Only synonymous out further the effects of mRNA secondary structure and stability on changes in the two codons (Thr6 and Ser7) preceding Pro8 and translation, we set up a genetic assay to look for more long-distance the Leu9 codon following Pro8 CCG resulted in restoration to effects of flgM mRNA changes affecting codons for amino acids levels near that of wild-type FlgM inhibitory activity. These results 2 through 25 on translation of the Pro8 CCG allele of flgM. A doped oligonucleotide mutagenesis of the region of flgM support our assertion that the efficiency of translation of a specific mRNA encoding amino acids 2–25 was performed. Technical codon is dependent on the flanking two codons and that our ob- details of the doped oligonucleotide synthesis limited our experi- served effects are caused primarily by codon context and not by ment to codons 2–25. The oligonucleotide mutagenesis was designed mRNA secondary structure. so that, on average, each oligo had one synonymous codon change The Effect of Gene Location on the CCU-to-CCG Translation Defect in for amino acids 2–25. We reasoned that if the 20-fold reduced the Context of Thr-Ser-Pro-Leu-Lys. Codon bias effects on translation FlgM activity of the Pro8 CCU-to-CCG change was caused by an – effect on mRNA secondary structure, suppressors might be obtained are enriched at the N terminus of genes (9 13). We wondered if throughout the amino acid 2–25 region that significantly increased the translating ribosome might transition from a translation initi- FlgM activity. Alternatively, if translation of a given codon is ation mode through the N terminus, which is strongly influenced influenced by the two adjacent codons, we expected significant by the presence of rare codons, codon context, mRNA secondary suppression of the Pro8 CCG translation defect to occur in the structure, tRNA modifications, and other translation factors, into region including the two codons before (Thr6 and Ser7) and the a more robust translation machine that is less affected by these two codons after (Leu9 and Lys10) Pro8. factors. Recently, Boël et al. determined that the codon choice in The results of the doped synonymous mutagenesis of amino the initial 20 codons of transcripts correlated with acids 2–25 in the Pro8 CCG background are presented in Fig. 4. translation efficiency and mRNA stability (14). Synonymous changes that significantly suppressed the FlgM To test whether the translational defect of Pro8 CCG in flgM inhibitory activity of Pro8 CCG are shown in red. Changes resulting was caused by its N-terminal location and whether translational in partial suppression are shown in blue. To show that the doped reinitiation affected mRNA stability, we constructed two fusion oligonucleotide mutagenesis covered the entire region, 10 colo- constructs with the Thr6-Ser7-Pro8-Leu9-Lys10 codons of flgM nies that showed no suppression of the Pro8 CCG allele were at different gene locations with either CCU or CCG at the Pro sequenced and are shown in black. position. One construct placed codons 1–85 of flgM to a lacZ The synonymous changes that suppressed the Pro8 CCG alleles under control of an arabinose-inducible promoter to remove any to give wild-type FlgM activity included all three possible synon- possible autoregulatory effects of flgM on its own transcription ymous substitutions at Thr6, all three possible synonymous sub- and lacked both the RBS region at the C terminus of flgM and a stitutions at Ser7, and the UUG-to-UUA substitution at Leu9, separate start codon for lacZ (Fig. 5). The second construct was which we earlier showed increased FlgM activity of the Pro8 CCG identical except that codons 1–60 of flgM in the fusion were allele sixfold. For technical reasons, the CUN Leu9 synonymous replaced with codons 1–60 of fliA. By measuring β-gal activity

4of6 | www.pnas.org/cgi/doi/10.1073/pnas.1614896114 Chevance and Hughes Downloaded by guest on September 28, 2021 under arabinose-inducing conditions, we are measuring a direct effects of small changes in flgM mRNA translation on the ex- effect of the Pro8 CCG change on flgM–lacZ translation. pression of reporter constructs. We analyzed synonymous codon- For the flgM(1–85)–lacZ construct, a change from Pro8 CCU context effects on translation and conclude that the translation to CCG reduced β-gal activity about sevenfold, but in this case of a given codon can be influenced significantly by up to two mRNA levels also were reduced by about fourfold (Fig. 5A). flanking codons. That is, when EF-Tu bound to an aminoacyl- Insertion of a duplicated copy of codons Thr6 through Lys10 of tRNA enters the A site at a particular codon, two preceding flgM after amino acid codon 26 still produced a translation defect codons can affect the correct-fit determination by EF-Tu and the of 30% when the middle Pro codon was changed from CCU to resulting translation efficiency of that codon. In a similar fashion, CCG in the inserted fragment. The insertion of flgM codons that codon can influence the translation of the next two codons Thr6 through Lys10 alone resulted in a twofold reduction in being translated. mRNA levels, but here the CCU to CCG change had no effect Since 1980 investigation of the effects of codon context on on mRNA levels. Inserting flgM codons Thr6 through Lys10 after translation has focused primarily on codon pairs. In their seminal codon 56 had no effect on either β-gal activity or mRNA levels. paper, Bossi and Roth determined that the translation of a UAG For the fliA(1–60)–flgM(61–85)–lacZ construct, flgM codons codon by an amber suppresser tRNA was highly influenced by Thr6 through Lys10 were separately inserted after amino acids 5, the adjacent codons (15). It has been known for decades that highly expressed genes in bacteria and yeast exhibit codon pair 26, and 56 with either the CCU or CCG proline. Both β-gal and biases. Modern high-throughput technologies (genetic selection mRNA levels were reduced similarly with Pro-CCG at all three being the classical high-throughput technology) can identify ad- positions. These results support our hypothesis that translation of jacent codon pairs through tens of thousands of combinations a given codon can be influenced by the two preceding codons and that modulate translation efficiency (16). Changing codon-pair that a given codon can influence the translation of the following biases on a genomic scale has proven an effective way to atten- two codons, at least within the first 60 codons of a gene. uate viruses in making effective live vaccines (17). We assume Discussion that the effects of biased codon pairs are caused by changes in the translation speed of various proteins (4). Similar methods We describe a sensitive translation assay involving a regulatory that include codon pair speeds in designing attenuated essential gene, flgM, and show that translation of a particular codon is genes may aid in the design of live bacterial vaccines. heavily influenced by two adjacent codons 5′ and one adjacent

Our data suggest that in thinking about context effects one GENETICS ′ codon 3 to the codon being translated. The expression assays should consider codon triplets in addition to codon pairs. Our used in this study took advantage of FlgM being a regulatory data demonstrate that the translation of a specific codon can 28 gene that works in a one:one stoichiometry with the σ tran- depend on the two codons flanking the codon being translated in scription factor. Thus, small changes in FlgM protein levels a nonadditive manner. After the second codon of a given gene, 28 would result in signal amplification on σ -dependent transcrip- both the P-site and E-site of the ribosome are occupied when tion. This amplification of signal allowed us to observe the elongation factor EF-Tu bound to a charged aminoacyl-tRNA species brings nascent charged tRNA to an empty A site. We propose that upon docking an aminoacyl-tRNA to the A site, the efficiency with which EF-Tu translates that codon is influenced flgM lacZ fliA flgM lacZ by adjacent codons and bound tRNA species in addition to the A Para B Para AA85 AA85 – Pro8 specific codon anticodon interaction at the A site. 26 56 5 26 56 (flgM Thr6-Lys10 codon insertion) (flgM Thr6-Lys10 codon insertion) *** ns * Our working model for the effect of two adjacent codons on specific codon translation is diagrammed in Fig. 1. In our model, *** GTP-bound elongation factor EF-Tu (purple) directs a charged * tRNA to the A site of the ribosome through Watson–Crick

* pairing of the first two bases in the translating triplet. The effi- ciency with which this codon is translated depends on the sta- bility of the anticodon–codon interactions. We propose that this stability is dependent on base-stacking energy, which is influ- enced by at least the two flanking codons in the mRNA. We also ns – *** ns propose that the stability of the anticodon codon interactions is * influenced by modifications of the tRNAs that affect interaction ns with the ribosomal A, P, and E sites and by protein or noncoding

** RNA translation factors. Synonymous changes will lead to differences in translation rates that, especially when different tRNAs are used, have different

FlgM FlgM FlgM26 FlgM26 FlgM56 FlgM56 FliA5 FliA5 FliA26 FliA26 FliA56 FliA56 binding efficiencies, abundances, and charging rates and result in CCU CCG CCU CCG CCU CCG CCU CCG CCU CCG CCU CCG differential mRNA stabilities. In addition, the same tRNA reads Fig. 5. Position effects of the Pro8 CCU-to-CCG change on the expression different codons with different efficiencies, as was determined in and mRNA stability of flgM(codons 1–85)–lacZ and fliA(codons 1–60)– an in vivo translational speedometer assay system (4). We con- flgM(codons 61–85)–lacZ gene fusions. (A) Fusions of codons 1–85 of flgM to lacZ clude that codon recognition is initiated by codon–anticodon hy- were made without and with duplicate insertions of codons Thr6-Ser7-Pro8- drogen bonding between the first and second bases of the Leu9-Lys10 after amino acid codons 26 and 56. (B) Fusions of codons 1–60 of translated codon followed by sensing the correct fit at the wobble fliA fused to codons 61–85 of flgM fused to lacZ were made with insertions base and base-stacking interactions contributed by the preceding of flgM codons Thr6-Ser7-Pro8-Leu9-Lys10 after amino acid codons 5, 26, two codons and bound tRNAs. and 56. The effects of the synonymous Pro8 CCU-to-CCG change on the The tRNA molecules of every organism are modified exten- expression and mRNA stability for all six constructs were determined. The sively, and the majority of modifications occur at the antiwobble fusions are deleted for the C-terminal region of flgM that contains an RBS, ′ and no ATG or GTG start codon precedes the lacZ-coding segment of each position of the anticodon loop and at the base immediately 3 to fusion. Expression of the different fusion constructs was determined by β-Gal the anticodon (18). [Thirteen other bases positions are modified assay under arabinose inducing conditions and mRNA levels were de- to a lesser extent in tRNA species of E. coli and Salmonella termined as described in SI Materials and Methods. enterica (7).] The base adjacent to the 3′ anticodon position, the

Chevance and Hughes PNAS Early Edition | 5of6 Downloaded by guest on September 28, 2021 “cardinal nucleotide,” also varies among species and is thought dependent on the 5′ and 3′ flanking codons, and the flanking to affect codon recognition significantly (19). These modifica- codons are dependent on their 5′ and 3′ flanking codons, then tions influence the stacking energy of the bases during codon– selection pressure on a single codon is exerted over five successive anticodon pairing (3). The translation proofreading steps cata- codons, which represent 615 or 844,596,301 codon combinations. “ ” lyzed by EF-Tu and EF-G, which sense hydrogen bonding and If modified tRNAs interact with bases in a codon context- – stacking energy to determine if the correct codon anticodon dependent manner that differs among species depending on dif- pairing has occurred, are influenced by the adjacent codons, ferences in tRNA modifications, ribosome sequences, and ribo- possibly resulting in the codon-context effects we observe. More- over, many tRNA-modifying proteins are present in only one of the somal and translation factor proteins, it is easy to understand why three kingdoms of life (1). Thus, specific tRNA modifications that many genes are poorly expressed in heterologous expression sys- affect wobble base recognition and contribute to the base-stacking tems in which codon use is the primary factor in the design of forces during translation can determine specific codon-context ef- coding sequences for foreign protein expression. The potential fects by adjacent synonymous codons on specific codon translation. impact of differences in tRNA modifications represents a signifi- Such effects of specific tRNA modifications on codon translation cant challenge in designing genes for maximal expression whether could account for the different codon pair biases observed in by natural selection or in the laboratory. species that are evolutionarily distant (possessing different specific tRNA modifications) and also could account for the difficulty in Materials and Methods expressing proteins in heterologous systems, i.e., expressing pro- Strains and oligonucleotides used in this work are listed in SI Materials and teins from plant and mammalian systems in bacteria. The MiaA Methods. Strain constructions and quantitative real-time PCR were performed (i6A37) modification has recently been shown to affect mRNA as described in SI Materials and Methods. Expression of luciferase was mea- translation in E. coli in a codon context-dependent manner, sup- sured as luminescence levels on a POLARstar OPTIMA luminometer from BMG porting our overall hypothesis (20). The translation of proline Labtech. β-Gal activities were determined as described previously (23). codons in the mgtL peptide transcript of Salmonella was recently shown to be affected by mutations defective in ribosomal proteins ACKNOWLEDGMENTS. We thank Premal Shah and Joshua Plotkin at the L27, L31, elongation factor EF-P, and TrmD, which catalyzes the University of Pennsylvania for the mRNA secondary structural analysis; John m1G37 methylation of proline tRNA (21). Modification of tRNA Roth, Lionello Bossi, Nara Figueroa-Bossi, Winfried Boos, Urs Jenal, and Jon species in E. coli also has been shown to vary with the growth phase Seger for helpful discussions; Henri Grosjean for educating us on the role of base stacking and tRNA modifications in mRNA translation; and the Univer- of the cell (22). Specific codon-context effects could represent sity of Utah 2014 Microbial Genetics BIO5255 class for their help in producing translation domains of life based on tRNA modifications. the data presented in Fig. 4. This work was supported by start-up funds from The difficulty for natural selection would be in finding codon the University of Utah and Public Health Service Grant GM056141 from the optimization for a given gene. If the speed through a codon is National Institutes of Health.

1. Grosjean H, de Crécy-Lagard V, Marck C (2010) Deciphering synonymous codons in the 18. El Yacoubi B, Bailly M, de Crécy-Lagard V (2012) Biosynthesis and function of post- three domains of life: Co-evolution with specific tRNA modification enzymes. FEBS transcriptional modifications of transfer RNAs. Annu Rev Genet 46:69–95. Lett 584:252–264. 19. Yarus M (1982) Translational efficiency of transfer RNA’s: Uses of an extended anti- 2. Steitz TA (2008) A structural understanding of the dynamic ribosome machine. Nat codon. Science 218:646–652. Rev Mol Cell Biol 9:242–253. 20. Aubee JI, Olu M, Thompson KM (2016) The i6A37 tRNA modification is essential for 3. Grosjean HJ, de Henau S, Crothers DM (1978) On the physical basis for ambiguity in proper decoding of UUX-Leucine codons during rpoS and iraP translation. RNA 22: – genetic coding interactions. Proc Natl Acad Sci USA 75:610–614. 729 742. 2+ 4. Chevance FF, Le Guyon S, Hughes KT (2014) The effects of codon context on in vivo 21. Gall AR, et al. (2016) Mg regulates transcription of mgtA in Salmonella Typhimu- translation speed. PLoS Genet 10:e1004392. rium via translation of proline codons during synthesis of the MgtL peptide. Proc Natl – 5. Chevance FF, Hughes KT (2008) Coordinating assembly of a bacterial macromolecular Acad Sci USA 113:15096 15101. machine. Nat Rev Microbiol 6:455–465. 22. Sakai Y, Miyauchi K, Kimura S, Suzuki T (2016) Biogenesis and growth phase- 6. Erhardt M, Namba K, Hughes KT (2010) Bacterial nanomachines: The flagellum and dependent alteration of 5-methoxycarbonylmethoxyuridine in tRNA anticodons. – type III injectisome. Cold Spring Harb Perspect Biol 2:a000299. Nucleic Acids Res 44:509 523. 7. Bjork GR, Hagervall TG (2014) Transfer RNA modifications: Presence, synthesis, and 23. Maloy SR (1990) Experimental Techniques in Bacterial Genetics (Jones and Bartlett, Boston). function. EcoSalPlus: Cellular and Molecular Biology of E. coli and Salmonella, and the 24. Chadsey MS, Hughes KT (2001) A multipartite interaction between Salmonella tran- Enterobacteriaceae, ed Lovett ST (ASM, Washington, D.C.). scription factor sigma28 and its anti-sigma factor FlgM: Implications for sigma28 8. Chevance FF, Karlinsey JE, Wozniak CE, Hughes KT (2006) A little gene with big ef- holoenzyme destabilization through stepwise binding. J Mol Biol 306:915–929. fects: A serT mutant is defective in flgM gene translation. J Bacteriol 188:297–304. 25. Lorenz R, et al. (2011) ViennaRNA Package 2.0. Algorithms Mol Biol 6:26. 9. Tuller T, et al. (2010) An evolutionarily conserved mechanism for controlling the ef- 26. Middleton SA, Kim J (2014) NoFold: RNA structure clustering without folding or ficiency of protein translation. Cell 141:344–354. alignment. RNA 20(11):1671–1683. 10. Pechmann S, Frydman J (2013) Evolutionary conservation of codon optimality reveals 27. Sanderson KE, Roth JR (1983) Linkage map of Salmonella typhimurium, Edition VI. – hidden signatures of cotranslational folding. Nat Struct Mol Biol 20:237 243. Microbiol Rev 47:410–453. 11. Li G-W, Oh E, Weissman JS (2012) The anti-Shine-Dalgarno sequence drives trans- 28. Karlinsey JE, Hughes KT (2006) Genetic transplantation: Salmonella enterica serovar – lational pausing and codon choice in bacteria. Nature 484:538 541. Typhimurium as a host to study sigma factor and anti-sigma factor interactions in 12. Bentele K, Saffert P, Rauscher R, Ignatova Z, Blüthgen N (2013) Efficient translation genetically intractable systems. J Bacteriol 188:103–114. initiation dictates codon usage at gene start. Mol Syst Biol 9:675. 29. Groisman EA (1991) In vivo genetic engineering with bacteriophage Mu. Methods 13. Goodman DB, Church GM, Kosuri S (2013) Causes and effects of N-terminal codon bias Enzymol 204:180–212. – in bacterial genes. Science 342:475 479. 30. Gillen KL, Hughes KT (1993) Transcription from two promoters and autoregulation 14. Boël G, et al. (2016) Codon influence on protein expression in E. coli correlates with contribute to the control of expression of the Salmonella typhimurium flagellar mRNA levels. Nature 529:358–363. regulatory gene flgM. J Bacteriol 175:7006–7015. 15. Bossi L, Ruth JR (1980) The influence of codon context on genetic code translation. 31. Karlinsey JE (2007) lambda-Red genetic engineering in Salmonella enterica serovar Nature 286:123–127. Typhimurium. Methods Enzymol 421:199–209. 16. Gamble CE, Brule CE, Dean KM, Fields S, Grayhack EJ (2016) Adjacent codons act in 32. Goodier RI, Ahmer BM (2001) SirA orthologs affect both motility and virulence. concert to modulate translation efficiency in yeast. Cell 166:679–690. J Bacteriol 183:2249–2258. 17. Coleman JR, et al. (2008) Virus attenuation by genome-scale changes in codon pair 33. Pfaffl MW (2001) A new mathematical model for relative quantification in real-time bias. Science 320:1784–1787. RT-PCR. Nucleic Acids Res 29:e45.

6of6 | www.pnas.org/cgi/doi/10.1073/pnas.1614896114 Chevance and Hughes Downloaded by guest on September 28, 2021