High Codon Bias Research Articles

Codon-tRNA Coadaptation Bias for Identifying Strong Native Promoters in Komagataella phaffii.

Promoters are crucial elements for engineering microbial production strains used in bioprocesses. For the increasingly popular chassis Komagataella phaffii (formerly Pichia pastoris), a limited number of well-characterized promoters constrain the data-driven engineering of production strains. Here, we present an in silico approach for condition-independent de novo identification of strong native promoters. The method relies on tRNA-codon coadaptation of coding sequences in the K. phaffii genome and is based on two complementary scores: the number of effective codons and the tRNA adaptation index. Genes with high codon bias are expected to be translated efficiently and, thus, also be under control of strong promoters. Using this approach, we identified promising strong promoter candidates and experimentally assessed their activity using fluorescent reporter assays characterizing 50 promoters spanning a 76-fold difference in expression levels in a glucose medium. Overall, we report several promoters that should be added to the molecular toolbox for engineering of K. phaffii and present an approach for identifying promoters in microbial genomes.

Analysis of evolutionary and genetic patterns in structural genes of primate lentiviruses.

BackgroundPrimate lentiviruses (HIV1, HIV2, and Simian immunodeficiency virus [SIV]) cause immune deficiency, encephalitis, and infectious anemia in mammals such as cattle, cat, goat, sheep, horse, and puma.ObjectiveThis study was designed and conducted with the main purpose of confirming the overall codon usage pattern of primate lentiviruses and exploring the evolutionary and genetic characteristics commonly or specifically expressed in HIV1, HIV2, and SIV.MethodsThe gag, pol, and env gene sequences of HIV1, HIV2, and SIV were analyzed to determine their evolutionary relationships, nucleotide compositions, codon usage patterns, neutrality, selection pressure (influence of mutational pressure and natural selection), and viral adaptation to human codon usage.ResultsA strong ‘A’ bias was confirmed in all three structural genes, consistent with previous findings regarding HIV. Notably, the ENC-GC3s plot and neutral evolution analysis showed that all primate lentiviruses were more affected by selection pressure than by mutation caused by the GC composition of the gene, consistent with prior reports regarding HIV1. The overall codon usage bias of pol was highest among the structural genes, while the codon usage bias of env was lowest. The virus groups showing high codon bias in all three genes were HIV1 and SIVcolobus. The codon adaptation index (CAI) and similarity D(A, B) values indicated that although there was a high degree of similarity to human codon usage in all three structural genes of HIV, this similarity was not caused by translation pressure. In addition, compared with HIV1, the codon usage of HIV2 is more similar to the human codon usage, but the overall codon usage bias is lower.ConclusionThe origin viruses of HIV (SIVcpz_gor and SIVsmm) exhibit greater similarity to human codon usage in the gag gene, confirming their robust adaptability to human codon usage. Therefore, HIV1 and HIV2 may have evolved to avoid human codon usage by selection pressure in the gag gene after interspecies transmission from SIV hosts to humans. By overcoming safety and stability issues, information from codon usage analysis will be useful for attenuated HIV1 vaccine development. A recoded HIV1 variant can be used as a vaccine vector or in immunotherapy to induce specific innate immune responses. Further research regarding HIV1 dinucleotide usage and codon pair usage will facilitate new approaches to the treatment of AIDS.

Evolutionary Dynamics of the SKN-1 → MED → END-1,3 Regulatory Gene Cascade inCaenorhabditisEndoderm Specification

Gene regulatory networks and their evolution are important in the study of animal development. In the nematode, Caenorhabditis elegans, the endoderm (gut) is generated from a single embryonic precursor, E. Gut is specified by the maternal factor SKN-1, which activates the MED → END-1,3 → ELT-2,7 cascade of GATA transcription factors. In this work, genome sequences from over two dozen species within the Caenorhabditis genus are used to identify MED and END-1,3 orthologs. Predictions are validated by comparison of gene structure, protein conservation, and putative cis-regulatory sites. All three factors occur together, but only within the Elegans supergroup, suggesting they originated at its base. The MED factors are the most diverse and exhibit an unexpectedly extensive gene amplification. In contrast, the highly conserved END-1 orthologs are unique in nearly all species and share extended regions of conservation. The END-1,3 proteins share a region upstream of their zinc finger and an unusual amino-terminal poly-serine domain exhibiting high codon bias. Compared with END-1, the END-3 proteins are otherwise less conserved as a group and are typically found as paralogous duplicates. Hence, all three factors are under different evolutionary constraints. Promoter comparisons identify motifs that suggest the SKN-1, MED, and END factors function in a similar gut specification network across the Elegans supergroup that has been conserved for tens of millions of years. A model is proposed to account for the rapid origin of this essential kernel in the gut specification network, by the upstream intercalation of duplicate genes into a simpler ancestral network.

Comparative genome‐wide analysis of codon usage of different bacterial species infecting Oryza sativa

Oryza sativa is vastly affected by microbial pathogen, causing blight-related diseases, which in turn deplete the growth and productivity of rice. In this study, we analyzed four bacterial rice pathogen genomes and reported on their codon usage that might have greater implication in mutation-related research. Differential codon usage indices, such as codon adaptation index (CAI), codon bias index (CBI), effective number of codons (ENc), relative synonymous codon usage (RSCU), correspondence analysis (COA), and parity plots, were applied on coding sequences of Pseudomonas fuscovaginae, Pseudomonas syringae, Xanthomonas oryzae, and Pseudomonas avenae speices. The RSCU results proposed a high-frequency usage of CUG and CGC that codes for leucine and arginine in all of the species. The CBI and CAI values between the genomes range from 0.17 to 0.3 and from 0.26 to 0.35, respectively, indicating a direct proportionality between these indexes. The mean ENc value of P. avenae coding sequence showed high codon bias compared with other genomes. The axis I variation from COA analysis shows a mean value of 42.28% codon variations in these bacterial species. Correlation studies between axis I and ENc-GC3, along with CAI and CBI, suggested the presence of nucleotide bias and mutational pressure as major forces for codon bias within these species. Hence, certain genes with high CAI-CBI have been correlated for better gene expression. Our study highlights the importance of nucleotide biasness, mutation pressure, and natural selection in shaping protein-coding genes in these four rice-affecting bacteria. This would further help in investigating the evolution of pathogenic gene families, which may direct research toward synthetic genes that could be suppressed or overrepresented based on their codon usage pattern toward pathogenicity.

Evidence that Natural Selection on Codon Usage inDrosophila pseudoobscuraVaries Across Codons

Like other species of Drosophila, Drosophila pseudoobscura has a distinct bias toward the usage of C- and G-ending codons. Previous studies have indicated that this bias is due, at least in part, to natural selection. Codon bias clearly differs among amino acids (and other codon classes) in Drosophila, which may reflect differences in the intensity of selection on codon usage. Ongoing natural selection on synonymous codon usage should be reflected in the shapes of the site frequency spectra of derived states at polymorphic positions. Specifically, regardless of other demographic effects on the spectrum, it should be shifted toward higher values for changes from less-preferred to more-preferred codons, and toward lower values for the converse. If the intensity of natural selection is increased, shifts in the site frequency spectra should be more pronounced. A total of 33,729 synonymous polymorphic sites on Chromosome 2 in D. pseudoobscura were analyzed. Shifts in the site frequency spectra are consistent with differential intensity of natural selection on codon usage, with stronger shifts associated with higher codon bias. The shifts, in general, are greater for polymorphic synonymous sites than for polymorphic intron sites, also consistent with natural selection. However, unlike observations in D. melanogaster, codon bias is not reduced in areas of low recombination in D. pseudoobscura; the site frequency spectrum signal for selection on codon usage remains strong in these regions. However, diversity is reduced, as expected. It is possible that estimates of low recombination reflect a recent change in recombination rate.

Molecular and physicochemical characterization of Metallothionein genes of mangroves–a computational approach

Metallothioneins (MTs) are a family of low-molecular-weight cysteine-rich multifunctional proteins involved in metal stress response. Heavy metals are non-biodegradable and are thus persistent in the marine environment. Despite the confirmation of the presence of MT genes in mangrove plants, relatively little is known about their expression, function and mechanisms. Therefore, a bio-computational study was carried out to understand the physicochemical properties and molecular characteristics of the MT gene of mangrove plant. The theoretical isoelectric points of the proteins were below 7.0, with no signal peptides and the instability index classified the protein as unstable. MTs were mainly localized in cell nucleus (over 52 %), relatively rich in random coil, and transmembrane helices were absent. High genetic diversity and strong codon bias was found. Moreover, the MT genes were found under positive selection during molecular evolution. The phylogenetic tree revealed that the genetic relationship coincided with class and type of metallothionein. These results could provide further insight in the evolution and molecular mechanism of MT gene and their biological nature.

Antagonistic relationships between intron content and codon usage bias of genes in three mosquito species: functional and evolutionary implications

Genome biology of mosquitoes holds potential in developing knowledge-based control strategies against vectorborne diseases such as malaria, dengue, West Nile, and others. Although the genomes of three major vector mosquitoes have been sequenced, attempts to elucidate the relationship between intron and codon usage bias across species in phylogenetic contexts are limited. In this study, we investigated the relationship between intron content and codon bias of orthologous genes among three vector mosquito species. We found an antagonistic relationship between codon usage bias and the intron number of genes in each mosquito species. The pattern is further evident among the intronless and the intron-containing orthologous genes associated with either low or high codon bias among the three species. Furthermore, the covariance between codon bias and intron number has a directional component associated with the species phylogeny when compared with other nonmosquito insects. By applying a maximum likelihood–based continuous regression method, we show that codon bias and intron content of genes vary among the insects in a phylogeny-dependent manner, but with no evidence of adaptive radiation or species-specific adaptation. We discuss the functional and evolutionary significance of antagonistic relationships between intron content and codon bias.

Rate-Limiting Steps in Yeast Protein Translation

SummaryDeep sequencing now provides detailed snapshots of ribosome occupancy on mRNAs. We leverage these data to parameterize a computational model of translation, keeping track of every ribosome, tRNA, and mRNA molecule in a yeast cell. We determine the parameter regimes in which fast initiation or high codon bias in a transgene increases protein yield and infer the initiation rates of endogenous Saccharomyces cerevisiae genes, which vary by several orders of magnitude and correlate with 5′ mRNA folding energies. Our model recapitulates the previously reported 5′-to-3′ ramp of decreasing ribosome densities, although our analysis shows that this ramp is caused by rapid initiation of short genes rather than slow codons at the start of transcripts. We conclude that protein production in healthy yeast cells is typically limited by the availability of free ribosomes, whereas protein production under periods of stress can sometimes be rescued by reducing initiation or elongation rates.

A unique DNA repair and recombination gene (recN) sequence for identification and intraspecific molecular typing of bacterial wilt pathogen Ralstonia solanacearum and its comparative analysis with ribosomal DNA sequences

Ribosomal gene sequences are a popular choice for identification of bacterial species and, often, for making phylogenetic interpretations. Although very popular, the sequences of 16S rDNA and 16-23S intergenic sequences often fail to differentiate closely related species of bacteria. The availability of complete genome sequences of bacteria, in the recent years, has accelerated the search for new genome targets for phylogenetic interpretations. The recently published full genome data of nine strains of R. solanacearum, which causes bacterial wilt of crop plants, has provided enormous genomic choices for phylogenetic analysis in this globally important plant pathogen. We have compared a gene candidate recN, which codes for DNA repair and recombination function, with 16S rDNA/16-23S intergenic ribosomal gene sequences for identification and intraspecific phylogenetic interpretations in R. solanacearum. recN gene sequence analysis of R. solanacearum revealed subgroups within phylotypes (or newly proposed species within plant pathogenic genus, Ralstonia), indicating its usefulness for intraspecific genotyping. The taxonomic discriminatory power of recN gene sequence was found to be superior to ribosomal DNA sequences. In all, the recN-sequence-based phylogenetic tree generated with the Bayesian model depicted 21 haplotypes against 15 and 13 haplotypes obtained with 16S rDNA and 16-23S rDNA intergenic sequences, respectively. Besides this, we have observed high percentage of polymorphic sites (S 23.04%), high rate of mutations (Eta 276) and high codon bias index (CBI 0.60), which makes the recN an ideal gene candidate for intraspecific molecular typing of this important plant pathogen.

Easy expression of the C-terminal heavy chain domain of botulinum neurotoxin serotype A as a vaccine candidate using a bi-cistronic baculovirus system

Clostridial botulinum neurotoxin (BoNT) is one of the most toxic proteins causing the food borne disease, botulism. In previous studies, recombinant BoNT production by Escherichia coli and yeast Pichia pastoris has been hampered by high AT content and codon bias in the gene encoding BoNT and required a synthetic gene to resolve this intrinsic bottleneck. This paper reports the simultaneous expression of the C-terminal heavy chain domain of BoNT (rBoNT/A-HC-6h) and enhanced green fluorescent protein (EGFP) using a bi-cistronic baculovirus-insect cell expression system. The expression of EGFP facilitated the monitoring of viral infection, virus titer determination, and isolation of the recombinant virus. Protein fusion with hexa-His-tag and one-step immobilized metal-ion affinity chromatography (IMAC) purification produced a homogenous, stable, and immunologically active 55-kDa rBoNT/A-HC-6h (about 3mg/L) with >90% purity. Furthermore, measured levels of serum titers were 8-folds for mice vaccinated with the purified rBoNT/A-HC-6h (2μg) than for mice administered with botulinum toxoid after initial immunization. Challenge experiment with botulinum A toxin demonstrated the immunoprotective activity of purified rBoNT/A-HC-6h providing the mice full protection against 102 LD50 botulinum A toxin with a dose as low as 0.2μg. This study provided supportive evidence for the use of a bi-cistronic baculovirus-Sf21 insect cell expression system in the facile expression of an immunogenically active rBoNT/A-HC.

Insight into pattern of codon biasness and nucleotide base usage in serotonin receptor gene family from different mammalian species

5-HT (5-Hydroxy-tryptamine) or serotonin receptors are found both in central and peripheral nervous system as well as in non-neuronal tissues. In the animal and human nervous system, serotonin produces various functional effects through a variety of membrane bound receptors. In this study, we focus on 5-HT receptor family from different mammals and examined the factors that account for codon and nucleotide usage variation. A total of 110 homologous coding sequences from 11 different mammalian species were analyzed using relative synonymous codon usage (RSCU), correspondence analysis (COA) and hierarchical cluster analysis together with nucleotide base usage frequency of chemically similar amino acid codons. The mean effective number of codon (ENc) value of 37.06 for 5-HT6 shows very high codon bias within the family and may be due to high selective translational efficiency. The COA and Spearman's rank correlation reveals that the nucleotide compositional mutation bias as the major factors influencing the codon usage in serotonin receptor genes. The hierarchical cluster analysis suggests that gene function is another dominant factor that affects the codon usage bias, while species is a minor factor. Nucleotide base usage was reported using Goldman, Engelman, Stietz (GES) scale reveals the presence of high uracil (>45%) content at functionally important hydrophobic regions. Our in silico approach will certainly help for further investigations on critical inference on evolution, structure, function and gene expression aspects of 5-HT receptors family which are potential antipsychotic drug targets.

Gene expression profile of the cynobacterium synechocystis genome

The expression of functional proteins plays a crucial role in modern biotechnology. The free-living cynobacterium Synechocystis PCC 6803 is an interesting model organism to study oxygenic photosynthesis as well as other metabolic processes. Here we analyze a gene expression profiling methodology, RCBS (the scores of relative codon usage bias) to elucidate expression patterns of genes in the Synechocystis genome. To assess the predictive performance of the methodology, we propose a simple algorithm to calculate the threshold score to identify the highly expressed genes in a genome. Analysis of differential expression of the genes of this genome reveals that most of the genes in photosynthesis and respiration belong to the highly expressed category. The other genes with the higher predicted expression level include ribosomal proteins, translation processing factors and many hypothetical proteins. Only 9.5% genes are identified as highly expressed genes and we observe that highly expressed genes in Synechocystis genome often have strong compositional bias in terms of codon usage. An important application concerns the automatic detection of a set of impact codons and genes that are highly expressed tend to use this narrow set of preferred codons and display high codon bias .We further observe a strong correlation between RCBS and protein length indicating natural selection in favor of shorter genes to be expressed at higher level. The better correlations of RCBS with 2D electrophoresis and microarray data for heat shock proteins compared to the expression measure based on codon usage difference, E(g) and codon adaptive index, CAI indicate that the genomic expression profile available in our method can be applied in a meaningful way to study the mRNA expression patterns, which are by themselves necessary for the quantitative description of the biological states.

Bioinformatics analysis of rabbit haemorrhagic disease virus genome

BackgroundRabbit haemorrhagic disease virus (RHDV), as the pathogeny of Rabbit haemorrhagic disease, can cause a highly infectious and often fatal disease only affecting wild and domestic rabbits. Recent researches revealed that it, as one number of the Caliciviridae, has some specialties in its genome, its reproduction and so on.ResultsIn this report, we firstly analyzed its genome and two open reading frameworks (ORFs) from this aspect of codon usage bias. Our researches indicated that mutation pressure rather than natural is the most important determinant in RHDV with high codon bias, and the codon usage bias is nearly contrary between ORF1 and ORF2, which is maybe one of factors regulating the expression of VP60 (encoding by ORF1) and VP10 (encoding by ORF2). Furthermore, negative selective constraints on the RHDV whole genome implied that VP10 played an important role in RHDV lifecycle.ConclusionsWe conjectured that VP10 might be beneficial for the replication, release or both of virus by inducing infected cell apoptosis initiate by RHDV. According to the results of the principal component analysis for ORF2 of RSCU, we firstly separated 30 RHDV into two genotypes, and the ENC values indicated ORF1 and ORF2 were independent among the evolution of RHDV.

Analysis of synonymous codon usage in hepatitis A virus.

BackgroundHepatitis A virus is the causative agent of type A viral hepatitis, which causes occasional acute hepatitis. Nevertheless, little information about synonymous codon usage pattern of HAV genome in the process of its evolution is available. In this study, the key genetic determinants of codon usage in HAV were examined.ResultsThe overall extent of codon usage bias in HAV is high in Picornaviridae. And the patterns of synonymous codon usage are quite different in HAV genomes from different location. The base composition is closely correlated with codon usage bias. Furthermore, the most important determinant that results in such a high codon bias in HAV is mutation pressure rather than natural selection.ConclusionsHAV presents a higher codon usage bias than other members of Picornaviridae. Compositional constraint is a significant element that influences the variation of synonymous codon usage in HAV genome. Besides, mutation pressure is supposed to be the major factor shaping the hyperendemic codon usage pattern of HAV.

Genomic heterogeneity within conserved metabolic pathways of Arthrobacter species - a bioinformatic approach

A comparative genomic analysis of three species of the soil bacterium Arthrobacter was undertaken with specific emphasis on genes involved in important and core energy metabolism pathways like glycolysis and amino acid metabolism. During the course of this study, it was revealed that codon bias of a particular species, namely Arthrobacter aurescens TC1, is significantly lower than that of the other two species A. chlorophenolicus A6 and Arthrobacter sp. FB24. The codon bias was also found to be negatively correlated with gene expression level which is determined by computing codon adaptation index of the genes. Uniformity in codon usage pattern among three species is evident in terms of genes which has high codon bias and multifunctional nature. Further, it was observed that this trend is present amongst the genes of important metabolic pathways, such as glycolysis and amino acid metabolism. The evolutionary divergence of the pathway gene sequences was calculated and was found to be equivalent in nature in the case of Arthrobacter sp. FB24 and Arthrobacter chlorophenolicus A6, but turned out to be dissimilar in the case of Arthrobacter aurescens TC1. A strong correlation between synonymous substitution rate and effective codon number or Nc was also observed. These observations clearly point out that the genes having low bias, in Arthrobacter aurescens TC1, and even of those that are part of highly conserved metabolic pathways like glycolysis and amino acid ensemble pathways have undergone a different type of evolution and might be subjected to positive selection pressure in comparison with Arthrobacter sp. FB24 and Arthrobacter chlorophenolicus A6.

Predictive role of mitochondrial genome in the stress resistance of insects and nematodes

Certain insects (e.g., moths and butterflies; order Lepidoptera) and nematodes are considered as excellent experimental models to study the cellular stress signaling mechanisms since these organisms are far more stress-resistant as compared to mammalian system. Multiple factors have been implicated in this unusual response, including the oxidative stress response mechanisms. Radiation or chemical-induced mitochondrial oxidative stress occurs through damage caused to the components of electron transport chain (ETC) leading to leakage of electrons and generation of superoxide radicals. This may be countered through quick replacement of damaged mitochondrial proteins by upregulated expression. Since the ETC comprises of various proteins coded by mitochondrial DNA, variation in the composition, expressivity and regulation of mitochondrial genome could greatly influence mitochondrial role under oxidative stress conditions. Therefore, we carried out in silico analysis of mitochondrial DNA in these organisms and compared it with that of the stress-sensitive humans/mammals. Parameters such as mitochondrial genome organization, codon bias, gene expressivity and GC(3) content were studied. Gene arrangement and Shine-Dalgarno (SD) sequence patterns indicating translational regulation were distinct in insect and nematodes as compared to humans. A higher codon bias (ENC≫35) and lower GC(3) content (≫0.20) were observed in mitochondrial genes of insect and nematodes as compared to humans (ENC>42; GC3>0.20), coupled with low codon adaptation index among insects. These features indeed favour higher expressivity of mitochondrial proteins and might help maintain the mitochondrial physiology under stress conditions. Therefore, our study indicates that mitochondrial genome organization may influence stress-resistance of insects and nematodes.

Codon usage bias in herpesvirus

In this study, I present a comprehensive analysis of codon usage bias in 43 herpesviruses for which the whole genome has been sequenced. The values of the effective number of codons revealed that the majority of the herpesviruses did not have high codon bias, with the exceptions of only simplexviruses and some varicelloviruses. The results of correspondence analysis, a plot of ENC and GC content at the third position of the codon and analysis of preferred codons all revealed that the codon bias was mainly influenced by nucleotide composition. This suggests that mutation pressure is the main factor influencing codon usage bias in herpesviruses.

Absence of Classical Heat Shock Response in the Citrus Pathogen Xylella fastidiosa

The fastidious bacterium Xylella fastidiosa is associated with important crop diseases worldwide. We have recently shown that X. fastidiosa is a peculiar organism having unusually low values of gene codon bias throughout its genome and, unexpectedly, in the group of the most abundant proteins. Here, we hypothesized that the lack of codon usage optimization in X. fastidiosa would incapacitate this organism to undergo quick and massive changes in protein expression as occurs in a classical stress response. Proteomic analysis of the response to heat stress in X. fastidiosa revealed that no changes in protein expression can be detected. Moreover, stress-inducible proteins identified in the closely related citrus pathogen Xanthomonas axonopodis pv citri were found to be constitutively expressed in X. fastidiosa. These proteins have extremely high codon bias values in the X. citri and other well-studied organisms, but low values in X. fastidiosa. Because biased codon usage is well known to correlate to the rate of protein synthesis, we speculate that the peculiar codon bias distribution in X. fastidiosa is related to the absence of a classical stress response, and, probably, alternative strategies for survival of X. fastidiosa under stressfull conditions.

Fine-Scale Crossover Rate Heterogeneity in Drosophila pseudoobscura

Broad-scale differences in crossover rate across the genome have been characterized in most genomes studied. Fine-scale differences, however, have only been examined in a few taxa, such as Arabidopsis, yeast, humans, and mice. No prior studies have directly looked for fine-scale recombination rate heterogeneity in Drosophila. We produced 370 Drosophila pseudoobscura containing a crossover event within the 2-megabase (MB) region between the genes yellow and white. We then examined 19 intervals within this region and determined where the crossovers occurred. We found that recombination events occur nonrandomly on a small scale and that mild "hotspots" of a few kilobases exist in Drosophila. Among the regions studied, recombination rates varied from 1.4 to 52 cM/MB. We also observed a trend toward high codon bias in regions of high recombination. Finally, we identified a significantly positive correlation between recombination rate and simple repeats, as well as the motif CACAC. These sequence features may contribute to broad-scale variation in crossover rate and, thus, shed light on features associated with crossover rate heterogeneity at a genome-wide scale.

Synonymous Codon Usage in Escherichia coli: Selection for Translational Accuracy

In many organisms, selection acts on synonymous codons to improve translation. However, the precise basis of this selection remains unclear in the majority of species. Selection could be acting to maximize the speed of elongation, to minimize the costs of proofreading, or to maximize the accuracy of translation. Using several data sets, we find evidence that codon use in Escherichia coli is biased to reduce the costs of both missense and nonsense translational errors. Highly conserved sites and genes have higher codon bias than less conserved ones, and codon bias is positively correlated to gene length and production costs, both indicating selection against missense errors. Additionally, codon bias increases along the length of genes, indicating selection against nonsense errors. Doublet mutations or replacement substitutions do not explain our observations. The correlations remain when we control for expression level and for conflicting selection pressures at the start and end of genes. Considering each amino acid by itself confirms our results. We conclude that selection on synonymous codon use in E. coli is largely due to selection for translational accuracy, to reduce the costs of both missense and nonsense errors.