Translational Selection and Yeast Proteome Evolution

Corresponding author: Hiroshi Akashi, 208 Mueller Laboratory, Pennsylvania State University, University Park, PA 16802., akashi{at}psu.edu (E-mail)

Communicating editor: W. STEPHAN

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

The primary structures of peptides may be adapted for efficient synthesis as well as proper function. Here, the Saccharomyces cerevisiae genome sequence, DNA microarray expression data, tRNA gene numbers, and functional categorizations of proteins are employed to determine whether the amino acid composition of peptides reflects natural selection to optimize the speed and accuracy of translation. Strong relationships between synonymous codon usage bias and estimates of transcript abundance suggest that DNA array data serve as adequate predictors of translation rates. Amino acid usage also shows striking relationships with expression levels. Stronger correlations between tRNA concentrations and amino acid abundances among highly expressed proteins than among less abundant proteins support adaptation of both tRNA abundances and amino acid usage to enhance the speed and accuracy of protein synthesis. Natural selection for efficient synthesis appears to also favor shorter proteins as a function of their expression levels. Comparisons restricted to proteins within functional classes are employed to control for differences in amino acid composition and protein size that reflect differences in the functional requirements of proteins expressed at different levels.

THE predominant view of protein evolution considers fitness effects of amino acid changes that arise from gene-specific relationships between the primary structures of encoded polypeptides and their particular function(s) (NEI 1975 ; KIMURA 1983 ; LI 1997 ). Critical properties of proteins (i.e., specificity, activity, or stability) depend on particular amino acids in specific regions of their structures. Mutation pressure and genetic drift determine encoded amino acids and their evolutionary divergence at sites where protein function is more tolerant to amino acid replacements.

Selection pressures related to efficient synthesis, rather than to proper function, of proteins are less firmly established. Amino-acid-altering mutations could affect fitness through physiological effects that are independent of their effects on protein function. Amino acids may vary in the energetic costs of their biosynthesis (RICHMOND 1970 ; KARLIN and BUCHER 1992 ; LOBRY and GAUTIER 1994 ; DUFTON 1997 ; CRAIG and WEBER 1998 ; GARAT and MUSTO 2000 ; JANSEN and GERSTEIN 2000 ; AKASHI and GOJOBORI 2002 ; ZAVALA et al. 2002 ), the complexity of their biosynthetic pathways (KARLIN and BUCHER 1992 ; DUFTON 1997 ; CRAIG and WEBER 1998 ), requirements for limiting resources (MAZEL and MARLIERE 1989 ; CRAIG et al. 2000 ; BAUDOUIN-CORNU et al. 2001 ), or the speed and accuracy with which their isoacceptors are translated (EIGEN and SCHUSTER 1979 ; TRIFONOV 1987 ; SHPAER 1989 ; LOBRY and GAUTIER 1994 ; AKASHI 1996 ; GUTIERREZ et al. 1996 ; PERCUDANI et al. 1997 ). Although individual amino acid mutations are likely to have small effects on overall cellular physiology, global evolutionary forces could underlie proteome-wide patterns of amino acid composition as well as variation in rates of protein evolution.

Among microbes, as well as multicellular eukaryotes, synonymous codon usage is coadapted with tRNA pools to enhance the efficiency of protein synthesis (reviewed in ANDERSSON and KURLAND 1990 ; SHARP et al. 1993 ; AKASHI 2001 ). Among codons recognized by different aminoacyl tRNAs (aa-tRNAs), translationally preferred codons tend to be recognized by the more abundant isoacceptor. Among codons recognized by the same isoacceptor (through "wobble" pairing), preferred codons generally have intermediate codon-anticodon stability (GROSJEAN and FIERS 1982 ; IKEMURA 1985 ; YAMAO et al. 1991 ; KANAYA et al. 1999 ; PERCUDANI and OTTONELLO 1999 ). In Escherichia coli, translation of major codons occurs 3- to 6-fold more quickly (ROBINSON et al. 1984 ; VARENNE et al. 1984 ; SORENSEN et al. 1989 ) and 10-fold more accurately (PRECUP and PARKER 1987 ) than translation of minor codons. Thus, major codons allow efficient use of ribosomes and reduce the cost of GTP-dependent "proofreading" or rejection of noncognate isoacceptors. In addition, accurate translation reduces the costs of producing dysfunctional peptides resulting from misincorporations and processivity errors (frameshifting and premature termination). Stronger codon usage bias in highly expressed genes reflects increases in the fitness benefits of major codons with the number of translation events at a given codon. Major codon preference among synonymous codons was apparent from early examinations of small numbers of yeast genes (BENNETZEN and HALL 1982 ; IKEMURA 1982 ; SHARP et al. 1986 ) and is consistent with large population sizes and a close relationship between growth rate and fitness in these microbes.

Among tRNAs carrying different amino acids, variation in either cellular concentrations or codon-anticodon stability could lead to translation selection both within and among synonymous families (SHPAER 1989 ; LOBRY and GAUTIER 1994 ; AKASHI 1996 ; PERCUDANI et al. 1997 ; MORTON and SO 2000 ; AKASHI 2001 ). Amino acid composition is related to gene expression in prokaryotes (SHPAER 1989 ; YAMAO et al. 1991 ; LOBRY and GAUTIER 1994 ; GUTIERREZ et al. 1996 ; AKASHI and GOJOBORI 2002 ; ZAVALA et al. 2002 ), yeast (IKEMURA 1982 ; PERCUDANI et al. 1997 ; JANSEN and GERSTEIN 2000 ), Giardia lamblia (GARAT and MUSTO 2000 ), Caenorhabditis elegans (DURET 2000 ), and plant chloroplasts (MORTON and SO 2000 ). Furthermore, amino acids represented by abundant tRNAs tend to be preferentially encoded in highly expressed genes (SHPAER 1989 ; YAMAO et al. 1991 ; LOBRY and GAUTIER 1994 ; PERCUDANI et al. 1997 ; DURET 2000 ). However, highly expressed proteins fall into particular functional categories (i.e., energy metabolism and protein synthesis) and tRNA pools could simply be adjusted to match the amino acid requirements for proper functioning of these proteins (GAREL 1974 ; IKEMURA 1982 ; YAMAO et al. 1991 ; XIA 1998 ; DURET 2000 ).

In multicellular eukaryotes, tissue-specific tRNA abundances have been found in tissues committed to high expression of a small number of genes. tRNA concentrations match amino acid usage of fibroin in the posterior silk gland of the silkworm Bombyx mori L. and crystallines in the calf lens (GAREL 1974 ) and hemoglobin in rabbit and human reticulocytes (HATFIELD et al. 1982 ). GAREL 1974 proposed "functional adaptation of tRNA"; selection for efficient translation regulates cellular tRNA isoacceptor concentrations to match the amino acid requirements of highly expressed proteins but does not affect their composition or evolutionary rates.

This study attempts to distinguish between unidirectional adjustments of tRNA pools to the amino acid requirements of highly expressed genes and coadaptation of both isoacceptor concentrations and amino acid usage in the budding yeast, Saccharomyces cerevisiae. Strong associations between synonymous codon usage and oligonucleotide DNA array estimates of mRNA levels suggest that estimates of transcript abundance provide informative predictors of the translation rates of genes. Usage of several amino acids shows associations with gene expression, and changes in amino acid composition result in stronger correlations between amino acid usage and tRNA abundances in highly expressed genes than in less expressed loci. Similar relationships within protein functional categories suggest that the primary structures of proteins reflect, at least in part, natural selection to enhance the rate and accuracy of their synthesis. Selection for efficient biosynthesis may also constrain protein size; among proteins in the same broad functional category, proteins encoded by highly expressed genes are consistently smaller than those encoded by less expressed loci.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Yeast gene sequences:
S. cerevisiae protein-coding sequences and descriptions (GOFFEAU et al. 1996 ) were obtained from ftp://genome-ftp.stanford.edu/pub/yeast/. Mitochondrial DNA-encoded genes, short coding regions (<100 codons), and genes identified as originating from phage or transposable elements were excluded from the analysis. In addition, genes with recent common ancestors (paralogs) were identified by performing unfiltered BLAST (ALTSCHUL et al. 1990 ) searches among all pairs of proteins encoded in the genome. Pairs of protein sequences showing alignments with 60% identity over 60 amino acids were formed into clusters and one gene from each cluster was included in the analysis. To maintain the sample size of highly expressed loci, the gene with the highest estimate of transcript abundance (see below) was chosen from each cluster.

Yeast expression data and functional categorization of proteins:
Transcript abundance data from HOLSTEGE et al. 1998 were obtained from http://web.wi.mit.edu/young/expression/transcriptome.html and protein abundance measures (from 2D gel data) were taken from FUTCHER et al. 1999 . Functional categorizations of gene products were obtained from the Yeast Protein Database (YPD; COSTANZO et al. 2000 ; https://www.incyte.com/proteome/YPDcategories/Functional_Categories.html). Composite categories were constructed for 1571 genes listed in more than one category. Using this criterion, I found that yeast proteins fall into 259 different functional categories (including "unknown") and 128 of the categories contain a single gene. Only genes listed in both the YPD and the Holstege transcript abundance database were included in the analyses. Of the 6310 predicted yeast protein-coding genes, 5483 were included in the final data set (see supplemental material at http://www.genetics.org/supplemental/).

Sequence and expression data for C. elegans, Drosophila melanogaster, Bacillus subtilis, and E. coli:
Coding sequences and estimates of transcript abundances [from matches to expressed sequences tag (EST) libraries] for C. elegans and D. melanogaster (MARAIS et al. 2001 ) were obtained from http://pbil.univ-lyon1.fr/datasets/Marais2001/data.html. Short coding regions (<100 codons) and genes identified as originating from phage or transposable elements were excluded from the analysis. Single members of each family of paralogs were included as described above. Genes that were not listed in the MARAIS et al. 2001 expression data files were not included in the analysis. A total of 11,546 and 11,864 predicted genes were analyzed from D. melanogaster and C. elegans, respectively. E. coli and B. subtilis data are described in AKASHI and GOJOBORI 2002 .

Identification of major codons:
Major codon usage (MCU) was calculated as (number of major codons)/(number of major codons + number of minor codons). Identities of major codons for S. cerevisiae were taken from KANAYA et al. 1999 except for glutamic acid [the major codon was identified as GAA in IKEMURA 1982 ]. Major codons were taken from AKASHI 1995 for D. melanogaster, DURET 2000 for C. elegans, and KANAYA et al. 1999 for E. coli and B. subtilis [with modifications described in AKASHI and GOJOBORI 2002 ]. Serine codons were divided into fourfold and twofold families so that each synonymous family is composed of codons that encode the same amino acid and that are connected by single mutational steps. Major codons for the two serine families were identified as those showing significant positive correlations with either gene expression [S. cerevisiae (Table 1) and C. elegans] or major codon usage for nonserine families (E. coli and B. subtilis).

Analyses of whole-genome data:
Spearman rank correlations were employed in the whole-genome analyses. Because abundances for some codons and amino acids are quite low, analyses were conducted on binned data. Genes were ranked by the HOLSTEGE et al. 1998 estimates of transcript abundance and data were pooled for genes with low to high transcript abundance until 5000 codons were reached for each bin (all genes with identical expression estimates were included in the same bin). The numbers of genes in low expression bins were elevated by large numbers of identical estimates of transcript abundance. For statistical analyses, codon and amino acid usages were compared among 65 expression classes containing an average of 84 genes each. Bins of 50,000 codons were employed for visualization of trends (Fig 2, Fig 3, Fig 4, and Fig 6).

View larger version (15K): In this window In a new window Download PPT slide   Figure 1. Distribution of transcript abundances of yeast genes. Data are from HOLSTEGE et al. 1998 . The first column plots the number of genes with no detectable transcripts. Other columns plot the numbers of genes with transcript abundances between a lower limit shown below the columns and an upper limit shown beneath the column to the right (the last column has no upper limit).

View larger version (34K): In this window In a new window Download PPT slide   Figure 2. Transcript abundance and major codon usage. Data are graphed for expression categories (bin sizes 50,000 codons). Abundances are within synonymous families.

View larger version (12K): In this window In a new window Download PPT slide   Figure 3. Transcript abundance and codon usage in sixfold redundant families. The numbers of tRNAs that recognize twofold and fourfold degenerate codons are shown (twofold/fourfold). Data are graphed for expression categories (bin sizes 5 x 104 codons). Abundances are within synonymous families.

View larger version (19K): In this window In a new window Download PPT slide   Figure 4. Transcript abundance and amino acid usage. Data are graphed for expression categories (bin sizes 5 x 104 codons). Abundances are among all codons.

View larger version (12K): In this window In a new window Download PPT slide   Figure 5. Correlations between amino acid usage and tRNA gene copy numbers. tRNA gene copy numbers are from PERCUDANI et al. 1997 . The numbers are pooled among genes encoding isoacceptors for each amino acid.

View larger version (15K): In this window In a new window Download PPT slide   Figure 6. Correlation between amino acid usage and tRNA gene copy numbers among expression classes. Data are graphed for expression categories (bin sizes 5 x 104 codons). Pearson product moment correlation coefficients between tRNA gene copy numbers and amino acid usage are plotted on the y-axis.

Analyses within functional categories:
Within functional categories, amino acid usage was compared between genes falling above and below a cutoff of one transcript per cell. Amino acid abundances were compared in 2 x 2 contingency tables; the columns of the tables were the high and low expression classes and the rows consisted of the counts of a particular amino acid and the pooled counts for all other amino acids. The Mantel-Haenszel procedure (SNEDECOR and COCHRAN 1989 ) was employed to calculate an overall probability for departures from equal amino acid usage among low and high expression genes across contingency tables from different functional categories. Thirty-one functional categories containing 10 genes in both expression classes were included in the analysis. "Unknown" was not included as a functional category.

The proportion of amino acids falling within "low complexity" regions was analyzed in a similar manner. The rows of 2 x 2 contingency tables consisted of the numbers of codons that fall within and outside of low complexity regions identified by the SEG and SEGN programs (WOOTTON and FEDERHEN 1996 ) using default parameter settings. The columns were the low- and high-expression classes and Mantel-Haenszel tests were conducted as described above.

Mann-Whitney U-tests (SNEDECOR and COCHRAN 1989 ) were employed to test for differences in the mean gene lengths of highly and lowly expressed genes within each functional category. Z values of Mann-Whitney U-tests were assigned positive and negative signs for higher and lower mean ranks of size among highly expressed proteins. To test for a consistent trend among categories, a Wilcoxon ranked signs test statistic was calculated for the signed Z values. A null distribution of the test statistic was generated by 10⁶ iterations of random assignment of sign (with P = 0.5) to each Z value and recalculation of the test statistic.

Yeast tRNA abundances:
IKEMURA 1982 quantified cellular abundances for 22 yeast tRNAs using 2D gel electrophoresis. The correlation between these measurements and the copy numbers of the corresponding tRNA genes in the S. cerevisiae genome (PERCUDANI et al. 1997 ) is remarkably high (r² = 0.803). To include all 41 isoaccepting tRNAs in the analyses, the gene copy numbers of PERCUDANI et al. 1997 were employed as estimates of tRNA abundances.

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Transcript abundance and translation rates in yeast:
Yeast cells growing at log phase under standard laboratory conditions contain 15,000 poly(A)-mRNA molecules (HEREFORD and ROSBASH 1977 ). Fig 1 plots the distribution of transcript abundances among yeast genes from the high-density oligonucleotide array data of HOLSTEGE et al. 1998 from yeast cells grown to mid-log phase in YPD media. The distribution of transcript abundance is strongly skewed toward low values; >80% of genes are represented by 2 mRNA molecules and only 3.5% of genes have transcript abundances of 10 mRNA molecules per cell. FUTCHER et al. 1999 found good correspondence between 2D gel quantifications of protein concentrations (ranging from 200 to 2 x 10⁶ molecules/cell) and estimates of transcript abundances. They estimate a rate of protein synthesis of roughly 4000 proteins/transcript for genes represented by 1 mRNA molecule/cell. For proteins represented by <1 mRNA/cell, they suggest post-transcriptional regulation; i.e., mRNA abundances are not informative predictors of translation rates.

Synonymous codon usage was examined among expression classes to determine the strength of correspondence between GeneChip estimates of transcript abundances and the translation rates of genes. Under major codon preference, the fitness benefit of a major codon is strongly dependent on the number of translation events experienced at a given codon. COGHLAN and WOLFE 2000 found that <40% of the variation in ranks of measures of codon bias among yeast genes was explained by transcript abundance (Spearman rank correlation, r_s = 0.62). However, correlations between major codon usage and transcript abundance are remarkably high in comparisons among bins of genes with similar expression estimates (AKASHI 2001 ; Table 1; Fig 2). Similar patterns among synonymous families that differ in the favored nucleotide in the third codon position (G-favored: Lys, Leu; A-favored: Pro, Gln; T-favored: Gly; C-favored: Phe, Tyr, His, Asn, Asp) suggest both that estimates of transcript abundance are informative predictors of average translation rates (among binned genes) and that translation selection is sufficient to overcome mutational biases associated with transcription (DATTA and JINKS-ROBERTSON 1995 ; MOREY et al. 2000 ) and substitutional biases associated with gene conversion (GERTON et al. 2000 ; BIRDSELL 2002 ). Codon preferences determined using microarray estimates of transcript abundance are consistent with those established through correspondence analysis of codon usage (SHARP and COWE 1991 ; KANAYA et al. 1999 ).

Seven of the twofold synonymous families (all NNY types) are recognized by a single isoacceptor through wobble pairing at the third codon position. Six of these families (Asn, Asp, Cys, His, Phe, and Tyr) show steady increases in usage of a single codon in highly transcribed genes (Table 1; Fig 2). Such patterns are consistent with translational selection for codon-anticodon stability (GROSJEAN et al. 1978 ; PERCUDANI and OTTONELLO 1999 ).

Codon usage for amino acids encoded by sixfold redundant codons provide the clearest evidence for translational preferences related to tRNA abundances. Third codon position wobble rules for eukaryotes are ambiguous (PERCUDANI 2001 ), but wobble pairing is not known to occur at the first codon position. Thus, fourfold and twofold redundant families for Leu, Arg, and Ser are recognized by nonoverlapping sets of tRNAs. Usage of twofold codons for Leu (r_s = 0.933, P < 10^-5) and Arg (r_s = 0.558, P < 10^-5) increases dramatically whereas twofold codons for Ser decrease (r_s = -0.897, P < 10^-5) within their synonymous families in highly expressed genes (Fig 3). These patterns correspond to differences in the numbers of tRNA genes (presumably resulting in higher tRNA abundances) that recognize the twofold and fourfold families for these amino acids. Translational selection appears to discriminate among synonymous codons recognized by nonoverlapping groups of tRNAs. Similar preferences may also bias the usage of codons that encode different amino acids.

Gene expression and amino acid composition in yeast:
Abundances for a number of amino acids are strongly correlated with gene expression levels (Table 2). The magnitude of changes in abundance can be quite large; alanine usage increases by greater than twofold in highly expressed genes and serine twofold codons are only one-third as abundant in highly expressed genes (Fig 4).

In S. cerevisiae, estimates of transcript abundance show strong positive associations with the frequency of meiotic double-strand breaks, a measure of local recombination rate (GERTON et al. 2000 ; BIRDSELL 2002 ). The latter is also positively correlated with both intergenic and third position GC content (BIRDSELL 2002 ), suggesting that biased gene conversion elevates G + C content. However, associations between amino acid usage and estimates of recombination rate are considerably smaller than associations with transcript abundance (data available from H. Akashi).

The interpretation of relationships between gene expression and amino acid composition is less straightforward than that for similar associations among synonymous codons. Changes in amino acid usage could reflect differences in the functional roles of proteins expressed at different levels (GAREL 1974 ; IKEMURA 1982 ; YAMAO et al. 1991 ; XIA 1998 ; DURET 2000 ). For example, JANSEN and GERSTEIN 2000 showed that transcript abundances are higher for cytosolic proteins than for membrane proteins in yeast. Greater usage of hydrophobic residues in less expressed proteins may reflect a greater abundance of transmembrane regions. To control for differences in the functional requirements of proteins, amino acid composition was compared among genes classified into common functional categories in the Yeast Proteome Database (COSTANZO et al. 2000 ). Table 3 shows the 31 different categories containing at least 10 genes in both the low (one or less transcript per cell) and the high (more than one transcript per cell) expression classes. There are clear differences in expression patterns among these categories; transcription factors and membrane proteins/transporters tend to be represented by few transcripts. Two classes of more abundant transcripts encode oxidoreductases, including enzymes of central metabolism and amino acid biosynthesis, and proteins involved in translation, such as ribosomal proteins and elongation factors. Transcript abundances for these classes are consistent with estimates of 200,000 ribosomes/cell in rapidly growing yeast (WARNER 1999 ) and 2,000,000 molecules/cell for some glycolytic enzymes (FUTCHER et al. 1999 ). Thus, amino acids that are employed more often in highly expressed genes may simply be those required for proper functioning of ribosomal proteins and cytosolic enzymes.

For analyses within the 31 functional categories, amino acid usage was compared between low- and high-expression classes in 2 x 2 contingency tables. Table 2 shows, for each amino acid, the number of individually significant 2 x 2 tests as well as the probability of the overall trend across tables. With cutoff values of two, three, and four transcripts per cell to divide high- and low-expression classes, the numbers of functional categories with at least 10 genes in each expression class reduce to 21, 15, and 11, respectively. However, the main trends of amino acid usage are robust to these cutoff values; Mantel-Haenszel test statistics remain significantly positive for Val, Ala, Gly, and Glu and negative for Phe, Leu, Ile, His, Asn, Cys, and Ser₂ for cutoff values between one and four transcripts per cell.

Several amino acids show strong statistical associations with expression levels in both whole-genome and within-category analyses. Ala, Val, and Gly show strong increases in abundance in highly expressed genes, whereas Leu, Ser₂, and Asn show strong declines (Fig 4). For such amino acids, changes in the relative abundances of different types of proteins in different expression classes are unlikely to explain relationships between amino acid usage and expression levels. These patterns are consistent with JANSEN and GERSTEIN's (2000) findings through comparison of amino acid composition of the yeast genome and transcriptome (amino acid usage for a given gene was weighted by estimates of its transcript abundance). However, results for some amino acids (Gln, Ser₄, Arg, Glu, and Tyr) differ between the all-gene and within-category analyses. Such patterns could reflect differences in the functional requirements of genes in different expression classes or differences in the statistical power of the two approaches. In either case, these amino acids show small differences in abundance between lowly and highly expressed genes.

Amino acid usage and tRNA abundances:
Relationships between amino acid usage and tRNA gene numbers for yeast genes with low, intermediate, and high transcript abundance are shown in Fig 5. Codons that experience few translation events will be under little or no selection for translationally preferred codons (among either synonymous or nonsynonymous alternatives). Thus, relationships between tRNA abundance and amino acid usage should show substantial scatter. However, under translational selection, the magnitude of fitness differences among codons recognized by rare and common isoacceptors should increase as a function of gene expression levels. Such fitness differences may exist among nonsynonymous as well as synonymous alternatives. In highly expressed genes, translational selection at positions of peptides otherwise determined largely by mutation drift will result in greater correspondences between amino acid usage and tRNA abundances.

Fig 6 shows that the Pearson product-moment correlation coefficient between amino acid usage and tRNA gene numbers increases steadily as a function of gene expression levels (5000 codons/bin, r_s = 0.68, Z = 6.39, P < 10^-5). Plots are shown for bins of 50,000 codons. Stronger correlations between amino acid usage and tRNA gene copy numbers in highly expressed proteins within functional categories (Table 3) support the contribution of translational selection in determining the amino acid composition of proteins (Wilcoxon ranked signs test, P < 10^-5).

Gene expression and protein size:
Given some tolerance of protein function to insertion/deletion variation, translational selection will favor reductions in protein size (AKASHI 1996 ). Eliminating codons from a given gene will reduce the amount of time that ribosomes spend translating its transcripts and enhance the overall rate of proteins synthesized per ribosome per time. Energy expenditure for both amino acid and protein synthesis will also be reduced. The magnitude of time and energy savings, and, consequently, the fitness advantage of protein size reduction will be a function of the number of times a given gene is translated.

Negative correlations between gene length and both synonymous codon bias (MORIYAMA and POWELL 1998 ) and transcript abundance (COGHLAN and WOLFE 2000 ; JANSEN and GERSTEIN 2000 ; PAL et al. 2001B ) have been found among yeast genes. However, previous studies did not control for differences in the functional requirements of proteins among expression classes, such as the lack of highly expressed transmembrane proteins. Among the 31 functional categories of yeast proteins, 27 show a higher mean rank of protein size among the less highly expressed proteins and 4 deviate in the opposite direction (Table 3). Mann-Whitney U-tests show significant departures at the 5% level (prior to correction for multiple tests) for 12 of the 31 classes; highly expressed proteins are smaller in 11 of these 12 classes. Overall, highly expressed genes tend to encode smaller proteins than do less expressed proteins in the same functional category (Wilcoxon ranked signs test, P < 10^-5). Protein length is also negatively correlated with 2D gel quantifications of abundance for 64 proteins (FUTCHER et al. 1999 ) from cells grown on both glucose (r_s = -0.47, P < 10^-4) and ethanol (r_s = -0.33, P < 0.005).

To determine whether low complexity nucleotide sequences (including homonucleotide runs and short repeats) contribute to differences in protein sizes among expression classes, simple sequences were identified using the SEGN software (WOOTTON and FEDERHEN 1996 ). Of the 31 functional categories, 21 show a higher percentage of simple sequences in less expressed proteins and 10 deviate in the opposite direction. A Mantel-Haenszel test shows significantly lower proportions of simple nucleotide sequences in highly expressed proteins (Z = 12.29, P < 10^-5). However, this reduction in simple sequences does not account entirely for the smaller sizes of highly expressed proteins; differences in the sizes remain highly significant after removal of low-complexity regions (Wilcoxon ranked signs test, P < 10^-4). Interestingly, the proportion of "low-complexity" amino acid sequences increases in highly expressed proteins within functional categories; 20 of 31 categories deviate in this direction and the overall pattern is highly statistically significant (Mantel-Haenszel test, Z = 20.06, P < 10^-5). This pattern may reflect a greater abundance of particular structural motifs (soluble folds with combinations of helices and sheets) represented among highly expressed proteins (JANSEN and GERSTEIN 2000 ).

Gene expression and GNN usage:
In yeast, GCN, GGN, and GTN codons for Ala, Gly, and Val, respectively, show the strongest increases in usage in highly expressed genes (Table 2; Fig 4). Fig 7 shows that such patterns are common to many prokaryotes and multicellular eukaryotes. GNN usage shows remarkably consistent increases in abundance with measures of translation rates (either estimates of transcript abundance or measures of synonymous codon usage bias) in C. elegans, D. melanogaster, B. subtilis, and E. coli (see also GUTIERREZ et al. 1996 ), as well as yeast. Similar patterns have been noted in the genomes of plant chloroplasts (MORTON and SO 2000 ) and Buchnera (PALACIOS and WERNEGREEN 2002 ). GNN increases occur among cytosolic and membrane proteins encoded in plant chloroplast genomes (MORTON and SO 2000 ) and within functional categories of yeast (Table 2), E. coli, B. subtilis (AKASHI and GOJOBORI 2002 ), and C. elegans (our unpublished data) proteins.

View larger version (16K): In this window In a new window Download PPT slide   Figure 7. Gene expression and GNN codon usage among highly diverged genomes. Data are described in the text. The pooled abundances of GNN codons (Val, Ala, Asp, Glu, and Gly) among all codons are plotted on the y-axis. mRNA abundances for C. elegans and D. melanogaster are counts of matches to EST libraries (DURET and MOUCHIROUD 1999 ) and are employed as estimators of translation rates. Bin sizes for ranking by mRNA abundances are 5 x 104 for each genome. MCU, major codon usage, is the number of major codons/(number of major + minor codons) and is also employed as an estimator of translation rates. Bins were constructed similarly to transcript abundance estimates (see text) for genes ranked by MCU. Bin sizes are 1 x 105 for S. cerevisiae, C. elegans, and D. melanogaster and 5 x 104 for B. subtilis and E. coli.

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Establishing translational selection in protein evolution:
Major codon preference posits adaptation of both tRNA concentrations and synonymous codon usage. Regulation of aa-tRNA abundances may result from relatively few, strongly selected mutations. However, codon usage bias results from weak selection at thousands of "silent" sites throughout the genome (reviewed in ANDERSSON and KURLAND 1990 ; SHARP et al. 1993 ; AKASHI 2001 ). Although relationships between amino acid composition and gene expression have been established in yeast (IKEMURA 1982 ; PERCUDANI et al. 1997 ; JANSEN and GERSTEIN 2000 ), translation selection in protein evolution has been difficult to substantiate because associations between amino acid composition and tRNA abundances can be explained by selection on tRNA concentrations to match the functional needs of highly expressed proteins (GAREL 1974 ; IKEMURA 1982 ; YAMAO et al. 1991 ; XIA 1998 ; DURET 2000 ).

Here, functional categorizations of proteins were employed to distinguish between associations between gene expression and amino acid composition that arise as a by-product of the functional requirements of proteins and those that reflect fitness benefits to translationally superior codons. Increases in the correlation between tRNA gene numbers and amino acid usage as a function of expression levels among all genes and similar patterns within broad functional categories support translational selection. Although the functional categorizations of proteins may be crude, it is unlikely that functional requirements explain consistent trends in amino acid usage in nonoverlapping groups of genes.

These findings do not exclude functional adaptation of isoacceptor abundances. tRNA pools may have been initially adjusted to match the functional requirements of highly expressed ribosomal proteins and enzymes of central metabolism. Translational selection would magnify amino acid usage biases beyond the initial functional needs of abundant proteins. The gradual increase in the correlation between amino acid usage and gene expression (Fig 6) supports a contribution of translational selection in the amino acid composition of even moderately expressed yeast genes. However, Table 2 shows relatively high overall usage of some amino acids that appear to be translationally less preferable (i.e., Leu, Asn, and Ile), suggesting a balance among forces including translational selection, functional constraint, and mutation pressure.

Associations between mutational processes and transcription rates could contribute to relationships between gene expression and codon and amino acid usage, as well as protein length. In E. coli, transcription induces C T transitions on the nontranscribed strand, presumably due to increased deamination of cytosine (FRANCINO et al. 1996 ; BELETSKII and BHAGWAT 1998 ). Genetic experiments in yeast have shown >10-fold increases in rates of -1 frameshift reversion mutations with transcription rates (DATTA and JINKS-ROBERTSON 1995 ). However, such experiments screen for particular types of mutations and the dependence of the overall spectrum of mutations on transcription has not been determined. In D. melanogaster (KLIMAN and HEY 1994 ) and C. elegans (DURET and MOUCHIROUD 1999 ), intron base composition has been examined to determine whether mutational processes are transcription dependent. However, the small numbers of introns in the yeast genome (DAVIS et al. 2000 ) and their greater abundance (ARES et al. 1999 ; LOPEZ and SERAPHIN 1999 ) and lengths (VINOGRADOV 2001 ) in highly expressed transcripts suggest that intron base composition does not reflect mutational equilibrium.

Population genetic analyses of putative fitness classes of nonsynonymous mutations (AKASHI 1995 ) may provide a means to distinguish between the contributions of translational selection and mutational biases in amino acid composition. In contrast to compositional studies that assume constant mutational processes among genes, such analyses assume constancy of mutational processes within genes over evolutionary time, and the statistical power to detect weak selection can be quite high (AKASHI 1999 ). The compositional analysis undertaken here may provide putative fitness classes of amino acid changes for such studies. Rigorous support for translational selection in yeast protein evolution may require sequence data from within and between closely related species.

Properties of translationally preferred aa-tRNAs:
The rate and accuracy of protein synthesis depend on both the abundances of aa-tRNAs and their intrinsic properties, such as the stability of codon-anticodon interactions. Recent studies have demonstrated conformational changes in aa-tRNAs, elongation factors, and ribosomes during protein synthesis (OGLE et al. 2001 ). Codon-anticodon interactions determine rate constants during the processes of tRNA selection, proofreading, and peptidyl transfer, which, in turn, determine both the speed and the accuracy of protein synthesis (reviewed in RODNINA and WINTERMEYER 2001 ). Processing of aa-tRNAs also affects translation of neighboring codons; near-cognate isoacceptors bound at the ribosomal P site induce frameshift events at the A site (FARABAUGH 2000 ). The analyses presented here have focused on aa-tRNA abundances, an extrinsic property of isoacceptors, because they can be estimated from gene numbers. However, codon preferences among nonsynonymous codons are likely to also reflect intrinsic properties of tRNAs or the amino acids that they carry. For example, the decline of Leu usage in highly expressed genes may reflect, in part, selection against usage of frameshift-prone codons (FARABAUGH 2000 ).

Increasing GNN usage is the most prominent feature of associations between amino acid usage and gene expression in yeast as well as in a number of distantly related organisms. Three base nucleotide periodicities in protein-coding genes (NASSAR and COOK 1976 ; TRIFONOV and SUSSMAN 1980 ; SHEPHERD 1981 ) have been explained in light of theoretical studies of the early evolution of the genetic code (CRICK et al. 1976 ; EIGEN and SCHUSTER 1979 ). TRIFONOV 1987 , TRIFONOV 1992 has argued that G:non-G:N codons may aid in maintenance of translational reading frame through interactions between mRNA and 16S rRNA during translation. Biochemical studies of protein synthesis will be required to determine whether GNN codons have special translational properties.

Intrinsic features of aa-tRNAs could also include properties of amino acids such as their requirements for limiting resources (MAZEL and MARLIERE 1989 ; CRAIG et al. 2000 ; BAUDOUIN-CORNU et al. 2001 ) or costs of biosynthesis or transport (RICHMOND 1970 ; KARLIN and BUCHER 1992 ; LOBRY and GAUTIER 1994 ; DUFTON 1997 ; CRAIG and WEBER 1998 ; GARAT and MUSTO 2000 ; JANSEN and GERSTEIN 2000 ; AKASHI and GOJOBORI 2002 ; ZAVALA et al. 2002 ). Natural selection may have elevated tRNA abundances for isoacceptors carrying metabolically favored amino acids so that translation selection acts in the same direction as such preferences (AKASHI and GOJOBORI 2002 ).

BAUDOUIN-CORNU et al. (2001) have shown that yeast proteins involved in sulfur and nitrogen transport and processing show reduced levels of amino acids requiring these atoms. Examination of the usage of S- and N-containing amino acids among proteins that are highly expressed during nitrogen or sulfur starvation would add support for nutrient limitation and protein evolution.

Calculations of the energetic costs of amino acid biosynthesis may differ between yeast and E. coli or B. subtilis due to both differences in amino acid biosynthetic pathways and alternative energy production pathways (alcohol fermentation and respiration). Such analyses may help to explain the identities of amino acids whose usage differs between highly and lowly expressed genes but are not undertaken here.

Translational selection and protein size:
In the yeast genome, the smaller sizes of proteins encoded by highly expressed genes are consistent with selection favoring reductions in the metabolic costs of protein and/or amino acid biosynthesis. Relationships between gene length and expression levels in multicellular eukaryotes are less clear. DURET and MOUCHIROUD 1999 found no association between protein length and transcript abundance (measured by the numbers of matches of ESTs to predicted gene sequences) in Arabidopsis thaliana and positive correlations between protein size and expression in the D. melanogaster and C. elegans. In contrast, CASTILLO-DAVIS et al. 2002 employed DNA array estimates of mRNA abundance and found strong statistical support for reductions in length among highly expressed C. elegans genes. Biases in methods for estimating gene expression will need to be explored and patterns will need to be studied among proteins of related function to determine whether gene length and expression level are related in these organisms. A lack of negative relationships would be at odds with strong evidence of selection for metabolic efficiency at silent sites in these multicellular eukaryotes (reviewed in SHARP et al. 1993 ; AKASHI 2001 ; DURET 2002 ).

Translational selection and protein divergence:
Rates of protein divergence are negatively correlated with expression levels among yeast genes (PAL et al. 2001A ). Translational selection in protein evolution could provide an explanation (AKASHI 2001 ); in highly expressed genes, amino acid changes that may be neutral with respect to protein function will be selected against if they decrease the rate or accuracy of protein synthesis. Translationally unpreferred amino acids that are maintained in these genes may be restricted to those that serve critical roles in protein function and may also be evolutionarily conserved.

Rates of protein evolution are also negatively correlated with gene expression in plants (WRIGHT et al. 2002 ), mammals (DURET and MOUCHIROUD 2000 ; IIDA and AKASHI 2000 ), and Drosophila (BETANCOURT and PRESGRAVES 2002 ), suggesting that translation selection may be a factor in protein divergence among multicellular eukaryotes. However, DURET and MOUCHIROUD 2000 proposed that expression patterns are related to functional constraint. Proteins expressed in multiple tissues encounter a large number of chemical environments and their primary structures may be constrained to avoid physical interactions with other proteins (HASTINGS 1996 ). Alternatively, proteins expressed in a greater number of tissues or in a greater number of developmental stages may be more likely than tissue-specific proteins to affect fitness. Both ideas relate rates of protein evolution to constraints on function. If the interpretations proposed here are correct, then fitness effects of amino acid changes related to the overall physiology of cells, rather than the specific functions of proteins, should also contribute to patterns of protein divergence and amino acid compositional differences among taxa.

	ACKNOWLEDGMENTS

I am grateful to C. Castillo-Davis, F. Holstege, T. Ikemura, and W.-Y. Ko for valuable discussions and comments on the manuscript. This work was supported by a grant from the Alfred P. Sloan Foundation.

Manuscript received October 11, 2002; Accepted for publication March 14, 2003.

	LITERATURE CITED

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

AKASHI, H., 1995  Inferring weak selection from patterns of polymorphism and divergence at "silent" sites in Drosophila DNA. Genetics 139:1067-1076.[Abstract]

AKASHI, H., 1996  Molecular evolution between Drosophila melanogaster and D. simulans: reduced codon bias, faster rates of amino acid substitution and larger proteins in D. melanogaster.. Genetics 144:1297-1307.[Abstract]

AKASHI, H., 1999  Inferring the fitness effects of DNA mutations from polymorphism and divergence data: statistical power to detect directional selection under stationarity and free recombination. Genetics 151:221-238.[Abstract/Free Full Text]

AKASHI, H., 2001  Gene expression and molecular evolution. Curr. Opin. Genet. Dev. 11:660-666.[Medline]

AKASHI, H. and T. GOJOBORI, 2002  Metabolic efficiency and amino acid composition in the proteomes of Escherichia coli and Bacillus subtilis.. Proc. Natl. Acad. Sci. USA 99:3695-3700.[Abstract/Free Full Text]

ALTSCHUL, S. F., W. GISH, W. MILLER, E. W. MYERS, and D. J. LIPMAN, 1990  Basic local alignment search tool. J. Mol. Biol. 215:403-410.[Medline]

ANDERSSON, S. G. and C. G. KURLAND, 1990  Codon preferences in free-living microorganisms. Microbiol. Rev. 54:198-210.[Abstract/Free Full Text]

ARES, M., JR., L. GRATE, and M. H. PAULING, 1999  A handful of intron-containing genes produces the lion's share of yeast mRNA. RNA 5:1138-1139.[Medline]

BAUDOUIN-CORNU, P., Y. SURDIN-KERJAN, P. MARLIERE, and D. THOMAS, 2001  Molecular evolution of protein atomic composition. Science 293:297-300.[Abstract/Free Full Text]

BELETSKII, A. and A. S. BHAGWAT, 1998  Correlation between transcription and C to T mutations in the non-transcribed DNA strand. Biol. Chem. 379:549-551.[Medline]

BENNETZEN, J. L. and B. D. HALL, 1982  Codon selection in yeast. J. Biol. Chem. 257:3026-3031.[Abstract/Free Full Text]

BETANCOURT, A. J. and D. C. PRESGRAVES, 2002  Linkage limits the power of natural selection in Drosophila. Proc. Natl. Acad. Sci. USA 99:13616-13620.[Abstract/Free Full Text]

BIRDSELL, J. A., 2002  Integrating genomics, bioinformatics, and classical genetics to study the effects of recombination on genome evolution. Mol. Biol. Evol. 19:1181-1197.[Abstract/Free Full Text]

CASTILLO-DAVIS, C. I., S. L. MEKHEDOV, D. L. HARTL, E. V. KOONIN, and F. A. KONDRASHOV, 2002  Selection for short introns in highly expressed genes. Nat. Genet. 31:415-418.[Medline]

COGHLAN, A. and K. H. WOLFE, 2000  Relationship of codon bias to mRNA concentration and protein length in Saccharomyces cerevisiae.. Yeast 16:1131-1145.[Medline]

COSTANZO, M. C., J. D. HOGAN, M. E. CUSICK, B. P. DAVIS, and A. M. FANCHER et al., 2000  The yeast proteome database (YPD) and Caenorhabditis elegans proteome database (WormPD): comprehensive resources for the organization and comparison of model organism protein information. Nucleic Acids Res. 28:73-76.[Abstract/Free Full Text]

CRAIG, C. L. and R. S. WEBER, 1998  Selection costs of amino acid substitutions in ColE1 and ColIa gene clusters harbored by Escherichia coli.. Mol. Biol. Evol. 15:774-776.[Medline]

CRAIG, C. L., C. RIEKEL, M. E. HERBERSTEIN, R. S. WEBER, and D. KAPLAN et al., 2000  Evidence for diet effects on the composition of silk proteins produced by spiders. Mol. Biol. Evol. 17:1904-1913.[Abstract/Free Full Text]

CRICK, F. H., S. BRENNER, A. KLUG, and G. PIECZENIK, 1976  A speculation on the origin of protein synthesis. Origins Life 7:389-397.[Medline]

DATTA, A. and S. JINKS-ROBERTSON, 1995  Association of increased spontaneous mutation rates with high levels of transcription in yeast. Science 268:1616-1619.[Abstract/Free Full Text]

DAVIS, C. A., L. GRATE, M. SPINGOLA, and M. ARES, JR., 2000  Test of intron predictions reveals novel splice sites, alternatively spliced mRNAs and new introns in meiotically regulated genes of yeast. Nucleic Acids Res. 28:1700-1706.[Abstract/Free Full Text]

DUFTON, M. J., 1997  Genetic code synonym quotas and amino acid complexity: Cutting the cost of proteins? J. Theor. Biol. 187:165-173.[Medline]

DURET, L., 2000  tRNA gene number and codon usage in the C. elegans genome are co-adapted for optimal translation of highly expressed genes. Trends Genet. 16:287-289.[Medline]

DURET, L., 2002  Evolution of synonymous codon usage in metazoans. Curr. Opin. Genet. Dev. 12:640-649.[Medline]

DURET, L. and D. MOUCHIROUD, 1999  Expression pattern and, surprisingly, gene length shape codon usage in Caenorhabditis, Drosophila, and Arabidopsis. Proc. Natl. Acad. Sci. USA 96:4482-4487.[Abstract/Free Full Text]

DURET, L. and D. MOUCHIROUD, 2000  Determinants of substitution rates in mammalian genes: expression pattern affects selection intensity but not mutation rate. Mol. Biol. Evol. 17:68-74.[Abstract/Free Full Text]

EIGEN, M., and P. SCHUSTER, 1979 The Hypercycle, a Principle of Natural Self-organization. Springer-Verlag, Berlin/New York.

FARABAUGH, P. J., 2000  Translational frameshifting: implications for the mechanism of translational frame maintenance. Prog. Nucleic Acid Res. Mol. Biol. 64:131-170.[Medline]

FRANCINO, M. P., L. CHAO, M. A. RILEY, and H. OCHMAN, 1996  Asymmetries generated by transcription-coupled repair in enterobacterial genes. Science 272:107-109.[Abstract]

FUTCHER, B., G. I. LATTER, P. MONARDO, C. S. MCLAUGHLIN, and J. I. GARRELS, 1999  A sampling of the yeast proteome. Mol. Cell. Biol. 19:7357-7368.[Abstract/Free Full Text]

GARAT, B. and H. MUSTO, 2000  Trends of amino acid usage in the proteins from the unicellular parasite Giardia lamblia.. Biochem. Biophys. Res. Commun. 279:996-1000.[Medline]

GAREL, J. P., 1974  Functional adaptation of tRNA population. J. Theor. Biol. 43:211-225.[Medline]

GERTON, J. L., J. DERISI, R. SHROFF, M. LICHTEN, and P. O. BROWN et al., 2000  Inaugural article: global mapping of meiotic recombination hotspots and coldspots in the yeast Saccharomyces cerevisiae. Proc. Natl. Acad. Sci. USA 97:11383-11390.[Abstract/Free Full Text]

GOFFEAU, A., B. G. BARRELL, H. BUSSEY, R. W. DAVIS, B. DUJON et al., 1996 Life with 6000 genes. Science 274: 546, 563–547.

GROSJEAN, H. and W. FIERS, 1982  Preferential codon usage in prokaryotic genes: the optimal codon-anticodon interaction energy and the selective codon usage in efficiently expressed genes. Gene 18:199-209.[Medline]

GROSJEAN, H., D. SANKOFF, W. M. JOU, W. FIERS, and R. J. CEDERGREN, 1978  Bacteriophage MS2 RNA: a correlation between the stability of the codon:anticodon interaction and the choice of code words. J. Mol. Evol. 12:113-119.[Medline]

GUTIÉRREZ, G., L. MARQUEZ, and A. MARIN, 1996  Preference for guanosine at first codon position in highly expressed Escherichia coli genes. A relationship with translational efficiency. Nucleic Acids Res. 24:2525-2527.[Abstract/Free Full Text]

HASTINGS, K. E., 1996  Strong evolutionary conservation of broadly expressed protein isoforms in the troponin I gene family and other vertebrate gene families. J. Mol. Evol. 42:631-640.[Medline]

HATFIELD, D., F. VARRICCHIO, M. RICE, and B. G. FORGET, 1982  The aminoacyl-tRNA population of human reticulocytes. J. Biol. Chem. 257:3183-3188.[Abstract/Free Full Text]

HEREFORD, L. M. and M. ROSBASH, 1977  Number and distribution of polyadenylated RNA sequences in yeast. Cell 10:453-462.[Medline]

HOLSTEGE, F. C., E. G. JENNINGS, J. J. WYRICK, T. I. LEE, and C. J. HENGARTNER et al., 1998  Dissecting the regulatory circuitry of a eukaryotic genome. Cell 95:717-728.[Medline]

IIDA, K. and H. AKASHI, 2000  A test of translational selection at ‘silent’ sites in the human genome: base composition comparisons in alternatively spliced genes. Gene 261:93-105.[Medline]

IKEMURA, T., 1982  Correlation between the abundance of yeast transfer RNAs and the occurrence of the respective codons in protein genes. Differences in synonymous codon choice patterns of yeast and Escherichia coli with reference to the abundance of isoaccepting transfer RNAs. J. Mol. Biol. 158:573-597.[Medline]

IKEMURA, T., 1985  Codon usage and tRNA content in unicellular and multicellular organisms. Mol. Biol. Evol. 2:13-34.[Abstract]

JANSEN, R. and M. GERSTEIN, 2000  Analysis of the yeast transcriptome with structural and functional categories: characterizing highly expressed proteins. Nucleic Acids Res. 28:1481-1488.[Abstract/Free Full Text]

KANAYA, S., Y. YAMADA, Y. KUDO, and T. IKEMURA, 1999  Studies of codon usage and tRNA genes of 18 unicellular organisms and quantification of Bacillus subtilis tRNAs: gene expression level and species-specific diversity of codon usage based on multivariate analysis. Gene 238:143-155.[Medline]

KARLIN, S. and P. BUCHER, 1992  Correlation analysis of amino acid usage in protein classes. Proc. Natl. Acad. Sci. USA 89:12165-12169.[Abstract/Free Full Text]

KIMURA, M., 1983 The Neutral Theory of Molecular Evolution. Cambridge University Press, Cambridge, UK/New York.

KLIMAN, R. M. and J. HEY, 1994  The effects of mutation and natural selection on codon bias in the genes of Drosophila. Genetics 137:1049-1056.[Abstract]

LI, W.-H., 1997 Molecular Evolution. Sinauer Associates, Sunderland, MA.

LOBRY, J. R. and C. GAUTIER, 1994  Hydrophobicity, expressivity and aromaticity are the major trends of amino-acid usage in 999 Escherichia coli chromosome-encoded genes. Nucleic Acids Res. 22:3174-3180.[Abstract/Free Full Text]

LOPEZ, P. J. and B. SERAPHIN, 1999  Genomic-scale quantitative analysis of yeast pre-mRNA splicing: implications for splice-site recognition. RNA 5:1135-1137.[Medline]

MARAIS, G., D. MOUCHIROUD, and L. DURET, 2001  Does recombination improve selection on codon usage? Lessons from nematode and fly complete genomes. Proc. Natl. Acad. Sci. USA 98:5688-5692.[Abstract/Free Full Text]

MAZEL, D. and P. MARLIÈRE, 1989  Adaptive eradication of methionine and cysteine from cyanobacterial light-harvesting proteins. Nature 341:245-248.[Medline]

MOREY, N. J., C. N. GREENE, and S. JINKS-ROBERTSON, 2000  Genetic analysis of transcription-associated mutation in Saccharomyces cerevisiae.. Genetics 154:109-120.[Abstract/Free Full Text]

MORIYAMA, E. N. and J. R. POWELL, 1998  Gene length and codon usage bias in Drosophila melanogaster, Saccharomyces cerevisiae and Escherichia coli.. Nucleic Acids Res. 26:3188-3193.[Abstract/Free Full Text]

MORTON, B. R. and B. G. SO, 2000  Codon usage in plastid genes is correlated with context, position within the gene, and amino acid content. J. Mol. Evol. 50:184-193.[Medline]

NASSAR, R. F. and R. D. COOK, 1976  Non-randomness of nucleotide bases in mRNA codons. Genet. Res. 27:353-362.[Medline]

NEI, M., 1975 Molecular Population Genetics and Evolution. North-Holland/American Elsevier, Amsterdam/New York.

OGLE, J. M., D. E. BRODERSEN, W. M. CLEMONS, JR., M. J. TARRY, and A. P. CARTER et al., 2001  Recognition of cognate transfer RNA by the 30S ribosomal subunit. Science 292:897-902.[Abstract/Free Full Text]

PAL, C., B. PAPP, and L. D. HURST, 2001a  Highly expressed genes in yeast evolve slowly. Genetics 158:927-931.[Free Full Text]

PAL, C., B. PAPP, and L. D. HURST, 2001b  Does the recombination rate affect the efficiency of purifying selection? The yeast genome provides a partial answer. Mol. Biol. Evol. 18:2323-2326.[Free Full Text]

PALACIOS, C. and J. J. WERNEGREEN, 2002  A strong effect of AT mutational bias on amino acid usage in Buchnera is mitigated at high-expression genes. Mol. Biol. Evol. 19:1575-1584.[Abstract/Free Full Text]

PERCUDANI, R., 2001  Restricted wobble rules for eukaryotic genomes. Trends Genet. 17:133-135.[Medline]

PERCUDANI, R. and S. OTTONELLO, 1999  Selection at the wobble position of codons read by the same tRNA in Saccharomyces cerevisiae.. Mol. Biol. Evol. 16:1752-1762.[Abstract]

PERCUDANI, R., A. PAVESI, and S. OTTONELLO, 1997  Transfer RNA gene redundancy and translational selection in Saccharomyces cerevisiae.. J. Mol. Biol. 268:322-330.[Medline]

PRECUP, J. and J. PARKER, 1987  Missense misreading of asparagine codons as a function of codon identity and context. J. Biol. Chem. 262:11351-11355.[Abstract/Free Full Text]

RICE, W. R., 1989  Analyzing tables of statistical tests. Evolution 43:223-225.

RICHMOND, R. C., 1970  Non-Darwinian evolution: a critique. Nature 225:1025-1028.[Medline]

ROBINSON, M., R. LILLEY, S. LITTLE, J. S. EMTAGE, and G. YARRANTON et al., 1984  Codon usage can affect efficiency of translation of genes in Escherichia coli.. Nucleic Acids Res. 12:6663-6671.[Abstract/Free Full Text]

RODNINA, M. V. and W. WINTERMEYER, 2001  Ribosome fidelity: tRNA discrimination, proofreading and induced fit. Trends Biochem. Sci. 26:124-130.[Medline]

SHARP, P. M. and E. COWE, 1991  Synonymous codon usage in Saccharomyces cerevisiae.. Yeast 7:657-678.[Medline]

SHARP, P. M., T. M. TUOHY, and K. R. MOSURSKI, 1986  Codon usage in yeast: cluster analysis clearly differentiates highly and lowly expressed genes. Nucleic Acids Res. 14:5125-5143.[Abstract/Free Full Text]

SHARP, P. M., M. STENICO, J. F. PEDEN, and A. T. LLOYD, 1993  Codon usage: Mutational bias, translational selection, or both? Biochem. Soc. Trans. 21:835-841.[Medline]

SHEPHERD, J. C., 1981  Periodic correlations in DNA sequences and evidence suggesting their evolutionary origin in a comma-less genetic code. J. Mol. Evol. 17:94-102.[Medline]

SHPAER, E. G., 1989  Amino acid composition is correlated with protein abundance in Escherichia coli: Can this be due to optimization of translational efficiency? Protein Seq. Data Anal. 2:107-110.[Medline]

SNEDECOR, G. W., and W. G. COCHRAN, 1989 Statistical Methods. Iowa State University Press, Ames, IA.

SORENSEN, M. A., C. G. KURLAND, and S. PEDERSEN, 1989  Codon usage determines translation rate in Escherichia coli.. J. Mol. Biol. 207:365-377.[Medline]

TRIFONOV, E. N., 1987  Translation framing code and frame-monitoring mechanism as suggested by the analysis of mRNA and 16 S rRNA nucleotide sequences. J. Mol. Biol. 194:643-652.[Medline]

TRIFONOV, E. N., 1992  Recognition of correct reading frame by the ribosome. Biochimie 74:357-362.[Medline]

TRIFONOV, E. N. and J. L. SUSSMAN, 1980  The pitch of chromatin DNA is reflected in its nucleotide sequence. Proc. Natl. Acad. Sci. USA 77:3816-3820.[Abstract/Free Full Text]

VARENNE, S., J. BUC, R. LLOUBES, and C. LAZDUNSKI, 1984  Translation is a non-uniform process. Effect of tRNA availability on the rate of elongation of nascent polypeptide chains. J. Mol. Biol. 180:549-576.[Medline]

VINOGRADOV, A. E., 2001  Intron length and codon usage. J. Mol. Evol. 52:2-5.[Medline]

WARNER, J. R., 1999  The economics of ribosome biosynthesis in yeast. Trends Biochem. Sci. 24:437-440.[Medline]

WOOTTON, J. C. and S. FEDERHEN, 1996  Analysis of compositionally biased regions in sequence databases. Methods Enzymol. 266:554-571.[Medline]

WRIGHT, S. I., B. LAUGA, and D. CHARLESWORTH, 2002  Rates and patterns of molecular evolution in inbred and outbred Arabidopsis. Mol. Biol. Evol. 19:1407-1420.[Abstract/Free Full Text]

XIA, X., 1998  How optimized is the translational machinery in Escherichia coli, Salmonella typhimurium and Saccharomyces cerevisiae? Genetics 149:37-44.[Abstract/Free Full Text]

YAMAO, F., Y. ANDACHI, A. MUTO, T. IKEMURA, and S. OSAWA, 1991  Levels of tRNAs in bacterial cells as affected by amino acid usage in proteins. Nucleic Acids Res. 19:6119-6122.[Abstract/Free Full Text]

ZAVALA, A., H. NAYA, H. ROMERO, and H. MUSTO, 2002  Trends in codon and amino acid usage in Thermotoga maritima.. J. Mol. Evol. 54:563-568.[Medline]

This article has been cited by other articles:

				Y. Hiraoka, K. Kawamata, T. Haraguchi, and Y. Chikashige Codon usage bias is correlated with gene expression levels in the fission yeast Schizosaccharomyces pombe. Genes Cells, April 1, 2009; 14(4): 499 - 509. [Abstract] [Full Text] [PDF]

				M. dos Reis and L. Wernisch Estimating Translational Selection in Eukaryotic Genomes Mol. Biol. Evol., February 1, 2009; 26(2): 451 - 461. [Abstract] [Full Text] [PDF]

				P. G. Higgs and W. Ran Coevolution of Codon Usage and tRNA Genes Leads to Alternative Stable States of Biased Codon Usage Mol. Biol. Evol., November 1, 2008; 25(11): 2279 - 2291. [Abstract] [Full Text] [PDF]

				A. K. Holloway, D. J. Begun, A. Siepel, and K. S. Pollard Accelerated sequence divergence of conserved genomic elements in Drosophila melanogaster Genome Res., October 1, 2008; 18(10): 1592 - 1601. [Abstract] [Full Text] [PDF]

				W. Jia and P. G. Higgs Codon Usage in Mitochondrial Genomes: Distinguishing Context-Dependent Mutation from Translational Selection Mol. Biol. Evol., February 1, 2008; 25(2): 339 - 351. [Abstract] [Full Text] [PDF]

				M. A. Gilchrist Combining Models of Protein Translation and Population Genetics to Predict Protein Production Rates from Codon Usage Patterns Mol. Biol. Evol., November 1, 2007; 24(11): 2362 - 2372. [Abstract] [Full Text] [PDF]

				R. Sasidharan and C. Chothia The selection of acceptable protein mutations PNAS, June 12, 2007; 104(24): 10080 - 10085. [Abstract] [Full Text] [PDF]

				N. Stoletzki and A. Eyre-Walker Synonymous Codon Usage in Escherichia coli: Selection for Translational Accuracy Mol. Biol. Evol., February 1, 2007; 24(2): 374 - 381. [Abstract] [Full Text] [PDF]

				J. K. Choi, S. C. Kim, J. Seo, S. Kim, and J. Bhak Impact of Transcriptional Properties on Essentiality and Evolutionary Rate Genetics, January 1, 2007; 175(1): 199 - 206. [Abstract] [Full Text] [PDF]

				J. D. Bloom, A. Raval, and C. O. Wilke Thermodynamics of Neutral Protein Evolution Genetics, January 1, 2007; 175(1): 255 - 266. [Abstract] [Full Text] [PDF]

				L. Nie, G. Wu, and W. Zhang Correlation of mRNA Expression and Protein Abundance Affected by Multiple Sequence Features Related to Translational Efficiency in Desulfovibrio vulgaris: A Quantitative Analysis Genetics, December 1, 2006; 174(4): 2229 - 2243. [Abstract] [Full Text] [PDF]

				B.-Y. Liao, N. M. Scott, and J. Zhang Impacts of Gene Essentiality, Expression Pattern, and Gene Compactness on the Evolutionary Rate of Mammalian Proteins Mol. Biol. Evol., November 1, 2006; 23(11): 2072 - 2080. [Abstract] [Full Text] [PDF]

				J. J. Elser, W. F. Fagan, S. Subramanian, and S. Kumar Signatures of Ecological Resource Availability in the Animal and Plant Proteomes Mol. Biol. Evol., October 1, 2006; 23(10): 1946 - 1951. [Abstract] [Full Text] [PDF]

				W.-Y. Ko, S. Piao, and H. Akashi Strong Regional Heterogeneity in Base Composition Evolution on the Drosophila X Chromosome Genetics, September 1, 2006; 174(1): 349 - 362. [Abstract] [Full Text] [PDF]

				E. M. Heizer Jr, D. W. Raiford, M. L. Raymer, T. E. Doom, R. V. Miller, and D. E. Krane Amino Acid Cost and Codon-Usage Biases in 6 Prokaryotic Genomes: A Whole-Genome Analysis Mol. Biol. Evol., September 1, 2006; 23(9): 1670 - 1680. [Abstract] [Full Text] [PDF]

				L. Xu, H. Chen, X. Hu, R. Zhang, Z. Zhang, and Z. W. Luo Average Gene Length Is Highly Conserved in Prokaryotes and Eukaryotes and Diverges Only Between the Two Kingdoms Mol. Biol. Evol., June 1, 2006; 23(6): 1107 - 1108. [Abstract] [Full Text] [PDF]

				C. E. Popescu, T. Borza, J. P. Bielawski, and R. W. Lee Evolutionary Rates and Expression Level in Chlamydomonas Genetics, March 1, 2006; 172(3): 1567 - 1576. [Abstract] [Full Text] [PDF]

				H. Akashi, W.-Y. Ko, S. Piao, A. John, P. Goel, C.-F. Lin, and A. P. Vitins Molecular Evolution in the Drosophila melanogaster Species Subgroup: Frequent Parameter Fluctuations on the Timescale of Molecular Divergence Genetics, March 1, 2006; 172(3): 1711 - 1726. [Abstract] [Full Text] [PDF]

				D. A. Drummond, A. Raval, and C. O. Wilke A Single Determinant Dominates the Rate of Yeast Protein Evolution Mol. Biol. Evol., February 1, 2006; 23(2): 327 - 337. [Abstract] [Full Text] [PDF]

				J. M. Comeron and T. B. Guthrie Intragenic Hill-Robertson Interference Influences Selection Intensity on Synonymous Mutations in Drosophila Mol. Biol. Evol., December 1, 2005; 22(12): 2519 - 2530. [Abstract] [Full Text] [PDF]

				D. A. Drummond, J. D. Bloom, C. Adami, C. O. Wilke, and F. H. Arnold Why highly expressed proteins evolve slowly PNAS, October 4, 2005; 102(40): 14338 - 14343. [Abstract] [Full Text] [PDF]

				M. L. Christianson Codon usage patterns distort phylogenies from or of DNA sequences Am. J. Botany, August 1, 2005; 92(8): 1221 - 1233. [Abstract] [Full Text] [PDF]

				T. M. Hambuch and J. Parsch Patterns of Synonymous Codon Usage in Drosophila melanogaster Genes With Sex-Biased Expression Genetics, August 1, 2005; 170(4): 1691 - 1700. [Abstract] [Full Text] [PDF]

				D. B. Carlini Context-Dependent Codon Bias and Messenger RNA Longevity in the Yeast Transcriptome Mol. Biol. Evol., June 1, 2005; 22(6): 1403 - 1411. [Abstract] [Full Text] [PDF]

				G. Marais, P. Nouvellet, P. D. Keightley, and B. Charlesworth Intron Size and Exon Evolution in Drosophila Genetics, May 1, 2005; 170(1): 481 - 485. [Abstract] [Full Text] [PDF]

				B. Lemos, B. R. Bettencourt, C. D. Meiklejohn, and D. L. Hartl Evolution of Proteins and Gene Expression Levels are Coupled in Drosophila and are Independently Associated with mRNA Abundance, Protein Length, and Number of Protein-Protein Interactions Mol. Biol. Evol., May 1, 2005; 22(5): 1345 - 1354. [Abstract] [Full Text] [PDF]

				D. P. Wall, A. E. Hirsh, H. B. Fraser, J. Kumm, G. Giaever, M. B. Eisen, and M. W. Feldman Functional genomic analysis of the rates of protein evolution PNAS, April 12, 2005; 102(15): 5483 - 5488. [Abstract] [Full Text] [PDF]

				J. Zhang and X. He Significant Impact of Protein Dispensability on the Instantaneous Rate of Protein Evolution Mol. Biol. Evol., April 1, 2005; 22(4): 1147 - 1155. [Abstract] [Full Text] [PDF]

				Z. Gu, L. David, D. Petrov, T. Jones, R. W. Davis, and L. M. Steinmetz Elevated evolutionary rates in the laboratory strain of Saccharomyces cerevisiae PNAS, January 25, 2005; 102(4): 1092 - 1097. [Abstract] [Full Text] [PDF]

				D. Wang, M. Hsieh, and W.-H. Li A General Tendency for Conservation of Protein Length Across Eukaryotic Kingdoms Mol. Biol. Evol., January 1, 2005; 22(1): 142 - 147. [Abstract] [Full Text] [PDF]

				H. Qin, W. B. Wu, J. M. Comeron, M. Kreitman, and W.-H. Li Intragenic Spatial Patterns of Codon Usage Bias in Prokaryotic and Eukaryotic Genomes Genetics, December 1, 2004; 168(4): 2245 - 2260. [Abstract] [Full Text] [PDF]

				S. Subramanian and S. Kumar Gene Expression Intensity Shapes Evolutionary Rates of the Proteins Encoded by the Vertebrate Genome Genetics, September 1, 2004; 168(1): 373 - 381. [Abstract] [Full Text] [PDF]

				S. I. Wright, C. B. K. Yau, M. Looseley, and B. C. Meyers Effects of Gene Expression on Molecular Evolution in Arabidopsis thaliana and Arabidopsis lyrata Mol. Biol. Evol., September 1, 2004; 21(9): 1719 - 1726. [Abstract] [Full Text] [PDF]

				J. M. Comeron Selective and Mutational Patterns Associated With Gene Expression in Humans: Influences on Synonymous Composition and Intron Presence Genetics, July 1, 2004; 167(3): 1293 - 1304. [Abstract] [Full Text] [PDF]

				H. B. Fraser, A. E. Hirsh, D. P. Wall, and M. B. Eisen Coevolution of gene expression among interacting proteins PNAS, June 15, 2004; 101(24): 9033 - 9038. [Abstract] [Full Text] [PDF]