Codon usage bias is correlated with gene expression levels in the fission yeast Schizosaccharomyces pombe

Transcription

1 Codon usage bias is correlated with gene expression levels Blackwell Y Hiraoka usage Publishing et al. bias in fission Inc yeast in the fission yeast Schizosaccharomyces pombe Yasushi Hiraoka 1,2,3, *, Kenichi Kawamata 1, Tokuko Haraguchi 1,2 and Yuji Chikashige 1,2 1 Department of Biology, Graduate School of Science, Osaka University, 1-1 Machikaneyama, Toyonaka, , Japan 2 Kobe Advanced ICT Research Center, National Institute of Information and Communications Technology, Iwaoka, Iwaoka-cho, Nishi-ku, Kobe , Japan 3 Graduate School of Frontier Biosciences, Osaka University, 1-3 Yamadaoka, Suita, , Japan Usage of synonymous codons represents a characteristic pattern of preference in each organism. It has been inferred that such bias of codon usage has evolved as a result of adaptation for efficient synthesis of. Here we examined synonymous codon usage in genes of the fission yeast Schizosaccharomyces pombe, and compared codon usage bias with expression levels of the gene. In this organism, synonymous codon usage bias was correlated with expression levels of the gene; the bias was most obvious in two-codon amino acids. A similar pattern of the codon usage bias was also observed in Saccharomyces cerevisiae, Arabidopsis thaliana and Caenorhabditis elegans, but was not obvious in Oryza sativa, Drosophila melanogaster, Takifugu rubripes and Homo sapiens. As codons of the highly expressed genes have greater influence on translational efficiency than codons of genes expressed at lower levels, it is likely that codon usage in the S. pombe genome has been optimized by translational selection through evolution. Introduction In protein synthesis, a triplet codon of mrna is translated to a respective amino acid. In 18 out of 20 amino acids (exceptions are Met and Trp), multiple codons are assigned for one amino acid in their translation. Such synonymous codons are not used equally, but are instead biased. One of the models explaining codon usage bias relies on selection for translational efficiency. It has been shown in some species that codon usage preference is correlated with the abundance of the respective trna (Ikemura 1985; Andersson & Kurland 1990; Sharp et al. 1993; Akashi 2001; Kanaya et al. 2001) which is expected to affect translational efficiency of the gene product (Akashi 2003). Although the question of codon usage bias has been extensively studied in many organisms, including humans as well as human infectious viruses (e.g. Plotkin & Dushoff 2003; Meintjes & Rodrigo 2005; Ren et al. 2007; van Hemert et al. 2007), unicellular organisms can provide a simpler model system for examining whether codon usage bias is related to cellular metabolism. In Escherichia coli, clear evidence has been obtained for correlation Communicated by: Masayuki Yamamoto (The University of Tokyo) *Correspondence: hiraoka@fbs.osaka-u.ac.jp between codon usage preference and trna abundance (Ikemura 1981a,b). Also, in the budding yeast Saccharomyces cerevisiae, it has been shown that codon usage bias is correlated with mrna copy numbers of the gene (Sharp & Cowe 1991; Akashi 2003). Translational efficiency can be directly related to growth rates, and thus it is expected that codon usage has been optimized for growth through evolution in such situations. Here we report codon usage in another unicellular eukaryote, the fission yeast Schizosaccharomyces pombe, in which the entire genome sequence has been determined (Wood et al. 2002). The usage of synonymous codons has been reported previously for three highly expressed genes and another 153 genes that are expressed at a lower level (Forsburg 1994). We examined codon usage for genes in the S. pombe genome, and found significant difference in codon usage bias with respect to expression levels of the gene. Thus, we propose that codon usage of the gene can predict its expression levels in S. pombe. Results Codon usage profiles in Schizosaccharomyces pombe Expression levels of 4932 open reading frames (ORFs) in the S. pombe genome were determined based on DNA DOI: /j x 499

2 Y Hiraoka et al. Figure 1 Histogram of mrna copy numbers. The majority of genes showed low levels of expression (A) whereas ribosomal were highly expressed (B). Thus, average codon usage is representative of with low expression levels. microarray analysis (Fig. 1A). Among these ORFs, genes for ribosomal showed high levels of expression (Fig. 1B). We first examined the usage of synonymous codons in the ribosomal protein genes, and found a significant difference compared with the genome average (Table 1). The difference can be seen clearly in twocodon amino acids (Lys, Asn, Phe, His, Glu, Cys, Tyr, Asp and Gln) where the usage bias of two codons for ribosomal was opposite to that of total (Fig. 2). In contrast, codon usage in meiotic was similar to that in total (Table 1; Fig. 2). The use of codons with G or C at the third position approximately corresponds to the GC content in the S. pombe genome (38%) in total and in meiotic, but significantly higher in ribosomal, with an exception of Gln (Fig. 2). Figure 2 Codon usage in two-codon amino acids. Occurrence of G or C nucleotides at the third position of the codon was approximately equal to the average GC content in the S. pombe genome (38%; indicated by the broken line). This was also the case in meiotic. In contrast, the G/C frequency was significantly higher (except for Gln) in ribosomal. Codon usage and gene expression in Schizosaccharomyces pombe To examine whether the codon usage bias is related to levels of gene expression or not, we compared codon 500 Genes to Cells (2009) 14, The Authors

3 Codon usage bias in fission yeast Table 1 Codon usage Codon Total Meiotic Ribosomal Codon Total Meiotic Ribosomal GGG (Gly) ACT (Thr) GGA (Gly) ACC (Thr) GGT (Gly) TGG (Trp) GGC (Gly) TGA (End) GAG (Glu) TAG (End) GAA (Glu) TAA (End) GAT (Asp) TGT (Cys) GAC (Asp) TGC (Cys) GTG (Val) TAT (Tyr) GTA (Val) TAC (Tyr) GTT (Val) TTG (Leu) GTC (Val) TTA (Leu) GCG (Ala) CTG (Leu) GCA (Ala) CTA (Leu) GCT (Ala) CTT (Leu) GCC (Ala) CTC (Leu) AGG (Arg) TTT (Phe) AGA (Arg) TTC (Phe) CGG (Arg) TCG (Ser) CGA (Arg) TCA (Ser) CGT (Arg) TCT (Ser) CGC (Arg) TCC (Ser) AAG (Lys) AGT (Ser) AAA (Lys) AGC (Ser) AAT (Asn) CAG (Gln) AAC (Asn) CAA (Gln) ATG (Met) CAT (His) ATA (Ile) CAC (His) ATT (Ile) CCG (Pro) ATC (Ile) CCA (Pro) ACG (Thr) CCT (Pro) ACA (Thr) CCC (Pro) usage in selected genes at seven levels of expression. We found that the codon usage varied according to level of gene expression (Table 2, Figs 3 5), and that codon usage in ribosomal was characteristic of that in highly-expressed genes. Correlation of codon usage with gene expression was particularly high in the two-codon amino acids (Fig. 3). Similar tendencies for correlation were also observed in a three-codon amino acid (Ile) and four-codon amino acids (Gly, Val, Ala, Thr and Pro) (Fig. 4). In Ile, the ATC codon showed codon usage increasing as a function of expression level whereas the ATT codon was used most frequently at all levels of gene expression. In Val, Ala, Thr and Pro, one of the four codons was most frequently used at all levels of gene expression although there was a tendency for a slight increase as a function of expression. One of the other three codons showed a significant increase in codon usage as a function of the level of gene expression. In contrast, in Gly, the most frequently used codon, GGT, showed a significant increase as a function of the expression. In six-codon amino acids, increasing usage as a function of gene expression was seen in codons CGT (Arg), TCG and TCC (Ser), and TTG, CTT and CTC (Leu) (Table 2). In these cases, six codons can be divided into two-codon and four-codon components depending on the first nucleotide of the codon. The usage of two-codon components showed a pattern similar to that in the two-codon amino acids (Fig. 5). Among four-codon components (Fig. 5), Ser and Leu showed a pattern similar to that in Val, Ala, Thr and Pro, in which one of the four codons was most frequently used at all levels of 501

4 Y Hiraoka et al. Table 2 Codon usage and gene expression levels mrna copy number per cell mrna copy number per cell Codon Codon GGG (Gly) ACT (Thr) GGA (Gly) ACC (Thr) GGT (Gly) TGG (Trp) GGC (Gly) TGA (End) GAG (Glu) TAG (End) GAA (Glu) TAA (End) GAT (Asp) TGT (Cys) GAC (Asp) TGC (Cys) GTG (Val) TAT (Tyr) GTA (Val) TAC (Tyr) GTT (Val) TTG (Leu) GTC (Val) TTA (Leu) GCG (Ala) CTG (Leu) GCA (Ala) CTA (Leu) GCT (Ala) CTT (Leu) GCC (Ala) CTC (Leu) AGG (Arg) TTT (Phe) AGA (Arg) TTC (Phe) CGG (Arg) TCG (Ser) CGA (Arg) TCA (Ser) CGT (Arg) TCT (Ser) CGC (Arg) TCC (Ser) AAG (Lys) AGT (Ser) AAA (Lys) AGC (Ser) AAT (Asn) CAG (Gln) AAC (Asn) CAA (Gln) ATG (Met) CAT (His) ATA (Ile) CAC (His) ATT (Ile) CCG (Pro) ATC (Ile) CCA (Pro) ACG (Thr) CCT (Pro) ACA (Thr) CCC (Pro) gene expression, and another codon showed an increase as a function of expression. In contrast, four-codon components in Arg showed a codon usage pattern similar to that in Gly, in which only one of the four codons increased as a function of expression. Taken together, these results showed that the usage of synonymous codons is correlated with expression levels of the gene in all amino acids in S. pombe. Codon usage profiles in other organisms The codon usage of ribosomal protein genes was compared to the genomic average in Saccharomyces cerevisiae, Arabidopsis thaliana, Caenorhabditis elegans, Oryza sativa, Drosophila melanogaster, Takifugu rubripes and Homo sapiens (Supporting Table S1). Codon usage of the ribosomal in Saccharomyces cerevisiae, A. thaliana and C. elegans was significantly different to the genomic average, which is especially clear in two-codon amino acids. In contrast, no significant difference in codon usage was observed between the ribosomal protein genes and the genomic average in O. sativa, D. melanogaster, T. rubripes and H. sapiens (Fig. 6). Codon usage of two-codon amino acids in ribosomal was plotted against that in total. Plots were lined along the orthogonal line for O. sativa, D. melanogaster, T. rubripes and H. sapiens (Fig. 6A), indicating that these two groups of have a similar usage of codons. In contrast, in Saccharomyces cerevisiae, A. thaliana and C. elegans, plots showed a scattered distribution from the orthogonal line (Fig. 6B), 502 Genes to Cells (2009) 14, The Authors

5 Codon usage bias in fission yeast Figure 3 Codon usage as a function of expression levels in two-codon amino acids. indicating that codon usage of ribosomal differs from that of total as in S. pombe. In contrast, a recent report shows that synonymous codon usage bias is only moderately correlated with expression levels of the gene in A. thaliana (dos Reis & Wernisch 2009). This apparent discrepancy may be because our results are based on only ribosomal protein genes. We examined ribosomal protein genes as representatives of highly-expressed genes to avoid possible complexity associated with tissuespecific expression levels in multi-cellular organisms. Discussion We demonstrated that codon usage is correlated with expression levels of the genes in S. pombe. In S. pombe, the majority of genes express less than 32 copies of mrna; only a small number of the genes (12%) express more than 32 copies of mrna, but the total number of mrna copies expressed from these genes occupies 50% of mrna in the cell (Fig. 1A). Thus, the S. pombe genes can be divided into two groups that produce one half of 503

6 Y Hiraoka et al. Figure 4 Codon usage as a function of expression levels in three- and four-codon amino acids. Figure 5 Codon usage as a function of expression levels in six-codon amino acids. Usage patterns of two-codon and four-codon components are shown separately in the upper and lower panels. 504 Genes to Cells (2009) 14, The Authors

7 Codon usage bias in fission yeast Figure 6 Codon usage profiles in ribosomal vs. total. Codon usage of two-codon amino acids (vertical axis) is plotted against that of total (horizontal axis). If their codon usage biases were close to each other, the plots were lined up along the orthogonal line. 505

8 Y Hiraoka et al. mrna copies in the cell: 12% of genes with high levels of expression (above 32 copies of mrna) and the rest of genes with low levels of expression (below 32 copies of mrna). Interestingly, codon usage preference switches at an expression level above and below 32 copies of mrna. An implication of this result is that major codons are allocated for the group of genes with high expression levels to allow efficient translation of the mrna, and that genes with low expression levels use minor codons to avoid competition with highly expressed genes. This usage allocation of codons may optimize overall translational efficiency to provide the best growth rate in S. pombe, and thus such codon usage bias may have been selected through evolution. From these results, we formulated a set of coefficients for each codon to determine a type of codon usage of a gene of interest (Table 3). This formula calculates codon usage score which provides a criterion for judgment of two types of codon usage: high or low expression type (above or below 32 copies of mrna, respectively) (see Experimental procedures). It should be mentioned that the formula does not predict expression levels of the gene, but instead predicts the types of codon usage. Examples for two S. pombe genes, cdc2 and act1, are shown in Table 3. The positive and negative values represent high- and low-expression types of codon usage, respectively. This formula may also be used to judge a type of codon usage of a foreign gene introduced to S. pombe cells. There are caveats to the analysis of data from host cells that have been transfected with a foreign gene. Examples for two versions of GFP are shown in Table 3. Aequorea victoria GFP has codon usage similar to that in low expression genes in S. pombe whereas humanized EGFP has codon usage of the high-expression type. Introduction of an extremely biased codon usage may affect growth rates through translational efficiency. High translational efficiency of foreign genes may not always be appreciated in the host cell as it may compete with synthesis of ribosomal. In choosing versions of GFP, translational efficiency might be optimized by matching the codon usage type with the target gene of fusion. Mutations causing amino acid substitution can directly affect functions of the protein, and thus have been effectively selected through evolution. In contrast, mutations only changing codon usage may be less effectively selected through evolution. However, in unicellular organisms such as yeasts, translational efficiency can be directly related to growth rates, and thus optimization of codon usage is expected to occur as a consequence of evolutional pressures. Experimental procedures Databases for codon usage and ribosomal Codon usages in the genome of the species examined were obtained from the web site Codon Usage Database, Table 3 Prediction of the types of codon usage Codon Expression coefficient (Ec) Cdc2 Act1 GFP-S65T EGFP Usage Σ Ec Usage Σ Ec Usage Σ Ec Usage Σ Ec GGG (Gly) GGA (Gly) GGT (Gly) GGC (Gly) GAG (Glu) GAA (Glu) GAT (Asp) GAC (Asp) GTG (Val) GTA (Val) GTT (Val) GTC (Val) GCG (Ala) GCA (Ala) GCT (Ala) GCC (Ala) Continued overleaf 506 Genes to Cells (2009) 14, The Authors

9 Codon usage bias in fission yeast Table 3 Continued Codon Expression coefficient (Ec) Cdc2 Act1 GFP-S65T EGFP Usage Σ Ec Usage Σ Ec Usage Σ Ec Usage Σ Ec AGG (Arg) AGA (Arg) AGT (Ser) AGC (Ser) AAG (Lys) AAA (Lys) AAT (Asn) AAC (Asn) ATG (Met) ATA (Ile) ATT (Ile) ATC (Ile) ACG (Thr) ACA (Thr) ACT (Thr) ACC (Thr) TGG (Trp) TGA (End) TGT (Cys) TGC (Cys) TAG (End) TAA (End) TAT (Tyr) TAC (Tyr) TTG (Leu) TTA (Leu) TTT (Phe) TTC (Phe) TCG (Ser) TCA (Ser) TCT (Ser) TCC (Ser) CGG (Arg) CGA (Arg) CGT (Arg) CGC (Arg) CAG (Gln) CAA (Gln) CAT (His) CAC (His) CTG (Leu) CTA (Leu) CTT (Leu) CTC (Leu) CCG (Pro) CCA (Pro) CCT (Pro) CCC (Pro) Codon usage score Codon usage type prediction Low (100%) High (100%) Low (75%) High (100%) 507

10 Y Hiraoka et al. (Nakamura et al. 2000). Genes for ribosomal were obtained from the web site Ribosomal Protein Gene Database, (Nakao et al. 2004). Analysis of codon usage and gene expression in Schizosaccharomyces pombe Levels of gene expression in S. pombe were determined using a DNA microarray containing 4932 ORFs (Chikashige et al. 2007). Relative amounts of mrna for each ORF were measured by DNA microarray using genomic DNA as a reference, and the copy number of mrna was calculated using an estimate of the total mrna number in the cell as copies. The entire DNA microarray data were deposited in gene expression omnibus (GEO, < accession number GSE13554). Expression levels of the ORFs are available through the web site: < CellMagic/index.html>. Thirty ORFs with the highest expression levels express an average of 256 copies of mrna, and contain a total of 7326 codons. Other groups of ORFs that express an average of 128, 64, 32, 16, 8 and 4 copies of mrna were selected such that each group contained at least 7000 codons or 20 ORFs (7263 codons of 34 ORFs, 7291 codons of 20 ORFs, 7234 codons of 20 ORFs, codons of 20 ORFs, codons of 20 ORFs, and codons of 20 ORFs, respectively). For S. pombe ribosomal, 141 ORFs containing codons were selected. For S. pombe meiotic, 18 ORFs containing 9039 codons were selected. These genes selected are listed in Supporting Table S2. Analysis of codon usage in other organisms Genes for ribosomal were obtained in Ribosomal Protein Gene Database, < and codon usage of these genes was examined by the Countcodon program in Codon Usage Database, < Formulation for codon usage score Codon usage score (S) was defined as an average of expression coefficients over all codons contained in the gene of interest: N Ec() r r S = = 1 N N is the number of codons in the gene, Ec is an expression coefficient calculated for each triplet codon as listed in Table 3, and the suffix r represents the array number of triplet codons from 1 to N. The expression coefficients were obtained by subtracting the codon usage of 32 copies of mrna from that of 256 copies of mrna for each triplet codon, and were normalized to make the SD 32 value equal to unity where the SD 32 is a standard deviation of the S-value for the group of genes expressing 32 copies of mrna. Positive and negative signs of codon usage score indicate the high-expression type (above 32 copies) and low-expression type (below 32 copies) of codon usage, respectively; its absolute value indicates a fold multiple of SD 32, which provides a measure for the likelihood of the prediction. Acknowledgements We thank Chihiro Tsutsumi for her technical support in DNA microarray analysis. This work was supported by grants from the Ministry of Education, Culture, Sports, Science and Technology of Japan (MEXT) to YH, TH and CY. References Akashi, H. (2001) Gene expression and molecular evolution. Curr. Opin. Genet. Dev. 11, Akashi, H. (2003) Translational selection and yeast proteome evolution. Genetics 164, Andersson, S.G. & Kurland, C.G. (1990) Codon preferences in free-living microorganisms. Microbiol. Rev. 54, Chikashige, Y., Tsutsumi, C., Okamasa, K., Yamane, M., Nakayama, J., Niwa, O., Haraguchi, T. & Hiraoka, Y. (2007) Gene expression and distribution of Swi6 in partial aneuploids of the fission yeast Schizosaccharomyces pombe. Cell Struct. Funct. 32, Forsburg, S.L. (1994) Codon usage table for Schizosaccharomyces pombe. Yeast 10, van Hemert, F.J., Berkhout, B. & Lukashov, V.V. (2007) Host-related nucleotide composition and codon usage as driving forces in the recent evolution of the Astroviridae. Virology 361, Ikemura, T. (1981a) Correlation between the abundance of Escherichia coli transfer RNAs and the occurrence of the respective codons in its protein genes. J. Mol. Biol. 146, Ikemura, T. (1981b) Correlation between the abundance of Escherichia coli transfer RNAs and the occurrence of the respective codons in its protein genes: a proposal for a synonymous codon choice that is optimal for the E. coli translational system. J. Mol. Biol. 151, Ikemura, T. (1985) Codon usage and trna content in unicellular and multicellular organisms. Mol. Biol. Evol. 2, Kanaya, S., Yamada, Y., Kinouchi, M., Kudo, Y. & Ikemura, T. (2001) Codon usage and trna genes in eukaryotes: correlation of codon usage diversity with translation efficiency and with CG-dinucleotide usage as assessed by multivariate analysis. J. Mol. Evol. 53, Meintjes, P.L. & Rodrigo, A.G. (2005) Evolution of relative synonymous codon usage in Human Immunodeficiency Virus type-1. J. Bioinform. Comput. Biol. 3, Nakamura, Y., Gojobori, T. & Ikemura, T. (2000) Codon usage tabulated from international DNA sequence databases: status for the year Nucleic Acids Res. 28, 292. Nakao, A., Yoshihama, M. & Kenmochi, N. (2004) RPG: the Ribosomal Protein Gene database. Nucleic Acids Res. 32, D168 D Genes to Cells (2009) 14, The Authors

11 Codon usage bias in fission yeast Plotkin, J.B. & Dushoff, J. (2003) Codon bias and frequencydependent selection on the hemagglutinin epitopes of influenza A virus. Proc. Natl. Acad. Sci. USA 100, dos Reis, M. & Wernisch, L. (2009) Estimating translational selection in Eukaryotic genomes. Mol. Biol. Evol. 26, Ren, L., Gao, G., Zhao, D., Ding, M., Luo, J. & Deng, H. (2007) Developmental stage related patterns of codon usage and genomic GC content: searching for evolutionary fingerprints with models of stem cell differentiation. Genome. Biol. 8, R35. Sharp, P.M. & Cowe, E. (1991) Synonymous codon usage in Saccharomyces cerevisiae. Yeast 7, Sharp, P.M., Stenico, M., Peden, J.F. & Lloyd, A.T. (1993) Codon usage: mutational bias, translational selection, or both? Biochem. Soc. Trans. 21, Wood, V., Gwilliam, R., Rajandream, M.A. et al. (2002) The genome sequence of Schizosaccharomyces pombe. Nature 415, Supporting Information/Supplementary materials The following Supporting Information can be found in the online version of the article: Table S1 Codon usage in ribosomal protein genes Table S2 List of genes selected Additional Supporting Information may be found in the online version of the article. Please note: Wiley-Blackwell are not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article. Received: 1 November 2008 Accepted: 15 January