Limits...
Selection on codon bias in yeast: a transcriptional hypothesis.

Trotta E - Nucleic Acids Res. (2013)

Bottom Line: We show that the most used codons in highly expressed genes can be predicted by mRNA structural data and that the codon choice at each synonymous site within an mRNA is not random with respect to the local secondary structure.Consistent with this, we report evidence supporting the adaptation of the tRNA pool to the codon profile of the most expressed genes rather than vice versa.We show that the correlation of codon usage with the gene expression level also includes the stop codons that are normally not decoded by aminoacyl-tRNAs.

View Article: PubMed Central - PubMed

Affiliation: Institute of Translational Pharmacology, Consiglio Nazionale delle Ricerche (CNR), Roma 00133, Italy.

ABSTRACT
Codons that code for the same amino acid are often used with unequal frequencies. This phenomenon is termed codon bias. Here, we report a computational analysis of codon bias in yeast using experimental and theoretical genome-wide data. We show that the most used codons in highly expressed genes can be predicted by mRNA structural data and that the codon choice at each synonymous site within an mRNA is not random with respect to the local secondary structure. Because we also found that the folding stability of intron sequences is strongly correlated with codon bias and mRNA level, our results suggest that codon bias is linked to mRNA folding structure through a mechanism that, at least partially, operates before pre-mRNA splicing. Consistent with this, we report evidence supporting the adaptation of the tRNA pool to the codon profile of the most expressed genes rather than vice versa. We show that the correlation of codon usage with the gene expression level also includes the stop codons that are normally not decoded by aminoacyl-tRNAs. The results reported here are consistent with a role for transcriptional forces in driving codon usage bias via a mechanism that improves gene expression by optimizing mRNA folding structures.

Show MeSH

Related in: MedlinePlus

Correlation of CAI score and mRNA level with the minimum free energy for intron sequences. Scatter plots showing the correlation of CAI score (panel A) and mRNA level (panel B) with the free energy of intron structures predicted using Quickfold from Mfold web server. Highly transcribed genes (mRNA per cell >32) are indicated by open circles. The broken line represents the linear regression line. The transcript concentration is expressed in log2-transformed molecule per cell.
© Copyright Policy - creative-commons
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC3814355&req=5

gkt740-F6: Correlation of CAI score and mRNA level with the minimum free energy for intron sequences. Scatter plots showing the correlation of CAI score (panel A) and mRNA level (panel B) with the free energy of intron structures predicted using Quickfold from Mfold web server. Highly transcribed genes (mRNA per cell >32) are indicated by open circles. The broken line represents the linear regression line. The transcript concentration is expressed in log2-transformed molecule per cell.

Mentions: To check whether the folding structural stability of introns is correlated with the PARS score of coding sequences, with transcript levels and with codon bias, we computed the minimum free energy (MFE) structure of intronic sequences using the Quikfold option of the DINAMelt web server (RNA 3.0) (28). The result shows that the minimum free energy (ΔG) for each intron is strongly and negatively correlated with the mRNA level (Rs = −0.693 and R2 = 0.461, N = 235, P < 10−5) (Figure 6B), and significantly correlated with the PARS score of coding regions (Rs = −0.362 and R2 = 0.09, N = 205; P < 10−5). Surprisingly, we also found that the minimum energy folding for each intron is strongly correlated with codon bias scores of coding sequences (Figure 6A): CAI (Rs = −0.687 and R2 = 0.515, N = 235, P < 10−5), CBI (Rs = −0.688 and R2 = 0.531, N = 235, P < 10−5), Fop (Rs = −0.686 and R2 = 0.515, N = 235, P < 10−5), CBI (Rs = −0.687 and R2 = 0.526, N = 235, P < 10−5) and Nc (Rs = 0.670 and R2 = 0.479, N = 235, P < 10−5). The MFE of a RNA sequence is a function of its length, nucleotide composition and nucleotide order. Thus, when intron is bound to the assembled spliceosome and, therefore, its short-range interactions with adjacent exon sequences are constrained by the ribonucleoprotein complex, its folding stability should increase with the length of its sequence. It has been previously reported that intron length in yeast genes is strongly and positively correlated with codon bias and the level of gene expression (36). Here, we found a strong correlation between the MFE and the length of intronic sequences (Rs = −0.959 and R2 = 0.479, N = 249, P < 10−5) showing that almost all the computed MFE is attributable to sequence size. We also found that the length as well as the MFE of intronic sequences is weakly correlated with CDS length: Rs = −0.192 and R2 = 0.03, N = 249, P = 0.0066 and Rs = 0.202 and R2 = 0.03, N = 249, P = 0.0059, respectively. Therefore, although the sequence length of introns and exons are very weakly and negatively correlated, their folding stabilities are strongly and positively correlated. The correlation of transcription level with the length-dependent MFE of introns and the PARS score of exons may involve long-range effects of pre-mRNA folding structures or structure-dependent interactions with the components of the transcription elongation complex. These results suggest that the association between the mRNA folding stability, codon bias and the transcript level occurs, at least partially, before the splicing of pre-mRNA.Figure 6.


Selection on codon bias in yeast: a transcriptional hypothesis.

Trotta E - Nucleic Acids Res. (2013)

Correlation of CAI score and mRNA level with the minimum free energy for intron sequences. Scatter plots showing the correlation of CAI score (panel A) and mRNA level (panel B) with the free energy of intron structures predicted using Quickfold from Mfold web server. Highly transcribed genes (mRNA per cell >32) are indicated by open circles. The broken line represents the linear regression line. The transcript concentration is expressed in log2-transformed molecule per cell.
© Copyright Policy - creative-commons
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC3814355&req=5

gkt740-F6: Correlation of CAI score and mRNA level with the minimum free energy for intron sequences. Scatter plots showing the correlation of CAI score (panel A) and mRNA level (panel B) with the free energy of intron structures predicted using Quickfold from Mfold web server. Highly transcribed genes (mRNA per cell >32) are indicated by open circles. The broken line represents the linear regression line. The transcript concentration is expressed in log2-transformed molecule per cell.
Mentions: To check whether the folding structural stability of introns is correlated with the PARS score of coding sequences, with transcript levels and with codon bias, we computed the minimum free energy (MFE) structure of intronic sequences using the Quikfold option of the DINAMelt web server (RNA 3.0) (28). The result shows that the minimum free energy (ΔG) for each intron is strongly and negatively correlated with the mRNA level (Rs = −0.693 and R2 = 0.461, N = 235, P < 10−5) (Figure 6B), and significantly correlated with the PARS score of coding regions (Rs = −0.362 and R2 = 0.09, N = 205; P < 10−5). Surprisingly, we also found that the minimum energy folding for each intron is strongly correlated with codon bias scores of coding sequences (Figure 6A): CAI (Rs = −0.687 and R2 = 0.515, N = 235, P < 10−5), CBI (Rs = −0.688 and R2 = 0.531, N = 235, P < 10−5), Fop (Rs = −0.686 and R2 = 0.515, N = 235, P < 10−5), CBI (Rs = −0.687 and R2 = 0.526, N = 235, P < 10−5) and Nc (Rs = 0.670 and R2 = 0.479, N = 235, P < 10−5). The MFE of a RNA sequence is a function of its length, nucleotide composition and nucleotide order. Thus, when intron is bound to the assembled spliceosome and, therefore, its short-range interactions with adjacent exon sequences are constrained by the ribonucleoprotein complex, its folding stability should increase with the length of its sequence. It has been previously reported that intron length in yeast genes is strongly and positively correlated with codon bias and the level of gene expression (36). Here, we found a strong correlation between the MFE and the length of intronic sequences (Rs = −0.959 and R2 = 0.479, N = 249, P < 10−5) showing that almost all the computed MFE is attributable to sequence size. We also found that the length as well as the MFE of intronic sequences is weakly correlated with CDS length: Rs = −0.192 and R2 = 0.03, N = 249, P = 0.0066 and Rs = 0.202 and R2 = 0.03, N = 249, P = 0.0059, respectively. Therefore, although the sequence length of introns and exons are very weakly and negatively correlated, their folding stabilities are strongly and positively correlated. The correlation of transcription level with the length-dependent MFE of introns and the PARS score of exons may involve long-range effects of pre-mRNA folding structures or structure-dependent interactions with the components of the transcription elongation complex. These results suggest that the association between the mRNA folding stability, codon bias and the transcript level occurs, at least partially, before the splicing of pre-mRNA.Figure 6.

Bottom Line: We show that the most used codons in highly expressed genes can be predicted by mRNA structural data and that the codon choice at each synonymous site within an mRNA is not random with respect to the local secondary structure.Consistent with this, we report evidence supporting the adaptation of the tRNA pool to the codon profile of the most expressed genes rather than vice versa.We show that the correlation of codon usage with the gene expression level also includes the stop codons that are normally not decoded by aminoacyl-tRNAs.

View Article: PubMed Central - PubMed

Affiliation: Institute of Translational Pharmacology, Consiglio Nazionale delle Ricerche (CNR), Roma 00133, Italy.

ABSTRACT
Codons that code for the same amino acid are often used with unequal frequencies. This phenomenon is termed codon bias. Here, we report a computational analysis of codon bias in yeast using experimental and theoretical genome-wide data. We show that the most used codons in highly expressed genes can be predicted by mRNA structural data and that the codon choice at each synonymous site within an mRNA is not random with respect to the local secondary structure. Because we also found that the folding stability of intron sequences is strongly correlated with codon bias and mRNA level, our results suggest that codon bias is linked to mRNA folding structure through a mechanism that, at least partially, operates before pre-mRNA splicing. Consistent with this, we report evidence supporting the adaptation of the tRNA pool to the codon profile of the most expressed genes rather than vice versa. We show that the correlation of codon usage with the gene expression level also includes the stop codons that are normally not decoded by aminoacyl-tRNAs. The results reported here are consistent with a role for transcriptional forces in driving codon usage bias via a mechanism that improves gene expression by optimizing mRNA folding structures.

Show MeSH
Related in: MedlinePlus