Modelling the Efficiency of Codon–tRNA Interactions Based on Codon Usage Bias

DNA RESEARCH 21, 511–525, (2014) Advance Access publication on 6 June 2014 doi:10.1093/dnares/dsu017 Modelling the Efficiency of Codon –tRNA Interactions Based on Codon Usage Bias RENANA Sabi1 and TAMIR Tuller1,2,* Department of Biomedical Engineering, Tel Aviv University, Tel Aviv, Israel1 and The Sagol School of Neuroscience, Tel-Aviv University, Tel-Aviv, Israel2 *To whom correspondence should be addressed. Tel. þ972-64058363. Fax. þ972-64058363. Email: Edited by Prof. Hiroyuki Toh (Received 10 January 2014; accepted 26 April 2014) Abstract The tRNA adaptation index (tAI) is a widely used measure of the efficiency by which a coding sequence is recognized by the intra-cellular tRNA pool. This index includes among others weights that represent wobble interactions between codons and tRNA molecules. Currently, these weights are based only on the gene expression in Saccharomyces cerevisiae. However, the efficiencies of the different codon–tRNA interactions are expected to vary among different organisms. In this study, we suggest a new approach for adjusting the tAI weights to any target model organism without the need for gene expression measurements. Our method is based on optimizing the correlation between the tAI and a measure of codon usage bias. Here, we show that in non-fungal the new tAI weights predict protein abundance significantly better than the traditional tAI weights. The unique tRNA –codon adaptation weights computed for 100 different organisms exhibit a significant correlation with evolutionary distance. The reported results demonstrate the usefulness of the new measure in future genomic studies. Key words: codon usage bias; tRNA adaptation index; protein levels; wobble interactions; ribosome 1. Introduction Allowed by the redundancy of the genetic code, coding regions exhibit non-uniform usage of synonymous codons. This deviation from uniform codon usage is termed codon usage bias (CUB) and is related among others to various aspects of gene translation (and more generally gene expression) and its efficiency;1 – 10 specifically, it was suggested that it regulates transcription and translation, but may also be related to recombination rate. Indeed, it is known for over 30 years that in most organisms the degree of CUB of genes correlates with their expression levels.11 – 14 Various approaches for quantifying the CUB of a gene have been suggested in the last decades: the effective number of codons, for instance, measures deviations from equal usage of synonymous codons,13 while other indices such as the frequency of optimal codons,15 the codon bias index,11 and the codon adaptation index (CAI)16 define a subset of ‘optimal’ codons and measure the frequency of these codons in the coding region of the gene. The CUB indices have been employed in hundreds of previous studies. For example, it is known that in many organisms (e.g. Escherichia coli) the CAI exhibits a very high correlation with protein levels (similar to the one obtained between mRNA levels and protein levels17); thus, CAI has been frequently used as a proxy of expression levels (see, for example,18 – 20). In addition, it has been employed in a vast number of key studies in the recent years.18,19,21,22 One disadvantage of measures that are based on a set of reference genes11,16,23 is the fact that in the case of organisms with poor genomic data and without large scale cellular measurements, creating a meaningful reference set is challenging. Another disadvantage of these measures is the fact that they cannot separate between the various possible causes of CUB in highly # The Author 2014. Published by Oxford University Press on behalf of Kazusa DNA Research Institute. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited. 512 Inference of Codon –tRNA Interaction Efficiencies expressed genes: some of them may be related directly to the translation process (e.g. co-evolution and/or adaptation to the tRNA pool8,24,25) and others may not be related to translation (e.g. GC content and folding9,26 – 28). In 2004, dos Reis et al. 8 proposed the tRNA adaptation index (tAI), which aims to estimate only the adaptation of codons/genes to some aspects directly related to the elongation step occurs in the ribosome via the adaptation to the tRNA pool, wobble interactions, and properties of the ribosome. Specifically, the tAI considers the fact that different tRNA species can recognize a codon with different affinities.2,8,29 Thus, the tAI is different than CUB-based measurements mentioned above and provides important information related to translation that is not necessarily covered by CUB measures. Indeed, measures of the adaptation of genes to the tRNA pool (such as the tAI) have extensively been used in the recent years for studying questions in diverse biomedical disciplines such as evolutionary biology, functional genomics, and systems biology (see, for example,3,30 – 35). Let ni be the number of tRNA isoacceptors that recognize the ith codon, the absolute adaptiveness value of the ith codon is defined in the following equation: Wi ¼ ni X ð1  Sij Þ  tGCNij ð1Þ j¼1 where tGCNij is the gene copy number of the jth tRNA that recognizes the ith codon (a proxy of the tRNA levels24,29,36), and Sij is a selective constraint on the efficiency of the interaction between the ith codon and the jth tRNA, which is scored between 0 ( perfect interaction) and 1 (no interaction);8 specifically, the Sij [Vol. 21, weights can be related to aspects of translation elongation (tRNA, wobble interactions, and properties of the ribosome), as these aspects are expected to affect the efficiency of the codon – anticodon interaction. Wi values are calculated according to Crick’s wobble rules for codon –anticodon pairing (Table 1). The codon relative adaptiveness value wi is obtained by dividing each Wi with the maximum Wi value over all codons.8 The tAI of a gene is defined as the geometric mean of the wi values of its codons. In 1966, Crick37 suggested that in some cases wobble pairing may occur in the third base of the codon. According to Crick, the pairing at the third codon position has to obey chemical constrains; thus, some of the optional parings will not occur. For example, the unlikely pairing of guanine – adenine is due to a side group of guanine, which cannot make one of its bonds. In addition to the four standard nucleotides, modified nucleosides often occupy the wobble position of the anticodon (usually position 34 of the tRNA). In fact, the wobble position is the most frequently modified nucleoside in tRNA.38,39 Inosine, for example, is a common modification of adenine that occurs in the wobble position of many tRNA species.39 – 43 Out of the eight Sij weights, four are related to (...truncated)