A Role for tRNA Modifications in Genome Structure and Codon Usage

Presentation on theme: "A Role for tRNA Modifications in Genome Structure and Codon Usage"— Presentation transcript:

1 A Role for tRNA Modifications in Genome Structure and Codon Usage
Eva Maria Novoa, Mariana Pavon-Eternod, Tao Pan, Lluís Ribas de Pouplana  Cell  Volume 149, Issue 1, Pages (March 2012) DOI: /j.cell Copyright © 2012 Elsevier Inc. Terms and Conditions

2 Cell  , DOI: ( /j.cell ) Copyright © 2012 Elsevier Inc. Terms and Conditions

3 Figure 1 Genome Phylogeny Based on tRNA Gene Content
(A) Distance-based phylogeny based on tRNA gene content, performed with equal number of species of each kingdom. The four phylogenetic clusters have been labeled accordingly. The phylogeny performed with the whole set of 527 species is consistent with these results (see Figure S1). (B) Diagram showing the increase in tRNA population complexity in the four main phylogenetic clusters found in this work (each tRNA is designated by its anticodon sequence). Each base at the wobble position is colored according to its chemical nature. Anticodons labeled with an asterisk (CGU, CAC, CCU) correspond to tRNA genes that are not found in all species comprising the ML-Archaea clade. Cell  , DOI: ( /j.cell ) Copyright © 2012 Elsevier Inc. Terms and Conditions

4 Figure 2 Unequal Enrichment of tRNA Isoacceptors Is Kingdom Specific
Mean tRNA abundances in the four phylogenetic clusters identified by gene content analysis: (1) Methanococcus-like Archaea, (2) non-Methanococcus-like Archaea, (3) Bacteria, and (4) Eukarya. Each tRNA anticodon is colored according to its average number of encoding tRNA genes. To deal with exceptional cases such as Ferroplasma acidarmanus, which is the sole archaea with a tRNALeu(AAG) gene (Marck and Grosjean, 2002), we have considered as absent those tRNA isoacceptors whose average tRNA gene copy number is between 0 and 0.05 (shown in yellow). Cell  , DOI: ( /j.cell ) Copyright © 2012 Elsevier Inc. Terms and Conditions

5 Figure 3 Identification and Quantification of Overrepresented tRNA Isoacceptors (A) Biplot of the scores after performing Principal Component Analysis (PCA). Archaea (red), Bacteria (purple) and Eukarya (green) are distinguishable clusters using this analysis. The archaeal outliers correspond to Methanococcus species, which were already identified as a separate cluster using the tRNA gene content analysis. (B) Biplot of the loadings, indicating the tRNA isoacceptors whose frequencies contribute the most to each of the clusters. Each anticodon has been colored depending on its wobble base. The ellipses surround those anticodons that are significantly associated to the PCs, either with PC1 negative values, which correspond to Bacteria (purple), or with PC2 negative values, which correspond to Eukarya (green) (see Table S1 for the individual correlation values). See also Figure S2 and Table S2. (C) Genome phylogeny based on tRNA-gene content. The distributions of the two wobble base modification enzymes that act upon the tRNA isoacceptors identified in the PCA are shown. Uridine methyltransferases (UMs, labeled in red) are exclusively distributed across the bacterial kingdom. Heterodimeric adenosine deaminases (ADATs, labeled in green) are exclusively distributed in eukaryotes. Homodimeric forms of ADATs (TadA) are found in bacteria, but they only increase the decoding capacity of tRNAArg, and for simplicity, are not shown in the phylogeny. Cell  , DOI: ( /j.cell ) Copyright © 2012 Elsevier Inc. Terms and Conditions

6 Figure 4 Match between Most Adapted Codons and Most Abundant Codons
(A) The match between the highest RSCU codon (green, most abundant codons) and the RGF value of its decoding tRNA (red, most adapted codons) is shown, for each kingdom, in the left column. The match after correcting the RGF values to account for the activity of UMs and ADATs is shown in the middle column. Archaea present neither ADATs nor UMs, and therefore the middle column is missing for this kingdom. The increase in the match score between RSCU and RGF after the correction is shown for each kingdom in the right histogram (except for Archaea). (B) Correlation between human tRNAArg isoacceptor abundance determined using tRNA microarrays and codon usage of ribosomal proteins (shown as RSCU), both for HeLa and HEK293T cell lines. The lack of correlation between these two parameters in the left plot is corrected in the right plot by the inclusion of the activity of ADATs. See also Figure S4 and Tables S3–S6. Cell  , DOI: ( /j.cell ) Copyright © 2012 Elsevier Inc. Terms and Conditions

7 Figure 5 Correlation between Preferred Codons and Protein Abundance
In both E. coli and S. cerevisiae, the abundance of preferred codons in a gene correlates with protein abundance (Spearman correlation: 0.44 and 0.70, with p values of 9.7e-20 and 5.1e-52, respectively). Complementarily, the frequency of nonpreferred codons in genes decreases proportionally to protein abundance. The local density of data points in the graph is signified by their color (darker corresponding to more populated areas of the plot). See also Figure S5. Cell  , DOI: ( /j.cell ) Copyright © 2012 Elsevier Inc. Terms and Conditions

8 Figure 6 Model for the Role of Modification Enzymes in the Evolution of Genome Compositions The emergence of the two tRNA modification enzymes (heterodimeric ADATs and UMs) was the main factor causing the divergence of decoding strategies between kingdoms. Archaea represents the most ancestral decoding strategy, where all isoacceptors are equally represented (and ANN anticodons are missing). ANN anticodons became overrepresented in eukaryotes due to the emergence of heterodimeric ADATs. Similarly, UNN anticodons became overrepresented in bacteria due to the appearance of UMs. Modification of the wobble position increased the decoding capacity of tRNAs, and consequently, translation efficiency. Thus, modifiable tRNAs were positively selected, causing a bias in tRNA gene content distribution which, in turn, caused the codon usage bias characteristic of the three main kingdoms. Cell  , DOI: ( /j.cell ) Copyright © 2012 Elsevier Inc. Terms and Conditions

9 Figure S1 Genome Phylogeny of the 527 Species Based on tRNA Gene Content, Related to Figure 1 Each identified phylogenetic cluster has been labeled accordingly, and is shown in green (Eukarya), black (Bacteria), red (ML-Archaea) and blue (NML-Archaea). Cell  , DOI: ( /j.cell ) Copyright © 2012 Elsevier Inc. Terms and Conditions

10 Figure S2 Identification and Quantification of Overrepresented tRNA Isoacceptors, Related to Figure 3 (A) Biplot of the scores after performing Principal Component Analysis. The left plot shows the PC1 versus PC2 loadings, the middle plot shows PC1 versus PC3, and the right plot shows PC2 versus PC3. The species have been colored according to their kingdom: Archaea (red), Bacteria (purple) and Eukarya (green). (B) Biplot of the loadings, indicating the tRNA isoacceptors whose frequencies contribute most to each of the clusters. The tRNA isoacceptors that are significantly associated to the PCs are circled, and colored according to the kingdom in which they are enriched: Eukarya (green) and Bacteria (purple). Cell  , DOI: ( /j.cell ) Copyright © 2012 Elsevier Inc. Terms and Conditions

11 Figure S3 Codon-Anticodon Recognition at the Wobble Position, Related to Table 1 Representation of all possible codon:anticodon pairings according to the extended wobble base pairing rules. The decoding capacity of both xo5U and I is increased in comparison to the other bases that can found at the wobble position of the anticodon. Cell  , DOI: ( /j.cell ) Copyright © 2012 Elsevier Inc. Terms and Conditions

12 Figure S4 Correlation between Codon Usage and tRNA Gene Copy Number, Related to Figure 4 (A) For each amino acid and each kingdom, the highest adapted codon, i.e., that corresponding to the highest tRNA gene copy number, has been computed and compared to the codon with highest RSCU. Correlation results have been presented as heat map, either as match (blue) or mismatch (orange-red). Interestingly, the majority of nonmatching codons are precisely those susceptible of being recognized by modified tRNAs. When taking into account the wobbling (right) facilitated by uridine methyltransferases (UMs) or adenosine deaminases (ADATs) we can observe that most nonmatching codons are now matching. (B) Correlation between the relative abundance of “nonpreferred” codons and GFP fluorescence values (Pearson's r: −0.51, p-value = 2.5e-11; Spearman's rho: −0.50, p-value = 6.2e-11). The local density of data points in the graph is signified by color (darker corresponding to more populated areas of the plot). (C) Correlation between the most adapted codon (highest RGF) and the most abundant codon (highest RSCU) for low and high expressed GFP sequences. Low expressed and high expressed GFP sequences were chosen as those having the top or bottom 5% fluorescence among the 154 GFP set. In the left, the direct correlation between most adapted codon and most frequent codon is shown; in the middle graph the increase in correlation due to the incorporation of UMs is shown. The histograms represent the percentage of match between RGF and RSCU with and without the inclusion of the activity of UMs in the calculation of RGF values. Arginine codons have been excluded from the analysis because all synthetic GFP sequences present extremely high RSCU values for tRNAArg(AGA) (ranging from 1.71 to 4.29), and therefore cannot be used to measure codon preferences or correlations. Similar results were obtained when using larger datasets (i.e., the top and bottom 10% fluorescence values in the GFP set). Cell  , DOI: ( /j.cell ) Copyright © 2012 Elsevier Inc. Terms and Conditions

13 Figure S5 Correlations between Codon Usage and Protein Expression Levels, Related to Figure 5 (A) Correlations between the most adapted codons (highest RGF) and protein expression levels without considering modifications are shown both for E. coli and S. cerevisiae (Spearman's rho: 0.20 and 0.55, respectively). Similarly, the correlations between the least adapted codons (all those not included before) and protein expression levels are shown. Only amino acids with modifiable tRNA isoacceptors have been considered in this data to make them comparable with those results shown in Figure 5. The significance of the differences between Figure 5 (with modifications) and this figure is p = 9.4e-5 for E. coli, and p = 6.3e-4 for S. cerevisiae data, respectively. (B) Correlation between the number of nonpreferred codons and protein abundance using diverse sources of experimentally determined protein abundances. For E. coli, the Spearman correlations are −0.48 (p-value = 8.5e-24) and −0.27 (p value = 6.5e-10) using experimental data from Lu et al. (2007) and Ishihama et al. (2008), respectively. For S. cerevisiae, the Spearman correlations are −0.76 (p value = 4.0e-68) and −0.74 (p-value = 1.2e-69) using data obtained by Lu et al. (2005) and Newman et al. (2006), respectively. (C) Correlation between “nonpreferred” codons and translation efficiency. The number of non-UM-preferred and non-ADAT-preferred codons has been computed for each of the genes whose mRNA and protein abundance data was available (Lu et al., 2007), both for E. coli and S. cerevisiae, respectively. Translation efficiency has been defined as the ratio between mRNA and protein abundance levels. Cell  , DOI: ( /j.cell ) Copyright © 2012 Elsevier Inc. Terms and Conditions