Association between translation efficiency and horizontal gene transfer within microbial communities

Abstract

Horizontal gene transfer (HGT) is a major force in microbial evolution. Previous studies have suggested that a variety of factors, including restricted recombination and toxicity of foreign gene products, may act as barriers to the successful integration of horizontally transferred genes. This study identifies an additional central barrier to HGT-the lack of co-adaptation between the codon usage of the transferred gene and the tRNA pool of the recipient organism. Analyzing the genomic sequences of more than 190 microorganisms and the HGT events that have occurred between them, we show that the number of genes that were horizontally transferred between organisms is positively correlated with the similarity between their tRNA pools. Those genes that are better adapted to the tRNA pools of the target genomes tend to undergo more frequent HGT. At the community (or environment) level, organisms that share a common ecological niche tend to have similar tRNA pools. These results remain significant after controlling for diverse ecological and evolutionary parameters. Our analysis demonstrates that there are bi-directional associations between the similarity in the tRNA pools of organisms and the number of HGT events occurring between them. Similar tRNA pools between a donor and a host tend to increase the probability that a horizontally acquired gene will become fixed in its new genome. Our results also suggest that frequent HGT may be a homogenizing force that increases the similarity in the tRNA pools of organisms within the same community.

Figure 1.

(A) The number of HGT events (six) as a function of RVtAI across five bins of equal size: a Whisker plot with the means marked with red circles. (B) Correlations with the number of HGT events when controlling for all the other factors; P denotes asymptotic P-value, pe denotes empirical P-value (‘Materials and Methods’ section). (C) Correlation given increasing number of factors (A) estimated mean expression levels, (B) variance in GC content and (C) variance in amino acid bias.

Figure 2.

(A and B) The tRs values of organisms that share more genes are more similar. (A) Correlation between number of shared genes and tRs for different cutoffs of gene sharing (1). (B) Whisker plot (five bins of equal size) of tRs versus number of shared genes (cut-off of 70%; five bins equal size). (C) Spearman correlations of various variables with the number of shared genes when controlling for all the other factors; P denotes asymptotic P-value, pe denotes empirical P-value (‘Materials and Methods’ section). (D) Correlation given an increasing number of factors (A) phylogenetic distance, (B) sum of genome sizes, (C) community co-membership, (D) selection for translation efficiency, (E) similarity in the GC content, (F) similarity in the amino acid bias, (G) similarity in the growth rate and (H) difference in genome sizes; most of the decrease in correlation is due to genome sizes, phylogenetic distance and similarity in amino acid bias. (E) Whisker plot of the tRs for pairs of ancestral organisms versus the mean number of ancestral HGT between them (x-axis) for five bins of equal size.

Figure 3.

(A) The tRs of organisms that live in the same community [(12), left] are higher than the tRs of organisms from different communities (right). The y-axis is the mean tRs of the organisms in each group; we used the Kolmogorov–Smirnov (KS)-test for computing P-values. (B) Non-parametric correlations of several variables with niche-sharing, when controlling for all the other variables. p denotes asymptotic P-value, pe denotes empirical P-value (‘Materials and Methods’ section). (C) Correlation given increasing number of factors (A) sum genome of sizes, (B) phylogenetic distance, (C) similarity in the amino acid usage, (D) selection for translation efficiency, (E) similarity in the GC content, (F) similarity in the growth rates and (G) similarity in genome sizes.

Figure 4.

(A) Correlation between fitness (growth rate in OD) and the tAI, across a set of GFPs with different codon bias that have been expressed in E. coli [based on data from (7)]. (B) HGT events from different source organisms to E. coli: the percent of genes that are non-transferable to E. coli versus the corresponding tRs between the source organism and E. coli [based on data of Sorek et al. (2)]. (C) Gene expression of recently transferred genes and endogenous genes. The red lines denote the mean expression level in each group, the purple lines mark a threshold denoting high expression level [log(mRNA levels) >2]; 11% of endogenous genes (that did not undergo recent HGT) are highly expressed and 6% of the genes that did undergo recent HGT are highly expressed. (D) The correlation between tAI and gene expression for the transferred genes, endogenous genes, and for all genes (blue) versus the correlation that is gained in each of these cases after optimizing the tRNA pool (brown); the optimal correlation when considering all the gene is closer to the actual one (a ‘difference’ of 0.03 between the optimal and actual correlations) than the correlation when considering only the non-transferred genes (a ‘difference’ of 0.04 between the optimal and actual correlations, 33% higher). (E) The distribution of tAI for endogenous genes (upper part: mean tAI is 0.25—the red line) and recently transferred genes (lower part: mean tAI is 0.22—the red line). (F) A schematic illustration of the possible bidirectional relation between HGT and similarity in the tRNA pools.