The chromosomal organization of horizontal gene transfer in bacteria

Abstract

Bacterial adaptation is accelerated by the acquisition of novel traits through horizontal gene transfer, but the integration of these genes affects genome organization. We found that transferred genes are concentrated in only ~1% of the chromosomal regions (hotspots) in 80 bacterial species. This concentration increases with genome size and with the rate of transfer. Hotspots diversify by rapid gene turnover; their chromosomal distribution depends on local contexts (neighboring core genes), and content in mobile genetic elements. Hotspots concentrate most changes in gene repertoires, reduce the trade-off between genome diversification and organization, and should be treasure troves of strain-specific adaptive genes. Most mobile genetic elements and antibiotic resistance genes are in hotspots, but many hotspots lack recognizable mobile genetic elements and exhibit frequent homologous recombination at flanking core genes. Overrepresentation of hotspots with fewer mobile genetic elements in naturally transformable bacteria suggests that homologous recombination and horizontal gene transfer are tightly linked in genome evolution.Horizontal gene transfer (HGT) is an important mechanism for genome evolution and adaptation in bacteria. Here, Oliveira and colleagues find HGT hotspots comprising ~ 1% of the chromosomal regions in 80 bacterial species.

Fig. 1

Scheme depicting key concepts used in this study. Intervals flanked by the same core gene families (C_X, C_Y) as those from pivot genome A were defined as syntenic intervals (i.e., the members of the core gene families X and Y were also contiguous in the pivot). The intervals that do not satisfy this constraint were classed as breakpoint intervals (green-shaded regions) and excluded from our analysis. For every interval in the pivot genome, we defined spot as the set of intervals flanked by members of the same core gene families (blue-shaded regions)

Fig. 2

Cumulative frequency of HTgenes. a Cumulative distribution of horizontally transferred genes (HTgenes, %) in spots for the 80 bacterial clades. b Histogram of the minimum number of spots needed to attain 50% of the total number of HTgenes (HTg₅₀ index). The average HTg₅₀ was only 1.9% (±0.2; standard deviation)

Fig. 3

Analysis of HTgenes and the abundance and distribution of hotspots. a 16S rRNA phylogenetic tree of the 80 bacterial clades. The tree was drawn using the iTOL server (itol.embl.de/index.shtml). The first column indicates the clade and is colored by phylum. The four subsequent columns correspond respectively to: the average number of HTgenes per genome computed using Count, the number of hotspots, the average Simpson dissimilarity index (β_SIM, accounting for turnover), and the average multiple-site dissimilarity index accounting only for nestedness (β_NES). These values are given in Supplementary Data set 1. b Distribution of the average number of hotspots per clade according to the average genome size (G_S). c Association between the number of hotspots and the number of HTgenes in the clade

Fig. 4

Functional characterization of hotspots. a Observed/expected (O/E) ratios of non-supervised orthologous groups (NOGs, shown as capitalized letters). The first two lines represent the values of HTgenes and accessory genes observed in hotspots when the null model was computed from the distribution of the same type of genes in coldspots. The last line shows the same type of analysis for the core genes flanking hotspots when the null model is computed using the core genes not flanking hotspots. Expected values were obtained by multiplying the number of HTgenes, accessory, or core genes in hotspots by the fraction of genes assigned to each NOG. *P < 0.05; **P < 10⁻²; ***P < 10⁻³, χ²-test. b Percentage of hotspots with antibiotic resistance genes (ARGs, top), and percentage of spots with ARGs that are hotspots (bottom). Note that hotspots are only 1.2% of all the spots

Fig. 5

Genetic mobility of hotspots. We represent the percentage of hotspots containing the different genetic elements (top) and the percentage of spots containing such elements that are hotspots (bottom). Note that hotspots are only 1.2% of all the spots. The analysis includes MGEs (IMEs, ICEs, prophages, integrons), mobility-associated proteins (MAPs) (ISs, TyrRec, SerRec), and tRNA/tmRNA, rRNA. Also, shown in colored bins are the observed/expected (log₁₀O/E) number of hotspots that contain the abovementioned elements, when the null model was computed from the distribution of coldspots containing the same type of elements. Expected values were obtained by multiplying the number of hotspots by the fraction of spots containing each type of element

Fig. 6

Chromosomal context of hotspots. a Number of hotspots containing prophages, ICEs/IMEs, and none of the above along the origin–terminus axis of replication. Linear regression and the confidence limits (shaded area) for the expected value (mean) were indicated for each category. The number of hotspots including prophages increases linearly with the distance to the origin of replication (Spearman’s ρ = 0.87, P < 10⁻³), but this is not the case for the other two categories (both P > 0.05). b Heatmap of odds ratios of co-localizations in hotspots of MGEs, mobility-associated proteins (MAPs) and RNAs. ***P < 10⁻³; **P < 10⁻²; *P < 0.05; Fisher’s exact test

Fig. 7

Genetic diversity of the accessory genes present in hotspots and coldspots. a Examples of gene nestedness and turnover in a spot. Turnover measures the segregation between intervals in terms of gene families, i.e., it accounts for the replacement of some genes by others. Nestedness accounts for differential gene loss and measures how the gene repertoires of some intervals are a subset of the repertoires of the others. It typically reflects a non-random process of gene loss. b Distributions of β diversity (β_SOR) in hotspots and coldspots. c Partition of β_SOR in its components of nestedness (β_NES) and turnover (β_SIM) for hotspots and coldspots (β_SOR = β_NES + β_SIM). ***P < 10⁻³; Mann–Whitney–Wilcoxon test; ns: not significant

Fig. 8

Evidence for more frequent homologous recombination in core genes flanking hotspots than in the other core genes. a Model for the creation and evolution of hotspots by homologous recombination at the flanking core genes. b We detected homologous recombination events in the core genes using ClonalFrameML, and searched for evidence of phylogenetic incongruence (P < 0.05) between each core gene family and the whole core genome tree of the clade using the Shimodaira–Hasegawa (SH) test. The observed–expected ratios (O/E) for these two analyses are significantly higher than one. ***P < 10⁻³; χ²-test. c Differences in nucleotide diversity between core genes flanking hotspots and the others. Nucleotide diversity was calculated using the R package “PopGenome” v2.1.6 by implementation of the diversity.stats() command. ***P < 10⁻³; Mann–Whitney–Wilcoxon test