The interplay of restriction-modification systems with mobile genetic elements and their prokaryotic hosts

Abstract

The roles of restriction-modification (R-M) systems in providing immunity against horizontal gene transfer (HGT) and in stabilizing mobile genetic elements (MGEs) have been much debated. However, few studies have precisely addressed the distribution of these systems in light of HGT, its mechanisms and its vectors. We analyzed the distribution of R-M systems in 2261 prokaryote genomes and found their frequency to be strongly dependent on the presence of MGEs, CRISPR-Cas systems, integrons and natural transformation. Yet R-M systems are rare in plasmids, in prophages and nearly absent from other phages. Their abundance depends on genome size for small genomes where it relates with HGT but saturates at two occurrences per genome. Chromosomal R-M systems might evolve under cycles of purifying and relaxed selection, where sequence conservation depends on the biochemical activity and complexity of the system and total gene loss is frequent. Surprisingly, analysis of 43 pan-genomes suggests that solitary R-M genes rarely arise from the degradation of R-M systems. Solitary genes are transferred by large MGEs, whereas complete systems are more frequently transferred autonomously or in small MGEs. Our results suggest means of testing the roles for R-M systems and their associations with MGEs.

Figure 1.

Quantification and distribution of R-M systems in 2261 prokaryotic genomes. (A) Amount of Types I, II, IIC, III R-M systems and Type IV REases found in genomes. Corresponding percentages are indicated. (B) Average R-M density (per genome per Mb) according to clade. The largest peak on R-M density observed for Epsilonproteobacteria (^a) results from the presence of multiple systems particularly among Helicobacter species. For comparison, we also show the density for Epsilonproteobacteria without Helicobacter (^b). Only clades with at least 10 different species were considered for comparison. The number of species within each clade is indicated next to its name. (C) Distribution of the average number of R-M systems per genome (upper graph) and average density per genome per Mb (bottom graph) according to genome size (Mb). Stippled line separates the regions having small and large genomes. Genomes of Helicobacter were not included to avoid obtaining extremely inflated values in the [1.5–2.2[ Mb genome size range.

Figure 2.

Box plots of the co-occurrence of R-M systems and natural competence machinery in small (<2 Mb) and large (≥2 Mb) genomes. Error bars represent standard deviations. Mann–Whitney–Wilcoxon test P value is indicated next to the box plots.

Figure 3.

Quantification and distribution of R-M systems in MGEs. (A) Amount of Types I, II, IIC, III R-M systems and Type IV REases found in plasmids. The latter were classified according to their transmissibility: plasmids encoding the entire conjugation machinery or at least the relaxase (MOB⁺, shown as +), and plasmids lacking even the relaxase (MOB⁻, shown as −). (B) Observed/expected (O/E) ratios of R-M systems in plasmids, prophages and ICEs/IMEs. Expected values were obtained by multiplying the total number of each type of R-M system by the fraction of R-M systems assigned to each MGE. (C) Co-occurrence of R-M systems and MGEs. Box plots of the genomic co-occurrence of R-M systems with plasmids, prophages, ICEs/IMEs and integrons in small (<2 Mb) and large (≥2 Mb) genomes. Error bars represent standard deviations. Mann–Whitney–Wilcoxon test P values are indicated.

Figure 4.

Associations between R-M and other defense systems. (A) Contextual analysis of R-M systems. Observed/expected (O/E) ratios of co-localized R-M systems (all types) and individual R-M types. Expected values were obtained by multiplying the total number of each type of R-M system by the fraction of R-M systems assigned to each MGE. (B) Box plots representing the co-occurrence of R-M and CRISPR-Cas systems in small (<2 Mb) and large (≥2 Mb) genomes. Error bars represent standard deviations. Mann–Whitney–Wilcoxon test P values are indicated.

Figure 5.

Evolution of R-M systems. Variation in dN/dS between REases, MTases and Type IIC systems. Error bars represent standard deviations. Significance was determined by computing Mann–Whitney–Wilcoxon test P values. For the sake of simplicity, we only show intra-REase and intra-MTase statistics. All remaining pairwise statistical comparisons are significant at P < 10⁻³ with the exception of Type I REases-Type III MTases (P < 0.05) and Type III REases-Type III MTases (P < 10⁻²). ***P < 10⁻³.

Figure 6.

Comparative analysis of complete R-M systems and solitary components. (A) Median size of regions (L, expressed in the number of genes) harboring complete volatile R-M systems versus solitary volatile R-M elements (indicated as X). Stippled line corresponds to the identity. (B) The number of solitary REases and MTases in phages. Over 90% of the total hits were found to correspond to solitary MTases. (C) Median plasmid size (kb) for plasmids containing only complete R-M systems, solitary components or both. Mann–Whitney–Wilcoxon test P value is indicated next to the box plots.