Pathways for horizontal gene transfer in bacteria revealed by a global map of their plasmids

Abstract

Plasmids can mediate horizontal gene transfer of antibiotic resistance, virulence genes, and other adaptive factors across bacterial populations. Here, we analyze genomic composition and pairwise sequence identity for over 10,000 reference plasmids to obtain a global map of the prokaryotic plasmidome. Plasmids in this map organize into discrete clusters, which we call plasmid taxonomic units (PTUs), with high average nucleotide identity between its members. We identify 83 PTUs in the order Enterobacterales, 28 of them corresponding to previously described archetypes. Furthermore, we develop an automated algorithm for PTU identification, and validate its performance using stochastic blockmodeling. The algorithm reveals a total of 276 PTUs in the bacterial domain. Each PTU exhibits a characteristic host distribution, organized into a six-grade scale (I-VI), ranging from plasmids restricted to a single host species (grade I) to plasmids able to colonize species from different phyla (grade VI). More than 60% of the plasmids in the global map are in groups with host ranges beyond the species barrier.

Fig. 1. Clustering of plasmid genomes along taxonomic boundaries.

All graphs represent AcCNET bipartite networks, with nodes representing plasmid genomes (P) or homologous protein clusters (HPC), extracted from (P) as indicated in Methods. Whenever a P node encodes a member of a given HPC, an edge between both nodes is drawn. Edge lengths depend on the similarity to the most representative member of the cluster, as defined by kClust. a, c–e panels show clustering by the host. Plasmid nodes were colored, as indicated in the respective color codes, by host phylum (a), host class within the Proteobacteria phylum (c), order-level within γ-proteobacteria (d), or genus-level, for the E. coli–Salmonella–Klebsiella group (e). b, f panels show clustering by MOB types in the overall plasmidome network (b) or in the family Enterobacteriaceae (f). Source data are provided as a Source Data file.

Fig. 2. Measuring plasmid similarity by AcCNET and ANI.

a Plasmids of the same taxon tend to be more alike. Violin plots showing the ratio of trajectories between plasmids from different taxa in the AcCNET network of Fig. 1. A trajectory is defined as a connection between a plasmid, a given HPC, and a second plasmid that also contains a member of the HPC. For each plasmid of each taxon considered, the ratio was obtained by dividing the number of trajectories linking plasmids hosted in different taxa (species, families, phyla…) by the total number of trajectories of that plasmid. White dots indicate the median connection ratio of trajectories between plasmids, black bars correspond to the Q1-Q3 interquartile range. b Density plot of ANI_L20 (x-axis) and AF (y-axis) values for pairwise comparisons between plasmids of the order Enterobacterales. The color code indicates the relative density of that particular ANI/AF area. Yellow areas correspond thus to more frequent ANI/AF scores. The red square highlights a high number of comparisons with ANI_L20 > 90% but AF < 10, corresponding to plasmids sharing only a small fragment of their genome with high homology (e.g., transposons or insertion sequences). Separate comparisons for MOB+ and MOB− plasmids with sizes higher and lower than 40 kb are shown in Supplementary Fig. 3. c Density plot of ANI_L50 (x-axis) versus AF (y-axis). Yellow areas correspond to more frequent ANI/AF scores. As shown in the figure, the L50 threshold on the ANI algorithm eliminates the high ANI/low AF comparisons. d Histogram of ANI_L50 scores obtained in comparisons between plasmids of the same PTU (blue) and between plasmids of different PTUs (red).

Fig. 3. ANI_L50 similarity network of the bacterial plasmidome.

Similarity networks of the RefSeq84 prokaryotic plasmidome obtained using the ANI_L50 algorithm as described in Methods. Nodes, corresponding to plasmid genomes, are colored according to their cognate host taxonomy (a) or MOB class, as defined by MOBscan (b). Source data are provided as a Source Data file.

Fig. 4. Conjugative and mobilizable PTUs of the order Enterobacterales.

The figure represents an integrative view of the PID+SBM methods. a, b ANI network of the Enterobacterales plasmidome, with plasmids colored according to their cognate host (a) or the MOB type of each plasmid (b). Source data are provided as a Source Data file. c A list of the most representative mobilizable and conjugative PTUs from Enterobacterales, according to their MOB type. Columns on the left indicate the MOB type and name of the PTU. Numbers in the boxes indicate the number of plasmids isolated from the different genera shown in the figure. Dark green squares indicate high plasmid prevalence, with >15 different plasmids retrieved from that genus. Boxes correspond, respectively, to Escherichia (Eco), Klebsiella (Kle), Salmonella (Sal), Shigella (Shi), Citrobacter (Cit), Serratia (Ser), Yersinia (Yer), Erwinia (Erw), and Proteus/Providencia (Pro). The number below each taxon represents the total number of plasmids from that taxon present in the database. Boxes squared with a black margin represent PTUs exclusively present in that particular host. The blue bar on the right indicates the most prevalent replicon structure, as determined by PlasmidFinder. The rightmost two columns of the figure represent, respectively, the intracluster density, and a prototype plasmid from the group.

Fig. 5. PTUs of different host range delineate pathways for HGT in Enterobacterales.

a The graph shows the host distribution of different PTUs. The distribution breath is the taxonomic depth between the two most distinct hosts in which that PTU is found. The left column shows percentages for nonmobilizable PTUs, while the right column shows data for mobilizable and conjugative PTUs. b Host range scale for mobilizable and conjugative PTUs from Enterobacterales, color-coded as in a. c Chord diagram showing the distribution of PTUs among the bacterial taxa shown in the outer rim of the diagram. Bars below each taxonomic group represent the total number of plasmids present in that taxon. Colored sectors indicate each of the mobilizable/conjugative PTUs defined in Fig. 4b. Inner ring is colored by PTU membership of plasmids. Whenever two bacterial taxa host plasmids from the same PTU, they are linked by an edge. This way, internal chords represent potential routes for plasmid-mediated HGT among different bacterial genera. Members of the Enterobacteriaceae family are gray-shadowed.

Fig. 6. PTUs identified in the entire bacterial plasmidome.

a A colormap of the PTUs identified by PID in the ANI_L50 network of the entire bacterial plasmidome, retrieved from RefSeq84. Plasmids are colored by PTU membership. A comprehensive list of all PTUs and their members can be found in Supplementary Data 2. Source data are provided as a Source Data file. b Characteristics of PTUs found in the six more abundant orders. The leftmost panel indicates the proportion of MOB+ and MOB− PTUs. The panel in the middle indicates the host range distribution (Grade I–VI). The rightmost panel shows the distribution of MOB types.

Fig. 7. HGT pathways in bacteria drawn by PTUs.

a A network showing possible plasmid-mediated pathways for genetic exchanges between different species. Each node in the network represents a bacterial species. The size of the node is proportional to the number of PTUs identified in that species, and the color code corresponds to its cognate taxonomical class or phylum (Cyanobacteria). b The exchange community formed by species from γ and β-proteobacteria. As in the panel above, the size and color of the nodes represent the number of PTUs and the taxonomic class of the species considered, respectively. c The exchange community formed by the orders Bacillales and Lactobacillales.