Diversity and evolution of an abundant ICEclc family of integrative and conjugative elements in Pseudomonas aeruginosa

ABSTRACT

Integrative and conjugative elements (ICEs) are widespread autonomous mobile DNA elements, containing the genes necessary for their excision, conjugative transfer, and insertion into a new host cell. ICEs can carry additional genes that are non-essential for their transfer but can confer adaptive phenotypes to the host. Our aim here was to better characterize the presence, distribution, and variation of ICEs related to the well-described ICEclc among Pseudomonas aeruginosa clinical isolates within a geographically restrained environment to understand the factors contributing to their evolution. We examined a total of 181 P. aeruginosa genome sequences obtained from patient or hospital environment isolates, most of which were obtained from a single hospital during 20 years of sampling. More than 90% of the isolates carried one or more ICEclc-like elements, with different degrees of conservation to the known ICEclc lifestyle and transfer genes. ICE clones closely matched their host clonal phylogeny, but not exclusively, indicating that both clonal evolution and ICE horizontal transfer are occurring in the hospital environment. ICEs from this singular hospital environment were mainly associated to three clone types found worldwide, suggesting an enrichment of local clones. Variable gene regions among the clinical P. aeruginosa ICEclc-type elements were notably enriched for heavy metal resistance genes, toxin-anti-toxin systems, potential efflux systems and multidrug resistance proteins, a metalloprotease and for a variety of regulatory systems, but not for specific recognizable antibiotic-resistance cassettes. Clonal persistence suggests adaptive benefits of these functional categories, and micro-patterns of gene gain and loss indicate ongoing ICE evolution within the P. aeruginosa hosts.

IMPORTANCE

Microbial populations swiftly adapt to changing environments through horizontal gene transfer. While the mechanisms of gene transfer are well known, the impact of environmental conditions on the selection of transferred gene functions remains less clear. We investigated ICEs, specifically the ICEclc-type, in Pseudomonas aeruginosa clinical isolates. Our findings revealed co-evolution between ICEs and their hosts, with ICE transfers occurring within strains. Gene functions carried by ICEs are positively selected, including potential virulence factors and heavy metal resistance. Comparison to publicly available P. aeruginosa genomes unveiled widespread antibiotic-resistance determinants within ICEclc clades. Thus, the ubiquitous ICEclc family significantly contributes to P. aeruginosa’s adaptation and fitness in diverse environments.

INTRODUCTION

Pseudomonas aeruginosa is a ubiquitous bacterium found in a diversity of environments, from soil to water, and can act as an opportunistic pathogen of a wide range of hosts including humans (1). P. aeruginosa isolates are well known as a major cause of infections in the cystic fibrosis lung (2) and are further responsible for the occurrence of nosocomial diseases, especially in intensive care units (3). The remarkable capability of P. aeruginosa to colonize a wide range of environments and hosts is reflected in its large genome with a high proportion of regulatory genes (4). Additionally, genomes of P. aeruginosa are mosaics of conserved core parts interspersed with highly variable accessory elements (5–8). The core genome, defined as the set of genes that are present in nearly all strains of P. aeruginosa, has an average inter-strain diversity of 0.5%–0.7% and makes up some 90% of the total genome of each strain (6, 9–11). The accessory genome consists of extrachromosomal elements like plasmids, as well as segments of DNA variable in size and content that are inserted into the chromosome at various loci (12–14). The most important process contributing to the evolution of the P. aeruginosa accessory genome is thought to be horizontal gene transfer (14).

The chromosomal loci where accessory genomic segments are frequently integrated have been called “regions of genome plasticity” (15). Some of these variable regions occur nearby genes coding for tRNAs and bear signatures of integrative and conjugative elements (ICEs) (14). ICEs are widespread mobile DNA elements in bacteria (16–20) that remain by integration in the host chromosome but can excise to form an autonomous molecule that can be conjugated to and inserted into the chromosome of new recipients (21). The notion of wide ICE distribution originated from a study that quantified the abundance and diversity of conjugative systems among prokaryotes, at that time covering 1,124 complete genomes (16). Notably, that study showed that all bacterial clades for which a significant number of sequenced genomes were available encode conjugative systems in their chromosomes. Classification of conserved conjugative marker genes suggested important abundance of ICEs, as well as conjugative plasmids (16). ICEs not only encode all the genes necessary for their integration, excision, and the conjugative machinery but also contain “cargo” genes that can confer adaptive phenotypes to the host (22). Known examples include antibiotic or heavy metal resistance, or the ability to metabolize specific or xenobiotic carbon sources (23–25).

Previously characterized ICEs in P. aeruginosa include elements related not only to ICEclc (of the taxonomic relative Pseudomonas knackmussii B13) (26) but also to PAGI-2 (in strain C), PAGI-3 (in strain SG17M), LESGI-3 (in strain LESB58) (5, 14), and PAPI-1 (in strain PA14) (8). The described P. aeruginosa ICEs share a set of highly conserved genes, while the primary difference between them is their cargo genes (12, 27). The ICEclc family is a loosely defined group of elements detected among Beta- and Gammaproteobacterial genomes (17, 28) with ICEclc as the main studied representative. ICEclc is 103 kb in size, occurs in two identical copies in the P. knackmussii B13 genome (28), and is characterized by the presence of the clc genes for chlorocatechol degradation (25). The element can transfer readily from strain B13 to P. aeruginosa (29) and to other recipients belonging to Beta- and Gammaproteobacteria lineages (30, 31). In addition to the clc genes themselves, ICEclc further carries genes for 2-aminophenol degradation (25), for a multidrug efflux pump (32), as well as for additional proteins of unknown function. A 50-kb region on ICEclc, with high sequence similarity to other (suspected) ICEs (25, 28, 33), contains the genes for ICE activation (33, 34), excision and integration (35), DNA processing (36) and conjugative transfer (37). A wide variety of ICEclc family elements have been identified in different bacterial hosts, and more recent surveys point to several instances of relatives carrying antibiotic-resistance determinants in clinical isolates (26, 38–43). In particular, the studies by Botelho et al. (38) and Hong et al. (42) showed ICEs related to ICEclc encoding carbapenem resistance in P. aeruginosa. These studies thus indicated that ICEs of the ICEclc family appear in various hosts and geographical locations worldwide and with different (adaptive) gene content. However, since they largely occur in very different isolates, it is difficult to understand ongoing ICE evolution and adaptation.

The main objective of our work was to study the micro-evolution of ICEs related to the ICEclc family over time within a coherent environment and to find potential evidence for enrichment of adaptive gene functions and selection. For this, we focused on clinical strains of P. aeruginosa isolated from patients and the hospital environment itself, across a period of 20 years. We included mainly isolates from a single hospital (Lausanne University Hospital) but also from nearby hospitals, with the idea that regional transmission might be detectable (Geneva, Montpellier, and Besançon). Genome sequences of 181 isolates were assembled, among which we searched for ICEs of the ICEclc family. On a smaller subset of non-redundant ICEs, we delineated the potential ICE boundaries wherever possible and analyzed conserved and variable (adaptive) gene contents within defined ICE regions. ICE and host phylogenies were compared from all or from single-copy orthologous genes, in order to delineate potential vertical and horizontal ICE transmission events. Local ICEs were then further phylogenetically compared to a set of recently identified ICEs among P. aeruginosa isolates in a worldwide survey (44). Selection for ICE adaptive functions was analyzed by comparison to random models, suggesting significant enrichment of a variety of gene functions in the local ICE pool, but (so far) not of specific antibiotic-resistance genes, although these are circulating in the worldwide pool.

RESULTS

Pseudomonas aeruginosa clinical isolates contain an abundance of ICEclc-like elements

To describe the micro-evolution of ICEs and, in particular, of ICEclc-type elements, we analyzed 181 P. aeruginosa genomes, obtained from strains isolated between 2003 and 2022 in a geographically constrained area. Most isolates originated from the Lausanne University Hospital (CHUV), further complemented with isolates from the Geneva University Hospital (HUG), University Hospital of Montpellier (CHU Montpellier), and Besançon Regional University Hospital Center (CHU Besançon), to survey potential regional transmission. The strains were isolated either from the hospital environment (e.g., contaminated hand soaps and door handles) or from infected patients (Table S1). Genomes were constructed by de novo assembly followed by reference-guided contigs scaffolding and patching (deploying the reference P. aeruginosa strain H26027 genome; National Center for Biotechnology Information (NCBI) accession number CP033684). The scaffolds have on average 69 gaps (Table S1). Strain genotyping by multilocus sequence typing (MLST, Table S1) showed that some of the analyzed strains belong to the epidemic high-risk clones (namely, ST395 and ST111) (45), which are characterized by frequent integron- or transposon-mediated acquisition of antibiotic-resistance elements (46). As expected, some MLST types are overrepresented in our data set (Table S1), and the strain collection represents a restricted set of the global distribution of sequence types (44).

First, we tried to obtain a broad overview of the presence of ICEs related to ICEclc. For this, we focused on 40 ICEclc genes, which we expected might be conserved: the intB13 integrase, the traI relaxase, the DNA topoisomerase topB, 13 open reading frames (ORFs) whose products are involved in regulation and activation of the element (33), and 24 ORFs whose products are implicated in the type IV ICE conjugative system (37). BLASTN searches with these queries and a 70% sequence identity cut-off against the 181 P. aeruginosa draft genomes returned hits to intB13 for 91% of the genomes (74%–91% nucleotide identity over 63%–94% of the gene length, Fig. 1A). Additionally, 92% of the isolates produced hits to traI (81%–85% nucleotide identity across 54%–97% of the gene length, Fig. 1A). Other ICEclc genes were, in contrast, less conserved among the P. aeruginosa genomes. For instance, among the regulation genes, we found hits for 91% of the genomes to bisD, but only 6% to bisR (Fig. 1A), and none of the isolates carried analogs of the ICEclc regulatory genes mfsR, marR, or tciR (47). A large proportion of the suspected ICE regions in the P. aeruginosa genomes carried homologs of parA. ParA-related proteins were previously found to be important for maintenance of the extrachromosomal circular form of the pathogenicity island PAPI-1 of P. aeruginosa PA14 (48) and causative of growth impairment linked to the presence of ICEclc in Pseudomonas putida (49). Finally, some genomes only carried a few ICEclc-related genes such as inrR, iceD4, and/or orf55476 analogs but no other detectable ICEclc gene homologies (Fig. 1A). These may thus not consist of bona fide complete ICE regions but of spurious hits. These results suggested ICEclc-like elements to be widely distributed among the P. aeruginosa isolates.

Fig 1.

ICEclc-like elements are abundant among P. aeruginosa clinical isolates. (A) BLASTN similarities (as fraction of 1 = 100%, blue hue color legend) for 40 ICEclc genes with 181 P. aeruginosa draft genomes (single top hit per ICEclc per genome, similarity over full gene lengths). ICEclc queries included intB13 integrase, traI relaxase, DNA topoisomerase topB, 13 ORFs of the regulation module, and 24 ORFs whose products form the type IV secretion system (ORF or gene names on top, horizontally). Each row corresponds to one genome. Genome keys are omitted for simplicity. Gray boxes indicate that no hits were found with default BLASTN parameters. (B) Example gene synteny comparisons of ICEclc with two recovered P. aeruginosa ICEs (ICE_19TCP and ICE_38987, extended list shown in Fig. S2). ORFs and their orientation are represented by blue boxes on top or bottom strand (reverse orientation). Known or genes with clear annotation are indicated (simplified per operon, e.g., ars for arsenic resistance operon). Bars highlight regions of homology among the three ICEs; darker hues indicating higher similarity (in BLASTN). Homology regions correspond to what is referred to as core of the ICEs. (C) Number of P. aeruginosa genomes with BLASTN hits above default thresholds to intB13, traI, iceB4, or bisD from ICEclc as proxies for the number of ICEclc-like elements per draft genome. (D) Multiple ICEclc-like elements per genome, shown here as regions with BLASTN similarities above default thresholds to ICEclc for genomes of P. aeruginosa strains 35969 and 17090.

To better delineate the continuity and gene content of the detected putative ICEs, we focused on a smaller set of sequences by excluding obvious redundancy. This was judged from similarity of a 200-kb region around the position of the identified traI gene(s) and the provenance of the strain (Table S1). If two or more 200-kb regions were the same between strains isolated from the same source, we assumed they originated from the same P. aeruginosa clone, and only one was taken for further analysis. This resulted in 85 regions with potential ICEs. Next, we manually searched for the ICE “boundaries,” consisting of regions carrying an expected 18-bp repeat sequence similar to that within the identified ICEclc attachment (att) sites (Table S2). We expected the att sites to be identical or closely related to the att sites previously identified in ICEclc as the newly identified putative ICEs share a closely related integrase to intB13 (35). Eighteen ICEs had identical direct repeats as in ICEclc (e.g., ICE_13520; both attR and attL contain the 18-bp “GTCTCGTTTCCCGCTCCA”; Table S2), whereas 50 showed one nucleotide mismatch in the putative attR repeat (but with attL still being identical to ICEclc, e.g., ICE_13524; Table S2). The att repeat sequences of these ICEs correspond to the 3′-end 18 bp of the second, third, and fourth copies of the tRNA^Gly gene in the dnaA-aligned reference genome (Fig. S1). One element (ICE_17090) contained a different set of direct repeats and was integrated into the first copy of the five tRNA^Gly genes on the P. aeruginosa genome (Fig. S1). The remaining 16 ICE regions (e.g., ICE_27983 and ICE_34204) have contig gaps at their end or beginning and could not be delineated further to include the 18-bp repeat sequences; the assembly gaps were used as fictive boundaries, and we can only make a minimum conservative size estimate (Table S2). Assuming the 18-bp repeats or contig gaps as ICE boundaries, their sizes range between 57 and 133 kb (Table S2). All except two identified ICEs encode a single relaxase (TraI) belonging to the MOB_H family (Table S2) (50). The other two ICEs (ICE_38702 and ICE_36935) encode one relaxase of the MOB_H family and a second one belonging to the MOB_P family. BLASTN comparisons of the presumed P. aeruginosa ICEs against ICEclc indicated that the gene synteny of the conserved region is broadly maintained, and most of the variable genes are located in between the integrase gene and traI (examples shown in Fig. 1B; Fig. S2).

To better understand if individual hosts could carry multiple ICEs of the same or of different types, we selected four ICEclc genes (intB13, traI, iceB4, and bisD) and used the number of BLASTN hits as a proxy for the number of ICEclc-like ICEs per P. aeruginosa genome (Fig. 1C). Assuming that one set of these four genes corresponds to one element, most genomes carried either one or two ICEclc-type ICEs, while few had three or none at all (Fig. 1C). To obtain a broader picture on the presence of other ICE types than ICEclc, we scanned a single representative of each of 21 different P. aeruginosa “clones” (see below) for ICEs and integrative and mobilizable elements (IMEs) using ICEfinder (https://bioinfo-mml.sjtu.edu.cn/ICEfinder/ICEfinder.html). IMEs are non-self-transmissible elements; instead, they rely on the presence of a conjugative element to mediate their horizontal transfer (51). On average, each genome carried 2.8 ICEs (between 1 and 5) and 0.67 IMEs (between 0 and 2) (Table S3), of which one ICEclc-like element and others belonged to a different category. Four of the 21 genomes carried two distinct ICEclc-like elements (Table S3). This suggests that the number of ICEclc-like elements found by single gene comparisons might be an overestimate and resulting from hits to ICEs more distantly related to ICEclc (as in Fig. 1D). Some isolates were found to carry two ICEclc-type elements in tandem, as in the case of strain 35969 (Fig. 1D), whereas others contained ICEclc-like elements in different loci of their genome (strain 17090, Fig. 1D). In addition, these strains showed presence of a phylogenetically more distant ICE, of which only certain ICEclc gene homologs were picked up above the BLASTN thresholds (70% sequence identity, Fig. 1D). To find further similarity between the 85 putative ICEs and other known ICE-like elements, we compared them by BLASTN against the previously characterized P. aeruginosa genomic islands PAGI-2, PAGI-3, PAPI-1, and LESGI-3 (5, 14, 48). We then filtered the results to keep only the hits with a query coverage higher than 60%. This resulted in 53 elements having 84%–97% nucleotide identity to LESGI-3, limited to a region of 61%–80% of their total length (Table S2). Accordingly, 54 elements had 84%–93% identity to PAGI-2, limited to a region of 61%–83% of their total length, while ICE_36935 is 100% identical to PAGI-2 over its whole length (Table S2). None of the identified ICEs had extensive homology to PAGI-3 or PAPI-1 elements (maximum coverage detected by BLASTN: 23% and 6%, respectively, of total ICE length).

Circulating ICE clones are phylogenetically congruent with their P. aeruginosa host

To assess the phylogenetic relationship among the 85 representative ICEs and between their hosts, we compared maximum-likelihood phylogenetic trees of both. The ICE tree was constructed from a multiple sequence alignment on the complete sequence of each element and bootstrapped (n = 1,000) to find majority branches (Fig. 2A). As outgroups, we added ICEs from two P. aeruginosa strains selected from the study of Botelho et al. (38). These ICEclc-like elements encode carbapenemases (bla_DIM-1 in IOMTU133, GenBank accession number GCA_001548335; and bla_NDM-1 and bla_PME-1 in SP4371, GenBank accession number GCA_003950255). The tree topology suggested there are eight major circulating clones (labeled a–h in Fig. 2A), plus several more distinct ICEs, with as low as 60% similarity (in terms of percentage of nucleotide identity, Fig. 2B). Six of the major ICE groups were found in similarly clustered host genome phylogeny, which was constructed from alignments of all the single-copy core orthologous genes (n = 4,260 genes; clusters a, b, d, e, f, and g; Fig. 2A). This suggests clonal divergence of those ICEs with their host genomes. The host phylogeny suggests that most of the isolates are clonal. Indeed, their nucleotide diversity is extremely low or inexistent (in terms of average nucleotide identity, Fig. 2C). In contrast, notably ICE groups c and h find themselves in different P. aeruginosa host lineages (a, b, and e, respectively; Fig. 2A), whereas additional distinct individual ICEs are found within clonal host lineages (as highlighted with brown labels and colored or gray lines in Fig. 2A). For example, P. aeruginosa strain 27167 is clonal to strains 28832, STH1, LTZ7, and LTZ1 (group f) but carries an ICE that is divergent from the ones carried by the other four strains (ICE group f). This strongly suggests that, in addition to clonal divergence, there is horizontal transfer of the ICEs in the hospital environment.

Fig 2.

Phylogenetic comparative analysis of P. aeruginosa hosts and their ICEclc family members. (A) Bootstrapped (n = 1,000) consensus phylogenetic trees of 85 representative P. aeruginosa genomes and ICEs from the ICEclc family. Two P. aeruginosa strains and their ICEs were added as outgroups (IOMTU133 and SP4371 from reference 38). The host “genome core” tree (right) is based on the concatenated alignment of all shared single-copy orthologous genes. ICE tree is based on a multiple sequence alignment of complete sequences. Gray and colored connecting lines indicate a different position in the ICE and genome core trees between the left and right, while color-shaded regions and letters (a–g) connect common host-ICE lineages. ICE names in red indicate those selected for further analysis. P. aeruginosa host names in brown fonts indicate those genomes harboring an ICE different from expected from their sublineage. The scale bar represents the expected average number of nucleotide substitutions per site. The scale of the ICE tree is broken between 0.6 and 2.56 nucleotide substitutions per site for visualization of the outlier “IOMTU133.” (B) Pair-wise percentage nucleotide identities (as heatmap according to colored scale) among 21 representative ICEs (i.e., those labeled red in panel A) from sequence alignments over the complete deduced ICE region. (C) Pair-wise average nucleotide identities among the 21 corresponding genomes to panel B (i.e., those labeled red in panel A) calculated across 30-bp sliding windows.

We next correlated the ICE phylogenetic signal with metadata such as the date and geographical location of sampling, isolation source, and MLST of the host (Fig. 3A). Some of the ICEs found in strains isolated in the 2000s are still found in more recent strains. For example, ICE_28832 isolated in 2008 is closely related to ICE_STH1, which is present in a P. aeruginosa strain isolated in 2017. Some others, like ICE_13524, isolated in 2003, are absent in more recent isolates from our data set. Interestingly, a few closely related ICEs are found in strains isolated in different geographical regions (e.g., ICE_26193, from a strain isolated in 2012 in Lausanne, is closely related to ICE_37295, present in a P. aeruginosa isolated in 2020 in Geneva). The strains carrying these ICEs are closely related in their host phylogeny (both P. aeruginosa 26193 and 37295 belong to group d, Fig. 2A). This suggests that the same P. aeruginosa clones are circulating within multiple hospitals. Also, the same ICEs are found in isolates from different origins, such as infected patients or hospital environment, potentially implying that contaminated surfaces in the clinics are a reservoir of the strains that infect patients and vice versa (e.g., ICE_34831, isolated from an infected patient, is closely related to ICE_36020 and ICE_35862, isolated from sink traps).

Fig 3.

Comparative phylogenetics of P. aeruginosa ICEclc family elements. (A) Correlation of ICE phylogenetic signal with the date and location of strain sampling, isolation source (light yellow, patients; other colors, various environmental sources; see Table S1) and MLST of the host. The phylogenetic tree is the same as in Fig. 2A except that the outgroups are omitted. (B) Phylogenetic comparative analysis of publicly available P. aeruginosa ICEclc family members. The bootstrapped (n = 1,000) consensus phylogenetic tree includes 139 P. aeruginosa ICEs from the ICEclc family (from reference 44) and the 21 ICEs analyzed in this study. For simplicity, only the ICE code numbers (IDs) of the latter and ICEclc are indicated on the tree; details of the former are available in Table S6. As outgroups, we included two P. aeruginosa ICEs randomly sampled from the sequences that were not considered of the clc family (namely, ICEs with a hit above 70% sequence identity to less than 20 of the ICEclc core genes; IDs PSAE.0321.00229_4970897.5065374 and PSAE.0321.00099_6407.111863 in Table S14 of reference 44). The tree is based on a multiple sequence alignment of complete ICE sequences. Gray tips represent ICEs without detectable anti-microbial resistance, while red tips indicate ICEs carrying anti-microbial resistance determinants (from Table S14 of reference 44). The scale bar represents the expected average number of nucleotide substitutions per site.

Micro-evolution of genes within the P. aeruginosa ICE conserved core

From the broad phylogeny of ICEs, we selected one element from each branch (red labels in Fig. 2A) to analyze conservation and micro-variations in their core genes. This resulted in 21 elements, which are a non-redundant representation of the diversity of circulating ICEs in the four hospitals. To place this local ICE diversity in perspective with worldwide circulating elements, we extracted 139 clc-like ICE sequences identified among 2009 publicly available and pruned P. aeruginosa genomes (44). We constructed a bootstrapped (n = 1,000) maximum-likelihood phylogenetic tree rooted with two outgroups that were randomly sampled from the same data set (44) but were not considered ICEs of the clc type (Fig. 3B). The ICEclc-type tree is composed of two to three major clades, one of which is broadly covered by most of the local ICEs (Fig. 3B). This indicates further ICEclc-type clonal diversification exists, which may become relevant in the context of anti-microbial resistance determinants (Fig. 3B, red symbols; see below).

We inferred the core- and pan-genome sizes (Fig. 4A through D) considering only the elements without assembly gaps (16 out of 21) and including ICEclc. The size of the core ICE genome, defined as the set of orthologous genes present in every element, saturates quickly when increasing the number of elements taken into consideration and tends toward 44 genes (Fig. 4A). The size of the soft core, defined as the set of genes present in 95% of the elements, shows a similar behavior but approaches 47 genes (Fig. 4B). The pan-genome size continually grows as more elements are added and does not seem to reach a saturation with the 17 selected ICEs (here at 350 genes, Fig. 4C), suggesting that new and unique variable genes are present in all the different elements. Accordingly, the cloud ICE genome, defined as the genes present only in one to two elements, is the most abundant class (Fig. 4D).

Fig 4.

Core- and pan-genome sizes of P. aeruginosa ICEs of the (local) ICEclc family. (A) Estimate of the number of genes making up the ICE core (n = 17 complete ICEs included). ICEs with assembly gaps (n = 6 ICEs) were omitted for a better estimation of true core-genome sizes. Red line represents the Tettelin fit (52). Each dot represents 1 of 18 independent subsampling of n = 1–17 ICEs. (B) Estimate of the number of genes making up the ICE soft core (i.e., the genes shared by 95% of the elements). Red line represents the Tettelin fit. (C) Estimate of the ICE pan-genome size (cutoff for similarity: maximum e value of BLAST comparisons = 0.00001 and minimum coverage length = 75%). (D) Attribution of the ICE genes into cloud (i.e., number of genes present in 1–2 of 17 ICEs), shell (i.e., number of genes present in 3–15 of 17 ICEs), soft-core (i.e., number of genes present in 16 of 17 ICEs), and core (i.e., number of genes present in 17 of 17 ICEs) compartments.

To address the extent of conservation in the defined core genome, we constructed a bootstrapped (n = 1,000) maximum-likelihood phylogenetic tree of the ICE core, inferred from all the single-copy orthologous core genes (n = 42 genes) (Fig. 5A). This list of core orthologous genes encompasses those that were experimentally confirmed for the ICEclc element and globally speaking codes for functions that control the lifestyle and transmission of the elements (Table S4). The core gene phylogeny mostly confirmed the tree topology of the dominant circulating ICEs based on their complete sequence (see indicated group names from Fig. 2A within names of Fig. 5A) but more clearly shows the divergence of the less common ICEs.

Fig 5.

Evolution of ICEclc family core and variable regions. (A) Consensus bootstrapped (n = 1,000) ICE phylogenetic tree based on established core genes (n = 42, single-copy orthologous genes) for the 16 complete ICEs selected from Fig. 2A (in red) with ICEclc as outgroup. Letters in brackets correspond to the subgroups of Fig. 2A. (B) Micro-divergence within the core conjugative gene region. Gene maps of the ICEclc conjugative locus versus three P. aeruginosa ICEs (ICE_24TCP, ICE_30429, and ICE_26797). ORFs and their orientation are represented by arrows; blue arrows indicate non-core genes; magenta arrows indicate core genes, and green arrows indicate gene insertions in two of the P. aeruginosa ICEs. Connecting colored blocks represent BLASTN similarities (darker colors indicate higher similarities). Gene annotation is based on published data (37) or BLASTP homologies. (C) Gene gain and losses (subset) across the 21 selected ICEs in conjunction with their phylogenetic positioning (consensus bootstrapped tree based on a multiple sequence alignment of complete sequences). (D) Enrichment or depletion in attributed functional classes of ICE-located genes in comparison to the P. aeruginosa genomes. Bars show mean proportions of cluster of orthologous groups (as indicated) class attribution from the annotated genes of the 21 P. aeruginosa genomes (ICE regions included) and their corresponding selected ICEs. Colors in the right panel indicate classes for which the observed mean proportion is bigger (high, red), smaller (low, green), or not significantly (blue) different from the 95% CI around the mean proportion attributed from 10 randomly sampled subsets of single-copy pan-genomic (all 21 P. aeruginosa isolates) chromosomal genes, which are shown as gray bars in the right panel. CBASS, cyclic-oligonucleotide-based anti-phage signaling system; TA, toxin-anti-toxin system.

At the level of individual conserved but non-core genes, the percentage of nucleotide identity was as low as 40%, as shown, e.g., for the integrase genes (Fig. S3). However, the integrase comparison further showed that ICE lifestyle modules comprise gene mosaics rather than strict coherent co-evolving entities. For example, the integrase tree has only 10 different integrase “branches” among the 21 groups (Fig. S3), and five putative ICEs carry two distinct integrases (magenta in Fig. S3) with ~70% nucleotide identity. Also, detailed inspection of gene order within core regions indicates that there are specific gene insertions and losses (Fig. S2). For example, we found an insertion of defense system genes within the conjugative locus of, e.g., ICE_24TCP and ICE_30429 (Fig. 5B, green arrows), with deletion of the ICEclc gene orf67800. This acquired locus contains a toxin-anti-toxin system (‘TA” in Fig. 5B) composed of a DNA-binding protein (anti-toxin) and a PIN domain containing protein (toxin). Additionally, it encodes a cyclic-oligonucleotide-based anti-phage signaling system, a defense system involving cell death prior to phage replication (53, 54). This system is composed of an oligonucleotide cyclase gene of the CD-NTase family (“cyclase” in Fig. 5B), an “effector” (endonuclease) gene, and additional associated Cap7- and Cap8-encoding genes. This locus further encodes a zinc metalloprotease (FtsH) and an ATPase of the AAA family. This suggests, from an evolutionary point of view, that the ICEclc-like elements in P. aeruginosa indeed carry a “core,” which undergoes clonal mutational drift and is subject to small indels, and a variable region that allows larger-scale gene insertions and deletions.

Variable region content indicates a wide variety of potentially adaptive genes among P. aeruginosa ICEclc-like elements

Since one would expect that the genes outside the immediate ICE lifestyle and transmission functions are potentially providing adaptive benefits to the host strains, we next analyzed in more detail the types of variable genes and their occurrences within the various predicted att borders of the ICEs. This interpretation is complicated by the large majority of hypothetical functions (i.e., without clear annotation and existing homologies; almost 75% of the genes, Table S5). However, a strikingly large abundance of predicted heavy metal resistance operons was detected among ICE variable regions (arsenic, cadmium, cobalt, copper, lead, mercury, nickel, and zinc; Table S5), with different patterns of co-occurrences among ICEs (Fig. 5C). For example, mercury resistance co-occurred with copper resistance, and in other cases with arsenic, or with cadmium resistance (Fig. 5C). In addition, we found a large fraction of annotated genes referred to (non-metal ion) efflux systems (e.g., BepE and BepF; enterobactin exporter EntS; toluene efflux pump TtgI; and multidrug resistance proteins MdtA, MdtE, and MexB) or transport energizing systems (e.g., potassium-transporting ATPase, 59 total occurrences in 85 ICEs), which may contribute to the capacity of the P. aeruginosa hosts to withstand antibiotic or anti-microbial compounds (55–59). Finally, some functions circulating in the ICE variable regions may point to adaptations to environmental or human-host conditions (Table S5). For example, a common ATP-dependent zinc metalloprotease (75 total occurrences in 85 ICEs) has been associated with virulence and survival of P. aeruginosa (60, 61). Other frequent, commonly found variable genes included a gene for polyphosphate ADP phosphotransferase, which may increase the capacity of the cell to scavenge phosphate; a fatty acid methyltransferase that could be involved in biotin synthesis; and quinone oxidoreductase 2, which associates to tricarboxylic acid (TCA) cycle components and 10-formyltetrahydrofolate synthesis (Table S5). Notably, local ICEs were remarkably depleted for clear anti-microbial resistance genes, which, however, are found in ICEs of the same major clade and the two other clades of the worldwide survey (Fig. 3B) (44).

In order to understand if ICEs are enriched for particular functional groups, which may be an indication of their selection, we compared the Cluster of Orthologous Groups (COG) distribution between the ICEs and their host genomes (Fig. 5D). We reasoned that if the cargo genes of the ICEs were under selection, their COG class distribution would be enriched compared to a random subselection of chromosomal genes. To test this hypothesis, we randomly sampled 106 genes (corresponding to the average number of genes present on the 21 analyzed ICEs) 10 times independently from the pool of all available chromosomal genes, classified them into COGs, and compared their average distribution and occurrence to those of the genes of the ICEs themselves (Fig. 5D). This showed that 11 out of the 25 COG categories are significantly underrepresented or overrepresented on the ICEs compared to the frequencies observed with randomly sampled chromosomal genes (Fig. 5D). For example, the class “inorganic ion transport and metabolism” is significantly enriched on the ICEs compared to the random sampling, which is in agreement with the noted abundance of heavy metal resistance operons on the elements. Also, genes in the COG classes “replication, recombination, and repair,” “transcription,” “cell cycle control,” “posttranslational modification,” and “general function prediction/unknown” were more represented than expected by chance (Fig. 5D). This suggests that these categories are under selection (and counterselection) on the local ICEs and may contribute to P. aeruginosa adaptation and survival under the Swiss hospital conditions.

DISCUSSION

The temporal tracing of a single focal species group in a consistent environment by full-genome sequencing is a great opportunity to study ongoing micro-evolution of mobile genetic elements such as ICEs. Genome mining and comparative genomic analysis of 181 P. aeruginosa isolates from hospital environments sampled over 20 years and a wider comparison to publicly available pruned P. aeruginosa genomes revealed a large diversity and abundance of ICEs with similarity to ICEclc of P. knackmussii B13 (25). This indicated and confirmed previous results from other hospital environments that ICEclc family elements are pervasive among P. aeruginosa strains (5, 12, 44). Our phylogenetic analysis showed that circulating local P. aeruginosa clones have low nucleotide diversity among their chromosomal core regions but primarily diversify by their accessory genomic segments. More than 90% of the analyzed genomes carried one or multiple ICEclc-like ICEs and frequently contained also ICEs or IMEs from other families, similarly as to recent conclusions from analysis of publicly available complete P. aeruginosa genomes (44). The ICEclc family members are characterized by a conserved core region that encodes basic lifestyle functions, such as the conjugative machinery, but they are diversified through different variable gene contents.

Phylogenetic comparison of the local P. aeruginosa chromosomal and ICE orthologous genes indicated both clonal ICE evolution as well as “non-matching” pairs. This is strong evidence that the ICEs are transferred horizontally among the different P. aeruginosa isolates in the hospital environment. Highly similar ICEs were also found in strains isolated from the other three hospitals, suggesting that both exchange of strains and of their ICEs can occur and contribute to P. aeruginosa diversification and adaptation. As far as we could analyze, the ICEclc family members appear to encode all genes necessary for their horizontal gene transfer, in comparison to a recent functional study of ICEclc conjugative transfer genes (37). This includes all genes for a complete type IV conjugative system, an integrase, and a relaxase gene. Some ICEs (e.g., ICE_38702 and ICE_36935) encoded two relaxases of different families, which may have consequences not only on the transfer dynamics of these ICEs themselves but also potentially on other co-mobilizable elements (IMEs) in P. aeruginosa.

In contrast, the characterized regulatory elements for ICEclc activation were only partly conserved among the P. aeruginosa ICEclc members. This suggests that control of ICE activation in P. aeruginosa may be distinct from that of ICEclc in P. knackmussii B13 and may respond to different environmental cues (ICEclc is activated in stationary-phase cells after growth on 3-chlorobenzoate). For instance, no homologs were detected among the P. aeruginosa ICEs for the mfsR and tciR genes, which are thought to provide the first level of control on ICEclc activation in P. knackmussii and P. putida during exponential growth (47). This confirms our previous assumption that (notably) insertion of the mfsR regulatory gene and the corresponding multidrug efflux system were an evolutionary novelty on ICEclc (32). In addition, only few P. aeruginosa ICEclc members encoded close homologs of alpA and bisR, which form the second regulatory layer of ICEclc activation (33)(47). Interestingly, however, the ICEclc bisDC regulatory genes responsible for formation of bistability and control of downstream ICE genes (such as the integrase and the conjugation system) were again conserved among all local P. aeruginosa ICEclc family members. This would indicate that the eliciting environmental or physiological cues influencing P. aeruginosa ICE activation are different from ICEclc, but once activated, these ICEs would also show bistable phenotypes of transfer-competent and “ICE-silent” cells (62). So far, however, the frequency of activation and ICE transfer from the P. aeruginosa hospital isolates has not been measured.

Although all local P. aeruginosa ICEclc members encompassed a conserved set of ~40 orthologous genes, the corresponding gene modules and gene syntenies showed evidence of mosaic structure. This indicates that they are not strictly coherent co-evolving entities but can undergo further gene insertions and deletions, as well as mutational drift. This may occur, for instance, by recombination between modular cores of different ICEs within the same host, which may also lead to the acquisition of new genes or regulatory elements and can contribute to the evolution of novel ICE variants. From an evolutionary point of view, this suggests that the core of the ICEs plays a critical role in their survival and persistence over time. In contrast, the permissiveness for gene insertions and deletions in their “variable” regions allows for even greater flexibility and adaptability and permits larger-scale changes that may result in the acquisition or loss of functional genes that confer a selective advantage to the host organism.

Not surprisingly, the uncovered P. aeruginosa ICEclc members carried a wide range of variable gene functions. Notably, genes not only for heavy metal tolerance but also for regulatory elements were commonly present and enriched with respect to random models, suggesting they are under positive selection. Accordingly, it seemed that, over time, the ICEs are acquiring more heavy metal resistance determinants, as these were less abundant in ICEs found in strains isolated earlier in time (Fig. 3A and 5C). Other ICEs have been detected in P. aeruginosa that encode heavy metal resistance, such as PAGI-2. Previously, a 99.972% ICE similar to PAGI-2 was found in Ralstonia metallidurans CH34 (5), which was described to grow in millimolar concentrations of toxic heavy metals. This suggests by analogy that the heavy metal resistance genes on the ICEs in the P. aeruginosa clones can contribute to their tolerance against heavy metal stress. Why this remains so strongly conserved and pervasive among the Lausanne/Geneva hospital clones is unclear. Potentially, their positive selection is related to the presence of metallic copper coating in the water systems of healthcare institutions, which are known to be frequently colonized by copper-tolerant P. aeruginosa and to act as a reservoir of opportunistic strains (63–66). Additionally, there may be hidden mechanisms of cross-resistance between heavy metals and other anti-microbial compounds, such as those described for copper and antibiotic cross-resistance (64).

Even though we did not detect any specific antibiotic-resistance genes among the local P. aeruginosa ICEclc members, notably, genes for carbapenem resistance have been acquired by ICEclc-like elements in other P. aeruginosa hospital strains, e.g., in Portugal and South Korea (38, 42). Furthermore, comparison to a worldwide survey of P. aeruginosa genomes carrying ICEclc-type elements that were identified in a 2023 study by Botelho and coworkers (44) showed that many related elements are present among the three major ICEclc clades with genes for anti-microbial compound resistance (Fig. 3B). It may thus just be a matter of time and selection pressure before such genes, either by recombination onto existing ICEs, by conjugative transfer of the corresponding ICEs, or by transmission of the P. aeruginosa host, find their way into the local hospital environment. Further inconspicuous potential resistance or virulence factors may be present on the local ICEs described here. These include a number of efflux systems which are associated with increased tolerance to anti-microbials, and an ATP-dependent Zinc metalloprotease FtsH, homologs of which have been involved in a variety of phenotypes associated to virulence and survival (60, 61). In conclusion, our findings support the importance of adaptive mobile genetic elements in the success of P. aeruginosa as an opportunistic pathogen.

MATERIALS AND METHODS

Strain isolations and sequencing

Strains from the CHUV were isolated and typed during ongoing infection control surveillance and investigations, some of which have been described elsewhere (67–69). Strains from CHU Besançon originated from a study on the evaluation of a new typing system (70). Strains from HUG were isolated during an outbreak of VIM-producing P. aeruginosa (71). Strains from the CHU Montpellier were isolated in the course of a research protocol (ClinicalTrials.gov ID: NCT02751658) that studied the dynamics of hospital environment contamination by waterborne pathogens associated to healthcare infections. All strains were sequenced using Illumina MiSeq technology.

Quality check and de novo assembly

The quality of the Illumina raw reads was assessed with FastQC (v.0.11.9) (72). Reads were trimmed with trimmomatic (v.0.39) (73) to remove short and low-quality reads and adapters. Genomes were de novo assembled with SPAdes (v.3.15.2) (74). The resulting contigs were filtered to keep only those longer than 500 bp and with a kmer coverage higher than 10. Single chromosomes were generated with RagTag (v.2.1.0) (75) from the filtered contigs of every genome by scaffolding and patching the filtered contigs against a reference complete genome of P. aeruginosa strain H26027 (NCBI accession number CP033684), also isolated from the Lausanne University Hospital. The genome scaffolds have been deposited with the European Nucleotide Archives under the accession study number PRJEB61470.

The sequence type of each strain was determined using MLST (v.2.19.0) (76), a program which extracts seven single-copy housekeeping genes (acsA, aroE, guaA, mutL, nuoD, ppsA, and trpE) and compares their sequence identity to previously deposited allele combinations in the P. aeruginosa PubMLST database (77).

Isolation, annotation, and comparison of ICE sequences

Forty ICEclc genes were used as query for BLASTN (v.2.13.0) (78) searches against the filtered contigs of each genome (using command line default parameters). To delineate the ICEclc-like elements from their draft P. aeruginosa genomes, we first defined regions of 100 kb upstream and downstream of the located traI-homolog using SeqKit (v.2.2.0) (79). Within these regions, potential ICE attachment sites were manually searched by homology to the ICEclc direct repeats (5′-GTCTCGTTTCCCGCTCCA-3′), one of which should be located at the end of a gene for tRNA^Gly and the other located within 50–130 kb from the first, allowing up to two mismatches (35, 80). Delineated “boundary”-defined ICEs were then annotated with PROKKA (v.1.14.6) (81), which was also used to annotate the rest of the draft genome. Sequence comparison maps (Fig. 1B and 4B; Fig. S2) were generated using genoplotR (82).

The presence of a tRNA-encoding gene at the site of insertion was detected with tRNAscan-SE (v.2.0) (83). Relaxase families were determined using MOBscan (84). Pair-wise average nucleotide identities were calculated using fastANI (v.1.33) (85) for the genomes and Bio3d (v.2.4–3) (86) for the aligned ICEs and integrases. Percentages of GC nucleotide content were determined employing a sliding window method (https://github.com/DamienFr/GC_content_in_sliding_window) by using default parameters for the genomes, a window size of 100 bp and a step size of 10 for the ICEs. Estimations of the total number and variety of ICEs and IMEs were made by using ICEfinder (https://bioinfo-mml.sjtu.edu.cn/ICEfinder/ICEfinder.html) (87).

To have a broader overview on globally circulating ICEclc-type elements, we downloaded all the P. aeruginosa genomes based on Table S13 of reference (44) (a total of 203 genomes, accession numbers in Table S6). Suspected ICE sequences were extracted from their respective genomes based on the coordinates given in Table S13 of reference (44) using seqkit subseq (v.2.2.0) (79). The 477 resulting ICEs were blasted against the 40 identified ICEclc core genes. Only the ICEs with a hit above 70% sequence identity to at least 20 of the ICEclc core genes were considered clc-like and kept for further analysis, resulting in 139 ICEs (Table S6 and presented in Fig. 3B).

Pan-genome and phylogenetic analysis

We calculated a pan-genome of all local ICEs (i.e., without including chromosomal genes), with the GET_HOMOLOGUES package (v.22082022) (88), under the following non-default parameters: the number of input sequences used for the generation of pan/core genomes was set to 18; and OrthoMCL (v1.4) (89) was used as clustering algorithm. Core and pan-genomes were computed with the compare_clusters.pl and parse_pangenome_matrix.pl scripts; rarefaction curves were generated using the plot_pancore_matrix.pl script (88).

For the comparative phylogenetic analysis of draft genomes of local P. aeruginosa clones and ICEs related to ICEclc, we concentrated on a set of 85 ICEs out of the initial set of 181 putative ICE regions. ICEclc family members were considered redundant if the 200-kb regions around the identified traI relaxase were similar among strains isolated from the same patient. To this set, we included ICEs from two external P. aeruginosa strains (NCBI accession numbers NZ_AP017302 and NZ_CP034369) from the study of Botelho et al. (38) that were found to encode carbapenem resistance genes. To build the species tree, with help of OrthoFinder (v.2.3.8) (90), we identified all single-copy orthologous genes among the 85 P. aeruginosa genomes and extracted all the corresponding nucleotide sequences using a custom python script, yielding a total of 4,260 ortholog genes. Next, the gene sequences were aligned with MAFFT (v.7.475) (91) and concatenated, from which a maximum-likelihood phylogenetic tree was built with IQ-TREE 2 (v.2.0.6) (92). The general time-reversible model, with unequal rates and unequal base frequencies, was used to compute the species tree (93). Trees were replicated in 1,000 bootstrappings to test the reliability of each branching. For the ICEclc family member and integrase trees, a multiple sequence alignment on the complete ICE or integrase sequence was produced with MAFFT, and the complete alignment was used to build the tree with IQ-TREE 2. For the ICEclc family member tree (Fig. 2A), the transversion model, with AG = CT and unequal base frequencies, was used to compute the tree. For the integrase tree (Fig. S3), the unequal transition/transversion rates and unequal base frequencies model were used to compute the tree (94). The tree of publicly available P. aeruginosa ICEclc family members (Fig. 3B) was computed using the transition model, with AC = GT, AT = CG, and unequal base frequencies. The ICE single-copy orthologous genes tree (Fig. 5A) was built with the same approach as the species tree [using the general time-reversible model of reference (93)]. Phylogenetic trees were displayed and compared with GGTREE (95) or dendextend (96).

Hierarchical orthologous groups were inferred using OMA standalone (v.2.5.0) (97), and the resulting phylogenetic profiles (Fig. 3A and 5C) were visualized using Phandango (98).

COG analysis

COG classes encoded by ICEs and P. aeruginosa genomes were attributed using WebMGA (99). To estimate the enrichment for COG attributable functions among the ICEs, we compared the actual COG class distributions to those from a randomly sampled distribution model. The model was produced by random sampling of 106 genes (the average number of genes in the 21 ICEs) from a “virtual pan-genome,” repeated 10 times independently. The virtual pan-genome was a list with every gene appearing in the annotation of the 21 host genomes (except for the ICE regions), filled up to 6,295 total genes (the average number of genes in the 21 analyzed genomes) with “hypothetical proteins.” The subsampled gene lists were again attributed to their COG class, and the average per-class distribution was calculated over the 10 random samples. Significant differences among actual and expected distributions (in Fig. 5D) are those in which the actual average class proportion is bigger or smaller than the 95% CI around the average of the corresponding randomly drawn class proportions.

DATA AVAILABILITY

The data sets generated and analyzed during the current study are available in the European Nucleotide Archives repository under the accession study number PRJEB61470.

SUPPLEMENTAL MATERIAL

The following material is available online at https://doi.org/10.1128/msphere.00517-23.

Fig. S1. msphere.00517-23-s0001.pdf.

Chromosome map of P. aeruginosa PAO1.

Fig. S2. msphere.00517-23-s0002.pdf.

Gene synteny comparisons among the 21 P. aeruginosa ICEclc-family elements.

Fig. S3. msphere.00517-23-s0003.pdf.

ICE relatedness inferred from similarities among their integrases.

Captions. msphere.00517-23-s0004.docx.

Captions to all supplemental material.

Table S1. msphere.00517-23-s0005.xlsx.

Characteristics of P. aeruginosa isolates.

Table S2. msphere.00517-23-s0006.xlsx.

Characteristics of the attributed ICEclc family elements in individual P. aeruginosa isolates from the regional hospital.

Table S3. msphere.00517-23-s0007.xlsx.

Total attribution of integrative and conjugative/mobilizable elements in P. aeruginosa isolates from the regional hospital.

Table S4. msphere.00517-23-s0008.xlsx.

Deduced core functions of ICEclc family members.

Table S5. msphere.00517-23-s0009.xlsx.

Gene function occurrences among the ICEclc family gene "cloud" in regional P. aeruginosa clinical isolates.

Table S6. msphere.00517-23-s0010.xlsx.

ICEclc family members identified among a worldwide survey of publicly available P. aeruginosa genomes.

ASM does not own the copyrights to Supplemental Material that may be linked to, or accessed through, an article. The authors have granted ASM a non-exclusive, world-wide license to publish the Supplemental Material files. Please contact the corresponding author directly for reuse.