Genomic structural plasticity of rodent‐associated Bartonella in nature

Abstract

Rodent‐associated Bartonella species have shown a remarkable genetic diversity and pathogenic potential. To further explore the extent of the natural intraspecific genomic variation and its potential role as an evolutionary driver, we focused on a single genetically diverse Bartonella species, Bartonella krasnovii, which circulates among gerbils and their associated fleas. Twenty genomes from 16 different B. krasnovii genotypes were fully characterized through a genome sequencing assay (using short and long read sequencing), pulse field gel electrophoresis (PFGE), and PCR validation. Genomic analyses were performed in comparison to the B. krasnovii strain OE 1–1. While, single nucleotide polymorphism represented only a 0.3% of the genome variation, structural diversity was identified in these genomes, with an average of 51 ± 24 structural variation (SV) events per genome. Interestingly, a large proportion of the SVs (>40%) was associated with prophages. Further analyses revealed that most of the SVs, and prophage insertions were found at the chromosome replication termination site (ter), suggesting this site as a plastic zone of the B. krasnovii chromosome. Accordingly, six genomes were found to be unbalanced, and essential genes near the ter showed a shift between the leading and lagging strands, revealing the SV effect on these genomes. In summary, our findings demonstrate the extensive genomic diversity harbored by wild B. krasnovii strains and suggests that its diversification is initially promoted by structural changes, probably driven by phages. These events may constantly feed the system with novel genotypes that ultimately lead to inter‐ and intraspecies competition and adaptation.

1. INTRODUCTION

The understanding of genetic mechanisms involved in bacterial diversification and adaptation is fundamental to identify imprints associated with the emergence of pathogenicity and antibiotic resistance. With the development of advanced sequencing platforms, trustworthy assemblies could be constructed from complex bacterial genomes (De Maio et al., 2019). Since genomic comparisons have been performed, the dynamic nature of bacterial genome organization was elucidated. In order to preserve important intrinsic mechanisms such as replication, transcription and translation, prokaryotic genomes endure high selection‐pressure to keep their structural stability through generations (Touchon & Rocha, 2016). However, the constant DNA exchange due to horizontal gene transfer (HGT) and exposure to fluctuating ecological niches, cause the instability of genomes. Genome instability is marked by evolutionary differences between gene repertories and genomic organizations (Cui et al., 2012; Iguchi et al., 2006; Liu et al., 2006; Scott et al., 2007; Yan et al., 2018). The evolution of genomes is shaped by the equilibrium between the maintenance of the genome organization and tolerance to instability. Instability plays a key role in adaptation and survival of bacteria, particularly in double niche bacteria, such as Bartonella species that constantly circulate between arthropod vectors and mammal hosts (Bochkareva et al., 2018; Darmon & Leach, 2014; Dobrindt & Hacker, 2001).

The genomic instability of bacterial chromosomes caused by structural variations (SVs) of large DNA fragments, including deletions, insertions, inversions and translocations, might result in altered genetic information and gene clusters (i.e., operons), and may interfere directly in the transcription and translation patterns of the bacterial machinery (Cerdeño‐Tárraga et al., 2005; Periwal & Scaria, 2015; Sheppard et al., 2018). Among the evolutionary elements involved in the genome instability, mobile elements can promote genomic changes directly through the exchange of genetic information between genomes (Cui et al., 2012; Nzabarushimana & Tang, 2018) and indirectly through the activation of DNA damage‐repair mechanisms by its integration or excision within the bacterial chromosome (Argueso et al., 2008; Hoff et al., 2018). Bacteriophages are a type of mobile elements that have a key role in the ecology and evolution of bacterial populations through the incorporation of external DNA into the bacterial chromosome, as well as selective weapons between populations (Bobay et al., 2014). Integrated bacteriophages, or prophages, have a relevant function in the bacterial adaptation, potentially enhancing bacterial fitness and resistance to foreign bacteriophages, as well as the dissemination of genes that confer virulence and antibiotic resistance (Howard‐Varona et al., 2018). Moreover, phages can induce SVs in the recipient bacteria through HGT (Ramisetty & Sudhakari, 2019) and vertically, without any influence of foreign genetic information, as shown in an evolutionary study with Bartonella spp. (Gutierrez, Markus, et al., 2018).

The Bartonella genus belongs to the order Rhizobiales, which comprises bacteria with diverse lifestyle (Neuvonen et al., 2016). It has been proposed that Rhizobiales evolved from free‐living ancestors to host‐associated bacteria (Wang et al., 2020). Bartonella species circulate between intracellular (within their mammalian hosts) and extracellular milieus (within arthropod guts), representing an intriguing genus‐model to study adaptation and evolution within the order Rhizobiales (Gutiérrez et al., 2015; Kosoy et al., 2012). Moreover, the diversity among Bartonella is notable, with 37 validated Bartonella species and three subspecies, and over 25 organisms with a Candidatus status or unclarified phylogeny (https://www.bacterio.net/genus/bartonella). Among the extensive list of Bartonella reservoirs, wild rodents and their associated fleas harbor one of the largest Bartonella diversity described to date (Gutiérrez et al., 2014, 2015; Inoue et al., 2009; Morick et al., 2010; Paziewska et al., 2011). The diversity observed in rodent‐associated Bartonella has been evident both at the nucleotide level (i.e. hundreds of genotypes identified based on genetic polymorphisms) and at the structural genomic level (Gutiérrez, Cohen, et al., 2018). Remarkably, recombination events between rodent‐associated Bartonella spp. have been extensively reported, incriminating it as the main mechanism for the diversity observed in nature (Gutiérrez, Cohen, et al., 2018; Morick et al., 2010; Paziewska et al., 2011). Hence, the vast evidence accumulated in the last two decades on rodent‐associated Bartonella, reflects their marked genomic plasticity, which deserves special attention due to their role as zoonotic and animal pathogens.

Our group has recently isolated and characterized four novel Bartonella species, represented by 38 different genotypes (Gutiérrez, Cohen, et al., 2018). These organisms were isolated from the blood and fleas of two sympatric rodent species, sampled in two sites in the Negev Desert, southern Israel, circumscribed within a 3‐km distance area (Gutiérrez, Cohen, et al., 2018). Coinfections with these rodent‐associated Bartonella spp. and genotypes were commonly observed among these hosts (in over 50% of the hosts), providing ideal conditions for HGT and recombination events (Gutiérrez et al., 2014; Gutiérrez, Cohen et al., 2018). From the Bartonella spp. identified, Bartonella krasnovii stands out as an example of the genomic plasticity within this genus, with 16 genotypes identified, all showing different genomic structures (i.e., different enzymatic restriction profiles) (Gutiérrez, Cohen, et al., 2018). Furthermore, under laboratory conditions, it was shown that B. krasnovii can undergo spontaneous SV events during a short period of evolution, while single‐point mutations are kept restricted (Gutierrez, Markus, et al., 2018). Interestingly, these SVs were associated with prophages, raising the hypothesis that these genetic elements promote rapid occurrence of genomic variation in this bacterial genus (Gutierrez, Markus, et al., 2018). In accordance with the latter, SVs have been previously observed among Bartonella genomes (Alsmark et al., 2004; Gutierrez, Cohen, et al., 2018; Lindroos et al., 2006), however, the genomic extension of this phenomenon in wild strains has been only partially explored.

Considering the importance of rearrangements for bacterial adaptation and variant emergence, the present study explored the extent of genomic variation events in nature, by comparing the diversity observed at nucleotide and gene content levels with structural genomic diversity. For this aim, a collection of B. krasnovii strains, isolated at the same period and location, was chosen, and the genetic content and genomic organization were evaluated both independently and by comparative analyses against OE 1–1 strain genome (isolated and sequenced 2 years earlier). In order to evaluate the association of the structural rearrangements and prophages, a genomic comparison with Bartonella quintana (an acknowledged Bartonella species lacking of prophages) was also performed. The results of this study shed light on the genetic elements involved in the variability and genomic plasticity in the Bartonella genus.

2. MATERIALS AND METHODS

2.1. Bacterial strains and culture conditions

Bartonella krasnovii strains were isolated from 10 wild Gerbillus andersoni and 10 Gerbillus pyramidum and 40 associated Synosternus cleopatrae fleas, captured during October 2016 in two different sand dune sites (~4 km apart) in the Negev Desert, Israel (34°23′E, 30°58′N and 34°23′E, 30°55′N) (Gutiérrez et al., 2018) (Figure 1a). These B. krasnovii strains were previously characterized into 16 genotypes (named A1‐A5, B, C1‐C3, D1‐D5, G and H) based on the identification of nucleotide polymorphisms at three genetic and intergenetic loci (i.e., gltA, rpoB and ITS); (Gutiérrez, Cohen, et al., 2018; Figure 1b). All strains were cultured on chocolate agar plates (Novamed, Ltd., Jerusalem, ISR), incubated at 37°C with constant 5% CO₂ atmosphere. Despite the fact all strains belong to the same species, they presented differences in colony morphotype and growth rates (Table S1). To further study their genomic diversity, one to three different B. krasnovii strains per genotype (for a total of 46 strains) were selected. The genomes of a total of 46 Bartonella strains were analysed by the SmaI digestion profiles using pulse field gel electrophoresis (PFGE), following previously described methods (Gutiérrez, Markus, et al., 2018). The observed SmaI restriction profiles were carefully and manually checked, using Salmonella ser. Braenderup H9812 (i.e., plugs digested with XbaI) as molecular marker in each run. Seventeen strains with different PFGE profiles, and at least one different Bartonella strain per genotype (for a total of 20 strains) were chosen for sequencing analyses (Figures 1b–c).

FIGURE 1.

Study design. (a) Sample processing: Isolation of rodent‐associated Bartonella from gerbils and their associated fleas and further DNA extraction. (b) Variant identification: Genotyping based on SNP of Bartonella conserved loci and SmaI‐PFGE profiles. (c) Whole‐genome sequencing: 20 different strains, comprising the 16 sequencing polymorphic genotypes and 17 different PFGE variants were selected to WGS through two different sequencing platforms, short reads (Illumina MiSeq sequencing) and long reads (MinION, Oxford Nanopore). (d) Genome assembly and validation: The genomes were assembled using pipelines which employ short and/or long reads, and the assembly outputs were validated by the comparison of in silico SmaI‐digested profiles and the “real” PFGE profiles. PCR assays of flanking regions were carried out. (e) Genomic analyses: The genomes were submitted to SNP calling analysis and only the validated assemblies were submitted to SV analysis. This figure was created using Biorender (http://BioRender.com; last accessed 29 December 2021)

Bartonella krasnovii strain OE 1–1 (Gutiérrez et al., 2020), isolated 2 years earlier, from a S. cleopatrae flea, which its genome was fully characterized and curated (CP031844.2, CP042965.1), was used as a reference for genome and genetic comparisons. This isolate was classified as genotype “A2” according to the polymorphism analysis of genetic and intergenic loci described above.

2.2. Whole genome sequencing

The DNA from each pure strain (retrieved after 3–4 consecutive single‐colony passages on chocolate agar plates) was extracted with DNeasy Blood and Tissue Kit (Qiagen), following the manufacturer's recommendations. The DNA concentration, as well as the presence of chemical contaminants, were measured using the spectrophotometer NanoDrop (Thermo Fisher Scientific), and the DNA integrity was checked on agarose gels. DNA extracted from the 20 selected strains were processed on two different sequencing platforms, namely short reads sequencing through Illumina MiSeq sequencing (150 bp paired), and long reads sequencing through MinION Mk1b (Oxford Nanopore Technologies, Oxford, GBR) using SpotON flow cell (see Methods S1).

2.3. Genome assembly validation

The genomes were assembled using both Unicycler version v0.4.9 (Wick et al., 2017), which combined short and long reads, and Canu version v2.2 (Koren et al., 2017), which uses long reads. Then, both resulting assemblies were compared to the PFGE profiles, as follows. First, the assembly outputs were characterized according to the quality and fragmentation status, using assembly‐stats (version 3). Then, the fasta files from the largest contig obtained from the assemblies (if this was above 2.0 Mb length) were analysed in Geneious version 7.1.9 (Kearse et al., 2012). The linear contigs were circularized in Geneious and we set (arbitrarily) as the starting site the chromosomal replication initiator protein dnaA gene. Once circularized, the assemblies were in silico SmaI‐digested in Geneious, producing a “virtual gel” of the restriction fragment profile. Genome assemblies were then validated (Figure 1d), testing whether their “virtual” restriction fragment profiles and estimated sizes match the “real” PFGE patterns. The resulting validated assemblies are shown in Table S2. Furthermore, a PCR assay per genome assembly was designed to amplify two flanking regions produced by SmaI restriction sites (Table S3). Therefore, only assemblies with matched PFGE and virtual gels and PCR‐positive for the confirmation of flanking regions, were considered as structural‐validated genomes.

All genome assemblies were included for nucleotide variation analyses, while only structural‐validated genomes were analysed for the presence of SVs (Figure 1e).

2.4. Average nucleotide identity

Average nucleotide identity (ANI) (Goris et al., 2007) was calculated between all the study B. krasnovii assemblies (i.e., validated and unvalidated) and the OE 1–1strain B. krasnovii genome (Gutiérrez et al., 2020), using FastANI (Jain et al., 2018).

2.5. Single nucleotide polymorphisms (SNP) calling analysis

The SNP calling analysis was performed using the Illumina sequenced short‐reads obtained from the 20 B. krasnovii strains mapped to the OE 1–1 strain genome (CP031844.2, CP042965.1), as previously described earlier (Gutiérrez, Markus, et al., 2018; see Methods S1).

2.6. Identification of selection signatures

The McDonald and Kreitman Test (MKT) was performed to evaluate non‐neutral evolution by comparing the ratio of nonsynonymous to synonymous polymorphisms within species (Pn/Ps) to the ratio of nonsynonymous to synonymous fixed differences between species (Dn/Ds), using the PopGenome package in R (McDonald & Kreitman, 1991; Pfeifer et al., 2014). The outgroup genome used for this test was the Bartonella birtlesii genome (GCA_000278095.1). In order to minimize any biases promoted by slightly deleterious mutations, low‐frequency polymorphisms (MAF < 0.1) were removed from this analysis, following recommendation described elsewhere (Charlesworth & Eyre‐Walker, 2008). The test was applied at the genomic level and for each coding‐gene separately, in which case p‐values were adjusted by false discovery rate (FDR).

2.7. Genome annotation and characterization

All genomes were annotated with Prokka (Seemann, 2014). Additional annotations were performed on all structural‐validated genomes using RASTtk (Brettin et al., 2015), implemented in the PATRIC portal (Wattam et al., 2017), and the Microbial Genomes Atlas Online (MiGA) (Rodriguez‐R et al., 2018).

Prophages were annotated on all structural‐validated genomes using Phaster annotation tool (Arndt et al., 2016; online server; http://phaster.ca, last accessed 3 November 2019). This tool identifies prophage regions from nucleotide sequences using gene prediction and local and GenBank annotation tools. In this way, Phaster is able to provide annotations regarding the phage position, length, boundaries, number of genes, attachment sites (att), tRNAs and phage‐like genes. The annotations allow the prediction of completeness or potential viability of identified prophages and classifies them as intact, incomplete or questionable. Additionally, clusters of at least 4 phage‐related genes were manually identified from the genome annotations (i.e., RASTtk), within a window of 5 up to 30 kb, and classified as “prophage‐remnants”. To identify possible relationships among the prophages, the principal component analysis (PCA) was done based on the distance matrix of the prophage nucleotide identities (see Methods S1).

Plasmids were screened in all genome assemblies as follows: (i) through the alignment of the short reads to the OE 1–1 strain plasmid (CP042965.1); (ii) searching conjugative transfer plasmid genes (traJ, trbG, trbH, trbI, trbG, traF, traD, trbB, trbC, trbD, trbJ, trbL) through tblastx (Boratyn et al., 2013) on the strain's assemblies; and (iii) through PlasmidVerify tool (Antipov et al., 2019).

2.8. Pangenomic analyses

To estimate the pangenome of B. krasnovii, the genomes were analysed following Panaroo (Tonkin‐Hill et al., 2020). To study further the sequence diversity at the gene level, predicted gene groups obtained from Panaroo were used to calculate ANI using pANIto (version 1), and the percentage of similarity of each core and soft genes was compared against the strain OE 1–1 genome.

2.9. Replication origin and termination loci's prediction

The replication origin (ori) and termination (ter) loci were predicted for each structural‐validated assembly using the GC skew analysis (Lobry, 1996) as previously described by Neuvonen et al. (2016) for Bartonella species, through the g‐language system (Arakawa et al., 2008), using the online server (http://soap.g‐language.org/gembassy/emboss_explorer; last accessed 22 October 2019) and through Oriloc program (Frank & Lobry, 2000). The latter procedure was validated through the GC skew analysis performed for a curated genome of Bartonella tribocorum (NC_010161.1) and compared to an updated database of replication origins in prokaryotic genomes DoriC 10.0 (Luo & Gao, 2019) available online (http://tubic.org/doric/public/index.php/index; last accessed 25 October 2019). The starting position and sense for all structural‐validated assemblies was defined using the annotation coordinates of the chromosomal replication initiator protein (dnaA) gene (position 1 of all genomes).

2.10. Structural variation analyses

To identify SVs in the structural‐validated genomes, pairwise whole genome alignments were obtained from each B. krasnovii assembly and the strain OE 1–1 genome using progressiveMauve (Darling et al., 2004, 2010), implemented in Geneious version 7.1.9 (Kearse et al., 2012). Accordingly, progressiveMauve alignments allowed the identification of inverted or translocated DNA regions (i.e., inversion and translocation events) within the genomes in comparison with the coordinates of its homologue regions within the strain OE 1–1 genome. On the other hand, deletions and insertions were identified using the short‐read alignment data obtained through GATK pipeline (McKenna et al., 2010; see Methods S1). To minimize false SV events, a cutoff of ≥1000 bp length per event was employed (i.e., all events <1000 bp length were excluded). The flanking regions (around 1 kb) of the SVs were explored from the genome annotation to identify the loci involved in the events.

To estimate the impact of the SVs on the study genomes, we explored the distance between the ori‐ and ter‐regions, which gave an estimate of the balance of the genomes. Accordingly, an off‐balanced or unbalanced genome was defined when the replicores lengths (i.e., ori‐to‐ter and ter‐to‐ori distances) resulted in a difference ≥ 40 kb, which represents a ≥2% difference of a 2.0 Mb circular genome (Kothapalli et al., 2005). In addition, using the genomic coordinates of the essential genes (n = 105) predicted by MiGA, we explored whether SVs affected their distribution (as inversions may cause the shifting on genes from leading to lagging strand and vice versa) and distance between consecutive essential genes (as translocations/deletions/insertions may impact the distance between these genes). For this aim, using the predicted ori‐ and ter‐regions, we defined the leading and lagging strands in the assemblies, being the leading strands of a circular genome the positive sense of the linear DNA segment from ori to ter, and the negative sense of the linear DNA segment from ter to ori. Then, using the genomic coordinates and strand directionality, we predicted the orientation on the leading or lagging strands of essential genes and the distance between the stop and starting position of consecutive essential genes.

Phylogenetic inference of the nucleotide‐based tree and the SV‐based tree was performed to estimate the implication of the SV events on the evolution of these B. krasnovii strains (see Methods S1). The plots corresponding to the genome comparisons and SVs phylogeny were generated through genoPlotR (Guy et al., 2011).

2.11. Bartonella quintana comparison

To evaluate the potential implication of the prophages on the structural organization of Bartonella genomes, a comparison of intraspecies variation with the B. quintana, an acknowledged Bartonella species deprived of prophages (Alsmark et al., 2004) was chosen to serve as a control. Accordingly, five fully circulated B. quintana genomes available in GenBank (https://www.ncbi.nlm.nih.gov/; last accessed 4 October 2021) were downloaded and annotated (following the above‐described methods). The included genomes represented a worldwide collection of B. quintana, as they were obtained from different hosts and geographical regions: B. quintana strain “Toulouse” (NC_005955), isolated from a human patient from France in 1993 (Alsmark et al. 2004), B. quintana strain “MF1‐1” (NZ_AP019773.1), isolated from a monkey from Japan in 2012 (Sato et al., 2015), B. quintana strain “RM‐11” (CP003784.1) a monkey, isolated from China in 2011 (Li et al., 2012), B. quintana strain KorN (CP041670), isolated from a human patient from South Korea in 2017; and B. quintana strain “NCTC12899” (NZ_LS483373.1), collection date 2016, for which isolation source is not available nor published.

3. RESULTS

3.1. Genome sequencing results

Illumina sequencing resulted in 2,889,782–5,647,059 paired‐end 150‐bp reads per sample (3,886,435 ± 828,284 reads; average ± SD). MinION Mk1b sequencing resulted in 90,011–501,921 long reads (with average of 200,463 ± 88,815) per sample, with an average read length of 6881.4 ± 1189.8 bp. The genome assembly lengths ranged from 2,023,468–2,260,108 bp, with an average of 2,143,012 ± 62,848 bp (±SD). GC% content varied from 37.9–38.2%, average 38.1 ± 0.11%.

3.2. Validation of genomic assemblies

Fourteen out of 20 genomes (70%) were structurally‐validated and confirmed as circular contigs. The structural‐validated genomes ranged between 2,047,607–2,210,336 bp of length, GC content at the range of 37.9–38.2%, number of CDS from 2021 to 2347, repeat regions from 52 to 136, and all genomes carried 42 tRNAs and 6 rRNAs (Table S2).

3.3. Nucleotide variation analyses

The average number of SNPs between the OE 1–1 strain and the B. krasnovii strains was 6381 ± 2125 SNPs per genome, ranging from 295–8667. On average, each genome carried 2756 ± 907 synonymous and 2015 ± 686 nonsynonymous mutations in their coding‐sequences (CDS) (Table 1).

Subsequent SNP allele frequency analysis highlighted seven regions with greater alternative allele frequency across the 20 genomes (Figure 2a), as follows: (i) hemin binding protein genes (peak 1); (ii) locus of the cytochrome C‐type biogenesis protein gene, DNA‐methyltransferase and pyruvate phosphate dikinase (PPDK) genes (peaks 2 and 3); (iii) Bartonella effector protein genes (Beps) for intracellular delivery domain (BID) containing VirB/VirD4 type‐IV‐secretion system (T4SS) (peak 4); (iv) VirB/VirD4‐homologous T4SS (vbh) regions (peaks 5 and 6); and (v) in Trw‐T4SS (trw) genes region (peak 7) (Figure S1).

FIGURE 2.

Nucleotide variation of wild Bartonella krasnovii strains is restricted to metabolic and virulence factor loci, exhibiting a minimal selection signature. (a) Frequency of SNPs on the studied genotypes per 20 KB window of the OE 1–1 B. kransnovii genome. Numbers refer to regions with greater alternative allele frequency (“peaks”) that are above the threshold of distribution, accordingly the following numbers represent: 1. Hemin binding proteins; 2 & 3. Cytochrome C‐type biogenesis proteins, DNA‐methyl transferase and pyruvate phosphate dikinase; 4. BID domain containing VirB/VirD4 type‐IV‐secretion system (T4SS) effectors; 5 & 6. VirB/VirD4‐homologous T4SS (Vbh) and 7. Trw‐T4SS (trw) genes region (Trw). (b) Distribution of α values identified in coding genes analysed separately across the genomes through McDonald and Kreitman Test (MKT) using Bartonella birtlesii as an outgroup. The red dots represent genes under negative selection (α ≤ 0) and the blue dots represent genes under positive selection (α ≥ 0). The big dots represent genes with significant values (p ≤ .05). All p‐values were adjusted by false discovery rate (FDR)

The McDonald and Kreitman test showed a low signal of positive selection with neutrality index (NI) of 0.91 (Fisher's test: p = 9.2E‐09). The analysis of all CDS separately across the genomes by MKT, after FDR adjustment, resulted in only two genes with a significant selection signal. The first gene (identified under positive selection) was a tyrosine‐type recombinase/integrase, located within an intact prophage (NI = 0.052, α = 0.9, adjusted‐p = 0.03). The second gene (identified under negative selection was brrF (Bartonella run‐off replication of host‐adaptability gene F; NI = 15.6, α = −14.6, adjusted‐p = 0.03; Figure 2b).

3.4. Pangenomic analysis

An average of 1780 CDS, 42 tRNA, 6 rRNA and 1 tmRNA were predicted in the genomes. The pangenome analysis showed that from 2464 total genes, 1391 (60.0%) were classified as core genes (in ≥99% of the strains), 55 (2.4%) as soft‐core genes (in ≥95 to <99% of the strains), 634 (27.4%) as shell genes (in ≥15 to <95% of the strains), and 403 (16.3%) as cloud genes (in <15% of the strains). The cloud gene pool was composed mainly by uncharacterized hypothetical proteins (37.5%) and phage associated genes (14.4%) followed by genes from other evolutionary origins, including genomic island proteins, transposon related proteins, T4SS‐related, NADH‐quinone related, adhesins, metabolic, transporters, filamentous haemagglutinin, and toxin‐antitoxin systems. The great majority of the cloud genes (67.5%) were unique to a single genome, followed by 98 (24.3%) found in two genomes and only 33 (8.2%) found in three genomes (Figure S2). From a total of 1446 core and soft genes, only 33 genes (2%) showed ANI values below 97% in comparison to the strain OE 1–1 genome (Figure S3).

3.5. Plasmid exploration

Plasmids were identified only in five strains (25%; BKF1, BKF2, BKR2, BKR8 and BKF9; PlasmidVerify log‐likelihood ratio > 65; all circular predicted by Unicycler) that belonged to three genotypes (A2, A3, and A5). All genotypes carried a circular plasmid‐contig of 29 kb length. The sequences showed 99.8%–100% ANI between strains and the B. krasnovii strain OE 1–1 plasmid (CP042965.1).

3.6. Bartonella prophages

The annotations allowed the identification of 120 prophages among the 14 B. krasnovii structurally‐validated genomes (Figure 3a). Most of the prophages were identified as “remnant prophage loci” (43.3%, 52/120), followed by “incomplete prophages” (25%, 30/120), “intact prophages” (17.5%, 21/120) and “questionable prophages” (2.5%, 3/120). Notably, the Bartonella gene transfer agent (BaGTA) locus was identified in all genomes (Figure 3a). The genetic distances between the prophages of the 14 B. krasnovii validated genomes, showed three distinct clusters on the PCA plot: cluster (i) composed by 15 intact prophages and one questionable prophage; cluster (ii) composed by BaGTA loci only (14); and, cluster (iii) composed by all remnant and incomplete prophages together with some intact (6) and questionable (2) prophages (Figure 3b). The ANI pairwise comparison between the prophages, ranged from 74.4–100%. The ANI between the intact and questionable prophages from cluster 1 was 95.6% (84.2–99.9%), cluster 2 was 87.3% (74.4–100%) and cluster 3 (composed only by BaGTA) was 99.7% (99.5–100%).

FIGURE 3.

Distribution of prophages and similarity across the study genomes. (a) Frequency and type of prophages identified among 14 Bartonella krasnovii genotypes. (b) Principal component analysis (PCA) of the identified prophages based on nucleotide similarities. Dots are coloured according to the prophage type

3.7. SV identification and effect estimation

From the 14 structural‐validated genomes, an average of 51 SV events were identified per genome in comparison to the OE 1–1 strain (Table 2, Figure 4a). Accordingly, all but one genome showed different profiles than strain OE 1–1 in their PFGE restriction pattern. Deletions and insertions, namely loci (≥1000 bp) without mapped reads from the studied strains or OE 1–1 strain genome, respectively, were the most frequent events identified (Table 2). No SV events were identified in the plasmids (Figure S4).

FIGURE 4.

Pairwise alignments of each variant and OE 1–1 genome, multiple alignments of B. krasnovii genotypes and phylogenetic nucleotide‐based tree. (a) Mauve alignment plots of each of the 14 B. krasnovii genotypes against the OE 1–1 B. krasnovii. (b) Mauve multiple alignments of B. tribocorum, B. krasnovii OE 1–1 genome and 14 B. krasnovii genotypes. The bars corresponding to the genomes were coloured according to the number of conserved segments or locally collinear blocks (LCBs) identified by the genome alignments, following a rainbow palette. The comparison features were coloured by shades of red and blue according to the length of the LCBs. The direct comparisons are coloured in red and the reversed ones in blue. (c) Phylogenetic ML tree of the nucleotides sequencing of the genomes assemblies, comprising the B. krasnovii OE 1–1 and 14 B. krasnovii genotypes, using bootstrap for supporting of clades and B. tribocorum as an outgroup. The three supported SV clades are coloured on the nucleotide‐based tree

The mauve‐based multiple alignment (Figure 4b) of 23 locally collinear blocks (LCB) (Darling et al., 2004, 2010), did not enable a conclusive Bartonella phylogeny based on the SVs. The phylogentic tree based on the SVs was poorly resolved and could support only three clades with high bootstrap values (Figure S5). The nucleotide‐based maximum likelihood (ML) phylogenetic tree based on 2042 orthologous groups (predicted by Microbializer; see Methods S1) showed high bootstrap values, except for the basal splits and the placements of BKF7, BKR3 and BKF3 strains (Figure 4c). The three clades supported by significant bootstrap values in the SV‐based tree showed the same topology of the supported clades of the nucleotide‐based tree. The SV tree suggests that strains BKR3 and BKF3, whose placement is poorly resolved in the nucleotide‐based tree, are closely related to strains BKF5, BKR6, BKR7 and BKR5 (Figure 4c, Figure S5).

The exploration of the flanking areas of the identified rearrangements showed that prophages were the loci mostly associated with the events (Figure 5). Accordingly, 48% (15/31) and 42% (16/38) of the inversion and translocation events were associated with prophages, respectively. Similarly, 62% (31/50) of insertions and translocations co‐occurring events were also associated with prophages. Uncharacterized hypothetical proteins were the second main element associated with 26% (8/31) of inversions, 34% (13/38) of translocations and 24% (12/50) of co‐occurring events, followed by association with 14 different genes. Forty percent (113/284) and 56% (178/317) of the insertion and deletion events were also found to be associated with prophages, respectively (Table S4). Deletion and insertion events were also noticed among virulence factors associated genes, being 7.5% of the total events associated with trw genes, followed by 3.4% associated with Bartonella adhesins associated genes (Figures S6 and S7). Regarding the association of virulence factors associated genes with inversions and translocations, only one inversion event was flanked by trwJ4 and trwI4 genes in one sample (BKR1) solely.

FIGURE 5.

Genes associated with SV events. Percentage of genes associated with SV events accessed through the analysis of the SVs flanking regions

The distribution of SV (especially inversions) and prophage insertions shows that most events occurred near the terminus (ter) region of replication of the chromosome (ter‐region) (Figure 6a and b, Figure S8). Notably, while “intact” prophages were distributed in three different regions across the chromosomes, and the BaGTAs were located out of the ter‐region, the majority of “remnant” and “incomplete” prophages were found in the ter‐region (Figure 6b).

FIGURE 6.

Distribution of all types of SV events and prophages across the 14 B. krasnovii genotypes. (a) Distribution of SV types across the genomes. (b) Distribution of prophage copies across the genomes

The SV events caused an effect in the balance level of some chromosomes (i.e., ori‐to‐ter distance). Accordingly, six genomes were classified as unbalanced, with up to 5.8% shifts between their theoretical middle position and the ter‐region, representing a difference of up to 126 kb (Table 3). Remarkably, the highest shifts were recorded among the A2 genotypes. From 105 essential genes analysed, only genes found near the ter‐region showed a re‐distribution between leading and lagging strands (Table S5). These genes were the gyrA, arginine‐‐tRNA ligase, aspartate‐‐tRNA ligase, proline‐‐tRNA ligase, tyrosine‐‐tRNA ligase, 30S ribosomal protein S4 and threonine‐‐tRNA ligase. Notably, the total number of essential genes in the leading and the lagging strands per genome was always found the same (78 and 27, respectable, Table 3). SVs also affected the distance between essential genes, increasing gene‐to‐gene distance in up to 120 Kb (Table S5).

3.8. SV events in Bartonella quintana

The comparison of five B. quintana genomes isolated from different hosts and geographical locations showed only one SV event, that is, a large inversion in the genomes of the strains MF1‐1, Korn and NCTC (Figure 7). The inverted segment was of approximately 700 Kb long and was flanked by an excinuclease subunit B and ribonuclease genes, in addition to repeated regions located upstream and downstream of the inverted event. Interestingly, Phaster annotation tool identified two small (6.7 and 7.1 kb) regions classified as “incomplete” prophages within the five B. quintana genomes studied (four genomes presented both regions, while one harbored only one). However, these regions contained a maximum of two genes with clear phage origin, namely genes showing similarities to the gp245 and gp59 proteins from the bacteriophage T4, and minor tail and baseplate wedge subunit related genes.

FIGURE 7.

Whole genome comparison of five different strains of Bartonella quintana. Pairwise alignments of five strains of B. quintana. The bars corresponding to the genomes were coloured according to the number of conserved segments or locally collinear blocks (LCBs) identified by the genome alignments, following a rainbow palette. The comparison features were coloured by shades of red and blue according the length of the LCBs. The direct comparisons are coloured in red and the reversed ones in blue. The alignment is based on SVs identified by progressiveMauve

4. DISCUSSION

This study offers a close look at the extent of the genomic diversity of co‐circulating wild B. krasnovii genotypes in discrete ecological niches in nature. The results presented here confirm the outstanding genomic structural diversity of Bartonella genomes with a conserved nucleotide variation and support the association of the former with prophage elements (Gutiérrez, Markus, et al., 2018).

SVs were previously observed among the zoonotic cat associated Bartonella, B. henselae (Lindroos et al., 2006), and the rodent‐associated Bartonella, B. grahamii (Berglund, Ehrenborg, et al., 2010a; Berglund, Ellegaard, et al., 2010b). The present study broadens and complements the knowledge on this subject by a detailed characterization of the genomic content of genotypes isolated at the same time and location, from two small communities of gerbils and their associated fleas. Particularly, this study reveals the implication of prophages and SVs at the ter‐region of the B. krasnovii chromosome, as a plastic zone for genomic rearrangements and prophage integration, reinforcing the evident association of prophages and SV events in rodent‐associated Bartonella. Furthermore, the evidence of six unbalanced genomes, and a strand‐shift in some essential genes near the ter‐region, suggests that some SV, probably driven by prophages, may lead to instability in Bartonella genomes. Given the differences in replication fidelity between the leading and the lagging strands, this observation may suggest a different rate of evolution in essential genes between different isolates (Fijalkowska et al., 1998).

The structural genomic diversity has been identified and well characterized in other bacterial species including Escherichia coli, Salmonella typhi, Staphylococcus aureus, Campylobacter jejuni, Mycobacterium spp. and other species (Cui et al., 2012; Garcia‐Betancur et al., 2012; Iguchi et al., 2006; Liu et al., 2006; Scott et al., 2007). Particularly, high rates of genome rearrangements were identified in Yersinia spp. (Darling et al., 2008). Yersinia pestis, like Bartonella, is a flea‐borne bacterium, that is constantly challenged to adapt to totally different niches. The genomic plasticity of this bacterium allowed the authors to perform a phylogenetic characterization, based on the SV data, of six genomes of Yersinia pestis, two Yersinia pseudotuberculosis and one genome of Yersinia enterocolitica (Darling et al., 2008). Contrary, the SVs‐based phylogeny of B. krasnovii in our study was not fully conclusive, while the nucleotide‐based tree was well supported, probably due to the fact that the phylogeny was done based on one single species, isolated from the same area at the same period, thus the events still are in progress to become fixed in the population. Studies involving isolates from small mammal communities, as the one presented here, represent an important addition of data to portray the ongoing genomic evolution in populations under constant natural selection. Under such conditions, confounding factors that could under‐ or overestimate the findings (e.g., different geographical locations or period of collection) are constrained, enabling a detailed genomic characterization, followed by a mechanistic investigation.

Rodent‐associated Bartonella constitute a polymorphic group of species in nature (Gutiérrez et al., 2014, 2015; Gutiérrez, Cohen, et al., 2018). This has been demonstrated by the characterization of hundreds of strains, sequenced from wild rodent communities (Berglund, Ehrenborg, et al., 2010a; Berglund, Ellegaard, et al., 2010b; Gutiérrez et al., 2014, 2015; Gutiérrez, Cohen, et al., 2018; Morick et al., 2010; Paziewska et al., 2011). Interestingly, the present study shows that, although the genomes of B. krasnovii contain regions with greater alternative allele frequency, point mutations were responsible for only 0.3% of the genomic variation, with only two genes showing significant signal of selection. Contrarily, structural diversity was observed in all genotypes, with an average of 51 SVs per genome recorded. The SV effect was initially noticed in the patterns of the SmaI‐restriction PFGE analysis, which correlated with several SV events identified through the comparison of each genome with the OE1‐1 strain genome. Moreover, the predicted phylogeny based on SVs showed agreement with the nucleotide tree in three supported clades, and enabled the support of one clade split in the nucleotide‐based tree. It is evident that SVs occur frequently in these genomes, affecting the overall Bartonella phylogeny. They may ultimately promote competitive interactions, resulting (in a later stage) in the selection of adaptive gene/nucleotide genotypes (Gutiérrez, Markus, et al., 2018) and in the speciation of Bartonella.

The genome plasticity observed in this study among B. krasnovii genotypes exceeded that of other Bartonella spp. previously investigated. A study that focused on B. grahamii showed high structural variation compared to nucleotide diversity (SNPs); however, the strains were isolated from different continents and from three different host families (Berglund, Ellegaard, et al., 2010b). Contrarily, another study based also on B. grahamii strains, but isolated from two different rodent families and collected from three near geographic locations (less than 30 kms apart) showed both low rates of SNPs and structural diversity (Berglund, Ellegaard, et al., 2010b). Structural variations were also observed among Bartonella henselae; however, in lower rates (106 events in 38 strains), including strains isolated from four different continents and collected from humans and cats (Lindroos et al., 2006). In the present study we identified a total of 702 SV events in 14 structurally‐validated genomes in comparison to the OE1‐1 genome. Remarkably, these strains were isolated from sympatric rodents that inhabit a restricted location. However, the higher diversity observed in our study might be a result of the smaller spatial scale. It is possible that in the comparison between Bartonella isolates across distant geographical locations and host groups, genetic drift played an additional role, reducing the observed genetic diversity.

In our study, most of the identified SV events were associated with prophages, confirming the implication of these elements in the onset of SVs, as previously shown in an in vitro evolutionary experiment with B. krasnovii (Gutiérrez, Markus, et al., 2018) and in other bacterial genera (Iguchi et al., 2006; Scott et al., 2007; Shen et al., 2020). Structural variations are a result of homologous and nonhomologous recombination events (Ottaviani et al., 2014; Treangen et al., 2009), that rather than point mutations, were previously suggested as the main source of genetic diversity among Bartonella spp. (Berglund, Ehrenborg, et al., 2010a; Berglund, Ellengaard, et al., 2010b; Gutiérrez, Cohen, et al., 2018; Guy et al., 2012; Paziewska et al., 2011). In E. coli, prophages play a key role in the host recombination events, shaping and affecting their evolvability (Bobay, Touchon, et al., 2013b). Prophages can employ the bacterial machinery to mediate their own recombination, replication and packing or impose their own recombination machinery by inhibiting the RecBCD enzyme subunit of the most important pathway of homologous recombination in the host (Bobay, Touchon, et al., 2013b). Interestingly, the prophages that confer the ability to promote recombination present mosaic genomes with a diverse gene content (Bobay et al., 2014). This pattern of mosaicism is the major outcome of recombination events due to HGT (Bobay, Touchon, et al., 2013b; Dion et al., 2020). The prophages identified in our study evidenced two clear genetic clusters, suggesting accumulation of mutations and participation of HGT in the sequence composition. In fact, one of the clusters harbored most of the incomplete and remnant prophages, suggesting prophage degradation across the genomes. In addition to driving SVs, the role of Bartonella prophages in promoting recombination deserves further investigation.

We found a high concentration of prophages and SVs in the ter‐region of the chromosome, suggesting that this region could be a plastic zone for integration and SV formation. Rearrangements at the ter‐region were already reported in other Bartonella species (Alsmark et al., 2004; Lindroos et al., 2006; Segers et al., 2017) and in a broad range of different bacterial species (Bowden et al., 2016; Esnault et al., 2007; Hoff et al., 2018; Kresse et al., 2003; Repar & Warnecke, 2017). Replication in circular bacterial chromosomes occurs bidirectionally around the chromosome, starting from the ori‐region and finishing in the ter‐region, located about 180 degrees apart (Kothapalli et al., 2005). When the replication forks encounter transcriptional elements, conflicts may happen activating the DNA repair mechanisms and consequently recombination events, driving instability and SVs near this region (Merrikh et al., 2012). In our study, the majority of prophages and prophage remnants were located in the ter‐region, and were associated with SVs. Prophages were suggested to provide conditions for bacteria to improve their fitness (i.e., protection against phages, provide virulence factors and/or antibiotic resistant genes) (Bobay et al., 2014). To keep this positive relationship, prophages may not integrate randomly, thus facilitating the coevolution of phages and bacteria to endure selection for integration sites (Bobay, Rocha, et al., 2013a). Temperate prophages show tropism to the ter‐region of the chromosome where less abundance of coding genes is present, reducing the impact on the genome organization and bacterial fitness (Bobay, Rocha, et al., 2013a). However, in our study six strains presented unbalanced genomes, and impacted the distribution of some essential genes that had changed the strand orientation. These changes may affect the translation pattern in the bacteria, being evidence that prophages have the potential to cause the disruption of the physical balance in Bartonella genomes. One of the major outcomes of genome instability is the phase variation that modulates the extension of expression of operons and genes that can lead to antigenic variation and consecutively resulting in new phenotypes (Darmon & Leach, 2014). In Salmonella typhi, the disruption caused by insertion sequences was demonstrated. The instability observed drove genome diversification, with the outcome of new phenotypes, being these events implicated in the evolution of the species (Liu et al., 2006).

Bartonella quintana, the agent of trench fever, presents a low degree of genomic polymorphism, that is probably due to a recent genetic drift caused by host switch (Berglund, Ehrenborg, et al., 2010a; Foucault et al., 2006). It has been proposed that this species has lost its single prophage during the evolution (Alsmark et al., 2004). Based on the latter, B. quintana served as a prophage‐free control for our study, to further investigate the association of SVs and prophages. In fact, the genome comparison of five different strains of B. quintana showed only one large inversion, demonstrating low degree of genomic plasticity compared with B. henselae (Lindroos et al., 2006) and rodent‐associated Bartonella (Berglund, Ehrenborg, et al., 2010a; Berglund, Ellengaard, et al., 2010b ; Gutiérrez, Markus, et al., 2018; Gutiérrez, Cohen, et al., 2018). However, prophage annotation revealed two small prophage‐related regions in the B. quintana genomes. These genomic loci may represent prophage vestiges or remnants, and not truly temperate prophages. The inspection of B. quintana genomes complemented our B. krasnovii findings and conclusion that prophages are SV drivers and consecutively implicated in the diversity of Bartonella genomes.

In conclusion, the present study supports the positive association of SVs and prophages in the generation of diversity in rodent‐associated Bartonella in nature, reflecting the dynamics of Bartonella lifestyle and the direct implication of these events in the adaptation and evolution of this genus.

AUTHOR CONTRIBUTIONS

The research was designed by Keyla Carstens Marques de Sousa, Ricardo Gutiérrez, Yaarit Nachum‐Biala, Shay Covo and Shimon Harrus. The collection and sampling of animals were performed by Hadas Hawlena. The research was performed by Keyla Carstens Marques de Sousa, Ricardo Gutiérrez and Shimon Harrus. The bioinformatic support was done by Dayana Yahalomi, Tali Shalit, Barak Markus and Haroldo Henrique de Rezende Neves. The phylogenetic analysis was performed by Einat Hazkani‐Covo. The analytic tools were assisted by Evgeniya Marcos‐Hadad and Shay Covo. Data analysis and writing of the article were carried out by Keyla Carstens Marques de Sousa, Ricardo Gutiérrez and Shimon Harrus.

CONFLICT OF INTEREST

The authors confirm no conflicting interests.

Supporting information

Data S1

de Sousa, K. C. M. , Gutiérrez, R. , Yahalomi, D. , Shalit, T. , Markus, B. , Nachum‐Biala, Y. , Hawlena, H. , Marcos‐Hadad, E. , Hazkani‐Covo, E. , de Rezende Neves, H. H. , Covo, S. , & Harrus, S. (2022). Genomic structural plasticity of rodent‐associated Bartonella in nature. Molecular Ecology, 31, 3784–3797. 10.1111/mec.16547

DATA AVAILABILITY STATEMENT

All genomes included in this study have been deposited in GenBank database under the BioProjet: PRJNA801099, genome accession numbers: CP093033 ‐ CP093046.