Genomic analysis of a Raoultella ornithinolytica strain causing prosthetic joint infection in an immunocompetent patient

We sequenced the genome of Raoultella ornithinolytica strain Marseille-P1025 that caused a rare case of prosthetic joint infection in a 67-year-old immunocompetent male. The 6.7-Mb genome exhibited a genomic island (RoGI) that was unique among R. ornithinolytica strains. RoGI was likely acquired by lateral gene transfer from a member of the Pectobacterium genus and coded for a type IVa secretion system found in other pathogenic bacteria and that may have conferred strain Marseille-P1025 an increased virulence. Strain Marseille-P1025 was also able to infect, multiply within, and kill Acanthamoaeba castellanii amoebae.

The thirteen studied strains exhibited a pangenome and a core genome of 9,815 and 3,822 genes, respectively (Fig. 1). Figure 1 shows the dispersion of the pangenome of R. ornithinolytica. The phylogenetic analysis based on accessory genes clustered strains Marseille-P1025 and NBRC 105727. demonstrated that 23 of the Marseille-P1025-specific genes were grouped in a 11,473-kb genomic island located in scaffold 21 (Fig. 2). This genomic island, which we named RoGI, exhibited a G + C content of 49.5% (vs 55 to 56% for the genomes of R. ornithinolytica strains, Table 1), and was absent from other R. ornithinolytica (Fig. 2). Of these 23 genes, nine coded type IVa secretion system proteins ( Table 2), including seven VirB proteins (VirB 4 to 11, Table 2) and two proteins related to bacterial conjugation, including a type IVa secretion system conjugative DNA transfer protein and a conjugal transfer protein (Table 2). Moreover, the RoGI island contained a gene coding a second CP4-57 prophage integrase (intA) ( Table 2). The genes coding the VirB1, VirB2, VirB3 and VirD4 proteins were identified at other locations of the genome from strain Marseille-P1025, thus supporting the assumption that this strain had a complete and putatively functional type IVa secretion system.
In addition, seven (7.4%) proteins were involved in intracellular trafficking and secretion, seven (7.4%) in replication and repair, four (4.2%) in cell wall/membrane/envelop biogenesis and four (4.2%) had a general functional prediction only (Table 2). Finally, three genes coded integrases including a CP4-57 prophage integrase (intA), two genes coded integrating conjugative element proteins, and two genes coded a CP4-57 prophage regulatory protein AlpA and a transposase, respectively ( Table 2). ClonalframeML and Phylogenetic Analysis. To verify whether the RoGI island was acquired by lateral gene transfer, we used a recombination and phylogenetic analysis. Figure 3 shows the recombination events of external origin marked by a dark blue horizontal line. ClonalFrameML identified 170 recombination events on all branches of the clonal genealogy, including 23 recombination events in the genome of strain Marseille-P1025 (Fig. 3). These 23 regions appeared to be possible recombination hotspots (Fig. 3). Three of these recombination hotspots (red circle) were located in scaffold 21 of strain Marseille-P1025 (located from nucleotides 5,425,000 to 5,612,500) (Fig. 3), close to the RoGI island that coded the type IVa secretion system (located from nucleotides 5,504,317 to 5,515,790, Fig. 3). The phylogenetic analysis of nucleotide sequences from the RoGI island supported the assumption that it was acquired by lateral gene transfer by identifying close phylogenetic neighbours in Pectobacterium atrosepticum strain JG10-08, Pectobacterium sp. strain SCC3193, two Pectobacterium wasabiae (strain CFBP-3304 and strain RNS08.42.1 A), Cedecea neteri strain ND14b and Citrobacter amalonaticus strain Y19 (Fig. S1).   12 . In order to confirm that strain Marseille-P1025 elaborates a conjugative pilus, electron microscopy was performed on cells after 24 h of incubation. Figure 4 shows that strain Marseille-P1025 possesses a conjugative pilus.

Interaction of R. ornithinolytica with A. castellanii trophozoites. Acanthamoeba castellanii
is a free-living amoeba that has previously been used as a eukaryote model to study the virulence of pathogenic microorganisms, including Acinetobacter baumannii, mycobacteria and streptococci [13][14][15][16] . To determine whether strain Marseille-P1025 can multiply in eukaryotic cells, triplicate co-culture assays were performed with Acanthamoeba castellanii amoebae. Raoultella ornithinolytica strain P2310, isolated from the feces of a healthy individual, was used as a control for this experiment (Figs 5 and 6). We observed that the numbers of both uninfected and infected A. castellanii trophozoites incubated into PAS at 32 °C decreased over time. However, the mean percentages of remaining live amoebae at day 3 were 45.0+/−1.33%, 19.69+/−1.44% and 27.83+/−4.82% for uninfected amoebae, amoebae infected with strain P2310 and amoebae infected with strain Marseille-P1025, respectively. Therefore, the number of infected amoebae decreased significantly more than those of uninfected amoebae (p < 0.05) in presence of both R. ornithinolytica strains (Figs 5A, S2), and strain Marseille-P1025 caused a higher amoebal mortality than strain P2310, although this difference was not statistically different (p = 0.17). We also evaluated the numbers of CFUs obtained from intra-amoebal bacteria at H0 and H72 of co-culture (Figs 5B, S2). At H0 and H72, a mean 3.57 × 10 5 CFUs/mL and 9.67 × 10 7 CFUs/mL, respectively, were cultivated for strain Marseille-P1025 versus 5.73 × 10 5 CFUs/mL and 9.33 × 10 7 CFUs/mL, respectively, for strain P2310. The growth rate of both strains (270+/−97.5 and 174+/−29.3, respectively) was significantly higher for strain Marseille-P1025 (p < 0.05). Hence, these experiment demonstrated that strain Marseille-P1025 exhibited a higher pathogenicity for amoebae than the control strain (Fig. 5B). To confirm these observations, we examined bacteria within amoebae by Gimenez staining. Optical microscopy observations were consistent with the CFU evaluations. We observed that after 5 h of co-culture, most A. castellanii cells were infected by R. ornithinolytica strain Marseille-P1025 (Fig. 6A). Not only was R. ornithinolytica strain Marseille-P1025 able to survive within A. castellanii, but it began to multiply after 24 hours of co-culture (Fig. 6B). At day 2 of co-culture, strain Marseille-P1025 continued to multiply within amoebae. Furthermore, at day 3 of co-culture, infected amoebae started to lyse (Fig. 6D) whereas strain Marseille-P1025 kept multiplying. The lysis of A. castellanii amoebae was complete after 5 days of co-culture with bacteria (Fig. S3). We also observed that R. ornithinolytica survived in PAS medium without amoebae but did not multiply from day 0 to day 3.

Discussion
Infections due to R. ornithinolytica are under-reported, possibly because this bacterium is difficult to identify using conventional phenotypic methods 17 . Raoultella ornithinolytica is currently regarded as an emerging hospital-acquired infection agent, particularly after invasive procedure 10 . Few pathogenic factors are recognized in R. ornithinolytica compared to other members of the family Enterobacteriaceae 10 . These include the ability to adhere to human tissues, to form biofilms in urinary catheters and to convert histidine to histamine in scombroid fishes, thus causing redness and flushing of the skin 10 .
By comparing the genome of strain Marseille-P1025 that had caused a chronic prosthetic joint infection in an immunocompetent patient, to those of other R. ornithinolytica strains, we identified a unique 11-kb genomic island (RoGI) among R. ornithinolytica strains. This genomic island coded nine proteins from the type IVa  Figure 2D shows an alignment of all four compared genomes. Common and specific genes are displayed in orange and red, respectively.
SCIEnTIFIC REPORtS | (2018) 8:9462 | DOI:10.1038/s41598-018-27833-z secretion system (SS) (Fig. 2, Table 2), four proteins related to bacterial conjugation and two integrases. In addition, the genome from strain Marseille-P1025 contained another four components of the type IVa SS, which suggests that this system was complete. The presence of a conjugative pili and the ability of strain Marseille-P1025 to invade, survive and multiply in an amoeba (Acanthamoeba castellanii) confirms the presence of type IVa secretion system.
The type IVa SS is involved in various mechanisms of bacterial pathogenesis such as the transfer of Agrobacterium tumefaciens oncogenic DNA into plants leading to tumours 18 . The Type IVa SS is also involved in conjugation and thus plays a crucial role in genomic plasticity, notably by enabling the transfer of plasmids carrying antibiotic resistance or virulence genes among pathogenic bacteria 19 . In addition, conjugation systems may also contribute significantly to the development of infections by promoting surface pili adhesion-mediated attachment, colonization and biofilm formation 20 . It is also reported that the type IVa secretion system, particularly the virB operon, is essential for survival and intracellular multiplication 21 . Siddiqui et al. have shown that pathogenic bacteria, but not those who are weakly or not pathogenic, can survive within A. castellanii cysts 16 . The ability of R. ornithinolytica to grow and cause cytopathic effect in A. castellanii seems to be correlated with its virulence 14,15 . We demonstrated that strain Marseille-P1025 was not only able to survive within amoebae but could also multiply more efficiently than the control strain and kill amoebae, supporting its virulent behavior.
We also detected the presence of three specific genes carried by the CP4-57 prophage, including two integrases (intA) and alpA, a transcriptional regulator of intA 24 . The IntA integrase has been shown to intervene in biofilm formation. In E. coli, the deletion of the intA gene reduces early biofilm formation 24 whereas the increased synthesis of IntA leads to excision of the CP4-57 prophage 24 which, in turn, increases biofilm formation 24 . Therefore, IntA may play a role in biofilm formation in strain Marseille-P1025, thus facilitating its adhesion to foreign material such as the patient's joint prosthesis.
In conclusion, R. ornithinolytica strain Marseille-P1025, that caused a rare case of chronic prosthetic joint infection in a 67-year-old immunocompetent male, exhibited a complete type IVa secretion system that was unique among R. ornithinolytica strains and was able to infect, multiply within, and kill amoebae. These properties may explain its particular virulence. In addition, this type IVa SS was mostly coded by a genomic island (RoGI) that was probably acquired by lateral gene transfer from Pectobacterium species.

DNA extraction and Genome sequencing. Strain Marseille-P1025 was cultivated on Columbia agar
(bioMérieux, Marcy-l'Etoile, France) at 37 °C in aerobic atmosphere for 24 hours. Then, after a pre-treatment with lysozyme at 37 °C for 2 hours, the DNA was extracted using an EZ1 biorobot and the EZ1 DNA tissue kit (Qiagen, hilden, Germany). The elution volume was 50 µL. Genomic DNA (gDNA) was quantified by a Qubit assay with the high sensitivity kit (Life technologies, Carlsbad, CA, USA) to 8 ng/µl, prior to being sequenced on a MiSeq sequencer (Illumina, San Diego CA, USA) with the Paired-End and barcode strategy in order to be mixed with 20 other projects constructed according to the Nextera XT library kit (Illumina).
One ng of gDNA was used as input and tagmented for the fragmentation step. Then, limited cycle PCR amplification completed the tag adapters and introduced dual-index barcodes. The libraries were then normalized on specific beads according to the Nextera XT protocol (Illumina), pooled into a single library and then loaded onto the reagent cartridge. Automated cluster generation and Paired-End sequencing with dual index reads was performed in a single 39-hour run in a 2 × 251-bp.
Total information of 9.8 Gb was obtained from a 1,165 K/mm 2 cluster density with 88% (18,993,000 clusters) of the clusters passing quality control filters. Within this pooled run, the index representation of R. ornithinolytica strain Marseille-P1025 was determined to be 5.51%. The 1,046,713 Paired-End reads were filtered according to the read qualities. Genome annotation and comparison. The sequencing reads were assembled using the A5 assembler 25 .
Then, a step of finishing was done using the Mauve software 26 and CLC bioserver. After assembly and finishing, the genome size was 5.6 Mb. Open reading frames (ORFs) were predicted using the Prodigal tool (http://prodigal. ornl.gov) with defaults parameters. The prediction of protein function was performed by searching against the GenBank database using BLASTP algorithm 27 . Functional classification of gene families (COG ID and Letters) was obtained using COGnitor against the COG database 28 . tRNAs and rRNAs were detected using tRNAscan-SE v.1.21 29 and RNAmmer v.1.2 30 , respectively. The presence or absence of plasmids was verified both by searching the gene annotation for any plasmid-related gene and by mapping all contigs against previously published Raoultella sp. plasmid sequences.
We compared the genome sequence of R. ornithinolytica strain Marseille-P1025 to those of other strains of this species found in public databases. As of August 30 th , 2016 13 R. ornithinolytica genomes were available in public databases. Of these, we used 12 genomes for comparative analysis and excluded that of strain S12 due to its insufficient quality. The twelve comparator genomes were those from strains 10-5246 (AGDM00000000), NBRC 105727 (BCYR00000000), B6 (CP004142), A14 (CP008886), CMUL058 (CVRH00000000), TNT (JHQH00000000), 2-156-04_S1_C1 (JNPC00000000), 2-156-04_S1_C2 (JNPD00000000), 811_RORN (JURX00000000), BAL286 (JXXF00000000), CB1 (LFBW00000000) and Yangling I2 (CP013338). All genomes were re-annotated using the Prokka software, version 1.11 31 . Comparisons between all selected genomes were done using Roary, a tool that rapidly builds large-scale pangenomes 32 , with a blast identity cut-off of 97% for the comparison between R. ornithinolytica species. In addition, Roary identified the specific and missing genes from strain Marseille-P1025. Specific genes were checked by BLASTP and TBLASTN against the other studied genomes. Missing genes were checked by TBLASTN against the genome of strain Marseille-P1025, using a coverage and identity of 60% and 80% as thresholds, respectively, as described by Kuenne et al. 33,34 . Easyfig. 35 was used to visualise the coding regions and colour the specific genes of strain Marseille-P1025.
Recombination and Phylogenetic analysis. The genome of strain Marseille-P1025 was used as a reference for whole-genome alignment 36 using Mugsy 37 . Then, a phylogenetic tree based on whole genome sequence alignment was done using the FastTree software 38 and the maximum likelihood method (Fig. 1). ClonalFrameML was used to search recombination hotspots in R. ornithinolytica genomes by analyzing both the whole genome alignment and the phylogenetic tree 39 .
Unique sequences were detected by a BLASTN search for homologous sequences and multiple sequence alignment using the Mafft software algorithm 40 . Phylogenetic analysis of these unique sequences was performed using MEGA version 7 41 and the maximum likelihood (ML) algorithm, with 1,000 bootstrap replicates. Electron microscopy. Electron microscopy was performed with detection Formvar coated grids. Forty 40 μL of bacterial suspension were deposited on a grid and incubated at 37 °C for 30 min, followed by a 10 sec incubation on ammonium molybdate 1%. Grids were then observed using a Morgagni 268D transmission electron microscope (Philips) at an operating voltage of 60 kV.

Culture of R. ornithinolytica and A. castellanii.
Raoultella ornithinolytica strain P2310, isolated from the feces of a healthy individual, was used as a control in co-culture experiments. Raoutella ornithinolytica strains Marseille-P1025 and P2310 were grown on 5% sheep blood-enriched Columbia agar (BioMérieux) at 35 °C for 24 hours in anaerobic atmosphere. Bacteria were then harvested, centrifuged at 4,000 × g during 5 minutes, washed twice and suspended in Page's modified Neff 's amoeba saline (PAS). The PAS medium was prepared as follows: solution A (for 100 mL of sterile distilled water), 1.2 g NaCl + 0.04 g MgSO 4 .7H 2 O + 1.42 g Na 2 HPO 4 + 1.36 g KH 2 PO 4 ; solution B (for 100 mL of sterile distilled water), 0.04 g CaCl 2 .2H 2 O; PAS solution −10 mL of solution A + 10 mL of solution B + 980 mL of sterile distilled water). The inoculum density was determined by the McFarland method.
Acanthamoeba castellanii strain Neff (ATCC 30010) was grown in 175 cm² culture flasks containing 30 mL peptone-yeast extract-glucose (PYG) at 28 °C. When a monolayer was formed, A. castellanii trophozoites were harvested by shaking the flasks and centrifuged at 500 × g for 10 min. The pellet was suspended in 30 mL PAS medium.
The quantification of the A. castellanii population was performed using a KOVA ® slide cell counting chamber.
Co-culture experiments. The amoebal trophozoite suspension (5 × 10 5 amoeba/mL) was inoculated in 24-well plates and allowed to adhere for 30 minutes at 32 °C. Then, R. ornithinolytica suspensions were inoculated on amoebae to achieve ratios of infection of 10 R. ornithinolytica cells/amoeba. As controls, A. castellanii and R. ornithinolytica strains were cultivated separately in PAS. After incubation for 2.5 h at 32 °C under a 5% CO 2 atmosphere, the co-culture wells were washed three times with PAS to remove any remaining extracellular or adherent bacteria. Incubation at 32 °C under 5% CO 2 was then performed for 3 days. The presence of viable Raoultella inside amoebal trophozoites was documented by sub-culturing at 0, 24, 48 and 72 h of incubation. For each time point, the A. castellanii monolayer from a well was lysed by three passages through a 25-gauge needle. Serial dilutions of the lysate were carried out, plated onto COS medium and incubated for 2 days at 32 °C under anaerobic atmosphere to determine the numbers of intracellular R. ornithinolytica colony forming units (CFU). Multiplication rate of the bacterial invasion was calculated as follows: recovered R. ornithinolytica (CFU)/R. ornithinolytica (CFU) at time 0. The A. castellanii population was also monitored during the 3-day experiment: counting and viability check of amoebae, cultivated alone and in co-culture, was performed using KOVA ® slides after Trypan Blue 0.4% coloration (Sigma-Aldrich, Taufkirchen, Germany). All experiments were reproduced three times, each time in duplicate. The standard error of the mean (SEM) was used to evaluate the experiment value distribution. To compare the intra-amoebal growth of the two tested bacterial strains, we also estimated the dayly multiplication rate of bacteria.
The presence of R. ornithinolytica within amoebae was also monitored for 5 days by Gimenez staining 42 . The observation was performed with a LEICA DM 2500 LED microscope.

Statistical analyses.
Statistical analyses mentioned in this study were performed using the Student's t-test and Chi-square test, with a significance level of P inferior or equal to 0.05.
Nucleotide sequence accession numbers. The genome sequence from R. ornithinolytica strain Marseille-P1025 was deposited in GenBank under accession number FTLF01000000.