Evolution of the Quorum network and the mobilome (plasmids and bacteriophages) in clinical strains of Acinetobacter baumannii during a decade

In this study, we compared eighteen clinical strains of A. baumannii belonging to the ST-2 clone and isolated from patients in the same intensive care unit (ICU) in 2000 (9 strains referred to collectively as Ab_GEIH-2000) and 2010 (9 strains referred to collectively as Ab_GEIH-2010), during the GEIH-REIPI project (Umbrella BioProject PRJNA422585). We observed two main molecular differences between the Ab_GEIH-2010 and the Ab_GEIH-2000 collections, acquired over the course of the decade long sampling interval and involving the mobilome: i) a plasmid harbouring genes for blaOXA 24/40 ß-lactamase and abKA/abkB proteins of a toxin-antitoxin system; and ii) two temperate bacteriophages, Ab105-1ϕ (63 proteins) and Ab105-2ϕ (93 proteins), containing important viral defence proteins. Moreover, all Ab_GEIH-2010 strains contained a Quorum functional network of Quorum Sensing (QS) and Quorum Quenching (QQ) mechanisms, including a new QQ enzyme, AidA, which acts as a bacterial defence mechanism against the exogenous 3-oxo-C12-HSL. Interestingly, the infective capacity of the bacteriophages isolated in this study (Ab105-1ϕ and Ab105-2ϕ) was higher in the Ab_GEIH-2010 strains (carrying a functional Quorum network) than in the Ab_GEIH-2000 strains (carrying a deficient Quorum network), in which the bacteriophages showed little or no infectivity. This is the first study about the evolution of the Quorum network and the mobilome in clinical strains of Acinetobacter baumannii during a decade.


Results
Genomic and phenotypic features of the isolates of A.baumannii from Ab_GEIH-2000/2010 collections. Table 1 shows the results of whole-genome sequencing (WGS) studies undertaken as part of the GEIH-REIPI Spanish Multicentre Acinetobacter baumannii Study II 2000-2010, Umbrella BioProject PRJNA422585 (Bioproject Acc.Number in NCBI server). In total, 44.44% of the Ab_GEIH-2000 strains contained the AbATCC329/pMMCU3 plasmid harbouring bla OXA 24/40 ß-lactamase and abKA/abkB genes, in contrast to 100% of the Ab_GEIH-2010 isolates. Moreover, the MICs for the strains under study are shown in Table 2. The MICs of imipenem (64 mg/L) and meropenem (>128 mg/L) were the same for all Ab_GEIH-2010 strains, but not for the Ab_GEIH-2000 strains. Interestingly, the Ab_GEIH-2010 strains showed higher susceptibility than the Ab_GEIH-2000 strains to amikacin (resistance which is associated with aminoglycoside-modifying enzymes).
The phylogenetic tree from both collections (Ab_GEIH-2000/2010) is shown in the Fig. 1. We observed the genomic similarity between strains of Ab_GEIH-2000 versus the genomic similarity of Ab_GEIH-2010 isolates.
Moreover, comparison of the chromosomal sequences ( Fig. 2) of the Ab_ GEIH-2000 and Ab_GEIH-2010 collections of strains revealed the following: i) the number of similar proteins (core-genome) was higher in the Ab_GEIH-2010 strains than in the Ab_GEIH-2000 strains (3654 proteins relative to 3301 proteins); ii) of these proteins (core-genome), 239 were located in bacteriophages in the Ab_GEIH-2010 strains, compared with 21 in the Ab_GEIH-2000 strains; and iii) of the 51 proteins present only in the Ab_GEIH-2010 strains (accessory genome), 40 proteins (79%) were located in the bacteriophages. Finally, comparison of the plasmidic sequences indicated the presence of 17 proteins only in the Ab_GEIH-2010 isolates (AbATCC329/pMMCU3 plasmid harbouring genes encoding OXA 24/40 ß-lactamase and AbKA/AbkB proteins) 19 .

Response of A. baumannii
Ab105_GEIH-2010 strain (carrying of Ab105-1φ and Ab105-2φ bacteriophages) under the SOS response (MMC). The SOS response (MMC) produced differences in microarray expression of the genes from the two bacteriophages of the Ab105_GEIH-2010. Under stress conditions by SOS response we observed an overexpression of 5% and 30% of genes by Ab105-1φ and Ab105-2φ bacteriophages respectively (Umbrella BioProject PRJNA422585). Only three ORFs were expressed in bacteriophage Ab105-1φ, while 28 ORFs were expressed in bacteriophage Ab105-2φ (30.10%) (Table 3). Moreover, up to 40 minutes after the addition of MMC (to induce the SOS response), overexpression of ORF27 (methyltransferase-SAM or AdoMet-MTase) with a relative expression (RE) of 200 times and ORF06 (MazG-like protein) with an RE of 81.3 times that of Ab105-2φ was observed by qRT-PCR. After 40 minutes, the RE of ORF 93 (DNA polymerase UmuC) increased by 10 times (phage Ab105-2φ).
On the other hand, in the arrays results from the Ab105_GEIH-2010, there were other overexpressed bacterial genes involved in the SOS response under MMC (Umbrella BioProject PRJNA422585). Among them, we highlight several genes related to DNA repair with high overexpression (>10 fold): i) Glutathione S-transferase ii) Deoxyuridine 5 -triphosphate nucleotidohydrolase (dUTPase); iii) ParB-like protein; and iv) Thymidylate synthase.
Finally, the bacterial response after incubation with MMC is shown in Fig. 5. Bacterial lysis occurred in the presence of MMC but not in the absence of this compound. Finally, Fig. 5 also shows the characteristic morphology of the phages in the Siphoviridae family, with a long tail and an icosahedral capsid of diameter~ 60 nm (http://viralzone.expasy.org/).  Table 4. Activation of the QS system by the ROS response (presence of H 2 O 2 ) produced overexpression of the abaI gene in all of the Ab_GEIH-2010 strains. In the presence of the 3-oxo-C12-HSL molecule (used to induce inhibition of the QS system), all of the Ab_GEIH-2010 strains overexpressed the aidA gene, while only 55.5% of the Ab_GEIH-2000 strains overexpressed this QQ enzyme and 44.4% of these strains overexpressed the abaI gene. Therefore, the Ab_GEIH-2010 strains showed QS-deficient cells relative to the Ab_GEIH-2010 strains, which had a functional QS system.

Responses of Ab_GEIH-2000/2010 collections under stress conditions (3-oxo-C12-HSL and
Finally, the absence of surface motility profile was also homogeneous (QQ phenotype) in the Ab_GEIH-2010 strains relative to the heterogeneity associated with the presence of surface motility in some isolates of the Ab_GEIH-2000 collection (Fig. 6).  (Table 4). Moreover, a statistically significant increase in the infective capacity of the bacteriophages (Ab105-1ϕ and Ab105-2ϕ) was found in the only strain in the study that did not possess AidA, AbaI or AbaR proteins (Ab166_GEIH-2000), in the presence of the external 3-oxo-C12-HSL molecule (Fig. 7).

Discussion
Viruses that infect bacteria (bacteriophages) can influence the dynamics of the bacterial community, the evolution of the bacterial genome and the biogeochemistry of the ecosystem. However, the degree of influence differs depending on whether the bacteriophages establish lytic, chronic or lysogenic infections. The analysis of different ecosystems by comprehensive modelling will help provide a better understanding of the diverse lifestyles and ecological impacts of lysogens in nature 9,20 . In this study of the genomic evolution in 18 A. baumannii strains belonging to ST-2 clone isolated in the same ICU in 2000 (9 strains referred to collectively as Ab_GEIH-2000) and in 2010 (9 strains referred to collectively as Ab_GEIH-2010), we found that the Ab_GEIH-2010 strains harboured two conserved temperate bacteriophages (Ab105-1φ and Ab105-2φ), which displayed lytic activity and activated the SOS response (in the presence of MMC). Microarray assays revealed overexpression of three viral proteins associated with protection of the virus against bacterial attack during the lytic activation of the bacteriophages. Moreover, qRT-PCR studies confirmed the overexpression of the genes coding the three proteins, 40 min after induction with MMC: i) the SAM-dependent methyltransferases or AdoMet-MTases (ORF27 from Ab 105-2φ), which have been associated with protection of the viral genome of the host restriction enzymes from the bacteria 21,22 ; and ii) MazG (ORF06), which is a pyrophosphohydrolase enzyme located in bacteriophages that infect Burkholderia cenocepacia and in marine bacteriophages (especially cyanophages), thus facilitating viral infection in the environment 23 . In Escherichia coli, the MazG protein has been associated with a decrease in activity of the MazF toxin (MazEF toxin-antitoxin system), a defence mechanism that inhibits the spread of phage P1 24 ; and iii) the mutation-inducing UmuC protein may have significant implications for the evolution of virulence and antibiotic resistance 25,26 . Several are the studies where has been demonstrated by NGS analysis the high degree of variation in A. baumannii clinical isolates, including single nucleotide polymorphisms (SNPs) and large DNA fragment variations in the resistance island (RI) regions, the type VI secretion system (T6SS) and the bacteriophages 27,28 . One of the key determinants of the size, composition, structure and development of a microbial community is the predation pressure exerted by bacteriophages 29 . Bacteria have accordingly evolved a battery of anti-phage defence strategies 30,31 . Evidence that QS signalling may be involved in regulating the response to phages is increasing 32 . For example, in Escherichia coli in the presence of QS signals, i.e. N-acyl-L-homoserine lactones (AHLs), there is a significant reduction in levels of the phage receptor LamB 33 , which protects the bacterium against attack from the λ phage. Moreover, in Vibrio anguillarum, mutants that are permanently locked in a high-cell density state are almost completely immune to the phage KVP40, due to the QS-mediated downregulation of the OmpK receptor used by the phage [34][35][36] . In other pathogens such as Pseudomonas aeruginosa, the QS system has also been associated with populations evolving with bacteriophages 37,38 . It has recently been demonstrated that the presence of lysogenic bacteriophages acts as a powerful driving force for the selection of functional bacterial QS systems both in vitro and in vivo in Pseudomonas aeruginosa strains, by stabilizing bacterial cooperation and therefore virulence 38 . Interestingly, in all strains of the Ab_GEIH-2010 collection, a new QQ enzyme (AidA) and the AbaI protein were overexpressed in the presence of exogenous 3-oxo-C12-HSL (a QS inhibitor) and H 2 O 2 (ROS response), respectively. These results indicated the presence of a functional QS/QQ network in these cells 13 . These molecular features were not present in the Ab_GEIH-2000 strains (lacking or displaying low expression of the AidA protein as well as mutations or deletions in AbaR protein), hence showing a QS/QQ-deficient 38 , which was not infected by the Ab105-1φ and Ab105-2φ bacteriophages.
Interestingly, in a single strain which did not possess AidA, AbaI or AbaR, the low infective capacity of the bacteriophages (Ab105-1ϕ and Ab105-2ϕ) increased significantly (**P < 0.05, Student's t-Test) in the presence of the external 3-oxo-C 12 -HSL molecule. The AidA enzyme was previously described by our research group in clinical strains of A. baumannii (non surface motile strains) associated with bacterial competition, as it is capable of hydrolyzing the signalling molecules (including 3-oxo-C12-HSL) mediated between species 13 . Weiland and colleagues described several QQ enzymes displaying hydrolytic activity against AHLs and AI-2 signals 39,40 . Therefore, several authors propose the use, together with lytic phage therapy, of QS modulators, i.e. quorum quenchers, to decrease the phage resistance mediated by the QS system 38,41,42 .
Another molecular difference between the strains in the two collections (Ab_GEIH-2000 and Ab_GEIH-2010) was the presence/absence of a plasmid harbouring genes encoding the OXA 24/40 ß-lactamase (resistance to carbapenems) and AbkA/AbkB proteins. This AbkA/AbkB toxin-antitoxin (TA) system was described by Mosqueda et al. in 2014 19 in all plasmids carrying the bla OXA 24/40 ß-lactamase gene that confers resistance to carbapenems in strains of A. baumannii in Spanish hospitals. In plasmids, TA systems have been associated with plasmid stabilization 43,44 , although it has been hypothesised that TA loci serve only to maintain plasmid DNA at the expense of the host organism 45 . Other authors suggest that these systems have evolved to favour the competitive ability of plasmids in cell progeny 46,47 . This hypothesis has been corroborated by the results of computer modelling 46 . The clinical antibiotic-resistant Acinetobacter baumannii strains have been shown to display higher susceptibility to environmental phages than antibiotic-sensitive strains 48 . Our results regarding the clinical multiresistant strains in the Ab_GEIH-2010 collection, which showed higher sensitivity to the Ab105-1φ and Ab105-2φ phages, is also consistent with the aforementioned finding.
In conclusion, two main molecular changes occurred during adaptation of clinical A. baumannii strains to a hospital environment during a decade: i) acquisition of a plasmid harbouring genes for OXA 24/40 ß-lactamase (associated with resistance to carbapenems) and AbKA/AbkB proteins (TA system), implicated in plasmid stabilization, and ii) acquisition of two temperate bacteriophages, Ab105-1φ (63 proteins) and Ab105-2φ (93 (2) genomes of both bacteriophages; (3) track of genes coloured as before; (4) similarity between the genomes of the two bacteriophages and the sequence of the other strains isolated in 2010. Red-coloured areas indicate no similarity in that region, i.e. this region is not present in the other genome or it was not detected with sufficient confidence. Green-coloured areas indicate a high level of confidence regarding the presence of the region; and (5) Equivalent plot for the strains isolated in 2000. In order to create the similarity tracks, we aligned the reads from each against the whole assembly, by using BWA software package. We used samtools to select those that map with high confidence in the phage regions and bedtools to measure the coverage. Those regions with coverage of more than 25× (expected around 150×) were flagged as very similar. proteins), containing important proteins associated with protection of the viral genome against bacterial attack (the SAM-dependent methyltransferases or AdoMet-MTases, as well as MazG and UmuC). Interestingly, the Ab_GEIH-2000 strains showed a QS/QQ-deficient network relative to the functional QS/QQ network observed in Ab_GEIH-2010 strains under stress conditions, which could indicate the molecular evolution to a functional network of QS and QQ cells by these temperate bacteriophages (Ab105-1φ and Ab105-2φ) in the Ab_GEIH-2010 collection. The functional QS/QQ network included a new QQ enzyme, AidA, which acts as a bacterial defence mechanism and was overexpressed in the presence of a QS inhibitor (exogenous 3-oxo-C12-HSL). The aforementioned molecular characteristics have an important influence on the evolution of bacterial pathogens and how these adapt to the host environment. We determined the antibiotic susceptibility profile by microdilution, according to CLSI recommendations 50 . We used PCR to determine the presence of the OXA 24/40 ß-lactamase and an AbkA/AbkB toxin-antitoxin system in all strains from the Ab_GEIH-2000 and Ab_GEIH-2010 collections.     Illumina MiSeq system. Isolates Ab155_GEIH-2000 and Ab105_GEIH-2010 reads were assembled using newbler Roche assembler. The other isolates were assembled using Velvet (Velvet v1.2.10 (https://www.ebi.ac.uk/~zerbino/ velvet/). Putative ORFs were predicted from assembled contigs using the GeneMarkS gene prediction program 52 , which was previously trained with the Acinetobacter baumannii genome (GI:83207914). Blast2Go 53 and RAST 54 were used for functional annotation of each predicted protein. rRNA and tRNA were identified using RNAmmer 55 and tRNAscan-SE 1.21 56 .
Construction of the pan genomes (all proteins), core genomes (similar proteins) and accessory genomes (no similar proteins) of the Ab_GEIH-2000 and Ab_GEIH-2010 strains was carried out using the PanSeq. 57 and Spine tools 58 .
The bacteriophage sequence was isolated from Ab105_GEIH-2010 (a strain representative of the Ab_GEIH-2010 collection) and manually assembled to improve the continuity of phage sequences. PCR amplification was used to confirm in silico assembly results used as a negative control for the Ab155_GEIH-2000 strain.  Reconstructed phage tools 59,60 . All phage proteins detected were manually annotated using the Protein BLAST 61 and InterProScan tools 62 and displayed ≥50% protein homology. The in silico assembly results were confirmed by PCR amplification.
The bacteriophage genomes of all strains analyzed in this study were compared following the indications of Krzywinski and collaborators 63 . Gene expression studies by microarrays and qRT-PCR. For the microarray studies, we obtained Dnase-treated RNA from a mid-exponential growth phase culture (optical density at 600 nm, 0.5). The cultures were treated with MMC (at a final concentration of 10 μg ml −1 ) and incubated for 1 h to induce an SOS response. The samples were removed for RNA extraction with the High Pure RNA Isolation Kit (Roche, Germany). The corresponding controls were processed in the same way but without addition of the above-mentioned compounds.

Study of the temperate bacteriophages of
The microarrays were specifically designed for the Ab 105_GEIH-2010 isolates by Bioarray Diagnostico Genético (Alicante, Spain) and using eArray (Agilent). The microarray assays were performed with 15,744 probes to study 4,017 genes. Labelling was carried out by two-colour microarray-based prokaryote analysis and Fair Play III labelling, version 1.3 (Agilent). Three independent RNAs per condition (biological replicates) were used in each experiment. Statistical analysis was carried out using Bioconductor, implemented in the RankProd software package for the R computing environment. A gene was considered induced when the ratio of the treated to the untreated preparation was ≥1.5 and the P value was <0.05 13 .
Quantitative real-time PCR (qRT-PCR) was used to examine the expression of bacteriophage genes in relation to interaction with the bacterial host (host-virus interactions). We used the Lightcycler 480 RNA MasterHydrolysis Probe (Roche, Germany) for the qRT-PCR studies. The UPL Taqman Probes (Universal Probe Library-Roche, Germany), Taqman probes and primers used are listed in Table S1 (Supplementary files). We adjusted the concentrations of the samples to achieve efficiencies of 90-110% and performed all experiments in triplicate from three RNA extractions (50 ng per RNA sample). For each strain, we normalized the expression of all genes relative to the single-copy rpoB housekeeping gene. We then calibrated the normalized expression of each gene of interest relative to its expression by untreated Ab_GEIHs RNA, which was assigned a value of 1.0.
Growth curves. About 100 ml LB broth was inoculated with 1:100 (v/v) of clinical strain Ab105_GEIH-2010 and was cultured overnight under shaking at 37 °C. In order to induce bacteriophages, one of the bacterial cultures, in which the optical density (OD) at 600 nm had reached 0.6, was exposed to MMC (Fischer Scientific, Loughborough, UK) at a final concentration of 10 µg/ml. Aliquots (1 ml) of the culture were removed for RNA extraction and the OD was measured every 20 minutes, and then every hour.
Isolation of the bacteriophages and TEM studies. Broth culture of strain Ab105_GEIH-2010 was induced as previously described for bacteriophage induction. Lysates were centrifuged at 3400 × g for 10 min and the supernatant was filtered through a 0.22 nm filter (Millipore). NaCl was added, to a final concentration of 0.5 M, and the suspensions were mixed thoroughly and left on ice for 1 h. The suspensions were then centrifuged at 3400 × g for 40 min at 4 °C, and the supernatants were transferred to sterile tubes. PEG 6000 (10% wt/vol) was added and dissolved by rocking the tubes at room temperature for 1 h and subsequent overnight incubation at 4 °C. Bacteriophages were then precipitated at 3400 × g for 40 min at 4 °C and resuspended in SM buffer (0.1 M NaCl, 1 mM MgSO 4 , 0.2 M Tris-HCl, pH 7.5) 64 . The samples were stored at 4 °C until being processed for TEM. The samples were negatively stained with 1% aqueous uranyl acetate before being examined in a JEOL JEM-1011 electron microscope.

QS/QQ network in Ab_GEIH-2000/2010 collections: QS/QQ genes.
We used qRT-PCR to examine expression of the network of QS/QQ genes (the abaI gene from the QS system and the aidA gene from the QQ system) 13 . We obtained DNAse-treated RNA from cultures treated with either non-native 3-oxo-C12-HSL (associated with inhibition of the QS system) 13,14 or H 2 O 2 (ROS response implicated in activation of the QS system) 13,65 . We used the Lightcycler 480 RNA MasterHydrolysis Probe (Roche, Germany) for the qRT-PCR studies. The UPL Taqman Probes (Universal Probe Library-Roche, Germany), Taqman probes and primers used are listed in Table S1 (Supplementary files). We adjusted the concentrations of the samples to yield efficiencies of 90-110% and normalized them with housekeeping gene (rpoB), as previously mentioned.
We also studied the surface motility (Quorum Sensing phenotype) 13 of all strains. The motility assays were performed in plates of modified LB-LN (nutrient depleted) 66 containing 2 g tryptone, 1 g yeast extract and 5 g NaCl. Assays were carried out with 0.25% Difco (Bacto TM agar) 13 .
Extraction of bacteriophages of Ab105_GEIH-2010 (representative strain from Ab_GEIH-2010 collection). The lysogenic bacteriophages (Ab105-1ϕ and Ab105-2ϕ) were isolated from a culture of A. baumannii strain Ab105_GEIH-2010, obtained by culturing the bacterium in LB medium at 37 °C until it reached the late log phase of growth. In order to lyse the cells and obtain the bacteriophages contained within, the culture was first incubated in the presence of 10% chloroform for 30 minutes and centrifuged at 3000 × g for 15 minutes before the supernatant was recovered and filtered through Millipore 0.22 μm membranes. The bacteriophages were then concentrated by the double agar overlay method 67  Ab177_GEIH-2000, which did not harbour any bacteriophages. The plates were washed with 3 ml of phage buffer (10 mM Tris-HCL pH 7.5 and 1.8 MgSO 4) under stirring for 3 hours, after which the buffer was recovered and the phages were refiltered.

Estimation of bacterial sensitivity to bacteriophages from Ab_GEIH-2000/2010 collections.
All strains from the Ab_GEIH-2000/2010 collections were grown overnight. The cultures were subsequently diluted 1:100 in LB medium and in LB medium supplemented with 10 μM 3-oxo-C12-HSL (exogenous AHL that inhibits the detection of quorum in A. baumannii). When the OD at 600 nm reached 0.1, the cultures were infected with the previously purified bacteriophages of strain Ab105_GEIH-2010 at a multiplicity of infection (MOI) of 10 (calculated by dividing the number of bacteria [colony-forming units CFUs] the number of phages [plaque-forming units, PFUs], in a given volume of infection mixture). The cultures were then incubated at 37 °C and shaken at 180 rpm for 4 hours. An aliquot of each culture was used to establish the number of CFUs by means of serial dilution. The bacteriophages were then extracted as described above and finally the PFUs were enumerated by the double layer agar method using strain Ab177_GEIH-2000 as substrate for infection.