The blaNDM-1-Carrying IncA/C2 Plasmid Underlies Structural Alterations and Cointegrate Formation In Vivo

In 2012, a carbapenemase-producing Salmonella enterica serovar Corvallis isolate carrying a blaNDM-1 multiresistance IncA/C2 plasmid, apart from IncHI2 and ColE-like plasmids, was detected in a wild bird in Germany. In a recent broiler chicken infection study, we observed transfer of this blaNDM-1-carrying IncA/C2 plasmid to other Enterobacteriaceae.

among S. Corvallis isolates, reisolates from three chicks belonging to four groups were selected. The selection of strains is shown in Table 1. To evaluate the in vivo stability of the plasmid content without selective pressure in isolation, 7 S. Corvallis reisolates detected on xylose lysine deoxycholate (XLD) agar from different chicks of group 1 were included. These are shown in Fig. S2 in the supplemental material. In total, 97 S. Corvallis reisolates were characterized in depth for the purpose of this investigation.
Isolation of S. Corvallis reisolates. S. Corvallis reisolates were isolated by suspending fresh fecal droppings from each chick in 4.5 ml of 0.85% (wt/vol) NaCl, from which a 100-l deposition volume was plated in duplicates onto XLD plates (Thermo Fisher Scientific, Germany) supplemented with 1 mg/liter cefotaxime (CTX) and 0.125 mg/liter meropenem (MEM). The addition of cefotaxime as a second antibiotic was to inhibit overgrowth by Pseudomonas spp. with intrinsic carbapenem resistance, which could hamper detection and quantification of the donor strain. Seven S. Corvallis strains from group 1 were isolated by plating onto XLD without selective supplementation. Strains were preserved at Ϫ80°C for later molecular analysis. Prior to molecular analysis, strains were serotyped. The designation is based on group (G1 to G4), day of isolation (1st to 29th day of life), and chick identifier (ID) (T1 to T10). S1-PFGE plasmid profiling and bla NDM-1 hybridization of S. Corvallis reisolates. All 97 reisolates of S. Corvallis were subjected to S1 pulsed-field gel electrophoresis (PFGE). Generated fragments were separated by the CHEF-DRIII system (Bio-Rad Laboratories, Spain) under running conditions as previously described (4). The S1-PFGE gels

TABLE 1
Distribution of 90 S. Corvallis strains which were selected for in-depth molecular analysis, including the strains isolated from the cecal content (29d*) a a Darker gray, strains selected for in-depth molecular analysis (all isolates, except n.a.); lighter gray, strains selected for WGS SNP analysis (see "Whole-genome SNP analysis" paragraph below). n.a, strain not available for analysis; *, postmortem cecal content isolates. of strains from group 1 and group 2 were further used for Southern blotting and bla NDM-1 hybridization.
In vitro conjugation experiments. Filter mating conjugation experiments with S. Corvallis strains harboring pSE12-01738-2 variants (D1, D2, D3, and D4) from the in vivo trial were conducted. Plasmid profiles of these are shown in Fig. S1 in the supplemental material. As recipients, nalidixic acid-resistant S. Paratyphi B (dTaϩ), S. Enteritidis, and S. Infantis were selected. From overnight cultures of selected strains, 500 l was inoculated into 25 ml (1:50) of Luria Bertani Bouillon-Miller liquid (LBL) and grown aerobically at 37°C with shaking (200 rpm), until optical density at 600 nm (OD 600 ) reached a value of 0.2. Afterward, donor and recipient strains were mixed 1:2 (100 l: 200 l) and centrifuged (16,000 rpm for 5 min). Here, 200 l of supernatant was discarded and the pellet was resuspended in the remaining 100 l and plated on an 0.22-m-pore-size mixed-cellulose ester membrane filter (Merck Millipore, Germany) placed on LB agar (Thermo Fisher Scientific, Germany). Conjugation experiments lasted 4 h and were repeated at room temperature (RT), 37°C, and 41.5°C. NDM-1-producing Salmonella transconjugants were selected on XLD containing 1 mg/liter CTX, 0.125 mg/ liter MEM, and 50 mg/liter nalidixic acid (NAL) and confirmed by serotyping, and conjugal transfer frequency (CTF) was calculated per donor.
WGS analysis. The whole-genome sequencing (WGS) analysis was conducted with Illumina MiSeq technology. Strains were grown overnight at 37°C in 4 ml of LBL with 1 mg/liter CTX, from which 1 ml was processed for DNA extraction using the PureLink genomic DNA minikit (Invitrogen, USA). DNA concentration (ng/l) was measured with the Qubit fluorometric quantitation (Invitrogen) system. Sequencing libraries were prepared with the Nextera XT DNA sample preparation kit (Illumina, San Diego, CA, USA). Paired-end sequencing was performed with the Illumina MiSeq benchtop (MiSeq Reagent v3 600-cycle kit, 2 ϫ 251 cycles). Raw reads were assembled de novo using CLC Genomics Workbench 9.5.2 (Qiagen, Hilden, Germany), and sequence types (STs), plasmid types, and resistance genes were detected using BatchUpload (5). Comparison of the pSE12-01738-2 variants (D3 and D4) was performed by mapping the raw reads to the reference pSE12-01738-2 plasmid (GenBank accession number CP027679) and visualizing them using BRIG (6).
For genome assembly, the hybrid assembly software Unicycler (v0.4.4) was used (7). It starts from an initial SPAdes short-read assembly and simplifies the assembly using information from short and long reads, thereby achieving a complete and accurate assembly (8). Assemblies were polished using Pilon (9). The cointegrated megaplasmid is represented using CLC Genomics Workbench 9.5.2.
Statistical analysis. For comparison of CTFs among donors (D1 to D4) under different temperature conditions (room temperature [RT], 37°C, and 41.5°C), statistical analysis with SPSS (ver. 21.0; SPSS Inc., USA) was performed. The distribution of the CTFs is presented by box-whisker plots with outliers and extreme outliers included. For the determination of statistical significance, one-way analysis of variance (ANOVA) was performed and least significant difference (LSD) was used as a post hoc test. The differences were considered significant if the P value was Ͻ0.05.
Following in vitro experiments, we observed variation in colony size and prolonged growth of Salmonella transconjugants after conjugation with S. Corvallis carrying the ϳ450-kb cointegrated megaplasmid. Therefore, 24 of these transconjugants (four small and four large colonies per recipient) from conjugation experiments at 41.5°C were analyzed by S1-PFGE. Analysis revealed that the size of the colony is not linked to full ϳ450-kb cointegrate acquisition. Additionally, resolution of the cointegrate (plasmids from ϳ170 to ϳ350 kb in size) in transconjugants was observed (data not shown).
Structural alterations of the bla NDM-1 -carrying pSE12-01738-2 plasmid were seen in ϳ10-kb and ϳ70-kb deletion and ϳ450-kb megaplasmid formation. The ϳ450-kb megaplasmid (462,435 bp) is a cointegrate of IncHI2 (pSE12-01738-1) and the multiresistance bla NDM-1 -carrying IncA/C 2 (pSE12-01738-2) plasmid (Fig. 4) and was detected in 2 out of 97 strains. The fusion was mediated by IS6-like family genetic elements. In a study of movement of IS26, which can be identical to IS6, it was observed that IS26 can form cointegrates between DNA molecules (10). Other studies have shown plasticity of IncHI2 and fusion with IncF plasmids (11,12). A fusion event can potentially facilitate dissemination of other genetic elements, such as heavy metal resistance in the case of tellurite (Ter cluster) present in pSE12-01738-1 (Fig. 4). A study by Lin et al. (13) revealed that spread of the bla CTX-M-17 gene present on a nonconjugative plasmid was due to fusion with a conjugative ϳ73-kb plasmid. As our IncHI2-IncA/C 2 cointegrate was detected in only two reisolates, we assume that such an S. Corvallis population persists in vivo but in lower numbers. This could be due to instability of the cointegrate, supported by our in vitro conjugation experiments where resolution of the IncHI2-IncA/C 2 cointegrate was observed. Plasmid resolution was observed by Xie et al. (14) in the case of the ϳ190-kb cointegrated multireplicon bla NDM-5 plasmid, suggesting plasmid instability in new recipients or during conjugation. Besides instability and decreased CTF effect, our cointegrate acquisition caused an elongated growth time for Salmonella recipients and variation of the colony size in vitro. In Alteration of bla NDM-1 -Carrying IncA/C 2 Plasmid Antimicrobial Agents and Chemotherapy a study on the transmission and burden of an ϳ1-Mb Pseudomonas syringae megaplasmid, pMPP1a107, a decrease in fitness was also observed (15). In another study, it was observed that the same plasmid can have up to 2.5-fold-higher fitness costs in different Pseudomonas species (16).
Recently, Paskova et al. (17) detected a bla NDM-1 -carrying ϳ300-kb multireplicon (IncA/C 2 and IncR) plasmid in an E. coli strain from human urine. The type I IncA/C 2 sequence part of this megaplasmid was 99% identical to pRH-1238, which is the same plasmid as pSE12-01738-2, with only a minor structural deletion in the latter (3). These findings confirmed our hypothesis of the broad host range and adaptation potential of this particular bla NDM-1 -carrying plasmid in vivo. This also suggests possible bla NDM-1 spillover from human clinical settings where carbapenems are an alternative to cephalosporin in cases of resistance (18).
We observed frequent loss of the IncHI2 pSE12-01738-1 plasmid, despite genes associated with the toxin-antitoxin system being present. Plasmids are undergoing selection pressure, and to control costs and maximize their spread, the host adapts strategies to cope with their presence (19). The cost of the pSE12-01738-1 plasmid might have outweighed the benefits for the host, leading to the plasmid loss (20,21). Structural alterations were more common in IncA/C 2 pSE12-01738-2 than in the pSE12-01738-1 and pSE12-01738-3 plasmids. These were seen in two deletion events of FIG 3 Visualization of the bla NDM-1 -carrying pSE12-01738-2 variants D3 and D4 compared to PacBio RSII reference sequence of pSE12-01738-2 plasmid (GenBank accession number CP027679) using BRIG (6) with resistance genes (red, beta-lactam genes; black, other resistance genes) as well as IS elements, transposase, and tra genes (all marked gray). the pSE12-01738-2 plasmid. The first, smaller deletion (ϳ10 kb) covers IS6 family transposase-flanked macrolide resistance genes (mphE and msrE), and a larger deletion (ϳ70 kb) included two tra clusters (traL-traK-traB-traV-traA and traC-traW-traU-traN) (Fig. 3). As tra genes are required for pilus assembly (traW), the structure of pilus (traC), and mating pair stabilization (traN) (22)(23)(24), the absence of some tra genes led to loss of the conjugation machinery in this pSE12-01738-2 variant (D3) (Fig. 2 and 3). The remaining traG, traH, traF, and traI genes did not maintain conjugation ability for this plasmid derivative in vitro.
Antimicrobial usage is the most common trigger for the spread of antimicrobial resistance (25); however, reducing antibiotic use alone is not sufficient to reverse resistance (26). Eliminating antimicrobial selection pressure alone does not lead to plasmid loss in all plasmid-host combinations (27). This was observed in our in vivo study. Therefore, insights into mechanisms which trigger and enhance plasmid loss might be an effective addition to support current knowledge as future intervention measures.
Our study revealed the most common structural alterations of a public-healthrelevant bla NDM-1 -carrying IncA/C 2 plasmid once carried with S. Corvallis into a broiler flock. Despite structural alterations and plasmid cointegration, the bla NDM-1 gene is maintained in different IncA/C 2 variants. For the future, synergy of reduction in antimicrobial usage and alternative approaches, such as promoting plasmid loss, might be an additional contribution aiming to slow the spread of resistance.

SUPPLEMENTAL MATERIAL
Supplemental material for this article may be found at https://doi.org/10.1128/AAC .00380-19.