Co-Occurrence of blaOXA-23 in the Chromosome and Plasmid: Increased Fitness in Carbapenem-Resistant Acinetobacter baumannii

This study aims to explore the co-occurrence of chromosomal and plasmid blaOXA-23 in carbapenem-resistant A. baumannii (CRAB) and its influence on phenotypes. A total of 11 CRAB isolates containing copies of blaOXA-23 on the chromosome and plasmid (CO), as well as 18 closely related isolates with blaOXA-23, located on either the chromosome or plasmid (SI), were selected for the determination of antibiotic susceptibility, virulence phenotype, and characteristic genomic differences. The co-occurrence of blaOXA-23 on the CRAB chromosome and plasmids did not enhance carbapenem resistance, but trimethoprim/sulfamethoxazole exhibited significantly reduced minimum inhibitory concentrations in CO. CO demonstrated a higher degree of fitness compared to SI. An increased biofilm formation ability and serum tolerance were also identified in CO, which may be associated with virulence genes, which include csuD, entE, pgaA, and plc. blaOXA-23-carrying transposons were found at different insertion sites on the chromosome. The most common site was AbaR-type genomic islands (50%). Two types of plasmids were found in CO. The co-occurrence of blaOXA-23 on the chromosome and a plasmid in CRAB had little effect on carbapenem susceptibility but was accompanied by increased fitness and virulence. Different origins and independent insertions of blaOXA-23-carrying transposons were identified in both the chromosomal and plasmid sequences.


Introduction
Acinetobacter baumannii is a Gram-negative, opportunistic, and nosocomial pathogen that causes hospital-and community-acquired infections. This pathogen is considered a global threat to public health because of the speed at which it develops resistance to antibiotics and because residual options for treating its infections are limited. A recent review reported a carbapenem-resistant Acinetobacter baumannii (CRAB) prevalence of about 80% in China [1]. Poor clinical outcomes have also been reported with CRAB infections [2].
Carbapenems are broad-spectrum antimicrobial agents. They are members of the β-lactam family that are active against most β-lactamase-producing organisms. Carbapenems are accepted as the first-line treatment for multidrug-resistant bacterial infections [3]. However, the increasing trend in the occurrence of CRAB indicates that last-resort treatments are increasingly ineffective. Many resistance mechanisms have been described for A. baumannii [3], including enhanced degradation of antibiotics by β-lactamases. OXA-type carbapenem-hydrolyzing class D β-lactamase, the universal β-lactamase in A. baumannii, plays a significant role in carbapenem resistance [4]. The bla OXA-23 gene is the most common determinant of carbapenem resistance in CRAB. Carbapenem resistance may also be attributed to modified porins, penicillin-binding proteins, and the resistance-nodulationdivision (RND) family efflux system [3].
A. baumannii shows high genetic plasticity, allowing for the accumulation of resistance determinants and highlighting the correlation between the horizontal transfer of

Prevalence of the Co-Occurrence of bla oxa-23 on Chromosomes and Plasmids
Between 1999 and 2018, positive rates of CO isolates among our sequenced isolates showed an increasing trend over time. Before 2013, no CO isolates were identified. However, 3, 2, and 6 CO isolates were identified out of a total of 28 isolates (10.7%, 7.1%, and 21.4% positive rates, respectively) in 2014, 2016, and 2018, respectively. We determined the clonal relatedness of the 11 isolates using MLST ( Table 1). All the isolates were identified as sequence type (ST) 2 using the Pasteur scheme. However, the Oxford scheme identified four ST208, three ST195, two ST368, one ST218, and one ST2143 isolate. In addition, aside from ST2143, all the other STs belonged to clone complex 208 (CC208). For a detailed analysis of isolate relationships and evolution, we developed a robust phylogeny based on single nucleotide variants present in the core regions of the genome that represent ancestral relationships between CO and SI isolate genomes ( Figure 1). Phenotypic characteristics were also attached to the phylogenetic tree (as described below), which might reveal the relevance of evolutionary history and phenotypic changes in CO and SI isolates. The analysis indicated that isolates were roughly grouped into three distinct clades, with little evidence showing that CO evolved from current SI isolates, implying that co-occurrence of bla OXA-23 on chromosomes and plasmids occurred previously. We found no significant evolutionary relationships or genomic characteristic differences between CO and SI isolates.

Resistance Gene Profiles and Correlation between Antibiotic Resistance and blaOXA-23 Location
We found all CO isolates were resistant to imipenem, meropenem, cefepime, ceftazidime, piperacillin/tazobactam (TZP), tetracyclines, doxycycline, ciprofloxacin, Figure 1. Phylogenetic tree of CO and SI isolates. The phylogenetic tree was constructed based on core single nucleotide variants of chromosomes. The internal colored ranges and the external colored stripes represented STs in Pasteur and Oxford typing schemes, respectively. The navy-blue left triangle and the orange right triangle indicated the location of bla OXA-23 in the chromosome and plasmid, respectively. The number of triangles indicated the number of copies of bla OXA-23 . The circle, square, and upward triangle indicated that bla OXA-23 was flanked by Tn2006, Tn2009, or ISAba1, while pink and azure classified gene location into chromosome or plasmid, respectively. The red and green stars represented minocycline and SXT resistant phenotypes. The purple and yellow bars represented the biofilm formation and serum tolerance virulence phenotypes.
To determine the correlation between bla OXA-23 co-occurrence and antibiotic resistance patterns, we performed antimicrobial susceptibility tests for SI isolates (Table S1). In addition, we used the unpaired Student's t-test to compare the resistance profiles of CO and SI isolates ( Figure S1). Among the 13 antibiotic agents listed above, only SXT showed significantly low MICs from the SI isolates (p < 0.05). The corrected Welch's t-test results indicated significant differences in the MICs of TZP and imipenem (p < 0.05). TZP and imipenem exhibited lower and more discrete MIC distributions in SI isolates as compared to CO isolates. We observed a similar trend in the MICs of ceftazidime, cefepime, meropenem, and minocycline, but there were no significant differences in MIC values (p > 0.05).
The MIC 50 and MIC 90 of minocycline in CO isolates were twice as high as those in the SI isolates. Moreover, the MIC 50 and MIC 90 of SXT in CO isolates were sixteen times and four times lower, respectively, than those in SI isolates ( Table 2). We further analyzed the MICs of minocycline and SXT using the Chi-square test, which showed significantly different MIC values in CO and SI isolates for these two agents (p < 0.05). However, no significant differences were observed with the other antibiotics. The WGS of the 11 isolates revealed 18 known resistance genes on the chromosome, consistent with the reduced susceptibility to β-lactams (bla OXA-23 and bla ADC-25 ), macrolides (mph(E) and msr(E)), aminoglycosides (armA, aph(3")-Ib, and aph(6)-Id), and tetracyclines (tet(B) observed. Among these genes, aph(3")-Ib, aph(6)-Id, bla ADC-25 , bla OXA-23 , bla OXA-66 , and tet(B) were detected in all isolates ( Figure 2). The SXT-related resistance genes, sul1 and sul2, had less distribution in CO isolates (45.5%) than in SI isolates (100%), leading to lower MICs. Plasmids were also found to contain no resistance genes except bla OXA-23 , indicating that the co-occurrence of resistance genes on the chromosome and plasmids only occurred with bla OXA-23 . To assess the impact of blaOXA-23 co-occurrence on chromosome and plasmid on the growth of A. baumannii isolates, we compared the growth rates of CO isolates with those of SI isolates. The k value for ATCC19606 was set at 1. However, no growth differences were observed between them (p > 0.05, Figure 3A), indicating that the number and position of blaOXA-23 had little effect on the growth rate. In the context of antibiotic existence (8 mg/L imipenem), the amount of blaOXA-23 in isolates had little influence on the growth rate (data not shown). In addition, we further investigated in vitro growth competition between CO and SI isolates with the same ST. Paired ST195 CO and SI isolates were coincubated, and the growth rate of the CO isolate was found to be higher than that of the SI isolate ( Figure 3B). We also observed a significant increase in the growth of CO isolates To assess the impact of bla OXA-23 co-occurrence on chromosome and plasmid on the growth of A. baumannii isolates, we compared the growth rates of CO isolates with those of SI isolates. The k value for ATCC19606 was set at 1. However, no growth differences were observed between them (p > 0.05, Figure 3A), indicating that the number and position of bla OXA-23 had little effect on the growth rate. In the context of antibiotic existence (8 mg/L imipenem), the amount of bla OXA-23 in isolates had little influence on the growth rate (data not shown). In addition, we further investigated in vitro growth competition between CO and SI isolates with the same ST. Paired ST195 CO and SI isolates were co-incubated, and the growth rate of the CO isolate was found to be higher than that of the SI isolate ( Figure 3B). We also observed a significant increase in the growth of CO isolates from 3 to 24 h, with the competitive index (CI) increasing from 6.93 to 35.0. The co-occurrence of bla OXA-23 offered a fitness advantage compared to the SI isolates. The fitness was enhanced by a longer co-incubation time, which suggested that the bla OXA-23 co-occurrence might be associated with a lower degree of fitness in vitro. Two asterisks (**) indicated a significant difference (p < 0.01). The ordinate represented the corrected adhesion unit. (D) Serum resistance of CO and SI isolates were shown by the survival of all isolates after incubation in normal human serum. Four asterisks (****) indicated a significant difference (p < 0.0001). The dots represented CO isolates and the triangles represented SI isolates.
We assessed the biofilm production capacity of SI and CO isolates using a biofilm assay. The results indicated that biofilm production varied between the CO and SI isolates (p < 0.01; Figure 3C). The mean values for the CO and SI isolates were 1.56 and 1.002, respectively, and their SD values were 1.185 and 0.6859, respectively. The adhesion unit of biofilm formation for ATCC19606 was set at 1. All isolates produced large quantities of biofilm, but CO isolates produced greater biofilm quantities in general. This significant variation suggests that biofilm formation capacity changing might be related to blaOXA-23 location diversity.
We determined the survival of CO and SI isolates after 3 h of incubation in NHS. Half of the CO isolates showed high survival rates of 10% or higher and were characterized as serum resistant. The percent survival for ATCC17978 was set at 100%. SI isolates remained sensitive to serum, with a survival rate of <1%. We examined the results using the Mann-Whitney test and that survival rates were significantly different between CO and SI isolates (p < 0.0001; Figure 3D). The ranges of the CO and SI isolates were distinct from 36.05 to 3.983, respectively. The mean survival rates of CO and SI were 11.35 and 0.4902, respectively, and the SD values were 12.21 and 0.9094, respectively. This indicated that blaOXA-23 We assessed the biofilm production capacity of SI and CO isolates using a biofilm assay. The results indicated that biofilm production varied between the CO and SI isolates (p < 0.01; Figure 3C). The mean values for the CO and SI isolates were 1.56 and 1.002, respectively, and their SD values were 1.185 and 0.6859, respectively. The adhesion unit of biofilm formation for ATCC19606 was set at 1. All isolates produced large quantities of biofilm, but CO isolates produced greater biofilm quantities in general. This significant variation suggests that biofilm formation capacity changing might be related to bla OXA-23 location diversity.
We determined the survival of CO and SI isolates after 3 h of incubation in NHS. Half of the CO isolates showed high survival rates of 10% or higher and were characterized as serum resistant. The percent survival for ATCC17978 was set at 100%. SI isolates remained sensitive to serum, with a survival rate of <1%. We examined the results using the Mann-Whitney test and that survival rates were significantly different between CO and SI isolates (p < 0.0001; Figure 3D). The ranges of the CO and SI isolates were distinct from 36.05 to 3.983, respectively. The mean survival rates of CO and SI were 11.35 and 0.4902, respectively, and the SD values were 12.21 and 0.9094, respectively. This indicated that bla OXA-23 co-location on plasmids and chromosomes might correlate with increased serum resistance. Therefore, bla OXA-23 co-occurrence may be relevant with fitness changes in A. baumannii in terms of biofilm formation and serum resistance.

Comparative Analysis of Virulence-Related Genes and Accessory Genome in CO and SI Isolates
To investigate differences in virulence capacity between CO and SI isolates, we identified virulence-associated genes using VFDB. The virulence gene profiles showed that most of the virulence genes were present in all isolates, except for the biofilm formation-related genes csuD and pgaA, the iron uptake-related gene entE, the LPS-related gene lpxA, and the enzyme-related gene plc ( Figure 4). The carriage ratios of csuD (93.1%), entE (93.1%), pgaA (81.8%), and plc (63.6%) were higher for CO isolates than for SI isolates. Due to the potential loss of defense systems in prokaryotic genomes during evolution, we censored the divergences of restriction-modification (R-M) system-related genes between CO and SI isolates. The R-M systems are an important component of defense systems [12]. BLASTbased on genes from the PADS Arsenal was performed [13], and all isolates were found to contain R-M-encoding genes ( Figure 4). Modification subunit (M subunit) encoding genes had a higher distribution in CO isolates, however, all others subunits had a lower distribution.

Comparative Analysis of Virulence-Related Genes and Accessory Genome in CO and SI Isolates
To investigate differences in virulence capacity between CO and SI isolates, we identified virulence-associated genes using VFDB. The virulence gene profiles showed that most of the virulence genes were present in all isolates, except for the biofilm formationrelated genes csuD and pgaA, the iron uptake-related gene entE, the LPS-related gene lpxA, and the enzyme-related gene plc ( Figure 4). The carriage ratios of csuD (93.1%), entE (93.1%), pgaA (81.8%), and plc (63.6%) were higher for CO isolates than for SI isolates. Due to the potential loss of defense systems in prokaryotic genomes during evolution, we censored the divergences of restriction-modification (R-M) system-related genes between CO and SI isolates. The R-M systems are an important component of defense systems [12]. BLAST-based on genes from the PADS Arsenal was performed [13], and all isolates were found to contain R-M-encoding genes (Figure 4). Modification subunit (M subunit) encoding genes had a higher distribution in CO isolates, however, all others subunits had a lower distribution. Moreover, we evaluated discrepancies between CO and SI isolates at the genetic level to determine if there were other genes responsible for the phenotypic changes. Using sequence BLAST and gene absence and presence identification, we compared the presence Moreover, we evaluated discrepancies between CO and SI isolates at the genetic level to determine if there were other genes responsible for the phenotypic changes. Using sequence BLAST and gene absence and presence identification, we compared the presence and absence of genes between CO and SI isolates. Genes and isolates were hierarchically clustered using one minus Pearson correlation metric, with CO and SI isolates clustered separately ( Figure S2). We further investigated gene presence and absence in isolates with the same ST (Oxford scheme). Five ST208 SI isolates were found to carry the XRE family transcriptional regulator encoding gene, which was absent in all ST208 CO isolates.

Mechanism of bla OXA-23 Multiplication
To obtain a general view of the mechanism underlying bla OXA-23 multiplication in A. baumannii, we investigated the bla OXA-23 transposon (Tn) types on the chromosomes and plasmids of each isolate separately ( Table 3). The bla OXA-23 on the chromosome was carried by either Tn2006 (8, 78.6%) or Tn2009 (3, 21.4%). Tn2009-harboring isolates were limited to ST208. Isolates 2018HLJAB1, 2018HLJAB2, and 2016BJAB1 harbored double copies of bla OXA-23 on the chromosome, with both copies carried by Tn2006. Isolate 2014TJAB1 was the only isolate with bla OXA-23 carried on different Tn types; the chromosomal and plasmid copies were carried on Tn2006 and Tn2009, respectively. We further investigated the bla OXA-23 location in SI isolates, aside from Tn2006 (9, 50.0%) and Tn2009 (2, 11.1%), bla OXA-23 was also flanked by ISAba1 (7, 38.9%) (Figure 1). We found that none of the plasmid-located bla OXA-23 from SI were situated at transposon. We also detected ISAba1 upstream of the bla OXA-23 gene in the CO and SI isolates.
Subsequently, we analyzed discrepancies between transposon insertion sites ( Table 3). The most common location of bla OXA-23 -containing Tn2006 in CO isolates was the AbaR4 resistance island (6, 42.9%). Isolates 2018HLJAB1, 2018HLJAB2, and 2016BJAB1 harbored duplicate copies of bla OXA-23 -containing Tn2006. One of the Tn2006 copies was inserted in the antibiotic resistance island AbaR, interrupted with comM. The other copy was inserted in the znuA promoter region. Furthermore, isolate 2018HBAB1 acquired bla OXA-23 through Tn2006 insertion upstream of flavin mononucleotide (FMN) reductase. Isolate 2014TJAB1 was longer and unique, and its insertion position was the same as that of AbaR25.
The insertion sites of bla OXA-23 -containing Tn2009 in isolates 2014BJAB1, 2016LNAB1, and 2014LNAB1 might be located inside the phage because there were phage-related proteins both upstream and downstream of this region. As previously reported [14], MDR-ZJ06 is one of the most representative isolates harboring Tn2009-located bla OXA-23 . The transposon is inserted in a gene cluster involved in the assembly of P pilus, with a 9-bp target site duplication (5 -CAAAAAATT-3 ). We found 11 bases (TTCTATCATAGG) inserted in the region between parA and YeeC in Tn2009. In addition, sequence profiles indicated a different direct repeated (DR) sequence from MDR-ZJ06. Tn2009 in CO isolates was located on the chromosomes and plasmids and was flanked by a 9-bp duplication (5 -AAAATATTT-3 ).

Plasmid Characteristics of CO Isolates
The evolutionary relationship between bla OXA-23 -harboring CO plasmids was generated through maximum likelihood phylogenetic analysis, in which the plasmids were clustered into two clades ( Figure 5A). We also performed pairwise blast analysis of the plasmids to analyze the differences in homology between them, as coverage and identity between plasmids were between 99.84 and 99.99%, and 11%, respectively. Thus, categorizing the 11 plasmids into two groups, CO plasmid type I and II, based on the consequences ( Figure S3). The earliest isolate in each group was used as the reference strain to perform circular comparisons of the two types of plasmids, and the overlapping regions of rings confirmed clustering. Plasmids from three 2014 isolates and 2016LNAB1 were more cognate and were classified as CO plasmid type I, which harbored Tn2009. Consequently, we investigated the Tn2009-harboring plasmids and found that they were significantly similar (99.98%) to the conjugative plasmid pABTJ1, which effectively contributed to the wide dissemination of bla OXA-23 in Acinetobacter spp. in China [15]. target site duplication (5'-CAAAAAATT-3'). We found 11 bases (TTCTATCATAGG) inserted in the region between parA and YeeC in Tn2009. In addition, sequence profiles indicated a different direct repeated (DR) sequence from MDR-ZJ06. Tn2009 in CO isolates was located on the chromosomes and plasmids and was flanked by a 9-bp duplication (5'-AAAATATTT-3').

Plasmid Characteristics of CO Isolates
The evolutionary relationship between blaOXA-23-harboring CO plasmids was generated through maximum likelihood phylogenetic analysis, in which the plasmids were clustered into two clades ( Figure 5A). We also performed pairwise blast analysis of the plasmids to analyze the differences in homology between them, as coverage and identity between plasmids were between 99.84 and 99.99%, and 11%, respectively. Thus, categorizing the 11 plasmids into two groups, CO plasmid type I and II, based on the consequences ( Figure S3). The earliest isolate in each group was used as the reference strain to perform circular comparisons of the two types of plasmids, and the overlapping regions of rings confirmed clustering. Plasmids from three 2014 isolates and 2016LNAB1 were more cognate and were classified as CO plasmid type I, which harbored Tn2009. Consequently, we investigated the Tn2009-harboring plasmids and found that they were significantly similar (99.98%) to the conjugative plasmid pABTJ1, which effectively contributed to the wide dissemination of blaOXA-23 in Acinetobacter spp. in China [15]. While the plasmids in all 2018 isolates and 2016BJAB1 were classified as CO plasmid type II and showed substantial similarities, they had different structural characteristics from type I plasmids. Notably, blaOXA-23 belonged to Tn2006, as described earlier. We used While the plasmids in all 2018 isolates and 2016BJAB1 were classified as CO plasmid type II and showed substantial similarities, they had different structural characteristics from type I plasmids. Notably, bla OXA-23 belonged to Tn2006, as described earlier. We used BLAST to evaluate the plasmid sequence in NCBI, and no sequences with high homology were found, suggesting that there may have been a new epidemic plasmid carrying bla OXA-23 in China in 2018. The circular map of p2018HLJAB1, as the representative of the CO type II plasmid, showed that the plasmid was 78,023 bp in size, with 262 open reading frames, and had an average GC content of 33.61% ( Figure 5B). The plasmid consisted of two main regions, genes associated with conjugation and bla OXA-23 -containing Tn2006. The VFDB analysis confirmed that there were no common virulence-related genes that were present on the CO plasmids. Meanwhile, three genes on the CO plasmids, which include a type II restriction enzyme methylase subunit encoding gene, a type II toxin-antitoxin system RelE/ParE family toxin ParE encoding gene in the CO type II plasmids, and an N-6 DNA methylase encoding gene in both CO type I and II plasmids were associated with virulence according to the clusters of orthologous genes functional categories that were identified using the eggNOG database [16]. Further, we inquired into the bla OXA-23 -carrying plasmids from SI simultaneously and discovered that all three plasmids could be characterized into CO plasmid type I with a few differences in bla OXA-23 location sites. The genomic environment showed that apart from bla OXA-23 , there were no other resistance genes located in SI plasmids.

Discussion
In many countries, including China, the emergence of CRAB-related nosocomial infection outbreaks has been rapidly increasing, leading to treatment failure and prolonged hospitalization [17,18]. As the most representative carbapenem resistance gene, the cooccurrence of bla OXA-23 on the chromosomes and plasmids may induce phenotypic changes in bacteria. In this study, we describe the carbapenem-resistant A. baumannii isolates that bla OXA-23 co-existed on the chromosome and plasmid for the first time.
Gene amplification is the method most used by bacteria to increase the expression of resistance genes and can, therefore, promote resistance to antibiotics. However, previous studies have shown that bla OXA-23 multiplication does not enhance carbapenem resistance in clinical CRAB [11]. Our study further proved that the presence of multiple copies and locations of bla OXA-23 had no significant effect on susceptibility to carbapenems and most antibiotics. However, we found that bla OXA-23 co-existence had a significant effect on SXT sensitivity. Our results showed that the MICs of SXT in CO isolates were lower than those in SI isolates. In an environment without antibiotics, most resistance mutations reduce bacterial fitness [19]. Based on this premise, bla OXA-23 amplification and multi-location should have reduced the fitness of the CO isolates. On the contrary, our results suggest that this may be correlated with fitness enhancement. bla OXA-23 co-occurrence on chromosomes and plasmids altered bacterial phenotypes that are important for bacterial fitness. CO isolates showed better competitive growth, serum tolerance, and biofilm formation capacity, indicating a lower fitness cost in the host and environment. We speculate that this increase in fitness may be concerned with the decreased resistance to SXT, which is related to bla OXA-23 co-existence. An SXT susceptible Staphylococcus aureus isolate was found to produce a lower bacterial load in infected mice than a resistant strain, indicating lower fitness cost [20]. Moreover, plasmids induce the accumulation of both positive and negative effects on bacterial fitness [21]. Plasmids carrying bla OXA-23 may influence fitness in a positive way. During the evolution of antimicrobial resistance, bacteria may acquire more and differentially located bla OXA-23 to better adapt to the environment, which implies a greater possibility of extending the host range and increasing spread.
High survival rate in serum and high biofilm formation capacity is closely related to bacterial virulence. This study showed that the CO isolates harbored more virulence genes (including pgaA and lpxA) than SI isolates. pgaA is related to β-1-6-poly-N-acetyl-Dglucosamine (PNAG) synthesis. PNAG is a surface polysaccharide that is indispensable for maintaining biofilm integrity [22]. The increase in pgaA in CO isolates may contribute to their higher biofilm production capacity. lpxA is a lipid A biosynthesis gene, and lipid A is one of the components of LPS. LPS is critical for in vitro serum resistance and contributes to the survival and fitness of A. baumannii [23]. Increased lpxA in CO isolates may help bacteria manifest a higher serum tolerance to escape the host's immune response. The modification subunit genes of the R-M system were different in CO and SI isolates. The R-M system functions by methylating or cleaving DNA to protect bacteria from invasion by foreign DNA, and is also involved in the adaptation of bacteria to changes in environmental conditions [24]. Gene presence-absence identification in ST208 CO and SI isolates revealed variations in an XRE family transcriptional regulator-encoding gene. In a previous study, the first member of the XRE family transcriptional regulator, StrT, was found to be involved in stress tolerance and virulence in Streptococcus suis [25]. Further research is needed to clarify the function and mechanisms of R-M system and XRE family transcriptional regulator in A. baumannii evolution.
Horizontal gene transfer (HGT), including plasmid conjugation, plays an important role in the dissemination of antibiotic resistance genes, allowing the transfer of genes between isolates and species [26]. Thus, HGT can accelerate the spread of resistance genes and increase global antimicrobial resistance. In conjugative plasmids, resistance genes are usually carried by a transposon, which can transfer genes between plasmids and chromosomes [27]. bla OXA-23 dissemination is mostly mediated by Tn2008, Tn2006, and Tn2009 in A. baumannii isolates [6]. Likewise, our results indicate that bla OXA-23 genes colocated on chromosomes and plasmids were either carried by Tn2006 or Tn2009. Tn2006 is the most frequently described bla OXA-23 -harboring transposon, and our results support this finding. Tn2006 has sufficient intercellular transfer efficiency in transferring genes between isolates via plasmids or itself. In contrast, the dissemination of Tn2009 was attributed to the clonal spread of the bacterial host [28], as Tn2009-harboring isolates in our study were limited to ST208. ISAba1 belongs to the IS4 insertion sequences family, which is associated with several antibiotic resistance genes in A. baumannii [29]. It plays an important role in bla OXA-23 gene expression, as the ISAba1 transposase is composed of two open reading frames and can produce a functional protein in response to a frameshift during protein translation [30]. We detected ISAba1 upstream of the bla OXA-23 gene and speculate that it might contain promoter sequences for bla OXA-23 expression; however, future expression studies will be required to fully unravel the influences of the co-occurrence of bla OXA-23 in chromosomal and on plasmid DNA. The CO plasmids can be classified into two types, where type II plasmids are speculated to be a new epidemic plasmid associated with bla OXA-23 occurrence in China. Future studies of molecular epidemiology and functional research into new plasmids are needed.
Different transposon insertion sites were identified. ZnuA is a substrate-binding protein of the zinc-uptake ABC transporter and is relevant in bacterial persistence [31]. Transposons with bla OXA-23 inserted in the ZnuA promoter region may influence gene synthesis, leading to decreased persistence. The bla OXA-23 gene of isolate 2018HBAB1 was obtained through Tn2006 insertion upstream of FMN reductase. FMN and the transcriptional regulator, LuxR, remained intact and had no effect on gene function. Furthermore, insertion of Tn2009 in the phage affected DNA polymerase V (UmuC), which is error-prone and depends on the lesion-bypass replication mechanism [32]. This might also affect upstream hypothetical proteins. Transposon insertion may result in the creation of novel resistance genes at the locus. Therefore, comM-AbaR, an antibiotic resistance island, is the preferred hotspot in A. baumannii [33]. The findings of this study were consistent with the previous study, which reported that the most common insertion site of Tn2006 in CO A. baumannii strains is located within comM-AbaR. The Tn2006 insertion site in CO isolates was found to be between the coding region of a hypothetical protein and the transcriptional regulator, GntR, and this location may affect transcription. In isolate 2014TJAB1, the different transposon types found on the chromosome and plasmid indicate that the origins of the multiple bla OXA-23 copies might not be the same. Further studies are needed to clarify the evolutionary mechanism of bla OXA-23 .
The co-occurrence of bla OXA-23 on chromosomes and plasmids was observed in Proteus mirabilis [34]. Two bla OXA-23 copies were detected, one on the chromosome and the other on a plasmid, indicating that P. mirabilis may be a reservoir for bla OXA-23 . bla OXA-23 was also carried on Tn2006 and inserted in AbaR4. The chromosome, Tn2006 alongside AbaR4, was integrated into the comM gene, while the plasmid copy was integrated into the putative colicin gene. bla OXA-23 amplification in the genome of P. mirabilis and E. coli increases the transmission risk of bacteria of the Enterobacteriaceae family [35]. We observed phenotypic changes related to bacterial fitness in CO isolates. In addition, changes may also increase persistence in the environment, which may be associated with the capacity enhancement of bacteria to develop antimicrobial resistance. Future studies that will analyze the mechanism of bla OXA-23 co-located on the chromosome and plasmids and interspecies transfer of bla OXA-23 will be required to fully elucidate its mechanism.

Bacteria Isolates
A total of 11 CRAB isolates harboring multiple copies of bla OXA-23 co-occurring on the chromosome and plasmids were selected from 135 A. baumannii representative isolates in China between 1999 and 2018, which we referred to as CO isolates. In addition, 18 closely evolutionary related isolates with bla OXA-23 located on the chromosome or plasmid were selected as controls, which we referred to as SI isolates. The principle for choosing these control isolates was based on single nucleotide variant phylogenetic tree analysis of the core genome and multilocus sequence typing using the Oxford scheme. Detailed information on these isolates is presented in Table 1 and Table S1.

Antimicrobial Susceptibility Testing
Antimicrobial susceptibilities were determined by disk diffusion on Mueller-Hinton agar (Bio-Rad, Marnes-La-Coquette, France) and the minimum inhibitory concentrations (MIC) of colistin and doxycycline were determined using the microbroth dilution method. E. coli strain ATCC25922 and P. aeruginosa strain ATCC27853 were used as reference control strains, and the aforementioned tests were performed following the Clinical and Laboratory Standards Institute (CLSI) guidelines [36]. Susceptibility was interpreted according to the criteria of the CLSI (M100-S29) [37], except for susceptibility to tigecycline, which was interpreted following the FDA criteria (susceptible: ≤2 mg/L; resistant: ≥8 mg/L) for Enterobacteriaceae.

Genome Sequencing and Analyses
Genomic DNA was extracted using QIAamp DNA Mini Kit (Qiagen, Venlo, Holland) and sequenced using PacBio RS II sequencer (Pacific Biosciences, Menlo Park, CA, USA). De novo assembly and resequencing were performed following the hierarchical genome assembly process workflow available in SMRT Analysis v2.3.0 (https://www.pacb.com/ products-and-services/analytical-software/smrt-analysis/, access at 2020/03/20) [38]. The long reads were combined with Illumina MiSeq paired-end reads of 300 bp for hybrid assembly using Unicycle [39]. Complete genomes were obtained when only one contig per chromosome/plasmid was used, and this contig could be circularized with overlapping reads of >100 bp at both ends.

Growth Curve Assays
Growth curve assays were performed in triplicate for 29 isolates; 11 CO and 18 SI isolates. Fresh bacterial cultures were grown overnight at 37 • C with shaking at 200 rpm in Luria-Bertani (LB) broth were diluted, separately transferred to 96-well microplates, and cultured as described above. Bacterial growth was determined by measuring the optical density at a wavelength of 600 nm (OD 600 ) for 24 h. Growth rates were determined by fitting the growth data to a logistic growth curve using GraphPad Prism8 and Equation (1), where Y 0 and Y M represent OD 600 values at time points 0 and M, respectively, and k is a constant calculated automatically by the program. The k value for ATCC19606 was set at 1.

In Vitro Competition Assay
The in vitro competition assay was performed as previously reported [48], with a few modifications. We chose a pair of SXT-susceptible CO and SXT-resistant SI isolates. Isolates cultures incubated separately overnight in LB broth at 37 • C were diluted to 5.0 × 10 5 colony-forming units (CFU)/mL, pooled together at a 1:1 ratio in LB broth, and co-cultured at 37 • C. At 0, 3, and 24 h, equivalent numbers of blend bacteria were collected, diluted, and plated on Mueller-Hinton plates with and without 16 mg/L SXT, separating SI from CO isolates. CFU numbers were determined after overnight incubation to assess the competitive growth of the CO and SI isolates. The competitive index (CI) was calculated as (CO CFU/SI CFU)/(inoculated CO CFU/inoculated SI CFU).

Biofilm Formation
Biofilm formation assays were performed as previously described [49]. Isolates were removed from the wells, and their optical densities were measured at OD 600 for growth. For biofilm staining, 150 µL of 10% crystal violet was added to each well and incubated for 30 min. The absorbance of the biofilms was measured at OD 585 . The adhesion unit was determined as OD 585 /OD 600 . The adhesion unit of biofilm formation for ATCC19606 was set at 1.

Serum Bactericidal Assay
Serum bactericidal assays were performed as previously described to determine serum sensitivity [50]. Aliquots of overnight LB broth diluted bacterial cultures in the mid-log growth phase were washed and resuspended in phosphate-buffered saline in triplicate. Then, the suspensions were added to normal human serum (NHS) and incubated at 37 • C for 3 h. Bacteria cultured in each suspension following inoculation and after incubation were serially diluted and plated. The serum bactericidal effect was expressed as percent survival. Survival rates were calculated as the ratio of the CFUs in the suspension of bacteria with and without NHS, and the percent survival for ATCC17978 was set at 100%.

Statistical Analysis
Statistical analyses were performed using GraphPad Prism 8. Comparisons between CO and SI isolates were performed using the two-tailed unpaired Student's t-test, the Chi-square test and the Mann-Whitney test. p values <0.05 were considered significant.

Conclusions
Our results suggest that bla OXA-23 co-occurrence on chromosome and plasmid does not enhance carbapenem resistance in A. baumannii but may result in fitness improvement. The different origins of bla OXA-23 and the different transposon insertion sites indicate that the co-existence of the resistance gene on chromosomes and plasmids may complicate its dissemination. We speculate that the simultaneous co-occurrence of multiple copies of bla OXA-23 on chromosomes and plasmids may have a relation with bacterial fitness increase in the host and environment, which will lead to severe conditions in patients, and increase the risk of the emergence and spread of CRAB.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/ 10.3390/antibiotics10101196/s1. Figure S1: correlation between antibiotics resistance and bla OXA-23 location analyzed by Student's t test, Figure S2: gene presence and absence variants of CO and SI isolates, Figure S3: blast analysis and visualization of bla OXA-23 -carrying plasmids, Table S1: phenotypic and genotypic characteristics of SI A. baumanii isolates.