Acinetobacter baumannii GC2 Sublineage Carrying the aac(6′)-Im Amikacin, Netilmicin, and Tobramycin Resistance Gene Cassette

ABSTRACT The aminoglycoside antibiotics amikacin, gentamicin, and tobramycin are important therapeutic options for Acinetobacter iinfections. Several genes that confer resistance to one or more of these antibiotics are prevalent in the globally distributed resistant clones of Acinetobacter baumannii, but the aac(6′)-Im (aacA16) gene (amikacin, netilmicin, and tobramycin resistance), first reported in isolates from South Korea, has rarely been reported since. In this study, GC2 isolates (1999 to 2002) from Brisbane, Australia, carrying aac(6′)-Im and belonging to the ST2:ST423:KL6:OCL1 type were identified and sequenced. The aac(6′)-Im gene and surrounds have been incorporated into one end of the IS26-bounded AbGRI2 antibiotic resistance island and are accompanied by a characteristic 70.3-kbp deletion of adjacent chromosome. The compete genome of the 1999 isolate F46 (RBH46) includes only two copies of ISAba1 (in AbGRI1-3 and upstream of ampC) but later isolates, which differ from one another by <10 single nucleotide differences (SND), carry two to seven additional shared copies. Several complete GC2 genomes with aac(6′)-Im in an AbGRI2 island (2004 to 2017; several countries) found in GenBank and two additional Australian A. baumannii isolates (2006) carry different gene sets, KL2, KL9, KL40, or KL52, at the capsule locus. These genomes include ISAba1 copies in a different set of shared locations. The distribution of SND between F46 and AYP-A2, a 2013 ST2:ST208:KL2:OCL1 isolate from Victoria, Australia, revealed that a 640-kbp segment that includes KL2 and the AbGRI1 resistance island replaces the corresponding region in F46. Over 1,000 A. baumannii draft genomes also include aac(6′)-Im, indicating that it is currently globally disseminated and significantly underreported. IMPORTANCE Aminoglycosides are important therapeutic options for treatment of Acinetobacter infections. Here, we show that a little-known aminoglycoside resistance gene, aac(6′)-Im (aacA16), that confers amikacin, netilmicin, and tobramycin resistance has been circulating undetected for many years in a sublineage of A. baumannii global clone 2 (GC2), generally with a second aminoglycoside resistance gene, aacC1, which confers resistance to gentamicin. These two genes are commonly found together in GC2 complete and draft genomes and globally distributed. One isolate appears to be ancestral, as its genome contains few ISAba1 copies, providing insight into the original source of this insertion sequence (IS), which is abundant in most GC2 isolates. Tracking ISAba1 spread can provide a simple means to track the development and ongoing evolution as well as the dissemination of specific lineages and detect the formation of many sublineages. The complete ancestral genome will provide an essential base point for tracking this process.


RESULTS
GC2 isolates containing aac(69)-Im. Initial screening using PCR had shown that four isolates, F4, F44, F46, and F48, from our small collection of multiply antibiotic-resistant isolates recovered in Brisbane between 1999 and 2002 belong to GC2 and carry the aac(69)-Im (aacA16) resistance gene, explaining an unusual combination of resistance to clinically relevant aminoglycoside antibiotics, namely, amikacin, netilmicin, and tobramycin. Additional resistance to gentamicin is supplied by the aacC1 gene. All isolates were also resistant to aminoglycosides, streptomycin, spectinomycin, and kanamycin, with resistance to neomycin variable. They were also resistant to ampicillin, ceftazidime, cefotaxime, sulfamethoxazole, tetracycline, trimethoprim, and the quinolones ciprofloxacin and nalidixic acid but were susceptible to the carbapenems imipenem and meropenem.
The draft genomes of these isolates were determined and their relationships examined (Table 1). For simplicity, the strain profiles include the sequence type (ST) in the Institut Pasteur (ST IP ) and the Oxford (ST Ox ) MLST schemes together with the identity of the gene clusters at the capsule (KL) and lipopolysaccharide outer core (OCL1) loci. Profiles are designated ST IP :ST OX :KL:OCL (24). F4, F44, F46, and F48 were distinct from the previously described GC2 (ST2:ST208:KL2:OCL1) members of the Brisbane collection (3,14) and formed a distinct GC2 group with the profile ST2:ST423:KL6:OCL1. Hence, they correspond to the ST69 group identified by Runnegar et al. (20). The oxaAb gene in the four isolates encodes the OXA-66 variant, as expected for GC2 isolates.
Complete genome of F46. To facilitate the examination of resistance islands and other features in the group, the complete genome of strain F46, recovered in 1999, was determined. F46 is the oldest strain in this group of aac(69)-Im strains and is among the earliest GC2 A. baumannii isolates recovered in Australian hospitals. The Unicycler hybrid assembly for strain F46 produced a 3,878,974-bp circular chromosome (GenBank accession number CP096575) and no plasmids.
A copy of ISAba1 in the appropriate orientation to enhance expression was found in the usual position (25) 9 bp upstream of the chromosomal ampC gene (allele 19), and this accounts for the resistance to ceftazidime and cefotaxime. An S81L substitution in GyrA together with an S84L substitution in ParC accounts for the nalidixic acid and ciprofloxacin resistance.
The acquired antibiotic resistance genes in the F46 genome (Table 1) were distributed between the two antibiotic resistance islands, AbGRI1 and AbGRI2, usually found in GC2 isolates. A previously described variant of AbGRI1 was present in F46. This variant, originally found in the genome of MDR-TJ (26) but now known as AbGRI1-3 (27), has a complex transposon structure (Fig. 1A). It interrupts the chromosomal comM gene and is surrounded by a 5-bp target site duplication (TSD). AbGRI1-3 contains the sul2 (sulfonamide), tetA(B) (tetracycline), and strA and strB (streptomycin) resistance genes (Fig. 1A). This 38,810-bp island differs from the proposed progenitor AbGRI1-0 (28) by a 2.85-kbp deletion in Tn6022, the addition of a 2.58-kbp fragment that includes tetA(B), and the incorporation of an additional copy of Tn6022 at the end of the tetA(B) gene.
The remaining acquired antibiotic resistance genes in F46, aacC1, aac(69)-Im (aacA16), aphA1b, and bla TEM-1D , were found in a version of AbGRI2, here called AbGRI2-Im (Fig. 1B), ISAba1  CR2Δ  CR2  sul2  tetR  tetA(B)  strA  strB  orf4b   IR 5393  IR L  IR R  IR L  IR R   IR L  IR R   orf4  sup  uspA  tniE  tniD  tniB  tniA  tniC   tniBΔ2  tniAb  tniCb  int  sup  uspA  tniEΔ  tniBΔ  tniA  tniC  orf   orf   orf4  orf5  orf6  orf7  orf8  orf9  orf10  orf11 AbGRI1-3 that has undergone multiple IS26-mediated rearrangements relative to the proposed precursor AbGRI2-0 (29). A 13.5-kbp fragment bounded by directly oriented copies of IS26 has been incorporated at the left end of the island, and 70.3 kbp of the adjacent chromosome corresponding to bases 1212167 to 1282446, locus tags A320_01150 to A320_01207, in the GC2 reference strain A320 (GenBank accession number CP032055) (29) has been deleted. The inserted fragment contains a partial class 1 integron with the aac(69)-Im gene cassette and an aadA1 cassette that has been split by an IS26-mediated inversion. The integron is next to a novel mercuric resistance operon (merDACPTR) that shares ,95% nucleotide identity with mer operons characterized in detail to date (30,31). The merD gene has been truncated by an IS26 and the IR mer is not present. In addition, multiple IS26-mediated deletions have removed the remnants of Tn9, Tn21, mer of Tn1696, and part of the class 1 integron that are found in AbGRI2-0. Antibiotic resistance islands in the genomes of F4, F44, F48, and K17. The enhanced draft genome sequences of F4, F44, and F48, which also contained aac(69)-Im, included 66 to 100 contigs (see Table S1 in the supplemental material). Screening draft genomes of our collection of over 200 Australian isolates for further isolates carrying KL6 revealed an additional ST2:ST423:KL6:OCL1 isolate, K17, carrying the aac(69)-Im gene that was recovered in Adelaide in 2002. K17 is isolate 6856390 in reference 32, and the enhanced draft genome of K17 is also reported here (Table S1).
The antibiotic resistance phenotypes of F48, F4, F44, and K17 and the resistance genes present were similar to those of F46. Like F46, they all included a copy of ISAba1 located 9 bp upstream of the ampC gene (allele 19) conferring resistance to ceftazidime and cefotaxime and the ParC and GyrA substitutions found in F46 accounting for the resistance to nalidixic acid and ciprofloxacin. F44 and F48 also include a copy of ISAba1 located 7 bp upstream of oxaAb (OXA-66). ISAba1 is in the correct orientation to drive expression of oxaAb, although F48 and F44 were susceptible to imipenem and only displayed reduced susceptibility to meropenem (2 mg/mL relative to 0.36/0.72 for F4 and F46; .8 mg/mL is considered resistant). This is consistent with reports that, even when overexpressed by the outward-facing promoter of ISAba1, OXA-66 does not confer resistance to meropenem and additional mutations are required to achieve this phenotype (33,34). The enhanced draft assemblies of F4, F44, F48, and K17 each also included AbGRI1-3 (Table 1).
Due to the presence of multiple copies of IS26, AbGRI2 was initially in 7 contigs in F48 and 6 contigs in F4, F44, and K17 which all lack the aphA1b kanamycin and neomycin resistance gene (Table 1). After assembly using a series of linkage PCRs, AbGRI2 in each of these isolates was found to be similar to AbGRI2-Im in F46 and all of them had an adjacent backbone deletion that was identical to the backbone deletion in F46 (Fig. 1C). In comparison to AbGRI2-Im in F46, F48 includes a 164-bp IS26-mediated deletion into aadA1. An additional copy of IS26 interrupting tnpA1000 was found in F4, F44, and K17, and homologous recombination between the two copies of IS26 bounding Tn6020 has removed aphA1b.
Relationships between Australian AbGRI2-Im isolates. The relationships were examined further by mapping ISAba1 positions and SND analysis. Only two copies of ISAba1 were detected in the chromosome of the 1999 isolate F46. One is in the AbGRI1-3 resistance region (position 1) and the second is the copy found upstream of the chromosomal ampC gene (position 2 [ Fig. 2A]). In contrast, we (35) and others (36,37) have found ISAba1 to be generally abundant in GC2 isolates. As the F46 genome is unusually free of ISAba1 copies, F46 likely represents the ancestor of the related isolates with additional ISAba1 copies in the chromosome.
In addition to ISAba1 copies at positions 1 and 2, F48 (date of isolation unknown) has an ISAba1 upstream of oxaAb (position 3) and only one further IS at position 4 in the chromosome (Fig. 2B). The ISAba1 at position 4 is also shared by the remaining isolates, which are all from 2002 and include ISAba1 copies at five further positions (5 to 9) as well as at positions 1 and 2 (Fig. 2C). Thus, it is likely that they evolved from F48. In addition, F44 has the ISAba1 upstream of oxaAb (position 3).
Comparison of the number of SNDs between the Australian KL6 isolates further supports the evolutionary links between the isolates. Compared to F46, F48 and each of the later isolates had acquired between 43 and 50 SNDs. Relative to one another, F48 and the 2002 isolates differed from one another only by between 5 and 11 SNDs, supporting a closer relationship.
Hence, it is possible that the strain imported from Indonesia in 2002 (20) resembled F48 and that the additional ISAba1 movement events occurred during the outbreak. Patient transfer could explain the strong similarity of the Adelaide isolate.
The aac(69)-Im gene in complete A. baumannii genomes from South Korea. The GenBank nonredundant nucleotide database was queried with the 552-bp aac(69)-Im gene sequence to determine how widespread the gene was among A. baumannii strains. Identical sequences were found in several complete GC2 genomes (Table 2). However, only two GC2 isolates recovered in South Korea, 1656-2 (38) and DU202 (39), also carried KL6 and shared the same profile (ST2:ST423:KL6:OCL1) as the Australian isolates reported here ( Table 2). The genome of 1656-2 (GenBank accession number CP001921) was one of the earliest A. baumannii complete genomes to be reported (38) and was used to define the KL6 capsule locus (40). However, an analysis of the resistance gene content was not reported. 1656-2 carries an unusual configuration of AbGRI1 (Fig. 3A), originally described by Huang et al. (26) and called RI 1656-2 but here designated AbGRI1 1656-2 . The span of AbGRI1-3 containing the tetA(B) gene is missing from the configuration in 1656-2 and an additional fragment containing the bla PER-1 gene and a second copy of strA has been acquired, likely via recombination within strA. 1656-2 also includes two plasmids (38).
Determining the original AbGRI2 configuration in 1656-2 ( Fig. 3B) was complicated by two inversions (Fig. 4A). One is an IS26-mediated inversion that has split the island, and the other is an inversion between two inversely oriented closely related (.98% nucleotide identity) integrated phage genomes that are not present in the F46 chromosome (Fig. 4B). Reversal of the inversions to give the presumed progenitor showed that it is derived from AbGRI2-Im in F46 (Fig. 3B), differing only by an IS26-mediated deletion of 8,084 bp that has removed the span between tnpR and the IS26 adjacent to intI1 (Fig. 3B). The chromosomal deletion to the left of AbGRI2 in 1656-2 is identical to the 70,280-bp deletion in F46, consistent with a shared origin. A second complete genome from South Korea, DU202 (GenBank accession number CP017152), which carries aac(69)-Im contains AbGRI1-3 as in F46 but has also acquired an AbGRI3 island (Table 2). Compared to AbGRI2 in F46, the corresponding island in DU202 differs by a further IS26-mediated deletion that has removed 943 bp at the left end of the island and has extended 7,501 bp further into the adjacent chromosome. At the right end of the island, an additional IS26-mediated inversion has occurred relative to F46, and an IS26-mediated deletion has removed the remaining span of the class 1 integron and 12,002 bp of the adjacent chromosome (Fig. 3B).
The IS26-bounded AbGRI2 island of the KL2 (ST208) isolates, including strain AYP-A2, recovered in Victoria, Australia, in 2013 (42), and six isolates, FDAARGOS_560, CFSAN093708, CFSAN093706, CFSAN093707, CFSAN093710, and CFSAN093709, from the United States (2004 to 2007), provides a pathway from the AbGRI2-0 progenitor to the AbGRI2-Im island in F46 (Fig. 5B). In addition to all the fragments found in F46, it also retains a larger span of the class 1 integron in AbGRI2-0, including the aadA1 and aacC1 gene cassettes and the sul1 gene, but this segment has been inverted by IS26 relative to AbGRI2-0. Reversing this inversion returns AbGRI2 in AYP-A2 to the configuration seen at the right end in AbGRI2-0, with the only difference being an IS26-mediated deletion that has removed the span between tnpA1 and IS6100 (Fig. 5B). The chromosomal deletion adjacent to the left end of the island in AYP-A2 is identical to the deletion seen in F46, again suggesting that the AbGRI2-Im-type island configurations in these isolates have a common origin. The remaining genomes, CFSAN093705, AB194-VUB, AB3-VUB, CUVET-MIC596, MDR-UNC, and OC043, all possess additional variations of the AbGRI2-Im island in AYP-A2 with larger adjacent chromosomal deletions (data not shown) but retain the aac(69)-Im gene cassette and novel mer regions.    Two alternate routes could link the KL2/ST208 and KL6/ST423 isolates. Either the AbGRI2-Im was shared at some stage between the ST2:ST423:KL6:OCL1 and ST2:ST208:KL2:OCL1 lineages or the K locus in the ancestral AbGRI2-Im strain was replaced by homologous recombination, as is known to occur (8,40,43,44). To examine this possibility, an SND analysis was used to detect recombination patches. Comparison of the AYP-A2 chromosome to that of F46 using Snippy revealed a large (640 kbp) recombination patch spanning across the origin and including both the KL and AbGRI1 (shown as an arc above the circular chromosome in Fig. 2A). This span in AYP-A2 contains 5,572 SNDs relative to the corresponding region in the F46 genome (0.87% difference). In contrast, the remaining ;3.4 Mb of the chromosome contains only 160 SNDs. However, this analysis did not identify which strain is ancestral.
ISAba1 distribution in complete genomes. One or more events involving movement of ISAba1 to a specific new location are, like single base changes, indicative of shared ancestry. Hence, the ISAba1 positions in the KL2, KL9, KL40, and KL52 isolates were also located. Insertion sequences (IS) shared by two or more genomes are shown in Fig. 6 together with the IS positions in the Australian and South Korean groups. The positions in the F46 genome of the 9-bp duplication surrounding each of these IS are recorded in Fig. 6. Only the ISAba1 in AbGRI1-3 and -5 (AbGRI1-1 lacks this ISAba1) and the copy upstream of ampC are shared   with either group of KL6 isolates. However, the genomes that include KL2, KL9, KL40, or KL52 all share ISAba1 copies at 8 to 17 additional positions (Fig. 6), consistent with a shared history combined with ongoing ISAba1 spread. Clearly, members of this group have also acquired new KL, presumably via introduction of a recombination patch.
ST2:ST1962:KL52:OCL1 isolates containing the aac(69)-Im gene in Australia. Granted the diversity of KL seen in the complete genomes, we searched our collection of genomes from Australian isolates again using the aac(69)-Im gene. This revealed that among 41 isolates recovered at the John Hunter Hospital in Newcastle, Australia, 2, H32 and H40 (GenBank accession numbers JASCXA000000000 and JASCXB000000000, respectively), recovered in 2006, carried aac(69)-Im.
H32 is resistant to amikacin, netilmicin, tobramycin, gentamicin, streptomycin, spectinomycin, ampicillin, ceftazidime, cefotaxime, sulfamethoxazole, tetracycline, trimethoprim, imipenem, meropenem, ciprofloxacin, and nalidixic acid. These resistances were due to the presence of the aac(69)-Im, aacC1, tet(B), strAB, sul2, bla TEM-1b , and aadA1 resistance genes in AbGRI1-5 and an AbGRI2-1m variant. In addition, the oxa23 gene confers resistance to carbapenems. An ISAba1 upstream of the chromosomal ampC gene is responsible for resistance to the third generation cephalosporins ceftazidime and cefotaxime, and the appropriate GyrA and ParC substitutions confer nalidixic acid and ciprofloxacin resistance. H40 had a resistance profile and resistance genes similar to those of H32 except for susceptibility to gentamicin due to loss of the aacC1 gene in H40. H40 also has AbGRI1-5. The AbGRI2-Im islands are deletion derivatives that still retain aac(69)-Im. In the case of H40, multiple IS26-mediated deletions have removed all of the island except for the IS26bounded aac(69)-Im fragment. However, the characteristic 70,280-bp adjacent deletion was present in both H32 and H40.
aac(69)-Im is found in many draft genomes. A search of the A. baumannii genomes in the GenBank WGS database (1 March 2023) with the aac(69)-Im gene sequence as a query retrieved a contig from 1,126 draft genomes of isolates that carried this gene (100% identity; 100% coverage). This is a significant proportion (;5%) of the currently available A. baumannii draft genomes. Granted that there are very few published reports of the aac(69)-Im gene (see Discussion), the number of genomes involved was unexpected and indicates that the presence of this gene is substantially underreported. Indeed, the presence in AYP-A2 was not reported previously (42). These isolates were recovered from 28 different countries over the time span 1999 to 2023 ( Table 3). The countries were mainly in Europe, the Middle East, Asia, and North America, but Australia (the isolates reported here and an additional Brisbane isolate from 2003) and Russia are also represented. This indicates a global dissemination. The sequence type of one representative genome from each country was examined, and all were ST2.

DISCUSSION
Following the original 1997 description of the aac(69)-Im in a class 1 integron-associated gene cassette in C. freundii (16) and its discovery in A. baumannii in South Korea in 2008 (15), it was rarely reported. It was found in a single GC2 A. baumannii isolate, said to be imported from Thailand, in a study of sporadic isolates recovered in Sweden (45) and also in a single GC2 A. baumannii isolate recovered between 2013 and 2014 in Nepal (46). The gene was added to the ResFinder database only in 2018 but as aac(69)-Ip and has been reported recently as -Ip in a small number of GC2 isolates from Israel and Italy (47). Granted the absence of a significant number of reports, our finding that aac(69)-Im is present in 15 complete and over 1,000 draft A. baumannii genomes, representing over 5% of currently available genomes, is surprising. This may reflect the fact that the focus of genomic studies is most often only on resistance to carbapenem antibiotics. However, the modest number of cases in which genomes have been analyzed in detail to document all of the resistance genes present and even fewer cases that assign them to specific known resistance island types is likely another factor. In addition, the aac(69)-Im gene appears to be almost exclusively present in A. baumannii and likely only in GC2 isolates, as searches of the GenBank nonredundant nucleotide database identified only one case of the gene in another species, namely, the original report (16). The source of the IS26-bounded AbGRI2 island has been identified (48). The proposed original configuration, AbGRI2-0, is present in A320 (RUH134), one of the earliest GC2 isolates for which genome data are available (29) (GenBank accession number CP032055). AbGRI2-0 includes further copies of IS26, and as a consequence of their action causing inversions and deletions, a large number of variants of this resistance island have been reported (3,29), and in many cases, IS26-mediated deletion of adjacent DNA is also observed. An inversion of the chromosome splitting AbGRI2, as seen in this study for 1656-2, is also seen in the A320 genome (29). However, the configurations reported here, which for simplicity and clarity we have designated collectively as AbGRI2-Im variants, are all derived from the configuration that arose as a consequence of the initial incorporation of an additional IS26-bounded segment. This AbGRI2 type had not been described previously. The addition also appears to have resulted from the action of IS26 in the targeted conservative mode (49,50) or via homologous recombination within the boundaries of IS26. Either prior or subsequent to this acquisition event, an IS26-mediated deletion of adjacent chromosomal DNA with a characteristic length occurred. Evolution of AbGRI2-Im in situ has continued to be shaped by the action of IS26, as evidenced by the several variations reported here (Fig. 1, 3, and 5).
Based on the analysis of ISAba1 distribution in the isolates described or analyzed in this study, the 1999 isolate F46 we sequenced appears to be close to the progenitor of all of the groups with different sets of shared ISAba1 locations. Indeed, it seems likely that ISAba1 entered the GC2 forebear in AbGRI1, as this IS is in the transposon that is the source of the ancestor of the AbGRI1-type resistance islands (28). If this is correct, an ISAba1 subsequently moved to the position upstream of ampC. This would parallel the situation in global clone 1, where it is known that the earliest isolates do not include any ISAba1 copies (51; unpublished observations). In one GC1 sublineage, the source of the ISAba1 has been traced to the entry of Tn6168 carrying a second copy of the ampC gene, and two further sublineages could be distinguished via the different patterns of shared ISAba1 positions (43). Further work tracking ISAba1 copies in genomes of isolates from different continents and countries may shed light on how this GC2 type has spread. For, example, the 1656-2 genome, which carries an AbGRI2-Im variant, was included in a phylogeny of GC2 isolates (52), and this revealed a close relationship with two isolates from an earlier study of a polyclonal outbreak in 2007 at the NIH Clinical Center, Bethesda, MD (44). Resistance genes were not examined in that study, but our examination of these draft genomes revealed that both isolates carried aac(69)-Im. The authors reported recombinational exchange involving the K locus, but the study predates the assignment of KL numbers (40). We found that one isolate, designated NIH-3, carried KL6 while the other, NIH-2, carried KL9. Analysis of the ISAba1 locations in the KL6 isolate relative to those found in the Australian and South Korean KL6 groups may provide insight into the source of this isolate.
It is well known that extensive recombinational exchange occurs in A. baumannii, and this complicates analysis of the relationships within clonal complexes. Among the earliest examples were exchanges of the region that includes that capsule locus (KL) (40), and the isolates that include AbGRI2-Im with different KL found in this study could have arisen by import of a different KL-containing segment. Recombination in this region was observed in the comparison of the F46 (KL6) and AYP-A2 (KL2) genomes and was reported previously for the NIH-3 (KL6) and NIH-2 (KL9) strains described above (44). Alternatively, import of the AbGRI2-Im region could also have occurred. Further work in the form of an in-depth analysis (ISAba1 distribution, SND analysis to detect large recombination patches, etc.) will be needed to determine if these two possibilities can be distinguished.
Finally, in the course of this study we detected a serious problem with the names assigned to aminoglycoside resistance genes, as two quite distinct genes are listed as aac(69)-Im. The gene described here is the first reported and hence retains the designation -Im. The second was published several years later (53) and by convention must be renamed. We suggest -In, as this designation is currently not assigned.

MATERIALS AND METHODS
Bacterial strains. Acinetobacter baumannii isolates used are listed in Table 1. F4, F44, F46, and F48 (kindly supplied by Mohammad Katouli) are part of the collection of isolates reported previously (22) that were collected between 1999 and 2002 at the Royal Brisbane and Women's Hospital (Brisbane, Australia). These isolates are also referred to as RBH4, RBH44, RBH46, and RBH48. Isolate K17 (original name isolate 6856390) was collected in 2002 at the Royal Adelaide Hospital (32) and was kindly supplied by Melissa Brown. Isolates H32 and H40 were collected at the John Hunter Hospital (Newcastle, Australia) in 2006 and were kindly supplied by John Ferguson.
PCR. The aac(69)-Im gene was amplified from whole-cell DNA using primers RH569 and RH570 (14) under standard conditions (60°C annealing temperature, 30-s extension time, 30 cycles) using Taq DNA polymerase (New England Biolabs) according to manufacturer's instructions. The PCR amplicon was sequenced by Sanger sequencing (Australian Genome Research Facility) to confirm the identity of the amplicon. PCR, carried out using standard conditions described above, was used to establish linkages between contigs in the draft genomes associated with resistance genes that were detected using Abricate version 1.0.0. Primers were described previously (29) or designed based on the F46 sequence to span the region between additional pairs of contigs.
Genome sequencing and assembly. Whole-cell DNA was prepared and sequenced on an Illumina HiSeq platform at the Wellcome Trust Sanger Institute (Cambridge, UK) as described previously (8). Sequencing generated 1,500,338 to 2,645,824 paired-end reads that were 100 bp in length after trimming to remove adapters. This represents 42-to 66-fold coverage of the entire genome (Table S1). Read quality was validated using FastQC (https://qubeshub.org/resources/fastqc). For draft genomes, the Illumina reads were assembled de novo using SPAdes (version 3.15.5) with default assembly parameters (k-mers of 21, 33, 55, and 77). Assemblies were checked for completeness and contamination using CheckM (54).
The antibiotic resistance islands were assembled from contigs in the draft using linkage PCR as described above, to link adjacent contigs generating enhanced draft assemblies containing 66 to 109 contigs (Table S1).
Whole-cell DNA from one isolate, F46, was also sequenced on a PacBio RS platform (DNA Link, South Korea) as described previously (35) to resolve complex repeat regions. This generated 182,344 reads with an average length of 6,028 bp, representing 283-fold coverage. The Illumina and PacBio reads from F46 were assembled de novo in a hybrid assembly using Unicycler version 0.4.0 (55) with default assembly parameters as described previously (35).
The protein-coding, rRNA, and tRNA gene sequences were annotated using Prokka version 1.14.5, the polysaccharide loci were identified and annotated using Kaptive version 2.0.0 (56), and the antibiotic resistance genes were annotated manually to correct names assigned by Abricate. Multilocus sequence typing (MLST) was done in silico using PubMLST (http://pubmlst.org). The ampC allele was identified using the ampC database on the PubMLST platform (57).
Antibiotic resistance and analysis of resistance regions. Resistance to antibiotics was determined using a standard disk diffusion assay. Meropenem MICs were determined by plating on doubling concentrations of meropenem or with meropenem Etest strips (bioMérieux, France) with a known meropenem-sensitive isolate, G13 (33), as a control. Antibiotic resistance regions in complete and draft genomes were identified as AbGRI1, AbGRI2, or AbGRI3 by their location and their content was typed using an in-house database of known AbGRI1, AbGRI2, and AbGRI3 structures (11,27,58) with stand-alone BLAST.
Identification of genomes carrying aac(69)-Im. Genomes containing aac(69)-Im were identified by performing a BLASTn search (https://blast.ncbi.nlm.nih.gov/Blast.cgi) of the GenBank nonredundant nucleotide or whole-genome sequence (WGS) databases with the aac(69)-Im gene as the query. Cutoffs of 99% identity and 100% coverage were applied. The original aac(69)-Im gene sequence (GenBank accession number Z54241) was found to contain a single nucleotide deletion that introduces a frameshift near the C terminus. Hence, the aac(69)-Im (but annotated as aacA16 [19]) gene sequence (GenBank accession number NG052380) which is derived from the complete genome of A. baumannii strain 1656-2 (GenBank accession number CP001921) was used as the query.
ISAba1 mapping. ISMapper (59) was used to map the locations of copies of ISAba1 in strains examined in this study. The complete genome of F46, the oldest isolate in the collection, was used as the reference sequence. The sequence of ISAba1, retrieved from ISFinder (https://isfinder.biotoul.fr/), was used as the query sequence, and paired-end Illumina reads were the input for the remaining isolates of interest. Where paired-end reads were not available for completed genomes in the public domain, one million MiSeq reads were simulated from the FASTA file of the complete genome using InSilicoSeq version 1.5.4 (60). After mapping, the individual output text files were compiled using a compiled_table.py script and were sorted by prevalence. The individual IS locations in the chromosome were compiled and identified by their location in the F46 reference genome.
Mapping of SNDs. Single nucleotide differences (SNDs) were mapped using Snippy version 3.0 (https://github.com/tseemann/snippy) with default parameters on the University of Sydney Artemis High Performance Computing cluster. The complete F46 genome was used as the reference sequence, with pairedend Illumina reads from F4, F44, F48, K17, and AYP-A2 as a query input. Illumina reads for the South Korean isolates 1656-2 and DU202 were not available, so their complete genomes were used as a reference with the F46 paired-end Illumina reads as a query. For comparisons between F4, F44, F48, and K17, the reads of each isolate and the draft sequence of each isolate were used sequentially as either the query or the reference.
Data availability. The draft sequences of F4, F4, F48, K17, H32, and H40 and the complete genome of F46 have been deposited in GenBank (BioProject PRJNA812061) under accession numbers JAKZLG000000000, JAQJIR000000000, JAQJIQ000000000, JAQJIP000000000, JASCXA000000000, JASCXB000000000, and CP096575, respectively. Reads for all samples are available under the BioSample and SRA accession numbers listed in the respective GenBank accessions.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. SUPPLEMENTAL FILE 1, DOCX file, 0.01 MB.