Genomic epidemiology of carbapenem resistance in Acinetobacter baumannii in Chile Authors: Pablo Araya-Vega, Maximiliano Matus-Köhler, Victoria Mardones-Verdugo, Sofía Gómez-Inostroza, Paulina González-Muñoz, Esteban Paredes-Osses, Freddy Roach-Poblete, Gerardo González-Rocha, Andrés Opazo-Capurro Published: 2026 ABSTRACT AbstractBackgroundCarbapenem-resistant Acinetobacter baumannii (CRAb) is a World Health Organization (WHO) critical priority pathogen, yet genomic data from Chile are scarce. Here, we investigated the genetic diversity of Chilean isolates, focusing on molecular epidemiology, carbapenemase genes, and mobile genetic elements.MethodsWe analysed 83 A. baumannii genomes (42 newly sequenced, 41 public). Newly sequenced isolates underwent antimicrobial susceptibility testing, and whole-genome sequencing was performed using Illumina, with hybrid assemblies for two CRAb strains. Genomes were assembled and analysed for STs, KL/OCL loci, resistance genes, plasmids and mobile elements, supported by phylogenomics.ResultsMLST revealed a diverse population structure dominated by ST15/IC4 (n = 26) and ST318/IC4 (n = 13), followed by ST109/singleton (n = 14), ST162/singleton (n = 13), ST1/IC1 (n = 9), ST79/IC5 (n = 5), and three additional singletons. Four isolates were carbapenem-susceptible, including three ST109 from the 1990s and one ST15 from 2016. Overall, 79/83 isolates were classified as carbapenem-non-susceptible based on phenotypic data and/or genomic determinants. Among these, blaOXA-58 (n = 24) and blaOXA-23 (n = 21) predominated, while blaNDM-1 was detected in a single isolate. Additionally, 33 genomes harboured the ISAba1–blaOXA-51-like arrangement, with the OXA-219 variant strongly associated with ST15 and ST318.ConclusionsThis three-decade genomic analysis shows sustained circulation of IC4 (ST15/ST318) in Chile. Carbapenem resistance was mainly driven by blaOXA-23 on conjugative Rp-T1 plasmids, blaOXA-58 on mobilizable R3-T14 plasmids, and frequent ISAba1–blaOXA-51-like arrangements, especially OXA-219. The persistence of plasmids and mobile elements highlights ongoing horizontal gene transfer and the importance of genomic surveillance for infection control in Chile. Introduction Carbapenem-resistant Acinetobacter baumannii (CRAb) is listed by the World Health Organization (WHO) as a critical priority pathogen list for research, discovery, and development of new antibacterial drugs.1 In 2024, the European Centre for Disease Prevention and Control (ECDC) highlighted CRAb as an emerging public health threat.2 Globally, CRAb imposes a substantial mortality burden, with thousands of deaths estimated in Latin America.3 In Chile, A. baumannii is a major cause of bloodstream infections and ventilator-associated pneumonia (VAP), with carbapenem resistance near ∼60%.4 Recent global reports underscore the endemicity of CRAb across regions and the prominence of plasmid-borne blaOXA-23 in international dissemination.5,6 Carbapenem resistance in A. baumannii is predominantly mediated by OXA-type class D carbapenemases (blaOXA-23/40/58/143/235).5 Moreover, overexpression of intrinsic OXA-51-like enzymes, classically associated to ISAba1 upstream the blaOXA-51-like gene, can also confer resistance, with variable activity among OXA-51-like variants.7 In contrast, other carbapenemases (e.g. NDM, IMP/VIM, KPC, GES) are less frequent in this species.4,8 Epidemiologically, CRAb is structured into globally disseminated international clones (IC1–IC9).4,9 In this context, MLST (Pasteur/Oxford) links clonal complexes to ICs, with the Pasteur scheme preferred for reduced recombination.9 While IC1/CC1Pas and IC2/CC2Pas dominate worldwide, Latin America shows greater representation of IC5/CC79Pas and IC4.4–6 In Chile, earlier studies documented IC4 circulation, particularly ST15 and the single-locus variant ST318, indicating divergence from broader regional trends.10 Complementing this, a Chilean study of 34 CRAb isolates (2010–2013) from Santiago hospitals reported ST1, ST15, ST79, ST162 and ST109, containing small Rep_3 (R3) plasmids carrying blaOXA-58 and aminoglycosides-modifying enzymes genes and novel transposon contexts (e.g. Tn6925), revealing multiple routes for resistance gene acquisition.6 Additionally, recent advances in whole-genome sequencing (WGS) have refined our understanding of replicon families that disseminate OXA-type carbapenemases—particularly Rp-T1 (formerly repAci6) and R3 types.11,12 Building on these advances, here we examined the molecular epidemiology, resistome, plasmids and mobile genetic elements of A. baumannii in Chile collected between 1990 and 2024, including both newly sequenced isolates and genomes deposited in public databases. Materials and methods Genomes and bacterial isolates collection This study comprises a total of 83 genomes, including 41 publicly available genomes and 42 newly sequenced clinical isolates (Table S1, available as Supplementary data at JAC-AMR Online). The 41 public genomes were retrieved from the Bacterial and Viral Bioinformatics Resource Centre (BV-BRC; https://www.bv-brc.org/) on 10 April 2025. In the case of newly sequenced isolates (n = 42), these were selected to capture a broad temporal range (1990–2024) and geographical distribution across the country, with emphasis on clinical relevance and representation of diverse STs. The temporal distribution of the complete dataset spans the following periods: 1990–1999 (n = 3), 2000–2009 (n = 5), 2010–2019 (n = 53), and 2020–2024 (n = 22). Geographically, isolates originated from multiple regions, including the Metropolitan Region (n = 48), Biobío (n = 12), Maule (n = 7), La Araucanía (n = 7), Los Lagos (n = 3), Ñuble (n = 3), Coquimbo (n = 2), and Valparaíso (n = 1), spanning ∼1300 km (Table S1). It is important to remark that our dataset does not represent a systematic or population-based sampling of A. baumannii across Chile, and overrepresentation of specific lineages cannot be excluded. For the 42 newly isolates included in this study, species identification was performed using multiplex PCR.13 Antibiotic susceptibility testing (AST) Antimicrobial susceptibility testing (AST) was performed only for the 42 isolates from our biorepository, which are predominantly distributed in the period 2010–2024. In the case of the 41 isolates publicly available, 34 had been previously characterized by Brito et al.6 including phenotypic AST confirming carbapenem resistance. For the remaining public genomes (n = 7), phenotypic data were not available, in consequence, carbapenem-resistance status was inferred from their genomic information. AST was carried out using the disk diffusion method, following the CLSI guidelines (CLSI-M100 Ed. 35).14 The antibiotics tested were imipenem, meropenem, durlobactam/sulbactam, ampicillin/sulbactam, piperacillin/tazobactam, ceftazidime, cefotaxime, cefepime, levofloxacin, ciprofloxacin, sulfamethoxazole/trimethoprim, gentamicin, and amikacin. For colistin and tigecycline, the minimum inhibitory concentration (MIC) was determined using the broth microdilution method. Due to the absence of breakpoints for tigecycline for Acinetobacter, we considered a MIC ≤ 1 mg/L as breakpoint for susceptible, based on pharmacokinetic/pharmacodynamic (PK/PD) properties.15 Although resulting resistance patterns indicate reduced susceptibility across multiple antimicrobial classes; no formal classification of multidrug-resistant (MDR), extensively drug-resistant (XDR) or pandrug-resistant (PDR) phenotypes were applied, and resistance is therefore described here in qualitative terms based on observed susceptibility profile or associated with the presence of resistant-determinant. Moreover, I and R isolates were classified as non-susceptible to both carbapenems. Whole-genome sequencing and bioinformatic analysis Short-read sequencing was carried out for all 42 A. baumannii from our biorepository. Genomic DNA (gDNA) was extracted using Chelex 100 resin (Bio-Rad®, USA). Library preparation was performed using the Illumina DNA Prep kit and sequenced on an Illumina NovaSeq X Plus platform. Raw reads were assembled using Skesa (v0.24)16 and annotated with Bakta web (v1.11.0).17 For all genomes, assembly quality was assessed using CheckM (v1.2.3) and Quast (v5.3.0).18,19 Only genomes meeting the criteria of ≥90% completeness and <5% contamination were considered for further analysis to ensure consistency and reliability. For species confirmation, fastANI (v1.3) was performed using the reference strain A. baumannii ATCC 19606 considering a minimum threshold of 97%.20 To homogenize the genome analysis, all 83 genomes were analysed for sequence type (ST) using MLST based on the Pasteur scheme (Pas) utilizing the PathogenWatch platform (https://pathogen.watch). We employed the Pas scheme due to its greater stability compared with the Oxford scheme, as it is less susceptible to homologous recombination.9 The capsular polysaccharide (KL) and the lipooligosaccharide outer core (OCL) were identified using Kaptive.9 Moreover, ABRIcate (v1.0.1) was used to screen for antimicrobial-resistance genes utilizing the NCBI Bacterial AMR Reference Gene Database (≥90% identity and coverage).21 Genomes carrying acquired carbapenemases were searched for mobile genetic elements (MGE) using the Mobile Element Finder (v1.0.3, 2020-10-09) tool.22 A900 (OXA-58+) and A914 (OXA-23+) isolates underwent long-read sequencing to resolve carbapenemase-encoding plasmid structures. For this purpose, gDNA was extracted with the QIAamp DNA Micro Kit, and DNA libraries were prepared using the Rapid Barcoding Kit (SQK-RBK114) and sequenced on the MinION platform utilizing a R10.4.1 flow cell. Hybrid assembly was performed using Unicycler (v0.5.1).23 Plasmids in silico analysis Carbapenemases-encoding plasmids were reconstructed in the CRAb isolates by MOB-Recon tool (v3.1.9).24 Then, the contigs predicted as plasmids were analysed by the Acinetobacter plasmid typing (APT) scheme.12 Next, these plasmids contigs were screened for carbapenemases genes by ABRicate v1.0.1 utilizing the NCBI Bacterial AMR Reference Gene Database. Moreover, we reconstructed the plasmids from the hybrid-assembled genomes of A900 (OXA-58+) and A914 (OXA-23+) and used these assemblies as reference sequences to evaluate the plasmids inferred by MOB-Recon from short-read data alone. Phylogenetic analysis Phylogenetic analysis included all genomes sequenced from this study (n = 42) and genomes from isolates previously reported in Chile and retrieved from BV-BRC (n = 41). Then, a core-genome alignment was carried out by Roary (v3.13.0),25 and SNPs sites within the core genome were extracted using SNP-sites (v2.5.1).26 To refine the phylogenetic tree analysis by excluding horizontally transferred sequences, Gubbins (v3.2.1)27 was employed to identify and remove high recombination regions from the alignments. Next, maximum likelihood (ML) phylogeny was inferred using IQTree (v2.4.0) under the GTR+F+I+R3 model with 2000 iterations suggested by the ModelFinder tool incorporated in IQtree (v2.4.0).28 The resulting tree was visualized and edited in iToL (v7.2).29 Results Multiplex PCR preliminary identified all 42 isolates from our collection as A. baumannii, whereas ANI analysis confirmed species assignment for all 83 genomes analysed. AST of the 42 A. baumannii isolates revealed high levels of resistance to most antibiotic classes (Figure 1). Carbapenem non-susceptibility was observed in 38/42 isolates (90.5%) for both imipenem and meropenem. Carbapenem non-susceptible isolates were defined as those classified as R or I according to CLSI criteria (see Materials and methods). High resistance rates were also detected for third-generation cephalosporins, including ceftazidime (42/42, 100%) and cefotaxime (42/42, 100%), as well as for piperacillin/tazobactam (41/42, 97.6%) and fluoroquinolones, including ciprofloxacin (40/42, 95.2%) and levofloxacin (40/42, 95.2%) (Figure 1). Cefepime showed limited activity, with only a small number of susceptible isolates (n = 2), while the majority were resistant or intermediate (R+I: 40/42, 95.2%). Among alternative agents, ampicillin/sulbactam showed retained activity in a subset of isolates (S: 14/42, 33.3%), and sulfamethoxazole/trimethoprim exhibited moderate activity (S: 19/42, 45.2%). Aminoglycosides showed variable activity, with gentamicin retaining greater activity (S: 15/42, 35.7%) compared with amikacin (S: 5/42, 11.9%). Notably, durlobactam/sulbactam showed high activity, with 40/42 (95.2%) isolates classified as susceptible. For last-line agents, most isolates showed intermediate susceptibility to colistin (41/42, 97.6%), while 10/42 (23.8%) isolates were resistant to tigecycline. As shown in Figure 1, resistance rates to most antibiotic classes are consistently high within this dataset, particularly for β-lactams and fluoroquinolones, whereas retained activity is mainly observed for durlobactam/sulbactam and, to a lesser extent, colistin and tigecycline. Antimicrobial susceptibility of Acinetobacter baumannii clinical isolates from Chile. AMK, amikacin; CAZ, ceftazidime; CIP, ciprofloxacin; CST, colistin; CTX, cefotaxime; FEP, cefepime; GEN, gentamycin; IMP, imipenem; LEV, levofloxacin; MEM, meropenem; SAM, ampicillin/sulbactam; SUD, durlobactam/sulbactam; SXT, sulphamethoxazole-trimethoprim; TGC, tigecycline; TZP, piperacillin/tazobactam. The CLSI guidelines consider resistance and intermediate categories for CST.14 The predominant lineage corresponded to the CC15/IC4, represented by ST15 (n = 26) and ST318 (n = 13). Additional STs comprised ST162 (CC162) (n = 13), ST79 (CC79/IC5) (n = 5), ST109 (n = 14), and ST1 (CC1/IC1) (n = 9). Moreover, singleton STs such as ST876, ST193 and ST241 were also detected (Figure 2). These STs exhibited a wide temporal and geographic distribution across multiple regions of Chile, including the Metropolitan Region, Biobío, Maule, La Araucanía, Los Lagos, Ñuble, Coquimbo, and Valparaíso (Figures 2 and 3), supporting the broad circulation of major lineages such as IC4 within the analysed dataset. Among the CSAb strains, three were isolated in the Biobío region during the 1990s, and one additional susceptible strain was collected in the Los Lagos region in 2016 (Figure 2). These CSAb isolates are clustered with the carbapenem-resistant ST109 isolate A375, which, in contrast, harbours the OXA-58-like carbapenemase mediating resistance to carbapenems (Figure 2). Eight distinct capsular polysaccharide loci (KL) were identified (Figure 2), with KL22 predominating (n = 36; 43.4%), followed by KL9 (n = 18; 21.7%), KL19 (n = 13; 15.7%), KL1 (n = 9; 10.8%) and KL49 (n = 4; 4.8%). KL11, KL48 and KL65 were detected at low frequencies. Six OCL variants were identified (Figure 2), with OCL7 being the most prevalent (n = 32; 38.6%), followed by OCL1 (n = 23; 27.7%), OCL3 (n = 21; 25.3%) and OCL10 (n = 5; 6%), while OCL6 and OCL8 were rare. Maximum-likelihood (ML) phylogenetic tree of 83 Acinetobacter baumannii isolates from Chile (1990–2024). The tree is based on core-genome SNPs and is annotated with sequence type (ST; Pasteur scheme), international clone (IC), capsular polysaccharide (KL) and lipooligosaccharide outer core (OCL) loci, intrinsic blaOXA-51-like allelic variants, acquired carbapenemase genes, predicted plasmid replicon types, and total number of antibiotic resistance genes (Sum ARG). Phenotypic susceptibility data were not available for most publicly retrieved genomes; therefore, absence of CSAb designation does not imply confirmed resistance. Temporal distribution of Acinetobacter baumannii isolates from Chile (1990–2024) according to sequence type (ST; Pasteur scheme), international clone (IC), and capsular (KL) and lipooligosaccharide outer core (OCL) loci. The blaOXA-51-like gene was detected in all genomes analysed, with allelic variants showing strong associations with specific lineages (Figure 2).30 The most prevalent variant was blaOXA-219, identified in 41/83 (49.4%) genomes and mainly associated with ST15, ST318/IC4 and ST1/IC1. This allele differs from the original blaOXA-51 by a L167 V substitution and was first described in 2014 in strains belonging to the WW4 lineage.31 Among previously deposited Chilean genomes, blaOXA-219 had been reported in a carbapenem-resistant isolate collected in 2017, in association with an upstream ISAba1 element (Figure 2). Notably, blaOXA-219 was also detected in ST1 strains, despite blaOXA-69 being typically linked to this lineage.32 Co-occurrence of blaOXA-219 with acquired OXA-type carbapenemases was observed in 10 genomes, including blaOXA-58 (n = 8) and blaOXA-23 (n = 2). The second most frequent allele was blaOXA-51, detected in 17 genomes, predominantly in ST162 (n = 13) and ST15 (n = 4). This variant, originally described in Argentina in 1996, is generally not associated with carbapenem resistance due to its limited hydrolytic activity.33 Sixteen of these genomes co-harboured an acquired carbapenemase gene (blaOXA-23, n = 14; blaOXA-58, n = 2), while a single carbapenem-susceptible isolate lacked additional OXA genes (Figure 2). The third most prevalent variant was blaOXA-67 (NCBI accession: NG_049807.1), identified in 14 ST109 genomes (Figure 2), a lineage previously reported in Argentina. Among these, 11 isolates also carried blaOXA-58, whereas three lacked acquired carbapenemase genes and resulted phenotypically susceptible to carbapenems. Less frequent alleles included blaOXA-65 in ST79/IC5 genomes (n = 5), all co-harbouring blaOXA-23, and blaOXA-69 in two ST1/IC1 genomes, both associated with blaOXA-58 (Figure 2). Single occurrences of blaOXA-106 (ST876, with blaOXA-23), blaOXA-91 (ST241, no additional carbapenemases), blaOXA-262 (ST15, no acquired carbapenemases), and blaOXA-120 (ST193, co-harbouring blaNDM-1) were also identified. Overall, these findings highlight substantial allelic diversity within the blaOXA-51-like family among our collection, with lineage-specific distributions and frequent co-occurrence with acquired carbapenemase genes. In 15 of the 22 blaOXA-23-positive genomes, this gene was located on contigs predicted to be plasmid-derived by MOB-Recon, while chromosomal localization was inferred for the remaining seven. The reconstructed blaOXA-23-carrying plasmids had a mean size of ∼82 kb and contained a replicase variant matching plasmid pACICU2 (CP031382.1), classifying them as Rp-T1 (repAci6) replicon type plasmids (Table 1). These plasmids encode conjugation-associated genes and were predicted to be conjugative, consistent with their recognized role in blaOXA-23 dissemination.12 Although previously reported mainly in IC2 and sporadically in IC4,12 in this study they were predominantly identified in ST162 (11/15), with additional occurrences in IC4 (ST15 and ST318) and single isolates belonging to ST109 and ST876 (Figure 2). Furthermore, blaOXA-23 was associated with transposable elements in 20/22 genomes, specifically transposons Tn2007 and Tn2008. Table 1.Summary of plasmid types carrying acquired carbapenemases in the Acinetobacter baumannii genomes analysed in this studyPlasmid typeAcquired carbapenemaseAverage plasmid size (bp)Mobility predictionSequence types (Pas)N° of genomesRP_T1OXA-2382 000ConjugativeST162 (11), ST15 (1), ST318 (1), ST109 (1) and ST876 (1)15R3_T14OXA-5817 000MobilizableST318 (6), ST109 (2) and ST162 (1)9 rep-less_1OXA-5820 068aMobilizableaST162a and ST318b24 497bNon-mobilizableb rep-less_2NDM-145 562ConjugativeST1931ST, sequence type (Pasteur scheme). aMobilizable plasmid. bNon-mobilizable plasmid. Summary of plasmid types carrying acquired carbapenemases in the Acinetobacter baumannii genomes analysed in this study ST, sequence type (Pasteur scheme). aMobilizable plasmid. bNon-mobilizable plasmid. For the blaOXA-58 carbapenemase, an even distribution between plasmid (n = 11) and chromosomal (n = 12) localization was observed based on MOB-Recon predictions. The blaOXA-58-carrying plasmids had a mean size of ∼17 kb, of which nine were classified as R3-T14 type, showing 100% identity and coverage with the replicase gene of plasmid pAb242 (accession number KY984056.1) (Table 1). Two additional plasmids lacked identifiable replicase genes and were classified as rep-less. R3-T14 plasmids correspond to the previously described GR2 group based on Aci1/Aci2 replicases.12 Although this plasmid group has been less extensively studied, related variants (e.g. R3-T19) have been implicated in blaOXA-58 dissemination.30 In this study, R3-T14 plasmids were predominantly identified in ST318/IC4, with additional detection in ST109 and ST162. In silico mobility prediction analysis indicated that 10 of these plasmids were mobilizable. One rep-less plasmid was predicted to be non-mobilizable and was notably smaller (4497 bp). The blaNDM-1 gene was detected in a single genome (UCO-558) (Figure 2) and was located on a plasmid of ∼45 kb. This plasmid was predicted to be conjugative but could not be assigned to a replicon type due to the absence of any rep genes. Comparative analyses of plasmid sequences revealed marked structural conservation across isolates collected more than a decade apart. Using the circularized plasmid from isolate A914 (OXA-23+, ∼92 kb) as reference, plasmids reconstructed from short-read assemblies showed 100% nucleotide identity but approximately 75% coverage, with uncovered regions mainly corresponding to putative integrons and insertion sequences. Comparison with fully closed blaOXA-23-carrying plasmids recovered from isolates collected in 2010 by Brito et al.6 demonstrated high similarity, with identities of 100, 92 and 92% and complete (100%) coverage of the reference sequences (UC24137, UC24371 and UC25604). These results indicate long-term stability of the plasmid backbone, with observed variation largely attributable to MGE rather than major structural rearrangements. A similar pattern was observed for the smaller plasmid harboured by isolate A900 (OXA-58+, ∼17 kb). Comparisons with plasmids recovered over multiple years, including a fully closed plasmid from 2010,6 revealed very high similarity, with BLASTn analyses showing 100% nucleotide identity and ∼97% coverage relative to the A900 reference. The minor uncovered regions likely reflect variation in short accessory segments rather than substantive structural divergence. As represented in the Figure 2, we identified several other antibiotic resistance genes, including those conferring resistance to aminoglycosides, fluoroquinolones, β-lactams, sulphonamides, and trimethoprim. Notably, aph(3’)-VIa was detected in 71/83 genomes (85.5%), primarily associated with resistance to amikacin; aac(3)-IIe in 34/83 (40.9%); ant(2'’)-Ia in 21/83 (25.3%); and ant(3'’)-IIa in all genomes (100%), along with other modifying enzymes such as aph(3'’)-Ib and aph(6)-Id. Specific mutations linked to fluoroquinolones-resistance were also observed, including gyrA_S81L (80/83, 96.4%), parC_S84F (14/83, 16.9%), and parC_S84L (61/83, 73.5%). From a clinical perspective, and consistent with the phenotypic resistance to ampicillin/sulbactam, we further detected the presence of other β-lactamase genes, with 30 isolates carrying a blaTEM-1. Finally, among the strains analysed, one isolate (A375) exhibited phenotypic resistance to colistin, which carried the pmrB_I164F mutation that has been linked to colistin resistance (Figure 2). In the case of the two durlobactam/sulbactam-resistant isolates, one (UCO558) harboured the blaNDM-1 gene, while the other presented the T526S mutation in PBP3. Both have been previously described as mechanisms conferring resistance to this drug.34 Additionally, we investigated the presence of metal resistance genes in the genomes included, mainly genes mediating resistance to mercury and nickel resistance. Notably, all genomes belonging to IC4 (ST15 and ST318) carried the mercury-resistance operon merA, merD, merR, and merT (Figure 2). Among ST1/IC1, 7/9 genomes carried five mercury resistance operon merA, merD, merE, merR, and merT, while the remaining two strains lacked any of them. In contrast, the nickel resistance gene nreB was present across all 83 genomes studied (Figure 2). Discussion In this study, we analysed two complementary sets of A. baumannii samples: (i) 42 isolates from our biorepository subjected to phenotypic AST and WGS, and (ii) 41 previously published genomes retrieved from public databases. A high prevalence of carbapenem non-susceptibility was observed among the 42 biorepository isolates, with 38/42 (R+I, 90.5%) resistant to both imipenem and meropenem. As illustrated in Figure 1, non-susceptibility to carbapenems, third-generation cephalosporins, piperacillin/tazobactam and fluoroquinolones was consistently high within this dataset. However, these phenotypic observations are primarily derived from isolates collected between 2010 and 2024 and therefore reflect resistance patterns in recent years rather than across the full 1990–2024 study period. The limited number of isolates from earlier decades (particularly the 1990s, n = 3, all carbapenem-susceptible) impedes robust inference regarding the emergence and long-term maintenance of multidrug resistance over the full temporal range. These results corroborate prior reports of extensive carbapenem and cephalosporin resistance in A. baumannii,10 while variable ampicillin/sulbactam susceptibility suggests limited temporal or regional shifts in an otherwise highly resistant phenotypes in recent decades. Resistance to third-generation cephalosporins, piperacillin/tazobactam, and fluoroquinolones remained consistent across the isolates, confirming their limited activity and the predominance of resistance to these drugs, despite occasional cefepime susceptibility. Notably, susceptibility to ampicillin/sulbactam in a subset of CRAb isolates supports its potential therapeutic role, in line with clinical evidence showing improved outcomes compared with colistin-based regimens. Susceptibility or intermediate profiles to colistin and TGC further indicate that these agents remain relevant options for severe A. baumannii infections. Importantly, 40/42 isolates (97.6%) from our biorepository were susceptible to durlobactam/sulbactam, which have been recently approved as an active drug against CRAb.35 One isolate (UCO558; Figure 2) was resistant to durlobactam/sulbactam, due to the production of NDM-1, which confers resistance to this drug,36 whereas another isolate (A375) presented the T526S mutation in PBP3, which has been associated to with elevated durlobactam/sulbactam MICs values (8–64 mg/L).34 These findings indicate extensive resistance across multiple antibiotic classes within the analysed isolates, consistent with a multidrug-resistant phenotype. However, it is important to note that formal MDR/XDR/PDR classifications were not applied, and therefore resistance is interpreted based on combined phenotypic profiles (Figure 1) and genomic determinants (Figure 2), rather than standardized surveillance definitions. Accordingly, comparisons with surveillance studies applying standardized MDR definitions should be made with caution. Importantly, a key limitation of our study is that phenotypic antimicrobial susceptibility data were only available for the biorepository isolates, whereas most publicly available genomes lacked corresponding AST data. As a result, carbapenem resistance across the full dataset could not be uniformly confirmed and is, in part, inferred from genomic determinants. An additional methodological consideration relates to the definition of carbapenem resistance. In this study, isolates categorized as I according to CLSI criteria were included within the CRAb group, reflecting a non-susceptibility-based definition. However, it should be noted that some studies apply stricter criteria (e.g. MIC ≥8 mg/L), which may affect direct comparisons of resistance rates across reports.37 Molecular epidemiological analysis confirms the predominance of IC4 in Chile, driven mainly by ST15 and ST318, with lower representation of ST162, ST109, and IC1, consistent with previous reports.6,10 Although IC2 predominates globally,5 the heterogeneous clonal structure observed here aligns with Latin American trends favouring IC4.5 Extending over three decades and spanning multiple regions, our data reveal the sustained detection and circulation of IC4 within the analysed dataset, KL/OCL diversification within ST15, and plasmid-associated dissemination of carbapenemases, suggesting long-term endemic circulation rather than repeated introductions. In contrast, ST162 appears persistently but less frequent, while ST1 and ST109 represent more transient lineages, underscoring the value of genomic surveillance to distinguish entrenched transmission from sporadic spread. Nevertheless, temporal patterns derived from genomic data (Figure 3) indicate that carbapenem-resistant lineages, particularly IC4, have been consistently detected since the mid-2000s, supporting the persistence of resistant clones over recent decade. The strong lineage-specific association between KL–OCL profiles observed here is consistent with global genomic studies of A. baumannii.29,30 IC4 (ST15 and ST318) was predominantly associated with KL22/OCL7, in agreement with reports from Brazil and Argentina, suggesting a conserved genomic background favouring persistence in South America.38 The presence of KL9 within some ST15 isolates parallels observations from Colombia, where capsular diversification within IC4 has been linked to horizontal gene transfer and immune selection.4 In contrast, ST1/IC1 and ST162 each displayed exclusive KL–OCL combinations (KL1/OCL1 and KL19/OCL1, respectively), although large-scale genomic analyses have shown that such stability may be lineage- and context-dependent.30 ST109 exhibited KL variability (KL9 and KL49) with a conserved OCL3 profile, indicating differential selective pressures on surface polysaccharide loci. The low prevalence of other KL and OCL variants (e.g. KL48, KL65, OCL6 and OCL8) likely reflects limited circulation or recent introduction into Chile.11,33 Considering the above, our findings suggest the long-term persistence and apparent predominance of IC4 (ST15/ST318) among carbapenem-resistant A. baumannii isolates in Chile. However, this observation should be interpreted with caution. Our dataset includes genomes derived from outbreak investigations and geographically clustered clinical isolates, which may disproportionately represent successful epidemic clones. Consequently, the observed lineage frequencies likely reflect cohort composition and local transmission dynamics rather than true national prevalence over time. Within this context, molecular epidemiological analysis indicates a strong representation of IC4 in the analysed collection, driven mainly by ST15 and ST318, with lower representation of ST162, ST109 and IC1, consistent with previous reports.6 Accordingly, the observed lineage distribution likely reflects sampling structure and local transmission dynamics rather than unbiased estimates of national lineage frequencies. In addition, the overrepresentation of isolates from the Metropolitan Region may disproportionately influence the apparent frequency of specific lineages, potentially reflecting local outbreak dynamics rather than geographically balanced national patterns. ST15 showed long-term circulation with a conserved KL22/OCL7 background and occasional KL9 variants (Figures 2 and 3), coupled with sustained multidrug resistance (Figure 1) and diverse carbapenemases, including intrinsic ISAba1-blaOXA-219, blaOXA-23, and blaOXA-58, carried on Rp-T1 and R3-T14 plasmids (Figure 2; Table 1), a combination that likely underpins its epidemiological success. ST318, appearing mainly in the 2010s–2020s, shared the same capsular profile but lacked KL diversification and was strongly associated with OXA-58 carried on R3-T14 plasmids, suggesting adaptive expansion driven by plasmid acquisition rather than surface polysaccharide variation (Figures 2 and 3; Table 1). Outside IC4, ST162 persisted at lower frequency from 2007–2024 with a stable KL19/OCL1 background, predominant OXA-23 carriage on Rp-T1 plasmids, and resistant phenotypes similar to IC4 clones (Figures 1–3; Table 1). In contrast, ST1 and ST109 appeared sporadically or were largely confined to earlier decades (Figures 2 and 3), the latter including the few carbapenem-susceptible isolates, suggesting that these lineages represent transient or declining clones overshadowed by the sustained dominance of IC4. Globally, CRAb populations are largely dominated by IC2, in which carbapenem resistance is mainly driven by the dissemination of chromosomally embedded blaOXA-23.4,5 In contrast, our collection reveals a markedly different epidemiological and genomic landscape. The Chilean isolates were dominated by non-IC2 lineages, particularly IC4 and IC1/IC5, and exhibited a heterogeneous array of carbapenem resistance mechanisms. A key distinguishing feature was the high prevalence of blaOXA-219, an intrinsic blaOXA-51-like variant strongly associated with IC4 (ST15 and ST318) and, unexpectedly, also detected in ST1/IC1. This contrasts with global datasets, where blaOXA-23 predominates and blaOXA-219 is rarely reported at comparable frequencies.4,5 The sustained detection of blaOXA-58, often carried on Rep_3 family plasmids and linked to specific STs such as ST318 and ST109, further differentiates our dataset from many regions where blaOXA-58 has been largely displaced. At the resistome level, however, our findings align with global trends, including the widespread presence of aminoglycoside-modifying enzymes and canonical gyrA/parC mutations conferring fluoroquinolone resistance. Notably, the high and lineage-specific prevalence of mercury and nickel resistance genes, particularly within IC4, represents a divergence from most global genomic surveys and suggests a potentially greater role for heavy metal–driven co-selection in the local ecology of A. baumannii. In conclusion, this 1990–2024 genomic investigation supports the sustained presence and epidemiological relevance of A. baumannii isolates belonging to IC4 (ST15/ST318) within the analysed Chilean genome collection, with carbapenem resistance driven by ISAba1–blaOXA-219, blaOXA-23 and blaOXA-58 on conserved plasmid backbones, underscoring the value of long-term genomic surveillance for regional antimicrobial resistance monitoring. Supplementary Material