We have performed a genome-wide comparison of 47 A. baumannii isolates reported from India. We screened the microbial database at NCBI for the complete genome sequences of A. baumannii isolates collected from Indian hospitals. We found a total of 278 A. baumannii genomes out of which 42 were complete. These genomes belonged to isolates reported from seven different hospitals across India during 2014-2020. In addition, five A. baumannii isolates, namely, AIIMS5, AIIMS7, MMC18, SRMC9, and SRMC18, collected in 2005 and maintained as a part of our laboratory collection were also sequenced in this study. The general information regarding the collection year, specimen type, and geographical location of all A. baumannii genomes involved in this study is given in Table 1 .

The genomic DNA from five A. baumannii isolates was extracted using the NaCl-CTAB method previously described (Wilson, 2001). The quality and quantity of the isolated DNA were checked on DeNovix DS-11 Spectrophotometer (DeNovix, Wilmington, DE). The QC passed DNA was used to prepare paired-end DNA libraries using the Illumina Nextera XT kit (Illumina, San Diego, CA) following the manufacturer’s guidelines. The DNA libraries were sequenced using Illumina NextSeq 500 platform with 2 x150 bp chemistry. The NGS QC toolkit v2.3.3 was used for quality check and filtering of raw reads using default parameters (PHRED quality score - 20) (Patel and Jain, 2012). Filtered high-quality reads were used to generate draft-genome using SPAdes de-novo assembler v3.11.1 (Bankevich et al., 2012). The genome coverage was calculated using BBMap v38.47 (https://sourceforge.net/projects/bbmap). The quality of assembled genome was checked using QUAST v5.0.2 (Gurevich et al., 2013). The annotations of all 47 A. baumannii genomes were carried out using Prokka v2.1.1 and comprehensive genome analysis service provided at Pathosystems Resource Integration Center (PATRIC) v3.6.5 (Seemann, 2014; Wattam et al., 2016).

The prediction of putative ARGs belonging to different resistance mechanisms was carried out using the Resistance Gene Identifier (RGI) tool provided at the Comprehensive Antibiotic Resistance Database (CARD) (Mcarthur et al., 2013). To identify ARGs, the A. baumannii genome assemblies were queried against the CARD database using perfect and strict criteria. The predicted ARGs showing sequence similarity between 97% to 100% were further classified into different resistance mechanisms (Alcock et al., 2020).

The putative VFGs present in A. baumannii genomes were predicted using an automatic pipeline called VFanalyzer provided at http://www.mgc.ac.cn/cgi-bin/VFs/v5/main.cgi (Liu et al., 2018). The VFanalyzer conducts a thorough search of the sequence similarity of VFGs with the known and predicted VFGs available in the virulence factors database for bacterial pathogens (VFDB). The identified VFGs were further classified into nine VF categories.

The prediction of plasmids in AIIMS5, AIIMS7, MMC18, SRMC9, and, SRMC18 genomes sequenced in this study was carried out using PlasmidSPAdes (Antipov et al., 2016). For analysis, the raw reads were used and based on the median chromosomal coverage the plasmid contigs were filtered. The filtered sequences were annotated using PATRIC v3.6.5. The plasmid sequences of 42 other A. baumannii genomes available at NCBI were downloaded. The presence of ARGs in all plasmid sequences was detected using CARD. The presence of insertion sequences and transposons was detected through a manual BLAST analysis of each A. baumannii genome against the ISfinder database (Siguier et al., 2006). The insertion sequences and transposons with an e-value equal to 0.0 were selected and further evaluated for the presence of ARGs. The prophage regions present within genome sequences were detected using the PHAge search tool enhanced release (PHASTER) database (Arndt et al., 2019). The putative prophage regions were further classified into bacteriophage families. The detection of AbaRIs was manually done by observing the genetic environments of the chromosomal comM gene using the interactive genome browser provided at the PATRIC server. The NCBI BLAST was used to align detected resistance island sequences against the sequences of previously known resistance islands (Altschul et al., 1990). The earlier proposed naming system for AbaRIs was used to classify the AbaRI variants predicted in this study (Bi et al., 2019). The annotations of detected AbaRIs regions obtained using PATRIC genome browser were used to construct AbaRIs gene maps in R 4.5.0.

The web-based sequence typing of the 47 A. baumannii genomes was carried out using MLST v2.0 services available at http://www.genomicepidemiology.org (Larsen et al., 2012). For the sequence typing, we used the Pasteur scheme which is designed to identify the seven house-keeping genes namely 60-kDa chaperonin (cpn60), elongation factor EF-G (fusA), citrate synthase (gltA), CTP synthase (pyrG), homologous recombination factor (recA), 50S ribosomal protein L2 (rplB), and RNA polymerase subunit B (rpoB). Furthermore, to understand the possible genetic relatedness between ST profiles we performed eBURST analysis using the goeBURST algorithm available at http://goeBURST.phyloviz.net (Francisco et al., 2009). To determine founding STs, we uploaded the numeric data of ST profiles obtained from MLST 2.0 in the flat text file format to goeBURST v1.2.1.

The phylogenetic relationship between 47 A. baumannii genomes was determined through multiple sequence alignment of genomes based on the single nucleotide polymorphisms (SNPs). The genome alignment and detection of SNPs in each genome were carried out using novel SNP procedure provided at http://cge.cbs.dtu.dk/services/CSIPhylogeny/ (Kaas et al., 2014). The novel SNP procedure filters, concatenates, aligns, and constructs a maximum likelihood phylogenetic tree with 100 bootstraps. The A. baumannii ATCC 19606 genome was used as a reference. The final version of the phylogenetic tree was visualized in iTOL v6 (Letunic and Bork, 2021).

The Welch’s t-test was performed to identify difference between the number of acquired ARGs predicted between genomes (n₁ = 18, x ¯ 2 = 8.83 , S₁ = 2.43, S 1 2 = 5.91 ) belonging to ST2 genotype and genomes (n₂ = 29, x ¯ 2 = 7.83 , S₂ = 5.21, S 2 2 = 28.15 ) belonging to other STs detected in the study. The difference was considered significant at p<0.05.

The assembled genomes of AIIMS5, AIIMS7, MMC18, SRMC9, and, SRMC18 showed an average of 38.95% GC content. The average assembly size of five sequenced genomes ranged between 3.7 to 3.8 Mb. The comprehensive genome analysis using PATRIC3.6.5 predicted a mean total of 3652 protein-coding sequences, 276 subsystems, and 62 RNAs. The Supplementary Table S1 represents the assembly statistics for five A. baumannii genomes. The Table 1 provides a summarized account of genomic features of all 47 Indian A. baumannii genomes analyzed in this study.

The web-based MLST analysis of 47 genomes using the Pasteur scheme predicted a total of 15 STs. Out of the 47 genomes analyzed, 18 (38.3%) belonged to ST2, while 10 (21.2%) belonged to ST622. We also detected presence of ST10 (n=5), ST25 (n=2), and ST1 (n=2) in the study. The remaining ten isolates belonged to the ten STs namely, ST85, ST94, ST150, ST285, ST374, ST494, ST575, ST1512, ST1545, and ST1549.

The eBURST analysis grouped STs into clusters based on single, double, and triple locus variation between ST profiles detected. A total of three clusters and five singletons were observed in the population of 15 STs ( Figure 1A ). The largest cluster included six STs namely, ST10, 494, 575, 1512, 1545, and 25. In this cluster, ST10 and ST575 were single locus variant while ST10 and ST1512 were double locus variants. The remaining STs in this cluster were triple variants. The second cluster included the double locus variant ST1 and ST94 while the third cluster included triple locus variants ST150 and ST374.

The SNP-based phylogeny showed presence of two main subclades ( Figure 1B ). The first subclade included eight genome sequences namely SP304 (ST285), VB35179 (ST1512), VB35435 (ST575), CIAT758 (ST10), PM194229 (ST10), PM194188 (ST10), PM2098 (ST10), and PM193665 (ST10) while the second subclade included rest of A. baumannii genomes. The phylogenetic analysis also showed that the A. baumannii genomes belonging to ST2 and ST622 formed two major monophyletic clades. The results of MLST and eBURST analysis when compared with the SNP phylogeny, showed that single and double locus variants were situated under same clades while the triple locus variants such as ST150 and ST374 were separated into different clades. The SNP data matrix showed variation between similar STs such as ST2, ST10, ST25 and ST1 ( Figure S1 ). In case of ST622, nine (KSK1, KSK2, KSK6, KSK7, KSK10, KSK11, KSK18, KSK19, and KSK20) out of 10 genomes showed very low variation, with the only exception of VB16141 which showed higher variation. Overall, the phylogenetic analysis based on the alignment of concatenated SNP showed accordance with the STs detected in this study.

The resistome analysis of 47 A. baumannii genomes showed presence of a total of 79 types of ARGs of intrinsic as well as acquired origin belonging to six resistance mechanisms, namely antibiotic efflux (n=21), antibiotic inactivation (n=51), antibiotic target alteration (n=3), antibiotic target protection (n=1), antibiotic target replacement (n=2), and reduced permeability to antibiotics (n=1) ( Figure 2 ). All predicted β-lactamase genes showed 100% sequence similarity with the respective β-lactamase alleles present in the CARD database. Apart from the β-lactamase genes, the remaining ARGs showed 97 to 100% sequence similarity with the ARG sequences present in the CARD database. Three types of efflux pump associated ARGs, namely the resistance-nodulation cell division (RND) family, the major facilitator superfamily (MFS), and the small multidrug resistance (SMR) family were predicted in the study. We predicted 15 alleles of OXA-51 type and 12 alleles of ADC-type β-lactamases in this study. Additionally, OXA-23 and its allele OXA-422 as well as OXA-58 and its allele OXA-420 were predicted in the study. The most dominant type of family encoding β-lactam inactivating enzymes predicted was blaOXA-23 (n=34/47), followed by blaADC-73 (n=13/47). Overall, 46 out of 47 A. baumannii genomes showed the coexistence of at least one type of the OXA-type-β-lactamases with one type of the ADC-type of cephalosporinases. The presence of putative genes for aminoglycoside acyltransferases (AAC family), aminoglycoside adenyltransferases (ANT family), and aminoglycoside phosphotransferases (APH family) was detected in 10, 46, and 38 genomes, respectively. The putative sulphonamide resistance conferring gene sul1 was predicted in 34 genomes, while sul2 was detected in 42 genomes. The presence of lpsB gene associated with reduced permeability to antibiotics was predicted in 42 genomes. However, the genomes of AIIMS5, AIIMS7, MMC18, SRMC9, and SRMC18 did not show the presence of the lpsB gene. The quinolone resistance-determining region (QRDR) associated parC and gyrA were detected in 47 and 43 genomes, respectively. However, the presence of nonmutated gyrA observed in SRM9, SP304, B8342, and B8300 indicates a potential susceptibility of these isolates to quinolones. In case of acquired ARGs no significant difference was found between acquired ARGs predicted in ST2 as compared to the acquired ARGs predicted in rest of the 14 STs detected in the study (t = 0.882, p = 0.191, critical t = 1.756). Amongst the 18 A. baumannii isolates belonging to ST2, 14 (77%) showed the presence of blaOXA-23. A variation in the number of acquired ARGs was observed in the study. The genomes of SRMC9, SP304, B8300, and B8342 showed absence of acquired ARGs. A lower number of acquired ARGs was observed in AIIMS5 (n = 4), MMC18 (n = 1), and SRMC18 (n = 1) which were collected during 2005 in comparison to the remaining genomes which showed a relatively higher number (13 ≥ n ≥ 5) of acquired ARGs ( Figure 3 ).

A total of 56 types of putative VFGs were predicted in 47 A. baumannii genomes ( Figure 4 ). The predicted VFGs were grouped under nine categories namely, adherence (n=4), biofilm formation (n=14), enzyme (n=2), immune evasion (n=8), iron uptake (n=19), regulation (n=4), serum resistance (n=2), antiphagocytosis (n=2), and stress adaptation (n=1). The putative VFGs predicted in all 47 genomes included outer membrane protein A (ompA) involved in adherence and invasion, efflux pump adeFGH involved in biofilm synthesis, pgaABCD locus encoding poly-β-1-6-N-acetylglucosamine required for biofilm formation, phospholipase C and D, the capsular polysaccharide biosynthesis gene cluster K, the lpx genes involved in lipid biosynthesis, the two-component signal transduction system BfmRS, and penicillin binding protein G (pbpG). Except for the basI in VB35179, all putative genes associated with acinetobactin biosynthesis and intake (barAB, basABCDFGHIJ, bauABCDEF, and entE) were detected in all genomes. The presence of putative bap gene encoding biofilm associated protein, the putative csuA/BABCDE operon associated with pili synthesis and assembly, the putative hemO encoding the heme oxygenase, and the putative abaI inducer and its receptor abaR involved in the quorum sensing were predicted in 37, 39, 38, and 40 genomes respectively. A noticeable variation was observed in the putative genes encoding virulence factors such as O-antigen polysaccharide (lpsO), type IV pilin protein (pilE), flagellar biosynthesis protein (flip), enzyme involved in biosynthesis of dTDP-l-rhamnose (rmlD), catalase (katA), and the bifunctional enzyme UDP‐N‐acetylglucosamine 2‐epimerase/ManNAc kinase associated with antiphagocytic capsular polysaccharide synthesis (wbjD/wecB). Altogether, a vast variety of VFGs were predicted in this study.

The plasmidSPAdes showed the presence of plasmid contigs in AIIMS5, AIIMS7, MMC18, and SRMC18. The CARD-based analysis of the detected plasmid contigs, as well as plasmids from other 42 A. baumannii genomes, predicted the presence of various putative ARGs associated with penems, aminoglycoside, tetracycline, macrolide, phenicol, rifamycin, sulfonamide, and heavy metal resistance ( Figure 3 ).

Amongst the 274 prophage regions predicted, 108 were intact, 120 were incomplete and 46 were questionable. Further analysis classified predicted prophage regions into 5 families namely, siphoviridae (n=191), myoviridae (n=49), podoviridae (n=32), autographiviridae (n=1), and inoviridae (n=1) ( Figure S2 ). The sequence showing similarity to PHAGE_Acinet_Bphi_B1251 was detected in all 47 genomes.

All 47 A. baumannii genomes analyzed in this study harbored numerous ISs of diverse bacterial origin. A total of 16 IS families, including Tn3, IS1, IS3, IS4, IS5, IS6, IS21, IS30, IS66, IS91, IS256, IS630, IS701, IS982, ISL3, and ISNCY, were detected in the study. We detected the presence of structures showing partial similarity to those of ISAba1-blaADC (n = 21/47) and ISAba1-blaOXA-23 (n = 36/47).

All Tn3-type transposons detected in this study showed presence of antibiotics and heavy metal resistance genes. A total of four types of Tn3 similar to Tn5393, TnAs3, TnAs2, and Tn3 were predicted ( Figure S3 ). Forty out of 47 genomes studied showed the presence of at least one type of Tn3 carrying at least one resistance gene. The Tn5393 detected in 35 genomes showed the presence of ARGs associated with aminoglycoside resistance (strA/aph(3’’)-Ib, strB/aph(6)-Id). In comparison, the TnAs3 detected in 29 genomes contained putative ARGs associated with an aminoglycoside, chloramphenicol, mercury, and quaternary ammonium compound resistance. The TnAs2, a truncated form of TnAs3, only showed presence of genes associated with mercury resistance. The structure of Tn3 showed presence of putative gene encoding β-lactamase-TEM-1b.

The insertion of AbaRIs at comM gene was confirmed in 31 out of 47 genomes. The predicted AbaRIs varied in the length. The longest resistance island of size 90.4 kb was detected in VB23193 (ST2). In contrast, the shortest resistance island of size 17.7 kb was detected in AIIMS7 (ST1). Based on the sequence similarities shown with previously known AbaRIs sequences, we classify predicted AbaRIs into four types namely, ABaR3 variant, AbGRI variant, AbaR4a variant, and AbaR4b variant ( Figure 5 ). The presence of truncated segment of Tn6022 carrying genes such as transposase (tniA), NTP-binding protein (tniB), universal stress protein (uspA), and sulfate permease (sulp) was found in all resistance islands except for the ABaR3 variant found in AIIMS7 which was a ST1 isolate ( Figure 5A ). The ABaR3 variant detected in ST1 (AIIMS7) possessed putative genes involved in cadmium and zinc resistance (czcD, czcR). However, it lacked the blaOXA-23 which was found in 27 out of 31 AbaRIs detected in this study. The ABaR3 variant predicted in AIIMS7 also showed presence of a class I integron gene cassette carrying putative genes conferring resistance to disinfectant (qacEΔ1), sulphonamide (sul1), and aminoglycoside (aadA, aac3). The two types of resistance islands were observed in ST2 genomes namely AbGRI variant and AbaR4a variant ( Figures 5B, C ). The complex AbGRI variants detected in VB23193 (ST2) and VB35575 (ST2) were longer in length as compared to the AbaR4a variant observed in the remaining ST2 genomes. Despite the complex nature and length variation in both ST2 variants (AbGRI and AbaR4) showed presence of certain common genes such as tetracycline resistance genes (tetB, tetR) as well as the truncated fragment of Tn5393 carrying strA and strB genes. Another difference we observed in AbaRI variants detected in ST2 genomes was, the presence of chloramphenicol (cmlA) and macrolides resistance (mphE, msrE) genes in AbGRI variant while their absence in AbaR4 variant. A fourth type of AbaRI variant namely AbaR4b was predicted in A. baumannii genomes belonging to ST10, ST85, ST575, ST25, ST622, and ST1512 ( Figure 5D ). Similar to AbGRI variant and AbaR4a variant predicted in ST2 isolates the AbaR4b showed presence of blaOXA-23. However, it lacked tetB, tetR and asrR genes. Overall, the comparative analysis of 47 A. baumannii genomes showed presence of multiple types of MGEs carrying putative resistance genes associated with antibiotics, disinfectants, and heavy metal resistance ( Figure 3 ).