Genomic insights into biofilm-associated virulence in extensively drug-resistant Acinetobacter baumannii

Highlights

• •WGS revealed biofilm-related virulome diversity in XDR A. baumannii ST2 and ST25 lineages.
• •ST2 isolates showed varied biofilm formation, while ST25 formed a strong biofilm.
• •Distinct bap (type-2/type-3) and ompA variants (V1/V3) were identified.
• •pilA nonsense mutation linked to loss of twitching motility in ST2 isolates.
• •Susceptible isolates demonstrated an increased propensity for strong biofilm formation.

Abstract

Acinetobacter baumannii is a notorious nosocomial pathogen known for its resistance to multiple antimicrobials, with biofilm formation contributing to its persistence in hospital environments. This study characterized biofilm-associated virulence genes in extensively drug-resistant (XDR) A. baumannii isolates from two distinct lineages, ST2 and ST25, to understand their roles in biofilm formation and antimicrobial resistance. From 135 non-repetitive multidrug-resistant (MDR) clinical isolates collected across Thailand, 15 XDR isolates (14 ST2 and 1 ST25) were selected for further analysis. Whole-genome sequencing (WGS) was performed to identify biofilm-associated genes and sequence polymorphisms. Biofilm formation and motility phenotypes were assessed, and gene expression analysis was evaluated by qRT-PCR. Most isolates (66.7 %) were moderate biofilm formers, and 80 % exhibiting higher biofilm biomass than the reference strain ATCC 19606. Notably, isolates with lower antimicrobial resistance profiles (i.e., relatively more susceptible among XDRs) tended to produce stronger biofilms. Significant variations in key biofilm genes were observed. Specifically, the abaIR quorum-sensing system was absent in 33.3 % of isolates. All ST2 strains carried bap type-2 with 4–11 BC repeats, while ST25 harboured bap type-3. ompA variants also showed lineage specificity (variant V1 in ST2 and V3 in ST25). All isolates harboured type 1 secretion system (T1SS) operon, however an ISAba1 insertion in the tolC in ST25 may impair protein secretion. Additionally, a nonsense mutation at codon 57 (TTA→TAA) in pilA was identified in all ST2 isolates, potentially accounting for the lack of twitching motility. These findings highlight the substantial genetic and phenotypic variability in biofilm-associated genes among XDR A. baumannii, providing insights into their persistence in healthcare settings.

Graphical abstract

1. Introduction

Acinetobacter baumannii is a member of the ESKAPE pathogens (Enterococcus faecium, Staphylococcus aureus, Klebsiella pneumoniae, Acinetobacter baumannii, Pseudomonas aeruginosa, and Enterobacter species), which are responsible for a significant portion of nosocomial infections and exhibit high levels of antimicrobial resistance (AMR) (De Oliveira et al., 2020). The global rise in carbapenem-resistant A. baumannii (CRAB) is particularly alarming due to the limited treatment options and increased mortality rates associated with these infections. International clones, including those with epidemiological importance in Thailand, have shown increasing resistance to carbapenems and colistin, which further complicates clinical management (Hamidian and Nigro, 2019; Hernandez-Gonzalez et al., 2022; Muller et al., 2023). The emergence of extensively drug-resistant (XDR) A. baumannii strains underscores the urgent need to understand the mechanisms driving this resistance and persistence in clinical settings (Jones et al., 2015; Kuo et al., 2012; Seruga Music et al., 2017).

In addition to AMR, biofilm formation enhances the persistence and adaptability of A. baumannii in hospital environments. Clinical strains from different clonal lineages exhibit varying capacities to form biofilms, which may contribute to their survival and persistence (Giannouli et al., 2013; Skerniskyte et al., 2018). Biofilm-forming strains persist longer on dry surfaces and can colonize a range of hospital materials, including stainless steel, polystyrene, and polycarbonate plastic (Espinal et al., 2012; Greene et al., 2016). This increases the risk of medical device-related infections, such as those involving cardiac devices, catheters, prosthetic joints, and shunts (Pompilio et al., 2021). Biofilm-associated infections are also increasingly identified in skin and soft tissue infections (SSTIs) (Dallo and Weitao, 2010), where biofilms impair wound healing and protect bacteria from immune clearance and antimicrobial agents. Outbreaks of A. baumannii in SSTIs are often linked to traumatic injuries, but cases in non-traumatic wounds, particularly among patients with underlying conditions such as diabetes mellitus and cirrhosis, have also been reported (Ali et al., 2014; Guerrero et al., 2010).

Despite the clinical importance of biofilm formation in A. baumannii, the genetic mechanisms underlying biofilm-associated virulence and resistance remain incompletely understood. Genomic plasticity contributes to variability in biofilm-related traits among clinical strains, highlighting the need to better characterize relevant virulence determinants (Imperi et al., 2011; Mangas et al., 2019). Several genes have been implicated in biofilm formation, including the chaperone-usher (Csu) pili gene cluster, particularly csuE, essential for initial attachment to abiotic surfaces (Harding et al., 2018; Pakharukova et al., 2018). Reports also suggest a functional trade-off between Csu pili and type IV pili during the initial surface attachment and maturation of biofilm structure (Ahmad et al., 2023; Ronish et al., 2019). Additional biofilm determinants include biofilm-associated protein (Bap), involved in biofilm maturation (Brossard and Campagnari, 2012); outer membrane protein A (OmpA), which helps to maintain biofilm structural integrity (Gaddy et al., 2009); and the BfmR/S two-component regulatory system, a central regulator of biofilm formation (Tomaras et al., 2008). Other factors such as RND efflux pumps and the abaI/abaR quorum-sensing system further modulate biofilm formation depending on environmental conditions (Mayer et al., 2020; Rumbo-Feal et al., 2013; Saipriya et al., 2020). Given the variability and dynamic nature of these genetic elements, continued studies are crucial to understand their impact on biofilm phenotypes, bacterial adaptability, and antimicrobial resistance.

Many virulence genes in A. baumannii, including those implicated in biofilm formation, have been identified as part of the conserved core genome (Antunes et al., 2014; Chan et al., 2015), and are thus commonly present in both biofilm-forming and non-forming strains. Previous studies have reported variability in the prevalence and expression of these genes among clinical isolates (Badmasti et al., 2015; Li et al., 2021; Zeighami et al., 2019), with higher expression levels often observed in epidemic clones compared to sporadic ones (Selasi et al., 2016). However, the relationship between the presence or expression of biofilm-associated virulence genes and biofilm or resistance phenotypes remains inconclusive, suggesting that other genomic feature may underlie biofilm variability. Moreover, detailed analyses of structural variations within these genes (e.g., gene truncations, domain polymorphisms, or insertion by mobile genetic elements) remain limited. This study explores these genomic features in XDR A. baumannii clinical isolates, focusing on two distinct lineages based on MLST allelic profiles: ST2, a dominant international clone circulating in Thailand, and ST25, an emerging but less prevalent clone worldwide. Particular attention is given to the structural diversity of key biofilm-related genes, including variant forms of ompA and bap between ST2 and ST25, polymorphism in pilA associated with loss of twitching motility, and insertion of ISAba1 within tolC of the type 1 secretion system (T1SS) locus. These genomic insights aim to improve understanding of biofilm-associated virulence and its complexity in XDR A. baumannii.

2. Materials and methods

2.1. Bacterial strain and culture conditions

Out of 135 non-repetitive multidrug-resistant (MDR) A. baumannii isolates collected from tertiary hospitals across Thailand, fifteen XDR isolates (12 ST2 and 3 ST25) were identified and selected for further analysis. After collection, the isolates were preserved in Mueller-Hinton broth (HiMedia Laboratories, India) supplemented with 15 % glycerol and stored at −40 °C until further use. Species identification was confirmed by microbiological and biochemical assays, as well as molecular detection of the intrinsic bla_OXA-51-like gene. (Khuntayaporn et al., 2021). Additionally, the Pasteur scheme of multilocus sequence typing (MLST) based on PCR amplification and sequencing had been performed to assign the sequence type and discriminate between A. baumannii and other closely related species within A. baumannii – A. calcoaceticus (ACB) complex (Diancourt et al., 2010). A. baumannii ATCC 19606 (Microbiologics, USA) was used as a reference strain for phenotypic-based testing. Unless otherwise specified, the bacterial cultures were prepared by streaking the glycerol stock onto Mueller-Hinton agar (HiMedia Laboratories, India) followed by overnight incubation at 37 °C. Antimicrobial susceptibility was assessed using disc diffusion or broth microdilution methods according to Clinical Laboratory Standards Institute (CLSI) (CLSI, 2022).

2.2. Biofilm formation assay

Semiquantitative measurement of biofilm biomass was performed based on a previously described protocol with slight modifications (Merritt et al., 2005). Briefly, overnight bacterial cultures grown in Mueller-Hinton Broth (HiMedia Laboratories, India) were adjusted to optical density (OD) of approximately 0.1 at 600 nm (equal to 0.5 McFarland, or approximately 1 × 10⁸ cfu/mL). The adjusted cultures were then diluted to a final concentration of approximately 1 × 10⁶ cfu/mL. Subsequently, bacterial inoculum was transferred to 96-well polystyrene flat-bottomed microtiter plate (SPL Life Sciences, South Korea) with minimum of four technical replicates per isolate. Wells containing sterile Mueller-Hinton broth served as negative controls.

After incubation at 37 °C for 24 h without shaking, planktonic cells were carefully removed, and wells were washed three times with sterile phosphate-buffered saline (PBS). Adherent biofilms were stained with 150 μL of 0.1 % crystal violet (HiMedia Laboratories, India) for 15 min. Excess stain was removed, wells were washed three times with sterile PBS, and plates were air-dried for 15 min. The crystal violet bound to the biofilm was solubilized by adding 150 μL of 95 % ethanol per well, and the absorbance was measured at 570 nm (OD₅₇₀) using CLARIOstar® Plus microplate reader (BMG Labtech, Germany). Assays were conducted in three independent experiments performed at intervals of at least 24 h.

The optical density cut-off value (OD_c) at 570 nm was claculated using the following formulas:

$$O{D}_c=averageOD{570}_{negativecontrol}+\left(3\times S{D}_{negativecontrol}\right)$$

$$O{D}_{strain}=averageOD{570}_{strain}-O{D}_c$$

The biofilm forming capacity was classified using criteria as described by Stepanovic et al. (2007) (Table 1).

Table 1.

Classification criteria for biofilm forming capacity.

Biofilm forming index	Definition
Not biofilm former	ODstrain$\le$ ODc
Weak	ODc$<$ ODstrain$\le$ 2ODc
Moderate	2ODc$<$ ODstrain$\le$ 4ODc
Strong	ODstrain$>$ 4ODc

2.3. Motility assay

Assessment of motility phenotypes were assessed according to previously described protocol with some modifications (Skerniskyte et al., 2018). Specifically, tryptic soya broth (TSB) (HiMedia Laboratories, India) supplemented with 0.25 % and 0.75 % agarose (Nippon Genetics, Japan) was used as the motility media. A single bacterial colony from freshly grown cultures was stabbed vertically into the center of agar plates using a sterile toothpick. The inoculated plates were incubated at 37 °C for 24 h (swarming motility) or 48 h (twitching motility) in a humidified airtight container. Motility was assessed by measuring the diameter (in millimeters) of the halo growth extending outward from the inoculation point. Additionally, twitching motility was visualized by carefully removing the agarose layer and staining the plate surface with 0.1 % crystal violet solution.

2.4. Genomic DNA extraction and whole genome sequencing

Genomic DNA was extracted from bacterial cultures grown to mid-log phase using the Gentra Puregene Yeast/Bact Kit (Qiagen, Germany), following the manufacturer’s instructions. The quality and concentration of extracted DNA samples were assessed using a QFX fluorometer (DeNovix, USA) and a Nanodrop nucleic acid analyzer (Hercuvan Lab System, UK). DNA integrity was further confirmed by electrophoresis on a 1 % agarose gel. For sequencing, approximately 400 ng of DNA per isolate was prepared into libraries using the Oxford Nanopore Rapid Barcoding Kit (SQK-RBK004). Sequencing was performed using a MinION Mk1c with an R9.4.1 flow cell. Raw sequencing data were basecalled using Dorado v0.5 (https://github.com/nanoporetech/dorado) with the high accuracy model.

2.5. Genome assembly and bioinformatic analysis

Raw sequencing reads were initially screened for potential contamination using Kraken2 v2.1.3 against the Standard-8 database (Wood et al., 2019). Long-read genome assembly was performed using Flye v2.9.3 (Kolmogorov et al., 2019) with five polishing iterations. Subsequently, the generated contigs were annotated using Prokka v1.14.6 (Seemann, 2014) and the NCBI Prokaryotic Genome Annotation Pipeline (PGAP). Assembly quality was assessed with QUAST v5.2 and CheckM v1.2.3 (Gurevich et al., 2013; Parks et al., 2015). Genomes with completeness greater than 85 % and contamination below 2 %, as determined by CheckM on PGAP-annotated assemblies, were considered sufficient for downstream analyses. The draft genomes have been deposited in the GenBank database under BioProject Accession PRJNA854605.

An MLST script (https://github.com/tseemann/mlst) was employed to screen contigs and assign the ST based on the Pasteur scheme. Virulence-associated genes were identified by screening assembled contigs against the Virulence Factor Database (VFDB) using ABRicate v1.0.1 (https://github.com/tseemann/abricate), applying thresholds of ≥80 % sequence identity and coverage. Further detailed annotation and manual comparative analyses at the gene level (e.g., presence/absence, polymorphism, and insertion sequences) were performed in Geneious Prime v2023.0.1 (https://www.geneious.com/). Pairwise and multiple sequence alignments were conducted using Clustal Omega and visualized with Jalview v2.11.3.3 (Sievers and Higgins, 2014; Waterhouse et al., 2009). Illustrations depicting genetic architectures of selected virulence genes were generated using Adobe Illustrator v26.4.1.

2.6. Total RNA extraction and gene expression analysis of biofilm genes

Total RNA was extracted from overnight planktonic bacterial cultures (16–18 h in Mueller-Hinton broth) using the FavorPrep™ Tissue Total RNA Mini Kit (Favorgen, Taiwan), following the manufacturer’s instructions. The concentration of the total RNA is determined spectrophotometrically using a Nanodrop nucleic acid analyzer (Hercuvan Lab System, UK). Approximately 2 µg of the RNA template is then reverse transcribed into complementary DNA (cDNA) using the ExcelRT™ Reverse Transcription Kit (Smobio Technology, Taiwan).

The cDNA was incorporated into KAPA SYBR FAST qPCR Master Mix (Kapa Biosystems, South Africa) in duplicate. The quantitative PCR (qPCR) was conducted using the Agilent Mx3000P. qPCR system under the following conditions: an initial enzyme activation step for 5 min at 95 °C, followed by 50 cycles of denaturation at 95 °C for 10 s and amplification at 60 °C for 30 s. One cycle for melt curve was added after the last cycle of amplification with the following conditions: 95 °C for 1 min, 55 °C for 30 s, and 95 °C for 30 s. The relative expression of the target gene was determined by the comparative cycle threshold method (2^−ΔΔCt) and visualized in GraphPad Prism v10.1.1 (https://www.graphpad.com/). The primers used for gene amplification are listed in Table 2. The rpoB gene was utilized for normalizing Ct values, with A. baumannii ATCC 19606 serving as the calibrator. Both the total RNA extraction and qPCR were carried out on three different occasions.

Table 2.

List of primers used for RT-qPCR.

Gene	Sequence	Ref.
rpoB	F: 5′-GAGTCTAATGGCGGTGGTTC-3′ R: 5′-ATTGCTTCATCTGCTGGTTG-3′	(Wiradiputra et al., 2023)
ompA	F: 5′-GGTATTCAGATAATTTTTCAGCAACTT-3′ R: 5′-AACAAATCAAACATCAAAGACCAA-3′	(Wiradiputra et al., 2023)
abaI	F: 5′-CCGCCTTCCTCTAGCAGTCA-3′ R: 5′-AAAACCCGCAGCACGTAATAA-3′	(Selasi et al., 2016)
bap	F: 5′-AATGCACCGGTACTTGATCC-3′ R: 5′-TATTGCCTGCAGGGTCAGTT-3′	(Selasi et al., 2016)
csuE	F: 5′-TCAGACCGGAGAAAAACTTAACG-3′ R: 5′-GCCGGAAGCCGTATGTAGAA-3′	(Selasi et al., 2016)

3. Results

3.1. Biofilm forming ability and AMR phenotypes of CRAB isolates

Microtiter plate biofilm formation assay was performed to determine the ability to produce biofilm biomass of fifteen CRAB isolates belonging to ST2 and ST25. A similar trend of biofilm biomass was observed between biofilm formation assays. The ST2 lineage isolates demonstrated a diverse capacity for biofilm formation, with 71.4 % being classified as moderate biofilm formers (Table 3, Supplementary file1 Table S1). The reference strain A. baumannii ATCC 19606 was also included and classified as a moderate biofilm former. Notably, the sole isolate from the ST25 clone exhibited a strong biofilm forming ability, and all but three ST2 isolates produced higher biofilm biomass than ATCC 19606.

Table 3.

Biofilm forming classification and susceptibility of CRAB isolates.

Isolate code	STP	Biofilm-forming index (BFI)	Twitching motility	Swarming motility	Antimicrobial susceptibility
					IMI	TZP	SAM	SXT	TET	CIP	LVX	CAZ	AMK	COL
ATCC 19606	52	Moderate	−	−	S	S	S	S	S	S	S	S	S	S
MTC0623	2	Strong	−	+	R	R	R	S	R	R	S	R	R	R
MTC0619	25	Strong	+	−	R	R	S	R	S	R	S	R	R	R
MTC0606	2	Strong	−	+	R	R	R	S	R	R	S	R	R	R
MTC0629	2	Moderate	−	+	R	R	I	R	R	R	S	R	R	R
MTC1120	2	Moderate	−	+	R	R	I	R	R	R	S	R	R	R
MTC0615	2	Moderate	−	+	R	R	R	R	R	R	I	R	R	R
MTC0610	2	Moderate	−	+	R	R	R	R	R	R	R	R	R	R
MTC1130	2	Moderate	−	+	R	R	I	S	R	R	S	R	R	R
MTC0620	2	Moderate	−	+	R	R	R	R	R	R	S	R	R	R
MTC0609	2	Moderate	−	+	R	R	R	R	R	R	S	R	R	R
MTC0608	2	Moderate	−	+	R	R	R	R	R	R	R	R	R	R
MTC1330	2	Moderate	−	+	R	R	R	R	R	R	S	R	R	S
MTC0617	2	Moderate	−	+	R	R	R	R	R	R	I	R	R	R
MTC1106	2	Weak	−	+	R	R	R	R	R	R	I	R	R	R
MTC0614	2	Non-biofilm former	−	+	R	R	R	R	R	R	S	R	R	R

AST and interpretation were performed by broth microdilution (IMI, CIP, LVX, CAZ, AMK, COL) and disk diffusion (TZP, SAM, SXT, TET) according to the guideline from CLSI M100 document. IMI, imipenem; TZP, piperacillin/tazobactam; SAM, ampicillin/sulbactam; SXT, trimethoprim-sulfamethoxazole; TET, tetracycline; CIP, ciprofloxacin; LVX, levofloxacin; CAZ, ceftazidime; AMK, amikacin; COL, colistin.

Given the notion that biofilm formation may be associated with resistance phenotype, antimicrobial susceptibility patterns of clinical isolates investigated in this study were determined. Susceptibility testing against ten antimicrobial agents across nine categories indicated that the isolates met criteria for XDR A. baumannii, defined as non-susceptibility to at least one agent in all but one or two antimicrobial categories (Magiorakos et al., 2012). Levofloxacin showed the highest susceptibility rate (66.67 % of the isolates) compared to other drugs, despite the pronounced resistance observed with ciprofloxacin, which belongs to the same fluoroquinolone drugs. Interestingly, grouping the isolates based on their susceptibility patterns showed that the more sensitive isolates tended to produce higher biofilm biomass (Fig. 1), although the difference in OD_570nm between the groups was not significant. Isolates with strong biofilm-forming capabilities remained sensitive to at least two drugs, whereas two moderate biofilm-forming isolates (MTC0610 and MTC0608) were resistant to all tested drugs. A. baumannii MTC0619 (ST25) was notably the most susceptible among the fifteen CRAB isolates, showing susceptibility to ampicillin/sulbactam, tetracycline, and levofloxacin, while being resistant to imipenem, ceftazidime, ciprofloxacin, gentamicin, trimethoprim-sulfamethoxazole, and colistin.

Fig. 1.

Distribution of XDR A. baumannii isolates based on their susceptibility pattern and biofilm forming capacity. The error bars indicate the mean and standard deviation of the mean. The difference of OD_570nm between groups was not statistically significant based on one-way ANOVA test.

In addition to assessment of biofilm formation, we examined the motility phenotypes of XDR A. baumannii isolates (Fig. 2, Supplementary file 2 Figure S1). Twitching motility, indicative of surface-associated movement through extension and retraction of type IV pili (T4P), was solely observed in the ST25 isolate. Conversely, swarming motility, a coordinated movement across solid surfaces, was evident in all ST2 isolates but not in ST25 (MTC0619). A. baumannii ATCC19606 notably did not demonstrate either swarming or twitching motility.

Fig. 2.

Twitching (A) and swarming motility (B) of XDR A. baumannii isolates. The dataset in ST2 represents the average from fourteen isolates, while ST25 includes the values from MTC0619. Horizontal lines indicate statistically different values (****P<0.0001) and non-significance (ns) based on one-way ANOVA test followed by Tukey HSD post-hoc test. The error bars show the standard error of the dataset in each group.

3.2. Overview of the virulome in clinical XDR A. baumannii isolates

The XDR isolates analysed in this study shared a comparable virulome, characterized by seven types of functional virulence factors identified through the screening of contigs against VFDB in ABRicate (Table 4). Notable variations were observed in genes associated with biofilm formation, with five isolates notably lacking the AbaIR quorum sensing system, known for its role in regulation of biofilm lifecycle. To investigate the possibility of technical artifacts, one of these isolates (MTC1120) was subjected to hybrid assembly approach combining both long- and short-read sequencing data, and the absence of abaIR was confirmed (unpublished data). Manual inspection of the assembled genome sequence was also did not detect these genes, further supporting their absence. Moreover, diversity of biofilm-associated proteins (bap) and bap-like proteins (blp) were detected in certain isolates.

Table 4.

Overview of virulence genes from XDR A. baumannii isolates in this study.

The virulome also revealed the presence of two effector delivery modules, the type 2 secretion system (T2SS) and type 6 secretion system (T6SS). Moreover, annotation of contigs identified a region harbouring a type 1 secretion system (T1SS). This secretion system was initially characterized in A. baumannii strain M2 and typically comprises three components: an outer membrane protein (TolC), an ATP-binding cassette (ABC) transporter (LssB), and a periplasmic adaptor protein (HlyD) (Harding et al., 2017). The intact T1SS locus observed in this study was flanked by lptCAD in the upstream region and a putative translation regulator in the downstream region (Fig. 3). However, disruption of the tolC sequence with ISAba1 element was observed in one isolate of ST25 clone (MTC0619). Sequence analysis confirmed the presence of right inverted repeat (IRR) of ISAba1 based on the ISfinder database (Siguier et al., 2006) and it was also predicted that the 3′-TTAAAATT-5′ as an insertion site of the IS element within tolC. Alignment of tolC from both isolates with a reference sequence from NCBI database (Accession No WP_269521631.1) indicated that the IS element contributed to the absence of 49 amino acid residues at the N-terminal region of tolC in A. baumannii MTC0619 (Supplementary file 2 Figure S2).

Fig. 3.

Schematic representation of T1SS cassette in A. baumannii MTC0619. The purple arrows indicate the genes constituting T1SS module, whereas the green arrows show genes that flanked T1SS region. The ISAba1 element is represented in yellow arrow and the black-shaded white text denotes the IRR of ISAba1.

3.3. Polymorphisms observed in the biofilm-associated protein (bap) gene

Most Bap proteins comprise of arrays of 80–110 amino acid modules featuring immunoglobulin-like (Ig-like) motifs. Based on a report regarding the diversity of proteins containing Ig-like motifs involved in biofilm formation among Acinetobacter spp. (De Gregorio et al., 2015), additional sequence analysis was conducted to investigate the occurrence of those genes in XDR A. baumannii isolates examined in this study. Amino acid residues in the C-terminal region indicated that A. baumannii MTC0619 (ST25) possessed bap type-3 with a sequence length of 15,195 bp (Fig. 4, Supplementary file 2 Fig. S3). In contrast, the other fourteen isolates (ST2) carried bap type-2, identified as two consecutive open reading frames (ORFs) with an overlap of 32 bases. ORF1 (2187 bp) contains a conserved N-terminal region and AB repeats, while ORF2 includes C-terminal domain and the larger repetitive region with varying numbers of BC repeats in some isolates (Supplementary file 1 Table S3, Supplementary file 2 Fig. S4). For instance, MTC0608 had ORF2 (15,999 bp) with eleven BC repeats as shown in Fig. 4. Moreover, two genes encoding the bap-like protein (blp) were identified in all ST2 isolates, whereas MTC0619 (ST25) possessed only one (Supplementary file 1 Table S2). The bap and blp1 genes were separated by intergenic region of approximately 31.4 kb. The blp1 sequences in all ST2 isolates belonged to type-2 based on the configuration of N-terminal, repetitive, and C-terminal regions (Supplementary file 2 Fig. S5). They also shared over 99 % amino acid sequence identity with the reference sequence (GenBank Accession No (WP_078207451.1). This gene was flanked by genes encoding for a putative metalloprotease in the upstream and GroES/EL chaperonin complex in the downstream region. The blp2 was flanked upstream by a gene encoding SohB protease and downstream by ISAba26 transposase. In MTC0619 (ST25), this gene had a similar length of 2187 bp and showed over 98 % identity with the corresponding gene found in reference strains MDR A. baumannii ACICU (GenBank Accession No NZ_CP031380.1) and AYE (GenBank Accession No NC_010410.1). The ST2 isolates had a slightly shorter blp2 sequence (2103 bp) but still demonstrated over 98 % identity to the reference strains (Supplementary file 1 Table S4). Notably, MTC0608 and MTC0615 had their blp2 sequences disrupted by ISAba36, resulting in only 82 % sequence identity with the other isolates and reference sequences.

Fig. 4.

Schematic representation of the bap gene in (A) A. baumannii MTC0619 and (B) A. baumannii MTC0608. The orange and blue arrows indicate bap genes in both isolates and blp1 gene in A. baumannii MTC0608, respectively. The overview of bap and blp1 exhibit the portion of N-terminal, repetitive, and C-terminal regions as noted in the figure.

3.4. Polymorphic variants of the outer membrane protein A (ompA) gene

All fourteen ST2 isolates possessed the outer membrane protein A (ompA) with 100 % coverage and identity to a reference sequence (Accession No WP 000777882). On the other hand, A. baumannii MTC0619 (ST25) had a lower value with 95.05 % coverage and 87.43 % sequence identity. The length of ompA gene in both isolates was 1029 bp for A. baumannii MTC0619 and 1071 bp for the ST2 isolates. These sequences were further investigated in accordance with the previously reported variant grouping of ompA in A. baumannii (Viale and Evans, 2020). Four external loops (EL) and the variable C-terminal regions were identified (Fig. 5), classifying the ompA genes of A. baumannii MTC0619 as variant 3 (V3) and that of fourteen A. baumannii in ST2 clone as variant 1 (V1). Moreover, ST2 isolates contained a longer and alanine-rich C-terminal tract as indicated by the presence of six amino acid stretch (AAAPAA).

Fig. 5.

Alignment of amino acid sequences showing ompA variant 3 (V3) in MTC0619 and variant 1 (V1) in all fourteen ST2 isolates. Bold and highlighted amino acids indicate the external loops (ELs) and the variable C-terminal region.

3.5. Distinct major pilin subunit (pilA) alleles linked to twitching motility phenotypes

Genes associated with type IV pili (TFP) components were found in all isolates. The major pilin subunit gene (pilA) with the size of 435 bp was initially identified in all ST2 isolates but absent in the ST25 isolate (MTC0619), based on VFDB screening. It showed 93.79 % amino acid sequence identity with a reference sequence (Accession No WP_000993715.1) and had a nonsense mutation in codon 57 (TTA → TAA) (Fig. 6). This gene was located adjacent to pilin-specific O-oligosaccharyl-transferase (tfpO). Further manual investigation revealed a slightly longer pilA sequence (486 bp) in isolate MTC0619. Although this pilA gene shared only 39 % sequence identity with those in ST2 isolates (Supplementary file 1 Table S5), BLASTx analysis showed it was 100 % identical to another reference sequence in the NCBI database (Accession No WP_000993726.1). The tfpO gene was also absent in MTC0619.

Fig. 6.

Multiple sequence alignment of amino acid sequences of pilA gene from XDR A. baumannii isolates investigated in this study. Conserved amino acid residues are color-highlighted with the minimum conservation threshold of 80 %. The red arrow marks the nonsense mutation at codon 57 present in all ST2 isolates.

3.6. Expression patterns of biofilm-associated virulence genes

To evaluate their expression in clinical XDR A. baumannii isolates compared to the laboratory strain A. baumannii ATCC19606, several genes associated with biofilm formation (i.e., abaI, bap, csuE, ompA) were selected, as shown in Fig. 7. No overall correlation was observed between the expression levels of these genes and their biofilm-forming capacity. The expression level of ompA was similar (mean 0.86 ± 0.24) across all fifteen isolates, regardless of the biofilm forming capacity. For abaI, five isolates demonstrated no expression, except for one (MTC1130), which lacked the AbaIR module (Table 4). A similar trend in expression levels was observed within the same bap types. Specifically, ten isolates carrying bap type-2 with a longer ORF2 (15,999 bp) exhibited similar expression levels (mean 0.90 ±SD 0.20), whereas four isolates with a shorter ORF2 (12,387 bp) had lower expression (mean 0.32 ± SD 0.04). The isolate MTC0619, which possessed bap type-3, notably showed no expression as it only shared approximately 30 % nucleotide sequence identity with bap type-2.

Fig. 7.

Expression of four genes associated with biofilm formation in XDR A. baumannii isolates relative to the calibrator strain A. baumannii ATCC19606. The expression levels were normalized to the housekeeping gene rpoB. The isolates in x axis are ordered based on the biofilm forming capacity. The bars indicate the means from three independent experiments and the error bars represent the standard deviations of the means.

4. Discussion

The genomic plasticity of A. baumannii enables rapid adaptation in its resistome and virulome, allowing the bacterium to maintain fitness under different selective pressures. Although biofilm-associated genes are generally conserved within clonally related A. baumannii isolates, notable variations exist across different clones (Antunes et al., 2014; Banerjee et al., 2024; Harding et al., 2018; Roy et al., 2022; Yakkala et al., 2019). In this study, XDR A. baumannii isolates demonstrated highly conserved virulence gene repertoires, but specific genetic polymorphisms and transcriptional variations were identified in genes related to biofilm formation, including the abaIR quorum-sensing system, biofilm-associated protein (Bap), outer membrane protein A (OmpA), the pilA gene encoding the major pilin subunit, and components of the type I secretion system (T1SS).

Biofilm formation typically involves surface attachment of planktonic cells, production of cell clusters in biofilm matrix, maturation of microcolonies, and dispersion of the biofilm (Roy et al., 2022; Sauer et al., 2022). The initial step requires expression of the csuA/BABCDE gene cluster, which encodes the components of the Csu pili for adhesion to abiotic surfaces. This cluster is highly conserved among clinical isolates, with csuE (encoding the tip adhesin) playing a crucial role in surface adhesion due to its unique hydrophobic finger-like structure (Harding et al., 2018; Pakharukova et al., 2018). While csuA/B encodes the major structural subunit of the pilus rod, csuE is functionally indispensable for pilus assembly and surface binding. Deletion of csuE abolishes pilus formation and biofilm development on plastic surfaces, and mutations in its hydrophobic tip structure similarly impair biofilm formation (Pakharukova et al., 2018). In contrast, mutations in csuA/B may still allow the formation of partial pilus-like structures composed of subunits CsuA, CsuB and CsuE (Pakharukova et al., 2022), suggesting that csuE-containing tip adhesins may form independently and retain adhesive capacity. Accordingly, csuE expression was selected as a representative marker of pilus-mediated adhesion.

All isolates in this study indeed carried a conserved Csu pili gene cluster, though csuE expression was varied among isolates. These findings suggest that csuE transcript levels alone may not directly predict biofilm-forming capacity in XDR isolates. Such transcriptional variability likely reflects differential regulatory activity of the BfmRS two-component system within each strain since the cultures were grown under basal condition. The BfmRS have been shown to modulate genes associated with biofilm and virulence genes under different environmental conditions, including Csu pili (Kishii et al., 2020; Tomaras et al., 2008). Moreover, it was shown that non-native acyl-homoserine lactone induction of BfmRS increases csuA/BABCDE expression, enhancing Csu pili assembly and biofilm biomass (Luo et al., 2015).

Bap and Blp, which contain bacterial immunoglobulin (BIg) domains, play critical roles in biofilm maturation (De Gregorio et al., 2015; Loehfelm et al., 2008). In this study, ST2 and ST25 isolates displayed distinct bap variants, with ST2 harboring bap type-2 and ST25 carrying bap type-3. Differences in BC repeat numbers contributed to length variation in bap type-2, which generally correlated with higher transcript levels, except in MTC1120. Expression of bap type-3 was absent in MTC0619 due to oligonucleotide specificity for the repetitive region of bap type-2. Pairwise alignment revealed only 30 % sequence identity between type-2 and type-3 variants, reflecting substantial genetic divergence. The bap type-3 has been reported as the least prevalent in A. baumannii, while bap type-2 is the most prevalent, occurring in 40 % of the analyzed STs (De Gregorio et al., 2015). Another study reported prevalence of bap type-2 in 68.44 % of A. baumannii isolates, primarily in ST2, while bap type-3 was found in only 6.91 % and was mainly associated with ST25 (Upmanyu et al., 2024). The blp1 and blp2 genes were highly conserved, although two ST2 isolates carried an ISAba36 insertion in blp2, whose functional significance remains unclear. Despite these genetic variations, no clear correlation between bap or blp sequence diversity and biofilm formation was established.

Additionally, insertion of ISAba1 was identified in the T1SS cassete of A. baumannii MTC0619 (ST25), contributing to the loss of 49 amino acid residues within the N-terminal region of the tolC gene. Since the T1SS is involved in secreting proteins like Bap (Harding et al., 2017), disruption of the T1SS gene cassette by this IS element could potentially alter biofilm architecture and stability through impaired protein secretion. Although structural implications of this disruption were not directly assessed here, this finding highlights how insertion elements may contribute to functional variability among clinical isolates. Detailed structural and functional studies would be necessary to gain a deeper understanding of the implications of this ISAba1 insertion on T1SS function and biofilm development.

OmpA also exhibited sequence polymorphisms between ST2 and ST25 isolates. Although it is unclear in which biofilm formation steps are involved, the effect of OmpA on membrane integrity was assumed to influence Csu pili attachment to abiotic surfaces (Gaddy et al., 2009). ST2 isolates in this study particularly carried hydrophobic alanine residues at the C-terminal region, which was presumed to provide the stability of cell envelope (Viale and Evans, 2020). However, sequence polymorphism observed in bacterial OmpA may also affect the cell surface properties as the higher surface hydrophobicity in some Escherichia coli OmpA variants was significantly found to reduce the biofilm biomass (Liao et al., 2022). Therefore, the previously proposed classification of A. baumannii OmpA alleles could serve as a basis for further investigations into its structural impacts on biofilm formation and other virulence aspects (Viale and Evans, 2020).

The correlation between biofilm production and antimicrobial resistance phenotypes remains controversial. A previous study showed that A. baumannii isolates with high levels of resistance were found to produce weak biofilms, whereas isolates with lower levels of resistance tended to form stronger biofilms (Qi et al., 2016). Later, a study by Al-Shamiri et al. revealed that while sensitive isolates could develop strong biofilms for 24 h, their biofilm-forming ability eventually waned, and they only produced weak biofilms. In contrast, resistant strains formed moderate to strong biofilms (Al-Shamiri et al., 2021).

Our results indicate that the majority (10/15) of XDR isolates formed moderate biofilms, consistent with previous observations that ST2 and ST25 clones typically exhibit robust biofilm formation (Giannouli et al., 2013; Kishii et al., 2020). Interestingly, isolates with higher antimicrobial susceptibility showed stronger biofilm formation, suggesting a possible phenotypic trade-off between resistance and biofilm capacity. Although the exact mechanisms behind such trade-offs are unclear, our findings support the notion that extensive antimicrobial resistance does not necessarily correlate with enhanced biofilm-forming ability.

In addition to biofilm genes, significant genetic variations were observed in pilA, encoding the major subunit of type IV pili (T4P). The amino acid sequence identity of pilA between ST2 and ST25 isolates was only 39 % (Supplementary file 1 Table S5), despite a conserved N-terminal region. A premature stop codon in pilA among ST2 isolates notably correlated with the lack of twitching motility, whereas the ST25 isolate retained twitching motility and robust biofilm formation. The tfpO gene, involved in PilA glycosylation, was found only in ST2 isolates, suggesting lineage-specific differences in pilin glycosylation patterns. Although PilA glycosylation has been proposed to influence pilus bundling and biofilm formation (Harding et al., 2015; Ronish et al., 2019), our data indicate that glycosylation alone might not fully explain differences in observed motility and biofilm phenotypes. These results suggest that T4P variations contribute differently to biofilm formation across clinical isolates, adding another layer of complexity to biofilm-associated virulence regulation.

There are several limitations to consider in this study. First, the sample size was confined to two sequence types (ST2 and ST25), which may limit the generalizability of the findings. Second, biofilm formation was assessed under static laboratory conditions, which may not fully replicate the dynamic environment in clinical settings. While the static microtiter plate assay provided valuable insights into initial biofilm formation capacity, future studies using continuous flow models and advanced microscopy (e.g., confocal laser scanning microscopy, scanning electron microscopy) could provide a more detailed understanding of biofilm architecture and dynamics. Third, the functional impact of gene disruption by insertion sequences remains unclear. Further investigation at both the transcriptional and protein structural levels is necessary to clarify how these disruptions influence biofilm formation. Finally, exploring biofilm evolution models could provide insights into how antimicrobial exposure drives biofilm-associated resistance and potential trade-offs between biofilm formation and resistance phenotypes. Adaptive responses within the biofilm microenvironment may lead to the evolution of tolerance and resistance, which could come at the cost of reduced biofilm biomass produced by subsequent planktonic cells after biofilm detachment. Similar trade-offs have been observed in A. baumannii AB5075-UW and Salmonella enterica serovar Typhimurium 14028S (Penesyan et al., 2019; Trampari et al., 2021).

5. Conclusion

In this study, we presented the genomic characterization of virulence genes related to biofilm formation in CRAB isolates displaying XDR phenotypes from two different clonal lineages, ST2 and ST25. While most isolates exhibited moderate biofilm formation, 80 % produced greater biomass than the reference strain ATCC 19606. Interestingly, those with lower antimicrobial resistance tended to form stronger biofilms. Genomic analysis revealed lineage-specific differences in key biofilm-related virulence genes, including bap, ompA, and pilA, as well as the absence of abaIR quorum-sensing system and disruption of T1SS gene cassette in some strains. Notably, all ST2 isolates carried a nonsense mutation in pilA, correlating with the defective twitching motility. These findings provide genetic and phenotypic diversity of biofilm determinants among XDR CRAB isolates in Thailand. The identification of clonal differences in biofilm-associated traits may have important implications for surveillance, infection control, and treatment outcomes, particularly given the role of biofilms in promoting environmental persistence and tolerance to antimicrobial agents. Further insights into the impact of these genes on biofilm architecture and capacity at the microscopic level will enhance our understanding of A. baumannii pathogenicity and resistance, as well as support to development of targeted interventions in high-risk clinical settings.

Declaration of competing interest

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests:

Mullika T. Chomnawang reports financial support was provided by Health Systems Research Institute, Thailand.

Data availability

The WGS data generated in this study have been deposited to the NCBI GenBank database and are accessible under the project accession number PRJNA854605. All other supporting data will be made available upon reasonable request.