Effect of polymyxin resistance evolution on the fitness costs of Acinetobacter baumannii and the underlying mechanisms of polymyxin resistance

Abstract

The increasing incidence and emergence of polymyxin-resistant Acinetobacter baumannii pose a significant threat to global public health. The presence of sub-minimal inhibitory concentration (sub-MIC) antibiotics during clinical treatment is inevitable and may drive the development of antibiotic resistance. Therefore, understanding the fitness costs and underlying mechanisms of resistance evolution under sub-MIC polymyxin B (PMB) is crucial for controlling the spread of PMB-resistant strains. In this study, PMB resistance evolved rapidly within 1–2 days in different A. baumannii strains and acquisition of PMB resistance altered the bacterial susceptibility to other antibiotics. All evolved strains suffered varying degrees of fitness costs, manifesting as a reduced growth rate, in vitro competitiveness, motility ability, biofilm formation, and adhesion and invasion to host cells. Proteomic and qRT–PCR analyses revealed upregulation of efflux pumps, two-component systems and outer membrane proteins in resistant strains, along with downregulation of genes involved in lipid A biosynthesis, motility and biofilm formation. Notably, the increasing PMB resistance was attenuated upon co-treatment with efflux pump inhibitor CCCP. This study indicates that evolution of PMB resistance incurs substantial fitness costs in A. baumannii, and the loss of LPS production and overexpression of efflux pumps may represent the key mechanisms underlying this evolutionary process.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-55660-0.

Introduction

Acinetobacter baumannii is a major opportunistic pathogen responsible for both nosocomial and community-acquired infections. It can cause ventilator-associated pneumonia, urinary tract infections, meningitis, and bacteremia, primarily affecting individuals with compromised or suppressed immune systems, particularly those admitted to intensive care units^1,2. The World Health Organization has classified A. baumannii as a critical-priority pathogen on its list of antibiotic-resistant priority pathogens, highlighting its serious threat to global public health. Currently, the emergence and spread of multidrug-resistant (MDR) and extensively drug-resistant (XDR) A. baumannii strains are increasing due to the organism’s high environmental adaptability and its ability to acquire resistance genes through horizontal gene transfer³. The high isolation and resistance rates have severely limited treatment options for MDR and XDR A. baumannii infections. Polymyxins are generally regarded as the last-resort antibiotics for severe infections caused by MDR A. baumannii. Unfortunately, clinical isolates of A. baumannii resistant to polymyxins have recently been reported⁴. Although the global prevalence of polymyxin resistance remains relatively low, higher rates with a continuing upward trend have been observed in Southeast Asian and Mediterranean regions, posing significant challenges for clinical management⁵. Therefore, elucidating the complex mechanisms underlying resistance development in A. baumannii is essential.

Effective antimicrobial therapy relies on adequate antibiotic concentrations to inhibit pathogen growth. However, during anti-infective treatment, antibiotic levels in target tissues are often below the minimum inhibitory concentration (MIC), allowing certain bacterial clones to continue growing under sub-MIC conditions⁶. It has been demonstrated that exposure to sub-MIC antibiotics can influence various bacterial functions, including antimicrobial resistance, expression of virulence factors, motility, adhesion and biofilm formation⁷. In our previous study, we found that sub-MIC concentrations of polymyxin B (PMB) could rapidly drive resistance in Klebsiella pneumoniae within 1–2 days⁸. Numerous studies have shown that the acquisition of antibiotic resistance can alter bacterial physiology and impose fitness costs related to growth capacity, virulence and antimicrobial susceptibility^9–11. Moreover, mutations associated with fitness costs are preferentially selected under sub-MIC antibiotic exposure¹². Therefore, understanding the fitness costs associated with resistant mutants derived by sub-MIC antibiotics is crucial for assessing the implications of resistance development and controlling the dissemination of resistant strains. In this study, five A. baumannii strains with distinct resistance profiles were subjected to in vitro continuous exposure to sub-MIC PMB to generate PMB-resistant derivatives. We subsequently evaluated the effects of evolution resistance on antimicrobial susceptibility, growth kinetics, in vitro competitive fitness, motility, morphological features, biofilm formation, and the capacity for host cell adhesion and invasion in A. baumannii, while also exploring the molecular mechanisms driving the evolution of PMB resistance.

Materials and methods

Bacterial strains and cell culture

A. baumannii standard strain ATCC 17978, carbapenem-sensitive clinical isolates (A1132, A1136) and carbapenem-resistant clinical isolates (A734, A736) were used in this study. The ATCC 17978 strain was obtained from the American Type Culture Collection (ATCC), and the clinical isolates were collected from Southwest Hospital of Army Medical University.

The A549 human alveolar epithelial cell line was purchased from the American Type Culture Collection and cultured in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum (Gibco, USA) and 1% penicillin-streptomycin at 37 °C with 5% CO₂.

In vitro resistance evolution experiments

Evolution of PMB resistance in A. baumannii was performed via continuous subculturing in cation-adjusted Mueller‒Hinton broth (CAMHB; Solarbio, China)⁸. Briefly, A. baumannii standard and clinical strains were streaked onto nutrient agar plates and incubated overnight at 37 °C. Individual colonies were suspended in sterile saline and adjusted to a density equivalent to the 0.5 McFarland standard. The bacterial suspension was then inoculated at a final concentration of 1% (v/v) into CAMHB that contained 1/2×MIC of PMB (MedChemExpress, China) for each strain. Subcultures were performed daily using the same method over seven consecutive days. Resistance development was monitored daily by determining the MIC values using the broth microdilution method. PMB resistance breakpoints were defined according to the Clinical and Laboratory Standards Institute (CLSI, 2022) guidelines: intermediate susceptibility, MIC ≤ 2 µg/ml; resistance, MIC ≥ 4 µg/ml¹³.

Stability evaluation of PMB resistance evolution

The evolved strains obtained on day 7 of PMB resistance evolution were passaged on nutrient agar plates and incubated overnight at 37 °C. Single colonies from five evolved strains were inoculated into CAMHB without antibiotics, respectively, and incubated at 37 °C for 24 h with shaking. Each strain had three biological replicates and then subcultured daily one per one sample under antibiotic-free conditions for a total of 15 days. The MIC of PMB against the 15th generation strains was detected by the broth microdilution method. Changes in MIC before and after 15 serial passages were analyzed to evaluate the stability of the PMB-resistant phenotype.

Antimicrobial susceptibility test

The MICs of various classes of antimicrobial agents against the original A. baumannii strains and their laboratory-evolved strains were determined by the broth microdilution method. In brief, overnight cultures of each strain were adjusted to the 0.5 McFarland standard and diluted 1:500 in CAMHB. A 100 µl aliquot of the bacterial suspension was added to each well of a 96-well microtiter plate containing 100 µl of two-fold serially diluted antibiotics. Following incubation at 37 °C for 18–20 h, MIC values were interpreted according to CLSI guidelines, except for tigecycline, which was assessed based on U.S. Food and Drug Administration (FDA) criteria. Escherichia coli ATCC 25922 was used as the quality control strain.

Growth curve assay

Overnight cultures of the original A. baumannii strains and their evolved variants were adjusted to the 0.5 McFarland standard and diluted 1:1000 in CAMHB. A 1 ml volume of the diluted suspension was transferred to individual wells of a 24-well microtiter plate and incubated at 37 °C with shaking. Bacterial growth was monitored by determining the optical density at 600 nm (OD₆₀₀) every 2 h and the growth curve assay was performed over a 24-h period.

Bacterial competition assay

The in vitro competition experiments were performed as previously described¹⁴. Briefly, the original A. baumannii strains and their evolved strains were cultured overnight in LB broth at 37 °C. The overnight cultures were adjusted to 0.5 McFarland standard and further counted on agar plates to ensure the consistency of initial bacterial numbers. Subsequently, 100 µl of each evolved strain and its parental strain were co-inoculated into 10 ml of LB broth and co-cultured at 37 °C with shaking. After 12 h and 24 h of incubation, bacterial suspensions were serially diluted and plated onto LB agar plates with or without 8 µg/ml PMB. Bacterial colonies were counted following incubation at 37 °C. The competition index (CI) was defined as follows: CI = colony number of PMB-resistant strain/colony number of PMB-sensitive parental strain. A CI > 1 indicates increased fitness of the resistant strain; a CI < 1 indicates reduced fitness; and a CI = 1 indicates no change in fitness¹⁵.

Bacterial motility assay

The swarming motility of A. baumannii parent and evolved strains were assessed on semisolid agar plates as previously described¹⁶. Briefly, A. baumannii strains were grown overnight on LB agar plates at 37 °C. Then, a single colony of each strain was picked and inoculated at the center surface of swarming plates (LB broth, 0.5% glucose, 0.5% agar). Plates were incubated at 37 °C for 24 h, and the diameter of the migration zone was measured to evaluate the surface-associated motility. For twitching motility, plates containing 1% tryptone, 0.5% yeast extract, 0.5% NaCl, and 1% agar were used¹⁷. A single colony of each strain was inoculated vertically to the bottom of the plate using a sterile toothpick to allow bacterial attachment to the dish surface. After incubation at 37 °C for 24 h, the twitching zone was measured to assess type IV pili-dependent motility.

Bacterial morphology assay

Morphological changes in A. baumannii parent strains and their evolved strains were evaluated by transmission electron microscopy (TEM). Strains were grown overnight in LB broth at 37 °C with shaking. Bacterial pellets were collected by centrifugation at 10,000 rpm for 10 min. After fixation in 2.5% glutaraldehyde overnight at 4 °C, the bacterial cells were washed with phosphate buffer and then post-fixed with 1% osmium tetroxide for 2 h. Samples were dehydrated through a graded acetone series, embedded in epoxy resin, and polymerized to form solid blocks. The blocks were cut into ultrathin sections and then stained with uranyl acetate and lead citrate. All the samples were observed and photographed under an HT7800 TEM (Hitachi High-Tech, Japan).

Biofilm formation assay

Biofilm formation capacity was assessed using the crystal violet staining method to compare evolved strains with their parental strains¹⁸. Briefly, overnight cultures of A. baumannii were 1:1000 diluted in fresh LB broth and 200 µl of the suspension was transferred into wells of a 96-well plate. After incubation at 37 °C for 24 h, non-adherent cells were gently washed away with phosphate buffered saline (PBS), and the attached biofilms were air-dried. Biofilms were stained with 1% crystal violet for 10 min, rinsed to remove excess dye, and solubilized with 30% acetic acid. Absorbance was measured at 570 nm using a Synergy H1 multimode microplate reader (BioTek, USA).

Furthermore, we evaluated the biofilm-forming capacity of evolved strains and their parental strains using confocal laser scanning microscopy (CLSM). Bacterial suspensions were inoculated into confocal dishes (NEST, China) and incubated at 37 °C for 24 h. Non-adherent bacteria were removed by washing with PBS, and the resulting biofilms were stained in the dark for 15 min with the Live/Dead BacLight Bacterial Viability Kit (Invitrogen, USA). Stained biofilms were subsequently visualized using an LSM780 CLSM (Carl Zeiss, Germany) at 40× magnification.

Bacterial adherence assay

Adherence and invasion of A. baumannii to eukaryotic cells were assessed as previously described^19,20. A549 human alveolar epithelial cells were seeded into 24-well plates at a density of 1.0 × 10⁵ cells/well and allowed to grow overnight at 37 °C under 5% CO₂. A. baumannii parental strains and their evolved strains were cultured overnight in LB broth at 37 °C. Following centrifugation, bacterial pellets were resuspended in antibiotic-free DMEM. A549 monolayers were washed three times with PBS, followed by addition of 1 ml of bacterial suspension per well at a multiplicity of infection (MOI) of 100:1. After co-incubation at 37 °C and 5% CO₂ for 6 h, nonadherent bacteria were removed by three PBS washes. Cells were then trypsinized and resuspended in serum-containing medium to collect adherent bacteria. Serial dilutions of the suspensions were plated onto MH agar plates, and colony forming units (CFUs) were enumerated after 24 h of incubation at 37 °C.

Bacterial invasion assay

A549 cells cultured in 24-well plates were infected with A. baumannii parental and evolved strains at an MOI of 100:1 for 6 h at 37 °C under 5% CO₂. Following infection, extracellular bacteria were eliminated by treatment with 1 mg/ml gentamicin or 300 µg/ml PMB for 1 h. PMB was used for gentamicin-resistant isolates. After three PBS washes, cells were lysed with 0.2% Triton X-100 in PBS. Aliquots (100 µl) of cell lysates were plated on MH agar plates and incubated at 37 °C for 24 h to quantify intracellular CFUs.

Proteomic analysis

Differentially expressed proteins (DEPs) between A. baumannii 17978 and its evolved strain were analyzed via data-independent acquisition (DIA) proteomics using the Orbitrap Astral platform. Bacterial pellets were lysed in protein extraction buffer containing 8 M urea, 1% SDS, and protease inhibitors, followed by sonication. After centrifugation at 16,000×g for 30 min at 8 °C, protein concentrations in the supernatants were determined using the BCA Protein Assay Kit (Thermo Scientific, USA) and verified by SDS-PAGE. Proteins were digested with trypsin, and resulting peptides were desalted using hydrophilic-lipophilic balance (HLB) solid-phase extraction cartridges and quantified by ultraviolet spectrophotometry. Based on peptide quantification, samples were fractionated using a Vanquish Neo UHPLC system and analyzed by an Orbitrap Astral mass spectrometer (Thermo Scientific, USA). Raw DIA data were processed using Spectronaut 19 software for database searching. Differential protein expression between groups was statistically evaluated using the t test in the R package; proteins exhibiting a fold change > 1.2 and a P value < 0.05 were considered significantly differentially expressed. Functional enrichment and clustering analyses of DEPs were performed using the Gene Ontology (GO) database (http://geneontology.org/) and pathway annotation via the KEGG database (https://www.genome.jp/kegg/).

PCR-Sanger sequencing

The original and evolved strains of A. baumannii 17978 and A734 were cultured overnight in LB broth at 37 °C with shaking. Genomic DNA was extracted from the samples using the TIANamp Bacteria DNA Kit (Tiangen, China). The pmrCAB genes were amplified using the Full pmrCAB primer pairs. The PCR products were purified and then sequenced using the following primers: pmrA-F and pmrA-R for pmrA gene; pmrB-F, pmrB-R, pmrB2-F and pmrB2-R for pmrB gene; pmrC-F, pmrC-R, pmrC2-F and pmrC2-R for pmrC gene (Supplementary Table S1)²¹.

Quantitative real-time PCR analysis

The gene expression associated with efflux pumps, motility, biofilm formation, and lipid A modification and biosynthesis in A. baumannii 17978 and its evolved strain was evaluated by quantitative real-time PCR (qRT-PCR). Total RNA was extracted and purified from bacterial samples using the RNAprep Pure Bacteria Kit (Tiangen, China), and RNA concentrations were measured using a DeNovix DS-11 spectrophotometer (DeNovix, USA). cDNA was synthesized using the PrimeScript FAST RT Reagent Kit (TaKaRa, China), followed by amplification with the QuantiNova SYBR PCR Mix Kit (QIAGEN, Germany). The primer sequences used in this study are listed in Supplementary Table S1²². The rpoB gene served as an internal reference control, and the relative expression levels of target genes were calculated using the 2^−ΔΔCT method.

Effect of the efflux pump inhibitor on PMB resistance

To assess the involvement of efflux pumps in PMB resistance evolution, the MICs of PMB against the original and evolved strains of A. baumannii were determined in the presence and absence of the efflux pump inhibitor carbonyl cyanide m-chlorophenylhydrazone (CCCP)²³. Briefly, PMB was serially diluted and added to each well of a 96-well microtiter plate containing bacterial suspension. CCCP was then added to each well to achieve a final concentration of 10 mg/L. A reduction of at least fourfold in the MIC of PMB upon addition of CCCP was considered indicative of a significant inhibitory effect²⁴.

Statistical analysis

All graphs, calculations, and statistical analyses were performed using GraphPad Prism 9.5, and data are presented as means ± standard deviation. Group comparisons were made using unpaired Student’s t-test or two-way ANOVA to evaluate statistical differences. A P value less than 0.05 was considered statistically significant.

Results

The susceptible changes of different A. baumannii strains to PMB during in vitro resistance evolution

Five PMB-sensitive strains with distinct resistance profiles were subjected to in vitro experimental evolution under sub-MIC concentrations of PMB for 7 days. The results showed that this evolutionary process led to significant alterations in the susceptibility of A. baumannii to PMB. Resistance evolved rapidly under antibiotic selection, with all strains exhibiting increased MIC values within 1 or 2 days. The MICs of carbapenem-sensitive strains 17978, A1132 and A1136 rose to 16–64 µg/ml after 7 days of evolution. In comparison, carbapenem-resistant strains A734 and A736 demonstrated more rapid resistance development, with MICs increasing to 32 µg/ml within 1 day and reaching 256 and 128 µg/ml, respectively, by day 7 (Fig. 1). All the evolved strains retained stable resistance levels even after consecutive 15 passages in antibiotic-free medium (data not shown).

Fig. 1.

The MIC changes of polymyxin B against A. baumannii standard strain 17978 and clinical strains (A1132, A1136, A734, A736) during in vitro resistance evolution with 1/2MIC polymyxin B.

Antimicrobial susceptibility assay

The antimicrobial susceptibility of conventional antibiotics against A. baumannii strains before and after evolution of resistance was evaluated, and the MIC results are presented in Table 1. All evolved strains exhibited a significant reduction in the MICs of β-lactams (ceftazidime, meropenem) and tigecycline compared with their parental strains. However, the MICs of aminoglycosides (gentamicin, amikacin) and fluoroquinolones (levofloxacin) showed no obvious changes following resistance evolution, with variations in susceptibility being strain-specific.

Table 1.

Antimicrobial activity of conventional antibiotics against A. baumannii original strains and their evolved strains.

Strain	MIC (µg/ml)
	Ceftazidime	Meropenem	Gentamicin	Amikacin	Levofloxacin	Tigecycline
17978	8/S	0.5/S	0.25/S	0.5/S	≤ 0.125/S	0.125/S
17978R	≤ 0.25/S	≤ 0.125/S	≤ 0.125/S	≤ 0.25/S	≤ 0.125/S	≤ 0.125/S
A1132	8/S	0.5/S	4/S	8/S	0.5/S	0.5/S
A1132R	2/S	≤ 0.125/S	4/S	4/S	0.125/S	≤ 0.125/S
A1136	8/S	2/S	2/S	4/S	0.25/S	0.5/S
A1136R	2/S	0.5/S	0.5/S	1/S	0.25/S	≤ 0.125/S
A734	> 256/R	32/R	> 256/R	> 256/R	8/R	2/S
A734R	8/S	1/S	> 256/R	> 256/R	8/R	0.5/S
A736	> 256/R	32/R	> 256/R	> 256/R	8/R	2/S
A736R	8/S	≤ 0.5/S	> 256/R	> 256/R	8/R	0.25/S

Fitness costs of A. baumannii strains after resistance evolution

Fitness costs associated with PMB resistance were assessed through growth curve analysis, in vitro competition assays and motility tests. All evolved strains exhibited fitness impairments relative to their parental counterparts. In growth curve assays, the growth capacity of all resistant strains was significantly reduced. PMB-resistant strains of 17978, A1132 and A1136 showed delayed growth onset after 6–8 h of incubation, while those of A734 and A736 exhibited delayed growth after 10–12 h hours (Fig. 2A-E). In competition assays, all evolved strains had reduced competitive fitness relative to sensitive strains, as evidenced by their CIs being far below 1 (Fig. 2F). The CI values declined with increasing coculturing time, and resistant strains evolved from carbapenem-sensitive strains (17978, A1132 and A1136) had lower CI values than those evolved from carbapenem-resistant strains (A734 and A736) after 24 h of co-incubation. Motility assays on agar plates revealed that PMB resistance evolution could significantly impair swarming and twitching motilities in all strains (Fig. 3A-B). As illustrated in Fig. 3C, the diameters of swarming and twitching zones for the evolved strain 17978 were markedly smaller than those of the parental strain.

Fig. 2.

Comparison of the growth curves (A-E) and competitiveness (F) for different A. baumannii strains before and after evolution of resistance with polymyxin B. *p < 0.05, **p < 0.01, ***p < 0.001.

Fig. 3.

Changes in the motility ability of different A. baumannii strains before and after evolution of resistance. (A, B) Diameters of the swarming and twitching zones for A. baumannii standard and clinical strains before and after the evolution of polymyxin B resistance. (C) Visualization of the diameter differences of swarming and twitching zones between A. baumannii 17978 original strain and its evolved strain. Data are presented as mean ± standard deviation; *p < 0.05, **p < 0.01, ***p < 0.001 compared with the polymyxin B sensitive strain.

Morphological characteristics of A. baumannii strains after resistance evolution

The morphological changes of evolved strains from A. baumannii standard strain 17978 and clinical isolates (A734, A1132) were observed by TEM. As shown in Fig. 4, the original control strains displayed regular cellular morphology, with thick and well-defined cell walls and densely packed cytoplasmic contents. Following resistance evolution, the overall cell integrity was maintained, whereas the cell walls displayed looseness and shrinkage. The resistant strains 17978 and A1132 exhibited reduced cytoplasmic density and resistant strain A734 showed folded cell walls and uneven staining.

Fig. 4.

Morphological changes in the original control strains and their evolved strains of A. baumannii 17978, A1132 and A734 observed by transmission electron microscopy.

Effect of resistance evolution on the biofilm formation of A. baumannii

We first evaluated the biofilm forming capacity of all evolved strains and their corresponding parental strains using crystal violet staining. The results showed that resistance evolution significantly reduced the biofilm formation in all tested strains to varying degrees (Fig. 5A). Subsequently, biofilms formed by evolved mutants of A. baumannii standard strain 17978 and clinical isolates (A734, A1132) were observed by CLSM. As shown in Fig. 5B, the original strains of 17978, A734 and A1132 formed thick and dense biofilms. In contrast, biofilm formation was markedly diminished and sparse following evolution of resistance, consistent with the crystal violet staining results.

Fig. 5.

Comparative difference analysis of the biofilm formation ability between A. baumannii original strains and their evolved strains. (A) Semiquantitative analysis of biofilm biomass in A. baumannii original and evolved strains determined by crystal violet staining. (B) Qualitative assessment of biofilm formation in original and evolved strains using confocal laser scanning microscopy. Biofilms were stained with the Live/Dead Bacterial Viability Kit, with live and dead bacteria labeled green and red, respectively. Data are presented as mean ± standard deviation; *p < 0.05, **p < 0.01, ***p < 0.001 compared with the polymyxin B sensitive strain.

Effect of resistance evolution on the adherence and invasion of A. baumannii

To assess changes in bacterial adhesion and invasion capabilities before and after evolution of resistance, the number of bacteria adherent to or invaded into A549 cells was quantified using plate colony counting (Fig. 6). The parental strains of 17978, A1132 and A734 exhibited varying degrees of adhesion to and invasion into A549 cells. However, both adhesion and invasion capacities were significantly reduced following evolution of resistance. Additionally, the intracellular bacterial load (CFU) was substantially lower than the number of adherent bacteria.

Fig. 6.

Comparison of the adhesion and invasion capabilities of A. baumannii strains to A549 cells. (A, B) Number of adherent A. baumannii original strains (17978, A1132, A734) and their evolved strains to A549 cells, as determined by plate counting. (C, D) Number of invading A. baumannii original strains (17978, A1132, A734) and their evolved strains into A549 cells, as determined by plate counting. Data are presented as mean ± standard deviation; *p < 0.05, **p < 0.01, ***p < 0.001 compared with the polymyxin B sensitive strain.

Effect of resistance evolution on the protein expression of A. baumannii

Proteomic analysis was performed to investigate alterations in protein expression between the A. baumannii 17978 parental strain and its evolved counterpart. Compared with the parental strain, a total of 370 DEPs were identified in the evolved strain, including 209 upregulated and 161 downregulated proteins. GO enrichment analysis revealed significant enrichment of DEPs in categories related to the outer membrane, cell adhesion, efflux transmembrane transporter activity, and establishment of localization. KEGG pathway analysis and heatmap construction indicated that upregulated DEPs were primarily enriched in pathways associated with resistance nodulation-division (RND) efflux pumps, ATP-binding cassette (ABC) efflux pumps, outer membrane proteins and two-component systems, whereas downregulated DEPs were mainly involved in lipid A biosynthesis, Csu pilus, type IV pilus, fimbrial biosynthesis and biofilm formation (Fig. 7, Supplementary Table S2).

Fig. 7.

Proteomic analysis of A. baumannii 17978 and its evolved strain. (A) Volcano plot showing differentially expressed proteins between A. baumannii 17978 and its evolved strain. (B) GO enrichment analysis of differentially expressed proteins; the size of each bubble corresponds to the number of genes enriched within the respective GO term. (C) KEGG pathway enrichment analysis of differentially expressed proteins, with the top 20 significantly enriched pathways displayed on the vertical axis. (D) Heatmap of differentially expressed proteins related to efflux pumps, outer membrane proteins, lipid A biosynthesis, bacterial motility and biofilm formation. Color intensity reflects the magnitude of protein expression changes, with red indicating upregulation and green indicating downregulation.

Mutational and expression analysis of the genes associated with resistance evolution

To investigate whether the evolution of PMB resistance was associated with mutations in the pmrCAB operon, sequence analysis of the pmrCAB genes was performed for both the parental and evolved strains of A. baumannii 17978 and A734, but no mutations in pmrA, pmrB and pmrC genes were found in these two PMB-resistant strains. Then, we studied the expression changes of pmrCAB genes in the PMB-resistant 17978 strain by qRT‒PCR analysis. As shown in Fig. 8A, only the pmrC gene showed upregulation in expression after resistance evolution. Furthermore, we also analyzed the expression levels of genes encoding efflux pump components, motility-associated proteins, biofilm-related factors and lipid A biosynthesis (Fig. 8A-C). The results showed that evolution of PMB resistance led to significant upregulation of genes associated with the RND family efflux pump (adeA, adeB) and the ABC family efflux pump (macA, macB). The regulatory genes adeR and adeS of the AdeABC efflux system were also significantly increased in the evolved strain. Conversely, genes associated with Csu pilus (csuA, csuC), type IV pilus (pilT, pilM), fimbrial synthesis (fimV), biofilm formation (pgaB, pgaD) and lipid A biosynthesis (lpxA, lpxC, lpxD) were notably decreased in the evolved strain. These gene expression changes were consistent with the proteomic findings.

Fig. 8.

Role of efflux pumps in the evolution of polymyxin B resistance. (A-C) qRT‒PCR analysis of differentially expressed genes associated with modification and biosynthesis of lipid A, efflux pumps, motility, and biofilm formation. (D) Changes in the MIC of polymyxin B against evolved resistant strains in the presence and absence of the efflux pump inhibitor CCCP. Data are presented as mean ± standard deviation; *p < 0.05, **p < 0.01, ***p < 0.001 compared with the parental 17978 strain.

Effect of efflux pumps on evolution of PMB resistance

Integrating the proteomic and qRT-PCR results, we found that the evolution of PMB resistance in these strains may be primarily attributed to the overexpression of efflux pumps. Therefore, the efflux pump inhibitor CCCP was used to validate the functional role of efflux pumps in PMB resistance. The results showed that the MICs of PMB decreased fourfold in the evolved strains 17978, A1136 and A736, while a eightfold reduction in PMB MIC was observed in the resistant strains A1132 and A734 upon co-incubation with CCCP (Fig. 8D).

Discussion

Polymyxins were introduced into clinical practice in the 1950s but were subsequently withdrawn due to their nephrotoxicity and neurotoxicity. In recent years, the rising global incidence of MDR bacterial infections, coupled with the limited development of new antimicrobial agents, has led to the reintroduction of polymyxins into clinical use^25,26. However, inappropriate clinical utilization of polymyxins, along with their widespread use as livestock feed additives, has contributed to the emergence of polymyxin resistance among Gram-negative bacteria, including A. baumannii²⁷. Previous studies have reported that polymyxin exposure is a significant risk factor for polymyxin- and carbapenem-resistant A. baumannii infections in hospitalized patients, and the development of polymyxin resistance in carbapenem-resistant strains is associated with poor clinical outcomes^28,29. Therefore, it is of great importance to simulate the in vivo evolution of resistance under laboratory conditions and to elucidate the underlying resistance mechanisms in evolved mutants. Several studies have investigated colistin resistance mechanisms in A. baumannii under increasing concentrations of colistin^30,31. However, considering the complexity of the clinical environment, we selected both standard and clinical A. baumannii strains with distinct resistance profiles and evolved PMB resistance under identical sub-MIC levels of PMB. To the best of our knowledge, this is the first study investigating the fitness costs and mechanisms associated with PMB resistance acquisition in multiple A. baumannii strains. Resistance evolution experiments demonstrated that all strains rapidly developed resistance to PMB within 1–2 days. Carbapenem-resistant strains exhibited more pronounced resistance development compared to carbapenem-susceptible strains, with MIC values reaching 256 and 128 µg/ml, respectively, after 7 days of evolution. The rapid emergence of high-level resistance under sub-MIC PMB exposure highlights a substantial risk of A. baumannii developing resistance during clinical PMB use. Therefore, exposure of pathogenic bacteria to sub-MIC antibiotic environments should be minimized as much as possible in anti-infective therapy. Evolution of resistance to one antibiotic often leads to cross-resistance to others due to shared resistance mechanisms³². In this study, the development of PMB resistance caused an increase in susceptibility to β-lactams and tigecycline, while no significant changes were observed in susceptibility to aminoglycosides or fluoroquinolones, which aligns with a previous report³¹. Morphological analysis revealed that the original control strains displayed thick, well-defined cell walls and dense cytoplasmic content. In contrast, evolved PMB-resistant strains exhibited loosened, folded, and unevenly stained cell walls, along with reduced cytoplasmic density in some strains. In addition, the expression of proteins involved in lipid A biosynthesis and lipopolysaccharide (LPS) transport was downregulated in the evolved strain, leading to defects in outer membrane structure and increased membrane permeability³³. The observed lower cross-resistance to multiple antibiotics may be attributed to increased outer membrane permeability in the evolved mutants, which needs further investigation to confirm this hypothesis.

Acquisition of antibiotic resistance is known to impose fitness costs on bacteria, resulting in reduced growth capacity³⁴. Consistent with these findings, all evolved strains in this study exhibited a significant reduction in growth rates compared to their parental strains. Furthermore, we investigated the in vitro competitive relationship between susceptible and resistant bacteria in the absence of PMB, revealing that resistant strains constituted only a minor proportion when competing with susceptible strains. The diminished competitiveness of resistant bacteria has also been reported in other bacterial species following resistance evolution under different antibiotics^32,35. In addition, PMB-resistant strains evolved from carbapenem-resistant strains had a lower CI values than those evolved from carbapenem-sensitive strains after 24 h of co-incubation. This may be attributed to the variation in the fitness cost of resistance evolution across different genetic backgrounds. These findings suggest that resistance evolution confers a substantial disadvantage in bacterial fitness under conditions lacking antibiotic selective pressure. However, the potential for dissemination of resistant variants persists even in the presence of trace amounts of antibiotics, which remains a significant concern. Although A. baumannii lacks genes required for flagellar biosynthesis, it can exhibit motility through twitching or surface-associated movement. Twitching motility is a coordinated, multicellular process driven by type IV pili, whereas surface-associated motility is an appendage-independent movement driven by the release of extracellular polymeric substances³⁶. In this study, swarming and twitching motilities were markedly decreased in the evolved PMB-resistant strains. This phenotypic observation was supported by proteomic and qRT‒PCR analyses, which demonstrated significant downregulation of genes associated with type IV pili, Csu pili and fimbrial assembly following resistance evolution. Biofilm formation is a multifactorial process, with initial adhesion requiring pilus expression to facilitate bacterial attachment to various abiotic surfaces. Pili-mediated motility plays a critical role in surface adherence, colonization and intra-biofilm spread³⁷. Biofilm formation assays revealed that all evolved strains exhibited impaired biofilm-forming capacity compared to their parental strains. The pgaABCD locus, involved in the synthesis of poly-β-1-6-N-acetylglucosamine in A. baumannii, is essential for biofilm development³⁸. Both transcriptional and protein level analyses showed significantly reduced expression of pgaABCD in the evolved strain, consistent with the observed motility defects. Adherence to host cells represents an early and important step in the pathogenesis process of A. baumannii³⁹. A. baumannii utilizes pili and fimbriae structures to mediate attachment to and invasion of host cells⁴⁰. Our results revealed that all strains were capable of adhering to A549 cells, with a limited number of bacteria able to invade intracellularly, these findings were not entirely consistent with the previous study indicating that A. baumannii strains were unable to invade HEp-2 cells⁴¹. This discrepancy may be attributed to differences in bacterial strains and host cell lines used. Nonetheless, all evolved strains displayed reduced adhesion to and invasion of A549 cells relative to their parental strains. These data indicate that acquisition of PMB resistance causes the decrease of motility and biofilm formation, ultimately impairing the ability of A. baumannii to adhere to and invade to host epithelial cells. This is the evolutionary cost that bacteria incur to survive under PMB pressure, at the expense of their colonization and spreading abilities.

Three main mechanisms are responsible for PMB resistance in A. baumannii, including modification or loss of LPS structure, plasmid-mediated dissemination of mobile colistin resistance genes and overexpression of efflux pumps. Modification of LPS is mediated by the PmrAB two-component system. Mutations in pmrA and pmrB induce the overexpression of pmrC, resulting in PMB resistance. Lipid A is the structural and functional core of bacterial LPS, mutations or disruptions in the lpxACD genes encoding lipid A biosynthesis lead to the complete loss of LPS⁴². In the present study, we didn’t found any mutations in pmrCAB operon in the evolved 17978 and A734 strains, further transcriptional analysis of the evloved 17978 strain showed weak upregulation of pmrC, bu not of pmrA and pmrB genes. Differences in the selection methods used for resistance evolution may explain the emergence of distinct resistance mechanisms. Unlike most studies, which used stepwise increasing concentrations of colistin^21,43, our evolved mutants were selected in the presence of a fixed sub-MIC concentration of PMB. Moreover, the proteomic data and qRT-PCR verification revealed downregulated expression of proteins involved in lipid A biosynthesis and LPS transport in the evolved strain, which could cause the decrease of LPS production and the obstruction of LPS transport and thereby contribute to PMB resistance²². Efflux pumps are transmembrane transport proteins that actively expel antibiotics from bacterial cells, thereby contributing to multidrug resistance⁴⁴. In A. baumannii, several major families of efflux pumps have been implicated in antimicrobial resistance, including the RND, ABC, major facilitator (MF), multidrug and toxic compound extrusion (MATE) and small multidrug resistance (SMR) families. Accumulating evidence indicates that the RND family efflux pumps—particularly AdeABC, AdeIJK, and AdeFGH are widely distributed in A. baumannii⁴⁵. The AdeABC efflux pump, first identified in A. baumannii, is considered the most clinically significant RND type system⁴⁶. The overexpression of AdeABC has been associated with the decrease in susceptibility to multiple antibiotic classes⁴⁷. Typically, the AdeABC system is a tripartite complex composed of three proteins, whereas 35% of A. baumannii clinical isolates, including ATCC17978, lack the adeC outer membrane protein component⁴⁸. Interestingly, AdeAB could recruit an alternative outer membrane protein to efflux antimicrobial compounds⁴⁹. The ABC family efflux system MacAB-TolC can cause multidrug resistance and pathogenicity by extruding antibiotics and virulence factors out of the cells^50,51. Previous studies have indicated that macAB expression is significantly upregulated in MDR clinical isolates and colistin-resistant strains of A. baumannii^52,53. In this study, our results showed increased expression levels of AdeAB, AdeIJK and MacAB-TolC in the PMB-resistant 17978 strain. Similar findings were reported in another study³¹. In addition, the expression of the two-component systems AdeRS and BaeSR-known positive regulators of AdeAB and AdeIJK/MacAB-TolC pumps, respectively-was also elevated in the resistant strain⁵⁴. A previous study reported that overexpression of AdeABC and AdeIJK pumps could alter the membrane-associated cellular functions of A. baumannii, resulting in decreased biofilm formation⁵⁵, which was consistent with our obtained results. To assess the functional contribution of efflux pumps to evolution of PMB resistance, the MIC changes of PMB were determined following exposure of the evolved strains to the efflux pump inhibitor CCCP. All evolved strains exhibited a 4- or 8-fold decrease in PMB MIC upon co-incubation with CCCP. Overall, these findings suggest that the loss of LPS production and overexpression of multiple efflux pumps may represent the predominant mechanisms in the development of PMB resistance in A. baumannii strains evolved under sub-MIC PMB selection.

Several limitations were present in this study. First, the sample size of bacterial isolates was relatively small, which may limit the generalizability of the findings regarding fitness costs and underlying mechanisms following resistance evolution across all clinical strains. Second, the whole genome sequencing and proteomic analysis were not performed on all evolved strains and their parental strains, which will lack a comprehensive understanding of the evolutionary mechanisms of PMB resistance. Third, although multiple efflux pumps were upregulated in the evolved PMB-resistant strains, knockout and complemented strains were not constructed to validate the specific regulatory role of each efflux pump in resistance evolution.

In summary, this study demonstrates that sub-MICs of PMB can rapidly drive the evolution of resistance within 1–2 days across various A. baumannii strains. The acquisition of PMB resistance was accompanied by a high fitness cost in bacteria, resulting in the decrease in growth rate, motility, competitiveness, biofilm formation and adherence and invasion to host cells. The loss of LPS production and overexpression of the AdeABC, AdeIJK and MacAB-TolC efflux pumps may represents the key mechanisms contributing to the evolution of PMB resistance in A. baumannii. Although PMB-resistant strains exhibited a reduction in certain virulence traits, they may regain adaptive advantages under antibiotic selective pressure, underscoring the importance of avoiding the long-term exposure of bacteria to sub-MIC levels of antibiotics. The results of this study provide deep insights into the interplay between the evolutionary acquisition of PMB resistance and biological fitness costs in different A. baumannii strains, and elucidate the potential mechanisms driving PMB resistance evolution, thereby informing clinical strategies to mitigate the emergence and dissemination of PMB-resistant strains.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Qianmei Wang and Wei Feng performed the protocol design and wrote this manuscript. Fengjun Sun and Qianmei Wang performed the collection and analysis of data. Ting Huang and Qian Yuan carried out the analysis and interpretation of data. Zhengze Shen and Wei Feng carried out the protocol design, wrote and revised the content of this manuscript. All authors have read and approved the final manuscript.

Funding

This work was supported by the Natural Science Foundation of Chongqing (CSTB2024NSCQ-KJFZZDX0034) and the Open project of Key State Laboratory of Sichuan province (ARRLKF24-01).

Data availability

The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium (https://proteomecentral.proteomexchange.org) via the iProX partner repository with the dataset identifier PXD075723.

Declarations

Competing interests

The authors declare no competing interests.