Conformational hijacking of lipoprotein transporter LolDF enables precision antimicrobial activity against Acinetobacter baumannii

Abstract

Acinetobacter baumannii is a critical-priority pathogen with increasing antibiotic resistance. Here, we define the mechanism of abaucin, a first-in-class narrow-spectrum antibiotic that selectively targets A. baumannii by inhibiting its essential lipoprotein transporter, LolDF. Extending prior studies of archived strains, we demonstrate potent activity against clinically isolated carbapenem-resistant A. baumannii (CRAB) strains both in vitro and in a murine pneumonia model. Cryo-EM structures of abaucin-bound LolDF reveal symmetric binding of two abaucin molecules within the LolDF cavity, which lock the transporter in a non-productive, outward-open conformation. Biochemical and structural analyses show that abaucin does not block substrate binding but instead traps the substrate-loaded transporter and prevents transfer to LolA. Together, these findings uncover a unique symmetry-enabled conformation-hijacking mechanism and establish LolDF as a tractable target for precision antibiotic development.

In this study, researchers reveal the mechanism of action as a narrow-spectrum antibiotic, abaucin, which selectively kills Acinetobacter baumannii by trapping an essential transporter in an inactive state.

Introduction

Carbapenem-resistant Acinetobacter baumannii (CRAB) exemplifies the alarming global surge of multidrug-resistant Gram-negative pathogens. Clinical isolates often display resistance to nearly all available antibiotics except tigecycline and polymyxins^1,2, leaving limited therapeutic options and resulting in considerable mortality and healthcare burden^3,4. Recognizing its critical threat, the World Health Organization has designated CRAB as a top-priority pathogen for the development of new antimicrobial agents⁵.

CRAB’s resilience largely relies on the low-permeability barrier provided by its outer membrane (OM), an asymmetric bilayer of lipooligosaccharide and phospholipids reinforced by β-barrel proteins and lipoproteins^6–11. Notably, OM lipoproteins are the most abundant envelope components in Gram-negative bacteria and participate in nearly every process that keeps the OM intact^7,10. They orchestrate β-barrel protein assembly^12–16, lipopolysaccharide insertion^17–21, and lipid asymmetry maintenance^22–27; tether the OM to the peptidoglycan layer to confer mechanical resilience^28–30; and promote the formation of outer-membrane vesicles (OMVs) to disseminate virulence and stress-response factors^30,31. Representative examples include OM lipoproteins in Acinetobacter baumannii (A. baumannii), such as OXA-type β-lactamases^32,33. Their incorporation into OMVs enables protection of both the producing cells and neighboring bacteria from β-lactam antibiotics, thereby promoting community-level tolerance in polymicrobial infections^32,34.

OM lipoproteins are universally trafficked via the localization of the lipoprotein (Lol) pathway^35–39. After maturation on the inner membrane⁴⁰, lipoproteins are extracted by the ATP-binding cassette (ABC) transporter complex and transferred to the periplasmic chaperone LolA. In organisms encoding the OM receptor LolB, LolA subsequently delivers these lipoproteins to LolB for insertion into the OM^35,41–43. Although the overall pathway is conserved, the ABC transport machinery differs among species: A. baumannii employs the LolDF complex, whereas Escherichia coli (E. coli) uses the LolCDE system^44,45. Disruption of this machinery leads to lipoprotein mislocalization, perturbing OM homeostasis and barrier integrity, ultimately leading to cell death⁴⁶. In A. baumannii, inhibiting LolDF would further block OM anchoring and OMV packaging of OXA β-lactamases, thereby reducing intrinsic β-lactam resistance and limiting vesicle-mediated dissemination of resistance traits. Thus, the Lol complex represents an attractive drug target for therapeutic intervention against multidrug-resistant infections^47–54.

A recent machine-learning-guided screen identified abaucin, an antibiotic that selectively targets A. baumannii by perturbing lipoprotein trafficking, providing a unique opportunity to develop a pathogen-selective therapeutic⁵⁴. However, its antibacterial efficacy and inhibitory mechanism remain insufficiently characterized. Further investigation using contemporary clinical isolates and structural elucidation of its target complex is essential to assess its therapeutic potential and guide future drug development. In this study, we validated the potency and selectivity of abaucin against clinically isolated CRAB both in vitro and in a murine lung infection model. Using cryo-electron microscopy (cryo-EM) and functional assays, we show that abaucin selectively inhibits A. baumannii LolDF by symmetrically engaging the central cavity and stabilizing a transport-incompetent, outward-open conformation. Mutagenesis and biochemical analyses identify key binding residues, and co-occupancy with substrate demonstrates that abaucin traps the transporter without displacing lipoprotein. This conformation-hijacking mechanism is structurally incompatible with the asymmetric LolCDE complex, providing a molecular explanation for abaucin’s species-specific activity. Together, these findings define the molecular basis of abaucin’s selectivity and establish a framework for rational design and optimization of next-generation therapeutics to combat emerging resistance.

Results

Abaucin selectively inhibits CRAB growth in vitro

To evaluate abaucin’s antibacterial spectrum in a clinically relevant context, we assessed its activity against a panel of antibiotic-resistant isolates freshly obtained from intensive care unit (ICU) patients. The collection comprised multiple β-lactamase-expressing CRAB isolates, as well as representative carbapenem-resistant Klebsiella pneumoniae (CRKP) and extended-spectrum β-lactamase-producing E. coli (ESBL-E. coli) strains. Abaucin consistently exhibited potent inhibitory activity against CRAB, with minimal inhibitory concentrations (MICs) comparable to or lower than those reported for reference laboratory strains according to CLSI guidelines, while showing negligible effects on CRKP or ESBL-E. coli (Fig. 1a, b). These results confirm abaucin’s narrow yet highly effective antibacterial spectrum against clinically relevant A. baumannii isolates. MIC profiling against an independently assembled representative panel of gut bacterial species revealed no detectable antibacterial activity, even at concentrations exceeding those required to inhibit CRAB (Fig. 1c), further supporting its selective activity. These data provide a clinically relevant narrow-spectrum profile for abaucin and extend the initial report by Liu et al.⁵⁴ through validation in freshly obtained ICU isolates and an expanded commensal panel.

Fig. 1. MICs of abaucin against clinically isolated Gram-negative pathogens and gut commensal bacteria.

MICs were determined for carbapenem-susceptible (HX-S9, HX-S19, and HX-S31) and carbapenem-resistant (HX-R7, HX-R9, and HX-R28) clinical isolates of A. baumannii in (a), extended-spectrum β-lactamase (ESBL)-producing E. coli (HX-1–7), and carbapenem-resistant Klebsiella pneumoniae (CRKP, HX-1–13) in (b), as well as representative gut commensal isolates (Bacteroides vulgatus, HX-25-48-A1, Bacteroides thetaiotaomicron, HX-25-52-D4, Bacteroides fragilis, HX-25-56-I2) in (c). All strains were isolated from West China Hospital, Sichuan University. MIC measurements were consistent across three independent experiments initiated from single colonies.

Abaucin shows efficacy in a murine pneumonia model

Having confirmed abaucin’s potent and selective activity against clinically isolated carbapenem-resistant A. baumannii strains, we next evaluated whether this efficacy extends to in vivo infection settings. In line with its previously reported in vivo activity⁵⁴, we employed a murine lung infection model that closely mimics clinical pneumonia caused by multidrug-resistant A. baumannii. This pathogen is a leading cause of hospital-acquired pneumonia in ICUs and is frequently associated with carbapenem resistance and high mortality. Assessing abaucin’s therapeutic performance in this model, therefore, provides a stringent evaluation of its translational potential and pathogen-specific efficacy.

An acute pneumonia model was established using neutropenic BALB/c mice, in which neutrophils were depleted to prevent immune-mediated clearance of the infection. To establish an infection model with an appropriate therapeutic window, mice were first challenged intratracheally with different inocula of CRAB HX-R7. Infection with 1 × 10⁶ CFU caused no mortality, whereas 1 × 10⁸ CFU resulted in greater than 50% mortality within 40 h of inoculation (Supplementary Fig. 1a). In contrast, infection with 1 × 10⁷ CFU led to a progressive increase in bacterial burden in both the lungs and blood after 24 h in the absence of abaucin treatment, while producing intermediate survival rate (Supplementary Fig. 1b), and was therefore selected for subsequent efficacy studies (Fig. 2a). Mice infected with 1 × 10⁷ CFU were treated with either vehicle, abaucin, or clinically used broad-spectrum-lactam antibiotics for resistant infections, including piperacillin (PIP, a β-lactam) and meropenem (MEM, a last-line carbapenem), administered at 20 mg kg⁻¹, at 2, 12, 24, and 36 h post-infection (p.i.). At 48 h p.i. of abaucin, body weight and pulmonary function were assessed, followed by euthanasia for further analysis. Lung tissues and bronchoalveolar lavage fluid (BALF) were collected for pathological and microbiological analysis. Histopathological examination of lung sections (Fig. 2b–e), including scanning electron microscopy (SEM) (Supplementary Fig. 1c), hematoxylin and eosin (H&E) staining (Fig. 2b), and lung injury scoring (Fig. 2c), revealed that infected mice displayed extensive alveolar wall thickening, severe structural damage, and pronounced inflammatory cell infiltration. Notably, these pathological changes were markedly alleviated in the abaucin-treated group, to a degree comparable to meropenem, whereas no improvement was observed with piperacillin treatment (Fig. 2b–e). Consistently, bacterial counts of BALF showed a significant reduction in A. baumannii burden following abaucin administration (Fig. 2d). Cytokine analysis further demonstrated substantial downregulation of interleukin (IL)-6 and IL-1β levels in treated mice (Fig. 2f, g). Moreover, abaucin treatment effectively restored both body weight and respiratory performance, both of which were impaired in infected mice (Fig. 2h, i). These findings demonstrate that abaucin exhibits in vivo efficacy comparable to clinically used last-line antibiotics and extend its activity from a wound infection model to a pneumonia setting, supporting its therapeutic potential against clinically relevant A. baumannii infections.

Fig. 2. Abaucin treatment ameliorates CRAB-induced pneumonia in neutropenic mice.

a Schematic of the experimental design. Neutropenia was induced by two doses of cyclophosphamide (CTX) administered 96 and 24 h before infection. At the time of infection, CRAB (HX-R7) was inoculated by intratracheal instillation (1 × 10⁷ CFU). Vehicle, abaucin, meropenem, or piperacillin (20 mg kg⁻¹) was administered in four doses at 2, 12, 24, and 36 h post infection, and mice were euthanized at 48 h post-infection. IP, intraperitoneal injection. b Representative haematoxylin and eosin (H&E)-stained lung sections from sham, vehicle-treated, abaucin-treated mice, meropenem and piperacillin-treated mice (scale bar, 20 μm; n = 5 mice per group). c Lung injury scores derived from histopathological assessment of lung sections shown in (b) (n = 5, ****p < 0.0001; ns, not significant). d Bacterial burden in BALF from vehicle-treated, abaucin-treated, meropenem-treated, and piperacillin-treated mice, determined by colony-forming unit (CFU) counts after serial dilution (n = 3, ****p < 0.0001; ns, not significant). e Lung tissue from vehicle-treated, abaucin-treated, meropenem-treated, and piperacillin-treated mice in the CRAB-induced pneumonia model. Concentrations of IL-1β (f) and IL-6 (g) in BALF from sham, vehicle-treated, abaucin-treated mice, meropenem-treated, and piperacillin-treated mice (n = 5, ****p < 0.0001; ns, not significant). h Body weight changes in sham, vehicle-treated, abaucin-treated, meropenem-treated, and piperacillin-treated mice over the course of infection (n = 5, ****p < 0.0001; ns, not significant). i Minute ventilation (MV; h) in sham, vehicle-treated, abaucin-treated mice, meropenem-treated, and piperacillin-treated with CRAB-induced pneumonia (n = 5, ****p < 0.0001; ns, not significant). Statistical significance was assessed by one-way ANOVA with Tukey’s multiple comparisons test. Data are shown as mean ± SEM. Source data are provided as a Source Data file.

While in vitro studies demonstrated abaucin’s target specificity, they do not predict its impact on the host gut microbiota. To address this, we analyzed fecal samples from the pneumonia model using full-length 16S rRNA sequencing before and after treatment. CRAB infection with vehicle treatment caused a marked shift in faecal microbial composition relative to sham controls (Supplementary Fig. 1d). Abaucin treatment did not further perturb the microbiota, as both the community composition and Shannon diversity remained broadly comparable to those of vehicle-treated mice. In contrast, meropenem and piperacillin induced distinct microbiota profiles accompanied by altered Shannon diversity (Supplementary Fig. 1d, e). Collectively, these results provide compelling in vivo validation of abaucin’s precision, demonstrating that its high target specificity is preserved within the complex host microbiome environment.

Cryo-EM structure of the abaucin-bound LolDF complex

Given abaucin’s exceptional species selectivity and potent in vivo efficacy against A. baumannii, we next sought to elucidate its molecular mechanism of inhibition. Abaucin was originally identified through a machine learning-guided screen for inhibitors of A. baumannii and was proposed to disrupt lipoprotein trafficking⁵⁴. In A. baumannii, this process is mediated by the noncanonical LolDF complex (Supplementary Fig. 2a), for which structural information was lacking. To uncover the basis of abaucin’s selective inhibition, we determined the cryo-EM structures of abaucin-bound A. baumannii LolDF, enabling direct visualization of drug binding and its conformational impact on the transporter. Recombinant LolDF from A. baumannii ATCC 17978 was expressed in E. coli C43 (DE3), and the purified complex was incubated with abaucin in the presence of ATP and MgCl₂ (Supplementary Fig. 2b, c). The cryo-EM structures of the abaucin-bound LolDF complex reveal two conformations, one featuring a wide cavity and the other a relatively narrow cavity, resolved at overall resolutions of 3.04 and 3.15 Å, respectively (Fig. 3a and Supplementary Fig. 2).

Fig. 3. Cryo-EM structure of abaucin-bound A. baumannii LolDF complex.

a Cryo-EM density maps and corresponding cartoon representation of abaucin-bound LolDF in the wide-cavity state (left) and narrow-cavity (right). LolF protomers are shown in blue and gold, LolD subunits in green and olive, abaucin in cyan, ATP in purple, and the magnesium ion as a green sphere. b Chemical structure of abaucin. Detailed interactions between abaucin and key residues in the wide-cavity state (c) and narrow-cavity state (d) of LolF.

In both states, LolDF adopts a typical type-VII ABC transporter architecture, with LolF constituting the homodimeric transmembrane (TMDs) and periplasmic (PDs) domains, and LolD forming the nucleotide-binding domains (NBDs) (Fig. 3a and Supplementary Fig. 3). Each LolF promoter consists of four transmembrane helices (TM1–TM4) and a large PD (residues G61–G257) (Supplementary Fig. 3a, d). The central cavity of the transporter is formed by the extended TM1 and TM2 helices from opposing protomers (Fig. 3a). These helices extend into the periplasm and connect with the PD, which consists of an α/β-rich extracytoplasmic (Sabre) subdomain and a β-α-β (Porter) subdomain (Supplementary Fig. 3a). The shorter TM3 and TM4 helices laterally flank the central cavity, with the intervening loop forming door-bar structures (residues S345 to R373) that project into the cavity through the lateral gates (Fig. 3a and Supplementary Fig. 3a). On the cytoplasmic side, the LolD dimer engages LolF through coupling helices located between TM2 and TM3, adopting the conserved ABC transporter NBD fold, comprising a RecA-like subdomain and an ATP-binding helical subdomain (Supplementary Fig. 3b).

This overall architecture of A. baumannii LolDF is broadly conserved relative to E. coli LolCDE, including the four-helix transmembrane core, the bipartite periplasmic Sabre and Porter subdomains, and the cytoplasmic NBD folds of LolD (Fig. 3a). However, in contrast to the asymmetric LolCDE complex, which contains distinct LolC and LolE subunits, A. baumannii LolDF forms a symmetric LolF homodimer. LolF shares ~38% sequence identity with E. coli LolC and LolE and retains structural features of both, resulting in a symmetric central cavity that is broader than that of LolCDE and flanked by door-bar elements on both sides (Supplementary Figs. 4 and 5a). Notably, whereas the LolCDE cavity is inherently asymmetric and functionally differentiated between LolC and LolE, the LolDF cavity appears structurally equivalent across protomers, suggesting a distinct mode of substrate engagement and translocation. Together, these features define a noncanonical architecture of the A. baumannii Lol system and provide a structural basis for the selective inhibition by abaucin.

Abaucin traps LolDF in an outward-open intermediate

Both abaucin-bound structures exhibit an outward-facing conformation with a central cavity open towards the periplasm (Fig. 3a). This architecture resembles the substrate-bound open state of LolCDE (PDB: 7ARH, 7MDX, and 7V8L), but with a substantially broader cavity (Supplementary Fig. 5a). Within this cavity, additional density is revealed that can be unambiguously assigned to two abaucin molecules with well-defined binding poses and interactions (Fig. 3a and Supplementary Fig. 2f). These abaucin molecules are occupying the hydrophobic pocket underneath the door-bars elements (Supplementary Fig. 5b). However, no lipoprotein density is observed in either structure.

The two conformations differ primarily in the degree of cavity compression. Relative to the wide state, the narrow state shows partial constriction of the central cavity, accompanied by inward movement of the PDs towards the central axis, but does not reach a fully closed state (Supplementary Fig. 5c). ATP and magnesium ions are resolved in the NBDs of both structures (Fig. 3a). However, unlike nucleotide-bound LolCDE (PDB: 9GVK), where nucleotide binding tightly couples to full NBD dimerization and subsequent cavity closure to expel the substrate (Supplementary Fig. 5d), the NBDs in both abaucin-bound LolDF structures remain undimerized (Supplementary Fig. 5e). This indicates that the observed cavity narrowing is not driven by the canonical ATP-coupled allosteric cycle, but instead reflects a local conformational response associated with abaucin binding.

Morphing between the wide and narrow states reveals a closing tendency of the TMDs and PDs (Supplementary Video 1). Similar closure transitions have been reported for LolCDE, although the conformational changes in LolDF occur symmetrically (Supplementary Video 1). This progression does not reach a fully closed state. Instead, the two abaucin molecules remain lodged within the central cavity, trapping LolDF in a partially open conformation.

Structural and functional characterization of the abaucin-LolDF interface

To elucidate the molecular basis of inhibition, we first characterized abaucin binding within the LolDF complex. Abaucin comprises three structural moieties: benzoxazine, piperidine, and trifluoromethyl-phenyl group (Fig. 3b). Two abaucin molecules adopt a C2-symmetric configuration, with their benzoxazine rings facing and interacting with each other at the center of the V-shaped cavity (Fig. 3a–d and Supplementary Fig. 6a). The pair is stabilized primarily by the door-bar loops that extend into the cavity from opposing LolF protomers (Fig. 3c and Supplementary Fig. 6a). The interacting benzoxazine rings are coordinated by the door-bar loop residues F360, D361, A362, and Y363 (Fig. 3c and Supplementary Fig. 6a). In the narrow-cavity conformation, abaucin establishes additional contacts with V47, M48, L260, N262, A263, and L270 within the TM helices of LolF (Fig. 3d and Supplementary Fig. 6a). Single substitutions at several interface residues (V47, M48, L260, A263, L270, F360, and Y363) caused bacterial lethality, underscoring their essential roles in LolDF function (Supplementary Fig. 6b, c). These TM residues are likely involved in lipoprotein recognition, as they align with the acyl chain-binding sites in LolCDE (Supplementary Figs. 4 and 5b).

For the non-lethal interface mutants D361A and A362D, which are located within the door-bar loop, we further examined their binding affinity and cellular response to abaucin. Purified LolDF^A362D exhibited an approximately tenfold reduction in abaucin binding relative to wild-type LolDF (Fig. 4a, b). This result is consistent with the previous report by Liu et al., showing that the LolDF^A362T mutant conferred a 16-fold increase in resistance to abaucin⁵⁴, supporting the door-bar region as a conserved abaucin-binding site. We next examined the morphology of wild-type cells and cells expressing mutant LolDF variants following abaucin treatment using SEM (Fig. 4c). Because the pathogenic nature of A. baumannii limits genetic manipulation, we used the closely related species Acinetobacter baylyi (A. baylyi) as a surrogate model. Notably, A. baylyi ADP1 exhibits an abaucin MIC comparable to that of A. baumannii ATCC 17978 (Fig. 4d), and their lolDF genes share 88% sequence identity (Supplementary Fig. 7a), supporting its suitability for functional characterization. A. baylyi strains with silenced lolDF (Supplementary Fig. 7b) were complemented with wild-type or non-lethal mutant A. baumannii LolDF variants (LolF^D361A and LolF^A362D) and then treated with or without abaucin. Cells expressing wild-type LolDF displayed extensive envelope collapse or rupture following abaucin treatment, accompanied by pronounced loss of viability (Fig. 4c). In contrast, cells expressing mutant LolDF variants exhibit improved survival and reduced morphological damage under the same conditions (Fig. 4c). Together, these findings indicate that weakening abaucin–LolDF interactions mitigates envelope damage and enhances bacterial tolerance, providing structural and functional support for the inhibitory mechanism and the species-selective activity of abaucin.

Fig. 4. Characterization of abaucin binding to LolDF.

a Bio-layer interferometry (BLI) sensorgrams showing abaucin affinity to wild-type LolDF or LolDF mutants carrying substitutions in the abaucin-binding sites. b Averaged K_D value from the BLI assay. Statistical significance was assessed by one-way ANOVA with Tukey’s multiple comparisons test. Data are shown as mean ± SEM. c Scanning electron microscopy (SEM) images of wild-type A. baylyi and the corresponding LolDF mutant strains with or without abaucin treatment. d MICs were determined for wild-type A. baumannii ATCC 17978 and A. baylyi ADP1. Data in (a–c) are representative of n = 3 independent experiments. Source data are provided as a Source Data file.

Abaucin blocks the transport activity of LolDF without preventing substrate binding

To determine whether abaucin interferes with lipoprotein binding or transport, we next examined its effects on substrate association with LolDF and transfer to LolA in vitro. Lipoprotein-associated LolDF complex was obtained by co-expressing A. baumannii LolDF with the lipoprotein BamE in A. baylyi. The purified complex was then incubated with LolA in the presence or absence of abaucin, ATP, and MgCl₂. Abaucin did not affect the amount of BamE associated with LolDF; however, it markedly impaired BamE transfer to LolA at concentrations ≥20 μM (Fig. 5a). Consistent with this, LolDF ATPase activity in the presence of abaucin (≥20 μM) was reduced to basal levels, which is comparable to those of the hydrolysis-deficient mutant LolD^S148EF (Fig. 5b). These results indicate that abaucin does not prevent substrate binding but instead directly inhibits LolDF ATP hydrolysis and transporter function.

Fig. 5. Abaucin inhibits lipoprotein transfer while sparing lipoprotein extraction.

a In vitro lipoprotein transport assay in the presence of ATP, MgCl₂, and abaucin. b ATPase activity of detergent-solubilised LolDF-BamE in the presence of LolA and abaucin (***p < 0.001). Abaucin reduced LolDF-associated ATPase activity to levels comparable to those of the ATPase-impaired LolD^S148EF control. Data in (a, b) are representative of n = 3 independent experiments; data in (b) are means ± SEM one-way ANOVA with Tukey’s multiple comparisons test. Statistical source data for (b) are provided in Source Data. c Cryo-EM structure of the lipoprotein-associated abaucin-bound A. baumannii LolDF complex. LolF protomers are shown in blue and yellow, and LolD protomers in green and olive green. The central inset shows density within the transmembrane cavity fitted with two abaucin molecules and the N-terminus of a lipoprotein substrate, including two acyl chains and a short peptide segment. d Comparison of the open cavity conformations of the lipoprotein-associated, abaucin-bound LolDF complex and the abaucin-bound LolDF structures in the absence of substrate. Source data are provided as a Source Data file.

To directly visualize this inhibited state, we further determined the cryo-EM structure of the lipoprotein-associated A. baumannii LolDF-BamE complex, prepared in A. baylyi and treated with abaucin, at an overall resolution of 3.34 Å (Supplementary Fig. 8). The structure adopts an outward-open cavity conformation closely resembling the narrow abaucin-bound state (Fig. 5c, d). Notably, the cryo-EM map reveals extra density within the central cavity consistent with both bound abaucin molecules and N-terminal features of a lipoprotein substrate (Fig. 5c). Although the local resolution does not permit detailed modeling, two acyl chains and part of the N-terminal peptide could be assigned based on the lipoprotein conformation observed in LolCDE structures (PDB: 7ARH, 7MDX, and 7V8L), including a characteristic straddled configuration of the N-amine-linked acyl chains (Fig. 5c).

The co-occupancy of lipoprotein and abaucin within the expanded central cavity indicates that abaucin does not compete with substrate binding but instead locks the transporter in a lipoprotein-accessible open state, preventing further transport. In LolCDE, substrate translocation is coupled to conformational rearrangements of the door-bar region following closure of the central cavity (Supplementary Fig. 9a–c). In contrast, abaucin binding in LolDF rigidly stabilizes the door-bar loops, thereby preventing the conformational changes required for lipoprotein translocation and blocking the substrate exit pathway (Supplementary Fig. 9d, e). Together, these findings support a model in which abaucin traps the transporter in a substrate-accessible but transport-incompetent state. This mechanism appears to depend on the symmetric engagement of the two door-bar loops contributed by the LolF homodimer, a structural arrangement absent from the asymmetric E. coli LolCDE complex, thereby providing a structural basis for the species selectivity of abaucin.

Discussion

Carbapenem-resistant A. baumannii represents a critical global health threat due to its formidable drug resistance and persistence in hospital environments^3,55. Abaucin, a compound recently identified by Liu et al. through a machine-learning-guided screen, emerged as a promising candidate with striking species selectivity against A. baumannii⁵⁴. Previous work established the activity of abaucin using archived isolates; here, we extend that analysis to freshly collected multidrug-resistant strains from ICU patients (Fig. 1). These contemporary clinical isolates reflect current selective pressures in healthcare settings and therefore provide a realistic assessment of abaucin’s therapeutic potential. By integrating susceptibility testing, animal infection models, and structural analyses, we establish a direct connection between abaucin’s molecular mechanism and its potent, species-selective activity in clinically relevant CRAB strains.

Abaucin has been reported to selectively inhibit the A. baumannii lipoprotein transporter LolDF, an essential component of the outer membrane biogenesis pathway^{37–39,46,48,54} (Supplementary Fig. 2a). Our structural and biochemical analysis define the molecular basis of this inhibition and its remarkable species selectivity. Central to this selectivity is a fundamental architectural divergence within the Lol transporter family: whereas A. baumannii employs a symmetric LolF homodimer, E. coli utilizes an asymmetric LolC-LolE heterodimer (Supplementary Fig. 5a). Cryo-EM structures reveal that two abaucin molecules symmetrically occupy the central cavity of LolDF, engaging both door-bar motifs and stabilizing the transporter in an inactive, substrate accessible outward-open conformation (Figs. 3a and 5c). Consistent with our observations, a recent study reported LolDF structures containing either two or four abaucin molecules, which adopt a similar conformational state⁵⁶. This binding mode is uniquely enabled by the symmetric architecture of LolDF and is not compatible with the asymmetric LolCDE complex, which lacks an equivalent composite pocket. This structural incompatibility provides a direct explanation for the intrinsic insensitivity of E. coli LolCDE to abaucin and highlights how species-specific transporter architectures can be exploited for precision antibacterial targeting.

Mechanistically, our data support a model in which abaucin co-occupies the transporter with its lipoprotein substrate and inhibits a post-binding step in transport (Supplementary Fig. 10). Unlike competitive inhibitors, abaucin does not displace the bulky lipoprotein cargo; instead, it binds adjacent to the acyl chains within the expanded central cavity, effectively acting as an allosteric wedge (Fig. 5c). Under physiological conditions, ATP binding and NBD dimerization drive inward movements of the TM2 helices and upward displacement of the periplasmic domains, leading to closure of the central cavity and extrusion of the lipoprotein substrate (Supplementary Fig. 5d). In the presence of abaucin, this conformational transition is arrested: the centrally bound abaucin pair immobilizes the door-bar motifs and prevents the closure required for substrate release (Extended Data Fig. 5c). This conformational hijacking effectively severs the allosteric communication between the transmembrane and nucleotide-binding domains, leading to a collapse of OM homeostasis.

This mechanism is supported by multiple lines of functional evidence. Abaucin does not reduce substrate association with LolDF but abolishes transfer to LolA, consistent with inhibition at a post-binding stage. ATPase activity is suppressed under transport-coupled conditions, indicating that abaucin prevents productive cycling rather than substrate engagement (Fig. 5). Structure-guided mutagenesis further corroborates this model. Morphological analyses revealed pronounced OM collapse and rupture upon abaucin treatment, hallmarks of defective lipoprotein trafficking and compromised membrane integrity (Fig. 4). Substitutions at key abaucin-interacting residues weaken binding, and the SEM of these mutant cells displaces improved envelope integrity under abaucin treatment (Fig. 4). These results establish a direct link between the structural binding site and antibacterial activity and define key determinants for future optimization.

Notably, the inhibitory mechanism of abaucin differs from other recently described strategies targeting bacterial ABC transporters. The small molecule SMT-738 inhibits LolCDE by binding at the interface between the transmembrane and periplasmic domains in the nucleotide-induced, NBD-dimerized state, inhibiting lipoprotein extraction rather than preventing lipoprotein transport⁵⁷. In contrast, macrocyclic peptides targeting the lipopolysaccharide transporter LptB₂FG function as substrate-dependent molecular glues, binding a pre-existing LPS-loaded intermediate through simultaneous interactions with both the transporter and its LPS substrate, thereby trapping LPS within the transport pathway and preventing transfer to LptA⁵⁸. By comparison, abaucin can bind the transporter in the apo state, without displacing substrate or interfering with lipoprotein extraction. Its inhibitory action is mediated through a unique form of conformational hijacking, in which symmetric engagement of the central cavity immobilizes the door-bar elements and locks the transporter in an outward-open, transport-incompetent conformation. This mechanism is structurally distinct from previously reported substrate- or ATP-dependent inhibitors, which provides a blueprint for precision, species-selective antibiotic design.

Together, these findings define abaucin as a prototype for conformation-specific inhibition of bacterial transporters, which exploits the intrinsic conformational cycle of LolDF and traps it in a nonproductive state. This mechanism underscores a broader principle: structural symmetry within essential transport complexes can be leveraged to achieve species-selective inhibition. Such precision targeting contrasts with the indiscriminate action of broad-spectrum antibiotics and may substantially reduce collateral damage to commensal microbiota, a critical consideration in preserving host-microbe balance. Beyond its mechanistic insights, this work establishes the LolDF complex as a validated therapeutic target. Its essentiality, coupled with species-specific architecture, makes it ideal for precision antibiotic development^{48,54,59–62}. This rational approach not only accelerates antibiotic discovery but also enables the design of compounds with minimized resistance potential and off-target effects.

Methods

Antimicrobial susceptibility assays

The collection and use of clinical isolates were approved by the Ethics Committee of West China Hospital, Sichuan University (2025 Review No. 390).

Susceptibility testing was performed in biological triplicate using the broth microdilution method recommended by the Clinical and Laboratory Standards Institute (CLSI M07-A112018). Aerobic bacterial strains were grown overnight in fresh Mueller-Hinton broth (MHB; Sigma-Aldrich, 90922). These included a panel of A. baumannii (SAMN34037527: HX_S9; SAMN34037535: HX_S19; SAMN34037546: HX_S31; SAMN34037486: HX_R7; SAMN34037488: HX_R9; SAMN34037507: HX_R28), ESBL-producing E. coli (SAMN28537734: HX-1-7), and carbapenem-resistant K. pneumoniae (CRKP, SAMN48633600: HX-1-13) clinical isolates from the West China Hospital Clinical Isolate Bank, A. baumannii ATCC 17978, and A. baylyi ADP1. The resulting suspensions were dispensed into 96-well flat-bottom plates containing abaucin across a concentration range of 128–1 µg mL⁻¹, and susceptibility was determined by standard broth microdilution.

For Bacteroides species (B. vulgatus (ATCC 29327: HX-25-48-A1), B. thetaiotaomicron (ATCC 29148: HX-25-52-D4), and B. fragilis (ATCC 25285: HX-25-56-I2)), susceptibility was assessed under anaerobic conditions on Brucella blood agar plates placed directly into an anaerobic chamber continuously flushed with an oxygen-free gas mixture (80% N₂, 10% H₂, 10% CO₂). After incubation at 37 °C for 16–20 h, cell viability was evaluated. Minimum inhibitory concentrations (MICs) were calculated using GraphPad Prism software.

Establishment of the mouse acute pneumonia infection model

All animal experiments were approved by the Laboratory Animal Ethics Committee of Sichuan University (protocol No. 20250313055). Specific-pathogen-free (SPF) 6–8-week-old male BALB/cJGpt mice were purchased from GemPharmatech Co., Ltd. (Nanjing, China). Male mice were used to avoid estrous cycle-related variability in inflammatory responses, ensuring consistent and reproducible results. Mice were housed under SPF conditions in ventilated cages (10–15 air changes/h) at 20–26 °C and 30–70% humidity with a 12 h light/dark cycle. Mice were rendered neutropenic by intraperitoneal administration of cyclophosphamide at 150 mg kg⁻¹ on day −4 and 150 mg kg⁻¹ on day −1 before infection. Mice were anaesthetized with isoflurane on day 0 and inoculated intratracheally with 100 μL of carbapenem-resistant A. baumannii HX-R7 suspension at 1 × 10⁶, 1 × 10⁷, or 1 × 10⁸ CFU mL⁻¹. Survival was monitored every 12 h for 72 h to estimate the median lethal dose. In parallel, lung tissues and whole blood were collected at 0, 12, 24, 48, and 72 h after infection to assess bacterial dissemination and infection progression. Based on these preliminary experiments, an inoculum of 1 × 10⁷ CFU mL⁻¹ was selected for subsequent therapeutic studies.

Therapeutic efficacy experiment

Male BALB/c mice aged 6–8 weeks were used to establish an acute pneumonia infection model, as described previously^63,64. Mice were rendered neutropenic by intraperitoneal administration of cyclophosphamide at 150 mg kg⁻¹ on day −4 and 150 mg kg⁻¹ on day −1. On day 0, mice were anaesthetized with isoflurane and inoculated intratracheally with 100 μL of CRAB (HX-R7) suspension (1 × 10⁷ CFU mL⁻¹). At 2, 12, 24, and 36 h after infection, animals received abaucin, piperacillin (PIP, a β-lactam), meropenem (MEM, a last-line carbapenem) (20 mg kg⁻¹; four doses given every 12 h) or vehicle by intraperitoneal injection. Mice were euthanized at the experimental endpoint (48 h post-infection), and faecal pellets, BALF, fresh lung tissue, and paraformaldehyde-fixed lung tissue were collected. Pulmonary function, including minute ventilation (MV), was measured as previously reported^65,66.

Bacterial load quantification

Bacterial loads in blood, lung-tissue homogenates and BALF were determined by serial dilution, plating on chromogenic agar selective for Acinetobacter baumannii (CHROMagar Acinetobacter) and overnight incubation at 37 °C. Colonies were counted from appropriate dilutions, and the mean colony-forming units (CFU) per mL were determined.

Histological analysis

Lung specimens were fixed in 4% paraformaldehyde for 48 h, embedded in paraffin, and sectioned at 4 μm using a microtome. Sections were stained with haematoxylin and eosin (H&E) and imaged with a digital pathology scanner (3DHISTECH). Lung injury was evaluated using a semiquantitative histopathological scoring system from a published method⁶⁷. Four indices were assessed: alveolar septal thickening, inflammatory cell infiltration, haemorrhage, and oedema. Alveolar septal thickening, haemorrhage, and oedema were each scored on a 0–4 scale, corresponding to none, mild, moderate, severe, and very severe changes. Inflammation was assessed by counting the total number of inflammatory cells per × 100 field. Scoring was performed in a blinded manner on randomly selected fields from each lung section. For each mouse, scores from five randomly selected fields per section were averaged to yield a final lung injury score.

Scanning electron microscopy

Lung tissue samples (~1 mm³) were fixed overnight in 2.5% glutaraldehyde, followed by post-fixation in 1% osmium tetroxide. Samples were then dehydrated through a graded ethanol series and dried under vacuum overnight. Dried tissues were sputter-coated with gold and examined using a field-emission scanning electron microscope (JSM-IT700HR InTouchScope^™, JEOL).

BALF collection

BALF was collected under terminal isoflurane anaesthesia. Immediately after cervical dislocation, the trachea was exposed and cannulated. The lungs were lavaged with a total of 1.2 mL PBS (0.4 mL × 3), yielding a recovery rate of >90%. The recovered fluid was centrifuged at 1000 × g for 10 min at 4 °C, and the supernatant was stored at −80 °C for cytokine quantification.

Enzyme-linked immunosorbent assay

Concentrations of IL-6 and IL-1β in BALF were measured using commercial ELISA kits (R&D Systems; M6000B Mouse IL-6 Quantikine ELISA Kit; MLB00C Mouse IL-1 beta/IL-1F2 Quantikine ELISA Kit) according to the manufacturer’s instructions. All biological samples were stored at −80 °C until analysis.

16S rRNA gene amplicon sequencing and bacterial community analysis

Fresh colonic contents (faecal samples) were collected after 2 days of treatment (day 2). Genomic DNA extraction and 16S rRNA gene sequencing of mouse faecal pellets were performed as described previously⁶⁸. The V3-V4 hypervariable regions of the 16S rRNA gene were amplified using primers F341 (5′–ACTCCTACGGGAGGCA GCAG-3′) and R806 (5′-GGACTACHVGGGTWTCTAAT-3′) and the KAPA HiFi HotStart ReadyMix PCR kit (KAPA Biosystems, Boston, MA, USA). Amplicons were separated on 2% agarose gels and purified with an AxyPrep DNA Gel Extraction Kit (Axygen Biosciences, Union City, CA, USA). Purified PCR products were used to construct Illumina DNA libraries, which were sequenced on an Illumina HiSeq PE250 platform (Illumina, San Diego, CA, USA).

Expression and purification of A. baumannii LolDF complex in E. coli

The operon encoding full-length LolDF from A. baumannii ATCC 17978 was polymerase chain reaction (PCR)-amplified from genomic DNA using primers DFF (5′-GATGGCTAAATAAGAAGTGTGAATGAACAC-3′) and DFR (5′-AGAACCGTTTA CCCACAGCCCATC-3′). The resulting amplicon was cloned into a modified pTrc99a vector engineered to append a C-terminal octa-histidine (8 × His) tag to LolD using a ClonExpress II One Step Cloning Kit (Vazyme, C112-01). The resulting plasmid, pTrc99a-His-lolDF, was transformed into E. coli C43 (DE3) cells (Novagen) for protein expression. Transformants were selected on Luria-Bertani (LB; Difco broth) agar plates supplemented with 100 µg mL⁻¹ ampicillin at 37 °C. A single colony was inoculated into liquid LB medium containing 100 µg mL⁻¹ ampicillin and grown to an optical density at 600 nm (OD₆₀₀) of 0.8, after which protein expression was induced with 100 µM isopropyl-beta-D-thiogalactopyranoside (IPTG) and cultures were incubated overnight at 20 °C.

Cells were harvested by centrifugation at 3000 × g for 15 min and resuspended in buffer A (20 mM HEPES, pH 7.8, 300 mM NaCl, 10% (vol/vol) glycerol) supplemented with 0.1 mM phenylmethylsulfonyl fluoride (PMSF). Cell lysis was performed by high-pressure homogenization at 900 bar, and incompletely lysed debris was removed by centrifugation (18,000 × g, 10 min, 4 °C). Membranes were collected by ultracentrifugation (100,000 × g, 1 h, 4 °C) and solubilized in buffer A containing 1% (wt/vol) n-dodecyl-β-d-maltopyranoside (DDM; Anatrace) for 1 h at room temperature. The solubilized fraction was loaded onto a 5 mL HisTrap HP column (GE HealthCare). The column was washed with buffer A containing 0.05% (wt/vol) DDM and 40 mM imidazole, and the protein was eluted with a buffer containing 0.05% (wt/vol) DDM and 280 mM imidazole. The eluate was concentrated and further purified by size-exclusion chromatography (SEC) using a Superdex 200 Increase 10/300 column (GE Healthcare) pre-equilibrated with buffer B (20 mM HEPES, pH 7.8, 150 mM NaCl, 0.05% (wt/vol) DDM). Protein-containing fractions were concentrated to ~7 mg mL⁻¹ using a 50-kDa cutoff Amicon Ultra-15 (Merck-Millipore) centrifugal filter, flash-frozen in liquid nitrogen, and stored at −80 °C for subsequent use.

Expression and purification of A. baumannii LolDF and LolDF-BamE complex in A. baylyi

The operon encoding full-length LolDF from A. baumannii ATCC 17978 was PCR-amplified from genomic DNA using primers DF-F (5′-GATGGCTAAATAAGAA GTGTGAATGAACAC-3′) and DF-R (5′-AGAACCGTTTACCCACAGCCCATC-3′). The resulting amplicon was cloned into a pJL03 vector (gifted by Prof. Zhaoqing Luo) carrying an araC-pBAD promoter and a gentamicin-resistant gene, using a ClonExpress II One Step Cloning Kit (Vazyme, C112-01), resulting in the recombination plasmid pJL03-lolDF. The bamE gene from A. baumannii ATCC 17978 was amplified using primers BamE-F (5′-AGAGTCGACCTGCAGGCATGCAATTA TTGGGATAGTGTAGCAAAATGC-3′) and BamE-R (5′-ATGGAAAAACGCCAG CTATAGTGGTGGGGCAGTTAAAATT GCTTC-3′) and cloned into the pJL03-lolDF vector, generating an arabinose-inducible LolDF-BamE co-expression construct. The recombination plasmids were further fused to LolD with a C-terminal octa-histidine (8 × His) tag, LolF with an N-terminal Flag tag, and BamE with a C-terminal Myc tag. Site-directed mutations were introduced using pJL03-lolDF as a template, as previously described⁶⁹.

Plasmids encoding wild-type, mutant LolDF, or LolDF-BamE were transformed into A. baylyi via natural transformation. Transformants were selected on LB agar supplemented with gentamicin (30 µg mL⁻¹). A single colony was inoculated into liquid LB medium containing gentamicin (30 µg mL⁻¹) and grown to an optical density at 600 nm (OD₆₀₀) of ~0.8, after which protein expression was induced with 0.2% L-arabinose and continued for 12 h overnight. Cells were harvested by centrifugation at 3000 × g for 15 min and resuspended in buffer A (20 mM HEPES, pH 7.8, 300 mM NaCl, 10% (vol/vol) glycerol) supplemented with 0.1 mM PMSF. Proteins were purified as described above. Purified wild-type LolDF and the LolDF mutants were used for bio-layer interferometry analysis. Purified LolDF-BamE complexes were used for the in vitro lipoprotein transfer assay, followed by immunoblot analysis to detect BamE transfer to LolA. Unprocessed scans of the blots are provided in the Supplementary Information file.

Expression and purification of A. baumannii LolA in E. coli

The DNA encoding the mature LolA protein (residues 29–231) from A. baumannii ATCC 17978 was PCR-amplified using primers AF (5′-GCACCCGTAGCAGCACCG GAGCAACAGGCAATTG-3′) and AR (5′-TTATTGATCAATGATATCGGTACCTTT CGGAGGCGTA-3′) and cloned into a modified pET28a (+) vector. The final construct encoded an N-terminal 8 × His tag followed by a small ubiquitin-like modifier (SUMO) fusion, a tobacco etch virus (TEV) protease cleavage site, as well as a C-terminal Strep tag, generating the fusion protein 8 × His-SUMO-TEV-LolA_29–231-Strep.

This construct was transformed into E. coli BL21 (DE3) cells (Novagen) for protein expression. Transformed cells were grown in LB medium supplemented with 50 µg mL⁻¹ kanamycin at 37 °C until the OD₆₀₀ reached 0.6. Protein expression was then induced with 1 mM IPTG for 3 h. Cells were harvested and lysed, and the lysate was subjected to Ni-NTA affinity chromatography. The protein was eluted with buffer B containing 300 mM imidazole, followed by desalting using a HiPrep 26/10 column (GE Healthcare) to remove imidazole. The 8 × His-SUMO tag was cleaved by TEV protease, and the reaction mixture was reapplied to a Ni-NTA column to remove the cleaved tag and any uncleaved protein. The purified Strep-tagged LolA was subsequently desalted, concentrated to 9 mg mL⁻¹, and stored at −80 °C for further use.

Cryo-electron microscopy sample preparation and data acquisition

The LolDF complex was freshly purified for cryo-EM analysis. To prepare the abaucin-bound LolDF sample, 25 µg of purified LolDF was incubated with 5 mM ATP, 2 mM MgCl₂, and 5 mM abaucin for 10 min at room temperature before application to glow-discharged Quantifoil R1.2/1.3 grids, 300 mesh copper. For the lipoprotein-associated abaucin-bound LolDF sample, the purified LolDF-BamE complex was incubated with 5 mM abaucin before grid preparation. Grids were blotted for 3–3.5 s under 95% humidity and vitrified in liquid ethane using a Vitrobot Mark IV (Thermo Fisher Scientific). Data were collected on a Titan Krios electron microscope operating at 300 kV, equipped with a K3 Summit direct electron detector (Gatan) or a Thermo Fisher Falcon 4i Summit direct electron detector coupled to an energy filter.

For the SerialEM dataset, micrographs were acquired using SerialEM at a nominal magnification of ×105,000, corresponding to a pixel size of 0.824 Å. Defocus values ranged from −1.0 to −3.0 µm. Each exposure lasted 8 s and was dose-fractionated into 40 frames, with a dose rate of ~7.8 counts s⁻¹ pixel⁻¹⁷⁰.

For the EPU dataset, images were collected at a nominal magnification of ×165,000, corresponding to a pixel size of 0.735 Å, with defocus values ranging from −1.0 to −2.0 µm. The total accumulated electron dose was set to 50 e⁻/Ų.

Image processing

For the SerialEM dataset, motion correction for image stacks was performed using MotionCor2⁷¹, producing dose-weighted, 2 × binned micrographs with a pixel size of 0.824 Å. Contrast transfer function (CTF) parameters were estimated with Gctf⁷², and subsequent data processing was carried out in RELION⁷³. 3,609,568 particles were selected by CTF and curated through four rounds of 3D classification in cryoSPARC⁷⁴, resulting in 438,065 particles without refinement. Additional 3D refinement yielded two classes: one with abaucin-bound LolDF, featuring a wide central cavity (3.04 Å resolution), and the other with a narrow cavity (3.15 Å resolution). These maps were generated after post-processing with B factors of −60 and −60 Å^2, respectively. C2 symmetry was imposed on the two maps of abaucin-bound LolDF. The final global resolutions were determined based on the gold-standard Fourier shell correlation (FSC) criterion at a cutoff of 0.143. Data-processing details are summarized in Supplementary Fig. 2.

For the EPU dataset, micrographs were motion-corrected and dose-weighted using MotionCor2, and CTF parameters were estimated with CTFFIND4. Initial particles were manually picked from a subset of high-quality micrographs and subjected to 2D classification to generate templates for automated particle picking in cryoSPARC. Multiple rounds of 2D classification yielded a curated dataset of 277,786 high-quality particles. These particles were subjected to ab initio reconstruction to generate initial 3D volumes, followed by heterogeneous refinement (bin2) to resolve distinct conformational states. Particles from the most homogeneous, well-resolved class were re-extracted at full pixel size (bin1) and subjected to a final round of 3D classification. The dominant class was retained for high-resolution refinement, resulting in a final dataset of 89,001 particles. Homogeneous refinement and subsequent non-uniform refinement in cryoSPARC yielded a map at a global resolution of 3.34 Å. C1 symmetry was imposed during refinement of the lipoprotein-associated abaucin-bound LolDF map. The final global resolution was determined based on the gold-standard FSC criterion at a cutoff of 0.143. Data processing details are summarized in Supplementary Fig. 8.

Model building and refinement

An AlphaFold3-predicted LolDF structure was used as the starting model and was docked into the density map using ChimeraX⁷⁵. Model building proceeded through iterative rounds of real-space refinement in PHENIX⁷⁶ (v1.20.1-4487) and manual adjustments in Coot⁷⁷ (v0.9.8.95). The final model was validated with MolProbity to assess stereochemistry and overall model quality⁷⁸, with detailed statistics provided in Table S1.

LolDF silencing in A. baylyi

A chloramphenicol-tetR-tetO cassette was inserted upstream of the lolF locus in A. baylyi ADP1 by the procedure described above⁷⁹, to generate a tetracycline-inducible tetO-controlled lolDF strain (designated PJW2125). An rrnB_T1-rrnB_T2 terminator was inserted upstream of the tetR-tetO cassette to terminate transcription from the original promoter (Supplementary Fig. 7b). Positive clones were confirmed by PCR (PCR primer: tetF: 5′-GGTATTCATTTGGGCATCTACACCCCTGG-3′; tetR: 5′-CGCAATATCAGATTTT TTGTCTGTGACGAC-3′) and gene sequencing. This strain exhibited inducer-dependent growth, with lethality occurring in the absence of anhydrotetracycline (aTc; MedChemExpress) due to the silencing of LolDF expression (Supplementary Fig. 7b). For genetic complementation, the lolDF-silenced A. baylyi strain was complemented with the pJL03-lolDF plasmid, which encodes A. baumannii ATCC 17978 LolDF under the control of an arabinose-inducible promoter. The transformants carrying pJL03-lolDF were selected in media supplemented with appropriate antibiotics or selection agents and induced by arabinose for protein expression, which was monitored by immunoblotting for His-tagged LolD or Flag-tagged LolF. Site-directed mutations were introduced using the pJL03-lolDF plasmid as a PCR template for mutagenesis

Site mutagenesis and cell viability assays

Site-directed mutations were introduced using pJL03-lolDF as a template, as previously described⁶⁹. Plasmids carrying gentamicin resistance and either wild-type or mutant genes were introduced into strain PJW2125. Transformants were selected on LB agar containing 30 µg mL⁻¹ gentamicin, 20 µg mL⁻¹ chloramphenicol, and 0.5 µg mL⁻¹ anhydrotetracycline, with incubation at 37 °C for 12 h. Control strains included PJW2125 transformed with the empty pJL03 plasmid (negative control) and plasmids expressing wild-type LolDF (LolD(His)F(Flag)) (positive controls). Cells were harvested, washed, and resuspended in sterile LB to an OD₆₀₀ of 0.5, followed by tenfold serial dilution (10⁻¹–10⁻⁵). Each dilution was spotted onto LB agar plates with gentamicin, with or without anhydrotetracycline, and incubated overnight at 37 °C. All functional assays were performed in triplicate.

Bio-layer interferometry (BLI) to assess binding kinetics

To characterize the binding kinetics of abaucin to LolDF complexes, purified wild-type and A. baumannii LolDF mutants were prepared using the A. baylyi ADP1 expression system as described above. Binding measurements were performed by biolayer interferometry (BLI) on an Octet K2 instrument (Sartorius, Göttingen, Germany). Desalted proteins were immobilized onto Ni-NTA-coated biosensors (Sartorius, 18-5101). After baseline equilibration, sensors were transferred to wells containing serially diluted abaucin in buffer B supplemented with 0.1% (wt/vol) DDM.

The assay protocol consisted of the following steps: (1) baseline equilibration for 5 min in assay buffer; (2) sensor loading with LolDF or the mutants for 5 min; (3) secondary baseline for 60 s; (4) association with abaucin for 100 s; and (5) dissociation in buffer for 80 s. Sensorgram data were processed using Octet Analysis Studio (v13.0) and fitted to a global 1:2 binding model.

Scanning electron microscopy for bacteria

Following washing with PBS, 10 µL of cell suspension was applied to 0.1% poly-L-lysine-coated coverslips and allowed to adhere for 1 h. Cells were then fixed overnight at 4 °C with 2.5% (vol/vol) glutaraldehyde in 0.1 M PBS. After fixation, samples were rinsed with PBS for 5–10 min and progressively dehydrated through an ethanol gradient (30, 50, 60, 75, 85, 95, and 100%), with each concentration step lasting 5–10 min. Coverslips were critical-point dried using a K850 system (Quorum Technologies, UK). For SEM imaging, dried coverslips were mounted on aluminum stubs with carbon-doped conductive tape and sputter-coated with gold using a Cressington 108 coater (Cressington Scientific Instruments, UK). Images were acquired on an EVO 10 scanning electron microscope (Zeiss) operating at an accelerating voltage of 20 kV.

In vitro lipoprotein binding and transportation assay

The single-cycle in vitro lipoprotein transport assay was adapted from a previously established E. coli LolCDE protocol³⁷. Purified A. baumannii LolDF-BamE complex was pre-incubated with abaucin (concentration range from 0 to 40 µM) for 10 min and reacted with excessive purified LolA (~1:21 molar ratio) for 20 min at 4 °C in the presence of 1 mM ATP and 2 mM MgCl₂. LolA was subsequently purified using a 5 mL StrepTrap column (GE Healthcare) as described above. Bound proteins were eluted with buffer B supplemented with 3 mM D-desthiobiotin (Sigma-Aldrich, D1411), concentrated, and analyzed by immunoblotting for LolA and captured BamE using anti-Strep (Sigma-Aldrich, SAB2702216, GT661) and anti-Myc antibodies (Sigma-Aldrich, M4439, 9E10), respectively. Samples taken prior to the transport reaction were also analysed for BamE binding and LolDF using anti-Myc antibodies and anti-His antibodies (Sigma-Aldrich, SAB2702218, GT359). Protein bands were visualized by chemiluminescence using an Amersham ImageQuant 800 imaging system. All assays were performed in triplicate.

ATPase activity assay

ATPase activity was measured using a commercial ATPase assay kit (Bioassay Systems) following the manufacturer’s protocol and previously established methods³⁷. Purified A. baumannii LolDF-BamE (1–2 mg mL⁻¹) were pre-incubated with A. baumannii LolA in the presence or absence of abaucin (20 or 40 µM). For the LolD^S148EF mutant, ATPase activity was measured directly by mixing the purified mutant protein with buffer B, assay buffer, and 4 mM ATP.

After 30 min of incubation at room temperature, the reaction was terminated by adding the stop reagent, followed by an additional 30 min incubation. Absorbance was recorded at 630 nm, and ATPase activity was quantified based on a phosphate standard curve. Phosphate standards and the blank control for colorimetric detection were prepared according to the manufacturer’s instructions (ATPase assay kit). ATPase activity for each sample was calculated from the mean of replicate measurements using the corresponding phosphate standard curve. All experiments were performed in at least three independent replicates.

Statistical analysis

All calculations or analyses were performed using the Prism 8 software package (GraphPad), or microbiome data were performed with the use of the ggpubr R software, version 4.3.1. All results shown were obtained by at least three biologically independent experiments, and the values are presented as means ± standard error of the mean (SEM). The Shapiro–Wilk test was applied to assess the normality of the data. The standard Student’s t test was applied when there were two groups with a normal distribution, while one-way ANOVA with Tukey’s multiple comparisons test for experiments containing more than two groups.

Author contributions

H.D. and X.T. conceived and designed the experiments. J.P. and Y.C. constructed the expression plasmids. J.P., Y.C., L.W., J.W., and H.C. expressed and purified the proteins and performed mutagenesis, ATPase assays, and functional analysis. D.Z., X.G., M.X., and B.D. established and evaluated the mouse pneumonia infection model. Y.X. and B.Y. isolated the clinic bacterial strains. J.P., Y.C., M.D., and L.Z. prepared the cryo-EM samples. H.D., Z.Z., Z.Z.R., A.W., Y.Z., and L.O. carried out cryo-EM data collection, processing, and structural reconstitution. H.D., X.T., Z.Z., and S.Z. performed model building and refinement. H.D., X.T., J.P., Y.C., X.W., and B.D. wrote the manuscript.

Peer review

Peer review information

Nature Communications thanks Konstantinos Beis, Wangxue Chen, and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Funding

This work was supported by the National Key Research and Development Program of China (2021YFA1301900 and 2021YFA1301203 to H.D.); the National Natural Science Foundation of China (31900039 and 32170029 to X.T., 81971974 and 82372297 to H.D.); the 1.3.5 Project for Disciplines Excellence of West China Hospital, Sichuan University (ZYYC23014 to H.D. and ZYGD23036 to B.Y.); and the Sichuan Provincial Natural Science Foundation Outstanding Young Scientist Fund Project (2024NSFJQ0002 to H.D.).

Data availability

The cryo-EM data generated in this study have been deposited in the EMDB and PDB under accession codes EMD-53518 and PDB 9R20 (Abaucin-bound LolDF, wide cavity), EMD-53522 and PDB 9R24 (Abaucin-bound LolDF, narrow cavity), and EMD-57887 and PDB 30NX (Abaucin-bound LolDF with lipoprotein). The 16S rRNA sequencing data generated in this study have been deposited in the NCBI Sequence Read Archive (SRA) under the following accession codes: PRJNA1477226. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains Supplementary material available at 10.1038/s41467-026-75428-4.