The VgrG2 effector of Acinetobacter baumannii mediates immune evasion by repressing Csu pilus assembly and triggering phagocyte methuosis Authors: Wenzeng Xu, Jiancang Zhou, Wang Zhang, Yan Jiang, Qingye Xu, Xiaochen Liu, Yisha Zhang, Xiaoxiang Tan, Junxin Zhou, Luyun Cui, Yake Yao, Liang Liang, Chengying Xu, Xiaoting Hua, Yunsong Yu, Hua Zhou Published: 2026 ABSTRACT AbstractAcinetobacter baumannii is a formidable nosocomial pathogen whose multidrug resistance and immune evasion capabilities present a major therapeutic challenge. This study identifies the type VI secretion system (T6SS) effector VgrG2 as a critical virulence regulator in A. baumannii. VgrG2 employs a two-pronged strategy to undermine host innate immunity. First, it transcriptionally represses the major adhesin Csu pilus, reducing bacterial recognition, phagocytic uptake, and neutrophils recruitment. Second, VgrG2 directly targets and hyperactivates the small GTPase Rac1 within phagocytes. This leads to actin cytoskeletal disarray, which drives excessive macropinocytosis, resulting in compromised phagocytosis and methuosis—a non-apoptotic, vacuole-dependent death specific to phagocytes. Thus, VgrG2 selectively eliminates key immune sentinels and alters pulmonary inflammation, promoting bacterial persistence and dissemination. These findings extend the role of T6SS effectors beyond interbacterial competition to include sophisticated eukaryotic-directed attacks, and highlight the VgrG2-Rac1 axis as a promising therapeutic target.Subject terms: Microbiology, Infection, Pathogens Introduction Acinetobacter baumannii is a gram-negative coccobacillus that has become one of the most frequently isolated bacteria from clinical specimens1. Besides, A. baumannii can persist on abiotic surfaces in the clinical environment due to its resistance to desiccation and its ability to form biofilms2,3. This opportunistic microorganism predominantly infects immunocompromised patients4,5. Nosocomial infections caused by A. baumannii occur at multiple anatomical sites, including ventilator-related pneumonia, central-line-related bloodstream infections, skin and soft tissue infections, surgical site infections, and catheter-related urinary tract infections6–8. Among these, ventilator-related pneumonia and sepsis are the most serious and are associated with a high mortality rate, causing a tremendous financial burden7. However, the treatment of A. baumannii infections has been a significant challenge due to its plastic drug-resistant phenotypes, particularly carbapenem resistance9,10. A. baumannii is also listed as one of the ESKAPE pathogens—Enterococcus faecium, Staphylococcus aureus, Klebsiella pneumoniae, Acinetobacter baumannii, Pseudomonas aeruginosa and Enterobacter spp.—the six leading antimicrobial-resistant bacteria for which effective treatments are urgently needed11–13. Carbapenem-resistant Acinetobacter baumannii (CRAB) isolates frequently exhibit co-resistance to multiple other antimicrobial classes. This propensity has contributed to the global dissemination of extensively drug-resistant (XDR) and, in some cases, pan-drug-resistant (PDR) strains14,15. Surveillance data indicate that carbapenem resistance in line-associated bloodstream infections ranged from 35% in general wards to 47% in ICUs and nearly 77% in long-term care facilities16. In Peru, over 97% of clinical A. baumannii isolates are carbapenem non-susceptible17. This highlights an urgent need for new therapeutic targets against this life-threatening disease. To effectively contain and treat multidrug-resistant A. baumannii, it is essential to fully elucidate the host immune response and the molecular mechanisms triggered by this bacterium. Research in mouse models has shed light on the host immune response to A. baumannii infection, characterizing it by the local recruitment of macrophages and neutrophils and the accompanying production of chemokines and pro-inflammatory cytokines18–20. Furthermore, investigating the key determinants of A. baumannii virulence and its coping mechanisms under host pressure is equally critical. Despite advances in virulence research, the interaction mechanisms between A. baumannii and its host remain incompletely understood21. The type VI secretion system (T6SS) is a contractile injection nanomachine widely distributed among Gram-negative bacteria, and its injection mechanism closely mimics that of contractile bacteriophage tails22. This system is composed of a membrane-spanning anchoring complex, a baseplate, an inner tube formed by stacked hexameric rings of Hcp protein and encased within a contractile sheath, and a puncturing spike complex located at the apex, which comprises a VgrG trimer and a conical, sharpened PAAR protein23,24 (Fig. 1A). Assembly of the T6SS apparatus initiates in an extended conformation on the baseplate platform. Upon contact with a target cell, the sheath undergoes a dramatic conformational transition and rapidly contracts in a spring-like manner, propelling the inner tube together with the spike complex out of the bacterial cell at high velocity. The spike complex, equipped with the sharp architecture of the PAAR protein, penetrates the target cell membrane. Effector proteins, which are either non-covalently associated with the VgrG, Hcp, or PAAR components or genetically fused as C-terminal extensions of these structural proteins, are thereby precisely delivered into the target cell, ultimately mediating toxic effects on bacterial competitors or physiological manipulation of eukaryotic host cells25. RNA sequencing showed significant differences between in vivo and in vitro A. baumannii and established VgrG2 as an essential virulence factor. A Schematic of T6SS, Created in BioRender. wenzeng, X. (2026) https://BioRender.com/yrs3rct. OM outer membrane, IM inner membrane, PAAR Proline-Alanine-Alanine-Arginine, VgrG Valine-glycine repeat protein G, Hcp Hemolysin co-regulated protein. B–D RNA-seq of A. baumannii from infected mouse lungs vs. LB broth (n = 3). B Study overview, Created in BioRender. wenzeng, X. (2026) https://BioRender.com/3kdg2rn. C Volcano plot depicting differentially expressed genes. vgrG3, p < 0.0001; tssA, p = 0.0487; tssC, p = 0.0008; hcp, p = 0.0017; vgrG2, p = 0.0002; V4Q77_RS06785, p = 0.0403. D Heatmap of acinetobactin biosynthesis genes. E T6SS and vgrG clusters in AB5075. F Relative expression of vgrG2, hcp and rhs in vivo versus in vitro (n = 5 biologically independent bacterial cultures per group). G Kaplan–Meier survival curves of Galleria mellonella (n = 30 per group). H Western blot of lysate (L) and supernatant (S) fractions from WT and mutants, probing for VgrG2. I Kaplan–Meier survival curves of mice (n = 10 per group). Lung photographs (J) and H&E-stained sections (K) from infected mice. L Bacterial loads in organs at 24 hpi (n = 5 biologically independent mice per group). M Live-cell imaging shows human lung organoids exposed to WT A. baumannii or the ΔvgrG2 mutant (10⁷ CFU/mL) for 24 h. N Survival of mice infected with WT, ΔvgrG2 or complemented ΔvgrG2::vgrG2 strains (n = 10 per group). O Western blot of lung lysates confirming VgrG2 expression in the complemented strain. The results are expressed as mean ± SD. p values were determined by two-sided edgeR quasi-likelihood F-tests with Benjamini–Hochberg adjustment for multiple comparisons (C), two-sided unpaired Student’s t test (F and L) and Logrank (Mantel–Cox) test (G, I, and N). Data are representative of at least three independent experiments with similar results. Source data are provided as a Source Data file. The Csu pilus is a critical adhesive structure assembled via the archaic chaperone-usher pathway in A. baumannii and plays a central role in biofilm formation. Studies have demonstrated that Csu pili mediate the initial adhesion of bacteria to abiotic surfaces, including glass, plastic, and medical materials, as well as to host epithelial cells such as the A549 cell line, and constitute an essential component for the development of mature biofilms and pellicles26. Bistable expression of Csu pili, in coordination with poly-N-acetylglucosamine, orchestrates the formation of characteristic “mountain-like” biofilm patches and enhances bacterial resistance to antimicrobial stressors such as colistin. Functionally, Csu pili not only contribute to the structural integrity of biofilms but also serve as important virulence factors that facilitate tissue colonization—including the lungs, spleen, and liver—during systemic infection. Deletion of the Csu operon markedly impairs biofilm formation capacity, whereas overexpression augments biofilm robustness27. Furthermore, Csu pili belong to the mannose-sensitive type 1 pilus family, and Csu-mediated adhesion and biofilm formation can be specifically inhibited by D-mannose and methyl-α-D-mannopyranoside26. Collectively, these attributes underscore the pivotal role of Csu pili in both environmental persistence and the pathogenic processes of A. baumannii. In this study, we constructed an acute murine lung infection model to examine the interactions between A. baumannii and the host. Our results demonstrated that VgrG2, a component of the T6SS, plays an essential role in the virulence of A. baumannii. Utilizing single-cell RNA sequencing, we further characterized how VgrG2 shapes the host response to infection and elucidated the underlying mechanisms through which VgrG2 facilitates immune evasion by A. baumannii, downregulates the expression of the Csu pilus, and triggers methuosis in phagocytes. Results RNA sequencing revealed significant differences between A. baumannii in vivo and in vitro, and VgrG2 was identified as an essential virulence factor To investigate the host-pathogen interaction during acute pulmonary infection, we selected A. baumannii strain AB5075 as our model pathogen. AB5075 is a highly virulent, multidrug-resistant U.S. military clinical isolate, widely used as a model strain for studying pathogen evolution in murine systems28. Based on a dose-escalation study (4 × 10⁶–4 × 10⁹ CFU) in C57BL/6 mice, an intranasal inoculum of 4 × 10⁷ CFU of AB5075 was selected to achieve reproducible acute lung infection without immediate lethality (Supplementary Fig. S1A, B). To delineate the transcriptional reprogramming of A. baumannii in response to the host milieu, we performed comparative RNA-seq analysis on bacterial populations harvested from infected murine lungs versus those grown in LB broth (Fig. 1B). This comparison revealed 1221 differentially expressed genes (DEGs) in the in vivo condition (Fig. 1C, Supplementary Data 1), comprising 587 up-regulated and 634 down-regulated genes, indicative of a profound shift in the transcriptional landscape under host-imposed constraints. Notably, the acinetobactin siderophore biosynthesis cluster was down-regulated in vivo (Fig. 1D), correlating with diminished iron uptake. This finding contrasts with a previous report of up-regulated siderophore genes during bacteremia29, suggesting niche-specific regulation. Concurrently, half of the genes involved in Csu pilus biosynthesis were suppressed (Supplementary Fig. S1C), a potential immune evasion strategy to reduce phagocytic recognition. While lipopolysaccharide (LPS) core biosynthesis genes remained unchanged (Supplementary Fig. S1D), genes responsible for LPS export were up-regulated (Supplementary Fig. S1E). Expression of capsule biosynthesis was largely unaltered. KEGG enrichment analysis of up-regulated DEGs highlighted elevated activities in central carbon metabolism (e.g., pyruvate metabolism), fatty acid catabolism, and the biosynthesis of branched-chain amino acids (valine, leucine, isoleucine) (Supplementary Fig. S1F). Furthermore, key regulatory systems, including two-component systems and quorum sensing, were up-regulated, likely facilitating adaptation to host pressures. To elucidate key virulence determinants active during host interaction, we analyzed differentially expressed genes encoding bacterial secretion system components. Up-regulated factors included OmpA of the type I secretion system (T1SS), and T6SS-associated genes tssC, hcp, vgrG2, and V4Q77_RS06785 (Fig. 1C, E). OmpA is a core outer membrane protein implicated in pathogenesis through maintaining membrane integrity, mediating adhesion and invasion, and modulating host autophagy and immune responses30. The T6SS is a contractile nanomachine pivotal in intermicrobial competition31. Within its architecture, VgrG2 forms the tip complex, Hcp constitutes the inner tube, and TssC is a structural component of the baseplate; notably, VgrG2 and Hcp are secreted effectors, while TssC is not. There are three vgrG paralogues encoded in the AB5075 genome. Transcriptomic analysis revealed that, among these, only the expression of vgrG2 was upregulated, whereas vgrG3 exhibited reduced transcript levels, and vgrG1 showed no significant change in expression (Fig. 1E). A gene encoding a putative Rhs protein, often secreted in a VgrG2-dependent manner, was located adjacent to vgrG2. The gene V4Q77_RS06785 was annotated as encoding a cold shock protein, a conserved family involved in stress adaptation. We confirmed the in vivo up-regulation of the secreted effectors vgrG2 and hcp gene by qRT-PCR (Fig. 1F). While the ability of certain T6SS effectors to target eukaryotic host cells has been demonstrated, the broader functional repertoire and mechanistic details of such targeting remain underexplored32. We constructed in-frame deletion mutants (ΔvgrG2, Δhcp, and Δrhs) to investigate their specific contributions to AB5075 virulence during infection. To assess the role of the identified T6SS effectors in virulence, we first utilized a Galleria mellonella infection model. Compared to the AB5075 wild-type (WT) strain, the Δrhs mutants exhibited no significant attenuation in virulence. In stark contrast, the Δhcp and ΔvgrG2 mutant was virtually avirulent, failing to cause any larval mortality (Fig. 1G). It is well established that VgrG2 requires Hcp for extracellular secretion. We therefore reason that the pronounced attenuation of the Δhcp mutant, similar to that of the ΔvgrG2 mutant, is likely attributable to the loss of Hcp-dependent secretion of VgrG2, which abrogates its delivery to the extracellular environment. To provide further independent evidence that VgrG2 secretion is strictly dependent on a functional Type VI secretion system, we additionally constructed a deletion mutant of tssM, which encodes an essential structural component of the T6SS apparatus. In both the Δhcp and ΔtssM strains, VgrG2 was detected exclusively in the cellular fraction and was completely absent from the supernatant (Fig. 1H). Collectively, these observations support the conclusion that VgrG2 plays a more prominent role in bacterial virulence than previously appreciated. We next validated these findings in a murine acute pneumonia model. Mirroring the results in G. mellonella, infection with the AB5075 ΔvgrG2 and Δhcp mutants resulted in no mortality, whereas the Δrhs mutant showed virulence comparable to the WT (Fig. 1I). Gross pathological examination of lungs harvested from mice infected with AB5075 WT revealed significant congestion and edema, changes that were markedly milder in ΔvgrG2-infected mice (Fig. 1J). Histopathological analysis (H&E staining) confirmed substantial neutrophil infiltration in both groups; however, alveolar architecture was more severely disrupted in WT-infected lungs compared to those infected with the ΔvgrG2 mutant (Fig. 1K). Quantification of bacterial loads showed that lung tissues from WT-infected mice harbored significantly higher counts than those infected with the ΔvgrG2 mutant (Fig. 1L). Furthermore, while ΔvgrG2 infection remained largely confined to the lungs, the WT strain disseminated systemically, breaching the alveolar-capillary barrier and establishing significantly higher bacterial burdens in distal organs, including the liver, kidney, spleen, and bloodstream. To further dissect the cell-autonomous impact, we employed a human induced pluripotent stem cell (iPSC)-derived lung organoid model that recapitulates the multicellular complexity of the native lung, encompassing epithelial, mesenchymal, and immune cell lineages33. After 24 h of co-culture, organoids infected with the WT strain suffered complete destruction of alveolar structures. In contrast, organoids infected with the ΔvgrG2 mutant largely maintained their structural integrity (Fig. 1M), indicating that VgrG2 plays a crucial role in damaging human lung tissue. To confirm that the observed avirulence was specifically due to the loss of vgrG2, we constructed two complemented strains (ΔvgrG2::vgrG2 and ΔvgrG2::Flag-vgrG2). In the mouse pneumonia model, ΔvgrG2::vgrG2 fully restored virulence to WT levels (Fig. 1N). Furthermore, using the Flag-tagged complementation strain, we confirmed that VgrG2 was detectable within the lungs of infected mice (Fig. 1O). To evaluate the conservation of this phenotype across strains, we constructed in-frame deletion mutants of vgrG2 and hcp in an additional A. baumannii strains, ATCC 17978. Interestingly, contrary to our observations in AB5075, neither the ΔvgrG2 nor the Δhcp mutant of ATCC 17978 displayed reduced virulence in the G. mellonella model (Supplementary Fig. S1G). This strain-specific difference may be linked to the reported repression of T6SS expression in some strains by TetR-family regulators encoded on plasmid pAB334. To further investigate the impact of VgrG2 on the virulence of the ATCC 17978 strain, we eliminated the endogenous plasmid pAB3. First, the Acinetobacter baumannii ATCC 17978 wild-type strain and the isogenic ΔvgrG2 mutant were serially passaged for 25 generations in antibiotic-free LB medium. Following plating on non-selective agar, individual colonies were isolated and screened by PCR to identify derivatives that had spontaneously lost the pAB3 plasmid, hereafter designated as pAB3(–) strains. Subsequent virulence assessment was performed using a mouse infection model with an inoculum of 5 × 10⁷ colony-forming units (CFU). Notably, the ATCC 17978 WT pAB3(–) strain exhibited a significantly enhanced virulence phenotype compared to the parental ATCC 17978 WT pAB3(+) strain (Supplementary Fig. S1H). In stark contrast, the ΔvgrG2 mutant remained avirulent regardless of pAB3 status, as no increase in virulence was observed even after the loss of the pAB3 plasmid. These findings demonstrate that under conditions in which T6SS activity is not suppressed by pAB3, VgrG2 exerts a marked and indispensable influence on bacterial virulence. In parallel, we also present the growth data for all strains examined (Supplementary Fig. S1I). These results confirm that the observed differences in virulence among the strains were not attributable to disparities in their growth. In summary, our data from multiple infection models establish that the vgrG2 gene, encoding a secreted structural component of the T6SS, is a critical determinant of A. baumannii virulence, whereas rhs appear dispensable under these conditions. The ΔvgrG2 mutant recruited more neutrophils and led to higher levels of inflammation To delineate how vgrG2 influences A. baumannii pathogenesis at the host-pathogen interface, we performed single-cell RNA sequencing (scRNA-seq) on lung tissues harvested from mice infected with PBS (control), AB5075 WT, or the ΔvgrG2 mutant (n = 3 per group) (Fig. 2A). This approach aimed to identify transcriptomic differences in the lung microenvironment associated with lethal infection versus host survival. Graph-based clustering of the integrated scRNA-seq dataset yielded a transcriptomic atlas comprising 10 distinct cell clusters (Fig. 2B, Supplementary Fig. S2A). Quantification of cell cluster abundances revealed a pronounced shift in lung cellular composition upon infection (Fig. 2C). While non-immune cells constituted approximately half of all cells in PBS-treated lungs, immune cells dominated the landscape in infected lungs. This shift was most extreme in ΔvgrG2-infected lungs, which exhibited the highest relative abundance of immune cells. Notably, neutrophils demonstrated the most substantial expansion, with their proportion being significantly higher in the ΔvgrG2 group compared to the WT group (Supplementary Fig. S2B). In contrast, the macrophage proportion remained relatively stable across all groups and was even slightly lower in the ΔvgrG2 group, possibly due to the overwhelming expansion of neutrophils. The ΔvgrG2 mutant enhanced neutrophil recruitment and elevated inflammatory responses. A scRNA-seq experimental design (lungs from PBS, WT, or ΔvgrG2; n = 3 per group), Created in BioRender. wenzeng, X. (2026) https://BioRender.com/tyyg3yu. UMAP of cell types (B) and their proportions (C). D Flow cytometry quantification of Ly6G⁺ neutrophils among CD45⁺ immune cells (biologically independent mice: PBS, n = 3; AB5075 WT, n = 5; ΔvgrG2 n = 5). E Survival of neutropenic and control mice after intratracheal challenge with a sublethal inoculum of 4 × 10⁶ CFU A. baumannii. F Pro-inflammatory scores from scRNA-seq (n = 3 biologically independent mice per group). CellChat intercellular communication networks: global ligand–receptor interactions (G), CXCL pathway (H) and TGFβ pathway (I) under WT or ΔvgrG2 infection. Red/blue lines indicate enhanced/weakened interactions; edge width reflects communication probability. J Relative mRNA expression of Il1b, Il6, Ccl3, and Cxcl2 in lung tissues at 24 hpi, quantified by qRT-PCR (biologically independent mice: PBS, n = 3; AB5075 WT, n = 6; ΔvgrG2, n = 6). K ELISA of TNF-α, IL-6, CCL3 and CXCL2 in lung homogenates (biologically independent mice: PBS, n = 3; AB5075 WT, n = 5; ΔvgrG2, n = 5). The results are expressed as mean ± SD. p values were determined by one-way ANOVA followed by two-sided Tukey’s post hoc test (D), the Kruskal–Wallis test with Dunn’s post-hoc correction (D), two-sided unpaired Student’s t test (J), Logrank (Mantel–Cox) test (E), the two-sided Wilcoxon rank sum test with continuity correction (F), and the two-sided Mann–Whitney U test (K). Data are representative of at least three independent experiments with similar results. Source data are provided as a Source Data file. We validated these compositional changes by flow cytometry. Consistent with the scRNA-seq data, neutrophil counts in the lungs were significantly higher in ΔvgrG2-infected mice than in WT-infected mice at 8, 16, and 24 h post-infection (hpi) (Fig. 2D). Macrophage proportions remained low at these time points, except for a significant increase observed in the WT group at 24 hpi (Supplementary Fig. S2C). These results suggest that neutrophils are the predominant responders during early acute infection and that the magnitude of neutrophil recruitment may be a critical determinant of disease outcome. To further clarify whether other immune cell populations play substantial roles in the host response to acute A. baumannii infection, we subsequently examined additional cell types by flow cytometry. Consistent with our prior single-cell RNA-seq analysis (Fig. 2C), dendritic cell numbers did not differ significantly between the lungs of WT- and ΔvgrG2-infected mice (Supplementary Fig. S2D). In contrast, the numbers of T cells and B cells were markedly higher in the WT group. Despite their elevated proportions in the WT group, T and B cells conferred limited protection against A. baumannii infection; mortality was significantly higher in WT-infected mice than in ΔvgrG2-infected animals. These findings further underscore the critical importance of neutrophils in the host defense against A. baumannii infection. We hypothesized that VgrG2 might promote virulence by dampening this neutrophilic response. The role of neutrophils in the host response to A. baumannii infection has been documented in multiple studies. In an immunocompetent mouse model of A. baumannii pneumonia, neutrophils are robustly recruited to the lungs within 6 h of infection, which constitutes a hallmark of effective bacterial clearance; moreover, distinct neutrophil subsets are mobilized to the site of infection35. When the neutrophil response is impaired (e.g., in immunodeficient mice), exacerbated lung injury and elevated early mortality ensue. Neutrophils eliminate bacteria directly through phagocytosis, and IL-17-deficient mice display enhanced neutrophil phagocytic activity, thereby exhibiting greater resistance to infection36. Furthermore, neutrophils exert antimicrobial functions through the production of reactive oxygen species (ROS) and IL-1β37. To functionally test the protective role of neutrophils, we depleted neutrophils in mice using an anti-Ly6G antibody. Subsequently, the mice were intratracheally challenged with a sublethal inoculum of 4 × 10⁶ CFU of A. baumannii. Consistent with the above studies, in mice rendered neutropenic by anti-Ly6G antibody pretreatment, even this relatively low bacterial burden resulted in a markedly elevated mortality rate (Fig. 2E). To investigate the mechanistic basis for the heightened neutrophil recruitment in ΔvgrG2-infected lungs, we conducted further transcriptomic analyses. At the transcriptional level, key pro-inflammatory cytokines and chemokines—including IL-1α, IL-1β, TNF, CCL3, CCL4, CCL6, and CXCL2—were significantly more upregulated in the ΔvgrG2 group compared to the WT group (Supplementary Fig. S2E). Consistently, gene set enrichment analysis confirmed a significantly higher pro-inflammatory signature score in the ΔvgrG2 group (Fig. 2F), corroborating a state of enhanced pulmonary inflammation. To map the intercellular signaling networks underpinning this inflammatory landscape, we performed a systematic analysis using CellChat38. Comparison of the global communication atlas between ΔvgrG2- and WT-infected lungs revealed an overall increase in the number of inferred interactions but a decrease in their interaction strength across most cell types (Fig. 2G). Notably, neutrophil autocrine signaling constituted an exception, exhibiting increases in both the number and strength of interactions. Quantification of pathway-specific signaling flux delineated the global immunocommunication profile. Pathways associated with pro-inflammatory responses, such as IFN-II, CXCL, and IL1, were markedly enhanced or activated. Concurrently, the anti-inflammatory TGFβ pathway was also upregulated. We next dissected the cellular contributors to these altered pathways. In the CXCL and IL1 pathways, the total interaction strength was increased in ΔvgrG2-infected lungs, with a particularly pronounced enhancement in neutrophil autocrine CXCL signaling (Fig. 2H, Supplementary Fig. S2F). Within this autocrine loop, CXCL2 signaling showed the most significant increase. Conversely, the overall interaction strength in the TNF and CCL pathways was reduced in the ΔvgrG2 group, yet neutrophil autocrine signaling within these pathways was intensified (Supplementary Fig. S2G, H), with CCL3 signaling being the most upregulated in the CCL autocrine circuit. These findings suggest a model wherein neutrophils in ΔvgrG2-infected lungs amplify their own recruitment through autocrine secretion of chemokines such as CXCL2 and CCL3. Simultaneously, the significant upregulation of the TGFβ pathway (Fig. 2I) indicates the activation of a counter-regulatory mechanism that may help constrain inflammation within a non-destructive range and prevent a pathological cytokine storm. Subsequently, we selected representative genes to verify the above transcriptional conclusions. Finally, we validated key transcriptional findings at the protein and RNA levels. Using qRT-PCR and ELISA, we confirmed that lung tissues from ΔvgrG2-infected mice exhibited higher levels of specific inflammatory mediators (Fig. 2J, K, Supplementary Fig. S2I, J). Enhanced neutrophil recruitment by the ΔvgrG2 mutant is associated with TLR4 activation We next sought to elucidate the mechanism underlying the heightened inflammation and increased neutrophil recruitment observed in ΔvgrG2-infected mice. Innate immune detection of infection is initiated by pattern recognition receptors (PRRs), which sense evolutionarily conserved microbial components known as pathogen-associated molecular patterns (PAMPs)39. A major family of PRRs is the Toll-like receptors (TLRs), which recognize a diverse array of PAMPs derived from bacteria, viruses, fungi, and parasites. Analysis of our single-cell transcriptome dataset revealed that both the proportion of Tlr4-positive cells and the average Tlr4 expression level in lung tissue were significantly higher in ΔvgrG2-infected mice compared to WT-infected and PBS control groups (Fig. 3A). This transcriptional finding was validated at the protein level by flow cytometry, which confirmed an increased number of TLR4-expressing cells in the lungs of ΔvgrG2-infected mice. In contrast, WT infection led to a decrease in TLR4⁺ cells relative to the control group (Fig. 3B). In the lung, TLR4 is expressed by both innate immune cells (e.g., alveolar macrophages, neutrophils) and structural cells (e.g., alveolar epithelial cells). We hypothesize that the reduction of TLR4⁺ cells in WT-infected lungs may reflect the exhaustion or depletion of this cellular pool—particularly neutrophils—during the active clearance of a high-burden, virulent infection. Conversely, the attenuated ΔvgrG2 mutant may fail to trigger this exhaustive consumption, resulting in the observed accumulation of TLR4-expressing cells, which in turn sustains a potent pro-inflammatory and neutrophilic response. To further determine whether TLR4 expression in the mouse lung is predominantly attributable to neutrophils, we performed flow cytometric analysis. Among all TLR4⁺ cells, the increases in both the number and frequency of neutrophils were most pronounced in the lungs of ΔvgrG2-infected mice (Supplementary Fig. S3A). We next probed the downstream signaling cascades of TLR4. Transcriptomic analysis indicated that both the MyD88-dependent and TRIF-dependent signaling pathways were significantly upregulated in the lungs of ΔvgrG2-infected mice compared to their WT-infected counterparts (Supplementary Fig. S3B, C). The concerted activation of these two major TLR4 signaling branches orchestrated the high expression of key downstream effector molecules, including pro-inflammatory cytokines (e.g., IL-1β, IL-6) and chemokines (e.g., CCL3, CCL5, CXCL2). This cascade ultimately culminated in the pronounced pulmonary inflammation observed in ΔvgrG2-infected mice. TLR4 activation links enhanced neutrophil recruitment to ΔvgrG2 mutant infection. A Dot plot showing the relative expression of Toll-like receptors in lung tissues from mice infected with PBS, AB5075 WT, or the ΔvgrG2 mutant (derived from scRNA-seq data). B Flow cytometry analysis of TLR4⁺ cell abundance in the lung tissues at 12 hpi and 24 hpi (biologically independent mice: PBS, n = 3; AB5075 WT, n = 5; ΔvgrG2 n = 5). C UMAP plot of neutrophil subsets from scRNA-seq data, with colors representing distinct transcriptional clusters. D Comparative analysis of pro-inflammatory scores among neutrophil subsets, stratified by infection group (PBS, AB5075 WT, ΔvgrG2), based on lung scRNA-seq data (n = 3 biologically independent mice per group). E Schematic of the in vivo infection model utilizing Tlr4−/− and wild-type (WT) mice challenged with either AB5075 WT or the ΔvgrG2 mutant, Created in BioRender. wenzeng, X. (2026) https://BioRender.com/sq07m1u. F Representative hematoxylin and eosin-stained lung sections at 12 hpi. G The percentage of Ly6G⁺ neutrophils and F4/80⁺ macrophages among CD45⁺ lung immune cells in Tlr4−/− mice across infection groups (PBS, AB5075 WT, ΔvgrG2) at 12 hpi, as quantified by flow cytometry (n = 5 biologically independent mice per group). H Protein levels of TNF-α, IL-6, CCL3, and CXCL2 in lung homogenates from Tlr4−/− mice at 12 hpi, measured by ELISA (biologically independent mice: PBS, n = 3; AB5075 WT, n = 5; ΔvgrG2 n = 5). I Relative mRNA expression of Il1b, Il6, Ccl3, and Cxcl2 in lung tissues from Tlr4−/− mice at 12 hpi, quantified by qRT-PCR (n = 5 biologically independent mice per group). J Bacterial burden in lung tissues of wild-type and Tlr4−/− mice at 12 hpi (n = 5 biologically independent mice per group). The results are expressed as mean ± SD. p values were determined by one-way ANOVA followed by two-sided Tukey’s post hoc test (B), the two-sided Wilcoxon rank sum test with continuity correction (D), two-sided non-parametric Mann–Whitney U test (G and H), and two-sided unpaired Student’s t test (G, J, I). Data are representative of at least three independent experiments with similar results. Source data are provided as a Source Data file. Given the central role of neutrophils in the observed hyperinflammation, we performed a comprehensive, cell-type-specific investigation. Neutrophils from our integrated dataset were isolated and subjected to sub-clustering (Fig. 3C). Notably, neutrophils from ΔvgrG2-infected lungs exhibited significantly higher Tlr4 expression levels and a greater overall inflammatory activity score compared to those from WT-infected lungs (Supplementary Fig. S3D, Fig. 3D). To functionally characterize this state, we performed Gene Ontology (GO) enrichment analysis on genes upregulated in ΔvgrG2 neutrophils versus WT. The analysis revealed a significant enrichment of pathways directly related to neutrophil recruitment, chemotaxis, and inflammatory response (Supplementary Fig. S3E), aligning with the observed histological and flow cytometry phenotypes. We next sought to understand the dynamic transition of neutrophils upon infection. Using pseudotemporal trajectory inference, we constructed a differentiation continuum, setting neutrophils from PBS-treated mice as the starting point. This analysis successfully captured a transcriptional trajectory from a baseline state toward an infection-activated state (Supplementary Fig. S3F). Examination of genes dynamically regulated along this pseudotime axis identified key effector molecules. Strikingly, the genes encoding the potent chemokines Ccl3 and Cxcl2 were among the top 20 most significantly upregulated genes as pseudotime progressed, underscoring their pivotal role in driving the self-amplifying inflammatory loop (Supplementary Fig. S3G). To directly establish TLR4’s functional role, we infected Tlr4−/− and wild-type (WT) C57BL/6 mice with AB5075 WT or ΔvgrG2 (Fig. 3E). Histopathological analysis at 12 hpi revealed markedly reduced neutrophil infiltration in Tlr4−/− mice, irrespective of the infecting strain, compared to WT mice (Fig. 3F). Flow cytometric quantification of lung leukocytes at 12 hpi showed low and comparable numbers of neutrophils and macrophages in Tlr4−/− mice infected with either strain (Fig. 3G). By 24 hpi, however, neutrophil numbers increased significantly in ΔvgrG2-infected Tlr4−/− mice, exceeding those in WT-infected Tlr4−/− mice, while macrophage numbers rose in the latter group (Supplementary Fig. S3H). Consistently, qRT-PCR and ELISA confirmed that inflammatory mediator levels in Tlr4−/− mice were comparable between infection groups at 12 hpi (Fig. 3H, I). At this early time point, the capacity of Tlr4−/− mice to clear or restrict bacterial growth was significantly impaired compared to wild-type mice, irrespective of infection with the AB5075 WT or ΔvgrG2 strain (Fig. 3J). In summary, our data reveal a mechanism whereby the attenuated ΔvgrG2 mutant sustains TLR4-MyD88/TRIF signaling in lung cells, leading to a chemokine-driven (CCL3/CXCL2) autocrine loop in neutrophils that amplifies inflammation. Crucially, in vivo studies using Tlr4−/− mice establish TLR4 as essential for the early (12 hpi) neutrophilic response and bacterial control against both bacterial strains. However, a delayed neutrophil influx in ΔvgrG2-infected Tlr4−/− mice by 24 hpi suggests the eventual engagement of compensatory, TLR4-independent pathways in this attenuated infection model. vgrG2 assists A. baumannii in immune escape by inhibiting the expression of Csu pilus To gain insight into the macrophage compartment in the context of the differential host response, we performed sub-clustering on macrophages from our single-cell dataset. This resolved two distinct populations: alveolar macrophages (AMs) and interstitial macrophages (IMs) (Fig. 4A). While AMs reside on the alveolar surface, IMs are located in the lung interstitium and perivascular spaces. In uninfected mice, AMs were the predominant population. However, upon infection with either WT or ΔvgrG2, IMs became the dominant subset. Notably, the proportion of IMs was significantly higher in WT-infected lungs compared to ΔvgrG2-infected lungs (Fig. 4B), reflecting a more severe depletion of AMs in the high-burden WT infection. Furthermore, IMs exhibited a significantly higher inflammatory activity score than AMs across infected samples (Fig. 4C). Comparative analysis revealed that, unlike neutrophils, macrophages from WT-infected lungs displayed a higher inflammatory signature than those from ΔvgrG2-infected lungs (Supplementary Fig. S4A). This may be attributed to the overwhelming neutrophilic response that predominates early in ΔvgrG2 infection. Intriguingly, macrophages from ΔvgrG2-infected mice still maintained significantly higher Tlr4 expression compared to both the WT-infected and control groups (Supplementary Fig. S4A), suggesting a retained sensing capacity. GO enrichment analysis of genes upregulated in ΔvgrG2 macrophages versus WT macrophages showed significant enrichment for pathways related to phagocytosis (Supplementary Fig. S4B). This transcriptional profile implied that the WT strain might evade macrophage phagocytosis more effectively than the ΔvgrG2 mutant. To test this hypothesis, we conducted an in vitro phagocytic assay using Raw264.7 macrophage-like cells (Fig. 4D). After two hours of co-culture, AB5075 WT bacteria were virtually non-phagocytosed, whereas the ΔvgrG2 mutant was readily internalized (Fig. 4E). This stark contrast in phagocytic susceptibility was independently confirmed using confocal microscopy after only 60 min of co-culture (Supplementary Fig. S4D). Together, these data indicate that the VgrG2-dependent phenotype is critical for A. baumannii to evade macrophage-mediated clearance. VgrG2 suppresses Csu pilus expression to evade immunity. A, B Macrophage subsets from lung scRNA-seq: UMAP (A) and proportions (B). C Pro-inflammatory scores: alveolar vs. interstitial macrophages (n = 3 biologically independent mice per group). D, E Phagocytosis assay in Raw264.7 cells – schematic (D) (Created in BioRender. wenzeng, X. (2026) https://BioRender.com/bwqxqy5) and intracellular CFU (E) (n = 3 biologically independent bacterial cultures per group). F RNA-seq volcano plot of ΔvgrG2 vs. WT: upregulation of Csu genes (n = 3). csuAB, p = 0.0297; csuA, p = 0.0110; csuB, p = 0.0075; csuD, p = 0.0267. G qRT-PCR validation of Csu operon in vitro (n = 3 biologically independent bacterial cultures per group). H qRT-PCR validation in vivo (n = 5 biologically independent bacterial cultures per group). I Flow cytometry of Ly6G⁺ neutrophils and F4/80⁺ macrophages (biologically independent mice: PBS, n = 3; ΔvgrG2, n = 5; ΔvgrG2ΔCsuE, n = 5). J Relative mRNA expression of Il1b, Il6, Ccl3, and Cxcl2 in lung tissues (n = 3 biologically independent mice per group). K ELISA of TNF-α, IL-6, CCL3, CXCL2 in lung homogenates (biologically independent mice: PBS, n = 3; ΔvgrG2, n = 5; ΔvgrG2ΔCsuE, n = 5). L Bacterial loads in lungs at 12 h (n = 5 biologically independent mice per group). M Mouse survival (n = 10 per group). N qRT-PCR of bfmR in WT vs. ΔvgrG2 (n = 3 biologically independent bacterial cultures per group). O Phos-tag immunoblot of total BfmR and p-BfmR (n = 3 biologically independent bacterial cultures per group). P qRT-PCR of Csu genes with/without BfmR inhibitor Cmp-98 (n = 3 biologically independent bacterial cultures per group). Q Biofilm formation with/without Cmp-98 (n = 9 biologically independent bacterial cultures per group). R Co-IP of VgrG2-BfmR interaction. The results are expressed as mean ± SD. p values were determined by the two-sided Wilcoxon rank sum test with continuity correction (C), two-sided edgeR quasi-likelihood F-tests with Benjamini–Hochberg adjustment for multiple comparisons (F), two-sided unpaired Student’s t test (G, H, J, K, L, N, O, P, and Q), one-way ANOVA followed by two-sided Tukey’s post hoc test (I), and Logrank (Mantel–Cox) test (M). Data are representative of at least three independent experiments with similar results. Source data are provided as a Source Data file. Given the distinct immune evasion phenotype of ΔvgrG2, we investigated whether VgrG2 intrinsically alters bacterial gene expression. We performed bulk RNA-seq on AB5075 WT and the ΔvgrG2 mutant cultured in vitro. Comparative transcriptomic profiling revealed a limited set of upregulated genes in ΔvgrG2, among which four were notably associated with the biosynthesis of Csu pili, a major adhesin and biofilm component in A. baumannii (Fig. 4F). We validated this finding by quantifying the transcription of all genes within the Csu pilus operon. Consistent with the RNA-seq data, Csu gene expression was significantly elevated in the ΔvgrG2 mutant under in vitro conditions (Fig. 4G). Intriguingly, analysis of bacteria isolated from infected lungs (in vivo) revealed a marked overall suppression of Csu pilus gene expression compared to in vitro levels (Fig. 4H). This downregulation in vivo likely represents a bacterial strategy to minimize surface exposure of adhesive structures, thereby reducing capture by patrolling immune cells. Strikingly, even within this suppressed in vivo context, the Csu genes remained significantly more highly expressed in the ΔvgrG2 mutant than in the WT strain (Fig. 4H). These data collectively indicate that the vgrG2 gene functions as an inhibitor of Csu pilus expression, both in vitro and in vivo. Consequently, the dysregulation of Csu pili in the mutant translated into measurable phenotypic changes. The ΔvgrG2 strain exhibited a significantly enhanced adhesive capacity toward lung epithelial cells (Supplementary Fig. S4E) and a robust increase in biofilm formation compared to the WT strain (Supplementary Fig. S4F). To directly test the functional link between VgrG2 and Csu pili, we constructed a ΔvgrG2 ΔcsuE double mutant and a ΔcsuE mutant. CsuE is a key structural component essential for pilus assembly and adhesion26. Phenotypic characterization confirmed that the double mutant exhibited a significant reduction in biofilm formation compared to the ΔvgrG2 single mutant (Supplementary Fig. S4G). In phagocytosis assays, the ΔvgrG2 ΔcsuE strain showed an intermediate phenotype: its resistance to macrophage uptake was stronger than that of ΔvgrG2 but did not fully revert to the high evasion level of the WT strain (Supplementary Fig. S4H). Scanning electron microscopy showed that surface pili were not obviously detected in both the WT and the ΔvgrG2 ΔcsuE strains, whereas clearly visible pili, especially between adjacent bacterial cells, were present in the ΔvgrG2 mutant (Supplementary Fig. S4I). We next assessed the contribution of Csu pili to the hyperinflammatory phenotype in vivo. In the murine lung infection model, the ΔvgrG2 ΔcsuE strain at 12 hpi induced a significantly attenuated neutrophilic recruitment compared to the ΔvgrG2 strain (Fig. 4I). Concomitantly, the proportion of TLR4⁺ cells in the lungs of mice infected with the double mutant was also significantly lower (Supplementary Fig. S4J). Consistently, qRT-PCR and ELISA analyses confirmed that the pulmonary inflammatory response was substantially reduced in ΔvgrG2 ΔcsuE-infected mice relative to those infected with ΔvgrG2 (Fig. 4J, K). Meanwhile, in the early stage of infection, the ΔvgrG2 ΔcsuE double-knockout strain exhibited a stronger ability to evade clearance by the host immune system compared with the ΔvgrG2 strain (Fig. 4L). Collectively, these results demonstrate that the elevated expression of Csu pili in the ΔvgrG2 mutant is a major driver of the early hyperinflammatory response, likely by enhancing bacterial visibility to innate immune surveillance via TLR4. This supports a model where, during early infection, WT A. baumannii utilizes VgrG2 to repress Csu pilus expression as an immune evasion strategy, reducing recognition by macrophages and neutrophils. However, the ΔvgrG2 ΔcsuE double mutant did not restore virulence (Fig. 4M), indicating that the avirulence of ΔvgrG2 is multifactorial and that VgrG2 contributes to pathogenesis through additional, Csu-independent mechanisms. Numerous previous reports have established that the Type VI secretion system (T6SS) and its secreted effectors can indeed exert regulatory effects on bacterial physiology, including biofilm formation and pilus expression. For instance, in the avian pathogenic Escherichia coli strain CE129, deletion of the T6SS core gene hcp2 results in a marked reduction in biofilm formation capacity and decreased expression of type 1 pilus genes (fimA, fimC, papC)40. Similarly, T6SS mutants of Rhizobium display defects in biofilm development, compromising their competitive colonization of legume rhizospheres41. In the rice pathogen Burkholderia glumae BGR1, deletion of the T6SS group_1 component tssD1 does not alter pathogenicity but impairs biofilm-associated competitive functions42. A study of 248 clinical Klebsiella pneumoniae isolates further revealed that T6SS-positive strains exhibit significantly higher biofilm biomass compared to T6SS-negative counterparts (0.3 ± 0.09 vs. 0.16 ± 0.06, P < 0.01)43. Additionally, deletion of T6SS2 (hcp2) in Vibrio parahaemolyticus leads to downregulation of flagellar filament and assembly proteins, thereby affecting motility and adhesion44. Next, we extended our investigation to elucidate the mechanism by which VgrG2 regulates the expression of Csu pili. Previous work in Acinetobacter baumannii has demonstrated that phosphorylated BfmR functions as a transcriptional activator that directly governs the expression of 303 genes and, importantly, acts as a key repressor of genes involved in pilus biogenesis, thereby constituting a critical mechanism for the suppression of pilus expression45. Examination of our transcriptomic data revealed that bfmR transcript levels were diminished in the ΔvgrG2 mutant compared to the WT strain. We hypothesized that VgrG2 might influence Csu pilus expression through modulation of the BfmR pathway. To verify this, we first confirmed by RT-qPCR that bfmR expression is indeed reduced in the ΔvgrG2 strain relative to WT (Fig. 4N). Subsequently, we generated an anti-BfmR antibody and employed Phos-tag SDS-PAGE to assess the abundance of both total BfmR and phosphorylated BfmR (p-BfmR). Consistent with our hypothesis, both BfmR and p-BfmR protein levels were substantially higher in the WT strain than in the ΔvgrG2 mutant (Fig. 4O). To further establish the functional relevance of BfmR in this context, we employed Cmp-98, a small-molecule BfmR inhibitor previously identified in our laboratory that acts by binding to the BfmR phosphorylation site46. Pharmacological inhibition of BfmR with Cmp-98 (16 μg/mL) in the WT strain resulted in a pronounced increase in the expression of Csu pilus-related genes and a concomitant enhancement of biofilm formation capacity (Fig. 4P, Q). A comparable effect was observed upon Cmp-98 treatment of the ΔvgrG2 strain. These results collectively indicate that BfmR serves as a critical intermediary in the VgrG2-mediated regulation of Csu pilus expression. To investigate whether VgrG2 directly interacts with BfmR, we performed exogenous co-immunoprecipitation assays. VgrG2 was expressed in 293T cells, and the resulting lysates were incubated with lysates prepared from the wild-type bacterial strain to test for a physical association with BfmR. No detectable direct interaction was observed (Fig. 4R). These results indicate that VgrG2 does not directly bind to BfmR. Consequently, the modulation of Csu pilus expression by VgrG2 cannot be attributed to a direct physical effect on BfmR. While our data show that VgrG2 influences the BfmR pathway, the molecular mechanism by which this occurs remains an open question; any link between VgrG2 and BfmR is indirect, and the nature of this indirect regulation has yet to be determined. VgrG2 triggers methuosis in phagocytes Based on the aforementioned findings, we hypothesized that the attenuated virulence of the ΔvgrG2 strain is attributable to the function of the VgrG2 effector protein itself. To test this, we purified recombinant VgrG2 protein, with its endotoxin content confirmed to be below 0.5 EU/mg, thereby excluding any confounding effects from LPS in subsequent assays. Treatment of Raw264.7 cells (mouse monocyte macrophage leukemia cell line) with purified VgrG2 revealed a potent cytotoxic effect in a dose- and time-dependent manner (Fig. 5A). Flow cytometric analysis corroborated this finding, showing significant cell death at concentrations ≥10 ng/mL after 24 h, with no concomitant increase in apoptosis (Supplementary Fig. S5A). In stark contrast, VgrG2 exhibited no significant cytotoxicity against lung epithelial cells (HBE and A549), even at a high concentration of 100 ng/mL (Fig. 5B). To extend these observations to primary immune cells, we treated mouse bone marrow-derived macrophages (BMDMs) and primary neutrophils with VgrG2. A significant reduction in cell viability was observed in both BMDMs (after 24 h) and neutrophils (after 8 h) at a concentration of 10 ng/mL (Fig. 5C). Consistently, in a bacterial co-culture system, wild-type AB5075 displayed significantly greater killing efficacy against Raw264.7 macrophages than against HBE epithelial cells over 12 h. The ΔvgrG2 mutant, in contrast, was largely devoid of this cytotoxic capacity against both cell types (Supplementary Fig. S5B, C). To definitively rule out LPS interference, Raw264.7 cells were treated for 24 h with buffer supplemented with an amount of LPS equivalent to that present in the VgrG2 preparation. This control treatment did not significantly affect cell viability compared with buffer alone (Supplementary Fig. S5D). Collectively, these in vitro data establish that the VgrG2 protein acts as a direct cytotoxin with selective potency against phagocytic cells (macrophages and neutrophils) but not lung epithelial cells. VgrG2 induces methuosis in phagocytes. A VgrG2 reduces Raw264.7 viability (CCK8, n = 6 biologically independent cell cultures per group). B No cytotoxicity in lung epithelial cells (CCK8, HBE and A549, n = 6 biologically independent cell cultures per group)). C Cytotoxicity in primary immune cells (CCK8, n = 6 biologically independent cell cultures per group). D Flow cytometry of Ly6G⁺ neutrophils and F4/80⁺ macrophages in lungs (biologically independent mice: Buffer A, n = 3; ΔvgrG2 + VgrG2, n = 5; ΔvgrG2 + Buffer A, n = 5). E Mouse survival with VgrG2 supplementation (n = 10 per group). F Vacuolization in Raw264.7 (bright-field microscopy). Red arrows indicate vacuoles. G Vacuoles accumulate Lucifer Yellow, not colocalize with ER/Mito/Lyso (live-cell imaging). ER endoplasmic reticulum, Mito mitochondria, Lyso lysosomes. H Transmission electron microscopy reveals single-membrane vacuoles coalesce and rupture (methuosis). The yellow and red arrows indicated the endoplasmic reticulum and mitochondria, respectively. I Vacuolization unaffected by apoptosis/autophagy/necroptosis inhibitors. J LC3B immunoblot in Raw264.7 with WT vs. ΔvgrG2 (n = 3 biologically independent cell cultures per group). K LC3B immunoblot with purified VgrG2 (n = 3 biologically independent cell cultures per group). The results are expressed as mean ± SD. p values were determined by two-sided unpaired Student’s t test (A, B, C, J, and K), one-way ANOVA followed by two-sided Tukey’s post hoc test (D), and Logrank (Mantel–Cox) test (E). Data are representative of at least three independent experiments with similar results. Source data are provided as a Source Data file. To validate the pathophysiological relevance of VgrG2’s cytotoxicity in vivo, we employed a virulence factor intervention model. Mice were first infected with the ΔvgrG2 strain for 12 h to establish an infection and recruit neutrophils to the lungs. Subsequently, these mice received an intranasal instillation of purified VgrG2 protein (1 μg/kg). Flow cytometric analysis of lung homogenates 8 h post-treatment revealed a significant decrease in the neutrophil population (Fig. 5D), directly demonstrating that VgrG2 retains potent neutrophil-killing activity within the complex host environment. Concurrently, the macrophage proportion showed a modest increase, which may reflect a relative change due to the acute depletion of neutrophils and the consumption of VgrG2 during its early cytotoxic activity. Critically, supplementing the ΔvgrG2 infection with a higher dose of VgrG2 protein (4 μg/kg) led to a significant recovery of virulence, as evidenced by increased death (Fig. 5E). Hematoxylin-eosin staining revealed lung tissue damage in mice treated with VgrG2 protein alone, marked by alveolar destruction, neutrophil infiltration, and congestion (Supplementary Fig. S5E). In mice infected with the ΔvgrG2 strain and subsequently treated with either VgrG2 protein or buffer, lung tissues exhibited extensive neutrophil accumulation and alveolar damage. However, the severity of these pathologies was not distinguishable by Hematoxylin-eosin staining. Collectively, this dose-dependent complementation confirms that the attenuated phenotype of the ΔvgrG2 mutant is directly attributable to the lack of the VgrG2 effector protein. Next, we investigated the cytopathic mechanism underlying VgrG2-induced phagocyte death. Microscopic examination of Raw264.7 cells treated with VgrG2 for 12 h revealed the formation of conspicuous cytoplasmic vacuoles. This vacuolization exhibited a clear dose-dependency: at 10 ng/mL, cells contained numerous small vacuoles (Fig. 5F), whereas at 100 ng/mL, these smaller vacuoles coalesced into a large vacuole with a diameter of ~5 μm. An identical phenotype, characterized by the formation of giant vacuoles, was recapitulated in primary BMDMs treated with VgrG2 (Supplementary Fig. S5F). Importantly, significant cytoplasmic vacuolization was also observed in Raw264.7 cells following 24 h of co-culture with wild-type AB5075 bacteria (Supplementary Fig. S5G). In stark contrast, co-culture with the isogenic ΔvgrG2 mutant failed to induce any detectable vacuolization under identical conditions. To elucidate the origin of the VgrG2-induced cytoplasmic vacuoles, we performed live-cell fluorescence imaging of Raw264.7 cells labeled with organelle-specific markers for the endoplasmic reticulum, mitochondria, and lysosomes. The vacuoles showed no colocalization with any of these intracellular compartments (Fig. 5G), indicating that the vacuoles were not derived from lysosomes, mitochondria or endoplasmic reticulum. Given that prominent vacuolation is a hallmark of macropinocytosis—a fluid-phase uptake mechanism that generates large vacuolar compartments bearing molecular signatures of late endosomes—we sought to determine whether the VgrG2-elicited vacuoles originated from this pathway47. To this end, we employed the fluid-phase marker Lucifer Yellow in conjunction with immunofluorescence staining for the late endosomal GTPase Rab7. Encouragingly, the fluid-phase endocytosis tracer Lucifer Yellow and the late endosome marker Rab7 were found to accumulate specifically within these vacuoles (Fig. 5G and Supplementary Fig. S5H), indicating that they might originate from macropinocytosis. This finding led us to investigate whether VgrG2 triggers methuosis, a distinctive, non-apoptotic form of cell death driven by the accumulation of macropinocytic vacuoles (which can exceed 5 μm in diameter) that fail to fuse with lysosomes47,48. Ultrastructural analysis by transmission electron microscopy provided definitive confirmation. VgrG2-treated Raw264.7 cells accumulated numerous single-membrane vacuoles that progressively coalesced into large vacuoles, culminating in vacuole rupture and consequent cell lysis (Fig. 5H). Concurrently, both the mitochondrial network and the endoplasmic reticulum preserved their structural integrity. Furthermore, the nuclear envelope exhibited no signs of disruption, and the chromatin within the nucleus remained in a dispersed, uncondensed state. TEM analysis of Raw264.7 cells co-cultured with bacterial strains for 24 h yielded results consistent with the aforementioned observations. Numerous vacuoles were evident in cells co-cultured with the WT strain, whereas no obvious vacuole formation was detected in cells co-cultured with the ΔvgrG2 mutant strain (Supplementary Fig. S5I). Given that methuosis is characterized as a cell death modality distinct from apoptosis, autophagy, and necrosis47, we proceeded to examine whether pharmacological inhibitors targeting these alternative pathways could suppress the vacuolation elicited by VgrG2. As hypothesized, the vacuolization induced by VgrG2 was not suppressed by inhibitors of apoptosis (ZVAD-FMK), autophagy (3-methyladenine), or necroptosis (necrostatin-1) (Fig. 5I). Correspondingly, VgrG2-mediated cytotoxicity was unaffected by ZVAD-FMK or 3- methyladenine but was partially attenuated by necrostatin-1 (Supplementary Fig. S5J), suggesting a potential contributory role for necroptotic signaling alongside the primary death mechanism. Prior reports have established that the type VI secretion system effector VgrG2 is capable of triggering autophagy in host cells, as demonstrated in several bacterial species such as Vibrio parahaemolyticus and Escherichia coli49,50. To further elucidate whether the vacuolation induced by VgrG2 is associated with autophagy, Raw264.7 cells were co-cultured with the wild-type Acinetobacter baumannii strain AB5075 or the isogenic ΔvgrG2 mutant for 12 h, after which the expression levels of LC3B, a key autophagic marker, were assessed. No significant differences in the abundance of either LC3B-I or LC3B-II were observed between the two groups (Fig. 5J). Consistent with this finding, treatment of Raw264.7 cells with purified VgrG2 protein for 24 h also failed to alter LC3B conversion (Fig. 5K). Moreover, the autophagy inhibitor 3-methyladenine neither abrogated VgrG2-mediated vacuolation nor attenuated its cytotoxic effects (Fig. 5I and Supplementary Fig. S5J). Collectively, these results indicate that, unlike its homologs in Vibrio parahaemolyticus and Escherichia coli, VgrG2 in A. baumannii does not possess the capacity to induce autophagy. To further characterize the mode of cell death induced by VgrG2, we examined markers of necroptosis and apoptosis, namely phosphorylated MLKL (p-MLKL) and poly (ADP-ribose) polymerase (PARP) cleavage, respectively. Treatment with purified VgrG2 protein did not result in detectable levels of p-MLKL (Supplementary Fig. S5K), nor did it induce PARP cleavage (Supplementary Fig. S5L). Although the necroptosis inhibitor necrostatin-1 conferred a modest alleviation of VgrG2-mediated cytotoxicity in Raw264.7 cells, it failed to suppress the accompanying cytoplasmic vacuolization (Fig. 5I and Supplementary Fig. S5J). Similarly, the pan-caspase inhibitor Z-VAD-FMK neither mitigated VgrG2-induced cytotoxicity nor attenuated vacuole formation (Fig. 5I and Supplementary Fig. S5J). Taken together, these findings indicate that the macrophage cell death triggered by VgrG2 proceeds independently of both necroptosis and apoptosis. Taken together, these findings indicate that methuosis is involved in the cell death elicited by VgrG2. VgrG2 elicits phagocyte death by targeting and activating Rac1 To elucidate the mechanisms by which VgrG2 triggers methuosis and impairs phagocytosis, we investigated its impact on the actin cytoskeleton—a master regulator of both macropinocytosis and phagocytic cup formation. Building on our initial phagocytosis assays, confocal microscopy of Raw264.7 cells after 2 h of bacterial co-culture revealed distinct morphological changes (Fig. 6A). Cells in the control group displayed actin filaments primarily organized perinuclearly. In contrast, cells co-cultured with the ΔvgrG2 mutant exhibited prominent membrane protrusions, forming filopodia enriched with long, bundled actin filaments. Importantly, internalized bacteria were visible within these actin-rich structures, indicating active phagocytosis. Strikingly, cells stimulated with the wild-type AB5075 strain showed a markedly different phenotype. Their filopodia contained actin filaments aggregated into small, discrete clusters, and no internalized bacteria were observed. This stark contrast led us to hypothesize that VgrG2 secreted by the WT strain disrupts normal actin remodeling, thereby compromising the cell’s phagocytic capacity. We next asked whether purified VgrG2 protein alone could recapitulate this aberrant cytoskeletal remodeling. Treatment of Raw264.7 cells with VgrG2 for 12 hours induced severe aggregation of the actin cytoskeleton. Actin filaments coalesced into pronounced aggregates, particularly surrounding characteristic cytoplasmic vacuoles (Fig. 6B). We postulate that this cytoskeletal rearrangement drives excessive macropinocytosis, thereby impairing phagocytic function in macrophages and ultimately leading to methuosis. Subsequently, fluorescence labeling showed that VgrG2 is internalized by Raw264.7 cells within 1 h of treatment (Supplementary Fig. S6A), indicating that Raw264.7 cells actively internalize VgrG2. However, the specific internalization pathway by which VgrG2 enters Raw264.7 cells—as well as the molecular basis for its selective activity toward macrophages and neutrophils—remains to be elucidated. To address this, we next investigated the uptake mechanism of VgrG2 by Raw264.7 cells. The cells were pretreated with inhibitors targeting distinct endocytic and phagocytic pathways, including chlorpromazine (an inhibitor of clathrin-mediated endocytosis, 25 μM), methyl-β-cyclodextrin (an inhibitor of caveolin-mediated endocytosis, 5 mM), 5-(N-ethyl-N-isopropyl)-amiloride (an inhibitor of micropinocytosis, 20 μM), and cytochalasin D (an inhibitor of phagocytosis, 5 μM). Additionally, the Rac1 inhibitor NSC23766 (100 μM) was employed, as Rac1 is a critical regulator of phagocytic uptake in macrophages. Following pretreatment, purified VgrG2 protein was added to the culture medium. After one hour of incubation, VgrG2 internalization was assessed by confocal microscopy, and after 24 h, the formation of intracellular vacuoles was examined by light microscopy. The results revealed that VgrG2 uptake and subsequent vacuolation were not abrogated by chlorpromazine, methyl-β-cyclodextrin, or 5-(N-ethyl-N-isopropyl)-amiloride. In stark contrast, both processes were markedly suppressed by cytochalasin D and NSC23766 (Fig. 6C, D). These findings indicate that VgrG2 enters Raw264.7 cells primarily via phagocytosis, rather than through clathrin- or caveolin-mediated endocytosis or macropinocytosis. This phagocytosis-dependent internalization mechanism provides a plausible explanation for the observed specificity of VgrG2 toward professional phagocytes, such as macrophages and neutrophils, and its lack of activity toward non-phagocytic epithelial cells. To definitively link this phenotype to intracellular VgrG2 activity, we constructed a recombinant adenovirus for VgrG2 overexpression in Raw264.7 cells. Successful transfection, confirmed at 72 h, resulted in complete remodeling of the actin network into dense aggregates (Fig. 6E). This genetic approach confirmed that intracellular VgrG2 expression is sufficient to induce profound actin cytoskeletal aggregation. VgrG2 impairs phagocytosis and induces phagocyte death via Rac1 targeting and activation. A Confocal microscopy (2 h co-culture): WT AB5075 shows clustered actin and no internalized bacteria; ΔvgrG2 enables active phagocytosis within filopodia. B Purified VgrG2 induces actin aggregation around vacuoles (12 h). C Confocal immunofluorescence image of Raw264.7 cells pretreated with chlorpromazine, methyl-β-cyclodextrin, 5-(N-ethyl-N-isopropyl)-amiloride, cytochalasin D and NSC23766. Then, these cells were treated with purified VgrG2 protein for 1 h. D Light microscopy image Raw264.7 cells pretreated with chlorpromazine, methyl-β-cyclodextrin, 5-(N-ethyl-N-isopropyl)-amiloride, cytochalasin D and NSC23766. Then, these cells were treated with purified VgrG2 protein for 24 h. E Adenoviral VgrG2 overexpression (72 h) induces actin aggregation (GFP⁺ cells). F Western blot: VgrG2 increases Rac1-GTP levels at 30 min (n = 3 biologically independent cell cultures per group). G NSC23766 (Rac1 inhibitor) abolishes VgrG2-induced vacuolization. H T NSC23766 rescues VgrG2 cytotoxicity in Raw264.7, BMDMs, and partially in primary neutrophils (CCK-8 assay, n = 6 biologically independent cell cultures per group)). I Co-immunoprecipitation confirms VgrG2-Rac1 interaction. The results are expressed as mean ± SD. p values were determined by two-sided unpaired Student’s t test (F, H). Data are representative of at least three independent experiments with similar results. Source data are provided as a Source Data file. This observation prompted us to investigate whether VgrG2 could exert a direct effect on actin dynamics. While previous studies have established that the C-terminal actin-crosslinking domain (ACD) of the Vibrio cholerae effector VgrG1 directly cross-links monomeric G-actin—leading to F-actin depolymerization, cytoskeletal collapse, and cell rounding51,52—a sequence analysis revealed that VgrG2 from A. baumannii AB5075 lacks this canonical ACD and is therefore unlikely to cross-link G-actin. We then considered upstream regulators. The rearrangement of the actin cytoskeleton and the formation of filopodia are precisely coordinated by the Rho family of small GTPases, which act as molecular switches cycling between active GTP-bound and inactive GDP-bound states53. Key members include Rac1, Cdc42, and RhoA. To identify the relevant GTPase, we assessed their activation status in Raw264.7 cells after 30 minutes of bacterial co-culture. Strikingly, Rac1 activation by the WT strain was significantly stronger than that by the ΔvgrG2 mutant strain (Supplementary Fig. S6B). In contrast, Cdc42 was not activated by either strain (Supplementary Fig. S6C), and while RhoA was activated, its activity level did not differ between the WT and mutant infections (Supplementary Fig. S6D). Consistent with this result, treatment of Raw264.7 cells with purified VgrG2 protein for 30 min significantly increased the level of active Rac1-GTP compared to the control (Fig. 6F). To establish functional necessity, we used the Rac1 activation inhibitor NSC 23766. This inhibitor completely abolished VgrG2-induced cytoplasmic vacuolization (Fig. 6G) and fully rescued macrophages (Raw264.7 and BMDMs) from VgrG2 cytotoxicity, with a partial rescue observed in neutrophils (Fig. 6H). Finally, co-immunoprecipitation assays confirmed a direct physical interaction between VgrG2 and endogenous Rac1 in adenovirus-transfected Raw264.7 cells (Fig. 6I). In summary, the Rac1 axis is thus established as the central signaling node through which VgrG2 executes its dual attack on phagocyte function and survival. Discussion This study implicates the T6SS effector VgrG2 as an important contributor to A. baumannii pathogenesis, acting through two potentially interconnected mechanisms: suppression of the Csu pilus assembly system at the transcriptional level and induction of methuotic cell death in phagocytes. These observations suggest a plausible link between VgrG2-mediated gene regulation, phagocyte dysfunction, and modulation of the inflammatory milieu during pulmonary infection. Within the Type VI Secretion System, Hcp and VgrG serve as core structural proteins that cooperatively assemble to form essential components of the secretion machinery54–57. Hcp proteins polymerize into the inner tube of the T6SS, which consists of stacked hexameric rings and functions as a conduit for effector protein translocation. VgrG constitutes the central element of the spike complex, forming a trimeric structure positioned at the distal tip of the Hcp tube to mediate penetration of target cell membranes. Collectively, the Hcp tube provides the structural foundation, while VgrG acts as the piercing tip, thereby forming a syringe-like nanomachine. Interestingly, deletion of the vgrG2 gene significantly attenuates A. baumannii virulence, whereas deletion of hcp does not produce a comparable effect. This disparity suggests that VgrG2 secretion may occur independently of a fully assembled T6SS apparatus. Notably, in the ATCC 17978 strain—where T6SS expression is regulated by TetR-family transcription factors encoded on plasmid pAB3—deletion of vgrG2 does not impact virulence34. This observation further indicates that VgrG2 secretion can be modulated by these plasmid-encoded TetR-like regulators. The functional relationship between VgrG2 and Hcp, the requirement for complete T6SS assembly in VgrG2 secretion, and the precise regulatory factors governing its export remain important subjects for future investigation. The discovery that VgrG2 represses Csu pilus expression expands the functional repertoire of T6SS effectors beyond classical toxin delivery58. While T6SSs are best known for bactericidal activity56,59, their role in modulating bacterial surface architecture to avoid immune recognition is less explored. The Csu pilus is a major adhesin and biofilm component26,60; its downregulation during infection likely minimizes opsonization and phagocytic uptake61. Notably, even in the suppressed in vivo context, Csu expression remained elevated in the ΔvgrG2 mutant, correlating with enhanced neutrophil recruitment and TLR4 activation. This suggests that fine-tuned pilus expression is critical for immune stealth, and that VgrG2-mediated repression represents a refined evasion strategy. The intermediate phenotype of the ΔvgrG2 ΔcsuE double mutant—partial restoration of phagocytosis resistance without full virulence recovery—indicates that VgrG2 operates through additional pathways beyond pilus regulation. The second major finding is that VgrG2 triggers methuosis, a non-apoptotic cell death pathway characterized by dysregulated macropinocytosis and cytoplasmic vacuolization62. Although methuosis has been previously documented in cancer cells, its induction by bacterial effectors in phagocytes remains scarcely reported48,63,64. VgrG2-induced vacuoles originate from macropinosomes that fail to mature or fuse with lysosomes, leading to osmotic imbalance and cell lysis. The specificity for phagocytes (macrophages and neutrophils) over epithelial cells underscores an evolved strategy to disable key immune sentinels. We further demonstrate that VgrG2 enters host cells via phagocytosis, providing a mechanistic explanation for its selective activity toward neutrophils and macrophages and its lack of effect on non-phagocytic epithelial cells. This cytotoxicity is independent of canonical death pathways and is instead driven by Rac1 hyperactivation. Rac1, a master regulator of actin dynamics and membrane trafficking65, emerges as the critical host target of VgrG2. By directly binding and activating Rac1, VgrG2 affects normal actin remodeling, impairing phagocytic cup formation while stimulating excessive macropinocytosis66. This dual action—blocking bacterial uptake while driving phagocyte self-destruction—represents a sophisticated offensive maneuver. The conservation of VgrG2-dependent virulence in AB5075, but not in ATCC 17978, aligns with known strain-specific T6SS regulation and underscores the importance of genetic context in effector function. Our study also highlights the role of TLR4 in orchestrating early neutrophil responses. The attenuated ΔvgrG2 mutant sustained higher TLR4 expression and signaling, leading to a chemokine-amplified neutrophil autocrine loop. Previous report indicates that cells can release the alarmin S100A8/A9, which stimulates neutrophil recruitment via TLR4 binding—an effect particularly pronounced in models of lung inflammation. TLR4 deficiency impairs neutrophil migration, leading to their accumulation within vasculature rather than alveolar infiltration, thereby compromising early inflammatory control67. Consistent with this, a recent study established a chronic A. baumannii infection model in Tlr4-knockout mice, demonstrating that TLR4 loss impairs bacterial clearance68. Efficient neutrophil recruitment is essential for pathogen elimination, especially in pulmonary infections where timely influx is critical for bacterial eradication69. In contrast, wild-type infection appears to deplete or downregulate TLR4-expressing cells—particularly macrophages and neutrophils—through VgrG2-mediated cytotoxicity, ultimately attenuating neutrophil recruitment and impairing bacterial clearance. This suggests that VgrG2 not only kills phagocytes but also modulates the inflammatory landscape to facilitate bacterial persistence. Our data demonstrate that VgrG2 from A. baumannii does not induce autophagy in host cells, in contrast to the autophagy-triggering capacity reported for VgrG2 homologs in Vibrio parahaemolyticus and Escherichia coli49,50. This functional divergence is not unexpected, as the T6SS effector VgrG can exhibit species-specific roles. Prior studies have established that VgrG in A. baumannii contributes to virulence, antimicrobial resistance, and host cell adherence, but its involvement in autophagy appears negligible70,71. It is noteworthy that, while A. baumannii as a species can activate autophagy through multiple pathways—including those mediated by the outer membrane protein OmpA72, the virulence factor Oma8773, and an isochorismatase-dependent, ubiquitin-mediated mechanism74—VgrG2 has not been observed to influence this process. In fact, the autophagy induced by A. baumannii is often characterized by incomplete autophagic flux, as exemplified by STX17 degradation leading to autolysosome dysfunction and by bacterial evasion of lysosomal degradation75. In our experiments, neither infection with the ΔvgrG2 mutant nor treatment with purified VgrG2 protein altered LC3B conversion, and the autophagy inhibitor 3-methyladenine failed to block VgrG2-mediated vacuolation or cytotoxicity. These findings strongly suggest that A. baumannii VgrG2 lacks the structural or functional motifs required for autophagy induction, unlike its counterparts in other bacteria. The precise molecular basis for this difference—such as variations in C-terminal effector domains or differential interactions with the host autophagy machinery—remains to be elucidated. Collectively, our results highlight the importance of species-specific functional characterization of T6SS effectors and caution against the extrapolation of autophagy-related roles across bacterial genera. While this study elucidates key mechanisms of VgrG2-mediated immune evasion, several limitations warrant consideration. First, the primary infection models employed—murine acute pneumonia and Galleria mellonella—though informative, may not fully recapitulate the complexity of human infections, particularly chronic or polymicrobial settings. Second, the phagocyte death and cytoskeletal remodeling assays were predominantly conducted using immortalized cell lines (e.g., Raw264.7) and primary cells ex vivo; these conditions might not entirely mirror the dynamic, multicellular microenvironment of an infected lung. Finally, while we identify Rac1 as a critical host target, the detailed biochemical mechanism of VgrG2-mediated Rac1 activation and its potential interplay with other host signaling pathways during infection merit deeper investigation. In conclusion, our work establishes the T6SS effector VgrG2 as a master regulator of A. baumannii virulence, orchestrating a two-pronged strategy to subvert host immunity (Fig. 7). When A. baumannii invades lung tissue, the bacterium upregulates expression of the vgrG2 gene in response to stress conditions imposed by the lung tissue environment. First, VgrG2 transcriptionally represses the Csu pilus, reducing bacterial visibility to innate immune surveillance and diminishing phagocytic recognition. Second, VgrG2 directly targets and activates the small GTPase Rac1 in phagocytes. This triggers profound actin cytoskeletal aggregation, which induces a dysregulated macropinocytic flux, culminating in impaired phagocytic function and methuosis—a non-apoptotic, vacuolization-dependent cell death. Together, these mechanisms enable A. baumannii to selectively deplete key immune sentinels (macrophages and neutrophils) while modulating the inflammatory landscape, thereby facilitating bacterial persistence and dissemination. Our findings not only expand the functional repertoire of T6SS effectors beyond interbacterial competition to include sophisticated eukaryotic-directed pathogenesis but also pinpoint VgrG2 and its host partner Rac1 as potential therapeutic targets for combating infections caused by this formidable multidrug-resistant pathogen. Proposed model of VgrG2-mediated immune evasion in A. baumannii. In wild-type bacteria, the effector VgrG2 exerts a dual function: it represses Csu pilus production and activates host Rac1 signaling, collectively dysregulating phagocyte function and dampening inflammatory responses. In the absence of VgrG2 (ΔvgrG2), derepressed Csu pili promote bacterial adhesion and TLR4-driven production of pro-inflammatory chemokines (CXCL2, CCL3), leading to enhanced neutrophil recruitment and reduced bacterial survival. Created in BioRender. wenzeng, X. (2026) https://BioRender.com/o34huti. WT wide type, BfmR Biofilm formation regulator R, TLR4 Toll-like receptor 4, CXCL2 C-X-C motif chemokine ligand 2, CCL3 C-C motif chemokine ligand 3, Rac1 Ras-related C3 botulinum toxin substrate 1. Methods Bacteria strains and growth conditions This study was approved by the Tab of Animal Experimental Ethical Inspection of the Sir Run Run Shaw Hospital, Zhejiang University School of Medicine (No. SRRSH2025-0039). The primary experiments in this study were performed using the well-characterized, carbapenem-resistant, and hypervirulent Acinetobacter baumannii strain AB5075 (AB5075-UW) and its mutants. The complete genome sequence of AB5075 is available under NCBI GenBank assembly accession GCF_036601155.1 (corresponding to GenBank accession CP008706), which served as the reference for genetic manipulations28. The strain was originally obtained from the BEI Resources repository (NR-17763) and maintained in our laboratory according to established protocols. In addition to the hypervirulent strain AB5075 and ATCC 17978 were also included in this study as reference strains; both were obtained from the American Type Culture Collection (ATCC). A comprehensive list of all bacterial strains employed is provided in Supplementary Table 1. Routine cultivation was performed in Mueller–Hinton broth (Oxoid) at 37 °C with vigorous shaking (200 rpm) (Supplementary Data 3). For long-term preservation, bacterial cultures were mixed with glycerol to a final concentration of 25% (v/v) and stored at –80 °C. Gene knockout was conducted using a homologous recombination-based approach as previously described with modifications76. In brief, upstream and downstream homology fragments flanking the target gene were amplified by PCR (see Supplementary Data 2 for primer sequences). The purified PCR products were then cloned into the BamHI/XbaI (Thermo Scientific) -digested pMo130-Hygr vector using a 2× Hieff Clone MultiS Enzyme Premix (Yeason), generating a knockout construct. This recombinant plasmid was subsequently introduced into A. baumannii strain AB5075 via electro-transformation. Transformants were selected on Mueller–Hinton agar (Oxoid) plates supplemented with 500 µg/mL hygromycin (MeilunBio). For the reference strains ATCC 17978, the hygromycin concentration was reduced to 100 µg/mL. To promote the second crossover event and subsequent allelic replacement, a selected transformant harboring the integrated plasmid was cultured in MH broth containing 10% sucrose. Putative knockout clones were first screened by spraying with 0.45 M pyrocatechol; colonies that turned white were considered candidate mutants. These candidates were further verified for hygromycin susceptibility. Finally, the successful deletion of the target gene was confirmed by diagnostic PCR followed by Sanger sequencing. For in situ genetic complementation, a similar procedure was followed with a key modification: the DNA fragment inserted into the pMo130-Hygr vector included not only the upstream and downstream homology arms of the vgrG2 locus but also the complete open reading frame of the vgrG2 gene itself. The resulting complementation plasmid was then electroporated into the ΔvgrG2 mutant strain of AB5075 to generate the complemented strain. Cells and organoids Murine leukemic monocyte macrophage cell line (Raw264.7), human lung carcinoma epithelial cell line (A549), and human bronchial epithelial cell line (HBE) were sourced from the Committee on Type Culture Collection of the Chinese Academy of Sciences (Shanghai, China). All cell lines were routinely maintained in Dulbecco’s Modified Eagle Medium (DMEM; Gibco) supplemented with 10% fetal bovine serum (FBS; VivaCell) at 37 °C in a humidified atmosphere containing 5% CO₂. Cells were routinely screened and confirmed to be free of mycoplasma contamination. Human iPSC-derived lung organoids were generated by Seven Plus Limited Company (Tianjin, China), following a previously established protocol33. Briefly, undifferentiated human iPSC colonies at ~75–85% confluence were detached by incubation with 1 mg/mL dispase solution for 5–10 min at 37 °C. The cells were then gently washed with DMEM/F12, and the colonies were scraped off in mTeSR™1 medium. Subsequently, the colonies were mechanically dissociated into smaller clumps using a serological pipette. Approximately 0.5 mL of these iPSC clumps were seeded into each well of a 24-well plate pre-coated with Matrigel. Upon reaching 50–75% confluence, the cells were subjected to a stepwise differentiation protocol to induce definitive endoderm, anterior foregut endoderm, and ultimately, lung organoids. Primary BMDMs were isolated from the femurs and tibiae of 6–8 week-old C57BL/6 mice77. Briefly, bone marrow was flushed from dissected bones using complete DMEM medium supplemented with 10% FBS, 1% penicillin-streptomycin, and 20 ng/mL recombinant murine M-CSF (MCE). The cell suspension was passed through a 70-μm strainer, pelleted by centrifugation (300 × g, 5 min), and resuspended in the same medium. Cells were seeded in non-tissue-culture-treated plates and incubated at 37 °C under 5% CO₂. On day 3, half of the medium was replaced with fresh M-CSF-containing medium. By day 6–7, adherent cells had differentiated into mature BMDMs, which were then gently detached using ice-cold PBS and seeded for subsequent experiments. Neutrophils were freshly isolated from mouse bone marrow using a density gradient method. Marrow cells were collected as described above and resuspended in ice-cold Hanks’ Balanced Salt Solution (HBSS). The cell suspension was layered onto a three-step Percoll (Absin) gradient (78%, 69%, and 52% Percoll in PBS) and centrifuged at 1500 × g for 30 min at 4 °C with minimal acceleration and no brake. Neutrophils were collected from the interface between the 69% and 78% layers, washed twice in HBSS, and treated with ACK lysis buffer (Solarbio) to remove residual erythrocytes. Neutrophil purity (>85%) was confirmed by flow cytometry based on CD11b⁺ Ly-6G⁺ expression, and viability exceeded 95% as assessed by trypan blue exclusion. Cells were kept on ice and used immediately for functional assays. Mice Female C57BL/6 and Tlr4−/− mice were purchased from Shanghai Model Organisms Center (Shanghai, China). All mice were housed under specific-pathogen-free (SPF) conditions with ad libitum access to food and water. The animal facility maintained a controlled environment at 20–22 °C, 50–60% relative humidity, and a standard 12-hour light/dark cycle. Experiments utilized mice at 8–10 weeks of age. The genotype of Tlr4−/− mice was identified by polymerase chain reaction (PCR). All animal experiments were approved by the Tab of Animal Experimental Ethical Inspection of the Sir Run Run Shaw Hospital, Zhejiang University School of Medicine (No. SRRSH2025-0039). To establish an acute pulmonary infection model, mice were lightly anesthetized and inoculated intranasally with A. baumannii strain AB5075. A dose-response study was first performed using a gradient of bacterial doses (4 × 10⁶, 4 × 10⁷, 4 × 10⁸, and 4 × 10⁹ CFU) to determine the optimal challenge dose. Based on survival, bacterial burden, and histopathological outcomes, an inoculum of 4 × 10⁷ CFU was selected for subsequent experiments. For most studies, mice were infected with either the AB5075 WT strain, the isogenic ΔvgrG2 mutant, or complemented strains as indicated. In specific experiments, mice were pretreated intranasally with a low dose of LPS (MCE; 1 μg) 24 h prior to bacterial challenge to induce pulmonary neutrophil recruitment. For virulence factor intervention studies, mice first infected with the ΔvgrG2 mutant for 12 h received a subsequent intranasal instillation of purified VgrG2 protein (1 or 4 μg/kg). At designated time points post-infection, mice were euthanized for sample collection. Lungs were harvested for bacterial burden quantification (homogenized and plated for CFU counts), histopathological analysis (H&E staining), flow cytometry, RNA extraction, scRNA-seq, and cytokine measurement (ELISA). In some experiments, distal organs (liver, spleen, kidney) and blood were also collected to assess systemic dissemination. Neutrophil depletion For neutrophil depletion studies, 8- to 12-week-old C57BL/6J mice were administered 200 µg of anti-Ly6G monoclonal antibody (clone 1A8, Bio X Cell, Cat# BE0075-1) or rat IgG2a isotype control via intraperitoneal injection 24 h prior to experimental manipulation. Depletion was maintained with subsequent doses of 100 µg every 48 h as needed. Neutrophil clearance was verified by flow cytometric analysis of peripheral blood, defining neutrophils as CD45⁺CD11b⁺Ly6G+ cells, with >90% depletion efficiency confirmed at 24 h post-injection. Galleria mellonella The virulence of bacterial strains was evaluated using the Galleria mellonella infection model, performed in accordance with established protocols78. Healthy final-instar larvae, each weighing between 200 and 300 mg, were selected for the assays. Exponential growth phase bacteria cultures (37 °C, 200 rpm, 4 h) were harvested by centrifugation, washed twice, and resuspended in sterile phosphate-buffered saline (PBS). Larvae were surface-disinfected with 70% ethanol and allowed to air dry. Using a sterile 1 mL syringe fitted with a 26-gauge needle, a 10 µL aliquot of the respective bacterial suspension was injected into the larval hemocoel via the last left proleg. Control groups were injected with an equal volume of sterile PBS. Following injection, groups of 10 larvae were placed in sterile Petri dishes and incubated at 37 °C in the dark. Serial dilutions were prepared to achieve the intended inoculum in a final volume of 10 µL for larval injection. The following challenge doses were administered: AB5075 at 1 × 10⁴ CFU per larva and ATCC 17,978 at 1 × 10⁵ CFU per larva per larva. Survival was recorded at 8-h intervals for a total of 120 h post-infection. Larvae were scored as dead when they showed no tactile response and displayed evident melanization. Isolation of bacteria from murine lung tissue Fresh lung tissues were rinsed twice on ice with 2 mL pre-chilled PBS supplemented with RNase inhibitor (RI; MCE) at a concentration of 40 U/mL, then minced into small pieces using sterile surgical scissors. The minced tissues were homogenized under the following conditions: 50 Hz, 30 s on/3 s off, 8 cycles, 4 °C. Immediately afterward, 1–5 mL of washing buffer (10 mM Tris-HCl, pH 7.5; 146 mM NaCl; 1 mM CaCl₂; 21 mM MgCl₂; 60 U/mL RI; 0.01% BSA) containing 0.05% Triton X-100 was added for lysis for 5–10 min. The reaction was terminated promptly by adding 3 volumes of the washing buffer. The resulting suspension was filtered through a 40 μm filter membrane, and the filtrate was centrifuged at 500 × g for 5 min at 4 °C. The supernatant was transferred to a new centrifuge tube and subjected to horizontal centrifugation at 4000 × g for 10 min at 4 °C to pellet the tissue-associated bacteria. The bacterial pellet was resuspended in PBS containing RI, followed by 2–3 rounds of centrifugation at 800 × g for 5 min at 4 °C; the supernatant was retained and the pellet was discarded. Finally, the target bacteria were pelleted again via centrifugation at 4000 × g for 10 min at 4 °C. Bacterial transcriptomic sequencing and analysis For transcriptional profiling of A. baumannii under host-mimicking conditions, bacterial RNA was isolated from two sources: (1) bacteria recovered from infected murine lung tissue (in vivo group) and (2) bacteria grown in LB broth to mid-exponential phase (in vitro control). To control for potential transcriptional perturbations introduced during tissue processing, the in vitro–grown bacteria were subjected to the same extraction and washing procedure used for lung-derived bacteria prior to RNA isolation. Total RNA was extracted using Total RNA Kit I (Omega) according to the manufacturer’s instructions, with additional on-column DNase I (Takara) treatment to remove genomic DNA. RNA quality and concentration were assessed using a Bioanalyzer (Agilent Technologies). Bacterial mRNA sequencing libraries were constructed and sequenced on an Illumina NovaSeq 6000 platform by Zhejiang Tianke (Hangzhou, China). Raw sequencing reads were quality-trimmed and aligned to the reference genome of A. baumannii AB5075 (GenBank assembly accession GCF_036601155.1) using Rockhopper v2.0.379 (Supplementary Table 2). Gene-level counts were obtained and imported into edgeR for differential expression analysis80. Differentially expressed genes (DEGs) were defined using a threshold of Benjamini–Hochberg adjusted P < 0.05 and an absolute fold-change ≥2. Functional annotation of genes was performed via eggNOG-mapper (http://eggnog-mapper.embl.de), and GO and KEGG enrichment analysis was conducted using the ClusterProfiler R package81, with terms considered significantly enriched at adjusted P < 0.05. Single-cell RNA sequencing and analysis To profile the host transcriptional landscape during A. baumannii infection, scRNA-seq was performed on lung tissues from experimentally infected mice. Animals were divided into three groups (n = 3 per group) and treated as follows: (1) infected with wild-type AB5075, (2) infected with the isogenic ΔvgrG2 mutant, or (3) mock-infected with PBS. At 24 h post-infection, lungs were harvested, minced, and subjected to dissociation into single-cell suspensions using a gentle tissue dissociation kit (Miltenyi Biotec) according to the manufacturer’s instructions. Cell viability was assessed by trypan blue exclusion and exceeded 85% in all samples. Single-cell libraries were constructed using the Chromium Next GEM Single Cell 3ʹ Reagent Kits v3.1 (10× Genomics) and sequenced on an Illumina NovaSeq 6000 platform by Beijing Novogene Co., Ltd. Raw sequencing data were processed with Cell Ranger (v7.1.0) using the mouse reference genome (mm10) for alignment and gene counting. Downstream analysis was performed in R (v4.3.1) using the Seurat package (v5.2)82. Cells were filtered based on the following quality-control thresholds: number of detected genes (nFeature_RNA) between 200 and 6000, mitochondrial gene percentage <20%, and hemoglobin gene percentage <5%. Data were normalized using the SCTransform method, and integration of all samples was performed to correct for batch effects. Principal component analysis (PCA) was conducted on the top 3000 variable genes, followed by uniform manifold approximation and projection (UMAP) for two-dimensional visualization. Cell type annotation was performed manually based on the expression of canonical marker genes. Differential gene expression between experimental groups was identified DESeq2; genes with adjusted P < 0.05 and |log₂(fold change)| > 0.5 were considered significant. Functional enrichment analysis of GO terms and KEGG pathways was performed using the clusterProfiler package81. Custom gene sets relevant to inflammation and M2 polarization were compiled from KEGG and GO Term gene sets as described. The activity of these gene sets in individual cells was scored using the AddModuleScore function in Seurat. Cell–cell communication networks were inferred using CellChat (v2.2.0)83 with the default mouse ligand–receptor interaction database. Signaling pathways showing differential interaction strength or number of links between groups were identified and visualized. For trajectory analysis, neutrophil and macrophage subsets were subsetted from the integrated object and analyzed using monocle3 (v1.3.4)84. Cells were ordered along pseudotime after dimensionality reduction (UMAP) and graph-based trajectory construction, and genes significantly associated with pseudotime progression were identified using the graph_test function. Bacterial phagocytosis assay The phagocytic activity of macrophages against Acinetobacter baumannii was evaluated using a modified version of an established protocol85. Briefly, the murine macrophage line Raw264.7 was seeded into 24-well plates at 2 × 10⁵ cells per well and maintained overnight in complete DMEM. Prior to bacterial challenge, the medium was replaced with antibiotic-free DMEM. A. baumannii strains were grown to mid-log phase, washed, and resuspended in PBS, then added to macrophage monolayers at a multiplicity of infection (MOI) of 50:1. To synchronize infection, plates were centrifuged at 500 × g for 5 min and incubated for 2 h at 37 °C under 5% CO₂. Following incubation, extracellular bacteria were removed by washing three times with PBS. Fresh medium containing 10 µg/mL tigecycline (MCE) was then added for 1 h to eliminate any remaining extracellular bacteria. After antibiotic treatment, cells were washed again and lysed with 0.1% Triton X-100 (Yeason) in PBS for 15 min. Cell lysates were serially diluted in PBS and plated on Mueller–Hinton agar to quantify internalized CFU. Bacterial adherence assay Bacterial adhesion to lung epithelial cells was assessed using the human bronchial epithelial line HBE. Cells were seeded in 24-well plates at 1.5 × 10⁵ cells per well and cultured until ~90% confluence. Before the assay, monolayers were washed with PBS and incubated in antibiotic-free DMEM. Exponential-phase bacteria were harvested, washed, and resuspended in PBS, then added to epithelial cells at an MOI of 50:1. Plates were centrifuged (500 × g, 5 min) and incubated for 1.5 h at 37 °C in 5% CO₂. Non-adherent bacteria were removed by three washes with PBS. Epithelial cells were subsequently lysed using 0.1% Triton X-100 for 15 min, and lysates were diluted and plated on Mueller–Hinton agar for CFU enumeration. Quantitative assessment of biofilm formation To assess biofilm production by A. baumannii strains, a static microplate-based assay was performed following an established protocol with slight adaptations86. Briefly, a single bacterial colony was inoculated into Mueller–Hinton broth and grown overnight at 37 °C with shaking. The overnight culture was diluted 1:100 in fresh MH broth, and 200 µL aliquots were dispensed into at least 6 replicate wells of a sterile 96-well flat-bottom polystyrene plate. After incubation for 24 h at 37 °C under static conditions, planktonic cells were gently removed by washing each well three times with PBS. Adherent biofilms were fixed with absolute methanol for 15 min, air-dried, and stained with 0.1% (w/v) crystal violet for 30 min at room temperature. Unbound dye was removed by thorough washing with PBS. The crystal violet retained by the biofilm was solubilized with 200 µL of 33% (v/v) glacial acetic acid under gentle shaking (200 rpm, 30 min). Absorbance of the solubilized dye was measured at 595 nm using a Multiskan Go microplate reader. The assay was conducted in at least three independent biological replicates, and the mean absorbance values were used for comparative analysis of biofilm formation capacity across strains. RNA extraction and RT-qPCR Total RNA was extracted from two sample types: (1) A. baumannii collected from infected murine lung tissue or cultured in vitro, and (2) whole lung tissue from infected or control mice. For bacterial RNA, pellets were lysed in TRIzol Reagent (Invitrogen) and purified using the E.Z.N.A. Total RNA Kit I (Omega Bio-Tek) with on-column DNase I digestion. For lung tissue RNA, ~30 mg of snap-frozen tissue was homogenized in TRIzol using a mechanical homogenizer, followed by chloroform phase separation and purification with the same RNA kit. RNA concentration and purity were determined spectrophotometrically (NanoDrop 2000). cDNA was synthesized from 500 ng total RNA using the PrimeScript™ RT Reagent Kit (Takara). Quantitative PCR was performed on a LightCycler 480 II system (Roche) with TB Green Premix Ex Taq II (Takara). Reactions were run in triplicate under standard cycling conditions (95 °C for 30 s, 40 cycles of 95 °C for 5 s and 60 °C for 30 s). Melt-curve analysis confirmed amplification specificity. Gene expression was normalized to the reference gene rpoB (bacteria) or B2m (mouse), and relative quantification was calculated using the 2−ΔΔCt method. Statistical significance was assessed by Student’s t test or one-way ANOVA. Primer sequences are listed in Supplementary Data 2. Cytokine and chemokine quantification by ELISA Levels of TNF-α, IL-6, CCL3, and CXCL2 in lung homogenates were measured using commercial enzyme-linked immunosorbent assay (ELISA) kits (Shanghai Enzyme-Linked Biotechnology Co., Ltd.) according to the manufacturer’s instructions. The respective product numbers were as follows: TNF-α (ml002095), IL-6 (ml002293), CCL3 (ml037861), and CXCL2 (ml037864). Briefly, samples were collected and centrifuged at 3000 × g for 10 min at 4 °C to remove debris. Standards and samples were added to pre-coated wells and incubated for 90 min at 37 °C. After washing, biotinylated detection antibody was added and incubated for 60 min, followed by streptavidin-HRP incubation for 30 min. Tetramethylbenzidine (TMB) substrate was added for color development, which was stopped with sulfuric acid. Absorbance was measured at 450 nm using a microplate reader (BioTek Synergy H1). Concentrations were interpolated from a standard curve generated with recombinant protein standards included in each kit. All samples were assayed in duplicate, and results from at least three independent experiments were analyzed. Collection of lung single cell and flow cytometric analysis Single-cell suspensions were prepared from murine lung tissue for immunophenotyping by flow cytometry. Briefly, freshly harvested lungs were placed in ice-cold Hank’s Balanced Salt Solution (HBSS) supplemented with 1× HEPES. Tissues were finely minced using sterile surgical scissors and transferred into digestion buffer consisting of HBSS with 0.4 mg/mL Collagenase I (Thermo Fisher Scientific). Digestion was carried out at 37 °C for 1 h with gentle agitation. The digested tissue was then gently dissociated by pipetting and passed through a 70-μm cell strainer to obtain a single-cell suspension. Erythrocytes were lysed using ACK lysing buffer (Solarbio) according to the manufacturer’s instructions. Cells were washed twice with cold PBS containing 2% FBS. Single-cell suspensions were stained with fluorochrome-conjugated antibodies against surface markers. Cells were first incubated with LIVE/DEAD™ Fixable Near-IR Dead Cell Stain (Invitrogen) for 20 min at 4 °C in PBS to discriminate viable cells. After washing, Fc receptors were blocked with anti-mouse CD16/32 antibody (BioLegend) for 10 min at 4 °C. Cells were then stained with the following antibody cocktail prepared in staining buffer (PBS with 2% FBS) for 30 min at 4 °C in the dark: anti-CD45-PE (BioLegend), anti-CD11b-Brilliant Violet 421 (BioLegend), anti-Ly-6G-PerCP/Cyanine5.5 (BioLegend), anti-F4/80-Alexa Fluor(R) 647 (BioLegend), anti-PE/Cyanine7-F4/80 (BioLegend), anti-Brilliant Violet 510-CD3ε (BioLegend), anti-Brilliant Violet 605-CD19 (BioLegend), anti-FITC-CD11c (BioLegend), anti-Alexa Fluor 700-I-A/I-E (BioLegend), and anti-TLR4-APC (BioLegend). Corresponding fluorescence minus one (FMO) controls and isotype controls were included for accurate gating. Samples were acquired on a CytoFLEX LX flow cytometer (Beckman Coulter) using CytExpert software (v2.4). Data analysis was performed using FlowJo software (v10.9.0). Gating strategy proceeded as follows: single cells were selected based on forward and side scatter properties, followed by exclusion of dead cells. Leukocytes were identified as CD45⁺ events. Neutrophils were defined as CD11b⁺ Ly-6G⁺, dendritic cells as CD11c⁺ I-A/I-E⁺, T cells as CD3⁺, B cells as CD19⁺, and macrophages as CD11b⁺ F4/80⁺. TLR4 expression was assessed within these populations. Median fluorescence intensity (MFI) and cell frequency data were exported for statistical analysis. Apoptosis in Raw264.7 macrophages and HBE epithelial cells was assessed following bacterial co-culture or treatment with purified protein. For bacterial co-culture, cells were infected with A. baumannii strains at an MOI of 50:1 for 12 h. For protein treatment, cells were incubated with purified VgrG2 (10–100 ng/mL) or buffer for 24 h. After treatment, cells were detached, washed, and stained using an Annexin V-FITC/PI Apoptosis Detection Kit (Yeasen) according to the manufacturer’s protocol. Briefly, cells were resuspended in binding buffer and incubated with Annexin V-FITC and propidium iodide (PI) for 15 min at room temperature in the dark. For apoptosis assays, Annexin V⁺ PI⁻ cells were considered early apoptotic, and Annexin V⁺ PI⁺ cells were considered necrotic. Cell viability assay (CCK-8) Cell viability after treatment with purified VgrG2 protein was assessed using the Cell Counting Kit-8 (CCK-8; MCE). Raw264.7, A549, or HBE cells were seeded in 96-well plates and allowed to adhere overnight. Cells were then treated with serial dilutions of VgrG2 (1–100 ng/mL) in serum-free medium for 24 h. Subsequently, 10 μL of CCK-8 reagent was added to each well, and plates were incubated for 2 h at 37 °C. Absorbance was measured at 450 nm using a microplate reader. Viability was calculated relative to buffer-treated controls. Experiments were performed with six technical replicates and repeated three times independently. Live-cell fluorescence imaging To investigate the origin and nature of VgrG2-induced vacuoles, Raw264.7 macrophages were plated in 24-well plates at a density of 5 × 10⁴ cells per well. After overnight adherence, cells were exposed to purified VgrG2 protein (10 ng/mL) for 24 h. Subsequently, live cells were labeled for 30 min at 37 °C with the following fluorescent probes according to the suppliers’ protocols: ER-Tracker Red (Beyotime), Lyso-Tracker Red (Beyotime), Mito-Tracker Green (Beyotime), or the fluid-phase tracer Lucifer Yellow CH dilithium salt (MCE). For nuclei visualization, cells were subsequently incubated with 10 µg/mL Hoechst (MCE) for 10 min at room temperature in the dark, followed by gentle washing. Following staining, cells were washed gently and maintained in pre-warmed phenol-red-free medium. Fluorescence images were acquired immediately using a ZEISS AXIO Observer A1 inverted fluorescence microscope (German) equipped with appropriate filter sets. Confocal microscopy For immunofluorescence analysis, Raw264.7 cells were seeded on glass coverslips and treated with purified His-tagged VgrG2 protein (10 ng/mL). Cells were fixed with 4% paraformaldehyde for 15 min, permeabilized with 0.1% Triton X-100 for 10 min, and blocked with 5% bovine serum albumin (BSA) for 1 h at room temperature. To visualize late endosomes and the His-tagged protein, samples were incubated overnight at 4 °C with primary antibodies against Rab7 (Cell Signaling Technology; 1:200) and 6×-His Tag (Invitrogen; 1:500). After washing, cells were incubated for 1 h at room temperature with Alexa Fluor 555-conjugated anti-rabbit IgG (Invitrogen; 1:1000) and Alexa Fluor 488-conjugated anti-mouse IgG (Invitrogen; 1:1000). In parallel, F-actin was labeled with TRITC-phalloidin (Yeasen, Cat# 40734ES75; 1:200). Nuclei were counterstained with DAPI (Beyotime; 1 µg/mL) for 10 min. To examine cytoskeletal rearrangements, Raw264.7 cells were either treated with purified VgrG2 (10 ng/mL) or co-cultured with WT or ΔvgrG2 A. baumannii (MOI = 50:1) for indicated durations. After fixation and permeabilization, F-actin was stained with TRITC-phalloidin (1:200) for 30 min, followed by DAPI nuclear labeling. For intracellular expression of VgrG2 protein, Raw264.7 cells were transduced with a recombinant adenovirus expressing Flag-tagged VgrG2, which was constructed by Shanghai GeneChem Co., Ltd. (China), at a multiplicity of infection (MOI) of 500 for 72 h. Successful transduction was confirmed by green fluorescence. Cells were then fixed and stained for F-actin using TRITC-phalloidin and for nuclei with DAPI as described above. All samples were mounted with antifade medium and imaged on a Nikon A1 Ti confocal microscope (Japan) equipped with 405 nm, 488 nm, and 561 nm lasers. Scanning and transmission electron microscopy Bacterial surface structures, including pili, were visualized by scanning electron microscopy. Bacterial samples were fixed overnight at 4 °C in 2.5% glutaraldehyde in 0.1 M sodium cacodylate buffer (pH 7.4). After washing with the same buffer, samples were dehydrated through a graded ethanol series (30%, 50%, 70%, 90%, and 100%), followed by critical-point drying. Dried specimens were mounted on aluminum stubs and sputter-coated with gold. Imaging was performed using a scanning electron microscope (Zeiss Gemini560) operated at 3 kV. To examine VgrG2-induced vacuole formation, Raw264.7 cells treated with purified VgrG2 protein or co-cultured with wild-type AB5075 were processed for TEM analysis. Briefly, cells were fixed with 2.5% (v/v) glutaraldehyde in 0.1 M PBS (pH 7.4) at room temperature for 30 min, followed by overnight incubation at 4 °C. After three washes with PBS, the samples were post-fixed with 1% osmium tetroxide in 0.1 M PBS (pH 7.4) at 4 °C for 2 h. Subsequently, cells were dehydrated through a graded ethanol series (30%, 50%, 70%, 90%, and two changes of 100%) and embedded in Spurr’s resin. Ultrathin sections (~70 nm) were cut and placed on copper grids. Sections were stained with uranyl acetate and lead citrate, then examined using a transmission electron microscope (Tecnai Spirit 120 kV). Recombinant VgrG2 protein purification To generate recombinant VgrG2 protein, the coding region of vgrG2 from A. baumannii AB5075 was inserted into the pET-28a(+) expression vector, resulting in an N-terminal fusion with an 8×His tag. The resulting plasmid was introduced into Escherichia coli BL21(DE3) and transformants were selected on MH agar plates supplemented with kanamycin (50 µg/mL). A single colony was used to inoculate a starter culture in LB containing kanamycin (Sangon), which was grown at 37 °C with shaking until the optical density at 600 nm (OD600) reached 0.6–0.8. Protein expression was induced by adding isopropyl-β-D-thiogalactopyranoside (IPTG; Sangon) to a final concentration of 0.5 mM, followed by further incubation at 18 °C for 16–18 h with agitation. Cells were harvested by centrifugation (4000 × g, 20 min, 4 °C) and resuspended in lysis buffer (50 mM Tris-HCl, 250 mM NaCl, pH 8.5). Bacterial suspensions were lysed by sonication on ice, and cell debris was removed by centrifugation (12,000 × g, 40 min, 4 °C). The supernatant was applied to a pre-equilibrated Ni-NTA affinity column. After extensive washing, bound protein was eluted with eluent buffer (lysis buffer supplemented with 300 mM imidazole). Protein purity, assessed by SDS-PAGE, exceeded 95%. Then, the sample was loaded onto a pre-equilibrated Superdex 200 column to eliminate imidazole. Endotoxin levels in both the purified VgrG2 protein and its storage buffer (50 mM Tris-HCl, 250 mM NaCl, pH 8.5) were quantified using a commercial endotoxin detection kit (MCE, Cat#HY-K1099), confirming a final concentration of less than 0.5 EU mg⁻¹. Western blotting To confirm the expression and secretion of VgrG2 in vitro and during infection, Western blotting was performed on three sample types: bacterial whole-cell lysates, culture supernatants, and infected lung homogenates. For bacterial pellets, bacteria were harvested by centrifugation (12,000 × g, 10 min, 4 °C), washed with PBS, and lysed in RIPA buffer (Beyotime) containing protease inhibitor cocktail. Lysates were clarified by centrifugation (12,000 × g, 15 min, 4 °C). For culture supernatants, proteins in sterile-filtered culture supernatants were precipitated with 10% (w/v) trichloroacetic acid overnight at 4 °C, washed twice with ice-cold acetone, and resuspended in 1× SDS loading buffer. For mouse lung homogenates, lung tissues were homogenized in RIPA buffer, incubated on ice for 30 min, and centrifuged (12,000 × g, 15 min, 4 °C) to collect the soluble fraction. Protein concentrations were determined using a BCA assay kit (Thermo Scientific). Equal amounts of protein were separated by SDS-PAGE on 4–12% Tris-Gly gels (Beyotime) and transferred to PVDF membranes (Merck Millipore, Cat# ISEQ00010). Membranes were blocked with 5% BSA in TBST for 1 h at room temperature and then incubated overnight at 4 °C with anti-Flag monoclonal antibody (Sigma-Aldrich; 1:2000 dilution). After washing, membranes were incubated with HRP-conjugated secondary antibody (Merck Millipore, 1:5000) for 1 h at room temperature. Signals were developed using an enhanced chemiluminescence kit (Merck Millipore) and imaged on a ChemiDoc MP system (Bio-Rad). Phos-tag SDS-PAGE and immunoblotting of BfmR To assess the phosphorylation status of BfmR, bacterial whole-cell lysates from the WT AB5075 and the isogenic ΔvgrG2 mutant were prepared from mid-logarithmic phase cultures grown in LB broth. Cells were harvested by centrifugation, washed with phosphate-buffered saline, and resuspended in lysis buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1 mM phenylmethylsulfonyl fluoride, and protease inhibitor cocktail). Lysates were generated by sonication on ice, clarified by centrifugation, and protein concentrations were determined by BCA assay. For Phos-tag SDS-PAGE, separating gels (e.g., 12% polyacrylamide) were polymerized in the presence of 50 µM Phos-tag acrylamide (Wako Chemicals) and 100 µM MnCl₂, while the stacking gel was prepared without Phos-tag. Equal amounts of protein (20–30 µg) were mixed with SDS-PAGE sample buffer lacking EDTA and incubated at room temperature for 10 min prior to loading. Electrophoresis was performed at 4 °C at a constant current of 30 mA per gel until the dye front reached the bottom of the gel. Following electrophoresis, the gel was soaked in transfer buffer (25 mM Tris, 192 mM glycine, 20% methanol) containing 1 mM EDTA for 15 min with gentle agitation to remove Mn²⁺ ions, and then rinsed with EDTA-free transfer buffer. Proteins were transferred to polyvinylidene difluoride (PVDF) membranes at 100 V for 90 min at 4 °C. Membranes were blocked with 5% bovine serum albumin in TBS containing 0.1% Tween-20 (TBST) for 1 h at room temperature and incubated overnight at 4 °C with a custom rabbit polyclonal anti-BfmR antibody (diluted 1:2000 in blocking buffer). After three washes with TBST, membranes were probed with horseradish peroxidase-conjugated anti-rabbit IgG secondary antibody and developed with enhanced chemiluminescence substrate. Both total BfmR and the slower-migrating phosphorylated BfmR (p-BfmR) bands were imaged on a ChemiDoc MP system (Bio-Rad). Activation assays for Rac1, RhoA, and Cdc42 The activation states (GTP-bound forms) of the small GTPases Rac1, RhoA, and Cdc42 in Raw264.7 macrophages were determined using commercial pull-down assay kits (Cytoskeleton), following the manufacturer’s protocols with adaptations for our experimental system. Active Rac1 and Cdc42 were captured using glutathione S-transferase (GST) fusion proteins containing the p21-binding domain (PBD) of PAK1, which specifically binds GTP-loaded Rac1/Cdc42. Active RhoA was captured using the Rho-binding domain (RBD) of Rhotekin. Raw264.7 cells were seeded and treated as indicated (e.g., with purified VgrG2 protein or bacterial co-culture). After treatment, cells were rapidly placed on ice, washed with ice-cold PBS, and lysed in the provided Cell Lysis Buffer supplemented with protease inhibitors. Lysates were clarified by centrifugation (10,000 × g, 1 min, 4 °C), and protein concentrations were determined using a BCA assay. Lysates were either used immediately or snap-frozen in liquid nitrogen and stored at −70 °C. Equal amounts of lysate protein (300–800 µg) were incubated with GST-PBD beads (for Rac1/Cdc42) or GST-RBD beads (for RhoA) at 4 °C for 60 min with gentle rotation. Beads were pelleted, washed thoroughly with Wash Buffer, and resuspended in 2× Laemmli sample buffer. Samples were boiled for 5 min, and the supernatant was collected for Western blot analysis. Proteins were separated by SDS-PAGE (12% gel) and transferred to PVDF membranes. Membranes were blocked with 5% BSA and probed with primary antibodies against Rac1 (1:500, kit provided), RhoA (1:1000, kit provided), or Cdc42 (1:500, kit provided), followed by HRP-conjugated secondary antibodies. Signals were developed using an enhanced chemiluminescence substrate and imaged. Total GTPase levels in whole-cell lysates (input) were analyzed in parallel for normalization. The amount of active GTPase was quantified by densitometry and expressed relative to the total protein and/or untreated controls. Co-immunoprecipitation assay To confirm the direct interaction between VgrG2 and endogenous Rac1, co-immunoprecipitation (Co-IP) was performed using adenovirus-transfected Raw264.7 cells. Raw264.7 cells were transduced with an adenovirus expressing Flag-tagged VgrG2 at an MOI of 100 for 72 h. Cells were then lysed in ice-cold NP-40 lysis buffer (50 mM Tris-HCl pH7.5, 150 mM NaCl, 1% NP-40) supplemented with protease inhibitor cocktail. Lysates were clarified by centrifugation (12,000 × g, 15 min, 4 °C), and protein concentrations were determined using a BCA assay. For immunoprecipitation, 500 μg of total protein was incubated with 20 μL of Pierce™ Anti-DYKDDDDK Magnetic Agarose (Thermo Scientific) overnight at 4 °C with gentle rotation. Beads were collected using a magnetic stand and washed four times with lysis buffer. Bound proteins were eluted by boiling the beads in 2× Laemmli sample buffer for 5 min. The eluted proteins were separated by SDS-PAGE and transferred to PVDF membranes. Western blotting was performed using anti-Rac1 antibody (1:500, Cytoskeleton) and anti-Flag antibody (1:2000). HRP-conjugated secondary antibodies and enhanced chemiluminescence were used for detection. Input lysates (10%) were included to verify total protein expression. Statistical analysis Statistical analyses were performed using GraphPad Prism 10 software. Data are presented as mean ± standard deviation (SD). Normality and homogeneity of variance were assessed using Shapiro–Wilk and Brown–Forsythe tests, respectively. For comparisons between two groups, a two-tailed unpaired Student’s t test was used for normally distributed data; otherwise, the non-parametric Mann–Whitney U test was applied. Comparisons among three or more groups were analyzed by one-way ANOVA followed by Tukey’s post-hoc test for parametric data, or by the Kruskal–Wallis test with Dunn’s post-hoc correction for non-parametric data. Survival curves in animal experiments were compared using the Log-rank (Mantel–Cox) test. A p-value < 0.05 was considered statistically significant. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary information Supplementary Information Description of Additional Supplementary Files Dataset 1 Dataset 2 Dataset 3 reporting summary Transparent Peer Review file Source data Source Data Author contributions W.Z.X., W.Z., and Y.S.Y. designed the research, analyzed the data, wrote the manuscript, and prepared the figures and tables. W.Z.X. and J.C.Z. conducted the experiment and analysis. Q.Y.X., L.Y.C., X.X.T., C.Y.X., L.L., and J.X.Z. contributed some critical experimental materials and guidance. Y.K.Y., Y.S.Z., and J.C.Z. provided comments and revisions on the manuscript. H.Z., X.T.H., Y.J., and Y.S.Y. supervised the research, coordinated the production of the data, and supervised the writing of the manuscript. W.Z.X. and H.Z. finalized the manuscript with input from all the authors. All authors reviewed and approved the final manuscript. Peer review Peer review information Nature Communications thanks the anonymous reviewers 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 (No. 2023YFC2307100), the Noncommunicable Chronic Diseases-National Science and Technology Major Project (No. 2025ZD0549000 and No. 2025ZD0549004), the National Natural Science Foundation of China (No. 82403516, No. 82272338, and No. 82400005) and the Zhejiang Provincial Natural Science Foundation of China (No. LMS26H190001 and No. LQN25H190006). Data availability The bulk RNA-seq of AB5075 generated in this study have been deposited in the Gene Expression Omnibus (GEO) database under accession code GSE318779. The scRNA-seq data of lungs from mice infected with AB5075 WT and ΔvgrG2 strains generated in this study have been deposited in the GEO database under accession code GSE319700. The remaining data are available within the article, Supplementary Information or Source Data file. Source data are provided with this paper. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. These authors contributed equally: Wenzeng Xu, Jiancang Zhou, Wang Zhang. Contributor Information Xiaoting Hua, Email: xiaotinghua@zju.edu.cn. Yunsong Yu, Email: yvys119@zju.edu.cn. Hua Zhou, Email: zhouhua1@zju.edu.cn. Supplementary information The online version contains supplementary material available at 10.1038/s41467-026-75868-y. References Associated Data Supplementary Materials Supplementary Information Description of Additional Supplementary Files Dataset 1 Dataset 2 Dataset 3 reporting summary Transparent Peer Review file Source Data Data Availability Statement The bulk RNA-seq of AB5075 generated in this study have been deposited in the Gene Expression Omnibus (GEO) database under accession code GSE318779. The scRNA-seq data of lungs from mice infected with AB5075 WT and ΔvgrG2 strains generated in this study have been deposited in the GEO database under accession code GSE319700. The remaining data are available within the article, Supplementary Information or Source Data file. Source data are provided with this paper.