Outer membrane vesicles containing OmpA induce mitochondrial fragmentation to promote pathogenesis of Acinetobacter baumannii

Acinetobacter baumannii is a highly antibiotic resistant Gram-negative bacterium that causes life-threatening infections in humans with a very high mortality rate. A. baumannii is an extracellular pathogen with poorly understood virulence mechanisms. Here we report that A. baumannii employs the release of outer membrane vesicles (OMVs) containing the outer membrane protein A (OmpAAb) to promote bacterial pathogenesis and dissemination. OMVs containing OmpAAb are taken up by mammalian cells where they activate the host GTPase dynamin-related protein 1 (DRP1). OmpAAb mediated activation of DRP1 enhances its accumulation on mitochondria that causes mitochondrial fragmentation, elevation in reactive oxygen species (ROS) production and cell death. Loss of DRP1 rescues these phenotypes. Our data show that OmpAAb is sufficient to induce mitochondrial fragmentation and cytotoxicity since its expression in E. coli transfers its pathogenic properties to E. coli. A. baumannii infection in mice also induces mitochondrial damage in alveolar macrophages in an OmpAAb dependent manner. We finally show that OmpAAb is also required for systemic dissemination in the mouse lung infection model. In this study we uncover the mechanism of OmpAAb as a virulence factor in A. baumannii infections and further establish the host cell factor required for its pathogenic effects.


Scientific Reports
| (2021) 11:618 | https://doi.org/10.1038/s41598-020-79966-9 www.nature.com/scientificreports/ wildtype A. baumannii using genetic recombineering 26 . The OmpA Ab ::Flag-tagged strain was verified by PCR and OmpA Ab ::Flag expression was validated by western blotting (Supplementary Fig. 2D,E). Next, A549 cells were infected with the OmpA Ab ::Flag-tagged strain and anti-flag antibody was used to perform immunofluorescence. Intriguingly, we detected Flag positive puncta directly colocalizing with mitochondria ( Fig. 2B). We further validated these results biochemically by fractionating mitochondria after infecting A549 cells with the OmpA Ab ::Flag-tagged strain. OmpA Ab ::Flag was exclusively present in the mitochondrial fraction and not in the cytosolic fraction (Fig. 2C), confirming our imaging results. In addition, we observed a morphological disruption of the mitochondrial network in A549 cells infected with A. baumannii. The morphology of mitochondria in infected cells looked highly fragmented compared to uninfected cells (Fig. 2B,D,E). Mitochondrial fragmentation is not a general hallmark of bacterial infections as the infection of epithelial cells with other extracellular bacteria E. coli or S. aureus did not trigger mitochondrial fragmentation (Fig. 2D,E). To further confirm the specificity of this phenotype, A549 cells were infected with different strains of A. baumannii (lab adapted strains Ab19606 and Ab17978; and the clinical isolate Ab5075). Interestingly all three A. baumannii strains induced mitochondrial fragmentation in A549 cells (Supplementary Fig. 2F,G). We then tested if mitochondrial fragmentation is specific to epithelial cells or if it also occurs in myeloid cells like macrophages upon A. baumannii infection. Murine RAW264.7 macrophages infected with A. baumannii revealed a similar pattern of mitochondrial fragmentation suggesting that A. baumannii infection induces a disruption in the mitochondrial network across different cell types ( Supplementary Fig. 2H,I). A few bacteria including Legionella pneumophila and Chlamydia trachomatis are known to perturb the morphology of mitochondria as well as the endoplasmic reticulum (ER) and the Golgi apparatus thereby influencing the eukaryotic secretory pathway 25,[28][29][30][31] . We performed immunofluorescence using the ER marker Calnexin and the Golgi marker GM130 and did not observe any differences in the ER and Golgi morphology upon infection with A. baumannii in A549 cells ( Supplementary Fig. 2J,K). Therefore, A. baumannii infection seems to alter only the mitochondrial network without affecting the morphology of the ER and the Golgi apparatus. Further we wanted to assess the role of OmpA Ab in the observed mitochondrial fragmentation phenotype. While the wildtype bacteria induced mitochondrial fragmentation, no difference was observed between the mitochondria of the uninfected cells and the cells infected with the isogenic ΔOmpA strains in the respective Ab19606 and Ab17978 genetic backgrounds (Fig. 2F,G; Supplementary Fig. 3A) confirming that OmpA Ab is required for inducing mitochondrial fragmentation in host cells upon infection with A. baumannii. The mitochondrial -infection, animals were euthanized and lungs, liver, kidneys and spleen were homogenized and CFUs were plated from these organs. Blue dashed line in (B) represents the CFU with which each mouse was infected with. Black dotted line in (C-E) represents the limit of detection. Shown is the representative data from two independent experiments. Two-tailed p value using Mann-Whitney test *p ≤ 0.05, **p ≤ 0.01.  (Fig. 2I), indicating that OmpA Ab is sufficient to induce mitochondrial fragmentation. Taken together, these data reveal that OmpA Ab is necessary and sufficient to induce mitochondrial fragmentation.

DRP1 is required for mitochondrial fragmentation and cytotoxicity induced by A. baumannii.
Mitochondrial fragmentation has been fairly well studied and the canonical pathway that mediates mitochondrial fragmentation is regulated by a GTPase domain-containing cytoplasmic protein belonging to the dynamin family of proteins called dynamin-related protein 1 or DRP1, also known as DNM1L. During mitochondrial fission that eventually leads to fragmentation of the mitochondrial network, DRP1 localizes to mitochondria and forms oligomeric structures wrapping around the mitochondrial fission sites and utilizes GTP to bring about a change in its conformation that ultimately causes scission of the mitochondrial tubules into fragmented mitochondria 22,32 (Fig. 3A). Using immunofluorescence and confocal microscopy we observed enhanced DRP1 co-localization with mitochondria in A549 cells infected with wildtype A. baumannii compared to cells infected with A. baumannii ΔOmpA (Fig. 3B). To further validate its role, we knocked-down DRP1 by siRNA in A549 cells (Fig. 3C) and subsequently infected these cells with wildtype A. baumannii. Consistent with our immunofluorescence data, DRP1 loss completely abolished mitochondrial fragmentation upon A. baumannii infection (Fig. 3D,E). Concordantly, we also observed the absence of mitochondrial fragmentation when cells were treated with Mdivi-1 (a DRP1 inhibitor) prior to wildtype A. baumannii infection (Fig. 3F,G). Since we observed enhanced cytotoxicity with wildtype A. baumannii infection compared to A. baumannii ΔOmpA and cytotoxicity and cell death are associated with mitochondrial fragmentation, we asked if we can reverse cytotoxicity by preventing mitochondrial fragmentation and preserving mitochondrial integrity. We knocked-down DRP1 by siRNA in A549 cells and subsequently infected these cells with A. baumannii and monitored cytotoxicity. Consistent with our hypothesis we observed a significant reduction in cytotoxicity in cells treated with DRP1 siRNA relative to scrambled siRNA control upon infection with wildtype A. baumannii (Fig. 3H). There was no difference in cytotoxicity in DRP1 siRNA treated cells relative to scrambled siRNA control when infected with Figure 2. OmpA Ab is necessary and sufficient to cause mitochondrial fragmentation. (A) Cytotoxicity was assessed by LDH release assay in A549 cells treated with the indicated bacteria at MOI 50 for 24 h. The experiments were done in triplicates. Error bars represent standard deviation. Two-tailed p value using unpaired t-test *p ≤ 0.05, ****p ≤ 0.0001. % cytotoxicity was calculated by normalizing the LDH release in the infected groups with uninfected cells (representing no cytotoxicity) and TritonX-100 treated cells (representing the highest cytotoxicity). (B) A549 cells were infected with the indicated bacteria for 6 h at MOI 50. Immunofluorescence was performed using anti-TOM20 antibody to stain mitochondria (red), anti-Flag antibody to stain OmpA (green) and DAPI to label nuclei (blue). Arrows indicate Flag positive OmpA Ab ::Flag puncta (green) colocalizing with mitochondria (red). Scale bar represents 10 µm. (C) After 6 h of infection of A549 cells with the indicated bacteria at MOI 50, cellular fractionation was performed. Western blot analysis was done on the mitochondrial and cytosolic fractions using anti-Flag, anti-Tim23 and anti-ß-actin antibodies.
One-way ANOVA with Tukey's multiple comparisons test ****p ≤ 0.0001. Mitochondrial area and perimeter quantifications were performed using an unbiased automated CellProfiler pipeline (see "Materials and methods" section for details). All the experiments shown here were performed three times independently. Immunofluorescence was performed using anti-TOM20 antibody to stain mitochondria (red) and anti-DRP1 antibody (green). Scale bar represents 10 µm. (C) Western blot on A549 cell lysates after indicated treatments. Anti-DRP1 antibody was used to examine DRP1 levels and ß-actin served as the loading control. (D) A549 cells treated with DRP1 siRNA or control (scrambled) siRNA were infected with wildtype A. baumannii (Ab19606) for 6 h at MOI 50. Immunofluorescence was performed using anti-TOM20 antibody to stain mitochondria (red), anti-DRP1 antibody (green) and DAPI to label the nucleus (blue). Scale bar represents 10 µm.
(E) The scatter plots represent the quantification of mitochondrial area (red) and perimeter (orange). The experiment was performed in triplicates, n = 75-132 cells. Error bars represent standard deviation. One-way ANOVA with Tukey's multiple comparisons test **p ≤ 0.01, ****p ≤ 0.0001 (F) A549 cells pre-treated with Mdivi1 (10 µM) or DMSO control were infected with wildtype A. baumannii (Ab19606) for 6 h at MOI 50. Immunofluorescence was performed using anti-TOM20 antibody to stain mitochondria (red) and DAPI to label the nucleus (blue). Two-tailed p value using unpaired t-test ****p ≤ 0.0001. % cytotoxicity was calculated by normalizing the LDH release in the infected groups with uninfected/untreated cells (representing no cytotoxicity) and TritonX-100 treated cells (representing the highest cytotoxicity). Mitochondrial area and perimeter quantifications were performed using an unbiased automated CellProfiler pipeline (see "Materials and methods" section for details). All the experiments shown here were performed three times independently. www.nature.com/scientificreports/ A. baumannii ΔOmpA suggesting that OmpA Ab mediates its cytotoxic effects through mitochondrial fragmentation (Fig. 3H). Taken together these data suggest that OmpA Ab induces canonical mitochondrial fragmentation which is driven by DRP1 and consistent with its effect on mitochondrial morphology, there is a corresponding increase in cytotoxicity in an OmpA Ab dependent manner.       www.nature.com/scientificreports/ leads to perturbation of mitochondrial function. We examined reactive oxygen species (ROS) levels using Cell-ROX fluorogenic probe and observed that green fluorescence indicative of ROS levels was significantly higher in wildtype A. baumannii infected A549 cells compared to cells infected with A. baumannii ΔOmpA (Fig. 4A,B). We next treated A549 cells with E. coli OmpA Ab and E. coli EV. The green fluorescence was significantly stronger in cells infected with E. coli OmpA Ab compared to cells infected with E. coli EV (Fig. 4C,D). These data show that OmpA Ab is necessary and sufficient to increase the cellular ROS levels during infection. We then examined the effect of OmpA Ab on mitochondrial membrane potential and ATP levels which are indicative of mitochondrial function. Tetramethylrhodamine ethyl ester (TMRE), a fluorescent cell-permeant dye that is sequestered in active functional mitochondria was used to assess mitochondrial membrane potential. We observed significantly reduced fluorescent signal, indicative of de-polarized mitochondria, in A549 cells infected with wildtype A. baumannii but not in cells infected with A. baumannii ΔOmpA (Fig. 4E). Moreover, ATP levels in A549 cells infected with wildtype A. baumannii were also significantly reduced compared to cells treated with A. baumannii ΔOmpA (Fig. 4F). Together these results show that A. baumannii infection not only causes changes in mitochondrial morphology but also leads to functional defects in mitochondria as evidenced by enhanced levels of ROS, reduced mitochondrial membrane potential and reduced ATP levels. All of these effects on host mitochondria are dependent on OmpA Ab suggesting that A. baumannii uses OmpA Ab as a potent virulence factor that induces major perturbations in host cellular homeostasis.

A. baumannii infection perturbs mitochondrial physiology in an OmpA
OMVs containing OmpA Ab trigger mitochondrial fragmentation. To explore the mechanistic details of how OmpA Ab triggers mitochondrial fragmentation, we first asked if bacterial contact with the host cells or bacterial internalization within the host cells are required to trigger mitochondrial fragmentation. To address this question, we used transwell plates where A549 cells were seeded in the bottom chamber while A. baumannii was added in the top chamber; these two compartments were separated by a 0.45 µm filter that prevented direct contact between the bacteria and the epithelial cells (Fig. 5A). The cells that were incubated with bacteria in the transwell plate exhibited mitochondrial fragmentation suggesting that bacterial internalization or bacterial contact with epithelial cells are not required to induce this phenotype (Fig. 5B,C). These data also indicate that OmpA Ab is able to pass through the 0.45 µm filter to interact with host cells and brings about mitochondrial fragmentation. This led us to wonder if OMVs, which are vesicles shed by all Gram-negative bacteria, could be the potential carriers of OmpA Ab since OMVs are predominantly loaded with outer membrane proteins 33,34 and given their size (typically 20-200 nm) 34 , they would easily pass through the 0.45 µm transwell filter. Using a standard OMV isolation protocol (see "Materials and methods" section) we isolated OMVs from the OmpA Ab ::Flag-tagged strain (described in Fig. 2B) and different control strains. The presence of OMVs from each preparation was confirmed by visualization using electron microscopy ( Supplementary Fig. 4C). For the OmpA Ab ::Flag-tagged strain, the presence of OmpA Ab in the isolated OMVs was confirmed by western blotting (Fig. 5D). OmpA Ab was exclusively membrane associated and not in the lumen of OMVs as assessed by the proteinase K protection assay ( Supplementary Fig. 4A,B). Next, we treated A549 cells with the isolated OMVs Immunofluorescence was performed using anti-TOM20 antibody to stain mitochondria (red) and DAPI to stain the nucleus (blue). Scale bar represents 10 µm. (I) The scatter plots represent the quantification of mitochondrial area (red) and perimeter (orange). Error bars represent standard deviation, n = 37-50 cells.
One-way ANOVA with Tukey's multiple comparisons test *p ≤ 0.05, **p ≤ 0.01. (J) Cytotoxicity was assessed by LDH release assay after 6 h of OMV treatment in A549 cells. The experiments were done in triplicates. Error bars represent standard deviation. Two-tailed p value using unpaired t-test **p ≤ 0.01. % cytotoxicity was calculated by normalizing the LDH release in the treated groups with untreated cells (representing no cytotoxicity) and TritonX-100 treated cells (representing the highest cytotoxicity). (K) A549 cells were treated with 400 µg/ml of the indicated OMVs (fluorescently labelled with Vybrant Dio) for 6 h. Immunofluorescence was performed using anti-TOM20 antibody to stain mitochondria (red) and OMVs were labelled in green. Z-stack confocal imaging was performed on the cells, orthogonal views presented here. Arrows indicate OMVs (green) colocalizing with mitochondria (red). Scale bar represents 10 µm. (L) The bar graph represents the % of indicated OMVs colocalizing with mitochondria. The cells were stained with phalloidin and intracellular OMVs were scored for their colocalization with mitochondria or not. Error bars represent standard deviation. Mitochondrial area and perimeter quantifications were performed using an unbiased automated CellProfiler pipeline (see "Materials and methods" section for details). All the experiments shown here were performed three times independently. www.nature.com/scientificreports/ (400 µg/ml; based on BCA protein estimation assay). Consistent with our previous results, treatment of A549 cells with OMVs isolated from wildtype A. baumannii induced mitochondrial fragmentation but not OMVs isolated from A. baumannii ΔOmpA (Fig. 5E,F).
To ascertain the sufficiency of OmpA Ab in inducing mitochondrial fragmentation, OMVs from E. coli ΔOmpA Ec ,OmpA Ab and E. coli ΔOmpA Ec EV were isolated and verified by visualization using electron microscopy ( Supplementary Fig. 4C) and western blotting using an antibody against BamA, an outer membrane protein in E. coli (Fig. 5G). The presence of OmpA Ab only in the OMV preparations from E. coli ΔOmpA Ec ,OmpA Ab but not from E. coli ΔOmpA Ec EV was confirmed by western blotting for the FLAG tag that is fused to the OmpA Ab protein; the FLAG signal was only present in the OMV preparation from E. coli ΔOmpA Ec ,OmpA Ab and not in the OMV preparation from E. coli ΔOmpA Ec EV (Fig. 5G). Furthermore, contamination of OMVs with bacterial inner membrane was ruled out by western blotting using an antibody against MsbA, an inner membrane protein ( Supplementary Fig. 4D). We then treated A549 cells with these OMVs (400 µg/ml; based on BCA protein estimation assay) and observed mitochondrial fragmentation in cells treated with OMVs from E. coli ΔOmpA Ec ,OmpA Ab (Fig. 5H,I). By contrast, there was no difference observed between the mitochondria of untreated cells and the cells treated with OMVs isolated from E. coli ΔOmpA Ec EV (Fig. 5H,I). We also assessed cytotoxicity and found that the presence of OmpA Ab in OMVs increased cytotoxicity in a concentration dependent manner (Fig. 5J). Together these data indicate that OmpA Ab is necessary and sufficient to cause mitochondrial perturbation in the context of infection by intact bacterial cells (Fig. 2F-I) as well as by its presence in OMVs (Fig. 5E,F,H,I).
We next monitored fluorescently-labelled OMVs intracellularly in host epithelial cells. Isolated OMVs from E. coli ΔOmpA Ec ,OmpA Ab were labelled with the fluorescent dye Vybrant-DiO and then incubated with A549 cells (400 µg/ml; based on BCA protein estimation assay) for 6 h. Fluorescently-labelled OMVs could be detected within A549 cells after 6 h and intriguingly a large proportion of OMVs directly colocalized with mitochondria as revealed by confocal Z-stack microscopy ( Fig. 5K; Supplementary Fig. 4E). We next asked if the presence of OmpA Ab was required for the localization of OMVs to mitochondria. Interestingly, OMVs from E. coli ΔOmpA Ec EV also colocalized with mitochondria suggesting that mitochondrial localization of OMVs did not depend on the presence of OmpA Ab (Fig. 5L; Supplementary Fig. 4E). However, as described earlier mitochondrial www.nature.com/scientificreports/ fragmentation was totally dependent on the presence of OmpA Ab in OMVs (Fig. 5H,I). Collectively, the data suggest that as an extracellular pathogen, A. baumannii utilizes OMVs as a delivery mechanism to target OmpA Ab into host cells to induce mitochondrial fragmentation that is dependent on DRP1.

OmpA Ab induces mitochondrial fragmentation in vivo.
Having established the role of OmpA Ab in mediating mitochondrial fragmentation in vitro, we wondered about the relevance of our findings in an in vivo infection. We infected BALB/c mice intranasally with 10 7 CFUs of wildtype A. baumannii Ab17978, isogenic A. baumannii ΔOmpA and the ΔOmpA + OmpA Ab ::Flag complemented strain. After 6 h of infection, we sacrificed the animals and collected the bronchoalveolar lavage (BAL) fluid from the trachea and isolated alveolar macrophages for immunofluorescence to examine their mitochondrial morphology (Fig. 6A). We observed significant mitochondrial fragmentation in macrophages isolated from the BAL fluid of mice infected with wildtype A. baumannii Ab17978 and the ΔOmpA + OmpA Ab ::Flag complemented strain (Fig. 6B,C). Macrophages isolated from the BAL fluid of mice infected with A. baumannii ΔOmpA displayed mitochondrial morphology similar to uninfected mice suggesting that OmpA Ab is required for inducing mitochondrial fragmentation in murine alveolar macrophages during a mouse lung infection (Fig. 6B,C). We further followed the localization of OmpA Ab ::Flag in alveolar macrophages from the BAL fluid and observed OmpA Ab ::Flag positive puncta colocalizing with mitochondria (Fig. 6D), similar to our initial observations in epithelial cells (Fig. 2B). Taken together these results suggest that similar to our in vitro data, OmpA Ab targets mitochondria and induces mitochondrial fragmentation in murine macrophages upon A. baumannii infection.

Discussion
A handful of primarily intracellular bacteria have been reported to perturb mitochondrial functions to disrupt host cell homeostasis 24 . In this study we describe how the extracellular bacterial pathogen A. baumannii uses the protein OmpA Ab to induce significant changes in mitochondrial morphology and function ultimately leading to host cell death and pathogenesis. Our results provide a mechanistic framework and suggest in vivo relevance to the previously reported observation of the localization of recombinant OmpA Ab protein to mitochondria 20 . We observed that A. baumannii infection induces mitochondrial fragmentation in host epithelial cells and macrophages in an OmpA Ab dependent manner. Furthermore, OmpA Ab also caused mitochondrial fragmentation in alveolar macrophages in a mouse lung infection model. Three genetically distinct isolates of A. baumannii including the clinically isolated strain Ab5075 cause mitochondrial fragmentation in host cells, demonstrating that this is a clinically relevant and conserved virulence mechanism of A. baumannii. OmpA Ab activates the canonical mitochondrial fragmentation pathway in lung epithelial cells which is driven by the host GTPase protein DRP1. Furthermore, OmpA Ab induces higher levels of ROS, dissipation of mitochondrial membrane potential and a significant reduction in ATP levels all indicative of disrupted mitochondrial function.
One of the open questions in the field is how extracellular bacteria can potentially target mitochondria and furthermore, it is also unknown if this virulence strategy is relevant in vivo. We describe the mechanism of OmpA Ab targeting mitochondria via OMVs released by A. baumannii and provide evidence that OmpA Ab is required for mitochondrial disruption in vivo. OMVs are secreted by all Gram-negative bacteria and their major components include the outer membrane and periplasmic proteins. Additionally, extracellular bacteria can secrete OMVs carrying toxins and virulence factors that can penetrate the host cells and cause damage without the actual bacteria getting internalized into host cells [34][35][36] . In accord with this we observed that physically separating host epithelial cells from A. baumannii did not prevent mitochondrial fragmentation and that the treatment of host epithelial cells with OMVs containing OmpA Ab was sufficient to induce mitochondrial fragmentation suggesting that OMVs carry OmpA Ab into cells where it causes mitochondrial fragmentation. OMVs were seen to colocalize directly with mitochondria. Furthermore, OmpA Ab positive puncta colocalized with mitochondria in alveolar macrophages derived from A. baumannii infected mice highlighting the relevance of our findings in A. baumannii pathogenesis in vivo. Since OmpA Ab is in the membrane of OMVs, OMVmitochondria colocalization potentially creates an interaction between OmpA Ab and mitochondria that might be required for the induction of mitochondrial fragmentation. Interestingly, a recent study reported the targeting of Neisseria gonorrhoeae outer membrane porin protein PorB to mitochondria via OMVs suggesting that the strategy of targeting mitochondria and perturbing mitochondrial functions might be a conserved virulence mechanism across bacteria 37 . Even though it is believed that OMVs are targeted to lysosomes for degradation, several studies have shown the presence of OMVs in the cytoplasm of host cells [36][37][38] . In agreement with these studies, we also observed OMVs in the cytoplasm colocalizing with mitochondria. Yet, it is not completely understood how the presence of OmpA Ab in OMVs induces mitochondrial fragmentation. It could be a direct insertion of the porin channel into mitochondria which leads to the release of cytochrome C into the cytoplasm thus disrupting mitochondrial function and leading to subsequent cell death. Alternatively, it could be an interaction of OmpA Ab with mitochondrial protein/s that activates the DRP1 mediated mitochondrial fission pathway leading to fragmented mitochondrial network and cell death. Future studies to determine what host-cell proteins interact with OmpA Ab , and what structural features of OmpA Ab are recognized and/or required for the phenotype will provide even greater mechanistic resolution of this process. We are intrigued by the fact that E. coli OmpA has none of the pathogenic functions of OmpA Ab , yet OmpA Ab expressed in E. coli can bestow its heterologous host with the molecular phenotypes characteristic of OmpA Ab . It is tempting to speculate that the pathogenic function of OmpA Ab described here may be a conserved mechanism of pathogenesis in other important human bacterial pathogens from the Moraxellaceae Family (e.g. Moraxella catarrhalis) and the Pseudomonadales Order (e.g. Pseudomonas aeruginosa).
A. baumannii infection often leads to severe pneumonia in immunocompromised people 39 . In the later stages of infection there is a disruption of the lung epithelial barrier due to cell death which allows the bacteria access to Scientific Reports | (2021) 11:618 | https://doi.org/10.1038/s41598-020-79966-9 www.nature.com/scientificreports/ blood and causes the infection to spread to other organs, a critical condition called bacteremia, which eventually leads to sepsis 2,39,40 . We studied bacterial dissemination to different organs using wildtype A. baumannii and A. baumannii ΔOmpA in vivo. Using a murine lung infection model, we show that OmpA Ab is required for bacterial colonization and dissemination to other organs demonstrating its crucial role in driving bacteremia during A. baumannii infections. Our work is in agreement with previous observations made in patients with A. baumannii infection where bacteremia and pneumonia were positively correlated with higher expression of OmpA Ab 14 . It has also been shown that the loss of OmpA Ab makes A. baumannii susceptible to human serum which could also explain why A. baumannii ΔOmpA is incapable of inducing bacteremia in patients 18 . Taken together our work complements previous findings and additionally provides mechanistic insights into the role of OmpA Ab , and the OMVs that carry it, as a major virulence factor in A. baumannii pathogenesis.

Materials and methods
Bacterial strains and culture conditions. The A. baumannii strains used in this study included the lab adapted strains Ab19606, Ab17978 and the clinical isolate Ab5075. S. aureus strain USA300 and E. coli Bw25113 were also used in this study. All the bacterial strains used in the study are listed in Supplementary Table 1. Bacterial strains were obtained from the American Type Culture Collection (ATCC). Bacterial cultures were grown by picking a single bacterial colony from tryptic soy agar (TSA) plates into tryptic soy broth (TSB) medium and the cultures were grown overnight at 37 °C.
Cell culture. A549 lung epithelial cells, HeLa cells and RAW264.7 macrophages were used in this study. Cloning and genetic recombineering. OmpA Ab was cloned using standard cloning techniques into Acinetobacter-E. coli shuttle vector pWH1266 41 . Briefly, C-terminally flag-tagged OmpA Ab was synthesized by Integrated DNA Technologies (IDT) and using PCR, terminal overhangs complementary to the vector pWH1266 were added. The vector was digested with BamH1 and the PCR amplified fragmented and the digested vector were ligated and finally transformed into DH5α competent cells and plated on LB plates containing 50 µg/ml carbenicillin for selection. Positive colonies were verified by PCR and Sanger sequencing.
Genetic recombineering to generate A. baumannii ΔOmpA and OmpA Ab ::Flag tagged strains was performed following the established recombineering method for A. baumannii 26,42 . Briefly, for OmpA Ab ::Flag strains constructs were designed using OmpA Ab coding sequence with upstream and downstream regions along with gentamicin resistance sequence for selection. For generating A. baumannii ΔOmpA, a kanamycin cassette containing upstream and downstream regions of OmpA Ab coding sequence were used. The constructs were synthesized by IDT and PCR amplification of the product was performed further. 5 µg of the amplified PCR products were electroporated into competent A. baumannii expressing Rec AB recombinase previously induced by 2 mM isopropyl ß-d-1-thiogalactopyranoside (IPTG). After electroporation the bacteria were outgrown in a total of 4 ml TSB media containing 2 mM IPTG for 4 h. Finally the bacteria were pelleted at 3000 rpm (956 g), most of the supernatant was removed and the final leftover volume containing the bacteria was plated on TSB plates containing 7.5 µg/ml gentamicin (for OmpA Ab ::Flag strains) and 50 µg/ml kanamycin (for A. baumannii ΔOmpA strain). Notably, it was challenging to generate the complemented strain in the A. baumannii 19606 genetic background. Despite multiple efforts, we were unable to complement A. baumannii ΔOmpA strain in the A. baumannii 19606 genetic background. Therefore, we show phenotypic rescue in A. baumannii 17978 genetic background. Of note, we did not observe any strain specific differences between A. baumannii 19606 and A. baumannii 17978 which were tested throughout this study.
Immunofluorescence, confocal microscopy and imaging data analysis. For immunofluorescence 10 5 cells/chamber were seeded in a 4 chamber slide a night before the experiment. The next day after the experimental treatment, the cells were washed twice with PBS and fixed with 4% (v/v) paraformaldehyde (PFA) and then permeabilized with PBS containing 0.1% (v/v) triton-X 100 for 15 min at room temperature. After blocking in 1% (w/v) bovine serum albumin (BSA) for an hour at room temperature, the samples were treated with primary antibody (diluted in BSA at 1:200 www.nature.com/scientificreports/ the samples were washed three times with PBS and treated with fluorescently labelled (Alexa488 or Alexa594) secondary antibodies for one-hour room temperature. The samples were then mounted with ProLong Gold mounting medium containing DAPI. Antibodies used for immunofluorescence: Tom20 (Proteintech, 11802-1-AP), DRP1 (Novus, NB110-55288), Flag (M2, Sigma Aldrich). For all the imaging Nikon Ti-E spinning disk confocal microscope was used. All the images were acquired using the 100X oil immersion objective. Image analysis and processing was performed using Fiji open source software. Mitochondrial imaging data analysis and quantification were performed in an unbiased automated way using an open source software called CellProfiler 43 . A CellProfiler pipeline was designed to capture mitochondria and the software added artificial masks and calculated area and perimeter of mitochondria from those masks. The quantification of mitochondrial fragmentation in alveolar macrophages from mouse BAL fluid was done as described before 44 . Briefly, the cells with a continuous peri-nuclear mitochondrial network were scored as normal while the cells displaying a discontinuous segmented peri-nuclear mitochondrial network were scored as fragmented. The quantification of mitochondrial fragmentation was performed in a completely blinded manner. Cytotoxicity analysis. Supernatants from 10 5 cells treated with bacteria or OMVs were collected at specific time points (see figure legends). Levels of released LDH in the supernatants was assessed using the cytotoxicity kit (11644793001 Roche). Detergent induced cell lysis was used as 100% cytotoxicity and untreated cells were used as background. Relative cytotoxicity was calculated by comparing treated cells with detergent lysed cells and the background from untreated cells was subtracted.
Western blotting. 10 6 cells/well were washed twice with PBS. The cells were lysed in 150 µl of RIPA buffer containing protease and phosphatase inhibitors (ThermoFisher Scientific). BCA protein estimation (Ther-moFisher Scientific) was performed to ascertain the protein concentration. Standard western blot protocol was followed further 45 . Briefly, the lysates were mixed with SDS-PAGE sample buffer, boiled at 95 °C and 20 µg protein was loaded on an SDS-PAGE precast gel (Invitrogen). The gel was blotted onto a nitrocellulose membrane and blocked for an hour in the intercept blocking buffer (LI-COR) at room temperature. Primary antibodies were used at 1:1000 dilution and incubated overnight. The next day after three washes in tris-buffered saline with 0.1% (v/v) tween 20 detergent (TBST), secondary antibodies were added at 1:10,000 dilution followed by an hour of incubation at room temperature. The membranes were imaged with LI-COR Odyssey CLx imaging system. For western blots using bacterial lysates, log phase bacterial cultures were centrifuged down at 10,000 rpm (10,621 g) for 10 min in an Eppendorf benchtop centrifuge and resuspended in sample buffer. The samples were boiled at 95 °C and 20 µg protein was loaded on an SDS-PAGE precast gel (Invitrogen). Antibodies used for western blotting: DRP1 (Novus, NB110-55288), Flag (M2, Sigma Aldrich), β-Actin (CST), GroEL (Enzo, ADI-SPP-610), MsbA (produced in-house), Tim23 (Proteintech, 11123-1-AP). Uncropped western blot images are presented in Supplementary Fig. 5. siRNA knockdown. 10 5 A549 cells/well were treated with 5 µM DRP1 siRNA (ON TARGETplus siRNA, Horizon Discovery). 5 µl of 5 µM siRNA (DRP1 or scrambled) was mixed with 95 µl serum free RPMI and further mixed with 100 µl of transfection reagent (5% v/v transfection formulation in serum free RPMI) and added onto cells. The cells were incubated for 48 h at 37 °C. Transfection efficiency was verified by western blotting.
OMV isolation and fluorescent labeling. OMVs were isolated as described previously with a few modifications 46 . One liter overnight grown bacterial cultures were centrifuged at 12,000 rpm (24,470 g) for 30 min at 4 °C in a Thermo Scientific Sorvall Lynx 6000 centrifuge. The supernatant was filtered through a 0.45 µm PVDF filter. The filtered supernatant was centrifuged at 40,000 rpm (185,511 g) for 2 h at 4 °C in a Beckman Coulter Optima L-100 K ultracentrifuge. The supernatant was removed and the pellet containing OMVs was resuspended in PBS containing 200 mM NaCl, 1 mM CaCl2, and 0.5 mM MgCl2. Protein concentration in OMVs was assessed by the BCA assay (ThermoFisher Scientific). For fluorescent labeling, OMVs were treated with the fluorescent dye Vybrant DiO (ThermoFisher Scientific) at 1:100 and incubated at 37 °C for 20 min. After the labeling, the samples were centrifuged at 10,000 rpm (10621 g) in an Eppendorf benchtop centrifuge to remove excess dye using Micron 10 kDa centrifugal filter units following manufacturer's protocol. The samples were washed three times with PBS and centrifuged at 3000 rpm (956 g) in an Eppendorf benchtop centrifuge to remove the unattached dye.
Mitochondrial assays: assessing ROS levels, mitochondrial membrane potential and ATP levels. ROS levels were assessed using the CellROX green reagent (ThermoFisher Scientific) following the manufacturer's protocol. 30% v/v H 2 O 2 used at a final concentration of 1:1000 for an hour at 37 °C served as the positive control. The cells were fixed with 4% v/v PFA, mounted with ProLong Gold mounting medium containing DAPI and imaged.
Mitochondrial membrane potential was examined using the TMRE based assay (ThermoFisher Scientific) following the manufacturer's protocol. The assay was performed in a 96 well dish and the fluorescence was read using a BioTek microplate reader. ATP levels were examined by using the ATPlite kit (Perkin-Elmer) following the manufacturer's protocol. The assay was performed in white walled 96-well dishes and the luminescence was read using a BioTek microplate reader. www.nature.com/scientificreports/ Mitochondrial isolation. Mitochondrial isolation was carried out using the mitochondrial isolation kit for cultured cells (ThermoFisher Scientific). Briefly, 10 7 cells after infection were washed three times with PBS and detached using TrypLE (ThermoFisher Scientific, Cat. No. 89874). The kit protocol was followed for cell fractionation. Mitochondrial and cytoplasmic fractions were collected at the end of the experiment and ran on an SDS-PAGE followed by western blot using antibodies Tim23 (Proteintech, 11123-1-AP) for the mitochondrial fraction and β-Actin (CST) for the cytoplasmic fraction. After a couple of more washes, cytospin was used to deposit the macrophages evenly on a microscopic slide. The deposited macrophages were encircled with a PAP pen in order to retain liquid on them during the immunofluorescence steps. Standard immunofluorescence (as described above) was performed on the slide directly taking care not to dislodge the cells. The antibodies used for immunofluorescence were anti-TOM20 and anti-Flag antibodies. The samples were finally mounted with ProLong Gold mounting medium containing DAPI.

Mouse infection.
Electron microscopy. The suspension of OMVs was adsorbed for 15 min to the surface of formvar and carbon coated transmission electron microscope (TEM) grids. After a shot rinse with distilled water, the sample was stained with 2% (v/v) uranyl acetate for 60 s and then air dried. Imaging was performed with a JEOL JEM-1400 TEM and a GATAN Ultrascan 1000 CCD camera at magnifications from 5000 × to 50,000 ×. Scale bars are indicated in the images.