1. Microbiology and Infectious Disease
Download icon

A novel decoy strategy for polymyxin resistance in Acinetobacter baumannii

  1. Jaeeun Park
  2. Misung Kim
  3. Bora Shin
  4. Mingyeong Kang
  5. Jihye Yang
  6. Tae Kwon Lee
  7. Woojun Park  Is a corresponding author
  1. Laboratory of Molecular Environmental Microbiology, Department of Environmental Science and Ecological Engineering, Korea University, Republic of Korea
  2. Department of Environmental Engineering, Yonsei University, Republic of Korea
Research Article
  • Cited 0
  • Views 397
  • Annotations
Cite this article as: eLife 2021;10:e66988 doi: 10.7554/eLife.66988

Abstract

Modification of the outer membrane charge by a polymyxin B (PMB)-induced PmrAB two-component system appears to be a dominant phenomenon in PMB-resistant Acinetobacter baumannii. PMB-resistant variants and many clinical isolates also appeared to produce outer membrane vesicles (OMVs). Genomic, transcriptomic, and proteomic analyses revealed that upregulation of the pmr operon and decreased membrane-linkage proteins (OmpA, OmpW, and BamE) are linked to overproduction of OMVs, which also promoted enhanced biofilm formation. The addition of OMVs from PMB-resistant variants into the cultures of PMB-susceptible A. baumannii and the clinical isolates protected these susceptible bacteria from PMB. Taxonomic profiling of in vitro human gut microbiomes under anaerobic conditions demonstrated that OMVs completely protected the microbial community against PMB treatment. A Galleria mellonella-infection model with PMB treatment showed that OMVs increased the mortality rate of larvae by protecting A. baumannii from PMB. Taken together, OMVs released from A. baumannii functioned as decoys against PMB.

eLife digest

Wrapped in a thick, protective outer membrane, Acinetobacter baumannii bacteria can sometimes cause serious infections when they find their way into human lungs and urinary tracts. Antibiotics are increasingly ineffective against this threat, which forces physicians to resort to polymyxin B, an old, positively-charged drug that ‘sticks’ to the negatively-charged proteins and fatty components at the surface of A. baumannii.

Scientists have noticed that when bacteria are exposed to lethal drugs, they often react by releasing vesicles, small ‘sacs’ made of pieces of the outer membranes which can contain DNA or enzymes. How this strategy protects the cells against antibiotics such as polymyxin B remains poorly understood.

To investigate this question, Park et al. examined different strains of A. baumannii, showing that bacteria resistant to polymyxin B had lower levels of outer membrane proteins but would release more vesicles. Adding vesicles from resistant strains to non-resistant A. baumannii cultures helped cells to survive the drugs. In fact, this protective effect extended to other species, shielding whole communities of bacteria against polymyxin B. In vivo, the vesicles protected bacteria in moth larvae infected with A. baumannii, leading to a higher death rate in the animals. Experiments showed that the negatively-charged vesicles worked as decoys, trapping the positively-charged polymyxin B away from its target.

Taken together, the findings by Park et al. highlight a new strategy that allows certain strains of bacteria to protect themselves from antibiotics, while also benefitting the rest of the microbial community.

Introduction

Pathogenic Acinetobacter baumannii strains cause urinary tract infections and ventilator-associated pneumonia (Wright et al., 2014). The incidence and prevalence of multidrug-resistant A. baumannii (MRAB), which are resistant to various antibiotics including sulbactam and tigecycline, threaten the effective prevention and treatment of MRAB-related infections (Brauers et al., 2005; Perez et al., 2007; Alsan and Klompas, 2010). The broad-spectrum β-lactam carbapenem antibiotics, such as imipenem, meropenem, and doripenem, are mainly used to treat severe MRAB infections (Papp-Wallace et al., 2011; Kang et al., 2012). Unfortunately, studies have reported the spread of MRAB clonal complex 92, which contains the blaOXA-23 gene encoding a carbapenem-hydrolyzing class D β-lactamase (Antunes et al., 2014; Yoon et al., 2017). Mobile genetic elements, particularly ISAba I upstream of the blaOXA-type genes, allow A. baumannii strains to transmit the antibiotic resistance (AR) gene to other bacteria and the environment through horizontal gene transfer (Bahador et al., 2015). Carbapenem-resistant A. baumannii (CRAB) is a top-priority pathogen, for which new antibiotics or combination therapies are desperately needed (Shlaes and Bradford, 2018). Lack of an efficacious treatment for MRAB and CRAB infections has caused a reversion to initial treatments involving drugs that are considered dated now, such as colistin and polymyxin B (PMB), despite the adverse effects of these drugs (Velkov et al., 2013).

Polymyxins are cationic polypeptide antibiotics that specifically kill Gram-negative bacteria by binding to the bacterial outer membrane (OM). These last-resort antimicrobial agents interact with the negatively charged lipopolysaccharide (LPS) components of the OM by displacing positively charged ions, such as Mg2+ and Ca2+. Consequently, the integrity of the cell membrane is impaired, leading to cell death (Jeannot et al., 2017; Moffatt et al., 2019). Modification of lipid A in LPS by pmrCAB and arnBCADTEF gene products, which are modulators of the bacterial OM surface charge and permeability, is generally recognized as a primary mechanism of resistance to polymyxins (Trent et al., 2001; Olaitan et al., 2014; Jeannot et al., 2017). The pmrCAB and arnBCADTEF operons regulated by a PmrAB two-component system modify lipid A by adding 4-amino-4-deoxy-l-arabinose (LAra4N) and phosphoethanolamine (PetN), respectively (Kline et al., 2008; Arroyo et al., 2011; Pelletier et al., 2013). Activation of the arnT operon (arnBCADTEF, also called as the pmrHFIJKLM operon) adds LAra4N groups to lipid A only in several Gram-negative bacteria such as Escherichia coli, Yersinia pestis, Pseudomonas aeruginosa, and Ralstonia solanacearum, but not in A. baumannii or Citrobacter rodentium (Trent et al., 2001; Sinha et al., 2019). Because A. baumannii lacks the genes for the biosynthesis of LAra4N, it uses PetN addition as the main polymyxin-resistance mechanism (Gerson et al., 2019). In PMB-resistant A. baumannii, the mutant PmrB protein is autophosphorylated, consequently phosphorylating the cytoplasmic transcriptional regulator PmrA. Phosphorylated PmrA then upregulates the PetN transferase PmrC and prevents PMB binding to lipid A (Cannatelli et al., 2014). These modifications in lipid A decrease the negative charge of the cell membrane, reducing the membrane-binding affinity of PMB. Consequently, the cell is protected from the PMB-mediated disruption of its membrane and survives (Trimble et al., 2016). In some A. baumannii clinical isolates, expression of the eptA gene is activated by the adjacent ISAba I sequence, and the gene product PmrC homolog contributes to PMB resistance (Trebosc et al., 2019). Interestingly, E. coli producing outer membrane vesicles (OMVs) can survive in the presence of PMB (Manning and Kuehn, 2011; Kulkarni et al., 2015; Kim et al., 2018).

Gram-negative bacteria release OMVs with diameters of 20–400 nm to their environments. These vesicles are composed of phospholipids (PLs), outer membrane proteins (OMPs), and LPSs or lipooligosaccharides (LOS), but also contain periplasmic proteins and cell wall components (Roier et al., 2016a). Modulation of peptidoglycan (PG) cross-links is associated with membrane stability and rigidity (Schwechheimer and Kuehn, 2015). Approximately one-third of the alpha-helical OM lipoprotein Lpp (also known as Braun’s lipoprotein) anchored to the outer cell membrane are covalently cross-linked with PGs through connection between C-terminal lysine residue of Lpp and the diaminopimelic acid (meso-DAP) site of PG (Dramsi et al., 2008). Reduction in Lpp production disrupts the cross-linkage of PG on the membrane, thereby significantly contributing to the formation of OMVs (Wessel et al., 2013). Differential expression of OM protein A (OmpA), which is a β-barrel porin that non-covalently interacts with the meso-DAP site of PG, can modulate OMV excretion (Schwechheimer and Kuehn, 2015). Many mutant studies have shown that the absence of inner membrane (IM) proteins, such as the Tol component of the Tol-Pal system and NlpA, can enhance OMV production (Bernadac et al., 1998; Toyofuku et al., 2015). Production of OMVs is also facilitated by a high periplasmic turgor pressure, which occurs when misfolded proteins, nucleotide fragments, or PG debris accumulate in the periplasm (Brown et al., 2015). Additionally, in Haemophilus influenzae, OMVs are also released by PLs that accumulate on the OM when the PL transporters of the VacJ/Yrb ABC transport system (also known as the Mla system) are not produced (Roier et al., 2016b). Changes in membrane fluidity and rigidity, determined by the compositions of acyl chains and head groups of PLs, affect OMV production (Mashburn-Warren et al., 2008). The proportion of saturated fatty acids appears to be higher in OMVs than in cellular membranes (Baumgarten et al., 2012).

In pathogenic bacteria, these spherical OMVs formed by various routes can have many different physiological functions, including biofilm formation, cell–cell communication, secretion of virulence factors, and survival under antibiotic stress (Toyofuku et al., 2015; Bonnington and Kuehn, 2016). OMVs have attracted interest as putative vaccine agents or vehicles for drug delivery (Jan, 2017). They can reduce the effectiveness of antibiotics and other bactericidal agents, thereby posing a serious threat to the treatment of bacterial infections (Koeppen et al., 2016; Jan, 2017). For example, the OMVs of β-lactam-resistant E. coli are enriched with proteins involved in degrading β-lactam antibiotics in comparison with those of susceptible bacteria (Kim et al., 2018). Accordingly, OMV supplementation of microbial cultures increases the microbial resistance to antimicrobial peptides. Similarly, hypervesiculating mutants show increased resistance (Manning and Kuehn, 2011). OMVs from A. baumannii are known to possess various virulence factors such as OmpA, proteases, phospholipases, superoxide dismutase, and catalase (Kwon et al., 2009). However, it is unclear whether A. baumannii can gain polymyxin resistance or confer such resistance onto its bacterial community through hypervesiculation. Here, the genomes and OMV-producing phenotypes of experimental evolution PMB-resistant A. baumannii cells were analyzed alongside other clinical isolates as well as the A. baumannii ATCC 17978 strain. Our multi-omics data and mutational studies suggested that OMVs functioned as decoys for PMB in A. baumannii. Accordingly, these vesicles can protect not only the OMV producer, but also the entire bacterial community from the bactericidal effects of PMB.

Results

OMV secretion and enhanced biofilm formation

PMB-resistant strains were generated by exposing wild-type cells (Lab-WT; minimum inhibitory concentration [MIC], 2 μg/mL) to increasing concentrations of PMB. Consequently, strains that had 2-fold (MIC, 4 μg/mL) and 64-fold (MIC, 128 μg/mL) higher PMB resistance than Lab-WT were generated (low and high PMB resistance, PMRLow and PMRHigh, respectively) (Figure 1A, Figure 1—source data 1). The PMRHigh strain had a lower growth rate (1.386 ± 0.17 h−1) than both Lab-WT (2.081 ± 0.92 h−1) and PMRLow (1.985 ± 0.86 h−1) strains (Figure 1—figure supplement 1A). Expression analyses of the pmrC gene, which encodes a PetN transferase in the absence of PMB, showed 1.2- and 4-fold higher expression in PMRLow and PMRHigh strains, respectively (Figure 1—figure supplement 1B). The zeta-potential analyses suggested that the average negative surface charge of PMRLow and PMRHigh decreased from −27 mV (Lab-WT) to −20 mV and −10 mV, respectively (Figure 1B, Figure 1—source data 1). The reduced surface charge, which is partly due to PmrC-driven incorporation of PetN into lipid A, presumably decreases the initial binding of PMB, a cationic amphiphilic cyclic decapeptide, to the cell surface. Consequently, cells with more PetNs in lipid A acquire resistance to PMB. PmrC-mediated reduction of surface charges was not feasible in PMRLow because the expression level of the pmrC gene was not significant. These observations could not account for all the reasons underlying the high PMB resistance of PMRHigh strain because short-term (10 min) PMB exposure (4 μg/mL, MIC of PMRLow) decreased the surface charge in all tested strains, albeit to different extents (Figure 1—figure supplement 1C, n = 10 per strain). OMV-associated biofilm was measured in the Lab-WT and PMRHigh strains (Figure 1—figure supplement 1D, E). Biofilm formation observed using confocal microscopy with FilmTracer indicated that the PMRHigh strain produced thick and condensed biofilm within 24 hr, and the addition of PMB induced biofilm formation in both strains (Figure 1—figure supplement 1D). Quantification of biofilm formation using the crystal violet assay also indicated that the PMRHigh strain produced more biofilm (twofold higher) than the Lab-WT strain (Figure 1—figure supplement 1E). Interestingly, addition of OMVs derived from PMRHigh increased the initial stage of Lab-WT biofilm formation (Figure 1—figure supplement 1F, G). The PMRLow and PMRHigh strains were resistant only to polymyxins, but not to the other nine tested antibiotics of different classes, including meropenem, rifampicin, chloramphenicol, and ampicillin (Figure 1—figure supplement 2). This PMB-specific phenotype prompted us to confirm hypervesiculation of the PMR-resistant strains through transmission electron microscopy (TEM). OMV production appeared to be twofold higher in PMRHigh than in Lab-WT, and the treatment with 1/2 MIC PMB (1 or 128 µg/mL, respectively) caused increased OMV production as quantified using the lipophilic dye FM4-64 (Figure 1C). We observed markedly larger and greater production of OMVs in the PMRHigh strain than in our Lab-WT strain using TEM (Figure 1D, Figure 1—figure supplement 1H). Dissociated OMVs from PMRHigh were larger and became entangled with other OMVs through hairy structured materials (Figure 1E).

Figure 1 with 2 supplements see all
Generation of experimentally evolved polymyxin B (PMB)-resistant strains and observation of outer membrane vesicle (OMV) production.

(A) PMB-resistant strains generated by exposing wild-type cells to increasing concentrations of PMB (N = 3 biological replicates). (B) Zeta-potential of the Lab-WT, PMRLow, and PMRHigh strains (n = 10 each) under the non-treated condition. (C) OMV production was quantified by using a lipophilic dye (FM4-64) under the non-treated condition or 1/2 minimum inhibitory concentration of PMB treatment condition (N = 3 biological replicates). Asterisks indicate statistical significance at p<0.01. (D) Transmission electron micrographs (TEMs) of the experimentally evolved strains. Cells were grown to the exponential phase in LB liquid medium. (E) Detached OMVs from each strain were observed using TEM. OMVs were connected in a row in the PMRHigh strain.

Figure 1—source data 1

The minimum inhibitory concentration of experimentally evolved polymyxin B-resistant strains and outer membrane vesicle production.

https://cdn.elifesciences.org/articles/66988/elife-66988-fig1-data1-v2.xlsx

Analyses of modified lipid A and Raman spectroscopy

To determine whether modified lipid A plays an important role in PMB resistance, lipid A was subjected to matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) analysis in the positive-ion mode. The lipid A extraction method is summarized in Materials and methods. The main peak of Lab-WT was measured at 1490 m/z, corresponding to bis-phosphorylated penta-acylated lipid A (Figure 2A, B). In the PMRHigh strain, the predominant spectrum was also measured at 1490 m/z, followed by 1630 m/z, which is predicted to be the peak when a PetN and a hydroxyl group were attached to the dominant lipid A (Figure 2A, B). The PMRHigh strain overexpressed pmrC (encoding a PetN transferase), which added a PetN to some lipid A molecules, but not all, leading to interference PMB binding to the OM (see Figure 1—figure supplement 1B). Raman spectroscopy is used for microbial analysis at a single-cell level since it can rapidly measure the cell components, including lipids, nucleic acids, and proteins, by detecting the inelastic scattering of a molecule irradiated by monochromatic lights. The phenylalanine peak (1001–1004 cm−1) appears to be a signature signal for the presence of peptides and proteins in a sample. This phenylalanine peak at 1002 cm−1 appeared to be similar in both strains (Figure 2C). However, the Raman peak corresponding to PL (at 1334, 1445, and 1666 cm−1) increased notably in PMRHigh (Figure 2C). The intensity of the other prominent peak at 1243 cm−1 assigned to an amide III β-sheet was higher in resistant strains (Figure 2C). Increased magnitude of Raman intensity at amide III, along with hydroxyl, carboxyl, and phosphoryl groups, was reported in bacterial biofilm, indicating profound changes in the bacterial cell membrane (Jung et al., 2014; Kim et al., 2019). The Raman peaks of amide III were increased in clinical quinolone-resistant E. coli (Kim et al., 2019). Principal coordinate analysis (PCoA) using the collected Raman spectra of bacterial cells revealed that dispersion of individual dots was clustered into two groups, with 38.2% explanatory power, although some dots were in shared areas between Lab-WT and PMRHigh (Figure 2D). Individual cell heterogeneity analysis showed that Lab-WT strains were more heterogeneous than resistant strains, suggesting that PMRHigh strains were specifically adapted and unified by PMB stress (Figure 2E).

Mass spectrometry and proposed structures of lipid A isolated from Lab-WT and PMRHigh.

(A) Lab-WT produced peaks at an m/z value of 1490, corresponding to bis-phosphorylated penta-acylated lipid A. (B) PMRHigh produced peaks at m/z values of 1490 and 1630, corresponding to bis-phosphorylated penta-acylated lipid A without/with one phosphoethanolamine group and -OH, respectively. (C) Raman spectra of single A. baumannii cells grown to the exponential phase in LB media. 15–20 single spectra were used to produce mean spectra. (D) Principal coordinate analysis (PCoA) demonstrated separation between Raman spectra of Lab-WT and PMRHigh. (E) Heterogeneity of the PCoA dots was analyzed.

Integration of genomic, transcriptomic, and proteomic analyses

We hypothesized that the higher production of OMVs secreted by the PMRHigh strain functioned as a decoy target for PMB. In this scenario, OMVs are expected to protect their mother cells as well as the entire bacterial community from the effect of PMB. To gain a better insight into the mechanism underlying PMB resistance and OMV biogenesis in the PMRHigh strain, whole-genome sequencing (WGS) analyses of PMRLow and PMRHigh were conducted (Supplementary file 1b). It has been known that A. baumannii ATCC 17978 strain possesses a single chromosome and two plasmids, namely pAB1 (13,408 bp) and pAB2 (11,302 bp). The same small plasmid, pAB2 (11,299 bp, three bases were deleted), was detected only in the PMRHigh. The large plasmid was missing in both evolved strains (Figure 3—figure supplement 1A). The chromosome sizes of ATCC 17978, PMRLow, and PMRHigh were 3,976,747 bp, 3,971,618 bp, and 3,955,017 bp, respectively (Figure 3—figure supplement 1A). Gene alignments and comparative genomic analyses were performed on the reference ATCC 17978 strain (Lab-WT). The WGS data were assembled to construct draft genomes with median N50 values of 3971 kb for PMRLow and 3955 kb for PMRHigh (Supplementary file 1b). The numbers of coding DNA sequences (CDSs) in the PMRLow and PMRHigh genomes were 3678 and 3762, with mean CDS lengths of 947.5 and 930.2, respectively (Supplementary file 1b). Both MUMmer program-based genomic and additional PCR analyses were performed to confirm mutational positions (Supplementary file 1a, b). No mutation was detected in the promoter (all intergenic) regions of either mutant strain genome. Two chromosomal regions mostly annotated as hypothetical proteins were lost during adaptive evolution of PMRLow and PMRHigh strains (Supplementary file 1c). All 20 antibiotic-resistant genes and 78 IS elements remain unchanged in both evolved strains (Supplementary file 1d, e). We focused on PMRHigh rather than PMRLow because the first strain produced more OMVs than the latter. In the PMRHigh strain, mutations were found in the protein damage repair gene surE, fimbrial gene fimT, cell division-related gene ftsL (nonsense mutation), DNA repair gene udg, DNA integration gene xerD, and fatty acid synthesis-related gene fabH (nonsense mutation) (Supplementary file 1a, Figure 3—figure supplement 1A). Additionally, comparative genomic analysis between Lab-WT and PMRHigh strains revealed point mutations in the pmrB gene of the PMRHigh strain. PmrB was the protein that showed different mutations in the PMRLow and PMRHigh strains, with amino acid substitutions at N353I and T2351, respectively (Supplementary file 1a, Figure 3—figure supplement 1A). The mutations in N353I and T2351 in PmrB are known to increase the expression of the pmrC, resulting in PetN modification of lipid A (Adams et al., 2009).

Due to high MIC under PMB and hypervesiculation of the PMRHigh strain, we focused on comparative transcriptome between lab WT and PMRHigh strain (Supplementary file 1f). Our RNA-seq analysis showed that 250 and 166 genes were up- and downregulated, respectively, in PMRHigh in comparison with the Lab-WT strain without PMB exposure (Figure 3—figure supplement 1B). In addition, 89.16% of all genes showed no change at the transcriptional level. The highly upregulated gene, encoding an aldehyde dehydrogenase (aldB), showed a 3.43-fold increase in the PMRHigh in comparison with that in the control (Figure 3—figure supplement 1B). It is worth noting that dynamic membrane lipid biosynthesis in the PMRHigh might generate long-chain fatty aldehydes that may be recycled through AldB (Asperger and Kleber, 1991; Ishige et al., 2000). 8 of the 16 most upregulated genes were hypothetical proteins, and the remaining genes were annotated as a glutamin-(asparagin)-ase (ansB), poly-beta-1,6-N-acetyl-d-glucosamine N-deacetylase (pgaB), ribosomal protein (rpmI), and an ABC transporter glutamine-binding protein (glnH) (Figure 3—figure supplement 1B). In particular, the PgaB belonging to the pgaABCD operon functions for exopolysaccharide export across the OM. The overexpressed pga operon appeared to facilitate the production of PNAG-attached OMVs in E. coli through deacetylation of PNAG by the PgaB activity, which resulted in enhanced biofilm formation (Stevenson et al., 2018). The PMRHigh strain showed upregulation of both pmrA and pmrC genes encoding a transcriptional regulator and a PetN transferase, respectively, which evidently affects the membrane charge leading to PMB resistance (Figure 3—figure supplement 1B). Notably, the most downregulated genes except the hypothetical gene were related to two genes encoding the OMPs ompA and bamE (Figure 3—figure supplement 1B). OmpA, a PG-linked non-specific OM porin, harbors a flexible periplasmic domain that is strongly associated with PG (Iyer et al., 2018). Lack of OmpA in the OM is known to promote bacterial OMV production (Schwechheimer et al., 2014). The bamE, encoding a β‐barrel assembly machine E, was shown to bind specifically to phosphatidylglycerol (PG) in the OM (Knowles et al., 2011). Both OmpA and BamE are determinants for cell shape and OM integrity (Ryan et al., 2010; Choi and Lee, 2019). Downregulation of both genes may lead to loosening of OM-PG linkages and is associated with hypervesiculation in the PMRHigh. An additional qRT-PCR assay confirmed that the pmr and pgaB genes showed more than fourfold higher expression, but all membrane-linkage-related genes (ompA, bamE, lpp, and mlaC) were downregulated in PMRHigh (Figure 3A, Figure 3—source data 1). Biogenesis of OMV in the PMRHigh was twice as much as that in Lab-WT and further increased under 1/2 MIC PMB (Figure 1C). To identify the commonly expressed genes in the high OMV production condition, we compared three conditions (Lab-WT + PMB vs. Lab-WT), (PMRHigh + PMB vs. PMRHigh), and (PMRHigh vs. Lab-WT) (Figure 3—figure supplement 2). Our RNA-seq analyses confirmed that 29 and 12 genes were up- and downregulated, respectively, in the hypervesiculation conditions (Figure 3—figure supplement 2A). High levels of gene expression involved in stress response and survival were also predominant in the transcriptomic analysis, in which ribosomal proteins (rplW and rplO) and chaperons (groL) were upregulated in the hypervesiculation conditions (Supplementary file 1g). The pgpA gene encoding a lipid phosphatase was also upregulated, which can dephosphorylate phosphatidylglycerophosphate to form PG, a major constituent in OMVs (Supplementary file 1g, Lu et al., 2011). Increased levels of PG in the OM induced changes in membrane integrity, causing higher OMV production (Sohlenkamp and Geiger, 2016). Notably, several genes belonging to the pmr and pga operons are included among the commonly upregulated gene categories, and the commonly downregulated genes included important membrane-associated genes (mlaC, ompA, lpp, and bamE) (Supplementary file 1g).

Figure 3 with 2 supplements see all
Integration of omics analysis in experimentally evolved strains and outer membrane vesicles (OMVs).

(A) The qRT-PCR assay was conducted in PMRHigh strain compared with Lab-WT. The expression of the strains was normalized to their respective 16S rRNA expression. All qRT-PCR procedures were conducted in triplicate from at least three independent cultures. (B) Two-dimensional gel electrophoresis image analysis of proteins expressing isolated OMVs from Lab-WT and PMRHigh strains (Figure 3—source data 2). The proteins were stained by colloidal Coomassie Brilliant Blue and analyzed by Image Master Platinum 5.0 image analysis program. The selected proteins indicated up- or down-expressed proteins in Lab-WT.

Figure 3—source data 1

The qRT-PCR assay in PMRHigh strain compared with Lab-WT.

https://cdn.elifesciences.org/articles/66988/elife-66988-fig3-data1-v2.xlsx
Figure 3—source data 2

Images of selected spots in proteomics analysis (Lab-WT, Lab-WT + 1/2 minimum inhibitory concentration [MIC], PMRHigh and PMRHigh + 1/2 MIC).

https://cdn.elifesciences.org/articles/66988/elife-66988-fig3-data2-v2.docx
Figure 3—source data 3

The data of two-dimensional gel electrophoresis analysis for identifying differentially expressed proteins in our isolated outer membrane vesicles collected from experimentally evolved strains.

https://cdn.elifesciences.org/articles/66988/elife-66988-fig3-data3-v2.xlsx

Purified OMVs were separated by two-dimensional gel electrophoresis (2-DE) followed by Coomassie blue staining. Gel-based proteomics with MALDI-TOF analysis were performed under the same four conditions (Figure 3B, Supplementary file 1h, Figure 3—source data 2 and 3). 43 out of the 61 up- or downregulated proteins were designated with the MS-Fit program by using the NCBInr and UniPortKB databases, and 18 spots were not identified (Supplementary file 1h). The commonly identified OMV components harbored several proteins that were noted in our RNA-seq analyses for OMV biogenesis, showing increased production (PmrC, alanine racemase [Alr], and PgaB) and decreased production (OmpA and OmpW, elongation factor Tu [TufB]). Both our RNA-seq and proteomics analyses strongly indicated that simultaneously identified genes and proteins are critical components of OMV biogenesis in both wild-type and PMRHigh strains. Interestingly, A. baumannii OMVs also contained many cytoplasmic cargo proteins such as ribosomal proteins and Ef-Tu, similar to the OMVs from other bacteria (Dallo et al., 2012; Deo et al., 2018). Elongation factor (Ef-Tu), observed in our proteomics analyses, is one of the most abundant proteins (>20%) in bacteria and has a canonical role in translation. However, it is also found on the surface of bacterial OMs, which suggests a possible noncanonical function on the membrane (Widjaja et al., 2017; Harvey et al., 2019). Accumulation of Ef-Tu in the periplasmic space might induce pressure, which promotes blebbing of the OM resulting in OMV secretion although lower level of TufB was detected in PMRHigh strain (Figure 3B, McBroom and Kuehn, 2007). Reduced expression of OmpA and OmpW leads to diminished linkage between the OM and PG and subsequently causes more OMV generation in the PMRHigh. Our RNA-seq and proteomics data implied that PMRHigh changes the OM surface charge and membrane integrity by altering the expression of many OMV-related genes, which leads to PMB resistance and higher OMV production. Decreased membrane-linkage proteins induce overproduction of OMVs, which is also linked to enhanced biofilm formation.

Mutational analyses for OMV production

OMV production and MIC were measured in all strains, including several Campbell-type and CRISPR-Cas9-based knockout strains (Figure 4—figure supplement 1A). We used several approaches for quantifying OMV production (lipid staining [FM4-64 dye], quantification of total proteins [Bradford assay], total DNA measurement, and LPS quantitation using the purpald assay) (Figure 4, Figure 4—source data 1, Figure 4—figure supplement 1B). OMV production appeared to be 1.5- to 2-fold higher in PMRHigh than PMRLow or Lab-WT, as quantified by using the lipophilic dye FM4-64 (Figure 4A, Figure 4—source data 1). Total protein, DNA, and LPS amounts in OMVs collected from the same amounts of cell cultures were higher in PMRHigh than Lab-WT (Figure 4B–D). Furthermore, when the pmrB gene was deleted in Lab-WT, OMV biogenesis was reduced (0.3-fold) (Figure 4A–D), indicating that OMV production is impaired in the absence of PmrB. This finding was supported by our observation in which an extra copy of pmrBH gene increased OMV production in Lab-WT (Figure 4A–D). However, the ΔftsL, Δudg, Δlpp, and ΔmlaC mutants exhibited higher OMV production (Figure 4A–D). As expected from our omics data, OMV production was reduced when the PNAG transporter-related gene, pgaB, was deleted in Lab-WT (Figure 4A–D). The relative OMV production measured by FM4-64, protein, DNA, and LPS quantification in the ΔfabH and ΔsurE mutants could be ignored because no statistically significant difference was found (Figure 4—figure supplement 1C–F). Our mutational study clearly demonstrated that mutant strains lacking genes involved in membrane cross-linkages and PL management (lpp, ftsL, and mlaC) showed increased OMV secretion. The MICs of the pmrB-deleted strain (ΔpmrB) appeared to be reduced (Figure 4E). Introduction of the complementary pmrB gene to Lab-WT (Lab-WT + pRK-pmrB) increased the MIC, whereas the empty vector (pRK) did not affect the MIC (Figure 4—figure supplement 1B). The MICs of ΔftsL, Δudg, Δlpp, and ΔpgaB remained unchanged in comparison to that of Lab-WT (Figure 4E). The MICs of all constructed mutants were not affected, although they showed higher OMV production due to exposure to PMB at the beginning stage of culture conditions, in which no OMV production was expected. However, tendency for protecting cells grown in stationary phase from high PMB concentration (1/2 MIC of each clinical strain) was observed in our tested higher OMV producers (Figure 4F).

Figure 4 with 1 supplement see all
Polymyxin B (PMB) susceptibility test and outer membrane vesicle (OMV) production in experimentally evolved stains and mutants.

OMV production was analyzed for (A) lipids (lipophilic dye), (B) total protein (Bradford assay), (C) total DNA amount, and (D) lipopolysaccharide amount (purpald assay) in the exponential phase. Each measurement was normalized by the equivalent OD600. (E) The minimum inhibitory concentration (MIC) of PMB in each strain was tested. (F) Tendency for protecting cells grown in stationary phase from high PMB concentration (1/2 MIC of each clinical strain). No protection was observed in ΔpmrB. In all panels, an asterisk indicates statistical significance at p<0.05 (N = 3 biological replicates).

Figure 4—source data 1

Mutational analysis and quantification of outer membrane vesicle in mutants.

https://cdn.elifesciences.org/articles/66988/elife-66988-fig4-data1-v2.xlsx

Protective effect of OMVs against PMB

We hypothesized that the OMVs secreted by PMRHigh and mutant strains act as decoys to mask the effect of PMB on cells. To assess this hypothesis, we extracted OMVs from PMRHigh culture (Figure 5—figure supplement 1A) and confirmed their presence using flow cytometry (Figure 5—figure supplement 1B). Concentrated OMVs were negatively stained with uranyl acetate and visualized using TEM (Figure 5—figure supplement 1C). The protective effect of these PMRHigh-derived OMVs was assessed using the Luria–Bertani (LB) broth (Figure 5A, Figure 5—source data 1) and LB agar plate tests (Figure 5B), after confirming that additional purified OMVs had no effect on the bacterial growth. Interestingly, supplementation of PMB-containing liquid cultures (Figure 5A) or agar (Figure 5B) with these OMVs could increase the survival of all tested strains except for PMRHigh on agar plates. To visualize direct binding PMB to OMVs, we developed a dansyl fluorophore-PMB, which was synthesized by forming a chemical bond between the primary γ-amines on the diaminobutyric-acid residues of PMB and dansyl-chloride (Soon et al., 2011). Confocal laser scanning microscopy (CLSM) images revealed that dansyl-PMB was less bound to PMRHigh cells than Lab-WT cells at their respective PMB MICs (2 or 4 µg/mL) due to the fact that the PmrC-driven lipid A modification of PMRHigh cells decreased PMB binding to OMs (Figure 5C). The lipid-selective dye (FM4-64) and dansyl-PMB were employed to visualize binding of PMB to purified OMVs of the PMRHigh stain (Figure 5D). Our data showed that the two dyes exhibited overlapping images, which clearly demonstrated the direct binding of PMB to OMVs. To test whether A. baumannii clinical isolates could be protected from PMB through OMV secretion, the linkages between PMB protection rates and OMV production in 40 MDR clinical isolates were assessed (Figure 6A, Figure 6—source data 1). The degree of OMV production in each clinical isolate varied. Although direct correlations between the MIC and OMV production could not be identified under our tested conditions due to the same reasons we provided earlier, higher OMV producers could show higher protection rates against PMB in the early stationary phase (Figure 6A). Particularly, tendency was greater in the six clinical isolates with red-colored area (Figure 6A). Interestingly, addition of PMRHigh-derived OMVs to the growth cultures of five randomly chosen MDR strains (Supplementary file 1i; F-1025, F-1208, F-1379, F-1410, and F-1629) enhanced the microbial survival in the presence of PMB (Figure 6B). PMB MIC tests were performed using four additional Gram-negative bacteria (P. aeruginosa PAO1, Pseudomonas putida KT2440, E. coli K12, and Acinetobacter oleivorans DR1) and the protective roles of their OMVs were tested (Figure 6—figure supplement 1). Our data suggested that OMV-driven protection under PMB conditions could be applicable to other Gram-negative bacteria.

Figure 5 with 1 supplement see all
Protective effects of outer membrane vesicles (OMVs) in A. baumannii measured in vitro and visualization of the decoy role of OMVs.

(A) The results are shown after 72 hr. The Lab-WT, PMRLow, and PMRHigh strains were used to measure the protective effects of OMVs (N = 3 biological replicates). (B) Protective effects were measured after the addition of cell suspensions in LB medium with/without OMVs to polymyxin B (PMB)-containing agar. Images were obtained after 14 hr of incubation at 37°C. (C) Confocal laser scanning microscopy images of Lab-WT and PMRHigh treated with dansyl-PMB for 30 min at 37°C. (D) Isolated OMVs from PMRHigh were stained with FM4-64 and dansyl-PMB.

Figure 5—source data 1

The Lab-WT, PMRLow, and PMRHigh strains were used to measure the protective effects of outer membrane vesicles after 72 hr.

https://cdn.elifesciences.org/articles/66988/elife-66988-fig5-data1-v2.xlsx
Figure 6 with 1 supplement see all
Outer membrane vesicle (OMV) production and protective effects in clinical isolates.

(A) Minimum inhibitory concentration value and OMV production were examined in 40 MDR clinical A. baumannii strains. Blue dots show the protection rate for treatment of 1/2 MIC polymyxin B (PMB) in each clinical isolate, and the bar graph shows the amount of OMVs in the early stationary phase (OD600 = 0.4). The colored area indicates degree of correlation between OMV production and protection rate in stationary phase-grown cells. Red area: strong correlation; green area: moderate correlation (N = 3 biological replicates). (B) Protective effects of OMVs in five clinical A. baumannii strains (F-1025, F-1208, F-1379, F-1410, and F-1629) measured using LB liquid medium without and with 1 μg/mL PMB after treatment with OMVs. The OD600 of the strains was measured in the wells of 96-well plates until 72 hr of incubation at 30°C (N = 3 biological replicates).

Figure 6—source data 1

Outer membrane vesicle production and protective effects in clinical isolates.

https://cdn.elifesciences.org/articles/66988/elife-66988-fig6-data1-v2.xlsx

Protective effects of OMVs in in vitro human gut microbiota

We next tested whether OMVs could mask the effect of PMB on the entire bacterial microbiota. To this end, we used human fecal microbiota and employed the survival of Galleria mellonella as an infection model (Figure 7). The fecal microbiota was incubated for 5 days under anaerobic conditions to verify the extended protective effect of OMVs. Cultivation was continuously performed after dividing the sample into three groups (a control group without PMB, a second bottle with PMB, and a third bottle with PMB after addition of OMVs 3 hr earlier). PMB and OMVs were added daily for 5 days. The environmental MIC of PMB was determined and 1/2 environmental MIC was used in the following experiments (~50% reduction of aerobic bacterial microbiota in 24 hr) (Figure 7—figure supplement 1A). After 5 days, the community cells were spotted on an LB plate and cultured under aerobic conditions to check PMB toxicity and the protective effect of OMVs (Figure 7—figure supplement 1B). The PMB-treated sample showed a 104-fold reduction in aerobic bacterial cells than the control, and the OMV plus PMB sample showed similar levels of cell counts as the control (Figure 7—figure supplement 1B).

Figure 7 with 1 supplement see all
Protective effects of outer membrane vesicles (OMVs) on human fecal microbiota and in G. mellonella infected with A. baumannii strains.

(A) The stacked bar plot shows the relative abundance of the microbiome at the genus level. The color keys indicate the 24 most abundant taxa. OMVs (25 μL/mL) were first added and allowed to stabilize, followed by incubation at 37°C after the addition of 4 μg/mL polymyxin B (PMB) for 3 hr. (B) The α-diversity analyses using rarefaction curves of operational taxonomic units for in vitro human fecal microbiota. (C) The principal coordinate analysis 3D plot of in vitro human fecal microbiota using log2-transformed Bray–Curtis distances. The distance between symbols on the ordination plot indicates the relative similarity in their community structures. (D) The growth curves of Enterococcus spp. under aerobic or anaerobic conditions. PMB treatment (4 μg/mL) was conducted daily for 5 days in identical conditions for analysis of the microbiome community (N = 3 biological replicates). (E) The G. mellonella-infection model was constructed to confirm the protective effects of OMV in vivo. 10 μL of OMVs, 106 CFU/mL of Lab-WT or PMRHigh, and 1 μg/mL PMB were mixed and injected through a syringe. For groups without OMVs, the same volume of phosphate-buffered saline was used. Mortality was defined as a lack of response or melanization for 96 hr in infected G. mellonella at 37°C (N = 3 biological replicates).

Figure 7—source data 1

Protective effects of outer membrane vesicles on human fecal microbiota and growth curve of Enterococcus spp. under aerobic or anaerobic conditions.

https://cdn.elifesciences.org/articles/66988/elife-66988-fig7-data1-v2.xlsx

Bacterial community analysis was conducted by extracting DNA from each sample anaerobically cultured for 5 days to determine the composition of the protected in vitro microbiota. Our taxonomic profile, along with rarefaction curves and the PCoA 3D plot, revealed that addition of OMVs extracted from the PMRHigh strain protected the bacterial community from PMB under our tested anaerobic conditions, resulting in recovery of a similar microbial composition (Figure 7A–C, Figure 7—source data 1). The total 16S rDNA genes per dried gram in the PMB-added sample (145,311 copies/dried gram) decreased by ~16% in comparison with that in the 5 days incubated control sample (173,403 copies/dried gram), while the corresponding value in the OMV-added sample was reduced by 14% (149,643 copies/dried gram) (Figure 7A). Surprisingly, the PMB-exposed sample showed significantly reduced numbers of 16S rDNA copies corresponding to the Enterococcus faecium group belonging to Gram-positive bacteria (75% lower than the control) and recovery of the E. faecium group occurred in the OMV-added PMB sample (Figure 7A). No growth inhibition with PMB was observed in both aerobically grown Enterococcus type strains (E. faecium and Enterococcus faecalis). Interestingly, severe growth retardation with PMB was especially observed in the anaerobic culture of E. faecium (Figure 7D, Figure 7—source data 1). Unlike aerobically grown cells, anaerobically grown Enterococcus cells might show the effects of PMB toxicity on membrane charges and composition, which warrants further examination. The survival of G. mellonella larvae infected with Lab-WT or PMRHigh was assessed in the presence or absence of OMVs (Figure 7E). Larvae were infected with either Lab-WT (106 CFUs/mL) or PMRHigh with or without PMB (1 μg/mL) after OMV addition. Comparison of the mortality rates demonstrated that PMRHigh, which possessed higher virulence than Lab-WT, induced faster death of all tested larvae within 72 hr (Figure 7E). In the presence of PMB, only 10% of the G. mellonella larvae infected with PMRHigh died within 24 hr, but 80% of the larvae died when OMVs were included (Figure 7E). This experiment suggested that A. baumannii survived longer due to the protective effect of OMVs against PMB, leading to the higher mortality of G. mellonella. Collectively, OMVs can protect the human pathogen A. baumannii from PMB under in vivo and in vitro conditions by functioning as decoys for PMB binding.

Discussion

Production of OMVs has been linked to many important cellular behaviors, including quorum-sensing, AR, and biofilm formation in pathogenic bacteria (Florez et al., 2017). OMV biogenesis was thought to be also linked to the elevated hydrophobicity of the cell surface in P. putida, which caused greater biofilm formation (Baumgarten et al., 2012). OMVs are known to function as cargoes for many extracellular or intracellular materials such as virulence factors, proteins, DNA, RNA, and signaling molecules (Bonnington and Kuehn, 2016). Interestingly, the OMVs produced by A. baumannii have been shown to have a virulence factor (OmpA) and an antibiotic-resistance factor (AmpC) (Weber et al., 2017). OMVs are composed of PLs, OMPs, periplasmic proteins, and LPS; however, the packaging of cytosolic molecules into OMVs remains unclear (Roier et al., 2016a; Jan, 2017). Interestingly, OMVs released from CRAB could be responsible for the horizontal transfer of the carbapenem-resistance-related gene blaNDM-1 to carbapenem-susceptible A. baumannii (Chatterjee et al., 2017). Notably, Klebsiella pneumoniae strains had OMs and larger OMVs harboring different lipid compositions under PMB treatment, which might contribute to the low permeability of PMB (Jasim et al., 2018). The following factors have been proposed to modulate OMV biogenesis: (i) diminished linkage proteins between OM and PG; (ii) increased turgor pressure due to the accumulation of cellular materials in the periplasm; (iii) modulation of the production of transmembrane proteins; and (iv) changes in membrane fluidity and rigidity (Toyofuku et al., 2015). The bacterial membranes of Gram-negative bacteria are also constantly modified via two-component systems, including PmrAB, in response to changes in the environment such as metals and PMB (Tsang et al., 2017). LptA is the periplasmic protein of the LPS transport (Lpt) complex, which bridges the IM and OM, transporting LPS across the periplasm (Powers and Trent, 2019). A recent study has shown that reduced LptA connections lead to overproduction of OMVs (Falchi et al., 2018). Decreased Lpp-PG covalent cross-linking or non-covalent OmpA-PG interactions also enhance the production of OMVs (Schwechheimer and Kuehn, 2015). Disruption of the Mla complex system, which transports mislocalized OM glycerophospholipids back to the IM, results in hypervesiculation by accumulation of glycerophospholipids in the outer leaflet (Kamischke et al., 2019). The Bam complex is an OM complex that mediates the folding and insertion of OM proteins into the membrane (Malinverni and Silhavy, 2011). DolP, a lipoprotein, is known to promote proper folding of BamA, one of the essential subunits of the BAM complex, which might also play a role in OM integrity (Ranava et al., 2021). Deletion of bamE in E. coli or in Salmonella enteritidis results in a more severe defect in OM and lower OM protein levels than does loss of bamC (Sklar et al., 2007; Fardini et al., 2009).

Taken together, the accumulated data suggest that OMVs can be generated when decreased cross-links between the OM, PG, and IM occur in the bacterial membrane. Our omics and mutational analyses substantiated that many genes involved in the structural components of OM and IM, such as ompA, lpp, mlaC, and ftsL, play important roles in OMV biogenesis (Figures 3 and 4). However, it is still not clear how the DNA repair-related udg gene is linked to OMV production (Figure 3). Linkage between those deleted regions in evolved strains and OMV biogenesis remains to be investigated (Figure 3—figure supplement 1A). It is worth noting that antibiotic-induced genomic alterations including deletions, rearrangements, and amplification have been reported in many laboratory-evolved antibiotic-resistant bacterial strains (Sandegren and Andersson, 2009). Several laboratory-evolved antibiotic-resistant E. coli strains also appeared to carry more than 20 mutations (Maeda et al., 2020). Tetracycline-induced genome rearrangements led to the loss of 6.1 kb having seven full-length and two truncated genes in E. coli strains (Hoeksema et al., 2018). Norfloxacin-treated E. coli had ~11.5 kb deletions possibly through IS1-mediated recombination (Long et al., 2016). Deducing a direct link between OMV production and a single OMV-related gene might be an oversimplification because of the multifaceted control of bacterial OMV production; however, our expression analyses clearly demonstrated that the hypervesiculating PMRHigh strain showed reduced expression of many OMV-related genes, which might cumulatively affect OMV production. Several other data have also shown that bacterial OMVs contained IM proteins as well as cytoplasmic proteins (Nagakubo et al., 2019). The unusual OMV biogenesis recently identified in a P. aeruginosa model system has clearly shown that OMV production through explosive cell lysis ended up including several cytosolic components of cells (Schwechheimer and Kuehn, 2015). Explosive cell lysis induced by phage-derived endolysin burst the outer and IMs, leading to the presence of cytoplasmic proteins and DNA in OMVs (Turnbull et al., 2016). Two known mechanistic factors (PQS and MucD) for OMV production were not required for cycloserine-induced OMV formation in P. aeruginosa, indicating the existence of multiple pathways for OMV biogenesis (Macdonald and Kuehn, 2013).

OMV biogenesis in PMRHigh causes PMB resistance via the decoy activity of OMVs. OMVs have been shown to protect bacterial cells against several classes of antibiotics and physical and chemical stresses (Kulkarni et al., 2015; Lee et al., 2016). The decoy activity of OMVs was proposed in an E. coli model under membrane-active antibiotics such as colistin and melittin (Kulkarni et al., 2015; Kim et al., 2018). Our study also supported the role of OMVs as decoys under PMB conditions in A. baumannii (Figure 5A, B), clinical isolates (Figure 6B), and in the microbial community (Figure 7A). The in vitro human microbiota experiment clearly showed that the presence of OMVs can modulate the bacterial community under PMB conditions. The reasons underlying our unexpected observation of sensitivity of the Gram-positive Enterococcus strains to PMB under anaerobic conditions remain to be elucidated (Figure 7A). Recently, more studies suggested that polymyxins can bind to teichoic acid, which is a negatively charged compound in the OM of Gram-positive bacteria (Brown et al., 2013; Rudilla et al., 2018; Yu et al., 2020). We speculated that different degrees of teichoic acid-like production might occur in anaerobic culture conditions. Our G. mellonella-infection experiments also suggested the protective role of OMVs on A. baumannii cells against PMB during infection (Figure 7C). Likewise, Vibrio cholerae cells use hyperproduction of OMVs to increase their colonization and adaptation during mammalian infection (Zingl et al., 2020). Interestingly, the OMVs induced by the pga operon can stimulate the production of broad antimicrobial antibodies against P. aeruginosa in a mouse infection model (Stevenson et al., 2018). In conclusion, our data suggested that greater attention must be paid to develop new antibiotics and vaccines because the multifaceted OMV production in pathogenic A. baumannii strains warrants extensive investigations.

Materials and methods

Key resources table
Reagent type
(species) or
resource
DesignationSource or
reference
IdentifiersAdditional
information
Strain, strain background (Escherichia coli)DH5aThermo Fisher Scientific18265017
Strain, strain background (Escherichia coli)S17-1λ pirThis laboratory
Strain, strain background (Acinetobacter baumannii ATCC 17978)A. baumannii Lab-WTThis study
Strain, strain background (Acinetobacter baumannii Lab-WT)A. baumannii PMRLowThis study
Strain, strain background (Acinetobacter baumannii Lab-WT)A. baumannii
(pRK-pmrB)
This study
Strain, strain background (Acinetobacter baumannii Lab-WT)A. baumannii ΔpmrBThis studyCRISPR-Cas9 method
Strain, strain background (Acinetobacter baumannii Lab-WT)A. baumannii ΔftsLThis studySingle-crossover recombination method
Strain, strain background (Acinetobacter baumannii Lab-WT)A. baumannii ΔlppThis studySingle-crossover recombination method
Strain, strain background (Acinetobacter baumannii Lab-WT)A. baumannii ΔmlaCThis studyCRISPR-Cas9 method
Strain, strain background (Acinetobacter baumannii Lab-WT)A. baumannii ΔpgaBThis studyCRISPR-Cas9 method
Strain, strain background (Acinetobacter baumannii Lab-WT)A. baumannii ΔudgThis studySingle-crossover recombination method
Strain, strain background (Acinetobacter baumannii Lab-WT)A. baumannii ΔfabHThis studyHomologous recombination method
Strain, strain background (Acinetobacter baumannii Lab-WT)A. baumannii ΔsurEThis studyHomologous recombination method
Recombinant DNA reagentpVIK112
(Plasmid)
This studyAddgene: Plasmid
#V002240
Produced by De Schamphelaire W et al.
Recombinant DNA reagentpRK415
(Plasmid)
This studyLife Science Market
Plasmid # PVT5709
Produced by Keen NT et al.
Recombinant DNA reagentpSGAb-kmWang et al., 2019Addgene: Plasmid
#121999
Kmr, ColE1, WH1266
Recombinant DNA reagentpCasAb-aprWang et al., 2019Addgene: Plasmid #121998Aprr, oriV, broad-host-range vector
Commercial assay or kitRNeasy Mini KitQIAGEN74104Accordance with the instructions of the manufacturer
Commercial assay or kitFastDNA spin kitMP Biomedicals116540600-CFAccordance with the instructions of the manufacturer
Chemical compound, drugPolymyxin B (PMB)Sigma-AldrichP4932-1MUTargets LPS
Chemical compound, drugKanamycinSigma-AldrichB5264Targets RNA
Chemical compound, drugApramycinSigma-AldrichA2024-1GTargets RNA
Chemical compound, drugIPTGGold BioI2481CInduces gene expression
Chemical compound, drugFM 4-64Thermo Fisher ScientificT3166Fluoresces when bound to phospholipids
Chemical compound, drugDanysl-chlorideSigma-AldrichD2625-1G
Chemical compound, drugFilmTracer SYPRO RubyInvitrogenF10318
Software, algorithmCLC Genomics Workbench v.10.0.1QIAGEN
Software, algorithmMUMmerhttp://mummer.sourceforge.net/
Software, algorithmIS finderhttps://isfinder.biotoul.fr/blast.php
Software, algorithmEzGenomehttp://ezgenome.ezbiocloud.net/ezg_browse
Software, algorithmSnapgenehttps://www.snapgene.com/

Experimentally evolved stains, clinical isolates, and determination of PMB susceptibility

Request a detailed protocol

The bacterial strains used in this study are listed in Supplementary file 1i. We named A. baumannii ATCC 17978 wild-type strain as Lab-WT and constructed PMB-resistant strains (PMRLow and PMRHigh) from Lab-WT. All the strains were grown at 37°C in LB broth with aeration by shaking at 220 rpm. The PMRLow and PMRHigh strains were constructed by transferring wild-type cells onto LB agar plates with serially higher PMB concentrations. In total, 3 A. baumannii experimentally evolved isolates and 40 clinical isolates were investigated. The clinical isolates were obtained from the Samsung Medical Center, Sungkyunkwan University School of Medicine, and National Culture Collection for Pathogens (NCCP). The complete genome sequence of the A. baumannii ATCC 17978 strain can be accessed in GenBank (accession number: CP000521, CP015122.1). Sequence reads for the experimentally evolved isolates are accessible from the NCBI sequence archives under accession numbers PRJNA530195 (Lab-WT), PRJNA530197 (PMRLow), and PRJNA530202 (PMRHigh). Nine different classes of antibiotics, including meropenem, rifampicin, chloramphenicol, and ampicillin, were selected to test the AR. The MICs of each antibiotic were examined in all tested cells (Lab-WT, PMRLow, PMRHigh, P. aeruginosa, P. putida, E. coli, and A. oleivorans) with the broth-dilution method using 96-well plates, as previously described (Pérez-Cruz et al., 2016). For determination of protection tendency with OMV in clinical isolates, 40 clinical A. baumannii isolates were grown at 37°C in LB broth at 220 rpm until each culture reached an OD600 of 0.8–1.0. The cell pellet was washed with phosphate-buffered saline (PBS) twice and each collected cells (106 CFU/mL) was inoculated separately into LB medium with or without 1/2 MIC of each isolate for 1 hr at 37°C. After grown cells were serially diluted to get 10−4 dilution with PBS, 10 µL of suspension spotted on an LB agar plates and incubated for 16 hr at 37°C. The protection rate (%) was calculated survived colonies (CFU) in comparison with untreated conditions of each strain.

Quantitative reverse transcription-PCR

Request a detailed protocol

The expression levels of our target genes were normalized using the expression level of 16S rDNA, as previously described (Shin et al., 2020). The total RNAs from different A. baumannii strains were isolated from 5 mL of cell cultures in the mid-exponential phase (optical density at 600 nm [OD600] of approximately 0.4) using the RNeasy Mini Kit (QIAGEN, Germany). For qRT-PCR, the cDNA was synthesized from 3 μg of RNA using the primers listed in Supplementary file 1j. All qRT-PCR procedures were conducted in triplicate from at least three independent cultures. The relative expression levels were compared with those of the Lab-WT strains.

Measurement of biofilm formation

Request a detailed protocol

Biofilms were grown in confocal dishes (SPL Life Sciences, South Korea) at 37°C for 24 hr in LB broth. The biofilm cells were stained with FilmTracer SYPRO Ruby (Invitrogen, USA) for 30 min at room temperature (RT), protected from light, and then washed with distilled water. The obtained CLSM (Carl Zeiss, Germany) images were analyzed and modified using the Zen 2.1 (Blue edition; Carl Zeiss, Germany) software. LB broth in sterile 96-well microtiter plates (SPL Life Sciences) was inoculated in triplicate with each overnight LB-grown culture and further diluted 1:100 with LB broth. The volume of the cells was determined by converting the OD600 value of the O/N cells. Uninoculated LB broth was used as the negative control. The microtiter plates were then incubated at 37°C for 24 hr. After removing planktonic cells, the biofilm biomass was stained with crystal violet and solubilized with 95% ethanol (v/v), after which its absorbance was measured at 595 nm. To measure OMVs-induced initial stage of biofilm formation in Lab-WT, cells were grown at 37°C in LB until an OD600 reached around 0.6 (for 6 hr). Biofilm formation was checked after additional 4 hr incubation with PMRHigh-driven OMVs (0, 12.5, or 25 μg/mL), then both crystal violet assay and CLSM observation were used.

Lipid A extraction and analysis with MALDI-TOF mass spectrometry

Request a detailed protocol

A. baumannii strains were grown at 37°C in LB broth until each culture reached an OD600 of 1.0. Cells were harvested via centrifugation at 7800 rpm for 15 min. Then, the cell pellet was washed with 30 mL of PBS. The supernatant was poured off and resuspended in 35 mL of a single-phase Bligh–Dyer mixture (chloroform:methanol:distilled water, 1:2:0.8; v/v). The solution was mixed by inversion and incubated at RT for >20 min to ensure complete cell lysis. Then, the mixture was centrifuged at 2000 rpm for 20 min. LPS will pellet along with proteins and nucleic acids; however, PLs, isoprenyl lipids, and small, hydrophobic peptides remain in the supernatant. The supernatant was discarded. The LPS pellet was washed with 35 mL of the single-phase Bligh–Dyer mixture and centrifuged at 2000 rpm for 20 min. The supernatant was discarded; 8 mL of mild acid hydrolysis buffer (50 mM sodium acetate, pH 4.5; 1% SDS) was added to the pellet, and it was mixed by pipetting. The solution was sonicated at a constant-duty 2× cycle for a 20 s/burst, with an interval of 5 s between bursts, at 50% output. Then, the sample was boiled in a water bath for 30 min, and the bottles were removed from the water bath and allowed to cool to RT. To extract lipids after hydrolysis, the mixture was formed by adding 9 mL of chloroform and 9 mL of methanol to the SDS solution. Then, the sample was mixed by inversion and centrifuged for 10 min at 2000 rpm. The lower phase was extracted into a clean 50 mL conical tube. A second extraction was performed by adding 30 mL of the lower phase from a pre-equilibrated two-phase Bligh–Dyer mixture (chloroform:methanol:DW, 2:2:1.8; v/v) to the upper phase from the previous step. The mixture was centrifuged at 2000 rpm for 10 min. The lower phase was extracted and pooled with the lower phase from the previous step. Then, the pooled lower phase (18 mL total) was washed by adding 34.2 mL of pre-equilibrated two-phase Bligh–Dyer upper phase to create a two-phase Bligh–Dyer mixture (chloroform:methanol:DW, 2:2:1.8; v/v). The solution was mixed and centrifuged at 2000 rpm for 10 min. The lower phase was removed to a clean glass rotary evaporator flask and the sample was dried using rotary evaporation. The lipid sample was added to 1.5 mL of chloroform:methanol (4:1, v/v) in a rotary flask, and ultrasonicated (40 s) to facilitate suspension of lipid from the sides of flask. The lipid A was dried in a nitrogen dryer and transferred to small glass sample vial. The dried sample was stored at 4°C. Lipid A was analyzed by MALDI-TOF mass spectrometry in the positive-ion mode. The matrix was a saturated solution of 2,5-dihydroxybenzoic acid (DHB). A 10 μL volume of the matrix solution was deposited on the sample plate, followed by 10 μL of the sample dissolved in chloroform:methanol (4:1, v/v).

WGS and data analysis

Request a detailed protocol

WGS analyses were performed on three selective A. baumannii isolates using the PacBio sequencing technique (ChunLab, South Korea). The genome sequences were downloaded from EzGenome (http://ezgenome.ezbiocloud.net/ezg_browse) and were analyzed using the CLgenomics program (ChunLab). Long-read NGS sequencing techniques, such as the PacBio used in this study, may produce sequencing errors (Rhoads and Au, 2015). To compensate for this shortcoming, additional PCR reactions using Pfu DNA polymerase were performed to confirm mutational positions listed in Supplementary file 1a (18 genes). The primers used for PCR reactions are listed in Supplementary file 1k. Genomes were analyzed by using ClustalW 2.0 software, while neighbor-joining trees and amino acid compositions were constructed and calculated using MEGA6 software. The origin of replication (oriC) was identified using Ori-Finder (http://tubic.tju.edu.cn/Ori-Finder/). In addition to the amino acid sequence of PmrB, the LPS/lipid A biosynthesis components were compared between the isolates and the reference strain ATCC 17978. Antibiotic-resistant genes were identified using the CLC Genomics Workbench v.10.0.1 (QIAGEN). We used the Find Resistance tool (Microbial Genomics Module) with the QMI-AR, CARD, ResFinder, and PointFinder. All parameters were set as default. Insertion sequence (IS) elements were analyzed with the IS finder (https://isfinder.biotoul.fr/blast.php).

RNA extraction, library construction, and sequencing

Request a detailed protocol

Both Lab-WT and PMRHigh strains were grown to exponential phase (OD600 ~ 0.5) in LB media. For antibiotic treatment conditions, both strains were grown to the exponential phase (OD600 ~ 0.25) in 1/2 MIC (1 µg/mL or 64 µg/mL, respectively) of PMB-supplemented LB media. Total RNA was isolated from 10 mL of cells by using the RNeasy Mini Kit (QIAGEN) according to the manufacturer’s instructions. All procedures for RNA sequencing were conducted by ChunLab. The RNA was subjected to a subtractive Hyb-based rRNA removal process using the MICROBExpress Bacterial mRNA Enrichment Kit (Ambion, USA). RNA sequencing was performed with two runs of the Illumina Genome Analyzer IIx to generate single-ended 100 bp reads. Quality-filtered reads were aligned to the reference genome sequence using the CLC Genomics Workbench v.10.0.1 (QIAGEN). Mapping was based on a minimal length of 100 bp with an allowance of up to two mismatches. Relative transcript abundance was measured in RPKM. The RNA-seq data have been deposited in NCBI under Gene Expression Omnibus (GEO) accession number GSE163581.

A. baumannii mutant construction

Request a detailed protocol

To construct ΔftsL, Δudg, and Δlpp mutants by homologous recombination, target genes were disrupted using a single-crossover recombination method with several vectors, such as pVIK112 and pRK415, as previously described (Shin and Park, 2015). The primers used in this study are listed in Supplementary file 1j. In general, EcoRI and KpnI restriction enzymes were used for all the knock-out and overexpression mutants. ΔpmrB is a pmrB mutant of Lab-WT strains. To construct the mutants, fragments from PCR amplification and gel extraction were inserted into the pVIK112 vector via ligation as the first step. The pVIK112-fragment unified vectors were then transformed into E. coli S17-1λ pir. Finally, the amplified plasmids were inserted into the Lab-WT and PMRHigh strains of A. baumannii ATCC 17978 by electroporation. To ensure that homologous recombination had occurred in A. baumannii ATCC 17978, PCR verification was conducted using the pmrB OE-F/MCS-R primer pairs. The MCS-R primer was designed based on the sequence of the pVIK112 plasmid. To construct ΔpmrB, ΔmlaC, and ΔpgaB by the CRISPR-Cas9 method, the pCasAb-apr (cat. no 121998) and pSGAb-km (cat. no 121999) plasmids were obtained from Addgene (Wang et al., 2019). A pair of 20 bp spacer oligos and 80-nt oligos were designed to target the genomic locus and donor repair template for editing. A designed 20 bp spacer was phosphorylated and annealed to the pSGAb-km plasmid with a Golden Gate assembly reaction. Then, the plasmids were transformed into E. coli DH5a-competent cells, followed by plating onto an LB agar plate containing 100 µg/mL, and incubated at 37°C for 16 hr. Successful cloning of the spacer was verified by PCR with the primers of Spacer-F/M13R and by sequencing with the primer of M13R (5′-CAGGAAACAGCTATGACC-3′). Then, 1 M of IPTG was added into A. baumannii Lab-WT harboring pCasAb-apr to induce the expression of the RecAb recombination system and Cas9 nuclease. After incubating at 37°C for 2 hr, the same method was used to prepare the competent cells as described previously (Wang et al., 2019). Then, 200 ng of the spacer-introduced pSGAb-km plasmid and 300 µM ssDNA (donor repair template) were co-transformed to the IPTG-induced A. baumannii Lab-WT cells harboring the pCasAb-apr plasmid by electroporation. The cells were plated onto an LB agar plate containing 100 µg/mL apramycin and 50 µg/mL kanamycin, and the plate was incubated at 37°C overnight. Successful editing was verified by PCR and sequencing. For plasmid curing, a colony was plated onto an LB agar plate containing 5% sucrose at 37°C overnight. Then, the colony were streaked onto LB agar plates without the antibiotics to confirm curing. A detailed genome-editing protocol has been provided in the file (Figure 4—figure supplement 1).

Transmission electron microscopy

Request a detailed protocol

TEM was conducted at the Korea University Medical Research Center (https://medicine.korea.ac.kr/web/msrc/home). The experimental procedure was in accordance with the instructions of the manufacturer. Briefly, for bacterial cell imaging, cells were grown for 18 hr in LB media. Aliquots were transferred onto carbon-coated copper grids (Formvar; Ted PELLA, Canada) and fixed with 2.5% glutaraldehyde diluted in 50 mM sodium cacodylate (pH 7.2). After washing three times with 3% saccharose, the cells were observed. For OMV imaging, aliquots of purified OMV samples were dispensed on carbon-coated copper grids. Excess liquid was discarded from the grids and the samples were negatively stained with phosphotungstic acid (3% [wt/vol]) for 5 min and then dried with filter paper. The specimens were examined using an H-7100 microscope (Hitachi, Tokyo, Japan) operating at an accelerating voltage of 75 kV and at magnifications of 40,000× for cells and 30,000× for OMVs.

OMV quantification

Request a detailed protocol

The lipid content of the OMV samples was indirectly quantified using the lipophilic dye FM4-64 (N-(3-triethylammoniumpropyl)−4-(4-(dibutylamino) styryl) pyridinium dibromide; Molecular Probes, Eugene, OR, USA, and Life Technologies, Carlsbad, CA, USA). The lipophilic dye FM4-64, which intercalates into the OM, is commonly used to stained and observed membrane of bacteria (Toyofuku et al., 2017; Rojas et al., 2018). To measure OMVs using this dye, cells were incubated for 16 hr at 37°C and 220 rpm in 5 mL of LB. The cultured cells were harvested, and the supernatant was filtered through a 0.22 μm syringe filter. Aliquots of the filtrate (100 μL) containing OMVs were transferred to the wells of a 96-well dark plate. The FM4-64 dye was added to each well at a final concentration of 1 μg/mL. After excitation at 510 nm, the fluorescence emission at 610 nm was measured using a fluorescence spectrophotometer (TECAN, Männedorf, Switzerland) using 5 nm slit widths for both excitation and emission. To quantify the protein contents of OMVs, all the samples were standardized using the Bradford assay (Bio-Rad Laboratories, Protein Assay Dye Reagent), and triplicates were used. For the DNA contents of OMVs, the NanoPhotometer (Implen, Germany) was used to quantify the concentration of DNA. To quantify the LPS content of OMVs, purpald assays using the 2-keto-3-deoxyoctonate ammonium salt (Kdo) standard (Sigma-Aldrich) were performed as previously described (Lee and Tsai, 1999).

Raman spectroscopy analysis

Request a detailed protocol

Isolates were grown in LB broth for the exponential phase at 37°C, and the cell pellet was washed with PBS twice. Each bacterial suspension was dropped into an aluminum dish (width, 1.75 in; depth, 0.38 in) for spectral collection. After drying the plate for 2 hr, each spot was washed with 10 µL of PBS. Raman spectra of the isolates were collected using a confocal Raman imaging system (XperRam35V; Nanobase) assembled with 3-port excitation 532 nm DPSS laser (LTL-532RL; Leading Tech), microscope body (Olympus BX43; Olympus), spectrometer (XPE-35 VPHG; Nanobase), and charge-coupled device (Atik 428EX; Atik). Each spectrum was the sum of 25 s of acquisition time with 2 mW of laser power and collected for 15–20 single cells of the A. baumannii isolate.

Isolation of pure OMVs and flow cytometric analysis for detecting OMVs

Request a detailed protocol

The procedure used to isolate the OMVs is presented in Figure 5—figure supplement 1. OMVs were isolated from PMRHigh bacteria. The isolation was confirmed using FM4-64 staining and flow cytometric analysis. To extract OMVs from PMRHigh strains, 600 mL of cells were grown for 16 hr to obtain a robust and quantifiable amount of OMVs. Overnight bacterial culture was grown to an OD600 of 0.8–1.2 and the culture was centrifuged at 7800 rpm at 4°C for 30 min. The supernatant of each culture was sequentially filtered through 5 μm hydrophilic polyvinylidene difluoride and 0.45 μm Millex membrane filters (Millipore, Billerica, MA, USA). The resulting filtrates were concentrated with centrifugation (7800 × g) at 4°C for 30 min using a 10 kDa molecular weight cutoff Amicon Ultra-15 centrifugal filter unit (Millipore). Each concentrated filtrate was centrifuged (100,000 × g, 4°C, 3 hr) in a tabletop Optima ultracentrifuge (Beckman Coulter). Pelleted OMVs were suspended in 100 μL of PBS. The OMVs were quantified by using Bradford assay. The confirmed OMV pellet was spread on LB agar plates to confirm the cell-free status. The OMVs were counted by using a flow cytometer (BD Accuri C6 Plus). The forward scatter threshold value was 10,000. Three independent experiments were conducted for each condition, with 50,000 cells typically analyzed per experiment.

Synthesis of dansyl-PMB and measurement of PMB binding affinity by CLSM

Request a detailed protocol

PMB sulfate (40 mg, Sigma-Aldrich) was dissolved in 1.2 mL of 0.1M NaHCO3, and dansyl-chloride (10 mg, Sigma-Aldrich) was dissolved in 0.8 mL of acetone. Dansyl-chloride was added to PMB and placed in the dark for 90 min at RT. After incubation, the mixture was loaded onto a Sephadex G-50 column (50 × 2.5 cm) equilibrated with 10 mM Na-phosphate buffer (pH 7.1) containing 0.145M NaCl, and 5–6 mL fractions from the column were assessed. The dansyl-PMB appears as a broad peak ahead of the unreacted dansyl-chloride peak. The location of the dansyl-PMB in the collected fractions was determined by holding a UV lamp over the fractions and looking for fluorescence. The fluorescence of the dansyl-PMB was yellowish, while the unreacted dansyl-chloride showed more blue-green fluorescence. The fractions containing dansyl-PMB were extracted into approximately 1/2 vol of n-butanol. The butanol was then evaporated to dryness in a glass Petri dish placed inside a desiccator, which was then evacuated and placed at 37°C for 24 hr. The dried dansyl-PMB was dissolved in 3 mL of buffer (5 mM HEPES, pH 7.0) and stored at −20°C. CLSM was performed using LSM 770 (Carl Zeiss Microscope, Jena, Germany). The exponential phase of the 5 mL bacterial culture was washed and resuspended with PBS. The cells were stained with dansyl-PMB (2 or 4 µg/mL) for 30 min at 37°C.

Protective effect of OMVs in the presence of PMB

Request a detailed protocol

To confirm the protective effect of OMVs on LB agar plates, cultured cells were inoculated on media containing 0, 0.5, 1, 2, or 4 μg/mL PMB in the absence or presence of OMVs from PMRHigh. Each culture was serially diluted 1:10 and inoculated beginning in the first spot. In the liquid medium, the protective effect of OMVs against PMB was monitored by assessing the growth curves of OMV-treated Lab-WT, PMRLow, and PMRHigh cells and clinical isolates (designated F-1025, F-1208, F-1379, F-1410, and F-1629) from patients treated at Sungkyunkwan University Hospital. PMB was used at concentrations of 0, 1, 2, and 4 μg/mL. The cultured cells (106 CFU/ml) were inoculated separately into LB medium containing different concentrations of PMB (0, 1, 2, and 4 μg/mL) and PMRHigh -derived OMVs (25 μL/mL). Bacterial growth was monitored at OD595 at 2 hr intervals for 72 hr. Experiments were performed using bacterial cultures from three independent batches. The bacterial cultures were shaken for 10 s inside the instrument (TECAN) every 10,000 s.

Analysis of fecal microbiota

Request a detailed protocol

The fecal sample was obtained from a woman aged 25 years, collected in sterilized plastic tubes, and suspended at a dilution of 1/10 in PBS (pH 7.4). The fecal solution (5 mL) was added to an anaerobic bottle, injected with 80% N2 + 20% CO2 gas, and sterilized. Subsequently, fecal microbiota (50 μL, 1/100 of the fecal solution) were inoculated and cultured at 37°C. Analysis of bacterial communities revealed whether the protected bacteria depended on the type of OMV in the bacterial community (Figure 6A). PMB (4 μg/mL) and OMVs (25 μL/mL) of equal concentrations were administered daily into the fecal samples suspended in PBS; the sample was stabilized for 3 hr by applying OMVs first, and then treated with PMB and incubated at 37°C for 5 days. Then, total community DNA was extracted using a FastDNA spin kit for soil (MP Biomedicals, USA), and the DNA yield was quantified using a NanoDrop spectrophotometer (BioTek, USA). The V3-V4 hypervariable region of 16S rRNA gene from the genomic DNA was amplified using the primers 341F (5′-TCGTCGGCAGCGTC-AGATGTGTATAAGAGACAG-CCTACGGGNGGCWGCAG-3′) and 805R (5′-GTCTCGTGGGCTCGG-AGATGTGTATAAGAGACAG-GACTACHVGGGTATCTAATCC-3′). The amplified products were confirmed by agarose gel electrophoresis. The amplicons were purified by Agencourt AMPure XP (Beckman Coulter, Republic of Korea) and quantified using a Quanti-iT Picogreen dsDNA Assay kit. Equimolar concentrations of each amplicon from the different samples were pooled and purified using Agencourt AMPure XP (Beckman Coulter). All sequencing procedures were conducted by ChunLab. The sequences obtained were compared and classified using the EzTaxon Database (http://www.ezbiocloud.net). The operational taxonomic units (OTUs) among the samples were obtained with a taxonomic composition using the CLcommunity program (ChunLab). The bacterial community data have been deposited in NCBI under Sequence Read Archive (SRA) accession numbers SRX9819399 PRJNA689940, SRX9819397 PRJNA689944, and SRX9819398 PRJNA689944.

G. mellonella-infection model

Request a detailed protocol

G. mellonella larvae were obtained from SWORM (Cheonan, Republic of Korea). Healthy G. mellonella larvae weighing 200 mg were starved for 4 hr at 20°C before injection. Then, the larvae were placed on ice and injected via the last left proleg with 10 µL of OMV solution (25 μL/mL) using 31-gauge, 6-mm-long needles (BD Ultra-Fine insulin syringes). Infected larvae were divided into the following four experimental groups with 15 larvae per group: (i) inactive control groups, which received 10 µL of PBS or OMVs; (ii) bacterial treatment groups, which received either 1 × 106 CFU/larvae of A. baumannii ATCC 17978 Lab-WT or PMRHigh; (iii) bacteria and antibiotic treatment groups, which received 1 × 106 CFU/larvae of Lab-WT or PMRHigh and 1 µg/mL PMB; and the (iv) OMV treatment group, which received 1 × 106 CFU/larvae of Lab-WT or PMRHigh, 1 µg/mL PMB, and OMVs. PBS was added to each group up to 10 µL. The larvae were incubated in a Petri dish (90 × 15 mm, SPL Life Sciences) with 100 mg of wheat bran powder (MG Natural, Republic of Korea) at 37°C in air for 96 hr and inspected and scored every 12 hr for death, failure to move in response to touch, and melanization.

Data availability

1. Sequence reads for the experimentally evolved isolates are accessible from the NCBI sequence archives under accession numbers PRJNA530195 (Lab-WT), PRJNA530197 (PMRLow), and PRJNA530202 (PMRHigh). 2. The RNA-seq data have been deposited in NCBI under Gene Expression Omnibus (GEO) accession number GSE163581. 3. The bacterial community data have been deposited in NCBI under Sequence Read Archive (SRA) accession number SRX9819399 PRJNA689940, SRX9819397 PRJNA689944 and SRX9819398 PRJNA689944.

The following data sets were generated
    1. Shin B
    (2018) NCBI BioProject
    ID PRJNA530195. Acinetobacter baumannii strain:Lab-WT Genome sequencing.
    1. Shin B
    (2018) NCBI BioProject
    ID PRJNA530197. Acinetobacter baumannii strain:PMR-Low Genome sequencing.
    1. Shin B
    (2018) NCBI BioProject
    ID PRJNA530202. Acinetobacter baumannii strain: PMR-High Genome sequencing.
    1. Park J
    (2020) NCBI Gene Expression Omnibus
    ID GSE163581. Transcriptome of Acinetobacter baumannii with or without polymyxin B.
    1. Kim M
    (2021) NCBI Sequence Read Archive
    ID SRX9819399+PRJNA689940. Fecal microbiota control.

References

    1. Alsan M
    2. Klompas M
    (2010)
    Acinetobacter baumannii: an emerging and important pathogen
    Journal of Clinical Outcomes Management: JCOM 17:363–369.
  1. Book
    1. Asperger O
    2. Kleber HP
    (1991)
    The Biology of Acinetobacter
    New York, USA: Plenum Press.

Decision letter

  1. María Mercedes Zambrano
    Reviewing Editor; CorpoGen, Colombia
  2. Gisela Storz
    Senior Editor; National Institute of Child Health and Human Development, United States

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Acceptance summary:

This study looks at the role of outer membrane vesicles produced by the major human pathogen Acinetobacter baumannii in the resistance to the antibiotic polymyxin. By doing extensive omics and mutational studies, the authors find that these vesicles can bind polymyxin B and protect both producing cells and entire bacterial communities from the bactericidal effect of this antibiotic.

Decision letter after peer review:

[Editors’ note: the authors submitted for reconsideration following the decision after peer review. What follows is the decision letter after the first round of review.]

Thank you for submitting your work entitled "A novel decoy strategy of Acinetobacter baumannii for polymyxin resistance" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, including María Mercedes Zambrano as the Reviewing Editor and Reviewer #1, and the evaluation has been overseen by a Senior Editor. The following individual involved in review of your submission has agreed to reveal their identity: John Boyce (Reviewer #3).

Our decision has been reached after consultation between the reviewers. Based on these discussions and the individual reviews below, we regret to inform you that your work will not be considered further for publication in eLife.

We found that while the work is both novel and very relevant to the field, there were several concerns regarding the interpretation of results and the subsequent claims that make this not suitable for publication. Some of the major issues involve the way in which data was collected and presented (such as MICs, evidence for OMVs, experiments with stool samples, and statistics). Most relevant, however, is the lack of convincing evidence for the role of OMVs in polymyxin resistance, such as polymyxin binding or changes in the LPS, to support the claim for OMV protection in this species.

Reviewer #1:

The authors explore the role of A. baumannii outer membrane vesicles (OMV) in resistance to polymyxin B (PMB). They obtain a laboratory-derived resistant strain that has altered expression of genes involved in OMV biogenesis and produces more OMVs that show protection against polymyxin B both in vitro and in vivo. This includes a MIC increase for several clinical strains and enhanced survival of A. baumannii in a Galleria mellonella assay. The results are consistent with the role of OMVs in other bacteria, contribute additional information regarding protection of A. baumannii in different experimental conditions, and suggest that OMVs protect by acting as decoys for PMB binding, although the precise mechanisms remain unclear.

1. The results are interesting, but I am concerned with the specificity of the observations. In the absence of experiments with OMVs from other strains, it is difficult to assess if the results are due to this strain or if OMVs from other strains could also work (such as OMVs from the low and WT strains, the latter used also in the fecal experiment). This is also important in terms of the mechanisms involved. Given the absence of changes in surface charge (zeta potential) when cells are exposed to PMB (Figure 1C, D), what do the authors suggest is the main reason for the observed protective effect on A. baumannii cells?

2. The work with the feline microbiota is confusing and raises many questions, partly because of the complexity of dealing with mixed communities. The results indicate that certain taxa/groups (as identified by peaks) change when PMB / OMVs are administered. Some peaks seem to increase upon treatment with OMVH, while at least one peak increased when treating with OMVW, yet this is not mentioned in the text. How can this be explained? Are there no Gram-negative bacteria in this fecal sample and if so, are these affected or not by PMB/OMVs? While these experiments could suggest that these OMVs affect various cell types, couldn't addition of any antibiotic and any OMVs also trigger changes? It is therefore unclear if the changes, assumed as protection, are specific to these OMVs.

3. The authors mention binding to PMB as a strategy (eg, Lines 39-40 and Line 278), even though there is no direct evidence provided. It would be best to tone down these claims. It is also unclear how the work leads to the conclusion regarding novel antibiotics and "multifaceted" OMV production (Line 338). Please revise the sentence to more accurately reflect the findings of this study.

Reviewer #2:

In the article submitted to eLife by Shin et al., the authors investigate mechanisms of Acinetobacter baumannii resistance to polymyxin B. They isolate random mutations that occur on the genome as a result if culturing A.b. in media with PMB. They find isolates that have low and high levels of PMB resistance. Whole genome sequencing identifies nearly 20 non-synonomous mutations in different genes. Substitutions in pmrB were common between two isolates. They generate deletion mutants and overexpression genotypes for PmrB, and also contruct deletions in the other genes. The authors move on to characterize PmrB-dependent production of OMVs. The reason for focusing on this gene in favor of the others is not firmly established. The data that follow are not conducted with a sufficient level of rigor to award acceptance to this journal. No statistical analysis is presented. Significance of the data is absent, making the impact of the study very minimal.

To my knowledge, the work demonstrating the involvement of PmrB in OMV formation for this important pathogen is novel and consistent with recent data that suggest modified LOS molecules might promote or prevent vesicle formation. In general the work is well written and easy to follow conceptually. However, the level of rigor applied to this study falls well short of the requirements for this journal. In its current condition, this article is not of sufficient impact to warrant publication in this journal.

Data regarding PMB resistance is presented as MICs and growth curve measurements. No bacteriacidal assay is used, which is a major weakness and oversight. The authors report no cfu data for the bacteria that have been treated. The data is over interpreted.

Measuring effects of PMB on bacterial growth on agar media is technically problematic due to the contact dependent killing activity of PMB.

No statistics are provided to support significance of the findings.

The authors do not use appropriate assays to assess whether the particles they are measuring are actual OMVs. Western blots and enzymatic assays for particular proteins that localize or do not localize to the OMV are absent and must be included.

The authors offer no explanation for why the PMB resistant mutant also acquire multidrug resistance (rif, cat, amp etc.). This line of interrogation is abruptly ended. Do the other mutations contribute to this mechanism?

The model does absolutely nothing for the paper as it does not even speculate as to how a PmrB modified envelope would cause vesicle formation.

The experiment where A. baumannii OMVs are added to stool samples of a cat is not rationalized sufficiently or conducted in a manner that one could specifically attribute the effects to the OMVs. A more appropriate study would involve a mouse model of disease. A. baumannii does not colonize the intestine so the relevance of this model is minimal.

Fm464 staining to confirm OMV isolation is insufficient to establish that these are actual OMVs. Blotting for OmpA an OM protein and some IM protein would provide stronger support.

Similarly quantifying the lipid content with FM4-64 is not rigorous as the dye is non-specific.

Reviewer #3:

This paper describes an analysis of the role of outer membrane vesicles (OMVs) in polymyxin resistance in Acinetobacter baumannii. Firstly, polymyxin-resistant A. baumannii strains were selected by growth on increasing concentrations of polymyxin and then genome sequencing showed the presence of a number of mutations. Both resistant strains had two pmrB mutations known to increase PmrC expression and give PEtn addition to LPS. The strains, especially PMRHigh, showed increased OMV production and addition of purified OMVs from this strain could increase polymyxin resistance of other sensitive strains. This is a very interesting paper worthy of further investigation. However, I feel that this paper falls short of showing a convincing role for OMVs in polymyxin resistance in this species. Without further support of presence/absence of PEtn on resistant membranes and OMVs released by the PMRHigh strain, the idea that OMVs are protective seems counterintuitive; if the LPS from this PMRHigh strain has PEtn addition (highly indicated by mutations and PmrC increased expression) then the released OMVs should not bind polymyxin due to the reduced charge. Indeed, it might be expected that OMVs from WT would be more protective as they would bind polymyxin.

1. Currently, the whole genome sequencing analysis is unconvincing. Firstly, the genome sizes and number of ORFs are quite different between parent and derivative resistant strains. This seems very unusual and is likely due to different success in de novo assembly. It would have been good to use reference assembly against the parent strain as the derivatives should be almost identical. Given the likely problems with assembly, at least a number of the proposed mutations should be checked by PCR and Sanger sequencing.

2. The two common mutations in PmrB, and the increased expression of PmrC, indicate that polymyxin resistance is likely to be primarily due to PEtn addition to LPS and concomitant reduction in neg charge and polymyxin binding. However, this needs to be shown by LPS MS. Indeed, I think it is essential to look at the LPS structure in both the parent sensitive cells, resistant derivatives and the OMVs from each strain.

3. Mutation of pmrB in the PMRHigh strain failed to return the strain to fully sensitive, which does argue in favor of PEtn addition not being the primary reason for resistance. However, this would be a clearer experiment if the actual transferase gene (pmrC) was inactivated.

4. Line 161-163. Wouldn't calculation of surface charge in the presence of polymyxin be confounded by cell death in sensitive strains?

5. Line 163-165. The fact that both resistant strains were not resistant to other antibiotics suggests in favor of PEtn addition as this is a polymyxin-specific mechanism. Why is this listed as "interesting".

6. What is the proposed explanation for the high number of released OMVs in the ftsL mutant but no increase in polymyxin resistance?

7. Line 22 suggests that, "as expected", mutations in genes associated with cell wall cross linking such as lpp showed reduced OMV secretion. Isn't this at odds with line 106 which say that reduction in Lpp production disrupts the cross-linkage of PG on the membrane, thereby significantly contributing to the formation of OMVs.

8. For conclusive assessment of the role of OMVs on altered microbiota, full metagenomic 16s rRNA should be conducted, rather than terminal restriction fragment length polymorphism analysis, which does not capture enough diversity.

9. For Galleria mortality experiments, in my opinion bacterial counts should be determined

10. In my opinion, for conclusive understanding of the role of OMVs in decoy resistance, polymyxin binding should be assessed.

[Editors’ note: further revisions were suggested prior to acceptance, as described below.]

Thank you for submitting your article "A novel decoy strategy for polymyxin resistance in Acinetobacter baumannii" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Gisela Storz as the Senior Editor. The reviewers have opted to remain anonymous.

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

Essential Revisions:

The manuscript has improved with the inclusion of new data, but the authors should include more detailed information and take care with the interpretation of some of their results.

1. Important discussion of other studies is missing, e.g., Jasim et al. about OMVs and polymyxins in Klebsiella (PMID: 30103446), or Macdonald et al. about OMVs in Pseudomonas (PMID: 23625841), etc.

Please include statistical information required to correctly interpret the results.

2. Line 152: it should be noted that the observed 1.2-fold higher pmrC expression in PMRLow was not statistically significant. Indeed, 1.2-fold increase is below the usual cut-off values used for transcriptomics.

3. Most of the reported transcriptomics data (e.g. Figure 3B) has no statistics associated with it, so it is not possible to know whether observed differences are significant. Add false discovery rate values or CVs to the fold-change data.

4. The number of replicates should be noted in the figure legend for all experiments. E.g. for Figure 4. it seems unlikely that there is a statistically significant difference between protein Bradford assay for LabWT and deltapmrB unless there are a large number of replicates here.

5. The addition of Raman spectroscopy and the specificity of its interpretation are unconvincing.

Several issues were brought up regarding the transcriptomics (in addition to statistics) and genomics data that require further clarification.

6. There is no clear explanation for why different SNVs were observed in the transcriptomic data compared to the genomic data. Were these mutations found in a high percentage of the mapped reads?

7. The rationale for the RNA-Seq experiments is unclear and there were only a few genes dysregulated. Therefore, the value of this experiment is unclear.

8. The mutations in mRNA are troublesome as e.g., a master regulator of stringent stress response and cell division were affected and it is not clear how this is further linked to OMVs and polymyxins. Many mutations (including master regulators) in the polymyxin-resistant mutant strain could affect the observed phenotypes.

9. Why do WT and PMR-resistant strains have different genome sizes? A difference of >28 kb and 11 genes sound a lot – what was deleted in these strains specifically? There are also a lot of other important mutations in the PMR high strain that could explain the growth phenotype and this makes the whole data interpretation regarding OMVs questionable.

10. The CRISPR-Cas9 method needs to be fully detailed in Materials and methods and it should be noted which mutants are from homologous recombination and which from CRISPR.

Biofilm experiments and interpretations should be further clarified. Please address the following regarding biofilm experiments:

11. The biofilm data is not convincing and should be quantified for multiple areas. There is no evidence that OMVs are associated with biofilms in this study.

12. Supplemental Figure 1. Was the increase in biofilm seen due to excess EPS formation or an increase in viable bacteria? Was any sonication done to see if the actual number of bacteria at the timepoint chosen increase in the evolved mutant with PMB?

13. There are other mechanisms of colistin (polymyxin) resistance seen including transmissible elements such as MCR-1 which encodes a phosphoethanolamine transferase leading to increased positive charge on the membrane. Is it known whether other mechanisms of resistance would also lead to increased OMV formation? If not, it might be good to mention in the discussion about whether the mechanism underlying increased OMV production would be applicable to other forms of polymyxin resistance.

14. Did the authors determine whether OMVs from other pathogens would result in a similar decoy effect or would this be expected to be specific to polymyxin stimulated OMVs in Acinetobacter?

15. It is not clear why the authors mention DNA relaxation, but do not really explain how this is linked to whole story.

16. The document includes a large amount of experimental data that makes it hard to follow and could therefore be improved for clarity. In addition, some of the experimental information is not reflected in the abstract. Finally, authors should take care to avoid speculations.

– For example, the Results section starts with a speculation about the lower growth rate, which is inappropriate but also makes the reader wonder whether other mutations could be the cause of the growth defect. It becomes clearer reading further when the WGS is described.

– There is further speculation in the results that should not be there and lots of the results/discussion feel out of context.

Please clarify some of the figures presented.

17. Figure 1. Panel D TEM is not particularly quantitative. It is difficult to claim that there a larger number of OMVs in the PMR(high) strain. Was this quantitated visually in some way? If not, it might be helpful to include supplemental TEM images at a lower magnification to help support this claim.

18. Figure 3B (expression vs genomic position image) should be relative to log base 2 fold-change so that up- and down-regulated genes are on a similar scale.

19. It would be useful to have a merged image of Figure 5D.

20. Figure 6 Panel A. This panel is a bit confusing. I am not sure what the number scale at the top of the graph represents versus the number scale at the bottom of the panel. The dots appear blue but the authors mention black dots? I assume these are the dots they are alluding to but this needs to be clarified.

https://doi.org/10.7554/eLife.66988.sa1

Author response

[Editors’ note: the authors resubmitted a revised version of the paper for consideration. What follows is the authors’ response to the first round of review.]

We found that while the work is both novel and very relevant to the field, there were several concerns regarding the interpretation of results and the subsequent claims that make this not suitable for publication. Some of the major issues involve the way in which data was collected and presented (such as MICs, evidence for OMVs, experiments with stool samples, and statistics). Most relevant, however, is the lack of convincing evidence for the role of OMVs in polymyxin resistance, such as polymyxin binding or changes in the LPS, to support the claim for OMV protection in this species.

Reviewer #1:

The authors explore the role of A. baumannii outer membrane vesicles (OMV) in resistance to polymyxin B (PMB). They obtain a laboratory-derived resistant strain that has altered expression of genes involved in OMV biogenesis and produces more OMVs that show protection against polymyxin B both in vitro and in vivo. This includes a MIC increase for several clinical strains and enhanced survival of A. baumannii in a Galleria mellonella assay. The results are consistent with the role of OMVs in other bacteria, contribute additional information regarding protection of A. baumannii in different experimental conditions, and suggest that OMVs protect by acting as decoys for PMB binding, although the precise mechanisms remain unclear.

1. The results are interesting, but I am concerned with the specificity of the observations. In the absence of experiments with OMVs from other strains, it is difficult to assess if the results are due to this strain or if OMVs from other strains could also work (such as OMVs from the low and WT strains, the latter used also in the fecal experiment). This is also important in terms of the mechanisms involved. Given the absence of changes in surface charge (zeta potential) when cells are exposed to PMB (Figure 1C, D), what do the authors suggest is the main reason for the observed protective effect on A. baumannii cells?

We agree with reviewer’s comments. The degree of OMV production in each clinical isolate varied. Although direct correlations between the MIC and OMV production could not be identified under our tested conditions. However, higher OMV producers could show higher protection rates against PMB in the early stationary phase (Figure 6A).

The MICs of all constructed mutants were not affected, although they showed higher OMV production due to exposure to PMB at the beginning stage of culture conditions, in which no OMV production was expected. However, tendency for protecting cells grown in stationary phase from high PMC concentration (1/2 MIC of each clinical strain) was observed in our tested higher OMV producers (Figure 4F).

We performed whole genome sequencing, RNA-sequencing, proteomics, microbial community analysis, mutational analyses, MALDI-TOF, transmission electron microscopy assays, and growth tests with purified outer membrane vesicles (OMVs). In our new manuscript, we offered profound understanding that OMVs released from Acinetobacter baumannii functioned as decoys against PMB, which explained why other clinical isolates and microbial communities with OMV were also protected against PMB treatment. Our data suggest that OMVs can act as an important survival strategy in complex bacterial communities as well as A. baumannii by reducing polymyxin B binding to bacterial membrane.

2. The work with the feline microbiota is confusing and raises many questions, partly because of the complexity of dealing with mixed communities. The results indicate that certain taxa/groups (as identified by peaks) change when PMB / OMVs are administered. Some peaks seem to increase upon treatment with OMVH, while at least one peak increased when treating with OMVW, yet this is not mentioned in the text. How can this be explained? Are there no Gram-negative bacteria in this fecal sample and if so, are these affected or not by PMB/OMVs? While these experiments could suggest that these OMVs affect various cell types, couldn't addition of any antibiotic and any OMVs also trigger changes? It is therefore unclear if the changes, assumed as protection, are specific to these OMVs.

We agree with the reviewers. We previously had the T-RFLP analysis with the cat feline microbiota but recognized its inaccuracy of peak identification. Therefore, we deleted our previous T-RFLP data and performed new illumine-based bacterial community analysis using with in vitro human fecal sample.

Line 396-398: Our taxonomic profile revealed that addition of OMVs extracted from the PMRHigh strain protected the bacterial community from PMB under our tested anaerobic conditions, resulting in recovery of a similar microbial composition (Figure 7A).

3. The authors mention binding to PMB as a strategy (eg, Lines 39-40 and Line 278), even though there is no direct evidence provided. It would be best to tone down these claims. It is also unclear how the work leads to the conclusion regarding novel antibiotics and "multifaceted" OMV production (Line 338). Please revise the sentence to more accurately reflect the findings of this study.

To address the binding issues raised in reviewer’s comments, we conducted several experiments and demonstrated direct binding of PMB to purified OMVs.

Line 358-366: To visualize direct binding PMB to OMVs, we developed a dansyl fluorophore-PMB, which was synthesized by forming a chemical bond between the primary γ-amines on the diaminobutyric-acid residues of PMB and dansyl-chloride. Confocal laser scanning microscopy (CLSM) images revealed that dansyl-PMB was less bound to PMRHigh cells than Lab-WT cells at their respective PMB MICs (2 or 4 µg/mL), due to the fact that the PmrC-driven lipid A modification of PMRHigh cells decreased PMB binding to OMs (Figure 5C). The lipid-selective dye (FM4-64) and dansyl-PMB were employed to visualize binding of PMB to purified OMVs of the PMRHigh stain (Figure 5D).

Reviewer #2:

In the article submitted to eLife by Shin et al., the authors investigate mechanisms of Acinetobacter baumannii resistance to polymyxin B. They isolate random mutations that occur on the genome as a result if culturing A.b. in media with PMB. They find isolates that have low and high levels of PMB resistance. Whole genome sequencing identifies nearly 20 non-synonomous mutations in different genes. Substitutions in pmrB were common between two isolates. They generate deletion mutants and overexpression genotypes for PmrB, and also contruct deletions in the other genes. The authors move on to characterize PmrB-dependent production of OMVs. The reason for focusing on this gene in favor of the others is not firmly established. The data that follow are not conducted with a sufficient level of rigor to award acceptance to this journal. No statistical analysis is presented. Significance of the data is absent, making the impact of the study very minimal.

To my knowledge, the work demonstrating the involvement of PmrB in OMV formation for this important pathogen is novel and consistent with recent data that suggest modified LOS molecules might promote or prevent vesicle formation. In general the work is well written and easy to follow conceptually. However, the level of rigor applied to this study falls well short of the requirements for this journal. In its current condition, this article is not of sufficient impact to warrant publication in this journal.

Data regarding PMB resistance is presented as MICs and growth curve measurements. No bacteriacidal assay is used, which is a major weakness and oversight. The authors report no cfu data for the bacteria that have been treated. The data is over interpreted.

We agreed with the reviewer. We conducted several bactericidal assays to address this issue.

Figure 4F and 6A, we conducted bactericidal assay of experimentally evolved strains, mutants and clinical isolates.

Line 387-391: After 5 days, the community cells were spotted on an LB plate and cultured under aerobic conditions to check PMB toxicity and the protective effect of OMVs (Supplementary figure 6B). The PMB-treated sample showed a 104-fold reduction in aerobic bacterial cells than the control, and the OMV plus PMB sample showed similar levels of cell counts as the control (Supplementary figure 6B).

Line 518-519: The protection rate (%) was calculated survived colonies (CFU) in comparison with untreated conditions of each strain.

Measuring effects of PMB on bacterial growth on agar media is technically problematic due to the contact dependent killing activity of PMB.

We considered that the technical problem that you mentioned would not be affected to conclude our results because of several other experimental evidences which include our data with in vitro human gut microbiota. Line 351-356: The protective effect of these PMRHigh-derived OMVs was assessed using the Luria–Bertani (LB) broth (Figure 5A) and LB agar plate tests (Figure 5B), after confirming that additional purified OMVs had no effect on the bacterial growth. Interestingly, supplementation of PMB-containing liquid cultures (Figure 5A) or agar (Figure 5B) with these OMVs could increase the survival of all tested strains except for PMRHigh on agar plates.

No statistics are provided to support significance of the findings.

Statistical analyses were added in Figure 1C, Figure 4, Supplementary Figure 1B.

The authors do not use appropriate assays to assess whether the particles they are measuring are actual OMVs. Western blots and enzymatic assays for particular proteins that localize or do not localize to the OMV are absent and must be included.

We thank the reviewer for suggesting various experimental methods. We conducted new proteomics analysis to supplement your proposal more specifically. The commonly identified OMV components were membrane-linkage proteins (OmpA, OmpW and BamE).

Line 291- 295: Gel-based proteomics with MALDI-TOF analysis were performed. Forty-three out of the 61 up- or downregulated proteins were designated with the MS-Fit program by using the NCBInr and UniPortKB databases, and 18 spots were not identified (Table S3).

Line 301-310: Interestingly, A. baumannii OMVs also contained many cytoplasmic cargo proteins such as ribosomal proteins and Ef-Tu, similar to the OMVs from other bacteria (Dallo et al., 2012; Deo et al., 2018). Elongation factor (Ef-Tu), observed in our proteomics analyses, is one of the most abundant proteins (>20%) in bacteria and has a canonical role in translation. However, it is also found on the surface of bacterial OMs, which suggests a possible noncanonical function on the membrane (Harvey et al., 2019; Widjaja et al., 2017). Accumulation of Ef-Tu in the periplasmic space might induce pressure, which promotes blebbing of the OM resulting in OMV secretion (Figure 3C, McBroon and Kuehn, 2007). Reduced expression of OmpA and OmpW leads to diminished linkage between the OM and peptidoglycan and subsequently causes more OMV generation in the PMRHigh

The authors offer no explanation for why the PMB resistant mutant also acquire multidrug resistance (rif, cat, amp etc.). This line of interrogation is abruptly ended. Do the other mutations contribute to this mechanism?

We stated that PMB resistant strains were not resistant to other antibiotics (rif, cat, amp etc.)

Line 165-168 : “The PMRLow and PMRHigh strains were resistant only to polymyxins, but not to the other nine tested antibiotics of different classes, including meropenem, rifampicin, chloramphenicol, and ampicillin (Supplementary figure 2).”

The model does absolutely nothing for the paper as it does not even speculate as to how a PmrB modified envelope would cause vesicle formation.

Our data suggested that either physical association of PmrB in the membrane or modification of the membrane charge by PmrB-activated PmrC might affect the production of OMVs in A. baumannii strains.

Lines 326-331, “when the pmrB gene was deleted in Lab-WT, OMV biogenesis was reduced 0.3-fold (Figures 4B–F), indicating that OMV production is impaired in the absence of PmrB. This finding was supported by our observation in which an extra copy of pmrBH gene increased OMV production in Lab-WT (Figure 4B-F). PmrB/A-activated PetN transferase or incorporation of PmrB into the IM region may modulate the OM charge and integrity, inducing more OMV production.”,

The experiment where A. baumannii OMVs are added to stool samples of a cat is not rationalized sufficiently or conducted in a manner that one could specifically attribute the effects to the OMVs. A more appropriate study would involve a mouse model of disease. A. baumannii does not colonize the intestine so the relevance of this model is minimal.

We deleted our previous T-RFLP data and performed new illumine-based bacterial community analysis using with in vitro human fecal sample.

Line 394-398: Our taxonomic profile revealed that addition of OMVs extracted from the PMRHigh strain protected the bacterial community from PMB under our tested anaerobic conditions, resulting in recovery of a similar microbial composition (Figure 7A).

Figure 7, We conducted bacterial community analysis with human fecal microbiota. Taxonomic profiling of in vitro human gut microbiomes under anaerobic conditions demonstrated that OMVs completely protected the microbial community against PMB treatment.

Fm464 staining to confirm OMV isolation is insufficient to establish that these are actual OMVs. Blotting for OmpA an OM protein and some IM protein would provide stronger support.

We developed a dansyl fluorophore-PMB and visualized direct PMB binding to purified OMVs by confocal microscopy images (Carl Zeiss, Germany). RNA-sequencing and proteomic analyses were conducted to understand contribution of genetic and proteins components (including OmpA) to OMV production. Our multi-omics studies and mutational analyses suggested that decreased membrane-linkage proteins induce overproduction of OMVs.

Similarly quantifying the lipid content with FM4-64 is not rigorous as the dye is non-specific.

We used several approaches for quantifying OMV production (lipid staining [FM4-64 dye], quantification of total proteins [Bradford assay], total DNA measurement, and LPS quantitation using the purpald assay) (Figure 4 and Supplementary figure 4B). We also developed a dansyl fluorophore-PMB and visualized direct PMB binding to purified OMVs by confocal microscopy images.

Line 637-639: The lipophilic dye FM4-64, which intercalates into the outer membrane, is commonly used to stained and observed membrane of bacteria (Rojas et al., 2018; Toyofuku et al., 2017).

Reviewer #3:

This paper describes an analysis of the role of outer membrane vesicles (OMVs) in polymyxin resistance in Acinetobacter baumannii. Firstly, polymyxin-resistant A. baumannii strains were selected by growth on increasing concentrations of polymyxin and then genome sequencing showed the presence of a number of mutations. Both resistant strains had two pmrB mutations known to increase PmrC expression and give PEtn addition to LPS. The strains, especially PMRHigh, showed increased OMV production and addition of purified OMVs from this strain could increase polymyxin resistance of other sensitive strains. This is a very interesting paper worthy of further investigation. However, I feel that this paper falls short of showing a convincing role for OMVs in polymyxin resistance in this species. Without further support of presence/absence of PEtn on resistant membranes and OMVs released by the PMRHigh strain, the idea that OMVs are protective seems counterintuitive; if the LPS from this PMRHigh strain has PEtn addition (highly indicated by mutations and PmrC increased expression) then the released OMVs should not bind polymyxin due to the reduced charge. Indeed, it might be expected that OMVs from WT would be more protective as they would bind polymyxin.

1. Currently, the whole genome sequencing analysis is unconvincing. Firstly, the genome sizes and number of ORFs are quite different between parent and derivative resistant strains. This seems very unusual and is likely due to different success in de novo assembly. It would have been good to use reference assembly against the parent strain as the derivatives should be almost identical. Given the likely problems with assembly, at least a number of the proposed mutations should be checked by PCR and Sanger sequencing.

Thank the reviewer for nice comments. We believed that genome sequencing is minor issue for claiming our conclusion because we performed transcriptomic and mutational analyses to understand the roles of several important genes in OMV production. The number of coding DNA sequences (CDSs) was incorrect. We corrected the number in Table S1 and sentences. The difference in the number of CDSs in resistant strains was not significant. Line 227-230: The numbers of coding DNA sequences (CDSs) in the PMRLow and PMRHigh genomes were 3625 and 3636, with mean CDS lengths of 930.2 and 947.5 respectively (Table S1).

2. The two common mutations in PmrB, and the increased expression of PmrC, indicate that polymyxin resistance is likely to be primarily due to PEtn addition to LPS and concomitant reduction in neg charge and polymyxin binding. However, this needs to be shown by LPS MS. Indeed, I think it is essential to look at the LPS structure in both the parent sensitive cells, resistant derivatives and the OMVs from each strain.

We conducted the MALDI-TOF analysis in agreement with the reviewer opinion

Figure 2, To determine whether modified lipid A plays an important role in PMB resistance, lipid A was subjected to matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) analysis in the positive-ion mode.

Line184-190, The main peak of Lab-WT was measured at 1,490 m/z, corresponding to bis-phosphorylated penta-acylated lipid A (Figure 2A and 2B). In the PMRHigh strain, the predominant spectrum was also measured at 1,490 m/z, followed by 1,630 m/z, which is predicted to be the peak when a PetN and a hydroxyl group were attached to the dominant lipid A (Figure 2A and 2B). The PMRHigh strain overexpressed pmrC (encoding a PetN transferase), which added a PetN to some lipid A molecules, but not all, leading to interference PMB binding to the OM (see Supplementary figure 1B).

3. Mutation of pmrB in the PMRHigh strain failed to return the strain to fully sensitive, which does argue in favor of PEtn addition not being the primary reason for resistance. However, this would be a clearer experiment if the actual transferase gene (pmrC) was inactivated.

We agree with reviewer’s comments. In our new manuscript, we offered profound understanding that OMVs released from Acinetobacter baumannii functioned as decoys against PMB. Role of pmrC for PMB resistance is well-known sot that we didn’t make a corresponding mutant. Our lipid A analysis (Figure 2A), RNA-sequencing (Figure 3B), qRT-PCR data (Supplementary Figure 1B) suggested that addition of PEtN to lipid A in resistant strains, which might be responsible for PMB resistance.

4. Line 161-163. Wouldn't calculation of surface charge in the presence of polymyxin be confounded by cell death in sensitive strains?

Line 153-155: The zeta-potential analyses suggested that the average negative surface charge of PMRLow and PMRHigh decreased from – 27 mV (Lab-WT) to -20 mV and -10 mV, respectively (Figure 1B).

Line 158-161: However, this observation could not account for all the reasons underlying the high PMB resistance of PMRHigh strain, because short-term (10 min) PMB exposure (4 μg/mL, MIC of PMRLow) decreased the surface charge in all tested strains, albeit to different extents (Supplementary figure 1C, n = 10 per strain).

5. Line 163-165. The fact that both resistant strains were not resistant to other antibiotics suggests in favor of PEtn addition as this is a polymyxin-specific mechanism. Why is this listed as "interesting".

Changes in surface charges on the membrane might affect the transport of other antibiotics into the membrane. “interesting” term was deleted in the manuscript.

6. What is the proposed explanation for the high number of released OMVs in the ftsL mutant but no increase in polymyxin resistance?

Figure 4, We added new protection rate analysis (tendency for protecting cells grown in stationary phase from high PMB concentration). The ftsL mutant was significantly protected under high PMB concentration. The MICs of all constructed mutants were not affected, although they showed higher OMV production due to exposure to PMB at the beginning stage of culture conditions, in which no OMV production was expected. However, tendency for protecting cells grown in stationary phase from high PMC concentration (1/2 MIC of each clinical strain) was observed in our tested higher OMV producers (Figure 4F).

7. Line 22 suggests that, "as expected", mutations in genes associated with cell wall cross linking such as lpp showed reduced OMV secretion. Isn't this at odds with line 106 which say that reduction in Lpp production disrupts the cross-linkage of PG on the membrane, thereby significantly contributing to the formation of OMVs.

The newly RNA-sequencing and mutational studies demonstrated that Lpp loss leads to overproduction of OMVs. We revised the sentence to clarify our intention.

Line 287-289: Notably, several genes belonging to the pmr and pga operons are included among the commonly upregulated gene categories, and the commonly downregulated genes included important membrane-associated genes (mlaC, ompA, lpp, and bamE) (Table S2).

Line 329-330: However, the ΔftsL,Δudg, Δlpp and ΔmlaC mutants exhibited higher OMV production (Figures 4A–D).

Line 334-336: Our mutational study clearly demonstrated that mutant strains lacking genes involved in membrane cross-linkages and phospholipid management (lpp, ftsL and mlaC) showed increased OMV secretion.

8. For conclusive assessment of the role of OMVs on altered microbiota, full metagenomic 16s rRNA should be conducted, rather than terminal restriction fragment length polymorphism analysis, which does not capture enough diversity.

We deleted our previous T-RFLP data and performed new illumine-based bacterial community analysis using with in vitro human fecal sample.

Figure 7, Analysis of fecal microbiota was conducted under aerobic/anaerobic conditions.

9. For Galleria mortality experiments, in my opinion bacterial counts should be determined

We determined the bacterial counts was 1 × 106 CFU/larvae and stated that in Materials and methods section.

10. In my opinion, for conclusive understanding of the role of OMVs in decoy resistance, polymyxin binding should be assessed.

We developed a dansyl fluorophore-PMB and visualized direct PMB binding to purified OMVs by confocal microscopy images

Figure 5C-D, Dansyl-PMB were employed to visualize binding of PMB to purified OMVs of the PMRHigh stain. Our data showed that the two dyes exhibited overlapping images, which clearly demonstrated the direct binding of PMB to OMVs.

Line 357-365: To visualize direct binding PMB to OMVs, we developed a dansyl fluorophore-PMB, which was synthesized by forming a chemical bond between the primary γ-amines on the diaminobutyric-acid residues of PMB and dansyl-chloride. Confocal laser scanning microscopy (CLSM) images revealed that dansyl-PMB was less bound to PMRHigh cells than Lab-WT cells at their respective PMB MICs (2 or 4 µg/mL), due to the fact that the PmrC-driven lipid A modification of PMRHigh cells decreased PMB binding to OMs (Figure 5C). The lipid-selective dye (FM4-64) and dansyl-PMB were employed to visualize binding of PMB to purified OMVs of the PMRHigh stain (Figure 5D).

[Editors’ note: what follows is the authors’ response to the second round of review.]

Essential Revisions:

The manuscript has improved with the inclusion of new data, but the authors should include more detailed information and take care with the interpretation of some of their results.

1. Important discussion of other studies is missing, e.g., Jasim et al. about OMVs and polymyxins in Klebsiella (PMID: 30103446), or Macdonald et al. about OMVs in Pseudomonas (PMID: 23625841), etc.

We have added the following information to address this point:

(Lines 451-454) Notably, Klebsiella pneumoniae strains had outer membranes and larger OMVs harboring different lipid compositions under PMB treatment, which might contribute to the low permeability of PMB (Jasim et al., 2018).

(Lines 498-500) Two known mechanistic factors (PQS and MucD) for OMV production were not required for cycloserine-induced OMV formation in P. aeruginosa, indicating the existence of multiple pathways for OMV biogenesis (Macdonald et al., 2013).

Please include statistical information required to correctly interpret the results.

2. Line 152: it should be noted that the observed 1.2-fold higher pmrC expression in PMRLow was not statistically significant. Indeed, 1.2-fold increase is below the usual cut-off values used for transcriptomics.

Thank you for pointing this out. The sentence has been revised to make it clear.

(Lines 160-162) Consequently, cells with more PetNs in lipid A acquire resistance to PMB. PmrC-mediated reduction of surface charges was not feasible in PMRLow because the expression level of the pmrC gene was not significant.

3. Most of the reported transcriptomics data (e.g. Figure 3B) has no statistics associated with it, so it is not possible to know whether observed differences are significant. Add false discovery rate values or CVs to the fold-change data.

To address the statistical issues raised in the reviewer’s comments, we have added the detailed information of our RNA-seq and RPKM values in Supplementary File 1f and Figure 3—figure supplement 1B, respectively. Our data were calculated using log2 fold changes. In Figure 3A, new qRT-PCR analyses were conducted to verify the expression of all candidate genes (10 genes) for OMV production, which confirmed that our qRT-PCR data were consistent with the RNA-seq data.

(Line 278-281) An additional qRT-PCR assay confirmed that the pmr and pgaB genes showed more than 4-fold higher expression, but all membrane-linkage related genes (ompA, bamE, lpp, and mlaC) were downregulated in PMRHigh (Figure 3A).

4. The number of replicates should be noted in the figure legend for all experiments. E.g. for Figure 4. it seems unlikely that there is a statistically significant difference between protein Bradford assay for LabWT and deltapmrB unless there are a large number of replicates here.

We agree with the reviewer’s comments. More than triplicate experiments were performed for all the data. We have also revised the legends of Figures and figure supplements and added statistics for all data.

5. The addition of Raman spectroscopy and the specificity of its interpretation are unconvincing.

Thank you for pointing this out. Although the Raman spectroscopy data were not convincing, this was the first time Raman spectroscopy has been used for analyzing OMVs. We removed all unnecessary information and only retained a short, important interpretation in the manuscript.

Several issues were brought up regarding the transcriptomics (in addition to statistics) and genomics data that require further clarification.

6. There is no clear explanation for why different SNVs were observed in the transcriptomic data compared to the genomic data. Were these mutations found in a high percentage of the mapped reads?

To avoid any confusion, we have deleted all SNV data and removed all the related information.

7. The rationale for the RNA-Seq experiments is unclear and there were only a few genes dysregulated. Therefore, the value of this experiment is unclear.

We thank the reviewer for pointing this out. We have identified the possible candidate genes for OMV biogenesis by using RNA-seq data and confirmed the levels of their expressions by further qRT-PCR assays. Additional mutational analyses also verified their roles in OMV biogenesis. We have moved all unnecessary information and reanalyzed the RNA-seq data using a different RPKM cutoff (>50 RPKM).

8. The mutations in mRNA are troublesome as e.g., a master regulator of stringent stress response and cell division were affected and it is not clear how this is further linked to OMVs and polymyxins. Many mutations (including master regulators) in the polymyxin-resistant mutant strain could affect the observed phenotypes.

Thank you for these comments. We believe that our omics and mutational analyses substantiated that all tested genes are involved in OMV biogenesis. However, deduction of a direct link between OMV production and a single OMV-related gene might be an oversimplification because of the multifaceted control of bacterial OMV production. Further analysis for understanding how other mutations affect PMB resistance and OMV production are warranted. To avoid any confusion in the manuscript, we deleted all SNV data and removed all the related information.

9. Why do WT and PMR-resistant strains have different genome sizes? A difference of >28 kb and 11 genes sound a lot – what was deleted in these strains specifically? There are also a lot of other important mutations in the PMR high strain that could explain the growth phenotype and this makes the whole data interpretation regarding OMVs questionable.

Thank you for pointing this out. We reanalyzed our genomic data using many software programs including CLC genomics Workbench v.10.0.1 (Qiagen, Germany). We corrected all errors. The corrected genome sizes and deleted regions were marked in Figure 3—figure supplement 1A. Long-read NGS sequencing techniques, such as the PacBio used in this study, may produce sequencing errors. To compensate for this shortcoming, additional PCR reactions using Pfu DNA polymerase were performed to confirm mutational positions listed in Supplementary File 1a (18 genes). The primers used for PCR reactions were listed in Supplementary File 1k. We removed all incorrect information. Our additional transcriptomic and mutational analyses revealed that upregulation of the pmr operon and decreased levels of membrane-linkage proteins are linked to overproduction of OMVs, which also promoted enhanced biofilm formation. Thus, we believe that the occurrence of errors in the genome sequencing is a minor issue in relation to our conclusions.

We have also added the following discussion to address this point and new references were added.

(Line 236-239) Two chromosomal regions mostly annotated as hypothetical proteins were lost during adaptive evolution of PMRLow and PMRHigh strain (Supplementary File 1c). Linkage between those deleted regions and OMV biogenesis remains to be investigated. All 20 antibiotic resistant genes and 78 IS elements remain unchanged in both evolved strains (Supplementary File 1c, 1d).

(Line 479-487) Linkage between those deleted regions in evolved strains and OMV biogenesis remains to be investigated (Figure 3—figure supplement 1A). It is worth noting that antibiotic-induced genomic alterations including deletions, rearrangements, and amplification have been reported in many laboratory-evolved antibiotic resistant bacterial strains (Sandergren and Anderson., 2009). Several laboratory-evolved antibiotic resistant Escherichia coli strains also appeared to carry more than 20 mutations (Maeda et al., 2020). Tetracycline-induced genome rearrangements led to the loss of 6.1 kb having 7 full-length and 2 truncated genes in E. coli strains (Hoeksema et al., 2018). Norfloxacin-treated E. coli had ~11.5 kb deletions possibly through IS1-mediated recombination (Long et al., 2016).

10. The CRISPR-Cas9 method needs to be fully detailed in Materials and methods and it should be noted which mutants are from homologous recombination and which from CRISPR.

In lines 662-683, we have included details of the CRISPR-Cas9 method in the Materials and methods and indicated which mutants were constructed using either homologous recombination or CRISPR-Cas9.

(Lines 662-683) To construct ΔpmrB, ΔmlaC, and ΔpgaB by the CRISPR-Cas9 method, the pCasAb-apr (cat. no 121998) and pSGAb-km (cat. no 121999) plasmids were obtained from Addgene (Wang et al., 2019). A pair of 20-bp spacer oligos and 80-nt oligos were designed to target the genomic locus and donor repair template for editing. A designed 20-bp spacer was phosphorylated and annealed to the pSGAb-km plasmid with a Golden Gate assembly reaction. Then, the plasmids were transformed into E. coli DH5a-competent cells, followed by plating onto an LB agar plate containing 100 µg/mL, and incubated at 37 °C for 16 h. Successful cloning of the spacer was verified by PCR with the primers of Spacer-F/M13R and by sequencing with the primer of M13R (5′-CAGGAAACAGCTATGACC-3′). Then, 1M of IPTG was added into A. baumannii Lab-WT harboring pCasAb-apr to induce the expression of the RecAb recombination system and Cas9 nuclease. After incubating at 37 °C for 2 h, the same method was used to prepare the competent cells as described previously (Wang et al., 2019). Then, 200 ng of the spacer-introduced pSGAb-km plasmid and 300 µM ssDNA (donor repair template) were co-transformed to the IPTG-induced A. baumannii Lab-WT cells harboring the pCasAb-apr plasmid by electroporation. The cells were plated onto an LB agar plate containing 100 µg/mL apramycin and 50 µg/mL kanamycin, and the plate was incubated at 37 °C overnight. Successful editing was verified by PCR and sequencing. For plasmid curing, a colony was plated onto an LB agar plate containing 5% sucrose at 37 °C overnight. Then, the colony were streaked onto LB agar plates without the antibiotics to confirm curing. A detailed genome-editing protocol has been provided in the Supplementary file (Figure 4—figure supplement 1).

Biofilm experiments and interpretations should be further clarified. Please address the following regarding biofilm experiments:

11. The biofilm data is not convincing and should be quantified for multiple areas. There is no evidence that OMVs are associated with biofilms in this study.

We would like to thank the reviewer for this important comment. To address the issues raised in the comment, we have added CLSM images (Figure 1—figure supplement 1D). Biofilm formation was quantified (Figure 1—figure supplement 1E).

(Lines 166-167) OMV-associated biofilm was measured in the Lab-WT and PMRHigh strains (Figure 1—figure supplement 1D and 1E).

(Lines 170-174) Quantification of biofilm formation using the crystal violet assay also indicated that the PMRHigh strain produced more biofilm (2-fold higher) than the Lab-WT strain (Figure 1—figure supplement 1E). In Figure 1—figure supplement 1G, we have included new CLSM images showing the initial stages of Lab-WT biofilm formation with OMVs.

Several published articles also have proven that OMV biogenesis is linked to enhanced biofilm formation (See our references, Baumgarten et al., 2012; Stevenson et al., 2018).

(Lines 441-443) “OMV biogenesis was thought to be also linked to the elevated hydrophobicity of the cell surface in P. putida, which caused greater biofilm formation (Baumgarten et al., 2012).”

(Line 264-266), “The overexpressed pga operon appeared to facilitate the production of PNAG-attached OMVs in E. coli through deacetylation of PNAG by the PgaB activity, which resulted in enhanced biofilm formation (Stevenson et al., 2018).”

In our study, enhanced biofilm formation was noticeable in all OMV overproducers and increased PgaB protein levels in the OMV were also observed in our proteomics analyses. Therefore, it is reasonable to speculate that OMV production partly contributed to the enhanced biofilm formation in our assay.

12. Supplemental Figure 1. Was the increase in biofilm seen due to excess EPS formation or an increase in viable bacteria? Was any sonication done to see if the actual number of bacteria at the timepoint chosen increase in the evolved mutant with PMB?

Biofilm formation was quantified (Figure 1—figure supplement 1E), and biofilm quantification (OD680) was normalized by the OD600 value, which showed that enhanced biofilm formation was not due to an increase in viable bacteria. Sonication was not used for biofilm assay.

(Lines 170-174) Quantification of biofilm formation using the crystal violet assay also indicated that the PMRHigh strain produced more biofilm (2-fold higher) than the Lab-WT strain (Figure 1—figure supplement 1E).

13. There are other mechanisms of colistin (polymyxin) resistance seen including transmissible elements such as MCR-1 which encodes a phosphoethanolamine transferase leading to increased positive charge on the membrane. Is it known whether other mechanisms of resistance would also lead to increased OMV formation? If not, it might be good to mention in the discussion about whether the mechanism underlying increased OMV production would be applicable to other forms of polymyxin resistance.

Both MCR-1 gene products and PmrC function as a phosphoethanolamine (PetN) transferase, although the mcr-1 gene is often located on the plasmid, which was not the case in our strains. Because deletion of the pmrB led to reduced OMV production, induction of the pmr operon might be linked to OMV biogenesis (Figure 4). However, there is no direct evidence for the linkage between PmrC and OMV formation. We removed unnecessary information and retained only valuable interpretative statements in the manuscript (line 324-326).

(Lines 334-336) Furthermore, when the pmrB gene was deleted in Lab-WT, OMV biogenesis was reduced (0.3-fold) (Figures 4A–D), indicating that OMV production is impaired in the absence of PmrB.

14. Did the authors determine whether OMVs from other pathogens would result in a similar decoy effect or would this be expected to be specific to polymyxin stimulated OMVs in Acinetobacter?

(Lines 386-390) PMB MIC tests were performed using four additional Gram-negative bacteria (Pseudomonas aeruginosa PAO1, Pseudomonas putida KT2440, E. coli K12, and Acinetobacter oleivorans DR1) and the protective roles of their OMVs were tested (Figure 6—figure supplement 1). Our data suggested that OMV-driven protection under PMB conditions could be applicable to other Gram-negative bacteria.

(Lines 140-143) Our multi omics data and mutational studies suggested that OMVs functioned as decoys for PMB in A. baumannii. Accordingly, these vesicles can protect not only the OMV producer, but also the entire bacterial community from the bactericidal effects of PMB

15. It is not clear why the authors mention DNA relaxation, but do not really explain how this is linked to whole story.

We agree with these comments. To avoid any confusion, we have deleted the sentence.

16. The document includes a large amount of experimental data that makes it hard to follow and could therefore be improved for clarity. In addition, some of the experimental information is not reflected in the abstract. Finally, authors should take care to avoid speculations.

– For example, the Results section starts with a speculation about the lower growth rate, which is inappropriate but also makes the reader wonder whether other mutations could be the cause of the growth defect. It becomes clearer reading further when the WGS is described.

– There is further speculation in the results that should not be there and lots of the results/discussion feel out of context.

We thank the reviewer for these encouraging comments. We have deleted and revised the unnecessary sentences and other information highlighted by the reviewer.

Please clarify some of the figures presented.

17. Figure 1. Panel D TEM is not particularly quantitative. It is difficult to claim that there a larger number of OMVs in the PMR(high) strain. Was this quantitated visually in some way? If not, it might be helpful to include supplemental TEM images at a lower magnification to help support this claim.

We have included new TEM images of PMRHigh at a lower magnification (Figure 1—figure supplement 1F).

18. Figure 3B (expression vs genomic position image) should be relative to log base 2 fold-change so that up- and down-regulated genes are on a similar scale.

We thank the reviewer for pointing this out. We have revised Figure 3—figure supplement 1B accordingly.

19. It would be useful to have a merged image of Figure 5D.

We have provided a merged image of Figure 5D.

20. Figure 6 Panel A. This panel is a bit confusing. I am not sure what the number scale at the top of the graph represents versus the number scale at the bottom of the panel. The dots appear blue but the authors mention black dots? I assume these are the dots they are alluding to but this needs to be clarified.

In the Figure 6 legend, “black” was rewritten as “blue.” Blue dots show the protection rate for treatment of 1/2 MIC PMB in each clinical isolate, and the bar graph shows the amount of OMVs in the early stationary phase (OD600 = 0.4).

https://doi.org/10.7554/eLife.66988.sa2

Article and author information

Author details

  1. Jaeeun Park

    Laboratory of Molecular Environmental Microbiology, Department of Environmental Science and Ecological Engineering, Korea University, Seoul, Republic of Korea
    Contribution
    Conceptualization, Investigation, Methodology, Writing - original draft, Writing - review and editing
    Contributed equally with
    Misung Kim and Bora Shin
    Competing interests
    No competing interests declared
  2. Misung Kim

    Laboratory of Molecular Environmental Microbiology, Department of Environmental Science and Ecological Engineering, Korea University, Seoul, Republic of Korea
    Contribution
    Conceptualization, Investigation, Methodology, Writing - original draft, Writing - review and editing
    Contributed equally with
    Jaeeun Park and Bora Shin
    Competing interests
    No competing interests declared
  3. Bora Shin

    Laboratory of Molecular Environmental Microbiology, Department of Environmental Science and Ecological Engineering, Korea University, Seoul, Republic of Korea
    Contribution
    Conceptualization, Investigation, Methodology, Writing - original draft
    Contributed equally with
    Jaeeun Park and Misung Kim
    Competing interests
    No competing interests declared
  4. Mingyeong Kang

    Laboratory of Molecular Environmental Microbiology, Department of Environmental Science and Ecological Engineering, Korea University, Seoul, Republic of Korea
    Contribution
    Investigation, Methodology
    Competing interests
    No competing interests declared
  5. Jihye Yang

    Laboratory of Molecular Environmental Microbiology, Department of Environmental Science and Ecological Engineering, Korea University, Seoul, Republic of Korea
    Contribution
    Investigation, Methodology
    Competing interests
    No competing interests declared
  6. Tae Kwon Lee

    Department of Environmental Engineering, Yonsei University, Wonju, Republic of Korea
    Contribution
    Methodology
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-3845-7316
  7. Woojun Park

    Laboratory of Molecular Environmental Microbiology, Department of Environmental Science and Ecological Engineering, Korea University, Seoul, Republic of Korea
    Contribution
    Conceptualization, Supervision, Funding acquisition, Writing - original draft, Project administration, Writing - review and editing
    For correspondence
    wpark@korea.ac.kr
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-3166-1528

Funding

National Research Foundation of Korea (NRF-2020M3A9H5104237)

  • Woojun Park

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank the group of Dr. Kwansoo Ko for providing the 5 clinical isolates from patients at Samsung Medical Center, Sungkyukwan University School of Medicine, and 35 clinical isolates from the National Culture Collection for Pathogens (NCCP). Funding: This work was supported by grants from the National Research Foundation of Korea (no. NRF-2020M3A9H5104237). The funders had no role in study design, data collection or interpretation, or the decision to submit the work for publication.

Senior Editor

  1. Gisela Storz, National Institute of Child Health and Human Development, United States

Reviewing Editor

  1. María Mercedes Zambrano, CorpoGen, Colombia

Publication history

  1. Received: January 28, 2021
  2. Accepted: June 27, 2021
  3. Accepted Manuscript published: June 28, 2021 (version 1)
  4. Version of Record published: July 30, 2021 (version 2)

Copyright

© 2021, Park et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 397
    Page views
  • 102
    Downloads
  • 0
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

    1. Epidemiology and Global Health
    2. Microbiology and Infectious Disease
    Barbara H Stokes et al.
    Research Article Updated

    The emergence of mutant K13-mediated artemisinin (ART) resistance in Plasmodium falciparum malaria parasites has led to widespread treatment failures across Southeast Asia. In Africa, K13-propeller genotyping confirms the emergence of the R561H mutation in Rwanda and highlights the continuing dominance of wild-type K13 elsewhere. Using gene editing, we show that R561H, along with C580Y and M579I, confer elevated in vitro ART resistance in some African strains, contrasting with minimal changes in ART susceptibility in others. C580Y and M579I cause substantial fitness costs, which may slow their dissemination in high-transmission settings, in contrast with R561H that in African 3D7 parasites is fitness neutral. In Cambodia, K13 genotyping highlights the increasing spatio-temporal dominance of C580Y. Editing multiple K13 mutations into a panel of Southeast Asian strains reveals that only the R561H variant yields ART resistance comparable to C580Y. In Asian Dd2 parasites C580Y shows no fitness cost, in contrast with most other K13 mutations tested, including R561H. Editing of point mutations in ferredoxin or mdr2, earlier associated with resistance, has no impact on ART susceptibility or parasite fitness. These data underline the complex interplay between K13 mutations, parasite survival, growth and genetic background in contributing to the spread of ART resistance.

    1. Genetics and Genomics
    2. Microbiology and Infectious Disease
    Joshua C D'Aeth et al.
    Research Article Updated

    Multidrug-resistant Streptococcus pneumoniae emerge through the modification of core genome loci by interspecies homologous recombinations, and acquisition of gene cassettes. Both occurred in the otherwise contrasting histories of the antibiotic-resistant S. pneumoniae lineages PMEN3 and PMEN9. A single PMEN3 clade spread globally, evading vaccine-induced immunity through frequent serotype switching, whereas locally circulating PMEN9 clades independently gained resistance. Both lineages repeatedly integrated Tn916-type and Tn1207.1-type elements, conferring tetracycline and macrolide resistance, respectively, through homologous recombination importing sequences originating in other species. A species-wide dataset found over 100 instances of such interspecific acquisitions of resistance cassettes and flanking homologous arms. Phylodynamic analysis of the most commonly sampled Tn1207.1-type insertion in PMEN9, originating from a commensal and disrupting a competence gene, suggested its expansion across Germany was driven by a high ratio of macrolide-to-β-lactam consumption. Hence, selection from antibiotic consumption was sufficient for these atypically large recombinations to overcome species boundaries across the pneumococcal chromosome.