The bacterial strains and plasmids used in this study are listed in Table 1. The primers used in this study are listed in Supplementary Table S1. Bacteria were routinely cultured at 37°C in Mueller-Hinton (MH) medium at 37°C with constant shaking unless otherwise indicated.

The whole genome sequence of A. baumannii ATCC17978 was obtained from GenBank (Smith et al., 2007) and adopted to manipulate the A. baumannii ATCC 17978 genome. The hfq::Km mutants were constructed by a previously described method (Aranda et al., 2010). Briefly, polymerase chain reaction (PCR) was used to amplify regions of the sequences upstream and downstream of the hfq gene using the primer pairs (Supplementary Table S1) Up_Hfq_F and Hfq_kan_R to amplify the 311-bp region upstream of the hfq gene and Kan_Hfq_F and Dw_Hfq_R to amplify the 220-bp region downstream of the hfq gene. The primers Hfq_kan_R and Kan_Hfq_F have approximately 22 nucleotides of overlapping sequence, resulting in fusion of the hfq upstream and downstream sequences with the kanamycin resistance gene after fusion PCR. The 311 and 220-bp DNA fragments containing the overlapping sequences from the km gene were fused together by PCR with a complementary km PCR fragment, resulting in the replacement of hfq with km. The resulting PCR product was gel purified using a gel extraction kit (Favorgen, PIF, Taiwan). The fused PCR fragment was electroporated into wild-type A. baumannii ATCC 17978. The Δhfq strains were identified by screening transformants on LB agar plates containing kanamycin (50 μg/μl). The deletion mutant was confirmed using conventional PCR for diagnostic size shifts by gel electrophoresis and was further confirmed by real-time reverse transcription PCR (RT-PCR).

The hfq parental allele was PCR amplified using the primers hfq_Expr_XhoI_F and hfq_Expr_XbaI_R and ligated into the A. baumannii–E. coli shuttle vector pWH1266_km (Lin et al., 2014). The recombinant DNA products were verified by sequencing. The resulting plasmid pHFQ was used to transform the Δhfq strain by electroporation with selection by tetracycline. The expression of the hfq gene in the Δhfq mutant and its complementation clone Δhfqc was verified using real-time RT-PCR by the method previously described (Chang et al., 2014). The relative expression level of hfq was normalized using that of 16s rRNA.

Overnight bacterial cultures were diluted 1:100 and subcultured for 3 h. The bacteria were recovered by centrifugation at 3,000g for 10 min, washed with deionized water, and suspended in deionized water at an OD₆₀₀ of 0.07. Droplets of each sample were deposited on glass slides for 10 min. Then, the droplets were removed, and the slides were dried for 20 min to immobilize the cells. The morphology of the bacteria was analyzed by atomic force microscopy (AFM; Dimension Icon, Bruker, Billerica, MA, United States) using an SNL-W-D triangular cantilever probe (Bruker) with a nominal 0.06 N/m spring constant and 2-nm tip curvature radius. Measurements were obtained in peak force tapping mode (or Quantitative NanoMechanics mode, QNM mode). The images obtained were analyzed using NanoScope Analysis 1.4 software (Bruker). The AFM scanning sizes were 3 μm × 3 μm with scan rates of 1 Hz. The spatial resolution of image is 5.86 nm/pixel. For each measurement, height sensor and peak force error images were recorded simultaneously. The height sensor image specifies fimbria morphology, while the peak force error image emphasizes edges of bacterial structures.

Outer membrane vesicles (OMVs) are isolated using the method of Rosen et al. (1995). Protein concentrations were determined and then equal volume of OMV content in each strain was analyzed by 10% SDS–PAGE.

A. baumannii strain ATCC17978 (wild-type) and the Δhfq and Δhfqc strains were cultured overnight in MH medium at 37°C, and the overnight culture was subcultured at 1:100 in 100 ml of MH medium. The cultures were incubated with shaking at 200 rpm in 250-ml Scott flasks at 37°C for 22 h. Every hour, the optical density was measured at 600 nm (OD₆₀₀).

For the stress tolerance test, overnight cultures were diluted to an OD₆₀₀ of 0.01 with MH broth. For temperature effects, cells were cultured at 20 and 50°C, respectively. For high osmolarity and oxidative stress effects, NaCl and H₂O₂ were added to the bacterial cell suspension at concentrations of 3% and 20 mM, respectively. In addition, ethanol and Triton X-100 were added to the bacterial suspension at concentrations of 3 and 1%, respectively. For acid and alkaline stress conditions, cells were cultured at pH 5.5 and pH 8.5, respectively. Bacterial growth with or without external stress was monitored by measuring the OD₆₀₀. After treatment with H₂O₂ for 30 min, the cells were diluted and plated on LB plates to determine the number of colony forming units (CFUs). The results represent the mean of at least three separate experiments.

To study the effect of the hfq mutation on the mRNA levels of basD, bauA, uspA, nlpE, A1S_0820, carO, groEL, and ompA, overnight cultures of the wild-type, Δhfq, and Δhfqc strains were washed, and total RNA was isolated for real-time RT-PCR.

Biofilm formation on polystyrene tubes was assessed by crystal violet staining of cells cultured in LB broth as previously described. The expression of the csuA/B gene (A1S_2218) and fim-like adhesion gene (A1S_1507) in the Δhfq mutant and its complementation clone Δhfqc was verified using real-time RT-PCR by the method previously described (Chang et al., 2014). The relative expression levels of A1S_2218 and A1S_1507 were normalized by that of 16s rRNA.

The assay was performed as described with some modifications (Wang et al., 2014). Briefly, mouse macrophage cells (RAW264.7) were cultured to 1 × 10⁶ cells/well in 24-well plates in RPMI 1640 medium. The overnight culture of A. baumannii was applied to each well at an MOI of 10. Bacteria were co-incubated with macrophages for 120 min at 37°C. After infection, the cells were washed and incubated for 1 h with streptomycin (250 μg/ml) containing medium. The CFU obtained from the lysate of the wild-type-infected macrophages was set as 100%, and other data were relative to this value.

Cell adhesion and invasion assays were performed as described with some modification (Wang et al., 2014). A549 (Human lung carcinoma cell line, epithelial cell; BCRC 60074) and NCI-H292 (human mucoepidermoid pulmonary carcinoma cell line, endothelial cell; BCRC 60372) were purchased from Food Industry Research and Development Institute, Hsinchu, Taiwan. Briefly, overnight bacterial cultures were diluted 100-fold and grown for 2 h. A549 cells and NCI-H292 cells were infected at 37°C for 2 h with bacterial suspensions at a multiplicity of infection (MOI) of 10. For the adhesion assay, infected monolayers were washed with PBS to remove non-adherent bacteria and lysed with 1% Triton X-100. The cell lysates were plated on LB agar plates to determine the total CFU associated with the cells. For the invasion assay, the infected cells were washed and further incubated at 37°C for 2 h in RPMI 1640 medium containing streptomycin (500 μg/ml) to kill extracellular bacteria. Cells were lysed and plated on LB agar to quantify viable invading bacteria. The adhesion ability was expressed as the percentage of adherent bacteria versus the total inoculum, and the invasion ability was expressed as the percentage of viable bacteria that survived the streptomycin treatment versus the total inoculum.

Overnight bacterial cultures were diluted 100-fold and grown for 2 h. A549 cells and H292 cells were infected at 37°C for 1 h at an MOI of 10. The infected cells were collected, and total RNA was isolated using an RNeasy minikit (Qiagen). The mRNA expression of IL-8 and IL-6 in A549 cells and H292 cells was determined by real-time RT-PCR as described (Kim et al., 2015). The relative amounts of cytokines in each infected cell line were normalized to the expression levels of the GAPDH gene.

The means of group differences were determined using a one-way ANOVA. A p-value < 0.05 was considered statistically significant. Data entry and analyses were performed using the Statistical Package for the Social Sciences software version 21.0 (SPSS Inc., Chicago, IL, United States).

Several studies have demonstrated that Hfq participates in the pathogenesis of many bacteria via different mechanisms. Here, we studied the effects of an Hfq-like protein on the pathogenesis of A. baumannii ATCC 17978. The gene (A1S_3785) homolog of the hfq gene, which is predicted to encode a protein of 168 amino acids, is located at 2,465,843–2,466,349 bp in the genome of A. baumannii ATCC17978 (Eijkelkamp et al., 2011). This gene (A1S_3785) and its encoded protein are almost twice the size of other gammaproteobacterial Hfqs due to an elongated C-terminus (Supplementary Figure S1). The N-terminal domains of Hfq in most bacteria are highly conserved between amino acid residues 1 and 66 (Supplementary Figure S1), a region that is responsible for RNA-binding and protein–protein interactions among Sm [the core of small nuclear ribonucleoprotein particles (snRNPs)] and Sm-like proteins (Moller et al., 2002). By contrast, the C-terminal end of A. baumannii Hfq contains a glycine-rich domain with a repetitive pattern, which was similar to the A. baylyi hfq gene encoding a usual C-terminus (Schilling and Gerischer, 2009). Compared to the Hfq protein lengths in several bacteria, the largest known Hfq proteins are annotated in members of the gammaproteobacterial family Moraxellaceae (Supplementary Table S1). The annotated Hfq lengths are between 168 and 174 amino acids for Acinetobacter and up to 210 amino acids for M. catarrhalis. Although Hfq shows a strong variation in its C-terminus (Vecerek et al., 2008), this C-terminus of Hfq in A. baylyi was not responsible for growth and cell phenotype (Schilling and Gerischer, 2009). To examine the role of Hfq in A. baumannii, we generated an isogenic mutant by the replacing the conserved hfq sequence with the kanamycin resistance gene (Δhfq) and an Hfq-complemented strain (Δhfqc), as shown in Figure 1A. The loss of hfq was verified by PCR and sequencing (Figure 1B). The loss and recovery of hfq expression was confirmed in the Δhfq and Δhfqc strains by real-time RT-PCR (Figure 1C).

The cell morphologies of the three A. baumannii strains (WT, Δhfq, and Δhfqc) were visualized from the AFM peak force error images, as shown in Figure 2A. Δhfq expressed very few fimbriae (white arrow), compared to WT and Δhfqc strains. Figure 2B shows the comparison of the cell sizes by estimating the bacterial spreading areas from the AFM height sensor images as shown in Figure 2A. The cell area (A) and number (n) are as following: WT (A: 1.82 ± 0.27 μm², n: 5) and Δhfqc (A: 2.03 ± 0.57 μm², n: 5) strains were both significantly larger than Δhfq (A: 0.98 ± 0.17 μm², n: 6) (p < 0.01). This result reveals that a lack of hfq could restrain the growth of A. baumannii. There was no significant size difference between WT and Δhfqc. OMVs (dashed white circle) were dominant in wild-type and were greatly reduced in the Δhfq strain.

To verify that Δhfq strain reduced OMVs secretion, the quantitative assay of OMVs was performed and the result was shown in Figure 2C. The total protein content of OMVs was largely reduced in Δhfq strain compared to WT and Δhfqc strains. Based on the study of Jin et al. (2011), OmpA of A. baumannii is the most abundant protein in OMVs with a molecular mass of 38 KDa. Taken together, the levels of OmpA were reduced in Δhfq strain compared to WT and Δhfqc strain.

Growth defects resulting from the loss of hfq have been demonstrated in several bacterial species (Sonnleitner et al., 2003; Geng et al., 2009; Wang et al., 2014). To evaluate the growth effects in hfq deletion mutants, we monitored the growth of the wild-type, Δhfq mutant, and Δhfqc strains in MH broth. The loss of hfq resulted in growth retardation of 7 h (Figure 3). On MH agar plates, except for colony size, other colony characteristics such as form, elevation, and margin did not differ between hfq mutant and wild-type strain (data not shown).

Acinetobacter baumannii has gradually become a leading nosocomial pathogen worldwide. The widespread dissemination of A. baumannii in hospital environments indicates that this organism might tolerate environmental stresses. To determine if Hfq contributes to external stress tolerance, the growth of the WT, Δhfq, and Δhfqc strains was assessed under various stress conditions. Hfq mutants showed defects in growth upon exposure to all stressors (Figure 4). The reduced tolerance of the Hfq mutant was restored to near wild-type levels by introduction of the Hfq-expressing plasmid. This result implies that Hfq contributes to the resistance to stresses such as temperature, pH, osmotic pressure, and oxidative stress in A. baumannii.

As Hfq contributes to external stress tolerance, the effect of Hfq on the mRNA expression of the genes involved in stress tolerance, including the acinetobactin-mediated iron acquisition system such as bas and bau, and growth-related stresses such as uspA, nlpE, A1S_0820, carO, groEL, and ompA, was evaluated (Figure 5). Our result showed that the expression of all genes, including basD, bauA, uspA, nlpE, A1S_0820, carO, ompA, and groEL, was significantly higher in Δhfq (p < 0.05) than in WT and was completely or partially restored in the Δhfqc strain. These data indicate that Hfq is required to modulate the expression of stress-related genes to survive stress.

Acinetobacter baumannii form biofilms, a phenotype that may explain its ability to survive in nosocomial environments and to cause device-related infections in compromised patients (Lee et al., 2008). As shown in Figure 6A, when hfq was deleted, biofilm formation was somewhat decreased compared with wild-type. Fimbria assembly plays an important role in biofilm initiation and maturation after initial attachment to abiotic surfaces (Gaddy and Actis, 2009). Two types of fimbria, the CsuA/B-A-B-C-D-E chaperone-usher secretion system and fim-like type 1 fimbria (A1S-1507 to A1S_1510), have been reported in A. baumannii (Tomaras et al., 2003; Nait Chabane et al., 2014). CsuA/BABCDE-dependent fimbriae are involved in biofilm formation, whereas the function of fim-like fimbria remains unclear. To determine if Hfq plays a role in fimbriae production, the expression levels of csuA/B and A1S_1507 in the wild-type and Δhfq and Δhfqc strains were examined by real-time RT-PCR as shown in Figure 6B. CsuA/B levels were significantly reduced (p < 0.01), whereas A1S_1507 levels were significantly increased (p < 0.01) in the Δhfq strain compared with the wild-type and Δhfqc strain, indicating that Hfq is required to modulate fimbriae production in different ways.

Acinetobacter baumannii colonizes biotic surfaces by adherence and invasion to host cells to evade immune attacks and subsequent bacterial persistence (Lee et al., 2006). To test the role of Hfq in A. baumannii virulence, the ability of wild-type and Δhfq mutant cells to adhere to and invade A549 lung adenocarcinoma cells and NCI-H292 human bronchial epithelial cells was assessed. Δhfq exhibited impaired abilities to adhere to and invade A549 and NCI-H292 cells (Figures 7A,B). IL-8 and IL-6, which cause neutrophil infiltration and non-resolving inflammation, are key cytokines contributing to bacterial elimination by host cells and are expressed in the epithelial lining (Oglesby et al., 2015). IL-6 acts as both a pro-inflammatory cytokine and an anti-inflammatory myokine (Scheller et al., 2011), and IL-8 plays an important role in acute inflammation (Baggiolini and Clark-Lewis, 1992). Because Hfq plays a role in A. baumannii adherence and invasion of lung epithelial cells, we examined the production of IL-8 and IL-6 by A549 and NCI-H292 cells challenged with wild-type or Δhfq mutants. As shown in Figure 7C, IL-8 expression was significantly increased (p < 0.001) in both cells upon exposure to ΔHfq. However, IL-6 expression was significantly decreased (p < 0.01) in A549 cells but increased (p < 0.05) in NCI-H292 cells upon exposure to the Δhfq mutant compared with the WT and Δhfqc strains. This result implies that hfq deletion may stimulate IL-8 in A549 and NCI-H292 cells. Moreover, the deletion of hfq on IL-6 expression in both cells is different.

In order to understand if Hfq is involved in the innate ability of macrophage to eliminate A. baumannii, we co-cultured macrophages with bacteria, killed external bacteria with streptomycin, and assessed the survival of internalized A. baumannii after streptomycin treatment by lysing macrophage cells. As shown in Figure 7D, the relative survival of hfq mutant was reduced to 0.1% at 2 h after streptomycin treatment. The Hfq-complemented strain was partially restored the survival pattern compared to the wild-type.