We analyzed the general biological characteristics of A. baumannii strains 119, 108, and 176 isolated from ICUs. These strains show perceptible differences in mucoid abundance by viscous string analysis (Figure 1A), classifying as hypermucoid (HM) (A. baumannii 119), medium mucoid (MM) (A. baumannii 108), and low mucoid (LM) (A. baumannii 176). The mucoviscosity were further confirmed based on the low-speed centrifugation method (Figure 1B). Besides, all three isolates were subject to the whole genome sequencing (WGS). Virulence factors of pathogenic bacteria (VFDB) analysis of the WGS data reveled the absence of csuA/BABCDE locus [relating to the capability of biofilm formation and immune evasion (18)] in the HM strain, which was consistent with the phenotype of poor biofilm-forming ability and high mortality to Galleria mellonella (Figures 1C,D and Supplementary Figure 1). However, although there is no resistance genes including aac(6’)-lb-cr, msr(E), mph(E), and sul1/sul2, the HM strain shows resistance to ciprofloxacin, clindamycin, erythromycin, and trimethoprim/sulfamethoxazole (Table 1, Supplementary Table 4, and Supplementary Figure 1A). It consists with the previous observation that decreased drug penetration contributes to antibiotic resistance in mucoid A. baumannii isolates (6). Taken together, the HM strain is MDR with poor biofilm-forming ability and high toxicity.

We evaluated the growth rate of LM, MM, and HM isolates under either acid or hydrogen peroxide (H₂O₂) conditions, respectively. Both the LM and the MM strains enter into the stationary phase after 20 h, whereas the LM strain shows a sharp logarithmic phase (Figure 2A). In contrast, the HM strain remains at the logarithmic growth phase at 24 h, which may be due to high metabolic cost of mucus production. Moreover, the LM strain is more sensitive to H₂O₂ than the MM strain (Figure 2B), consisting with the lower transcript levels of catalase associated genes katE and katG in LM than MM (Figure 2C). Meanwhile, the expression level of katE and katG are highly associated with the mucoid phenotype (Figure 2C). These results demonstrate that the mucoid A. baumannii isolates are tolerance to the exposure of either acid or hydrogen peroxide.

To compare the invasion of these A. baumannii isolates, we co-cultured the strains with mouse lung macrophages (MH-S), mouse monocyte macrophages (RAW 264.7), and human alveolar basal epithelial cells (A549), respectively. Given that the growth of these strains shows no difference in cell culture media (Supplementary Figure 2), the LM, MM, and HM strains were first incubated with MH-S and RAW 264.7, respectively. Colistin (100 μg/mL) was used to eradicate the extracellular bacteria without causing cytotoxicity to mammalian cells (Supplementary Figure 3). The intracellular bacteria always appeared early in the macrophages infected with LM (Figures 3A–C), suggesting a positive relevance between mucoid phenotype and anti-phagocytic ability. The viable counts of internalized LM is higher than the others. Meanwhile, we excluded the explanation that the cytotoxicity of LM, MM, and HM to cells is response for such difference (Supplementary Figure 4). Moreover, we found that there is less intracellular HM in epithelial cells as well (Figure 3D). To further explore whether mucoid bacteria are resistance to phagocytosis, we examined the anti-phagocytic ability of K. pneumoniae. Compared to the low mucoid K. pneumoniae WNX-2, high mucoid K. pneumoniae ATCC 43816 hardly invade macrophages (Supplementary Figure 5), indicating a general behavior of anti-phagocytosis in pathogens with the mucoid phenotype. Altogether, we find that the mucoid A.baumannii shows anti-phagocytic activity to both macrophages and epithelial cells.

The increased C3b deposition on bacterial surface facilities phagocytosis through enhancing opsonization (19). We evaluated the relative abundance of C3b deposition on mucoid bacteria using flow cytometry. Results show that the strain with lower mucus has a higher abundance of C3b deposition among the LM, MM, and HM strains. C3b on LM was nearly three-time higher than the others (Figure 4A). Given that the HM and MM strains show no difference in C3b positive signals, we hypothesized that the consumption of C3b could also reduce its deposition on bacteria. Compared to the MM strain, the transcriptional level of translation elongation factor (tuf) (relating to C3b hydrolysis) is relatively high in the HM isolate (Figure 4B), indicating the greater consumption of C3b in HM (20). Besides, the similar content of lipooligosaccharide (LOS) in the LM, MM, and HM isolates, implied that the release of LOS is not a major cause of the different anti-phagocytic activities (21, 22). These results suggest that the HM strain shows potent anti-phagocytic activity through the reduction of C3b deposition and high potential of C3b hydrolysis.

Capsular polysaccharide mediates anti-phagocytic activities by reducing the C3b deposition on bacterial surface (23, 24). Therefore, we investigated the capsular difference in LM, MM, and HM using the zwitterionic TPE-Pn⁺⁺ (with strong membrane-penetrating capability) and monocharged TPE-N⁺ (unable to stain bacteria with capsular) (25). Results indicate that the HM strain carries a thick capsule (Figure 5A and Supplementary Figure 7). The capsular differences are further evaluated using the alcian staining. The MM strain produces the highest amount of capsule, and the HM strain shows the highest molecular weight of the produced capsule (Supplementary Figure 11), indicating that the yield and composition of capsule varies in mucoid isolates.

The MM isolate has less C3b deposition with high capsular production, consisting with the decreased C3b deposition potential in high capsular strains (Figure 4). To further decipher the correlation between C3b deposition and capsular production, we introduced a capsular reversible and hyper-production A. baumannii model (Figure 5B). The inducing agent (chloramphenicol) has no cytotoxicity to macrophages (Supplementary Figures 8B, 9). We observed that the decreased C3b deposition potentiated the anti-phagocytic activity in capsular hyper-produced A.baumannii (Figures 5C,D).

The efficiency of C3b depositionis modulated by the capsular structure including the hydroxyl group and the backbone length of polysaccharide chain (26–28). According to the capsular classification database (27–29), the HM, MM, and LM isolates are classified as types of K2, K82, and K14 (Figure 5E), respectively. Compared to K82, K2 has shorter backbone while longer length of polysaccharide (10), resulting in a better anti-phagocytic activity through effectively inhibiting the deposition of C3b (Figure 3). Altogether, the abundance and composition of capsule reduces the deposition of C3b, resulting an anti-phagocytic activity in the hyper mucoid A. baumannii.

All bacterial strains used in this study were listed in Supplementary Table 1. Routinely, bacteria were cultured at 37°C in brain heart infusion (BHI) (Beijing Land Bridge Technology, Shanghai, China) medium with shaking at 200 rpm (revolution per minute). A549 and RAW 264.7 cells (Supplementary Table 2) were cultured in Dulbecco’s modified Eagle’s medium (DMEM) (Thermo Fisher Scientific, Waltham, MA, United States). MH-S cells (Supplementary Table 2) were cultured in Roswell Park Memorial Institute (RPMI)-1640 medium (Thermo Fisher Scientific, Waltham, MA, United States). All the media were supplemented with 10% heat inactivated fetal bovine serum (FBS) (Invitrogen, Thermo Fisher Scientific, Waltham, MA, United States) and 1% (w/v) penicillin-streptomycin (Solarbio Life Science, Shanghai, China).

The A. baumannii isolates were evaluated by string test as described previously (41). All tested strains were cultured on sheep blood agar plate (5%) overnight at 37°C, then a single bacterial colony was stretched with an inoculation loop. The mucoviscosity assay was performed by low-speed centrifugation (39). Briefly, the tested strains were incubated in Luria-Bertani broth (LB) (Beijing Land Bridge Technology, Shanghai, China) at 37°C with shaking overnight. Then cultures were centrifuged at 1,000 g for 5 min. The absorption of supernatant was measured under the wavelength of 600 nm using an Infinite M200 Microplate reader (Tecan).

The biofilm abundance was detected following a previously described method with some modifications (42). Briefly, 100 μL of 1 × 10⁶ CFUs/mL A. baumannii strains was cultured in Mueller–Hinton broth (MHB) (Land Bridge Technology, Beijing, China) at 37°C for 24 h. The densities of bacteria transferred to new wells were measured under the wavelength of 600 nm by Infinite M200 Microplate reader. Then the original wells were washed three times with sterile phosphate buffered saline (PBS), following the air-drying, and 1% crystal violet stanning. The bound dye was resolubilized in 95% ethanol and the absorption of the dye solution was measured under the wavelength of 600 nm by Infinite M200 Microplate reader.

The virulence of A. baumannii isolates were evaluated in vivo using the G. mellonella larvae infection model as previously described (43). The healthy larvae (0.25–0.35 g) of G. mellonella (purchased from Huiyude Biotech Company, Tianjin, China) were randomly divided into four groups (n = 7 per group) and infected with 10 μL of A. baumannii strains suspension (1.0 × 10⁶CFUs) at the right pleopod and the other groups were injected with an equal volume of PBS. Survival rates of G. mellonella were recorded for 2 days.

The antibacterial susceptibility test was performed by broth microdilution according to the Clinical and Laboratory Standards Institute (CLSI) guideline (44). Briefly, antibiotics were two-fold diluted in MHB and mixed with an equal volume of bacterial suspensions in MHB containing approximately 1.5 × 10⁶ CFUs/mL in a clear, UV-sterilized, 96-well plate. After 16–20 h incubation at 37°C, minimal inhibit concentration (MIC) were defined as the lowest concentrations of antibiotics with no visible growth of bacteria.

For acid tolerance assay, fresh prepared bacteria were dilution by 1:100 in LB broth (pH 7.0), mixed with an equal volume of pH 4.0 medium in a 96-well microplate. The growth dynamics were recorded under the wavelength of 600 nm with an interval of 1h at 37°C measured by Infinite M200 Microplate reader.

Fresh prepared cultures were adjusted to McFarland turbidity of 0.5 and diluted in 4 mL BHI broth. The bacterial suspensions were treatment with 50 mmol/L H₂O₂ (Sinopharm Chemical Reagent Co., Shanghai, China) for 30 min, following by plating serial dilution on BHI agar plates. Then the CFUs were counted after incubating at 37°C for 24 h.

MH-S cells were plated on glass coverslips (14 mm, NEST Life and Science Technology Co., Wuxi, China) in 24-well culture plates to form monolayers. Then the cells were infected with pHrodo Green-labeled (Invitrogen, Thermo Fisher Scientific, Waltham, MA, United States) A. baumannii strains [multiplicity of infection (MOI) = 10] for 4 h, following fixating in 4% paraformaldehyde for 20 min. F-actin and nuclei were labeled with ActinRed⁵⁵⁵ ReadyProbes (Invitrogen, Thermo Fisher Scientific, Waltham, MA, United States) and DAPI (Beyotime Biotechnology, Shanghai, China), respectively. Images were captured using a Leica SP8 confocal microscope, and Z-axis sections were cut every 3 μm to analyze the location of internalized bacteria. Images were analyzed and merged by LAS AF Lite software (Leica Biosystems, Germany).

Bacterial imaging assay was performed as described previously (25). Specifically, fresh prepared bacterial cultures were washed and resuspended with 200 μL PBS. Then the bacterial solutions were transferred into a sterilized EP tube with probe solutions (20 μmol/L) and incubated at room temperature for 30 min. After that, 10 μL of the stained bacteria was transferred to a piece of clean glass slide and then covered by a coverslip for fixation. Images were collected and analyzed by LAS AF Lite software.

The cell infection assay was performed as described previously, with some modifications (45). Mammalian cells with 4 × 10⁵ were seeded at 24-well plates to form monolayers. Then, bacterial resuspensions were diluted in DMEM or RPMI-1640 medium supplemented with 1% FBS and cocultured with cells at an MOI of 10. At the end of the trials, cells were incubated for an additional 30 min with 100 μg/mL colistin to remove the extracellular bacteria. After washing with PBS, the cells were lysed by DMEM or RPMI-1640 medium supplemented with 0.1% Triton-X 100 (Beyotime Biotechnology Co., Shanghai, China). The harvested bacteria were plating on BHI agar plates with different dilutions for the Colony-count technique to quantify the number of internalized bacteria. In the capsule-induced model, chloramphenicol (10 μg/mL) was added to the bacterial suspensions for the stress maintenance.

Extraction of A. baumannii capsule was performed as described previously with some modifications (46). Briefly, cultures were resuspended with 200 μL lysis buffer (60 mmol/L Tris, pH8; 10 mmol/L MgCl₂; 50 μml/L CaCl₂; 20 μg/mL DNase and RNase; and 3 mg/mL lysozyme), then incubated at 37°C for 1 h. Following vortex and three repeated liquid nitrogen/37°C freeze-thaw cycles, additional DNase and RNase were added and incubated at 37°C for 30 min. About 10 μL 10% SDS was then added and incubated at 37°C for another 30 min. The suspensions were boiled at 100°C for 10 min and then incubated at 60°C with protease K for 1 h. After centrifugation, the supernatants were retained and precipitated overnight in pre-cooling 75% ethanol, followed by pelleting, air-drying, resuspending with SDS sample buffer at a volume normalized based on OD₆₀₀ and boiling for 5 min.

Samples were separated on 4–20% BioRad TGX Tris-glycine gels (Bio-RAD, Hercules, CA, United States). After electrophoresis, the gel was washed with deionized water and stained with a solution of 0.1% (w/v) of Alcian Blue 8GX (Sigma-Aldrich, Merck, Germany) for 1 h. Gels were decolorized by placing in a pH 4.75 solution containing 40% ethanal and 60% 20 mmol/L sodium acetate for overnight.

Method of capsule induction was performed as described previously, with some modifications (10). About 10 μg/mL chloramphenicol was added to logarithmic phase bacteria. After overnight incubation, the capsule of A. baumannii strains was extracted and analyzed with alcian blue staining.

For quantifying the C3b deposition, previously described method was used with some modification (47). Briefly, A. baumannii isolates were cultured overnight and adjusted to McFarland turbidity 0.5, then 100 μL bacterial suspension was mixed with 100 μL mouse serum and incubated at 37°C for 30 min. After PBS washing, samples were incubated with antibodies against mouse complement factor C3b (Thermo Fisher Scientific, United States) and incubated with a secondary fluorescent antibody for another 30 min subsequently. Samples were then resuspended with PBS and analyzed using Becton-Dickinson FACS Canto II flow cytometer. The gating on single cells with positive gates established at a fluorescence excluding 99% of the isotype control samples.

Bacterial total RNA was extracted and examined using M5 EASYspin Plus kit (Mei5bio, Beijing, China) and Nanodrop spectrophotometer (Thermo Scientific, MA, United States), respectively. Reverse transcription was performed using a PrimeScript RT reagent Kit with gDNA Eraser (Takara, Beijing, China) with the manufacturer’s protocol. The messenger RNA levels relative to those of the control genes 16S were determined by real-time PCR tests with PowerUp SYBR Green Master Mix (Applied Biosystems, Thermo Fisher Scientific, Carlsbad, CA, United States). RT-PCR tests were performed using the ABI Quantstudio 7 detection system (Applied Biosystems, Thermo Fisher Scientific, Carlsbad, CA, United States). The fold changes in gene expression were determined using the 2^–ΔΔCt method. Primers used in this study were listed in Supplementary Table 3.

As previously described (48), the capsular K-type of related A. baumannii isolates were analyzed using wzc gene BLAST.

Statistical analysis was performed using GraphPad Prism 7.0 (GraphPad Software, Inc.). All data were expressed as the mean ± SD and unless otherwise noted, unpaired t-test between two groups were used to calculate p-values.