All A. baumannii strains were isolated from a tertiary, first-class teaching hospital in East China, and the strain number, ST type, etc. are shown in Table S1 . 10 strains of ST191, 12 strains of ST195, 11 strains of ST208, 10 strains of ST462, and 9 strains of ST547, total 52 strains were used for each biological experiments. Two strains from each ST type above, plus two STn strains, a total of 12 strains were selected for subsequent genomic and transcriptomic analysis based on their ability to resist complement killing. The antimicrobial susceptibility of all A. baumannii isolates was determined using agar dilution according to the Clinical and Laboratory Standards Institute (CLSI) guidelines from 2020. Resistance breakpoints determined by the European Committee on Antimicrobial Susceptibility Testing (EUCAST) were used for polymyxin B.

Growth curves were performed as described previously with slight modifications (Hall et al., 2014) Briefly, A. baumannii in the log-phase of growth were adjusted to 0.5 absorbance units (OD600) and diluted 50-fold with LB medium. Aliquots (200 μl) were added to a 96-well plate, placed in a microplate reader at 37°C, and shaken for 5 s every 30 min, and the OD600 was measured for 24 h.

Desiccation resistance assays were completed as previously described (Boll et al., 2015) Log-phase A. baumannii were harvested and washed twice with an equal volume of LB medium. Each sample was adjusted to 1×108 CFU/ml and then serially diluted and plated to determine the input CFU. Bacterial suspension (10 μl) was spotted onto 96-well polystyrene plates and desiccated at 30°C at a humidity of 40%. PBS (200 μl) was added to each well to resuspend the bacteria after 48 h and 96 h, respectively, and the resuspended bacteria were serially diluted and plated to calculate the output CFU. The percent of survival was defined as the ratio of output to input CFUs.

A. baumannii in the log-phase of growth were adjusted to OD₆₀₀ = 0.5 and spread evenly on an MH agar plate. A sterile filter paper disk with a diameter of 6 mm was placed in the middle of the inoculated agar, and 10 μl 20% (v/v) hydrogen peroxide was pipetted onto the filter paper. The plates were incubated at 37°C for 18 h, after which the growth inhibition zone around the disk was measured using Vernier calipers.

Blood collected from healthy volunteers was centrifuged and filtered with a 0.22 μm pore size syringe filter to obtain sterilized serum. Half of the serum was inactivated by heating in a 56°C water bath for 30 min. A. baumannii in the log-phase of growth were adjusted to 2 × 10⁶ CFU/ml and mixed with normal or inactivated serum at a ratio of 1:9, respectively. After incubation for 1 h at 37°C, samples were serially diluted and spread on MH agar plates. After overnight incubation at 37°C, bacteria colonies were counted, and the bacterial survival rate was calculated by the following formula: Bacterial survival rate = (number of colonies in normal serum/number of colonies in inactivated serum) × 100%.

A. baumannii in the log-phase of growth were adjusted to 0.5 absorbance units (OD₆₀₀), before 200 μl aliquots were pipetted into the wells of a 96-well plate. Bacteria were cultivated in an incubator for 24 h at 37°C and 5% CO₂, to allow biofilm formation, before nonattached bacteria were removed from the wells by aspiration. Biofilms were gently washed with ddH₂O, after which the 96-well plate was inverted on absorbent paper for 15 min at room temperature. Biofilm cells were fixed by adding 4% (v/v) paraformaldehyde to each well for 20 min and were then stained with 1% (w/v) crystal violet for 15 min. Biofilms were washed thoroughly with ddH₂O, before the bound crystal violet was eluted with 200 μl 95% ethanol, and were then incubated at 37°C for 30 min. The crystal violet eluted from each biofilm was measured colorimetrically at OD₅₇₀.

A. baumannii in the log-phase were adjusted to 1× 10⁷ CFU/ml. G. mellonella larvae weighing ~300 mg was divided randomly into groups, with each group containing 15 larvae. Each larva was infected with 10 μl of the adjusted bacterial suspension, incubated at 37°C, and observed once every 12 h for 3 days. Larval survival status was assessed using the acupuncture method, and no response was considered as death.

A. baumannii genomes were extracted using QIAamp DNA Mini Kit (QIAGEN, Valencia, CA), whole genome sequencing was performed by Illumina HiSeq 2500 (Illumina, San Diego, CA), and the raw sequencing results were quality-controlled using FastQC v.0.11.5. Trimmomatic v.0.40 was used to trim splice regions. The trimmed reads were assembled and annotated by SPAdes(http://cab.spbu.ru/software/spades/) v.3.6 and RAST(https://rast.nmpdr.org/), respectively. Antibiotic resistance genes and virulence genes were identified using the ResFinder (Bortolaia et al., 2020) and VFDB databases (Chen et al., 2016), respectively, with 80% identity and 80% query coverage as the cutoff values.

A. baumannii RNA library preparation, construction, sequencing, and processing of reads were performed at the Novogene Co., Ltd. (Beijing, China). The raw data were first processed by a Perl script to remove reads with adapters, those containing poly-N and low-quality reads. The sequencing error rate for a single base position should be less than 1% and no more than 6% ( Table S2 ). All the analyses below were based on high-quality clean data.

Clean reads were mapped to A. baumannii MDR-ZJ06 (accession No. CP001937.2) using Bowtie2-2.2.3. Counting the reads numbers mapped to each gene was achieved using HTSeq v0.6.1. Differential expression analysis for the DST and NST groups was performed using the DESeq R package (1.18.0), and p-values were corrected using the Benjamini–Hochberg method to control for false discovery rate (Glickman et al., 2014). Genes for which corrected p-values <0.05 were obtained, after DESeq processing, were defined as differentially expressed.

Gene ontology (GO) enrichment analysis of DEGs was performed using the GOseq R package. GO terms with p-values less than 0.05 were considered significantly enriched for DEGs. Subsequently, the COG database was used to identify the functions of the proteins encoded by the DEGs. KOBAS software was used to test the statistical enrichment of DEGs in the KEGG pathway.

For analysis of expression of specific genes, A. baumannii RNA was first stabilized with RNA protection solution (Qiagen, 74124) and then extracted and purified according to the instructions of the RNeasy Mini Kit (Qiagen, 74104). One microliter RNA was added to nine microliter reaction buffer from the PrimeScript™ RT reagent Kit (Takara, RR037A) in the appropriate ratio, and then reverse transcribed to generate cDNA as follows: 37°C for 15 min, 85°C for 5 sec, and 4°C. cDNA was added to the reaction system of the TB Green^® Premix Ex Taq™ (Takara, 420A), and PCR was carried out as follows: 95°C, 30 s for one cycle (stage 1); 95°C, 5 s, 60°C, 34 s for 40 cycles (stage 2); and 95°C, 15 s, 60°C,1 min, 95°C 15 s (stage 3). The genes and primers used are shown in Table S3 , and recA was selected as the internal reference. The relative expression of different genes was calculated by the 2^-ΔΔCt method.

Statistics were performed using prism9. Two groups were compared using the Mann-Whitney t-test, and the five groups were compared using a one-way ANOVA, with horizontal lines representing means ± standard error of means; *p < 0.05, **p < 0.01, ***p < 0.001, **** p < 0.0001.The biological experiments were repeated at least three times, except for the growth curve experiment and anti-desiccation assay, where two biological replicates were done, and the biofilm experiment, where four biological replicates were done.

We first compared their growth and there was no difference in the growth rates of strains in the two groups ( Figure S1 , p = 0.42).

As resistance to desiccation and oxidative stress are essential for the persistence of A. baumannii in the medical environment, we compared their abilities in these two aspects. After 96 h of desiccation, strains in the DST group had a mean survival rate of 10.38%, while strains in the NST group had a mean survival rate of 5.13% (p < 0.0001; Figure 1A ). Moreover, the oxidative stress analysis showed that the hydrogen-peroxide-induced growth inhibition zone for the DST group was lower than that for the NST group (zone diameter 36.29 mm vs. 37.16 mm; p =0.027), indicating that the DST strains were more resistant to oxidative stress than the NST strains ( Figures 1B, C ).

The DST group had higher resistance to complement-mediated killing (mean survival 11.00% vs. 5.31%; p < 0.0001; Figure 1D ), but lower biofilm formation ability (OD₅₇₀ 0.31 vs. 0.42; p < 0.0001), than the NST group ( Figure 1E ).

When G. mellonella was used as an in vivo bacterial infection model, larvae infected with the DST group had a higher mean survival rate than those infected with the NST group in the early (12 h), middle (36 h), and late (72 h) stages of infection, however, all this difference was not statistically significant ( Figures 1F , S2 ).

All strains belonging to the DST group were resistant to all three cephalosporins, while the strains of the NST group displayed resistance rates of 33.34%, 66.66%, and 33.34% to ceftazidime, cefepime, and ceftriaxone, respectively. Similarly, the DST strains were all resistant to carbapenems, whereas 66.66% of the NST group strains showed resistance. However, for aminoglycosides, the resistance rate was 66.66% for the DST group and 16.66–33.34% for the NST group. All the DST group strains were resistant to quinolones, while the NST group strains had a resistance rate of 66.66%. The overall MICs of the DST group were also higher than those of the ST group for minocycline and combined antibiotics ( Table 1 ).

The general features of the twelve genomes were shown in Table S4 . The phylogenetic tree built using roary software (default parameters) indicates that strains of the same ST type were closely related, but ST208 of the DST group was evolutionarily distant from ST191 and ST195. ST547 of the NST group was also distant from ST462 and the STn type strains ( Figure S3 ).

The results appeared that the two groups shared significant virulence genes. The encoded products of which included adhesion-related outer membrane protein OmpA, serum resistance-associated penicillin-binding protein PbpG, cleavage cell membrane related phospholipase Plc and PlcD, iron acquisition protein BarAB, BauABCDE, AbaI and AbaR related to quorum sensing, PgaABCD involved in biofilm formation, the CPS synthesis component, Wza, Wzb and Wzc related to immune escape, etc ( Figure 2 ). This is consistent with the similar virulence seen for strains in the two groups in the G. mellonella infection model.

Twenty-four virulence genes associated with CPS were found only in strains of the DST group. These include glycosyltransferase-related genes itrB2, gtr3/4/5/8/20/21/22, kpsS, CMP-glycan pathway-related pseB/C/F/H/I, UDP-glycan pathway-related fnlA/fnlB, dehydrogenase pgt1, and isomerase wecB. However, 17 virulence-associated genes were found exclusively in strains from the NST group. These include 15 genes related to the CPS (CMP-glycan pathway-related legC, lgaA/F/G gene, glycosyltransferase gtr14/15/52, M3Q_295/296, and weeH) and two genes related to iron acquisition.

Overall, strains of the DST group harbored a greater number of CPS synthesis-related genes than those of the NST group, consistent with the increased resistance to complement killing and desiccation previously reported for DST group strains (Russo et al., 2010; Tipton et al., 2018). Many different genes related to the K locus, which is associated with CPS synthesis in A. baumannii, were unique to members of each of the two groups, consistent with the high variability previously reported for the middle region at the K locus (Kenyon and Hall, 2013).

To explore why the DST group strains had higher antibiotic resistance rates than those of the NST group, the profile of resistance genes in each strain was analyzed using the ResFinder database. Although strains in the two groups had several resistance genes in common, including genes encoding aminoglycoside, β-lactam, and sulfonamide resistance, many other resistance genes were found only in the DST group, correlating with the differences in drug susceptibility observed for the two groups. These unique genes included ant(3”)-Ia_1 and armA_1 (aminoglycoside resistance), blaTEM-1D_1 (β-lactam resistance), mph(E)_1 (macrolide resistance), and sul1_5 (sulfonamide resistance; Figure 3 ). Interestingly, the only resistance genes unique to the NST group were β-lactam-related genes, such as blaOXA-120_1, blaOXA-51_1, and blaOXA-531_1 ( Figure 3 ).

Transcriptome analysis showed that a total of 620 protein-coding genes were differentially expressed between the two groups (padj < 0.05), with 492 genes upregulated and 128 genes downregulated, in DST compared with NST, accounting for approximately 15% of all protein-coding genes ( Figure 4 ). After screening by log₂ FoldChange (DST/NST) ≥ 2, 128 genes were identified as being significantly upregulated in the DST group, including nine CPS synthesis-related genes, one efflux pump gene, and three T6SS secretion system-related genes. Loci corresponding to gene_ids for all these genes were identified in the genomes of the DST strains being analyzed ( Table S5 ). After screening by log₂ FoldChange (DST/NST) ≤-2, 53 genes were identified as being significantly downregulated in the DST group, including three genes related to CPS biosynthesis and six genes related to iron uptake ( Table S6 ). Meanwhile, we randomly selected two genes for Q-PCR validation, respectively. The results showed that the relative trend in the expression of these genes, in strains from the different groups, was consistent with the transcriptomic data ( Figures 4B, C ).

GO analysis showed that the most dramatically enriched gene set related to “lipid biosynthesis process,” which contains the wecB, capD, and galE genes involved in CPS synthesis, as well as the fabB, fabF, and psd genes involved in lipid syntheses. Other functional groups enriched in the DST upregulated genes group included “lipid metabolism process” and “carbohydrate transmembrane transport activity” ( Figure 5A ). The most significantly enriched, downregulated genes in the DST group were “potassium-transporting ATPase activity” and “ATPase activity, coupled to transmembrane movement of ions, phosphorylative mechanism,” with both of these functional sets containing the potassium ion transport genes kdpA and kdpC. Other functional sets with higher levels of enrichment in the DST downregulated genes group included “chemotaxis” and “taxis,” both of which include the chpA and pilI genes ( Figure 5B ).

Similar to the GO analysis, the COG analysis revealed that the “lipid transport and metabolism” category was enriched in the upregulated genes in the DST group ( Figure 6 ). Additionally, the functional categories with the next largest differences in the number of genes expressed between the two groups were “Transcription” and “Replication,”. Although upregulated genes outnumbered downregulated genes in many functional categories, the “intracellular trafficking, secretion, and vesicular transport” category only contained downregulated genes. This suggests that strong secretion and vesicular transport abilities may be less important to DST strains.

The KEGG pathway enrichment analysis indicated that the most significant pathway enriched for upregulated genes was “starch and sucrose metabolism” (p = 0.0054), which contains three genes, galU, ugd, and pgi, related to the CPS biosynthesis. This is in agreement with the fact that the most significant terms in the GO analysis also contain three CPS synthesis-related genes. Other enriched pathways were “Pentose phosphate pathway,” “Biotin metabolism,” and “Fatty acid biosynthesis” ( Figure 7 ). The most significant of the downregulated gene-enriched pathways were the “two-component systems” (p=1.20E-07), which include proteins associated with potassium uptake (kdpA, kdpB, and kdpC) and motility (pilJ, pilH, pilI, pilG, PilR).

Differences in DEGs associated with virulence and antibiotic resistance were next investigated since these two phenotypes are closely linked to the clinical significance of bacteria and to their survival in the environment.

The VFDB database and the one-way blast approach for SwissProt annotation were combined to identify 42 virulence-related genes among the upregulated genes in strains of the DST group. The genes identified were similar to those identified by genomic analysis, the functions of which related mainly to CPS synthesis and transport (wza/wzb/wzc, etc.), iron uptake (Heme utilization genes, TonB-dependent iron carrier receptors, and iron regulatory proteins), type VI secretion systems (vgrg, vgrgA, contractile small subunits, and PAAR domain-containing proteins), and biofilm formation (pilin/pgaB, ompW, oprM, abaR). In addition, the toxin-antitoxin cdiA/cdiB, RNA chaperone hfq, sensor protein pilS, and the redox-sensitive transcriptional activator soxR were also highly expressed in strains of the DST group ( Table S7 ). The functions of 36 virulence genes identified as downregulated in the DST groups mainly involved type IV pilus synthesis, and included genes encoding pilus assembly proteins (pilC, pilB, pilJ, pilG, pilH, pilI, pilT) and the pilus regulatory protein gene (PilR). In addition, genes downregulated in the DST group contained genes related to iron uptake, CPS synthesis, potassium transport, chemotactic and competition functions, type II secretion system, and phospholipase synthesis ( Table S8 ).

Alignment of DEGs with the antibiotic resistance database ResFinder identified that genes encoding resistance to several antibiotic classes, including aminoglycosides, tetracyclines, β-lactam, chloramphenicol, and cationic antimicrobial peptide, were upregulated in the DST group ( Table S9 ), while chlorhexidine, chloramphenicol, and aminoglycoside resistance genes were downregulated in the DST group ( Table S10 ).