Sixty-four patients admitted to the ICU of secondary referral hospital “Dr Laza K. Lazarević” Šabac, Serbia, due to severe coronavirus disease 2019 (COVID-19) were included in this study. Patients were hospitalized during the study period from December 2020 to February 2021, and COVID-19 was confirmed by real-time PCR. All patients (n = 64, male:female ratio 1:1) required mechanical ventilation (MV). Besides, Acinetobacter spp. infection was confirmed based on bacteriological analyses of different samples recovered from COVID-19 patients on MV (Supplementary Table 1). The criteria to include patient in study protocol were (a) severe form of the COVID-19, (b) need for MV, and (c) confirmed Acinetobacter spp. infection. Information on patients (gender, age, and comorbidities) are presented in Supplementary Table 1. Patients admitted to ICU during the same period, without confirmed Acinetobacter spp. infection (n = 48), were not included in the study. Surveillance program through rectal swabs, was not carried out within ICU, so there is no available data if patients excluded from this study were also colonized by A. baumannii but did not develop the infection. The study protocol was approved by the Ethical Committee of the General hospital “Dr Laza K. Lazarević” Šabac (Approval No. 08-1/2).

Bacterial isolates (one bacterial isolate per patient), belonging to Acinetobacter spp. (n = 64) were collected from blood (n = 28), tip of the central venous catheter (CVC; n = 22), tracheal aspirate (TBA; n = 10), tip of the aspirator (n = 3), and sputum (n = 1; Supplementary Table 1). Identification of the isolates was performed using standard microbiological procedures (cultivation, staining, and microscopy), followed by sequencing of 16S rRNA gene amplicons (Macrogen DNA sequencing service, Netherlands; Jovcic et al., 2009). 16S rRNA gene sequences were deposited to NCBI database (accession numbers ON705785–ON705830). Identification by 16S rRNA sequence was assessed using Basic Local Alignment Search Tool (BLAST, http://blast.ncbi.nlm.nih.gov/Blast.cgi) for searches against GenBank database. The strains labels were assigned randomly to preserve the anonymity of the patients (Supplementary Table 1).

Genetic relatedness among analyzed Acinetobacter spp. was determined by PFGE of ApaI digested (3 h at 30°C) genomic DNA as previously described (Vukotic et al., 2020). Electrophoresis was run at LKB hexagonal electrode array (2015 Pulsafor unit, LKB Instruments, Bromma, Sweden) for 18 h at 300 V in 0.5 × Tris-Borate-EDTA at 9°C with pulse times of 8 s for 8 h and 18 s for 10 h during electrophoresis. Gels (1.2% agarose) were stained with ethidium bromide (SERVA Electrophoresis GmbH) and gel images were captured under UV light. Cluster analysis, using Pearson correlation coefficient with a 1.0% optimization and a hierarchic UPGMA algorithm, was used to generate a dendrogram describing the relationship among A. baumannii pulsotypes.

The clonal lineages determination of tested A. baumannii isolates was performed by multiplex PCR as previously described (Turton et al., 2007). Identification of an isolate as a member of sequence type Group 1 (international clone 2, IC2) and Group 2 (international clone 1, IC1) required amplification of all three fragments in the corresponding multiplex PCR. As positive PCR controls, A. baumannii strains 6077/12 and 4,031 were used for Group 1 (IC2) and Group 2 (IC1), respectively, (Novovic et al., 2015).

Multilocus sequence typing of four strains selected according to PFGE were determined using MLST 2.0 resource of Center for Genomic Epidemiology.¹

Antimicrobial susceptibility testing was performed using the broth microdilution method according to the European Committee on Antimicrobial Susceptibility Testing breakpoints (EUCAST, Version 11.0, 2021, https://eucast.org/). Minimum inhibitory concentrations (MICs) were determined for meropenem (1–8 μg/ml), imipenem (1–8 μg/ml), colistin (1–8 μg/ml), gentamicin (1–8 μg/ml), tobramycin (1–8 μg/ml), and levofloxacin (0.25–2 μg/ml). Experiments were done in triplicate. After 24 h incubation at 37°C, cell density was monitored by OD₅₇₀ measurements using Multiscan FC Microplate Photometer (Thermo Scientific, United States) and MIC values were determinate as the lowest concentration of antibiotic that inhibited bacterial growth. Acinetobacter baumannii strain ATCC19606 was used as a control strain.

The presence of the genes encoding for different oxacillinases (bla_OXA-23, bla_OXA-24, bla_OXA-51, and bla_OXA-58), cephalosporinase (bla_AmpC), and metallo-β-lactamase (bla_NDM-1) in genomes of analyzed isolates (n = 64) was monitored by PCR amplification method as previously described (Woodford et al., 2006; Beceiro et al., 2009; Teo et al., 2012). Using the same method were amplified carO and omp33-36 genes (Mussi et al., 2011; Novovic et al., 2015). Localization of the insertion sequence ISAba1 upstream of oxacillinase genes was performed by combining a reverse primer from the pairs used for detection of each oxacillinase gene with ISAba1-F primer (Beceiro et al., 2009). Genomic DNA of four A. baumannii isolates was sequenced using Illumina HiSeq (MicrobesNG, IMI-School of Biosciences, University of Birmingham, Birmingham, United Kingdom). The Invitrogen Collibri Library Prep kits for Illumina HiSeq was used for the library preparation. The quality of each sequenced genome was checked using FastQC (Andrews, 2010). Genetic determinants of antibiotic resistance of four strains selected according to PFGE were determined using ResFinder 3.1. and threshold of ID = 98.00% was selected. In addition, genes determinants of virulence factors were screened in four sequenced genomes using VFDB: Virulence factors database² (Chen et al., 2005). The threshold for VFDB was set at 70%. In order to analyze sugars presented in capsule, Kaptive 2.0, a tool for rapidly identifying and typing capsule (K) and outer LPS (O) loci from whole genome sequence data was used (Lam et al., 2022).

Pan-genome analysis approach was used in order to compare genomes of four sequenced A. baumannii isolates. Genome sequences were firstly annotated using Prokka (version 1.13) and obtained annotated GFF files were further clustered by Roary (version 3.13.0) into core genes (selection threshold for hard core genes was presence in >99% of the isolates and for soft core genes threshold was presence in 95–99% of isolates) and accessory genes (further subdivided into shell genes—present in 15–95% of isolates; and cloud genes—present in less than 15%). Phylogenetic tree was created by Roary and visualized by Phandango.

Collected A. baumannii isolates (n = 64) were tested for the ability to form biofilm following the protocol described previously (Stepanović et al., 2007) with slight modifications. Aliquots (20 μl) of overnight cultures adjusted to the 0.5 McFarland units, were transferred into 96-well microtiter plates containing 180 μl of Tryptone Soya Broth (TSB). Microtiter plates were incubated aerobically for 48 h at 35°C and after incubation were washed three times with phosphate-buffer saline (PBS, pH 7.2). Remaining bacteria were fixed to the surface of microtiter wells by drying plates at 65°C for 30 min. Visualization of the biofilm formation was performed by staining with 0.1% crystal violet 30 min at room temperature (RT). The excess dye was removed by washing three times with 1 × PBS. Quantification of biofilm formation was done by resolubilisation of remaining dye in mixed solution of 96% ethanol and acetone (4:1) and determination of absorbance at 570 nm using Multiscan FC Microplate Photometer (Thermo Scientific, United States). Interpretation of obtained results was performed according to recommendations of Stepanović et al. (2007) and based on previously calculated OD values for each strain. All isolates were tested in triplicate. Sterile medium tested in triplicate was used as a negative control, while Pseudomonas aeruginosa PAO1 strain was used as a positive control. The cut-off value (ODc) was established as three SDs above the mean OD of the negative control: ODc = average OD of negative control + (3 × SD of negative control) (Stepanović et al., 2007). ODc is calculated for each microtiter plate separately. Final OD of each tested strain was calculated as OD average of the strain minus ODc. According to calculated values, isolates were divided into four groups: no (N) biofilm producer (OD ≤ ODc), weak (W) biofilm producer (ODc < OD ≤ 2 × ODc), moderate (M) biofilm producer (2 × ODc < OD ≤ 4 × ODc), and strong (S) biofilm producer (4 × ODc < OD).

Mucus consists primarily of water (~95%), and the major non-aqueous component is mucin, while proteoglycans, lipids, proteins, and DNA are also present in smaller quantities (Ohar et al., 2019). The binding of A. baumannii isolates to mucin was tested as described previously (Muñoz-Provencio et al., 2009), with the following modifications. Briefly, the wells of microtiter plates were coated with 200 μl of porcine stomach mucin type II (Sigma, Germany) resuspended in 50 mM carbonate buffer (30 mg/ml, pH 9.6). The same volume of carbonate buffer (without mucin) was added to the control wells. Plates were stored at 4°C for 48 h. Wells were washed three times with 1 × PBS, uncoated binding places were saturated with PBS containing 1% Tween 20, incubated 1 h at RT, and washed once more with 1 × PBS. Suspensions of overnight bacterial cultures were prepared in 1 × PBS and adjusted to 0.5 McFarland. Subsequently, 200 μl of each bacterial suspension was added in both coated and uncoated control wells (test was done in triplicate). After 2 h incubation at 35°C, nonadherent cells were removed using 1 × PBS with 0.05% Tween 20 and fixation of mucin-bound bacterial cells was performed by drying plate at 65°C for 1 h. 200 μl of 0.1% crystal violet (HiMedia Labs Pvt. Ltd., India) was added to the wells, plates were incubated at RT for 45 min and unbound stain was removed by three washes in 1 × PBS. Finally, citrate buffer (50 mM, pH 4) was added to dissolve the stain bound to the bacterial cells and the absorbance was measured at 570 nm using Multiscan FC Microplate Photometer (Thermo Scientific, United States).

All 64 A. baumannii isolates were tested for two types of motility: swarming and twitching. Freshly grown cultures were stabbed with sterile toothpick on the surface of modified Luria Bertani (LB) agar (tryptone—10 g/L; NaCl—5 g/L; yeast extract—5 g/L) with 0.4% agar for swarming motility and 0.8% agar for twitching motility, as described previously (Clemmer et al., 2011). Acinetobacter baumannii ATCC19606 strain was used as negative control as it is non-motile strain. In, addition, modified LB broth was used as a base medium for all motility assays, and plates were used on the same day that they were prepared. All isolates were tested for swarming and twitching motility under the same condition (light, temperature etc.) as it was reported that motility can be decreased in the presence of light (Mussi et al., 2010). After inoculation, the plates were incubated at 37°C for 48 h. In the case of swarming motility test, motile positive isolates were defined if the zone around the point of inoculation was greater than 10 mm, and for twitching motility isolates were characterized as non-motile (zone < 5 mm), intermediary-motile (zone 5–20 mm), or highly motile (zone ˃ 20 mm; Vijayakumar et al., 2016).

The phenotypic assay of gelatinase activity was performed as described previously (Su et al., 1991). A drop of overnight bacterial culture (10 μl) was transferred on a Gelatin agar plate containing peptone 5 g/L, yeast extract 3 g/L, gelatin 30 g/L, and agar 15 g/L (pH 7). Plates were incubated at 35°C for 48 h and after incubation flooded with a saturated solution of ammonium sulfate (550 g/L). Gelatinase activity was assessed as positive if transparent halo appeared around bacterial cells growth.

For comparing a difference in biofilm formation and mucin-adhesion ability among A. baumannii isolates obtained from different patient groups (males and females) Mann–Whitney test was used. Analyses were performed in Prism 9 version 9.3.1 and results were visualized as scatter and whicker plots. Dendrogram for analyses of the PFGE patterns was created in SPSS 28.0 for Windows.

Virulence potential data of tested isolates were summarized and visualized by a heat map created using Rx64 3.5.1 software. The results were approximated on the relative scale ranging from blue as the lowest values, progressing to white, then to red as the highest value.

Duration of patient hospitalization, time points of specimen collection, and origin of bacterial isolates are presented in Figure 1. All 64 isolates analyzed in this study were confirmed as A. baumannii according to 16S rRNA gene sequencing. According to Blast results of selected isolates the percentage of similarity of 16S rRNA genes were in the range 97.82–99.02%. The incidence of A. baumannii infection in COVID-19 patients was 57.14% during the study period.

Ethyl alcohol (70%) was used as disinfectant for surfaces and items which may be in contact with the patient, while chlorine-based agents were used for disinfection of floors. Healthcare workers had disposable, single use protective gloves, masks, and gowns to protect patients at ICU from possible further infection. The use of closed aspiration systems was implemented in order to prevent nosocomial infections.

The average duration of hospitalization for all patients was 16.86 days, and male COVID-19 patients were hospitalized 16.22 days on average, while female patients spent 17.5 days in the hospital. Mortality rate was 100%.

The multiplex PCR performed for identification of international clone belonging revealed that all tested A. baumannii isolates (n = 64) belong to international clone 2, IC2 (sequence type Group 1).

According to the obtained PFGE ApaI fingerprint, a dendrogram was constructed for all isolates and shown in Figure 2. Additionally, PFGE profiles of tested isolates obtained by ApaI digestion are presented in Supplementary Figure 2.

Of the 64 isolates, 60 (93.75%) possessed an identical ApaI profile of DNA fragments indicating the dominant presence of one pulsotype among the isolates, while only four (6.25%) isolates were different. Although these four isolates (isolates 1, 39, 54, and, 55) had a very similar ApaI profile to the dominant strain, they differed in at least one ApaI fragment; isolate 1 lacks an ApaI fragment of 242 kb, while a unique fragment of 228 kb is present; isolate 39 has an extra ApaI fragment of 250 kb that could be a plasmid, while isolates 54 and 55 have the same profile, differing from others in that they lack a fragment of 300 kb, and have a new unique fragment of 250 kb (Supplementary Figure 2). The obtained results strongly indicate the clonal distribution of the dominant strain in the hospital. Considering the slightly different genomic pattern of four isolates (differ in only one ApaI fragment), it can be assumed that they also originated from the dominant strain during propagation in different patients.

According to PFGE four isolates were sequenced: isolate 1, isolate 2 (as representative of all other strains), isolate 39, and, isolate 54 (as it has the same PFGE pattern as isolate 55). Sequences were deposited to NCBI databank: JAPCYJ01.1 (A. baumannii 1), JAPCYK01.1 (A. baumannii 2), JAPCYL01.1 (A. baumannii 39), and JAPCYM01.1 (A. baumannii 54).

MLST analyses revealed that all isolates were identified as ST195 (ST1816) according to Oxford nomenclature, and ST2 according to Pasteur nomenclature.

All tested A. baumannii isolates (n = 64) were sensitive to colistin, while resistant to meropenem, imipenem, gentamicin, tobramycin, and, levofloxacin according to the broth microdilution method. According to the available data, resistance to different classes of antibiotics (carbapenem antibiotics, aminoglycosides, and, fluoroquinolone) confirmed multidrug resistance (MDR) phenotype in tested isolates (Magiorakos et al., 2012). Susceptibility of A. baumannii isolates was not tested to novel antibiotics, such as ceftazidime/avibactam, ceftolozane/tazobactam, and cefiderocol due to unavailability of these antibiotics. Obtained MIC values are presented in Supplementary Table 2.

Overview of antimicrobial resistance genes detected in genomes of four representative Acinetobacter baumannii isolates (1, 2, 39, and, 54) using ResFinder is presented in Supplementary Table 3. Screening of virulence factors genes in sequenced genomes was performed using VFDB database and results are presented in Supplementary Table 4. VFDB revealed that isolate A. baumannii one had slightly higher virulence potential comparing to other sequenced isolates (Supplementary Table 4). Analyses of sugar presented in capsule were performed using Kaptive 2.0 program. Obtained results revealed that there is no differences in K locus neither in O locus among sequenced genomes (Supplementary Figures 3A,B).

Pan-genome reflects the total number of genes that are present in a given dataset and the main goal of pan-genome analysis is genomic comparison of different isolates of the same species (Muzzi et al., 2007). Pan-genome analysis revealed a total of 4,096 gene clusters, which were separated into the core genome, comprised of 3,597 genes (3,597 hard core and 0 soft core genes) and accessory genome containing 499 genes in the shell and 0 genes in the cloud. The core genome of four A. baumannii isolates revealed that the analyzed genomes are phylogenetically related, as they share a high number of common genes (Supplementary Figure 4).

Typically for A. baumannii, intrinsic genes encoding for oxacillinase OXA-51 and cephalosporinase AmpC were detected in all tested isolates by PCR method. Additionally, all isolates gave a positive PCR signal for the bla_OXA-23 gene as well as ISAba1 insertion sequence upstream of this gene. The gene encoding for OXA-24 oxacillinase was identified in 18 isolates (28.12%), while the bla_OXA-58 gene was found in four isolates (6.25%; Supplementary Table 1). ISAba1 insertion sequence was not detected upstream of bla_OXA-24, bla_OXA-51, and, bla_OXA-58 genes in analyzed isolates. The bla_NDM-1 gene was not present in genomes of tested isolates. The analysis of the carO nucleotide sequences from tested A. baumannii isolates revealed that all isolates had the carO gene identical to corresponding gene of A. baumannii MS14413 (CP054302.1). Further, the gene encoding for another porin included in carbapenem resistance in A. baumannii, Omp33-36, of all isolates was almost identical to the omp33-36 gene of A. baumannii MS14413 with only one nucleotide substitution at position 113 resulting in amino acid substitution Lys38Thr.

Under the tested conditions only one isolate (1/64; 1.56%) was classified as weak (W) biofilm producer. Fifteen isolates (23.44%) were moderate (M) biofilm producers; of those 53.33% (8/15) were recovered from male and 46.67% (7/15) from female patients (Supplementary Table 5). Most of the tested isolates (75%; 48/64) were strong (S) biofilm producers recovered from female (52.08%; 25/48) and male (47.92%; 23/48). According to the Mann–Whitney test, there is no statistically significant difference in biofilm formation ability between female and male isolates. OD values of the biofilm producers by sample type (blood, tip of the aspirator, tracheal aspirate, tip of the CVC, and sputum) are presented in Figure 3.

Mucin-adhesion ability expressed relatively to control (non-coated wells in the microtiter plate) of 64 A. baumannii isolates is shown in Figure 4. The isolates recovered from female patients showed lower binding affinity to mucin compared to isolates from male patients (p < 0.05) according to the Mann–Whitney test.

All A. baumannii isolates were tested for twitching and swarming type of motility under the same temperature and light. According to twitching phenotype isolates were divided into three groups (based on the zone diameter around the inoculation point). Only 9.38% (6/64) isolates were characterized as non-motile, while the highest number of tested isolates (82.81%; 53/64) were intermediary motile. The 7.81% (5/64) isolates were highly motile, with a zone diameter up to 45 mm. Relative to the type of the sample, the number of motile strains were higher for isolates from blood (hemoculture and tip of the CVC) compared to motile isolates originated from the respiratory tract (tracheal aspirate, tip of the aspirator, and sputum, respectively) and the data are presented in Supplementary Figure 1.

Examination of swarming-like motility revealed that only 17.19% (11/64) of tested isolates were classified as a motile category. It is interesting that all isolates with the proven swarming ability, had also the ability of twitching motility, and all were originated from blood.

None of the isolates showed gelatinase activity.

In order to summarize the overall virulence potential, phenotypic characteristics and antibiotic resistance of 64 A. baumannii isolates from COVID-19 patients on MV, heat map was constructed (Figure 5). All tested isolates, according to the heat map, can be divided into three distinct clusters. Cluster three comprised the strains considered as the most virulent, including six isolates recovered from male patients (strains 5, 6, 14,15, 22, and 35) and one isolate recovered from female patient (strain 13; Figure 5). In addition, heatmap displaying the results of hierarchical clustering among 64 A. baumannii isolates was constructed and presented in Supplementary Figure 5.