Genomic Investigation of Two Acinetobacter baumannii Outbreaks in a Veterinary Intensive Care Unit in The Netherlands

Acinetobacter baumannii is a nosocomial pathogen that frequently causes healthcare-acquired infections. The global spread of multidrug-resistant (MDR) strains with its ability to survive in the environment for extended periods imposes a pressing public health threat. Two MDR A. baumannii outbreaks occurred in 2012 and 2014 in a companion animal intensive care unit (caICU) in the Netherlands. Whole-genome sequencing (WGS) was performed on dog clinical isolates (n = 6), environmental isolates (n = 5), and human reference strains (n = 3) to investigate if the isolates of the two outbreaks were related. All clinical isolates shared identical resistance phenotypes displaying multidrug resistance. Multi-locus Sequence Typing (MLST) revealed that all clinical isolates belonged to sequence type ST2. The core genome MLST (cgMLST) results confirmed that the isolates of the two outbreaks were not related. Comparative genome analysis showed that the outbreak isolates contained different gene contents, including mobile genetic elements associated with antimicrobial resistance genes (ARGs). The time-measured phylogenetic reconstruction revealed that the outbreak isolates diverged approximately 30 years before 2014. Our study shows the importance of WGS analyses combined with molecular clock investigations to reduce transmission of MDR A. baumannii infections in companion animal clinics.


Introduction
Acinetobacter baumannii (A. baumannii) is an opportunistic pathogen commonly associated with nosocomial infections and poses a critical threat in healthcare settings. It can cause fatal infections such as bloodstream infections and pneumonia in humans and animals [1][2][3][4][5]. The emergence of multidrug resistance (MDR) of A. baumannii in nosocomial infections was reported for the first time in the early 1980s [6]. Antimicrobial resistance (AMR) in A. baumannii is evolving rapidly, leading to extensive drug resistance against available antimicrobials, including carbapenems and third-generation cephalosporins, the last time points in the caICU. Each patient was admitted to the ICU at different time points, and there was no overlap of ICU stay among the four dogs. An additional 25 environmental samples from the caICU were obtained. However, they were all negative for A. baumanii. The second outbreak took place in March 2014 in the caICU, on which two dogs were infected with A. baumanii. One patient (214030705701) was admitted and stayed in the ICU for complications following surgery, and another one (214031705301) was never admitted to the ICU but stayed in a medium care ward for recovery across from the ICU ward. In the 2014 outbreak, A. baumanii was recovered from 18 of 28 environmental screening samples, including the caICU treatment table, a cage, the operating table, the preparation room, and the fur of a hospitalized dog. The antimicrobial susceptibility tests (ASTs) revealed that all clinical isolates from both 2012 and 2014 outbreaks and one environmental sample from 2014 (214032504901) were multidrug-resistant (MDR), conferring resistance to aminoglycosides, cephalosporins, chloramphenicol, enrofloxacin, penicillins, tetracycline, and trimethoprim/sulfamethoxazole. The phenotypic resistance of other environmental isolates from 2014 were identical, showing resistance to third-generation cephalosporins (3GC), chloramphenicol, and penicillins. The MLST analysis using the Pasteur scheme revealed that all outbreak isolates and one surface isolate from the ICU treatment table belonged to the same sequence type (ST2). In contrast, other environmental isolates displayed ST241, ST239, and ST837. One isolate (214032504501) collected from a cage had an unknown or untypeable sequence type (ST). The description of bacterial isolates and genomic characterization are summarized in Table 1. AST results can be found in Supplementary Table S1, and the epidemiological features of the two outbreaks are visualized in Supplementary Figure S1.  ST2  MDR  92x  152  3,894,627  212062205001  dog 2-urinary tract  2012  22 June 2012  ST2  MDR  145x  154  3,902,040  212082004201  dog 3-urinary tract  2012  20 August 2012  ST2  MDR  144x  171  3,876,024  212090506901  dog 4-wound  2012  5 September 2012  ST2  MDR  148x  321  3,962,219  214030705701  dog 5-respiratory tract  2014  7 March 2014  ST2  MDR  97x  199  3,939,663  214031705301  dog 6-wound  2014  17 March 2014  ST2  MDR  104x  362  3,912,898  214032504901  ICU treatment table  2014  25 March 2014  ST2  MDR  162x  169  3,823,448  214032504501  medium care cage 9  2014  25 March 2014  -3GC  144x  87  4,004,712   MCD-Ch-OK3-2PV-1  dog 7-commensal  skin carriage  2014  25 March 2014  ST241  3GC  131x  73  3,932,237   UKG-Inl-T1-1N-1  clinic, operating table  2014  25 March 2014  ST239  3GC  145x  324  3,889,664  UKG-Inl-T1-4N-2  preparation room  2014  25 March 2014  ST837  3GC  88x

Outbreak Investigation Using Core-Genome MLST (cgMLST)
Based on the MLST finding of the same sequence type (ST2) with identical resistance patterns in outbreak isolates, it was assumed that the A. baumannii 2012 outbreak strains somehow thrived in the caICU for protracted times. To confirm the relatedness of the outbreak strains, the whole-genome sequences of outbreak isolates were compared with sequences of three human reference strains (Table 1), which are dominant in Europe. All genomes contained >93% of 2390 alleles defined in the cgMLST scheme. The 14 isolates were grouped into 9 distinct clusters based on cgMLST complex types (CT) (Figure 1). The cgMLST analysis identified two clonal clusters (C1 and C2) with different complex types (CT1695 and CT1425) in which only 1 or 2 alleles differences were found within each cluster. C1 consisted of four clinical isolates from the 2012 outbreak, whereas C2 was formed by two clinical isolates and one surface isolate derived from the caICU treatment table (214032504901) from the 2014 outbreak. Both clusters were closely related to RUH-134, the European Clone-II human reference strain, in which its alleles differed by 47 and 43 single nucleotide substitutions (SNPs) from C1 and C2, respectively. The environmental isolates belonged to distinct ST types separated by >2080 SNP differences from the C1 and C2 and reference strains (Figure 1).

Estimation of Divergence Date of ST2 Outbreak Isolates
A SNP-based molecular clock analysis was performed to estimate the mutation rate on an evolutionary time scale of the ST2 outbreak isolates. The maximum SNP difference between outbreak-related ST2 isolates was 421 SNPs, but most SNPs were obtained by recombination events that took place in the ST2 outbreak isolates, indicated by the fact that 75% (317/421) SNPs were located in regions < 1 kb apart (Supplementary Table S3). The exclusion of these recombination regions revealed that the 2012 and 2014 outbreak isolates differed by only 84 SNPs. A Bayesian molecular clock analysis allows indicating the time of divergence of the ST2 outbreak isolates (i.e., the mutation rate of the SNPs difference identified from the ST2 outbreak isolates). The molecular clock was estimated at 1.286 × 10 −6 (95% highest posterior density (HPD) interval 1.125 × 10 −6 −1.449 × 10 −6 ) substitutions per site per year. This estimated substitution rate referred to approximately five to seven SNPs per year. Based on this analysis, the divergence date of the 2012 and 2014 outbreak isolates was 30 years ago (95% HPD interval [25][26][27][28][29][30][31][32][33][34][35], suggesting that the ancestor of the 2012 and 2014 outbreak isolates may have appeared around the 1980s.

Comparative Genome Analysis
The phylogenetic analysis based on core genome alignment with SNP detection demonstrated that the outbreak isolates belonged to different cgMLST clusters and were genetically closely related to the EC-II human reference strain. Two environmental isolates from 2014 (one from the preparation room (UKG-Inl-T1-4N-2) and one from the clinic operating table (UKG-Inl-T1-1N-1) belonged to a distinct cluster from the rest of the isolates ( Figure 2). The pangenome analysis showed that all genomes shared 2588 core genes with differences in gene presence and absence between outbreak isolates. The gene differences between the 2012 and 2014 outbreak isolates included several phage components, a potential capsular biosynthesis region, several genomic islands, and mobile genetic elements containing antimicrobial-resistance genes (not shown). Serum resistance gene (traT) was present only in a dog genome with wound infection from 2012 (212090506901). There were no hostassociated genes identified in the human and animal isolates. However, we identified some virulent genes associated with ST2 isolates. Virulence genes such as biofilm-associated protein (bap) and TonB dependent siderophore receptor (bauA) were only found in ST2 isolates. The environmental isolates carried unique accessory genes that were not detected in clinical isolates. For example, genes coding for a type IV secretion system protein complex was detected only in one sample isolated from the clinic operating table (UKG-Inl-T1-1N-1). The variation of gene content differences between isolates is displayed in Figure 2. Figure 2. The differences in gene content between genomes included in this study. The pan-genomic matrix (right block) shows the absence and presence of core and accessory genes corresponding to mid-rooted phylogenetic dendrogram (left) (blue = presence of genes, white = absence of genes). The red line (top) indicates the size of contigs with different kilobase (kb). The blue line curve underneath the matrix displays the frequency of the presence of genes in each genome.
The isolates of the two outbreaks carried different mobile elements carrying resistance genes, and different mobile genetic elements such as insertion sequences (IS) and unit transposons were identified. The 2012 clinical isolates had IS6100 insertion sequence carrying AMR genes aadA1, aac(3)-Ia, sul1 and qacE, and Tn 6207 transposon carrying tetracycline-and streptomycin-resistant genes (tet(B) and aph(6)-Id). The 2014 clinical isolates had a different IS type (ISVsa3) and AbaR4 transposon carrying tetracycline-and streptomycin-resistant genes (tet(B) and aph(6)-Id).
With RFPlasmid and Plasmidfinder, contigs containing replication (rep) genes were detected in two environmental samples (UKG-Inl-T1-1N-1, UKG-Inl-T1-4N-2), but not in the clinical isolates, indicating that only the environmental samples contain putative plasmids. One environmental sample (UKG-Inl-T1-1N-1) collected in 2014 from the operating room of the clinic carried a putative plasmid containing both a rep 7a gene and chloramphenicol resistant gene (cat pC221 ). The other environmental isolate (UKG-Inl-T1-4N-2), collected from the preparation room, carried a putative plasmid containing a replication and tet(R) gene. Other environmental isolates had only beta-lactamase genes and an efflux pump (AdeABC) and did not carry any mobile genetic elements associated with resistance mechanisms. The antimicrobial genes and relevant MGEs are displayed in Figure 3.

Discussion
The current study findings captured the genomic epidemiology of two MDR A. baumannii outbreaks in 2012 and 2014 at the caICU in the Netherlands. The canine isolates from both outbreaks shared the same MLST sequence type (ST2) and identical phenotypic resistance pattern suggesting a protracted outbreak. The canine outbreak isolates are genetically similar to the European clone (EC-II), one of the most prevalent clones globally. Even though this study could not identify how MDR A. baumannii strains were introduced to the veterinary clinic, we proved that MDR clones were shared among humans, companion animals, and the environment. One environmental sample from 2014 (214032504901) shared the same genotype as ST2 clinical isolates, but other samples from the ICU environment were genetically distinct from clinical and reference strains. The WGS analysis also revealed that the MDR-A. baumannii isolates from both outbreaks in the caICU diverged 30 years before 2014, consistent with the spread of MDR A. baumannii in the early 1980s [29].

Two Independent MDR A. baumannii Outbreaks Confirmed by WGS-Based Analysis
The two outbreaks might have been mistaken as a single protracted one if the interpretation was solely based on the same MLST sequence type (ST2) and phenotypic resistance profiles. Indeed, two different clonal clusters of the 2012 and 2014 outbreaks computed by the cgMLST analysis confirmed that the MDR A. baumannii outbreaks in the caICU were two independent events. These findings suggested that the cgMLST study and SNP phylogeny provided the optimal resolution in differentiation outbreak strains. Similar to our findings, the cgMLST analysis of nosocomial infections associated with carbapenemresistant A. baumannii from an Italian ICU was able to show two clonal clusters, whereas their traditional typing results suggested one cluster [30]. The authors also agreed that the cgMLST is a valuable tool that provides the highest discriminatory power in studying clonal relations among outbreak strains. From our study, both SNP-based results with filtered recombination and cgMLST analyses are compatible in outbreak investigations. In the case of the SNP-based analysis, it required an additional step to filter the recombination that can affect the conclusion of the outbreak investigation. Thus, cgMLST covers the limitation of an SNP-based approach by reducing the effect of recombination. It can be beneficial to cooperate with the cgMLST scheme in the WGS-based routine surveillance since the software used for cgMLST is user-friendly and does not require in-depth bioinformatics skills to compute the analysis. These findings indicate that the interpretation based on conventional MLST and phenotypic resistance profiles are insufficient to study the epidemiology and transmission chain of a limited number of A. baumannii infections in a veterinary healthcare setting.

Genetic Differences between Clinical and Environmental A. baumanni Isolates
The comparative genome analysis of outbreak isolates and reference genomes demonstrated the differences in gene content, including antimicrobial resistance genes, virulence factors, and mobile genetic elements. We identified biofilm-associated virulent genes such as bap and bauA only in ST2 isolates. In addition, the dog genome obtained from the wound infection from 2012 carried a virulence gene (traT), which encoded the R6-5 plasmid-specified outer membrane protein that was demonstrated to mediate serum resistance in bloodstream infections [31]. This gene is not universally present [32], but a recent study showed traT was found in 80% of carbapenemase-producing A. baumannii isolates in Iran [33]. The differences in gene content among clinical isolates might be due to phage insertion or deletion since we observed variation in phage components among outbreak isolates. Our findings cannot explain the mechanisms in which these genetic differences impact the pathogenesis in dog patients.
Although the MDR phenotypes were identical between the 2012 and 2014 outbreak isolates, only the catA1 gene encoding for chloramphenicol resistance was present in the 2014 isolates. However, in A. baumannii, the catA1 gene is redundant, as all isolates are intrinsically resistant to chloramphenicol due to the CraA efflux pump [34]. All isolates carried resistance-nodulation-division (RND) efflux pumps (AdeABC) that previously showed that over-expression of efflux pumps had significant effect on susceptibility to some antimicrobials including beta-lactams, fluoroquinolones, and aminoglycosides [35]. However, a recent study showed that efflux-pump overexpression played a less significant role in the development of carbapenem resistance in A. baumannii, whereas biofilm production was strongly associated with carbapenem resistance phenotype [36]. In the current study, all MDR ST2 isolates carried biofilm-associated protein (bap), and there was no molecular detection of carbapenem resistance genes.
To explore the potential of A. baumannii as a zoonotic pathogen, we tried to identify differences in gene contents based on the host (i.e., human, dog, and environment). We could not identify host-specific genes. Our results implicate that A. baumannii may freely transmit between the animal and human host and cause infection without the requirement of host-specific factors. Host-specific genes may exist, and we might not have detected them with the limited number of isolates included in this study. From this study, we urge to include more isolates of animal origin in future research to carefully investigate the human-animal interplay of A. baumannii.
Environmental sampling from both years added additional value to the outbreak investigation. In 2012, A. baumannii was not traceable from the environment in the ICU, whereas it was recovered from surfaces in the ICU, neighboring rooms, and fur from a hospitalized dog in 2014. One surface isolate from the ICU treatment table was genetically identical to the 2014 patient isolates and belonged to the same clonal cluster (C2). This demonstrates the risk of environmental contamination and highlights the pre-existing challenge in eradicating A. baumannii from surfaces. Other environmental isolates displayed different sequence types without other ARGs besides beta-lactamase genes and mobile genetic elements except in a sample derived from the clinic operating room in the 2014 outbreak. This isolate belonged to ST239 carrying the blaOXA-51 carbapenemase gene and the putative plasmid harboring genes for type B chloramphenicol acetyltransferase. [37]. This putative plasmid type has never been reported in Acinetobacter species before. This plasmid-mediated chloramphenicol resistance mechanism is different from the 2014 outbreak isolates in which catA1 is chromosomally located and encoded type A chloramphenicol acetyltransferase. The exchange of such plasmids between humans and companion animals is still unknown. ST239 was reported before in pets from France and a child from Tanzania; however, just as in our study, only single isolates of ST239 were described [38,39]. The detection of carbapenemase genes and putative plasmids encoding for drug resistance in the healthcare environment in this study is noteworthy. In addition, we identified the putative plasmids only in two environmental isolates (UKG-Inl-T1-1N-1, UKG-Inl-T1-4N-2) from 2014. All patient isolates carried mobile elements carrying aminoglycosides resistance genes (aac(3)-Ia, ant(3"), sulfonamide resistance gene (sul1), and blaTEM, suggesting the horizontal gene transfer of ARGs.

Methodological Considerations
The plasmid analysis in this study remained limited, given that the WGS was based on short paired-end sequencing. A combination of short contigs assembly with longread sequencing can precisely determine whether the genes identified are encoded by chromosome or plasmid. In addition, we did not perform phenotypic analysis such as broth dilution and carbapenem inactivation methods to detect carbapenemase production of the isolates in this study [40]. However, whole-genome sequencing analysis revealed that there were no carbapenemase genes in patient genomes from both outbreaks. Another limitation was the interpretation of drug resistance in clinical isolates. Currently, there are no established veterinary-specific clinical breakpoints for Acinetobacter species and the standardized definitions for multi drug resistance are not widely available [41]. Thus, we used the general definition of multidrug resistance that has been widely used to characterize MDR in animal isolates [42] as opposed to the more comprehensive guideline developed for human medicine [43]. Thus, there is a need to develop a definition of drug resistance in veterinary medicine that can be used universally. Further work is required to study the diversity and abundance of A. baumannii in animal species and the horizontal transfer dynamics of virulence and AMR plasmids between the pathogenic and commensal strains.

The Importance of WGS-Based Surveillance in Animals and the Environment
In veterinary medicine, A. baumannii remains a neglected pathogen with limited data from strains originating from animals and their environment. Our study demonstrated that animals and humans share identical clones (ST2) and the same B-lactamase (blaOXA-66). Our study underlines the importance of genomic investigation combined with molecular clock determination in studying A. baumannii from animal origins. This easy-to-use and relatively cheap WGS and cgMLST platform could benefit WGS-based routine typing in outbreak management and surveillance in hospital settings. The implications from this study can increase awareness and help reduce transmission of MDR A. baumannii infections in small animal veterinary clinics.

Selection of Strains
A total of 11 Acinetobacter baumannii strains were isolated from dogs from the outbreaks in 2012 (n = 4), 2014 (n = 3), and environmental samples (n = 4) in 2014 in the calCU at the faculty of Veterinary Medicine of Utrecht University. Three reference strains belonging to European clones EC-I, EC-II, and EC-III were also sequenced and included in this study for comparison with study isolates for outbreak investigation.

Antimicrobial Susceptibility Testing (AST)
Antimicrobial susceptibility tests (ASTs) of the isolates were performed to determine the minimum inhibition concentration (MIC), using the microbroth dilution assay MICRONAUT-S (Merlin Diagnostika Gmbh, Bornheim, Germany). AST was performed as recommended by the manufacturer for inoculum preparation, broth composition, and incubation conditions. Customized MIC plates were used and read both visually and with a microplate reader (ThermoFisher scientific multiskan TM [44,45]. The multidrug resistance (MDR) was classified if the isolate was resistant to 3 or more antimicrobial classes [42] while isolates were defined as third-generation cephalosporins (3GC) resistant if resistance was found for third-generation cephalosporins (Ceftiofur, CET).

DNA Isolation and Quantification
DNA isolation was performed using the DNeasy ® UtraClean ® Microbial kit (Qiagen Gmbh, Germany) according to the manufacturer's protocol. A total of 50 µL of DNA concentration was collected after DNA isolation and stored at 4 • C. A total of 1 µL of DNA from the sample with 199 µL of Qubit ® working solution from Qubit™ dsDNA HS Assay Kit (Invitrogen, Thermo Fisher Scientific, Waltham, MA, USA) was mixed to measure the concentration of DNA using Invitrogen Qubit Fluorometers (Thermo Fisher Scientific, Waltham, MA, USA).

Whole-Genome Sequencing and Genome Analysis
The sequencing of the A. baumannii isolates was performed using Illumina Miseq sequencing using 2 × 250 bp reads and 300 bp insert size by the Utrecht Sequencing Facility (USEQ, Utrecht, The Netherlands). The Illumina library was prepared with the final DNA concentration of 2 ng/µL using the Nextera XT Library Prep Kit (Illumina, San Diego, CA, USA) according to the manufacturer's protocol.

Multi-Locus Sequence Typing
The sequence types (STs) of the genomes were assigned according to the Pasteur multilocus sequence typing (MLST) schemes as previously described [28]. Subsequently, the core genome MLST (cgMLST) was performed using the cgMLST scheme [58] cooperated in the Ridom SeqSphere + v8.0.2 software (Ridom GmBH, Münster, Germany) accessed on 1 October 2021. The clonal relationship of outbreak strains was visualized by a minimum spanning tree based on 2390 target alleles by the cgMLST scheme (paired-wise ignored missing values). The clonal cluster (CC) was defined based on the cgMLST scheme [58] where isolates sharing ≤10 different alleles in target genes were considered highly related (CC).

Time-Resolved Phylogeny Reconstruction
Relevant A. baumannii genomes with known isolation dates (n = 159) were obtained from Genbank (Supplementary Table S2). Among them, 6 genomes related to the outbreak isolates were selected from a phylogenetic tree based on SNP detection of the downloaded genomes (Supplementary Figure S2). This analysis included these additional ST-2 genomes (n = 6) from relevant literature [59][60][61][62][63][64] dated between 1982, when the EC-1 reference strain was isolated, and 2012 (Supplementary Table S2). Firstly, Gubbins v1.4.5 [65] predicted the recombination events in core genome alignment. Subsequently, the recombination regions were filtered, and the resulting super alignment of the 3144 genes without recombination signature was used in BEAST v1.8.4 [66] with the isolates dates as tip dates. BEAST was used to estimate the divergence dates using the BEAST XML generated by BEAUti [67]. The analysis was based on the generalized time-reversible (GTR) model without rate variation between sites and gamma correction as distance model, a Bayesian Skyline plot with 4 groups as demographic models, and a strict clock model. BEAST was run for 10,000,000 iterations with sampling. Tracer was used to evaluate the Effective Sample Sizes (ESS). ESS values > 200 were obtained.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/ 10.3390/pathogens11020123/s1, Figure S1: the description of A. baumannii cases in a companion animal intensive care unit (caICU) in 2012 and 2014.; Figure S2: the phylogenetic tree of isolates with different sequence types (ST); Table S1: antimicrobial susceptibility test results of the outbreak isolates from 2012 and 2014; Table S2: relevant genomes with known isolation date from litera-ture; Table S3: the recombination regions and the single nucleotide polymorphism (SNP) difference between outbreak isolates. Institutional Review Board Statement: Ethical review and approval were waived for this study since the bacterial isolates used in this study were obtained as part of routine diagnostics.
Informed Consent Statement: Not applicable.

Data Availability Statement:
The sequence data were deposited in the European Nucleotide Archive (ENA) under study accession number PRJEB19153.