Acinetobacter baumannii from Samples of Commercially Reared Turkeys: Genomic Relationships, Antimicrobial and Biocide Susceptibility

Acinetobacter baumannii is especially known as a cause of nosocomial infections worldwide. It shows intrinsic and acquired resistances to numerous antimicrobial agents, which can render the treatment difficult. In contrast to the situation in human medicine, there are only few studies focusing on A. baumannii among livestock. In this study, we have examined 643 samples from turkeys reared for meat production, including 250 environmental and 393 diagnostic samples, for the presence of A. baumannii. In total, 99 isolates were identified, confirmed to species level via MALDI-TOF-MS and characterised with pulsed-field gel electrophoresis. Antimicrobial and biocide susceptibility was tested by broth microdilution methods. Based on the results, 26 representative isolates were selected and subjected to whole-genome sequencing (WGS). In general, A. baumannii was detected at a very low prevalence, except for a high prevalence of 79.7% in chick-box-papers (n = 118) of one-day-old turkey chicks. The distributions of the minimal inhibitory concentration values were unimodal for the four biocides and for most of the antimicrobial agents tested. WGS revealed 16 Pasteur and 18 Oxford sequence types, including new ones. Core genome MLST highlighted the diversity of most isolates. In conclusion, the isolates detected were highly diverse and still susceptible to many antimicrobial agents.


Introduction
Acinetobacter baumannii are nonmotile, oxidase-negative, aerobic, Gram-negative coccobacilli [1]. These bacteria are associated with nosocomial infections worldwide [2]. Although A. baumannii is an opportunistic pathogen, it has led to many outbreaks in hospitals and care-facilities with high morbidity and mortality rates [3]. These infections are mainly caused by outbreak strains, which can spread rapidly between patients [4]. Many disease conditions, including ventilator-associated pneumonia, bloodstream infection, urinary tract infection, wound infection and meningitis have been described [3], and A. baumannii has been shown to be a common co-infecting agent in COVID-19 patients in intensive care units [5,6]. A. baumannii can rapidly develop antimicrobial resistance [7,8] due to various resistance mechanisms, such as β-lactamase production, efflux pump overexpression,

Sample Collection and Isolation
In total, 250 samples from 95 different farms were collected from allegedly healthy commercial fattening turkey flocks distributed all over Germany (n = 94) and the Czech Republic (n = 1) as part of a Salmonella surveillance in 2019. This included 118 chick-boxpapers (paper with wood shavings on which the turkey chicks were transported from the hatchery to the production house containing meconium) from one-day-old turkey chicks taken on arrival at the production house from 81 farms (with 24 farms providing more than one sample). Six unused chick-box-papers were also examined as negative controls. In addition, 50 boot swab samples (containing one pair of boot swabs each) taken during the rearing period and 82 boot swab samples from turkeys leaving for the slaughterhouse were investigated. Data and subsequent results were compiled, assessed, and evaluated using Microsoft Excel (Microsoft Office 2019). After pre-enrichment in buffered peptone water (Thermo Scientific, Wesel, Germany) at 37 • C for 16 to 18 h, approximately 10 µL enrichment broth was streaked on chromogenic media Brilliance UTI Clarity agar (Thermo Scientific, Wesel, Germany) and incubated at 37 • C for 24 h. Buffered peptone water without any supplements was analysed as sterility control. In addition, 393 diagnostic samples sent to the Institute of Poultry Diseases, Freie Universität Berlin, Berlin, Germany between 2018-2020 were examined. These included liver and yolk sac samples from 88 one-to sixday-old commercial turkey chicks, as well as 217 lung-and heart-swabs from commercial turkeys. Cultivation was performed on Columbia agar with 5% sheep blood (Thermo Scientific, Wesel, Germany) and Brilliance UTI Clarity agar at 37 • C for 24 h.
Presumptive colonies were selected, sub-cultured, and confirmed to species level by matrix-assisted laser desorption/ionisation time of flight mass spectrometry (MALDI-TOF MS) (Bruker Daltonic GmbH, Bremen, Germany). All isolates were stored at −20 • C in brain heart infusion (BHI) medium (Roth, Karlsruhe, Germany) until further use.

Antimicrobial Susceptibility Testing
Antimicrobial susceptibility testing was performed by broth microdilution according to the instructions of the Clinical and Laboratory Standards Institute (CLSI, 2022) [10]. The Acinetobacter isolates were tested with custom-made microtiter plates (MCS Diagnostics, Swalmen, The Netherlands) for their susceptibility to 18 antimicrobial agents or combinations: colistin, streptomycin, neomycin, trimethoprim/sulfamethoxazole, gentamicin, nalidixic acid, ciprofloxacin, enrofloxacin, marbofloxacin, tetracycline, doxycycline, florfenicol, imipenem, ceftiofur, cefquinome, cefotaxime, cefoperazone, and tiamulin. This test panel was the same as used in the GERM-Vet programme, the German national resistance monitoring programme of veterinary pathogens, for Gram-negative bacteria. The reference strain Escherichia coli ATCC ® 25922 served as quality control. The minimal inhibitory concentration (MIC) values were interpreted as susceptible, intermediate, or resistant using the human-specific clinical breakpoints from CLSI [10], as veterinary-specific clinical breakpoints are not available for Acinetobacter spp.

Macrorestricton Analysis with Subsequent Pulsed-Field Gel Electrophoresis
Macrorestricton analysis using the enzyme ApaI (New England Biolabs, Frankfurt, Germany) and subsequent pulsed-field gel electrophoresis (PFGE) were performed for a preliminary characterisation of the 99 A. baumannii isolates as previously published [35], with a minor modification: for restriction analysis with ApaI (30U), the plug slices were incubated overnight at 25 • C. A Lambda PFGE ladder (New England Biolabs, Frankfurt, Germany) with a size range from 48.5 to 1018 kb served as size marker. Electrophoresis was performed using the CHEF-DR III system (Bio-Rad Laboratories, Düsseldorf, Germany). Gels were stained with GelRed (Biotium, San Francisco, CA, USA) and scanned with the laboratory's imaging system (BIO RAD Molecular Imager GelDoc TM XR+ with Image Lab TM Software, Düsseldorf, Germany). An isolate from diagnostics (141_Diagnostik) served as internal control on each gel. Cluster analysis concerning the percentage similarity was performed with BioNumerics software, version 7.6.3 (Applied Maths, bioMérieux). Similarities were calculated with the dice coefficient (optimization 1.5%, tolerance 1.5%) and the unweighted pair group method with arithmetic mean (UPGMA) [35]. Pulsotypes were defined at a threshold value of ≥80% (named alphabetically) and at a threshold value of ≥87% (additional numeric marking) [35,36].

Whole-Genome Sequencing
For whole-genome sequencing (WGS), 26 isolates were selected, including at least one isolate per PFGE pulsotype (cut off level of ≥80%). DNA was isolated using the Master Pure DNA Purification Kit for Blood Version II (Epicentre Biotechnologies) as published by the manufacturer. The libraries were prepared using the Nextera XT DNA Library Preparation Kit (Illumina Inc., San Diego, CA, USA) according to the manufacturer's instructions. The 2 × 300 bp paired-end sequencing in 40-fold multiplexes was performed on the Illumina MiSeq platform (Illumina) with MiSeq Reagent Kit v3 (600-cycle) (Illumina). For sequence assembly, the Illumina reads were trimmed by Trim Galore v0.6.6 (RRID:SCR_011847) and quality checked by FastQC [37]. De novo assembling was carried out using Unicycler v0.4.9. [38]. Antimicrobial resistance genes were detected using ABRicate [39] with NCBI AMRFinderPlus [40], ResFinder [41], and CARD [42] databases. Plasmid replicons were searched for using ABRicate [39] applied to the PlasmidFinder 2.1 database (https:// cge.food.dtu.dk/services/PlasmidFinder/ accessed on 21 February 2023). The databank PubMLST (https://pubmlst.org/ accessed on 21 February 2023 [43]) was used to confirm the species with ribosomal multilocus sequence typing (rMLST) [44] and to compare and identify sequence types (ST) using both the Pasteur [45] and the Oxford [46,47] scheme. New STs and new alleles were submitted to PubMLST [43]. The generated genomes were used for core genome multilocus sequence typing (cgMLST) with SeqSphere + v7.5.5 (Ridom GmbH, Münster, Germany) [48]. This typing scheme is based on a core genome of 2390 alleles. However, the calculations for the minimum spanning tree presented here were done on the basis of only 1943 alleles as all missing values were excluded. Detected β-lactamases were compared with those listed in the Beta-Lactamase DataBase (www.bldb.eu accessed on 21 February 2023) [49]. Accession numbers and bioproject number are presented in the Data Availability section.

Isolation
Ninety-nine A. baumannii isolates were collected during the study period. A. baumannii was detected in 79.7% (n = 94) of the 118 chick-box-papers. In two chick-box-papers, two morphologically different A. baumannii isolates were recovered. Two further A. baumannii isolates (2.4%) were found among the 82 boot swab samples tested from turkeys before slaughter. None of the 50 boot swab samples taken during the rearing period were positive for A. baumannii (Table 1). Taken together, 1.5% of the boot swab samples contained A. baumannii. The six unused chick-box-papers tested negative. A. baumannii was detected in one of the 217 swabs (0.5%) sent in for bacteriological diagnostics. The single positive pooled heart-lung-swab originated from a seven-week old turkey (isolate 141_Diagnostik). All of the 88 one-to six-day-old commercial turkey chicks were negative for A. baumannii in their liver and in their yolk sac (Table 1).
In total, there were 13 farms from which several A. baumannii isolates were detected (minimum two isolates, maximum five isolates). Only in one of them (farm 13) A. baumannii was detected in a chick-box-paper (isolate 16_W23.1) as well as in a boot swab sample before slaughter (isolate 98_E23.3) ( Figure S1).

Antimicrobial Susceptibility Testing
The results of the antimicrobial susceptibility testing are displayed in Table 2. As there are no CLSI-approved veterinary-specific clinical breakpoints currently available for A. baumannii, human clinical breakpoints were applied. Using these interpretive criteria, all tested isolates were susceptible to imipenem and gentamicin. A high percentage of the tested isolates was susceptible to doxycycline (98%), trimethoprim/sulfamethoxazole (98%), and tetracycline (96%). Concerning cefotaxime, 31% of the isolates were classified as susceptible, 67% as intermediate, and 3% as resistant, despite the fact that the MICs of cefotaxime revealed a unimodal distribution with a mode MIC value of 16 mg/L. For ciprofloxacin, 83% of the isolates were susceptible and 17% were resistant. Bimodal MIC distributions, with two peaks representing a "susceptible" wildtype population and a non-wildtype population with acquired resistance properties, were seen for all the (fluoro)quinolones, including nalidixic acid, ciprofloxacin, enrofloxacin, and marbofloxacin. The same 17 isolates classified as ciprofloxacin-resistant also showed elevated MIC values for nalidixic acid as well as the veterinary fluoroquinolones enrofloxacin and marbofloxacin. All isolates were classified as intermediate to colistin. For the other tested antimicrobial agents there were no clinical breakpoints available. The MIC values were high especially for tiamulin, cefoperazone, and florfenicol, which is in accordance with the intrinsic resistance properties of A. baumannii. Bimodal MIC distributions were also seen for the tetracyclines, namely tetracycline and doxycycline, and also for trimethoprim/sulfamethoxazole.

Macrorestricton Analysis with Subsequent Pulsed-Field Gel Electrophoresis
At the cut off level of ≥ 80%, there were 21 PFGE pulsotypes (A-U) containing up to 21 isolates (pulsotype P). At the cut off level ≥ 87%, there were 33 PFGE pulsotypes. These included up to 11 isolates per pulsotype and up to nine isolates with indistinguishable PFGE patterns (pulsotype B1), including isolates from different farms and different arrival dates ( Figure S1).

Whole-Genome Sequencing
The whole-genome sequencing results are listed in Table 4. rMLST confirmed the assignment to the species A. baumannii with 100% support in all sequenced isolates.
The sul2 gene was detected in one of the two isolates, which showed resistance to trimethoprim/sulfamethoxazole. In the ten sequenced ciprofloxacin-resistant isolates, two gene mutations were detected in gyrA and parC genes resulting in the amino acid substitutions Ser81Leu (GyrA) and Ser84Leu or Ser84Phe (ParC), respectively. Isolate 71_W90.3, which showed an elevated MIC value for nalidixic acid but not for ciprofloxacin, had only a mutation in gyrA, which resulted in the amino acid substitution Ser81Leu (Table 4). Multilocus sequence typing (MLST) analysis using the Pasteur scheme revealed 16 different STs (Table 4). Four STs (2157, 2158, 2159, and 2160) were newly described, and a new fusA allele (detected in isolate 71_W90.3 with the new ST2159), namely Pas_fusA-407, was newly added to the PubMLST database. By far the most commonly detected ST was ST25, comprising nine isolates (35%), followed by ST241 and ST374 with two isolates each (8%), respectively. The 12 remaining isolates all showed individual allelic profiles and different STs (Figure 1).
The sul2 gene was detected in one of the two isolates, which showed resistance to trimethoprim/sulfamethoxazole. In the ten sequenced ciprofloxacin-resistant isolates, two gene mutations were detected in gyrA and parC genes resulting in the amino acid substitutions Ser81Leu (GyrA) and Ser84Leu or Ser84Phe (ParC), respectively. Isolate 71_W90.3, which showed an elevated MIC value for nalidixic acid but not for ciprofloxacin, had only a mutation in gyrA, which resulted in the amino acid substitution Ser81Leu (Table 4).
cgMLST using 1943 alleles for distance calculation showed a wide distribution of the 26 isolates tested. Most of these isolates showed a distinct allelic profile and were not closely related. They showed differences between 1775 and 1820 alleles. There was one cluster with ten related isolates (only up to 96 alleles apart). These ten isolates belonged to the Pasteur  In the MLST analysis using the Oxford scheme, 18 different STs were present. Ten of these STs (2769,2771,2772,2773,2774,2775,2776,2777,2778, and 2779) were newly described, including three new alleles, which were added to the PubMLST database. ST1588 was the most common, including five isolates (19%), followed by ST229 including three isolates (12%), and ST1416 and ST2774 with two isolates each (8%) ( Table 4).
cgMLST using 1943 alleles for distance calculation showed a wide distribution of the 26 isolates tested. Most of these isolates showed a distinct allelic profile and were not closely related. They showed differences between 1775 and 1820 alleles. There was one cluster with ten related isolates (only up to 96 alleles apart). These ten isolates belonged to the Pasteur STs 25 and 2159 (new) and the Oxford STs 229, 1588, 2778 (new), and 2779 (new). The corresponding isolates all showed fluoroquinolone resistance. Otherwise, only two isolate pairs had closely related allelic profiles: isolates 48_W24.2 and 95_W75.1 with only two alleles difference, and isolates 17_W63.2 and 59_W118.3 with three alleles difference ( Figure 2).

Discussion
In the chick-box/meconium samples from one-day-old turkey chicks, there was a very high presence of A. baumannii. Overall, 79.9% of the chick-box-papers contained A. baumannii isolates. Intriguingly, the highest detection rate in birds (25% from n = 661) up till now was also found in white stork nestlings. Other findings in goslings and chickens seem to have also been especially prevalent in younger birds [18]. Interestingly, the detection rate of A. baumannii found in boot swab samples (n = 132) taken during rearing and before slaughter was low, with only 1.5%. Our results, therefore, highlight that the presence of A. baumannii in samples from poultry can vary considerably with the age of the birds and is transient. In another study, for example, A. baumannii was not isolated in bioaerosols from a housing with 7-week-old turkeys [54]. The detection of only one A. baumannii isolate in 217 lung-heart swabs (0.5%) and in none of the yolk sac and liver samples from one-to six-day-old turkey chicks during diagnostics, additionally points towards a generally low presence of A. baumannii in fattening turkeys. The data, therefore, suggest, as in wild birds [24], that there is no evidence for a general preference of A. baumannii for avian hosts. With regard to the diagnostic samples in this study, there was also no evidence of A. baumannii playing a role in diseased turkeys.
The preliminary characterisation via PFGE revealed that in total, the A. baumannii isolates found in this study were very heterogenous, forming 21 pulsotypes at a cut off level of ≥80% and 33 pulsotypes at a cut off level of ≥87%, comprising between one and eleven isolates. Core genome MLST highlighted the diverse population of A. baumannii isolates found in this study. However, as anticipated, the PFGE results did not completely correspond with the core genome MLST data of the 26 isolates subjected to WGS. Due to the very heterogenous isolates, which were not closely related, an environmental origin as discussed in the studies on storks [18] and cattle [16] seems likely. The source of the A. baumannii isolates is not clear. To investigate possible reservoirs in future studies, the environment of one-day-old chicks, i.e., hatcheries and transport vehicles, should be analysed. In other studies, A. baumannii isolates have been found in the air of a duck hatchery [27] and non-sterile water (which is used for humidity regulation during the brood), which has been suggested to be a possible source of contamination in hatcheries [26]. Moreover, feather down has also been considered as a potential carrier [26,55]. In general, Acinetobacter spp. are widespread environmental microorganisms [56] and A. baumannii, for example, can be detected in soil [57,58].
Antimicrobial susceptibility testing revealed that the majority of isolates were susceptible to a wide range of antimicrobial agents, which is comparable with the results obtained from cattle and white storks as well [16,18]. Multidrug resistance properties were not detected. In addition to the species-specific intrinsic resistance properties, only two isolates showed acquired resistance to two different classes of antimicrobial agents, namely (fluoro)quinolones and tetracyclines. The highest resistance rates were detected for ciprofloxacin. The MIC values of the other tested quinolones, for which no clinical breakpoints exist, confirmed this finding. The detected mutations in gyrA and parC genes, respectively, are linked to fluoroquinolone selection of resistance, which suggests that the isolates have been circulating in an environment where fluoroquinolones have been used [59]. Interestingly, the isolate showing only one mutation in the gyrA gene and none in the parC gene revealed only an elevated MIC value for nalidixic acid, but not one for ciprofloxacin. The tetracycline resistance in two isolates could be attributed to the tet(39) gene, which encodes an active efflux mechanism and has been described in Acinetobacter spp. found frequently in the aquatic environment [60,61]. As A. baumannii is intrinsically resistant to trimethoprim, resistance to trimethoprim/sulfamethoxazole could only be attributed to the gene sul2, which confers resistance to sulfonamides [62] in one isolate. The cause of the resistance in the other trimethoprim/sulfamethoxazole-resistant isolate (36_W51.1) was not resolved as no further sul genes nor mutations in the genes folA and folP could be detected. The aminoglycoside nucleotidyltransferase gene ant(3 )-IIa was present in all isolates as described in other studies [63,64]. Phosphotransferase aph(3 )-Ib and aph(6)-Id, which mediate streptomycin resistance, were both detected in isolate 35_W50.1 (streptomycin MIC value 64 mg/L). The streptomycin MIC values of the remaining 25 solates, which did not harbour these two resistance genes, varied between 4 mg/L and ≥128 mg/L. One of the three isolates resistant to cefotaxime (isolate 17_W24.2) was examined by WGS and no additional beta-lactamase gene, except the intrinsic ones, could be identified. The classification of the isolates as cefotaxime-resistant may be due to the unimodal MIC distribution of the tested A. baumannii isolates around the clinical breakpoint. In other studies, cefotaxime resistance has been described in association with the production of the CTX-M-2 extended spectrum class A β-lactamase [65,66]. In general, bla OXA genes are found on both chromosome and plasmids, and it might also be possible that more than one copy of bla OXA is on the chromosome [67]. The isolates tested in this study showed a diverse selection of intrinsic bla OXA β-lactamase genes. It is important to say that we did not detect acquired β-lactamase genes, such as bla OXA-23 or bla OXA-58 , which are associated with carbapenem resistance [2] in any of the 26 sequenced isolates. The gene bla OXA-64 , which was the most frequently detected bla OXA gene in this study, has previously been found in feather down and dust from turkey and goose hatcheries [18]. It correlates with the Pasteur ST25 [68], except in the case of isolate 71_W90.3, which interestingly showed a new Pasteur type, ST2159 (with a new fusA allele), but a known Oxford type ST229. Some other bla OXA β-lactamase genes found in our study have been detected in samples from other avian species as well, i.e., bla OXA-51 (white stork choana), bla OXA-68 (chicken choana, feather down and dust from a chicken hatchery), bla OXA-104 (white stork choana), bla OXA-208 (white stork pellet), and bla OXA-385 (1-day-old chicken choana) [18]. Interestingly, Wilharm et al. could assign two chicken samples with bla OXA-68 (Pasteur ST23) to the international clone 8 (IC8) [18], which includes outbreak strains in human medicine (Pasteur STs 10 and 157) [69]. Only one isolate (16_W23.1) in this study had bla OXA-68 . This isolate showed new Pasteur and Oxford STs. Furthermore, a variety of different bla ADC β-lactamase genes were found. The most common, bla ADC-26 , was mostly present in isolates that also carried bla OXA64 , except for isolate 37_W52.1, which was characterised by bla OXA-104 , Pasteur ST46 and Oxford ST1557, and the isolates 48_W63.2 and 95_W118.3, characterised by bla OXA-259 , Pasteur ST374 and Oxford ST1416.
Biocide susceptibility testing revealed that the MIC values for the four biocides were all distributed unimodally. There is not much data concerning biocide susceptibility available for comparison. In one study, 14 A. baumannii isolates from dogs and cats were examined using the same protocol [70]. Interestingly, the MIC ranges were generally wider in our study presented here, often starting at lower dilution steps. However, it has to be kept in mind that this could be due to the number of isolates tested (n = 99 vs. n = 14).
MLST analysis revealed many different STs, which highlights the heterogenous nature of the isolates. Pasteur ST25, which was detected in 35% of the sequenced isolates, was most prevalent. In humans, ST25 is a successful lineage, which can lead to epidemics, has spread worldwide, and belongs to the international clone 7 (IC7) [68,69,[71][72][73][74][75]. All ST25 isolates described here carry bla OXA-64 and were resistant to ciprofloxacin, which corresponds to the results of other studies [68,71]. Only one of the ten ciprofloxacin-resistant isolates examined by WGS (isolate 66_W83.1) belonged to Pasteur ST333, which was first described in China [76]. Two of our isolates were susceptible to ciprofloxacin and belonged to Pasteur ST374, which, according to the PubMLST database, occurs worldwide. The ST374 lineage is grouped into the clonal complex CC3 belonging to the international clone IC3 [77]. Two further isolates belonged to Pasteur ST241, which has been detected in human samples across the world according to PubMLST database. In Germany, it has been found in a cattle faecal sample [16] and in milk powder [63]. Interestingly, our two ST241 isolates, which were only three alleles apart in cgMLST, had a new Oxford sequence type ST2774, harboured bla OXA-91 and bla ADC-52 , and showed an elevated cefquinome MIC of ≥64 mg/L. None of the Pasteur STs in this study corresponded to those found in chicken and turkey meat in Switzerland [29]. More Oxford STs (n = 18) were found in this study in comparison to the Pasteur STs (n = 16). The Pasteur ST25 comprised four different Oxford STs (ST1588, ST229, ST2779 (new), and ST2778 (new)).
In general, it can be concluded for commercial turkeys, as it has been for cattle [16] and storks [18], that the population of A. baumannii is highly diverse and still susceptible to many antimicrobial agents. The overall occurrence of A. baumannii in samples from commercially reared turkeys seems to be very low. Only chick-box-papers were found to harbour large numbers of A. baumannii isolates. Although Acinetobacter isolates have been obtained from rhizospheric soil, tomato, and cauliflower roots [78], a transfer from these sources to the animals investigated in this study can be excluded as the turkey chicks/turkeys did not have contact to these matrices. Thus, the possible origin of the A. baumannii isolates found in this study remains to be elucidated and will be a subject for further investigation.