Antibiotic-resistant bacteria, antibiotic resistance genes, and antibiotic residues in wastewater from a poultry slaughterhouse after conventional and advanced treatments

Slaughterhouse wastewater is considered a reservoir for antibiotic-resistant bacteria and antibiotic residues, which are not sufficiently removed by conventional treatment processes. This study focuses on the occurrence of ESKAPE bacteria (Enterococcus spp., S. aureus, K. pneumoniae, A. baumannii, P. aeruginosa, Enterobacter spp.), ESBL (extended-spectrum β-lactamase)-producing E. coli, antibiotic resistance genes (ARGs) and antibiotic residues in wastewater from a poultry slaughterhouse. The efficacy of conventional and advanced treatments (i.e., ozonation) of the in-house wastewater treatment plant regarding their removal was also evaluated. Target culturable bacteria were detected only in the influent and effluent after conventional treatment. High abundances of genes (e.g., blaTEM, blaCTX-M-15, blaCTX-M-32, blaOXA-48, blaCMY and mcr-1) of up to 1.48 × 106 copies/100 mL were detected in raw influent. All of them were already significantly reduced by 1–4.2 log units after conventional treatment. Following ozonation, mcr-1 and blaCTX-M-32 were further reduced below the limit of detection. Antibiotic residues were detected in 55.6% (n = 10/18) of the wastewater samples. Despite the significant reduction through conventional and advanced treatments, effluents still exhibited high concentrations of some ARGs (e.g., sul1, ermB and blaOXA-48), ranging from 1.75 × 102 to 3.44 × 103 copies/100 mL. Thus, a combination of oxidative, adsorptive and membrane-based technologies should be considered.

Molecular typing of ESBL-producing E. coli and MRSA. The majority of E. coli isolates were assigned to phylogroups B1 (50.0%) and E (27.3%), which are commonly associated with commensal strains. Less abundant phylogroups were represented by A and F (each 9.1%) as well as C (4.5%). Notably, extraintestinal pathogenic (ExPEC) groups B2 and D were not detected.

Figure 2.
Abundance of "frequent" antibiotic resistance genes in influent and effluent after physicochemical and biological treatments and subsequent ozone treatment. Displayed are mean values with standard deviation. Significance is given by one-tailed nonparametric Mann-Whitney U test calculation and is shown by asterisks (*p ≤ 0.05, **p ≤ 0.001). Compared are gene copies per 100 mL in (i) raw influent with effluent after physicochemical and biological treatments and (ii) in effluent after physicochemical and biological treatments with effluent after subsequent ozone treatment. Significance is given by one-tailed nonparametric Mann-Whitney U test calculation and is shown by asterisks (**p ≤ 0.001). Compared are gene copies per 100 mL in (i) raw influent with effluent after physicochemical and biological treatments and (ii) in effluent after physicochemical and biological treatments with effluent after subsequent ozone treatment. LOD -limit of detection.  (Fig. 4). E. coli followed by enterococci and A. baumannii were the most predominant target species in the influent, with abundances of 6.90 × 10 4 , 8.41 × 10 3 and 7.47 × 10 3 cell equivalents per 100 mL, respectively. P. aeruginosa exhibited the lowest concentration of 4.32 × 10 1 cell equivalents per 100 mL. After the physicochemical and biological treatments, all target species were reduced significantly (p < 0.001). Moreover, K. pneumoniae was reduced < LOD. After the subsequent ozonation, a nonsignificant reduction in E. coli and enterococci was observed, whereas P. aeruginosa was significantly reduced < LOD (p < 0.001).

Discussion
This study provides insights regarding the occurrence and impact of ESKAPE bacteria, including ESBL-producing E. coli, ARGs and antimicrobial residues, in wastewater samples from a German poultry slaughterhouse that might be disseminated to the environment. Furthermore, we evaluated the efficacy of ozonation treatment of wastewater for a reduction in bacterial loads and ARGs. Significance is given by one-tailed nonparametric Mann-Whitney U test calculation and is shown by asterisks (**p ≤ 0.001). Compared are gene copies per 100 mL in (i) raw influent with effluent after physicochemical and biological treatments and (ii) in effluent after physicochemical and biological treatments with effluent after subsequent ozone treatment. LOD -limit of detection. Table 1. Antimicrobial residues detected in wastewater samples from the in-house WWTP of the examined poultry slaughterhouse. a Abbreviations for antimicrobial agents: TYL tylosin, SXM sulfamethoxazole, SDD sulfadimidine, ENR enrofloxacin, AMP ampicillin. b N-Sampling number. c Only samples with concentrations > LOQ were used for the evaluation. www.nature.com/scientificreports/ The high number of notified 3MDRO E. coli and K. pneumoniae isolates might be caused by the use of β-lactams (e.g., cephalosporines, penicillins) and fluoroquinolones for the treatment of animal infections in the veterinary medicine 20 . Cephalosporines and fluoroquinolones belong to the veterinary critically important antimicrobial agents (VCIAs) and are crucial for combating specific infectious diseases (e.g., septicaemias and respiratory and enteric diseases) in livestock. However, the use of VCIA also affects public health, as these antimicrobials are also considered the highest priority critically important antimicrobials (HPCIAs) for humans 21 .
The detected concentrations of enrofloxacin in untreated wastewater exceeded its Predicted No Effect Concentrations (PNECs) for resistance selection of 0.064 µg/L 25 , indicating that the withdrawal period of enrofloxacin might not be strictly adhered to. This selective pressure might further facilitate the development of fluoroquinolone resistance in clinically relevant bacteria. Thus, the use of fluoroquinolones should be restricted to individual treatments and no longer permitted for general medication of poultry herds. Thus, we strongly recommend forcing the development of alternative antimicrobial-free treatment strategies to avoid the use of similar antimicrobials in different compartments.
The high abundance of β-lactam (bla) resistance genes in the investigated cephalosporin-resistant E. coli correlates well with the detected prevalence of bla TEM-1 , bla TEM-52 , bla TEM-116 , bla SHV-1 , bla SHV-12 and bla CTX-M-1 in isolates of livestock wastewater and poultry products 26,27 . In general, ESBL-producing E. coli carrying diverse virulence determinants (i.e. for human colonization and infection) are widely disseminated in German poultry production 16,28 . While other reports have detected up to 16.2% ExPEC isolates from poultry wastewater, no isolate of this study was assigned to the ExPEC phylogroups B2 or D 16 .
The development and dissemination of ESBL-producing Enterobacteriaceae are not only promoted by the use of β-lactams in veterinary medicine but also facilitated by co-and cross-resistance through disinfectants, heavy metals, and other antimicrobials, since corresponding genes are often located on the same plasmid 29 . The cooccurrence of bla and mcr genes is of particular importance, since colistin was reintroduced into human therapy to treat infections caused by carbapenemase-producing Enterobacteriaceae (CPE) or multidrug-resistant A. baumannii and P. aeruginosa 30 . Moreover, mobile colistin resistance genes (esp. mcr-1) is considered to originate from livestock 31 . Colistin-resistant Enterobacteriaceae carrying mcr-1 on a wide variety of transmissible plasmids have already been reported from German process waters and wastewaters of livestock slaughterhouses, municipal WWTPs, and hospitals 18,32 .
Reports on the occurrence of ARGs in livestock wastewater conferring resistance to HPCIA have already been published 33 . However, there is a lack of qualitative and quantitative data for German slaughterhouses. We detected high abundances of the mcr-1, bla CMY-2 and bla CTX-M genes in untreated wastewater, which is worrying since they confer resistance to HPCIA in humans 20 . Their potential incorporation into clinically relevant bacteria might narrow antimicrobial treatment options and compromise their efficacy in cases of infection, especially if resistance against different last resort antibiotics (e.g., carbapenemases, mcr) occurs 34,35 . The human impact of livestock as a reservoir for combinations of antimicrobials of the last resort (i.e. bla NDM-1 /mcr-1) have already been described 36,37 .
Of special concern is the occurrence of the carbapenemase gene bla OXA-48 among investigated wastewater samples detected by molecular-based methods. As selective cultivation did not yield any carbapenem-resistant isolates, culturable carbapenemase-producing Enterobacteriaceae can be excluded as predominant hosts for bla  . Thus, molecular detection of carbapenemase genes in environmental samples has only limited value, as the host species is not identified. However, other studies demonstrated different transferable plasmids carrying the bla OXA-48 gene in Enterobacteriaceae, which might not be cultivable or underrepresented in the analysed isolates of our study 38,39 . A probable reservoir for bla OXA-48 might be Shewanella spp., as this gene naturally occurs in these bacteria. As Shewanella spp. are not of particular clinical relevance, they might play a role as drivers of bla OXA-48 spread to clinically relevant bacteria (i.e. ESKAPE) 40 . The slight increase of bla OXA-48 in effluent samples treated with zone might indicate its selective properties regarding antibiotic-resistant, and more ozone-tolerant bacterial species 41 .
The presence of the vanA gene in untreated wastewater is worrying, as vancomycin also represents an antimicrobial of the last resort that is critically important for human medicine 21 . However, cultural analysis for vancomycin-resistant enterococci (VRE) was negative, which is in line with other studies reporting the absence or rare occurrence of VRE in livestock wastewater from German slaughterhouses 17,19 . This might indicate the presence of vancomycin-variable enterococci (VVE), which are vanA-positive but phenotypically susceptible to vancomycin 42 . Because of their susceptibility to vancomycin, traditional methods fail to detect VVE. vanA encodes inducible high-level resistance to glycopeptides, and VVE have the ability to revert into vancomycinresistant phenotypes upon vancomycin exposure 42 . However, glycopeptides are not approved for the treatment of livestock in the EU, and the use of avoparcin in feed as a growth promoter was banned in Germany in 1996 43 . Moreover, after treatments in WWTPs, vanA and enterococci were reduced below the detection limit. Thus, the role of agriculture in Germany in the development of glycopeptide-resistant enterococci is less significant than that in the general community and hospitals. Pseudomonas spp., Aeromonas spp. and Raoutella spp. isolated from surface water have also been reported to carry the vanA gene 44 48 . In 2018, 571.7 tons of antimicrobials of these classes were sold to veterinarians in Germany, making up 79.1% of the total amount 49 . Interestingly, the supplied quantities of these antimicrobials correlate well with the abundances of particular ARGs in the investigated wastewater samples. Moreover, high abundances of ermB, sul1 and tetM genes are in line with other studies reporting a high prevalence of these resistance determinants in isolates recovered from livestock wastewater and poultry products, inter alia, in ExPEC and MDR E. coli isolates 16 . The high abundance of ermB in wastewater is critical since ermB encodes resistance to macrolides, which are HPCIAs for human medicine and are crucial for the treatment of severe Campylobacter infections, particularly in children 21 . Furthermore, erm genes could be transferred to gram-positive pathogens (e.g., staphylococci, streptococci, enterococci), resulting in MLS B (macrolide, lincosamide, streptogramin B) cross-resistance, compromising the efficacy of further antibiotics such as erythromycin and clindamycin 50 . This might be the reason for the high resistance rates of MRSA to erythromycin and its combination with clindamycin. MRSA of CC9 and CC398, which were detected in this study, have already been reported in wastewater from poultry slaughterhouses and in retail poultry products 17,51 . They were also isolated from human infections, especially in regions with high livestock production, underlying their zoonotic potential 52 . Interestingly, the detected concentration of tylosin (12 µg/L) might exert selective pressure on species with resistance to macrolides, including MRSA, as it exceeded its PNEC of 4 µg/L. However, detected antimicrobials were eliminated after ozonation, which is in line with other studies reporting high degradation rates (70-100%) of macrolides as well as sulfonamides and fluoroquinolones depending on ozone dose and reaction time 53 . The removal of various organic micropollutants, which might not be reduced by other advanced treatment technologies, e.g., ultrafiltration, is an advantage of ozonation.
However, additional ozonation following conventional treatment did not significantly reduce the loads of facultative pathogenic bacteria, which might be due to different factors, such as ozonation time, concentration of the applied ozone, wastewater composition and physical characteristics of the examined organisms 41 . It is important to note that a high number of bacterial cells in wastewater might be aggregated in flocs, protecting the inner cells from the damaging effects of ozone. Comparable results regarding microbial reduction rates were observed in German municipal WWTPs after conventional treatment 54 .
Conventional treatment significantly reduced the abundances of ARGs conferring resistance to clinically relevant antimicrobials. Some of them (i.e., mcr-1, bla CTX-M-32 ) were further reduced below the limit of detection after subsequent additional treatment with ozone, underlying the importance of advanced treatment. However, despite the significant reduction through conventional and advanced treatments, effluents still exhibited relatively high concentrations of some ARGs, e.g., sul1, ermB and bla OXA-48 (> 10 2 copies/100 mL). Similar reduction rates were reported by Czekalski and colleagues 41 . Ozonation does not show the best effect on the reduction of ARGs in comparison to other methods, e.g., chlorination or ultrafiltration. This might be due to low ozone dosages frequently used in practice 13,55 . However, increasing the amount of ozone might lead to the release of harmful products to the environment and have an ecotoxicological effect 56 . Applying additional filtration steps (e.g., charcoal or sand filtration) after ozonation to remove remaining ozone and unwanted byproducts may result in bacterial regrowth 13 . Moreover, the receiving water bodies can support bacterial regrowth if they contain the necessary nutrients. Iakovides et al. reported the regrowth of total and antibiotic-resistant E. coli after the stress caused by ozone was relieved, indicating the importance of proper setting of ozonation parameters 53 . For wastewater exhibiting high concentrations of ARGs and antibiotic-resistant bacteria, a combination of oxidative, adsorptive (charcoal or sand filtration), and membrane-based technologies should be considered to interrupt dissemination of (facultative) pathogenic antibiotic-resistant bacteria to the environment.

Material and methods
Examined poultry slaughterhouse and its wastewater management. The investigated poultry slaughterhouse exhibited a capacity of > 100,000 slaughtered chickens per day by producing 3,600 m 3 wastewater that was treated in an in-house wastewater treatment plant (WWTP) before being discharged into a preflooder and a receiving waterbody (i.e., river).
On-site treatment of wastewater is based on physicochemical, biological, and advanced oxidation processes. First, the wastewater was mechanically pretreated using screeners and sieves. Fat and greases are removed using grease traps. Afterwards, the wastewater is treated by dissolved air flotation with a subsequent biological treatment by activated sludge. Additionally, the in-house WWTP was equipped with an ozone system (Xylem Water Solutions Deutschland GmbH, Großostheim, Germany). The ozone dosage used was 75 g/m 3 , and the contact time varied between 15 and 30 min, depending on the water flow rate.
Sampling and sample preparation. In a two-month period, the in-house WWTP of a poultry slaughterhouse was sampled six times with a minimum time interval of 1 week. During sampling, a total of 18 samples were collected representing the influent (n = 6), the effluent after physicochemical and biological treatments (n = 6) and the effluent after ozone treatment (n = 6). The samples were taken as qualified samples according to the German standard methods for the examination of water, wastewater, and sludge (DIN 38402-11:2009-02) 57 . Therefore, five subsamples of 200 mL in two-minute sampling intervals were collected and mixed. The samples were transported to the laboratory in a Styrofoam box cooled to 5 ± 2 °C and were further processed within 24 h. Influent samples were manually filtered using stomacher strainer bags with a tissue filter (pore size, 0.5 mm; VWR, Radnor, PA, USA) to remove large particles.
Cultivation, identification and susceptibility testing of target bacteria. The  www.nature.com/scientificreports/ and P. aeruginosa, as well as MRSA and vancomycin-resistant enterococci (VRE). Detailed information on their cultivation procedures has already been published 17 . Species identification was performed by MALDI-ToF MS (bioMérieux, Marcy-l'Étoile, France) equipped with Myla software. The isolates were purified on Columbia agar supplemented with 5% sheep blood, and cryopreservation at − 80 °C in cryotubes (Mast Diagnostics, Reinfeld, Germany) was used for storage. Isolated target bacteria were further subjected to antimicrobial susceptibility testing by the microdilution method according to protocols of the European Committee on Antimicrobial Susceptibility Testing (EUCAST v 11.0) using the Micronaut-S MDR MRGN-Screening system for Gram-negatives and the MICRONAUT-S MRSA/GP testing panel for Gram-positive bacteria (MERLIN, Gesellschaft für mikrobiologische Diagnostika GmbH, Bornheim-Hersel, Germany) 58 . The results were evaluated based on clinical cut-off values provided by EUCAST 59 . The multidrug-resistance phenotype (3MDRO) was defined based on combined resistance to piperacillin (PIP), cefotaxime (CTX) and/or ceftazidime (CAZ), and ciprofloxacin (CIP) as previously described 60 . Detection and analysis of selected antibiotic resistance genes (ARGs) in target bacteria. Template DNA for PCRs was prepared by boiling bacterial suspensions in 10 mM Tris-EDTA pH 8.0 (Sigma-Aldrich, St. Louis, MO, USA) according to Aldous et al. 61 . Enterobacterales with phenotypic resistance to 3rd-generation cephalosporins were screened by PCR for β-lactamase (bla) genes encoding enzymes SHV, TEM, and CTX-M (groups 1, 2, 8 and 9) as previously described [62][63][64] . Isolates of the A. calcoaceticus-baumannii complex and P. aeruginosa were examined for the presence of bla PER , bla GES , and bla VEB by PCR as described 65 . Colistin-resistant isolates (MIC > 2 mg/L) were screened for mcr-1 to mcr-5 as well as mcr-6 to mcr-9 genes using multiplex PCR protocols as described by Rebelo et al. 66 and Borowiak et al. 67 , respectively. The obtained PCR amplicons were purified with the innuPREP DOUBLEpure Kit (Analytik Jena AG, Jena, Germany) and subjected to Sanger sequencing at Microsynth Seqlab (Göttingen, Germany).

Molecular typing of resistant bacterial isolates.
Phylotyping of E. coli isolates (A, B1, B2, C, D, E, F, clade I-V) was conducted as previously described 68 . MRSA isolates were spa-typed by amplifying and sequencing the Staphylococcus protein A repeat region according to Harmsen et al. 69 . The Ridom spa server database (http:// www. spase rver. ridom. de) was used for assignment of spa types.
DNA extraction from water samples, quantification of antibiotic resistance genes (ARGs) and taxonomic gene markers. Volumes of 50 mL, 200 mL and 400 mL of influent, effluent after physicochemical and biological treatments and effluent after ozone treatment were subjected to DNA isolation after treatment with 0.25 mM propidium monoazide (PMA) (BLU-V viability PMA-kit, Qiagen GmbH, Hilden, Germany) as previously described 12,14 .
Determination of antimicrobial residues. Water samples were analysed for 45 antibiotics and two metabolites (N-acetylsulfamethoxazole and anhydroerythromycin) by high-performance liquid chromatography (HPLC) coupled to tandem mass spectrometry (MSMS) after dilution and filtration through hydrophilic PTFE filters (Macherey-Nagel, Düren, Germany) as previously described 71 . The analysed antibiotics belong to the following substance classes: β-lactams (i.e. penicillins, cephalosporins and carbapenems), tetracyclines, fluoroquinolones, sulfonamides (as well as their synergist trimethoprim), macrolides including tylosin and spiramycin, lincosamides, glycopeptides, oxazolidinones, and nitroimidazoles. All analysed antibiotics, including their limit of quantification (LOQ), are given in Supplemental Material Table S3. The LOD of each individual analyte was one-third of the respective LOQ.