Antibiotic Resistance Patterns of Pseudomonas spp. Isolated From Raw Milk Revealed by Whole Genome Sequencing

Psychrotrophic bacteria in raw milk are most well known for their spoilage potential and the economic losses they cause to the dairy industry. Food-related psychrotrophic bacteria are increasingly reported to have antibiotic resistance features. The aim of this study was to evaluate the resistance patterns of Pseudomonas spp. isolated from bulk-tank milk. In total, we investigated the antibiotic susceptibility profiles of 86 Pseudomonas spp. isolates from raw milk. All strains were tested against 15 antimicrobial agents. Pseudomonas isolates were most highly resistant to imipenem (95.3%), followed by trimethoprim-sulfamethoxazole (69.8%), aztreonam (60.5%), chloramphenicol (45.3%), and meropenem (27.9%). Their multiple antibiotic resistance (MAR) index values ranged from 0.0 to 0.8. Whole-genome sequencing revealed the presence of intrinsic resistance determinants, such as BcI, ampC-09, blaCTX-M, oprD, sul1, dfrE, catA1, catB3, catI, floR, and cmlV. Moreover, resistance-nodulation-cell division (RND) and ATP-binding cassette (ABC) antibiotic efflux pumps were also found. This study provides further knowledge of the antibiotic resistance patterns of Pseudomonas spp. in milk, which may advance our understanding of resistance in Pseudomonas and suggests that antibiotic resistance of Pseudomonas spp. in raw milk should be a concern.


INTRODUCTION
Pseudomonas spp., which have been identified as predominantly psychrotrophic bacteria, are important spoilage bacteria in food (Marchand et al., 2009). The Pseudomonas genus is found extensively in environments such as water, soil, and sediment (Devarajan et al., 2016). They are ubiquitous Gram-negative bacteria that have a wide metabolic versatility, which allows them to acclimate to different habitats with temperatures ranging from 4 to 42 • C (Quigley et al., 2013;Molina et al., 2014;Chakravarty and Gregory, 2015). Moreover, Pseudomonas can multiply at cold temperatures and accounts for more than half of all bacteria found in milk (Munsch-Alatossava and Alatossava, 2006;Fricker et al., 2011;Von Neubeck et al., 2015). Pseudomonas strains produce heat-stable extracellular peptidases and/or lipases that spoil raw milk, and some strains are opportunistic pathogens that are found in the environment, such as Pseudomonas aeruginosa (Jeukens et al., 2017).
The widespread administration of antimicrobial agents to animals used for food products has imposed a strong selective pressure that increases resistance among known pathogens and commensal bacteria (Beena et al., 2011;Decimo et al., 2016). Antimicrobial resistant Pseudomonas is also a concern as antimicrobial resistance is a key factor in the emergence of infectious diseases (Jones et al., 2008;Gomi et al., 2017). A study based on phenotypes and 16S rDNA gene sequences identified Pseudomonas isolates that were categorized as being a high risk for antibiotic resistance (Munsch-Alatossava et al., 2012). Moreover, Pseudomonas spp. have the ability to remain viable in the aquatic environment for long periods owing to innate resistance mechanisms, thereby increasing the risk of spreading antibiotic resistance genes and mobile genetic elements (Devarajan et al., 2016). The exchange of genetic material encoding resistance genes via mobile genetic elements, plasmids, or transposons can result in transfer of resistance between pathogenic and non-pathogenic bacteria (Mathur and Singh, 2005;Panelists, 2006;Decimo et al., 2016). Yomoda et al. (2003) reported that 22 of 32 Pseudomonas putida strains isolated from Japanese hospitals carried plasmids that could be transferred to P. aeruginosa by conjugation or transformation. Therefore, the transfer of antibiotic resistance determinants from nonpathogenic species to pathogens is a serious concern (Yomoda et al., 2003;Molina et al., 2014).
Antibiotic resistance has been a focus of research on foodrelated and mastitis pathogens in milk for many years. The potential for non-pathogenic commensal foodborne bacteria to become a biological hazard has also been investigated as resistance genes can be spread to humans via food (Decimo et al., 2016). The aim of this study was to evaluate the resistance of Pseudomonas spp. isolated from bulk-tank milk in China to different antibiotics. This study represents an extensive investigation of resistance of Pseudomonas spp. to a number of human and veterinary antimicrobial agents.

Bacterial Isolates
A total of 143 Pseudomonas strains were isolated from raw milk samples from 87 bulk tanks in Shaanxi province, China, in spring 2016 (average daily temperature >20 • C) for a previous study of their proteolytic properties . All isolates were re-identified using the rpoD gene as rpoD provides higher resolution than 16S rDNA gene sequences for Pseudomonas species (Rajwar and Sahgal, 2016). In total, 86 isolates representing 11 different Pseudomonas spp. were subjected to antimicrobial susceptibility testing. A brief description of these 86 isolates is provided in Supplementary Table S1.

Analysis of Sequence Data
Data generated from the Illumina platform were used for bioinformatics analysis. All analyses were performed using the I-Sanger Cloud Platform 1 from Shanghai Majorbio BioTech Co., Ltd. (Shanghai, China). Briefly, the original image data were transferred into sequence data, defined as raw data or raw reads, by base calling and saved as a FASTQ file. A quality statistic was then applied for quality trimming to remove low-quality data from the clean data. Assembly of the clean reads was performed using SOAPdenovo2 (Luo et al., 2015). Finally, resistance genes were detected using the ResFinder web server and ARG-ANNOT database (Zankari et al., 2012;Gupta et al., 2014). All whole genome sequence data from this study have been deposited in the NCBI Sequence Read Archive database (accession numbers PRJNA523883, PRJNA523885, and PRJNA593738).

Antimicrobial Susceptibility Profiles of
Pseudomonas spp. of Pseudomonas strains were susceptible to gentamicin and ciprofloxacin, followed by 93.0% that were susceptible to tetracycline and 91.9% that were susceptible to levofloxacin. However, a high percentage of Pseudomonas strains were resistant to imipenem (95.3%). Resistance to trimethoprimsulfamethoxazole (69.8%), aztreonam (60.5%), chloramphenicol (45.3%), and meropenem (27.9%) was also widespread in this study. The distribution of resistance patterns to the 10 antibiotics tested is shown in Figure 1. In total, 48 (55.8%) isolates were resistant to three or more drug classes and were defined as MDR strains. One isolate was not resistant to all eight classes of antibiotics, and one isolate could resist seven classes of antibiotics. Moreover, the 86 tested isolates had MAR index values ranging from 0.0 to 0.8 (Supplementary

Resistance Determinants
Overall, more than 100 resistance determinants were found in 44 isolates (Table 2), and all of the isolates carried resistance determinants. Eleven different aminoglycoside resistance genes were detected. The most prevalent aminoglycoside resistance gene, aph(6)-Ic, is one of three resistance genes present in the composite transposon Tn5, which is found in Gram-negative bacteria (Steiniger-White et al., 2004). The ksgA gene, which confers resistance to kasugamycin (Zou et al., 2018), was also found in seven isolates. Genes ant(3")-Ia, aph(6)-Id and aph(3')-IIc as well as rpsL mutations, known to be responsible for streptomycin resistance (Ramirez and Tolmasky, 2010), were also found in these isolates (Supplementary Table S3).
Another important resistance gene group was found to confer resistance to lipopeptides. In total, 21 resistance genes have been detected with 12 genes conferring resistance to polymyxin, such as phoP and phoQ. The most common gene, bacA, is a bacitracin resistance gene (Shaaly et al., 2013), and bcrA is another one, but it is only found in two isolates. floR is a resistance gene against florfenicol (Arcangioli et al., 1999). Additionally, mutations in liaR, liaS, cls, pgsA, and rpoC, known to be responsible for daptomycin resistance (Yang et al., 2010;Arias et al., 2011;Peleg et al., 2012;Miller et al., 2013), were found in a few strains ( Table 2 and Supplementary Table S3).

DISCUSSION
Bacterial antibiotic resistance is considered a worldwide problem in the medical, environmental, and agricultural fields. Many researchers have focused on antibiotic resistance in pathogenic bacteria, which pose immediate risks to human health. However, more and more interests are focusing on commensal bacteria associated with food (Munsch-Alatossava et al., 2012;Decimo et al., 2016). Pseudomonas spp. isolated from raw milk shows extensive resistance to many antibiotics, and some even exhibit the highest known levels of antibiotic resistance (Straley et al., 2006;Munsch-Alatossava and Alatossava, 2007;Decimo et al., 2016). EUCAST and CLSI criteria are the standards for bacterial resistance assessment. However, clinical breakpoints are based on parameters that are only relevant for therapeutic purposes. Therefore, assessing the antibiotic susceptibility of environmental bacteria using only CLSI criteria is inadequate. In contrast, EUCAST criteria are clinically and/or epidemiologically based and are more reliable for the interpretation of the antibiotic Frontiers in Microbiology | www.frontiersin.org susceptibility of environmental bacteria (Berendonk et al., 2015;Gomi et al., 2017). Therefore, European Committee on Antimicrobial Susceptibility Testing [EUCAST] (2018) criteria were used in the present study to assess the resistance phenotypes of Pseudomonas spp. to aztreonam, ciprofloxacin, levofloxacin, gentamicin, imipenem, and meropenem. However, Clinical and Laboratory Standards Institute [CLSI] (2016, 2018) criteria were used to interpret chloramphenicol, polymyxin B, tetracycline, and trimethoprim-sulfamethoxazole resistance as there is no MIC value for these antibiotics in the European Committee on Antimicrobial Susceptibility Testing [EUCAST] (2018) criteria.
According to a previous survey, β-lactam antibiotics are commonly used for dairy mastitis therapy . Therefore, seven β-lactam antibiotics (ampicillin, aztreonam, cefoxitin, ceftiofur, imipenem, meropenem, and penicillin) were tested in the present study. Around 60% of Pseudomonas strains were resistant to aztreonam, which is higher than described by Decimo et al. (2016). A study conducted between 2010 and 2012 reported that 51-100% of Pseudomonas spp. in non-turbid drinking water were resistant to aztreonam (Flores Ribeiro et al., 2014). In contrast to our results, which showed that 95.3% of Pseudomonas strains were resistant to imipenem, Decimo et al. (2016) found that only 11% of Pseudomonas spp. from bulk-tank milk were resistant to imipenem, and Arslan et al. (2011) found that all Pseudomonas spp. isolates from cheese were susceptible to imipenem. In the present study, the meropenem resistance rate was only 27.9% (n = 24), which is similar to resistance rates reported in previous studies of Pseudomonas spp. from bulk-tank milk and the Danube River (Decimo et al., 2016;Kittinger et al., 2016). Although no breakpoints were available for the four other β-lactam antibiotics, we found that 59.3, 98.8, 2.3, and 97.7% of isolates had MIC value ≥256 µg/mL for ampicillin, cefoxitin, ceftiofur, and penicillin, respectively.
In the 1980s, studies reported that P. aeruginosa that were the dominant cause of nosocomial infections were resistant to almost all aminoglycosides and β-lactams with the exception of cephalosporins and carbapenems (Neu, 1983;Acar, 1985). Penicillin G is still the most commonly used antibiotic for dry cow therapy (United States Department of Agriculture, Animal Plant Health Inspection Service National Animal Health Monitoring System [USDA APHIS], 2008), and one study found that 100% of Pseudomonas spp. isolates from cheese were resistant to penicillin G (Arslan et al., 2011).
In the present study, 14 different β-lactamase genes were detected in approximately 20 of the Pseudomonas isolates, which were resistant to at least one β-lactams antibiotic from among aztreonam, imipenem, and meropenem except 149-2 and 152-2. Gram-negative bacteria produce AmpC-type β-lactamases that can hydrolyze amino-and ureido-penicillins, cephamycin, andat low levels-cephalosporin and monobactams (Rodriguez-Martinez et al., 2009). Four isolates carrying ampC-09 were all resistant to aztreonam, MIC value for cefoxitin ≥256 µg/mL, and MIC value for cefoxitin ≥64 µg/mL except 198-5. Two isolates 141-6 and 151-5 harbored bla CMY−51 , an ampC variant first described in Citrobacter freundii (Porres-Osante et al., 2014), and were resistant to aztreonam, MIC value for cefoxitin ≥256 µg/mL, and MIC value for cefoxitin ≥128 µg/mL. The isolate 114-2 was intermediately resistant to aztreonam with the gene bla CTX−M , which has been demonstrated to hydrolyze aztreonam (Wang et al., 2011). All five isolates with OCH-8 and LRA-2 showed MIC value 128 µg/mL for ceftiofur. However, five isolates taking BcI, bla CTX−M , and LRA-13 showed a ceftiofur MIC value for ceftiofur ≤64 µg/mL. The resistance genes that have been found were much lower than the β-lactam resistance rates. Meanwhile, Pseudomonas spp. are reported to be naturally resistant to penicillin G as well as the majority
Fluoroquinolone resistance has been associated with mutations in the genes gyrA and parC (Aoike et al., 2013). The mutations were found in two isolates, 141-6 and 151-5, that were resistant to ciprofloxacin and levofloxacin. Moreover, these two isolates also carried eight other quinolones resistance determinants, including mutations in the quinolone resistance-determining regions of DNA gyrase (gyrB) and DNA topoisomerase IV (parE). Isolate 103-1, which carried a gyrA mutation from E. coli, also exhibited resistance to ciprofloxacin and levofloxacin. Five isolates harbored the plasmid-mediated quinolone resistance gene qnrB; however, these isolates were all susceptible to ciprofloxacin and levofloxacin. Moreover, 60 isolates were resistant to trimethoprim-sulfamethoxazole. However, only five isolates carried the trimethoprim resistance gene dfrE or the sulfamethoxazole resistance gene sul1 (Toleman et al., 2007;Chen et al., 2019). Resistance to ciprofloxacin and levofloxacin in the other isolates may have been due to RND efflux pumps (Kohler et al., 1996;Li et al., 2015).
Polymyxin lipopeptide antibiotics are currently last-resort antibiotics for the treatment of MDR Gram-negative bacterial infections. Of 21 lipopeptide resistance genes, 12 confer resistance to polymyxin, including arnA, phoP, phoQ, pmrA, pmrB, pmrC, pmrE, pmrF, rosA, rosB, lpxA, and lpxC, with phoP and phoQ being the most common (McPhee et al., 2003;Gatzeva-Topalova et al., 2005;Lee and Ko, 2014;Zhang et al., 2017). Polymyxin B resistance has been found to be conferred by mutations in two-component regulatory systems of phoPQ, which are well known to contribute to polymyxin resistance in P. aeruginosa (Barrow and Kwon, 2009;Miller et al., 2011;Lee and Ko, 2014). The PmrF operon also plays an essential role in resistance to polymyxin and can be upregulated by PhoPQ (Marceau et al., 2004). In the present study, the number of mutations in phoPQ regions was higher than polymyxin B resistance rates, and a few polymyxin B-resistant isolates carried the fosmidomycin resistance genes rosA and rosB, which encode an efflux pump that confers resistance to polymyxin B in response to polymyxin B (Bengoechea and Skurnik, 2000).
In our study, resistance rates of Pseudomonas isolates to chloramphenicol and tetracycline were 45.3% (39/86) and 4.7% (4/86), respectively, which was much lower than the ratios of Pseudomonas aeruginosa isolated from milk samples and from major hospitals and laboratories in Jamaica (Brown and Izundu, 2004;Decimo et al., 2016). The chloramphenicol resistance is mainly caused by enzymatic inactivation by chloramphenicol acetyltransferases (CAT) and also efflux mechanisms mediated by cml genes (Schwarz et al., 2004). In total, cat and cml genes were found in five isolates, which is less than the number of chloramphenicol resistance isolates (n = 39). However, the effects of simultaneous expression of multidrug transporters, such as AcrAB-TolC or MexAB-OprM, can increase the chloramphenicol resistance of P. aeruginosa (Lee et al., 2000). Seven different tetracycline resistance genes have been found in this study, of which otrA, tetB(P), tetM, and tetT could code ribosomal protection proteins and tet42, tetA, and tetG are tetracycline efflux pump genes (Thaker et al., 2010). However, only two tetracycline-resistant isolates harbored the tetG gene. The other two resistant isolates harbored RND efflux pumps, such as MexCD-OprJ, which could help the isolates resistant to tetracycline . The isolates that carried the resistance determinants were not resistant to tetracycline, especially 114-6 and 115-5, which had five and six different resistance genes, respectively. Moreover, MIC values of 83 isolates were ≥256 µg/mL for clindamycin. Gram-negative aerobic organisms are reported to be uniformly resistant to clindamycin, whereas only a few resistance genes have been detected (Ras et al., 1992).
One interesting finding was the presence of a triclosan efflux pump. The genes triA, triB, and triC, which were present in three isolates, are tightly coupled transcripts that form the TriABC protein and associate with OpmH to assemble a functional triclosan efflux pump (Mima et al., 2007). Triclosan is now used in many consumer products, including toothpastes, soaps, cosmetics, cutting boards, and mattress pads (Bhargava and Leonard, 1996;Chuanchuen et al., 2002). Triclosan contamination is raising concerns due to the high risk of it converting into toxic dioxin in aquatic environments, and attention should be paid to its usage (Yan et al., 2019).
In total, 48 isolates were MDR strains with 22 different combined groups (Supplementary Table S1). Although the isolates harbor ARGs, RND efflux pumps may play a role in resistance because they have wide antibiotic substrate spectrums (Depardieu et al., 2007). The MAR index is considered a good risk assessment tool. A MAR index value of 0.20 used to differentiate between low and high risk of contamination (Krumperman, 1983;Riaz et al., 2011). The MAR index values of the 86 isolates tested in this study were between 0.0 and 0.8 with 59.3% (51/86) of isolates having MAR indices >0.20, suggesting high antibiotic use and high selective pressure. Swetha et al. (2017) found that 19 P. aeruginosa isolates from raw milk samples had MAR index values ranging from 0.33 to 1.0 and were resistant to between four and 12 antibiotics. Pseudomonas isolates from wastewater treatment facilities in South Africa had MAR index values ranging from 0.26 to 0.58 and were resistant to between five and 11 antibiotics (Odjadjare et al., 2012). Moreover, some Pseudomonas isolates acquire resistance to antibiotics through horizontal gene transfer (HGT) of plasmids carrying genetic material encoding for antibiotic resistance. We found 16 resistance genes from E. coli, S. aureus, and M. tuberculosis, indicating that HGT played a key role in the development of these Pseudomonas strains. HGT can provide genes necessary for survival more quickly than spontaneous mutations (Charpentier et al., 2012). Although antibiotic resistance genes are an important public health concern in clinical and veterinary environments, it is unclear whether resistance genes of pathogenic microorganisms are transmitted from environmental bacteria or elsewhere (Forsberg et al., 2012). Therefore, HGT between Pseudomonas isolates and other bacteria should be further investigated.

CONCLUSION
Knowledge of the antibiotic resistance profiles of food-related bacteria, including Pseudomonas spp., is currently needed. In this study, we successfully characterized the antibiotic susceptibility profiles of Pseudomonas spp. isolates from raw milk and assessed their antibiotic resistance determinants using whole genome sequencing. Our results indicate that MDR Pseudomonas species were prevalent in raw milk in Shaanxi province, China. Moreover, whole genome sequencing revealed the presence of different resistance determinants in all of the isolates as well as acquired resistance genes from other bacteria, demonstrating that HGT occurred between these Pseudomonas isolates and other bacteria. Although this study was conducted in only one region, the emergence of MDR Pseudomonas species and HGT is an important public health issue. Therefore, our data support the need for further investigations into Pseudomonas species to prevent the spread of resistance to other pathogenic bacteria.

DATA AVAILABILITY STATEMENT
The datasets generated for this study can be found in the NCBI Sequence Read Archive database (accession numbers PRJNA523883, PRJNA523885, and PRJNA593738).

AUTHOR CONTRIBUTIONS
LM and HL contributed equally, designed and performed the research. TL, LD, and HH helped with antimicrobial susceptibility testing. SZ and YZ helped with the data analysis.
NZ gave advices to the researchers. JW gave opinions on the research design.