# GENETICS OF ACQUIRED ANTIMICROBIAL RESISTANCE IN ANIMAL AND ZOONOTIC PATHOGENS

EDITED BY: Axel Cloeckaert, Michel Stanislas Zygmunt and Benoît Doublet PUBLISHED IN: Frontiers in Microbiology

#### *Frontiers Copyright Statement*

*© Copyright 2007-2018 Frontiers Media SA. All rights reserved. All content included on this site, such as text, graphics, logos, button icons, images, video/audio clips, downloads, data compilations and software, is the property of or is licensed to Frontiers Media SA ("Frontiers") or its licensees and/or subcontractors. The copyright in the text of individual articles is the property of their respective authors, subject to a license granted to Frontiers.*

*The compilation of articles constituting this e-book, wherever published, as well as the compilation of all other content on this site, is the exclusive property of Frontiers. For the conditions for downloading and copying of e-books from Frontiers' website, please see the Terms for Website Use. If purchasing Frontiers e-books from other websites or sources, the conditions of the website concerned apply.*

*Images and graphics not forming part of user-contributed materials may not be downloaded or copied without permission.*

*Individual articles may be downloaded and reproduced in accordance with the principles of the CC-BY licence subject to any copyright or other notices. They may not be re-sold as an e-book.*

*As author or other contributor you grant a CC-BY licence to others to reproduce your articles, including any graphics and third-party materials supplied by you, in accordance with the Conditions for Website Use and subject to any copyright notices which you include in connection with your articles and materials.*

> *All copyright, and all rights therein, are protected by national and international copyright laws.*

*The above represents a summary only. For the full conditions see the Conditions for Authors and the Conditions for Website Use.*

ISSN 1664-8714 ISBN 978-2-88945-394-8 DOI 10.3389/978-2-88945-394-8

# About Frontiers

Frontiers is more than just an open-access publisher of scholarly articles: it is a pioneering approach to the world of academia, radically improving the way scholarly research is managed. The grand vision of Frontiers is a world where all people have an equal opportunity to seek, share and generate knowledge. Frontiers provides immediate and permanent online open access to all its publications, but this alone is not enough to realize our grand goals.

## Frontiers Journal Series

The Frontiers Journal Series is a multi-tier and interdisciplinary set of open-access, online journals, promising a paradigm shift from the current review, selection and dissemination processes in academic publishing. All Frontiers journals are driven by researchers for researchers; therefore, they constitute a service to the scholarly community. At the same time, the Frontiers Journal Series operates on a revolutionary invention, the tiered publishing system, initially addressing specific communities of scholars, and gradually climbing up to broader public understanding, thus serving the interests of the lay society, too.

## Dedication to Quality

Each Frontiers article is a landmark of the highest quality, thanks to genuinely collaborative interactions between authors and review editors, who include some of the world's best academicians. Research must be certified by peers before entering a stream of knowledge that may eventually reach the public - and shape society; therefore, Frontiers only applies the most rigorous and unbiased reviews.

Frontiers revolutionizes research publishing by freely delivering the most outstanding research, evaluated with no bias from both the academic and social point of view. By applying the most advanced information technologies, Frontiers is catapulting scholarly publishing into a new generation.

# What are Frontiers Research Topics?

Frontiers Research Topics are very popular trademarks of the Frontiers Journals Series: they are collections of at least ten articles, all centered on a particular subject. With their unique mix of varied contributions from Original Research to Review Articles, Frontiers Research Topics unify the most influential researchers, the latest key findings and historical advances in a hot research area! Find out more on how to host your own Frontiers Research Topic or contribute to one as an author by contacting the Frontiers Editorial Office: researchtopics@frontiersin.org

# **GENETICS OF ACQUIRED ANTIMICROBIAL RESISTANCE IN ANIMAL AND ZOONOTIC PATHOGENS**

Topic Editors:

**Axel Cloeckaert,** ISP, INRA, Université François Rabelais de Tours, UMR 1282, Nouzilly, France **Michel Stanislas Zygmunt,** ISP, INRA, Université François Rabelais de Tours, UMR 1282, Nouzilly, France **Benoît Doublet,** ISP, INRA, Université François Rabelais de Tours, UMR 1282, Nouzilly, France

Antimicrobial susceptibility testing by the disk diffusion method of a b-lactam resistant *E. coli* isolate.

Image: Axel Cloeckaert, INRA.

Development and spread of antimicrobial resistance is the result of an evolutionary process by which microorganisms adapt to antibiotics through several mechanisms including alteration of drug target by mutation and horizontal transfer of resistance genes. The concomitant occurrence of independent antimicrobial resistance mechanisms is a serious threat to human health and has appeared in several emerging epidemic clones over the past decade in humans and also in animals. The increasing prevalence of antimicrobial drug resistance among animal and zoonotic foodborne pathogens is of particular concern for public health. In this Ebook, we gathered a collection of articles which deal with the most important aspects of the genetics of acquired antimicrobial resistance extending from medically-important resistance, emerging epidemic resistant clones, main mobile genetic elements spreading resistance, resistomes, dissemination between animals and humans, to the "One Health" concept.

**Citation:** Cloeckaert, A., Zygmunt, M. S., Doublet, B., eds. (2018). Genetics of Acquired Antimicrobial Resistance in Animal and Zoonotic Pathogens. Lausanne: Frontiers Media. doi: 10.3389/978-2-88945-394-8

# Table of Contents

*07 Editorial: Genetics of Acquired Antimicrobial Resistance in Animal and Zoonotic Pathogens*

Axel Cloeckaert, Michel S. Zygmunt and Benoît Doublet

# **Section 1. The global spread of emerging antibiotic-resistant bacteria in the "One Health" perspective**

*10 Addressing the Antibiotic Resistance Problem with Probiotics: Reducing the Risk of Its Double-Edged Sword Effect*

Ivan C. V. J. Imperial and Joyce A. Ibana

# **Section 2. Antimicrobial resistance in zoonotic pathogens from food-producing animals**

*20 Characterization of Integrons and Resistance Genes in* **Salmonella** *Isolates from Farm Animals in Shandong Province, China*

Xiaonan Zhao, Jie Yang, Baozhen Zhang, Shuhong Sun and Weishan Chang


Xianhui Huang, Linfeng Yu, Xiaojie Chen, Chanping Zhi, Xu Yao, Yiyun Liu, Shengjun Wu, Zewen Guo, Linxian Yi, Zhenling Zeng and Jian-Hua Liu

# **Section 3. Antimicrobial resistance in companion animal pathogens**

*44 Occurrence of OXA-48 Carbapenemase and Other b-Lactamase Genes in ESBL-Producing Multidrug Resistant* **Escherichia coli** *from Dogs and Cats in the United States, 2009–2013*

Xiaoqiang Liu, Kamoltip Thungrat and Dawn M. Boothe


Fengru Deng, Huiwen Wang, Yifei Liao, Jun Li, Andrea T. Feßler, Geovana B. Michael, Stefan Schwarz and Yang Wang

# **Section 4. Diversity of chromosomal mutation-mediated antimicrobial resistances**


Su Xu, Zhuyingjie Fu, Ying Zhou, Yang Liu, Xiaogang Xu and Minggui Wang

*94 Antimicrobial Susceptibility Profiles of Human* **Campylobacter jejuni** *Isolates and Association with Phylogenetic Lineages*

Wonhee Cha, Rebekah Mosci, Samantha L. Wengert, Pallavi Singh, Duane W. Newton, Hossein Salimnia, Paul Lephart, Walid Khalife, Linda S. Mansfield, James T. Rudrik and Shannon D. Manning

*106 Characteristics of Quinolone Resistance in* **Escherichia coli** *Isolates from Humans, Animals, and the Environment in the Czech Republic* Magdalena Röderova, Dana Halova, Ivo Papousek, Monika Dolejska, Martina Masarikova, Vojtech Hanulik, Vendula Pudova, Petr Broz, Miroslava Htoutou-Sedlakova, Pavel Sauer, Jan Bardon, Alois Cizek, Milan Kolar and Ivan Literak

*118 Characteristics of Quinolone Resistance in* **Salmonella** *spp. Isolates from the Food Chain in Brazil*

Bruno R. Pribul, Marcia L. Festivo, Marcelle S. Rodrigues, Renata G. Costa, Elizabeth C. dos P. Rodrigues, Miliane M. S. de Souza and Dalia dos P. Rodrigues

# **Section 5. Conjugative plasmids and integrative elements in multidrug resistance in animal and human pathogens**


*145 Prevalence and Molecular Characteristics of Extended-Spectrum b-Lactamase Genes in* **Escherichia coli** *Isolated from Diarrheic Patients in China* Li Bai, Lili Wang, Xiaorong Yang, Juan Wang, Xin Gan, Wei Wang, Jin Xu, Qian Chen, Ruiting Lan, Séamus Fanning and Fengqin Li


Janine T. Bossé, Yanwen Li, Jon Rogers, Roberto Fernandez Crespo, Yinghui Li, Roy R. Chaudhuri, Matthew T. G. Holden, Duncan J. Maskell, Alexander W. Tucker, Brendan W. Wren, Andrew N. Rycroft and Paul R. Langford on behalf of the BRaDP1T Consortium

# *167 Single-Molecule Sequencing (PacBio) of the* **Staphylococcus capitis** *NRCS-A Clone Reveals the Basis of Multidrug Resistance and Adaptation to the Neonatal Intensive Care Unit Environment*

Patrícia Martins Simões, Hajar Lemriss, Yann Dumont, Sanâa Lemriss, Jean-Philippe Rasigade, Sophie Assant-Trouillet, Azeddine Ibrahimi, Saâd El Kabbaj, Marine Butin and Frédéric Laurent

# **Section 6. Resistomes of humans and farm animals in frame of the "One Health" concept**

# *178 Integron Digestive Carriage in Human and Cattle: A "One Health" Cultivation-Independent Approach*

Delphine Chainier, Olivier Barraud, Geoffrey Masson, Elodie Couve-Deacon, Bruno François, Claude-Yves Couquet and Marie-Cécile Ploy

*185 The Resistome of Farmed Fish Feces Contributes to the Enrichment of Antibiotic Resistance Genes in Sediments below Baltic Sea Fish Farms* Windi I. Muziasari, Leena K. Pitkänen, Henning Sørum, Robert D. Stedtfeld, James M. Tiedje and Marko Virt

# Editorial: Genetics of Acquired Antimicrobial Resistance in Animal and Zoonotic Pathogens

Axel Cloeckaert, Michel S. Zygmunt and Benoît Doublet\*

ISP, Institut National de la Recherche Agronomique, Université François Rabelais de Tours, UMR 1282, Nouzilly, France

Keywords: one health, integron, mobile genetic element, antibiotic resistance, dissemination

**Editorial on the Research Topic**

### **Genetics of Acquired Antimicrobial Resistance in Animal and Zoonotic Pathogens**

Antimicrobial resistance has become a global public health concern due to multidrug-resistant (MDR) bacteria and to the lack of novel antibiotics. Resistant bacteria, including zoonotic pathogens, can be exchanged between animals and humans through direct contact, the food chain, or contamination of the shared environment. Resistance to medically-important antibiotics such as extended spectrum beta-lactams, carbapenems, fluoroquinolones, or aminoglycosides, is of increasing magnitude among zoonotic pathogens. Acquired antimicrobial resistance is the result of an evolutionary process by which microorganisms adapt to antibiotics through several mechanisms including alteration of drug target by mutations and horizontal transfer of novel/foreign genes, referred to as resistance genes. Acquired resistance genes coding for any of the three major resistance mechanisms, i.e., enzymatic inactivation, reduced intracellular accumulation, or modification of the cellular target sites are associated with mobile genetic elements including plasmids, transposons, gene cassettes, integrative, and conjugative elements or other mobile elements (Schwarz et al., 2017).

This Research Topic is focused on acquired antimicrobial resistance mechanisms in animal and zoonotic pathogens isolated from food-producing animals, food products, companion animals, humans, and the environment. Different sets of articles document the most important aspects of the genetics of acquired antimicrobial resistance extending from medically-important resistance genes such as those conferring resistance to extended spectrum cephalosporins and more recently those conferring resistance to colistin, emerging epidemic clones, primary mobile genetic elements such as class 1 integrons, culture-independent approach of resistomes, dissemination between animals and humans, to the "One Health" concept.

First, in an elegant review article, Imperial and Ibana addressed the global problem of the spread of emerging antibiotic-resistant bacteria in the "One Health" perspective. They summarized the processes that govern the spread of antibiotic resistance in relation to resistance genes, their horizontal transfer through mobile genetic elements, the microbial ecology of resistant bacteria in the human gut microbiota, and their dissemination by international travel of humans, animals, or food. Finally, they discussed the probiotic use in both human and veterinary applications to tackle the antibiotic resistance threat and concluded on the "double-edged sword" with potential risk in propagating antibiotic resistance by probiotics.

Salmonella enterica spp. are important zoonotic pathogens related to foodborne diseases worldwide. In the present Research Topic, Zhao et al. and Zhao et al. contributed with two research articles dealing with antimicrobial resistance of S. enterica serotypes isolated from food-producing animals (chickens, ducks, and pigs) in farms and slaughterhouses in the Shandong province, China.

#### Edited by:

Daniela Ceccarelli, Wageningen Bioveterinary Research (WBVR), Netherlands

Reviewed by: Catherine M. Logue, University of Georgia, United States

> \*Correspondence: Benoît Doublet benoit.doublet@inra.fr

#### Specialty section:

This article was submitted to Antimicrobials, Resistance and Chemotherapy, a section of the journal Frontiers in Microbiology

Received: 25 October 2017 Accepted: 23 November 2017 Published: 05 December 2017

#### Citation:

Cloeckaert A, Zygmunt MS and Doublet B (2017) Editorial: Genetics of Acquired Antimicrobial Resistance in Animal and Zoonotic Pathogens. Front. Microbiol. 8:2428. doi: 10.3389/fmicb.2017.02428 They reported a higher prevalence of Salmonella in chickens (∼24%) compared to pigs (∼9%). The serotype distribution and multidrug resistance phenotypes were dependent on the animal species, serotypes Indiana, and Enteritidis being prevalent in chickens and strongly associated with MDR phenotypes. High antimicrobial resistance rates were observed for old class antibiotics such as nalidixic acid, ampicillin, and tetracyclines (>80%). The plasmid-mediated quinolone resistance genes qnr were also frequently detected in isolates from farm animals and in slaughterhouses contributing probably with target gene mutations to a rate of around 40% of ciprofloxacin resistance. A relatively high resistance rate to cefotaxime (∼30%) was also described, mostly Salmonella isolates harboring the ESBL gene blaCTX-M-55, largely predominant in Asia. In another study from China, Huang et al. investigated the prevalence of colistin resistance and associated mcr-1 gene on a large collection of commensal E. coli isolates (>4,000) from food-producing animals during the period 2013–2014. They reported an overall resistance rate of 18.7% with MIC colistin ≥ 4 mg/L, with a higher frequency of colistin-resistant E. coli isolated from pigs (24%) compared to chickens (14%). Among 200 randomly selected colistin-resistant E. coli isolates, they found 182 positive mcr-1 positive isolates. Colistin being in some cases the last therapeutic option to treat infections due to carbapenemaseproducing Gram-negative bacteria, the high prevalence of MCR-1-mediated colistin resistance among commensal E. coli recovered from food-producing animals is worrisome in China. These MCR-1-producing Enterobacteriaceae might transfer to humans through the food chain or to farmers via direct animal contact, and furthermore transfer the mcr-1 resistance gene to human pathogens. Since the recent ban of colistin as animal feed additive in China, future surveillance programmes will be necessary to follow the evolution of colistin resistance in foodproducing animals (Walsh and Wu, 2016).

This Research Topic includes also a panel of articles focused on resistance genes in different companion animal pathogens. Resistance to extended spectrum cephalosporins is a major concern for companion animals due to the close contact with their owners. Liu et al. investigated the occurrence of Extended Spectrum Beta Lactamase (ESBL)-producing clinical E. coli recovered from dogs and cats in the United States, from 2009 to 2013. They reported a prevalence of 3.8% ESBL-producing E. coli, the blaCTX-<sup>M</sup> resistance genes being the most prevalent ones followed by the cephalosporinase and carbapenemase genes, blaCMY-<sup>2</sup> and blaOXA-48, respectively. CTX-M-producing MDR E. coli isolates were primarily of sequence types ST131, ST648, and ST405 suggesting the possible transfer of predominant human clones with companion animals in the U.S. In a similar large epidemiological study dealing with Morganellaceae of animal origin in France, Schultz et al. investigated the prevalence of ESBL and carbapenemase genes and their genetic supports. They found a similar prevalence of ESBL producers (∼4%) among Proteus mirabilis isolates from dogs, cats, and horses. Interestingly, molecular characterization showed the spread of a clonal population harboring the MDR integrative mobilizable element SGI1-V (Salmonella genomic island 1—variant V) carrying the ESBL gene blaVEB-6, previously identified in human isolates in France. Xia et al.reported a study focused on 16S rRNA methylases conferring high-level resistance to aminoglycosides in Klebsiella pneumoniae isolates from diseased dogs and cats in a veterinary hospital in China. They highlighted the spread of genetically-related K. pneumoniae ST37 isolates as well as the horizontal transfer of a conjugative IncF33:A-:B- plasmid carrying the 16S rRNA methylase gene rmtB and the ESBL gene blaCTX-M-55. Finally, Deng et al. described the multiresistance genes vga(A), vga(A)LC, sal(A), and lsa(E) carried by plasmids and chromosomal genomic islands and conferring resistance to pleuromutilins, lincosamides, and streptogramin A in different Staphylococci species isolated from pet animals and sometimes their owners in China. In regard of all these articles, it is important to consider the current role of companion animals as a reservoir of resistant bacteria and their mobile resistance determinants that may be exchanged in either direction between animals and humans, warranting the prudent use of all antibiotics in veterinary medicine of companion animals.

Acquired resistance mechanisms are based on resistance mutations of a chromosomal gene or on the acquisition of mobile resistance genes. Several articles exemplified both phenomena of molecular resistance mechanisms, whose relative importance depends on the antibiotic family and the implicated bacterial species. With the recent description of plasmidmediated resistance to colistin (mcr genes), a renewed interest of research took place for molecular resistance mechanisms to polymixins. Prasgasam et al. characterized carbapenemand colistin-resistant K. pneumoniae isolates from bacteremia cases in India by whole genome sequencing. They described multiple mutations in the chromosomal genes coding for lipopolysaccharide (LPS) lipid A modifications including silent mutations, point mutations, insertions and/or deletions. The most significant were mutations in the mgrB gene. The resistance genes mcr were not detected in this study. The significance of other mutations observed in this study needs to be confirmed for conferring polymixin resistance. Another example of resistanceconferring mutations in chromosomal genes is given in the study by Xu et al. dealing with fosfomycin resistance in Staphylococcus aureus. The key resistance mechanisms to fosfomycin include the production of fosfomycin-modifying enzymes (FosA, FosB, FosC, and FosX), modifications of the target enzyme MurA or the membrane transporters GlpT and UhpT. The GlpT and UhpT transporters are responsible for fosfomycin uptake. They previously showed that mutations in glpT and uhpT genes were common in fosfomycin- and methicillin-resistant S. aureus (Fu et al., 2016). Using uhpT and/or glpT deletion mutants in S. aureus, they confirmed the role of mutations or insertional inactivation in these transporter or their regulatory genes in highlevel fosfomycin resistance (MIC > 1,024 µg/ml). Cha et al. investigated the phylogenetic lineages of fluoroquinolone- and macrolide-resistant Campylobacter jejuni isolates recovered from Michigan patients. Resistance mechnisms to fluoroquinolones and macrolides involve acquisition of mutations of the target sites of the antibiotics, i.e., the DNA gyrase/topoisomerase IV and 23S rRNA subunit, respectively. They identified clonal spread of specific fluoroquinolone-resistant C. jejuni lineages, like ST464, associated with history of foreign travel and also a significant association between tetracycline-resistant C. jejuni ST982 and contact with cattle, chickens and drinking well water at home. Linking genetic diversity and antimicrobial resistance profiles of C. jejuni from various sources is needed to better understand transmission dynamics to humans.

Acquired resistance genes are not by themselves mobile but are carried by various genetic structures allowing their horizontal mobility at the molecular or cellular level. The spread of successful multidrug resistance (MDR) mobile genetic elements (plasmids and genomic islands) between bacteria is the main driving force in the dissemination of acquired antibiotic resistance genes. The involvement of plasmids in the zoonotic spread of ESBL- and carbapenemase-resistance genes between human and animal reservoirs are documented by several articles in the present Research Topic. Seiffert et al. described the occurrence of a new plasmid variant IncK2 carrying the cephalosporinase gene blaCMY-<sup>2</sup> in E. coli isolated from poultry, poultry meat and humans. The identification of very close CMY-2-encoding IncK2 plasmids in genetically diverse E. coli from poultry and humans suggested an important ability of horizontal transfer by conjugation in different reservoirs. In another article also dealing with ESBL-producing Enterobacteriaceae, Moremi et al. analysed ESBL producers in wild fish obtained from Lake Victoria as well as in environmental samples obtained from the Mwanza city in Tanzania. Lake Victoria is the major source of fish consumed by Mwanza residents and also receives treated wastewater effluents from Mwanza city. They previously described high rates of CTX-M15-producing Enterobacteriaceae in the city hospital as well as in animals and humans from the community (Mshana et al., 2016). In the present study, they reported a significant proportion of ESBL-producing Enterobacteriaceae in fish gut and environmental samples involving both clonal spread of resistant strains and dissemination of CTX-M15 encoding IncY plasmids. This study suggests that transmission of both ESBL-producing clones and plasmids may occur between humans and wild fish, and reciprocally, through environmental contamination by anthropogenic activity and the food chain, respectively. An additional article by Bai et al. highlights the diversity of ESBLcarrying plasmids, blaCTX-<sup>M</sup> genes and ESBL-producing E. coli isolated from diarrheic patients in China. Thanks to whole genome sequencing, additional articles emphasize the important role and diversity of integrative elements implicated in multidrug

# REFERENCES


resistance in animal and human pathogens, two articles from Bossé et al. and Bossé et al. and one from Simões et al., respectively.

Finally, two research articles investigated the resistomes of humans, cattle and fish in different environmental settings, by using a culture-independent approach on targeted resistance genes and/or associated mobile elements like integrons. Chainier et al. reported a high frequency of integron carriage in the gut of cattle and humans living in the same geographic area, in France and conversely Muziasari et al. described a low occurrence of resistance genes in the gut of farmed fish and in associated sediment samples in fish farms in the Northern Baltic Sea, in Finland. Using high throughput culture independent methods to study resistomes in different microbiota will provide useful information in the future to manage the spread of antimicrobial resistance.

In summary, this Research Topic addresses various issues related to the genetics of antimicrobial resistance in frame of the "One Health" concept. They emphasize the huge diversity of resistance genes, mobile genetic elements, and multidrugresistant clones in different microbial environments. In the era of increasing antimicrobial resistance, judicious use of antibiotics is an absolute necessity in veterinary and human medicine to prolong antibiotic usefulness. On the other hand, bacteria have developed a large set of genetic weapons specialized in the acquisition and spread of acquired resistance genes. Successful MDR mobile genetic elements are one of the main driving forces in the antibiotic resistance burden. A research effort is needed to decipher the molecular basis of resistance dissemination within and between microbial niches to manage the evolution toward resistance.

# AUTHOR CONTRIBUTIONS

All authors listed have made a substantial, direct and intellectual contribution to the work, and approved it for publication.

# ACKNOWLEDGMENTS

We warmly thank Dr. Daniela Ceccarelli as associate editor and all reviewers of this Research Topic. This work was supported by public funds from the French National Institute of Agricultural Research.

Walsh, T. R., and Wu, Y. (2016). China bans colistin as a feed additive for animals. Lancet Infect. Dis. 16, 1102–1103. doi: 10.1016/S1473-3099(16)30329-2

**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Cloeckaert, Zygmunt and Doublet. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Addressing the Antibiotic Resistance Problem with Probiotics: Reducing the Risk of Its Double-Edged Sword Effect

### Ivan C. V. J. Imperial and Joyce A. Ibana\*

Immunopharmacology Research Laboratory, Institute of Biology, College of Science, University of the Philippines Diliman, Quezon City, Philippines

Antibiotic resistance is a global public health problem that requires our attention. Indiscriminate antibiotic use is a major contributor in the introduction of selective pressures in our natural environments that have significantly contributed in the rapid emergence of antibiotic-resistant microbial strains. The use of probiotics in lieu of antibiotic therapy to address certain health conditions in both animals and humans may alleviate these antibiotic-mediated selective pressures. Probiotic use is defined as the actual application of live beneficial microbes to obtain a desired outcome by preventing diseased state or improving general health. Multiple studies have confirmed the beneficial effects of probiotic use in the health of both livestock and humans. As such, probiotics consumption is gaining popularity worldwide. However, concerns have been raised in the use of some probiotics strains that carry antibiotic resistance genes themselves, as they have the potential to pass the antibiotic resistance genes to pathogenic bacteria through horizontal gene transfer. Therefore, with the current public health concern on antibiotic resistance globally, in this review, we underscore the need to screen probiotic strains that are used in both livestock and human applications to assure their safety and mitigate their potential in significantly contributing to the spread of antibiotic resistance genes in our natural environments.

### Edited by:

Axel Cloeckaert, French National Institute for Agricultural Research (INRA), France

### Reviewed by:

Atte Von Wright, University of Eastern Finland, Finland Francisco Dionisio, University of Lisbon, Portugal

> \*Correspondence: Joyce A. Ibana jaibana@up.edu.ph

### Specialty section:

This article was submitted to Antimicrobials, Resistance, and Chemotherapy, a section of the journal Frontiers in Microbiology

Received: 25 September 2016 Accepted: 28 November 2016 Published: 15 December 2016

### Citation:

Imperial ICVJ and Ibana JA (2016) Addressing the Antibiotic Resistance Problem with Probiotics: Reducing the Risk of Its Double-Edged Sword Effect. Front. Microbiol. 7:1983. doi: 10.3389/fmicb.2016.01983 Keywords: antibiotic resistance, probiotics, mobile genetic elements, veterinary medicine, livestock production

# INTRODUCTION

Since its advent, antibiotics remain as the major therapeutic strategy that is used to address numerous diseases of infectious etiologies both in human and veterinary medicine (Drago et al., 2011; Schjørring and Krogfelt, 2011; Verraes et al., 2013; Allen and Stanton, 2014; Card et al., 2015). However, the indiscriminate and improper use of antibiotics has led to the decreased susceptibility and increased resistance rates observed not only in disease-causing microbes but in commensal microbes as well (Rosander et al., 2008; Drago et al., 2011; Allen and Stanton, 2014; von Wintersdorff et al., 2014; Card et al., 2015). Rampant antibiotic use has pushed microbes to adapt and survive by acquiring antibiotic resistance genes that led to antibiotic-resistant strains (Schjørring and Krogfelt, 2011; Forslund et al., 2013; Fouhy et al., 2013; Ghosh et al., 2013; Verraes et al., 2013; Allen and Stanton, 2014; Card et al., 2014). Antibiotic resistance genes are then vertically passed on to the next generation of microbes; in some cases, they are acquired through

horizontal transfer from one microbe to another, when thriving in the same microbial environment (Schjørring and Krogfelt, 2011; Broaders et al., 2013; Forslund et al., 2013; Fouhy et al., 2013; Ghosh et al., 2013; Verraes et al., 2013; Allen and Stanton, 2014; Card et al., 2014, 2015).

In the human clinical setting, these antibiotic-resistant pathogens have caused numerous treatment failures that eventually led to both hospital morbidities and mortalities (Vankerckhoven et al., 2008; Schjørring and Krogfelt, 2011; Lu et al., 2014). Overall, the prevalence of antibiotic resistance has now become a global health problem that needs urgent attention from the world health authorities (Egervärn et al., 2010; Schjørring and Krogfelt, 2011; Ghosh et al., 2013; Penders et al., 2013; Hu et al., 2014; Lu et al., 2014; von Wintersdorff et al., 2014; Card et al., 2015; van Schaik, 2015). In addressing the problem on antibiotic resistance, the use of probiotics in lieu of antibiotics for treating certain diseases of host organisms has been investigated (Rosander et al., 2008; Muñoz-Atienza et al., 2013). Numerous studies have shown that instead of killing pathogenic microbes through antibiotics, the establishment of commensal and sometimes mutualistic microbes may hinder the growth of disease-causing microbes found in the same host microbial environment (Saarela et al., 2007; Hammad and Shimamoto, 2010; Klein, 2011; Nueno-Palop and Narbad, 2011; Wei et al., 2012; Varankovich et al., 2015). In addition, it has also been demonstrated that maintaining what is considered "normal" microbiota for certain host microbial environments may prevent diseased conditions that are not necessarily of infectious etiology and may improve general health outcome (Franz et al., 2011; Nueno-Palop and Narbad, 2011; Wei et al., 2012; Téllez et al., 2015; Varankovich et al., 2015). As a result, probiotic use, defined as the application of actual live beneficial microbes, has been increasingly practiced for both human and veterinary applications (Tompkins et al., 2008; Vankerckhoven et al., 2008; Sanders et al., 2010; Xiao et al., 2010; Nueno-Palop and Narbad, 2011; Songisepp et al., 2012; Devi et al., 2015; D'Orazio et al., 2015; Fuochi et al., 2015; Senan et al., 2015; Varankovich et al., 2015). Among the modes of probiotic use, the consumption of probiotics through the gastrointestinal route may be considered the most common application in both human and veterinary uses.

However, microbes used as probiotics are not exempted from acquiring antibiotic resistance genes. Given their shared microbial environment in the gastrointestinal tract, a risk of pathogenic microbes acquiring antibiotic resistance genes from probiotic microbes exists, and vice versa (Mater et al., 2007; Rosander et al., 2008; Liu et al., 2009; Egervärn et al., 2010; Drago et al., 2011; Nueno-Palop and Narbad, 2011; Gueimonde et al., 2013; Varankovich et al., 2015). If improperly cooked, livestock treated with probiotics that are consumed by humans as food may also pose as a possible source of antibiotic resistance genes for the human gut microbiota (Devirgiliis et al., 2011; Schjørring and Krogfelt, 2011; Forslund et al., 2013; Verraes et al., 2013; Allen and Stanton, 2014; Hu et al., 2014; Woolhouse et al., 2015). To complicate the aforementioned risks, some probiotic microbes are even screened specifically for antibiotic resistance to be used concomitantly with antibiotics in treating certain medical conditions (Galopin et al., 2009; Hammad and Shimamoto, 2010). As such, there is a need to review existing studies to clarify the safety of increasing probiotic use in relation to the existence of antibiotic resistance genes.

This review aims to describe the processes that govern the spread of antibiotic resistance in relation to antibiotic resistance genes. Antibiotic resistance gene transfer in the absence of probiotics is discussed first to elucidate the ongoing problem of the prevalence of antibiotic-resistant bacterial strains. Probiotic uses in both human and veterinary applications are then described and reviewed to reaffirm their beneficial use. Screening of probiotic bacterial strains for antibiotic resistance genes is then discussed to evaluate the safety of probiotic use. Finally, probiotic use in relation to the spread of antibiotic resistance genes is tackled to clarify the potential role of probiotics in propagating antibiotic resistance.

# ANTIBIOTIC RESISTANCE

Although the remarkable increase in the incidence and prevalence of antibiotic resistance were observed after the introduction and widespread use of antibiotics (Datta and Hughes, 1983; Hughes and Datta, 1983), antibiotic resistance is believed to have existed long before human antibiotic use (Hughes and Datta, 1983; Broaders et al., 2013). It is evident in multiple ecological interactions, wherein many organisms, may they be microbes or macro-organisms, have the ability to produce natural antibiotics that ultimately increase their chances of survival (Samuels et al., 2013; Cawoy et al., 2014; Timbermont et al., 2014; Pinchas et al., 2015; Sherpa et al., 2015; Téllez et al., 2015). Organisms use antibiotics to kill or inhibit growth of pathogenic microbes, while some microbes use antibiotics to compete for the same resources that other microbes consume as well (Samuels et al., 2013; Timbermont et al., 2014; Pinchas et al., 2015; Sherpa et al., 2015; Téllez et al., 2015). As a natural evolutionary response, microbes that are able to adapt to and survive these natural antibiotics gain advantage in producing the next generation of microbes. Hence, antibiotic resistance is a natural phenomenon (Sherpa et al., 2015).

Taking advantage of these ecological interactions, we have learned to harness the use of antibiotics for a variety of applications (Scanlon et al., 2014; Sherpa et al., 2015). We have used antibiotics to treat infectious diseases that were once considered very fatal, until the advent of readily available antibiotic medications (Sherpa et al., 2015). In agriculture, we have learned that the use of antibiotics greatly increases yield in rearing animals primarily as food source (Allen and Stanton, 2014; Woolhouse et al., 2015; Xiao et al., 2015). Although farmers were not entirely clear as to how antibiotic use increases growth rates of livestock animals, they have continually used antibiotics in sub-therapeutic doses as feeding supplements with observable results (Allen and Stanton, 2014). In addition, infectious diseases that once plagued farm animals are now treated successfully by antibiotics and even prevented by prophylactic use (Allen and Stanton, 2014; Woolhouse et al., 2015). As such, humans have multiple reasons to use antibiotics.

However, with rampant antibiotic use, the natural way of developing antibiotic resistance has also changed in both manner and rate. Due to different human applications, multiple avenues for microbes to encounter antibiotics have put immense selective pressure on microbes to develop antibiotic resistance (Rosander et al., 2008; Devirgiliis et al., 2011; Drago et al., 2011; Schjørring and Krogfelt, 2011; Forslund et al., 2013; Fouhy et al., 2013; Ghosh et al., 2013; Verraes et al., 2013; Allen and Stanton, 2014; Hu et al., 2014; von Wintersdorff et al., 2014; Card et al., 2015). In turn, these have hastened the development of antibiotic-resistant microbial strains. The rate of developing antibiotic-resistant strains is now occurring much faster than the rate of discovering new antibiotics (Sherpa et al., 2015).

# Antibiotic Resistance Genes and the Influence of Ecology

Just like any evolutionary response, adaptive changes of an organism are passed on to the next generation via the organism's genetic material. Some microbes possess antibiotic resistance genes that confer their ability to survive exposure to antibiotics (Rosander et al., 2008; Drago et al., 2011; Schjørring and Krogfelt, 2011; Forslund et al., 2013; Fouhy et al., 2013; Ghosh et al., 2013; Verraes et al., 2013; Allen and Stanton, 2014; Hu et al., 2014; von Wintersdorff et al., 2014; Card et al., 2015). According to the Antibiotic Resistance Gene Database, there are at least 23,317 antibiotic genes established so far, which are effective against at least 249 known antibiotics (Hu et al., 2013). In several recent studies, investigators have shown that the presence of antibiotic resistance genes in microbial organisms is greatly affected by their microbial ecology (Forslund et al., 2013; Hu et al., 2013, 2014; Gibson et al., 2015). This observation reiterates the natural selection pressure toward antibiotic resistance brought about by ecological interactions found in the natural microbial environment (Sherpa et al., 2015).

In the study of Hu et al. (2013), investigators studied the presence of antibiotic resistance genes in the human gut microbiota. A total of 162 individuals from China, Denmark, and Spain were screened for antibiotic resistance genes. It was found that a total of 1,093 antibiotic resistance genes were present in the sample population. Antibiotic resistance genes against three antibiotic classes – tetracycline, macrolides, and beta-lactams – accounted for more than 75% of the total antibiotic resistance genes. Interestingly, antibiotic resistance genes found in Chinese individuals were statistically clustered in terms of similarity, when compared to the two European populations that had statistically more similar antibiotic resistance profiles. This suggests that the actions of human hosts as a population also affect the microbial ecology and, ultimately, the antibiotic resistance gene profiles of the gut microbiota. In support of the previous statement, a study of Lu et al. (2014) demonstrated that human populations of different age groups also produced different antibiotic resistance genes profiles that were statistically clustered by age. Hence, different activities of different human age groups also dictate antibiotic resistance profiles of human gut microbiota.

In addition, Forslund et al. (2013) showed that countryspecific antibiotic use influences the antibiotic resistance genes found in the human gut microbiota. In addition to actual medical application of antibiotics, the investigators also considered the antibiotics used in livestock agriculture for food production. Investigators found that veterinary use of antibiotics also influenced antibiotic resistance genes profile of the human gut microbiota. This reaffirms the role of ecological interactions among humans, animals, and the microbial environment in influencing antibiotic resistance genes, ultimately found in the human gut microbiota (Schjørring and Krogfelt, 2011; Hu et al., 2014). As such, many studies recommend that in addressing the problem on antibiotic resistance, an ecological approach and perspective are needed as evidence suggests that antibiotic resistance is not entirely confined to human medical antibiotic use alone (Devirgiliis et al., 2011; Schjørring and Krogfelt, 2011; Forslund et al., 2013; Verraes et al., 2013; Allen and Stanton, 2014; Hu et al., 2014; Woolhouse et al., 2015). Agricultural use of antibiotics should be properly regulated as well as clinical prescription of antibiotics in humans (Verraes et al., 2013; Allen and Stanton, 2014; Woolhouse et al., 2015).

# Antibiotic Resistance Genes and Their Spread through Mobile Genetic Elements

As mentioned earlier, antibiotic resistance genes are vertically transferred from one generation to another and favor the survival of resistant microbes (Schjørring and Krogfelt, 2011; Penders et al., 2013; Verraes et al., 2013; Allen and Stanton, 2014; Fouhy et al., 2014; Hu et al., 2014; Card et al., 2015; van Schaik, 2015). However, it is the ability of microbes to conduct horizontal gene transfer that somehow complements the rampant human antibiotic applications in hastening the development of antibiotic-resistant strains. Horizontal gene transfer in microbes is made possible through mobile genetic elements (Karami et al., 2007; Mater et al., 2007; De Vries et al., 2011; Drago et al., 2011; Haug et al., 2011; Broaders et al., 2013; Machado and Sommer, 2014; von Wintersdorff et al., 2014). The gastrointestinal tract is a prime candidate for conducting studies on horizontal gene transfer (Broaders et al., 2013).

In the human gastrointestinal tract, four mobile genetic elements are considered: plasmids, conjugative transposons, integrons, and bacteriophages (Broaders et al., 2013). Among mobile genetic elements, only bacterial plasmids, conjugative transposons, and integrons are considered well-established and well-documented to cause antibiotic genes transfer in both the natural environment and the clinical setting. It has been reported that antibiotic resistance genes were successfully transferred from a commensal bacterial strain to a pathogenic bacterial strain, and vice versa, using plasmids as vectors of horizontal gene transfer (Broaders et al., 2013). Using conjugative transposons, antibiotic resistance to tetracycline, chloramphenicol, kanamycin, and erythromycin are reported to be transferred in certain species of bacteria. The loss of efficacy of tetracycline against opportunistic Bacteroides spp. infection is specifically attributed to conjugative transposons (Broaders et al., 2013). Integrons, which function very similarly, like transposons, have been identified as the

cause of spread of antibiotic resistance in Vibrio cholerae (Broaders et al., 2013). On the other hand, the role of bacteriophages in promoting bacterial horizontal gene transfer is only suggested with the small but significant presence of phage-related proteins in the human gut microbial community (Broaders et al., 2013). Given the genetic-hacking machinery of viruses, bacterial horizontal gene transfer brought about by bacteriophage infection from one bacterium to another cannot be simply ruled out.

Besides mobile genetic elements, it was suggested that perhaps the most mobile that antibiotic resistance genes can be is through actual travel of human hosts carrying antibiotic-resistant strains. In the study of von Wintersdorff et al. (2014), the investigators demonstrated that international travel of 122 healthy Dutch travelers with documented fecal antibiotic resistance genes profiles caused significant increases in extended spectrum betalactamase-encoding genes and quinolone resistance-encoding genes found in their gut microbiota, immediately after their return to the Netherlands. Travel to Southeast Asia and travel to the Indian subcontinent were the most associated with increase in quinolone resistance-encoding genes alone, and increase in both beta-lactamase-encoding genes and quinolone resistanceencoding genes, respectively. In sum, these alarming findings suggest that human travel can contribute to the global spread of antibiotic resistance genes (von Wintersdorff et al., 2014).

# Current Efforts to Control Antibiotic Resistance

Given the ecological nature of the problem on antibiotic resistance, several nations have adopted policies to address the issue. Perhaps the most notable is that of the member states of the European Union (EU). Since 2006, all EU countries have prohibited the use of antibiotics for the sole purpose of growth promotion in agricultural livestock industry (Cogliani et al., 2011). In addition, the European Food Safety Authority (EFSA) has instituted guidelines on the use of food additives in animal products that may potentially spread antibiotic resistance genes (Panel, 2012). However, other developed nations, such as the United States (US), has not imposed strict regulatory policies on antibiotic use for livestock growth promotion. Animal food and pharmaceutical industries in the US have strongly opposed restrictions in antibiotic use and have argued that such policies have been detrimental to food production in places where they were implemented. Furthermore, some countries do not have any known or established policy regarding the use of antibiotics as growth promoters, nor even require veterinary prescription for animal antibiotic use (Maron et al., 2013).

With the current paucity of clear regulatory policies in many countries, a global effort is needed in order to consolidate an effective strategy in controlling antibiotic resistance. Without a unifying regulatory guideline to follow, the problem on antibiotic resistance will most likely persist given the interweaving ecological interactions that govern the spread of antibiotic resistance. Furthermore, the spread of antibiotic resistance genes through human travel or trading of animal products between nations may make instituted regulatory policies of some countries less effective.

# USE OF PROBIOTICS

As the development of antibiotic resistance strains continues, the use of probiotics as a substitute for antibiotics is becoming more popular in both the medical field and livestock agriculture (Collins and Gibson, 1999; Collado et al., 2012; Muñoz-Atienza et al., 2013; D'Orazio et al., 2015; Téllez et al., 2015; Varankovich et al., 2015). Probiotic use is defined as the actual application of live beneficial microbes to obtain a desired outcome, may it be prevention of a diseased state or improvement in general health outcome observed in the host organism (Collins and Gibson, 1999). The basis of its efficacy relies on the symbiosis of an established microbial ecology that resists the intrusion or overproduction of pathogens that lead to a diseased condition of the host organism (Catanzaro and Green, 1997). By limiting the use of antibiotics, probiotic use may help decrease the rate of development of antibiotic-resistant strains secondary to widespread and rampant antibiotic use (Collins and Gibson, 1999; Collado et al., 2012; Muñoz-Atienza et al., 2013; Varankovich et al., 2015).

# Probiotics and Livestock Benefits

Many farm animals are considered to be adapted in a symbiotic relationship with microbes as shown by specialized gastrointestinal organs that enable microbial fermentation (Téllez et al., 2015). The capacity of farm animals to ferment complex polysaccharides with the help of their intestinal microbiota gives as much as 70% of energy acquired by ruminants and up to 30% of energy acquired by monogastric animals (Téllez et al., 2015). However, it is only fairly recent that this symbiotic relationship between farm animals and their microbiota is taken advantage of by farmers with the use of probiotics. With the increasing problem on antibiotic resistance, studies have shown that probiotic use can replace antibiotics in preventing diseased conditions and promoting growth in livestock animals (Muñoz-Atienza et al., 2013; Téllez et al., 2015).

Probiotic use in livestock agriculture of chickens and turkeys has been demonstrated to confer increased resistance in Salmonella spp. infections through accelerated establishment of what is considered normal and healthy microbiota for the aforementioned birds (Téllez et al., 2015). In addition, the incidence of idiopathic diarrhea in commercial turkey brooding houses was reported to be decreased by probiotic use (Téllez et al., 2015). As a whole, large-scale commercial trials of appropriate probiotic administration in chickens and turkeys demonstrated increased performance and reduced overall costs of production (Téllez et al., 2015).

In cattle raising, it was reported that probiotic use provided no significant effects in reducing cattle pathogens (Téllez et al., 2015). However, Escherichia coli O157:H7, a food-borne pathogen capable of causing severe hemorrhagic illness in humans, was known to be shed by livestock animals in their feces. With the use of different combinations of bacteria used

as probiotics, it was shown that there was decreased shedding of E. coli O157:H7 in both cattle and sheep livestock that may translate to decreased risk of E. coli O157:H7 human infections (Téllez et al., 2015).

Even in a different habitat such as in aquaculture, probiotic use has also been gaining ground. In the study of Muñoz-Atienza et al. (2013), it was demonstrated that lactic acidproducing bacteria of aquatic origin can exhibit antimicrobial activity against established gram-positive and gram-negative fish pathogens. However, no comprehensive in vivo assessment has yet to determine the beneficial and possible harmful effects of probiotic use in aquaculture (Muñoz-Atienza et al., 2013).

# Probiotics and Human Health

For humans, probiotic use has become a popular practice for promoting good health. Many commercial food products are now being supplemented by probiotic bacteria with claims of promoting good health (Songisepp et al., 2012). It is no surprise that there is an increasing number of probiotic users as we begin to understand better the role and importance of human gut microbiota, as well as what is considered to be "normal" and symbiotic human gut microbiota (Catanzaro and Green, 1997; Collins and Gibson, 1999; Collado et al., 2012).

The complex microbiota of the human gut is considered to play important roles in several gastrointestinal functions. These functions include host nutrition, regulation of gut epithelial development, regulation of fat storage, stimulation of intestinal angiogenesis, inflammatory immune response, and pathogen resistance (Collins and Gibson, 1999; Ventura et al., 2009). In parallel comparisons with the macro-ecosystems of the natural environment, we can deduce that the more stable the composition of the human gut microbiota is, the more beneficial it is for the human host habitat. Indeed, there are expected human gut microbiota profiles that are considered to be symbiotic and are indicators of good health (Varankovich et al., 2015). This is supported by associations demonstrated between deviations in the composition of symbiotic adult pattern of human gut microbiota and a variety of human medical conditions such as inflammatory bowel disease, allergy, obesity, and atopic disease (Ventura et al., 2009).

Looking at current human medical applications, probiotic use is clinically proven to modulate infant gut microflora disturbance after antibiotic use (Collado et al., 2012). The use of antibiotics early in the infant's life is considered detrimental to infant gut microbial diversity, as antibiotics do not discriminate in killing microorganisms that are affected by the antibiotics' mechanism of action. As such, normal gut microbiota composed of commensal and beneficial microbes is decreased in the infant's gut, allowing the potential increase in the population of harmful microbes that are not affected by the antibiotics' mechanism of action. This process leads to diseased conditions associated with antibiotic use such as antibiotic-induced diarrhea (Collins and Gibson, 1999; Collado et al., 2012; Varankovich et al., 2015). In addition, perinatal and postnatal probiotic use are reported to have potential benefits in preventing future developments of allergies (Kukkonen et al., 2007), asthma (Luoto et al., 2010), gastrointestinal infections (Johnston et al., 2007), and obesity (Vliagoftis et al., 2008), as the aforementioned diseased conditions are all associated with the interplay of gut microbiota and the development of the host immune system (Collado et al., 2012). By disturbing the infant's gut microbiota, antibiotics pose a potential risk in pre-disposing infants to the aforementioned diseased conditions.

Unlike in infants, human adults with consistent oral intake of established probiotics have not yet shown clinically significant changes in adult gut microbiota composition, structure, and gene content (Ursell et al., 2013). However, for a specific medical condition, such as Clostridium difficile-associated diarrhea, studies have shown that probiotic use, sometimes in the extreme form of fecal transplantation, effectively addresses recurrent C. difficile-associated diarrhea (Ursell et al., 2013; Van den Abbeele et al., 2013; Varankovich et al., 2015). C. difficileassociated diarrhea is a disease associated with prolonged antibiotic use which kills normal human gut microbiota that supposedly hinders the growth of the C. difficile population (Van den Abbeele et al., 2013; Varankovich et al., 2015).

There are other human medical conditions now being connected to the human gut microbiota. Interestingly, these medical conditions associated with human gut microbiota are not confined to diseases with infectious etiologies (Ventura et al., 2009). Common medical conditions with established non-microbial pathophysiologies, such as obesity and diabetes mellitus, and much rarer gastrointestinal disorders, such as irritable bowel syndrome, Crohn's disease, and necrotizing enterocolitis, are presently being associated with dysbiosis in the human gut microbiota (Peterson et al., 2008; Emami et al., 2009; Serino et al., 2009; Salonen et al., 2010; Angelakis et al., 2012; Snedeker and Hay, 2012; Jeffery and O'Toole, 2013). Deviation from what is accepted to be "normal" human gut microbial ecology appears to be part of the aforementioned diseases' pathophysiologies. Therefore, more medical applications of probiotic use to maintain normal human gut microbial ecology are anticipated to appear in the near future.

# SAFETY OF PROBIOTIC USE

Microbes used as probiotics are not exempted from the natural processes governing antibiotic resistance (Mater et al., 2007; Rosander et al., 2008; Chang et al., 2009; Egervärn et al., 2010; Drago et al., 2011; Nueno-Palop and Narbad, 2011; Gueimonde et al., 2013; Varankovich et al., 2015).

As such, it is imperative to screen microbes effectively for antibiotic resistance genes before using them as probiotics. A crucial aspect in studying antibiotic resistance in probiotic bacteria is to separate intrinsic resistance from acquired resistance (Sanders et al., 2010). Mechanisms of intrinsic resistance, such as active efflux of antibiotics by a bacterial outer membrane, is not governed by acquired antibiotic resistance genes (Chang et al., 2009; Hammad and Shimamoto, 2010). Focusing on acquired antibiotic resistance, random genetic changes on chromosomal genes should be further distinguished from the more likely transmissible type of antibiotic resistance.

However, to our knowledge, no unified world-wide health authority has taken full responsibility in screening for antibiotic resistance genes in probiotic microorganisms. Fortunately, research projects, such as the Biosafety Assessment of Probiotics used for Human Consumption (PROSAFE), the Assessment and Critical Evaluation of Antibiotic Resistance Transferability in Food Chain (ACE-ART), and the Joint International Organization for Standardization-International Dairy Federation Action Team on Probiotics (ISO-IDF) have individually contributed to address the issue (Sanders et al., 2010). In evaluating the safety of potential probiotic strains, two statuses are currently acceptable – Qualified Presumption of Safety (QPS) by the EFSA and Generally Recognized as Safe (GRAS) by the US-FDA. It is noteworthy that GRAS status is applied to microorganisms and microbial-derived ingredients used in food products while QPS is applied to any biological agent in the form of bacteria, fungi, or virus, that is intentionally added at different stages into the food chain. However, the QPS is considered by many as the more applicable and flexible criteria, given the emerging risk of spreading antibiotic resistance genes through probiotic strains (Sanders et al., 2010).

In the paper of Sanders et al. (2010), a case was discussed wherein the outcomes of aforementioned standards for evaluating potential probiotics were not consistent. The case involved the use of a probiotic bacteria intended for infantformula that claims improved growth in developing infants. Both standards were aware that the probiotic strain involved carries a chromosomal gene for tetracycline resistance. Using the GRAS criteria, the US FDA evaluated the "safety of probiotic use" with "reasonable certainty," while the European counterpart cited "safety of probiotic use in light of the strength or weakness of the evidence for benefit" and "the lack of knowledge necessitates application of precautionary principles." As such, the US FDA granted GRAS status to the involved probiotic strain, while the EFSA did not grant QPS status.

Given the separate approaches in screening probiotic bacterial strains, multiple independent studies have demonstrated several methods in screening different probiotic strains for antibiotic resistance genes (Tompkins et al., 2008; Vankerckhoven et al., 2008; Chang et al., 2009; Galopin et al., 2009; Xiao et al., 2010; Klein, 2011; Nueno-Palop and Narbad, 2011; Cebrián et al., 2012; Songisepp et al., 2012; Wei et al., 2012; Tan et al., 2013; Devi et al., 2015; Senan et al., 2015). In places where it is not bound to either US FDA or EFSA, screening for antibiotic resistance genes in probiotic strains becomes even more important as rigid guidelines and regulations for probiotic use are lacking (Chang et al., 2009).

# Lactic Acid Bacteria and Their Safety Profile

The most common group of bacteria used as probiotics belongs to the group of lactic acid bacteria (Gueimonde et al., 2013; Varankovich et al., 2015). Included in this group are Lactobacillus and Enterococcus. Both genera are currently extensively screened for species that can be used as probiotic bacteria (Tompkins et al., 2008; Vankerckhoven et al., 2008; Chang et al., 2009; Klein, 2011; Nueno-Palop and Narbad, 2011; Cebrián et al., 2012; Songisepp et al., 2012; Tan et al., 2013; Devi et al., 2015; Senan et al., 2015).

Lactobacillus probiotic strains are capable of improving digestion, absorption, and availability of nutrients in both livestock animals and humans (Téllez et al., 2015; Varankovich et al., 2015). They are also known to inhibit and kill Helicobacter pylori, a pathogen regarded as the major cause of gastritis and peptic ulcers, and is a risk factor for gastric malignancy in humans (Varankovich et al., 2015). In addition, the risk of human infectious disease due to Lactobacillus is considered negligible at less than one case per million individuals (Sanders et al., 2010).

Enterococcus probiotic strains are also known to be effective in reducing recovery periods of acute diarrhea in both animals and humans (Vankerckhoven et al., 2008; Varankovich et al., 2015). However, unlike Lactobacillus, the genus Enterococcus has member strains that are considered opportunistic pathogens and are sometimes the etiologic agents of some human nosocomial infections, such as bacteremia and infective endocarditis (Vankerckhoven et al., 2008; Sanders et al., 2010; Franz et al., 2011; Varankovich et al., 2015).

It has also been reported in several studies that some species of lactic acid bacteria have intrinsic resistance to bacitracin, kanamycin, teicoplanin, vancomycin, and betalactams (Varankovich et al., 2015). Given the beneficial effects of lactic acid bacteria, intrinsic resistance to certain antibiotics may be considered advantageous if an antibiotic-probiotic combination therapy is desired (Hammad and Shimamoto, 2010; Varankovich et al., 2015). In the study of Hammad and Shimamoto (2010), investigators deliberately screened for antibiotic-resistant probiotic strains in 40 commercially available Japanese probiotic supplements, which are to be used with a possible probiotic-antibiotic combination therapy. However, results showed no antibiotic resistance genes found in their isolates. As such, no isolated probiotic strain was reported feasible for a probiotic-antibiotic combination therapy.

Due to horizontal gene transfer, concerns are still raised particularly in lactic acid bacteria strains that carry mobile genetic elements such as plasmids (Varankovich et al., 2015). In the study of Mater et al. (2007), in vivo transfer of the vancomycin resistance gene, a plasmid encoded gene, was successfully demonstrated between lactic acid bacteria Enterococcus faecium strains and Lactobacillus acidophilus probiotic strains during digestive transit in mice. The results highlighted the risk of probiotics being a conduit for the spread of antibiotic resistance (Mater et al., 2007).

In the study of Chang et al. (2009), it was found that commercially available food and drugs with probiotic additives contained lactic acid bacteria strains positive for antibiotic resistance genes. Although the incidence of the antibiotic resistance genes was relatively low among the sample population, the antibiotic resistance genes found in the probiotic strains were located in mobile genetic elements such as plasmids and transposons. Despite the GRAS status of the specified lactic acid bacteria strains, the study's findings confirm the threat of spreading antibiotic resistance genes through the use of probiotics. This is especially the case in countries without established guidelines and regulations in biosafety testing and

rigid post-marketing surveillance. As such, a regulatory body, if not a united world-wide health authority, should always be in place to rigidly monitor probiotic use in every country (Chang et al., 2009).

# Bifidobacteria and Their Safety Profile

Another group of bacteria commonly used as probiotics belong to the genus Bifidobacterium (Gueimonde et al., 2013; Varankovich et al., 2015). Bifidobacterium is a known major constituent of the gut microbiota of both animals and humans. Bifidobacteria have the capacity to metabolize non-digestable host dietary carbohydrates such as plant-derived dietary fiber.

As a probiotic, strains of Bifidobacterium are known to inhibit adherence of enterotoxigenic E. coli, enteropathogenic E. coli, and C. difficile to intestinal epithelial cells (Varankovich et al., 2015). In combination with Lactobacillus, Bifidobacterium was shown to alleviate side effects of H. pylori eradication therapy as some Bifidobacterium strains also suppress H. pyloriinduced genes in human epithelial cells (Varankovich et al., 2015). In addition, Bifidobacterium strains are also known to alleviate infectious diarrhea as well as inflammatory bowel disease (Varankovich et al., 2015). Like Lactobacillus, infectious diseases due to Bifidobacterium are extremely rare (Gueimonde et al., 2013).

Bifidobacterium strains are known to have intrinsic resistance against ciprofloxacin, nalidixic acid, mupirocin, streptomycin, and aminoglycosides (Wei et al., 2012; Gueimonde et al., 2013; Varankovich et al., 2015). However, resistance genes for lincosamides, macrolides, streptogramin B, and tetracycline are reported to be located in transposons (Gueimonde et al., 2013). In the study of Xiao et al. (2010), investigators tested the antibiotic susceptibility of Bifidobacterium strains distributed in the Japanese market. A total of 23 Bifidobacterium strains were isolated and tested for 15 antibiotics. Results confirmed the intrinsic resistance of Bifidobacterium against aminoglycosides. However, Bifidobacterium animalis subsp. lactis was found to be resistant to tetracycline and was in fact harboring a tetracycline resistance gene in its chromosome. The study concluded that although there is no risk factor for safety found in Bifidobacterium strains distributed in the Japanese market, the presence of the tetracycline resistance gene stresses the need for future evaluation (Xiao et al., 2010).

# Miscellaneous Concerns

The problem with the absence of a unified world-wide health authority that assumes responsibility for regulating probiotic use is exemplified by the case mentioned in the paper of Sanders et al. (2010). In countries with no established regulating body for probiotic use, the result is even worse as exemplified by the study of Chang et al. (2009). Keeping in mind that the likelihood of the spread of antibiotic resistance genes through human international travel was clearly indicated in the study of von Wintersdorff et al. (2014), we believe that a united global effort to screen probiotics that are marketed for human consumption is imperative.

In addition, given the established ecological nature of the emergence of antibiotic resistance problem, it is deemed logical to deduce that regulating probiotic use should also involve an

ecological approach. It should be noted that studies involving the regulation of probiotic use in animals are lacking when compared to those of human applications. We should be reminded that numerous studies have already shown the connection between antibiotic resistance genes in animals and those in humans (Devirgiliis et al., 2011; Schjørring and Krogfelt, 2011; Forslund et al., 2013; Verraes et al., 2013; Allen and Stanton, 2014; Hu et al., 2014; Woolhouse et al., 2015). Without addressing regulations in animal probiotic applications, efforts in regulating human probiotic use might be considered inadequate in the end.

# CONCLUSION

The emergence of antibiotic resistant pathogens through the spread of antibiotic resistance genes is an ecological problem that is exacerbated by the widespread indiscriminate use of antibiotics in livestock agriculture, and in veterinary and human medicine. The use of probiotics in lieu of antibiotics to control some diseases in animals and humans may reduce the antibiotic selective pressures on microorganisms in our natural environments and contribute in reducing the problem of the rapid emergence of antibiotic resistant pathogens. However, probiotic bacterial strains used in both animal and human applications also have risks in becoming conduits themselves in spreading antibiotic resistance genes. We conclude that the use of probiotics to address the global problem of emerging antibiotic

## REFERENCES


resistant microorganisms is a "double-edged" sword – with both beneficial effects and associated risks as depicted in **Figure 1**. Therefore, although probiotics are currently generally regarded as safe, we think that it is imperative to implement proper regulation on their use in both livestock and human applications globally to effectively mitigate their potential contribution in the spread of antibiotic resistance genes in our natural environments.

## AUTHOR CONTRIBUTIONS

II and JI conceptualized and wrote the manuscript.

# ACKNOWLEDGMENTS

This work was funded by the UP System Enhanced Creative Writing and Research Grant (ECWRG-2016-1-034). JI's probiotics research at the Immunopharmacology Research Laboratory (IRL) is supported by the UP System Balik-PhD Grant (OVPAA-BPhD-2015-01). We thank the Natural Sciences Research Institute and the Institute of Biology, College of Science, University of the Philippines Diliman, and Dr. Ernelea P. Cao for their support to II and JI in writing this review. We also thank Ms. Sandra Jelyn Cutay for her technical help in the preparation of this manuscript.


the gastrointestinal tract of humans. Antonie Van Leeuwenhoek 97, 189–200. doi: 10.1007/s10482-009-9401-0


MTCC 5463. Probiotics Antimicrob. Proteins 7, 249–258. doi: 10.1007/s12602- 015-9199-1


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Imperial and Ibana. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Characterization of Integrons and Resistance Genes in Salmonella Isolates from Farm Animals in Shandong Province, China

Xiaonan Zhao<sup>1</sup> , Jie Yang<sup>1</sup> , Baozhen Zhang<sup>2</sup> , Shuhong Sun<sup>1</sup> \* and Weishan Chang<sup>1</sup> \*

<sup>1</sup> College of Animal Science and Technology, Shandong Agricultural University, Tai'an, China, <sup>2</sup> Laboratory for Reproduction Health Institute of Biomedicine and Biotechnology Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, Shenzhen University Town, Shenzhen, China

A total of 154 non-duplicate Salmonella isolates were recovered from 1,105 rectal swabs collected from three large-scale chicken farms (78/325, 24.0%), three large-scale

duck farms (56/600, 9.3%) and three large-scale pig farms (20/180, 11.1%) between April and July 2016. Seven serotypes were identified among the 154 isolates, with the most common serotype in chickens and ducks being Salmonella enteritidis and in pigs Salmonella typhimurium. Antimicrobial susceptibility testing revealed that high antimicrobial resistance rates were observed for tetracycline (72.0%) and ampicillin (69.4%) in all sources. Class 1 integrons were detected in 16.9% (26/154) of these isolates and contained gene cassettes aadA2, aadA1, drfA1-aadA1, drfA12-aadA2, and drfA17-aadA5. Three β-lactamase genes were detected among the 154 isolates, and most of the isolates carried blaTEM−1(55/154), followed by blaPSE−1(14/154) and blaCTX−M−<sup>55</sup> (11/154). Three plasmid-mediated quinolone resistance genes were detected among the 154 isolates, and most of the isolates carried qnrA (113/154), followed by qnrB (99/154) and qnrS (10/154). Fifty-four isolates carried floR among the 154 isolates. Multilocus sequence typing (MLST) analysis showed that nine sequence types (STs) were identified; ST11 was the most frequent genotype in chickens and ducks, and ST19 was identified in pigs. Our findings indicated that Salmonella was widespread, and the overuse of antibiotics in animals should be reduced considerably in developing countries.

### Edited by:

Axel Cloeckaert, Institut National de la Recherche Agronomique (INRA), France

### Reviewed by:

Sebastian Guenther, Freie Universität Berlin, Germany Séamus Fanning, University College Dublin, Ireland

#### \*Correspondence:

Shuhong Sun jqybfkyjs@163.com Weishan Chang changweishan@yeah.net

#### Specialty section:

This article was submitted to Antimicrobials, Resistance and Chemotherapy, a section of the journal Frontiers in Microbiology

> Received: 31 March 2017 Accepted: 27 June 2017 Published: 12 July 2017

#### Citation:

Zhao X, Yang J, Zhang B, Sun S and Chang W (2017) Characterization of Integrons and Resistance Genes in Salmonella Isolates from Farm Animals in Shandong Province, China. Front. Microbiol. 8:1300. doi: 10.3389/fmicb.2017.01300 Keywords: Salmonella, antimicrobial susceptibility, class 1 integron, antimicrobial resistance gene, MLST

# INTRODUCTION

Salmonella is an important source of foodborne diseases that cause morbidity and mortality worldwide. Among 94 million cases of non-typhoid Salmonella infections, it was presumed that approximately 85% of the cases were induced by food origin Salmonella (Chiu et al., 2010). In China, Salmonella causes an estimated 22.2% of foodborne diseases (Wang et al., 2007). Many Salmonella serovars exist. More than 2,600 serovars are classified based on the reactivity of antisera to O and H antigens (Stevens et al., 2009), and the serovars from farms have a significant overlap with those causing illnesses in humans (Alcaine et al., 2006). Animals have been recognized as an important reservoir for Salmonella, and this pathogen can be transferred to humans via the food chain, posing a serious threat to human health (Vo et al., 2006).

Zhao et al. Prevalence of Salmonella from Animals

The use of antimicrobials is important for the control and treatment of Salmonella. However, antimicrobial- and multidrug-resistant Salmonella strains have emerged, leading to treatment failure (Gong et al., 2013). The increasing prevalence of multidrug-resistance among Salmonella, not only against the front-line antimicrobials, chloramphenicol and trimethoprim/sulfamethoxazole but also against clinically important antimicrobial agents, such as β-lactams and fluoroquinolones, is also an emerging problem (Lunguya et al., 2013).

The spread of the antibiotic resistant potential in Salmonella is mainly attributed to integrons. Integrons are DNA elements, capable of capturing antimicrobial resistant genes and disseminating them using a mobile genetic element (MGE) such as a plasmid among bacteria. The class I integron is the most common integron type identified in multidrugresistant (MDR) Salmonella and plays an important role in the dissemination of resistance genes among pathogens (Wright, 2010).

In developed countries, many surveys have been conducted at the molecular level to monitor the incidence of antibioticresistant Salmonella in animal farms (Melendez et al., 2010; Graciela et al., 2016). However, the extent of antibioticresistant Salmonella in many developing countries and the molecular mechanisms underlying this resistance remain unclear. Therefore, we selected large-scale animal farms as sample sites, collected swab samples, isolated Salmonella and characterized the molecular mechanisms of antimicrobial resistance.

# MATERIALS AND METHODS

# Samples and Salmonella Isolation

From April to July 2016, rectal swabs were collected from healthy animals on farms in Qingdao, Jinan and Zibo regions in Shandong Province, China. All of the sampling sites were visited only once. In total, 1,105 samples were collected in a random manner from chickens (n = 325), ducks (n = 600), and pigs (n = 180). The samples were independently collected from individual animals, and the sample collection conformed to the cluster random sampling principle. Farms were chosen based on their scale with the following requirements: for chickens, the breeding stock was >150,000 heads; for ducks, the breeding stock was >100,000 heads, and for pigs, the breeding stock was >1,000 heads. The owners of each farm gave permission for rectal swab samples to be collected. The animals from which samples were extracted remained alive and did not undergo any surgery. Therefore, ethical approval was not required for the study because the sampling process did not harm the animals. All of the collected samples were transported in an ice box to our laboratory within 6 h for further bacteriological analysis.

Isolation and identification of Salmonella were performed as described previously (Yan et al., 2010), with some modifications. Briefly, swabbing samples were placed into a sterile plastic bag containing 100 ml of buffered peptone water (BPW) and mixed vigorously for 3 min. The BPW mixture was then incubated for 24 h at 37◦C for pre-enrichment. Approximately 1 ml of preenrichment cultures were incubated in 10 ml of selenite cysteine (SC) broth and 10 ml of rappaport-vassiliadis (RV) broth at 42◦C for 24 h, respectively. After selective enrichment, a loop-full of SC and RV broth cultures were streaked onto xylose lysine tergitol 4 (XLT4) agar and incubated at 37◦C overnight. A minimum of two presumptive Salmonella colonies was confirmed by PCR using a previously described method (Malorny et al., 2003).

# Salmonella Serotyping

According to the manufacturer's instructions, the serogroup and serovars of Salmonella isolates were determined according to the Kauffmann-White scheme by slide agglutination with O and H antigens (Tianrun Bio-Pharmaceutical, Ningbo, China).

# Antimicrobial Susceptibility Testing

A minimal inhibition concentration (MIC) assay, as described by the Clinical and Laboratory Standard Institute (Clinical Laboratory Standards Institute, 2013), was used in this study to test the susceptibility of 12 commonly used antibiotics (**Table 1**), including ampicillin (AMP), amikacin (AMK), enrofloxacin (ENO), ciprofloxacin (CIP), nalidixic acid (NA), florfenicol (FFN), tetracycline (TET), ceftiofur (CEF), gentamicin (GEN), neomycin (NEO), levofloxacin (LVX), and fosfomycin (FOS). Escherichia coli (ATCC 25922) and Klebsiella pneumoniae (ATCC 700603) were used as the quality control strains in this study. Salmonella isolates resistant to more than three classes of antimicrobials were defined as MDR isolates.

# Detection of Class I Integrons and Antimicrobial Resistance Genes

Bacterial DNA was extracted using a TIANamp Bacteria DNA Kit (Tiangen, Beijing, China) according to the manufacturer's instructions. Conserved primers were used for the detection and identification of class I integrons using previously described primers and procedures (Kerrnet et al., 2002). PCR screening for β-lactamase-encoding genes blaTEM, blaPSE−1, blaSHV, and blaCTX−<sup>M</sup> was performed as previously described (Li et al., 2013).



Furthermore, PCR amplification was used to screen for plasmidmediated quinolone resistance genes, qnrA, qnrB, qnrC, qnrD, and qnrS, which were the most frequently observed in China, using previously described primers (Ahmed et al., 2013). Finally, the florfenicol resistance gene, floR, was detected using previously described primers (Ahmed et al., 2013). The PCR products were purified and subsequently sequenced (Invitrogen, Beijing, China). The obtained DNA sequences were compared with those in GenBank using Basic Local Alignment Search Tool (BLAST).

# MLST

Seven housekeeping genes (aroC, dnaN, hemD, hisD, purE, sucA, and thrA) were used to characterize Salmonella by MLST. MLST was performed as described online (http://mlst.warwick. ac.uk/mlst/dbs/Senterica/documents/primersEnterica\_html). All polymerase chain reaction products were purified and sequenced (Invitrogen, Beijing, China), and the alleles and STs were assigned according to the MLST scheme at http://mlst.warwick.ac.uk/ mlst/dbs/Senterica.

# Data Analysis

The statistical package SPSS (version 15.0, SPSS, Chicago, IL, USA) was used to compare the prevalence and MDR resistance rate of Salmonella isolated from chickens, ducks and pigs, and a P-value less than 0.05 was considered significant.

# RESULTS

# Prevalence and Serotypes of Salmonella

In this study, a total of 154 non-duplicate Salmonella isolates (154/1105, 13.9%) were recovered. From chickens, 78 Salmonella isolates were recovered (78/325, 24.0%) (**Table 2**), which was significantly higher than the Salmonella isolated from ducks and pigs (P < 0.05). Seventy-eight Salmonella isolates were divided into six serovars. The most common serovar was Salmonella enteritidis (69/78, 88.5%) (**Table 3**).

From ducks, 56 Salmonella isolates were recovered (56/600, 9.3%) (**Table 2**), and they were divided into two serovars. The most common serovar was Salmonella enteritidis (38/56, 67.9%) (**Table 3**).

From pigs, 20 Salmonella isolates were recovered (20/180, 11.1%) (**Table 2**), and they were divided into three serovars. The most common serovar was Salmonella typhimurium (13/20, 65.0%) (**Table 3**).

# Antimicrobial Susceptibility Testing

Among 78 isolates from chickens, they were susceptible to amikacin, levofloxacin and fosfomycin. Most isolates were resistant to ampicillin (69/78, 88.5%) and tetracycline (61/78, 78.2%). In addition, 63 isolates (63/78, 80.8%) exhibited MDR (**Table 3**).

Among 56 isolates from ducks, they were susceptible to amikacin, levofloxacin and fosfomycin. Most isolates were resistant to tetracycline (52/56, 92.9%) and ciprofloxacin (45/56, 80.4%). In addition, 50 isolates (50/56, 89.3%) exhibited MDR (**Table 3**), which was significantly higher than the Salmonella isolated from chickens and pigs (P < 0.05).

Among 20 isolates from pigs, they were susceptible to amikacin and levofloxacin. Most isolates were resistant to ampicillin (15/20, 75.0%) and tetracycline (9/20, 45.0%). In addition, 9 isolates (9/20, 45.0%) exhibited MDR (**Table 3**).

# Characteristics of Class I Integrons and Antimicrobial Resistance Genes

Among the 78 isolates recovered from chickens, 17 isolates (17/78, 21.8%) contained four groups of resistance gene cassettes, consisting of drfA1-aadA1 (1.7 kb, n = 7), aadA2 (1.2 kb, n = 5), drfA17-aadA5 (2 kb, n = 3), and aadA1 (1.2 kb, n = 2). Three β-lactamase genes were detected among the isolates, and blaTEM−<sup>1</sup> (n = 25) was the most commonly isolated βlactamase gene, followed by blaPSE−<sup>1</sup> (n = 7) and blaCTX−M−<sup>55</sup> (n = 4). Three plasmid-mediated quinolone resistance genes were detected among the isolates. qnrA (n = 53) was the most commonly isolated plasmid-mediated quinolone resistance gene, followed by qnrB (n = 44) and qnrS (n = 7). In addition, 23 isolates carried floR (**Table 3**).

Among the 56 isolates recovered from ducks, eight isolates (8/56, 14.3%) contained three groups of resistance gene cassettes, consisting of aadA2 (1.2 kb, n = 4), drfA1-aadA1 (1.7 kb, n = 3), and drfA12-aadA2 (2 kb, n = 1). Three β-lactamase genes were detected among the isolates. blaTEM−<sup>1</sup> was the most commonly isolated β-lactamase gene (n = 20), followed by blaPSE−1(n = 2) and blaCTX−M−<sup>55</sup> (n = 1). Three plasmid-mediated quinolone resistance genes were detected among the isolates. qnrA was the most commonly isolated plasmid-mediated quinolone resistance gene (n = 44), followed by qnrB (n = 40) and qnrS (n = 2). In addition, 13 isolates carried floR (**Table 3**).

Among the 20 isolates recovered from pigs, one isolate (1/20, 5.0%) contained one group of a resistance gene cassette, consisting of aadA2 (1.2 kb, n = 1). Three β-lactamase genes were


TABLE 3 | Resistance phenotype, ST, incidence of class I integron, and resistance gens in Salmonella isolated from animals in farms.


#### TABLE 3 | Continued


#### TABLE 3 | Continued


TABLE 3 | Continued


detected among the isolates. blaTEM−<sup>1</sup> was the most commonly isolated β-lactamase gene (n = 10), followed by blaCTX−M−<sup>55</sup> (n = 6) and blaPSE−<sup>1</sup> (n = 5). Three plasmid-mediated quinolone resistance genes were detected among the isolates. qnrA was the most commonly isolated plasmid-mediated quinolone resistance gene (n = 16), followed by qnrB (n = 15) and qnrS (n = 1). In addition, 18 isolates carried floR (**Table 3**).

# MLST

A total of nine STs among the 154 isolates were found. ST11 was the most common ST in both chickens and ducks, and it was represented by 69 and 38 Salmonella isolates, respectively. ST19 was the most common ST in pigs, and it was represented by eight Salmonella isolates (**Table 3**). The STs in this study were correlated with specific serovars, such as ST11 with Salmonella enteritidis, ST19 and ST34 with Salmonella typhimurium, and ST40 with Salmonella derby.

# DISCUSSION

In this study, Salmonella spp. were recovered from chickens, ducks and pigs in Qingdao, Jinan and Zibo regions. For the chickens, the prevalence (24.0%) was significantly higher than that reported in Shanghai, China (4.5%) (Liu et al., 2010) but was lower than that reported from chicken farms in Egypt (41.0%) (Hanem et al., 2017). The prevalence (9.3%) in ducks was similar to that obtained from duck farms in Sichuan province (12.0%) (Li et al., 2013) but was lower than those reported in Penang, Malaysia (39.0%) (Adzitey et al., 2012), and in South Korea (65.2%) (Cha et al., 2013). For pigs, the occurrence ratio (11.1%) was similar to those reported in previous studies of Salmonella spp. in food products of animal origin in China (Jiang et al., 2006; Li et al., 2013) but was higher than that reported from conventional farms (3.5%) in Korea (Migma et al., 2015). Data on the prevalence of Salmonella in different studies were difficult to compare based on differences in regions, collection seasons, sample types, isolation methodologies, culture methods, culture media, and environmental conditions.

For serotyping, a total of seven serovars were found among the 154 isolates, including six from chickens, two from ducks, and three from pigs. The most common serotype in chickens and ducks was Salmonella enteritidis. This result was consistent with those from Shanxi province (Yang et al., 2010), but it was different from other reports that the dominant serotype in chicken farms was Salmonella Colindale in Chad (Tabo et al., 2013). The most common serotype in duck farms was Salmonella typhimurium (Martelli et al., 2016). The dominant serotype in pigs was Salmonella typhimurium, which was the most common serovar isolated from humans and it can lead to severe human and animal diseases (Deng et al., 2012), but it was different from other studies, where the dominant serotype in pig farms was Salmonella IIIb in Henan province (Kuang et al., 2015), and Salmonella derby in England and Wales (Miller et al., 2011). The difference in dominant serotype among animals may be due to differences in the pathogenicity of two serovars, geographical regions and diversities (Volf et al., 2010; European Centre for Disease Prevention Control, 2013).

Antimicrobial resistance in Salmonella is a threat to human public health. As shown in **Table 3**, the high rates of antimicrobial resistance were against tetracycline (72.0%) and ampicillin (69.4%) in all sources, which was similar to reports of Salmonella isolates from Africa, in which chickens exhibited resistance to tetracycline (93.0%) and ampicillin (47.0%) (Zishiri et al., 2016). These high resistance rates are due to its wide use in animal feed and were consistent with other reports (Piras et al., 2011; Shao, 2011; Bai et al., 2015). In addition, resistance to ciprofloxacin in 35.9% of chickens, 80.4% of ducks, and 30.0% of pigs deserves our attention because resistance to this antimicrobial agent may lead to the delay or failure of fluoroquinolone therapies (Van et al., 2007). In this study, all of the isolates were susceptible to amikacin, which may be because this antimicrobial is not used for therapeutic purposes in veterinary medicine or as a growth promoter in conventional animal fattening, and the result was consistent with other reports (Eva et al., 2015). In this study, MDR Salmonella isolates were frequently observed among chickens, ducks and pigs. In addition, MDR Salmonella is serotype-dependent (Clemente et al., 2014): the data provided evidence that Salmonella indiana, Salmonella typhimurium and Salmonella enteritidis were strongly associated with MDR phenotypes. Of particular concern, MDR strains could transfer to humans via animal or animal-derived products and pose a great risk to public health (Rosangela et al., 2016).

In this study, our results related to the incidence of class I integrons (26/154, 16.9%) were similar to the report in Sichuan (Li et al., 2013) but were higher than those reported in the USA, as class I integrons were identified in only 2.8% of the Salmonella isolates from bulk milk and milk filters (Van et al., 2013). In the present study, the incidence of class I integrons was significantly higher in Salmonella from chickens (21.8%) than Salmonella from pigs (5.0%). In addition, in this study, the Salmonella isolates carrying class I integrons included Salmonella enteritidis, typhimurium and indiana.

Production of β-lactamases is considered to be the main mechanism of resistance in Gram-negative bacteria to overcome penicillin-derived antibiotics, and the blaTEM and blaCTX−<sup>M</sup> ESBLs can hydrolyse third and fourth generation cephalosporins. In this study, a total of three β-lactamase genes were detected among the Salmonella isolates recovered from chickens, ducks and pigs: blaTEM−1, blaPSE−1, and blaCTX−M−55. Most of the isolates carried blaTEM−1, which was similar to the report in South Africa that blaTEM−<sup>1</sup> was the most commonly identified β-lactamase gene in Salmonella isolates from food-producing animals (Igbinosa, 2015). In addition, in this study, most isolates carried blaTEM−<sup>1</sup> and blaCTX−M−55, which confer resistance to ampicillin.

Quinolones are the first choice for the treatment of invasive and systemic salmonellosis that occurs in humans and animals (Dimitrov et al., 2007). A total of three quinolone resistance genes were detected among the Salmonella isolates recovered from chickens, ducks and pigs: qnrA, qnrB and qnrS. qnrA was the most commonly isolated plasmid-mediated quinolone resistance gene consistent with a report in Henan, where qnrA, qnrB and qnrS were identified in Salmonella strains isolated from retail food with an incidence of 46.6, 12.7, and 19.5%, respectively (Yang et al., 2013). It is well known that qnr genes confer only low-level resistance to fluoroquinolones, and accumulation of quinolone resistance-determining region (QRDR) mutations is necessary for S. enterica to be resistant to fluoroquinolone, especially ciprofloxacin (Eaves et al., 2004). In this study, most Salmonella isolates containing a plasmid-mediated quinolone resistance gene were resistant to ciprofloxacin, nalidixic acid and gentamicin.

Florfenicol, a new chemosynthesis broad spectrum antibiotic of chloramphenicol analogs, is a fluorinated derivative of thiamphenicol. It is not approved for human use. In this study, floR was identified in 35.1% of Salmonella strains isolated from chickens, ducks and pigs, which was significantly higher than that reported in Egypt (1.0%) (Ahmed and Shimamoto, 2012). In addition, floR was identified in 90.0% of Salmonella strains isolated from pigs in this study. In this study, most Salmonella isolates containing the floR gene were resistant to florfenicol.

# REFERENCES


MLST results reveal that a total of nine STs were identified in this study. ST11 was the most frequent genotype that was recovered in chickens and ducks, and ST19 was the most frequent genotype that was recovered in pigs. ST11 belongs to Salmonella enteritidis, and ST19 belongs to Salmonella typhimurium; they all have continually been reported to cause human salmonellosis in recent years (Cai et al., 2016; Kang et al., 2017). In addition, our results revealed that the MLST patterns were generally associated with serotypes and provided a reliable prediction of the Salmonella serovars, which was consistent with previous research (Achtman et al., 2012).

# CONCLUSION

The prevalence of Salmonella was higher in the animal farms. Moreover, many serovars reported in humans and MDR Salmonella were recovered in this study. The high rates of MDR Salmonella, class I integrons and antibiotic resistance gene positive isolates detected suggest that measures must be taken to facilitate the reasonable use of antimicrobials in animal husbandry. Therefore, continuous surveillance of Salmonella and associated antimicrobial resistance in Salmonella of animals is essential to detect emerging Salmonella serovars and associated resistance genes.

# AUTHOR CONTRIBUTIONS

SS, XZ, contributed to the conception of the study; WC, XZ; contributed significantly to analysis and manuscript preparation; XZ, performed the data analyses and wrote the manuscript; XZ, JY, BZ: helped perform the analysis with constructive discussions.

## ACKNOWLEDGMENTS

This work was supported by the National key R&D project (2016YFD0501608) and (2016 YFD0500510). Taishan Scholars Program (201511023); Funds of Shandong "Double Tops" program.

hypothesis that cow- and human-associated Salmonella isolates represent distinct and overlapping populations. Appl. Environ. Microb. 72, 7575–7585. doi: 10.1128/AEM.01174-06


different broiler chickens associated with those of human isolates. BMC Microb. 10:86. doi: 10.1186/1471-2180-10-86


from the Democratic Republic of the Congo: emergence of decreased fluoroquinolone susceptibility and extended-spectrum beta lactamases. PLoS Negl. Trop. Dis. 7:e2103. doi: 10.1371/journal.pntd.0002103


of Salmonella isolates from retail foods in Henan Province, China. Food Control 32, 228–235. doi: 10.1016/j.foodcont.2012.11.022


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Zhao, Yang, Zhang, Sun and Chang. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Serotype Distribution, Antimicrobial Resistance, and Class 1 Integrons Profiles of Salmonella from Animals in Slaughterhouses in Shandong Province, China

Xiaonan Zhao, Chaoqun Ye, Weishan Chang\* and Shuhong Sun\*

*College of Animal Science and Technology, Shandong Agricultural University, Tai'an, China*

### Edited by:

*Michel Stanislas Zygmunt, Institut National de la Recherche Agronomique (INRA), France*

#### Reviewed by:

*Olivier Barraud, University of Limoges, UMR Inserm 1092 and CHU Limoges, France Kristina Kadlec, Friedrich Loeffler Institute Greifswald, Germany*

> \*Correspondence: *Shuhong Sun jqybfkyjs@163.com Weishan Chang sdwschang@sina.com*

#### Specialty section:

*This article was submitted to Antimicrobials, Resistance and Chemotherapy, a section of the journal Frontiers in Microbiology*

Received: *02 October 2016* Accepted: *26 May 2017* Published: *21 June 2017*

#### Citation:

*Zhao X, Ye C, Chang W and Sun S (2017) Serotype Distribution, Antimicrobial Resistance, and Class 1 Integrons Profiles of Salmonella from Animals in Slaughterhouses in Shandong Province, China. Front. Microbiol. 8:1049. doi: 10.3389/fmicb.2017.01049* The current study aimed to analyze the prevalence and characterization of *Salmonella* enterica isolated from animals in slaughterhouses before slaughter. A total of 143 non-duplicate *Salmonella* were recovered from 1,000 fresh fecal swabs collected from four major pig slaughterhouses (49/600, 8.2%) and four major chicken slaughterhouses (94/400, 23.5%) between March and July 2016. Among *Salmonella* isolates from pigs, the predominant serovars were *Salmonella* Rissen (28/49, 57.1%) and Typhimurium (14/49, 28.6%), and high antimicrobial resistance rates were observed for tetracycline (44/49, 89.8%) and ampicillin (16/49, 32.7%). Class 1 integrons were detected in 10.2% (5/49) of these isolates and all contained gene cassettes *aadA2* (0.65 kb). Two β-lactamase genes were detected among these isolates, and most of these isolates carried *bla*TEM-<sup>1</sup> (46/49), followed by *bla*OXA-1(4/49). Seven STs (MLST/ST, multilocus sequence typing) were detected in these isolates, and the predominant type was ST469 (19.6%). Among *Salmonella* isolates from chickens, the predominant serovars were *Salmonella* Indiana (67/94, 71.3%) and Enteritidis (23/94, 24.5%), and high antimicrobial resistance rates were observed for nalidixic acid (89/94, 94.7%), ampicillin (88/94, 93.6%) and tetracycline (81/94, 86.2%). Class 1 integrons were detected in 23 isolates (23/94, 24.5%), which contained empty integrons (0.15 kb, *n* = 6) or gene cassettes *drfA17-aadA5* (1.7 kb, *n* = 6), *aadA2* (1.2 kb, *n* = 5), *drfA16-bla*PSE*-*1*-aadA2-ereA2* (1.6 kb, *n* = 5) or *drfA1-aadA1* (1.4 kb, *n* = 1). Three β-lactamase genes were detected, and all 94 isolates carried *bla*TEM-1, followed by *bla*CTX-M-<sup>55</sup> (*n* = 19) and *bla*SPE−<sup>1</sup> (*n* = 3). Five STs were found in these isolates, and the predominant type was ST17 (71.3%). Our findings indicated that *Salmonella* was widespread in animals at slaughter and may be transmitted from animal to fork.

Keywords: Salmonella, animal slaughterhouses, antimicrobial resistance, class 1 integrons, MLST/ST

# INTRODUCTION

Salmonella enterica is a major global foodborne pathogen (Chiu et al., 2010; Scallan et al., 2011). More than 2,600 different serovars have been identified worldwide (Guibourdenche et al., 2010). In China, Salmonella causes an estimated 22.2% of foodborne diseases, and the majority of diseases are associated with the ingestion of contaminated meat products (Wang et al., 2007). Both pigs and chickens have been recognized as an important reservoir for antibiotic resistant Salmonella, and the resistance genes can be transferred to other bacteria via mobile genetic elements, such as plasmids and gene cassettes (Vo et al., 2006).

Agents of different antimicrobial classes, such as β-lactams or fluoroquinolones, are frequently used in clinical practice for Salmonella enterica infections. Unfortunately, Salmonella has gradually developed a high resistance rate to these antimicrobials, leading to the increase of healthcare costs and even clinical treatment failure (Cui et al., 2009; Gonzalez-Sanz et al., 2009). To date, numerous studies have been conducted to monitor antimicrobial resistance and molecular epidemiology of Salmonella isolated from pigs and chickens in slaughterhouses (Van et al., 2013; Mohamed et al., 2014).

However, little information concerning prevalence and characterization of Salmonella from animals in slaughterhouses in China is available. Shandong province, as a major breeding region, is the main producer of meat products in China. Therefore, major animal slaughterhouses in Shandong province, China were singled out as sampling sites to analyze the prevalence and characterization of Salmonella from animals in slaughterhouses.

# MATERIALS AND METHODS

# Description of Sampling Sites

From March to July 2016, 1,000 fresh fecal swabs were collected from four major pig slaughterhouses with process capacity of 1,500–2,500 pigs per day (150 samples per pig slaughterhouse) and four major chicken slaughterhouses with process capacity of 10,000–40,000 chickens per day (100 samples per chicken slaughterhouse). The animal slaughterhouses are respectively located in Weihai, Ciyao, Zhucheng, and Yantai regions in Shandong province, China. Sampling was carried out before slaughter, and at the time animals from different farms has been mixed.

# Identification and Sreotyping of Salmonella

From each animal slaughterhouse, fresh fecal swabs were randomly collected from different individual animals, and transported in an ice box to our laboratory within 6 h for further bacteriological analysis. Each swab sample was added into 50 mL buffered peptone water (BPW) and was incubated at 37◦C for 16 to 18 h. After that, 0.1 mL of the BPW suspensions was sub-cultured in 10 mL subpackaged Rappaport-Vassiliadis (RV) broth at 42◦C for 24 h. One loopful of each RV broth culture was then plated onto xylose lysine tergitol 4 agar plates, and was incubated at 37◦C for 24 to 48 h (Yan et al., 2010). Presumptive Salmonella colonies were identified using both the VITEK system (BioMerieux, Marcy 1'Etoile, France) and polymerase chain reaction (PCR) amplification of the inherent gene invA (Malorny et al., 2003).

All Salmonella isolates were serotyped according to the Kauffmann-White scheme by slide agglutination with O and H antigen-specific sera (Tianrun Bio-Pharmaceutical, Ningbo, China) (Grimont and Weill, 2007).

# Antimicrobial Susceptibility Testing

The Kirby-Bauer disk diffusion method was used in this study to examine resistance of Salmonella to 10 commonly used antibiotics, including amoxicillin/clavulanic acid (AMC, 20/10µg), ampicillin (AMP, 10µg), cefotaxime (CTX, 30µg), ciprofloxacin (CIP, 5µg), florfenicol (FFC, 30µg), gentamicin (GEN, 10µg), nalidixic acid (NAL, 10µg), spectinomycin (SPT, 10µg), tetracycline (TET, 30µg), and sulfamethoxazole/trimethoprim (SXT, 1.25/23.75µg). Escherichia coli (ATCC25922) was used as a quality control. The results were interpreted based on the Clinical and Laboratory Standards Institute (CLSI) standards guidelines (CLSI, 2013). Salmonella isolates resistant to more than three classes of antimicrobials were defined as multidrug resistance (MDR) isolates.

# Detection of Class I Integrons and β-Lactamase-Encoding Genes

Bacterial DNA was extracted using a TIANamp bacteria DNA kit (Tiangen, Beijing, China) according to the manufacturer's instructions. The gene cassettes within the variable region of class I integrons were detected via polymerase chain reaction (PCR), using previously described primers and procedures (Kerrn et al., 2002). The PCR products were cloned into the pMD18-T vector using the pMD18-T cloning kit (Takara, Dalian, China) and submitted for sequencing (Invitrogen, Beijing, China).

PCR screening for β-lactamase-encoding genes blaTEM, blaPSE-1, blaCMY-2, blaSHV, blaDHA-1, blaOXA, and blaCTX-<sup>M</sup> was performed as previously described (Guerra et al., 2001; Chen et al., 2004; Batchelor et al., 2005; Hasman et al., 2005; Li et al., 2013). The PCR products were purified and subsequently sequenced.

# MLST

The MLST analysis was performed by sequencing the fragments of seven housekeeping genes (aroC, dnaN, hemD, hisD, purE, sucA, and thrA), and the alleles and STs were assigned according to the MLST scheme at http://mlst.warwick.ac.uk/ mlst/dbs/Senterica. A minimum spanning tree was created using Bionumerics software 6.5 (Applied Maths, Kortrijk, Belgium), according to the instructions (the unweighted pair group method of arithmetic averages method).

# Statistical Analyses

All statistical analyses were performed using package SPSS 15.0 (SPSS Inc., Chicago, IL, USA). The chi-square test was used to compare the prevalence, multidrug resistance rate and carriage of class 1 integron of Salmonella isolated from pigs and chickens, and P < 0.05 was considered difference significant.

# RESULTS

# Prevalence and Serotypes of Salmonella

From pig slaughterhouses, 49 Salmonella isolates were recovered (49/600, 8.2%), including 13 from Weihai (13/150, 8.7%), 9 from Ciyao (9/150, 6.0%), 11 from Yantai (11/150, 10.7%), and 16 from Zhucheng (16/150, 10.7%) (**Table 1**). In terms of isolation rate of Salmonella, no significant difference was found between the pig slaughterhouses (P > 0.05). The 49 Salmonella belonged to 6 serovars, including Salmonella Rissen (n = 28), Typhimurium (n = 14), Grampian (n = 3), Derby (n = 2), Indiana (n = 1), and Enteritidis (n = 1). The most common serovars were Salmonella Rissen (28/49, 57.1%) and Typhimurium (14/49, 28.6%) (**Table 2**).

From chicken slaughterhouses, 94 Salmonella isolates were recovered (94/400, 23.5%), including 23 from Weihai (23/100, 23.0%), 33 from Ciyao (33/100, 33.0%), 17 from Yantai (17/100, 17.0%), and 21 from Zhucheng (21/150, 21.0%) (**Table 1**). In terms of isolation rate of Salmonella, no significant difference was found between the chicken slaughterhouses (P > 0.05). These 94 Salmonella isolates belonged to 4 serovars, including Salmonella Indiana (n = 67), Enteritidis (n = 23), Typhimurium (n = 3), and Hadar (n = 1). The dominant serovars were Salmonella Indiana (67/94, 71.3%) and Enteritidis (23/94, 24.5%) (**Table 2**).

# Antimicrobial Susceptibility Testing

All 49 isolates from pig slaughterhouses were susceptible to amoxicillin/clavulanic acid and cefotaxime. But most isolates were resistance to tetracycline (44/49, 89.8%) and ampicillin (16/49, 32.7%). In addition, 7 isolates (7/49, 14.3%) exhibited MDR (**Table 2**). In addition, 4 isolates were susceptible to all antibiotics used in this study.

All 94 isolates from chicken slaughterhouses were susceptible to amoxicillin/clavulanic acid and sulfamethoxazole/trimethoprim. But most isolates were resistant to nalidixic acid (89/94, 94.7%), ampicillin (87/94, 92.6%), and tetracycline (81/94, 86.2%). Eighty-six isolates (86/94, 91.5%) exhibited MDR (**Table 2**). Of note, MDR rate of Salmonella from chickens was higher than that from pigs (P < 0.05). In addition, 2 isolates were susceptible to all antibiotics used in this study.

# Characteristics of Class 1 Integrons and β-Lactamase-Encoding Genes

Among the 49 isolates recovered from pigs, class 1 integrons were found in 5 isolates (5/49, 10.2%), including 4 Salmonella Typhimurium and 1 Enteritidis. The 5 isolates only contained the single resistance gene cassette aadA2 (0.65 kb). Two β-lactamase genes were detected among the isolates, most of the isolates carried blaSPE-<sup>1</sup> (n = 46) and blaOXA-1(n = 4) (**Table 2**).

Among the 94 isolates recovered from chicken, class 1 integrons were found in 23 isolates (23/94, 24.5%), including 16 Salmonella Indiana, 5 Enteritidis and 2 Typhimurium. Of these isolates, 5 groups of resistance gene cassettes were detected: empty integrons (0.15 kb, n = 6), drfA17-aadA5 (1.6 kb, n = 6), aadA2 (1.2 kb, n = 5), drfA16-blaPSE-1-aadA2-ereA2 (1.7 kb, n = 5), and drfA1-aadA1 (1.4 kb, n = 1). Three β-lactamase genes were detected among these isolates. Most of the isolates carried blaTEM-<sup>1</sup> (n = 94), followed by blaCTX-M -55 (n = 19) and blaSPE-1(n = 3) (**Table 2**).



# MLST

One hundred and forty-three Salmonella isolates were divided into 9 STs, including 7 STs from pigs (ST11, ST17, ST19, ST34, ST40, ST358, and ST469), and 5 STs from chickens (ST11, ST17, ST19, ST33, and ST3172). The STs identified in the present study showed the following correlations with Salmonella serovars: ST11 with Salmonella Enteritidis, ST17 with Indiana, and ST469 with Rissen.

BioNumerics software version 6.5 was used to generate a minimum-spanning tree based on all the sources of STs (**Figure 1**). The dominant ST was ST17 (68/143, 47.6%), with most of isolates from chickens and only one from pigs, followed by ST469 (28/143, 19.6%), with all isolates from pigs. ST34 and ST19 belonged to one clone complex and had the same serovars of Salmonella Typhimurium. ST11 and ST3172 belonged to one clone complex, and had the same serovars of Salmonella Enteritidis.

# DISCUSSION

In this study, Salmonella isolation rate from pigs (8.2%) was much lower than that (71.8%) in Jiangsu province, China (Cai et al., 2016), and the most common serotype in pigs was Salmonella Rissen, which in consistent with the result from the retail pork products in Thailand (Prapas et al., 2016). However, this finding was different from that reported in EU in which Salmonella Typhimurium was the most common serotype (European Food Safety Authority, 2014). Of note, Salmonella Rissen isolates from pigs only showed resistant to tetracycline (85.7%), which may be associated with the fact that the antimicrobial is frequently used in pig farms in China (Bai et al., 2015).

The Salmonella isolation rate from chickens (23.5%) was similar to the result reported for frozen chicken meat in Shandong province (26.3%), China (Cui et al., 2016). However, the result in this study was much lower than that (45.2%) from chickens in Henan province, China (Bai et al., 2015) and was higher than that (4.5%) from large-scale chicken farms in Shanghai, China (Liu et al., 2010). The difference of these isolation rates may be related with collection seasons, culture methods, and local environments. In the present study, the most common serotypes identified in chickens were TABLE 2 | Resistance phenotype, incidence of class 1 integron, and resistance gens in Salmonella isolated from animals in slaughterhouses.


### TABLE 2 | Continued


#### TABLE 2 | Continued


*amoxicillin/clavulanic acid (AMC), ampicillin (AMP), cefotaxime (CTX), ciprofloxacin (CIP), florfenicol (FFC), gentamicin (GEN), nalidixic acid (NAL), spectinomycin (SPT), tetracycline (TET,) and sulfamethoxazole/trimethoprim (SXT).*

Salmonella Indiana and Enteritidis, consistent with findings reported in Henan, China (Bai et al., 2015). However, this finding differed from the result reported in Sichuan province, China, in which Salmonella Derby and Typhimurium were the most common serotypes (Li et al., 2013). Additionally, Salmonella Kentucky and Enteritidis were the most common serotypes in the USA (National Antimicrobial Resistance Monitoring System, 2011), and Salmonella Typhimurium in the EU (European Food Safety Authority, 2014). This difference may be associated with geographical regions. In the present study, Salmonella Indiana showed a high MDR rate (61/68, 89.7%), similar with the result conducted in China (Lu et al., 2011), which demonstrated that most of Salmonella Indiana showed MDR, and these bacteria were not only resistant to streptomycin and tetracycline but also were resistant to chloramphenicol, fluoroquinolones and cephalosporin antibiotics.

In the current study, most Salmonella isolates showed high resistance to tetracycline, ampicillin, and nalidixic acid, similar to the report on slaughterhouses in Italy (Piras et al., 2011), suggesting that these drugs may have been widely used on animals during disease control and prevention. A high resistance rate (63.4%) of nalidixic acid was observed in Salmonella isolates, consistent with other reports (Piras et al., 2011; Siriken et al., 2015). The resistance rate to ciprofloxacin was up to 42.7%. The results may be related with the fact that fluoroquinolone antibiotics are the most common treatment for Salmonella infections. A relatively high resistance rate to cefotaxime (29.1%) was observed in this study, which may be associated with the fact that third-generation cephalosporins have become the primary drugs for the treatment of salmonellosis because of the increase in fluoroquinolone resistance. In addition, the results of the present study showed the high prevalence of multidrug resistant Salmonella isolates in chickens (91.5%), much higher than those reported in Henan province (46.0%), China (Bai et al., 2015) and in central China (34.7%) (Kuang et al., 2015). In this study, MDR isolate rate of Salmonella (91.5%) from chickens were higher than that (14.3%) from pigs, and the higher occurrence of MDR Salmonella isolates from chickens likely reflects the extensive use of antibiotics during intensive rearing. In addition, MDR Salmonella is serotypedependent (Clemente et al., 2014): the data provided evidence that Salmonella Indiana, Typhimurium and Enteritidis were strongly associated with MDR phenotypes. However, these findings were different from a previous study showing that Salmonella Derby is commonly associated with MDR (Newell et al., 2010).

In the present study, PCR identified class 1 integrons in 19.6% of Salmonella isolates, which was similar to the 15.0% reported from retail meat products in the USA (Zhao et al., 2009) but higher than that of (2.8%) reported from milk products (Van et al., 2013). In the present study, the incidence of class 1 integrons was higher in Salmonella from chickens (24.5%) than Salmonella from pigs (10.2%) (P < 0.05). Class 1 integrons are often associated with MDR Salmonella isolates, consistent with the result of the present study. In addition, the Salmonella isolates carrying class 1

integrons included Salmonella Typhimurium, Enteritidis, and Indiana.

Four β-lactamase genes were detected among Salmonella isolates recovered from pigs and chickens: blaTEM-1, blaPSE-1, blaOXA-1, and blaCTX-M-55. Most isolates carried blaTEM-1, consistent with the report from meat and milk products in Egypt (Ashraf et al., 2014), but different from the report from animal slaughterhouses and retail meat products in Sichuan, China, which showed the dominant β-lactamase gene was blaOXA-1, followed by blaTEM-1, blaPSE-1, and blaCMY-<sup>2</sup> (Li et al., 2013). The fact that 46 Salmonella from pigs carried blaTEM-<sup>1</sup> whereas only 16 were resistant to ampicillin, and only 88 out of 94 Salmonella carrying blaTEM-<sup>1</sup> from chickens showed resistant to ampicillin may be associated with the expression status of blaTEM-<sup>1</sup> genes and is needed to be further studied.

In addition, blaCMY-<sup>2</sup> encodes resistance to third-generation cephalosporins, an important class of antibiotics used to treat complicated cases of salmonellosis (Gonzalez-Sanz et al., 2009). The incidence of blaCMY-2-positive Salmonella in China was low and was only reported in Shanxi and Sichuan (Yang et al., 2010; Li et al., 2013).

The MLST results revealed 9 STs identified in Salmonella from pigs and chickens. ST19 and ST34 have continually been reported to cause human salmonellosis in recent years, and these bacteria belong to the same serotype, Salmonella Typhimurium (Cai et al., 2016), and this circumstance was also true for Salmonella Enteritidis, represented by ST11 and ST3172. These findings suggested that serovars and STs were tightly coupled (Sukhnanand et al., 2005). ST358 is rare in China and corresponds to Salmonella Grampian, which causes an unusual increase in human cases of Salmonella Grampian infections (Horvath et al., 2013). This observation indicates that Salmonella could spread from animals to humans via pork and chicken products (Osman et al., 2014).

# CONCLUSIONS

Collectively, our findings exhibit the prevalence and characteristics of Salmonella isolated from animals in slaughterhouses in Shandong province, China. In addition, this study highlights the necessity to carry out the long-term surveillance for Salmonella recovered from food-producing animals.

# REFERENCES


# AUTHOR CONTRIBUTIONS

WC and SS: conceived and designed the study. XZ and CY: performed the experiments and analyzed the data. XZ, WC, and SS: wrote and revised the manuscript.

# ACKNOWLEDGMENTS

This work was supported by the Taishan Scholars Program (201511023) and the National key R&D project of China (2016YFD050 1608).


resistance in the Salmonella enterica isolated from clinically diarrhoeic humans in Egypt. Eur. J. Clin. Microbiol. Infect. Dis. 33, 1323–1337. doi: 10.1007/s10096-014-2072-4


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Zhao, Ye, Chang and Sun. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# High Prevalence of Colistin Resistance and mcr-1 Gene in Escherichia coli Isolated from Food Animals in China

Xianhui Huang† , Linfeng Yu† , Xiaojie Chen, Chanping Zhi, Xu Yao, Yiyun Liu, Shengjun Wu, Zewen Guo, Linxian Yi, Zhenling Zeng and Jian-Hua Liu\*

National Reference Laboratory of Veterinary Drug Residues, College of Veterinary Medicine, South China Agricultural University, Guangzhou, China

### Edited by:

Axel Cloeckaert, Institut National de la Recherche Agronomique (INRA), France

#### Reviewed by:

Laurent Poirel, Institut National de la Santé et de la Recherche Médicale (INSERM), France Maria Soledad Ramirez, California State University, Fullerton, USA

\*Correspondence:

Jian-Hua Liu jhliu@scau.edu.cn

†These authors have contributed equally to this work.

### Specialty section:

This article was submitted to Antimicrobials, Resistance and Chemotherapy, a section of the journal Frontiers in Microbiology

Received: 30 September 2016 Accepted: 20 March 2017 Published: 04 April 2017

### Citation:

Huang X, Yu L, Chen X, Zhi C, Yao X, Liu Y, Wu S, Guo Z, Yi L, Zeng Z and Liu J-H (2017) High Prevalence of Colistin Resistance and mcr-1 Gene in Escherichia coli Isolated from Food Animals in China. Front. Microbiol. 8:562. doi: 10.3389/fmicb.2017.00562 The objective of this study was to determine the minimal inhibitory concentration of colistin for Escherichia coli from food animals and the possible underlying colistin resistance mechanisms. During 2007–2014, 4,438 E. coli isolates of food animal origins were collected. The susceptibility of colistin was tested by the agar dilution method. Mutations in pmrA, pmrB, and mgrB and the presence of mcr-1 gene were determined by PCR and DNA sequencing. Complementation experiments were carried out to evaluate the contribution of the mutations to colistin resistance. There was a high frequency of colistin resistance in E. coli from pigs on farm (24.1%) and at slaughter (24.3%) in 2013–2014, followed by chickens on farm (14.0%) and at slaughter (9.5%). The resistance frequency of E. coli in cow isolates was the lowest (0.9%). MIC distribution for colistin showed that most isolates (75.2%) were distributed at 0.25 mg/L– 0.5 mg/L, followed by 4 mg/L–8 mg/L (16.8%). Compared with the isolates from pigs and chickens recovered during 2013–2014, E. coli isolates collected during 2007– 2008 (5.5%) and 2010–2011 (12.4%) showed significantly lower frequency of colistin resistance (P < 0.05). DNA sequencing and complementation experiments failed to detect any insertion inactivation or mutation in pmrA, pmrB, and mgrB associated with colistin resistance. However, 91.0% colistin-resistant isolates were positive for mcr-1. The high frequency of colistin resistance and mcr-1 gene among E. coli isolates from food animals in China urged the need to minimize potential risks of colistin resistance development and the spread of mcr-1 gene.

Keywords: Escherichia coli, colistin, food animals, resistance, mcr-1

# INTRODUCTION

The rising prevalence of multidrug-resistant (MDR) gram-negative Enterobacteriaceae (GNB), especially carbapenem- resistant, has resulted in a renewed interest in polymyxins, especially polymyxin E (colistin), for the management of gram-negative infections in many countries (Falagas and Michalopoulos, 2006; Cassir et al., 2014). Despite their relatively recent reintroduction in clinical practice, reports on colistin resistant isolates are on the rise (Falagas et al., 2010; Olaitan et al., 2014a). Resistance to polymyxins has been traditionally regarded as occurring via mutations

in genes regulating the synthesis of LAra4N (Falagas et al., 2010; Olaitan et al., 2014b). However, we recently described for the first time the emergence of plasmid-mediated colistin resistance gene, mcr-1, which now has been identified in several Enterobacteriaceae species from various sources (environment, food, animal and humans) (Liu et al., 2016).

Colistin has been used in veterinary medicine through prophylactic or metaphylactic practices, but the prevalence of colistin resistance in bacteria isolated from animals in many countries was still low (Kempf et al., 2013; Wasyl et al., 2013; Quesada et al., 2014). In China, colistin has been widely used in veterinary medicine, especially in swine and poultry for many years. We previously detected high prevalence of mcr-1 among E. coli isolates from pigs at slaughter in Guangzhou (Liu et al., 2016). Soon after, mcr-1 gene has been reported to be present in Enterobacteriaceae from animals, food and humans worldwide (Schwarz et al., 2001; Quan et al., 2017; Wang et al., 2017a). However, little is known about the prevalence of colistin resistance and mcr-1 gene among commensal E. coli isolates from other food animals in China. The aim of this study was to investigate the frequency of colistin resistance among commensal E. coli isolates from farm animals (chicken, cattle, and pig) and food animals at slaughter recovered from 12 provinces of China and to determine the possible underlying mechanisms among part of colistin-resistant isolates.

# MATERIALS AND METHODS

# Origin of E. coli Isolates

Cloacal samples from chickens (laying hens, chickens, and broilers) and rectal swabs from pigs (piglets, weaned pigs, fattening pigs, and sows) and cattle were collected from 107 food animal farms located in different geographic areas of China (Guangdong, Henan, Jiangxi, Ningxia, Jilin, Qinghai, Sichuan, Shanghai, Jiangsu, Shandong, Beijing, and Neimeng) from May 2013 to August 2014 (**Table 1**). Animals were randomly selected for sampling on each farm based on their age and stage of production. Ten to thirty samples per stage of production per farm were collected. In addition, cecal contents of chickens from seven farmers markets and two live-bird markets and rectal swabs of pigs from two live pig markets and eight abattoirs located in Guangdong, Henan, Shandong, Liaoning, and Sichuan province were collected at slaughter between April 2013 and August 2014. No more than five animal samples per farm were analyzed. All samples were seeded on MacConkey agar plates and were incubated at 37◦C for 24 h. One presumptive colony with typical E. coli morphology and size was selected and then inoculated on eosin-methylene blue agar. After incubation, suspected E. coli colony was identified using classical biochemical methods. In addition, 349 E. coli isolates (91 were from chicken during 2007–2008, 86 from chicken during 2010–2011, and 172 from pigs during 2010–2011) from healthy food animals mentioned in our previous study were also included in this study for comparison (Yang et al., 2014).

# Antimicrobial Susceptibility Testing

The minimal inhibitory concentration (MIC) of colistin was determined by the agar dilution method according to the protocols recommended in M100-S25 of the (Clinical and Laboratory Standards Institute, 2013). For isolates from pigs at slaughter, MICs of ampicillin, cefotaxime, imipenem, gentamicin, amikacin, neomycin, apramycin, florfenicol, tetracycline, ciprofloxacin, and fosfomycin were also determined. The results were interpreted according to epidemiological cut-off (ECOFF) values recommended by EUCAST<sup>1</sup> (colistin, florfenicol, and neomycin) and the interpretative criteria recommended by CLSI (M100-S25) (ampicillin, cefotaxime, gentamicin, amikacin, fosfomycin, and ciprofloxacin) (Clinical and Laboratory Standards Institute, 2013).

Statistical significance for the comparison of resistance prevalence data was determined by the χ 2 test. P values less than 0.05 were considered statistically significant.

# PCR Amplification and Sequencing

A total of 200 colistin-resistant E. coli isolates of different origins (127 from pigs, 70 from chickens, and 3 from cows) were randomly selected for PCR amplification of mcr-1 (Liu et al., 2016). In addition, 50 of them were randomly selected for sequencing for genes encoding PmrA, PmrB, and MgrB. pmrA were amplified using primers described previously (Quesada et al., 2014). The primers used for amplification of entire mgrB and pmrB genes were as follows: EmgrB-F (5<sup>0</sup> - CCGCTGAGTAATAATCCTAT -3<sup>0</sup> ) and EmgrB-R (5<sup>0</sup> - TACAACCAAAGACGCAAT -3<sup>0</sup> ), EpmrB-F (5<sup>0</sup> -

<sup>1</sup>http://www.eucast.org



ATAAGCTGAAACGGATGGC -3<sup>0</sup> ) and EpmrB-R (5<sup>0</sup> - CATA ATAATCAGGGCGAAAGT -3<sup>0</sup> ). PCR products of pmrA, pmrB, and mgrB were sequenced and the nucleotides and deduced protein sequences were analyzed at the National Center for Biotechnology Information web site<sup>2</sup> . In addition, the pmrA, pmrB, and mgrB sequences of five colistin-susceptible E. coli isolates were determined as control.

# Complementation Experiments

The wild-type mgrB and pmrB genes from an E. coli reference strain ATCC 25922 (colistin MIC of 0.25 µg/ml) were amplified by PCR using primers EmgrB-F/EmgrB-R and EpmrB-F/EpmrB-R, respectively. The non-coding mdh sequence was amplified by PCR using primers described previously (Jayol et al., 2014). The PCR products were cloned into the plasmid pCR-BluntII-TOPO (Invitrogen) encoding resistance to kanamycin and zeocin. The resulting plasmids pTOPO-mgrB, pTOPO-pmrB, and pTOPO-mdh were separately transformed into E. coli TOP10 strains by electroporation. Transformants were selected by on Mueller–Hinton agar supplemented with 50 mg/L of kanamycin. The recombinant plasmids were isolated and transformed into electrocompetent colistin-resistant E. coli with mgrB mutation (pTOPO-mgrB) or E. coli with pmrB mutation (pTOPOpmrB) via electroporation. The transformants were selected on Mueller–Hinton agar supplemented with zeocin (25 mg/L) and the presence of the cloned gene was confirmed by PCR. The colistin MICs of the transformants were determined by the agar dilution method.

# RESULTS

# Antimicrobial Susceptibility

Overall, 4089 commensal E. coli collected from 973 chickens on farm, 1392 pigs on farm, 325 chickens at slaughter, 1063

<sup>2</sup>http://www.ncbi.nlm.nih.gov

pigs at slaughter, and 336 cows on farm during 2013–2014 were recovered from 4888 samples. Among them, 763 (18.7%) isolates showed resistance to colistin (MIC ≥ 4 mg/L) (**Table 1**). The MIC values were shown in **Figure 1**. MICs of most isolates (75.2%) were distributed at 0.25 mg/L–0.5 mg/L, followed by 4 mg/L–8 mg/L (16.8%). Only MICs of 6.1% isolates and 1.9% isolates were distributed at 1 mg/L–2 mg/L and ≥16 mg/L, respectively.

There was a high frequency of colistin resistance in E. coli from pigs on farm (24.1%) and at slaughter (24.3%), followed by chickens on farm (14.0%) and at slaughter (9.5%). The resistance frequency of cow isolates was the lowest (0.9%). Compared with the isolates recovered during 2013-2014, E. coli isolates collected during 2007–2008 and 2010–2011 showed significantly lower frequency of colistin resistance (P < 0.05, **Table 1**).

Of the 258 colistin resistant isolates from pig at slaughter, 76.7% showed resistance to 3–9 other antimicrobial agents, including tetracycline (93.8%), ampicillin (79.5%), florfenicol (64.3%), cefotaxime (13.2%), neomycin (56.6%), gentamicin (29.1%), ciprofloxacin (28.3%), apramycin (12.4%), fosfomycin (7.8%), and amikacin (0.4%) (**Table 2**). The frequencies of antimicrobial resistance to other antimicrobial agents among colistin resistant isolates were significantly higher than those of colistin susceptible isolates (P < 0.01), except to amikacin and gentamicin. All of the isolates were susceptible to imipenem.

# mcr-1 Detection and Sequence of pmrA, pmrB, and mgrB Genes

Of the 200 randomly seiected colistin-resistant isolates, 182 (91.0%) were positive for mcr-1. The sequences of the pmrA, pmrB, and mgrB genes known to be involved in polymyxin resistance were determined in 50 isolates. For PmrA, no amino acid substitution was observed among the 50 isolates except one isolate that had the G144S substitution. However, G144S substitution was found to be present in colistin-susceptible

our previous study were reviewed for colistin resistance. By comparison, colistin resistance among E. coli isolated from chicken raised nearly three times from 2007/2008 to 2013/2014 and that among E. coli isolated from pigs raised nearly two times from 2010/2011 to 2013/2014. Though we could not obtain the amount of colistin consumption on each farm sampled



isolates (Quesada et al., 2014). For MgrB, 4 isolates from different regions possessed D31G substitution. For PmrB, one isolate had two amino acid substitutions (T246I and D282N). However, the MICs of colistin remained unchanged upon transformation with plasmid pTOPO-mgrB or pTOPO-pmrB.

# DISCUSSION

Despite the frequent use of colistin in animal farming for over 50 years, the occurrence of colistin resistance among E. coli strains isolated from food animals remains low (<1%) (Kempf et al., 2013; Kieffer et al., 2015). However, in this study, we found a very high prevalence of colistin resistance (18.7%) among commensal E. coli isolates from food animals, especially pigs. The frequency of resistance in commensal intestinal E. coli is considered to be a good marker for the selection pressure exerted by antibiotic use in the host animals and the resistance problems to be predicted in pathogenic bacteria (van den Bogaard and Stobberingh, 2000). This high prevalence of colistin resistance may be due to the increasing use of colistin in food animals in recent years. Our previous studies showed that most E. coli strains from chicken and pigs in China showed resistance to fluoroquinolones and florfenicol, and over 20% isolates exhibited resistance to third-generation cephalosporins, amikacin and fosfomycin (Chen et al., 2014; Rao et al., 2014). Thus, in recent years, the lack of effective drugs against E. coli might be attributed by the increased consumption of colistin in veterinary medicine, especially in piglets which are frequently treated with colistin sulphate for colibacillosis. This high selective pressure might result in the highest prevalence (64.7%) of colistin resistance among E. coli isolates from weaned piglets found in this study. Compared with pig and chicken isolates, the prevalence of colistin resistance among E. coli from cows was very low (0.9%) which might be associated with the infrequent use of this drug on dairy-farm.

To determine whether there was an increase of colistin resistance from 2007 to 2014, E. coli isolates collected in

in this study, data from China Veterinary Drug Association showed that the volume of colistin sales increased significantly from 2011 to 2013 (China Veterinary Drug Association, 2014). Taken together, our results revealed that colistin resistance in food animals was correlated with the consumption of colistin. Interestingly, MICs of most colistin- resistant isolates were 4 or 8 mg/L. The emergence and spread of E. coli with low level of colstin resistance (MIC = 4 or 8) might lead to the treatment failure of diarrhea with standard colistin dosage (2–20 mg/kg, ppm). Thus, farmers have to illegally use medicated feed added with increased dosage of

> communication). Colistin resistance among commensal E. coli isolates recovered from pigs at slaughter was also worrisomely high. These resistant bacteria might contaminate meat during slaughtering procedures and transfer to humans by food chain or to workers via direct animal contact as indicated in some previous studies (Angulo et al., 2004; Liebana et al., 2013). Recently, several studies suggested the possibility of the transference MCR-1-producing Enterobacteriaceae to humans via food chain (Campos et al., 2016; Carnevali et al., 2016; Figueiredo et al., 2016; Wang et al., 2017b). Hence it is urgent to limit the usage of polymyxins (colistin) in veterinary medicine especially as feed additives in China. Fortunately, following our discovery of mcr-1, the Chinese Government has banned the use of colistin in animal feed since Nov 1, 2016 (Walsh and Wu, 2016).

colistin (80–100 ppm) to prevent diarrhea in piglets (personal

Similar to our previous results, colistin resistance is mainly caused by mcr-1 gene. Quesada et al. have recently found mutations in PmrAB that confer resistance to polymyxins in E. coli (Quesada et al., 2014). However, in this study, we failed to detect any meaningful mutation in pmrAB and mgrB conferring resistance to colistin. Further studies are needed to understand the possible mechanism mediating colistin resistance among mcr-1-negative isolates.

# CONCLUSION

We have detected a high prevalence of colistin resistance and mcr-1 gene in E. coli from food animals. Though colistin exhibited high antimicrobial activities against GNB, including E. coli, A. baumannii, Pseudomonas aeruginosa isolates, and K. pneumonia in human (Chen et al., 2015), the frequent presence of mcr-1-positive E. coli and in food animals might be a threat to human. As colistin is the last therapeutic option against infections caused by MDR GNB, careful monitoring of the evolution of colistin resistance and the spread of mcr-1 gene in isolates from humans in China is urgently needed.

# AUTHOR CONTRIBUTIONS

fmicb-08-00562 March 31, 2017 Time: 16:28 # 5

Conceived and designed the experiments: J-HL, XH, and ZZ. Performed the experiments: XH, LY, XC, CZ, XY, YL, SW, ZG, and LY. Analyzed the data: J-HL, XH, LY, YL, and ZZ. Wrote the paper: J-HL and XH.

# REFERENCES


# ACKNOWLEDGMENT

This work was supported in part by the National Key Basic Research Program of China (No. 2013CB127200) and the Guangdong Natural Science Foundation (No. S2012030006590).

animals and human beings in China: a microbiological and molecular biological study. Lancet infect. Dis. 16, 161–168. doi: 10.1016/S1473-3099(15)00424-7


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Huang, Yu, Chen, Zhi, Yao, Liu, Wu, Guo, Yi, Zeng and Liu. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Occurrence of OXA-48 Carbapenemase and Other β-Lactamase Genes in ESBL-Producing Multidrug Resistant *Escherichia coli* from Dogs and Cats in the United States, 2009–2013

#### Xiaoqiang Liu<sup>1</sup> \*, Kamoltip Thungrat <sup>2</sup> and Dawn M. Boothe<sup>2</sup>

### *Edited by:*

*Benoit Doublet, National Institute for Agricultural Research, France*

#### *Reviewed by:*

*Jean-Yves Madec, ANSES-French Agency for Food, Environmental and Occupational Health & Safety, France Angela Novais, UCIBIO-Research Unit on Applied Molecular Biosciences, Portugal*

### *\*Correspondence:*

*Xiaoqiang Liu liuxiaoqiang142@163.com*

### *Specialty section:*

*This article was submitted to Antimicrobials, Resistance and Chemotherapy, a section of the journal Frontiers in Microbiology*

> *Received: 15 March 2016 Accepted: 23 June 2016 Published: 11 July 2016*

### *Citation:*

*Liu X, Thungrat K and Boothe DM (2016) Occurrence of OXA-48 Carbapenemase and Other* β*-Lactamase Genes in ESBL-Producing Multidrug Resistant Escherichia coli from Dogs and Cats in the United States, 2009–2013. Front. Microbiol. 7:1057. doi: 10.3389/fmicb.2016.01057*

Objective: The aim of this study was to explore the occurrence and molecular characterization of extended-spectrum β-lactamases (ESBL), plasmid-mediated AmpC β-lactamase (pAmpC) and carbapenemases among ESBL-producing multidrug resistant (MDR) *Escherichia coli* from dogs and cats in the United States.

*<sup>1</sup> Department of Basic Veterinary Medicine, College of Veterinary Medicine, Northwest A&F University, Yangling, China, <sup>2</sup> Department of Anatomy, Physiology and Pharmacology, College of Veterinary Medicine, Auburn University, Auburn, AL, USA*

Methods: Of 2443 *E.*coli isolated from dogs and cats collected between August 2009 and January 2013, 68 isolates were confirmed as ESBL-producing MDR ones. PCR and sequencing were performed to identify β-lactamases and plasmid-mediated quinolone resistance (PMQR) genes, and shed light on the virulence gene profiles, phylogenetic groups and ST types.

Results: Phylogenic group D and B2 accounted for 69.1% of the isolates. 50 (73.5%) isolates carried CTX-M ESBL gene, and the most predominant specific CTX-M subtype identified was *bla*CTX−M−<sup>15</sup> (*n* = 33), followed by *bla*CTX−M−<sup>1</sup> (*n* = 32), *bla*CTX−M−<sup>123</sup> (*n* = 27), *bla*CTX−M−<sup>9</sup> (*n* = 19) and *bla*CTX−M−<sup>14</sup> (*n* = 19), and *bla*CTX−M−<sup>123</sup> was firstly reported in *E. coli* isolates in the United States alone or in association. Other β-lactamase genes *bla*TEM, *bla*SHV, *bla*OXA−48, and *bla*CMY−<sup>2</sup> were detected in 41.2, 29.4, 19.1, and 17.6% of 68 ESBL-producing MDR isolates, respectively. The *bla*TEM and *bla*SHV genes were classfied as ESBLs with the exception of the *bla*TEM−<sup>1</sup> gene. Additionally, 42.6% (29/68) of isolates co-expressed *bla*CTX−M−<sup>15</sup> and PMQR gene *aac(6*′ *)-Ib-c*. The overall occurrence of virulence genes ranged from 11.8 (*ireA*) to 88.2% (*malX*), and most of virulence genes were less frequent among CTX-M-producing isolates than non-CTX-M isolates with the exception of *malX* and *iutA*. The 68 isolates analyzed were assigned to 31 STs with six being novel. Three pandemic clonal lineages ST131 (*n* = 10), ST648 (*n* = 9), and ST405 (*n* = 9) accounted for more than 41% of the investigated isolates, and ST648 and ST405 of phylogenetic D were firstly reported in *E. coli* from dogs and cats in the United States. Conclusion: *bla*CTX−M−<sup>123</sup> of ESBLs and carbapenemase *bla*OXA−<sup>48</sup> were firstly reported in ESBL-producing MDR *E.*coli from dogs and cats in the United States, and ST131, ST648, and ST405 were the predominant clonal groups.

Keywords: *Escherichia coli*, ESBL, OXA-48 carbapenemase, multidrug resistance, companion animals

# INTRODUCTION

Extraintestinal pathogenic strains of Escherichia coli (ExPEC) are the most important dogs and cats bacterial pathogens associated with extraintestinal infections (Beutin, 1999). However, extended-spectrum β-lactamase (ESBL)-producing ExPEC are isolated worldwide with increasing frequency from human and animal clinical isolates (Pitout, 2012). The occurrence of β-lactamases, including ESBLs, plasmidmediated AmpC β-lactamases (pAmpC) and carbapenemases among E. coli pose serious challenges to the use of penicillins, extended-spectrum cephalosporins (3rd and 4th generation cephalosporins), monobactams, and carbapenems (Karisik et al., 2008; Geser et al., 2012). Furthermore, ESBL-producing isolates are often cross-resistant to fluoroquinolones and other antimicrobial agents, thus expressed multidrug resistance (MDR). This combination of properties can significantly affect the course and outcomes of infections. β-lactamase genes commonly located on mobile genetic elements, such as plasmids, transposons, or integrons, and the resistance plasmids can easily be transferred between bacterial isolates by conjugation mechanism. Accordingly, transmission of βlactamase genes between companion animals and owner has become a subject of active discussion as companion animals could be potential sources of ESBL-producing E. coli isolates causing community-acquired infections (Schmiedel et al., 2014).

Although the ESBLs, pAmpCs and carbapenemases in E. coli isolates from humans and animals have been characterized in various studies around the world, knowledge about the βlactamases and population structure in MDR E. coli isolates from companion animals in the United States is limited. Prior to the current study only two studies have described the occurrence and the diversity of ESBLs in E. coli from dogs and cats in the United States (O'Keefe et al., 2010; Shaheen et al., 2011), and the isolates were collected from September 2004 to December 2007, and May 2008 to May 2009, respectively. However, the β-lactamases, particularly CTX-M-type ESBLs, are characterized by ongoing and complex evolution. Currently, greater than 150 variants have been identified, and several chimeras, e.g., blaCTX−M−<sup>64</sup> and blaCTX−M−<sup>123</sup> have been reported since 2009 (He et al., 2013). Moreover, several novel β-lactamases, e.g., blaKPC, blaNDM−1, and blaOXA−<sup>48</sup> are emerging worldwide in E. coli isolated from humans or animals.

The aim of the present study was to (i) investigate the occurrence and molecular characterization of ESBL-producing MDR E. coli recovered from clinical cases of infection in dogs and cats in the United States, over a period of time ranging from August 2009 to January 2013, and (ii) characterize the association between β-lactamases, phylogenetic groups, virulence genes and the ST types.

# MATERIALS AND METHODS

## Bacterial Isolates

Between August 2009 and January 2013, a total of 2443 E. coli isolates from urine, wound, ear, genital tract, anal sac, nasal structure, and soft tissue samples of dogs and cats with presumed naturally-occurring infection in six geographical regions of the United States: West (California), South (North Carolina), Central (Missouri), Midwest (Ohio and Illinois), and Southeast (Alabama), Northeast (Massachusetts) were received from a nationally recognized veterinary diagnostic laboratory. Isolates were reconfirmed to be E. coli upon receipt by the Clinical Pharmacology Laboratory (CPL) at Auburn University based on reculture overnight on CHROMagar Orientation (BD Diagnostics, Franklin Lakes, NJ) at 37◦C, and then the isolates were harvested and stored in tryptic soy broth containing 30% glycerol at −80◦C until studied.

# Susceptibility Testing and Initial ESBL Identification

Antimicrobial susceptibility testing was performed for all 2443 isolates using 96 well custom microdilution susceptibility plates according to the manufacturer's protocol (Trek Diagnostic Systems, Inc., Cleveland, OH). Susceptibility testing was performed using 16 antimicrobials representing six antimicrobial classes and classified into 12 antimicrobial categories: penicillins: ampicillin; penicillins + β-lactam inhibitor: amoxicillin-clavulanic acid; anti-pseudomonal + β-lactam inhibitor: ticarcillin-clavulanic acid; non-extended spectrum cephalosporins (1st generation cephalosporins): cephalothin; extended-spectrum cephalosporins (3rd and 4th generation cephalosporins): cefotaxime, ceftazidime, and cefpodoxime; cephamycins: cefoxitin; carbapenems: meropenem; tetracyclines: doxycycline; phenicols: chloramphenicol; fluoroquinolones: enrofloxacin and ciprofloxacin; aminoglycosides: gentamicin and amikacin; and folate pathway inhibitor: sulfamethoxazoletrimethoprim (Magiorakos et al., 2012; Thungrat et al., 2015). All MIC determinations were performed in triplicates and E. coli ATCC 25922 was used for quality control. The results were interpreted according to the guidelines of Clinical Laboratories Standards Institute (CLSI; CLSI, 2013). The MICs were recorded using the Sensititre Vizion system (Trek Diagnostic Systems), and each isolate was categorized in terms of its resistant phenotype as to: susceptible (S), non-multidrug resistance (DR) or MDR. DR was defined as resistance to 1 or 2 antimicrobial classes, and MDR was defined as resistance to three or more antimicrobial classes.

Additionally, all the 2443 E. coli isolates were screened for ESBL production using microdilution-based Sensititre (TREK diagnostic systems, Cleveland, Ohio) with ESBL Confirmatory MIC plates (ESB1F) as described previouly (Aly et al., 2012). Finally, the ESBL-producing isolates expressed MDR phenotype were used in the current study.

# Phylogenetic Grouping and Virulence Genotyping

The distribution of phylogenetic groups amongest the ESBLproducing MDR isolates was determined by the new quadruplex PCR as recently described by Clermont et al. (Clermont et al., 2013). Genomic DNA were extracted from bacterial preparations using the PreMan <sup>R</sup> Ultra Preparation Reagent according to the manufacturer's protocol. The presence of 17 virulence genes [fimH, sfa/focDE, afa/draBC, papA, papC, papG alleles (I, II, III), hlyA, cnf1, kpsM II, fyuA, iutA, ireA, iroN, traT, and malX] known for their association with pathogenicity ExPEC isolates was ascertained in each isolate by use of established PCR assay as reported previously (Johnson and Stell, 2000; Liu et al., 2015).

# Identification of β-Lactamase Genes and Other Resistance Genes

The occurrence of β-lactamase genes blaCTX−Ms, blaTEM, blaSHV, blaCMY−2, blaKPC, blaNDM−1, and blaOXA−<sup>48</sup> were identified by PCR and subsequent sequencing using specific primers and conditions previously described (Yan et al., 2004; Poirel et al., 2011; Shaheen et al., 2011). Furthermore, the identification of plasmid-mediated quinolone resistance (PMQR) genes [qnrA, qnrB, qnrC, qnrD, qnrS, aac(6′ )-Ib-cr and qepA] was carried out as described previously (Liu et al., 2012).

# Transfer of Resistance Genes by Conjugation

We tested whether the ESBL-producing E. coli isolates harboring blaCTX−Ms, blaTEM, blaSHV, or blaOXA−<sup>48</sup> enzymes were transferable. Conjugation was performed by broth mating at 37◦C on 10 ESBL-producing MDR isolates using plasmid-free sodium azide resistant E. coli J53 (J53 AZ<sup>r</sup> ) as recipient as described previously (Shaheen et al., 2011). Transconjugants were selected on tryptic soy agar plates supplemented with sodium azide (150 µg/ml) and cefotaxime (2 µg/ml). Antimicrobial susceptibility, confirmatory tests for ESBL production, and PCR detection of ESBL genes were performed on all transconjugants as described to confirm transfer of ESBL genes.

# Multilocus Sequence Typing (MLST)

MLST was performed using seven conserved housekeeping genes of E. coli (adk, fumC, gyrB, icd, mdh, purA, and recA). A detailed scheme of the MLST procedure, including the primers, PCR conditions, allelic type and sequence type assignment methods, is available at MLST databases at the at the Warwick University website (http://mlst.warwick.ac.uk/mlst/dbs/Ecoli).

# Statistical Analysis

Significance was determined by Pearson's Chi-squared test with Yates continuity correction using "R" software (version 3.0.1), and the level of significance was set at P < 0.05.

# RESULTS

# Antimicrobial Susceptibility

Among the 2443 investigated E. coli isolates, 92 isolates (3.8%) were ESBL producers, including 68 (73.9%) MDR isolates, 20 (21.5%) DR isolates, and 4 (4.4%) S isolates. Among the 68 ESBLproducing MDR isolates (including 52 dog and 16 cat isolates), 97.1% (66/68) isolates were resistance to cephalothin, followed by cefotaxime (94.1%), ampicillin (92.6%), cefpodoxime (91.2%), amoxicillin-clavulanic acid (86.8%), ticarcillin-clavulanic acid (85.3%), ceftazidime (69.1%), cefoxitin (44.1%), and meropenem (17.6%). Moreover, some of the investigated ESBLproducing MDR isolates were also resistant to non-β-lactam agents, including ciprofloxacin (91.2%), doxycycline (88.2%), enrofloxacin (82.4%), sulfamethoxazole-trimethoprim (50%), chloramphenicol (44.1%), gentamicin (39.7 %), and amikacin (30.9%).

# Phylogenetic Groups and the Virulence Genes Distribution

Phylogenetic analysis showed that the predominant phylogenetic groups were D (35.3%) and B2 (33.8%), followed by C (11.8%), A (10.3%), B1 (5.9%), E (1.5%), and F (1.5%). Fourteen of seventeen investigated virulence genes were detected, with the overall occurrence ranging from 11.8% (ireA) to 88.2% (malX) with the exception of papG I and papG II, which were not detected in any isolate. The isolates of phylogenic group B2 harbored more virulence genes (mean 7.3), and followed by group B1 (mean 5.8), group D (mean 5.6), group A (mean 4.7), and group C (mean 3.8). Furthermore, CTX-M-producing isolates possessed more virulence genes (mean 8.4) than did non-CTX-M isolates (mean 4.0; P < 0.0001). Several virulence genes, including sfa/focDE, afa/draBC, papA, papC, papG III, hlyA, cnf1, and iroN, were significantly more common or even exclusively present in non-CTX-M-producing isolates, whereas traT was significantly more common in CTX-M-producing isolates than in non-CTX-M isolates (72 vs. 22.2%, P = 0.0008). Additionally, the occurrence of virulence genes among ESBL-producing MDR E. coli was significantly lower than among non-ESBL isolates with the exception of malX (P < 0.01; **Table 1**).

# Distribution of β-Lactamases and PMQR Genes

The distribution of β-lactamase and PMQR genes among the 68 ESBL-positive MDR E. coli isolates was shown in **Table 2**. The results showed that blaTEM, blaSHV, blaCTX−M, blaCMY−2, and blaOXA−<sup>48</sup> were detected in 28 (41.2%), 20 (29.4%), 50 (73.5%), 12 (17.6%), and 13 (19.1%) isolates, respectively. 94.1% (64/68) of the isolates harbored two or more β-lactamase genes, and one isolate from dog with severe urinary tract infection coharbored eight tested genes [blaTEM−5, blaSHV−12, blaCMY−2, blaCTX−M−15, blaCTX−M−1, blaCTX−M−14, blaCTX−M−123,and aac(6′ )-Ib-cr; **Table 2**]. For the blaCTX−<sup>M</sup> positive isolates, CTX-M enzymes were clustered in CTX-M-1 (n = 35), CTX-M-9 (n = 22), and hybrid β-lactamases (n = 27) clusters. CTX-M-1 and CTX-M-9 double-positive group accounted for 10.3%


of isolates, and three isolates co-harbored CTX-M-1, CTX-M-9 as well as hybrid β-lactamase. blaCTX−M−<sup>15</sup> (n = 33) was the predominant genotype in blaCTX−<sup>M</sup> positive isolates, and followed by blaCTX−M−<sup>1</sup> (n = 32), blaCTX−M−<sup>123</sup> (n = 27), blaCTX−M−<sup>9</sup> (n = 19), and blaCTX−M−<sup>14</sup> (n = 19). Sequencing of blaTEM gene revealed 24 blaTEM−1, three blaTEM−5, and one blaTEM−30, whereas sequencing of blaSHV gene revealed 17 blaSHV−<sup>12</sup> and two blaSHV−3. All blaTEM and blaSHV genes were classfied as ESBLs with the exception of the blaTEM−<sup>1</sup> gene based on the sequencing. Moreover, 48.5% (33/68) of investigated isolates harbored aac(6′ )-Ib-cr, while none of the isolates carried qnr and qepA genes. The vast majority of aac(6′ )-Ib-cr-producing isolates were positive for blaCTX−M−15, blaCTX−M−1, and blaCTX−M−123, but negative for CTX-M 9 group enzymes despite blaCTX−M−<sup>14</sup> and aac(6′ )-Ib-cr coexisted in three isolates.

# Conjugation Experiments

We tested whether blaCTX−<sup>M</sup> genes or other β-lactamase genes in 10 selected isolates were transferable by conjugation experiments, and seven out of the 10 ESBL-producing isolates successfully transferred the β-lactamase genes to the recipient E. coli. PCR analysis showed the presence of respective blaCTX−<sup>M</sup> genes and other β-lactamase genes, including two blaOXA−48-carrying plasmids from all the transconjugants (**Table 3**). Meanwhile, PMQR gene aac-(6')-Ib-cr was co-transferred with β-lactamase genes. Generally, all donors and their transconjugants were resistant to amoxicillin-clavulanic acid, ampicillin, cefotaxime, cefoxitin, cefpodoxime, cephalothin, and ticarcillin-clavulanic acid, and all transconjugants exhibited an increase of at least eight-fold in MICs compared to the recipient, E. coli J53 AZ<sup>r</sup> . The ciprofloxacin MICs for four transconjugants harboring aac-(6′ )- Ib-cr ranged from 0.06 to 0.125 mg/L, representing an increase of two-fold to four-fold compared with the recipient (**Table 3**). Additionally, the transconjugants remained susceptible to meropenem, ciprofloxacin, gentamicin, chloramphenicol, and doxycycline, whereas one transconjugant harboring blaOXA−<sup>48</sup> was resistant to sulfamethoxazole-trimethoprim and reduced the susceptibility to meropenem.

# MLST

The MLST investigation revealed that the 68 isolates were assigned to 31 STs, including six new STs (**Table 2**). Twelve STs were represented by more than two isolates, and other 19 STs contained a single isolate each. ST131 (n = 10), ST648 (n = 9), and ST405 (n = 9) accounted for more than 41% (28/68) of investigated isolates and 54% (27/50) of CTX-Mproducing isolates, respectively. 74.2% (23/31) of STs, especially ST131, ST648, and ST405 were positively associated with CTX-M-producing isolates, while other STs, including ST10, ST5232, ST1722, ST2175, ST1800, ST73, ST372, and ST127 seem to have no relationship with CTX-Ms. Vast majority of ST131 and ST648 isolates were positively associated with blaCTX−M−<sup>15</sup> and/or blaCTX−M−<sup>1</sup> as well as blaCTX−M−123, whereas 77.8% of ST405 isolates were negatively associated with blaCTX−M−1, blaCTX−M−15, and blaCTX−M−<sup>123</sup> genes. Moreover, 55.6% of ST648 isolates were positively associated with blaOXA−48, and 12


 SXT

ENR,

 CEP,

*(Continued)*

 TIM,

 TIM,


*aThe antibiotics in parentheses*

 *indicated that the antibiotics were variability among the isolates.*

pAmpC genes blaCMY−<sup>2</sup> were distributed in nine STs. Notably, all ST131, ST405, and ST648 isolates expressed resistance to ciprofloxacin and 3rd generation cephalosporins, whereas all ST131 isolates remained susceptible to cefoxitin. A strong correlation was revealed between the virulence gene profiles and STs, and the same STs showed the similar virulence gene profiles. Among the three most common STs, ST405 isolates harbored more virulence genes (mean 4.6), followed by ST131 (mean 4.4), and virulence genes were less abundant in ST648 isolates (mean 3.4). Almost all of the ST131 and ST405 isolates were positive for afa/draBC, traT, and malX genes, ST648 isolates were significantly associated with fimH, malX and traT, but negative for afa/draBC.

# DISCUSSION

ESBLs, pAmpC and carbapenemases are mostly responsible for the emerging resistance to the β-lactam antibiotics, especially the 3rd generation cephalosporins and carbapenems in E. coli (Pitout, 2012). In the present study, we conducted a molecular detection and characterization of the β-lactamase genes in ESBLproducing MDR E. coli isolates from dogs and cats in the United States over a period of time ranging from August 2009 to January 2013, and also revealed the association between the phylogenetic groups, virulence gene profiles, genetic backbones and β-lactamase types.

The prevalence of 3.8% ESBL-producing E. coli found in this study is similar to that recorded in a recent study (3%; Shaheen et al., 2011) but higher than the first survey (1%; O'Keefe et al., 2010) among E. coli from dogs and cats in the United States. Surprisingly, 73.9% (68/92) of the ESBL-producing E. coli exhibited MDR phenotype, and 75% of MDR isolates were resistant to more than 10 antimicrobial agents tested. Phylogenetic groups D and B2 were the main phylogenetic groups in this study, and it was similar to the phylogenetic subtype distribution of the ESBL-producing isolates from human patients (Hu et al., 2013), which further demonstrated that isolates in phylogenetic groups D and B2 were associated with extraintestinal infections. Among the 68 ESBL-producing MDR isolates, blaCTX−<sup>M</sup> was prominent and detected in 73.5% (50/68) of isolates, whereas two previous similar surveys carried out in different states in the United States showed that the corresponding prevalence of blaCTX−<sup>M</sup> were 16.7% and 89.7%, respectively (O'Keefe et al., 2010; Shaheen et al., 2011). It is indicated that the geographical regions, time, resistant phenotype and the history of antimicrobial treatment of the animals can affect the prevalence of blaCTX−<sup>M</sup> gene. The high prevalence of blaCTX−<sup>M</sup> strongly suggests a significant role for E. coli isolates from companion animals as ESBL gene reservoirs, which poses an additional risk to humans. Therefore, monitoring of the spread of blaCTX−<sup>M</sup> genes in E. coli isolates in dogs and cats is urgently needed. Althouth blaCTX−M−<sup>15</sup> was still the most frequently encountered gene, the specific genotype of blaCTX−<sup>M</sup> is undergoing changes, which was supported by available evidence from the occurrence of CTX-M-9 group as well as the occurrence of a novel hybrid β-lactamase gene blaCTX−M−123. blaCTX−M−<sup>123</sup> was firstly discovered in E. coli from pig feces in China in 2013 (He et al., 2013), and afterward in human specimen (Hu et al., 2013). It is interesting to note that blaCTX−M−<sup>15</sup> is also the most widely distributed ESBL gene among human-associated Enterobacteriaceae (Cantón and Coque, 2006). These finding revealed the possibility of crosstransmission between animals and humans. Moreover, several isolates appear only with blaTEM−1, blaCMY−2, or blaOXA−48, suggesting that these isolates perhaps carry other ESBL genes, which will require further studies.

blaCMY−<sup>2</sup> was the most prevalent pAmpC, and it not only confer resistance to a wide range of extended-spectrum cephalosporins but also are not affected by β-lactamase inhibitors. blaCMY−<sup>2</sup> was dected in 17.6% of the isolates in our study, and it was significantly lower than the occurance of blaCMY−<sup>2</sup> (89%) in E. coli from companion animals in a previous study in the United States (Shaheen et al., 2011). We supposed that the occurance of blaCMY−<sup>2</sup> might be underestimated since only the ESBL-producing MDR isolates were characterized in this study. Meanwhile, our results showed that majority (58.3%) of CMY-2-producing isolates belonged to phylogenetic group D, consistent with a previous study in E. coli from human in Australia (Sidjabat et al., 2014). This similar distribution of phylogenetic group further certified that blaCMY−<sup>2</sup> can also be transferred between different bacterial species and between animals and humans (Li et al., 2007; Shaheen et al., 2011). blaOXA−<sup>48</sup> was initially reported in Klebsiella pneumoniae isolates in Turkey in 2001(Poirel et al., 2004) and afterward in other Mediterranean countries (Spain, France, Italy, Egypt, and Lebanon Turkey) (Girlich et al., 2014). In 2013, it was firstly discovered in E. coli from dogs in Germany (Stolle et al., 2013). blaOXA−<sup>48</sup> can hydrolyze carbapenems and βlactamase inhibitors but has no activity toward broad-spectrum cephalosporins (Mathers et al., 2013). Our data showed that about 19% of the isolates carried the blaOXA−48, and they were mostly associated with meropenem resistance, sequence types ST648, ST405, and ST131 as well as different combinations of β-lactamase genes. To our knowledge, blaOXA−<sup>48</sup> was firstly reported in the United States in 2012 (Poirel et al., 2012), and the present study is the first report of blaOXA−<sup>48</sup> in E. coli from dogs and cats in the United States. Moreover, blaOXA−<sup>48</sup> can transfer with other β-lactamases and aac(6′ )-Ib-cr. This finding also revealed possibility of the transfer between humans and companion animals appears highly probable through multiple potential pathways although blaOXA−<sup>48</sup> is still sporadic occurrence in animals.

aac(6′ )-Ib-cr was the exclusive PMQR gene in this study, and CTX-M-producing isolates (particularly blaCTX−M−<sup>15</sup> positive isolates) showed significantly higher occurrence of aac(6′ )-Ibcr compared to non-CTX-M or non-ESBL isolates (62 vs. 11.1 vs. 10%, P < 0.001). The frequent combination of blaCTX−M−<sup>15</sup> and aac(6′ )-Ib-cr in this study further supported the previous studies that coproduction of β-lactamases and PMQR genes could conduce to the dissemination of MDR isolates, and also reflect the fact that genes encoding resistance to β-lactams and quinolones are located on the same plasmid. Although it was not the primary focus of this study, our results coincided



with a previous study (Qin et al., 2013) that ESBL-producing isolates presented a lower occurrence of studied virulence genes compared with non-ESBL isolates (the data from another study in our laboratory) with the exception of malX gene, and CTX-M-producing E. coli harbored fewer virulence genes than non-CTX-M isolates (P<0.0001). A possible reason why individual virulence gene increased among ESBL-producing is that might be a fitness trade-off for the ESBL to survive antibiotics exposure (Qin et al., 2013) and the difference source of E. coli. The exact explanation needs additional study in the future.

A previous review suggested that attention should be paid to the rising of E. coli ST131, ST648, ST405, and ST38 isolates as they can play an important role in the worldwide distribution of CTX-M-producing E. coli (Pitout, 2012). It was further confirmed by our results since ST131, ST648, and ST405 accounted for 54% of the CTX-M-producing MDR isolates. ST131 was the predominant clone in this study, and all ST131 isolates remained susceptible to cefoxitin, which has been recently suggested as an alternative carbapenems for the treatment of infections by ESBLproducing E. coli (Guet-Revillet et al., 2014). It is noteworthy that nine ST648 isolates were strongly associated with blaCTX−M−<sup>15</sup> (88.9%, 8/9), blaOXA−<sup>48</sup> (55.6%, 5/9), and severe clinical signs. The zoonotic potential of ST648 ESBL-producing isolates has been indicated in the isolates from humans, domestic and wild animals in previous studies (Nicolas-Chanoine et al., 2008; Cortes et al., 2010), and two recent studies in Europe further suggested that ST648 clone may represent a novel genotype that combines MDR phenotype, extraintestinal virulence and zoonotic potential in companion animals (Huber et al., 2013; Ewers et al., 2014). Furthermore, ST131, ST648, and ST405 isolates have the similar β-lactamase gene combinations and resistance profiles, respectively. While it is alarming that other STs have various β-lactamase gene combinations, especially one ST38 isolate, which was associated with the highest frequency of β-lactamases and aac(6′ )-Ib-cr, high level cephalosporins resistance (MICs ≥ 32 µg/ml), lowest frequency of virulence genes and severe clinical signs. Nevertheless, constant attention and further investigations for ST648 and ST38 isolates in companion animals are necessary as they are now rapidly and globally disseminated as well as the companion animals are more and more considered an important source of human infections as the physical closeness.

# CONCLUSION

CTX-M-producing E. coli tend to have less virulent properties compared with the non-CTX-M isolates. CTX-Ms represented by blaCTX−M−1, blaCTX−M−15, and blaCTX−M−<sup>123</sup> have spread rapidly. The occurrence of blaCTX−M−<sup>123</sup> of ESBLs and blaOXA−<sup>48</sup> carbapenemase were particularly striking, being reported here for the first time in E. coli from dogs and cats in the United States. ST131, ST648, and ST405 were the predominant clonal groups among the ESBLproducing E. coli, and all ST131 isolates remained susceptible to the cefoxitin. This information will be useful for assessing the epidemiological risk factors and appropriate use of antimicrobials for ESBL-producing E. coli infections of dogs and cats.

# AUTHOR CONTRIBUTIONS

All authors listed, have made substantial, direct and intellectual contribution to the work, and approved it for publication.

# REFERENCES


# ACKNOWLEDGMENTS

The authors wish to acknowledge the assistance of laboratory staff in the Clinical Pharmacology Laboratory of Auburn University. This study was supported by the Fundamental Research Funds for the Central Universities (no. 2452016044), National Science Foundation of Shaanxi province (no. 2014JM3071), and partially supported by from Morris Animal Foundation (no. D07- MS006).


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Liu, Thungrat and Boothe. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Detection of SGI1/PGI1 Elements and Resistance to Extended-Spectrum Cephalosporins in Proteae of Animal Origin in France

Eliette Schultz1,2, Axel Cloeckaert<sup>1</sup> , Benoît Doublet<sup>1</sup> , Jean-Yves Madec<sup>2</sup> and Marisa Haenni<sup>2</sup> \*

1 Infectiologie et Santé Publique, Institut National de la Recherche Agronomique, Université François Rabelais de Tours, UMR 1282, Nouzilly, France, <sup>2</sup> Université Lyon-Agence Nationale de Sécurité Sanitaire de l'Alimentation, de l'Environnement et du Travail, Unité Antibiorésistance et Virulence Bactériennes, Lyon, France

Proteae, and especially Proteus mirabilis, are often the cause of urinary tract infections (UTIs) in humans. They were reported as carriers of extended-spectrum β-lactamase (ESBL) genes, and recently of carbapenemases, mostly carried by the Salmonella genomic island 1 (SGI1) and Proteus genomic island 1 (PGI1). Proteae have also lately become an increasing cause of UTIs in companion animals, but antimicrobial susceptibility data in animals are still scarce. Here, we report the characterization of 468 clinical epidemiologically unrelated Proteae strains from animals collected between 2013 and 2015 in France. Seventeen P. mirabilis strains (3.6%) were positive for SGI1/PGI1 and 18 Proteae (3.8%) were resistant to extended-spectrum cephalosporins (ESC). The 28 isolates carrying SGI1/PGI1 and/or ESC-resistance genes were isolated from cats, dogs, and horses. ESBL genes were detected in six genetically related P. mirabilis harboring blaVEB−<sup>6</sup> on the SGI1-V variant, but also independently of the SGI1-V, in 3 P. mirabilis strains (blaVEB 6 <sup>−</sup> and blaCTX M 15 − − ) and 1 Providencia rettgeri strain (blaCTX M 1 − − ). The AmpC resistance genes blaCMY 2 <sup>−</sup> and/or blaDHA 16 <sup>−</sup> were detected in 9 P. mirabilis strains. One strain presented both an ESBL and AmpC gene. Interestingly, the majority of the ESBL/AmpC resistance genes were located on the chromosome. In conclusion, multiple ESC-resistance genetic determinants are circulating in French animals, even though SGI1-V-carrying P. mirabilis seems to be mainly responsible for the spread of the ESBL gene blaVEB−<sup>6</sup> in dogs and horses. These results are of public health relevance and show that companion animals in close contact with humans should be regarded as a potential reservoir of ESC-resistant bacteria as well as a reservoir of ESC-resistance genes that could further disseminate to human pathogens.

Keywords: SGI1, PGI1, Proteus, ESBL, AmpC, animal, dog

# INTRODUCTION

Proteae are specific Enterobacteriaceae comprising bacterial species present in water, soil or in the intestinal tract of humans and animals. In humans, Proteae, and especially Proteus mirabilis, are often the cause of urinary tract infections (UTIs; Schaffer and Pearson, 2015). They are otherwise opportunistic pathogens responsible for various infections ranging from minor to life-threatening issues. In veterinary medicine, P. mirabilis and other Proteae are rarely found

### Edited by:

Daniela Ceccarelli, Wageningen Bioveterinary Research, Netherlands

#### Reviewed by:

Nicolas Carraro, University of Lausanne, Switzerland Genevieve Garriss, Karolinska Institutet, Sweden

> \*Correspondence: Marisa Haenni marisa.haenni@anses.fr

#### Specialty section:

This article was submitted to Antimicrobials, Resistance and Chemotherapy, a section of the journal Frontiers in Microbiology

Received: 06 October 2016 Accepted: 06 January 2017 Published: 19 January 2017

#### Citation:

Schultz E, Cloeckaert A, Doublet B, Madec J-Y and Haenni M (2017) Detection of SGI1/PGI1 Elements and Resistance to Extended-Spectrum Cephalosporins in Proteae of Animal Origin in France. Front. Microbiol. 8:32. doi: 10.3389/fmicb.2017.00032

as pathogens except as a cause of UTIs in companion animals (Bubenik et al., 2007). This pathology is only rarely treated with extended-spectrum cephalosporins (ESC), and the recommended antibiotics are sulphonamides, aminoglycosides, or fluoroquinolones.

Proteus mirabilis is naturally susceptible to β-lactams and β-lactamases inhibitors (Stock, 2003). In the late 1990s, the emergence of P. mirabilis isolates expressing acquired β-lactamases was first reported in France (Chanal et al., 2000). Extended-spectrum (ESBL) and AmpC β-lactamases are of a critical importance because they both confer resistance to nearly all β-lactams, including ESC. Interestingly, even though the corresponding genes (mostly blaCTX−M−types and blaVEB−<sup>6</sup> for ESBL and blaCMY−<sup>2</sup> and blaDHA−<sup>16</sup> for AmpC) are generally located on plasmids which allow an easy intra- or interspecies dissemination, several studies revealed chromosomeencoded ESBL/AmpC genes in P. mirabilis (Song et al., 2011; Harada et al., 2012). These chromosome-encoded genes are often carried by genomic islands (such as the Salmonella Genomic Island 1, SGI1) or integrating conjugative elements (ICEs) that may also be transmitted (Harada et al., 2010; Mata et al., 2011). While ESBL-producing P. mirabilis are nowadays commonly isolated from humans, the first CTX-M-55-producing P. mirabilis in animals was only reported in 2011 from a macaque imported from Vietnam to France (Dahmen et al., 2013).

Besides, P. mirabilis can also carry various genomic islands conferring multidrug resistance. For example SGI1, the genomic island widely disseminated in Salmonella, was first identified in a clinical P. mirabilis from a diabetic patient from Palestine in 2006 (Ahmed et al., 2007). SGI1 is a site-specific integrative mobilizable element conferring multidrug resistance initially described in Salmonella enterica serovar Typhimurium DT104 (Boyd et al., 2001).

Salmonella genomic island 1 is the first MDR genomic island identified in S. enterica, and contains a complex class 1 integron, named In104 (Hall, 2010). Since the identification of SGI1 in S. Typhimurium DT104, more than 30 different SGI1 variants carrying different combinations of antimicrobial resistance genes were described so far (Hall, 2010). The complex In104 integron variants classically possess one or two cassette attachment sites (attI1) carrying various resistance gene cassette arrays, contain an IS6100 element, may contain additional resistance genes, and are bound by 25-bp inverted repeats IRi and IRt (Boyd et al., 2001) (**Figure 1**). In the great majority of these variants, the complex In104 integron or its variants are found always at the same position in the SGI1 scaffold, i.e., between the resolvase gene res (also named tnpR) and the open reading frame (ORF) S044 at the 3<sup>0</sup> end of SGI1 (Hall, 2010). A few additional variations occurred in the SGI1 backbone of some variants, especially the insertion/deletion created by ISVch4 between ORF S005 and S009 that was found in several SGI1 variants (SGI1- H, -Ls, -Ks, -Ps, Qs, -PmABB, and -PmMAT) in S. enterica and P. mirabilis (Doublet et al., 2008; Siebor and Neuwirth, 2013).

Salmonella genomic island 1 is found integrated most of the time within the last 18 bp of the well-conserved chromosomal trmE gene (also named thdF). SGI1 is specifically mobilized in trans by conjugative plasmids of the IncA/C family (Doublet et al., 2005). Only this plasmid family has been shown to be able to mobilize SGI1 (Douard et al., 2010). The main reason of this specificity is that the SGI1 excision from the chromosome is triggered by the master activator AcaDC encoded by IncA/C conjugative plasmids (Carraro et al., 2014; Kiss et al., 2015). Then, as an extrachromosomal form, SGI1 is able to hijack the conjugative apparatus encoded by IncA/C plasmids to be conjugally transferred to a recipient cell (Carraro et al., 2014).

Since 2006, P. mirabilis strains carrying different SGI1 variants have only been reported in China and France (Boyd et al., 2008; Siebor and Neuwirth, 2013; Qin et al., 2015). Importantly, the SGI1-V variant, which is specifically found in P. mirabilis, harbors the blaVEB−<sup>6</sup> gene and was first reported in a lethal human case in France (Siebor and Neuwirth, 2011).

Recently, a new multidrug resistant genomic island named Proteus Genomic Island 1 (PGI1) was described in human P. mirabilis isolates in France (Siebor and Neuwirth, 2014). A specific PGI1 variant, PGI1-PmPEL, was shown to harbor both the blaVEB−<sup>6</sup> gene and the carbapenemase-encoding gene blaNDM−<sup>1</sup> (Girlich et al., 2015). P. mirabilis isolates of animal origin were also shown to carry SGI1 or PGI1. Indeed, SGI1 positive P. mirabilis isolates were reported in poultry and swine farms in China (Lei et al., 2014, 2015). In France, we recently described the very first cases of SGI1 (including the VEB-6 producing SGI1-V variant) or PGI1-positive P. mirabilis in dogs (Schultz et al., 2015).

The SGI1/PGI1-positive P. mirabilis isolates reported in animals and humans so far were sporadic cases. Considering the apparent emergence of these genetic determinants in P. mirabilis, our aim was to investigate the prevalence of SGI1 and PGI1 in Proteae of animal origin in France, and to characterize molecularly the collected strains. In line with recent observations that a SGI1-V-carrying P. mirabilis clonal population was shared between humans and animals, we also investigated the genetic relationship of those isolates in order to draw hypotheses on a possible transfer between the two populations. Finally, as these islands in P. mirabilis isolates were also shown to occasionally capture ESBL or carbapenemase genes, we investigated the global prevalence of those genes in the same collection.

# MATERIALS AND METHODS

# Bacterial Strains and Antibiotic Susceptibility Testing

Between April, 2013 and February, 2015, a total of 468 clinical non-duplicate Proteae isolates (P. mirabilis, n = 459; P. vulgaris, n = 1; P. penneri, n = 1; Morganella morganii, n = 2; Providencia rettgeri, n = 4; P. stuartii, n = 1) were collected from various animal hosts corresponding to distinct and epidemiological unrelated individuals (dogs, n = 411; cats, n = 25; horses, n = 13; bovine, n = 7; rabbit, n = 3; ovine, n = 3; ferret, n = 2; snake, n = 2; chicken; n = 1;

bird, n = 1) and originating from distant geographical areas throughout France. These isolates were collected through the Resapath, the French antimicrobial resistance surveillance network in animal pathogens<sup>1</sup> . Identification was performed by peripheral veterinary laboratories and then confirmed at the Anses laboratory in Lyon, France, using Api20E strips (bioMérieux, Marcy l'Etoile, France). Isolates were screened for antibiotic resistance by the disk diffusion method according to the guidelines of the CA-SFM with 30 antibiotics of veterinary or human interest<sup>2</sup> .

# SGI1/PGI1 Detection and PCR Mapping

Salmonella genomic island 1 detection was performed by PCR amplification with degenerated primers designed in order to amplify all known SGI1 and PGI1 integrase genes (FwintSGI1HR, 50ATGTTGCGTCAGGCYGAGGC; RvintSGI1HR, 50GAGTGYCCAAGAAGSCGAGAG). The chromosomal location was confirmed by PCR amplification of the left and right junctions in the chromosome, as previously described (Schultz et al., 2015). The genetic diversity of SGI1 was assessed by PCR mapping covering the entire island using primers previously described (Siebor and Neuwirth, 2011, 2013). PGI1 detection, chromosomal location, backbone mapping and resistance gene detection were performed by PCR, as previously described (**Figure 1**) (Siebor and Neuwirth, 2014).

β-lactamase genes detection was performed by PCR, as previously described (Dallenne et al., 2010). For the CTX-M-1 group, an additional PCR was performed using external primers (ISEcp1L1, 5<sup>0</sup> CAGCTTTTATGACTCG; P2D, 5<sup>0</sup> CAGCGCTTTTGCCGTCTAAG) and the amplicons were sequenced.

The chromosomal location of these genes was assessed with the I-CeuI (New England Biolabs, Hertfordshire, UK) technique (Liu et al., 1993). After digestion of the complete DNA, DNA fragments were separated by Pulsed-Field Gel Electrophoresis (PFGE) in TBE 0.5 X at 14◦C using a CHEF Mapper (Bio-Rad Laboratories, Richmond, CA, USA). Running conditions were 6 V/cm with a switch time of 5.3–49.9 s for 19.7 h. Southern blots were performed by transferring the DNA on a Hybond-N<sup>+</sup> membrane and hybridizing the membrane with DIG-labeled probes specific for the 23S rDNA, the ESBL (blaVEB−6, blaCTX−M−1, or blaCTX−M−15) and AmpC (blaCMY−<sup>2</sup> or blaDHA−16) genes of interest (see detailed primers in Supplementary Table S1). Probes were prepared by PCR using labeled DIG-dUTP (PCR DIG probe synthesis kit; Roche Diagnostics, Indianapolis, IN, USA). Detection was performed using the DIG DNA Labeling and Detection Kit (Roche Diagnostics) according to the manufacturer's instructions.

The plasmidic location of these genes was assessed by Southern blots on S1-PFGE gels (New England Biolabs, Hertfordshire, UK). Running conditions were 6 V/cm with a

Identification and Genetic Location of ESBL and AmpC Genes

<sup>1</sup>www.resapath.anses.fr

<sup>2</sup>www.sfm-microbiologie.org

switch time of 1–30 s for 20 h. Hybridization was performed as described above for Southern blots on I-CeuI gels, using the same ESBL/AmpC probes and the DIG DNA Labeling and Detection Kit.

Finally, all SGI1/PGI1-positive, ESBL and AmpC strains were analyzed by PCR-based replicon typing (PBRT kit; Diatheva, Fano, Italy) in order to type the plasmids carrying the ESBL/AmpC genes.

# Genetic Relationship of the Isolates

The genetic diversity was investigated by PFGE. DNA plugs were digested with SmaI (Promega, Madison, WI, USA) and genomic DNA was separated by CHEF Mapper gel electrophoresis in TBE 0.5 X at 14◦C. The running conditions were 6 V/cm with a switch time of 5–20 s for 22 h. XbaI-digested S. enterica serovar Braenderup strain H9812 was used as size ladder. The DNA patterns were analyzed using BioNumerics software version 4.5 (Applied Maths, Sint-Martens-Latem, Belgium) to construct a phylogenetic tree. Analysis was performed using the Dice coefficient with optimization set at 0.5% and tolerance at 1%.

# RESULTS AND DISCUSSION

# Prevalence and Molecular Characterization of the SGI1/PGI1 Genomic Islands

Salmonella genomic island 1/Proteus genomic island 1-carrying P. mirabilis are emerging pathogens in humans and animals. In a recent study, SGI1/PGI1-carrying P. mirabilis were reported in animals in France but no large-scale non-biased data on their prevalence were available (Schultz et al., 2015). Here, we investigated a large collection of 468 Proteae isolates of animal origin in France to estimate the prevalence of the SGI1/PGI1 elements. We also assessed the prevalence of resistance to broadspectrum cephalosporins in those Proteae isolates considering that the blaVEB−<sup>6</sup> gene was recurrently reported on the SGI1-V variant in P. mirabilis (Siebor and Neuwirth, 2011; Schultz et al., 2015). Among the 468 isolates studied, 17 P. mirabilis (17/468, 3.6%) were positive for SGI1 (11 isolates) or PGI1 (6 isolates) by PCR (**Figure 2** and **Table 1**). They were mostly recovered from dogs (n = 13; 13/411, 3.2%), but also from cats (n = 2; 2/25, 8%), and horses (n = 2; 2/13, 15.4%). While Proteae are more

FIGURE 2 | Analysis of SmaI-Pulsed-Field Gel Electrophoresis (PFGE) patterns obtained from the relevant P. mirabilis isolates. PFGE profiles were compared using BioNumerics software version 4.5 (Applied Maths) with settings of 0.5% optimization and 1.0% tolerance. DNA of S. enterica serovar Braenderup strain H9812 was used as standard size marker in PFGE experiments and as outgroup in analysis (Hunter et al., 2005). The isolates in bold correspond to previously published profiles included for comparison purposes (Schultz et al., 2015).



<sup>a</sup>ESBL, extended-spectrum beta-lactamase; Amx, amoxicillin; Chl, chloramphenicol; Ffc, florfenicol; Amk, amikacin; Apr, apramycin; Gen, gentamicin; Tob, tobramycin; Net, netilmicin; Kan, kanamycin; Str, streptomycin; Sss, sulfonamides; Tmp, trimethoprim; Nal, nalidixic acid; Enr, enrofloxacin. <sup>b</sup>French department: The department is the administrative subdivision of France. <sup>c</sup>P. rettgeri strain.

frequently isolated from dogs, SGI1/PGI1 elements were more prevalent in cats and horses. This would deserve confirmation due to the low number of isolates collected from these two animal categories.

Among the 11 SGI1-positive P. mirabilis, six showed an ESBL phenotype using the double disk synergy test by antibiogram. PCR mapping of the SGI1 resistance gene cluster and backbone revealed that all six ESBL-positive isolates harbored the SGI1- V variant carrying the blaVEB−<sup>6</sup> gene (**Figure 1**). In the five other SGI1-positive P. mirabilis isolates, PCR mapping of the antimicrobial resistance gene cluster gave positive results for different complex class 1 integrons, i.e., InSGI1-H, InPmABB and InPmMAT (**Figures 1**, **2** and **Table 1**). PCR mapping of the SGI1 backbone confirmed that SGI1 elements with InSGI1- H, InPmABB, and InPmMAT harbored the same 2789 bp deletion at position 5340-8120 (GenBank accession number AF261825) in the region spanning from ORF S005 to ORF S009, as previously described. Except for SGI1-InPmMAT, this deletion was replaced by the insertion of ISVch4 (also called IS1359) (**Figure 1**). Interestingly, this region contains two ORFs S007 and S006 coding for homologs of the master activator AcaDC of IncA/C conjugative plasmids and named SgaDC (SGI1 activator, subunits D and C; Poulin-Laprade et al., 2015; Muranyi et al., 2016). The SgaDC activator (also named FlhDCSGI1) was shown to be active on the same AcaDC-dependent promoter regions, i.e., Pxis, and thus should be also implicated in transfer and/or maintenance of SGI1 (Muranyi et al., 2016). The absence of flhDCSGI1 in these variants and the partial deletion of ORF S005 (traN) may have implications in their spread that needs to be further studied. In addition to the classical pentaresistance of SGI1 to amoxicillin, chloramphenicol, streptomycin and sulphonamides, other SGI1/PGI1-conferred resistances were amikacin, apramycin, gentamicin, tobramycin, netilmicin, kanamycin, nalidixic acid, trimethoprim, and enrofloxacin.

The six PGI1-positive P. mirabilis isolates were characterized by PCR mapping of the whole backbone. PCR products of the expected sizes were obtained for the complete mapping indicating a conserved PGI1 genetic structure in these isolates (**Figure 1**). All six isolates showed similar antibiotic resistance profiles as the ones previously described (Siebor and Neuwirth, 2014; Schultz et al., 2015). Interestingly PGI1, which was only recently reported in P. mirabilis, was represented in more than one-third of the genomic islands characterized here, thus confirming the proportion observed in a recent study on a much

smaller number of isolates (Schultz et al., 2015). This may suggest either a previously undetected situation in the Proteae population or a recent and rapid spread of PGI1 elements, which were first described in 2014 (Siebor and Neuwirth, 2014). Our data also indicate that the dissemination of SGI1/PGI1 in multidrugresistant P. mirabilis in animals is not a sporadic phenomenon and should be considered with great attention.

# Extended-Spectrum and AmpC β-Lactamases

Eighteen strains (18/468, 3.8%) harbored an ESBL profile (n = 9), an AmpC profile (n = 8), or a combination of both (n = 1) after antibiotic susceptibility analysis. Only one AmpC-producing strain did not present any additional resistance (**Table 1**). Otherwise, proportions of strains resistant to non-β-lactam antibiotics were particularly high for chloramphenicol (14/18, 77.8%), streptomycin (10/18, 55.6%), trimethoprim/sulphonamides (13/18, 72.2%), nalidixic acid (13/18, 72.2%), and enrofloxacin (10/18, 55.6%). Of note, none of the strains were resistant to carbapenems.

In addition to the six SGI1-V positive P. mirabilis described above, the ESBL phenotype was detected in four other strains. Two P. mirabilis isolates (38327 and 39081) harbored the blaVEB−<sup>6</sup> gene independently of the SGI1 element (**Figure 2**), whereas the two last ESBL-producing isolates harbored the blaCTX−M−<sup>15</sup> gene (P. mirabilis 34146) and the blaCTX−M−<sup>1</sup> gene (P. rettgeri 39465). The prevalence of ESBL producers was lower than the one recently reported in France. However, the previous sampling was based on isolates conserved by the veterinary laboratories and a bias toward non-susceptible isolates cannot be excluded (Schultz et al., 2015). On the contrary, no ESBL-producing isolate was detected in Japan (Harada et al., 2014). In any case, Proteae from animal origin present much less ESC-resistant isolates than those from human origin, since the nowadays ESBL rate in animals is very close to the French situation in human medicine in the late 1990s (Chanal et al., 2000).

Concerning AmpC resistance phenotypes, the blaCMY−<sup>2</sup> gene was detected alone in 7 P. mirabilis isolates whereas another P. mirabilis isolate (38375) presented a blaDHA−<sup>16</sup> gene (**Figure 2**). Finally, one P. mirabilis isolate (39214) possessed both blaCMY−<sup>2</sup> and blaDHA−<sup>16</sup> genes (**Figure 2**). The presence of multiple ESC-resistance genes in two strains (blaVEB−6/blaCMY−<sup>2</sup> in 38327 and blaCMY−2/blaDHA−<sup>16</sup> in 39214) shows the capacity of P. mirabilis to accumulate redundant resistance genes and thus acting as a potential reservoir.

The chromosomal location of genes was assessed by Southernblots on I-CeuI-PFGE using the probes corresponding to the ESBL/AmpC genes carried by the studied isolates, as well as probes specific for the 23S rDNA. This method revealed that all blaCMY−<sup>2</sup> (**Figure 3**) and blaDHA−<sup>16</sup> genes (Supplementary Figure S1) were located on the bacterial chromosome. All blaVEB−<sup>6</sup> genes carried by the SGI1-V variant as well as the blaVEB−<sup>6</sup> gene identified in isolate 39081 were also proved to be encoded by the chromosome (Supplementary Figure S2). On the contrary, the plasmidic localization of genes was proved by Southern-blots on S1-nuclease-PFGE using adequate probes. The last non-chromosomal blaVEB−<sup>6</sup> (38327; Supplementary Figure S3) and blaCTX−M−<sup>1</sup> (P. rettgeri 39465; **Figure 4**) genes were thus shown to be carried on plasmids, which were considered as non-typable because of the total absence of amplification using the PBRT kit. Finally, the P. mirabilis isolate 34146 carried two copies of the ESBL gene blaCTX−M−15, one

FIGURE 3 | Chromosomal localization of blaCMY−<sup>2</sup> in the relevant P. mirabilis isolates. (A) Whole genomic DNAs of isolates 34381 (lane1), 37665 (lane 2), 38327 (lane 3), 38368 (lane 4), 39165 (lane 5), 39175 (lane 6), 39193 (lane 7), and 39214 (lane 8) were digested with I-CeuI, and the restricted fragments subjected to PFGE. DNA fragments were transferred to a nylon membrane and hybridized with probes specific to blaCMY−<sup>2</sup> (B), and the 23S rRNA gene (C). The arrows indicate the bands of interest.

located on the chromosome and the other on a non-typable plasmid (**Figure 4**). Whereas ESBL genes were mostly reported on plasmids in Enterobacteriaceae, it is noteworthy to observe that it is seemingly not the case in Proteae. These data strongly suggest that Proteae may be more prone than other Enterobacteriaceae to integrate resistance determinants into the chromosome, mostly on SGI1/PGI1 or other genomic islands.

# Genetic Diversity of the Isolates

arrows indicate the bands of interest.

SmaI PFGE revealed that all SGI1/PGI1-positive isolates were genetically unrelated as were the ESBL/AmpC producing isolates (**Figure 2**), except the six SGI1-V positive isolates that belonged to the same cluster but interestingly differed from the ones previously reported (Schultz et al., 2015). This clustering suggests the spread of a clonal population among different and unrelated individuals. Such a dominance of a clonal population of SGI1- V/blaVEB−<sup>6</sup> P. mirabilis strains in humans and companion animals may result from close contacts between these two populations, as also demonstrated for the transfer of other multidrug resistant bacteria or plasmids, such as those carrying ESBL genes. Indeed, the transmission of bacterial clones from animals to humans (or vice-versa) through physical contacts or contact with contaminated saliva or feces has been described (Espinosa-Gongora et al., 2015; Ljungquist et al., 2016).

# CONCLUSION

In this study, we showed a significant prevalence rate (∼4%) of ESBLs/AmpC in P. mirabilis of animal sources. A longterm survey is now needed to decipher whether these are emerging phenotypes or only sporadic cases. The multidrug resistance genomic islands SGI1 and PGI1 play a major role in the dissemination of ESBLs/AmpC genes as well as other nonβ-lactam resistance genes. Moreover, we report the spread of the SGI1-V/blaVEB−6-carrying P. mirabilis clonal population to horses. This peculiar ESBL-producing P. mirabilis population was previously recognized in humans and dogs and highly suspected in poultry isolates (Seiffert et al., 2013). Altogether, these data suggest an inter-transmission pathway of public health relevance that needs further investigations to be clarified. Therefore, P. mirabilis should be regarded as a potential reservoir of resistance traits in companion animals making us believe that veterinarians should pay more attention to P. mirabilis as an opportunistic multidrug-resistant pathogen.

# AUTHOR CONTRIBUTIONS

ES, MH, J-YM, BD, and AC designed the experiments. ES did the experiments. ES, MH, and BD analyzed the data. ES and MH drafted the manuscript. BD, J-YM, and AC actively contributed to the manuscript's writing. All authors approved the final version of this manuscript.

# FUNDING

This work was supported by the French Agency for Food, Environmental and Occupational Health and Safety (ANSES). ES is supported by a Ph.D. fellowship from ANSES and INRA.

# ACKNOWLEDGMENTS

We would like to thank Cécile Ponsin for the technical help in the susceptibility testing, Pierre Châtre for performing the BioNumerics analyses, and Karine Praud for the help in the characterization of the genomic islands.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb.2017. 00032/full#supplementary-material

# REFERENCES

fmicb-08-00032 January 18, 2017 Time: 17:9 # 8


Stock, I. (2003). Natural antibiotic susceptibility of Proteus spp., with special reference to P. mirabilis and P. penneri strains. J. Chemother. 15, 12–26. doi: 10.1179/joc.2003.15.1.12

**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Schultz, Cloeckaert, Doublet, Madec and Haenni. This is an openaccess article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Clonal Spread of 16S rRNA Methyltransferase-Producing Klebsiella pneumoniae ST37 with High Prevalence of ESBLs from Companion Animals in China

Jing Xia1,2, Liang-Xing Fang1,2, Ke Cheng1,2, Guo-Hao Xu1,2, Xi-Ran Wang1,2 , Xiao-Ping Liao1,2, Ya-Hong Liu1,2,3 and Jian Sun1,2 \*

<sup>1</sup> National Risk Assessment Laboratory for Antimicrobial Resistance of Animal Original Bacteria, South China Agricultural University, Guangzhou, China, <sup>2</sup> Guangdong Provincial Key Laboratory of Veterinary Pharmaceutics Development and Safety Evaluation, South China Agricultural University, Guangzhou, China, <sup>3</sup> Jiangsu Co-Innovation Centre for Prevention and Control of Important Animal Infectious Diseases and Zoonoses, Yangzhou, China

### Edited by:

Benoit Doublet, National Institute for Agricultural Research, France

### Reviewed by:

Bruno Gonzalez-Zorn, Complutense University of Madrid, Spain Liang Li, LA BioMed, USA

> \*Correspondence: Jian Sun jiansun@scau.edu.cn

#### Specialty section:

This article was submitted to Antimicrobials, Resistance and Chemotherapy, a section of the journal Frontiers in Microbiology

Received: 04 January 2017 Accepted: 14 March 2017 Published: 29 March 2017

### Citation:

Xia J, Fang L-X, Cheng K, Xu G-H, Wang X -R, Liao X-P, Liu Y-H and Sun J (2017) Clonal Spread of 16S rRNA Methyltransferase-Producing Klebsiella pneumoniae ST37 with High Prevalence of ESBLs from Companion Animals in China. Front. Microbiol. 8:529. doi: 10.3389/fmicb.2017.00529 We screened 30 Klebsiella pneumoniae isolates from dogs and cats at a single animal hospital in Guangdong Province, China. Among them, 12 K. pneumoniae strains possessed high-level resistance to amikacin and gentamicin and these were screened for 16S rRNA methyltransferase (16S-RMTase) genes. And then the genes positive isolates were detected for ESBLs (extended spectrum β-lactamases) and analyzed by pulsed-field gel electrophoresis, multilocus sequence typing, PCR-based replicon typing and plasmid analysis. The genetic profiles of rmtB were also determined by PCR mapping. The twelve 16S-RMTase gene-positive isolates were rmtB (11/30) and armA (2/30) with one isolate carrying both genes. Extended spectrum β-lactamases genes were represented by blaCTX-M-<sup>55</sup> (9/12), blaCTX-M-<sup>27</sup> (2/12) and blaCTX-M-<sup>14</sup> (1/12). The twelve 16S-RMTase containing strains were grouped into five clonal patterns and ST37 was the most prevalent sequence type. Ten rmtB-bearing plasmids conjugated successfully and all belonged to IncN and IncF (F33:A-:B-) incompatibility groups. Nine of the transconjugants carried a 97 kb plasmid and the other harbored both ∼60 and ∼200 kb plasmids. rmtB and blaCTX-M-<sup>55</sup> were present on the same plasmid and indicated the co-transfer of these two genes, with the rmtB gene showing highly relevant relationships with IS26 and Tn3. Our findings suggested a high prevalence of 16S-RMTase genes in K. pneumonia ST37 from dogs and cats. Additional studies are needed to trace the evolutionary path of this type of resistance among the K. pneumonia isolates, and to determine whether they have been transferred to humans.

Keywords: Klebsiella pneumoniae, 16S rRNA methyltransferases, companion animals, MLST

# INTRODUCTION

Klebsiella pneumoniae is an opportunistic pathogen associated with a wide spectrum of community- and hospital-acquired infections (Lee et al., 2011). Increasing resistance to multiple antimicrobial agents has compromised the effectiveness of K. pneumoniae treatment options (Lee et al., 2011). Coexistence of resistance genes on the same plasmid appears to be the primary cause

of the spread of resistance determinants (Luo et al., 2011). This has resulted in the appearance of multidrug resistant (MDR) or even pan drug-resistant (PDR) strains (Luo et al., 2011). Acquisition of MDR by Enterobacteriaceae members has become a global concern especially for K. pneumonia PDR strains (Zacharczuk et al., 2011a; Wasyl et al., 2015; Grevskott et al., 2017).

Aminoglycoside resistance is due primarily to aminoglycoside-modifying enzymes. Acetyltransferases, nucleotidyltransferases, and phosphotransferases inactivate commonly used aminoglycosides such as gentamicin (GEN) and tobramycin (Yang et al., 2011). Recently, plasmid-encoded 16S rRNA methyltransferases (16S-RMTase) have emerged in the Enterobacteriaceae family and in a group of glucose-nonfermentative microbes (Zacharczuk et al., 2011b). This is a new resistance mechanism to 4,6-disubstituted 2-deoxystreptamines and 4,5-disubstituted 2-deoxystreptamines. These structures encompass the majority of clinically important aminoglycosides (Zacharczuk et al., 2011b). Since the first report in 2003, ten 16S-RMTase-encoding genes, rmtA, rmtB, rmtC, rmtD, rmtE, rmtF, rmtG, rmtH, armA, and npmA, have been identified (O'Hara et al., 2013).

Reports on the prevalence of the 16S-RMTases have increased in the past years with the majority focused on the human clinical isolates including K. pneumoniae (Mezzatesta et al., 2013; Belbel et al., 2014; Bartoloni et al., 2016). Moreover, the majority of reports on dissemination of armA and rmtB to various Enterobacteriaceae species were focused on E. coli isolated from chickens and pigs (Yao et al., 2011; Yang Y. et al., 2015). Previous research has reported the occurrence of the ArmA methyltransferase in an ST11 clone of K. pneumoniae isolated from cats and dogs in Spain (Hidalgo et al., 2013a). Another research focused on the dissemination of rmtB-blaCTX-M-<sup>9</sup> group genes and rmtB-qepA in Enterobacteriaceae isolates of dogs and cats in China mainly related to E. coli (Deng et al., 2011a). However, researches on prevalence and mechanisms of 16S-RMTase spread in pets via K. pneumonia strains in China are lacking.

Because of the common use of antimicrobial agents and the close contact with humans, companion animals may become potential sources for dissemination of antimicrobial resistance (Lloyd, 2007). Thus, we decided to identify the prevalence of 16S-RMTase genes among K. pneumonia isolates from dogs and cats in a veterinary hospital in Guangdong Province, China.

# MATERIALS AND METHODS

# Bacterial Isolates

Thirty clinical K. pneumoniae strains were recovered from diseased dogs and cats during July, August, and September in 2010 and from July to October in 2012 at one animal hospital in Guangdong Province, China. Further information about the antimicrobial treatments of these pets were unfortunately not available. Samples were collected from the exudates of the infection areas (urinary tract infections, skin infections, or intraabdominal infections) or feces, then seeded on MacConkey agar at 37◦C. Each isolate was from a single animal. All bacterial species were identified with classical biochemical methods and confirmed using a matrix-assisted laser desorption and ionization time-of-flight mass spectrometry (MALDI-TOF MS) method (Shimadzu, Japan). As 16S-RMTase can confer highlevel resistance to both GEN and amikacin (AMK), all isolates were subcultured in MacConkey agar containing 64 mg/L GEN and 64 mg/L AMK to screen for GEN/AMK-resistant isolates.

# Detection of Resistance Genes and Antimicrobial Susceptibility Testing

All the strains exhibiting aminoglycoside resistance were screened for 16S-RMTase genes including rmtA, rmtB, rmtC, rmtD, rmtE, rmtF, rmtG, rmtH, armA, and npmA (Hidalgo et al., 2013b; Xia et al., 2016). Aminoglycoside-resistant isolates were further analyzed for the presence of extended spectrum βlactamase (ESBL) genes (blaTEM, blaCTX-M, and blaSHV) using previously described primers (Liao et al., 2015).

The minimal inhibitory concentration (MIC) of ampicillin (AMP), cefotaxime (CTX), meropenem, ciprofloxacin (CIP), norfloxacin (NOR), GEN, kanamycin (KAN), AMK, streptomycin (STR), neomycin (NEO), apramycin (APR), chloramphenicol (CHL), florfenicol (FFC), tetracycline (TET), sulfamethoxazole/trimethoprim (SXT) were determined by the agar dilution method according to Clinical and Laboratory Standards Institute guidelines (Clinical and Laboratory Standards Institute, 2015). E. coli ATCC 25922 was used as a quality control strain.

# Clonal Relatedness

Chromosomal DNA digested with XbaI restriction enzyme was used for pulsed-field gel electrophoresis (PFGE) to analyze the genetic relatedness of all isolates containing 16S-RMTase genes (CHEF Mapper1, Bio-Rad Laboratories, Hercules, CA, USA) as previously described (Fang et al., 2015). PFGE patterns were analyzed with the Dice coefficient and the unweighted pair group method with average linkages (UPGMA) clustering method using BioNumerics software (Applied Maths, Ghent, Belgium). PFGE types were defined with >90% similarity between clusters. Multilocus sequence typing (MLST) of 16S-RMTase genes containing K. pneumoniae was performed according to the protocols described on the K. pneumoniae database website<sup>1</sup> . Seven chromosomal genes were PCR amplified and sequenced. Then the sequences were compared with those reference sequences, submitted to the MLST database to determine allele numbers and STs (Belbel et al., 2014).

# Transfer of the 16S-RMTase Genes and Plasmids Replicon Typing

Isolates positive for 16S-RMTase genes were selected for conjugation experiments by the broth-mating method using STR-resistant E. coli C600 as the recipient to determine the

<sup>1</sup>http://bigsdb.pasteur.fr/klebsiella/

transferability of 16S-RMTase genes. Transconjugants were selected on MacConkey agar plates supplemented with STR (1000 µg/mL) and AMK (64 µg/mL). Transconjugants harboring 16S-RMTase genes were confirmed by PCR and antimicrobial susceptibility testing as described above (Fang et al., 2015). Antimicrobial susceptibility testing of the transconjugants and co-transfer of other resistance genes as mentioned above were also determined.

Plasmids were preliminarily classified according to their incompatibility group by using the PCR-based replicon typing (PBRT) scheme described previously (Yao et al., 2011). To better characterize IncF plasmids, replicon sequence typing of IncF plasmids was performed according to the protocol previously described (Yang Q.E. et al., 2015). The A- and Bsymbols indicate the absence of the FIA and FIB replicons, respectively.

# Plasmid Analysis

PFGE with S1 nuclease (TakaRa Biotechnology, Dalian, China) digestion of genomic DNA was performed for all the transconjugants as described previously (Xia et al., 2015). After Southern blotting, fix the DNA to a Hybond-N<sup>+</sup> membrane (GE Healthcare, Little Chalfont, UK), the plasmids were probed with the 16S-RMTase genes and blaCTX-M-1G/9G gene (DIG High Prime DNA Labeling and Detection Starter Kit I, Roche Applied Science, Mannheim, Germany). All the transconjugants were subjected to restriction enzyme digestion (EcoRI) analysis to clarify whether a specific plasmid had been disseminated among the isolates (Deng et al., 2011b). EcoRI-RFLP types were defined with >80% similarity between clusters using the method described as PFGE analysis.

# Genetic Profiles of 16S-RMTase Genes

Reports on the genetic contexts of the 16S-RMTase genes indicated that IS26, Tn3, ISCR1, and ISCR3 played an important role in the dissemination of these genes. These transfer events enabled horizontal spread among different bacterial lineages (Deng et al., 2011b). Here we focused on the link between 16S-RMTase genes and mobile elements by PCR mapping. The primers using in PCR mapping were listed in Supplementary Table S1.

# Ethics Statement

In this study, the owners of the companion animals from which fecal swabs and the exudates of the infection areas were taken gave permission for their animals to be used in this study. Written informed consent was obtained, and the study protocol in our research was approved by the South China Agriculture University

TABLE 1 | 16S rRNA methyltransferases gene-positive K. pneumoniae isolates from this study.


a Isolates with transconjugants are underlined.

<sup>b</sup>Resistance phenotypes transferred to the recipient by conjugation are underlined. GEN, gentamicin; CTX, cefotaxime; SXT, sulfamethoxazole/trimethoprim; AMP, ampicillin; FFC, florfenicol; AMK, amikacin; CIP, ciprofloxacin; TET, tetracycline; STR, streptomycin; KAN, kanamycin; CHL, chloramphenicol; NEO, neomycin; APR, apramycin; NOR, norfloxacin.

<sup>c</sup>Genes that were co-transferred by conjugation are underlined.

<sup>d</sup>ND, not determined; NT, non-typeable.

Animal ethics committee and carried out in accordance with relevant guidelines.

# RESULTS

# Prevalence of 16S-RMTase Genes and Antimicrobial Susceptibility Testing

Among 30 K. pneumoniae isolates from the hospital, 12 were resistant to AMK and GEN. However, all K. pneumoniae strains contained the 16S-RMTase genes rmtB (11/30) and/or armA (2/30) and a single isolate carried both genes (1/30). The AMK/GEN-resistant K. pneumoniae isolates showed an MDR phenotype (resistant to three or more classes of antimicrobials). Apart from aminoglycoside antibiotics, there was a very high frequency (>90.0%) of resistance to AMP, CTX, SXT, TET, CIP, NOR, CHL, and FFC. The most frequently observed pattern of MDR was GEN-CTX-SXT-AMP-CIP-TET-STR-KAN-AMK-CHL-NEO-NOR (**Table 1**).

# Characterization of Resistance Genes

We examined the prevalence of ESBL and AME genes in the 16S-RMTase gene-positive isolates. Among these, nine isolates contained blaCTX-<sup>M</sup> that were distributed as blaCTX-M-<sup>55</sup> (6), blaCTX-M-<sup>55</sup> and blaCTX-M-<sup>27</sup> (2), and blaCTX-M-<sup>55</sup> and blaCTX-M-<sup>14</sup> (1). Two other ESBLs were also present in these isolates as blaTEM-<sup>1</sup> (10/12) and blaSHV-<sup>1</sup> (10/12) (**Table 1**).

# PFGE Typing and MLST

PFGE was successfully performed in all twelve 16S-RMTase genes positive K. pneumoniae isolates and were grouped into five clonal patterns designated PFGE types A to E (**Figure 1**). The predominant PFGE type was type C that accounted for 67% (8/12) of the isolates, suggesting that one small clonal outbreak had occurred. The MLST results revealed that one ST was the most prevalent in this collection: ST37 (9/12). ST2018, ST395, and ST147 each had one isolate (**Table 1**). The results obtained by MLST were consistent with those obtained by PFGE; eight rmtBpositive K. pneumoniae strains showing similar PFGE patterns were belonged to ST37.

# Transfer of the 16S-RMTase Genes and Plasmid Analysis

Plasmids carrying rmtB from 10 isolates were successfully transferred to recipients by conjugation. All the transconjugants were highly resistant to AMK, GEN, and KAN (MICs ≥ 512 µg/mL). PBRT of the plasmid incompatibility groups showed that nine rmtB-bearing plasmids from K. pneumoniae isolates carried two replicons. The two replicons were confirmed to be IncFII in combination with IncN, and the other rmtB-bearing plasmid was non-typeable. All the IncFII plasmids were further determined to be F33:A-:B-.

Nine of the 10 transconjugants carried one approximately 97 kb plasmid. The other one harbored both ∼60 kb (rmtB positive) and ∼200 kb (armA positive) plasmids at the same time. Moreover, rmtB and blaCTX-M-<sup>55</sup> genes were located on the same 97 kb plasmid. The RFLP analysis indicated that 7 out of the 10 transconjugants had identical plasmid restriction patterns. The two remaining plasmids shared the same RFLP type (**Figure 2**).

# Genetic Profile Analysis

The genetic profiles were successfully obtained in the 10 rmtB-positive transconjugants. We found four new genetic profiles in the 10 rmtB-positive transconjugants and all possessed Tn3 and blaTEM-<sup>1</sup> directly upstream of rmtB. Tn3 was truncated by IS26 in 9 of the 10 transconjugants, which were inserted at the same site (3573 of Tn3). The other isolate KP01 harbored the intact tnpA–tnpR genetic structure of Tn3, with a Na+/H+-exchanging protein gene and ISCR1 downstream rmtB. IS26 was also found directly downstream of rmtB in seven plasmids, which was located in the same orientation to the 5<sup>0</sup> IS26 elements. Moreover, the Na+/H+-exchanging protein gene located between rmtB and IS26 (in the opposite orientation) occurred in strain KP30 plasmid (**Figure 3**). The four unique sequences were deposited in GenBank under accession numbers KY402260, KY402261, KY402262, and KY402263.

# DISCUSSION

The 16S-RMTase genes confer broad and high-level resistance to most clinically available aminoglycosides (Habeeb et al., 2013). This has become a worldwide concern especially when these genes are combined with other classes of antimicrobial resistance determinants. This has led to the emergence of MDR strains (Habeeb et al., 2013).

In the present study, our results highlighted a high prevalence of 16S-RMTase genes in K. pneumoniae isolates from pets compared with other reports of human infections (Yan et al., 2004; Yu et al., 2009). Similarly, these 16S-RMTase-positive K. pneumoniae isolates were all MDR strains and mostly resistant to β-lactams, aminoglycosides, TETs, amphenicols, sulfonamides, and fluoroquinolones. Fortunately, the isolates we investigated in this study were susceptible to carbapenem antibiotics, which was different from other studies (Zacharczuk et al., 2011b; Li et al., 2012).

The twelve 16S-RMTase-positive K. pneumoniae isolates were successfully typed into five different groups by PFGE. Group C predominated and contained eight isolates indicating that the rmtB-positive isolates were spread by clonal dissemination. MLST analysis indicated four STs: ST37, ST2018, ST395, and ST147. The most prevalent ST in this collection was ST37 containing eight isolates belonging to the same PFGE type C group.

Previous reports had documented that ST11 was the most prevalent 16S-RMTase-positive K. pneumoniae in human isolates. ST101 was the second frequently reported ST type, which often co-harbored KPC-2 (K. pneumoniae carbapenemase) in 16S-RMTase-positive K. pneumoniae (Li et al., 2012; Mezzatesta et al., 2013; Hu et al., 2014; Seiffert et al., 2014). However, the sharing of ST37 between isolates in our study is important because KPC-producing

K. pneumoniae ST37 of human origin has been detected only sporadically in China (Yang et al., 2013; Liu et al., 2015). This observation might provide further support in concerns about transmission of resistance between companion animals and humans.

ESBL genes, especially CTX-M genes, frequently coexist and co-transfer with 16S-RMTase genes in Enterobacteriaceae members including K. pneumoniae (Yan et al., 2004; Bogaerts et al., 2007; Zacharczuk et al., 2011a). In this study, PCR and sequence analyses revealed that blaCTX-M-<sup>55</sup> and blaTEM-<sup>1</sup> were co-transferred with rmtB in 9 and 10 transconjugants of K. pneumoniae strains, respectively. This confirmed an intimate association between ESBLs and 16S-RMTase genes (Deng et al., 2011a; Xia et al., 2016). In addition, PCR-based plasmid replicon typing indicated that the nine transconjugants were in the FII and N incompatibility groups. All IncFII plasmids belonged to the same F33:A-:B- subgroup. This finding is similar with reports of rmtB-positive E. coli from food animals, pets and humans in China (Yu et al., 2010; Deng et al., 2011a,b; Yao et al., 2011; Xia et al., 2016).

This data indicated the horizontal transmission of the prevailing plasmids from variety bacterial species between different hosts. Moreover, our plasmid digestion profiles and S1-PFGE analysis confirmed that the prevalence of the rmtB-positive K. pneumoniae isolates was due to the dissemination of a ∼97 kb plasmid containing blaCTX-M-55.

Previous studies indicated that Tn3, IS26 and ISCR3 were associated with the transmission of rmtB (Doi and Arakawa, 2007; Deng et al., 2011b). In this study, the rmtB gene showed highly relevant relationships with IS26 and Tn3. The presence of a truncated Tn3 inserted via IS26 upstream of rmtB, and IS26 inserting in the same direction located downstream of rmtB, was the most prevalent structure in seven transconjugants. Within this group, five plasmids of the corresponding transconjugants shared similar EcoRI-RFLP patterns (**Figures 2**, **3**). This indicated a plasmid with this particular genetic profile was the origin of rmtB in the K. pneumoniae isolates from this group of pets.

In conclusion, rmtB was the most prevalent 16S-RMTase gene of K. pneumoniae that originated in pets. The dissemination of rmtB-producing K. pneumoniae isolates in this study appeared to be clonal spread and horizontal transfer. ST37 K. pneumoniae and F33:A-:B- plasmids sized ∼97 kb with the same genetic profile mediated by IS26 transmission both contributed to the dissemination of rmtB. In this study, we investigated only a single veterinary hospital. An organized surveillance effort is required to better understand the scope of this problem and identify control measures among hospitals.

# AUTHOR CONTRIBUTIONS

fmicb-08-00529 March 27, 2017 Time: 18:29 # 7

Y-HL and X-PL designed and organized the study. JX and L-XF did the research. KC, G-HX, and X-RW did the assisted help. JS analyzed the data and wrote the paper.

# FUNDING

This work was supported by the Program for Changjiang Scholars and Innovative Research Team in University of Ministry of

# REFERENCES


Education of China (Grant No. IRT13063), the Special Fund for Agro-scientific Research in the Public Interest (Grant No. 201203040), and the Natural Science Foundation of Guangdong Province (Grant No. S2012030006590).

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2017.00529/full#supplementary-material

methyltransferase from a patient in Miami, Florida. Antimicrob. Agents Chemother. 58, 4980–4981. doi: 10.1128/AAC.02632-14


Taiwanese hospitals. J. Antimicrob. Chemother. 54, 1007–1012. doi: 10.1093/jac/ dkh455


Diagn. Microbiol. Infect. Dis. 64, 57–63. doi: 10.1016/j.diagmicrobio.2009. 01.020


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Xia, Fang, Cheng, Xu, Wang, Liao, Liu and Sun. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Detection and Genetic Environment of Pleuromutilin-Lincosamide-Streptogramin A Resistance Genes in Staphylococci Isolated from Pets

Fengru Deng1,2† , Huiwen Wang<sup>1</sup>† , Yifei Liao<sup>1</sup>† , Jun Li<sup>1</sup> , Andrea T. Feßler<sup>3</sup> , Geovana B. Michael<sup>3</sup> , Stefan Schwarz1,3 \* and Yang Wang<sup>1</sup> \*

<sup>1</sup> Beijing Key Laboratory of Detection Technology for Animal-Derived Food Safety, College of Veterinary Medicine, China Agricultural University, Beijing, China, <sup>2</sup> Guangdong Provincial Key Laboratory of Protein Function and Regulation in Agricultural Organisms, College of Life Sciences, South China Agricultural University, Guangzhou, China, <sup>3</sup> Institute of Microbiology and Epizootics, Centre for Infection Medicine, Department of Veterinary Medicine, Freie Universität Berlin, Berlin, Germany

### Edited by:

Axel Cloeckaert, French National Institute for Agricultural Research, France

#### Reviewed by:

Séamus Fanning, University College Dublin, Ireland Lei Dai, Iowa State University, USA

#### \*Correspondence:

Yang Wang wangyang@cau.edu.cn Stefan Schwarz stefan.schwarz@fu-berlin.de

†These authors have contributed equally to this work.

#### Specialty section:

This article was submitted to Antimicrobials, Resistance and Chemotherapy, a section of the journal Frontiers in Microbiology

Received: 22 December 2016 Accepted: 02 February 2017 Published: 14 February 2017

#### Citation:

Deng F, Wang H, Liao Y, Li J, Feßler AT, Michael GB, Schwarz S and Wang Y (2017) Detection and Genetic Environment of Pleuromutilin-Lincosamide-Streptogramin A Resistance Genes in Staphylococci Isolated from Pets. Front. Microbiol. 8:234. doi: 10.3389/fmicb.2017.00234 Increasing emergence of staphylococci resistant to pleuromutilins, lincosamides, and streptogramin A (PLSA) and isolated from humans and pets is a growing public health concern worldwide. Currently, there was only one published study regarding one of the PLS<sup>A</sup> genes, vga(A) detected in staphylococci isolated from cat. In this study, eleven pleuromutilin-resistant staphylococci from pets and two from their owners were isolated and further characterized for their antimicrobial susceptibilities, plasmid profiles, genotypes, and genetic context of the PLS<sup>A</sup> resistance genes. The gene sal(A) identified in 11 staphylococcal isolates was found for the first time in Staphylococcus haemolyticus, Staphylococcus epidermidis, and Staphylococcus xylosus. Moreover, these 11 isolates shared the identical regions flanking the sal(A) gene located in the chromosomal DNA. Two S. haemolyticus isolates from a cat and its owner carried similar vga(A)LC plasmids and displayed indistinguishable PFGE patterns. A novel chromosomal multidrug resistance genomic island (MDRGI) containing 13 resistance genes, including lsa(E), was firstly identified in S. epidermidis. In addition, vga(A)LC, sal(A), and lsa(E) were for the first time identified in staphylococcal isolates originating from pet animals. The plasmids, chromosomal DNA region, and MDRGI associated with the PLS<sup>A</sup> resistance genes vga(A), vga(A)LC, sal(A), and lsa(E) are present in staphylococci isolated from pets and humans and present significant challenges for the clinical management of infections by limiting therapeutic options.

Keywords: PLS<sup>A</sup> genes, staphylococci, pet, human, multidrug resistance

# INTRODUCTION

Transferable resistance to three chemically distinct classes of antimicrobial agents (pleuromutilins, lincosamides, and streptogramin A; PLSA) in staphylococci has been attributed to ABC transporters of the Vga, Lsa, or Sal families. All of the corresponding resistance genes, including vga(A), vga(A)V, vga(A)LC, vga(B), vga(C), vga(E), vga(E)V, lsa(E), and sal(A), were mainly identified in staphylococci from food-producing isolates (Allignet and El Solh, 1997;

Haroche et al., 2000; Kadlec and Schwarz, 2009; Jung et al., 2010; Lozano et al., 2012; Hauschild et al., 2012; Wendlandt et al., 2013; Li et al., 2014; Hot et al., 2014). However, the vga(A) gene was also identified in Staphylococcus epidermidis isolates originating from one cat, clinic environment and an employee (Weiß et al., 2013).

To date, the antimicrobial resistance mechanisms of staphylococci in pets have received less attention than those in food-producing animals. Given the close relationship between pets and their owners, resistant bacteria and their mobile resistance determinants may be exchanged in either direction between pets and humans (Schwarz et al., 2016). The increasingly frequent isolation of methicillin- and multidrugresistant staphylococci from pets, particularly from dogs with pyoderma and otitis (Cain, 2013), and the occurrence of the same resistance genes in pets and in humans (Butaye et al., 2001; Simjee et al., 2002), may underline the potential role of pets in the transmission of antimicrobial-resistant staphylococci to humans. Currently, there is one published study regarding one of the PLS<sup>A</sup> genes detected in staphylococci isolated from a cat (Weiß et al., 2013). In the present study, we investigated the distribution of PLS<sup>A</sup> resistance genes in staphylococci of pet origin and analyzed the locations and genetic environments of these genes.

# MATERIALS AND METHODS

# Ethics Statement

This research was carried out according to the principles of the Declaration of Helsinki. The involved pet animals were treated with the best practice veterinary care and the informed consent was obtained from pets' owners. The study protocol was approved by the Ethics Committee of China Agricultural University.

# Bacterial Strains and Antimicrobial Susceptibility Testing

A total of 300 anal swabs or nasal samples were used in this study which were collected from dogs (n = 269), cats (n = 10), and some of their owners (n = 21) in the Veterinary Teaching Hospital of China Agricultural University, Beijing in 2013. The pleuromutilin-resistant isolates were selected by incubation on mannitol salt agar (Luqiao, Beijing, China) supplemented with 2 µg/ml of valnemulin at 37◦C for 16–24 h.

The MIC (Minimal Inhibitory Concentration) determinations were done by a standard broth microdilution test following the recommendations of the Clinical and Laboratory Standards Institute [CLSI] (2015). Staphylococcus aureus ATCC 29213 was served as the quality control strain.

# Molecular Methods

The genomic DNAs of the valnemulin-resistant isolates were obtained using a Wizard <sup>R</sup> Genomic DNA Purification Kit (Promega, Madison, WI, USA), and were screened for the vga, lsa, and sal gene types by PCR assays (Novotna and Janata, 2006; Hot et al., 2014; Li et al., 2014; Wendlandt et al., 2015). Furthermore, the species assignment of the resistant isolates was done by 16S rDNA sequencing and confirmed by MALDI-TOF MS (Bruker Daltonik, Bremen, Germany).

The clonality of the valnemulin-resistant isolates was analyzed by pulsed-field gel electrophoresis (PFGE) as described previously (Wang et al., 2011), and the PFGE results were analyzed using InfoQuestFP software (version 4.5).

The S1 nuclease-PFGE (S1-PFGE) and subsequent Southern blot hybridization were performed to locate the valnemulin resistance genes as described before (Barton et al., 1995). The Low Range PFGE Marker (New England BioLabs, Beverly, MA, USA) served as the size marker.

# Characterization and Sequence Analysis of the PLS<sup>A</sup> Genes

Genomic DNAs of S. sciuri isolate 100N carrying sal(A) and S. epidermidis isolate 138N carrying sal(A) and lsa(E) were submitted to high-throughput whole-genome sequencing (WGS), and preceded by library construction on a HiSeq 2500, which produced 150 bp paired-end reads (Berry Genomics Company, Beijing, China).

Draft assembly of the genomic DNA sequences was analyzed by CLC Genomics Workbench 5 (CLC Bio, Aarhus, Denmark). All contigs with the average coverage of >100-fold were searched for the PLS<sup>A</sup> genes using BLAST analysis. The regions flanking the PLS<sup>A</sup> genes were identified using de novo assembly as earlier described (Zerbino and Birney, 2008) and the random primer walking strategy (Zhang et al., 2009). Sequence analysis was conducted using the ORF Finder (http://www.ncbi.nlm.nih.gov/gorf/gorf.html) and BLAST functions (http://blast.ncbi.nlm.nih.gov/Blast.cgi). To obtain more information about the genetic environments of the PLS<sup>A</sup> genes in valnemulin-resistant isolates, the random primer walking strategy and inverse PCR were performed.

# Nucleotide Sequence Accession Numbers

The PLSA-carrying segments of various isolates in this study have been deposited in GenBank, and their accession numbers are KX712120 (Staphylococcus haemolyticus plasmid p131A carrying vga(A)LC), KX712121 (S. epidermidis plasmid p132R carrying vga(A)), KX712119 (S. epidermidis 138N carrying sal(A)), and KX712118 (S. epidermidis 138N carrying lsa(E)).

# RESULTS

# Characterization of Valnemulin-Resistant Staphylococcal Isolates and Antimicrobial Resistance Profiles

Amongst the 300 anal swabs and nasal samples, 13 (4.3%) valnemulin-resistant isolates with valnemulin MICs of ≥8 µg/ml were detected, which were identified to species level as Staphylococcus sciuri (n = 8), Staphylococcus haemolyticus (n = 2), S. epidermidis (n = 2), and Staphylococcus xylosus (n = 1) (**Table 1**).


TABLE 1 | Characteristics of the 13 valnemulin-resistant staphylococci isolates identified in this study.

<sup>a</sup>The PLS<sup>A</sup> resistance gene is located on the chromosome (C), on a plasmid (P), or both (P, C). <sup>b</sup>VAL, valnemulin; TIA, tiamulin; RET, retapamulin; LIN, lincomycin; VIR, virginiamycin M1; SPE, spectinomycin; GEN, gentamicin; ERY, erythromycin. The MICs that classify an isolate as resistant or high MICs (in cases where no applicable breakpoints are available to classify an isolate as resistant) are displayed in bold. <sup>c</sup>PFGE was performed only for ≥ two isolates of the same species.

Nine (eight S. sciuri and one S. xylosus) of the 13 isolates carried only sal(A), one S. haemolyticus and one S. epidermidis isolate carried vga(A)LC and vga(A), respectively, one S. haemolyticus isolate carried both vga(A)LC and sal(A), and one S. epidermidis isolate carried both lsa(E) and sal(A) (**Table 1**). Notably, the lsa(E) gene was only detected in one S. epidermidis isolate of dog origin, and the vga genes were detected in staphylococcal isolates originating from both humans and cats. In contrast, sal(A) was widespread among the valnemulin-resistant staphylococcal isolates from pets. As the sal(A) gene has been previously found exclusively in S. sciuri isolates (Hot et al., 2014; Wendlandt et al., 2015), this finding describes for the first time the occurrence of sal(A) in S. haemolyticus, S. epidermidis, and S. xylosus.

All 13 valnemulin-resistant isolates displayed resistance or high MICs to pleuromutilins (valnemulin, tiamulin, and retapamulin), lincomycin, virginiamycin M1 and spectinomycin (**Table 1**).

# Genotyping by PFGE

Using a cut-off of 90% similarity, the eight S. sciuri isolates, which carried highly conserved (>99% nucleotide identity) sal(A) genes, clustered into five PFGE types (A–E), with type A accounting for the majority (3/8) of the isolates (**Table 1**; **Figure 1**). Remarkably, the feline S. haemolyticus isolate 131A and the human S. haemolyticus isolate 131R from the corresponding pet owner, which carried identical copies of vga(A)LC, showed the same PFGE type H, indicating that they were closely related. This finding supports the idea that these isolates might have been exchanged between the cat and its owner.

# Genetic Environment of PLS<sup>A</sup> Genes

S1-PFGE and Southern blot hybridization indicated that all vgatype genes in the three isolates were located on plasmids, while all lsa(E) and sal(A) genes in the respective staphylococci were located in chromosomal DNA (**Table 1**; **Figure 2**).

To determine the regions flanking the vga-type genes in isolates 131R, 131A, and 132R, inverse PCRs were performed using the primers vga-F (5<sup>0</sup> -CAAGCTGAAAAGCCAACAAGG-3 0 ) and vga-R (5<sup>0</sup> -CCTCGTCAATTTCCCATATAGT-3<sup>0</sup> ), which are located inside the conserved regions of vga(A) and vga(A)LC genes. Three amplicons of 6056-bp (plasmids p131A and p131R) and 7209-bp (p132R) were obtained (**Figure 3A**). Plasmids p131A and p131R from the feline and human S. haemolyticus isolates differed in only nine base pairs, and showed 99% nucleotide sequence identity to plasmid pUR2355 from human S. aureus (accession no. JQ312422). Plasmid p132R shared 99% nucleotide identity with the S. epidermidis plasmid pUR3036 of cat origin (accession no. JQ312423). The most notable difference between p132R and p131A/p131R was the presence of 13 amino acid substitutions in the Vga protein. Surprisingly, the lincomycin MICs of isolates 131A, 131R, and 132R were 16, 16, and 64 µg/ml, respectively, which is in contrast to the finding that vga(A)LC confers higher MICs to lincosamides than does vga(A) (Novotna and Janata, 2006).

Genomic DNA of S. sciuri isolate 100N carrying sal(A) and S. epidermidis isolate 138N carrying sal(A) and lsa(E) were sequenced by WGS. Comparative analysis of the draft genomes of isolates 100N and 138N with the corresponding sequenced genomes of staphylococci from NCBI revealed the presence of a very similar 12.5-kb DNA segment (97% nucleotide identity) containing the sal(A) gene inserted between the housekeeping genes aspS and alaS of isolates 100N and 138N (**Figure 3B**). This DNA segment contains 10 ORFs, of which only sal(A) is an antimicrobial resistance gene. Additionally, two pairs of primers, aspS-F (5<sup>0</sup> -CGTTGTTGAAGATGGTGCGTCT-3 0 ), salA-R (5<sup>0</sup> -GGACCGAACCTTGAAATGATTG-3<sup>0</sup> ), and salA-F (5<sup>0</sup> -GATGGATACCTTATAGAAGGTG-3<sup>0</sup> ), alaS-R (50 -GTCTGTATCATAGTTCGTTGG-3<sup>0</sup> ), which are located

FIGURE 2 | Localization of pleuromutilins, lincosamides, and streptogramin A (PLSA) in the valnemulin-resistant staphylococcal isolates by S1 nuclease-PFGE (A) and Southern blot hybridization (B). M indicates the Low Range PFGE Marker. The black arrows indicate the location of PLS<sup>A</sup> gene in Southern blot hybridization and white arrows indicate the position of PLS<sup>A</sup> gene in genome according to the S1-PFGE and marker. 131A and 131R (lanes 1–2, respectively) were vga(A)LC-positive S. haemolyticus isolates. 132R (lane 3) was a vga(A)-positive S. epidermidis isolate. 138N (lane 4) was a lsa(E)-positive S. epidermidis isolate. 90A, 95N, 96A, 96N, 100N, 123N, 131A, 138N, 139N, 140N, and 145N (lanes 5–15, respectively) were sal(A)-positive staphylococcal isolates.

in the conserved regions of the aspS, sal(A), and alaS genes (**Figure 3B**), respectively, were designed, and long-range PCRs were preformed to investigate the genetic environments of sal(A) in the other nine sal(A)-positive staphylococcal isolates (**Table 1**). Two amplicons of 7.4 and 7.7 kb were obtained, and identical sequences were found in all nine isolates.

A 20,154-bp lsa(E)-carrying fragment containing 21 ORFs was obtained in the canine S. epidermidis isolate 138N (**Figure 3C**) after the conjoint analysis of the results of WGS and modified random primer walking strategy. Similar to the corresponding regions in porcine E. faecalis E533 (accession no. KX156278), this lsa(E)-carrying fragment also contained the macrolide-lincosamide-streptogramin B resistance gene erm(B), the aminoglycoside resistance genes aacA-aphD, aadE, and aphA-3, the spectinomycin resistance gene spw, the streptothricin resistance gene sat4, and the lincosamide resistance gene lnu(B). Further downstream of the lnu(B) gene, a second copy of the genes aadE-sat4-aphA-3 was detected (**Figure 3C**). This multidrug resistance gene region confers resistance to seven classes of antimicrobial agents and explained why isolate 138N exhibited resistance not only to PLS<sup>A</sup> antibiotics, but also to spectinomycin, gentamicin and erythromycin (**Table 1**). In addition, the hp4 gene encoded a predicted protein related to an acetyltransferase enzyme GNAT (GCN5-related N-acetyltransferase, with no confirmed function) from Enterococcus.

# DISCUSSION

Increasing emergence of staphylococci that are resistant to PLS<sup>A</sup> and are isolated from humans and pets is a growing public health concern worldwide. Each of these three classes of antimicrobial agents contains important drugs for human and veterinary medicine. The pleuromutilin antibiotics valnemulin and tiamulin are frequently used in veterinary medicine. Retapamulin was the first pleuromutilin approved for topical use in skin infections of humans caused by staphylococci (Jacobs, 2007). Lincosamides, such as clindamycin and lincomycin, have generally been indicated for the treatment of osteomyelitis but also skin and soft tissue infections caused by staphylococci in humans and pets (Xue et al., 1996; Zetner et al., 2003). Virginiamycin

Deng et al. PLS<sup>A</sup> Genes in Staphylococci from Pets

is a two-component streptogramin with a broad spectrum of activity against Gram-positive bacteria. Some antimicrobial agents approved for use in human medicine are also applied to nonfood-producing animals under specific regulations, such as the Animal Medicinal Drug Use Clarification Act (AMDUCA) in the USA. In this study, all 13 valnemulinresistant isolates displayed resistance or high MICs to pleuromutilins (valnemulin, tiamulin, and retapamulin), lincomycin, virginiamycin M1, indicating that all PLS<sup>A</sup> genes including vga(A), vga(A)LC, sal(A), and lsa(E) of these isolates were functional. Interestingly, genes sal(A) and vga(A)LC coexisted in one feline S. haemolyticus isolate 131A, and genes sal(A) and lsa(E) were co-located in the chromosomal DNA of the canine S. epidermidis isolate 138N. Isolate 138N exhibited significant differences in the MICs of valnemulin, lincomycin, and virginiamycin M1 compared with other isolates that carried only one type of PLS<sup>A</sup> genes (**Table 1**). There have been no studies regarding to sal, vga, and lsa coexisting in the same isolate, while both vga(A) and lsa(E) were identified in four bovine staphylococcal isolates (Wendlandt et al., 2015). Notably, vga(A)LC, sal(A), and lsa(E) were for the first time identified in staphylococcal isolates originating from pet animals.

The presence of PLS<sup>A</sup> resistance genes in pet- and humanassociated staphylococci indicates the adaptation of these bacteria to antibiotic pressure, given that PLS<sup>A</sup> resistance genes confer resistance to antimicrobial agents used in both pets and humans. Additionally, S. epidermidis 138N carried at least 13 resistance genes conferring resistance to seven classes of antimicrobial compounds (pleuromutilins, lincosamides, streptogramins, macrolides, aminoglycosides, aminocyclitols, and streptothricins), indicating that the use of any of the abovementioned antimicrobial agents may lead to co-selection of the MDR genomic island (MDRGI) in isolate 138N. The plasmids, chromosomal DNA region, and MDRGI associated with the multiresistance genes vga(A), vga(A)LC, sal(A), and lsa(E) exist in staphylococci isolated from pets and humans, which enhances the dissemination of PLSA-resistant staphylococci among pets and humans and presents significant challenges for the clinical management of infections by limited therapeutic options.

# REFERENCES


The three canine sal(A)-positive S. sciuri isolates 139N, 140N, and 145N not only shared the identical regions flanking the sal(A) gene, but also had the uniform PFGE pattern (subtype A), suggesting that these isolates have originated from a single clone. Remarkably, plasmids p131A and p131R, the two almost identical vga(A)LC-carrying plasmids present in isolates with the same PFGE subtype (subtype H), were derived from a cat and its owner. This fact suggests that these two isolates might have been exchanged between the cat and its owner due to their extensive contact. In this regard, it is important to consider the current role of dogs and cats as actual family members in many households and, consequently, the close contact to their owners and other family members (Schwarz et al., 2016).

# CONCLUSION

This is the first description of the PLS<sup>A</sup> genes in staphylococci of pet origin and also the description of a novel lsa(E)-carrying MDRGI in S. epidermidis. Moreover, we identified the sal(A) gene for the first time in S. haemolyticus, S. epidermidis, and S. xylosus. Pets are likely reservoirs of antimicrobial-resistant bacteria, warranting the prudent use of all antimicrobials in pet animal medicine.

# AUTHOR CONTRIBUTIONS

YW and SS designed research; FD, HW, and YL performed research; FD, JL, AF, and GM analyzed data; FD, SS, and YW wrote the paper. All authors listed have approved research for publication.

# ACKNOWLEDGMENTS

This study was funded by grants from National Natural Science Foundation of China (31422055 and 31370046), and German Federal Ministry of Education and Research (MedVet-Staph 2, grant no. 01KI1301D) provided through the German Aerospace Center.



**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Deng, Wang, Liao, Li, Feßler, Michael, Schwarz and Wang. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Molecular Mechanisms of Colistin Resistance in *Klebsiella pneumoniae* Causing Bacteremia from India—A First Report

Agila K. Pragasam<sup>1</sup> , Chaitra Shankar <sup>1</sup> , Balaji Veeraraghavan<sup>1</sup> \*, Indranil Biswas <sup>2</sup> , Laura E. B. Nabarro<sup>1</sup> , Francis Y. Inbanathan<sup>1</sup> , Biju George<sup>3</sup> and Santhosh Verghese<sup>4</sup>

# *Edited by:*

Axel Cloeckaert, French National Institute for Agricultural Research (INRA), France

#### *Reviewed by:*

Séamus Fanning, University College Dublin, Ireland Juan Wang, University College Dublin, Ireland Raffaele Zarrilli, University of Naples Federico II, Italy

> *\*Correspondence:* Balaji Veeraraghavan vbalaji@cmcvellore.ac.in

#### *Specialty section:*

This article was submitted to Antimicrobials, Resistance and Chemotherapy, a section of the journal Frontiers in Microbiology

*Received:* 11 October 2016 *Accepted:* 19 December 2016 *Published:* 09 January 2017

### *Citation:*

Pragasam AK, Shankar C, Veeraraghavan B, Biswas I, Nabarro LEB, Inbanathan FY, George B and Verghese S (2017) Molecular Mechanisms of Colistin Resistance in Klebsiella pneumoniae Causing Bacteremia from India—A First Report. Front. Microbiol. 7:2135. doi: 10.3389/fmicb.2016.02135

<sup>1</sup> Department of Clinical Microbiology, Christian Medical College, Vellore, India, <sup>2</sup> Department of Microbiology, Molecular Genetics and Immunology, University of Kansas Medical Centre, Kansas, KS, USA, <sup>3</sup> Department of Haematology, Christian Medical College, Vellore, India, <sup>4</sup> Department of Nephrology, Christian Medical College, Vellore, India

Colistin has long been a reserve drug used for the treatment of carbapenem resistant Klebsiella pneumoniae. Carbapenem resistance in K. pneumoniae has been increasing and is as high as 44% in India. Although a reserve agent, with rise in rates of resistance to carbapenems, the usage of colistin has increased over the years leading to slow emergence of resistance. Colistin resistance is mainly mediated by the alteration in the LPS of bacterial outer membrane with the addition of L-Ara4-N and PEtN molecules. These alterations are mediated by mutations in several genes involved in lipidA modifications and most commonly mutations in mgrB gene has been reported. Recently there is emergence of plasmid mediated resistance due to mcr-1 and mcr-2 genes which poses a threat for the rapid global spread. This study aims at characterizing eight colistin resistant K. pneumoniae from bacteremia by whole genome sequencing. Eight K. pneumoniae were isolated from blood culture during 2013 and 2014 at the Department of Clinical Microbiology, Christian Medical College, India. Antimicrobial susceptibility testing was performed and minimum inhibitory concentration (MIC) was determined for colistin and polymyxin B by broth-micro dilution method. Whole genome sequencing was performed using Ion Torrent and the genome of all eight isolates was analyzed. The eight isolates were resistant to all the antimicrobials expect tigecycline. MIC of colistin and polymyxin B were ranged from 4 to 1024µg/ml and 0.5 to 2048µg/ml respectively. Multiple mutations were observed in the chromosomal genes involved in lipid A modifications. mcr-1 and mcr-2 gene was absent in all the isolates. The most significant were mutations in mgrB gene. Among the eight isolates, four, three and one were belonged to sequence types ST 231, ST14 and ST147 respectively. Seven isolates had blaOXA−48 like, one co-expressed blaNDM−<sup>1</sup> and blaOXA−48 like genes leading to carbapenem resistance. Overall, multiple numbers of alterations have been observed. This includes silent mutations, point mutations, insertions and/or deletions. Mutations in mgrB gene is responsible for resistance to colistin in this study. Due to emergence of resistance to reserve drugs, there is a need for combination therapies for carbapenem resistant K. pneumoniae and colistin must be judiciously used.

Keywords: colistin, polymyxin B, resistance, *Klebsiella pneumoniae*, whole genome sequencing

# INTRODUCTION

Klebsiella pneumoniae is a gram negative bacteria and is a common cause of severe community and hospital acquired infections. It readily colonizes human mucosal surfaces especially the gastrointestinal (GI) tract and oropharynx (Bagley, 1985; Dao et al., 2014; Rock et al., 2014). In immunocompromised hosts, it can invade other sites leading to severe infections (Paczosa and Mecsas, 2016). In India, carbapenems are commonly used for treatment of K. pneumoniae due to high rates of ESBL producers (Ah et al., 2014). Increasing carbapenem resistance has resulted in colistin (polymyxin E) and tigecycline being used as last resort drugs for treatment of multidrug resistant (MDR) isolates (Carmeli et al., 2010).

Polymyxins are cationic polypeptides with five different compounds, known as polymyxin A–E (Storm et al., 1977). However, only polymyxin B and polymyxin E (colistin) are used in clinical practice to treat bacterial infections. When initially introduced, polymyxins were not widely used due to their adverse effects of nephrotoxicity and neurotoxicity (Falagas and Kasiakou, 2006). However, with the rise of carbapenem resistant organisms, polymyxins have gained attention as one of the few agents available for treatment (Falagas et al., 2005). Polymyxins are positively charged antimicrobial peptide molecules, which act by biding to negatively charged phosphate groups in the lipidA moiety of lipopolysaccharides (LPS). This causes disruption and loss of cell membrane integrity, leading to cell death (Yahav et al., 2012).

Due to its rising and sometimes inappropriate use, colistin resistance is increasing. Resistance to colistin is mainly mediated by modification of the LPS moiety with the addition of positively charged L-Ara4-N and PEtN molecules (Falagas et al., 2010). This modification decreases the negative charge of the outer membrane, reducing its interaction with colistin (Velkov et al., 2013). It is mainly due to mutations in the two component regulatory systems. The most common mutations are in mgrB, phoP/phoQ, pmrA, pmrB, pmrC, and crrABC (Cheng et al., 2010; Poirel et al., 2015; Wright et al., 2015).

Recently, plasmid mediated colistin resistance named mcr-1 (plasmid Mediated Colistin Resistance) encoding for phosphoethanolamine transferase has been reported in Escherichia coli and K. pneumoniae from China (Liu et al., 2016). Following this, mcr-1.2, a variant of mcr-1, was identified in KPC producing K. pneumoniae in Italy (Di Pilato et al., 2016) followed by the mcr-2 gene reported from Belgium, This had 76% identity with the mcr-1 gene previously reported (Xavier et al., 2016). The increasing number of resistance mechanisms reported in such a short time period is alarming.

The emergence of colistin resistance is a serious cause for concern for both clinicians and patients, particularly in countries with high rates of carbapenem resistant Enterobacteriaceae such as China, India, and Greece. Hence, it is necessary to perform surveillance studies for colistin resistance organisms in Indian hospitals. In this study, we report eight colistin resistant K. pneumoniae isolated from bloodstream infections from Christian Medical College, Vellore, India. Whole genome sequencing was employed to study the mechanisms of colistin resistance such as chromosomal genes and plasmid mediated mcr-1 and mcr-2 genes.

## MATERIALS AND METHODS

## Bacterial Strains

A total of eight clinical isolates of colistin resistant (CR) K. pneumoniae (CRKP1–CRKP8) were included in this study. These CR-K. pneumoniae were isolated in blood culture from patients admitted at Christian Medical College, Vellore, South India between 2013 and 2015. All the eight isolates were identified up-to species level as per standard biochemical tests (Versalovic et al., 2011) and by Vitek2 (bioMerieux)

# Demographic and Clinical Details of Patients

These were obtained from electronic medical records available in the hospital intranet.

## Antimicrobial Susceptibility Testing

The Kirby Bauer disk diffusion method was performed to ascertain antimicrobial susceptibility to commonly used agents. This included ceftazidime (30 µg), cefotaxime (30 µg), cefoxitin (30 µg), cefepime (30 µg), piperacillin/tazobactam (100/10 µg), imipenem (10 µg), meropenem (10 µg), ciprofloxacin (5 µg), gentamicin (10 µg), netilmicin (30 µg), amikacin (30 µg), polymyxin B (300 units), and tigecycline (15 µg). The susceptibility results were interpreted as per CLSI (2013, 2014, 2015) guidelines, respectively (CLSI M100S-23; CLSI M100S-24; CLSI M100S-25). Tigecycline results were interpreted according to FDA criteria (http://www.accessdata.fda.gov/drugsatfda\_docs/ label/2009/021821s016lbl.pdf).

The minimum inhibitory concentration (MICs) were determined by broth micro dilution method (BMD) for both polymyxin agents, colistin and polymyxin B. The results were interpreted as per EUCAST guidelines (EUCAST, 2013, 2014, 2015, respectively), (EUCAST susceptibility breakpoint v 3.0, v 4.0 and v 5.0). E. coli ATCC 25922 was used as a quality control.

# Carbapenemase Detection by CarbaNP and Multiplex PCR

The CarbaNP test was performed for all isolates by the previously described method (Nordmann et al., 2012). K. pneumoniae ATCC BAA 1705 and K. pneumoniae ATCC BAA 1706 were used as positive and negative controls, respectively.

For molecular testing, all the isolates were grown overnight on blood agar and DNA was extracted using Qiagen kit as per manufacturer's protocol (Qiagen Bacterial DNA mini kit). The presence of carbapenemase genes such as blaSPM, blaIMP, blaVIM, blaNDM, blaKPC, and blaOXA−48 like genes were determined by conventional multiplex PCR as described previously (Anandan et al., 2015). The products were analyzed on 2% agarose gel stained with ethidium bromide. Respective positive controls were used for each run (Courtesy: IHMA, Inc., USA).

# Molecular Characterization of Colistin Resistance

### PCR and Sequencing for LPS Modifications Genes

Genes involved in LPS modification including mgrB, phoP/phoQ, and pmrA/pmrB/pmrD were amplified and sequenced as previously described (Jayol et al., 2015; Wright et al., 2015).

## Whole Genome Sequencing (WGS) for CR- K. pneumoniae

WGS was performed using the Ion Torrent PGM platform using 400 bp read chemistry. Sequencing was performed as per the protocol recommended by Life Technologies. Raw reads were assembled de novo using Assembler SPAdes software v4.4.0.1 in Torrent suite server version 4.4.3. The genome was annotated using the PATRIC (Pathosystems Resource Integration Center -https://www.patricbrc.org) and RAST (Rapid Annotations using Subsystems Technology) databases. Upon annotation, antimicrobial resistance genes were found as per ARDB and CARD, respectively. These genes were further taken for mutational analysis. The whole genome sequences were deposited at GenBank under the following accession numbers: CRKP1 (MOXM00000000); CRKP2 (MIEJ00000000); CRKP3 (MPCT00000000); CRKP4 (MDZG00000000);

TABLE 1 | Minimum inhibitory concentration (MIC) for colistin and polymyxin B by broth micro dilution (BMD) for *K. pneumoniae.*


CRKP5 (MOXL00000000); CRKP6 (MEBR00000000); CRKP7 (MOXN00000000); CRKP8 (LZYN00000000).

# Molecular Typing of CR K. pneumoniae by MLST

The MLST database at the Center of Genomic Epidemiology was used to identify the sequence types (ST) of the study isolates using WGS data (https://cge.cbs.dtu.dk/services/MLST/).

### eBURST

The program eBURST v 3.0 was used to identify clonal complexes.

# RESULTS

# Demographic and Clinical Details of Patients

The demographic and clinical details of each patient were obtained from electronic medical records. The patients' ranged from 25 to 59 years in age. Five were male. Six patients were under the care of the hematology team, one was under the care of neurology, and one under nephrology. All but one patient was immunocompromised. K. pneumoniae sepsis was hospital acquired in all cases and the sources of infection were neutropenic sepsis (n = 4), pneumonia (n = 1), urosepsis (n = 1), intra-abdominal infection (n = 1), and line infection (n = 1). Meropenem was administered to all patients before K. pneumoniae was isolated and colistin was administered to five patients. Four patients expired within 30 days of infection and seven within 90 days of infection.

# Antimicrobial Susceptibility Testing

All isolates were extremely drug resistant (XDR) as per the definition of Magiorakos et al. (2012) All isolates were resistant to cephalosporins, carbapenems, fluoroquinolones, aminoglycosides, and polymyxin B by disk diffusion but were susceptible to tigecycline. The MIC-values determined by BMD for colistin and polymyxin B ranged from 4 to 1024 and 0.5 to 2048 µg/ml, respectively. The MIC results of the eight isolates are summarized in **Table 1**.

# Carbapenemase Detection

CarbaNP test was positive for all the eight isolates tested indicating the production of carbapenemases. Multiplex PCR for carbapenamase genes revealed the presence of blaOXA−48 like gene in seven isolates. One isolate co-expressed blaNDM−<sup>1</sup> <sup>+</sup> OXA−<sup>232</sup> genes.

# Whole Genome Sequence Analysis

For whole genome sequencing analysis, PATRIC, and RAST databases have been used to analyse the annotated sequences (http://patricbrc.org/) (http://rast.nmpdr.org/). As these eight isolates were XDR strains, multiple resistance genes against various classes of antimicrobial agents have been identified. These are summarized in **Table 2**. For cephalosporin resistance, the most common resistance genes were blaTEM, blaSHV, and blaCTX−M−15. For carbapenem resistance, blaOXA−232, a variant of blaOXA−48 like gene was found in six of the study isolates, one


2|AntimicrobialresistancegenesidentifiedagainstvariousantimicrobialagentsinCR




isolate co-producing blaOXA−<sup>232</sup> and blaNDM−1, while one isolate had blaOXA−181, also a variant of the blaOXA−48 like gene.

For colistin resistance, RAST classified 11 genes involved in lipid A modification with L-Ara4N pathway. These included ugd, arnA\_DH, arnA\_FT, arnB, arnC, arnT, pmrJ, pmrL, pmrD, pmrM, and pmrG. In addition, 10 genes were classified under genes involved in Lipid A modifications. These include pagP,pagL, lpxO, phoQ, phoP, pmrA, pmrB, eptA, eptB, and yijP. Of these 21 genes, a few gene sequences could not be retrieved from the annotated genomes of the study isolates from RAST/PATRIC databases and hence analysis was not done. However, for the available genes, sequences were retrieved and analyzed for mutations. Upon analysis, multiple mutations were observed in various genes. In seven isolates, there was a change from arginine to alanine at position 114 of phoP. In one isolate, there was a change from arginine to alanine in position 128. In phoQ, five isolates had a substitution of glycine for aspartic acid at position 146 and three isolates had the same substitution at position 150. The mutations in other genes such as pmrB, pmrC, phoB, arnB, arnC, arnT, arnA\_FT, and pagP are listed in **Table 3** and have not been reported in other studies to date. No mutations were found in pmrA, eptB, and phoR genes in any of the isolates analyzed. mcr-1 and mcr-2 genes were absent in all the study isolates.

# Targeted Gene Sequencing Analysis of Mutations in Genes Involved in LPS Modifications

Of the eight isolates screened for mutations, CRKP3 and CRKP6 had a premature stop codon in mgrB gene resulting in a truncated protein of 27 amino acids. Whereas, a frame shift mutation in CRKP4 and deletion of nucleotide A at 10th position of mgrB gene in CRKP7 was noted, this could be responsible for the resistant phenotype. Three isolates, CRKP1, CRKP2, and CRKP5 did not have any mutations in the mgrB gene. Sequencing of two component systems phoP/phoQ and pmrA/pmrB and pmrD found multiple silent mutations. Since targeted sequencing results did not reveal much information, WGS data analysis was performed to further understand the additional resistance mechanisms.

# Molecular Typing of CR *K. pneumoniae* by MLST

MLST database available at https://cge.cbs.dtu.dk/services/ MLST/ by the Center of Genomic Epidemiology was used to identify the sequence type (ST) of the study isolates using WGS data. Of the eight CR-K. pneumoniae, three different sequence types were identified (ST231, ST14, and ST147). Four isolates belonged to ST231 (CRKP3, CRKP4, CRKP6, and CRKP7), three to ST14 (CRKP1, CRKP2, and CRKP5) and one to ST147 (CRKP8).

## eBURST

eBURST analysis found no clustering of colistin resistant isolates into a clonal complex. The result is depicted in **Figure 1**.

# DISCUSSION

Antimicrobial resistance is on the rise globally. The limited treatment options available for carbapenem resistant K. pneumoniae have resulted in increasing use of colistin, despite its adverse effects. Colistin resistance is now emerging, leaving clinicians with very few options to treat these extensively drug resistant bacteria.

The main challenge in screening colistin resistance lies in determining polymyxin susceptibility (Balaji et al., 2011). Polymyxins are large cationic peptide molecules. They do not diffuse through agar well, hence disk diffusion is neither reliable nor recommended. A reliable screening test is awaited. Other susceptibility methods produce variable and discrepant results. Further, polymyxin B and colistin produce discrepant results (Hindler and Humphries, 2013; Humphries, 2015). Colistin susceptibility testing by disk diffusion has resulted in very major errors of up to 32% when compared with broth or agar dilution methods for colistin (Tan and Ng, 2006; van der Heijden et al., 2007). E-test may also produce discrepant results, hence broth microdilution is considered the reference method (Tan et al., 2007).

From this study results, the MIC was found to be two to eight fold higher for polymyxin B when compared to colistin by BMD method. Such discordant results could be due to the nonreliability of MIC testing for polymyxin B. These findings concur with previous studies where the MIC was found to be higher for polymyxin B than colistin (van der Heijden et al., 2007).

A number of studies have reported discrepant results when testing polymyxin B and colistin susceptibility by disk diffusion and MIC methods (Behera et al., 2010). Similar observations were noted in this study; all eight CR-K. pneumoniae were found to be resistant on BMD testing with variable MIC values for polymyxin B and colistin as shown in **Table 1**. Detection of molecular resistance mechanisms are based on preliminary screening with polymyxins. This emphasizes the need for a reliable screening test.

mgrB is the commonest chromosomal gene mutation involving modification of LPS which includes insertional inactivation causing colistin resistance. The second commonest cause is multiple mutations in the pmrAB and phoPQ, the two component systems involved in the LPS modifications.

Of the genes analyzed, mutations in mgrB with truncation to 27 amino acids concurs with previous studies (Poirel et al., 2015). The T157P mutation in pmrB has also been previously identified (Jayol et al., 2014, 2015; Olaitan et al., 2014). Multiple new mutations were identified in various genes in this study. These include mutations in phoP (R114A), phoQ (D146G), pmrB (L334P, A246T), pmrC (V39L, R152H, A279G, D477N, D480N, S260L, C27F, V50L, A235P), phoR (R69C), arnB (A112D, D285E, G47D, I126V), arnC (S30T), arnA\_FT (I260L, N442K, L161C, S18A, T185A), and pagP (F170I). However, the significance of these mutations and their contribution to colistin resistance needs to be studied further.

One isolate (CRKP8) had very high MICs of 1024 µg/ml and 2048µg/ml by BMD for colistin and polymyxin B respectively. This is a huge variation in MIC between two different methodologies.

The clonality distribution of colistin resistant K. pneumoniae is of great interest. Worldwide, sequence types (ST) associated with colistin resistant K. pneumoniae are ST258, ST512, and ST147 (Mammina et al., 2012; Cannatelli et al., 2014). However, in our study it was ST231, ST14 and ST147. Our study agrees with Jayol et al who reported ST14 K. pneumoniae producing blaOXA−48 like (Jayol et al., 2014, 2015). However, in this study, the resistance mechanisms were mgrB mediated rather than due to mutation in pmrB. One of the study isolates belonging to ST14 (CRKP6) was a blaNDM−<sup>1</sup> producer. Isolates with mutated and/or truncated mgrB were of different sequence types. This further indicates that resistance determinants are irrespective of the sequence types.

Interestingly, isolates CRKP3 and CRKP6, both of which had a mutation in mgrB gene resulting in a truncated product, were not of the same clone. However, isolate CRKP4 and CRKP7 which had a frame shift mutation in mgrB gene, both belonged to ST231.

# CONCLUSION

Inappropriate usage of colistin in the clinical setting should be avoided, as selection pressure may contribute to colistin resistance due to mutations. Further, use of colistin in agricultural practices should be forbidden to prevent the further spread of resistance. The global demand for colistin in agriculture is estimated to reach 16,500 tons by 2021 (QYResearch Medical Research Centre, 2015). India is also one of the leading producers of colistin for veterinary use (Liu et al., 2016). In this study, multiple mutations in the genes coding for LPS modification have been observed. These include silent mutations, point mutations, insertions, and/or deletions. Mutation profiles were observed to be divergent. This signifies the fact that clinical isolates do develop mutations for survival. Altogether, mutations in the mgrB gene were found to be the most common, followed by mutations in the other genes such as pmrB, pmrC, arnB, and

# REFERENCES


arnAFT. This observation emphasizes that colistin resistance is multifactorial. However, the significant effect of these mutations observed in this study needs to be validated. Therefore, further studies should be focused on the role of these genes in pathways involving LPS modifications conferring resistance to colistin.

# ETHICS STATEMENT

Ethical approval was obtained for the study from the Ethical Committee of Christian Medical College, Vellore, India.

# AUTHOR CONTRIBUTIONS

AP, CS, FI: lab methods, data analysis, manuscript writing. BV, IB: study design and manuscript writing. LN, BG, SV: study design and data collection.

# FUNDING

Fluid Research Grant of Christian Medical College, Vellore, India.


and molecular biological study. Lancet Infect. Dis. 16, 161–168. doi: 10.1016/S1473-3099(15)00424-7


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Pragasam, Shankar, Veeraraghavan, Biswas, Nabarro, Inbanathan, George and Verghese. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

fmicb-08-00914 May 24, 2017 Time: 14:54 # 1

# Mutations of the Transporter Proteins GlpT and UhpT Confer Fosfomycin Resistance in Staphylococcus aureus

Su Xu1,2, Zhuyingjie Fu1,2, Ying Zhou1,2, Yang Liu1,2 \*, Xiaogang Xu1,2 \* and Minggui Wang1,2

1 Institute of Antibiotics, Huashan Hospital, Fudan University, Shanghai, China, <sup>2</sup> Key Laboratory of Clinical Pharmacology of Antibiotics, National Health and Family Planning Commission, Shanghai, China

#### Edited by:

Axel Cloeckaert, Institut National de la Recherche Agronomique (INRA), France

#### Reviewed by:

Makoto Kuroda, National Institute of Infectious Diseases, Japan Fangyou Yu, First Affiliated Hospital of Wenzhou Medical University, China

#### \*Correspondence:

Xiaogang Xu xuxiaogang@fudan.edu.cn Yang Liu liuyang@fudan.edu.cn

#### Specialty section:

This article was submitted to Antimicrobials, Resistance and Chemotherapy, a section of the journal Frontiers in Microbiology

Received: 01 December 2016 Accepted: 04 May 2017 Published: 19 May 2017

#### Citation:

Xu S, Fu Z, Zhou Y, Liu Y, Xu X and Wang M (2017) Mutations of the Transporter Proteins GlpT and UhpT Confer Fosfomycin Resistance in Staphylococcus aureus. Front. Microbiol. 8:914. doi: 10.3389/fmicb.2017.00914 With the increasing spread of methicillin-resistant Staphylococcus aureus worldwide, fosfomycin has begun to be used more often, either alone or in combination with other antibiotics, for treating methicillin-resistant S. aureus infections, resulting in the emergence of fosfomycin-resistant strains. Fosfomycin resistance is reported to be mediated by fosfomycin-modifying enzymes (FosA, FosB, FosC, and FosX) and mutations of the target enzyme MurA or the membrane transporter proteins UhpT and GlpT. Our previous studies indicated that the fos genes might not the major fosfomycin resistance mechanism in S. aureus, whereas mutations of glpT and uhpT seemed to be more related to fosfomycin resistance. However, the precise role of these two genes in S. aureus fosfomycin resistance remains unclear. The aim of the present study was to investigate the role of glpT and uhpT in S. aureus fosfomycin resistance. Homologous recombination was used to knockout the uhpT and glpT genes in S. aureus Newman. Gene complementation was generated by the plasmid pRB473 carrying these two genes. The fosfomycin minimal inhibitory concentration (MIC) of the strains was measured by the E-test to observe the influence of gene deletion on antibiotic susceptibility. In addition, growth curves were constructed to determine whether the mutations have a significant influence on bacterial growth. Deletion of uhpT, glpT, and both of them led to increased fosfomycin MIC 0.5 µg/ml to 32 µg/ml, 4 µg/ml, and >1024 µg/ml, respectively. By complementing uhpT and glpT into the deletion mutants, the fosfomycin MIC decreased from 32 to 0.5 µg/ml and from 4 to 0.25 µg/ml, respectively. Moreover, the transporter gene-deleted strains showed no obvious difference in growth curves compared to the parental strain. In summary, our study strongly suggests that mutations of uhpT and glpT lead to fosfomycin resistance in S. aureus, and that uhpT mutation may play a more important role. The high resistance and low biological fitness cost resulting from uhpT and glpT deletion suggest that these strains might have an evolutionary advantage in a fosfomycin-rich clinical situation, which should be closely monitored.

Keywords: Staphylococcus aureus, fosfomycin, resistance, membrane transporter, glpT, uhpT

# INTRODUCTION

fmicb-08-00914 May 24, 2017 Time: 14:54 # 2

Staphylococcus aureus is one of the most common bacterial pathogens worldwide in both community and hospital settings. Methicillin-resistant Staphylococcus aureus (MRSA) is an important multi-resistant pathogen. To date, vancomycin has remained the cornerstone drug in the management of invasive MRSA infections. However, the continuous rise in the vancomycin minimum inhibitory concentration (MIC), known as the "vancomycin MIC creep" phenomenon, poses a significant challenge to MRSA therapy; therefore, fosfomycin has recently been used alone or in combination with other antibiotics in treating MRSA infections (del Rio et al., 2016; VanEperen and Segreti, 2016). Nevertheless, this situation has inevitably led to the emergence of fosfomycin-resistant MRSA strains. In a recent review, the susceptibility of S. aureus to fosfomycin ranged between 33.2 and 100% in the nine available studies (frequency = 91.7%, 95% confidence interval 88.7–94.9%); in seven of the studies susceptibility rate was >90% (Vardakas et al., 2016). According to the CHINET surveillance program in China in 2010, 29.5% of the MRSA clinical isolates were resistant to fosfomycin (Guo et al., 2013). And Yu et al. (2010) reported a fosfomycin susceptible rate of 33.2%.

The mechanisms of action and resistance of fosfomycin have been studied for decades. Fosfomycin was first discovered in 1969 as an effective bactericidal agent against Grampositive and Gram-negative organisms. The mechanism of action of fosfomycin differs from that of most commonly used antimicrobials. In general, fosfomycin is transported across the bacterial wall primarily with the help of the glycerol-3-phosphate (G-3-P) transport (GlpT) system. In the presence of glucose-6-phosphate (G-6-P), the hexose phosphate uptake transport (UhpT) system is induced, and provides an alternative route to the GlpT system. UhpT are important membrane transporter proteins for small molecules, including fosfomycin (Castaneda-Garcia et al., 2009). When transported into the cytosol of a bacterium, fosfomycin deactivates the target protein UDP-N-acetylglucosamine-3 enolpyruvyltransferase (MurA, encoded by the murA gene), thereby preventing the formation of N-acetylmuramic acid from N-acetylglucosamine and phosphoenolpyruvate, which is the initial step in peptidoglycan chain formation of the bacterial wall (Kahan et al., 1974). The key resistance mechanisms to fosfomycin include the loss or reduced production of transporters, reduced affinity to MurA, and production of fosfomycin-modifying enzymes (Sastry and Doi, 2016).

However, to date, the mechanisms contributing to fosfomycin resistance have been mostly studied in Gram-negative bacteria, with few related studies on Gram-positive bacteria. We have conducted several studies to investigate the fosfomycin resistance mechanisms in Gram-positive cocci, including Enterococcus faecium and S. aureus (Xu et al., 2013; Chen et al., 2014; Fu et al., 2016a,b). These previous studies indicated that the fos gene was not the major mechanism of fosfomycin resistance in MRSA isolates from our hospital, whereas mutations of glpT and uhpT seemed to be more closely related to fosfomycin resistance. However, the exact roles of these two genes in S. aureus fosfomycin resistance remain unclear. Thus, we designed the present study to investigate the roles of glpT and uhpT in S. aureus fosfomycin resistance.

# MATERIALS AND METHODS

# Bacterial Strains and Plasmids

The strains and plasmids used in this study are presented in **Table 1**. The clinical fosfomycin-resistant MRSA strains SA2, SA94, and SA30 were collected from the blood or cerebral spinal fluid of patients at Huashan Hospital and were characterized previously (Fu et al., 2016b). And the strain names are in accordance with that in the **Supplementary Table S1** of the previous article (Fu et al., 2016b). Each of the clinical MRSA strains was with a different type of transporter gene mutation (**Table 1**). The S. aureus strains Newman and RN4220, and the plasmid pKOR1 were used in the homologous recombination assay (Bae and Schneewind, 2005; Wang et al., 2015). In addition, S. aureus ATCC29213 (American Type Tissue Culture Collection, Manassas, VA, United States) was used for the quality control of susceptibility testing. These strains and plasmids were laboratory collection.

# Allelic Gene Deletion by Homologous Recombination

Knockout of the transporter genes glpT and uhpT was conducted as previously described (Bae and Schneewind, 2005; Wang et al., 2015). The plasmids and primers used are listed in **Tables 1**, **2**, respectively. Proper gene deletion was verified by analytical polymerase chain reaction (PCR) and sequencing of the genomic DNA at the borders of the PCR-derived regions. Sequencing was then performed to confirm the nucleotides. The amplified fragments were used to construct the homologous recombinant pKOR1-1uhpT/glpT with Gateway <sup>R</sup> BP ClonaseTM II Enzyme mix (Thermo Fisher Scientific, Waltham, MA, United States).

pKOR1-1uhpT and pKOR1-1glpT were introduced into S. aureus RN4220 by electroporation for modification. The plasmid extracted from strain RN4220 was then introduced into S. aureus Newman. The desired uhpT and glpT deletion mutants were selected as described previously (Bae and Schneewind, 2005).

The successful generation of the Newman-1uhpT and Newman-1glpT strains was further confirmed by PCR and sequencing. PCRs were performed using the primers attB1-uhpTup-F/attB2-uhpT-CF and attB1-glpT-up-F/attB2-glpT-CF in the strains S. aureus Newman, Newman-1uhpT, and Newman-1glpT, respectively.

# Construction of the Complemented Strain

Fragments were PCR-amplified from S. aureus Newman using the primers C-uhpT-F/R and C-glpT-F/R. The PCR products and vector pRB473 were double-digested with the designed restriction enzymes BamHI and EcoRI (for uhpT), or BamHI and KpnI (for glpT), and ligation was performed with T4 fmicb-08-00914 May 24, 2017 Time: 14:54 # 3

ligase. The resulting plasmids were transferred into S. aureus RN4220, and then introduced into the deletion and clinical strains with defects on uhpT and/or glpT, SA2, SA94, and SA30.

# Antimicrobial Susceptibility Testing

Fosfomycin susceptibility of the knockout and clinical strains with defects on uhpT and/or glpT, and their complemented strains were tested with the E-test (BioMerieux SA, La Balme Les Grotts, France), according to the manufacturer's guidance. Results were interpreted according to European committee on antimicrobial susceptibility testing criteria (European Committee on Antimicrobial Susceptibility Testing [EUCAST], 2017) (susceptible, ≤32 mg/L; resistant, ≥64 mg/L).

# Measurement of Growth Curves

To evaluate the influence of deletion of the transporter genes on bacterial growth, we measured the in vitro growth curves of S. aureus Newman, Newman-1uhpT, Newman-1glpT, Newman-1uhpT&glpT, and the clinical strains. The strains were cultivated in tryptic soy broth overnight at 37◦C. The bacterial solution was diluted to an optical density at 600 nm (OD600) of 0.1 and cultivated again. The OD<sup>600</sup> was then measured at 0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, and 24 h for each strain, by


### TABLE 2 | Primers for PCR and sequencing.


fmicb-08-00914 May 24, 2017 Time: 14:54 # 4

spectrophotometer (UNICO, Shanghai, China). The procedure was repeated three times for each strain, and the mean OD<sup>600</sup> values were used to draw the growth curves.

# Phenotype Microarray (PM) Analysis

Phenotype Microarray analysis was performed using BIOLOG Phenotype MicroarrayTM (BIOLOG, Hayward, CA, United States) according to the manufacturer's recommendations. The deletion mutants, namely Newman-1uhpT, Newman-1glpT, Newman-1uhpT&glpT, and the parental strain Newman, were tested with the 96-wells plates PM1 and PM2, containing 190 carbon substrates, including G-6-P (PM1 plate, well C1). To assess the altered phenotypes in carbon metabolism of the deletion mutants, the growth was compared to the parent S. aureus Newman.

# RESULTS

The deletion mutants showed considerably increased MIC values to fosfomycin compared to that of the parental strain. The Newman-1uhpT&glpT strain, in which both transporter genes were knocked out, showed high-level resistance (MIC > 1024 µg/ml) to fosfomycin, as determined by the E-test (**Table 3**). When uhpT or glpT was knocked out from S. aureus Newman, the fosfomycin MICs increased from 0.5 to 32 µg/ml or 4 µg/ml, respectively.

Complementing uhpT and glpT led to a reduced fosfomycin MIC in the deletion mutants and the clinical fosfomycinresistant S. aureus strains with defects at both sites. By complementing plasmid pRB473-uhpT into Newman-1uhpT, the strain's fosfomycin MIC decreased from 32 to 0.5 µg/ml. Similarly, by complementing glpT into Newman-1glpT, strain's fosfomycin MIC decreased from 4 to 0.25 µg/ml. S. aureus SA2, SA94, and SA30 were clinical fosfomycin-resistant strains, with mutations of both uhpT and glpT, uhpT only, and glpT only, respectively. When complemented with the functional

TABLE 3 | Fosfomycin MIC (µg/ml) of S. aureus mutant strains and complemented strains.


transporter genes, the fosfomycin MICs decreased considerably, as shown in **Table 3**.

In vitro bacterial growth curves of the wild-type strain S. aureus Newman and the deletion mutants were compared to evaluate the potential fitness cost of these resistant-conferring mutations. As shown in **Figure 1**, no significant depression in growth was observed in Newman-1uhpT and Newman-1glpT compared to the wild-type strain. However, the strain Newman-1uhpT&glpT presented slight growth inhibition compared to the wild-type.

Phenotype Microarray analysis was performed using carbon utilization panels, PM1 and PM2, in 190 carbon substrates. The changes in carbon metabolism were listed in **Figure 2**. S. aureus Newman showed metabolic advantage over Newman-1uhpT and Newman-1uhpT&glpT in wells containing G-6-P (**Figure 2**, PM1, well C1). G-3-P was not included in the substrate list, and there was no obvious change found in Newman-1glpT.

# DISCUSSION

Intravenous fosfomycin is broadly used in the treatment of multidrug-resistant pathogens in Europe and Asia owing to its unique antibiotic mechanism, high permeability, and high susceptibility rate (Falagas et al., 2009). Falagas et al. (2010) performed fosfomycin susceptibility testing in non-urinary MRSA isolates, among which 99.2% (129/130) were found to be susceptible. The same group reviewed the susceptibility data of Gram-positive cocci, and reported a cumulative susceptibility rate of 87.9% (4240/4892) in S. aureus (Falagas et al., 2009).

Fosfomycin-resistance mechanism has been well described for Gram-negative bacteria such as E. coli (Kim et al., 1996; Horii et al., 1999; Huang et al., 2003; Takahata et al., 2010). In E. coli, GlpT and UhpT are responsible for fosfomycin uptake. Mutations or insertional inactivation in the glpT and/or uhpT genes or their regulatory genes lead to the loss of function of the transporters and fosfomycin resistance. The inactivation of either uhpT or glpT conferred a moderate fosfomycin resistance, (MICs increased from 2 to 8 µg/ml and 32 µg/ml, respectively, compared to the wild type) (Takahata et al., 2010). In P. aeruginosa, the inactivation of glpT produced significant decrease in fosfomycin MIC, from 8 to 1024 µg/ml (Castaneda-Garcia et al., 2009; Takahata et al., 2010). Modification or overexpression of murA, production of fosfomycin-modifying enzymes, are also associated with fosfomycin resistance (Garcia et al., 1995; Kim et al., 1996; Bernat et al., 1997; Horii et al., 1999; Fillgrove et al., 2003; Roberts et al., 2013).

There is less known of fosfomycin resistance mechanism in Gram-positive cocci. In previous works, we collected MRSA clinical strains, and found that only the minority of the fosfomycin-resistant MRSA strains carried the fos gene or murA mutation, while glpT and uhpT mutations were common (82.1%, 55/67, vs. 77.6%, 52/67, respectively) (Fu et al., 2016a,b). This fact indicated that, fosB or murA mutation is not the major contributor to fosfomycin resistance in MRSA, while mutations within the glpT and/or uhpT genes might play an important

FIGURE 1 | The in vitro growth curves of S. aureus strains. The strains were cultivated in tryptic soy broth overnight at 37◦C. The bacterial solution was diluted to an optical density at 600 nm (OD600) of 0.1 and cultivated again. The OD<sup>600</sup> was then measured at 0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, and 24 h to draw the curves. (A) Newman (◆), Newman-1uhpT (), Newman-1glpT (N), and Newman-1uhpT&glpT (×). (B) Newman (◆) and the clinical uhpT/glpT mutants: SA2 (N), with mutations of both glpT and uhpT; SA94 (), with mutation of uhpT; SA30 (×), with mutation of glpT.

role in S. aureus fosfomycin resistance. In the present study, we established uhpT and/or glpT deletion mutants. Knocking out both genes resulted in high-level fosfomycin resistance (MIC > 1024 µg/ml). Complementing either of the two genes into the deletion mutants and clinical mutated strains resulted in a decreased fosfomycin MIC. Direct comparison of uhpT and glpT according to the level of increase of the MIC suggested that uhpT has a greater effect on the strain's MIC to fosfomycin than glpT.

To evaluate the possible, the in vitro fitness cost of the transporter gene mutation, we compared growth curves between the fosfomycin-sensitive wild-type strain, laboratory deletion mutant strain, and clinical strains with defects on uhpT and/or glpT. Previous reports have shown that mutations of uhpT and glpT can compromise the growth of strains of E. coli, Klebsiella pneumoniae, and Proteus mirabilis (Li Pira et al., 1987; Marchese et al., 2003). A probable mechanism might be that mutations of the glpT and/or uhpT transporting systems prevent carbon

fmicb-08-00914 May 24, 2017 Time: 14:54 # 5

fmicb-08-00914 May 24, 2017 Time: 14:54 # 6

source getting into the cytoplasm, and therefore disturb cell metabolism (Venkateswaran and Wu, 1972; Nilsson et al., 2003). But in P. aeruginosa, the glpT mutation was found to lead to fosfomycin resistance with no obvious fitness cost (Castaneda-Garcia et al., 2009). In the present study, there was only a slight reduction of growth observed in the strain Newman-1uhpT&glpT compared to the wild type, and no significant growth suppression was observed either in the laboratory deletion mutant strains or in clinical strains with defects on uhpT and/or glpT, which is similar as observed in P. aeruginosa. So S. aureus might also compensate the disadvantage in energy obtainment caused by uhpT and/or glpT mutation through other transporting systems. But further study is still in need for verification.

We observed that G-6-P utilization was defected in both Newman-1uhpT and Newman-1uhpT&glpT. UhpT is the membrane transporter of this substrate, deletion mutants of uhpT showed defects in G-6-P metabolism is as expected. The G-3-P metabolism in S. aureus seems to be more complicated. G-3- P seems to be an intermediate product in carbon/phosphorus metabolism pathway. As another low G+C Gram-positive bacteria, B. subtilis shares similar carbon metabolism pattern as S. aureus. In B. subtilis, G-3-P is produced from glycerol with glycerol kinase. And G-3-P dehydrogenase can oxidize G-3-P to dihydroxyacetone phosphate, an intermediated in glycolysis (Holmberg et al., 1990). We have not observed significant change in metabolism. This may be because that G-3-P utilization defect is easily compensated by other pathways.

In summary, the results of our study strongly suggest that mutations of uhpT and glpT lead to fosfomycin resistance in S. aureus, and that the uhpT mutation may play a more important role. The high resistance and low fitness cost resulting from uhpT and glpT mutations suggest that these mutated strains might have an evolutionary advantage in a fosfomycin-rich clinical situation.

# REFERENCES


The widely observed uhpT or glpT mutation in S. aureus might be a threat in hospital settings. Further studies are needed to evaluate the frequency of S. aureus fosfomycin mutants, and virulence of these mutants.

# AUTHOR CONTRIBUTIONS

Designed and conceived the experiments: YL, XX, and MW. Performed the experiments: SX, ZF, and YZ. Analyzed the data: SX, ZF, and YZ. Wrote and reviewed the manuscript: SX, ZF, XX, and YL.

# FUNDING

This study was supported by grants from the National Natural Science Foundation of China (81171613 to XX and 81120108024 to MW).

# ACKNOWLEDGMENT

We thank Professor M. Li, Renji Hospital, Shanghai Jiaotong University, for providing strain S. aureus RN4220 and plasmid pKOR1.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2017.00914/full#supplementary-material

### TABLE S1 | Carbon sources tested in PM1 and PM2 plates.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Xu, Fu, Zhou, Liu, Xu and Wang. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

fmicb-08-00914 May 24, 2017 Time: 14:54 # 7

# Antimicrobial Susceptibility Profiles of Human Campylobacter jejuni Isolates and Association with Phylogenetic Lineages

Wonhee Cha<sup>1</sup> , Rebekah Mosci<sup>1</sup> , Samantha L. Wengert<sup>1</sup> , Pallavi Singh<sup>1</sup> , Duane W. Newton<sup>2</sup> , Hossein Salimnia3,4, Paul Lephart<sup>4</sup> , Walid Khalife<sup>5</sup> , Linda S. Mansfield1,6, James T. Rudrik<sup>7</sup> and Shannon D. Manning<sup>1</sup> \*

<sup>1</sup> Departments of Microbiology and Molecular Genetics, Michigan State University, East Lansing, MI, USA, <sup>2</sup> Department of Pathology, University of Michigan, Ann Arbor, MI, USA, <sup>3</sup> School of Medicine, Wayne State University, Detroit, Michigan, USA, <sup>4</sup> Detroit Medical Center University Laboratories, Detroit, MI, USA, <sup>5</sup> Sparrow Hospital, Lansing, MI, USA, <sup>6</sup> Departments of Large Animal Clinical Sciences, Michigan State University, East Lansing, MI, USA, <sup>7</sup> Bureau of Laboratories, Michigan Department of Health and Human Services, Lansing, MI, USA

### Edited by:

Benoit Doublet, National Institute for Agricultural Research, France

#### Reviewed by:

Catherine M. Logue, Iowa State University, USA Séamus Fanning, University College Dublin, Ireland

> \*Correspondence: Shannon D. Manning mannin71@msu.edu

#### Specialty section:

This article was submitted to Antimicrobials, Resistance and Chemotherapy, a section of the journal Frontiers in Microbiology

Received: 11 March 2016 Accepted: 11 April 2016 Published: 26 April 2016

#### Citation:

Cha W, Mosci R, Wengert SL, Singh P, Newton DW, Salimnia H, Lephart P, Khalife W, Mansfield LS, Rudrik JT and Manning SD (2016) Antimicrobial Susceptibility Profiles of Human Campylobacter jejuni Isolates and Association with Phylogenetic Lineages. Front. Microbiol. 7:589. doi: 10.3389/fmicb.2016.00589 Campylobacter jejuni is a zoonotic pathogen and the most common bacterial cause of human gastroenteritis worldwide. With the increase of antibiotic resistance to fluoroquinolones and macrolides, the drugs of choice for treatment, C. jejuni was recently classified as a serious antimicrobial resistant threat. Here, we characterized 94 C. jejuni isolates collected from patients at four Michigan hospitals in 2011 and 2012 to determine the frequency of resistance and association with phylogenetic lineages. The prevalence of resistance to fluoroquinolones (19.1%) and macrolides (2.1%) in this subset of C. jejuni isolates from Michigan was similar to national reports. High frequencies of fluoroquinolone-resistant C. jejuni isolates, however, were recovered from patients with a history of foreign travel. A high proportion of these resistant isolates were classified as multilocus sequence type (ST)-464, a fluoroquinolone-resistant lineage that recently emerged in Europe. A significantly higher prevalence of tetracycline-resistant C. jejuni was also found in Michigan and resistant isolates were more likely to represent ST-982, which has been previously recovered from ruminants and the environment in the U.S. Notably, patients with tetracycline-resistant C. jejuni infections were more likely to have contact with cattle. These outcomes prompt the need to monitor the dissemination and diversification of imported fluoroquinolone-resistant C. jejuni strains and to investigate the molecular epidemiology of C. jejuni recovered from cattle and farm environments to guide mitigation strategies.

Keywords: Campylobacter jejuni, antimicrobial resistance, multilocus sequence typing, epidemiology

# INTRODUCTION

Campylobacter spp. are Gram negative bacteria responsible for the greatest number of cases of bacterial gastroenteritis worldwide (World Health Organization [WHO], 2013). It is estimated that 1.3 million Campylobacter infections occur every year in the U.S., resulting in 13,000 hospitalizations and 120 deaths, with C. jejuni comprising almost 90% of the cases

(Centers for Disease Control and Prevention [CDC], 2015). Furthermore, previous studies have demonstrated an association between campylobacteriosis and autoimmune diseases such as Guillain Barré syndrome (Nachamkin et al., 1998), reactive arthritis (Zia et al., 2003), and irritable bowel syndrome (Connor, 2005). Campylobacter spp. have a broad host range and can colonize the intestinal tracts of chickens, turkeys, pigs and ruminants without causing clinical signs (Altekruse et al., 1999; Stanley and Jones, 2003). These pathogens can also survive in water and soil for extended periods of time, up to several months (Jones, 2001; Pitkänen, 2013). The consumption of contaminated poultry is the primary source of sporadic human Campylobacter infections (Kaakoush et al., 2015), while approximately 66% of Campylobacter outbreaks are attributed to dairy products, mostly raw milk or cheese (Interagency Food Safety Analytics Collaboration Project, 2015). Direct transmission from animal sources including household pets such as dogs and cats, has also been reported (Damborg et al., 2004).

The most common clinical presentation of campylobacteriosis is self-limiting gastroenteritis with vomiting, cramping, and diarrhea lasting for 7–10 days in most cases (Peterson, 1994). Many individuals develop more severe and prolonged infections. Extraintestinal spread of the bacterium, for example, can lead to bacteremia and infection of other organs in some individuals (Man, 2011). Infants, geriatric patients, and immunocompromised individuals more commonly require treatment with antibiotics to combat C. jejuni infections (Guerrant et al., 2001). Ciprofloxacin, a fluoroquinolone that inhibits DNA synthesis by targeting gyrA and macrolides such as azithromycin and erythromycin, which hinder bacterial protein biosynthesis by targeting 23S rRNA, have been recommended as the first line antimicrobials. Yet, resistance to both antimicrobials has emerged and increases in resistance frequencies have been reported (Zhou et al., 2015). Indeed, the proportion of Campylobacter isolates resistant to fluoroquinolones has increased from 14.2% in 1998 to 25.3% in 2012 in the U.S. (Centers for Disease Control and Prevention [CDC], 2014).

Campylobacter resistance to fluoroquinolones and macrolides is conferred by point mutations in their target sites, the gyrA and 23S rRNA genes, respectively (Payot et al., 2006). The C257T point mutation in gyrA that yields a Thr-86-Ile amino acid change is the most frequently observed mutation that leads to high-level resistance to quinolone, i.e., nalidixic acid, and fluoroquinolone (Iovine, 2013). For macrolides, point mutations of A2074C, A2074G, and A2075G in domain V of the 23S rRNA gene, have been found to confer a high-level of resistance [Minimal inhibitory concentration (MIC) >128 µg/mL], while A2074T has been shown to confer a low-level of resistance (MIC = 8 µg/mL) (Vacher et al., 2005). The presence of the 23s rRNA methyltransferase gene ermB and an amino acid change in L4/L22 ribosomal proteins also have been described as resistance mechanisms for macrolides (Pérez-Boto et al., 2010; Wang et al., 2014). In addition, the active efflux pump, CmeABC, works synergistically with point mutations in these gene targets to simultaneously resist the action of fluoroquinolone, macrolide, tetracycline, beta-lactam, and ketolide antimicrobials (Zhang, 2008).

Tetracycline has been suggested as an alternative treatment for patients with systemic Campylobacter infections (Guerrant et al., 2001), but it is rarely used in practice. On the other hand, tetracycline is widely used in food animals like chickens and cattle for preventive and growth promoting purposes and treatment, e.g., in sheep for abortion (Giguere et al., 2006). In Campylobacter, resistance to tetracycline is conferred by tet(O), which has been found widely in isolates recovered from various sources (Connell et al., 2003). tet(O) encodes a ribosomal protection protein that induces a conformational change upon binding to the bacterial ribosome, the target site for tetracycline. This binding results in the release of the bound tetracycline molecule. A prior study has identified the tet(O) gene in plasmids in the majority of isolates from campylobacteriosis cases (Gibreel et al., 2004), though it has also been identified in the bacterial chromosome (Wu et al., 2014).

In the U.S., the FoodNet surveillance system was designed to monitor the incidence of common foodborne pathogens, including Campylobacter, while the National Antimicrobial Resistance Monitoring System (NARMS) was implemented to examine antimicrobial resistance trends. Michigan is not one of the 10 states included in the FoodNet system, however, Campylobacter was the most common foodborne pathogen reported through the Michigan Disease Surveillance System (MDSS) in the past decade (2004–2013). As a result, this study was undertaken to determine the frequency of antimicrobial resistance in a subset of C. jejuni isolates collected in Michigan between 2011 and 2012, and to estimate the genetic diversity of both susceptible and resistant isolates using multilocus sequence typing (MLST). We hypothesized that the frequency of antimicrobial resistance in C. jejuni isolates recovered from Michigan patients is similar to national frequencies, but that specific risk factors, which may be unique to this patient population, are associated with resistant infections. Identifying which lineages are associated with resistance and estimating their frequency in different populations is important for disease prevention efforts aimed at controlling resistance emergence and more rapidly detecting resistant infections.

# MATERIALS AND METHODS

# Study Population and Campylobacter Isolation

From 2011 to 2012, we obtained C. jejuni isolates from patients with campylobacteriosis identified via the Michigan State University (MSU) Enteric Research Investigative Network (ERIN) surveillance system. This system was established in collaboration with the Michigan Department of Health and Human Services (MDHHS) and four major hospitals in Michigan. All protocols were approved by the Institutional Review Boards at Michigan State University (IRB# 10-736SM), the MDHHS (842-PHALAB), and each participating hospital.

Isolates were cultured on tryptone soy agar plates with sheep blood and cefoperazone (20 µg/mL), amphotericin B (4 µg/mL), and vancomycin (20 µg/mL) using microaerophilic conditions at 37◦C for 48 h. Multiplex PCR was performed on the extracted DNA to confirm the species following previously described protocol (Yamazaki-Matsune et al., 2007). The isolates were stored in tryptone soy broth with 10% glycerol at −80◦C until further testing.

# Epidemiological Data

fmicb-07-00589 April 22, 2016 Time: 11:38 # 3

Demographic and epidemiological data were retrieved from the MDSS<sup>1</sup> and managed using Microsoft Access and Excel. Three cases were not Michigan residents, but developed campylobacteriosis while traveling in Michigan and thus, epidemiological data was transferred to the respective states. The C. jejuni isolates from these cases were included in the overall genetic diversity and resistance prevalence estimates, but were excluded from further analyses. A patient was considered to have a history of travel only when the traveling period was within 1 week prior to the onset of symptoms. For animal contact, birds were defined as poultry, including chickens, ducks, and turkeys, while ruminants were defined as cattle, sheep and goats, and domestic animals were defined as household pets such as dogs and cats. The season was classified based on the onset date of symptoms: spring (March, April, May), summer (June, July, August), fall (September, October, November), and winter (December, January, February). Food history represents consumption of specific foods (e.g., frozen chicken, homeprepared chicken, etc.) within a week prior to onset of symptoms.

# Phenotypic Antimicrobial Susceptibility Profiling

The MICs of nine antimicrobials were determined by a standard broth microdilution test following the guidelines of the Clinical and Laboratory Standards Institute [CLSI] (2015). The Sensititre system (Trek Diagnostic Systems, Thermo Fisher Scientific Inc., Cleveland, OH, USA) was used for each isolate following the manufacturer's instructions. Antimicrobials included ciprofloxacin (fluoroquinolone), nalidixic acid (quinolone), azithromycin (macrolide), erythromycin (macrolide), tetracycline, florfenicol, telithromycin, clindamycin, gentamicin, and C. jejuni ATCC 33560 was used as the quality control strain for every batch. The breakpoints for each antimicrobial were determined using epidemiologic cutoff values (ECOFFs), following the guidelines of European Committee on Antimicrobial Susceptibility Testing [EUCAST] (2015), per current protocol of the NARMS (Centers for Disease Control and Prevention [CDC], 2014). If bacterial growth was observed at the highest MIC tested (e.g., 64 µg/mL for tetracycline), then the MIC for the isolate was interpreted as greater than the highest MIC, i.e., > 64 µg/mL.

# Whole Genome Sequencing

DNA was extracted from all 94 C. jejuni isolates using the Wizard genomic DNA purification kit (Promega, Madison, WI, USA) and the concentrations were measured using a Qubit fluorometer (Life Technologies; Invitrogen, Carlsbad, CA, USA). A total of 1ng of DNA per isolate was used for library preparation with the Nextera XT kit (Illumina, San Diego, CA, USA) following manufacturer's instruction. Validation of the library size and quantity was performed using Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA) and KAPA library quantification kit (Kapa Biosystems, Woburn, MA, USA), respectively. The libraries were pooled together for denaturing and sequencing on a MiSeq (Illumina) platform for 2 × 250 reads at the Research Technology Support Facility at MSU. Genomic assemblies were performed de novo using Velvet, 1.2.07 (Zerbino and Birney, 2008) after trimming with Trimmomatic (Bolger et al., 2014), followed by quality checking with FastQC<sup>2</sup> . Assemblies were constructed using different kmer values (31, 33, and 35), and the assembly yielding the best N50 value for each isolate was used for downstream analyses.

# Multilocus Sequence Typing

The MLST profile of each sample was initially determined using the web-based server<sup>3</sup> with both the assembled contigs and extracted alleles from whole genome sequences. Each gene sequence was also confirmed by PCR-based MLST, as described previously (Dingle et al., 2001). Allele, sequence type (ST), and clonal complex (CC) assignments were made using the PubMLST database<sup>4</sup> (Jolley and Maiden, 2010). New alleles (n = 4) and STs (n = 6) identified in this study were deposited in the PubMLST database (id 24892- 28175).

# In Silico Analysis of Resistance Genes

Sequences specific for gyrA and 23S rRNA were extracted from the draft genomes based on reference sequences available via the National Center for Biotechnology Information using the Basic Local Alignment Search Tool (BLAST) (Altschul et al., 1990) (gyrA: GenBank KU693342-KU693435; 23S rRNA: GenBank KU891692-KU891785). Regions of the 23S rRNA and gyrA genes, which include the typical point mutation sites associated with resistance to macrolides and fluoroquinolones, respectively, were aligned by MegAlign (DNAstar, Madison, WI, USA). Sanger sequencing was used to confirm the point mutations identified in 47 strains relative to the ATCC 33560 C. jejuni strain (gyrA: GenBank KU693436; 23S rRNA KU891786), as described in prior studies (Vacher et al., 2003; Said et al., 2010). The presence of tet(O) was determined from the genome sequences and further confirmed by PCR using a published protocol (Gibreel et al., 2004).

# Data Analysis

The frequency map of all campylobacteriosis cases reported in Michigan between 2011 and 2012 (n = 1,449) was generated using ArcMap GIS software (version 10.2; ESRI, Redlands, CA, USA) using the data extracted from MDSS.

To identify evolutionary relationships between C. jejuni isolates, a Neighbor-joining phylogeny (p-distance) with 1,000

<sup>1</sup>http://www.michigan.gov/mdhhs/0,5885,7-339-71550510431274---,00.html

<sup>2</sup>http://www.bioinformatics.babraham.ac.uk/projects/fastqc/

<sup>3</sup>www.cbs.dtu.dk/services/MLST

<sup>4</sup>http://pubmlst.org/campylobacter/

bootstrap replications was constructed in MEGA6 (Tamura et al., 2013) based on seven MLST loci. Clusters or CCs were classified as STs that grouped together with >70% bootstrap support, and parsimonious informative sites were further evaluated for evidence of genetic recombination using Splitstree4 (Huson and Bryant, 2006).

Statistical analyses were performed using SAS version 9.3 (SAS Institute, Cary, NC, USA). Differences in the frequencies of antimicrobial resistance across ST, CC, and other variables including disease presentation, were examined using χ 2 and Fisher's exact tests for dichotomous variables, and the student's t-test for continuous variables; p < 0.05 was considered significant. Multivariate analyses were performed using logistic regression with any independent variables that yielded a p-value of less than 0.2 or were plausibly linked to resistant C. jejuni infections, e.g., age, sex. The model was built using a forward stepwise method with the requirement for a significance level of ≤0.1 to remain in the model.

# RESULTS

# Description of Campylobacter Cases Identified in Michigan

Ninety four C. jejuni isolates were recovered from clinical cases of campylobacteriosis identified at four hospitals between January 2011 and December 2012. Among the 94 cases, more than 50% (n = 52) were male, while 39 cases were female patients; the sex was not known for three cases. Children 2 years and younger (22.3%; n = 21) and adults older than 50 years (25.5%; n = 24) comprised about half of the total cases. Race was available for 80 of the 94 cases, and the majority identified as Caucasian (n = 60; 75.0%).

A total of 1,449 laboratory confirmed Campylobacter cases were reported in Michigan in 2011 and 2012 and the frequency of reported cases varied across counties (**Figure 1**). The participating hospitals and most of the residences of the cases included in this study were in counties with higher population densities and a greater frequency of campylobacteriosis. Particularly, cases residing in Wayne, Washtenaw and Oakland counties, which represent the counties with the highest disease frequency, comprised 66.7% (n = 58) of the total cases.

Sixty-eight cases had travel history information, among which nine (13.2%) had a history of foreign travel, while 13 (19.1%) had a history of domestic travel in the week prior to the onset of symptoms (**Table 1**). Four cases had a history of domestic travel outside Michigan, while three cases from other states (Ohio, New Jersey, and Georgia) developed symptoms and were diagnosed with campylobacteriosis while traveling in Michigan. Among the 64 cases with a history of animal contact, 38 (59.4%) had contact with domestic animals including dogs and cats. Contact with ruminants, which were all identified as cattle, was reported in seven cases (10.9%), while six cases (9.4%) reported contact with birds such as chickens and ducks. Six of the cases reporting contact with cattle (85.7%) and five of the cases with bird contact (83.3%) also reported contact with dogs and cats.

# Antimicrobial Resistance Profiles of C. jejuni Isolates and Mechanisms of Resistance

Thirty isolates (31.9%) were susceptible to all nine antimicrobial agents tested, while 64 isolates (68.1%) were resistant to one or more agents. The highest frequency of resistance was observed for tetracycline (n = 58 isolates; 61.7%), followed by resistance to both ciprofloxacin and nalidixic acid (n = 18 isolates; 19.1%) (**Table 2**). All C. jejuni isolates resistant to ciprofloxacin and nalidixic acid showed high MICs (4–32 µg/mL, ≥64 µg/mL, respectively). Resistance to florfenicol was detected in one isolate (1.1%), and all isolates were susceptible to gentamicin. A total of 15 isolates (16.0%) were resistant to two or more classes of antibiotics. Thirteen (13.8%) of these isolates were resistant to ciprofloxacin, nalidixic acid, and tetracycline, while two isolates (2.1%) were resistant to both azithromycin and erythromycin as well as ciprofloxacin, nalidixic acid, telithromycin, and clindamycin.

All 18 C. jejuni isolates that were phenotypically resistant to ciprofloxacin had a point mutation at 257 in gyrA; 17 isolates had the C257T mutation resulting in an amino acid change of Thr-86-Ile, while one isolate had double mutations (C257G and A258G) resulting in a Thr-86-Arg change. Two isolates that were resistant to azithromycin and erythromycin had an A2074T point mutation in 23S rRNA, and all 58 tetracycline resistant isolates harbored tet(O).

# Epidemiological Associations with Antimicrobial Resistant C. jejuni Infections

To identify factors associated with antimicrobial resistant C. jejuni infections, we conducted univariate analyses using demographic and epidemiological data (Supplementary Table S1). Notably, cases reporting a history of foreign travel had a higher likelihood of fluoroquinolone-resistant C. jejuni infections (Fisher's p < 0.0001) with an odds ratio (OR) of 35.7 (exact 95% CI; 4.6, 395.3). In detail, among the nine cases with foreign travel history, seven were resistant to ciprofloxacin and nalidixic acid; six of these isolates were also resistant to tetracycline, yielding a significant association between foreign travel and the following resistance profile: ciprofloxacin-, nalidixic acid-, tetracyclineresistance (CipNalTet) (Fisher's p < 0.0001; OR = 35.3). Cases with CipNalTet infections were significantly more common during the winter months of December, January, and February (Fisher's p < 0.05; OR = 5.7), while eating chicken prepared at home was protective for both CipNalTet infections (p < 0.01; OR = 0.0) and fluroquinolone-resistant infections (p < 0.01; OR = 0.08). By contrast, contact with cattle was associated with tetracycline-resistant infections (Fisher's p < 0.05; OR = infinity) and cases reporting consumption of frozen chicken were less likely to have tetracycline-resistant infections (p < 0.05; OR = 0.2).

Multivariate analysis was conducted to identify predictors of fluoroquinolone-resistant infections in all 94 cases using the factors with significant associations (p < 0.2) identified in the

univariate analyses as well as biologically plausible factors such as age, sex (**Table 3**). History of foreign travel was associated with fluoroquinolone-resistant infections regardless of age and sex; the OR was 33.4 (95% CI = 3.9–285.2) in the final model. When the same multivariate analysis was performed for CipNalTet infections, history of foreign travel (OR = 54.2; 95% CI = 4.1–717.1) and winter (OR = 25.3; 95% CI = 1.6– 405.7) were independently associated with infection. However, there was no association between foreign travel history and season, including winter. Although having chicken prepared at home was a protective factor in the univariate analysis, this variable could not be examined in the multivariate models as it significantly reduced the sample size due to missing data in 37 of the cases.

# Genetic Diversity and Phylogenetic Structure of C. jejuni

A total of 49 different STs, including six novel STs, were represented among the 94 C. jejuni isolates recovered in Michigan (**Figure 2**). These STs were assigned to 17 CCs, while 11 STs were singletons. The six new STs were assigned to ST-6749 (CC-353), ST- 6751 (CC-61), ST-6752 (CC-353), ST-6788 (CC-1332), ST-7009 (CC unassigned), and ST-7010 (CC unassigned). The most prevalent STs were ST-982 (n = 10; 10.6%) and ST-353 (n = 9; 9.6%), followed by ST-45 (n = 7; 7.4%), ST-50 (n = 5; 5.3%) and ST-48 (n = 4; 4.3%). Thirty four of the remaining STs had only one isolate assigned to each ST.

The MLST-based Neighbor-joining phylogeny for all 94 isolates showed that some STs were closely related, though the bootstrap values were low, which is likely due to the high diversity and frequent recombination among STs in this isolate population. Indeed, an evaluation of the 144 parsimonious informative sites provided evidence of significant recombination [pairwise homoplasy index (PHI) = 0.0] among the STs via a Neighbor-net analysis (Supplementary Figure S1). In order to elucidate the ST distribution and evolutionary relationships of isolates that were restrictively derived from Michigan, we excluded cases with foreign travel and history of travel outside of Michigan as well as cases with missing data. A total of 36 different STs, including four novel STs, represented the 52 Michigan cases without any history of travel (n = 46) or travel only within Michigan (n = 6). Although there was still evidence of recombination between these isolates, the phylogenetic tree

showed enhanced bootstrap support, and six distinct clusters with bootstrap support values exceeding 70% were identified (**Figure 3**, Supplementary Figure S2).

# Association between Phylogenetic Lineage, Epidemiologic Data, and Resistance

Multiple epidemiological factors were associated with specific C. jejuni genotypes. For example, individuals reporting a history of foreign travel were more likely to have infections caused by ST-464 (CC-464) isolates (Fisher's p < 0.05), while infection with ST-982 was associated with cattle contact (Fisher's p < 0.05). Drinking water obtained from a well at home (Fisher's p < 0.05), and contact with birds, i.e., chickens (Fisher's p < 0.01) were also associated with ST-982 infections, while contact with birds, i.e., chickens (Fisher's p < 0.05) and female gender (χ <sup>2</sup> p < 0.05) were both associated with CC-21 (n = 25) infections. In addition, cases older than 50 years of age were more likely to have isolates


† Counts for sex, age group, race, travel history, and water source at home are mutually exclusive for each category; counts for animal contact and food consumption are repeated across categories as they are reported. ‡The percentages are based on the number of cases for which information was available. Total number of cases varies between variables due to the difference in missing information.

#### TABLE 2 | Frequency (%) of antimicrobial resistance and minimum inhibitory concentration (MIC) of human C. jejuni isolates in Michigan.


† CLSI: Clinical and Laboratory Standards Institute.


TABLE 3 | Univariate and multivariate analyses of factors associated with fluoroquinolone resistant C. jejuni infections among all cases (n = 94).

<sup>∗</sup>Age (years) is a continuous variable; not proper for univariate analysis used (χ <sup>2</sup> or Fisher's exact test). ∗∗Consumption of home prepared chicken was a significant protective factor by univariate analysis; however, it was not used for multivariate analysis because of the small sample size left when the characteristic was included in the base model. †Wald Confidence interval

belonging to CC-45 (n = 9) relative to all other CCs (Fisher's p < 0.05).

Among the 52 cases without any history of travel or travel only within Michigan, similar associations were observed. Notably, contact with birds, i.e., chickens, was associated with both ST-982 (Fisher's p < 0.01) and CC-21 (Fisher's p < 0.05) infections. Because the phylogenetic clusters were better defined in the isolate population recovered exclusively from Michigan cases (**Figure 3**), we also analyzed epidemiologic associations by clusters identified in the Neighbor joining phylogeny (**Figure 3**). Isolates belonging to Cluster V, which includes ST-982 and the 4 other STs, all representing CC-21, were significantly associated with contact with birds, i.e., chickens (Fisher's p < 0.05). Additionally, when compared to other clusters, the two cases reporting contact with cattle were included in Cluster V; however, the association was not significant (Fisher's p = 0.0571), which may be due to the small sample size.

It is also notable that the three isolates assigned to ST-464 had the same profile with resistance to ciprofloxacin, nalidixic acid and tetracycline (CipNalTet). A significant association was observed between ST-464 and resistance to ciprofloxacin and nalidixic acid (Fisher's p < 0.01), as well as CipNalTet (Fisher's p < 0.01). Similarly, the isolates belonging to ST-982 had a higher likelihood of having resistance to tetracycline when compared to all other lineages (Fisher's p < 0.05). No other significant associations were observed for other STs or CCs with specific resistance profiles.

# DISCUSSION

Increasing frequencies of resistance to antimicrobials in C. jejuni is a growing public health concern worldwide. In the U.S., NARMS monitors resistance trends by testing C. jejuni isolates collected from 10 states across the country; however, Michigan is not included in this system. Hence, this represents the first study reporting the antimicrobial resistance profiles and genetic diversity of C. jejuni recovered from patients in Michigan collected via active surveillance. The high frequency of resistance to several important antimicrobials including the fluoroquinolones and macrolides, which are commonly used to treat human infections, is of concern as are the risk factors for resistant infections. Indeed, we found that a history of foreign travel is the most important risk factor for fluoroquinoloneresistant C. jejuni infections in Michigan, which is consistent with prior reports in other geographic locations in the U.S. (Smith et al., 1999; Kassenborg et al., 2004) and Europe (Engberg et al., 2001; Rautelin et al., 2003; Norström et al., 2006). A case-control study conducted using FoodNet surveillance sites in the U.S. during 1998–1999, for instance, demonstrated that history of foreign travel resulted in a higher likelihood of acquiring fluoroquinolone-resistant Campylobacter infections (OR = 7.6; 95% CI = 4.3-13.4) (Kassenborg et al., 2004). This same study also found that consumption of poultry prepared outside of the home was a risk factor for domestically acquired fluoroquinolone-resistant infections (OR = 10.0, 95% CI = 1.3– 78). The latter result is similar to our finding that chicken prepared at home was protective and resulted in a decreased likelihood of having a fluoroquinolone-resistant infection. These data are important as they demonstrate that the major sources of fluoroquinolone-resistant C. jejuni infections are similar in different parts of the U.S. even though disease frequencies may vary. They also emphasize the importance of monitoring patients with foreign travel history and considering individual treatment schemes in order to more readily combat resistant infections.

Prior studies have suggested that a high frequency of travelassociated fluoroquinolone-resistant Campylobacter infections may be due to a high prevalence of fluoroquinolone resistance in the destination areas (Piddock, 1995; Smith et al., 1999). Recently, clonal spread of specific fluoroquinolone-resistant lineages, e.g., ST-464, have been reported in Europe (Cody et al., 2012; Kittl et al., 2013). According to the PubMLST database (Jolley and Maiden, 2010), ST-464 has been frequently reported in Europe and Asia, however, no prior reports have been noted in the U.S. All three isolates assigned to ST-464 had the same resistance profile (CipNalTet) and two of these isolates were from cases that had traveled to Cambodia and to Italy, Turkey, Greece, and France; travel status was missing for the remaining case. These data suggest the dissemination of a fluoroquinolone-resistant ST-464 lineage outside of Europe, warranting the need to enhance surveillance efforts and identify factors associated with global spread.

From a total of 94 isolates, thirty six (38.3%) had at least one point mutation in gyrA compared to the ATCC 33560 reference strain that is susceptible to all antimicrobials tested (McDermott et al., 2004). Eighteen isolates had non-synonymous point mutations conferring resistance to fluoroquinolones; the amino acid changes, Thr-86-Ile and Thr-86-Arg, were observed in 17 isolates and 1 isolate, respectively. Although the Thr-86-Arg mutation was described as one of the mutations inducing high MIC values for fluoroquinolones in Campylobacter spp. (Rossi et al., 2015), this mutation is rarely reported. It is also notable that only this isolate with Thr-86-Arg had transversion mutations, C257G and A228C in gyrA (**Figure 2**). Interestingly, among the 17 isolates with a Thr-86-Ile mutation, 13 different STs were represented and five of these were recovered from patients with a

history of foreign travel to multiple geographic locations. These data demonstrate that fluoroquinolone resistance can develop via similar point mutations in gyrA among phylogenetically distinct Campylobacter lineages. No other pattern or association was observed in the type or frequency of synonymous point mutations in gyrA and specific resistance profiles or increased MICs. However, notable associations were observed between STs and the types of point mutations. For example, all seven isolates assigned to ST-45 had a T234C point mutation, while the two closely related lineages, ST-2875 and ST-7009, ST-468 and ST-4185, respectively, contained identical synonymous gyrA point mutations (**Figure 2**). Indeed, prior studies have suggested the

Frontiers in Microbiology | www.frontiersin.org April 2016 | Volume 7 | Article 589 |

use of gyrA as a genetic marker to investigate the relatedness of C. jejuni strains (Ragimbeau et al., 2014). It is also important to note that not all mutations were restricted to closely related lineages. The C330T synonymous mutation in gyrA, for instance, was found in ST-353, ST-6, ST-2310, and ST-21, distinct lineages located in different branches of the phylogenetic tree.

Prior studies have detected variation in the level of resistance to macrolides based on the type of mutation in 23S rRNA as well. The A2074T mutation in 23S rRNA, for example, has been linked to low macrolide resistance levels previously (Vacher et al., 2005), while a recent study detected a high resistance level (MIC > 512 µg/ml) in a C. jejuni strain from a human clinical case (Ohno et al., 2016). Similar to the latter finding, the two macrolide resistant isolates recovered from Michigan had the same A2074T 23S rRNA mutation with high resistance levels to both erythromycin and azithromycin (MIC > 64 µg/ml). Notably, these isolates were multiple drug resistant (MDR) with resistance to four classes of antimicrobials including macrolides, a fluoroquinolone, a lincosamide (clindamycin) and a ketolide (telithromycin), suggesting that other resistance mechanisms could impact the overall level of resistance. One example is the cmeABC efflux pump, which has been linked to resistance to multiple antimicrobials, including fluoroquinolones, macrolides, and tetracycline. Both of the Michigan strains, however, were susceptible to tetracycline. It is important to note that these isolates, representing ST-5221 and ST-6, respectively, clustered together in the Neighbor-joining phylogeny with ST-7010 (**Figure 2**), all of which contained a high frequency of similar synonymous point mutations in 23S rRNA (Supplementary Table S2). This finding suggests that some C. jejuni lineages may be more susceptible to point mutations, which may lead to antimicrobial resistance. Further studies are required to determine whether specific synonymous mutations impact the mutability of some isolates resulting in greater mutation rates or enhanced fitness in vivo.

A significantly higher tetracycline resistance rate (61.7%) was observed in the study samples when compared to the 2012 NARMS report (47.8%) (χ <sup>2</sup> p < 0.01). All of the tetracyclineresistant C. jejuni isolates (n = 58) harbored tet(O), which was confirmed by both PCR and in silico analysis. Although we could not determine whether tet(O) was plasmid-derived in all of the isolates examined, the gene was found in multiple lineages throughout the phylogeny, suggesting that plasmid-mediated resistance and horizontal gene transfer across lineages is likely. In addition, a significant association was observed between ST-982 and tetracycline resistance as well as cattle contact. According to the PubMLST database, ST-982 has previously been reported from cattle (n = 14), cow's milk (n = 4), the farm environment (n = 3) and a lamb (n = 1) in the U.S., while it was linked to human clinical cases in Canada and the U.K. Another study conducted in the state of Washington (Davis et al., 2013), however, recovered ST-982 isolates from both humans and cattle. Taken together, these data suggest that this resistant genotype has the ability to cross species and may be more likely to acquire resistance to tetracycline. A significant association was also observed between ST-982 and contact with chickens and drinking well water at home, suggesting that this lineage may be readily adaptable to different hosts and environments. Further investigation of the genetic diversity and antimicrobial resistance profiles of C. jejuni recovered from reservoir hosts and the environment, however, is needed to better understand transmission dynamics and host specificity.

Recent studies in Europe show mounting evidence that clonal spread of specific fluoroquinolone-resistant C. jejuni lineages may be contributing to the increasing trend of fluoroquinolone resistance observed worldwide (Cody et al., 2012; Wimalarathna et al., 2013; Kovac et al., 2014 ˇ ). Our study affirms foreign travel as a major risk factor for fluoroquinolone-resistant infections in the U.S., and further suggests that specific resistant lineages, like ST-464, may be emerging in the country. With continuing globalization through imported food and foreign travel, unified diagnostic guidelines on resistance detection and global surveillance is warranted to track the emergence and spread of these resistant lineages so effective mitigation strategies can be implemented.

# AUTHOR CONTRIBUTIONS

WC, JR, SM designed the study; WC, RM, SW, PS, SM generated data; WC, DN, HS, PL, WK, LM, JR, and SM analyzed the data; and WC, SM drafted the manuscript. All authors read and approved the final manuscript.

# FUNDING

This study was funded by the National Institutes of Health Enterics Research Investigational Network (ERIN) Cooperative Research Center at Michigan State University (SM, grant number U19AI090872) and the United States Department of Agriculture, National Institute of Food and Agriculture (SM, grant number 2011-67005-30004). The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication. Graduate student support (WC) was provided in part by the Michigan State University College of Veterinary Medicine.

# ACKNOWLEDGMENTS

We thank Sandip Shah, Ben Hutton, Jason Wholehan, Sheri Robeson, Tiffany Henderson, and James Collins at the MDHHS and the laboratory staff at each participating hospital for help with specimen processing and culture as well as Jonathan Lehnert and Jessica St. Charles for technical assistance. We also thank Dr. Vilma Yuzsbasyian-Gurkan, the Director of the Comparative Medicine and Integrative Biology graduate program at Michigan State University, for her continued support of the project.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb.2016. 00589

# REFERENCES

fmicb-07-00589 April 22, 2016 Time: 11:38 # 11


Listeria monocytogenes (Lm), and Campylobacter using Outbreak Surveillance Data. Available at: http://www.cdc.gov/foodsafety/ifsac/index.html


Great Britain. J. Clin. Microbiol. 52, 1853–1861. doi: 10.1128/JCM. 00355–314


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Cha, Mosci, Wengert, Singh, Newton, Salimnia, Lephart, Khalife, Mansfield, Rudrik and Manning. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

fmicb-07-00589 April 22, 2016 Time: 11:38 # 12

# Characteristics of Quinolone Resistance in *Escherichia coli* Isolates from Humans, Animals, and the Environment in the Czech Republic

Magdalena Röderova<sup>1</sup> \*, Dana Halova<sup>2</sup> , Ivo Papousek <sup>2</sup> , Monika Dolejska2, 3 , Martina Masarikova3, 4, Vojtech Hanulik 1, 5, Vendula Pudova<sup>1</sup> , Petr Broz <sup>6</sup> , Miroslava Htoutou-Sedlakova1, 5, Pavel Sauer 1, 5, Jan Bardon<sup>1</sup> , Alois Cizek 3, 4, Milan Kolar <sup>1</sup> and Ivan Literak 2, 3

### *Edited by:*

Benoit Doublet, National Institute for Agricultural Research (INRA), France

#### *Reviewed by:*

Isabelle Kempf, French Agency for Food (ANSES), Environmental and Occupational Health & Safety, France Marisa Haenni, French Agency for Food (ANSES), Environmental and Occupational Health & Safety, France

*\*Correspondence:*

Magdalena Röderova m.roderova@gmail.com

### *Specialty section:*

This article was submitted to Antimicrobials, Resistance and Chemotherapy, a section of the journal Frontiers in Microbiology

*Received:* 19 October 2016 *Accepted:* 20 December 2016 *Published:* 09 January 2017

#### *Citation:*

Röderova M, Halova D, Papousek I, Dolejska M, Masarikova M, Hanulik V, Pudova V, Broz P, Htoutou-Sedlakova M, Sauer P, Bardon J, Cizek A, Kolar M and Literak I (2017) Characteristics of Quinolone Resistance in Escherichia coli Isolates from Humans, Animals, and the Environment in the Czech Republic. Front. Microbiol. 7:2147. doi: 10.3389/fmicb.2016.02147

<sup>1</sup> Department of Microbiology, Faculty of Medicine and Dentistry, Palacký University Olomouc, Olomouc, Czechia, <sup>2</sup> Department of Biology and Wildlife Diseases, Faculty of Veterinary Hygiene and Ecology, University of Veterinary and Pharmaceutical Sciences Brno, Brno, Czechia, <sup>3</sup> Central European Institute of Technology (CEITEC), University of Veterinary and Pharmaceutical Sciences Brno, Brno, Czechia, <sup>4</sup> Department of Infectious Diseases and Microbiology, Faculty of Veterinary Medicine, University of Veterinary and Pharmaceutical Sciences Brno, Brno, Czechia, <sup>5</sup> Department of Microbiology, University Hospital Olomouc, Olomouc, Czechia, <sup>6</sup> Institute of Applied Biotechnologies (IAB), Prague, Czechia

Escherichia coli is a common commensal bacterial species of humans and animals that may become a troublesome pathogen causing serious diseases. The aim of this study was to characterize the quinolone resistance phenotypes and genotypes in E. coli isolates of different origin from one area of the Czech Republic. E. coli isolates were obtained from hospitalized patients and outpatients, chicken farms, retailed turkeys, rooks wintering in the area, and wastewaters. Susceptibility of the isolates grown on the MacConkey agar with ciprofloxacin (0.05 mg/L) to 23 antimicrobial agents was determined. The presence of plasmid-mediated quinolone resistance (PMQR) and ESBL genes was tested by PCR and sequencing. Specific mutations in gyrA, gyrB, parC, and parE were also examined. Multilocus sequence typing and pulsed-field gel electrophoresis were performed to assess the clonal relationship. In total, 1050 E. coli isolates were obtained, including 303 isolates from humans, 156 from chickens, 105 from turkeys, 114 from the rooks, and 372 from wastewater samples. PMQR genes were detected in 262 (25%) isolates. The highest occurrence was observed in isolates from retailed turkey (49% of the isolates were positive) and inpatients (32%). The qnrS1 gene was the most common PMQR determinant identified in 146 (56%) followed by aac(6′ )-Ib-cr in 77 (29%), qnrB19 in 41 (16%), and qnrB1 in 9 (3%) isolates. All isolates with high level of ciprofloxacin resistance (>32 mg/L) carried double or triple mutations in gyrA combined with single or double mutations in parC. The most frequently identified substitutions were Ser(83)Leu; Asp(87)Asn in GyrA, together with Ser(80)Ile, or Glu(84)Val in ParC. Majority of these isolates showed resistance to beta-lactams and multiresistance phenotype was found in 95% isolates. Forty-eight different sequence types among 144 isolates analyzed were found, including five major clones ST131 (26), ST355 (19), ST48 (13), ST95 (10), and ST10 (5). No isolates sharing 100% relatedness and originating from different areas were identified. In conclusion, our study identified PMQR genes in E. coli isolates in all areas studied, including highly virulent multiresistant clones such as ST131 producing CTX-M-15 beta-lactamases.

Keywords: *Escherichia coli*, quinolone resistance, human, animals, wastewater, MLST, PFGE

# INTRODUCTION

Escherichia coli is a common part of the gastrointestinal flora of humans and animals but some strains can cause serious diseases. Quinolones are the antimicrobial agents of choice for treatment of various infections caused by E. coli or other Gramnegative bacteria. Because of extensive use for multiple clinical indications in human or veterinary medicine, bacterial resistance to quinolones has developed over the time (Andriole, 2005). The World Health Organization indicated fluoroquinolones as critically important drugs for human medicine due to a strong correlation observed between the consumption and an increase in the resistance, and they recommended prudent use of fluoroquinolones both in human and veterinary medicine (World Health Organization, 2012).

According to the European Antimicrobial Resistance Surveillance Network (EARS-Net), an interactive database, an alarming emergence of fluoroquinolone-resistant invasive E. coli isolates is evident in European countries (European Centre for Disease Prevention and Control, 2015). Quinolone antibiotics are one of the most widely used antimicrobial agents in the treatment of urinary tract infections. Unfortunately, the extensive use has led to the increase of the rate of E. coli isolates resistant to fluoroquinolones all over the world (Lautenbach et al., 2004; Dalhoff, 2012). Resistance to nalidixic acid and ciprofloxacin was reported as high or extremely high in isolates from broilers in European countries (ECDC, EFSA, and EMA, 2015). E. coli isolates resistant to fluoroquinolones has been described in food-producing and companion animals (Platell et al., 2011; Gosling et al., 2012; Literak et al., 2013; Agabou et al., 2016) as well as in wild animals and the environment (Jiménez Gómez et al., 2004; Colomer-Lluch et al., 2013).

Target sites for quinolones in E. coli strains are the bacterial topoisomerases, namely DNA gyrase (topoisomerase II) as the primary site and topoisomerase IV as a secondary target. Both enzymes are essential for bacterial DNA replication. Mutations in specific domains of gyrA, gyrB, parC, and parE cause single aminoacid changes in either gyrase or topoisomerase IV that contribute to quinolone resistance. Multiple mutations in the quinolone-determining resistant regions (QRDR) of topoisomerase enzymes are usually associated with high-level of fluoroquinolone resistance in E. coli strains (Hopkins et al., 2005). Target-mediated resistance represents the most common and clinically most significant form (Ruiz, 2003; Aldred et al., 2014).

The first identified plasmid-mediated quinolone resistance (PMQR) gene was discovered in 1998 and termed qnrA1 (Martínez-Martínez et al., 1998). So far, three families of plasmidmediated mechanisms associated with quinolone resistance have been identified: i/Qnr proteins protecting target enzymes DNA gyrase and topoisomerase IV from quinolone inhibition ii/ aminoglycoside acetyltransferase Aac(6′ )-Ib-cr acetylating several fluoroquinolones such as ciprofloxacin and norfloxacin and iii/ efflux pumps QepA and OqxAB removing antibiotics from bacterial cells. PMQR provide only a low level of quinolone resistance, not reaching the clinical breakpoints defined by the Clinical and Laboratory Standard Institute criteria (CLSI, 2015). However, PMQR genes may facilitate the selection of higher-level resistance in the presence of quinolones and lead to treatment failure (Strahilevitz et al., 2009; Jacoby et al., 2014).

The aims of this study were to determine the occurrence of PMQR in E. coli isolates from humans, food-producing animals, wild animals and wastewater samples from one defined area in the Czech Republic and to compare genetic characteristics of PMQR-positive E. coli isolates from various sources as well as to outline possible ways of their transmission between humans and animals.

# MATERIAL AND METHODS

## Sample Collection

Between May 2013 and December 2014, E. coli isolates were obtained from human clinical materials (urine, stool, sputum, blood, bile, endotracheal aspirate, samples from cannulas, and bronchoalveolar lavages) of patients hospitalized at the University Hospital Olomouc (UHO) in the Czech Republic and from urine samples from outpatients in the community of Olomouc Region. The community subjects had neither been hospitalized in the previous 3 months nor had been living in nursing homes.

In the same period (May 2013 and December 2014), E. coli isolates were obtained from environmental samples from chicken farms (n = 2628) and cloacal swabs from market-weight turkeys at slaughterhouses (n = 120) in the eastern part of the Czech Republic (Olomouc and South Moravian Region). Environmental samples were taken from bedding using gauze shoe covers worn by a worker who walked through a poultry house. Turkey cloacal swabs were collected at slaughterhouses and placed into Amies transport medium.

Wastewaters were taken during 2013–2015 in six sampling sites that included wastewater treatment plants (WWTP) from four hospitals located in three towns of the Olomouc Region (Olomouc, Prostejov and Sternberk) (WWTP1-4), the abattoir in Prerov (WWTP5) and WWTP near Henclov (WWTP6). ˇ In total, 124 wastewater samples were examined including 21 samples from WWTP1 (in 2013-6 samples; 2014-13; 2015-2), 25 from WWTP2 (2014-20, 2015-5), 19 from WWTP3 (in 2013-6; 2014-11, 2015-2), 19 from WWTP4 (in 2013-6; 2014-11, 2015- 2), 21 from WWTP5 (collected weekly for a period of 4 months in 2013/2014, plus 3 control samplings were made in May and August 2014) and 19 from WWTP6 (collected weekly for a period of 4 months in 2013/2014). Cellulose swabs were immersed into wastewater at the inflow and/or outflow of sampling points for 48 h according to the procedure described (Moore et al., 1952).

Faecal rook (Corvus frugilegus) samples (n = 595) were collected from roosting places used by rook flocks in the Olomouc Region. The samples were collected in Prerov in October, November and December 2012, in Tovacov from January till March 2013, and in Troubky in November 2013. The sampling method has been described previously (Literak et al., 2012). Smears from fresh feces were taken by cotton swabs tampons and inserted into Amies transport medium.

# Selective Cultivation of *E. coli* and Detection of PMQR Genes

After delivery to the laboratory, collected samples were handled according to the following procedures in order to isolate E. coli.

Samples from the environment of chicken farms were placed into peptone water and incubated aerobically for 24 h at 37◦C. Subsequently, the peptone water was inoculated onto MacConkey agar with ciprofloxacin (0.05 mg/L; MCACIP) and cultivated overnight. Swabs from retailed turkeys were subcultivated directly on MCACIP overnight. One isolate per sample was selected for further phenotypic and genetic analysis.

Swabs taken from wastewaters were inserted into a sterile bottle with peptone water and incubated at 37◦C for 24 h. The enriched samples of peptone water were subcultivated on MCACIP overnight. One to ten colonies of lactose-positive colonies showing different morphology were taken from each MCACIP and subjected to further examination.

Faecal swabs from rooks were incubated in buffered peptone water at 37◦C overnight and subsequently cultivated on MCACIP overnight. In this type of sample, also only one isolate per sample was selected.

Human clinical material was handled according to the types of material and cultivated aerobically for 24 h at 37◦C. A total number of 3521 E. coli isolates were collected and from these, three hundred and three isolates were randomly selected for detection of PMQR genes. The collection analyzed included 270 resistant (MIC of ciprofloxacin >0.5 mg/L) and 33 sensitive (MIC of ciprofloxacin ≤0.5 mg/L) isolates according to the European Committee in Antimicrobial Susceptibility Testing (EUCAST) criteria.

The species identification of all isolates was confirmed using matrix-assisted laser desorption/ionization-time of flight mass spectrometer (MALDI-TOF MS) (Biotyper Microflex, Bruker Daltonik GmbH, Bremen, Germany).

Genomic DNA of all E. coli isolates obtained using heat lysis was used as a template for PCR detection of PMQR genes (aac(6′ )-Ib-cr, qepA, qnrA, qnrB, qnrC, qnrD, qnrS, oqxAB), followed by sequencing of the amplicons. As positive controls, well-known characterized strains were included in each reaction (Table S1).

# Antimicrobial Susceptibility Testing of PMQR-Positive Isolates

For each PMQR-positive E. coli isolate, susceptibilities to ampicillin (the ranges of tested concentrations for the antimicrobial substance were 0.5–64 mg/L, ampicillin/sulbactam (0.5–64), cefazolin (0.5–64), cefuroxime (0.5–64), gentamicin (0.25–32 or 0.5–64), trimethoprim/sulfamethoxazole (1–128), colistin (0.25–32), oxolinic acid (0.25–32 or 0.5–64), ofloxacin (0.125–16 or 0.25–32), tetracycline (0.25–32), aztreonam (1–64), piperacillin (2–256), piperacillin/tazobactam (2–256), cefoperazone (0.25–32), cefotaxime (0.125–16), ceftazidime (0.125–16), cefepime (0.125–16), cefoperazone/sulbactam (0.5–64), meropenem (0.125–16), ciprofloxacin (0.125–16 or 0.25–32), tigecycline (0.06–8), tobramycin (0.25–32), and amikacin (0.5–32) were tested using the standard microdilution method according to the EUCAST (European Committee on Antimicrobial Susceptibility Testing) breakpoint criteria (European Centre for Disease Prevention and Control, 2015). Microdilution antibiotic panels were prepared using dispensing instrument DYNAMIC 3000 automated system (DYNEX, Czech Republic). E. coli ATCC 25922 and E. coli ATCC 35218 were used as reference strains for quality control.

# Detection of ESBL Genes in PMQR-Positive Isolates

All PMQR-positive isolates of E. coli with the minimum inhibitory concentration (MIC) of the tested 3rd and 4th generation cephalosporins ≥1 mg/L were screened for extendedspectrum beta-lactamase production (ESBL) using Jarlier's double-disk synergy test (DDST) which was modified by including a disk with cefepime and another disk with ceftazidime and ceftazidime/clavulanic acid (Jarlier et al., 1988; Htoutou Sedlakova et al., 2011). Each ESBL-producing isolate was screened for blaCTX-M, blaTEM and blaSHV genes by PCR followed by sequencing of amplicons of genes responsible for ESBL phenotype (Table S1).

# Detection of Mutations in the Topoisomerase Genes and MLST Analysis of PMQR-Positive Isolates

From a group of 262 PMQR-positive E. coli isolates, 144 isolates from all of the studied areas were selected for further characterization that included the detection of specific mutations in gyrA, gyrB, parC and parE genes and multilocus sequence typing (MLST). Isolates were selected randomly in order to cover all the studied areas.

Total genomic DNA from E. coli strains was prepared from an overnight culture (16 h, 37◦C) grown on meat-peptone agar using DNeasy Blood & Tissue kit (QIAGEN, Germany) according to the manufacturer's recommendations.

PCR amplification of the part of gyrA, gyrB, parC and parE genes that included the sequence of QRDRs was performed using previously described primers (Oram and Fisher, 1991; Vila et al., 1994, 1996; Ruiz et al., 1997). The reaction mixture contained complete reaction buffer with MgCl<sup>2</sup> (containing 100 mmol/L Tris-HCl (pH 8.8), 500 mmol/L KCl, 1% Triton X-100, 15 mmol/L MgCl<sup>2</sup> (Top-Bio, Czech Republic), 0.5 U of Taq DNA polymerase (Top-Bio), 0.4 µmol/L primer concentration for each primer, 40 µmol/L concentration of deoxynucleoside triphosphates and 1 µL of template DNA. The PCR was run in the Light Cycler96 instrument (Roche, USA) under the following conditions: initial denaturation at 95◦C for min, 30 cycles of denaturation (95◦C for 30 s), annealing (58◦C for 30 s), extension (72◦C for 60 s), and final extension at 72◦C for 7 min.

MLST of seven housekeeping genes (adk, fumC, gyrB, icd, mdh, purA, recA) was performed according to the MLST protocol standardized for E. coli (http://mlst.warwick.ac.uk/mlst/).

For each of the 144 samples, 11 amplicons (4 topoisomerase and 7 MLST amplicons) were pooled in an equimolar ratio to generate one single sample and sent to the laboratories of the Institute of Applied Biotechnologies (Prague, Czech Republic) for sequencing. DNA libraries were constructed using the Nextera XT DNA Sample Preparation Kit (Illumina, Inc., San Diego, CA) and two multiplex sequencing assays (1 × 96 samples, 1 × 48 samples) were performed on the Illumina MiSeq platform (Illumina, Inc. San Diego, CA).

# NGS Data Analysis

A quality control check of pair-end FASTQ files was performed as a first step. Bases with Phred Quality Score lower than defined threshold (threshold = 30) were filtered out. Quality control included check for possible adapter contamination. Then all variants of reference sequences of single housekeeping genes of E. coli (adk, fumC, gyrB, icd, mdh, purA, recA) in FASTA file were downloaded (http://mlst.warwick.ac.uk/mlst/dbs/Ecoli) and indexed by Burrows-Wheeler Aligner (BWA version 0.7.13; Li and Durbin, 2010). Paired-end FASTQ files were aligned against all indexed reference FASTA files using BWA-mem algorithm with minimum seed length = 19, matching score = 1, mismatch penalty = 4, gap open penalty = 6, and gap extension penalty = 1. Generated Sequence Alignment Map (SAM format) was converted into its binary form (BAM format) and sorted by coordinates. Unmapped sequences reads were filtered out from BAM files. Algorithms Samtools (samtools version 0.1.18) and VarScan (Koboldt et al., 2009) were applied for variant calling using the following parameters: minimum coverage = 10, minimum mapping quality = 10, minimum reads = 2, minimum variant frequency = 0.05, and p = 0.05 were used for variant calling. Generated Variant Calling Format (VCF format) was statistically analyzed and a minimum number of variants (required 0, this number means that our FASTQ files have 100% match with a variant of the reference gene) was evaluated like a gene form of selected housekeeping gene of E. coli. All bioinformatics analyses were performed on Operating System Linux (Ubuntu 14.04 LTS) and using programming language Python.

Single point mutations in selected part of gyrA, gyrB, parC and parE including QRDRs were evaluated using the Integrative Genomics Viewer IGV version 2.3. (Broad Institute, Cambrige, UK) (https://www.broadinstitute.org/software/igv). The genome sequence E. coli MG1655 (NC\_000913.2) was selected as a reference sequence for alignment.

Sanger sequencing was performed with 11 samples to confirm the results of MLST analysis carried out by NGS. This was done because some problems occurred in the assignment of the appropriate allele in MLST analysis. Gene sequences were subsequently submitted to the E. coli MLST Database, University of Warwick (http://mlst.warwick.ac.uk/mlst/dbs/ Ecoli/). Obtained MLST allele sequences were compared with the results generated by NGS.

# Pulsed-Field Gel Electrophoresis

E. coli isolates belonging to the same sequence type (ST) according to MLST analysis and originating from different sources (human, animal or wastewater) were subjected to examination of their epidemiological relationship using pulsedfield gel electrophoresis (PFGE). The total genomic DNA was isolated from overnight bacterial culture according to the procedure published (Husickova et al., 2012). All samples were digested with XbaI (30 U for 3 h) (Takara, Bio, Otsu, Shiga, Japan) and subjected to PFGE. The resulting restriction profiles were analyzed with the GelCompar II software (Applied Maths, Kortrijk, Belgium) using the Dice coefficient (1,5%) for comparing similarity and unweighted pair group method using arithmetic averages for cluster analysis. The results were interpreted according to criteria described by Tenover et al. (1995).

# RESULTS

# PMQR Genes in *E. coli* Isolates with Reduced Susceptibility or Resistance to Ciprofloxacin

A collection of 1050 isolates of E. coli with reduced susceptibility or resistance to ciprofloxacin was obtained from humans, food producing, or wild animals and wastewater samples from the area of the Olomouc and South Moravian Regions, Czech Republic (**Table 1**). The highest occurrence of E. coli strains isolated from MCACIP of animal originwas detected from turkey samples (105 isolates per 120 samples), the lowest from chicken samples (156 isolates per 2628 samples). In human population, 303 E. coli isolates were selected according to the procedure described in material and methods section. From the total amount of 124 wastewater samples, 372 E. coli isolates were analyzed.

PMQR genes were identified in 262 (25%) E. coli isolates (**Table 1**). The gene qnrS1 was the most prevalent, found in 56% of PMQR-positive isolates from all studied areas. It was identified in 76% PMQR-positive isolates from poultry farms, turkey, rooks and wastewater. The second most common gene was aac(6′ )- Ib-cr found in 82% of human PMQR-positive clinical isolates followed by qnrB19 that dominated in isolates from turkey and wastewater samples. The qnrA, qnrC, and qepA genes were not found.

Twenty-three isolates carried two different PMQR genes of the following combination: qnrB19+qnrS1 (12 isolates), qnrB1+aac(6′ )-Ib-cr (7), qnrB4+aac(6′ )-Ib-cr (2), qnrS1+oqxAB (1) qnrB10+qnrS1 (1).


Röderova et al. Quinolone Resistance in Escherichia coli

In case of wastewaters where more than one E. coli colony per sample was examined, the occurrence of several E. coli isolates carrying different PMQR genes and originating the same water sample was detected. Six wastewater samples contained E. coli isolates carrying qnrS1 together with E. coli carrying qnrB19 were detected. In one sample E. coli harboring qnrB1 and E. coli with qnrS1 were found. The last sample contained two different types of E. coli, one carrying aac(6′ )-Ib-cr and the other with qnrS1 .

# Antimicrobial Susceptibility of PMQR-Positive Isolates

For each PMQR-positive E. coli isolate, susceptibility to 23 antimicrobial agents was tested (Table S2). Isolates from hospitalized patients showed the highest level of resistance including ampicillin (97%), piperacillin (96%), third generation cephalosporins (63% for cefotaxime and 43% for ceftazidime), tetracycline (83%), and tobramycin (86%). Isolates from the community displayed high level of resistance to ampicillin (93%), tetracycline (93%), piperacillin (93%), tobramycin (64%), and resistance to third generation cephalosporins (50.0% for cefotaxime and 43% for ceftazidime). On the other hand, wastewater isolates were less resistant to tested antibiotics. High level of resistance was detected only for ampicillin (79%), tetracycline (83%) and piperacillin (78%).

Resistance to ampicillin was common also among isolates from retailed turkeys (96%), chicken farms (71%) and rook feces (75%). Other frequent resistance phenotypes included tetracycline and piperacillin found in 77 and 59% of chicken, in 86 and 90% from turkey and in 75 and 70% from rook isolates, respectively. The majority of E. coli isolates from food-producing animals were susceptible to third generation cephalosporins while 20% of rook isolates showed resistance. Most isolates from the inpatients and the community showed resistance to ciprofloxacin. A total of 270 isolates were resistant (MIC of ciprofloxacin >0.5 mg/L) and 33 sensitive (MIC of ciprofloxacin ≤0.5 mg/L) according to the EUCAST criteria. However, this results is influenced by different selection criteria for human isolates since isolates with MIC of ciprofloxacin >0.5 mg/L were preferably included in the study. In contrast, only 50 of all PMQR-positive isolates from other studied areas showed resistance to ciprofloxacin with the highest level observed in isolates from chicken farms (41%). A total of 36% of animal E. coli isolates (chicken, turkey, and rook) were resistant to oxolinic acid, 32% to ofloxacin and 28% to ciprofloxacin. In the group of wastewater samples, 28% of isolates displayed resistance to oxolinic acid as well as to ciprofloxacin, and 34% to ofloxacin. The distribution of MIC values of tested quinolones in selected isolates ( n = 144) and the comparison with the results of genetic detection is summarized in the **Table 2** .

A total number of 66 isolates was screened for ESBLs using phenotypic and genetic methods. Production of ESBL enzymes was found in 61 E. coli isolates. Subsequent genetic analysis revealed the presence of blaCTX - M -<sup>15</sup> in 51 isolates (hospital, community, wastewater). Two isolates from chicken and wastewater carried the gene blaCTX - M - 1 , blaCTX - M -<sup>14</sup> was found in two isolates from hospital and wastewater and blaSHV -12

 a

 one


Distributionofsubstitutionsintopoisomerasesubunits,PMQRdeterminantsandMICoftestedquinolonesin

was detected in three isolates from wastewater. Three isolates showing ESBL phenotype were negative for all tested bla genes.

# Mutations in the Topoisomerase Genes in PMQR-Positive Isolates

Various mutations in gyrA, parC or parE genes were found in 60 out of 144 PMQR-positive isolates while no isolate with mutation in gyrB was observed (**Table 2**).

The most common GyrA substitution was Ser(83)Leu found in 59 isolates (41%), followed by Asp(87)Asn, detected in 49 isolates (34%). A total of 49 isolates carried double or triple substitution in GyrA.

Two aminoacid substitutions in ParC were found in 25 isolates (17%) with Ser(80)Ile and Glu(84)Val being the most common combination (n = 23). The most common substitution in ParE was Asp(475)Glu found in 18 isolates (13%).

All the E. coli isolates with a high level of ciprofloxacin resistance (>32 mg/L) were shown to carry double mutations in gyrA in combination with single or double mutations in parC. Substitutions at codon 83 and 87 in GyrA along with substitutions at codon 80 and 84 in parC gene were the most common. One isolate carried three mutations in gyrA with aminoacid changes at positions 83, 84, and 87. The majority of high-level ciprofloxacin resistant isolates were obtained from hospitalized patients and possessed the gene aac(6′ )-Ib-cr. Other isolates with high level ciprofloxacin resistance originated from community and WWTP3 also carried aac(6′ )-Ib-cr.

In most isolates susceptible to ciprofloxacin according to the EUCAST criteria (≤0.5 mg/L), no mutations in gyrA, gyrB, or parC were detected. However, five isolates possessed single aminoacid substitution at the codon 83 in GyrA (MIC of ciprofloxacin was 0.5 mg/L). All isolates were susceptible to almost all tested antimicrobial agents except ofloxacin (the MIC was 16–64 mg/L) and three of them were resistant to ofloxacin (MIC = 2 mg/L). Substitutions at position 378 [Arg(378)His)] or 475 [Asp(475) Glu] in ParE were found in 18 isolates susceptible to ciprofloxacin (MIC of ciprofloxacin was 0.2 or 0.5 mg/L). These strains displayed high level of resistance to ampicillin and piperacillin.

# Clonal Relationship of PMQR-Positive Isolates

In the group of 144 isolates, analyzed by MLST, 48 different STs were identified (Table S3) including five major clones ST131 (n = 26), ST355 (19), ST48 (13), ST95 (10), and ST10 (5). Isolates of ST131 were obtained from various types of human clinical materials and from one wastewater sample, while the second most common clone ST355 carrying the qnrS1 or the combination of qnrS1+qnrB19 genes was found predominantly from retailed turkey. MLST analysis also showed the presence of identical STs in samples of different origin. Isolates of ST88 or ST226 were identified from chicken farms as well as from retailed turkeys. ST10, ST48, and ST95 were found in the poultry and wildlife (chicken farms, retailed turkey, and rooks) and in wastewater samples. Other sequence types identified in samples of different origin included ST428, ST533, and ST617. The determination of relevant ST failed in four samples. In these samples bioinformatics analysis did not assign appropriate allele variant and Sanger sequencing did not provide relevant sequences, suggesting the isolates belonged to atypical E. coli isolates untypable by MLST.

ST131 was most often associated with blaCTX-M-<sup>15</sup> and aac(6′ )-Ib-cr, however one strain harbored the combination of blaCTX-M-<sup>15</sup> a qnrB1. The production of CTX-M-15 enzyme was also detected together with qnrB1 in one ST393 and two ST410 strains. In three ST405 and one ST617 a combination of blaCTX-M-<sup>15</sup> and aac(6′ )-Ib-cr was identified. One ST48 E. coli isolate harbored blaCTX-M-15, together with qnrB1 and aac(6′ )- Ib-cr. In one qnrS1-postitive ST43 strain, blaCTX-M-<sup>1</sup> gene was detected. Two qnrS1-positive and SHV-12 producing strains, one belonging to ST540 and the other to ST58 were identified. The other CTX-M-14 or CTX-M-15 producing strains were not analyzed using MLST.

E. coli isolates belonging to the same ST and originating from different sources (human, animal or environmental) were subjected to PFGE to determine the level of their genetic relatedness (**Figure 1**). Based on these criteria a total of 68 isolates of the following STs were analyzed: ST10 (n = 5), ST48 (n = 13), ST58 (n = 2), ST88 (n = 3), ST95 (n = 10), ST131 (n = 26), ST226 (n = 2), ST428 (n = 2), ST533 (n = 2), and ST617 (n = 3). Overall, a high diversity of restriction patterns was observed in isolates of the same ST. No isolates with 100% identity of PFGE profiles originating from different areas (human, animal, or environment) were identified.

Globally disseminated ST131 was predominant in samples from hospitalized patients as well as from community subjects while no isolates of this clonal group were found in foodproducing animals or rooks. The majority (69%) of E. coli ST131 isolates collected from these samples displayed high-level of ciprofloxacin resistance (>32 mg/L), harbored double or triple mutations in gyrA along with double mutations in parC gene and the PMQR gene aac(6′ )-Ib-cr. The overall similarity of PFGE patterns of ST131 isolates was <50% confirming high genetic diversity of this clonal group. Only two E. coli ST131 isolates originating from the same patient from which the samples were taken 17 days apart shared 100% identity.

Significant clonal similarity was detected only in isolates originating from single collection areas, including ST10, ST48, ST88, and ST95 from retailed turkeys. In the ST10 group two isolates shared 100% similarity of their PFGE profiles. Both isolates were susceptible to ciprofloxacin, carried qnrS1, and did not contain gyrase or topoisomerase mutations. A group of six closely related strains was detected in ST48 group. All of them were collected from turkeys, displayed susceptibility or a low level of resistance to ciprofloxacin (MIC 1 - 2 mg/L), harbored qnrS1 and were originated from the same turkey farm. Another three ST48 isolates from WWTP5 displayed 100% identity of their PFGE patterns, originated from wastewater samples obtained from the same collection date and shared only 52% similarity of the PFGE patterns with ST48 isolates from turkey.

In the ST95 group, two and three strains from turkeys showed identical PFGE profiles. All isolates of the ST95 group were susceptible to ciprofloxacin, did not possess any mutations in gyrase or topoisomerase genes and most of them carried both qnrB19 and qnrS1 genes.


FIGURE 1 | Dendrogram of *E. coli* isolates including information about origin, PMQR genes, gyrase and topoisomerase substitutions, MIC of ciprofloxacin and beta-lactamases.

# DISCUSSION

Quinolones are antimicrobial agents widely used in hospitals, community and veterinary practice. According to the combined data of ECDC, EFSA, EMA (ECDC, EFSA, and EMA, 2015), the consumption rate of fluoroquinolones reached 3.2 tons in primary care sector in the Czech Republic in 2012. The same data reported that 1.2 tons of fluoroquinolones were used in food producing animals in the country. However, the consumption of this class of antibiotics is still relatively low compared to other countries such as Germany, France, or Italy (ECDC, EFSA, and EMA, 2015).

The first aim of this study was to investigate the occurrence of E. coli isolates with PMQR genes in humans, food-producing animals, wild animals and wastewater samples from a defined area of the Czech Republic. The results pointed at 25% occurrence of E. coli with PMQR genes in all areas studied. The most common PMQR gene in E. coli from humans was aac(6′ )-Ib-cr found mainly in ST131 clone with a high level of fluoroquinolone resistance. It has been demonstrated that aac(6′ )-Ib-cr gene is more prevalent in E. coli ST131 than other PMQR genes (Cerquetti et al., 2009; Peirano et al., 2010; Hussain et al., 2012) and it is carried mainly by plasmids of incompatibility group F (IncF) along with genes encoding extended-spectrum beta-lactamase CTX-M or resistance to other antimicrobial classes (Woodford et al., 2009; Nicolas-Chanoine et al., 2014).

Regarding other types of PMQR determinants, qnrS1 gene was identified in all areas examined. It was found (alone or in combination with qnrB19) in 73% of isolates that showed reduced susceptibility to ciprofloxacin and originated from wastewater samples or food-producing animals. Currently, this PMQR variant seems widespread in E. coli isolates from various types of food-producing animals in Europe (Forcella et al., 2010; Veldman et al., 2011). The second most common qnr variant, qnrB19, was found predominantly in the isolates from turkeys and wastewater samples. Both qnrB19 and qnrS1 were previously detected in chicken broilers in the Czech Republic (Literak et al., 2013). In E. coli isolates from rooks, qnrB19, qnrD1, qnrS1 and aac(6′ )-Ib-cr were identified. These PMQR variants have been previously detected in E. coli from rooks sampled in several European countries including the Czech Republic (Literak et al., 2012; Jamborova et al., 2015).

All E. coli isolates with high levels of ciprofloxacin resistance (MIC >32 mg/L), including ST131 clone, carried double or triple mutations in gyrA in combination with single or double mutations in parC. E. coli ST131 with specific substitutions in GyrA or ParC as found in our study have been previously described (Cerquetti et al., 2010; Platell et al., 2011; Paltansing et al., 2013). Isolates from poultry and rooks that displayed clinical levels of ciprofloxacin resistance linked to single or double mutations in gyrA and parC were found in our study. Strains of various animal origin with these types of mutations have been recently found worldwide (Sáenz et al., 2003; Liu et al., 2012; Castillo et al., 2013; Johnning et al., 2015; Balakrishnan et al., 2016).

ST131 was the predominant clone in our study found in the hospitalized patients as well as in the community subjects and also in one sample from the hospital WWTP. E. coli ST131 is a worldwide multidrug resistant clone responsible mainly for urinary but also for other types of infections and widely disseminated in hospital and community patients worldwide (Johnson et al., 2003; Cerquetti et al., 2010; Lee et al., 2010; Nicolas-Chanoine et al., 2014). In the Czech Republic, isolates of this ST producing CTX-M beta-lactamases were found in patients suffering from community or hospital acquired infections (Dolejska et al., 2012; Micenková et al., 2014; Papagiannitsis et al., 2015). Our results also pointed at the occurrence of CTX-M-15-producing ST131 clone in UHO. The ST131 clone has also been isolated from wildlife, companion animals, and retail chicken meat (Ewers et al., 2010; Ghodousi et al, 2016; Oh et al., 2016) as well as from wastewater or water samples (Dolejska et al., 2011a; Colomer-Lluch et al., 2013; Varela et al., 2015).

ST10 clonal complex (including ST10, ST34, ST43, ST48, ST167, and ST617) was the second most widespread in our study. It represents one of the largest clonal complexes within the E. coli MLST database, generally being antimicrobial-susceptible, with low virulence (Manges and Johnson, 2012) but it has also been associated with strains producing ESBL or AmpC betalactamases (Alcalá et al., 2016; Ben Said et al., 2016; Wang et al., 2016). Based on our results, one ST48 E. coli producing CTX-M-15 from wastewater, one ST617 strain with CTX-M-15 from wound swab and one ST43 E. coli with production of CTX-M-1 beta-lactamase from chicken were identified. In contrast to other studies (Huijbers et al., 2014; Agabou et al., 2016), no epidemiological link between human and animal isolates of this clonal group was observed in our study.

ST95 was the third largest clonal group in our study including isolates from cloacal swabs from market-weight turkeys and one isolate from a wastewater sample. This clone has been associated with infections in domestic (Mora et al., 2013) and wild birds (Jamborova et al., 2015) as well as urinary tract infections in humans (Riley, 2014). In contrast to ST131, ST95 lineage includes less multidrug resistant strains (Gibreel et al., 2012; Adams-Sapper et al., 2013). ST355 prevalent in our samples from marketweight turkey, seems to be not as widespread in the world as previous ones. It has been documented in E. coli isolates from urine and rectal samples in a multinational report by Adler et al. (2014) and in a study from the U.S. (Tartof et al., 2007).

Several authors suggested potential transmission of E. coli clones from animals to humans (Platell et al., 2011; Agabou et al., 2016; Nüesch-Inderbinen and Stephan, 2016) and the role of contaminated food in the local spread of resistant E. coli strains (Vincent et al., 2010). In a study from Platell et al. (2011), a high level of relatedness of E. coli ST131 strains from companion animals and humans was observed. Moreover, Agabou et al. (2016) demonstrated a clonal relationship between human and chicken ciprofloxacinresistant E. coli isolates in North-Eastern Algeria. In contrast to these studies, we did not find any correlation between samples of human, animal, or environmental origin. No strains sharing 100% relatedness and originating from different areas of collection were identified using PFGE. However, some limitations of our study might affect the final results. These include the preferable selection of E. coli isolates resistant to ciprofloxacin from clinical materials of hospitalized patients and community subjects, the preferable selection of urine samples and the absence of rectal samples from a healthy population.

# AUTHOR CONTRIBUTIONS

Conception and design of the study: JB, AC, MK, IL. Sample collection: MR, DH, IP, MM, VH, VP, MH, PS, JB, and AC. Data analysis and interpretation, Final approval of the version to be published and drafting the article: MR, DH, IP, MD, MM, VH, VP, PB, MH, PS, JB, AC, MK, and IL. Critical revision of the article: MR, DH, IP, MD, MK, and IL. Agreement to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved: MR, DH, IP, MD, MM, VH, VP, PB, MH, PS, JB, AC, MK, and IL.

# FUNDING

This work was supported by the grant Internal Grant Agency of Ministry of Health of the Czech Republic (NT/14398), Internal Grant Agency of the Palacký University Olomouc (IGA\_LF\_2016\_022), Internal Grant Agency of the University of Veterinary and Pharmaceutical Sciences Brno, Brno, Czech Republic (217/2015/FVHE) and CEITEC 2020 - Central European Institute of Technology (CZ.1.05/1.1.00/02.0068) from European Regional Development Fund (LQ1601) from the Czech Ministry of Education, Youth and Sports within the National Programme for Sustainability II.

# ACKNOWLEDGMENTS

The authors would like to thank L. Cekanova, P. Julinkova, E. Sotova, K. Konopacová, D. Lencianova, V. Kindlova, L. Smetanova, V. Hruby, N. Hostakova, J. Klimes and the stuff of

# REFERENCES


single wastewater treatment plants for their kind cooperation in the field and in the laboratory. Our special thanks go to Lars Hansen (University of Copenhagen, Denmark) and Lina Cavaco (National Food Institute, Copenhagen, Denmark) for control strains.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2016.02147/full#supplementary-material


Escherichia coli clones causing urinary tract infections in Korea. J. Infect. 60, 146–153. doi: 10.1016/j.jinf.2009.11.004


the Chicago area: high prevalence of ST131 producing CTX-M-15 in community hospitals. Int. J. Antimicrob. Agents. 36, 19–23. doi: 10.1016/j.ijantimicag.2010.02.016


streams and municipal and hospital effluents. FEMS Microbiol. Ecol. 91:fiv015. doi: 10.1093/femsec/fiv015


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Röderova, Halova, Papousek, Dolejska, Masarikova, Hanulik, Pudova, Broz, Htoutou-Sedlakova, Sauer, Bardon, Cizek, Kolar and Literak. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Characteristics of Quinolone Resistance in Salmonella spp. Isolates from the Food Chain in Brazil

Bruno R. Pribul 1, 2 \*, Marcia L. Festivo<sup>1</sup> , Marcelle S. Rodrigues <sup>1</sup> , Renata G. Costa<sup>1</sup> , Elizabeth C. dos P. Rodrigues <sup>1</sup> , Miliane M. S. de Souza<sup>2</sup> and Dalia dos P. Rodrigues <sup>1</sup>

<sup>1</sup> National Reference Laboratory for Enteric Diseases, Oswaldo Cruz Institute(FIOCRUZ), Rio de Janeiro, Brazil, <sup>2</sup> Laboratory of Veterinary Bacteriology, Federal Rural University of Rio de Janeiro, UFRRJ, Rio de Janeiro, Brazil

#### Edited by:

Axel Cloeckaert, Institut National de la Recherche Agronomique, France

#### Reviewed by:

Séamus Fanning, University College Dublin, Ireland Iddya Karunasagar, Nitte University, India

> \*Correspondence: Bruno R. Pribul bpribul@gmail.com

#### Specialty section:

This article was submitted to Antimicrobials, Resistance and Chemotherapy, a section of the journal Frontiers in Microbiology

Received: 29 September 2016 Accepted: 14 February 2017 Published: 14 March 2017

#### Citation:

Pribul BR, Festivo ML, Rodrigues MS, Costa RG, Rodrigues ECdP, de Souza MMS and Rodrigues DdP (2017) Characteristics of Quinolone Resistance in Salmonella spp. Isolates from the Food Chain in Brazil. Front. Microbiol. 8:299. doi: 10.3389/fmicb.2017.00299 Salmonella spp. is an important zoonotic pathogen related to foodborne diseases. Despite that quinolones/fluoroquinolones are considered a relevant therapeutic strategy against resistant isolates, the increase in antimicrobial resistance is an additional difficulty in controlling bacterial infections caused by Salmonella spp. Thus, the acquisition of resistance to quinolones in Salmonella spp. is worrisome to the scientific community along with the possibility of transmission of resistance through plasmids. This study investigated the prevalence of plasmid-mediated quinolone resistance (PMQR) in Salmonella spp. and its association with fluoroquinolone susceptibility in Brazil. We evaluated 129 isolates, 39 originated from food of animal sources, and 14 from environmental samples and including 9 from animals and 67 from humans, which were referred to the National Reference Laboratory of Enteric Diseases (NRLEB/IOC/RJ) between 2009 and 2013. These samples showed a profile of resistance for the tested quinolones/fluoroquinolones. A total of 33 serotypes were identified; S. Typhimurium (63) was the most prevalent followed by S. Enteritidis (25). The disk diffusion test showed 48.8% resistance to enrofloxacin, 42.6% to ciprofloxacin, 39.53% to ofloxacin, and 30.2% to levofloxacin. According to the broth microdilution test, the resistance percentages were: 96.1% to nalidixic acid, 64.3% to enrofloxacin, 56.6% to ciprofloxacin, 34.1% to ofloxacin, and 30.2% to levofloxacin. Qnr genes were found in 15 isolates (8 qnrS, 6 qnrB, and 1 qnrD), and the aac(6′ )-Ib gene in 23. The integron gene was detected in 67 isolates with the variable region between ±600 and 1000 bp. The increased detection of PMQR in Salmonella spp. is a serious problem in Public Health and must constantly be monitored. Pulsed-field gel electrophoresis was performed to evaluated clonal profile among the most prevalent serovars resistant to different classes of quinolones. A total of 33 pulsotypes of S. Typhimurium were identified with a low percentage of genetic similarity (≤65%). This result demonstrates the presence of high diversity in the resistant clones evaluated in this study.

Keywords: foodborne diseases, Salmonella spp., quinolone resistance, plasmid mediated quinolone resistance, clonal profile

# INTRODUCTION

Foodborne diseases caused by Salmonella spp. are a serious public health problem in many parts of the world. The variety of food sources, particularly foods of animal origin, and routes of transmission can lead to human infection (Scallan et al., 2011).

In addition, the progressive increase of antimicrobial resistance in foodborne Salmonella isolates is observed as due to the uncontrolled use of these drugs for therapeutic and prophylactic purposes in foods of animal origin such as poultry, pigs, and cattle. These events have reinforced the need for epidemiological studies describing the prevalence and patterns of resistance in these bacteria (Yang et al., 2010; Tamang et al., 2011). Antimicrobial-resistant bacteria emerge from the use of antimicrobial drugs to treat and prevent diseases and promote growth in large-scale animal production.

Quinolones, particularly fluoroquinolones, are commonly used for the treatment of multi-drug resistant salmonellosis "in human and veterinary medicine" because of their broad spectrum antimicrobial activity (Dalhoff, 2012).

Point mutations in DNA gyrase and topoisomerase IV genes are directly related to quinolone resistance in Enterobacteriaceae "by changes in the action target site called quinolone resistancedetermining regions (QRDR)." "In Salmonella spp., these mutations are related to resistance to nalidixic acid (NAL) and reduced susceptibility to FQs such as that of ciprofloxacin (Cip) (Cavaco and Aarestrup, 2009)." It is believed that the resistance to quinolones is mediated only by this mechanism. However, the situation changed with the discovery of a variety of determinants of plasmid-mediated quinolone resistance (PMQR).

Currently three mechanisms are recognized as PMQRs. The qnr genes with five different qnr families, each with different numbers of alleles "(qnrA1–7, qnrS1–4, qnrB1–31, qnrC, and qnrD) (Jacoby et al., 2009)"; "a modified aminoglycoside acetyltransferase gene [aac(6′ )-1b-cr] (Robicsek et al., 2005)"; and a specific quinolone efflux pump (qepA) (Yamane et al., 2007) and multidrug resistance pumps such as oqxAB (Zhao et al., 2010). PMQR-positive isolates present a low-level of resistance to quinolones (only a small reduction in susceptibility to nalidixic acid). However, the ability to highlight pre-existing resistance mechanisms, such as chromosomal mutations in target regions of quinolones that still allow the selection of resistant mutants to quinolone concentrations (therapeutic doses), emphasize the importance of studying these genes (Cui et al., 2014).

The present study identified the occurrence of some PMQR in Salmonella spp. isolated between 2009 and 2013 from the food chain in Brazil, and characterized the genetic similarity profile of serovars of greatest importance in the dispersion of resistance to quinolones in Brazil.

# MATERIALS AND METHODS

# Bacterial Isolates

129 Salmonella spp. strains with resistance to quinolone and/or fluoroquinolone were evaluated. Of these, 51.9% (67/129) were from human clinical isolates, 30.2% (39/129) from food products for human consumption (beef, eggs, milk), 7.1% (9/129) from food of animal origin for human consumption (poultry, swine, cattle), and 10.8% (14/129) from environmental samples (water and drag swabs); all samples were selected from a database. The studied strains were sent to the National Reference Laboratory of Enteric Diseases (NRLEB/IOC/RJ) between 2009 and 2013 and stored in phosphate-buffered agar at room temperature and/or in BHI/glycerol broth −70◦C. The isolates were inoculated in Nutrient Broth (DIFCO) and incubated at 37◦C for 12–18 h for subsequent tests, as confirmation of the biochemical, serological and antimicrobial resistance profile.

# Antigenic Characterization

The serological determination of Salmonella serotypes was determined according to the Kauffmann-White scheme using slide agglutination with O and H antisera prepared in the LRNEB/IOC/RJ.

# Antimicrobial Susceptibility

The obtained resistance profiles were confirmed by the disk diffusion test according to Clinical and Laboratory Standards Institute (2013, 2014). According to Pribul et al. (2016), this test was performed using representatives of the quinolone class (OXOID) for human and veterinary therapeutic use, such as Nalidixic Acid (NAL), Ciprofloxacin (CIP), Enrofloxacin (ENO), Ofloxacin (OFL), and Levofloxacin (LVX).

MIC determinations for Nalidixic Acid (SIGMA), Ciprofloxacin (SIGMA), Enrofloxacin (SIGMA), Levofloxacin (SIGMA), and Ofloxacin (SIGMA) were performed in 96-well-microplates and according to the Clinical and Laboratory Standards Institute (2013) broth microdilution assay.

# Detection of PMQR

Total DNA was extracted using the DNEASY Tissue Qiagen <sup>R</sup> kit. The studied genes were detected by PCR amplification using primer sequences reported in Pribul et al. (2016). The qnrA, qnrB, and qnrS genes were amplified through multiplex PCR reactions; the rrs gene was used as the reaction control. The qnrC, qnrD, aac(6′ )-Ib, integrase, and variable integron region genes were amplified by simplex PCR.

# PFGE

The isolates from serovars S. Typhimurium, S. Muenchen, S. Infantis, and S. Heidelberg were subjected to molecular typing by pulsed-field gel electrophoresis, which clonally evaluates isolates. The PulseNet protocol was used in this study including DNA preparation according to Heir et al. (2000) and digestion with XbaI restriction enzyme, according to Pfaller et al. (1992), Tenover et al. (1997), and Cooper et al. (2006). The definition of clones was based on the recommendations of Tenover et al. (1997) and Barrett et al. (2006). S. Braenderup H9812, which is considered the universal strain for PulseNet (Hunter et al., 2005), was used as the standard. The restriction patterns were analyzed in the BioNumerics software IV (Applied Maths).

#### Pribul et al. Quinolone Resistance in Salmonella spp. in Brazil

# RESULTS

## Serovar Identified

Altogether, 26 different Salmonella serovars were identified. Salmonella Typhimurium (48.8%, 63/129) was the predominant serovar followed by Salmonella Enteritidis (19.4%, 25/129). The prevalent serovars associated with resistance to quinolones are presented in **Table 1**.

Most of the studied samples were isolated in 2012 (88 of 129).

Among these 129 isolates that were previously resistant to Nalidixic Acid, five were sensitive to all tested quinolones (including Nalidixic Acid), 55 (42.6%) were resistant to Ciprofloxacin, 63 (48.8%) to Enrofloxacin, 51 (39.53%) to Ofloxacin, and 48 (37.2%) to Levofloxacin in the disc diffusion test.

The broth microdilution test identified 36.4% (47/129) isolates with decreased susceptibility to Ciprofloxacin (MICs between 0.125 and 0.5 mg/ml), 20.1% (26/129) to Enrofloxacin, 9.3% (12/129) to Ofloxacin, and 6.2% (8/129) to Levofloxacin (MICs between 0.5 and 1 mg/ml). The decreased susceptibility breakpoint to Nalidixic Acid is not reported by Clinical and Laboratory Standards Institute (2015). Seventy-three (56.6%) isolates were resistant to Ciprofloxacin, 83 (64.3%) to Enrofloxacin, 44 (34.1%) to Ofloxacin, and 39 (30.2%) to Levofloxacin. A total of 124 isolates (96.1%) were resistant to Nalidixic Acid.

The resistance profile obtained with the microdilution test showed that 37 (28.7%) isolates were resistant to all tested quinolones, 30 (23.2%) to Ciprofloxacin, Enrofloxacin, and Nalidixic Acid, 16 (12.4%) to Enrofloxacin and Nalidixic Acid, 2 (1.5%) to Ciprofloxacin and Nalidixic Acid, and 39 (30.2%) to Nalidixic Acid only.

The detection of resistance genes showed six isolates carrying the qnrB gene, eight the qnrS gene, and one the qnrD gene. Among these 15 positive isolates, 10 strains were recovered from human samples, 3 from food of animal origin, 1 from environmental samples, and 1 from animal samples. The most qnr-positive prevalent serovar was S. Typhimurium followed by S. Saintpaul and S. Livingstone. None of the isolates presented the qnrA or qnrC genes.

A total of 23 isolates showed the aac(6′ )-Ib gene, which is prevalent in S. Typhimurium (14/23). The most prevalent source of isolation was human (10/23), followed by foodborne (7/23), animal (3/23), and amibental (3/23). Thirteen isolates aac(6′ )-Ib positive were resistant to all tested quinolones.

Three qnr-positive isolates presented the aac(6′ )-Ib gene in association: two S. Typhimurium and one S. Saintpaul. These two Salmonella ser. Typhimurium were resistant to all tested quinolones/fluoroquinolones in the broth microdilution assay at the highest concentration.

Sixty-seven isolates showed the presence of integrase gene within 600 to 1000 bp variable region range and were mainly identified in human samples (38/67) followed by food samples (15/67), ambiental (8/67), and animal (6/67). The S. Typhimurium serovar was the most frequent (39/67) isolate with the conserved region of class 1 integron and variable regions between ±900 and >1000 bp. Nine isolates of serovar S. Typhimurium carrying the aac(6′ )-Ib gene were positive for TABLE 1 | Distribution of quinolone-resistant Salmonella spp. serovars isolated from food chain diseases.


<sup>a</sup>Non-typhoidal Salmonella.

the integron with ±900 bp; two of these were also positive for qnr.

**Figure 1** shows the clonal profile comparison between the quinolones/fluoroquinolones resistant strains and other sensitive strains.

The genetic similarity among isolates with resistance to quinolones was approximated 84% in S. Infantis, 92% in S. Heidelberg, 88% in S. Muenchen, and 63% in S. Typhimurium despite their isolation in different periods, regions, and sources.

Thirty-three distinct pulsotypes were identified among strains with low percentage genetic similarity in serovar S. Typhimurium (≤65%), representing the highest diversity among resistant clones.

The **Table 2** presents resistance profiles obtained with the microdilution test, detection of PMQR, size of variable integron region, and the pulsotype identified by the pfge technique.

# DISCUSSION

The variation in resistance to the different tested quinolones can be explained by the mechanism of resistance when the resistance level depends on the affected target enzyme, the number of accumulated mutations, and presence of PMQRs. Furthermore, there is a relationship between the level of specific resistance and potency of each drug, especially in the newest quinolones (Sanders, 2001; Ruiz et al., 2012).

Chong et al. (2010) reported that an increased resistance to fluoroquinolones based on the acquisition of qnr genes could be related with reduction in the clinical efficacy "of this class" of antimicrobial. "However, Jacoby et al. (2009)" argue that the genes involved in plasmid-level resistance to fluoroquinolones are still poorly understood when compared to other resistance mechanisms.

A high prevalence of isolates carrying PRQM genes is reported in the present study (27%, 35/129). The most prevalent serovar associated with the presence of PMQR genes was Salmonella ser. Typhimurium (18/35). A high level of detection of S. Typhimurium was expected because this serovar is directly related to detected genotypic and phenotypic profiles of antimicrobial resistance (Herrero et al., 2008; Kingsley et al., 2009). The presence of PMQR genes is related "to decreased susceptibility to fluoroquinolones," accelerating the selection of

Infantis, Heidelberg, Muenchen, and Typhimurium according to PFGE. Isolates with resistance to quinolones; <sup>a</sup> ID of IOC/year of isolation; <sup>b</sup>region; <sup>c</sup> state; <sup>d</sup>source of isolation; <sup>e</sup> serovar; <sup>f</sup>pulsotype.

fluoroquinolone-resistant mutants (Rodríguez-Martínez et al., 2011).

Three isolates presented association between the qnr and aac(6′ )-Ib genes. A similar association has been reported by Park et al. (2006) in the United States, Xiong et al. (2011) while investigating the aac(6′ )-Ib and qnr genes in Enterobacter cloacae "in China, and" Kim et al. (2013) in enterobacteria isolated from clinical samples in Korea. Not sequencing the aac(6′ )-Ib gene to determine the cr variant was one limitation in the present study.

Regardless that some authors recognize the location of the aac(6′ )-Ib gene mostly in class 1 integrons, our results show the absence of this gene in all analyzed strains [12 aac(6′ )-Ib positive isolates without the integron region] (Rodríguez-Martínez et al., 2011; Kim et al., 2013).

The Enteritidis serovar was not assessed by PFGE because, according to the literature, these isolates have low clonal diversity (Spiliopoulou et al., 2007).

Six distinct pulsotypes were detected in S. Infantis serovar isolates. Those with resistance to quinolones are placed in two separate pulsotypes (BRJFXX01.13 and BRJFXX01.12) with a genetic similarity of ∼85%. The quinolone resistance isolates were obtained from different sources, regions, and periods, and the resistance to quinolones showed variations. The 6754/12 isolate showed resistance to ciprofloxacin, nalidixic acid, and enrofloxacin and carried the qnrD gene; the 9606/10 isolate showed resistance to nalidixic acid only and did not carry resistant genes.

The S. Heidelberg serovar presented eight distinct pulsotypes. The resistant isolates showed a clonal ratio of 100% similarity between the isolates 5/12 and 19/12, and ∼94% between them and isolate 11394/11. The 11394/11 isolate (environmental source from the Southern region) was detected in the BRJF6X01.004 pulsotype (pulsotype with 13 susceptible isolates). Isolates 5/12 and 19/12, within the same pulsotype, were foodborne and originated in the Southern region.

Three pulsotypes were identified in serovar Muenchen isolates with quinolone resistance profiles (pulsotypes BRJJ6X01008, BRJJ6X01007, and BRJJ6X01006), showing 88% of genetic similarity. Among the isolates resistant, the isolates of human origin presenting a profile similarity of ∼97%. The isolates of foodborne origin presenting the same origin clonal being from different periods and states. The 2128/12, 2120/12, and 1192/12 isolates show similar quinolone resistance profiles. However, isolate 851/11 show a resistance profile to ciprofloxacin and ofloxacin. The 2120/12 and 2128/12 isolates show similar resistance profile and are carriers of the qnrS gene.

The detection of 33 different pulsotypes of S. Typhimurium indicates that different clones with resistance to quinolones are circulating in Brazil. The most prevalent pulsotype (BRJPXX01.090) is mainly represented in samples from the Southern region and are related to food and human sources. Most isolates of this pulsotype show the same resistance profile to quinolones/fluoroquinolones (except isolates 3309/11, 5179/10, 591/12, and 667/12) demonstrating a resistance profile to all tested quinolones/fluoroquinolones. Out of the 25 isolates showing resistance to quinolones, 4 did not carry resistance genes. Twelve isolates show an integron with a variable region of

#### TABLE 2 | Resistance profile and resistance genes in isolates evaluated by PFGE.


#### TABLE 2 | Continued


<sup>a</sup>CIP, Ciprofloxacin; ENO, Enrofloxacin; NAL, Nalidixic Ácid; LVX, Levofloxacina; OFL, Ofloxacin; <sup>b</sup>MIC, Minimum Inhibitory Concentration; <sup>c</sup>PMQR, Plasmid-Mediated Quinolone Resistance; <sup>d</sup> IOC ID/Year, Institut Oswaldo Cruz Identification by Year; <sup>e</sup>F, Food; A, Animal; E, Environment; H, Human.

900 bp and one with >1000 bp. One isolate shows the qnrB gene and, four show the aac(6′ )-Ib gene. Two isolates show the 900 bp integron and the aac(6′ )-Ib gene; one isolate shows the 900 bp integron and the qnrB, acc(6′ )-Ib gene.

The profiles identified in the PFGE analysis show relatively high diversity in all serovars and, indicate that cases of resistance to quinolones are probably sporadic. This interpretation is in

## REFERENCES


accordance with other results reported in the literature (Feasey et al., 2012).

# AUTHOR CONTRIBUTIONS

All authors listed, have made substantial, direct and intellectual contribution to the work, and approved it for publication.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Pribul, Festivo, Rodrigues, Costa, Rodrigues, de Souza and Rodrigues. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Plasmids Carrying blaCMY−2/<sup>4</sup> in Escherichia coli from Poultry, Poultry Meat, and Humans Belong to a Novel IncK Subgroup Designated IncK2

Salome N. Seiffert1,2,3, Alessandra Carattoli2,4, Sybille Schwendener<sup>1</sup> , Alexandra Collaud<sup>1</sup> , Andrea Endimiani<sup>2</sup> and Vincent Perreten<sup>1</sup> \*

1 Institute of Veterinary Bacteriology, Vetsuisse Faculty, University of Bern, Bern, Switzerland, <sup>2</sup> Institute for Infectious Diseases, University of Bern, Bern, Switzerland, <sup>3</sup> Graduate School for Cellular and Biomedical Sciences, University of Bern, Bern, Switzerland, <sup>4</sup> Department of Infectious, Parasitic and Immune-Mediated Diseases, Istituto Superiore di Sanità, Rome, Italy

### Edited by:

Benoit Doublet, National Institute for Agricultural Research, France

### Reviewed by:

Christopher Morton Thomas, University of Birmingham, UK Michael Brouwer, Wageningen University and Research Centre, Netherlands

### \*Correspondence:

Vincent Perreten vincent.perreten@vetsuisse.unibe.ch

### Specialty section:

This article was submitted to Antimicrobials, Resistance and Chemotherapy, a section of the journal Frontiers in Microbiology

Received: 27 October 2016 Accepted: 27 February 2017 Published: 15 March 2017

#### Citation:

Seiffert SN, Carattoli A, Schwendener S, Collaud A, Endimiani A and Perreten V (2017) Plasmids Carrying blaCMY−2/<sup>4</sup> in Escherichia coli from Poultry, Poultry Meat, and Humans Belong to a Novel IncK Subgroup Designated IncK2. Front. Microbiol. 8:407. doi: 10.3389/fmicb.2017.00407 The blaCMY−2/4-carrying IncB/O/K-like plasmids of seven Escherichia coli strains from poultry, poultry meat and human urine samples were examined using comparative analysis of whole plasmid sequences. The incompatibility group was determined by analysis of the incRNAI region and conjugation assays with strains containing the IncK and IncB/O reference plasmids. Strains were additionally characterized using MLST and MIC determination. The complete DNA sequences of all plasmids showed an average nucleotide identity of 91.3%. Plasmids were detected in E. coli sequence type (ST) 131, ST38, ST420, ST1431, ST1564 and belonged to a new plasmid variant (IncK2) within the IncK and IncB/O groups. Notably, one E. coli from poultry meat and one from human contained the same plasmid. The presence of a common recently recognized IncK2 plasmid in diverse E. coli from human urine isolates and poultry meat production suggests that the IncK2 plasmids originated from a common progenitor and have the capability to spread to genetically diverse E. coli in different reservoirs. This discovery is alarming and stresses the need of rapidly introducing strict hygiene measures throughout the food chain, limiting the spread of such plasmids in the human settings.

### Keywords: ESBL, pAmpC, food, antibiotic resistance, animals

# INTRODUCTION

The worldwide dissemination of Enterobacteriaceae along with resistance plasmids carrying β-lactamase-encoding genes (bla) is posing an increasing threat to the public health system (Mathers et al., 2015). Poultry and poultry meat have been revealed as a large reservoir of extended-spectrum β-lactamases (ESBL) and plasmid-mediated AmpC (pAmpC) enzymes in several countries (Endimiani et al., 2012; Dierikx et al., 2013; Nilsson et al., 2014; Hansen et al., 2016; Mo et al., 2016). These resistance genes are carried by various mobile genetic elements; primarily located on plasmids which are self-replicating extrachromosomal elements (Couturier et al., 1988; Carattoli, 2009). Such plasmids not only vary in their size and compatibility to other plasmid groups, but also in their capability of transferring to various bacterial species (Carattoli, 2009, 2013).

The involvement of plasmids in the zoonotic spread of β-lactamase- and carbapenemaseencoding genes between human and animal reservoirs has already been emphasized

based on the presence of similar plasmids in Escherichia coli from humans, poultry, and poultry meat (Leverstein-van Hall et al., 2011; Börjesson et al., 2013; Seiffert et al., 2013b; Voets et al., 2013; Guerra et al., 2014; Wang et al., 2014). However, there is a lack of firm evidence since there are only few completely sequenced and assembled plasmids to demonstrate the spread and exchange of the same plasmids in E. coli of human and poultry origin.

By comparing datasets from poultry, poultry meat, and clinical human isolates from Switzerland, we noticed that the pAmpC blaCMY−<sup>2</sup> and blaCMY−<sup>4</sup> genes were associated with plasmids belonging to the IncB/O/K groups (Endimiani et al., 2012; Seiffert et al., 2013a; Vogt et al., 2014). Other studies investigating pAmpC-producing E. coli have detected the nearly identical kind of plasmid in healthy humans (Baudry et al., 2009; Porres-Osante et al., 2015; Hansen et al., 2016), poultry (Dierikx et al., 2010; Hansen et al., 2016; Mo et al., 2016), poultry meat (Egervärn et al., 2014), and dogs (Hansen et al., 2016). This prompted us to perform an in-depth analysis of blaCMY−2/4-carrying IncB/O/K plasmids from E. coli isolated in three settings in Switzerland to determine their genetic relatedness, revealing a novel IncK-group variant, the IncK2.

# MATERIALS AND METHODS

# Bacterial Strains

All strains were obtained from previous studies characterizing 3rd-generation cephalosporin-resistant E. coli (3GC-R-Ec) from poultry, poultry meat, and humans isolated at the Institute of Veterinary Bacteriology and at the Laboratory of Microbiology of the Institute of Infectious Diseases, University of Bern, Switzerland (Endimiani et al., 2012; Seiffert et al., 2013a; Vogt et al., 2014). Strains were collected between 2011 and 2014 (**Table 1**) and were selected based on the presence of blaCMY−2/<sup>4</sup> by PCR/DNA sequencing (Seiffert et al., 2013a; Vogt et al., 2014) and on ambiguous results with the PCR-based replicon typing (PBRT) of plasmids, giving cross-reaction with either IncK and/or IncB/O PCRs (Carattoli et al., 2005).

Plasmids were transformed by electroporation into either E. coli ElectroMax DH10B competent cells or into E. coli One Shot <sup>R</sup> TOP10 Electrocomp cells (Invitrogen, Life Technologies) as previously described (Seiffert et al., 2013a; Vogt et al., 2014). The MIC of antibiotics were obtained from donors and transformants using the microdilution Sensititre panels ESB1F and EUMVS2 (Trek Diagnostics) and interpreted following the CLSI guidelines (CLSI, 2016). Sequence types (ST) were assigned following the multi-locus sequence typing (MLST) Achtman scheme<sup>1</sup> .

# Whole Plasmid Sequencing (WPS)

Plasmids were purified using the Genopure Plasmid Midi Kit (Roche Diagnostic) according to the manufacturer's procedures and used for Whole-Plasmid Sequencing (WPS). WPS was performed using the 454-Junior Genome Sequencer procedure (Roche Diagnostic). The libraries of plasmid DNA were constructed using the GS FLX Titanium Rapid Library Preparation Kit.

# De novo Assembly of DNA Reads, Gap-Closure and Annotation

Contigs with at least a 70-fold coverage were obtained using the GS de novo Assembler software (Roche Diagnostics). Scaffolding was first done in silico by using the 454 ReadStatus output file to identify reads with overlapping adjacent contigs. The assembly and gap-closure, as well as insertion/deletion events, single nucleotide polymorphisms and shufflons were confirmed and verified by PCR followed by Sanger DNA sequencing. The annotation was done using the Artemis Version 8 (Sanger Institute) in combination with a pairwise alignment using a BLASTN and BLASTP homology search<sup>2</sup> .

# Phylogenetic Analysis of IncB/O/K Plasmids

For the phylogenetic analysis of the IncB/O/K-like plasmids, two fully sequenced plasmids, namely pCT belonging to IncK (GenBank acc. No. FN868832.1; Cottell et al., 2011) and p3521 belonging to IncB (GU256641; Papagiannitsis et al., 2011) were included as reference. These plasmids were chosen as they represent IncK and IncB plasmids whose complete nucleotide sequence is available in the GenBank. The phylogenetic trees and the homology matrices were constructed by UPGMA method and Jukes-Cantor correction [Multiple alignment (open gap penalty (OG): 100%, unit gap penalty (UG): 0%, gap penalty: 100%), 2000 Bootstrap trials] using Bionumerics 7.6 (Applied Maths, Kortrijk, Belgium).

The analyses were performed using the full length sequences of the IncK2 plasmids with and without shufflon region. The deleted shufflon region started at the shufflon-specific DNA recombinase (rci) and ended with the last shufflon gene which varies from plasmid to plasmid (position 62686–65450 in pTMSA970). In the case of the reference IncK and IncB plasmids, in addition to the shufflon region, the 3,144-bp blaCTX−M−14-carrying element was removed from the IncK reference plasmid (position 68837– 71981 in FN868832) and the 26,548-bp large blaACC−<sup>4</sup> containing element was removed from the IncB reference plasmid (position 59569–86117 in GU256641).

# Conjugation for Incompatibility Testing

Mating experiments were performed crossing the transformant strain containing plasmid p5312.29 (strain SNS5312.29) or pTMSA970 (strain TMSA970) with reference plasmids Rhh72 (IncB/O), and R387 (IncK; provided by the Istituto Superiore di Sanità, Rome, Italy) which are used for incompatibility testing (Couturier et al., 1988). Conjugations were performed on Luria-Bertani (LB) agar solid media streaking two crossing lines, one for each mating strain, and incubating the mating plates overnight at 37◦C.

<sup>1</sup>http://mlst.warwick.ac.uk/mlst/dbs/Ecoli

<sup>2</sup>http://blast.ncbi.nlm.nih.gov/Blast.cgi


TABLE 1 | Origin, characteristics and resistance profile of the original E. coli strains and of the E. coli strains transformed with the blaCMY−2/<sup>4</sup> containing IncK2 plasmids.

The three isolates from broiler cloacae were from three different herds. The three isolates from poultry meat were from three different lots originating from the meat packing plant (MPP) of slaughterhouse A; The three isolates from human infections were from patients hospitalized at the same hospital. CEF, cephalotin; CFZ, cefazolin; FEP, cefepime; CTX, cefotaxime; CTX-CLA, cefotaxime/clavulanic acid; FOX, cefoxitin; CPD, cefpodoxime; CAZ, ceftazidime; CRO, ceftriaxone; CHL, chloramphenicol; CIP, ciprofloxacin; CST, colistin; GEN, gentamicin; IPM, imipenem; KAN, kanamycin; MEM, meropenem; NAL, nalidixic acid; TZP, piperacillin/tazobactam; STR, streptomycin; SMZ, sulfamethoxazole; TET, tetracycline; TMP, trimethoprim.

Donor and transconjugant colonies were counted after plating 10-fold serial dilutions on LB plates containing 50 µg/ml of ampicillin (as resistance marker for p5312.29 and pTMSA970), 50 µg/ml kanamycin (as resistance marker for Rhh72), 25 µg/ml chloramphenicol (as resistance marker for R387), 50 µg/ml of ampicillin plus 50 µg/ml kanamycin, and 50 µg/ml of ampicillin plus 25 µg/ml chloramphenicol.

Stability of the plasmids within the conjugant bacterial cells was determined as follows: three independent colonies resulting from each mating and exhibiting the double resistance (ampicillin/kanamycin, or ampicillin/chloramphenicol) were grown 24 h at 37◦C in 50 ml of LB without antibiotics. Tenfold dilutions of the conjugant cultures were plated on LB plates containing either ampicillin, kanamycin or chloramphenicol alone, and on plates containing both ampicillin and kanamycin, or both ampicillin and chloramphenicol. Colonies were counted after an overnight incubation at 37◦C. Stability was determined as the ratio between the number of ampicillin-resistant colonies and the number of ampicillin/kanamycin-resistant colonies for the p5312.29 × Rhh72 and pTMSA970 × Rhh72 transconjugants, and between the number of ampicillin-resistant colonies and the number of ampicillin/chloramphenicol-resistant colonies for the p5312.29 × R387 and pTMSA970 × R387 transconjugants, respectively.

# Nucleotide Sequence Accession Numbers

The complete nucleotide sequence of the seven IncK2 plasmids have been deposited into the GenBank database under accession numbers KR905385 (p5312.29), KR905389 (p4809.66), KR905387 (pTMSA992), KR905388 (pTMSA970), KR905386 (pTMSA1088), KR905384 (pDV45), and KR905390 (pDV10).

# RESULTS AND DISCUSSION

## Plasmid Sequencing

Whole plasmid sequencing of seven blaCMY−2/4-containing plasmids [poultry cloacae: pTMSA970, pTMSA992, pTMSA1088;

poultry meat: pDV10, pDV45; human urine isolates: p4809.66, p5312.29] revealed the presence of highly related plasmids in both genetically related and diverse E. coli from poultry (ST38, ST420), poultry meat (ST38, ST1564) and human clinical samples (ST131, ST1431).

Plasmids mainly differed from each other within the highly variable shufflon region, and a region situated between impC and nikA (**Figure 1**). A major deletion of approximately 6 kb, that starts after yfbA and ends before psiB, distinguished plasmids pDV10, pTMSA1088, pTMSA970, pTMSA992 from the larger plasmids pDV45, p4809.66, and p5312.29 (**Figure 1**).

Another fragment of 2.3 kb situated between yggA and nikA distinguished plasmids p5312.29 and pTMSA992 from the other plasmids (**Figure 1**). Plasmid pTMSA970 also had an extra insertion of a previously unreported IS-element which is closely related to ISSbo1 (GenBank acc. no. CP001062) and belongs to the IS91 family. Integrations or losses of these regions might have occurred as single events during the transfer of the plasmids from one bacterial host to the other rather than during a longer evolutionary process. Outside of these regions, the plasmid sequences were virtually identical, only differing by a few SNPs, which have all been confirmed by Sanger sequencing.

Couturier et al. (1988). Due to the absence of the complete nucleotide sequence of the IncB/O plasmid, pMU720 was not included into Figure 3. The phylogenetic tree was constructed by UPGMA method and Jukes-Cantor correction [Multiple alignment (open gap penalty (OG): 100%, unit gap penalty (UG): 0%, gap penalty: 100%), 2000 Bootstraps] using Bionumerics 7.6 (Applied Maths, Kortrijk, Belgium). H, human urine sample; P, poultry; PM, poultry meat.

Of note, except for the variable shufflon regions, plasmid pDV45 from E. coli from poultry meat was identical to plasmid p4809.66 from human E. coli, and plasmid pDV10 E. coli from poultry meat was identical to plasmid pTMSA1088 from poultry (**Figure 1**), which strongly suggests transfer of plasmids between genetically diverse E. coli along the food chain.

The variable region between impC and nikA also harbors, along with many genes encoding for proteins with unknown function the single-stranded DNA-binding (Ssb) protein. One of the main functions of the Ssb protein is binding of ssDNA and subsequent protection from degradation by nucleases (Meyer and Laine, 1990). The direct consequence of the loss of Ssb for the IncB/O/K-group plasmids is difficult to determine and needs further investigations.

The shufflon is a biological switch that controls the C-terminal end of the PilV proteins via site-specific recombination. This mechanism has been described in R64 as well as in pCT where the shufflon determines the recipient specificity in liquid mating despite being part of a minor component of the thin pilus (Komano, 1999; Cottell et al., 2014). The shufflon of R64 consists of four invertible DNA fragments embedded with seven 19-bp repeat sequences. These 19-bp repeats can be subdivided into a 7-bp core-site region where the DNA crossover occurs and the highly conserved 12-bp right arm region (Komano, 1999; Brouwer et al., 2015).

The shufflon of the seven plasmids is highly diverse and contains between two and eight recombination sites. Four of the 19-bp repeat sequences found in R64 were also


identified in the plasmids. Interestingly, the 7-bp core-site region is identical to the ones identified in the IncI1/I2 plasmids. The terminology of 19-bp repeat sequences follows the designation of the repeats in R64 (Komano, 1999) (i.e., repeat-e: GTGCCAATCCGGTtcgTGG; repeat-f: GTGCCAATCC GGTatcTGG; repeat-g: GTGCCAATCCGGTGtgtGG; repeat-h: GTGCCAATCCGGTttgTGG; letters in italic represent the coresite, small letters for the variable bases). The three plasmids pTMSA970, pTMSA992, and p5312.29 have only two repeat-e sequences, followed by p4809.66 with four repeat sequences in the following order: e-e-e-g. The most complex shufflons with eight repeat sequences each were detected in pDV10 (g-g-e-e-ee-h-f), pDV45 (e-e-e-e-g-g-h-f), and pTMSA1088 (e-e-h-f-e-e-gg). Although, no apparent correlation between the numbers of recombination sites could be detected (neither with the size of the plasmid nor with their origin), the observation of such great diversity may have an impact on the mating preferences of the various plasmids.

# Incompatibility Assays

Maths, Kortrijk, Belgium).

Sequence analysis of the region upstream of the repA including the incRNAI revealed that the DNA sequences of the seven plasmids diverged from the previously reported incRNAI of the IncK, IncB, and IncB/O plasmids (**Figure 2**).

This region before the repA is highly conserved among the specific plasmid Inc groups and is crucial for the incompatibility behavior among plasmids of the same group (Couturier et al., 1988; Praszkier et al., 1991). The key element for this behavior is the incRNAI that forms a stem and loop structure and negatively regulates the transcription of repA. Consequently, if two plasmids of the same or similar Inc group(s) reside in the same bacterial cell the incRNAIs hybridize with each other and prohibit translation (Couturier et al., 1988). Additionally, a phylogenetic tree constructed using only the incRNAI sequences containing the stem-and loop structure strengthened the previous observations that the seven plasmids are highly similar to each other, and are more closely related to the IncK group than the IncB/O groups (**Figure 2**).

The divergent incRNAI sequences suggested that these plasmids could be compatible with the IncK and IncB/O reference plasmids. Therefore, incompatibility assays were performed crossing SNS5312.29 and TMSA970 strains with both IncB/O Rhh72 and IncK R387 reference strains, respectively. Transconjugants were obtained from all mating experiments, but at different frequency. Both SNS5312.29 and TMSA970 showed higher frequencies of conjugation when crossed with IncB/O (1 × 10−<sup>5</sup> transconjugant/donor cell) than when crossed with the IncK (1 × 10−<sup>7</sup> transconjugant/donor cell) reference plasmid. This suggests that the plasmids under study are more closely related to IncK than to IncB/O. Furthermore, IncB transconjugants demonstrated higher stability (>90%) when grown in liquid medium without antibiotics then the IncK

transconjugants (>50%). Based on both conjugation results and incRNAI sequence analysis, the plasmids were assigned to a novel variant of the IncK group designated as IncK2.

# Phylogenetic Analysis of the IncK2 Plasmids

The 4,198-bp long blaCMY−2/4-like-carrying mobile element (ISEcp1-blaCMY−2/4-blc-sugE1) which is flanked by two 13 bp imperfect inverted repeats (IV-R and IV-L) was integrated between the traT and traU in all seven plasmids. Integration of ISEcp1-blaCMY−2/4-blc-sugE1 at the same site inside the tra locus is a solid indication that all the IncK2 plasmids analyzed have a common ancestor (**Figure 1**).

The same structure (ISEcp1-blaCMY−2/4-blc) has already been reported in IncK plasmid from E. coli of human origin in Spain and from E. coli from chicken meat in Norway (Porres-Osante et al., 2015; Mo et al., 2016), as well as in IncA/C plasmids from Salmonella enterica subsp. enterica serotype Newport in the USA (Cao et al., 2015). Notably, the only blaCMY−4-carrying plasmid pDV10 found has also developed from the same progenitor as the other plasmids and was subjected to subsequent evolution via a Trp221Arg substitution in CMY-2.

The phylogenetic analysis of the IncK2 plasmids supports the hypothesis of a common ancestor. All the IncK2 plasmids share a nucleotide identity ranging from 86.1 to 98.0% which increases to 86.5 to 100% once the shufflon region has been removed. This similarity is remarkable given that the percentage of identity for the two reference plasmids IncK and IncB is maximally 78.7% and without the shufflon region 80.1% (**Figure 3**). The IncK2 plasmids differed from the IncK plasmid pCT not only by their incompatibility and replication gene, but also in their structure and base composition (**Figure 1**).

Taken together, these results lead to the conclusion that the IncK2 plasmids are an overall highly homogenous cluster although they are distributed in genetically diverse E. coli isolated from different sources and timeframes. The phylogenetic analysis points toward a common ancestor that has subsequently spread between the various reservoirs. Indeed, the most striking characteristic of the IncK2 group is its occurrence in different reservoirs (including poultry, poultry-meat, and humans isolates) from which they were isolated. There is strong evidence that all IncK2 plasmids originated from a common source due to: (i) the identical integration site of the blaCMY-carrying mobile element; (ii) the overall 91.3% identity among all the plasmids.

Publications from Denmark, Japan, the Netherlands, Norway, Spain, and Sweden investigating pAmpC-producing E. coli isolates from healthy and hospitalized humans as well as strains from poultry and poultry-meat have reported a high prevalence of blaCMY−<sup>2</sup> in both reservoirs (Dierikx et al., 2010; Börjesson et al., 2013; Voets et al., 2013; de Been et al., 2014; Egervärn et al., 2014; Huijbers et al., 2015; Porres-Osante et al., 2015; Hansen et al., 2016; Mo et al., 2016). These publications were all using the PBRT kit for characterization of their blaCMY−2-carrying plasmids, whereby a substantial part of them was either identified as IncK plasmids or as IncB/O plasmids.

Based on the present study, it has to be noted that the IncK2 plasmid-carrying strains may have been falsely identified as positive for the IncK or IncB/O plasmid groups using the PBRT. Indeed, sequence alignment of our IncK2 plasmids with DNA sequences from the GenBank revealed that such plasmids have already been identified in E. coli from poultry origin in different countries. One plasmid (GenBank acc. no. JXMX01000007.1) isolated from E. coli strain 53C from retail chicken meat in the Netherlands (de Been et al., 2014) was almost identical to pDV45 (also from chicken meat) with 98.2% nucleotide identity and 100% without considering the shufflon. Another plasmid, pNVI1292, from E. coli strain 2012-01-1292 (GenBank acc. no. KU312044) from retail chicken meat in Norway (Mo et al., 2016) was nearly identical to pDV10 (also from poultry meat) and to pTMSA1088 (poultry) with 98.7 and 97.8% nucleotide identity, respectively, and 100% without considering the shufflon. Of note, strain 2012-01-1292 also belonged to ST38 like both strains harboring pDV10 and to pTMSA1088 indicating that specific clones may also contribute to the spread of the IncK2 plasmids.

The novel IncK2 group seems to be an emerging epidemic plasmid with a potential for dissemination in E. coli from different reservoirs including human and poultry. To our knowledge, this is the first time that a highly conserved plasmid family is described in isolates from human clinical samples, poultry, and poultry meat all coming from the same geographic region. However, further epidemiological studies are necessary to assess the scope of the spread of this plasmid in both human and animals. Nevertheless, the presence of a common plasmid in three different settings is alarming and should stimulate rapid introduction of adequate measures in the meat production chain to limit the spread of drug-resistant plasmids.

# AUTHOR CONTRIBUTIONS

Conception and design (VP, AE, SNS, AC); acquisition of data (VP, AE, SNS, AC, ACo, SS); analysis and interpretation of data (VP, AE, SNS, AC); drafting the work (VP, AE, SNS, AC); critical revison for important intellectual content (all authors); final approval of the version to be published (all authors); agreement to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved (all authors).

# FUNDING

This work was supported by Grant 1.12.06 from the Federal Food Safety and Veterinary Office (FSVO) and internal funds of Institute of Infectious Diseases (IFIK) and Institute of Veterinary Bacteriology (IVB), University of Bern.

# ACKNOWLEDGMENTS

We thank Matt Riley and Stephen Kania for advice and support allocated to VP during his sabbatical leave from March to July 2015 at the Department of Biomedical and Diagnostic Sciences, College of Veterinary Medicine, University of Tennessee, Knoxville.

# REFERENCES

fmicb-08-00407 March 13, 2017 Time: 16:43 # 8


ESBL phenotype isolated from food-producing animals and healthy humans. J. Antimicrob. Chemother. 69, 2658–2668. doi: 10.1093/jac/ dku206

**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Seiffert, Carattoli, Schwendener, Collaud, Endimiani and Perreten. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Predominance of CTX-M-15 among ESBL Producers from Environment and Fish Gut from the Shores of Lake Victoria in Mwanza, Tanzania

Nyambura Moremi <sup>1</sup> , Elizabeth V. Manda<sup>1</sup> , Linda Falgenhauer 2, 3, Hiren Ghosh2, <sup>3</sup> , Can Imirzalioglu2, 3, Mecky Matee<sup>4</sup> , Trinad Chakraborty 2, 3 and Stephen E. Mshana<sup>1</sup> \*

<sup>1</sup> Department of Microbiology and Immunology, Weill Bugando School of Medicine, Mwanza, Tanzania, <sup>2</sup> Institute of Medical Microbiology, Justus Liebig University, Giessen, Germany, <sup>3</sup> German Center for Infection Research, Partner site Giessen-Marburg-Langen, Campus Giessen, Giessen, Germany, <sup>4</sup> Department of Microbiology/Immunology, Muhimbili University of Health and Allied Sciences, Dar es Salaam, Tanzania

#### Edited by:

Benoit Doublet, National Institute for Agricultural Research, France

### Reviewed by:

Nilton Lincopan, University of São Paulo, Brazil Andrea Endimiani, University of Bern, Switzerland

\*Correspondence: Stephen E. Mshana stephen72mshana@gmail.com

### Specialty section:

This article was submitted to Antimicrobials, Resistance and Chemotherapy, a section of the journal Frontiers in Microbiology

Received: 13 July 2016 Accepted: 04 November 2016 Published: 01 December 2016

#### Citation:

Moremi N, Manda EV, Falgenhauer L, Ghosh H, Imirzalioglu C, Matee M, Chakraborty T and Mshana SE (2016) Predominance of CTX-M-15 among ESBL Producers from Environment and Fish Gut from the Shores of Lake Victoria in Mwanza, Tanzania. Front. Microbiol. 7:1862. doi: 10.3389/fmicb.2016.01862 Extended-Spectrum Beta-Lactamase (ESBL)-producing bacteria are a common cause of healthcare and community-associated infections worldwide. The distribution of such isolates in the environment and their presence in fish as a result of sewage contamination is not well-studied. Here we examined fish and environmental samples from Mwanza city for the presence of ESBL-producing bacteria. From 196 fish sampled from local markets, 26 (13.3%) contained lactose-fermenting ESBL-producing bacteria, while 39/73 (53.4%) environmental samples from the same area were ESBL producers. Antibiotic resistance genes, multi locus sequence types (MLST) and plasmid replicon types in 24 selected isolates from both populations were identified with whole genome sequencing using Illumina MiSeq. Nine of eleven sequenced fish isolates had the blaCTX-M-15 gene whereas 12/13 from environment carried blaCTX-M-15. Antibiotic resistance genes encoding resistance to sulfonamides (sul1/sul2), tetracyclines [tet(A)/tet(B)] fluoroquinolones [e.g., aac(6′ )-Ib-cr, qnrS1], aminoglycosides [e.g., aac(3)-lld, strB, strA,] and trimethoprim (e.g., dfrA14) were detected. E. coli sequence type ST-38 (2) and ST-5173 (2) were detected in isolates both from the environment and fish. IncY plasmids carrying blaCTX-M-15, qnrS1, strA, and strB were detected in five environmental E. coli isolates and in one E. coli isolate from fish. Our data indicate spillage of resistant environmental isolates into Lake Victoria through the sewage system. Persistence of blaCTX-M-15 in the Mwanza city environment is complex, and involves both clonal spread of resistant strains as well as dissemination by commonly occurring IncY plasmids circulating in isolates present in humans, the environment as well as in the food chain.

Keywords: blaCTX-M-15, fish, environment

# INTRODUCTION

The extended-spectrum beta-lactamases (ESBL)-producing Gram-negative bacteria Escherichia coli and Klebsiella spp., particularly those producing CTX-M enzymes, have emerged as important causative agents of healthcare-associated infections across the world (Hawkey and Jones, 2009). Apart from being present in hospitals and clinics, ESBL-producing E. coli strains are prevalent in the community and are reported to be responsible for community-acquired bacterial infections (Arpin et al., 2005; Pitout et al., 2005). Studies from different areas of the world show an association between the presence of ESBL-producing E. coli and exposure to either food or contaminated water (Ho et al., 2011; Laube et al., 2013; Xi et al., 2015). The presence of ESBL-producing E. coli in the community has led to the hypothesis that there could be a transmission of these strains from human waste to the environment (Martinez, 2009a). Evidence in the literature has also documented the probable horizontal transfer of resistance genes from either human sewage or clinical isolates to fish in rivers or lakes in which drainage of wastewater from treatment plants occurs (Kümmerer, 2009; Martinez, 2009b; Jiang et al., 2012; Blaak et al., 2014). Multiple genotypes of ESBL-producing E. coli have been found in animals and humans in Mwanza (Mshana et al., 2011, 2016; Seni et al., 2016). There are few studies from developed countries on the presence of ESBL-producing isolates in fish and environment samples and the role played by the food chain in transmission of resistance genes through contamination by human and animals sewage and anthropogenic activities in relation to water bodies (Zurfluh et al., 2013; Abgottspon et al., 2014). In Mwanza, a port city with a population of 750,000 located on the southern shores of Lake Victoria (Fitzpatrick et al., 2015), effluents from wastewater treatment plants and pit-latrines used by most of households located in the hills drain into the lake, which is also a source of fish consumed by residents as staple food. We previously used conventional phenotypic characterization and whole genome sequencing to examine samples obtained from hospitals, from rural farming communities including animals, farmers and households, in Mwanza for the presence of ESBLproducing Enterobacteriaceae. Our data indicate clonal spread of bacteria belonging to a small number of STs present in all populations investigated, but also suggest that commonly occurring promiscuous plasmids are involved in resistance dissemination (Fortini et al., 2015; Mshana et al., 2016). Here we examined for the presence of ESBL-producing bacteria in fish obtained from Lake Victoria as well as environmental samples obtained from the city. Our results indicate that E. coli genotypes that were observed in humans and animals before are also present in environment and Fish.

# MATERIALS AND METHODS

# Isolation and Identification of Bacterial Isolates

Ten fish markets located in different urban and rural sections of the Mwanza region were randomly selected for the study. These markets receive fishes obtained from different fishing sites within Lake Victoria. A total of 196 Nile tilapia (Oreochromis niloticus) fish from randomly picked vendors were sampled with an average of 15 to 20 fish at each market between July and September 2015. Fish were first washed with saline, following which a sterile surgical blade was used to open the carcass and to make a longitudinal incision along the gut. The incision was opened and the gut contents were swabbed using a sterile swab.

In addition, at ten sites including the Ngerengere River that crosses the city, drain waste and possible sewage from households located in the surrounding hills were sampled (environmental samples). Environmental samples included dirty muddy water samples from different location in the city (**Figure 1**). For each site between six to seven samples were taken from different locations (**Figure 1**). About 3 ml of each sample was collected using sterile 25 ml falcon tube (BD, Nairobi, Kenya). A sample was mixed with sterile 0.9% saline at ratio of 1:1 and vortexed to produce a homogenous solution.

Using swabs, MacConkey agar (Oxoid, Basingstoke, UK) plates supplemented with 2 mg/L cefotaxime (Medochemie Ltd, Limassol, Cyprus) were inoculated to screen for ESBL-producing Enterobacteriaceae. Enteric bacilli were identified using colony morphology and differentiated based on lactose fermentation on MacConkey agar. Single colonies from predominant lactose fermenting bacteria were picked for further identification using several biochemical tests (Triple Sugar Iron Agar, Simmons' citrate Agar, Sulfur-Indole-Motility test and Urease test) (Murray et al., 1995). In case of ambiguous results, the VITEK <sup>R</sup> 2 system (BioMérieux, Marcy l'Etoile, France) was used to confirm identification. A confirmed ESBL-producing E. coli isolate was used as a positive and E. coli ATCC 25922 as a negative control. ESBL isolates were stored as glycerol cultures at −80◦C and used for further characterization.

# Antimicrobial Susceptibility Testing

Antimicrobial susceptibility testing was done by disk diffusion method as recommended by the Clinical Laboratory Standards Institute (CLSI) guidelines (Wayne, 2012). A bacterial suspension of 0.50 McFarland standard turbidity was prepared from pure culture. An inoculum was then plated on Mueller-Hinton Agar plates (HiMedia, Mumbai, India) and the following antibiotic disks were set: Tetracycline (2 µg), ciprofloxacin (5 µg), gentamicin (10 µg), or trimethoprim/sulphamethoxazole (1.25/23.75 µg) (Oxoid, Hampshire, UK). The plates were incubated aerobically at 37◦C for 18–24h. The diameters of the respective zone of inhibitions were measured and interpreted following CLSI 2012 guidelines (Wayne, 2012). Disk approximation method based on CLSI guidelines was used to confirm ESBL production and for selected isolates ESBL production was further identified using VITEK <sup>R</sup> 2 system (BioMérieux, Marcy l'Etoile, France) and in addition the MIC for cefepime, carbapenems and colistin were determined in all selected isolates.

Using STATA version 11 the two-sample test of proportion was done to compare the rates of resistance between ESBL isolates from fish and those from environment. A p-value < 0.05 was used to indicate a significant difference at 95% confidence interval.

# Whole Genome Sequencing

Eleven ESBL-producing isolates obtained from fish and 13 E. coli from the environmental samples were chosen for whole genome sequencing (WGS). DNA was isolated using the Purelink Genome DNA Mini kit (Invitrogen, Darmstadt, Germany) according to the manufacturer's instruction. WGS was carried

out using an Illumina Nextera XT library with 2x300bp pairedend reads on an Illumina MiSeq instrument (Illumina, San Diego, CA, USA). The raw data was assembled using SPAdes (version 3.0) (Bankevich et al., 2012). Contigs from E. coli isolates were ordered by using MAUVE (Rissman et al., 2009) and E. coli MG1655 (accession number U00096.3) were chosen as a reference for all E. coli isolates, Citrobacter freundii strain P10159 for Citrobacter braakii isolates (accession number CP012554; no complete genome available for C. braakii), Klebsiella pneumoniae strain ATCC BAA-2146 (accession number CP006659) for K. pneumoniae and Enterobacter cloacae strain 34977 (accession number CP010376) for E. cloacae isolates. Pseudogenomes were created and whole genome phylogenetic analysis was subsequently performed by using ParSNP package of Harvest Suite (Treangen et al., 2014). The raw sequencing data of the sequenced isolates are available at the European Nucleotide Archive (ENA) under the project number PRJEB12361.

# In silico Analyses of Resistance Genes, MLST, Plasmid Replicon Types, and Quinolone Resistance-Determining Regions (QRDR) Mutations

Sequences were analyzed for their multi locus sequence types, transferrable resistance genes, plasmid replicon types and pMLST using MLST 1.8, ResFinder, Plasmidfinder and pMLST software from the Center for Genomic Epidemiology (Larsen et al., 2012; Zankari et al., 2012; Carattoli et al., 2014). Search for plasmid-encoded heavy metal resistances and detergence resistance was performed using blastn with the references given in Supplementary Table 1.

The location of blaCTX-M-15 was determined by analyzing the contigs harboring blaCTX-M-15 using blastn. The whole genome sequences were compared with the plasmid pPGRT46 (accession number KM023153.1) using BRIG and blastn (Alikhan et al., 2011; Fortini et al., 2015). Quinolone resistance-determining regions (QRDR) mutations were identified by silico analysis by comparing the sequence using a reference sequence from a quinolone-susceptible Enterobacteriaceae strain (Weigel et al., 1998; Liu et al., 2012).

# Ethical Approval

The protocol of this study was approved by CUHAS/BMC joint ethics and scientific review committee with reference CREC/019/2014.

# RESULTS

# Bacterial Isolates

A total of 26 (13.3%) lactose-fermenting ESBL-producing bacteria were isolated from gut samples of 196 wild Nile tilapia fish from Lake Victoria. Diverse bacterial species were detected and included C. braakii (11/26, 42.3%), E. cloacae (5/26, 19.2%), K. pneumoniae (5/26, 19.2%), E. coli (4/26, 15.4%) and Klebsiella oxytoca (1/26, 3.9%). From 73 environmental samples 39 (53.4%) ESBL-producing enteric bacteria were isolated. Of these 39 isolates, 20 (51.3%) were E. coli and 19 (48.7%) were K. pneumoniae.

# Antimicrobial Susceptibility Testing

All isolates had MICs for cefotaxime and cefepime of >32 and >8 µg/ml respectively. High resistance rates to co-trimoxazole (n = 19, 73.1%), ciprofloxacin (n = 19, 73.1%), and gentamicin (n = 19, 73.1%) and tetracycline (n = 16, 61.5%) were observed among the 26 ESBL-producing isolates from fish. Environmental isolates were resistant to co-trimoxazole (n = 34, 87.2%), ciprofloxacin (n =14, 38%), and gentamicin (n = 18, 46.1%). ESBL-producing isolates from fish were significantly more resistant to gentamicin and ciprofloxacin (p < 0.01: **Table 1**). In addition, all isolates were sensitive to meropenem and imipenem (MIC <0.25 µg/ml), and all were sensitive to colistin.

# Analysis of Antibiotic Resistance Genes

Each of the bacterial genome-sequenced fish and environmental isolates carried up to four different β-lactamase genes (ESBL, AmpC and other β-lactamases, **Table 2**). The most common ESBL gene was blaCTX-M-15, present both in fish (9/11) and environmental isolates (12/13). The non-ESBL β-lactamase gene blaTEM-1 was present in 7/11 fish isolates and 9/13 environmental isolates, whereas blaOXA-1 was present in 6/11 fish isolates and 1/13 environmental isolates. K. pneumoniae isolates from fish were found to harbor blaSHV-11 and blaSHV-1 due to the fact that K. pneumoniae chromosomally possesses blaSHV-1/-11. In addition, the fish isolates also harbored a number of AmpC type β-lactamase genes (blaACT-15, blaMIR-3, blaCMY-37, blaCMY-49), which were not present in the environmental isolates. The latter two results might simply reflect the fact that while different types of enterobacterial species were isolated from fish, only E. coli were obtained from the environmental samples studied.

Commonly occurring aminoglycoside resistance genes detected in both fish and environmental isolates were aac(6′ )- Ib-cr (6/24), strA/strB (15/24), aac(3)-IId (5/24), and aadA1 (4/24) (**Table 2**). Other aminoglycoside resistance genes were present either only in fish isolates (aadA2, aac(3)-IIa) or in environmental samples (aadA5). Quinolone resistance genes were detected in 8/11 (72.7%) of fish isolates and in 7/13 (53.7%)

TABLE 1 | Rate of resistance of ESBL-producing isolates from fish and environment to SXT, TET, CIP, and CN.


SXT, sulphamethoxazole/trimethoprim; TET, tetracycline; CIP, ciprofloxacin, CN, gentamicin.

of environmental isolates that were sequenced. Quinolone resistance genes detected in both populations comprised of the aac(6′ )-Ib-cr (6/24), qnrB1 (5/24), and qnrS1 alleles. Other quinolone resistance genes were only present in fish isolates (qnrB29, qnrB48, oqxA, oqxB).

Of the 24 sequenced isolates, 20 (83.3%) harbored sulfonamide resistance genes (sul1 or sul2), and 19 of these isolates harbored in addition a resistance gene encoding trimethoprim resistance (dfrA14, n = 11; dfrA17, n = 3; dfrA1, dfrA18, dfrA30, dfrA5, dfrA7, each n = 1). Tetracycline resistance genes, [tet(A), n = 15; tet(D), n = 3], were present in 18/24 isolates. Altogether, there was a direct correlation between resistance phenotype and resistance genotype in all 24 isolates.

Using in silico analysis parC mutations were detected in 7 E. coli, only 2 isolates had S80I mutation that is associated with fluoroquinolone resistance, while gyrA mutations were observed in 9 isolates (2 E. cloacae and 7 E. coli). S83 (S83T, S83L, and S83A) mutations that are associated with fluoroquinolone resistance were observed in 5/9 isolates. One isolate had both S83L and D87N.

# Plasmid-Encoded Heavy Metal Resistance Operons and Detergence Genes

An analysis for the presence of plasmid-located heavy metal- and detergent- resistance genes revealed, that 8/24 isolates harbored the qacEdelta gene, conferring resistance to tertiary ammonium compounds. Seven (29.2%) of isolates sequenced carried a mercury resistance operon originally described in plasmid R478, and 4/24 (16.6%) isolates had genes mediating resistance to silver. Only one isolate carried genes involved in resistance to copper together with nickel/cobalt efflux system (Supplementary Table 2).

# Location of the blaCTX-M Gene and AmpC Genes

Analysis of the whole genome sequences was used to identify and map the location of blaCTX-M alleles. For 8/24 isolates, the blaCTX-M-15 gene was located in the chromosome. In two isolates blaCTX-M-15 was located on a phage-like plasmid, similar to the E. coli phage-like plasmid p1303\_95 isolated from wound swab (isolate SO007, accession number of the reference: CP009168.1) and pECOH89 (isolate SO053, accession number of pECOH89: HG530657.1) (Falgenhauer et al., 2014). The isolate SO069 harbored blaCTX-M-15 on an IncI1 plasmid similar to pESBL-EA11 (Ahmed et al., 2012). Six of 24 sequenced isolates (one fish and five environmental isolates, F044, SO005, SO008, SO025, SO037, SO063, all E. coli) harbored a resistance cassette similar to the one present in plasmid pPGRT46 (accession number KM023153.1) that included both qnrS1 as well as blaCTX-M-15. The other isolates displayed blaCTX-M-15 or blaCTX-M-55 containing resistance cassettes with similarities to other ESBL-encoding plasmids (F025: pEC\_L46, accession number GU371929.1; F102: pSTm-A54650, accession number LK056646.1; SO042: pSKLX3330, accession number KJ866866.1; SO060: pCA14, accession number CP009231.1). The isolates F016 and F017 did not harbor any ESBL


isolates.

resistance gene. It should be noted that because conjugation experiments were not done plasmid location could only be suggested.

In 13 of 22 ESBL-encoding isolates, the environment of the ESBL gene was characterized by a Tn3 transposon deletion in the vicinity of the blaCTX-M gene (**Figure 2**). This included four isolates harboring a chromosomally located blaCTX-M-15 allele. The other isolates, including two chromosomally located blaCTX-M-15 and the single blaCTX-M-55 isolate, did not have any Tn3-related sequences. All ampC genes detected were located in the chromosomes (F006, F009, F016, F025, F102).

# Multi Locus Sequence Types and Whole Genome Phylogeny

E. coli isolates sequence type (ST) ST-38 and ST-5173 were present in both fish and environmental isolates. Among the 13 E. coli environmental isolates 12 different STs were observed. The fifteen E. coli isolates, 13 from the environment and 2 from fish, grouped into five clusters using genome-based phylogenetic analysis. The tree was generated on the basis of a core genome accounting for 78% (3.62 Mbp) of the reference genome (E. coli MG1655). The two E. coli isolates from fish have 100% similarity with respective ST isolates derived from the environment (SO025 and F044, ST-38; F080 and SO038, ST-5137; **Figure 3**). E. cloacae

FIGURE 2 | Schematic depiction of the genetical environment of blaCTX-M genes in the blaCTX-M-encoding isolates. (A) Genetical environments associated with Tn3 transposon deletion, (B) gentical environments not associated with Tn3 transposon deletion.

number U00096.3), (B) Klebsiella pneumoniae (reference Klebsiella pneumoniae strain ATCC BAA-2146, accession number CP006659), (C) Citrobacter braakii (using Citrobacter freundii P10159, accession number CP012554.1 as a reference as no complete genome of Citrobacter braakii is presently available), (D) Enterobacter cloacae (E. cloacae strain 34977, accession number CP010376). The phylogenetic analysis was performed using Harvest Suite.



\*There is no MLST scheme available for Citrobacter braakii.

isolates from fish displayed three different STs (ST-91 n = 2; ST-422 n = 1; ST-500 n = 1), based on an analysis of core genome content of around 68% (3.33 Mbp) using E. cloacae strain 34977 as reference. The two ST-91 isolates were more closely related to each other, than the other two isolates. K. pneumoniae isolates were typed as ST-37 (2/3) with a single isolate as ST-280 (1/3). The core genome with K. pneumoniae strain ATCC BAA-2146 (accession number CP006659) as reference accounted for 88% (4.78 Mbp) of its genome size. The ST-37 and ST-280 are quite different from one other. The C. braakii isolates could not be assigned to any ST, because there is no typing scheme available. Based on 82% (4.17 Mbp) core genome size using C. freundii strain P10159 as a reference genome, these two isolates are very distinct from each other (**Figure 3**). E. coli were grouped in different phylogroups based on scheme of Clermont et al. (2000). A total of 7/15(46%) of E. coli isolates belong to the newly described phylogroups E (**Table 3**).

# Plasmid Replicon Types

Of the eleven fish isolates tested, seven (63.6%) were found to carry IncF plasmid replicon types as compared to 5/13 (38.5%) of environmental isolates (**Table 4**). The IncF pMLST types detected in fish isolates were F-: A-:B36, K4:A-:B- and K5:A-:B-. For the environmental isolates IncF pMLST types detected were F2: A- :B-, F29:-A-:B10 and F31:A4:B1. The IncI1 plasmids detected in E. coli from environment were classified using pMLST either as ST-31 and unknown pMLST while that in E. coli from fish had an unknown IncI1 pMLST. Two out of four E. cloacae isolates from fish harbored IncF plasmids with identical pMLST (F-:A- :B36). We extracted contig sequences of the F044, SO005, SO008, SO025, SO037, SO063, carrying the resistance cassette present in pPGRT46 (Fortini et al., 2015) (accession no. KM023153) and compared them to this plasmid to examine for overall homology. These isolates showed an overall overlap of between 76 and 90% in their nucleotide sequences with plasmid pPGRT46 (**Figure 4**).

## DISCUSSION

Lake Victoria is Africa's largest lake by area, and borders three East-African countries: Tanzania (49%), Uganda (45%), and Kenya (6%). It is not only the major source of fish consumed by Mwanza residents but also receives biologically treated wastewater effluent from Mwanza city, where a tertiary hospital serving eight regions in Tanzania is located. Studies performed in this hospital have shown high rates of ESBL-producing K. pneumoniae and E. coli carrying blaCTX-M-15 in IncF plasmids (Mshana et al., 2011, 2013). In addition recent studies involving isolates from animals and humans from community have shown the blaCTX-M-15 allele to be predominant (Mshana et al., 2016; Seni et al., 2016).

In the present study, a significant proportion of fish gut and environmental samples were positive for ESBL-producing bacteria. As previously observed in isolates from humans in hospitals (Mshana et al., 2011, 2013), the majority of these isolates were multiply resistant to co-trimoxazole, gentamicin, tetracycline, and ciprofloxacin. In addition, it was observed that ESBL isolates from fish were significantly more resistant to ciprofloxacin and gentamicin. This could be explained by the fact majority of the environment isolates were E. coli while other genera such as Citrobacter and Enterobacter which tend to be resistant to ciprofloxacin and gentamicin (Dos Santos et al., 2015) formed majority of isolates from Fish. Such high AMR rates detected in the isolates from gut contents of wild fish are of concern and indicate strong anthropogenic environmental contamination of the Lake Victoria. The presence of ESBL isolates in the environmental samples suggests that the Lake Victoria is being contaminated by sewage from the hospitals and from animals (Mshana et al., 2009; Kayange et al., 2010; Moremi et al., 2014; Seni et al., 2016). Farming activities that use of animal manure together with agricultural waste from the community generate steady contaminated effluents along the borders of Lake Victoria. Climatic conditions including sustained periods of rain or man-made erosion may further contribute to increasing the presence of these bacteria in the lake.

The prevalence of ESBL-producing E. coli from wild fish was 2% (4/196) which is comparable to the results previously reported from China (Jiang et al., 2012) but lower than in Egypt which reported a prevalence of 4% (Ishida et al., 2010). Unlike the two studies from Egypt and China which were conducted among farmed fish with antibiotic exposure, the investigated isolates here were from uncultured free-living fish and emphasize the


TABLE 4 | Plasmid characteristics of the sequenced isolates, n.t.: not typable using IncHI1 pMLST scheme (no pMLST alleles present); # variants with homology or coverage less than 100%.

role of environmental pollution with AMR isolates as previously published (Kümmerer, 2009; Guenther et al., 2011).

As was already noted in studies from the Asian, European and the African continents which documented the emergence and spread of blaCTX-M-15 in humans and cattle (Madec et al., 2012; Zhang et al., 2013; Rafaï et al., 2015), CTX-M-15 was also found to be the predominant allele in this study. The gene was located on the chromosome or on different plasmids indicating extreme environmental mobility. The blaCTX-M-15 allele has been reported in other studies (Dobiasova et al., 2015; Fortini et al., 2015; Ibrahim et al., 2015) of Gram-negative bacterial isolates from humans, fish, and animals, suggesting a transmission/circulation of this gene among different settings.

Two plasmid-mediated AmpC enzymes were detected in this study from fish isolates: blaMIR-1 and blaACT-15, this is in contrast with a previous report that showed the presence of plasmid-mediated blaCMY-2 allele in fish (Welch et al., 2009). No isolates carrying blaCMY-2 were detected in this study. However, a chromosomally located blaCMY-37 which had previously been described in C. freundii, was detected in a C. braakii, (Ahmed and Shimamoto, 2008). One resistant isolate harbored neither a plasmid-based AmpC nor a detectable ESBL enzyme (F017) suggesting the resistance observed might be due to unknown mechanisms or the presence of resistance genes in the chromosome. The chromosomally located genes cannot be detected using the software from the Center for Genomic Epidemiology because the database targets for transferable resistance genes only. As in previous studies performed in China and Egypt among farmed fish (Ishida et al., 2010; Jiang et al., 2012) plasmid-mediated quinolone resistance genes such as aac(6′ )-Ib-cr, qnrB and qnrS were also detected in this study in both sampled populations.

E. coli ST-38 which was also observed in humans in hospitals, humans in the community and in livestock/ companion/domestic animals in Mwanza city (Mshana et al., 2011, 2016; Seni et al., 2016) was detected in this study in both fish and environmental samples. According to the new phylogroup scheme by Clermont et al. (2000), all these Isolates (human, livestock/companion animal, fish/environment) were grouped as the new phylogroup E emphasizing the uniformity of these strains in our setting. The ST-38 E. coli from fish and the environment as well as other ST types from environment harbor resistance cassettes that include CTX-M-15 and QnrS1. A similar resistance cassette was previously detected on an IncY plasmid detected in isolates from healthy pregnant women in Nigeria, and from ESBL-producing isolates from livestock and companion animal isolates in the Mwanza region (Fortini et al., 2015; Seni et al., 2016). The sequence overlap with the plasmid pPGRT46 was very high (up to 90%). These findings draw attention to the likelihood that this resistance cassette and its

plasmid have spread centrally across the African continent into multiple E. coli genotypes and in different environments.

The E. coli ST-38 was observed in both fish and environment isolates as documented above was previously detected in humans and animals in the same setting. This finding is worrisome as zoonotic transmission is possible and its association with human colonization and infection might pose treatment challenges to human health. The transmission and persistence of ESBLproducing bacteria through the food chain and environment through sewage contamination has been documented previously (Beninati et al., 2015; van Hoek et al., 2015). The findings of this study suggest the possibility of the transmission of both ESBL-producing clones and plasmids between humans and wild fish through environmental contamination and indicate that anthropogenic activities and the food chain as potential factors for the persistence and dissemination of blaCTX-M-15 in Mwanza City. The current study did not observed E. coli ST-131 from fish however it was obtained from environment; this could be due to that only 2 out 4 isolates from fish were sequenced. Further studies with large number of E. coli from fish are needed to confirm the role of food chain in the persistence of blaCTX-M-15 in Mwanza city.

# CONCLUSION

This is the first report on the epidemiology of ESBL-producing Gram-negative bacteria from fish in Lake Victoria and its surrounding environment. More than 70% of the sequenced bacterial isolates carried quinolone and aminoglycoside resistance genes. Detection of isolates/plasmids carrying blaCTX-M-15 which have been also found in humans and companion/livestock animals in the same region suggest environmental contamination with sewage effluents from humans an animal sources. Our data suggest that additional efforts to implement better sanitation and control sewage management are warranted. Many common illnesses in particular diarrhea, can be attributed to poor sanitation and unsafe water (Cheng et al., 2005) which exacerbates antibiotic use and contributes significantly to the problem of antibiotic resistance. Policy measures to improve water and sanitation quality could greatly contribute to the reduction of ESBL-resistance in this region.

# AUTHOR CONTRIBUTIONS

NM, MM, CI, TC, and SM conceived, designed and executed the study; EM, NM, and SM collected the data and samples; NM, EM, and SM performed laboratory analysis; LF and TC performed WGS; LF, HG, TC, and SM analyzed the data; NM, LF, TC, and SM wrote the manuscript which was critically reviewed by all authors. All authors have read and approved the final draft of the manuscript.

# REFERENCES


# ACKNOWLEDGMENTS

We are grateful to all of the fish vendors and fishermen who helped us to get useful information during sample collection. Our heartfelt thanks also go to Mr. Vitus Silago for the diligent technical laboratory assistance he provided during the study period. We thank Christina Gerstmann from the Institute of Medical Microbiology in Germany for excellent technical assistance. This study was supported by a grant from the Wellcome Trust (WT087546MA) to SACIDS and also by grants from the Bundesministerium fuer Bildung und Forschung (BMBF, Germany) within the German Center for Infection Research [DZIF/grant number 8000 701–3 (HZI) to CI and TC; TI06.001/8032808811 to TC] and the research network RESET (grant number 01KI1313G to CI and TC).

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2016.01862/full#supplementary-material


Gram-negative bacteria isolated from fish farms in Egypt. J. Vet. Med. Sci. 72, 727–734. doi: 10.1292/jvms.09-0538


in the community. J. Antimicrob. Chemother. 56, 52–59. doi: 10.1093/jac/ dki166


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Moremi, Manda, Falgenhauer, Ghosh, Imirzalioglu, Matee, Chakraborty and Mshana. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Prevalence and Molecular Characteristics of Extended-Spectrum β-Lactamase Genes in Escherichia coli Isolated from Diarrheic Patients in China

Li Bai<sup>1</sup>† , Lili Wang<sup>2</sup>† , Xiaorong Yang<sup>3</sup> , Juan Wang<sup>4</sup> , Xin Gan<sup>1</sup> , Wei Wang<sup>1</sup> , Jin Xu<sup>1</sup> , Qian Chen<sup>2</sup> , Ruiting Lan<sup>5</sup> , Séamus Fanning1,6 \* and Fengqin Li<sup>1</sup> \*

### Edited by:

Axel Cloeckaert, Institut National de la Recherche Agronomique (INRA), France

#### Reviewed by:

Stefan Paul Schwarz, Friedrich Loeffler Institute, Germany Marisa Haenni, Agence Nationale de Sécurité Sanitaire de l'Alimentation, de l'Environnement et du Travail (ANSES), France

### \*Correspondence:

Fengqin Li lifengqin@cfsa.net.cn Séamus Fanning sfanning@ucd.ie †These authors have contributed

equally to this work.

#### Specialty section:

This article was submitted to Antimicrobials, Resistance and Chemotherapy, a section of the journal Frontiers in Microbiology

Received: 08 July 2016 Accepted: 19 January 2017 Published: 13 February 2017

#### Citation:

Bai L, Wang L, Yang X, Wang J, Gan X, Wang W, Xu J, Chen Q, Lan R, Fanning S and Li F (2017) Prevalence and Molecular Characteristics of Extended-Spectrum β-Lactamase Genes in Escherichia coli Isolated from Diarrheic Patients in China. Front. Microbiol. 8:144. doi: 10.3389/fmicb.2017.00144 <sup>1</sup> Key Laboratory of Food Safety Risk Assessment, Ministry of Health, China National Center for Food Safety Risk Assessment, Beijing, China, <sup>2</sup> Institute for Nutrition and Food Hygiene, Beijing Key Laboratory of Diagnostic and Traceability Technologies for Food Poisoning, Beijing Center for Disease Prevention and Control, Beijing, China, <sup>3</sup> Center for Disease Control and Prevention of Sichuan Province, Sichuan, China, <sup>4</sup> College of Veterinary Medicine, Northwest A&F University, Xianyang, China, <sup>5</sup> School of Biotechnology and Biomolecular Sciences, University of New South Wales, Sydney, NSW, Australia, <sup>6</sup> UCD-Centre for Food Safety, School of Public Health, Physiotherapy and Sports Science, University College Dublin, Belfield, Ireland

Background: The emergence and spread of antimicrobial resistance has become a major global public health concern. A component element of this is the spread of the plasmid-encoded extended-spectrum β-lactamase (ESBL) genes, conferring resistance to third-generation cephalosporins. The purpose of this study was to investigate the molecular characteristics of ESBL-encoding genes identified in Escherichia coli cultured from diarrheic patients in China from 2013 to 2014.

Materials and Methods: A total of 51 E. coli were confirmed as ESBL producers by double-disk synergy testing of 912 E. coli isolates studied. Polymerase chain reaction (PCR) and DNA sequencing were performed to identify the corresponding ESBL genes. Susceptibility testing was tested by the disk diffusion method. Plasmids were typed by PCR-based replicon typing and their sizes were determined by S1-nuclease pulsed-field gel electrophoresis. Multi-locus sequence typing (MLST) and phylogrouping were also performed. Broth mating assays were carried out for all isolates to determine whether the ESBL marker could be transferred by conjugation.

Results: Of the 51 ESBL-positive isolates identified, blaCTX-M, blaTEM, blaOXA, and blaSHV were detected in 51, 26, 3, 1 of these isolates, respectively. Sequencing revealed that 7 blaCTX-M subtypes were detected, with blaCTX-M-14 being the most common, followed by blaCTX-M−<sup>79</sup> and blaCTX-M-28. Of the 26 TEM-positive isolates identified, all of these were blaTEM−<sup>1</sup> genotypes. All isolates contained one to three large plasmids and 10 replicon types were detected. Of these, IncFrep (n = 50), IncK/B (n = 31), IncFIB (n = 26), IncB/O (n = 14), and IncI1-Ir (n = 8) replicon types were the predominating incompatibility groups. Twenty-six isolates demonstrated the ability to transfer their cefotaxime resistance marker at high transfer rates. MLST typing identified 31 sequence types and phylogenetic grouping showed that 12 of the 51 donor strains belonged to phylogroup B2.

Conclusion: This study highlights the diversity of the ESBL producing E. coli and also the diversity of ESBL genes and plasmids carrying these genes in China, which poses a threat to public health.

Keywords: Escherichia coli, human, extended-spectrum β-lactamase genes, horizontal gene transfer, conjugation, S1 nuclease digestion, China

# INTRODUCTION

fmicb-08-00144 February 9, 2017 Time: 13:46 # 2

Extended-spectrum β-lactamase (ESBL) producing Escherichia coli are a frequent cause of community- and hospital-acquired infections and one of the leading causative agents of infections worldwide (Hampton, 2013; Nischal, 2014). E. coli can become resistant to extended-spectrum cephalosporins by mutational overproduction of AmpC and/or by expression of acquired ESBLs. The latter emerged in the 1980s as derivatives of TEM (named after the patient Temoneira) and SHV (sulfhydryl reagent variable) enzyme types (Gutkind et al., 2013). The genes encoding these acquired enzymes are associated with plasmids with the potential for horizontal dissemination, and the most widespread ESBL type identified is CTX-M (Gutkind et al., 2013). These ESBL genotypes have spread in a pandemic manner and been associated with outbreaks in hospitals and communities worldwide (Gutkind et al., 2013).

Plasmid-mediated transfer of drug resistance-encoding genes among bacterial species is considered to be one of the most important mechanisms driving the dissemination of multi-drug resistance (MDR). An important element of this MDR dissemination is the transmission of the plasmidencoded ESBL genes, which confer resistance to third-generation cephalosporins. Infection with ESBL-positive bacteria frequently results from inappropriate use of antimicrobial compounds in human and veterinary medicine a feature that also constitutes a risk factor for selection and dissemination of resistant clones (Gutkind et al., 2013; Hamprecht et al., 2016). Consequently, the treatment options for infections caused by ESBL-producing microorganisms are limited.

ESBL-producing E. coli has been reported from China, mostly from extra-intestinal E. coli cultured from patient specimens such as bloodstream infections (BSIs), along with intra-abdominal infections (IAIs) and urinary tract infections (UTIs). Some 67.8% of the isolates from BSIs were identified to be positive for the predominant CTX-M enzymes, CTX-M-14 and CTX-M-15 (Wang S. et al., 2016), 68.2% isolates were from UTIs and ICU and which carried the CTX-M-15 gene (Liu et al., 2015) and 69.6% were from IAIs and were also found to be ESBLproducing E. coli (Fan et al., 2016). However, there is limited epidemiological information on ESBL-producing E. coli from diarrheic patients. In this paper, we report on the isolation and characterization of a collection of ESBL-producing E. coli isolated from diarrheic patients in China. The aims of this study were (i) to identify the plasmids carrying ESBL-encoding genes, (ii) to determine the phylogenetic grouping, the multi-locus sequence typing (MLST), and the pathogenic grouping, and (iii) to determine the plasmid replicon types and S1 nuclease-based plasmid profiles.

# MATERIALS AND METHODS

# Ethics Statement

Fecal samples were acquired with the written consent from all patients. This study was reviewed and approved by the ethics committee of China Center for Disease Prevention and Control, according to the medical research regulations of the Ministry of Health, in China. All of this research work was conducted within China.

# Bacteria and Growth Conditions

Fifty-one isolates including 27 from Beijing, 9 from Guangxi province, 9 from Henan province, and 6 from Sichuan province were confirmed as ESBL producers by the double-disk synergy test from 912 E. coli recovered from diarrhea cases. All human samples were collected in urban areas, and each individual was tested only once. The protocols for isolating E. coli were described previously (Wang et al., 2013).

Isolates from patients with typical E. coli phenotypes were confirmed by the API 20E biochemical gallery (bioMérieux, Beijing, China). Those E. coli expressing an ESBL phenotype were selectively enriched for study using Luria Bertani (LB) supplemented with 2 mg/L cefotaxime. All E. coli isolates that were recovered were further screened for ESBL production by determination of synergy between 0.25 and 128 mg/L ceftazidime or cefotaxime and 4 mg/L clavulanic acid (CLSI, 2012). Isolates showing a ≥3 twofold concentration decrease in an minimal inhibitory concentration (MIC) for either ceftazidime or cefotaxime tested in combination with clavulanic acid versus the MIC of the agent when tested alone were considered as ESBLproducing (for example, ceftazidime MIC = 8 mg/L; ceftazidime– clavulanic acid MIC = 1 mg/L; CLSI, 2012). E. coli ATCC 25922 and Klebsiella pneumoniae ATCC 700603 were used as quality control microorganisms in antimicrobial susceptibility tests (CLSI, 2012). All confirmed isolates were maintained in brain heart infusion broth containing 50% [v/v] glycerol in a −70◦C freezer for subsequent genotypic and phenotypic characterization.

# Antimicrobial Susceptibility Testing

Antimicrobial susceptibility of all E. coli isolates was determined using the agar dilution method and interpreted according to the Clinical and Laboratory Standards Institute guidelines

(CLSI, 2012) and the European Committee on Antimicrobial Susceptibility Testing (EUCAST-2012)<sup>1</sup> . The following antimicrobial compounds were assessed: ampicillin (AMP, 1–128 mg/L), cefazolin (CZO, 0.5–64 mg/L), cefotaxime (CTX, 0.015–128 mg/L), cefotaxime–clavulanic acid, ceftazidime (CAZ, 0.03–64 mg/L), ceftazidime–clavulanic acid, chloramphenicol (CHL, 1–128 mg/L), ciprofloxacin (CIP, 0.015–512 mg/L), gentamicin (GEN, 0.125–64 mg/L), imipenem (IMP, 0.03– 16 mg/L), tetracycline (TET, 0.25–64 mg/L), tigecycline (TGC, 0.015–32 mg/L), and trimethoprim–sulfamethoxazole (SXT, 0.06/1.19-16/304 mg/L). MDR was defined as resistance to three or more different classes of agent (Wang et al., 2013).

# PCR Amplification and DNA Sequence Analysis of ESBLs

The following ESBLs resistance determinants were investigated by polymerase chain reaction (PCR): blacarB, blaCMY−2, blaCTX-M, blaOXA, blaSHV, blaTEM (Xu et al., 2005; Karczmarczyk et al., 2011). All PCR products amplified from blaCTX and blaTEM genes were commercially sequenced (Takara Biotechnology Cooperation, Dalian, China) and subsequently analyzed using the BLAST program<sup>2</sup> .

# Plasmid Characterization: S1 Nuclease Digestion and PCR-Based Plasmid Replicon Typing

S1-nuclease (Promega, Madison, WI, USA) digestion as well as pulsed-field gel electrophoresis (PFGE) analysis was performed for all 51 isolates. Briefly, the procedure included a lysis step of the bacterial cells embedded in agarose plugs followed by digestion with 8 U S1 nuclease at 37◦C for 30 min. Finally, each plasmid sample was resolved by PFGE in a Chef-Mapper <sup>R</sup> XA System (Bio-Rad, USA) at 14◦C, with a switch time between 1 and 12 s, at 6 V/cm on a 120◦ angle in 0.5× tris-boric acid-ETDA (TBE) buffer for 18 h. Each DNA band identified was considered as a unit length of linear plasmid. The approximate sizes of plasmids were determined by comparing profiles with XbaI-digested DNA from Salmonella serotype Braenderup H9812. Plasmids were assigned to incompatibility groups on the basis of the presence of specific replicon sequences identified by PCR using the primers previously designed and the corresponding amplification protocols described (Wang et al., 2013).

# Conjugation-Based Mating Experiments and Verification

Conjugation experiments were performed using a broth mating protocol to determine if plasmids coding for CTX-M and TEM enzymes could be transferred (Barton et al., 1995). Fifty-one donors were used for mating with a sodium azide-resistant E. coli J53 as previously described (Wang et al., 2013). Transconjugants were selected on LB agar plates containing sodium azide

<sup>1</sup>http://www.eucast.org

<sup>2</sup>http://www.ncbi.nlm.nih.gov/BLAST/

(100 mg/L) and cefotaxime (2 mg/L). PCR amplification, antimicrobial susceptibility testing, plasmid replicon typing, and S1-PFGE were performed for all the transconjugations to determine the presence of ESBL genes, antibiotic phenotypes and incompatibility groups, respectively (Wang et al., 2013).

# Phylogenetic Grouping

Phylogenetic groups were determined for each E. coli isolate using an established multiplex PCR targeting chuA (279 bp), yjaA (211 bp), and tspE4 (152 bp) according to the protocol of Clermont et al. (2000). The method was previously developed to classify E. coli into four phylogenetic groups designated A, B1, B2, and D.

# Multi-locus Sequence Typing of Isolates

MLST analysis was conducted by sequencing fragments of seven housekeeping genes (adk, fumC, gyrB, icdF, mdh, purE, recA) and sequence types (STs) were assigned using the E. coli MLST website (Wirth et al., 2006). eBURST was performed for MLST analysis.

# Detection of Virulence Genes of Diarrheagenic E. coli

Five distinct classes of diarrheagenic E. coli (DEC) are recognized as being associated with diarrheal disease: enteropathogenic E. coli (EPEC), Shiga toxin-producing E. coli (STEC), enteroaggregative E. coli (EAEC), enteroinvasive E. coli (EIEC), and enterotoxigenic E. coli (ETEC). PCR was used to distinguish these E. coli pathotypes by amplification of the following gene targets: typical EPEC (eae and bfp), atypical EPEC (eae or bfp), STEC (eae and stx1 and/or stx2), ETEC (elt and/or estIa or estIb), EIEC (invE and ipaH), and EAEC (aggR and/or aatA or aaiC) (Imuta et al., 2016; Tian et al., 2016).

# RESULTS

# Antimicrobial Susceptibility Testing of 51 ESBLs Isolates

All 51 ESBL-producing E. coli isolates were found to be resistant to ampicillin, cefazolin, cefotaxime, ceftazidime, and susceptible to imipenem and tigecycline. Resistance to tetracycline (27/51, 53.0%) was most common, followed by resistance to trimethoprim–sulfamethoxazole and gentamicin (49.0%, 25/51). All E. coli isolates expressed a MDR phenotype (being resistant to at least three different classes of antimicrobial compounds; **Figure 1**).

# Identification of β-Lactamase Resistance-Encoding Genes

Of the 51 isolates, blaCTX-M, blaTEM, blaOXA, and blaSHV were detected in 51, 26, 3, and 1 isolates, respectively. Sequencing revealed seven blaCTX-M subtypes: blaCTX-M-14 (n = 15), blaCTX-M-79 (n = 14), blaCTX-M-28 (n = 10), blaCTX-M-27 (n = 5), blaCTX-M-15 (n = 3), blaCTX-M-24 (n = 2), and blaCTX-M-3 (n = 2). Of the 26 TEM-positive strains, all were blaTEM−1.


FIGURE 1 | A heat-map summary of the sources; a schematic showing the S1 nuclease plasmid profile; ESBL markers identified by PCR and sequencing, the resistance profile, the corresponding Inc plasmid type(s), MLST and DEC type for all 51 ESBL-positive E. coli previously reported from humans in China. Black squares shown indicate a feature present in that isolate denoting its original source the ESBL marker(s) detected, its corresponding antibiogram profile and the Inc types detected. White squares denote features that are lacking in the corresponding bacterial isolate. Antimicrobial compounds used are abbreviated as follows: AMP, ampicillin; CAZ, ceftazidime; CHL, chloramphenicol; CIP, ciprofloxacin; CTX, cefotaxime; CZO, cefazolin; GEN, gentamicin; IMP, imipenem; SXT, trimethoprim/sulfamethoxazole; TET, tetracycline, TGC, tigecycline.

fmicb-08-00144 February 9, 2017 Time: 13:46 # 4

# S1 Nuclease Digestion and PCR-Based Plasmid Replicon Typing

S1 nuclease-based plasmid analysis showed that all 51 ESBLpositive isolates contained detectable large plasmids; most possessed two plasmids (n = 24, 47.1%) and some (n = 12, 23.5%) had three plasmids. All of the remaining E. coli isolates harbored a single plasmid. Heterogeneity among the profiles was a common feature noted (**Figure 1**). Eighteen plasmid replicons were detected by qualitative PCR-based plasmid replicon typing (PBRT) among those 51 isolates carrying large plasmids. PBRT typing identified 10 of 18 replicons. Interestingly, IncA/C, IncT, IncW, IncFIIA, IncX IncHI1, IncHI2, or IncL/M types were not be detected by PCR analysis in our collection. IncFrep (n = 50), IncK/B (n = 31), IncFIB (n = 26), IncB/O (n = 14), and IncI1-Ir (n = 8) replicon types were the predominated types. Several isolates were positive for more than one replicon type (n = 46). A summary of these features along with the corresponding antimicrobial resistance profiles for all 51 isolates is shown as a heat-map in **Figure 1**.

# Conjugational Transfer of Resistance Carried by the 26 E. coli Isolates

Each of the 51 E. coli isolates expressing an ESBL phenotype, was tested for its ability to transfer the phenotype, by conjugation under laboratory conditions. Twenty-six isolates transferred the cefotaxime resistance marker to a susceptible E. coli recipient with frequencies ranging from 3.8 × 10−<sup>2</sup> to 4.2 × 10−<sup>1</sup> transconjugants per donor cell (**Figure 2**). Following conjugation the antimicrobial susceptibility profiles of all 26 transconjugants was determined by the disk diffusion method. In addition to cefotaxime resistance, resistance to several non-β-lactam-based antimicrobial compounds such as gentamicin, tetracycline, and trimethoprim–sulfamethoxazole were also transferred to the recipient, suggesting that these might be located on the same plasmid. For transconjugants, the most commonly detected blaCTX-M-encoding genes included blaCTX-M-79 followed by blaCTX-M-14 and blaCTX-M-28. It should be noted that nearly half of those genes both in the blaCTX-M-1 group and the blaCTX-M-9 group were transferable to the recipient bacterium, under laboratory conditions. S1-PFGE analysis showed that the donor isolates carried multiple plasmids with sizes ranging from 34- to 294-kb.

# Phylogenetic Grouping and MLST

In all 51 isolates tested, 12 were assigned to B2 and 10 assigned to group D. Twenty isolates were assigned to group A and 9 were assigned to group B1 (**Figure 1**). MLST sub-typing identified 27 types along with 5 new STs not previously registered in the E. coli MLST database [including the isolates BJ6 (58-53-12-58-24-01- 42), BJ13 (43-41-15-08-11-08-06), BJ19 (261-160-02-63-55-04), BJ20 (261-160-02-63-55-04), BJ20 (56-11-04-10-07-08-06)]. The most prevalent ST was ST266, followed by ST10, ST38, ST131, ST2178, and ST4591. Most of these strains belonged to clonal complexes (cc) of ST10cc, ST266cc, ST131cc, and ST38cc in **Figure 1**.

# Occurrence of Pathogens

PCR screening of the typical virulence genes, revealed that 20 of the 51 ESBL-producing study isolates can be assigned to a pathotype with seven EPEC, six ETEC, six EAEC, and one ETEC isolate, being noted (**Figure 1**).

# DISCUSSION

ESBL-producing E. coli are important causative agents of foodborne infections worldwide (Nischal, 2014). In China, 60.7, 88.8, and 57.1% ESBL-positive E. coli were detected from isolates sampled from chicken farms in Henan province (Yuan et al., 2009), healthy broilers in Shandong province (Li et al., 2016), and in piglets (Zhang et al., 2015) separately which in turn may transmit to the human population causing food-borne infections (Leverstein-van Hall et al., 2011). Therefore, tracking the transmission routes of food-borne E. coli in community settings is a necessary step required as efforts are made to limit its continuous spread. In China, there is limited information available to describe the nature of ESBL-positive bacteria cultured from patients with diarrhea (Liu et al., 2015; Fan et al., 2016).

In this study, the observed a prevalence of 5.6% (51/912) that was similar to data published previously (Valenza et al., 2014; Hamprecht et al., 2016). Plasmids confer positively selectable phenotypes including antimicrobial resistance genes among others (Andersson and Hughes, 2010). These mobile genetic elements represent an important pool of adaptive and transferable genetic information (with large plasmids defined as being >30-kb in size; Andersson and Hughes, 2010).

Resistance to third-generation cephalosporins was most commonly associated by ESBLs of the CTX-M-1 group in the present study (54.9%), followed by ESBLs of the CTX-M-9 group (45.1%). When comparing these data with other studies from China investigating only E. coli, the prevalence of the CTX-M-1 group (55.7%) and CTX-M-9 group (43.3%) among E. coli ESBLs was similar to those values reported in an earlier study detected from BSIs (Wang S. et al., 2016), but different compared with the prevalences in UK, the Netherlands, and Germany in which CTX-M-1 was the dominant group compared to CTX-M-9 not only from human isolates but also in those cultured from food-producing animals (Wu et al., 2013; Hamprecht et al., 2016). Genotypes corresponding to several CTX-M enzymes were identified, the majority of which were found to be of the blaCTX-M-14 sub-type, compared with other CTX-M types (such as blaCTX-M-15) identified in hospitals (Liu et al., 2015; Wang L.H. et al., 2016) which differed from data reported from a single survey of retail foods (which identified blaCTX-M-55) in China (Xu et al., 2014).

As TEM-1 is the most common plasmid-mediated β-lactamase identified in enteric Gram-negative bacilli, high rates of ampicillin-resistant, corresponding with blaTEM−<sup>1</sup> positive isolates are to be expected. In the current study, the blaTEM−<sup>1</sup> gene was also identified in six transconjugants (**Figure 2**). Particular plasmid incompatibility groups are more frequently encountered among E. coli and these are thought to play a major role in the dissemination of specific resistance


FIGURE 2 | A heat-map showing the comparison of ESBL-positive E. coli donors and the resultant transconjugants, characterized on the basis of their plasmid profiles; ESBL-markers identified by PCR; antimicrobial resistance profile; and plasmid replicon type(s). Antimicrobial compounds are abbreviated as described in the legend to Figure 1.

genes. Interestingly, blaCTX-M genes were found to be located predominantly on IncFrep or IncK/B types and conjugation experiments showed that blaCTX-M from 26 isolates were transferable, with IncFrep and IncFIB being the most prevalent replicon types, a feature that differed from data reported previously (Wang et al., 2013; Xie et al., 2016). Most of the transferable isolates belonged to the phylogenetic group A followed by groups-B2, -B1, and -D. Since virulent organisms causing infections are commonly associated with phylogroup B2, and to a lesser extent group D, most of the isolates in this study belonged to phylogroups-B1 and -A, which are regarded as commensal bacteria (Wirth et al., 2006). Based on data obtained from the in vitro conjugation experiments, some 51% of the isolates that were capable of transferring one or more plasmids were phylogrouped as commensal isolates. This finding is consistent with previous reports showing that commensal E. coli in food-producing animals were likely to be a reservoir of ESBL-encoding genes and these bacteria played a role in the dissemination of such mobile elements (Smet et al., 2010). To our surprise, nearly half of the isolates are DEC. It would appear that the CTX-M resistance mechanism has spread to several types of E. coli. Moreover EAEC O25:H4/CG131 is thought to have evolved to harbor CTX-M plasmids (Imuta et al., 2016).

The diversity of MLST profiles identified suggested that the blaCTX-M genes were acquired as a result of horizontal transfer from existing resistant organisms, rather than clonal expansion of specific resistant strains (Adler et al., 2016). ST10 clone complex was the most prevalent STcc detected in our study and which are associated with some are pathogenic E. coli types such as ETEC, EAEC, EPEC. Also this ST10cc was reported earlier in Europe (Day et al., 2016) and was associated with the production of different CTX-M enzymes. Further it has been observed that the latter can harbor the recently reported mcr-1 gene (Bernasconi et al., 2016). ST266 was also detected and this is a ST type that had not previously been reported, therefore it may represent a new MLST type identified in China. Most of the ST266 types observed were identified as ETEC isolates, which is a major cause of acute secretory diarrhea and often referred to as the cause of travelers' diarrhea, being one of the major causes of diarrheal disease in children from developing countries (Joffre et al., 2016). ST131

# REFERENCES


is a globally spread clonal group of extraintestinal pathogenic E. coli (ExPEC), comprising different sub-lineages with the ability to acquire and spread antibiotic resistance and virulence genes via plasmids. It is seldom detected in companion and food-producing animals (Lanza et al., 2014; Day et al., 2016; Imuta et al., 2016; Wang L.H. et al., 2016). In this study, only three ST131-B2 isolates associated with the production of different CTX-M enzymes (including the genotypes blaCTX-M-14 or blaCTX-M-79 with blaTEM−1) were identified in contrast to earlier reports (Lanza et al., 2014). This ST poses a major threat to public health because of its global distribution. Identification of ST131 isolates that produce CTX-M has important implications for the future treatment of community-associated infections. The potential widespread dissemination of ST131 with its distinctive combination of resistance and virulence requires further investigation.

# CONCLUSION

The blaCTX-M positive E. coli characterized in this study pose a serious challenge for the treatment of human infections. These isolates were found to be not only resistant to cephalosporins but to other classes of antimicrobial compound. Continuous investigation and surveillance will extend our understanding of the transmission dynamics and the evolution of these isolates.

# AUTHOR CONTRIBUTIONS

LB, RL, SF, and FL designed the experiments. LB, SF, and FL wrote the paper. LB, LW, and JW did the experiments and analyzed the data. XY, XG, WW, and QC collected the strains and analyzed the data. JX discussed the results.

# FUNDING

This work was supported by National Natural Science Foundation of China (81402685) and Beijing Talents Fund (2015000021223ZK35)


identified in multi-drug resistant Escherichia coli isolated from food-producing animals and healthy humans. Front. Microbiol. 4:188. doi: 10.3389/fmicb.2013. 00188


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Bai, Wang, Yang, Wang, Gan, Wang, Xu, Chen, Lan, Fanning and Li. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

fmicb-08-00144 February 9, 2017 Time: 13:46 # 8

## Edited by:

Benoit Doublet, National Institute for Agricultural Research, France

#### Reviewed by:

Vincent Burrus, Université de Sherbrooke, Canada Sophie Payot, INRA – Nancy Centre, France Xavier Bellanger, Université de Lorraine, France

#### \*Correspondence:

Janine T. Bossé j.bosse@imperial.ac.uk; Paul R. Langford p.langford@imperial.ac.uk

#### †Present address:

Roy R. Chaudhuri, Department of Molecular Biology and Biotechnology, University of Sheffield, Western Bank, Sheffield, UK; Matthew T. G. Holden, School of Medicine, University of St Andrews, St Andrews, UK ‡These authors have contributed equally to this work. §Members of the BRaDP1T consortium are listed in the acknowledgments.

#### Specialty section:

This article was submitted to Antimicrobials, Resistance and Chemotherapy, a section of the journal Frontiers in Microbiology

Received: 10 March 2016 Accepted: 12 May 2016 Published: 15 June 2016

#### Citation:

Bossé JT, Li Y, Fernandez Crespo R, Chaudhuri RR, Rogers J, Holden MTG, Maskell DJ, Tucker AW, Wren BW, Rycroft AN, Langford PR and the BRaDP1T Consortium (2016) ICEApl1, an Integrative Conjugative Element Related to ICEHin1056, Identified in the Pig Pathogen Actinobacillus pleuropneumoniae. Front. Microbiol. 7:810. doi: 10.3389/fmicb.2016.00810

# ICEApl1, an Integrative Conjugative Element Related to ICEHin1056, Identified in the Pig Pathogen Actinobacillus pleuropneumoniae

Janine T. Bossé<sup>1</sup> \* ‡ , Yanwen Li<sup>1</sup>‡ , Roberto Fernandez Crespo<sup>1</sup> , Roy R. Chaudhuri<sup>2</sup>† , Jon Rogers<sup>3</sup> , Matthew T. G. Holden<sup>4</sup>† , Duncan J. Maskell<sup>2</sup> , Alexander W. Tucker<sup>2</sup> , Brendan W. Wren<sup>5</sup> , Andrew N. Rycroft<sup>6</sup> , Paul R. Langford<sup>1</sup> \* and the BRaDP1T Consortium§

<sup>1</sup> Section of Paediatrics, Department of Medicine, Imperial College London, London, UK, <sup>2</sup> Department of Veterinary Medicine, University of Cambridge, Cambridge, UK, <sup>3</sup> Animal and Plant Health Agency Bury St Edmunds, Suffolk, UK, <sup>4</sup> The Wellcome Trust Sanger Institute, Wellcome Trust Genome Campus, Hinxton, Cambridge, UK, <sup>5</sup> Faculty of Infectious and Tropical Diseases, London School of Hygiene and Tropical Medicine, London, UK, <sup>6</sup> Department of Pathology and Pathogen Biology, The Royal Veterinary College, Hatfield, UK

ICEApl1 was identified in the whole genome sequence of MIDG2331, a tetracyclineresistant (MIC = 8 mg/L) serovar 8 clinical isolate of Actinobacillus pleuropneumoniae, the causative agent of porcine pleuropneumonia. PCR amplification of virB4, one of the core genes involved in conjugation, was used to identify other A. pleuropneumoniae isolates potentially carrying ICEApl1. MICs for tetracycline were determined for virB4 positive isolates, and shotgun whole genome sequence analysis was used to confirm presence of the complete ICEApl1. The sequence of ICEApl1 is 56083 bp long and contains 67 genes including a Tn10 element encoding tetracycline resistance. Comparative sequence analysis was performed with similar integrative conjugative elements (ICEs) found in other members of the Pasteurellaceae. ICEApl1 is most similar to the 59393 bp ICEHin1056, from Haemophilus influenzae strain 1056. Although initially identified only in serovar 8 isolates of A. pleuropneumoniae (31 from the UK and 1 from Cyprus), conjugal transfer of ICEApl1 to representative isolates of other serovars was confirmed. All isolates carrying ICEApl1 had a MIC for tetracycline of 8 mg/L. This is, to our knowledge, the first description of an ICE in A. pleuropneumoniae, and the first report of a member of the ICEHin1056 subfamily in a non-human pathogen. ICEApl1 confers resistance to tetracycline, currently one of the more commonly used antibiotics for treatment and control of porcine pleuropneumonia.

Keywords: animal infections, antibiotic resistance, respiratory tract, conjugation, Pasteurellaceae

# INTRODUCTION

Actinobacillus pleuropneumoniae is a major contributor to swine respiratory disease worldwide, causing considerable economic losses. Isolates can be differentiated into 15 established serovars, based on capsular polysaccharides, and a recently proposed serovar 16 identified on the basis of serology alone (Sárközi et al., 2015). There are geographical differences in the distribution of serovars. Within the UK, clinical isolates are predominantly serovar 8, with serovars 2, 6, 7, and 12 also represented (O'Neill et al., 2010).

There is growing concern regarding antimicrobial resistance in bacteria from food-producing animals (Michael et al., 2015). In Europe, tetracyclines are still the most commonly used antimicrobial for treatment of swine pleuropneumonia (European Medicines Agency, 2012). The genes tetB, tetH, tetL, and tetO, reported to mediate tetracycline resistance in A. pleuropneumoniae, are usually carried on small plasmids (Blanco et al., 2006, 2007). We recently sequenced the genome of MIDG2331, a serovar 8 UK clinical isolate of A. pleuropneumoniae (Bossé et al., 2016), and identified chromosomally encoded tetracycline resistance genes within a putative integrative conjugative element (ICE). Similar to genomic islands, ICE have the ability to integrate into bacterial chromosomes at specific sites, often in tRNA loci, via the action of an integrase (predominantly tyrosine recombinases) (Boyd et al., 2009; Wozniak and Waldor, 2010). However, ICE differ from genomic islands in that they are self-mobilizing, encoding all of the genes necessary for excision from the chromosome and conjugal transfer (Boyd et al., 2009). The core genes of ICE tend to group into functional modules, with syntenic regions responsible for maintenance, dissemination and regulation, which may be interspersed with accessory genes carried on transposons or other insertion elements (Burrus and Waldor, 2004; Wozniak and Waldor, 2010). The genes encoding the type 4 secretion


<sup>a</sup>All isolates except MIDG2427 and MIDG3458 were cultured from pigs submitted to the then Veterinary Laboratories Agency (now Animal and Plant Health Agency) regional laboratories in the UK, as indicated.

<sup>b</sup>The length of each ICEApl1 sequence, calculated as the predicted circular form, is shown for each isolate. Size variation from that in MIDG2331 is indicated in bold. The sequence in MIDG2648 is missing 3 genes and has a truncated copy of the site-specific recombinase gene; whereas small deletions in the other sequences are all intergenic.

<sup>c</sup>Order of tRNA genes upstream of ICEApl1: G, tRNA-Gly (GCC), L, tRNA-Leu (TAA). Variation from the order seen in MIDG2331 is indicated in bold.

d ICEApl1 identified in MIDG2331 in previous study,<sup>4</sup> all other ICEApl1 sequences were identified in this study.

system (T4SS), required for transport of DNA into recipient cells, include a ubiquitous ATPase encoded by virB4 or traU (Guglielmini et al., 2011).

ICEs are the most abundant conjugative elements identified in prokaryotes, and there is evidence of cross-clade transfer (Guglielmini et al., 2011). Within the Pasteurellaceae, ICE have been identified and characterized in Haemophilus influenzae and Haemophilus parainfluenzae (Juhas et al., 2007b), Pasteurella multocida (Michael et al., 2012), and Mannheimia haemolytica (Eidam et al., 2015). Here we report characterization of ICEApl1, to our knowledge the first ICE described in A. pleuropneumoniae.

# MATERIALS AND METHODS

# Comparative Sequence Analysis

The full sequence of ICEApl1, identified within the genome of MIDG2331 (accession number LN908249) was analyzed using BLASTn and BLASTx<sup>1</sup> . A comparative alignment was generated for sequences most similar to ICEApl1 using Mauve version 2.3.1<sup>2</sup> . Default parameters were used for all programs.

# Detection of Other Isolates Containing ICEApl1

We screened 185 isolates of A. pleuropneumoniae (clinical isolates collected between 1995 and 2012 from the UK, Denmark, the Czech Republic, Cyprus, and Greece) for virB4 by PCR using primers virB4\_for (CCTTCACGGTTAAAGAATCG AC)/virB4\_rev (GCATCGTTTATTGGAAATGGAT). Primers were designed based on the virB4 gene in MIDG2331, amplifying the region from 1532511 to 1532894 in the genome sequence. Serovars 1 (1.2%), 2 (11.7%), 5 (2.4%), 6 (4.7%), 7 (10.6%), 8 (58.8%), 9/11 (2.4%), 10 (2.9%), and 12 (5.3%), were represented, and 84% of the isolates were from the UK. Genome sequence data was generated and assembled as previously described (Howell et al., 2013; Bossé et al., 2015) for 31 virB4 positive isolates. Sequences matching ICEApl1 were identified by BLASTn, assembled using Geneious 9.0.4, and deposited to Genbank (see **Table 1** for accession numbers).

Minimum Inhibitory Concentrations (MICs) for tetracycline were determined for isolates containing ICEApl1, according to the CLSI M37-A3 guidelines (Clinical and Laboratory Standards Institute [CLSI], 2008).

# Conjugal Transfer of ICEApl1

MIDG23311ureC::nadV was used as the conjugal donor, with matings performed as previously described (Bossé et al., 2015). Plasmid-free, tetracycline-sensitive, nalidixic acid-resistant clinical isolates of serovars 6 (MIDG3376), 7 (MIDG2465), 8 (MIDG3217), and 12 (MIDG3347) were used as recipients. Transconjugants were selected on Brain Heart Infusion agar supplemented with 0.01% NAD, 5 mg/L tetracycline and 40 mg/L nalidixic acid. PCR was used to confirm the presence of the virB4 gene (as above), as well as serovar of, and the absence of

<sup>1</sup>http://blast.ncbi.nlm.nih.gov/

<sup>2</sup>http://darlinglab.org/mauve/mauve.html

nadV, in selected transconjugants using previously described primers (Bossé et al., 2014, 2015). Chromosomal insertion sites in transconjugants were determined by PCR using primers ICE50\_out1 (TGAGGGAGTAACAAGCAACACAG)/mfd30\_out (TTTACCGCTTGCCGATAATGCG) for the 5<sup>0</sup> junction, and ICE30\_out1 (CAATGGAGAAAGAGAGTTGTTGGAC) /hybF50\_out (GACATCTCGTGCATAACCATTCC) for the 3 0 junction, respectively. Amplicons were sequenced using internal primers ICE50\_out2 (GGAAGGTTCAATATCA CGACGG) or ICE30\_out2 (AGGCATACAGCAGCAACAAATC), as appropriate. For comparison, the region between mfd and hybF in the conjugal recipients (prior to conjugation) was amplified using mfd30\_out/hyb50\_out and sequenced in both directions.

# Confirmation of the Circular Extrachromosomal form of the ICE by Nested PCR

DNA was extracted from MIDG23311ureC::nadV and selected transconjugants, and nested PCR was performed as previously described (Eidam et al., 2015), using primers ICE50\_out1/ICE30\_out1 followed by primers ICE50\_out2/ICE30\_out2. Amplicons were sequenced using primers ICE50\_out2 and ICE30\_out2.

# RESULTS AND DISCUSSION

# Sequence of ICEApl1 and Comparative Analysis

ICEApl1, a 56083 bp element, is inserted into a tRNA-Leu (TAA) gene, a common insertion site for ICE in the family Pasteurellaceae (Dimopoulou et al., 2002; Michael et al., 2012; Eidam et al., 2015), in a tRNA cluster located between genes hybF and mfd. In the MIDG2331 genome (Bossé et al., 2016), this tRNA-Leu (TAA) gene is annotated as MIDG2331\_01481, and is located between bases 1570419 and 1570505. Although all of the tRNA genes in the cluster, as well as hybF and mfd, are on the complement strand in the MIDG2331 genome, all further references to these genes, and the location of ICEApl1, will be with respect to the forward orientation. Insertion of ICEApl1 generated 66 bp imperfect direct repeats (DRs) at the left and right attachment sites, attL and attR (**Figure 1A**). The three bases that differ in the DRs reflect sequence variation in the tRNA-Leu (TAA) genes in A. pleuropneumoniae and H. influenzae (**Figure 1B**). The attI site in the closed circular form of ICEApl1 (confirmed by PCR; see below) is identical to that in ICEHin1056 (**Figure 1C**), both having 100% identity with the last 65 bases of the H. influenzae tRNA-Leu (TAA) gene followed by a T. These data suggest that insertion of ICEApl1 in MIDG2331 has generated an altered tRNA-Leu (TAA) gene, resulting from cross-over of the circular plasmid form of the ICE into the chromosome between the G at position 25 and the A at position 54 of MIDG2331\_01481.

FIGURE 1 | Analysis of ICEApl1 insertion site. The imperfect direct repeats (DRs) (A) flanking ICEApl1 share sequence identity with the last 65 bases of tRNA-Leu genes (B) from Actinobacillus pleuropneumoniae (e.g., MIDG2331\_01482 and \_01484) and Haemophilus influenzae (e.g., accession number LK008335), which differ at 3 bases (in light blue for bases normally found in A. pleuropneumoniae and red for bases normally found in H. influenzae). The attI sequences (C) in the closed circular forms of ICEApl1 and ICEHin1056 are identical, and match the end of the H. influenzae tRNA-Leu (TAA) gene sequence, with an additional T

which is also present in the DRs (A). Insertion of ICEApl1 in the A. pleuropneumoniae tRNA-Leu (TAA) gene MIDG2331\_01481 has resulted in an altered sequence,

indicating cross-over of the circular ICEApl1 into the chromosome occurred between the G at position 25 and the A at position 54.

Comparative sequence analysis revealed that ICEApl1 is related to the ICEHin1056 subfamily of elements (**Figure 2**) found in H. influenzae and H. parainfluenzae (Mohd-Zain et al., 2004; Juhas et al., 2007b). ICEApl1 encodes 67 genes that share extensive sequence homology and gene order with ICEHin1056 and other members of this subfamily. The first 14687 bp of ICEApl1 shares 99% identity with the region of ICEHin1056 reported to contain replication and stabilization genes (Juhas et al., 2007a,b, 2013). The 8933 bp Tn10 element in ICEApl1, although in the same location and orientation as that in ICEHin1056, more closely resembles that in ICEHpaT3T1 (99% identity, but inverted), with tetracycline resistance genes tetR, tetB, tetC, and tetD; and gltS encoding glutamate permease (Juhas et al., 2007b). The Tn10 element in ICEHin1056 has a further IS5 insertion (encoding chloramphenicol resistance) within it that is not seen in ICEApl1 (Juhas et al., 2007b). The 20466 bp following the Tn10 insertion shares 99% identity with the region containing genes encoding components of the type IV secretion system (required for conjugal transfer) in ICEHin1056 (Juhas et al., 2007a,b, 2013). This region is well conserved in all of the members of the ICEHin1056 family (Juhas et al., 2007a,b, 2013). The gene order in ICEApl1 remains syntenic with that of ICEHin1056 up to traC, where in ICEHin1056 there is a Tn3 insertion (encoding beta-lactamase resistance) that is not found in ICEApl1. From traC in ICEApl1, the nucleotide sequence and gene order more closely resemble those in ICEHin2866 up to the site-specific tyrosine recombinase gene, which is the final gene in ICEApl1 on the attR side. This 8.5 kb region shares 98% identity with sequences in ICEHin2866, and includes accessory genes encoding a type I restriction enzyme M subunit, and a

transposon gamma-delta resolvase, as well as four hypothetical genes of unknown function (Juhas et al., 2007b).

Surprisingly, the ICEHin1056 subfamily of conjugative elements has previously only been reported in H. influenzae and H. parainfluenzae, two human species of Haemophilus, where they appear to be evolving by descent (Dimopoulou et al., 2007; Juhas et al., 2007b). To our knowledge, this is the first report of a member of the ICEHin1056 subfamily in a Pasteurellaceae species that infects livestock. There have been ICE reported for bovine isolates of P. multocida (ICEPmu1) and M. haemolytica (ICEMh1), as well as an uncharacterized putative ICE in Histophilus somni strain 2336, which are related and appear to have evolved from a common ancestor, but are part of a different subfamily than ICEHin1056 elements (Juhas et al., 2007b; Michael et al., 2012; Eidam et al., 2015). The identity of the attI sites in both ICEApl1 and ICEHin1056 with the last 65 bases of the H. influenzae tRNA-Leu (TAA) gene would suggest more recent acquisition of an ICEHin1056 element in A. pleuropneumoniae.

# Distribution of ICEApl1 in A. pleuropneumoniae Isolates

PCR analysis revealed the presence of a virB4 amplicon in 32/185 A. pleuropneumoniae isolates, including MIDG2331. All 32 (31 from the UK, 1 from Cyprus) were serovar 8 and had an MIC for tetracycline of 8 mg/L, i.e., above the CLSI breakpoint of ≥ 2 mg/L for A. pleuropneumoniae (Clinical and Laboratory Standards Institute [CLSI], 2008). ICEApl1 sequences were detected in the whole genomes of the 32

isolates, and comparative analysis revealed that, other than minor nucleotide differences in some, all of the sequences were complete except the element from MIDG2648, which was lacking three genes (encoding a putative DNA-binding protein and two hypothetical proteins) in the accessory gene region, and has a truncated copy of the tyrosine recombinase gene (**Table 1**).

# Conjugal Transfer of ICEApl1 and Detection of Circular Intermediate Form

As ICEApl1 appeared to be present only in serovar 8 isolates of A. pleuropneumoniae, it was possible that other serovars blocked conjugal entry of the ICE, either due to restriction modification systems or CRISPR mediated restriction (Elhai et al., 1997; Garneau et al., 2010). We therefore tested the ability to conjugally transfer ICEApl1 to clinical isolates of A. pleuropneumoniae representing serovars 6, 7, 8, and 12 (MIDG3376, MIDG2465, MIDG3217, and MIDG3347, respectively) that are commonly found in the UK (O'Neill et al., 2010). All tested isolates produced transconjugants, as initially confirmed by PCR (data not shown). The frequencies of conjugation were similar for the serovar 7, 8, and 12 recipients (between 10−<sup>4</sup> and 10−<sup>5</sup> ), but much lower (5 × 10−<sup>8</sup> ) for the serovar 6 isolate tested. Similar frequencies have been reported for ICEHin1056 elements in H. influenzae, with strain related differences also noted (Juhas et al., 2007b). More serovar 6 isolates would need to be tested in order to determine if the difference in frequency of conjugation for ICEApl1 is serovar-specific. The presence of a circular intermediate form of ICEApl1 was confirmed in the donor strain and in transconjugants by nested PCR. Sequenced amplicons confirmed a single copy of the 66 bp attI (**Figure 1C**) at the closed junction of the circular intermediates.

Sequencing PCR products generated at both the attL and attR ends in the transconjugants confirmed insertion of ICEApl1 in the same tRNA cluster between mfd and hybF as in the donor strain (MIDG23311ureC::nadV). However, in all transconjugants tested, the sequences between mfd and hybF contained only the altered tRNA-Leu (TAA) gene in which the ICE inserted, flanked by tRNA-Gly (GCC) genes at either end of the cluster (**Figure 3B**). This is in contrast to the cluster in the donor strain where there are 5 tRNA genes on the attL side, and a tRNA-Gly (GCC) gene on the attR side (**Figure 3C**). Sequencing across the tRNA cluster in the recipient strains prior to conjugation revealed that, although a different order of tRNA genes was present in MIDG3376 compared to the other strains (**Figure 3A**), all contained 3 copies each of tRNA-Gly

tRNA-Leu (TAA) gene shown in red has the altered bases of the attL DRs in Figure 1A.

(GCC) and tRNA-Leu (TAA) genes. These results indicate that in all transconjugants tested, a deletion of 3 tRNA genes was associated with ICEApl1 insertion. However, examination of the insertion sites in the serovar 8 clinical isolates with endogenous ICEApl1 (**Table 1**) revealed conservation of the 6 tRNA genes normally found in this cluster – i.e., 3 copies each of tRNA-Gly (GCC) and tRNA-Leu (TAA) – with the order of genes showing one of 3 patterns (see **Figure 3C** and **Table 1** for details). In MIDG2648, the truncated element present in this isolate is located in a different copy of the tRNA-Leu (TAA) gene than seen in isolates with intact ICEApl1 (**Figure 3C**), with 4 tRNA genes on the attL side, and a tRNA-Leu (TAA) followed by tRNA-Gly (GCC) gene on the attR side. It would appear that, although the majority of clinical isolates show integration in the same tRNA-Leu (TAA) gene (i.e., the final copy in the tRNA cluster in the forward orientation), ICEApl1 has the ability to integrate into different copies of this gene, as has been reported for ICEHin1056 in H. influenzae (Dimopoulou et al., 2002). In Pseudomonas knackmussii, ICEclc was found to insert into different copies of the tRNA-Gly (GCC) gene, with double integration in some transconjugants (Sentchilo et al., 2009). In that study, excision and reintegration was associated with generation of a heterogenous population in which ICEclc was found to move from its original insertion site to alternate tRNA-Gly genes, but only those with the GCC anticodon (Sentchilo et al., 2009). Similarly, following in vitro conjugal transfer of ICEKp1 into a recipient strain of Klebsiella pneumoniae, integration was found at any of four tRNA-Asn genes, with insertion in multiple copies in some transconjugants (Lin et al., 2008). Furthermore, in some transconjugants, there was evidence of deletions between copies of the tRNA-Asn genes associated with integration of ICEKp1, which may have been due to recombination between multiple insertion sites (Lin et al., 2008). We did not detect multiple insertions of ICEApl1 in our transconjugants, as only one PCR product was generated using the outward facing primers designed to amplify the closed junction of the circular ICE. These primers would also have generated a secondary product in the presence of tandem insertions, given the proximity of the copies of the tRNA-Leu (TAA) genes in A. pleuropneumoniae. It is also possible that recombination may occur between the DRs found in alternate copies of the target tRNA, with or without the presence of an integrated ICE, resulting in deletion of the intervening sequences. However, given the results of Lin et al. (2008), it is likely that conditions during in vitro conjugation favor integration of multiple copies of ICE leading to deletions, whereas this does not appear to be common amongst clinical isolates. Little is known regarding the signals that govern initiation of horizontal transfer of ICE between pathogens in a host animal environment.

# CONCLUSION

fmicb-07-00810 June 14, 2016 Time: 16:33 # 7

Identification of ICEApl1 in only serovar 8 clinical isolates of A. pleuropneumoniae may simply be a reflection of this being the most common in the UK (O'Neill et al., 2010), and thus in our collection. It may also indicate a tendency for ICE to be inherited by vertical transmission rather than horizontal transfer. The similar in vitro conjugation frequencies of ICEApl1 into isolates of serovars 7, 8, and 12 suggests there are no restriction endonuclease or CRISPR barriers to transfer between these serovars. Futhermore, the variation in order of tRNA genes flanking ICEApl1 suggests that horizontal transfer may have occurred independently into different isolates, as a similar variation in tRNA gene order was also seen in clinical isolates lacking ICEApl1. As reported for other ICE, ICEApl1 has the ability to integrate into different copies of the target tRNA gene, in this case tRNA-Leu (TAA). Although we did not detect multiple insertions following in vitro conjugal transfer of ICEApl1, it is likely that recombination between tandem insertions was responsible for the deletions detected in transconjugants.

To our knowledge, this is the first description of an ICE identified in A. pleuropneumoniae, and the first report of a member of the ICEHin1056 subfamily found in a non-human pathogen. The presence of ICEApl1 in isolates of A. pleuropneumoniae confers resistance to tetracycline, which is commonly used for treatment and control of porcine pleuropneumonia (European Medicines Agency, 2012). Although currently only found in serovar 8 isolates, the ability to transfer to other serovars was confirmed in vitro, and has implications for the spread of antimicrobial resistance in this important pig pathogen.

# REFERENCES


# AUTHOR CONTRIBUTIONS

JB, PL, AR, BW, DM, and AT conceived the study; JB, YL, RFC, RRC, MH, and JR produced the data; JB, YL, RFC, and RRC analyzed the data; JB and PL wrote the paper.

# FUNDING

This work was supported by a Longer and Larger (LoLa) grant from the Biotechnology and Biological Sciences Research Council (BBSRC grant numbers BB/G020744/1, BB/G019177/1, BB/G019274/1, and BB/G018553/1), the UK Department for Environment, Food and Rural Affairs, and Zoetis (formerly Pfizer Animal Health) awarded to the Bacterial Respiratory Diseases of Pigs-1 Technology (BRaDP1T) Consortium. MTGH was supported by the Wellcome Trust (grant number 098051). JR was funded from the former AHVLA's Research and Development Internal Investment Fund (grant number RD0030c).

# ACKNOWLEDGMENTS

The BRaDP1T Consortium comprises: DM, AT, Sarah E. Peters, Lucy A. Weinert, Jinhong (Tracy) Wang, Shi-Lu Luan, RRC (University of Cambridge; present address for RRC is: Department of Molecular Biology and Biotechnology, University of Sheffield, Firth Court, Western Bank, Sheffield, S10 2TN, UK), AR, Gareth A. Maglennon, Jessica Beddow (Royal Veterinary College); BW, Jon Cuccui, Vanessa S. Terra (London School of Hygiene and Tropical Medicine); and PL, JB, YL (Imperial College London). The authors wish to thank Susanna Williamson and Chris Teale from the APHA for their advice and input.



**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Bossé, Li, Fernandez Crespo, Chaudhuri, Rogers, Holden, Maskell, Tucker, Wren, Rycroft, Langford and the BRaDP1T Consortium. This is an openaccess article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

fmicb-08-00311 March 2, 2017 Time: 16:34 # 1

### Edited by:

Benoit Doublet, National Institute for Agricultural Research, France

#### Reviewed by:

Liang Li, Los Angeles Biomedical Research Institute, USA Geovana Brenner Michael, Friedrich-Loeffler-Institut, Germany

#### \*Correspondence:

Janine T. Bossé j.bosse@imperial.ac.uk Paul R. Langford p.langford@imperial.ac.uk

#### †Present address:

Roy R. Chaudhuri, Department of Molecular Biology and Biotechnology, University of Sheffield, Firth Court, Western Bank, Sheffield, UK; Matthew T. G. Holden, School of Medicine, University of St Andrews, St Andrews, UK

‡These authors have contributed equally to this work.

> § Members are listed in the Acknowledgments section.

#### Specialty section:

This article was submitted to Antimicrobials, Resistance and Chemotherapy, a section of the journal Frontiers in Microbiology

Received: 15 November 2016 Accepted: 15 February 2017 Published: 06 March 2017

#### Citation:

Bossé JT, Li Y, Rogers J, Fernandez Crespo R, Li Y, Chaudhuri RR, Holden MTG, Maskell DJ, Tucker AW, Wren BW, Rycroft AN, and Langford PR on behalf of the BRaDP1T Consortium (2017) Whole Genome Sequencing for Surveillance of Antimicrobial Resistance in Actinobacillus pleuropneumoniae. Front. Microbiol. 8:311. doi: 10.3389/fmicb.2017.00311

# Whole Genome Sequencing for Surveillance of Antimicrobial Resistance in Actinobacillus pleuropneumoniae

Janine T. Bossé<sup>1</sup> \* ‡ , Yanwen Li<sup>1</sup>‡ , Jon Rogers<sup>2</sup> , Roberto Fernandez Crespo<sup>1</sup> , Yinghui Li<sup>1</sup> , Roy R. Chaudhuri<sup>3</sup>† , Matthew T. G. Holden<sup>4</sup>† , Duncan J. Maskell<sup>3</sup> , Alexander W. Tucker<sup>3</sup> , Brendan W. Wren<sup>5</sup> , Andrew N. Rycroft<sup>6</sup> , and Paul R. Langford<sup>1</sup> \* on behalf of the BRaDP1T Consortium§

<sup>1</sup> Section of Paediatrics, Department of Medicine, Imperial College London, London, UK, <sup>2</sup> Animal and Plant Health Agency, Bury St Edmunds, UK, <sup>3</sup> Department of Veterinary Medicine, University of Cambridge, Cambridge, UK, <sup>4</sup> The Wellcome Trust Sanger Institute, Cambridge, UK, <sup>5</sup> Faculty of Infectious and Tropical Diseases, London School of Hygiene and Tropical Medicine, London, UK, <sup>6</sup> Department of Pathology and Pathogen Biology, The Royal Veterinary College, Hatfield, UK

The aim of this study was to evaluate the correlation between antimicrobial resistance (AMR) profiles of 96 clinical isolates of Actinobacillus pleuropneumoniae, an important porcine respiratory pathogen, and the identification of AMR genes in whole genome sequence (wgs) data. Susceptibility of the isolates to nine antimicrobial agents (ampicillin, enrofloxacin, erythromycin, florfenicol, sulfisoxazole, tetracycline, tilmicosin, trimethoprim, and tylosin) was determined by agar dilution susceptibility test. Except for the macrolides tested, elevated MICs were highly correlated to the presence of AMR genes identified in wgs data using ResFinder or BLASTn. Of the isolates tested, 57% were resistant to tetracycline [MIC ≥ 4 mg/L; 94.8% with either tet(B) or tet(H)]; 48% to sulfisoxazole (MIC ≥ 256 mg/L or DD = 6; 100% with sul2), 20% to ampicillin (MIC ≥ 4 mg/L; 100% with blaROB−1), 17% to trimethoprim (MIC ≥ 32 mg/L; 100% with dfrA14), and 6% to enrofloxacin (MIC ≥ 0.25 mg/L; 100% with GyrAS83F). Only 33% of the isolates did not have detectable AMR genes, and were sensitive by MICs for the antimicrobial agents tested. Although 23 isolates had MIC ≥ 32 mg/L for tylosin, all isolates had MIC ≤ 16 mg/L for both erythromycin and tilmicosin, and no macrolide resistance genes or known point mutations were detected. Other than the GyrAS83F mutation, the AMR genes detected were mapped to potential plasmids. In addition to presence on plasmid(s), the tet(B) gene was also found chromosomally either as part of a 56 kb integrative conjugative element (ICEApl1) in 21, or as part of a Tn7 insertion in 15 isolates. Our results indicate that, with the exception of macrolides, wgs data can be used to accurately predict resistance of A. pleuropneumoniae to the tested antimicrobial agents and provides added value for routine surveillance.

Keywords: animal infections, antimicrobial resistance genes, integrative conjugative elements, plasmids, genomics, respiratory tract, Pasteurellaceae

# INTRODUCTION

fmicb-08-00311 March 2, 2017 Time: 16:34 # 2

Antimicrobial resistance (AMR) in bacteria from food-producing animals is a growing concern (Michael et al., 2015). Extensive use of antimicrobial agents for treatment and prevention of diseases fosters an environment in which resistance determinants are acquired and maintained by pathogens, as well as commensal bacteria. In the UK, the swine industry accounts for a large proportion of the antimicrobial agents sold for use in food-producing animals; with tetracyclines, beta-lactams, and trimethoprim/sulphonamides being the most commonly used antimicrobial agents (Burch, 2005; Borriello, 2013).

Actinobacillus pleuropneumoniae is a major contributor to swine respiratory disease, causing considerable economic losses worldwide. Strategies to reduce the incidence and severity of disease include good husbandry, vaccination, and antibiotic treatment. The latter is essential to limit the severity and spread of pleuropneumonia. Knowledge of resistance profiles for A. pleuropneumoniae is required to inform treatment decisions. Furthermore, this information contributes to the larger picture of AMR in bacteria of animal origin (Hendriksen et al., 2008; El Garch et al., 2016).

Typically, determination of antimicrobial susceptibility is done either by disk diffusion or minimum inhibitory concentration (MIC) assays. Identification of the genetic determinants of resistance not only corroborates phenotypic results, but is also useful for epidemiological purposes, as there are often multiple different genes that can confer resistance to a given antimicrobial agent. Both PCR and microarray hybridization have been used to detect the presence of genes encoding resistance phenotypes, but these assays are limited to detecting the sequences tested, and do not allow direct detection of mutations conferring resistances (Frye et al., 2010; Ledeboer and Hodinka, 2011). It has recently been proposed that whole genome sequencing (wgs) may be an alternative for routine surveillance of resistance profiles and for identification of emerging resistances (Zankari et al., 2013). We recently used wgs to identify (for the first time) dfrA14 as the genetic determinant of trimethoprim resistance detected in 16 clinical isolates of A. pleuropneumoniae (Bossé et al., 2015b). In this study, we compare the MIC profiles for nine antimicrobial agents with detection of resistance genes in wgs data of 96 isolates.

# MATERIALS AND METHODS

# Bacterial Strains and Antimicrobial Susceptibility Testing

A total of 96 clinical A. pleuropneumoniae isolates from the UK, previously tested for trimethoprim resistance (Bossé et al., 2015b), were analyzed using the agar dilution susceptibility assay, according to the CLSI VET01-A4 guidance (Clinical and Laboratory Standards Institute [CLSI], 2013), for determination of MICs for tetracycline, ampicillin, sulfisoxazole, enrofloxacin, erythromycin, tilmicosin, tylosin, and florfenicol. Some samples were re-tested for resistance using the disk diffusion susceptibility test according to the CLSI VET01-A4 guidance (Clinical and Laboratory Standards Institute [CLSI], 2013). A. pleuropneumoniae ATCC 27090 and Histophilus somni ATCC 70025 were used as controls for all susceptibility tests. All clinical A. pleuropneumoniae isolates had been cultured from pneumonic lungs of pigs submitted for post-mortem to the then Animal Health and Veterinary Laboratory Agency (now Animal and Plant Health Agency) diagnostic laboratories in England (see Supplementary Table S1 for details). The serovar of each isolate was determined by PCR as previously described (Bossé et al., 2014).

# Genome Sequencing and Analysis

Genomic DNA was prepared from all isolates using the FastDNA Spin kit (MP Biomedicals), and 0.5 µg were used for library preparation and paired-end sequencing (Illumina HiSeq 2000), as previously described (Howell et al., 2013; Bossé et al., 2015b; Weinert et al., 2015). Cutadapt (Martin, 2011) was used to remove Illumina adapter sequences, and Sickle<sup>1</sup> was used to trim low-quality sequences from the ends of reads, prior to assembly into contigs using Velvet 1.2.08 (Zerbino and Birney, 2008) and VelvetOptimiser 2.2.5 (Gladman and Seemann, 2012). Assemblies with N50 < 10000 were excluded from further analysis. The draft genome sequence for each isolate in this study has been deposited in the European Nucleotide Archive<sup>2</sup> , and accession numbers are listed in Supplementary Table S1.

AMR genes were identified in the draft genomes using ResFinder (Zankari et al., 2012), with a threshold of 98% identity and minimum length of 60%. Alternatively, the genomes were queried by BLASTn and tBLASTn using sequences of known resistance genes from A. pleuropneumoniae and other members of the Pasteurellaceae found in GenBank (Supplementary Table S2). Clustal alignments were used to compare the gyrA, gyrB, parC, and parE genes in isolates with elevated MICs for enrofloxacin (≥0.25 mg/L) with those genes in all remaining isolates in order to identify any mutations that could contribute to enrofloxacin resistance. The profiles of resistance genes identified were then compared to the results of the MIC assays, and where discrepancies were found, isolates were re-tested for susceptibility by disk diffusion, and by PCR for the presence of the specific AMR genes using primers listed in **Table 1**. Correlations between resistance phenotypes and genotypes were calculated using Fisher's exact test function ('fisher.test') in R (version 3.3.2).

# RESULTS

Full results of the testing and identification of AMR genes for each isolate are also shown in Supplementary Table S1. The correlations between presence of identified AMR genes and resistance phenotype are shown in **Table 2**. The majority of the 96 isolates tested in this study, representing serovars 2 (11.5%), 6 (7.3%), 7 (10.4%), 8 (68.8%), and 12 (2.1%), were collected between 2005 and 2009, with smaller numbers representing the other years between 1998 and 2011 (Supplementary Table S1).

<sup>1</sup>https://github.com/najoshi/sickle

<sup>2</sup>http://www.ebi.ac.uk/ena

#### TABLE 1 | Primers used in this study.

fmicb-08-00311 March 2, 2017 Time: 16:34 # 3


TABLE 2 | Correlation of phenotypic resistance to selected antimicrobial agents and presence of specific resistance genes detected in draft genomes of 96 Actinobacillus pleuropneumoniae isolates from the UK.


<sup>a</sup>No isolates were resistant to florefenicol, and no florfenicol resistance genes were detected in the genomes. No specific macrolide resistance genes were identified, and phenotypic resistance to erythromycin, tilmicosin and tylosin was not clear from the MIC values obtained (see Supplementary Table S1 for further details). <sup>b</sup>Specific genes detected for each resistance phenotype: tet(B) or tet(H) for tetracycline, blaROB-1 for ampicillin, sul2 for sulfisoxazole; dfrA14 for trimethoprim, and GyrAS83F for enrofloxacin resistant isolates. <sup>c</sup>Correlation between resistant phenotype and genotype was calculated using Fisher's exact test function ('fisher.test') in R (version 3.3.2). d Initially, two isolates with sul2 identified in their genomes had an MIC ≤ 16 mg/L, however re-test by disk diffusion showed a zone of inhibition of 6 mm, indicative of resistance. Poor solubility of sulfisoxazole may have caused the discrepancy.

Only 33% of the isolates did not have any AMR genes, whereas 20% carried a single, and 47% two or more AMR genes. None of the isolates tested were resistant to florefenicol, and no florfenicol resistance genes were detected in the genomes.

The MICs for ampicillin showed clear separation into low MIC (MIC ≤ 2 mg/L) for isolates with no resistance gene detected, and high MIC (MIC ≥ 8 mg/L) for 18 isolates with blaROB−<sup>1</sup> detected by ResFinder. One isolate (MIDG3354) with an MIC = 4 mg/L also had blaROB−<sup>1</sup> detected by ResFinder. Thus, there was 100% correlation between the presence of blaROB−<sup>1</sup> and an MIC ≥ 4 mg/L for ampicillin. In all cases, the blaROB−<sup>1</sup> gene identified showed 99.9% identity with that found in the plasmid pB1000 (accession number DQ840517) from Haemophilus parasuis (San Millan et al., 2007). For all but one isolate, the full 918 bp blaROB−<sup>1</sup> sequence was identified on a single contig. In MIDG3443, where only 857/918 bp were identified by ResFinder, BLASTn confirmed that the entire blaROB−<sup>1</sup> sequence was present, but split over two contigs.

For tetracycline, 94.8% of isolates considered phenotypically resistant (MIC ≥ 4 mg/L) were found to carry either tet(B) or tet(H). A 100% correlation was found between the presence either of these genes and an MIC ≥ 8 mg/L. We found of four isolates with an MIC = 4 mg/L, only one (MIDG2567) carried a resistance gene, tet(B), whereas three other isolates (MIDG3342, MIDG3352, and MIDG3356) had no detectable tetracycline resistance gene. The tet(B) gene identified by ResFinder showed 99.9% to 100% identity with a 1206 bp gene found in the Shigella flexneri 2a SRL pathogenicity island (accession number AF326777) or the S. flexneri 2b plasmid R100 (accession number AP000342). In some cases, the gene was split over two contigs (with only one part identified by ResFinder), as confirmed by BLASTn. In five isolates, tet(H) genes were identified that had either 99.9% identity with the 1179 bp gene found in Pasteurella aerogenes plasmid pPAT1 (accession number AJ245947) or 100% identity with the 1203 bp sequence found in Pasteurella multocida plasmid pPMT1 (accession number Y15510). These two genes differ at their 3<sup>0</sup> ends, and BLASTn analysis confirmed the presence of the 1203 bp gene in four A. pleuropneumoniae isolates, and the 1179 bp sequence in one.

For trimethoprim, there was a 100% correlation between an MIC ≥ 32 mg/L and the presence of dfrA14 in 16 isolates, as previously described (Bossé et al., 2015b). Isolates lacking dfrA14 had an MIC ≤ 4 mg/L, showing clear separation from the trimethoprim-resistant isolates. There was also a 100% correlation between an MIC ≥ 256 mg/L for sulfisoxazole and the presence of the sul2 gene in 44 isolates (MIC ≥ 512 mg/L for 39 of these). In two isolates, MIDG2652 and MID3362, initial results (MIC ≤ 16 mg/L) indicated susceptibility despite the presence of the sul2 gene. PCR confirmed the presence of sul2 in these isolates, and resistance to sulfisoxazole was confirmed by disk diffusion (zone of inhibition = 6 mm). The sul2 gene identified by ResFinder shared 100% with an 816 bp gene from Acinetobacter fmicb-08-00311 March 2, 2017 Time: 16:34 # 4

bereziniae (accession number GQ421466). In cases where less than 816 bp of the gene were identified by ResFinder, BLASTn confirmed the presence of the remaining sequence on a separate contig.

The six isolates with elevated MICs for enrofloxacin (≥0.25 mg/L) all had identical sequences for their gyrA, gyrB, parA, and parE genes. The amino acid sequence for each conserved protein encoded by these genes was used to search (by tBLASTn) the remaining genomes in order to determine if any mutations could account for the increased resistance to enrofloxacin. Compared to the sequences found in the enrofloxacin-sensitive isolates, no mutations were found in GyrB, ParA or ParC, however, a substitution in the GyrA sequence (S83F) was found only in the six isolates with MICs for enrofloxacin ≥ 0.25 mg/L.

All isolates had an MIC ≤ 16 mg/L for both erythromycin and tilmicosin, whereas there was a clear separation into low (MIC ≤ 2 mg/L) and high (MIC ≥ 32) levels of resistance to tylosin. However, no macrolide resistance genes were identified by ResFinder in any of the isolates, nor were any mutations detected in the 23S rRNA gene, or in the genes encoding ribosomal proteins L4 and L22 (rplD and rplV, respectively).

# DISCUSSION

Recently, several groups have investigated the use of genome sequencing as an alternative/adjunct to phenotypic testing for detection and surveillance of AMR in different bacteria (Zankari et al., 2013; Gordon et al., 2014; Köser et al., 2014; Walker et al., 2015; Zhao et al., 2016). For each bacterium, it is necessary to determine the correlation between genetic detection of a resistance gene with phenotypic determination of resistance in order to validate the usefulness of a wgs approach.

In this study, we investigated 96 clinical isolates of A. pleuropneumoniae, an important respiratory pathogen of pigs, and found a high correlation between the presence of specific AMR genes and elevated MICs for the corresponding antimicrobial agents tested. However, the results of macrolide testing are inconclusive. Although 20 isolates had an MIC ≥ 32 mg/L for tylosin (seven of which had an MIC ≥ 64 mg/L), all isolates had an MIC ≤ 16 mg/L for both erythromycin and tilmicosin, and no macrolide resistance genes were detected by ResFinder. Furthermore, no point mutations known to confer resistance to macrolides (e.g., in rplV, rplD, rumA, or the 23S rRNA genes) were detected by BLASTn in any of the isolates with elevated MICs for tylosin. In other studies, where specific macrolide resistance genes were identified in members of the Pasteurellaceae (Olsen et al., 2015; Dayao et al., 2016), higher MIC values (≥64 mg/L) were reported for tilmicosin and erythromycin. There is no established CLSI clinical MIC breakpoint for tylosin resistance in A. pleuropneumoniae, and a recent VetPath survey (El Garch et al., 2016) also showed that although only 1/158 A. pleuropneumoniae isolates tested had an MIC ≥ 32 mg/L for tilmicosin (considered resistant), 144/158 had an MIC ≥ 32 mg/L for tylosin (none considered resistant due to lack of established breakpoint). From these and our results, it is not clear if there is a specific resistance to tylosin associated with an MIC ≥ 64 mg/L, and if so, what is the mechanism of this resistance.

All six isolates with an MIC ≥ 0.25 mg/L for enrofloxacin had an identical amino acid exchange (S83F) in the GyrA protein, which is in agreement with a previous study in A. pleuropneumoniae (Wang et al., 2010). Wang et al. (2010) also reported that combinations of this and other mutations in gyrA, parC, and parE were associated with higher levels of resistance to enrofloxacin (MIC ≥ 1 mg/L), but these were not detected in our isolates. The six isolates with reduced susceptibility to enrofloxacin were isolated from samples submitted to the same APHA laboratory (Langford), and five were in the same year. Enrofloxacin resistance does not appear to be widespread in A. pleuropneumoniae isolates in the UK.

For trimethoprim resistance, there was a 100% correlation between the presence of dfrA14 and an MIC ≥ 32 mg/L, as previously reported (Bossé et al., 2015b). Although only 16 of the tested isolates were resistant to trimethoprim, all of these also carried the sul2 gene and had an MIC ≥ 512 mg/L for sulfisoxazole. The sul2 gene was detected in a total of 46 isolates, however, phenotypic detection of resistance to sulfisoxazole was problematic in some of these. Although 85% of the isolates carrying sul2 had an MIC ≥ 512 mg/L, and 96% an MIC ≥ 256 mg/L, two isolates initially showed an MIC ≤ 16 mg/L. When re-tested by disk diffusion, these two isolates showed a zone of 6 mm, consistent with a resistant phenotype (Clinical and Laboratory Standards Institute [CLSI], 2013). Issues with the solubility of sulfisoxazole may account for the discrepancies. Co-location of the dfrA14 and sul2 genes in two distinct plasmids (Bossé et al., 2015b) accounts for the 16 isolates with resistance to both trimethoprim and sulfisoxazole. Analysis of the sequences flanking sul2 in the isolates lacking dfrA14 indicate other possible plasmids that will require further investigation. Only two of the isolates carrying sul2 had no other detectable AMR genes. The remaining 44 all had tet(B) or tet(H), with eighteen also carrying blaROB−1, and eight of these with dfrA14 as well. A further eight isolates carried a combination of tet(B), sul2 and dfrA14, but lacked the blaROB−<sup>1</sup> gene (Supplementary Table S1).

There was a 100% correlation between the presence of blaROB−<sup>1</sup> in 19 isolates and an MIC ≥ 4 mg/L for ampicillin; with an MIC ≥ 8 mg/L in 18 of these. The sequences flanking blaROB−<sup>1</sup> indicate possible plasmids. Further investigation is required in order to determine if these represent known plasmid(s) described in other Pasteurellaceae, such as pB1000 (San Millan et al., 2007) or pJMA-1 (Moleres et al., 2015) or unique plasmids not yet described.

In 98% of isolates carrying either tet(B) or tet(H), an MIC ≥ 8 mg/L for tetracycline was detected. One isolate with an MIC = 4 mg/L contained the tet(B) gene, whereas a further three isolates with an MIC = 4 mg/L did not have any tetracycline resistance genes detected. The reason for this discrepancy is not clear. The sequences flanking tet(H) indicate possible plasmid(s), whereas tet(B) was found either on small contigs with sequences related to possible plasmid(s), or on larger contigs with sequences indicating a chromosomal insertion fmicb-08-00311 March 2, 2017 Time: 16:34 # 5

site. In 21 isolates with the latter, we recently identified a 56 kb integrative conjugative element (ICE), ICEApl1, inserted in a tRNA-Leu (TAA) gene, which contains tet(B) as part of a Tn10 insertion (Bossé et al., 2016). In 15 other isolates with tet(B) on large contigs, the sequence appears to part of a Tn7 insertion disrupting the competence related gene, comM (see Supplementary Table S1). The majority (55/96) of isolates tested were resistant to tetracycline, and only 11 of these carried no other detectable AMR gene.

Reproducibility of phenotypic results has been an issue with surveillance of AMR, with calls for a more standardized method to allow direct comparisons between labs (Michael et al., 2015; El Garch et al., 2016). Our results confirm that wgs is a valuable method that can be used, either in combination with phenotypic testing or on its own, for surveillance of AMR in A. pleuropneumoniae. In each isolate where an AMR gene was identified, phenotypic results confirmed resistance. In the case of macrolides, no genes were identified and it is not clear what MIC for tylosin should be considered as a breakpoint indicating resistance. Genome sequencing not only identifies specific AMR genes, but also gives an indication of their locations, either within the chromosome or on plasmids. ResFinder was useful for identifying and locating known AMR gene sequences within the draft genomes, though in cases where a sequence was split over multiple contigs, only the contig with the larger proportion of the gene was identified. ResFinder is also not capable of detecting point mutations in genes such as gyrA that lead to a resistance phenotype. In these cases, BLASTn or tBLASTn can be used to either detect mutations or to determine the location of missing sequences not identified by ResFinder.

Plasmids are the biggest contributors to the spread of AMR, with dissemination occurring not only through clonal expansion of the isolates harboring them, but also by conjugal transfer when genes expressing the required machinery are present. Tracking the spread of AMR genes/plasmids provides important epidemiological information that can best be provided by wgs data. Plasmid sequences are often distributed over multiple small contigs within draft genomes, suggesting either repeat sequences within the plasmid, or the presence of multiple plasmids sharing common backbone sequences. Plasmid isolation and sequencing can determine if a single or multiple small plasmids are present, as has been described for P. multocida (San Millan et al., 2009), especially in the case of isolates resistant to multiple antimicrobial agents. Most small plasmids that have been characterized in A. pleuropneumoniae and other Pasteurellaceae are mobilisable, but do not harbor genes encoding conjugal transfer machinery (Blanco et al., 2007; Matter et al., 2008; Kang et al., 2009; San Millan et al., 2009; Bossé et al., 2015a,b). These plasmids can be disseminated either via clonal expansion or conjugation. The latter mechanism requires the presence of other replicons encoding the required machinery, such as the recently described ICE (ICEApl1, ICEMh1, ICEPmu1), discovered in the genomes of different Pasteurellaceae species (Michael et al., 2012; Eidam et al., 2015; Bossé et al., 2016). As more wgs become available for members of the Pasteurellaceae, it is likely that more ICE will be identified, and their role in dissemination of small plasmids can be further investigated.

In summary, genome sequencing is likely to be used increasingly in the surveillance of AMR, providing a standard method that can be easily compared between different laboratories in different countries. This method can be used on its own for identification of known resistance genes, or in conjunction with sensitivity testing where resistance mechanisms have yet to be identified. Furthermore, wgs data can provide more information than just AMR genotypes, it can also give insights into the mechanisms allowing spread of resistance amongst isolates.

# MEMBERS OF BRaDP1T CONSORTIUM

The BRaDP1T Consortium comprises: Duncan J. Maskell, Alexander W. (Dan) Tucker, Sarah E. Peters, Lucy A. Weinert, Jinhong (Tracy) Wang, Shi-Lu Luan, Roy R. Chaudhuri (University of Cambridge), Andrew N. Rycroft, Gareth A. Maglennon, Jessica Beddow (Royal Veterinary College); Brendan W. Wren, Jon Cuccui, Vanessa S. Terra (London School of Hygiene and Tropical Medicine); and Paul R. Langford, Janine T. Bossé, Yanwen Li (Imperial College London).

# AUTHOR CONTRIBUTIONS

JB, PL, AR, BW, DM, AT conceived the study; JB, YaL, JR, RF, YiL, RRC, and MH produced the data; JB, YaL, RF, RRC analyzed the data; JB, PL wrote the paper.

# FUNDING

This work was supported by a Longer and Larger (LoLa) grant from the Biotechnology and Biological Sciences Research Council (BBSRC grant numbers BB/G020744/1, BB/G019177/1, BB/G019274/1, and BB/G018553/1), the UK Department for Environment, Food and Rural Affairs, and Zoetis (formerly Pfizer Animal Health) awarded to the Bacterial Respiratory Diseases of Pigs-1 Technology (BRaDP1T) consortium. MH was supported by the Wellcome Trust (grant number 098051). JR was funded from the former AHVLA's Research and Development Internal Investment Fund (grant number RD0030c).

# ACKNOWLEDGMENT

The authors wish to thank Susanna Williamson and Chris Teale from the APHA for their advice and input.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb.2017. 00311/full#supplementary-material

# REFERENCES

fmicb-08-00311 March 2, 2017 Time: 16:34 # 6


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Bossé, Li, Rogers, Fernandez Crespo, Li, Chaudhuri, Holden, Maskell, Tucker, Wren, Rycroft, and Langford on behalf of the BRaDP1T Consortium. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Single-Molecule Sequencing (PacBio) of the *Staphylococcus capitis* NRCS-A Clone Reveals the Basis of Multidrug Resistance and Adaptation to the Neonatal Intensive Care Unit Environment

Patrícia Martins Simões 1, 2, 3 \*, Hajar Lemriss <sup>4</sup> , Yann Dumont <sup>5</sup> , Sanâa Lemriss <sup>6</sup> , Jean-Philippe Rasigade2, 3, Sophie Assant-Trouillet 1, 2, Azeddine Ibrahimi <sup>4</sup> , Saâd El Kabbaj <sup>6</sup> , Marine Butin1, 7 and Frédéric Laurent 1, 2, 3

#### *Edited by:*

Benoit Doublet, National Institute for Agricultural Research (INRA), France

### *Reviewed by:*

Olin Silander, Massey University, New Zealand Ravi Ranjan, University of Illinois at Chicago, USA

#### *\*Correspondence:*

Patrícia Martins Simões patricia.martins-simoes01@chu-lyon1.fr

#### *Specialty section:*

This article was submitted to Evolutionary and Genomic Microbiology, a section of the journal Frontiers in Microbiology

*Received:* 07 September 2016 *Accepted:* 28 November 2016 *Published:* 15 December 2016

### *Citation:*

Martins Simões P, Lemriss H, Dumont Y, Lemriss S, Rasigade J-P, Assant-Trouillet S, Ibrahimi A, El Kabbaj S, Butin M and Laurent F (2016) Single-Molecule Sequencing (PacBio) of the Staphylococcus capitis NRCS-A Clone Reveals the Basis of Multidrug Resistance and Adaptation to the Neonatal Intensive Care Unit Environment. Front. Microbiol. 7:1991. doi: 10.3389/fmicb.2016.01991 <sup>1</sup> Department of Clinical Microbiology, Northern Hospital Group, Hospices Civils de Lyon, Lyon, France, <sup>2</sup> International Centre for Research in Infectious Diseases, Institut National de la Santé et de la Recherche Médicale U1111, University of Lyon, Lyon, France, <sup>3</sup> National Reference Center for Staphylococci, Hospices Civils de Lyon, Lyon, France, <sup>4</sup> Biotechnology Laboratory (Medbiotech), Medical and Pharmacy School, University Mohammed V de Rabat, Rabat, Morocco, <sup>5</sup> Department of Clinical Microbiology, Eastern Hospital Group, Hospices Civils de Lyon, Lyon, France, <sup>6</sup> Department of Biosecurity PCL3, Laboratory of Research and Medical Analysis of the Fraternal of Gendarmerie Royale, Rabat, Morocco, <sup>7</sup> Neonatal Intensive Care Unit, Eastern Hospital Group, Hospices Civils de Lyon, Lyon, France

The multi-resistant Staphylococcus capitis clone NRCS-A has recently been described as a major pathogen causing nosocomial, late-onset sepsis (LOS) in preterm neonates worldwide. NRCS-A representatives exhibit an atypical antibiotic resistance profile. Here, the complete closed genome (chromosomal and plasmid sequences) of NRCS-A prototype strain CR01 and the draft genomes of three other clinical NRCS-A strains from Australia, Belgium and the United Kingdom are annotated and compared to available non-NRCS-A S. capitis genomes. Our goal was to delineate the uniqueness of the NRCS-A clone with respect to antibiotic resistance, virulence factors and mobile genetic elements. We identified 6 antimicrobial resistance genes, all carried by mobile genetic elements. Previously described virulence genes present in the NRCS-A genomes are shared with the six non-NRCS-A S. capitis genomes. Overall, 63 genes are specific to the NRCS-A lineage, including 28 genes located in the methicillin-resistance cassette SCCmec. Among the 35 remaining genes, 25 are of unknown function, and 9 correspond to an additional type I restriction modification system (n = 3), a cytosine methylation operon (n = 2), and a cluster of genes related to the biosynthesis of teichoic acids (n = 4). Interestingly, a tenth gene corresponds to a resistance determinant for nisin (nsr gene), a bacteriocin secreted by potential NRCS-A strain niche competitors in the gut microbiota. The genomic characteristics presented here emphasize the contribution of mobile genetic elements to the emergence of multidrug resistance in the S. capitis NRCS-A clone. No NRCS-A-specific known virulence determinant was detected, which does not support a role for virulence as a driving force of NRCS-A emergence in NICUs worldwide. However, the presence of a nisin resistance determinant on the NRCS-A chromosome, but not in other S. capitis strains and most coagulase-negative representatives, might confer a competitive advantage to NRCS-A strains during the early steps of gut colonization in neonates. This suggests that the striking adaptation of NRCS-A to the NICU environment might be related to its specific antimicrobial resistance and also to a possible enhanced ability to challenge competing bacteria in its ecological niche.

Keywords: bacteremia, multiple drug resistance, late-onset sepsis, SMRT, nisin, comparative genomics, *Staphylococcus capitis*

# INTRODUCTION

Coagulase-negative staphylococci (CoNS) are common commensals of the human skin and mucosa and are also opportunistic pathogens responsible for infections associated with indwelling medical devices and bloodstream infections (Otto, 2009). Among CoNSs, Staphylococcus epidermidis is the most prevalent species and is classically involved in nosocomial bacteremia in patients with co-morbidities including very low-birth weight preterm infants (Otto, 2009). However, the Staphylococcus capitis has also been incriminated in sepsis throughout the world in neonatal intensive care units (NICUs) (Van Der Zwet et al., 2002; Venkatesh et al., 2006; D'mello et al., 2008; Rasigade et al., 2012; Boghossian et al., 2013; Cui et al., 2013). We recently reported that a group of closely related multidrug-resistant (MDR) S. capitis strains belonging to the same clone, named NRCS-A, were present in most French NICUs but absent in the other pediatric or adult ICUs (Rasigade et al., 2012). In subsequent work, we demonstrated the dissemination of NRCS-A strains in distant NICUs throughout the world (Butin et al., 2016). This dissemination is of major concern because of the extensive drug resistance of NRCS-A as well as its ability to become highly endemic in some NICUs, representing up to 40% of all bacteremia cases (Stoll et al., 2002, 2005; Rasigade et al., 2012). Due to the presence of the type V-related staphylococcal chromosome cassette mec (SCCmec), all S. capitis NRCS-A strains are resistant to beta-lactams and exhibit reduced susceptibility to other antimicrobial agents in common use in NICUs, including resistance to aminoglycosides and resistance or heteroresistance to vancomycin (Venkatesh et al., 2006; Rasigade et al., 2012). Overall, the worldwide diffusion of a multidrug-resistant S. capitis clone that is highly adapted to NICU antibiotic selective pressure raises questions about the potential genetic support of its epidemiological success.

Few genome sequences of S. capitis are publicly available to date. In addition, only a single study using the genome of a S. capitis isolate collected from a bloodstream infection in an adult has reported the genomic comparison of virulence factors between S. capitis and S. epidermidis (Cameron et al., 2015). Considering the specific importance NRCS-A in the NICU setting and its international dissemination, we present here the first closed genome sequence of the French NRCS-A prototype strain CR01, a comparison of the whole-genome sequences (WGS) of three other NRCS-A strains collected from distant countries (Australia, Belgium, the United Kingdom), and a description of the genomic features of this NRCS-A clone.

# METHODS

## Genome Sequencing and Assembly

The four strains of S. capitis, CR01, CR03, CR04, and CR05, used in this study were part of a previously published collection, with all strains being of the international S. capitis NRCS-A clone (Butin et al., 2016). The strains were isolated from blood cultures of preterm infants diagnosed with neonatal sepsis in Neonatal Intensive Care Units (NICUs) from four different countries: France (isolate CR01), Belgium (isolate CR03), Australia (isolate CR04) and the United Kingdom (isolate CR05) (for further details see Supplementary Data).

Initially, whole-genome pyrosequencing (454 Life Sciences/Roche) was performed for all 4 isolates, as previously reported (Lemriss et al., 2014, 2015). Single-molecule real-time (SMRT) sequencing was then performed for strain CR01 to obtain a closed reference genome for this clone (Supplementary Data).

Sequencing for de novo assembly was performed using PacBio RS II (Menlo Park, CA, USA). High molecular weight DNA was sheared in a Covaris g-TUBE (Covaris, Woburn, MA, USA) to obtain 20 kb fragments. After shearing, the DNA size distribution was checked using a Fragment Analyzer (Advanced Analytical Technologies, Ames, IA, USA), and 5µg of the sheared DNA was used to prepare a SMRTbell library with PacBio SMRTbell Template Prep Kit 1 (Pacific Biosciences, Menlo Park, CA, USA) according to the manufacturer's recommendations. The resulting library was size selected using a BluePippin system (Sage Science, Inc. Beverly, MA, USA) for molecules larger than 11 kb. The recovered library was sequenced using 1 SMRT cell with P6- C4 chemistry and MagBeads with a PacBio RSII system (Pacific Biosciences, Menlo Park, CA, USA) at 240 min movie length.

Based on the SMRT cell sequencing, we generated 50,876 postfilter polymerase reads with an average read length of 15.6 kb and an N50 read length of 21.1 kb. The average coverage was 265X; for de novo assembly, the PacBio module "RS\_HGAP\_Assembly.2" in SMRTpipe version v2.3.0 was used for continuous long reads (CLR) after polishing and error correction with Quiver, as described previously (Chin et al., 2013). DNA methylation was determined using the RS\_Modification\_and\_Motif\_Analysis protocol within SMRT Portal v2.30, with a standardized in silico false positive error of ∼1%. Only motifs with a mean modification quality value (QV) >50 and a mean coverage of >100X were validated as being modified (https:// github.com/PacificBiosciences/Bioinformatics-Training/wiki/

Methylome-Analysis-Technical-Note). Two polished contigs,

one corresponding to the chromosome and the other to the plasmid, were obtained. Circularization of both contigs was achieved by manual comparison and removal of regions of overlaps.

# Computational Analysis

Automatic syntactic and functional annotation of the closed, reference genome of strain CR01 [EMBL accession number: LN866849 (chromosome) and LN866850 (plasmid)] was performed using the MicroScope platform pipeline (Vallenet et al., 2006, 2013). The annotations of the draft genomes of strains CR03 (EMBL accession number: CVUF01000001-CVUF01000031), CR04 (EMBL accession number: CTEM01000001-CTEM01000038) and CR05 (EMBL accession number: CTEO01000001-CTEO01000039) were also performed using the MicroScope platform pipeline, as published previously (Lemriss et al., 2014). Visual inspection and manual curation were carried out using the MaGe platform and BLAST searches against NCBI databases.

Virulence factors were identified using the Virulence Factors database (VFDB; http://www.mgc.ac.cn/VFs/, Chen et al., 2012) and tblastn (using the ncbi-blast-2.2.27+ suite, Camacho et al., 2009) searches against in-house databases of known virulence genes of staphylococci.

Prophage regions were identified using the PHAST server (http://phast.wishartlab.com, Zhou et al., 2011). The predicted results were confirmed using the MaGe platform.

For comparison, we obtained publicly available assemblies for six additional whole-genome sequences of S. capitis: strains QN1 (NCBI Ref. Seq.: NZ\_AJTH00000000.1), VCU116 (NCBI Ref. Seq.: NZ\_AFTX00000000.1), SK14 (NCBI Ref. Seq.: NZ\_ACFR00000000.1), LNZR-1 (NCBI Ref. Seq.: NZ\_JGYJ00000000.1), C87 (NCBI Ref. Seq.: NZ\_ACRH00000000.1), and AYP1020 (NCBI Ref. Seq.: NZ\_CP007601.1). All six genomes were aligned against the closed genome of isolate CR01 using Mauve Progressive v.2.3.1 (Darling et al., 2010) with default settings.

Resistance genes were identified by a combination of ResFinder v.2.1 (Kleinheinz et al., 2014) and tblastn searches using the complete, curated Antibiotic Resistance Proteins database (multifasta file) available at the Comprehensive Antibiotic Resistance Database (CARD) (McArthur et al., 2013). For ResFinder, cutoffs of 80% minimum length and 90% identity were used to search for resistance genes in CR03, CR04, and CR05 assemblies previously performed against the reference CR01 using Mauve Progressive to determine their location and genomic context. For detection of resistance genes using the CARD complete, curated Antibiotic Resistance Proteins database, tblastn searches against a database composed of the four genomes of strains CR01, CR03, CR04, and CR05 were performed.

Genes specific to the NRCS-A clone were identified using the MaGe platform "Gene Phyloprofiler" tool (Vallenet et al., 2013) and the four annotated genomes of NRCS-A and public S. capitis genomes. Briefly, genomes were compared in terms of gene content using pre-computed homologies and synteny groups. Pairwise comparisons between predicted protein sequences of the studied genome and the proteins of another genome allowed computation of ranked hits and determination of bidirectional best hits (BBH) (for each protein, the three best hits were retained). Putative orthologous relationships between two genomes were defined as gene couples satisfying the BBH criterion and an alignment threshold for homology set as a minimum of 50% sequence identity along 80% of the length of the smallest protein. Mauve alignments were also used to complement the genomic context, followed by manual curation of the putative genes exclusively found in only the four NRCS-A genomes.

Insertion sequences (IS) in the genomes were identified using the ISfinder analysis tool (Siguier et al., 2006). Both blastn and blastp queries were performed with default settings. Matches with an e-value smaller than 0.5 were manually curated. To confirm whether a given IS position was conserved in the other genomes analyzed, the closed genome of isolate CR01 was aligned with the three other NRCS-A genomes and with the six public genomes using Mauve Progressive v.2.3.1 (Darling et al., 2010) with default settings. ORFs within immediate proximity of conserved IS loci in the NRCS-A genomes were compared with the S. capitis reference genome to check for potential alterations in coding sequences, using both Mauve Progressive and Mage platform. genomic islands (GIs) were predicted using IslandViewer 3 (Dhillon et al., 2015) and confirmed by alignment with the three NCRS-A genomes and the six public S. capitis genomes using Mauve Progressive. IS insertions within GIs were predicted as mentioned above. NRCS-A-specific GIs were searched in the NCBI database using megablast (Camacho et al., 2009) query, limiting to the Staphylococcus taxon (taxid: 1279). Published GIs from reference genomes were searched against Staphylococcus aureus Mu50 (NCBI Reference Sequence: NC\_002758.2), COL (NCBI Reference Sequence: NC\_002951.2), MW2 (NCBI Reference Sequence: NC\_003923.1), N315 (NCBI Reference Sequence: NC\_002745.2), FPR3757 (NCBI Reference Sequence: NC\_007793.1) and S0385 (NCBI Reference Sequence: NC\_017333.1), and S. epidermidis RP62A (NCBI Reference Sequence: NC\_002976.3) using the Mage platform [15]. Syntonomes of more than 3 ORFs and with identity greater than 33% were considered to be GIs.

# Investigation of *nsr* Gene Presence in Clinical Bacterial Strains

A panel of 15 strains (**Table 1**) comprising clinical strains of S. capitis (NRCS-A and non-NRCS-A) were used for nisin susceptibility testing. All staphylococcal species were originally isolated from blood, and species assignment was determined using Matrix-Assisted Laser Desorption/Ionization Time of Flight Mass Spectrometry (MALDI-TOF Vitek MS, Biomerieux, France). After isolation, the strains were cultured on Horse Blood Agar (Biomerieux) and then stored at −80◦ . Testing for susceptibility to nisin was performed using the disk diffusion test adapted from EUCAST<sup>1</sup> . Briefly, nisin solutions were

<sup>1</sup>European Committee on antimicrobial Susceptibility Testing, 2015, http://www. eucast.org/fileadmin/src/media/PDFs/EUCAST\_files/Disk\_test\_documents/ Manual\_v\_5.0\_EUCAST\_Disk\_Test.pdf

TABLE 1 | Bacterial strains used in the nisin susceptibility test.


SC, Staphylococcus capitis; FR, France; BE, Belgium; UK, United Kingdom; AUS, Australia; South K, South Korea; NO, Norway; NZ, New Zealand; USA, United States of America; DK, Denmark; GR, Greece; Sing, Singapore; NICU, neonatal intensive care unit; ICU, intensive care unit (adults).

prepared using 2.5% nisin powder (Sigma Aldrich, USA) according to a previously described protocol (Piper et al., 2009). Sterile paper disks (Thermofisher Diagnostics, France) were soaked with 30µg of nisin solution. After an overnight incubation at 36◦C on Horse blood Agar (Biomerieux, France), a 0.5 McF was seeded on Mueller Hinton agar (MHE) plates (Biomerieux, France). A disk containing 30µg of nisin was placed in the middle of the plate. The diameter of the inhibition zones was measured after overnight incubation of the inoculated MHE plates at 36◦C. The average diameter of inhibition and SEM were determined from four independent experiments.

PCRs for detecting the nsr gene were performed using consensus primers: pF—5′GGAGATATGGGACCTATGA TTGCA 3′ and pR—5′GCTGTAAkTTCwCCkGAACTkG C 3′ . These primers were designed based on alignment of the nsr gene sequence found in strain CR01 (S. capitis clone NRCS-A) and sequences retrieved from NCBI after a blastn (ref) search for homologous sequences in staphylococcal WGS genomes [S. hyicus ATCC 11249 (acc. num. CP008747), Staphylococcus sp. TE8 (acc. num. NZ\_JMGB00000000.1), S. epidermidis MC28 (acc. num. NZ\_ATCZ00000000.2), S. epidermidis VCU128 (acc. num. NZ\_AHLI00000000.1)]. PCR amplification was conducted in a final volume of 25µL, with 2.0µL bacterial DNA (QuickExtract kit, Qiagen), 2.5 µL 10x Reaction Buffer, 0.75µL 50 mM MgCl2, 1 µL 10 mM forward primer, 1µL 100 mM reverse primer, 4µL 5 mM dNTPs, 1 µL each primer and 0.125µL Taq polymerase (Eurobio, France). The cycling parameters were as follows: (i) initial denaturation for 3 min at 95◦C; (ii) 34 cycles, with 1 cycle consisting of 30 s at 95◦C, 60 s at 60◦C, and 90 s at 72◦C; and (iii) a final extension for 10 min at 72◦C.

# RESULTS

WGSs of four strains belonging to S. capitis clone NRCS-A and originating from NICUs in four distant countries were obtained to investigate lineage-specific virulomes, resistomes, and mobilomes as well as the presence of unshared genes using 454 pyrosequencing technology, as reported previously (Lemriss et al., 2014, 2015). Depending on the isolate, assembly of the draft genomes produced 26 to 39 contigs. Re-sequencing of strain CR01 using SMRT technology (PacBio) allowed for the generation of closed complete sequences for the chromosome and the single plasmid that was subsequently used to produce scaffolds for the three remaining WGSs.

The length of CR01's chromosome is 2,522,871 bp and exhibits a low G + C content (33.02%), as is expected for staphylococci. We identified and annotated 2419 protein-coding regions, 19 rRNAs, 62 tRNAs, 34 ncRNAs (including RNAIII), and 20 transposases associated with insertion sequence elements and transposons. Of note, gaps (identified in the draft genome assemblies of the four NRCS-A isolates (based on 454 pyrosequencing)) frequently occur within the vicinity of transposases, suggesting that some of the 454 short read assemblies were unable to bridge the gaps associated with repetitive genomic elements. These data clearly emphasize the suitability of thirdgeneration sequencing technologies, such as PacBio SMRT, for obtaining fully defined de novo assemblies and, thus, for overcoming the issue of both local and global repeats (Koren and Phillippy, 2015). The genomic characteristics of the NRCS-A isolates tested in this study are summarized in **Table 2**.

# Virulome

The presence of virulence-associated genes in the closed genome of strain CR01 and subsequently the draft genomes of strains CR03 (Be), CR04 (Aus), and CR05 (UK) were inferred by comparison with known virulence factors previously reported for S. aureus, S. epidermidis, and S. haemolyticus using the VFDB database (Chen et al., 2012). The findings were augmented with BLAST searches of well-characterized staphylococcal virulence factors and key regulators (**Table 3**). Comparison of the four NRCS-A genomes with the six published genomes of non-NRCS-A S. capitis, including QN1, VCU116, SK14, LNZR-1, C87, and AYP1020 strains, showed that none of the known staphylococcal virulence genes are exclusively carried by the NRCS-A clone.

Interestingly, the virulome of the NRCS-A S. capitis genome was quite similar to that of S. epidermidis RP62A, including the icaABDCR and capABC biofilm-related operons, Clp proteases and multiple copies of PSM beta type 1b in tandem (75% aa identity to PSMbeta1b of strain RP62) and one PSM alpha (59% aa identity) (Otto, 2012). All PSMs were found in the same genetic environment as in the S. epidermidis RP62A genome, with conservation of upstream and downstream genes.

As previously observed for the fully closed S. epidermidis RP62a and S. epidermidis ATCC 12228 genomes (Zhang et al., 2003; Gill et al., 2005), no S. aureus toxins, except beta- and deltahemolysins, were identified in any of the 10 S. capitis genomes


TABLE 2 | General genomic features of *S. capitis* strain CR01 compared with another three WGS NRCS-A genomes.

<sup>a</sup>Putative plasmid, no definitive data about circularization.

b Insertion sequence isolated in a small contig, indicating probable multiple insertions in the complete genome.

(four NRCS-A and six public non-NRCS-A). Similarly, none of the genes associated with secretion systems (esxA, esxB, esaA, esaB, esaC, essA, essB, and essC) or S. aureus serine proteases (splA, B, C, D, E, and F) are present in the S. capitis genomes.

# Resistome

Several antibiotic resistance-associated genes were identified and correlated with the specific resistance phenotype previously reported for the NRCS-A clone (Rasigade et al., 2012; **Table 4**). All are located on mobile genetic elements (MGEs). Resistance to aminoglycosides is related to the bifunctional aminoglycosidemodifying gene aacA-aphD, which is carried on the transposon Tn4001 (GenBank accession no. AB682805.1; Lyon et al., 1984). Resistance to methicillin is due to the presence of a composite SCCmec-SCCcad/ars/cop cassette exclusively present in the NRCS-A clone, as recently published by our team (Martins Simões et al., 2013). Genomic comparison between the four NRCS-A strains showed that this mobile element is nearly identical: all ORFs are conserved (100% aa homology), and the few nucleotide sequence variations occur only in the CRISPR repeat region, as is expected for these regions. As previously reported (Martins Simões et al., 2013), no other antibiotic resistance genes were detected for this composite SCC element. Finally, the plasmidic bla operon, coding for blaZ beta-lactamase and its regulators, is present in all NRCS-A strains, except strain CR05 (UK).

With regard to other antimicrobial families, for which variability in resistance profiles have been observed among NRCS-A isolates, phenotypic antimicrobial susceptibility testing matched the specific resistance gene contents of each NRCS-A strain. Plasmidic msrA and tetK genes, respectively involved in erythromycin resistance (with a negative D-test) and tetracycline resistance, were only identified in strain CR04 (Australia). In strain CR05 (UK), the gene far1, which is responsible for fusidic acid resistance, is carried by a putative phage.

# Mobile Elements

As mentioned previously, the shortest circular sequence obtained by SMRT for strain CR01 corresponds to a plasmid of 26.140 bp (GC content of 29.25%). It harbors 35 ORFs, including (i) the bla operon coding for penicillinase resistance to penicillins (see above) and (ii) copper resistance-related genes such as copZ, copA, and csoR (copper transcriptional repressor; **Figure 1**). Genome comparison of strain CR01's plasmid with the three other NRCS-A draft genomes revealed that the plasmid is not conserved in all NRCS-A strains (**Figure 1**). In strain CR03 (Belgium), one contig presents 31 of the 35 ORFs identified on the CR01 plasmid. Conversely, in strain CR04 (Australia), only 9 of the 35 ORFs (corresponding to the blaZRI operon and copper resistance operon) were detected in a contig corresponding to a putative plasmid. Finally, no putative plasmid was detected in strain CR05 (UK).

One intact prophage region was identified in all four NRCS-A genomes. This region shows 100% aa identity to the genomes of strains CR01, CR03, and CR04, but it is degenerate in CR05 (**Figure 2**). Moreover, this phage exhibits strong homology to a phage present in strain VCU116 [57.1, 90.2, 76.2% aa homology for regions 1, 2 and 3, respectively (**Figure 2**)] and less than 30% homology with phage Staphy\_StB20\_(NCBI Ref. Seq. NC\_019915), the closest phage found in the NCBI database. Of note, two additional incomplete prophage regions were also detected in strain CR05.

Nineteen distinct insertion sequence (IS) elements, clustering into three distinct families (type IS256, n = 4; type IS1272, n = 5; IS431mec, n = 10), were identified in the CR01 genome. The presence of type IS256 was confirmed in all four NRCS-A strains, with two IS256 elements being associated with the aminoglycosides resistance gene aacA-aphD (′ aac(6′ )-aph(2′′) on transposon Tn4001. Type IS1272 was also found in strains CR03 and CR04 but not in strain CR05, and their genomic locations are not conserved. Two IS431mec cassettes are present in all NRCS-A genomes, as they are carried within the SCCmec-SCCcad/ars/cop element that is specific to and conserved among the NRCS-A lineage (Martins Simões et al., 2013).

Finally, other GIs (n = 5) were predicted for CR01 (**Table 2**) and found in the CR03, CR04, and CR05 strains. However, only one was found exclusively in the NRCS-A lineage. This GI (bp position: 150,780–158,140 in the strain CR01 genome) contains six ORFs: one putative phosphoglycolate phosphatase, one conserved protein of unknown function and three pseudogenes of phage proteins (bp position: 151,470–152,238). Unexpectedly, two of these pseudogenes present 40 and 46.5% aa identity to Listeria monocytogenes phage B025 protein gp27 and the third pseudogene presents 60.6% aa identity to L. monocytogenes phage B025 protein gp28. Both gp27 and gp28 proteins have unknown functions.


TABLE 3 | Comparison of virulence factors in *S. capitis* NRCS-A, non-NRCS-A, and *S. epidermidis* after exclusion of virulence factors present only in *S. aureus*.

The presence of virulence genes is indicated by the "+" sign.

#### TABLE 4 | Genomic profiles of antibiotic resistance for clone NRCS-A strains from four different countries.


All antibiotic resistance genes were found in MGEs. Putative regions in the draft genomes are indicated by an \*.

gene; blaR, regulatory protein BlaR1; blaI, penicillinase repressor; mntH, divalent metal cation transporter MntH, tetK, tetracycline resistance gene.

# Methylome and Restriction Modification Systems

SMRT technology (PacBio) allows for genome-wide detection of modified nucleotides based on the rate at which DNA polymerase incorporates bases during sequencing (Flusberg et al., 2010, Nat. Methods; Roberts et al., 2013). Analysis of polymerase kinetic profiles in strain CR01 identified 1333 methylated positions (Table S2) including 94% (n = 1253) corresponding to adenine methylations (m6A) and 0.006% (n = 8) to cytosine methylations (m4C). The adenine modifications observed correlate with the presence of two adenine restrictionmodification systems (hsdMSR 1 pos: 476,473–482,360 bp;

hsdMSR 2 pos: 686,750–691,866 bp) in the strain CR01 genome, whereas the presence of an mcrBC 5-methylcytosine restriction system (bp position 487,030–492,276) correlates with the cytosine modifications detected. Interestingly, the latter is associated in tandem with the specific additional hdsMSR 1 operon located immediately after the SCCmec-SCCcad/ars/cop element (see above).

# NRCS-A Clone Specific Genes

Comparison of the closed genome of strain CR01 with the draft genomes of the three other NRCS-A genomes and with the six public non-NRCS-A genomes revealed a unique set of 63 genes present exclusively in the NRCS-A clone genome (Table S1). Of these, 28 ORFS are carried by the composite SCCmec-SCCcad/ars/cop mobile element, which is related to the acquisition of methicillin resistance (mecA gene). Interestingly, within this cassette, the CRISPR element is specific to the NRCS-A lineage (Martins Simões et al., 2013). Of note, both the clone-specific cassette (28 genes) and its genetic environment (downstream and upstream regions) are fully conserved in the four NRCS-A strains, which confirmed that the four isolates belong to the same clonal population, even though they were isolated from distant countries/continents. Indeed, it is highly unlikely that the same SCCmec element (100% identity) could be acquired in four independent events in four distant countries.

Among the 35 remaining genes found exclusively in the NRCS-A lineage, 25 are of unknown function, and 10 correspond to the following: (i) an additional type I restriction modification system (hsdMSR, n = 3 genes), (ii) a cytosine methylation operon (mcrBC, n = 2), (iii) a cluster of four genes known to be involved in the biosynthesis of teichoic acids [ispD (2-C-methyl-D-erythritol 4-phosphate cytidylyltransferase), tarJ (ribitol-5 phosphate dehydrogenase), tarK (Glycosyl glycerophosphate), and tagF (CDP-glycerol glycerophosphotransferase), n = 4]. Of note, the tenth gene encodes a 316 aa protein presenting 41% aa homology with the determinant for resistance to nisin (nsr gene) that is present in some Lactococcus lactis strains (Froseth and McKay, 1991; Liu et al., 2005) and which shows protease activity. Contrary to the plasmidic localization of the nsr gene in nisin-resistant L. lactis strains, the nsr gene in the NRCS-A clone is located on the chromosome (position: 521,747– 522,694 bp) immediately upstream from the potassium (K+) transporter kdpEDABC operon conserved in all S. capitis isolates and downstream from the four teichoic acid biosynthesis genes (ispD, tarJ, tarK, and tagF; see above). No IS or transposon-like region was identified near the nsr gene.

# Phenotypic Assay of Clone NRCS-A Nisin Resistance

To assess functional expression of the nisin resistance gene (nsr) found exclusively in the NRCS-A clone, we performed a disk diffusion test with nisin-charged disks using a collection of S. capitis strains, including nine strains belonging to the NRCS-A clone and six nsr-negative, non-NRCS-A strains. The S. capitis strains belonging to the NRCS-A clone presented significantly lower nisin inhibition zones (mean of four independent measurements ± SEM: 8.81 mm ± 1.31) than the strains isolated from adults that do not carry the nsr gene (16.5 ± 1.5; the Student t-test, p < 0.001 vs. 8.8 mm ± 1.3 in NRCS-A). This confirms that the nsr gene found exclusively in S. capitis strains

belonging to the neonatal NRCS-A clone is functional and confers resistance to nisin. Nonetheless, it is known (i) that S. aureus (SA) can modulate its resistance to nisin and other small antibacterial peptides via two-component systems (TCSs) GraSR, NsaSR, BraSR in association with ABC transporters (Howden et al., 2010; Blake et al., 2011; Falord et al., 2011; Kolar et al., 2011; Kawada-Matsuo et al., 2013) and (ii) that SA strains resistant or heteroresistant to vancomycin (VISA and hVISA) usually present increased cell wall thickening (Cui et al., 2003). The same mechanisms have been observed in L. lactis resistant to nisin (Kramer et al., 2006; Shin et al., 2016). Thus, to assess whether the resistance to nisin observed in the NRCS-A strains is due exclusively to expression of the nsr gene and not to a cell wall-thickening adaptation, complementary experiments are required using nsr+/nsr− isogenic isolates. Such analyses are beyond the scope of the present paper.

# DISCUSSION

The NRCS-A clone is of major interest due to its worldwide high dissemination in NICUs and its prevalence as an agent of neonatal sepsis in preterm newborns (Butin et al., 2016). Comparison between the complete genome of the prototype strain belonging to the S. capitis NRCS-A clone (strain CR01, France) with the draft genomes of clinical S. capitis isolates belonging or not to the NRCS-A clone revealed the presence of multiple MGEs mediating the atypical antibiotic resistance profile of this clone. These data confirm (i) the ability of the NRCS-A clone to adapt to the specific selective pressure of the antibiotics used in NICUs and (ii) emphasize the ability of WGS to provide accurate predictions of the resistance phenotypes of staphylococci and to become a promising alternative for culture methods.

Based on genomic comparisons using previously characterized staphylococcal virulence genes, the NRCS-A virulome is highly similar to that of other non-NRCS-A S. capitis as well as S. epidermidis RP62A strains, which suggests that the success of this clone in neonates is likely not due to increased virulence. Nonetheless, we identified the exclusive presence of a gene encoding a functional nisin resistance (nsr) gene in the NRCS-A lineage that is seldom found in staphylococci.

Nisin, a 34 aa antimicrobial peptide produced by a group of gram-positive Lactococcus and Streptococcus species (Shin et al., 2016), is active against a wide range of gram-positive bacteria, including staphylococci. Its mode of action is dual, involving (i) inhibition of cell wall biosynthesis (transglycosylation step), as it binds to and sequesters lipid II from its functional location, and (ii) pore formation (Wiedemann et al., 2001; Peschel and Sahl, 2006; Egan et al., 2016). Nisin has been described as a key player of the gut barrier, and it is widely used as food preservative due to its potent bactericidal activity. Moreover, it has been shown that (i) expression of the nsr gene, identified in Streptococcus agalactiae, in L. lactis not producing nisin induces resistance to nisin and (ii) the presence of human nisin-producing lactic acid bacteria in the gut reduces intestinal colonization by vancomycin-resistant enterococci (Millette et al., 2008).

Taken together, these data and our results strongly suggest that nisin resistance might be responsible for the potential increase in the ability of the NRCS-A clone to establish itself as part of the initial microflora of neonates, in which the Lactococcus genus is one of the dominant bacterial taxa (Park et al., 2005; Morelli, 2008). Expression of the NSR peptidase might enable NRCS-A isolates to establish themselves in the gut of neonates and also to colonize it and survive longer than nisin-susceptible bacteria. This may explain the over-representation of these isolates in sepsis processes, either through direct translocation in blood (Taft et al., 2015) or via colonization of indwelling devices. Although it remains unknown how S. capitis NRCS-A strains are able to translocate into the bloodstream of infants, it is hypothesized that the entry point might be the digestive tract, which is immature in very preterm infants (Taft et al., 2015). Further studies are needed to fully characterize the digestive microflora of very low-weight preterm-infants and its potential impact on the selection of and the fitness advantage to NRCS-A isolates.

# AUTHOR CONTRIBUTIONS

PM: work, study design, data analysis, and manuscript preparation. HL: data analysis and manuscript preparation. YD: data analysis, work, and manuscript preparation. SL: work and manuscript preparation. JR, SA, AI, and SE: manuscript preparation. MB and FL: study design and manuscript preparation.

# FUNDING

The present work was financed by the French Ministry of Health and the French Institute for Public Health Surveillance (INVS) - Santé publique France in the framework of the National Reference Center of Staphylococci and by the grant ING20111223510 from the Fondation pour la Recherche Medical (FRM).

# ACKNOWLEDGMENTS

We would like to thank Michele Bes and Helene Meugnier at the National Reference Center for Staphylococci in Lyon (France), and our colleagues at the National Reference Center of Staphylococci in Belgium (Olivier Denis), in the United Kingdom (Angela Kearns) and in Australia (Margaret Deighton).

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2016.01991/full#supplementary-material

# REFERENCES


strain with heteroresistance to vancomycin in a neonatal intensive care unit. J. Clin. Microbiol. 40, 2520–2525. doi: 10.1128/JCM.40.7.2520-2525.2002


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Martins Simões, Lemriss, Dumont, Lemriss, Rasigade, Assant-Trouillet, Ibrahimi, El Kabbaj, Butin and Laurent. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Integron Digestive Carriage in Human and Cattle: A "One Health" Cultivation-Independent Approach

Delphine Chainier<sup>1</sup>‡ , Olivier Barraud<sup>1</sup>‡ , Geoffrey Masson<sup>1</sup>† , Elodie Couve-Deacon<sup>1</sup> , Bruno François1,2, Claude-Yves Couquet<sup>3</sup> and Marie-Cécile Ploy<sup>1</sup> \*

1 INSERM, CHU Limoges, UMR 1092, Université Limoges, Limoges, France, <sup>2</sup> INSERM, CIC1435, CHU Limoges, Limoges, France, <sup>3</sup> Laboratoire Départemental d'Analyses et de Recherches de la Haute-Vienne, Limoges, France

#### Edited by:

Benoit Doublet, National Institute for Agricultural Research, France

Reviewed by: Xian-Zhi Li, Health Canada, Canada

Krassimira Hristova, Marquette University, United States

> \*Correspondence: Marie-Cécile Ploy marie-cecile.ploy@unilim.fr

#### †Present address:

Geoffrey Masson, LBM de Saint-Benoît, Saint-Benoît, Ile de La Réunion, France

‡These authors have contributed equally to this work.

#### Specialty section:

This article was submitted to Antimicrobials, Resistance and Chemotherapy, a section of the journal Frontiers in Microbiology

Received: 03 May 2017 Accepted: 15 September 2017 Published: 27 September 2017

#### Citation:

Chainier D, Barraud O, Masson G, Couve-Deacon E, François B, Couquet C-Y and Ploy M-C (2017) Integron Digestive Carriage in Human and Cattle: A "One Health" Cultivation-Independent Approach. Front. Microbiol. 8:1891. doi: 10.3389/fmicb.2017.01891 Objectives: Dissemination of antimicrobial resistance (AMR) is a global issue that requires the adoption of a "One-Health" approach promoting integration of human and animal health. Besides culture-dependent techniques frequently used for AMR surveillance, cultivation-independent methods can give additional insights into the diversity and reservoir of AMR genetic determinants. Integrons are molecular markers that can provide overall and reliable estimation of AMR dissemination. In this study, considering the "One-Health" approach, we have analyzed the integron digestive carriage from stools of humans and cattle living in a same area and exposed to different antibiotic selection pressures.

Methods: Three collections of human [general population (GP) and intensive care unit patients (ICUs)] and bovine (BOV) stool samples were analyzed. The three main classes of integrons were detected using a multiplex qPCR both from total DNA extracted from stools, and from Gram-negative bacteria obtained by culture after an enrichment step.

Results: With the cultivation-independent approach, integron carriage was 43.8, 52.7, and 65.6% for GP, ICU, and BOV respectively, percentages being at least twofold higher to those obtained with the cultivation-dependent approach. Class 1 integrons were the most prevalent; class 2 integrons seemed more associated to cattle than to humans; no class 3 integron was detected. The integron carriage was not significantly different between GP and ICU populations according to the antibiotic consumption, whatever the approach.

Conclusion: The cultivation-independent approach constitutes a complementary exploratory method to investigate the integron digestive carriage of humans and bovines, notably within subjects under antibiotic treatment. The high frequency of carriage of integrons in the gut is of clinical significance, integrons being able to easily acquire and exchange resistant genes under antibiotic selective pressure and so leading to the dissemination of resistant bacteria.

Keywords: integrons, digestive carriage, antimicrobial resistance, humans, bovine, One Health

# INTRODUCTION

fmicb-08-01891 September 25, 2017 Time: 10:59 # 2

Antimicrobial resistance (AMR) has increased to a dramatic extend and is a public health threat (Roca et al., 2015). This is particular true for Gram-negative bacteria (GNB) with the expansion of extended-spectrum beta-lactamase (ESBL) or carbapenemase-producing Enterobacteriaceae and with the emergence of new resistance mechanisms, as plasmid-mediated colistin resistance (Meletis, 2016). AMR is a global issue encompassing human, animal, and the environment. Efforts to tackle AMR dissemination thus require the adoption of a "One-Health" approach that promotes integration of public health and veterinary disease, food, and environmental surveillance (Sikkema and Koopmans, 2016). Culture-dependent techniques are frequently used for AMR surveillance in GNB. However, given the role of horizontal gene transfer in AMR dissemination, cultivation-independent methods give additional insight into the diversity and reservoir of AMR determinants.

Several studies indicated that integrons are molecular markers that can provide overall and reliable estimation of AMR dissemination (Amos et al., 2015; Gillings et al., 2015). Integrons are bacterial genetic elements capable of promoting the acquisition and expression of antibiotic resistance genes embedded within gene cassettes (Cambray et al., 2010). Three main classes of integrons are involved in the spread of AMR and several studies have underlined the link between integrons and multidrug resistance in GNB (Leverstein-van Hall et al., 2003; Barraud et al., 2014). In humans and animals, class 1 integrons are the most prevalent, followed by class 2 integrons (Barlow et al., 2004; Cocchi et al., 2007; Vinue et al., 2008; Barraud et al., 2014), whereas very few class 3 integrons are found, this class being mostly associated with environmental samples (Simo Tchuinte et al., 2016).

Nowadays, it is impossible to separate the animal and human resistance issues as genetic transfers exist between these two entities. In this study, considering a "One-Health" approach and the integron biomarker, we have taken benefit from a large collection of stools to analyze the digestive carriage of integrons in humans and bovines, in a cattle-rearing area. This study highlights the added-value of using two complementary approaches for detecting integrons, a cultivation-independent one performed directly from stools and a cultivation-dependent one from GNB and emphasizes the importance of the gut as a reservoir of genetic elements encoding antibiotic resistance.

# MATERIALS AND METHODS

# Stool Samples

A collection of 567 stool samples from human and bovines has been previously established by our team. These samples were consecutively collected from 2006 to 2010 and stored at −80◦C (human collection number DC-2008-604). All subjects originated from the same cattle-rearing area located in the southwestern France (Haute-Vienne, 5.520 km<sup>2</sup> ). The stools were obtained from (i) subjects from the general population working in non-clinical units of Limoges hospital and attending industrial medicine clinics ('GP,' n = 194), (ii) patients hospitalized in the intensive care unit of the hospital ('ICU,' n = 245), (iii) bovines, mainly cows, from 55 different farms ('BOV,' n = 128). Only one stool sample per subject was collected. For the two human stool collections, on-going antibiotic therapies and previous antibiotic therapies during the last 3 months were reported; for bovines, no data regarding antibiotic consumptions were available.

# Bacterial Culture

Two hundred milligrams of stools were inoculated into 5 ml of Brain Heart Infusion broth (AES Laboratory, France) at 37◦C overnight without agitation. The day after, 10 µl were inoculated onto Drigalski agar plates (bioMérieux, France) incubated for 18–24 h at 37◦C. GNB with different morphotypes were selected for identification (ID) and antibiotic susceptibility testing (AST). ID was performed using the Vitek 2 system (Vitek2-ID-GNB, bioMérieux, France). AST was determined using the disk diffusion method according to the recommendations of the French Society for Microbiology with the 21 following drugs: amoxicillin (AMX), amoxicillin clavulanic acid (AMC), ticarcillin (TIC), ticarcillin clavulanic acid (TCC), cefotaxim (CTX) ceftazidim (CAZ), imipenem (IPM), kanamycin (K), gentamicin (GM), tobramycin (TM), amikacin (AN), streptomycin (S), spectinomycin (SPT), nalidixic acid (NA), pefloxacin (PEF), ciprofloxacin (CIP), sulfamethoxazole (SSS), trimethoprim (TMP), trimethoprim-sulfamethoxazole (SXT), chloramphenicol (C), and tetracycline (TET). Results categorized intermediate were reported as resistant in the results section.

# Integron Detection

Total DNA was extracted from 200 mg of stool samples using the QIAamp DNA stool mini kit (Qiagen, Courtaboeuf, France) following manufacturer's instructions.

Class 1, 2, and 3 integrons were detected both from stools and from GNB using a multiplex qPCR, as previously described (Barraud et al., 2010). For all qPCR runs, plasmids pBAD18::intI1, pGEM <sup>R</sup> -T Easy::intI2 and pBAD18::intI3 as positive controls. So as to check for PCR inhibitors, a qPCR targeting the 16S DNA gene was performed (Relman, 1993). All qPCR assays were performed on the MX3005P system (Agilent <sup>R</sup> ).

For each stool sample, all isolates belonging to a same species but with different AST and/or integron profiles were considered as different strains.

# Statistical Analysis

Mc Nemar Chi<sup>2</sup> test for paired data was used to evaluate the proportions of integrons (all integrons, class 1 integrons only, class 2 integrons only and class 1 and 2) detected with the cultivation-dependent approach and the cultivationindependent approach. The association between previous or ongoing antibiotherapy and integron content was tested using Chi<sup>2</sup> test or Fisher exact test according to the theoretical effectives.

Link between integron content and AMR phenotypes was analyzed using the Chi<sup>2</sup> test and the Fisher's exact test for group smaller than 5, as well as to compare integron carriage results of different groups.

# Ethics Statement

fmicb-08-01891 September 25, 2017 Time: 10:59 # 3

For this study, an ethics review process was not needed as stool samples were obtained through standard care. The stools samples from GP patients working in non-clinical units of Limoges hospital and attending industrial medicine clinics were sampled for detection of pathogens like Salmonella and stool samples from ICU patients were sampled for digestive screening of multidrug-resistant bacteria carriage. All subjects were informed that their stool samples could be used for a clinical research purpose and they were able to express their opposition. We declared the human stool collections to the French "Ministère de l'Enseignement et de la Recherche" under CODECOH number DC-2008-604. Moreover, this study was approved by the scientific committee of the sponsor, Limoges hospital.

# RESULTS

# Human and Animal Populations

Characteristics of the three populations (number of subjects, sex ratio, mean age and on-going or previous antibiotic therapy) are described in Supplementary Table S1. Antibiotic consumption was predominant in ICU patients.

# Bacterial Culture

Culture was negative for 30% of ICU stools, whereas only for 2% of GP and BOV. A total of 1012 different GNB were isolated from the 488 stools with a positive culture (Supplementary Table S2). The average number of GNB per stool was 1.9, 1.4, and 2.3 for GP, ICU, and BOV, respectively. Escherichia coli was the most frequent species: 67.1% of all GNB (81.1, 55.5 and 63.0% for GP, ICU, and BOV respectively). Other major species were Citrobacter freundii (n = 10) and Klebsiella pneumoniae (n = 9) for GP, K. pneumoniae (n = 32) followed by Pseudomonas aeruginosa (n = 23) and K. oxytoca (n = 22) for ICU, E. fergusonii (n = 25) and Acinetobacter lwoffi (n = 22) for BOV.

# GNB and Integron Content

Prevalence of integrons was 16.2, 23.8, and 14.2% among GNB isolated from GP, ICU, and BOV respectively, with mainly class 1 integrons (Supplementary Table S2). E. coli was the main species carrying integrons in the three populations with respectively 57, 52, and 36 integron-positive isolates in GP, ICU and BOV. In GP and BOV, respectively 3 and 6 GNB other than E. coli were integron-positive: Proteus mirabilis (3), Proteus vulgaris (1), Providencia rettgeri (1), Enterobacter cloacae (1), Citrobacter braakii (1), Salmonella Typhimurium (1) and Acinetobacter lwoffi (1). In ICU, 32 other GNB were integron-positive, mainly K. pneumoniae (n = 11), K. oxytoca (n = 8), and P. aeruginosa (n = 5).

Analysis of the link between integrons and antibiotic resistance was determined for E. coli. We observed a gradual increase in the percentage of integron-positive E. coli correlated with the number of acquired resistances (Supplementary Table S3). Globally, the percentage of E. coli with resistance to at least two antibiotics was higher among integron-positive isolates than among integron-negative isolates (95.3% versus 23.9%; p < 0.001). Moreover, a statistically significant link was observed between integron and acquired resistance to most antibiotics: penicillins, penicillins and inhibitors, almost all aminoglycosides (except amikacin), quinolones, sulfamethoxazole, trimethoprim, tetracycline, and chloramphenicol (Supplementary Table S4). In all three populations, negative predictive values (NPVs) of integron carriage for antibiotic resistance were up to 97% for third generation cephalosporins, up to 98% for aminoglycosides used in therapeutics (i.e., gentamicin, amikacin, and tobramycin), up to 93% for fluoroquinolones and up to 92% for SXT.

# Digestive Carriage of Integrons

With the cultivation-dependent approach based on integron detection from GNB isolates (number of patients with at least 1 GNB carrying an integron), the integron carriage of the subjects was 23.2, 26.9% for GP and ICU patients respectively and 24.2% for bovines (**Table 1**). With the cultivation-independent approach (number of patients with integron detection by qPCR directly from stools), the integron carriage was at least twofold higher in the three populations: 43.8, 52.7, and 65.6% for GP, ICU, and BOV respectively (**Table 1**). These differences between the two approaches were statistically significant for the three populations (p < 0.01). In humans (GP/ICU), integron carriage was mainly due to class 1 integrons, class 2 integrons being present in only 5–6% of the subjects. In bovines, integron carriage was due to both class 1 and class 2 integrons, 57.1 and 39.1% respectively. No class 3 integron was detected. We did not find any significant difference in integron carriage along the 4 years of the study (data not shown).

# Concordance between the Two Approaches

Results between the cultivation-dependent and the cultivationindependent approaches were compared for all the 567 samples. Results were concordant for 383 samples (67.5%) with 255 being integron-negative and 128 integron-positive with the two approaches. For the remaining samples, the cultivationindependent approach allowed the detection of integrons in 170 samples for whom the culture did not find integronpositive isolates. On the other hand, for 14 patients, we detected integrons with the cultivation-dependent approach and not with the cultivation-independent approach.

# Integron Carriage and Antibiotic Selection Pressure

Differences of integron carriage between the two human populations (GP and ICU) were not significant whatever the approach (p = 0.37 with data from the cultivation-dependent approach; p = 0.065 with data from the cultivation-independent approach). The integron carriage was also not significantly different between GP and ICU populations according to the antibiotic consumption, whatever the approach (**Tables 2A,B**).

#### TABLE 1 | Digestive carriage of integrons.

fmicb-08-01891 September 25, 2017 Time: 10:59 # 4


Numbers indicate subjects being integron carriers among each population. Class 3 integrons were not detected. <sup>∗</sup>Test used was the Mc Nemar Chi<sup>2</sup> test for paired data.

TABLE 2 | Integron digestive carriage and antibiotic therapy with data from the cultivation-dependent approach.


Numbers indicate subjects being integron-positive or integron-negative carriers. <sup>∗</sup>Patients with no data for antibiotic therapy were excluded. Test used was the Chi<sup>2</sup> test# or the Fisher's exact test¤.

# DISCUSSION

In this study based on a "One Health" approach, we have retrospectively taken benefit from a stool collection from human

and bovine subjects living in a cattle-rearing area in France, to check for their digestive carriage of integrons. Indeed, integrons are now well-recognized as genetic biomarkers of acquired resistance (Leverstein-van Hall et al., 2003; Barraud et al., 2014) and they can provide a reliable estimation of AMR dissemination (Amos et al., 2015; Gillings et al., 2015). The "One Health" approach is now required to tackle AMR, human and animal health being interconnected, and international declarations on AMR pay particular attention to this aspect in action plans. In our work, the integron carriage was studied with both

methods: a cultivation-dependent approach based on GNB, as classically determined (Skurnik et al., 2005, 2008; Vasilakopoulou et al., 2009; Bailey et al., 2010; Kheiri and Akhtari, 2016) and a cultivation-independent approach with a qPCR directly applied from stool DNA. This cultivation-independent approach is novel and complementary to the classical approach. To our knowledge, such an approach was previously used only once for determination of the integron carriage in cattle, not directly from stools but from an enrichment broth (Barlow et al., 2004). Even though their methodology was slightly different than ours, the authors observed integron carriage rates from 50% for class 1 integrons and 28% for class 2 integrons with the cultivationdependent approach to 86 and 94% using a direct detection of integrons from stool DNA after an enrichment step. In our study, we also observed an increase of the integron carriage with the cultivation-independent approach, with at least a twofold increase (**Table 1**) and the cultivation-independent approach allowed detecting 170 integron carriers that were considered as non-carriers with the cultivation-dependent approach. This increase was expected. First, some bacteria may have been not collected during the culture process: (i) GNB colonies were visually selected and some strains may have been missed, notably if morphotypes were similar; (ii) culture of one integron-positive GNB could have been not observed due to competition with other more predominant GNB; (iii) not all GNB were able to grow on Drigalski agar plates, like Campylobacter or Pasteurella (Lee et al.,

2002; Kehrenberg and Schwarz, 2011). Moreover, it has been shown that Gram-positive bacteria, especially Corynebacteria, may contain integrons (Nesvera et al., 1998; Nandi et al., 2004; Barraud et al., 2011). However, our method using Drigalski agar plates only allowed isolating GNB. Using the cultivationindependent method, these Gram-positive bacteria containing integrons can be detected. Lastly, bacteria from patients under antibiotics may have been unable to grow. This is in agreement with the fact that in ICU patients exposed to a high antibiotic selective pressure, culture was negative for 30% of stools. Lastly, we know that culturing allows detection of a very limited part of all the species present in the gut microbiota (Harmsen and de Goffau, 2016) and we can expect that non-cultivable bacteria can also contain integron. All these explanations argue that the cultivation-independent approach would constitute a complementary approach for the determination of the digestive carriage of integrons than cultivation-dependent approaches, notably when an antibiotic treatment is ongoing. The benefit of cultivation-independent approaches for the diagnostic of antibiotic resistance has been previously shown (Decousser et al., 2017; Dunne et al., 2017). However, for 14 subjects, a signal was positive via the cultivation-dependent approach but negative via the cultivation-independent approach. Hypothesis of inhibitors from stools was excluded thanks to correct amplification of the 16S-DNA from all the samples. For these patients, the number of integron-positive strains was probably very low, under the threshold of detection by qPCR. The culture method with an enrichment broth allowed to isolate these bacteria. This confirms that both cultivation-dependent and cultivation-independent approach are complementary.

For humans, independently of the method used, integron carriage was mainly due to class 1 integrons (79 out of 85 integron carriers for GP, and 122 out of 129 for ICU), class 2 being present in only 11% of the integron carriers. This is in agreement with previous studies on integron detection in commensal E. coli (Bailey et al., 2010; Copur-Cicek et al., 2014; Kheiri and Akhtari, 2016) or in clinical isolates (Barraud et al., 2014; Malek et al., 2015; Xu et al., 2017). For bovines, results were different from humans, notably when looking at the cultivationindependent approach. Even if class 1 integrons remained the most prevalent (73 out of the 84 carriers), class 2 integrons were detected among 50 out of the 84 bovine integron carriers. These results were not obtained when looking at the culture with only 2 out of the 36 bovine carriers being positive for class 2 integrons. This discrepancy was also retrieved by Barlow et al. (2004) who obtained 94% of cattle fecal samples positive for class 2 integrons from enrichment broths versus 28% from GNB. Other studies indicated that class 2 integrons seemed more related to animals than to humans (Lapierre et al., 2008; Barlow et al., 2009). One can hypothesize that bacteria from the bovine microbiota constitute a yet unknown reservoir for class 2 integrons. Animal and human gut constitute a huge reservoir of resistances genes (Fitzpatrick and Walsh, 2016). If animals are reservoirs for class 2 integrons, one can expect that presence of such a class within humans may be related to contacts with animals or via the food chain. It is noteworthy that no class 3 integron was detected even using the cultivationindependent approach. It has been previously hypothesized that class 3 integrons are mainly distributed in the environment (Stalder et al., 2014; Simo Tchuinte et al., 2016), and exceptionally within humans or animals.

Rates of integron carriage in human obtained with the cultivation-dependent approach are similar to data found in previous studies, between 15 and 30% (Leverstein-Van Hall et al., 2002; Vinue et al., 2008; Vasilakopoulou et al., 2009). In previous studies, antibiotic therapy was identified as a risk factor for integron acquisition, both in humans (Nijssen et al., 2005; Skurnik et al., 2005; van der Veen et al., 2009) and in animals (Barlow et al., 2009; Wu et al., 2011). Skurnik et al. (2005, 2008) first reported different E. coli integron carriage in animals according to their degrees of exposure to humans. Authors observed a gradual increase in the prevalence of class 1 integrons correlated with the increase in exposure to human. In our study, antibiotic selection pressure was very different between GP and ICU with 8% of GP subjects exposed to antibiotics versus 83% of ICU patients. However, in our study, we did not find statistical difference in integron carriage between GP and ICU, whatever the approach.

Escherichia coli was the predominant species in all three populations. A significant link was observed in E. coli between presence of integrons and acquired resistance to at least two antibiotics (p < 0.001). This was demonstrated previously whatever the definition of the multiresistance (Martinez-Freijo et al., 1998; Leverstein-van Hall et al., 2003; Shahi and Kumar, 2015; Ochoa et al., 2016). Here, we observed that when an integron was present in an E. coli isolate, the most frequent phenotype was the association of a low level penicillinase conferring resistance to amoxicillin and ticarcillin with coresistances to trimethoprim, sulfamethoxazole and tetracycline (data not shown). As we showed recently (Barraud et al., 2014), integrons alone clearly cannot explain the resistance to all of the tested antibiotics, but integrons are usually carried by plasmids or transposons, which can also host other resistance genes. It has been shown in humans and animals that E. coli can harbor plasmids with different genes encoding these resistances (to ampicillin, tetracycline, trimethoprim, and sulfonamides) and that some of these plasmids often carry integrons (Kirchner et al., 2014; Jackson et al., 2015). A statistically positive significant link was also observed between integrons and each antibiotic, except for carbapenems, third generation cephalosporins and amikacin. Finally, NPVs – i.e., when an integron is absent, the strain has a high probability to be susceptible – were >97% for third generation cephalosporins, up to 98% for aminoglycosides used in therapeutics, up to 93% for fluoroquinolones and up to 92% for SXT. This confirms our previous data aiming at considering integrons as relevant biomarkers of antibiotic resistance with a good NPV (Barraud et al., 2014). On the other hand, most positive predictive values (PPV) – i.e., when an integron is present, the strain has a high probability to be resistant – were below 95% or even lower for all the antibiotics (Supplementary Table S3). This may be explained by the low prevalence of multidrug resistant enterobacteria in our county.

Our study aimed at determining the digestive carriage of integrons among human and bovine populations. We showed that the cultivation-independent approach increased significantly, by twofold, the rates of integron carriers compared to the cultivation-dependent approach. It is likely that noncultivable bacteria of the gut harbor integrons. This high frequency of carriage of integrons in the gut is of clinical significance. Indeed, integrons are known to be very efficient genetic elements for capture, expression and dissemination of antibiotic resistance gene cassettes and they can harbor a wide diversity of gene cassettes encoding resistance to almost all antibiotic families. Antibiotics are known to alter the gut microbiota and lead to selection and emergence of resistant bacteria (Ianiro et al., 2016). Furthermore, it is known that antibiotics induce the mechanism of acquisition of antibiotic resistance gene cassettes by integrons (Guerin et al., 2009). Our findings thus highlight that the gut is a reservoir for integrons that can easily acquire and exchange resistant genes under antibiotic selective pressure leading to the dissemination of resistant bacteria.

Here, we showed that both cultivation-dependent and independent methods are complementary for the detection of integrons. However, although the cultivation-independent approach gives new insights about the epidemiology of integrons, we advocate for the maintenance of the cultivation-dependent approach that is the sole approach able to provide, at the species level, the link between integrons and antibiotic resistance. This study also highlights that the gut, in both humans and animals, is a high potential reservoir of genetic determinants for antibiotic resistance, that can be further acquired by pathogens.

# REFERENCES


# AUTHOR CONTRIBUTIONS

OB performed analysis, wrote the manuscript; DC performed analysis, wrote the manuscript; GM performed analysis; EC-D reviewed the manuscript; BF enrolled ICU patients, reviewed the manuscript; C-YC collected animal stool samples; M-CP designed the study, wrote and reviewed the manuscript.

# FUNDING

This work was supported by grants from Ministère de l'Enseignement Supérieur et de la Recherche, Institut National de la Santé et de la Recherche Médicale (INSERM) and Direction de la Recherche et Innovation du CHU de Limoges (APIHA project).

# ACKNOWLEDGMENTS

Authors would like to thank Cécile Duchiron for her help in collecting human clinical data, Marie-Jeanne Cornuejols for her help in collecting bovine samples and Anaïs Labrunie for her precious help in the statistical analysis.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2017.01891/full#supplementary-material

of animal and human origin. FEMS Immunol. Med. Microbiol. 50, 126–132. doi: 10.1111/j.1574-695X.2007.00242.x


Escherichia coli isolated from healthy companion animals. Zoonoses Public Health 62, 479–488. doi: 10.1111/zph.12178


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Chainier, Barraud, Masson, Couve-Deacon, François, Couquet and Ploy. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# The Resistome of Farmed Fish Feces Contributes to the Enrichment of Antibiotic Resistance Genes in Sediments below Baltic Sea Fish Farms

Windi I. Muziasari <sup>1</sup> \*, Leena K. Pitkänen<sup>1</sup> , Henning Sørum<sup>2</sup> , Robert D. Stedtfeld<sup>3</sup> , James M. Tiedje<sup>3</sup> and Marko Virta<sup>1</sup> \*

<sup>1</sup> Department of Food and Environmental Sciences, University of Helsinki, Helsinki, Finland, <sup>2</sup> Department of Food Safety and Infection Biology, Norwegian University of Life Sciences, Oslo, Norway, <sup>3</sup> Department of Civil and Environmental Engineering, Center for Microbial Ecology, Michigan State University, Michigan, MI, USA

#### Edited by:

Axel Cloeckaert, French National Institute for Agricultural Research (Institut National de la Recherche Agronomique), France

#### Reviewed by:

Krassimira Hristova, Marquette University, USA Felipe C. Cabello, New York Medical College, USA

#### \*Correspondence:

Windi I. Muziasari windi.muziasari@helsinki.fi Marko Virta marko.virta@helsinki.fi

#### Specialty section:

This article was submitted to Antimicrobials, Resistance and Chemotherapy, a section of the journal Frontiers in Microbiology

Received: 07 September 2016 Accepted: 19 December 2016 Published: 06 January 2017

#### Citation:

Muziasari WI, Pitkänen LK, Sørum H, Stedtfeld RD, Tiedje JM and Virta M (2017) The Resistome of Farmed Fish Feces Contributes to the Enrichment of Antibiotic Resistance Genes in Sediments below Baltic Sea Fish Farms. Front. Microbiol. 7:2137. doi: 10.3389/fmicb.2016.02137 Our previous studies showed that particular antibiotic resistance genes (ARGs) were enriched locally in sediments below fish farms in the Northern Baltic Sea, Finland, even when the selection pressure from antibiotics was negligible. We assumed that a constant influx of farmed fish feces could be the plausible source of the ARGs enriched in the farm sediments. In the present study, we analyzed the composition of the antibiotic resistome from the intestinal contents of 20 fish from the Baltic Sea farms. We used a high-throughput method, WaferGen qPCR array with 364 primer sets to detect and quantify ARGs, mobile genetic elements (MGE), and the 16S rRNA gene. Despite a considerably wide selection of qPCR primer sets, only 28 genes were detected in the intestinal contents. The detected genes were ARGs encoding resistance to sulfonamide (sul1), trimethoprim (dfrA1), tetracycline [tet(32), tetM, tetO, tetW], aminoglycoside (aadA1, aadA2), chloramphenicol (catA1), and efflux-pumps resistance genes (emrB, matA, mefA, msrA). The detected genes also included class 1 integron-associated genes (intI1, qacE11) and transposases (tnpA). Importantly, most of the detected genes were the same genes enriched in the farm sediments. This preliminary study suggests that feces from farmed fish contribute to the ARG enrichment in farm sediments despite the lack of contemporaneous antibiotic treatments at the farms. We observed that the intestinal contents of individual farmed fish had their own resistome compositions. Our result also showed that the total relative abundances of transposases and tet genes were significantly correlated (p = 0.001, R <sup>2</sup> = 0.71). In addition, we analyzed the mucosal skin and gill filament resistomes of the farmed fish but only one multidrug-efflux resistance gene (emrB) was detected. To our knowledge, this is the first study reporting the resistome of farmed fish using a culture-independent method. Determining the possible sources of ARGs, especially mobilized ARGs, is essential for controlling the occurrence and spread of ARGs at fish farming facilities and for lowering the risk of ARG spread from the farms to surrounding environments.

Keywords: rainbow trout, whitefish, fish intestinal content, class 1 integrons, transposons, mobile genetic elements, qPCR array, culture-independent method

# INTRODUCTION

Fish farms have been suggested as one reservoir of antibiotic resistance genes (ARGs) in the environment due to the prophylactic and therapeutic use of antibiotics (Cabello et al., 2013; Miranda et al., 2013). Because the occurrence of ARGs at fish farms leads to inefficiencies in antibiotic treatments and there is a potential risk of ARG spread from the farms to the surrounding environments, determining the sources of ARGs at fish farms is essential.

The use of antibiotics in fish farming in Finland is controlled by the Finnish Food Safety Authority (EVIRA) and requires a guideline from veterinary professionals. Oxytetracycline, a combination of sulfonamide-trimethoprim, and florfenicol are the antibiotics authorized for use in Finnish fish farms (EVIRA, 2015a,b). From 2001 to 2014, ∼2.3 metric tons of sulfonamide, 0.6 metric ton of trimethoprim, 1.2 metric tons of oxytetracycline, and 0.04 metric ton of florfenicol were used in fish farming in Finland (EVIRA, 2015a,b). The antibiotics are used mainly against fish pathogens, such as Aeromonas salmonicida, Flavobacterium psychrophilum, and Flavobacterium columnare, and pathogens which occur only in sea farms, such as Vibrio anguillarum, Yersinia ruckeri, Pseudomonas anguilliseptica, and Pseudomonas edwardsielloosi (Viljamaa-Dirks, 2016).

Previously, we investigated the presence of ARGs in sediments below two fish farms in the Northern Baltic Sea, Finland (Tamminen et al., 2011a; Muziasari et al., 2014, 2016). Both farms raise rainbow trout (Oncorhynchus mykiss, Walbaum) and European whitefish (Coregonus lavaretus, Linnaeus). In the studied farms, the combination of sulfonamide-trimethoprim mixed in with fish feed was used against fish pathogens such as furunculosis-causing A. salmonicida. Occasionally, florfenicol was also used. Oxytetracycline had not been used since year 2000. However, records of the amount of antibiotics used at the studied farms were not available. Our previous study showed that certain ARGs encoding resistance to tetracycline, sulfonamide, trimethoprim, and aminoglycoside, an antibiotic that has never been used at the farms, were enriched in the two farm sediments even though the concentration of antibiotics in the sediments was low (1–100 ng/g of sediment). The ARGs were abundant and persistent in the farm sediments for several years but were not detected in sediments at a distance of 200 m from the fish farms. Genes associated with class 1 integrons and transposons were also the genes enriched in the farm sediments. The presence of class 1 integrons and transposons of mobile genetic elements (MGEs) in the environment can contribute to the spreading of ARGs via horizontal gene transfer systems (Aminov, 2011). Sources of the gene enrichment in the farm sediments have yet to be elucidated.

We assumed that a constant influx of farmed fish feces contributes to the enrichment of ARGs in the farm sediments. Farmed fish consume antibiotic-containing feed and selection for resistant bacteria may occur in the fish intestine during antibiotic treatment (Giraud et al., 2006). Thus, resistant bacteria carrying ARGs in the fish intestine may be secreted into fish feces and released to the surrounding water and finally to the sediments (Kümmerer, 2009). Both farms use an open cage system in which fish are kept in a net that allows free transfer of fish farming waste, containing fish feces, uneaten feed, and antibiotics, from the farms to surrounding water environments and eventually to sediments. Since the Baltic Sea has no tides and water circulation is slow (Ojaveer et al., 2010), the fish farming waste directly impacts the sediments beneath the farms.

The antibiotic resistome consists of all the existing ARGs that are capable of conferring resistance to antibiotics (Wright, 2007). Previous studies on the resistome of farmed fish detected up to 10 ARGs in cultured resistant-bacteria from skin, gills, intestinal contents, and meat (Schmidt et al., 2001; Furushita et al., 2003; Akinbowale et al., 2007; Jacobs and Chenia, 2007; Ndi and Barton, 2011). To study the antibiotic resistome in environmental samples, in which the majority of bacteria are not cultivable, culture-independent methods can be used (Perry and Wright, 2013). Highly parallel qPCR arrays provide a cultureindependent method that permits the combination of hundreds of assays to detect and quantify selected known genes in a single experiment. This method has been used to analyze the resistome composition in environmental samples (Looft et al., 2012; Zhu et al., 2013; Wang et al., 2014; Su et al., 2015; Karkman et al., 2016) and fish farming environments (Muziasari et al., 2016).

In the present study, we used a highly parallel qPCR array with 364 primer sets for the detection and quantification of ARGs, other genes encoding resistance to antibacterial compounds, a mercury resistance gene, MGE associated genes, and the 16S rRNA gene. The qPCR array was used to analyze the composition of intestinal content resistomes of rainbow trout and whitefish farmed at the two Baltic Sea fish farms. The parts of farmed fish that are exposed to the surrounding water environments, such as fish skin and gills, were also analyzed to detect the possible risk of ARG spread in the fish farming environments.

# MATERIALS AND METHODS

## Sampling

Rainbow trout, O. mykiss (Walbaum), and European whitefish, C. lavaretus (Linnaeus) were sampled from two fish farms in the Northern Baltic Sea, Finland, in September 2014. Each farm produces ∼50 metric tons of both rainbow trout and whitefish annually. The two farms used an open cage system in which each cage was 20 m in diameter and 5 m deep. The average depth was 7.5 m (±SD 1.3 m) and the average water temperature was 14.2◦C (±0.85◦C) during the sampling time. Detailed information of the two farms has been described previously (Pitkänen et al., 2011; Tamminen et al., 2011a,b; Muziasari et al., 2014, 2016).

Five fish from each group of small rainbow trout, big rainbow trout, small whitefish and big whitefish were sampled (**Table 1**). The 20 sampled fish were randomly picked from healthy fish in the net cages. The fish were bought and slaughtered using the normal fish farming process. The fish were put in boxes of ice, transported to the laboratory (duration of transport 6 h), and directly processed. The fish were individually measured and weighed. Mucosal skin and gill filaments were sampled from each fish using sterile cotton swabs before the fish abdominal cavity was opened using sterile surgical tools. The tissues and visceral

#### TABLE 1 | Farmed fish samples from the Northern Baltic Sea farms.


\*The Baltic Sea farms also raise fish that are already mature fish.

\*\*History of antibiotic treatments before entering the Baltic Sea farms was not available.

fat surrounding the digestive system were removed. The portion of the digestive system consisting of the small and large intestines was removed. The content of the fish intestinal tract was gently squeezed into a zipper plastic bag and homogenized manually. The intestinal contents of each fish were analyzed individually as biological replicates. The mucosal skin and gill filaments from groups of five fish were pooled. Five fish per group were chosen to account for the minimum number of incremental samples to be taken, in line with EU commission regulation No 252/2012. The sample materials were kept at −80◦C until DNA extraction.

# DNA Extraction

Total DNA was extracted directly from 200 mg of each homogenized intestinal content sample using the QIAamp DNA Stool Mini Kit (Qiagen Sciences, Germantown, MD) and the pre-treatment using FastPrep method (MP Biomedicals, Irvine, CA, USA) following the manufacturer's instructions. In addition, total DNA from a swab of mucosal skin and gill filaments was extracted using the Cador Pathogen Mini Kit (Qiagen Sciences, Germantown, MD) and the pre-treatment B2 according to the manufacturer's instructions. The pre-treatment step was added in each DNA extraction method to improve the quality and yield of DNA for the qPCR array measurement. The DNA quality and concentration were analyzed with a Nanodrop 1000TM spectrophotometer (Thermo Scientific, Wilmington, DE, USA).

# The qPCR Array

The qPCR reactions were performed using 364 primer sets (**Table S1**). Validation of the primer sets was confirmed in the previous studies (Looft et al., 2012; Pitkänen et al., 2011; Tamminen et al., 2011a; Zhu et al., 2013; Muziasari et al., 2014). Of the 364 primer sets; 307 primer sets were used for ARGs encoding resistance to the nine main classes of antibiotics (aminoglycoside, beta-lactam, (flor)/(chlor)/(am)phenicol, macrolide (MSLB), multidrug-efflux, sulfonamide, tetracycline, trimethoprim, and vancomycin) and covering the three major mechanisms of antibiotic resistance (antibiotic efflux-pumps, cellular protection and antibiotic deactivation); 21 primer sets for other genes encoding resistance to antibacterial compounds, such as antiseptic and antibacterial peptides; 34 primer sets for genes associated with MGEs, such as plasmids, transposons, and insertion sequences (IS) as well as integrons; one primer set for a mercury resistance gene (merA) to study a possible co-selection of ARGs with mercury resistance gene; and one primer set for the 16S rRNA genes. Each primer set was designed to target sequence diversity within a gene to assess the environmental resistome and was therefore analyzed independently.

The qPCR array was performed in the SmartChip Real-time PCR system (WaferGen Biosystem, Freemont, CA, USA). In short, the SmartChip has 5184 reaction wells with a volume of 100 nL each, filled using the SmartChip Multisample Nanodispenser (WaferGen Biosystem, Freemont, CA, USA). PCR cycling conditions and initial data processing were as previously described (Wang et al., 2014). However, the detection limit was at a threshold cycle (CT) of 27 (Zhu et al., 2013; Karkman et al., 2016; Muziasari et al., 2016). The melting curve of each PCR product was also analyzed to monitor the specificity of the primer sets. The abundance of the detected genes in proportion to the 16S rRNA gene in each DNA sample was calculated using relative quantification with the 2−1CT method, in which 1C<sup>T</sup> = (C<sup>T</sup> detected gene − C<sup>T</sup> 16S rRNA gene) (Schmittgen and Livak, 2008).

# Data Analysis

To see if the MGEs are involved in the prevalence of ARGs, the correlation between the total relative abundances of transposases and the most prevalent ARG in the intestinal contents of the farmed fish were analyzed using linear regression. A p < 0.05 were considered to be significant. In addition, the notched boxplot was used to see the difference between the detected genes' relative abundances. The detected genes were grouped based on the three mechanisms of ARGs (antibiotic deactivation, cellular protection, and efflux-pumps), resistance genes to antiseptics, and MGE-associated genes. The notch in the box-plot displays the 95% confidence interval of the median (McGill et al., 1978). The linear regression and notched box-plot was plotted using a graphing package for R, ggplot2 v.0.9.3.1.

The 20 farmed-fish were grouped based on fish species (rainbow trout and whitefish), size, and age (small and big fish), farming time (3 and 15 months), and the history of antibiotic treatment at the farms (treated and non-treated) as shown in **Table 1**. To compare the resistome composition of the intestinal contents between the farmed fish groups, the genes' presence in each sample and their relative abundance values were used to calculate the distance matrix of non-metric multidimensional scaling (NMDS) or matrix of dissimilarities.

The NMDS distance matrix between samples was performed based on the Bray-Curtis method using function metaMDS from the Vegan package in R (Oksanen et al., 2014). Values missing from the data due to no detection in the qPCR array were replaced with zero for the data analysis. Function ordiplot and ordiellipse from Vegan were used, respectively, to plot the NMDS distance matrix results and to add the 95% confidence region of the farmed fish groups. All analyses were performed using RStudio v.0.98.501 (RStudio, Boston, MA, 2012).

# RESULTS

# Antibiotic Resistance Genes in the Intestinal Contents

We analyzed the ARG composition in the intestinal contents of 20 fish from two Northern Baltic Sea farms. The 20 farmedfish consisted of five fish from each group of small rainbow trout, big rainbow trout, small whitefish, and big whitefish. The WaferGen qPCR array used 307 primer sets to detect and quantify ARGs, 21 primer sets for antibacterial or antiseptic resistance genes, 34 primer sets for MGE-associated genes, and one primer set for a mercury resistance gene. Altogether 28 out of the 363 genes were detected in the intestinal contents of the farmed-fish (**Figure 1**). The detected genes included transposases (tnpA) associated with IS21, IS6, IS6100, IS1216, and ISEcp1, an integrase gene of class 1 integron (intI1), a gene encoding resistance to an antiseptic which is also known as a backbone gene of class 1 integrons (qacE11), a sulfonamide resistance gene (sul1), a trimethoprim resistance gene (dfrA1), tetracycline resistance genes [tet(32), tetM, tetO, tetT, and tetW], aminoglycoside resistance genes (aadA1 and aadA2), a chloramphenicol resistance gene (catA1), a multidrug-efflux resistance gene (emrB), and macrolide (MLSB)-efflux resistance genes (matA, mefA, and msrA).

Three of the 28 detected genes were found in all four groups of the farmed-fish: small rainbow trout (small\_rt), big rainbow trout (big\_rt), small whitefish (small\_wf), and big whitefish (big\_wf) (**Figure S1**). The three genes were all transposases that associated with IS1216, IS6, and ISEcp1. The transposase associated with IS1216 was the most abundant gene detected in almost every intestinal content with a relative abundance of ca. 10−3–10−<sup>1</sup> in proportion to the 16S rRNA gene (**Figure 1**). The tetM-01 and tetM-02, targeting variants of the tetracycline resistance gene, tetM, showed these genes to be the most prevalent and abundant among the detected genes with relative abundances of ca. 10−2–10−<sup>1</sup> in proportion to the 16S rRNA gene. Other tetracycline resistance genes [tet(32), tetO, tetT, and tetW] were found in at least one of the five intestinal contents from small\_rt, small\_wf, and big\_wf. The total relative abundances of the tet genes and transposases were significantly correlated [F(1, 9) = 22.3, p = 0.001, R <sup>2</sup> = 0.71] indicating that the transposases could be connected to the prevalence of the tet genes in the intestinal contents (**Figure 2**). tet genes were not found in the big\_rt samples. Two samples of the big rainbow trout, big\_rt1 and big\_rt5, differed from the rest of the intestinal

content samples (**Figures S1**, **S2**) and carried genes that were not detected in the other intestinal content samples (**Figure 1**). These genes were transposases associated with IS21 and IS6100, intI1, qacE11, sul1, dfrA1, aadA1, aadA2, and catA1. The efflux-pump resistance genes, emrB, matA, mefA, and msrA, were found in at least one of the intestinal contents. The mercury resistance gene (merA) was not detected in any of the samples.

In all, the detected ARGs in the intestinal contents of the farmed fish included the three major resistance mechanisms: antibiotic deactivation, cellular protection, and efflux pumps (**Table S4**, **Figure 3A**). The detected genes' relative abundances were ca. 10−4–10−<sup>1</sup> in proportion to the 16S rRNA gene (**Figure 3B**, **Tables S2**, **S3**). Relative abundances of the MGE, cellular protection and antiseptic resistance genes were similar and higher compared to the efflux-pump and antibiotic deactivation resistance genes (**Figure 3B**).

We compared the resistome composition of the intestinal contents between groups of the farmed-fish (**Table 1**). The groups based on fish species, size, farming time, and the history of antibiotic treatment at the farms did not separate in the nonmetric Bray-Curtis dissimilarity (**Figures S2A–D**). However, the intestinal content resistome of each individual fish within the groups seemed to have its own composition, as shown in **Figure 1**.

# Antibiotic Resistance Genes in the Mucosal Skin and Gill Filaments

The 364 assays of the WaferGen qPCR array were used to observe and quantify the resistome not only in the intestinal contents of the farmed fish but also in the fish mucosal skin and gill filaments. The skin and gill samples were pooled according to fish species and size to obtain a sufficient amount of DNA for the qPCR array analysis. In total, two genes were detected in the mucosal skin and only one gene in the gill filaments (**Figure 1**). In the mucosal skin, the two detected genes were rep, a specific replicon region of broad-host-range IncN plasmid, and emrB, a multidrug-efflux resistance gene. In the mucosal gill filaments, the emrB was the only detected gene. The emrB was the only gene found in each sample type: the intestinal contents, mucosal skin and gill filaments of the farmed fish.

# DISCUSSION

# Farmed Fish Feces as a Source of Gene Enrichment in Sediments

There is growing concern that antibiotic use in fish farming promotes the enrichment of ARGs in the environment (Cabello et al., 2013). The prevalence of resistant bacteria carrying ARGs (Nonaka et al., 2007; Shah et al., 2014) and enrichment of ARGs in sediments below fish farms have been observed globally (Tamminen et al., 2011a; Gao et al., 2012; Muziasari et al., 2014, 2016; Xiong et al., 2015). However, the causes of ARG enrichment in the farm sediments have received very little attention. We assumed that farmed fish feces are a plausible source of ARG enrichment in sediments below fish farms in the Northern Baltic Sea, Finland. In this study, we analyzed the resistome composition of the intestinal contents of 20 fish raised at the Baltic Sea farms. We used the WaferGen qPCR array with 364 primer sets of known target genes. We compared the composition of the genes detected in the intestinal contents to our previous data of the sediment resistome at the farms (Tamminen et al., 2011a; Muziasari et al., 2014, 2016). We found that most of the detected genes in the intestinal contents (20 of the 28 genes) reflected the composition of the genes enriched in the farm sediments. **Table 2** shows the 20 same genes that were detected in both farmed fish intestinal contents and farm sediments including transposases, class 1 integron-associated genes, and ARGs encoding resistance to tetracycline, sulfonamide, trimethoprim, and aminoglycoside. Average relative abundance of the genes in the intestinal contents was ca. 10−2–10−<sup>1</sup> in proportion to the 16S rRNA gene and ca. 10−4–10−<sup>3</sup> in proportion to the 16S rRNA gene in the farm sediments (**Table 2**). Bacterial community changes in the sediments in response to fish farming in the Baltic Sea also have been reported, with Actinobacteria, Chloroflexi, and Firmicutes prominent in the farm sediments (Tamminen et al., 2011b). Actinobacteria and Firmicutes are known as the core intestinal microbiota of salmonids (Pond et al., 2006; Navarrete et al., 2010; Wong et al., 2013). These findings provide indirect evidence supporting our assumption that certain ARGs are being introduced into the sediments below the fish farms in the

ARGs are grouped by the mechanism of resistance: antibiotic deactivation, cellular protection, and efflux-pumps. (B) Relative abundances of the genes detected in the farmed fish in proportion to the 16S rRNA gene (as log values). The box-plot presents the biological replicates of the farmed fish (n = 20) and the notch the 95% confidence intervals of the median.

Northern Baltic Sea via the bacteria contained in the feces of the farmed fish.

The enrichment of ARGs also could be caused by the presence of antibiotics in the sediments. Although concentrations of tetracycline, oxytetracycline, sulfamethoxazole, sulfadiazine, and trimethoprim in the sediments were very low (ca. 1–100 ng/g of sediment), selection of resistance by antibiotics might have occurred in the past with the ARGs persisting in the sediments (Tamminen et al., 2011a; Muziasari et al., 2014). Enrichment of ARGs in the absence of selective antibiotic pressure could also be caused by co-selection with heavy metals (Baker-Austin et al., 2006). While mercury can be found in fish feed (Choi and Cech, 1998), biological and chemical analyses of the sediments revealed only very low background concentrations of this metal (Pitkänen et al., 2011). However, other heavy metals (copper, cadmium, zinc) can potentially be present in fish feed or antifouling substances and exert co-selective pressures in these sediments (Seiler and Berendonk, 2012).

On the other hand, the mercury resistance gene merA was not detected in the intestinal contents, but it was enriched in the farm sediments (Pitkänen et al., 2011). Similarly, not all sediment-enriched ARGs were detected in the intestinal contents. This might be due to the amount of the genes in the intestinal contents that were below the detection limit of the qPCR array, which varies between the primer sets (ca. 10−2–10−<sup>4</sup> copy number of target genes). Alternatively, the enriched genes in the farm sediments might be coming from other sources such as uneaten medicated feed (Kerry et al., 1995). The deposition



<sup>a</sup>Muziasari et al. (2016).

<sup>b</sup>Muziasari et al. (2014).

<sup>c</sup>Tamminen et al. (2011a).

<sup>a</sup>, <sup>c</sup>The quantification of the genes using standard qPCR.

The 20 genes were the same genes found to be enriched in the sediments below fish farms in the Northern Baltic Sea, Finland. The table also shows the average relative abundance of the genes to the 16S rRNA gene in the intestinal contents and in the farm sediments. The gene assays of the qPCR array grouped by classification of the antibiotics the genes confer resistance to, class 1 integron and transposon.

of uneaten feed that increases the amount of organic matter in the farm sediments might be also resulting in the enrichment of indigenous sediment bacteria carrying the enriched ARGs (Tamminen et al., 2011b; Buschmann et al., 2012).

Because the ARGs were detected in the intestinal contents of the farmed fish with no history of antibiotic treatment at the farms, the ARGs could potentially be derived from resistant bacteria in the surrounding water environment that were able to be transferred to the intestinal microbiota of fish (Giatsis et al., 2015). The ARGs also may have already been selected in the fish intestine before entering the Baltic Sea farm cages, possibly during the processes of hatching or rearing juvenile fish in freshwater ponds. Specific record of antibiotics use in the juvenile fish prior to entering the Baltic Sea farms was not available. However, it has been reported that antibiotics were used against F. psychrophilum and F. columnare in Finnish hatchery farms (EVIRA, 2007). Antibiotic resistant bacteria can be found in fish eggs and juvenile fish due to the routine use of antibiotics during the rearing process (Hansen and Olafsen, 1999). The ARGs are still present in the fish intestine even after transferred to different water environments at the Baltic Sea farms, maybe because the resistant bacteria carrying the ARGs are the core intestinal microbiota of the farmed fish (Ghanbari et al., 2015). Although our data does not refute the prior understanding on the occurrence of antibiotic resistance in the fish farming environment due to antibiotic use at farms, it suggests that the juvenile fish may have a greater role as the source of antibiotic resistance than has been expected before. The routine use of antibiotics in fish hatchery and rearing farms should not be recommended because it may affect the natural microbiota of fish larvae and juvenile (Ringø et al., 1995). It is, therefore, important to monitor the antibiotic treatments of juvenile fish to minimize the potential risk of ARG emergence in fish farm environments.

# Resistome Composition of the Intestinal Contents

Tetracycline resistance (tet) genes were the most prevalent ARGs in the intestinal contents of the farmed fish. The detected tet genes were genes which only encode ribosomal protection proteins (RPP), the tet(32), tetM, tetO, tetT, and tetW. tetM was found to be the most prevalent of the detected ARGs. This was expected, since tetM has the widest host range of all known tet genes and is well known to be associated with a very wide range of conjugative transposons (Roberts, 2005). Beside tetM our results were in contrast to other studies in which the efflux-pump genes, tetA, tetD, and tetE were commonly found in resistant bacteria isolated from intestines of farmed fish (Furushita et al., 2003; Akinbowale et al., 2007). The transposase-IS1216 is predominant in the intestinal contents of the Baltic Sea farmed fish and known to be associated with conjugative transposons (Ciric et al., 2011). The RPP genes are usually associated with conjugative transposons, while the efflux-pump genes are generally associated with large plasmids (Roberts, 2012). Moreover, we observed a significant correlation between the total relative abundances of the tet genes and transposases (p = 0.001, R <sup>2</sup> = 0.71). Transposases are, therefore, likely involved in the prevalence of the tet genes in the Baltic Sea farmed fish intestinal contents. A highly significant correlation between the abundance of transposases and tet genes has been observed also in swine manure (Zhu et al., 2013).

In this study, the tet genes were not found in the intestinal contents of the big rainbow trout. However, some genes were only found in the intestinal contents of two big rainbow trout. Those genes were the transposases associated with IS21 or Tn21 and IS6100, genes associated with class 1 integrons, intI1 and qacE11, and the ARGs sul1, dfrA1, aadA1, aadA2, and catA1. Class 1 integrons are known to be able to capture and incorporate gene cassettes consisting of sul1, dfrA1, aadA1, aadA2 (Partridge et al., 2009), and catA1 (Soto et al., 2003). In the farm sediments, class 1 integrons carry the sul1 and aadA1 genes (Pärnänen et al., 2016). This explains the presence of aadA1 aminoglycoside resistance genes at the farms even though aminoglycosides have never been used there. The prevalence of class 1 integrons could be cause by the association of class 1 integrons with Tn21 (Liebert et al., 1999) or with IS6100 (Partridge et al., 2001). The presence of sul1, dfrA1, aadA1, aadA2, and catA1 in the intestinal contents of two big rainbow trout may have been mediated by the class 1 integrons carried by MGEs such as Tn21 or IS6100.

The chloramphenicol and macrolide resistance genes were found in the intestinal contents and are also known to be associated with conjugative transposons (Hegstad et al., 2010). This suggests that the presence of ARGs in the farmed fish intestinal contents may be connected to the prevalence of transposases. Otherwise, the ARGs, especially efflux-pump resistance genes, could be intrinsically possessed by fish bacteria since the efflux-pump resistance genes are also known to have other functions beside antibiotic resistance, such as intracellular metabolite detoxification (Martinez et al., 2009).

We observed that the resistome composition of the intestinal contents varied between individual farmed fish. Beside the connection with the MGEs, the resistome composition in the fish intestine may be also correlated with the composition of bacterial communities (Forsberg et al., 2014). It has been reported that the bacterial communities in the intestines of individual farmed fish varies greatly and stable regardless of the variations of diet and rearing system (Wong et al., 2013; Lyons et al., 2017). Our results also show that the resistome composition of the intestinal contents might not connected to the types of fish species, size, farming time, and antibiotic treatments at the farms. This suggests that the resistome of the farmed-fish intestinal contents has its own composition and is influenced less by the fish farming processes. However, these results may vary if more farmed fish are sampled and included in the future studies. To our knowledge, this is the first study reporting the resistome composition of farmed-fish using a culture-independent method.

The intestinal content resistome of the farmed-fish covered the three mechanisms of resistance genes: antibiotic deactivation, cellular protection, and efflux-pumps. The relative abundance of cellular protection resistance genes was higher than the other two mechanisms of resistance genes. This is because the cellular protection resistance genes included all the detected RPPtetracycline resistance genes that were the most abundant ARGs in the intestinal contents of the farmed-fish. In the farm sediment resistome, however, the efflux-pump resistance genes were the most abundant (Muziasari et al., 2016). The low frequency of detection of efflux-pump resistance genes in the resistomes of farmed-fish intestines in this study might be due to the qPCR array targeting only known ARGs since the qPCR array requires prior knowledge of ARG sequences for primer design. Insight on the ARG mechanisms within fish farming-associated resistomes is essential for predicting the emergence of ARGs at the fish farm facilities (Miranda et al., 2013).

# Resistome Composition in the Mucosal Skin and Gill Filaments

In the mucosal skin and gill filaments of the farmed-fish, only one of the 307 targeted ARGs, the multidrug-efflux resistance gene, emrB, was detected. The emrB was also detected in the intestinal contents. Since the emrB was found in the three different sample types taken from the farmed fish, the emrB gene maybe intrinsically possessed by fish bacteria (Martinez et al., 2009). In other studies, ARGs encoding resistance to tetracycline, sulfonamide, and trimethoprim have been frequently found in resistant bacteria isolated from the skin and gills of farmed fish (Schmidt et al., 2001; Akinbowale et al., 2007). The different results might be due to the difference between ARG detection methods, as the qPCR array detection limit may prevent detection of a low abundance of ARGs in fish mucosal skin and gill filaments.

# CONCLUSION

This study indicates that the resistomes of the intestinal contents of farmed fish contribute to the enrichment of ARGs in sediments below fish farms in the Northern Baltic Sea. Using high-throughput methods to study fish resistomes, we found a significant correlation between the relative abundances of the transposases in them and tetracycline resistance genes. This might lead to the risk of ARG mobilization from piscine bacteria to bacteria in the surrounding environments. The presence of transposases and class 1 integrons might also affect the prevalence of certain ARGs in the intestinal contents of individual fish and shape the composition of their resistomes. Determining the sources of ARGs at fish farms is thus important for managing and minimizing the emergence of ARGs at the fish farms and for lowering the risk of ARG spreading to the surrounding farm environments.

# AUTHOR CONTRIBUTIONS

WM and MV: contributions to the experimental design. WM and LP: sampling. WM, LP, and RS: acquisition and analysis of data for the work. HS, JT, and MV: analysis, interpretation and adding important intellectual content. WM, LP, HS, RS, JT, and MV: drafting and revising the manuscript.

# FUNDING

Academy of Finland grants, Maj and Tor Nessling Foundation and with support from the Center for Microbial Ecology and the Center for Health Impacts of Agriculture at Michigan State University.

# ACKNOWLEDGMENTS

The authors thank Johanna Muurinen and Harri Urponen for technical support on sampling at the fish farms.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2016.02137/full#supplementary-material

Figure S1 | Venn diagram. Number of genes shared by the groups of the Baltic Sea farmed fish: small rainbow trout (small\_rt), big rainbow trout (big\_rt), small

whitefish (small\_whitefish), and big whitefish (big\_wf). Each group consisted of five farmed fish.

Figure S2 | Non-metric multidimensional scaling (NMDS) of the intestinal content resistomes of fish farmed at the Baltic Sea farms. The distance matrix between the intestinal content resistomes was based on Bray-Curtis dissimilarity. The farmed fish (n = 20) were grouped based on (A) Fish species: rainbow trout (dark red) and whitefish (rose). (B) Fish size: big fish (dark red) and small fish (rose). (C) Farming time at the farms: 3 months (dark red) and 15 months (rose). (D) The history of antibiotic treatment at the farms: Non-treated (dark red) and treated with a combination of sulfonamide-trimethoprim 1 month

# REFERENCES


before the fish were sampled (rose). Ellipse line indicates 95% confidence regions of the farmed fish groups.

Table S1 | The list of primer sets used in the qPCR array.

Table S2 | Gene abundance values relative to the 16S rRNA gene in Rainbow trout raised at the Northern Baltic Sea farms.

Table S3 | Gene abundance values relative to the 16S rRNA gene in European whitefish raised at the Northern Baltic Sea farms.

Table S4 | Number of positive qPCR assays and the total assays.


aquaculture and non-aquaculture sites. Environ. Microbiol. 16, 1310–1320. doi: 10.1111/1462-2920.12421


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Muziasari, Pitkänen, Sørum, Stedtfeld, Tiedje and Virta. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.