# EVOLUTION OF GENETIC MECHANISMS OF ANTIBIOTIC RESISTANCE

EDITED BY : Silvia Buroni, Simona Pollini, Gian Maria Rossolini and Elena Perrin PUBLISHED IN : Frontiers in Genetics and Frontiers in Microbiology

#### Frontiers eBook Copyright Statement

The copyright in the text of individual articles in this eBook is the property of their respective authors or their respective institutions or funders. The copyright in graphics and images within each article may be subject to copyright of other parties. In both cases this is subject to a license granted to Frontiers. The compilation of articles constituting this eBook is the property of Frontiers.

Each article within this eBook, and the eBook itself, are published under the most recent version of the Creative Commons CC-BY licence. The version current at the date of publication of this eBook is CC-BY 4.0. If the CC-BY licence is updated, the licence granted by Frontiers is automatically updated to the new version.

When exercising any right under the CC-BY licence, Frontiers must be attributed as the original publisher of the article or eBook, as applicable.

Authors have the responsibility of ensuring that any graphics or other materials which are the property of others may be included in the CC-BY licence, but this should be checked before relying on the CC-BY licence to reproduce those materials. Any copyright notices relating to those materials must be complied with.

Copyright and source acknowledgement notices may not be removed and must be displayed in any copy, derivative work or partial copy which includes the elements in question.

All copyright, and all rights therein, are protected by national and international copyright laws. The above represents a summary only. For further information please read Frontiers' Conditions for Website Use and Copyright Statement, and the applicable CC-BY licence.

ISSN 1664-8714 ISBN 978-2-88963-222-0 DOI 10.3389/978-2-88963-222-0

#### 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

# EVOLUTION OF GENETIC MECHANISMS OF ANTIBIOTIC RESISTANCE

Topic Editors: Silvia Buroni, University of Pavia, Italy Simona Pollini, University of Florence, Italy Gian Maria Rossolini, University of Florence, Italy Elena Perrin, University of Florence, Italy

Image: created by Silvia Buroni with BioRender.com

Since the first introduction of antibiotics into clinical practice, microbial drug resistance has emerged as a major obstacle in the treatment of infections. Recently, the combination of emergence of a complex variety of multidrug resistant strains and the dearth of newly discovered molecules to effectively target and eliminate these strains, has made antibiotic resistance one of the major public health problems of this century.

Although different strategies can be adopted to contain the emergence and spread of antibiotic resistance, including (i) antimicrobial stewardship, (ii) infection control, and (iii) tighter control over the use of antibiotics in agriculture and breeding, a better understanding of the dynamics that lead to the evolution of antibiotic resistance remains essential for the development of more efficient strategies to combat this phenomenon.

The recent developments in genomics have greatly contributed to expand our knowledge on the mechanisms of microbial resistance, and of the processes by which they emerge, develop and spread. Different approaches and expertise can be used to accelerate advances in this area, ranging from clinical studies on the evolution of resistance in vivo, to theoretical modeling and the study of evolution in the laboratory.

Citation: Buroni, S., Pollini, S., Rossolini, G. M., Perrin, E., eds. (2019). Evolution of Genetic Mechanisms of Antibiotic Resistance. Lausanne: Frontiers Media SA. doi: 10.3389/978-2-88963-222-0

# Table of Contents

*05 Editorial: Evolution of Genetic Mechanisms of Antibiotic Resistance* Silvia Buroni, Simona Pollini, Gian Maria Rossolini and Elena Perrin

### CHAPTER 1

#### ANTIBIOTIC RESISTANCE GENES AS NEW ANTIBIOTIC TARGET


Diana Machado, Emmanuel Lecorche, Faiza Mougari, Emmanuelle Cambau and Miguel Viveiros

*29 Effect of a Point Mutation in* mprF *on Susceptibility to Daptomycin, Vancomycin, and Oxacillin in an MRSA Clinical Strain*

Feng-Jui Chen, Tsai-Ling Lauderdale, Chen-Hsiang Lee, Yu-Chieh Hsu, I-Wen Huang, Pei-Chi Hsu and Chung-Shi Yang

#### CHAPTER 2

#### ANTIBIOTIC RESISTANCE GENES AS GENETIC MARKERS FOR BACTERIAL POPULATION DIVERSITY AND FOR MONITORING ANTIBIOTIC RESISTANCE

*41 Genetic Diversity of* norA*, Coding for a Main Efflux Pump of*  Staphylococcus aureus

Sofia Santos Costa, Benjamin Sobkowiak, Ricardo Parreira, Jonathan D. Edgeworth, Miguel Viveiros, Taane G. Clark and Isabel Couto


Jie Gong, Meng Xiao, He Wang, Timothy Kudinha, Yu Wang, Fei Zhao, Weiwei Wu, Lihua He, Ying-Chun Xu and Jianzhong Zhang

#### CHAPTER 3

#### WHOLE GENOME SEQUENCING ANALYSIS OF MOBILE GENETIC ELEMENTS ASSOCIATED WITH ANTIBIOTIC RESISTANCE

*74 Characterization of NDM-Encoding Plasmids From* Enterobacteriaceae *Recovered From Czech Hospitals*

Veronika Paskova, Matej Medvecky, Anna Skalova, Katerina Chudejova, Ibrahim Bitar, Vladislav Jakubu, Tamara Bergerova, Helena Zemlickova, Costas C. Papagiannitsis and Jaroslav Hrabak


Francesco Santoro, Alessandra Romeo, Gianni Pozzi and Francesco Iannelli

## CHAPTER 4

#### WHOLE GENOME SEQUENCING ANALYSIS AND RESISTOME


Yao-Ting Huang, Jia-Min Chen, Bing-Ching Ho, Zong-Yen Wu, Rita C. Kuo and Po-Yu Liu

### CHAPTER 5

#### DIRECT EVOLUTION EXPERIMENTS TO STUDY HOW ANTIBIOTIC RESISTANCE EMERGES, EVOLVES AND SPREADS

*134 Mutational Evolution of* Pseudomonas aeruginosa *Resistance to Ribosome-Targeting Antibiotics*

Fernando Sanz-García, Sara Hernando-Amado and José L. Martínez


Johannes Cairns, Roosa Jokela, Jenni Hultman, Manu Tamminen, Marko Virta and Teppo Hiltunen

## Editorial: Evolution of Genetic Mechanisms of Antibiotic Resistance

*Silvia Buroni1, Simona Pollini2,3, Gian Maria Rossolini2,3 and Elena Perrin4\**

*1 Department of Biology and Biotechnology, University of Pavia, Italy, 2 Department of Experimental and Clinical Medicine, University of Florence, Italy, 3 Microbiology and Virology Unit, Careggi University Hospital, Italy, 4 Department of Biology, University of Florence, Italy*

Keywords: antibiotic resistance, evolution, genetic mechanisms, mobile genetic elements, whole genome sequencing

**Editorial on the Research Topic**

#### **Evolution of Genetic Mechanisms of Antibiotic Resistance**

Almost simultaneously with the introduction of antibiotics into clinical practice, antibiotic resistance (AR) emerged as a crucial problem for the treatment of infections (Gould, 2016; Podolsky, 2018). While initially the problem could be solved by new antibiotics, it became relevant with the dearth of the discovery of new classes of molecules (Podolsky, 2018), up to the present days in which, accordingly to World Health Organization (WHO), we are close to a "Post-Antibiotic" era (Reardon, 2014).

While different strategies could be used to try to counteract AR (Lee et al., 2013), a better understanding of the dynamics that lead to its evolution remains essential for the development of more efficient strategies to combat this phenomenon (Davies and Davies, 2010; Lukacisinova and Bollenbach, 2017).

This Research Topic collects 15 articles focused on different aspects of the genetic mechanisms and on the evolution and spread of AR. Starting from the analysis of single genes related to resistance, the topic moves to the analysis of resistance at whole genome and population levels thanks to the recent advances in genome sequencing and analysis, and also covers direct evolution experiments that allowed to follow the emergence, evolution and spread of resistance in defined laboratory conditions.

An in-depth knowledge of the genetic mechanisms of AR allows, for example, to identify new possible antibiotic targets. Two different examples have been reported for mycobacterial species (Machado et al.; Sanz-Garcia et al.). In both cases, characterization of the physiological function of an AR determinant revealed an important role that goes beyond resistance. Sanz-Garcia et al., through a review of the literature on aminoglycoside acetyltransferases in mycobacteria, highlighted the roles of these enzymes in resistance to muramidase enzymes, thanks to their contribution to acetylation of peptidoglycan, and as virulence determinants, since, for example, they control the production of pro-inflammatory cytokines. The authors suggested that these enzymes are promising drug targets rather than a resistance mechanism to aminoglycoside. The work by Machado et al. is instead focused on efflux systems of *Mycobacterium leprae*. The fact that several efflux systems are conserved, despite the drastic genome reduction occurred in this species, indicates a probable role in intracellular survival, making them suitable as possible new antibiotic targets.

#### *Edited and reviewed by:*

*Ludmila Chistoserdova, University of Washington, United States*

> *\*Correspondence: Elena Perrin elena.perrin@unifi.it*

#### *Specialty section:*

*This article was submitted to Evolutionary and Genomic Microbiology, a section of the journal Frontiers in Genetics*

*Received: 30 July 2019 Accepted: 17 September 2019 Published: 11 October 2019*

#### *Citation:*

*Buroni S, Pollini S, Rossolini GM and Perrin E (2019) Editorial: Evolution of Genetic Mechanisms of Antibiotic Resistance. Front. Genet. 10:983. doi: 10.3389/fgene.2019.00983*

1 **5**

Another example of the roles of AR genes in cell biology is reported in the work of Chen et al. They demonstrated that a mutation in the *mprF* gene conferring resistance to daptomycin and vancomycin in methicillin-resistant *Staphylococcus aureus* (MRSA), not only decreases the resistance to oxacillin, but has also a pleiotropic effect on cell membrane and cell wall and on the doubling time.

AR genes can also be used as genetic markers for bacterial population diversity and to monitor AR at different levels. For example, three different alleles of the gene coding for the *norA* efflux pump have been identified in *S. aureus*, and an analysis of the variability of these genes in 112 strains suggests that different variants reflect the population structure of this species (Costa et al.). Three different tetracycline resistance genes together with three tetracyclines molecules have been instead used to monitor the presence of antibiotics and of AR in the longshore sediments of the Three Gorges Reservoir in China (Lu et al.), revealing a significant seasonal variation of both that correlates with a higher use of antibiotics in winter. Finally, Gong et al. developed genetic markers that can be used for population structure analysis and to identify the populations which are most frequently associated with AR.

Beyond single genes involved in AR, it is well known that a fundamental role in the evolution and spread of resistance is played by mobile genetic elements (MGE) that are often associated to more than one mechanism of resistance to different classes of antibiotics. The next-generation and third generation sequencing (long reads sequencing) technologies allowed a better characterization of these elements.

For example, in three different works reported in this research topic, whole genome sequencing (WGS) analysis of different plasmids from clinical isolates allowed a better understanding of the mechanisms of resistance of the associate strains and of the evolution and spread of resistance in hospital environments. In the first article, Paskova et al. reconstructed the important role played by IncX3 plasmids in the dissemination of *bla*NDM-like (New Delhi Metallo-Beta-lactamases) genes in both sporadic cases and in an outbreak of NDM-like-producing *Enterobacterales* in Czech hospitals. Di Pilato et al. have instead identified a plasmid lineage that, circulating for over 20 years in different countries in various environmental and clinical *Pseudomonas* spp., contributed to the spread of different Metallo-Beta-lactamases (MBL) encoding genes. Finally, Morroni et al. identified and sequenced a plasmid in *Enterococcus faecium* isolated from a patient in Italy, which is a serious concern since it carries and could co-spread two genes involved in resistance to last resort agents such as oxazolidinones.

Among MGE, also integrative conjugative elements (ICE) have an important role in AR, like for example the Tn*5253* ICE that confers resistance to tetracycline and chloramphenicol in *Streptococcus pneumoniae*. Santoro et al. elucidated the mechanisms of excision and circularization that allowed the transfer of this ICE.

WGS data are useful not only to study MGE associated with AR genes, but also to study the complete resistome of a certain species. A review of the WGS data obtained in different kind of experiments in *Pseudomonas aeruginosa*, for example, highlighted how these data not only allowed to understand the evolutionary dynamics of AR, but are also useful to design specific therapeutic strategies (Lopez-Causape et al.). Similar results have been obtained in a comparative analysis between the genomes of some *Stenotrophomonas acidaminiphila* strains that revealed the evolution of sulfamethoxazole resistance in this species (Huang et al.).

WGS and all the recent -omic technologies have also been the driving force for the development of several direct evolution experiments that allowed to follow and study in a controlled laboratory environment how AR emerges, evolves and spreads. Two examples of these experiments are reported in this research topic. The first one suggested that the development of resistance is in some ways predictable, since *P. aeruginosa* populations evolved in parallel in the presence of antibiotics showed similar patterns of resistance mutations (Sanz-Garcia et al.). Moreover the authors demonstrated that bacteria are able to develop higher levels of resistance to antibiotics to which they are considered intrinsically resistant (Sanz-Garcia et al.). Similar results have been obtained by Lasek-Nesselquist et al. that, using a continuous culture bioreactor model to avoid the problem of nutrients depletion, found that daptomycin resistance in *S. aureus* arises from a combination of mutations localized in few loci, and that most of these mutations appeared early, at low frequency, within the population.

A further step forward in this field is the construction of synthetic bacterial communities to be used for evolution experiments closer to real natural conditions than the use of single strains. An example of the construction and the characterization of a synthetic bacterial community for studies of experimental evolution is reported in the work of Cairns et al.

To conclude, in this Research Topic different aspects of the genetic mechanisms of AR have been addressed. New possible antibiotic targets have been identified and new information on how AR appears and diffuses have been obtained. We think that this aspect of AR is important and need to be further investigated, since trying to avoid the emergence and spread of AR could be essential to face the "Post-Antibiotic" era.

## AUTHOR CONTRIBUTIONS

EP wrote the manuscript; SB, SP, and GMR participated in manuscript correction.

## FUNDING

EP was funded by an "Assegno Premiale" from the Department of Biology, University of Florence. SB is supported by the Italian Ministry of Education, University and Research (MIUR): Dipartimenti di Eccellenza Program (2018–2022) − Department of Biology and Biotechnology "L. Spallanzani," University of Pavia.

## REFERENCES


Reardon, S. (2014). WHO warns against "post-antibiotic" era. *Nature News.* doi: 10.1038/nature.2014.15135

**Conflict of Interest:** 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 © 2019 Buroni, Pollini, Rossolini and Perrin. 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) and the copyright owner(s) 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.*

# Mycobacterial Aminoglycoside Acetyltransferases: A Little of Drug Resistance, and a Lot of Other Roles

Fernando Sanz-García<sup>1</sup>†‡, Ernesto Anoz-Carbonell1,2‡ , Esther Pérez-Herrán<sup>1</sup>† , Carlos Martín1,3, Ainhoa Lucía1,3, Liliana Rodrigues1,3,4† and José A. Aínsa1,3 \*

<sup>1</sup> Departamento de Microbiología, Facultad de Medicina – Instituto Universitario de Investigación de Biocomputación y Física de Sistemas Complejos, Instituto de Investigación Sanitaria Aragón, Universidad de Zaragoza, Zaragoza, Spain, <sup>2</sup> Departamento de Bioquímica y Biología Molecular y Celular, Facultad de Ciencias – Instituto Universitario de Investigación de Biocomputación y Física de Sistemas Complejos, Universidad de Zaragoza, Zaragoza, Spain, <sup>3</sup> Centro de Investigación Biomédica en Red Enfermedades Respiratorias, Instituto de Salud Carlos III, Madrid, Spain, <sup>4</sup> Fundación Agencia Aragonesa para la Investigación y el Desarrollo, Zaragoza, Spain

Aminoglycoside acetyltransferases are important determinants of resistance to aminoglycoside antibiotics in most bacterial genera. In mycobacteria, however, aminoglycoside acetyltransferases contribute only partially to aminoglycoside susceptibility since they are related with low level resistance to these antibiotics (while high level aminoglycoside resistance is due to mutations in the ribosome). Instead, aminoglycoside acetyltransferases contribute to other bacterial functions, and this can explain its widespread presence along species of genus Mycobacterium. This review is focused on two mycobacterial aminoglycoside acetyltransferase enzymes. First, the aminoglycoside 2<sup>0</sup> -N-acetyltransferase [AAC(2<sup>0</sup> )], which was identified as a determinant of weak aminoglycoside resistance in M. fortuitum, and later found to be widespread in most mycobacterial species; AAC(2<sup>0</sup> ) enzymes have been associated with resistance to cell wall degradative enzymes, and bactericidal mode of action of aminoglycosides. Second, the Eis aminoglycoside acetyltransferase, which was identified originally as a virulence determinant in M. tuberculosis (enhanced intracellular survival); Eis protein in fact controls production of pro-inflammatory cytokines and other pathways. The relation of Eis with aminoglycoside susceptibility was found after the years, and reaches clinical significance only in M. tuberculosis isolates resistant to the second-line drug kanamycin. Given the role of AAC(2<sup>0</sup> ) and Eis proteins in mycobacterial biology, inhibitory molecules have been identified, more abundantly in case of Eis. In conclusion, AAC(2<sup>0</sup> ) and Eis have evolved from a marginal role as potential drug resistance mechanisms into a promising future as drug targets.

Keywords: mycobacteria, aminoglycoside antibiotics, aminoglycoside acetyltransferase, drug target, pathogenicity, aminoglycoside resistance

### BACTERIAL RESISTANCE TO AMINOGLYCOSIDE ANTIBIOTICS

Aminoglycoside (AG) antibiotics (**Box 1**) have not been an exception to the fact that after their introduction in clinical practice, resistance has been recorded (Waglechner and Wright, 2017). In fact, the three classes of aminoglycoside-modifying enzymes, aminoglycoside acetyltransferases (AACs), aminoglycoside phosphotransferases (APHs), and aminoglycoside nucleotidyltransferases

#### Edited by:

Silvia Buroni, University of Pavia, Italy

#### Reviewed by:

Peter Sander, University of Zurich, Switzerland Roland Brosch, Institut Pasteur, France

> \*Correspondence: José A. Aínsa ainsa@unizar.es

#### †Present address:

Fernando Sanz-García, Departamento de Biotecnología Microbiana, Centro Nacional de Biotecnología, Consejo Superior de Investigaciones Científicas, Madrid, Spain Esther Pérez-Herrán, GlaxoSmithKline, Tres Cantos Medicines Development Campus, Tres Cantos, Spain Liliana Rodrigues, Unit of Medical Microbiology, Global Health and Tropical Medicine, Instituto de Higiene e Medicina Tropical, Universidade Nova de Lisboa, Lisbon, Portugal ‡These authors have contributed

equally to this work

#### Specialty section:

This article was submitted to Evolutionary and Genomic Microbiology, a section of the journal Frontiers in Microbiology

Received: 05 October 2018 Accepted: 11 January 2019 Published: 30 January 2019

#### Citation:

Sanz-García F, Anoz-Carbonell E, Pérez-Herrán E, Martín C, Lucía A, Rodrigues L and Aínsa JA (2019) Mycobacterial Aminoglycoside Acetyltransferases: A Little of Drug Resistance, and a Lot of Other Roles. Front. Microbiol. 10:46. doi: 10.3389/fmicb.2019.00046

(ANTs), have been widely detected in most pathogenic bacteria as a major determinant of resistance; in these bacteria the presence of aminoglycoside-modifying enzymes correlated with patterns of AG susceptibility (Smith and Baker, 2002). The modified AG (either by acetylation, phosphorylation or nucleotidylation) fails to inhibit their bacterial target, the 30S ribosomal subunit (Smith and Baker, 2002). Most genes encoding aminoglycosidemodifying enzymes are plasmid-located (indicative of a potential acquisition by horizontal gene transfer processes) and confer the bacterial hosts with high levels of AG resistance (Davies and Wright, 1997).

In mycobacteria, however, resistance to AGs resulted mainly from mutations of the ribosome components that prevent the drugs from inhibiting its function (Jugheli et al., 2009; Zhang and Yew, 2009; Shcherbakov et al., 2010). This is due to the fact that most mycobacterial species have either one (like Mycobacterium. tuberculosis) or two (like Mycobacterium fortuitum) ribosomal operons, hence making dominant those mutations acquired in their components (Magnet and Blanchard, 2005; Shi et al., 2013). The presence of aminoglycoside-modifying enzymes, mostly AAC, in mycobacterial species has been reported over the years (as detailed in the following sections), and the role of such AACs has been explored, originally for their contribution to AG resistance, and more recently for their role in other bacterial processes, which has resulted in the interest of developing inhibitors of these enzymes (Jana and Deb, 2005; Labby and Garneau-Tsodikova, 2013). In this review, we will summarize major findings on two mycobacterial AACs, the AAC(2<sup>0</sup> )-I and Eis enzymes, that have resulted in a Copernican turn for AACs in mycobacteria, from being putative drug resistance mechanisms, to reach the status of novel drug targets.

#### AACs IN NON-TUBERCULOUS MYCOBACTERIA

The first detection of AACs in mycobacteria (Hull et al., 1984) was reported in a group of M. fortuitum isolates, an opportunistic fast-growing mycobacteria. Biochemical assays of crude extracts from M. fortuitum strains revealed the presence of AAC activity, strongly acetylating gentamicin and kanamycin A, along with other AGs. This substrate profile was consistent with that of AAC(3) enzymes that had been previously described in Pseudomonas and Enterobacteriaceae (Angelatou et al., 1982), although confirmation at the genetic or molecular levels were not done at that time. Surprisingly, the AG susceptibility profile of M. fortuitum could not be correlated with the activity of AACs, indicating that in this species AACs were not the major responsible for AG resistance; it was neither correlated with the presence of plasmids, hence suggesting a chromosomal location (Hull et al., 1984). In fact, the frequency of resistant mutants to kanamycin and amikacin in M. fortuitum and the related species Mycobacterium chelonae ranged between 10−<sup>4</sup> and 10−<sup>7</sup> (Wallace et al., 1985). This relatively high frequency of mutations, along with the fact that AAC activity was detected at similar levels between susceptible and resistant strains, led the authors to suggest that ribosome alterations were the main factor responsible of AG resistance in these species (Wallace et al., 1985). In another study (Udou et al., 1986), altered transport or permeability of AGs was identified as a contributor to AG resistance in M. fortuitum, since ribosomes from a clinical isolate were inhibited by one tenth of the MIC of AGs: for example, the MIC of kanamycin for M. fortuitum was 50 µg/ml, and in cell-free systems, 5 µg/ml of kanamycin reduced the activity of ribosomes to 13% in comparison with drug-free controls; similar results were obtained when using gentamicin or paromomycin (Udou et al., 1986).

The biochemical analysis of crude extracts from other nontuberculous mycobacteria such as Mycobacterium smegmatis, Mycobacterium phlei, Mycobacterium vaccae, and Mycobacterium kansasii, from both clinical and environmental origins, revealed similar characteristics to those found in M. fortuitum: crude extracts from all strains contained AAC enzymatic activity, but no correlation with AG susceptibility profile could be established (Udou et al., 1987; Ho et al., 2000). In other mycobacterial species such as Mycobacterium avium and Mycobacterium intracellulare, however, AAC activity could not be detected (Ho et al., 2000). In a recent study, using cell-free translation assays, ribosomes of Mycobacterium abscessus and M. smegmatis were inhibited by both AGs having a 2<sup>0</sup> -amino group (tobramycin, dibekacin, and kanamycin B) and by those having a 2<sup>0</sup> -hydroxyl group (amikacin, and kanamycin A). However, in M. abscessus, those AGs having a 2<sup>0</sup> -amino group (tobramycin, dibekacin, and kanamycin B) were less efficient in killing the cells and inhibiting its ribosomes, which is consistent with the presence of a highly active AAC(2<sup>0</sup> ) activity in this species. These findings suggested that in M. abscessus, AACs could somewhat mitigate the bactericidal effect of its AGs substrates (Maurer et al., 2015).

Interestingly, when characterizing the enzymatic activity of crude extracts, early reports detected that substrates such as amino sugars, malonyl-CoA, propionyl-CoA or butyryl-CoA inhibited the enzymatic activity of AACs from mycobacterial species. Such effect of non-acetyl CoA donors had never been described for the AAC(3) enzymes from other bacteria (Udou et al., 1987). Altogether, these findings strongly suggested for the very first time that, in mycobacteria, AACs could have important metabolic functions, and their contribution to AG resistance could be marginal (Udou et al., 1989).

#### AAC(2<sup>0</sup> ), A NEW ENZYME COMES INTO SCENE

In Providencia stuartii, a Gram-negative species, phylogenetically distant from mycobacteria, a novel class of AAC was identified as AAC(2<sup>0</sup> ), because of its acetylating activity against gentamicin and lack of enzymatic activity against kanamycin A. The gene encoding AAC(2<sup>0</sup> )-Ia was cloned from P. stuartii (Rather et al., 1993) and found to be present in the chromosome of all isolates of this bacteria. In P. stuartii, the expression of the aac(2<sup>0</sup> )- Ia gene was controlled by several transcriptional regulators (Macinga and Rather, 1999), suggesting that this enzyme could play an important role, beyond its contribution to drug resistance (Franklin and Clarke, 2001). In fact, AAC(2<sup>0</sup> )-Ia contributes to

#### BOX 1 | Aminoglycosides: origins, structure, mode of action, resistance and clinical use.

Most aminoglycoside antibiotics are produced by bacterial species of the genus Streptomyces, such as the antitubercular streptomycin that is produced by S. griseus being the first antibiotic identified from bacteria. Other genera producing aminoglycosides are Micromonospora and Bacillus. Many semi-synthetic aminoglycosides, such as amikacin, have also been produced (Mingeot-Leclercq et al., 1999; Magnet and Blanchard, 2005; Shi et al., 2013).

Structurally, aminoglycosides are formed by an aminocyclitol (commonly, 2-deoxystreptamine) with additional amino sugars bound by glycosidic bonds. There are two large families of 2-deoxystreptamine aminoglycosides, those carrying substitutions at positions 4 and 5 of the 2-deoxystreptamine ring (including neomycin, paromomycin, lividomycin, ribostamycin and butirosin) and those being substituted at positions 4 and 6 of the 2-deoxystreptamine (including kanamycin, amikacin, tobramycin, dibekacin, arbekacin, gentamicin, isepamicin, sisomicin, and netilmicin). The carbon atoms in the sugar bound to position 4 of the 2-deoxystreptamine ring are named with primed numbers ('), and those in the sugar bound to positions 5 or 6 of the 2-deoxystreptamine ring are named with double-primed numbers ("). Other aminoglycosides contain aminocyclitols distinct to 2-deoxystreptamine (this is the case of streptomycin or apramycin) or are formed by fused amino sugar rings (i.e., spectinomycin) (Magnet and Blanchard, 2005; Shi et al., 2013). The following figure shows the structure of kanamycin A, an example of 4,6 di-substituted 2-deoxystreptamine aminoglycoside antibiotic.

The bacterial target of aminoglycosides is the 30S small ribosomal subunit, and their global effect is the interference with protein synthesis. In bacterial cells, translation is initiated when the 30S ribosomal subunit binds the Shine-Dalgarno sequence (because of sequence complementarity between the Shine-Dalgarno sequence and the 16S rRNA molecule of the 30S ribosomal subunit), which is normally present in the 5<sup>0</sup> untranslated region of mRNAs. Next, initiation factors and fMet-tRNA will join the complex that finally will recruit the 50S ribosomal subunit in order to start translation. Aminoglycoside binding to the 30S subunit does not affect the binding of mRNA and the large 50S subunit, so translation can proceed. However, aminoglycosides differ in their binding site at the 30S subunit, hence affecting the protein production at different levels. Whereas spectinomycin blocks translocation (hence being bacteriostatic), streptomycin and most 2-deoxystreptamine aminoglycosides lock the ribosome in a conformation that is prone to introducing erroneous aminoacyl-tRNAs. The accumulation of aberrant proteins in the bacteria results in cell death (Davies and Wright, 1997; Magnet and Blanchard, 2005; Shell et al., 2015).

Resistance to aminoglycosides may be due to several mechanisms (Davies and Wright, 1997; Magnet and Blanchard, 2005). Reduced uptake (which can be a consequence of alterations in the composition of bacterial membrane or to metabolic conditions like anaerobiosis) or the action of efflux pumps can lead to limited intracellular concentration of aminoglycosides, hence causing resistance. Mutations in 16S ribosomal RNA or certain ribosomal proteins such as S12 (encoded by rpsL gene) lead to aminoglycoside resistance through target modification; this is also achieved after the action of methyltransferases, which introduce methyl groups in guanine or adenine nucleotides of 16S ribosomal RNA. The presence of aminoglycoside-modifying enzymes is, however, the most prevalent mechanism of aminoglycoside resistance; there are three types of aminoglycoside-modifying enzymes: aminoglycoside N-acetyltransferases (AAC), O-phosphotransferases (APH), and O-nucleotidyltransferases (ANT). Aminoglycoside-modifying enzymes are named by using the aforementioned abbreviations, followed by a number in brackets indicating the site of modification in the aminoglycoside molecule (as explained above), a Roman numeral related with substrate profile, and a lower-case letter for differentiating isoenzymes, i.e., AAC(2<sup>0</sup> )-Ib.

Clinically, due to their oto- and nephrotoxicity and the rising prevalence of resistance, aminoglycosides are commonly reserved as a second line of treatment of serious infections. Due to their low absorption when given orally, aminoglycosides need to be administered through injections. Streptomycin was a first-line drug in the treatment of tuberculosis, and in our days, kanamycin and amikacin are listed as second line drugs against this disease. Spectinomycin is used against Neisseria gonorrhoeae infections (Suay-Garcia and Perez-Gracia, 2018). Pseudomonas aeruginosa infections in cystic fibrosis patients, septicemia, endocarditis and several other infections caused by non-tuberculous mycobacteria, Gram-positive or Gram-negative bacteria can be treated efficiently by using aminoglycosides, either alone or in combinations with other antibacterials such as the beta-lactam antibiotics (Mingeot-Leclercq et al., 1999; Magnet and Blanchard, 2005; Bassetti et al., 2018).

O-acetylation of peptidoglycan affecting cell morphology and the expression of autolysins, and can use acetylated peptidoglycan precursors as donors of acetyl groups, another indication about the role of this enzyme beyond resistance (Payie et al., 1995, 1996; Clarke et al., 1996; Payie and Clarke, 1997).

#### AAC(2<sup>0</sup> ) in Mycobacteria

Given the high AG acetylating activity against gentamicin, tobramycin, netilmicin and its derivatives 2<sup>0</sup> -N-ethyl- and 6 0 -N-ethyl-netilmicin, kanamycin A and kanamycin B found in M. fortuitum (Hull et al., 1984; Ainsa et al., 1996), we launched a molecular approach aimed at characterizing the determinant of AG resistance in this species. A genomic library of M. fortuitum was transformed in M. smegmatis, and characterization of AG resistant clones allowed the identification of a M. fortuitum gene showing sequence similarity to aac(2<sup>0</sup> )-Ia of P. stuartii. The enzyme of M. fortuitum was named AAC(2<sup>0</sup> )-Ib (Ainsa et al., 1996) and was capable of acetylating gentamicin, but not kanamycin A, hence indicating that another AAC enzyme should be present in M. fortuitum. Similarly to P. stuartii, the aac(2<sup>0</sup> )-Ib gene was found in all strains of M. fortuitum regardless of the phenotype of AG resistance, suggesting other roles for the AAC(2<sup>0</sup> )-Ib in this species. Further studies (done by database searching or by southern blot analysis using the probe of aac(2<sup>0</sup> )-Ib gene) demonstrated the presence of aac(2<sup>0</sup> )- I genes in other mycobacterial species, including M. tuberculosis (the major pathogenic species in this genus), Mycobacterium leprae, and M. smegmatis, indicating thus that the presence

of aac(2<sup>0</sup> )-I genes in mycobacteria could be universal (Ainsa et al., 1997). Interestingly, the expression of the aac(2<sup>0</sup> )-Id gene in M. smegmatis was driven from two promoters, and the strongest one produced a leaderless transcript having a GTG translation start codon at its 5<sup>0</sup> end (Mick et al., 2008). Leaderless transcripts are those in which the transcription start site coincides with the translation start codon; although representing a rather unusual feature in the model organism E. coli, they are quite common in mycobacteria, where 25% of all transcripts are predicted to be leaderless (Shell et al., 2015); in fact, eis gene of M. tuberculosis (see the section "The Eis Protein Becomes a Novel Aminoglycoside Acetyltransferase," below) is transcribed also as a leaderless mRNA (Zaunbrecher et al., 2009). In leaderless mRNAs, there is no Shine-Dalgarno sequence, and 70S ribosomes bind directly to the 5<sup>0</sup> -end of the mRNA in order to initiate translation (Shell et al., 2015).

#### New Roles for AAC(2<sup>0</sup> ) in Mycobacteria

Two major evidences supported that in mycobacteria, AAC(2<sup>0</sup> )- I enzymes could have additional roles other than acetylation of AGs containing 2<sup>0</sup> -amino group: first, the capability of amino sugars and diverse acyl-CoA molecules to inhibit the acetylating activity of mycobacterial crude extracts (Udou et al., 1989); and second, the implication of AAC(2<sup>0</sup> )-Ia (an enzyme of the same class) in acetylation of cell wall substrates in P. stuartii. In a series of laboratory mutants of P. stuartii expressing aac(2<sup>0</sup> )-Ia gene at different levels, it was found that the extent of peptidoglycan acetylation correlated with the activity of AAC(2<sup>0</sup> )-Ia, hence suggesting a partial contribution of this enzyme (along with other enzymes) to peptidoglycan acetylation. Many bacterial pathogens acetylate their peptidoglycan as a way to resist the action of muramidase enzymes. Under in vitro conditions, AAC(2<sup>0</sup> )-Ia was able to acetylate tobramycin having acetylated peptidoglycan as donor of acetyl groups (Payie et al., 1995, 1996; Payie and Clarke, 1997).

We investigated the hypothesis of mycobacterial AAC(2<sup>0</sup> )-I enzymes having also a role in cell wall metabolism, and a gene knock-out mutant of M. smegmatis deleted in aac(2<sup>0</sup> )-Id gene was constructed. This mutant (named EP-10) defective in AAC(2<sup>0</sup> )- I activity was more susceptible to gentamicin, tobramycin, dibekacin, and netilmicin than the parental wild-type strain, and crude extracts of M. smegmatis EP-10 failed to acetylate 2<sup>0</sup> -amino group containing AGs, whereas wild-type strains of M. smegmatis readily acetylated such AGs (Ainsa et al., 1997; Maurer et al., 2015). M. smegmatis EP-10 was also twofold more susceptible to lysozyme, a feature that has been associated with the extent of peptidoglycan acetylation (Ainsa et al., 1997). Hence, we concluded that in M. smegmatis, AAC(2<sup>0</sup> )-Id enzyme could also contribute to acetylation of peptidoglycan, since a less extensively acetylated peptidoglycan in the knock-out strain would be more susceptible to lysozyme degradation affecting cell viability.

#### Biochemical Analysis of Mycobacterial AAC(2<sup>0</sup> )-I Enzymes

The presence of an AG (2<sup>0</sup> )-N-acetyltransferase gene in the genome of M. tuberculosis was intriguing. In this species, AAC activity had never been reported (Mitsuhashi et al., 1977) and the expression of the aac(2<sup>0</sup> )-Ic gene [annotated as Rv1258c gene in the M. tuberculosis H37Rv genome (Cole et al., 1998)] in the surrogate host M. smegmatis could not be associated with any change in the levels of susceptibility to AGs (Ainsa et al., 1997). We constructed a knock-out mutant of M. tuberculosis H37Rv deleted in the aac(2<sup>0</sup> )-Ic gene and observed that the mutant (named M. tuberculosis B1) was twofold more susceptible than the original wild-type strain to AGs containing a 2<sup>0</sup> -amino group such as gentamicin, tobramycin and dibekacin, and fourfold more susceptible to 6<sup>0</sup> -N-ethyl-netilmicin. This indicated that the aac(2<sup>0</sup> )-Ic gene was being expressed in M. tuberculosis although at a very low level, and that the AAC(2<sup>0</sup> )-Ic enzyme in M. tuberculosis would acetylate all these four AGs in the wild type strain; hence, acetylated AGs would bind less efficiently to the ribosome, and the AAC(2<sup>0</sup> )-Ic enzyme would contribute to basal AG resistance in this species.

In order to find the physiological role of this enzyme in M. tuberculosis, recombinant E. coli-produced AAC(2<sup>0</sup> )-Ic enzyme from M. tuberculosis was studied and found to efficiently acetylate AG antibiotics containing an amino group at the 2<sup>0</sup> position (as expected; for example, K<sup>m</sup> for kanamycin B was 1.4 µM), and surprisingly, it was also capable of acetylating (although to a much lesser extent) other AGs such as kanamycin A (K<sup>m</sup> 320 µM) and amikacin (K<sup>m</sup> 968 µM) that have a hydroxyl group in the 2<sup>0</sup> position, hence suggesting this enzyme to be capable of both N- and O-acetylation (Hegde et al., 2001; Draker et al., 2003). This residual activity of AAC(2<sup>0</sup> )- Ic against kanamycin A and amikacin does not affect their bactericidal activity, and these two antibiotics are used as second line drugs against drug resistant M. tuberculosis (World Health Organization [WHO], 2010). The activity of M. tuberculosis AAC(2<sup>0</sup> )-Ic is dependent on metal ions, being inhibited by Cu2<sup>+</sup> and Au3<sup>+</sup> (Li et al., 2015).

Another study consisted in binding covalently the AGs kanamycin A, tobramycin, neamine and neomycin B to an agarose matrix in order to quantify the extent of AG acetylation by AAC enzymes and their subsequent ability to bind an artificial probe mimicking the A-site of the ribosome (Barrett et al., 2008). In such experimental model system, the AGs acetylated by M. tuberculosis AAC(2<sup>0</sup> )-Ic enzyme (only tobramycin, neamine and neomycin B) maintained their binding affinity with the probe mimicking A-site of the ribosome at detectable levels, maybe because the percent of AG acetylation by AAC(2<sup>0</sup> )-Ic was low. In contrast, these AGs were very efficiently acetylated by E. coli AAC(3) (at a different amino group) and had readily lost their affinity for binding this artificial probe. These experiments suggested that subtle differences in the structure of modified AGs (i.e., acetylation at the amino group in 2<sup>0</sup> or 3 position) are sufficient to drastically affect their capability of binding to the A-site probe, and this would be expected to correlate with their activity as ribosome inhibitors. In consequence, in mycobacteria, AAC(2<sup>0</sup> )-I enzymes would not play a major role in resistance to these drugs (Barrett et al., 2008). High resolution crystal structures of AAC(2<sup>0</sup> )-Ic complexed with AGs demonstrated that this enzyme is a member of the GNC5 acetyltransferase superfamily and suggested a role in the synthesis of mycothiol,

a metabolite that has a key role in regulating redox potential in mycobacteria (Vetting et al., 2002, 2003).

In agreement with the preferential N-acetylating activity over the O-acetylating activity that was found in the M. tuberculosis enzyme (Hegde et al., 2001; Draker et al., 2003), M. abscessus clinical isolates were found to be more susceptible to AGs containing a hydroxyl group at the 2<sup>0</sup> position (such as amikacin and kanamycin A) than to AGs with a 2<sup>0</sup> -amino group (such as tobramycin, dibekacin and kanamycin B), as the latter group would be substrates of M. abscessus AAC(2<sup>0</sup> )-I enzyme (Maurer et al., 2015). In fact, crude extracts of M. abscessus efficiently acetylated kanamycin B, whereas kanamycin A was not acetylated at detectable levels. In this species, deletion of the aac(2<sup>0</sup> )-I gene resulted in increased susceptibility to kanamycin B, tobramycin, dibekacin and gentamicin C (all of them containing a 2<sup>0</sup> -amino group) (Rominski et al., 2017). These two reports demonstrate that in M. abscessus, the presence of an AAC(2<sup>0</sup> )-I enzyme contributes to decreased innate susceptibility to AGs containing a 2<sup>0</sup> -amino group (Luthra et al., 2018).

#### Becoming a Drug Target: Developing Inhibitors of AAC(2<sup>0</sup> )-I

The interest in developing inhibitors against AAC(2<sup>0</sup> )-I enzymes came from a study in the non-tuberculous species M. abscessus (Maurer et al., 2014). It was found that AGs such as amikacin, gentamicin or tobramycin, which are normally bactericidal against E. coli, do not have such activity against M. abscessus or M. smegmatis. However, disruption of the chromosomally encoded aac(2<sup>0</sup> )-I gene in these species restored the bactericidal activity of these AGs (Maurer et al., 2014). Given that AGs are used as second-line drugs in treatment of multidrug resistant (MDR) tuberculosis infections, and also in the treatment of other infections caused by non-tuberculous mycobacteria, the possibility of developing compounds that could enhance bactericidal activity of current antimycobacterial treatments became an interesting approach.

To date, the only putative inhibitor of M. tuberculosis AAC(2<sup>0</sup> )-Ic is andrographolide, a natural product that was identified in methanolic extracts of a plant that were capable of inhibiting growth of M. tuberculosis strains (Prabu et al., 2015). In silico analysis predicted that this compound could potentially bind with a high affinity the AAC(2<sup>0</sup> )-Ic enzyme, as well as isocitrate lyase (a metabolic enzyme of the glyoxylate shunt, involved in persistence and virulence of M. tuberculosis) and other M. tuberculosis proteins (Prabu et al., 2015). However, the specificity of this binding and the ability to really inhibit such putative target proteins were not tested. Given that the gene encoding AAC(2<sup>0</sup> )-Ic is not essential in M. tuberculosis, a direct link between AAC(2<sup>0</sup> )-Ic inhibition and bacterial growth inhibition could be discarded. Hence, the ability of plant extracts containing andrographolide to inhibit growth of M. tuberculosis could be due to the presence of additional compounds in the extract, or to multiple effects on M. tuberculosis cells. Other in silico analysis revealed that AAC(2<sup>0</sup> )-Ic enzyme from M. tuberculosis could interact with ten other proteins (including a protein of a putative RND-like efflux pump), suggesting that inhibition of AAC(2<sup>0</sup> )-Ic could also impact many other metabolic processes, hence conferring this enzyme with a relevant role in drug discovery of antituberculosis agents (Joshi et al., 2013).

#### THE Eis PROTEIN BECOMES A NOVEL AMINOGLYCOSIDE ACETYLTRANSFERASE

Investigation of M. tuberculosis virulence factors lead to the identification of a protein, that was required for infecting and survival in human macrophages; this protein was named Eis for enhanced intracellular survival (**Box 2**) (Wei et al., 2000). Bioinformatic analysis revealed that Eis protein of M. tuberculosis was an acetyltransferase of the GCN-5 family (Samuel et al., 2007).

Later on, the analysis of kanamycin resistant M. tuberculosis laboratory and clinical strains revealed mutations in the −10 and −35 regions of the eis gene promoter, which resulted in increased levels of eis mRNA and Eis protein. These mutations were related to low-level resistance to kanamycin (MIC 25 µg/ml), but not to amikacin (MIC < 4 µg/ml), whereas 16S rRNA mutations confer higher levels of resistance to kanamycin (MIC > 80 µg/ml) and frequently cross-resistance to amikacin. These eis mutants also displayed increased levels of AAC activity, hence demonstrating that Eis was a novel class of AAC, highly divergent from all other previously known AACs. Eis is capable of acetylating kanamycin more efficiently than amikacin, and streptomycin was not found to be a substrate of Eis (Zaunbrecher et al., 2009). From then on, detection of mutations in eis promoter has become a relevant assay in clinical microbiology laboratories for determining susceptibility to kanamycin (Georghiou et al., 2012); the diagnostic and clinical implications of these tests are beyond the scope of this review.

Formation of stable hexamers by Eis is required for its AAC activity (Ganaie et al., 2011; Anand and Sharma, 2018), which can acetylate multiple amine groups of different AG antibiotics, including netilmicin, sisomicin, neamine, ribostamycin, paromomycin, neomycin B, kanamycin, amikacin, tobramycin and hygromycin, resulting in mono-, di-, tri, and tetraacetylated products (Chen et al., 2011; Houghton et al., 2013a), being able to use not only acetyl-CoA but also other acyl-CoA derivatives (Chen et al., 2012a). The Eis protein works by a randomsequential bisubstrate mechanism of acetylation (Tsodikov et al., 2014). Interestingly, Eis is also able to acetylate capreomycin, a polypeptide second-line antituberculosis drug commonly used in the treatment of MDR tuberculosis infections (Reeves et al., 2013). Several metal ions such as Au3+, Cd2<sup>+</sup> and Zn2<sup>+</sup> inhibited Eis activity in vitro (Li et al., 2015).

Other mycobacterial species such as M. smegmatis and M. abscessus have ortholog (and even paralog) eis genes, although the Eis proteins have distinct biochemical features and impact on AG susceptibility in comparison with Eis of M. tuberculosis (Chen et al., 2012b; Rominski et al., 2017). For example, M. abscessus has two eis genes, and the deletion of one (but not the other one) resulted in altered AG susceptibility (Rominski et al., 2017; Luthra et al., 2018). Consistently with these findings, mutational

BOX 2 | Discovery and characterization of enhanced intracellular survival (Eis) protein of M. tuberculosis.

A genome wide investigation of the ability of M. tuberculosis for infecting macrophages resulted in the identification of a coding sequence that, once cloned in the non-pathogenic M. smegmatis species, conferred capacity for infecting the human macrophage-like cell line U937. This gene was named eis (Rv2416c in the M. tuberculosis genome; Cole et al., 1998) for enhanced intracellular survival (Wei et al., 2000), and was detected only in pathogenic species of mycobacteria. The promoter of the eis gene is similar to consensus E. coli sigma-70 dependent promoters (Roberts et al., 2004) and it is recognized by M. tuberculosis SigA sigma factor (Wu et al., 2009). The Eis protein was found to be mostly hydrophilic, but having a hydrophobic N-terminal end, so that it could be found in the cytosolic but also in other cell fractions such as the membrane, cell wall or among the secreted proteins of M. tuberculosis (Dahl et al., 2001). In fact, antibodies against Eis could detect this protein in the sera of tuberculosis patients (Dahl et al., 2001) and in the cytoplasm of infected macrophages (Samuel et al., 2007). Also, it was found that Eis protein, directly added to cultures of human monocytes, modulated the secretion of pro-inflammatory cytokines in a similar way to that found in M. tuberculosis infected cells (Samuel et al., 2007), hence suggesting a role of Eis as an effector protein. Further studies demonstrated that Eis inhibits the extra-cellular signal-regulated kinase 1/2 (ERK1/2) and JAK pathways, and in consequence it inhibited the production of TNF-alpha and IL-4, and stimulated the production of IFN-gamma and IL-10 (Lella and Sharma, 2007). The effect of Eis in increasing production of IL-10 was found to be related to Eis-mediated acetylation of histone H3, which binds the promoter of the human IL-10 gene (Duan et al., 2016). Hence, by disturbing cross-regulation of T-cells and impairing TH1 and TH2 response, Eis could mediate M. tuberculosis pathogenicity (Lella and Sharma, 2007). In fact, an isolate of the Beijing family (which are more transmissible and virulent than other M. tuberculosis genetic lineages (Hanekom et al., 2011) was found to contain elevated levels of Eis protein, mediated by increased expression of SigA (Wu et al., 2009). Other key factors in host immune response to tuberculosis are also mediated by Eis, which increased production of ROS and consequently modulated processes such as autophagy, inflammation, and cell death (Shin et al., 2010). These processes are started by Eis-dependent acetylation of dual-specificity phosphatase-16 (DUSP16)-mitogen-activated protein kinase phosphatase-7 (MKP-7), which dephosphorylates the JNK protein leading to its inactivation (Kim et al., 2012; Yoon et al., 2013). Other studies have revealed the activity of Eis for acetylating arylalkylamines such as histamine, octopamine, or tyramine, suggesting novel roles for this protein in M. tuberculosis pathogenicity (Pan et al., 2018).

changes in the amino acid residues lining the substrate binding site of M. tuberculosis Eis altered its substrate specificity (Jennings et al., 2013).

It is important to note that in M. tuberculosis, transcription of the eis gene is activated by the regulator WhiB7 (Reeves et al., 2013). Mutations in the promoter of whiB7 gene that led to increase in the mRNA of this gene resulted in increased expression of eis gene, along with other genes such as rv1258c (encoding the Tap efflux pump; Ainsa et al., 1998), hence resulting in cross resistance to several drugs including kanamycin (mediated by Eis protein) or streptomycin (mediated by Tap efflux pump). Similarly, in M. abscessus, WhiB7 controlled the expression of one of the two eis genes in this species and also that of the erm(41) gene, which encodes a ribosomal methyltransferase that by altering target structure is associated with resistance to macrolide antibiotics (Pryjma et al., 2017; Luthra et al., 2018). Subinhibitory concentrations of clarithromycin induced the whiB7 gene and consequently decreased Eis-mediated susceptibility to AGs, such as amikacin that is currently used in the treatment of M. abscessus infections (Pryjma et al., 2017).

#### Aminoglycosides and Beyond. . .

The unusual properties of Eis acetyltransferase include its capability for acetylating peptides and proteins (Kim et al., 2012; Houghton et al., 2013b; Yoon et al., 2013), in contrast with other AACs. The M. tuberculosis nucleoid-associated protein HU (encoded by Rv2986c gene) can be acetylated by Eis on multiple lysine residues, hence decreasing its ability to interact with DNA, and altering its DNA compactation activity (Ghosh et al., 2016; Green et al., 2018). Overexpression of Eis led to a hyperacetylation of HU protein, and consequently, to a decompactation of the genome (Ghosh et al., 2016). The reverse effect (condensation of relaxed DNA) could be reached through the deacetylation of HU protein, which is mediated by a Sir2 family protein from M. tuberculosis encoded by the Rv1155c gene (Anand et al., 2017; Green et al., 2018). Controlling the architecture of DNA is a key process in any bacteria, and so, the HU protein is essential for M. tuberculosis. In fact, inhibitors of HU have been discovered (Bhowmick et al., 2014), which could act in synergy with potential inhibitors of Eis, as described in the next section.

#### Finding Inhibitors of Eis Protein

The crystal structure of M. tuberculosis Eis protein was determined by several groups (Chen et al., 2011; Kim et al., 2012), which has been useful for determining docking properties of potential inhibitory compounds; also, its comparison with the crystal structure of M. smegmatis Eis protein revealed several distinct structural features that may account for the biochemical and substrate differences between the two proteins (Kim et al., 2014). A first screening of potential inhibitors of Eis from M. tuberculosis resulted in the identification of 25 molecules (including the antiseptic chlorhexidine) that inhibited Eis with IC<sup>50</sup> values in the low micromolar range. In addition, this inhibition was specific to the M. tuberculosis Eis protein, since these molecules could not inhibit significantly AACs from AAC(2<sup>0</sup> ), AAC(3), and AAC(6<sup>0</sup> ) families (Green et al., 2012), nor Eis protein from Bacillus anthracis (Green et al., 2015a). Later studies revealed the presence of eis-like genes in many pathogenic and non-pathogenic bacteria (remarkably, many mycobacterial species have two or even three paralogs of the eis gene), and chlorhexidine was capable of inhibiting (to different levels) all Eis proteins that were capable of acetylating AGs (Green et al., 2015b).

A second screening of a larger collection of small-molecule compounds resulted in the identification of several families of compounds capable of inhibiting Eis activity. These contained diverse chemical scaffolds (Garzan et al., 2016a,b, 2017; Willby et al., 2016; Ngo et al., 2018). Besides, these inhibitors were highly selective for Eis, and did not inhibited AACs from other families (Garzan et al., 2016b). More importantly, as these inhibitors bound in the AG pocket of the Eis protein, they were able to reverse kanamycin resistance of a M. tuberculosis isolate (Garzan et al., 2016b; Willby et al., 2016; Ngo et al., 2018).

Some of these inhibitors, such as those based on a pyrrolo[1,5 a]pyrazine scaffold, also lacked any toxicity on mammalian cell lines (Garzan et al., 2017).

#### OTHER AMINOGLYCOSIDE-MODIFYING ENZYMES IN MYCOBACTERIA

Genome-wide analysis of M. tuberculosis genome identified only one other potential AAC (encoded by the Rv1347c gene), although such enzymatic activity could not be detected on the recombinant protein (Draker et al., 2003). Later studies related the product of the Rv1347c gene with a role in the synthesis of mycobactin, the mycobacterial siderophore (Card et al., 2005).

Leaving apart AACs, only a few reports of other classes of aminoglycoside-modifying enzymes have been done in mycobacteria. An APH enzyme of the APH(300) family, conferring resistance to the AG streptomycin only, has been characterized in M. fortuitum (Ramon-Garcia et al., 2006) and M. abscessus (Dal Molin et al., 2018; Luthra et al., 2018); the latter species encodes up to 11 additional putative APH enzymes (Ripoll et al., 2009). Furthermore, a putative APH of the APH(3<sup>0</sup> ) class, encoded by the Rv3168 gene of M. tuberculosis, was identified and expressed as a recombinant enzyme in E. coli, being related to kanamycin phosphotransferase activity (Ahn and Kim, 2013).

## BACK TO THE START POINT: Eis IN M. fortuitum

We started this review referring to previous work that has shown that crude extracts of M. fortuitum harbored AAC activity having gentamicin, tobramycin, netilmicin and its derivatives 2<sup>0</sup> -Nethyl- and 6<sup>0</sup> -N-ethyl-netilmicin, kanamycin A and kanamycin B as substrates. So far, the only AG acetyltransferase identified in this species has been AAC(2<sup>0</sup> )-Ib (Ainsa et al., 1996), which cannot explain the acetyltransferase activity against kanamycin A and 2<sup>0</sup> -N-ethyl-netilmicin detected in M. fortuitum crude extracts. Later studies characterized Eis AAC in diverse mycobacterial species, but no report were been done on a putative Eis protein in M. fortuitum. In view of the universal presence of Eis proteins in mycobacteria, and its activity as AAC, we hypothesized that M. fortuitum could also have a putative Eis protein that would be responsible for the acetyltransferase activity against kanamycin A and 2<sup>0</sup> -N-ethyl-netilmicin detected in crude extracts of this species, as it was reported earlier (Hull et al., 1984; Ainsa et al., 1996).

Thus, we first ascertained the existence of an eis gene in the genome of M. fortuitum (Ho et al., 2012), which presented a 94% of identity with respect to the one reported in M. tuberculosis; the sequence of both Eis proteins from M. tuberculosis and M. fortuitum also presented high levels of identity (**Figure 1**). We designed two oligonucleotides for amplifying specifically a DNA fragment from M. fortuitum genome containing eis gene. This DNA fragment was subsequently cloned in pMV261 vector (Stover et al., 1991), which expresses genes constitutively under the control of the hsp60 gene promoter, resulting in plasmid pFS2. Given that this plasmid still contains the Tn903-derived aminoglycoside-3<sup>0</sup> -phosphotransferase (aph) gene (present in the original pMV261 cloning vector) conferring kanamycin resistance as selection marker, we anticipated that M. smegmatis strains harboring pMV261 vector or its derivatives would be intrinsically resistant to kanamycin A; this would prevent from determining whether this antibiotic is a substrate of the M. fortuitum Eis protein. Additionally, the selection marker could have cross-effect with other AGs and thus interfering with the resistance phenotype conferred by eis gene. To circumvent these problems, we generated a derivative of pFS2 through the disruption of the aph gene with the ampicillin resistance cassette (bla gene) from pGEM <sup>R</sup> -T easy (Promega). The resistance to 2 0 -N-ethylnetilmicin conferred by eis gene, as demonstrated for the parental plasmid pFS2 (see **Table 1**) was used as resistance marker for transformant selection and plasmid maintenance. This process resulted in plasmid pEAC which contains the M. fortuitum eis gene and no other determinant of AG resistance.

The three plasmids [original empty vector pMV261 as a control, and the two plasmids (pFS2 and pEAC) containing M. fortuitum eis gene] were introduced in M. smegmatis mc<sup>2</sup> 155 in order to over-express the ortholog eis gene from M. fortuitum and to elucidate its hypothetical implication in AG susceptibility. The antibiotic susceptibility assay was made, based on a double dilution protocol with the addition of resazurin dye (Palomino et al., 2002).

We observed that plasmids pFS2 and pEAC produced detectable changes in AG susceptibility of M. smegmatis mc<sup>2</sup> 155, which can be attributed to the expression of the plasmid-borne M. fortuitum eis gene. The major shift in the MICs was detected for 2<sup>0</sup> -N-ethylnetilmicin (**Table 1**), since the MIC increased from 3.12 to 6.25 µg/ml in the control strains to 50–100 µg/ml in the strains containing M. fortuitum Eis, accounting for a 8- to 32-fold increase; similar changes were observed for 6<sup>0</sup> -N-ethylnetilmicin (8-fold increase in the MIC). A moderate decrease in the susceptibility to kanamycin A (fivefold), hygromycin (twofold to fourfold) and gentamicin (twofold to fourfold) was also observed. Finally, slight changes (twofold) in the MICs were detected for kanamycin B and capreomycin; this finding is consistent with previous reports on the activity of M. tuberculosis Eis against capreomycin (Reeves et al., 2013). The levels of susceptibility to the AGs amikacin, streptomycin and spectinomycin, and to other non-AG compounds tested (isoniazid, rifampicin, ethambutol, ciprofloxacin, tetracycline, chloramphenicol) were not altered significantly by the presence of the plasmid-encoded eis gene (**Table 1** and data not shown), suggesting that either these antimicrobials are not substrates of the Eis enzyme or the corresponding acetylations (if any) might just not affect antibacterial activity.

## CONCLUDING REMARKS

The presence and activity of AAC(2<sup>0</sup> )-I and Eis AACs in mycobacteria have clearly demonstrated that their primary role is little related with susceptibility to AGs.


FIGURE 1 | Comparison of the amino acid sequences of Eis proteins from M. fortuitum and M. tuberculosis. Symbols under the sequence alignment: Asterisks (<sup>∗</sup> ) indicate positions with identical amino acid in both proteins. Colons (:) indicate positions which have amino acids with strongly similar properties. Periods (.) indicate positions which have amino acids with weakly similar properties.



The values separated by a dash indicate that growth was detected in both concentrations. The range of antibiotic concentrations spanned from 0.78 µg/ml to 100 µg/ml, and from 0.25 µg/ml to 125 µg/ml; a twofold difference in the MIC was not considered as significant. aph: aminoglycoside 3<sup>0</sup> -phosphotransferase from Tn903; bla: beta-lactamase.

In the case of AAC(2<sup>0</sup> )-I enzymes, its presence in phylogenetically distant genera as Providencia and Mycobacterium remains to be an evolutionary mystery. In contrast with this restricted distribution of AAC(2<sup>0</sup> )-I enzymes among bacteria, Eis enzymes seem to be more widely distributed, being present even in non-pathogenic and environmental species, which suggest a general function of Eis-like enzymes in bacterial metabolism, and virtually excludes any potential selection of eis genes due to the use of AGs, or its horizontal transfer from other species.

It is clear that in mycobacteria ribosomal modifications constitute the major mechanism of AG resistance, given that only one or two copies of ribosomal RNA operons are present in these species, hence making likely the acquisition of mutations conferring high levels of AG resistance. Then, AAC(2<sup>0</sup> )-I enzymes only contribute modestly to innate low level susceptibility to AGs, and despite other roles have been suggested in the literature for mycobacterial AAC(2<sup>0</sup> )-I enzymes, their relevance as potential drug targets is still modest, especially in comparison with Eis acetyltransferase. The contribution of Eis acetyltransferase to virulence of M. tuberculosis, and the finding that Eis is related with resistance to kanamycin (a second line drug for the treatment of tuberculosis) in clinical isolates has greatly attracted the attention and promoted the interest in developing Eis inhibitors. In some way, Eis inhibitors would fall into the class of anti-virulence and antiresistance mechanisms compounds, which is a trending topic

in the age of antimicrobial resistance. Globally, antimicrobial resistance is a major public health threat, and multi and extensively drug resistant (MDR, XDR) tuberculosis is a case of particular concern, so progress and major advances that can be expected from the coming years will be greatly welcomed.

#### AUTHOR CONTRIBUTIONS

JA, LR, and CM contributed to the conception and design of the study. FS-G, EA-C, AL, LR, and EP-H carried out the experimental work. FS-G, EA-C, and JA wrote the manuscript. All authors contributed to manuscript revision, read and approved the submitted version.

#### REFERENCES


#### FUNDING

FS-G and EA-C are recipients of FPU fellowships from the Spanish Ministry of Economy and Competitivity. JA acknowledges funding from the European Comission [More Medicines for Tuberculosis (MM4TB) grant 260872] and the Spanish Ministry of Economy and Competitivity (grants SAF-2013-48971-C2-2-R and SAF2017-84839- C2-1-R).

#### ACKNOWLEDGMENTS

Dessi Marinova was acknowledged for critical reading of the manuscript.

resistance in XDR-TB. Proc. Natl. Acad. Sci. U.S.A. 108, 9804–9808. doi: 10. 1073/pnas.1105379108


as kanamycin adjuvants against drug-resistant M. tuberculosis. ACS Med. Chem. Lett. 7, 1219–1221. doi: 10.1021/acsmedchemlett.6b00261



**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 © 2019 Sanz-García, Anoz-Carbonell, Pérez-Herrán, Martín, Lucía, Rodrigues and Aínsa. 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) and the copyright owner(s) 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.

# Insights on *Mycobacterium leprae* Efflux Pumps and Their Implications in Drug Resistance and Virulence

Diana Machado1,2, Emmanuel Lecorche3,4,5, Faiza Mougari 3,4,5, Emmanuelle Cambau2,3,4,5 \* and Miguel Viveiros 1,2 \*

<sup>1</sup> Global Health and Tropical Medicine, Instituto de Higiene e Medicina Tropical, Universidade Nova de Lisboa, Lisbon, Portugal, <sup>2</sup> Study Group for Mycobacterial Infections (ESGMYC), European Society for Clinical Microbiology and Infectious Diseases (ESCMID), Basel, Switzerland, <sup>3</sup> Université Paris Diderot, INSERM IAME UMR1137, Sorbonne Paris Cité, Paris, France, <sup>4</sup> APHP, Groupe Hospitalier Lariboisière Fernand-Widal, Laboratoire de Bacteriologie, Paris, France, <sup>5</sup> Centre National de Référence des Mycobactéries et Résistance des Mycobactéries aux Antituberculeux, Paris, France

#### *Edited by:*

Silvia Buroni, University of Pavia, Italy

#### *Reviewed by:*

Katiany Rizzieri Caleffi Ferracioli, Universidade Estadual de Maringá, Brazil Yusuf Akhter, Babasaheb Bhimrao Ambedkar University, India

#### *\*Correspondence:*

Emmanuelle Cambau emmanuelle.cambau@aphp.fr Miguel Viveiros mviveiros@ihmt.unl.pt

#### *Specialty section:*

This article was submitted to Evolutionary and Genomic Microbiology, a section of the journal Frontiers in Microbiology

*Received:* 18 July 2018 *Accepted:* 28 November 2018 *Published:* 13 December 2018

#### *Citation:*

Machado D, Lecorche E, Mougari F, Cambau E and Viveiros M (2018) Insights on Mycobacterium leprae Efflux Pumps and Their Implications in Drug Resistance and Virulence. Front. Microbiol. 9:3072. doi: 10.3389/fmicb.2018.03072 Drug resistance in Mycobacterium leprae is assumed to be due to genetic alterations in the drug targets and reduced cell wall permeability. However, as observed in Mycobacterium tuberculosis, drug resistance may also result from the overactivity of efflux systems, which is mostly unexplored. In this perspective, we discuss known efflux pumps involved in M. tuberculosis drug resistance and virulence and investigate similar regions in the genome of M. leprae. In silico analysis reveals that the major M. tuberculosis efflux pumps known to be associated with drug resistance and virulence have been retained during the reductive evolutionary process that M. leprae underwent, e.g., RND superfamily, the ABC transporter BacA, and the MFS P55. However, some are absent (DinF, MATE) while others are derepressed (Mmr, SMR) in M. leprae reflecting the specific environment where M. leprae may live. The occurrence of several multidrug resistance efflux transporters shared between M. leprae and M. tuberculosis reveals potential implications in drug resistance and virulence. The conservation of the described efflux systems in M. leprae upon genome reduction indicates that these systems are potentially required for its intracellular survival and lifestyle. They potentially are involved in M. leprae drug resistance, which could hamper leprosy treatment success. Studying M. leprae efflux pumps as new drug targets is useful for future leprosy therapeutics, enhancing the global efforts to eradicate endemic leprosy, and prevent the emergence of drug resistance in afflicted countries.

Keywords: antimicrobial resistance, efflux pumps, leprosy, mycobacteria, tuberculosis, virulence

### INTRODUCTION

Leprosy and tuberculosis are public health threatening infectious diseases with similar problems of ongoing human-to-human transmission, inherent drug resistance to several antimicrobial agents, propensity to develop resistance to antimycobacterial drugs, and virulence (Singh et al., 2016; Dheda et al., 2017). Whilst there is extensive knowledge about the mechanisms of M. tuberculosis drug resistance, less is known about the mechanisms by which M. leprae develops drug resistance. M. leprae is an obligate intracellular pathogen and one of the few known microorganisms that still cannot be cultured in vitro Machado et al. Efflux Pumps in M. leprae

which have been hindering the study of the mechanisms of drug resistance by biochemical and functional studies. The overexpression of multidrug (MDR) efflux pump genes is a common mechanism of antimicrobial resistance in M. tuberculosis (da Silva et al., 2016; Machado et al., 2017). Likewise, efflux pumps certainly contribute to drug resistance in M. leprae, which is mostly unexplored.

Efflux pumps are one of the most widespread resistance determinants in bacteria. Usually, they are chromosomally encoded and are greatly conserved at both gene and protein level across bacterial species. More than 50 putative efflux pumps have been associated with the transport of several drugs in M. tuberculosis (De Rossi et al., 2006; Louw et al., 2009; Kapopoulou et al., 2011; Black et al., 2014; da Silva et al., 2016). Although they are mostly known due to their role in the efflux of antimicrobials, efflux pumps are mainly involved in physiological processes such as cell-to-cell communication, bacterial virulence, cellular homeostasis, detoxification of intracellular metabolites, and intracellular signal trafficking (De Rossi et al., 2006; Martinez et al., 2009; Viveiros et al., 2012; Black et al., 2014; da Silva et al., 2016; Li et al., 2016; Sandhu and Akhter, 2018). Recently, it was shown that the loss of the efflux pump AcrAB in Salmonella enterica serovar Typhymurium reduces virulence leading to the accumulation of noxious molecules inside the bacteria reducing the bacterial factors required for infection (Wang-Kan et al., 2017). From a biological point of view, drug resistance and virulence are required for pathogen survival. In normal conditions, the expression of these systems is tightly downregulated by specific transcriptional regulators (Grkovic et al., 2001) and their overexpression is achieved only in the presence of specific stressors capable of binding to the transcriptional regulators. The induction of efflux systems in the presence of inducers such as antimicrobials or host factors during infection promote a low-level resistance phenotype that allows the bacteria to survive during prolonged periods in the presence of drugs contributing for the development and stabilization of resistant phenotypes (Machado et al., 2012; Schmalstieg et al., 2012).

In this perspective, we compared in silico M. tuberculosis efflux pumps involved in drug resistance and virulence with those of M. leprae investigating their possible involvement in antimicrobial resistance and virulence in M. leprae.

#### REDUCTIVE EVOLUTION

The genome of the non-pathogenic Mycobacterium smegmatis mc<sup>2</sup> 155 has 7 Mb in size; the genomes of the pathogenic M. tuberculosis and M. leprae are much smaller in length and for M. leprae this is even more dramatic with almost 2000 genes lost in comparison with M. tuberculosis. Compared to non-pathogenic mycobacteria, M. tuberculosis and M. leprae evolved by extensive reductive evolution suggesting that pathogenic mycobacteria evolved toward pathogenicity by the loss of genetic material as the result of niche adaptation. Contrary to M. tuberculosis, M. leprae is an obligate intracellular pathogen. Adaptation to a permanent pathogenic lifestyle in constant association with the host led to gene loss toward a minimal gene set, such as those coding for metabolism and respiration, needed for a successful obligate intracellular parasitism (Moran, 2002; Scollard et al., 2006) and limited capacity to survive extracellularly (Cole et al., 2001; Eiglmeier et al., 2001). Of interest is the fact that the G+C content is lower in M. leprae pseudogenes (56.5%) than in its active ORFs (60.1%) (Cole et al., 2001). Changes in G+C content during the path of evolution may confer an advantage in response to environmental changes (Mann and Chen, 2010). Free-living organisms have an average G+C content higher than obligatory pathogens and symbionts. The shift toward lower G+C contents and smaller genomes in obligate pathogenic mycobacteria seems to occur in response to environmental adaptation where they encounter low selective pressure (Mann and Chen, 2010). In this context some genes became inactivated, as they are not required in these highly specialized niches meeting the theoretical principles of Morris, Lenski, and Zinser's Black Queen Hypothesis for the symbiotic reductive genome evolution of microorganisms (Morris et al., 2012), now applied to a bacterium and his long-lasting and almost exclusive host—the human being. In this case, M. leprae relies on his host functions to live efficiently, losing burdensome genes for functions it does not have to perform for itself (Morris et al., 2012). Nevertheless, it is not clear why M. leprae maintains such high number of pseudogenes in the genome. It has been hypothesized that the maintenance of pseudogenes is due to the slow-growth rate (McLeod et al., 2004) or lack of recombination (Bolotin and Hershberg, 2015). Additionally, pseudogene maintenance may allow the bacteria to revert back and forward from a nonfunctional protein to a functional one (Bolotin and Hershberg, 2015). If true, this may explain why some genes are loss and others are maintained as pseudogenes in M. leprae genome.

In this evolutionary context, where an obligate intracellular pathogen evolved to become dependent on his host, the central question of this work is the impact of M. leprae genome downsizing on antimicrobial resistance. This can be viewed as the time when the host decides that he no longer wants to maintain this intimate relation and starts antibiotic treatment with the assistance of his clinician and the health system.

#### DRUG RESISTANCE AND EFFLUX SYSTEMS IN *M. LEPRAE*

Several mycobacterial drug efflux pumps have been described in M. tuberculosis (**Table 1**). Comparative analysis of M. leprae genome shows the presence of approximately half of these transporters while several others are inactivated or absent, probably lost as consequence of reductive evolution. In silico, M. tuberculosis H37Rv genome encodes 267 putative transporters, of which 129 belong to the ATP-binding cassette (ABC) superfamily, 31 to the major facilitator superfamily (MFS), 14 to the resistance nodulation and cell division (RND) superfamily, 1 to the small multidrug resistance (SMR) family, and 1 to the multidrug and toxic compound extrusion (MATE) family. M. leprae genome encodes for 114 transporters, of which 62 corresponds to ABC transporters, 6 MFS, 5 RND, and 1 SMR TABLE 1 | Putative drug membrane transporters encoded by M. tuberculosis H37Rv and its orthologous in M. leprae TN.


(Continued)

#### TABLE 1 | Continued


\* CDS, coding DNA sequence; \*\*determined at protein level; pseudogenes are underlined. ABC, ATP-binding cassette; ACR, acriflavine; AGs, aminoglycosides; AP, antimicrobial peptides; AU, adamantyl ureas; AZ, azoles; BCECF, 2',7'-bis-(2-carboxyethyl)-5(6)-carboxyfluorescein; BDQ, bedaquiline; BL, β-lactams; BP, biarylpiperazines; BSP, bisanilinopyrimidines; CFZ, clofazimine; CHL, chloramphenicol; CMB, carboxymycobactins; CPC, cetylpyridinium chloride; DAU, daunorubicin; DOX, doxorubicin; EMB, ethambutol; ERY, erythromycin; EtBr, ethidium bromide; FQs, fluoroquinolones; IA, indoleamides; INH, isoniazid; KAN, kanamycin; MAC, macrolides; MATE, multidrug and toxic compound extrusion; MB, mycobactins; MFS, major facilitator superfamily; NOR, norfloxacin; NOV, novobiocin; PA, pyrazolones; PD, pyridines; Phleo, phleomycin; PR, pyrroles; PUR, puromycin; PY, pyronin Y; RIF, rifampicin; RND, resistance nodulation division; SKI, imidazoline SKI-356313; SMR, small multidrug resistance; SPE, spectinomycin; STR, streptomycin; TET, tetracycline; TPP, tetraphenylphosphonium; VAN, vancomycin.

(Elbourne et al., 2017). Those that have been associated with drug resistance in M. tuberculosis are discussed below. Alignment visualization of the M. tuberculosis and the M. leprae whole genome sequences, with the predicted CDS regions of the efflux transporters of M. leprae highlighted is shown in **Figure 1**.

#### ABC Transporters

ABC transporters can be divided in those encoded by "one," "two," and "three genes" (**Table 1**). Among those coded by "one gene" in M. leprae is ML2084, homologue of BacA that is involved in virulence of M. tuberculosis (Domenech et al., 2009) and in the active transport of drugs across the membrane (**Table 1**). Absent from M. leprae genome is the transporter Rv0194, which was the first to be associated with β-lactam transport in M. tuberculosis -important taking into account since β-lactams can be useful in M. leprae chemotherapy (Danilchanka et al., 2008; Garima et al., 2015). The phosphate-specific ABC transporter (PstB) is known to be operative in M. tuberculosis during phosphate limiting conditions during infection (Banerjee et al., 2000). The pst operon encodes pseudogenes in M.

BRIG software (Alikhan et al., 2011). CDS, coding DNA sequence.

leprae. As PstB is non-functional in M. leprae and active in M. tuberculosis this may indicate that M. leprae encounters stable phosphate content within the host, making the presence of PstB unnecessary. Besides its role on phosphate uptake, PstB is also associated to the extrusion of antibiotics (**Table 1**). Rv1473 and Rv2477c encode putative macrolide transporters with functional orthologues in M. leprae genome. To our knowledge, there is no evidence for M. leprae isolates resistant to clarithromycin so far, stressing the therapeutic usefulness of this antibiotic against M. leprae.

ABC transporters encoded by "two genes" are Rv1218c-Rv1217c, Rv1273c-Rv1272c, Rv1668c-Rv1667c, and Rv1687c-Rv1686c. Although nothing is known about their expression in M. tuberculosis, the orthologous genes of Rv1668c-Rv1667c and Rv1687c-Rv1686c in M. leprae, ML1240c-ML1239c, and ML1350c-ML1349c, respectively, are non-functional. The Rv1218c-Rv1217c efflux pump has been associated with M. tuberculosis resistance to a wide variety of chemical classes of compounds (**Table 1**). Orthologues of Rv1273c-Rv1272c can be found in M. leprae genome but because there are no studies about their role on M. tuberculosis drug resistance, nothing can be anticipated for M. leprae.

ABC transporters encoded by "three genes" are the operons Rv1458c-Rv1457c-Rv1456c, DrrABC and Rv2688c-Rv2687c-Rv2686c associated with the extrusion of several drugs in M. tuberculosis (**Table 1**). M. leprae orthologue of Rv1458c-Rv1457c-Rv1456c is ML0590c-ML0589c-ML0587c, which encodes a functional efflux transporter. In relation to Rv2688c-Rv2687c-Rv2686c, M. leprae chromosome has non-functional orthologues of two components, ML1034- ML1035 (positive strand), while the orthologue of Rv2688c is absent. The genes Rv2686c, Rv2687c, and Rv2688c are co-transcribed. Rv2686c and Rv2687c proteins possess six transmembrane segments, whereas Rv2688c has a nucleotidebinding domain and is likely involved in ATP hydrolysis. In consequence, Rv2688c probably coordinates the functionality of the whole operon. In this case, it is not surprising that the absence of the Rv2688c orthologue in M. leprae renders non-functional the other genes within this operon, contributing to the increased susceptibility of M. leprae to fluoroquinolones.

#### MFS Transporters

M. leprae genome possess 11 MFS drug transporters, whereas 20 can be detected in M. tuberculosis (**Table 1**). Among the nine transporters absent from M. leprae genome, nothing is known about their role in M. tuberculosis for five of them, while for the remaining four it has been described an association with resistance to several drugs (**Table 1**). Concerning the 11 MFS efflux transporters present in M. leprae, eight are nonfunctional of which some were found to be upregulated in M. tuberculosis in response to antibiotics (**Table 1**). Among these is noted that Rv1258c, also known as Tap-like efflux pump, is a pseudogene (ML1104c) in M. leprae presenting 58.4% similarity at nucleotide level with Rv1258c. Rv1258c is associated with reduced susceptibility to several drugs, namely to rifampicin and clofazimine (**Table 1**) and has an essential role in physiology, growth, and cell morphology (Ramón-García et al., 2012). These findings emphasize the important role of the Rv1258c efflux pump in the oxidative stress response, cell wall assembly and growth, intrinsic drug resistance (Ramón-García et al., 2012) and macrophage tolerance (Adams et al., 2011). Since M. leprae orthologue is non-functional, it is unlikely that Rv1258c play any role in M. leprae intrinsic drug resistance and virulence. This can be one more genomic trait of M. leprae that contributes to its notable susceptibility to rifampicin and clofazimine in vivo.

M. leprae chromosome encodes only three functional MFS efflux pumps associated with drug transport in M. tuberculosis (**Table 1**). Of these, the M. tuberculosis P55 efflux pump (Rv1410c), orthologue of M. leprae ML0556c, is one of the most relevant and well-studied efflux pumps of M. tuberculosis and has been associated with the resistance to several drugs (**Table 1**) and virulence. P55 forms an operon with LprG (Rv1411c), a conserved lipoprotein, which is required for in vivo growth of M. tuberculosis (Bigi et al., 2004; Farrow and Rubin, 2008), virulence (Bianco et al., 2011a) and accurate cellwall assembly (Bigi et al., 2004; Bianco et al., 2011b). M. leprae encodes both proteins presenting high similarity with those of M. tuberculosis. P55 is also associated with cholesterol transport, carbon metabolism, and oxidative stress, which are of major importance for mycobacterial optimal survival and pathogenesis (Ramón-García et al., 2015). Contrary to that observed for Rv1258c (Tap-like efflux pump), the presence of P55 in M. leprae genome indicates a vital role of this transporter in M. leprae for which a significant contribution in providing intrinsic antibiotic resistance is plausible.

#### MATE Transporters

DinF (Rv2836c) is the only MATE transporter that M. tuberculosis genome encodes. The M. tuberculosis homologue in M. smegmatis (Mmp) is involved in the resistance to multiple drugs (**Table 1**). Importantly, DinF is absent from M. leprae genome which can be related with the Na+-dependent nature of the MATE transporters that may not exist in the environment where M. leprae resides. Youm and Saier (2012) also noted the absence of other NA<sup>+</sup> transporters in M. leprae that are present in M. tuberculosis and suggested that these facilitators probably contribute to the maintenance of ion homeostasis and adaptation to several stress conditions.

#### SMR Transporters

M. tuberculosis genome harbors only one gene belonging to the SMR family, the mmr gene, orthologue of ML1756 of M. leprae. Mmr overexpression was showed to decrease susceptibility of M. smegmatis and M. tuberculosis to intercalating dyes, quaternary ammonium compounds and antibiotics (**Table 1**). Mmr is controlled by the TetR-like transcriptional repressor Rv3066 (Bolla et al., 2012) whose orthologue in M. leprae is ML1757. In both species, the transcriptional repressor is located immediately downstream of mmr or ML1756. However, while in M. tuberculosis the Rv3066 gene encodes a 202-aminoacidic protein, its orthologue in M. leprae is a pseudogene. This means that mmr transcription is no longer repressed in M. leprae. Nothing is known about M. leprae susceptibility to biocides and dyes thus the significance of mmr depression cannot be unveiled.

#### RND Transporters

M. tuberculosis genome contains 13 genes that encode MmpLs (Mycobacterial membrane protein, Large), and five auxiliary proteins, the MmpSs (Mycobacterial membrane protein, Small) (**Table 1**). The MmpLs efflux pumps are responsible for the transport of lipids, mainly mycolic acids, essential for mycobacterial survival and pathogenesis, and heme transport (Cox et al., 1999; Camacho et al., 2001; Converse et al., 2003; Domenech et al., 2005; Tullius et al., 2011; Grzegorzewicz et al., 2012; Tahlan et al., 2012; Rodríguez et al., 2013). The expression of M. tuberculosis MmpL proteins is controlled by a complex regulatory network that includes orthologues of TetR (Rv1816 and Rv3249c) and MarR (Rv0678) transcriptional regulators (Radhakrishnan et al., 2014; Delmar et al., 2015). The transcriptional regulator Rv0678 has no orthologue in M. leprae and Rv3249c and Rv1816, whose M. leprae counterparts are ML0770 and ML0933, are non-functional. This indicate that some of the M. leprae mmpL genes are out of regulation and are being constitutively expressed or MmpLs regulation in M. leprae involves a different regulatory network from that found in M. tuberculosis. Of the 13 MmpLs encoded in M. tuberculosis, only MmpL3, MmpL4, MmpL7, MmpL10, and MmpL11 are functional proteins in M. leprae, while the remaining are absent or non-functional (**Table 1**). Of these, only MmpL3 is essential for M. tuberculosis survival (Domenech et al., 2005). Moreover, MmpL3 has emerged as a novel therapeutic target in M. tuberculosis (Li et al., 2014). Due to high degree of similarity between M. leprae MmpL7, MmpL11, and MmpL3 and their orthologous in M. tuberculosis (**Table 1**), we can anticipate a similar function for both proteins in M. leprae.

MmpS5-MmpL5, one of the most important RND transporters of M. tuberculosis, is absent in the M. leprae genome. During the reductive evolutionary process that M. leprae experienced, the MmpS5-MmpL5 efflux transporter was eliminated probably to maintain only the pathways required for a strict intracellular lifestyle, typical of M. leprae. The overexpression of the MmpS5-MmpL5 efflux transporter was shown to be associated with resistance of M. tuberculosis to azoles (Milano et al., 2009) and bedaquiline and cross-resistance to clofazimine (Andries et al., 2014). The absence of the MmpS5- MmpL5 explains way clofazimine is so efficient against M. leprae. So far, very rare M. leprae strains were described with clofazimine resistance reinforcing the connection between MmpS5-MmpL5 and clofazimine resistance as well as its unique hypersusceptibility in M. leprae.

#### CONCLUSIONS AND FUTURE PRESPECTIVES

The occurrence of shared multidrug resistance efflux transporters between M. leprae and M. tuberculosis reveals implications for drug resistance and virulence. Multidrug resistance efflux pumps are ubiquitous in nature. Some efflux pumps exhibit a dual role in M. tuberculosis contributing to both drug resistance and virulence. Here, we have shown that the major M. tuberculosis efflux pumps that are associated with drug efflux and virulence have been retained during the reductive evolutionary process that M. leprae underwent. These efflux pumps are not only important for substrate transport across the inner membrane but are also responsible for drug resistance by extruding drugs from the periplasm to the outside of the cell. They may confer a selective advantage in hostile environments, therefore contributing to M. leprae pathogenicity and acquired drug resistance to therapy as seen in M. tubreculosis. It has been recently shown that resistance to effective multidrug therapy, especially in the high burden countries such as Brazil and India, is on rise, with

#### REFERENCES


noteworthy rates of resistance especially against rifampicin and dapsone (Cambau et al., 2018). Rifampicin resistance was found in new cases of leprosy that may relate to individual abuse of this antibiotic usage for treating other bacterial infections as it was also seen with ofloxacin resistance although an antibiotic not used for the first-line treatment of leprosy (Cambau et al., 2018). Future work should focus on efflux pumps, as those mentioned above, as new drug targets for new leprosy therapeutics. A comparative transcriptomic profile of these transporters may provide additional insights, since differences are expected in the efflux pump expression due to pathogen specificity as consequence of the obligate intracellular lifecycle of M. leprae. The modulation of these novel targets will enhance the eradication efforts of endemic leprosy and prevent emergence of drug resistance in afflicted countries. This comparative and perspective study identified these new targets using biological information gathered from M. tuberculosis and constitutes the first step for a more detailed computational studies to bring more mechanistic insights and biological analyses to be applied to M. leprae, susceptible and drug resistant clinical strains, similar to what have been done for M. tuberculosis (Sandhu and Akhter, 2016). The increase in the number of available sequenced genomes and structural data of these proteins together with the advances on experimental and computational biology will improve our knowledge on the relationship between M. leprae protein sequence, structure, dynamics and function (Li et al., 2017).

#### AUTHOR CONTRIBUTIONS

DM and MV designed this conceptual study. DM performed the analyses. DM, EL, FM, EC and MV have made substantial contributions to the work and approved its final version.

#### FUNDING

DM and MV work is partially supported by the Global Health and Tropical Medicine (GHTM) Research Center (Grant UID/Multi/04413/2013) from Fundação para a Ciência e a Tecnologia (FCT), Portugal and own funding. DM is supported by grant SFRH/BPD/100688/2014 from FCT, Portugal. EL, FM, and EC work on leprosy is supported by the 2016-2022 grant from Santé Publique France.


resistance in Mycobacterium smegmatis. Mol. Gen. Genet. 262, 949–956. doi: 10.1007/PL00008663


bovis and Mycobacterium tuberculosis. Antimicrob. Agents Chemother. 45, 800–804. doi: 10.1128/AAC.45.3.800-804.2001


multidrug resistance in Mycobacterium tuberculosis. PLoS ONE 7:e34538. doi: 10.1371/journal.pone.0034538


in experimental latent tuberculosis infection. PLoS ONE 8:e58378. doi: 10.1371/journal.pone.0058378


**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 © 2018 Machado, Lecorche, Mougari, Cambau and Viveiros. 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) and the copyright owner(s) 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.

# Effect of a Point Mutation in mprF on Susceptibility to Daptomycin, Vancomycin, and Oxacillin in an MRSA Clinical Strain

Feng-Jui Chen<sup>1</sup> \*, Tsai-Ling Lauderdale<sup>1</sup> , Chen-Hsiang Lee<sup>2</sup> , Yu-Chieh Hsu<sup>1</sup> , I-Wen Huang<sup>1</sup> , Pei-Chi Hsu<sup>1</sup> and Chung-Shi Yang<sup>3</sup>

<sup>1</sup> National Institute of Infectious Diseases and Vaccinology, National Health Research Institutes, Zhunan, Taiwan, <sup>2</sup> Division of Infectious Diseases, Kaohsiung Chang Gung Memorial Hospital, College of Medicine, Chang Gung University, Kaohsiung, Taiwan, <sup>3</sup> Institute of Biomedical Engineering and Nanomedicine, National Health Research Institutes, Zhunan, Taiwan

We previously reported the sequential recovery of daptomycin-nonsusceptible MRSA clinical isolates with an L431F substitution in the MprF protein. The aim of the present study is to determine the effect of this mutation by replacing the mprF gene on the chromosome of a daptomycin-susceptible progenitor strain, CGK5, to obtain CGK5mut having the L431F MprF mutation. Compared to CGK5, the daptomycin and vancomycin MICs of CGK5mut increased from 0.5 to 3 µg/ml and from 1.5 to 3 µg/ml, respectively; however, its oxacillin MIC decreased from 128 to 1 µg/ml in medium without added 2% NaCl. The expression levels of vraSR and several other cell-wall synthesis-related genes were significantly increased in CGK5mut, and the mutant also had significantly reduced negative cell membrane charge, thicker cell wall, and longer doubling time. These features were abolished in the reverse mutant carrying F431L MprF, confirming the pleiotropic effects of the L431F MprF mutation. We believe that this is the first work that shows a single MprF missense mutation can lead to not only changes in the cell membrane but also increased expression of vraSR and subsequently increased resistance to daptomycin and vancomycin while simultaneously conferring increased susceptibility to oxacillin in an isogenic MRSA strain.

#### Keywords: MRSA, evolution, drug resistance, daptomycin, vancomycin, oxacillin

#### INTRODUCTION

Daptomycin, a cyclic lipopeptide antibiotic, is one of the last-line agents for the treatment of certain severe multidrug-resistant Staphylococcus aureus infections, including those caused by methicillinresistant S. aureus (MRSA) (Enoch et al., 2007). Daptomycin functions by inserting itself into the bacterial cell membrane in a calcium-dependent manner to cause membrane depolarization, leading to cell death (Enoch et al., 2007; Baltz, 2009). Daptomycin-nonsusceptible (DAP-NS) MRSA isolates, although still uncommon, have emerged during daptomycin treatment of patients (Lee et al., 2010; Boyle-Vavra et al., 2011). DAP-NS MRSA mutants have also been generated in the laboratory by serial passage of isolates in sublethal concentrations of daptomycin (Enoch et al., 2007; Camargo et al., 2008; Mishra et al., 2009, 2012; Rubio et al., 2012).

The exact mechanisms giving rise to daptomycin-non-susceptibility in S. aureus are not fully elucidated but appear to involve diverse genetic events and several genetic loci, including

#### Edited by:

Elena Perrin, Università degli Studi di Firenze, Italy

#### Reviewed by:

Daniel Haeusser, Canisius College, United States Balaji Veeraraghavan, Christian Medical College & Hospital, India

> \*Correspondence: Feng-Jui Chen frchen@nhri.org.tw

#### Specialty section:

This article was submitted to Evolutionary and Genomic Microbiology, a section of the journal Frontiers in Microbiology

Received: 28 February 2018 Accepted: 07 May 2018 Published: 25 May 2018

#### Citation:

Chen F-J, Lauderdale T-L, Lee C-H, Hsu Y-C, Huang I-W, Hsu P-C and Yang C-S (2018) Effect of a Point Mutation in mprF on Susceptibility to Daptomycin, Vancomycin, and Oxacillin in an MRSA Clinical Strain. Front. Microbiol. 9:1086. doi: 10.3389/fmicb.2018.01086

mprF, yycG (walK), vraSR, tagA, and dltABCD (Friedman et al., 2006; Baltz, 2009; Bertsche et al., 2011; Song et al., 2013). These loci are all part of the cell wall stimulon in S. aureus and include genes encoding proteins involved in the production of membrane phospholipids. The mprF (multiple peptide resistance factor) gene seems to be especially critical, as mprF mutations are the most frequently reported genetic lesions in DAP-NS MRSA isolates (Friedman et al., 2006; Lee et al., 2010; Boyle-Vavra et al., 2011; Mehta et al., 2012a). MprF is a bi-functional membrane protein with lysylphosphatidylglycerol (LPG) synthase and flippase activities (Peschel et al., 2001; Ernst et al., 2009; Ernst and Peschel, 2011). Different point mutations in mprF have been associated with elevated LPG synthesis. This results in increased amounts of LPG relative to phosphatidylglycerol (PG) on the outer leaflet of the cytoplasmic membrane and an accompanying reduction in cell membrane negative charge (Baltz, 2009; Rubio et al., 2012).

A feature that has been seen in both clinical and laboratorygenerated DAP-NS MRSA isolates is a concomitant vancomycin intermediate or heterogeneous intermediate resistance (VISA or hVISA) phenotype (Camargo et al., 2008; Mishra et al., 2009); VISA has moderate resistance to vancomycin and hVISA has varying sub-population of cells resistant to vancomycin, thus exhibit mixed susceptibility to vancomycin. Mutations in the vraSR, graSR, or walKR (yycGF) two-component systems have been associated with the VISA/hVISA phenotype (Howden et al., 2010). Among these 3 two-component systems, the VraSR system is particularly important in maintaining cell wall integrity. It serves as a sentinel in response to cell wall damage by positively regulating a unique set of genes involved in cell wall synthesis, resulting in the generation of a resistant phenotype (Kuroda et al., 2003; Gardete et al., 2006; McAleese et al., 2006). Another unusual feature is the so-called "seesaw" effect on β-lactam susceptibility, wherein DAP-NS isolates exhibit reduced β-lactam MICs (Mishra et al., 2009; Lee et al., 2010; Yang et al., 2010; Mehta et al., 2012a). However, these phenomena are not observed in all DAP-NS MRSA isolates.

Our previous study of eight sequential clinical MRSA isolates from a patient with persistent bacteremia revealed an L431F amino acid substitution in the MprF protein of DAP-NS isolates (Lee et al., 2010). Since this mutation had not been reported previously, we undertook the present study to determine the effect of this mutation on the cellular response to daptomycin. To eliminate the possibility of unknown genetic changes that might have occurred in paired clinical strains, we used a base-substitution method to replace a single nucleotide (from CTT to TTT) within the chromosomal mprF gene of the daptomycin-susceptible (DAP-S) progenitor of the DAP-NS strains. To confirm the results obtained by this single amino acid exchange (L431F) in MprF, we also constructed a reverse mutant carrying F431L MprF as well as a silent EcoRV site. Our study demonstrated that this single amino acid change (L431F) confers increased resistance to both daptomycin and vancomycin, with a concurrent decrease in oxacillin MIC. The phenotype and genetic factors associated with these changes were investigated.

## MATERIALS AND METHODS

#### Bacterial Strains

The bacterial strains, plasmids, and primers used are listed in **Tables 1**, **2**. Unless stated otherwise, Luria-Bertani (LB) broth and plates were used for growth of Escherichia coli and S. aureus at 37◦C. E. coli strain XL10-Gold (Stratagene, La Jolla, CA, United States) and GeneHogs <sup>R</sup> (Invitrogen, Carlsbad, CA, United States) were used for cloning. S. aureus cells were transformed by electroporation, as described previously (Schenk and Laddaga, 1992). Ampicillin (100 µg/ml), chloramphenicol (5 µg/ml), erythromycin (5 µg/ml), spectinomycin (100 µg/ml), and tetracycline (5 µg/ml) were used for plasmid selection in E. coli and S. aureus.

### Construction of MprF Mutant Derivatives

To add new selection markers and cloning sites, the bgaB cassette located between the HindIII sites was removed from pMAD by digesting with HindIII, and religated. A modified bgaB, with extra restriction sites, was cloned by PCR (using primers bgaBnew-F and bgaBnew-R) from the original pMAD into the pMAD lacking the bgaB cassette to generate new pMAD. Two selection markers, as sat4 and tetM cassettes, were incorporated by cloning PCR fragments of the sat4 and tetM genes from the chromosomal DNA of 2V076 and RN6911, respectively, into the StuI-SalI sites of the new pMAD using primers sat4-F and sat4- R, and tetM-F and tetM-R to produce pMAD-SAT4-tetM, which confers resistance to nourseothricin and tetracycline. To assess the effect of MprF L431F on daptomycin non-susceptibility, an mprF-bearing fragment from CGK6 (the first DAP-NS isolate containing the MprF F431 mutant) was amplified by PCR using the primers mprF-F and mprF-R, and ligated into the BamHI-SmaI sites of pMAD-SAT4-tetM. This recombinant plasmid, pMprFmut, was used as the allelic exchange vector for mutation of the mprF gene in CGK5 to create CGK5mut. To validate the phenotypes in CGK5mut, the mutant strain was reverted to wild type by allelic exchange again. The mprF fragment was amplified from CGK5 and then cloned into pMAD-SAT4-tetM. This recombinant plasmid, pMprF5, was used as the template DNA for site-directed mutagenesis to introduce a new EcoRV site<sup>1</sup> into the complementation construct (and without altering the coding sequence) to allow it to be differentiated from the CGK5 parental strain. After PCR using the mutant primers mprF(EcoRV)-F and mprF(EcoRV)-R, the mprF reverse mutant was treated with DpnI, ligated into a circle with T4 DNA ligase, and then used as the template DNA. The entire mprF gene was then cloned from the mutated plasmid into the pMAD-SAT4 tetM vector again to eliminate potential mutation of the vector sequence. The constructs were verified by restriction analysis and DNA sequencing. Sequencing was performed at the DNA Sequencing Core Lab of our institutes.

The allele replacement procedure was applied to create a single base replacement in the mprF gene of CGK5, as described previously (Arnaud et al., 2004). Briefly, the pMprFmut plasmid was electroporated into S. aureus strain RN4220, and then

<sup>1</sup>http://emboss.bioinformatics.nl/cgi-bin/emboss/silent

electroporated into CGK5. Transformants were selected at 30◦C on LB plates containing tetracycline and X-Gal (150 µg/ml). One blue colony was inoculated in Trypticase Soy broth (TSB) containing tetracycline and incubated with shaking for 2 h at 30◦C followed by 6 h at 43◦C, serial diluted then plated on Trypticase Soy agar (TSA) plates containing tetracycline and X-Gal and incubated at 43◦C overnight to obtain light blue colonies caused by a single crossover event. One light blue colony was inoculated in TSB without antibiotic and incubated with shaking at 30◦C overnight, then diluted 1:100 and incubated at 43◦C for 6 h; serial dilutions were plated on TSA plates in the absence of antibiotics and incubated at 37◦C overnight. Several white colonies were selected to verify for tetracycline sensitivity, which indicates loss of the integrated vector resulting from a double crossover event. To confirm the double crossover, PCR amplifications were performed with multiplex primers hybridizing outside and inside of the mprF gene and vector sequences. A colony with a single crossover was used as a negative control. The resulting mutant was verified by DNA sequencing, including the adjacent region of the mprF gene. The complementation construct (CGK5mutR) was constructed by the same procedure.

#### Antimicrobial Susceptibility Testing

The MICs of daptomycin, oxacillin and vancomycin were determined by Etest <sup>R</sup> (bioMérieux SA, Marcy l'Étoile, France) following the manufacturer's instructions and using Mueller Hinton II agar (MHA) (Becton Dickinson, Cockeysville, MD, United States). The daptomycin Etest strips were overlaid with 40 µg/ml of calcium (Package insert) and the MHA contained 2.9–5.9 µg/ml of calcium. The MICs of oxacillin were also determined by broth microdilution (BMD) (CLSI, 2013). The BMD method was performed in Mueller Hinton II broth (MHB) (Becton Dickinson) with and without 2% NaCl from an inoculum of 5 × 10<sup>5</sup> CFU/ml, and the MIC was read after incubation at 35◦C for 24 h. S. aureus ATCC 29213 and Enterococcus faecalis ATCC 29212 were used as quality control organisms for Etest. S. aureus ATCC 29213 and ATCC 43300 were used as quality control organisms for BMD.

#### Population Analysis Profiles

A population analysis profile (PAP) for vancomycin was performed on CGK5 and CGK5 mutant derivatives following protocols previously described (Hiramatsu et al., 1997; Howden et al., 2006). Briefly, overnight cultures of test isolates were serially diluted in TSB and inoculated onto brain–heart infusion (BHI) agar plates containing 0–8 µg/ml vancomycin. After 48 h incubation at 35◦C, colonies were counted and plotted. Mu3 (ATCC 700698) and N315 S. aureus strains were tested in parallel as hVISA positive and negative controls, respectively. The area under the curve (AUC) values of the test strains were compared to that of Mu3.

#### Cell Wall Thickness

Bacterial cells for transmission electron microscopy were prepared following previously described protocols (Hanaki et al., 1998). Photographic images were obtained at a final magnification of 15,000× using a Hitachi H-7650


tsStands for thermosensitive.

microscope (Hitachi High-Technologies Corporation, Tokyo, Japan). Fifty measurements of equatorially cut cells were taken for the calculation of cell wall thickness, and results were expressed as mean ± SD following previously described protocols (Cui et al., 2000).

#### Growth Rate

Overnight fresh cultures of bacteria were adjusted in 0.85% NaCl to 0.5 McFarland turbidity, then diluted 1:200 in MHB to obtain 5 × 10<sup>5</sup> CFU/ml starting inoculum. The inoculum was dispensed at 120 µl per well in triplicates into a 100-well plate and incubated at 37◦C in Bioscreen C MBR (Oy Growth Curves Ab Ltd, Helsinki, Poland) (Richardson et al., 2008). Triplicate medium-only blank wells were included in each plate. The OD<sup>600</sup> of each well was read every 30 min for 24 h. The average OD of the blank wells was subtracted from the average of the triplicate test wells at each time point and plotted. Doubling times were calculated using the exponential growth phase following a previously described protocol (Cui et al., 2003). To verify the OD measurements, the CFU counts were checked by the shaker flask method.

#### Cytochrome c Binding Assay

Cytochrome c binding assay was performed following an approach similar to that of Mukhopadhyay et al. (2007) with slight modification. Briefly, bacteria grown overnight at 35◦C were harvested and washed twice with 20 mM MOPS buffer (pH 7.0) and resuspended in the same buffer to a final OD<sup>578</sup> of 7.

#### TABLE 2 | Primers used in this study.


<sup>a</sup>Restriction enzyme sites are underlined. P represents phosphorylation. Boldface type indicates a mutant nucleotide.

Cytochrome c (Sigma Chemicals, St. Louis, MO, United States) was prepared in the same buffer. The bacterial suspension was incubated with 1 mg/ml cytochrome c for 10 min, and then centrifuged at 3000 g at 4◦C for 10 min. The supernatant containing unbound cytochrome c was collected and measured spectrophotometrically at OD530. The cytochrome c was serially diluted (0.1–0.6 mg/ml) to create a standard curve to measure the cytochrome c concentration of the supernatant.

### Western Blot Analysis for PBP2a Detection

Detailed descriptions of mouse monoclonal anti-PBP2a antibody (2F6F) production and detection of PBP2a were presented previously (Chen et al., 2014). However, in that study, PBP2a was detected in the total protein preparation. In the present study, for more precise detection of PBP2a in the membrane, membrane fraction was prepared following previously described protocol with slight modifications (Downer et al., 2002). Briefly, strains were grown in TSB overnight at 37◦C with shaking at 200 rpm, then diluted 1/100 in 50 ml fresh medium and further incubated until cells reached an OD<sup>600</sup> of 1. Ten milliliter of the bacterial cultures was then centrifuged, and the pellet was resuspended in 1 ml lysis buffer. Cells were disrupted by FastProteinTM Blue Matrix using a FastPrep-24 homogenizer (MP Biomedicals) at 6 M/s for 4 cycles of 20 s with a 2-min ice incubation in between each cycle. Protein samples (1 µg of protein per lane) were separated in 10% sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE), and then transferred to a polyvinylidene difluoride membrane (Immobilon, Millipore Corp., Bedford, MA, United States). PBP2a was probed with mouse monoclonal anti-PBP2a antibody (2F6F), and bands were visualized with HRP-conjugated secondary antibodies (Abcam Inc., Cambridge, MA, United States) followed by incubation in Western lightning chemiluminescence reagent plus (Perkin Elmer Life Sciences, Boston, MA, United States). Sortase A, a transpeptidase, was used as an internal control, identified by rabbit polyclonal anti-Sortase A primary antibody (Abcam Inc.).

### Real-Time Quantitative PCR Analysis for mprF, vraSR, and Cell Wall Synthesis-Related Genes

For RNA isolation, strains were grown overnight with shaking at 37◦C in MHB and diluted with 5 ml fresh medium to 1/100. They were then grown at 37◦C with shaking at 200 rpm, and samples were collected from four time points (2, 3, 4, 5 h) for analysis. Approximately 2 × 10<sup>9</sup> cells were harvested for RNA isolation. Total RNA isolation was performed as described previously (Chen et al., 2009). Real-time quantitative PCR analysis was performed using Maxima SYBR Green qPCR Master Mix (Thermo Fisher Scientific, United States) on a Roche LightCycler 480 II Real-Time PCR System (Roche). Results were normalized to the expression of gyrB. The expression levels of mprF, vraSR, and several selected cell wall synthesis-related genes (fmtA, murZ, pbp2, sgtB, tagA) were determined using the primers listed in **Table 2**. The expression of ldh1, a gene involved in nitric oxide resistance in S. aureus (Richardson et al., 2008), was used as a negative control. The target gene transcripts were quantified by using the basic relative quantification method of the LightCycler 480 Software v1.5.1.62 (Roche). Three independent experiments were performed in duplicate and results are shown as mean ± SD. The mRNA expression levels of genes from CGK5 were defined as 1.

## Construction of mprF and vraSR Promoter-Reporter Plasmids

For introducing the pluxT2 reporter into CGK5, the selection marker spc was amplified by PCR from the Z172 strain (Chen et al., 2013), digested with SmaI and then cloned into pluxT2 (Chen et al., 2014) to generate pluxT2-SPC. The mprF and vraSR promoter fragments were PCR-amplified from the chromosome of CGK5 using the primers PmprF-F, PmprF-R and PvraSR-F, PvraSR-R. These amplified promoter fragments were fused upstream of the lux reporter genes of pluxT2-SPC using the BamHI and SalI sites, and the resulting plasmids were used for transformation of the S. aureus strain RN4220 and then of CGK5 and CGK5 mutant derivatives. The resulting promoter-reporter fusion constructs were confirmed by restriction enzyme analysis and DNA sequencing.

## Luciferase Reporter Assays

Bioluminescence was measured with a SpectraMax L microplate reader (Molecular Devices, Sunnyvale, CA, United States) to determine the transcriptional level of the mprF and vraSR promoter constructs. Three independent transformants harboring promoter-reporter fusion plasmids were grown overnight with shaking at 37◦C in TSB containing spectinomycin for reporter plasmid maintenance, diluted 1:100 in the same medium, and aliquots (200 µl) of the cultures were transferred into 96-well plates in duplicate and incubated at 37◦C. Both OD<sup>595</sup> and bioluminescence (in relative light units, RLU) were monitored every hour for 7 h.

## Statistical Analysis

Assay results are reported as mean ± SD where appropriate. For comparison of differences between CGK5 and CGK5 mutant derivatives, the Student's t-test was performed using GraphPad Prism 6 software (GraphPad, La Jolla, CA, United States). A P value less than 0.05 was considered statistically significant.

## RESULTS

To examine the effects of the MprF L431F amino acid substitution in an isogenic background, we replaced the chromosomal wild-type mprF gene of the DAP-S parental strain CGK5 with the mutant mprF gene from the CGK6 DAP-NS strain, creating the new strain CKG5mut. A single base change (from CTT to TTT) causes the L431F mutation in MprF. Furthermore, we generated a reverse mutation complementation construct from CGK5mut, called CGK5mutR, which carries a wild-type mprF with a new EcoRV restriction site that can be used to differentiate CGK5mutR from CGK5. CGK5mut and CGK5mutR were verified by PCR and sequencing of the entire

mprF gene. We then carried out a phenotypic analysis to compare CGK5mut and CGK5mutR against their parental strain CGK5.

The daptomycin MIC of CGK5mut was 6-fold higher (3 µg/ml) compared to that of the wild-type parental strain CGK5 (0.5 µg/ml), a level of increase similar to what we previously reported for DAP-NS clinical isolates (Lee et al., 2010). To our surprise, the vancomycin MIC of CGK5mut was also increased compared to CGK5 (3 µg/ml vs. 1.5 µg/ml). The reverse complementation construct, CGK5mutR, showed the same level of susceptibility as CGK5. PAP results showed that, of the three strains, only CGK5mut displayed a hVISA phenotype (**Figure 1A**). The bacterial feature that is often associated with DAP-NS and VISA or hVISA strains is increased cell wall thickness (Camargo et al., 2008; Mehta et al., 2012a). Transmission electron microscopy analysis revealed that CGK5mut has a significantly thicker cell wall (38.09 ± 5.67 nm) than CGK5 (30.65 ± 2.58 nm) and CGK5mutR (25.94 ± 3.75 nm) (p < 0.05) (**Figure 1B**). It has also been reported that clinical and in vitro-derived S. aureus strains exhibiting increased vancomycin MICs grow slower than their progenitors (Cui et al., 2003; Camargo et al., 2008). This phenomenon has been attributed to increased cell wall synthesis at a biological cost to the resistant strains (Cui et al., 2003). We also found that CGK5mut (doubling time = 240 ± 9.5 min) grows slower than both CGK5 (doubling time = 158 ± 3.5 min) and CGK5mutR (doubling time = 170 ± 4.2 min) in MHB (**Figure 2**).

Altered cell membrane charge has been found in some DAP-NS S. aureus isolates with point mutations in mprF (Peschel et al., 2001; Oku et al., 2004). Cytochrome c is a cationic protein that has been used to estimate the relative bacterial cell surface charge of S. aureus (Mukhopadhyay et al., 2007). We detected a significant increase in unbound cytochrome c for CGK5mut (0.676 ± 0.056 mg/ml) compared to CGK5 (0.393 ± 0.058 mg/ml) and CGK5mutR (0.417 ± 0.02 mg/ml) (p < 0.05) (**Figure 3**), indicating an increased positive charge density on the cell surface of CGK5mut and suggesting a mechanism by which the L431F mutation of MprF might contribute to daptomycin resistance.

The mechanism of oxacillin resistance in MRSA is thought to be mainly due to the production of penicillin-binding protein 2a (PBP2a, encoded by the mecA gene) with reduced affinity for β-lactams (Hiramatsu et al., 2001). Although first observed

negative controls, respectively. (B) Transmission electron microscopy images of CGK5 and its mutant derivatives. Cell wall thickness (in nanometers) was measured at 150,000× magnification. Fifty measurements of equatorially cut cells were taken for calculation of cell wall thickness, and results are expressed as mean ± SD. CGK5, progenitor of CGK5mut; CGK5mut, derivatives of CGK5 with L431F mutation in mprF; CGK5mutR, reverse complementation construct of CGK5mut.

between glycopeptide and β-lactams (Sieradzki and Tomasz, 1999), the "seesaw" inverse relationship between daptomycin and β-lactam susceptibility has been reported by several groups (Mishra et al., 2009; Lee et al., 2010; Yang et al., 2010; Mehta et al., 2012a). We determined the level of PBP2a production in CGK5mut and found it to be comparable to that of CGK5 and CGK5mutR (**Figure 4A**).

However, concurrent to the increased daptomycin MIC, the oxacillin MIC of CGK5mut decreased to 1 µg/ml from 128 µg/ml in CGK5, as determined by the BMD method in medium without 2% NaCl, whereas CGK5mutR showed a similar MIC as CGK5 (**Figure 4B**). Thus, CGK5mut exhibits the "seesaw-like" effect similar to which we and others have previously observed in DAP-NS in vitro-selected and clinical S. aureus isolates (Lee et al., 2010; Mehta et al., 2012a,b). Interestingly, the oxacillin MICs of CGK5, CGK5mut and CGK5mutR were the same (256 µg/ml) in medium containing 2% NaCl (**Figure 4B**). Therefore, we further measured the level of PBP2a of these strains grown in oxacillin (0.25 µg/ml) containing medium with and without 2% NaCl. The level of PBP2a of CGK5mut showed a marked reduction after exposure to oxacillin, and 2% NaCl counteracted this effect (**Figure 4A**).

Mutation of mprF has been shown to impact the expression of mprF itself (Yang et al., 2009b). In addition, VraSR, a twocomponent regulatory system that plays a major role in cell wall synthesis and is a key player in the cell-wall stress response, has often been implicated in vancomycin and daptomycin resistance (Kuroda et al., 2003; Camargo et al., 2008; Muthaiyan et al., 2008; Mehta et al., 2012a). We therefore compared the expression of mprF and vraSR-related genes in CGK5 and its mutant derivatives by real-time quantitative PCR analysis. We found there was no significant difference in the mprF expression among them (**Figure 5**). However, significant increases (>4 fold increase, P < 0.01) in the expression of vraSR were seen in CGK5mut compared to CGK5 and CGK5mutR at the 4-h time point. In addition, the expression levels of four examined cell wall synthesis-related genes (fmtA, murZ, pbp2, sgtB) were also significantly increased in CGK5mut (>2∼3-fold increase, P < 0.01) at this time point.

To clarify whether the up-regulation of gene expression in CGK5mut is due to promoter activity, we constructed mprF and vraSR promoter-reporters. The transcriptional activity of the mprF and vraSR promoters was compared in CGK5 and its mutant derivatives by using a bioluminescence assay. mprF promoter activity among the tested strains showed no obvious differences (**Figure 6A**). However, the vraSR promoter showed significantly higher promoter activity in CGK5mut compared to the other two strains and the maximum discrepancy was at the 4-h time point (**Figure 6A**), indicating that the up-expression of vraSR in CGK5mut was mediated by its promoter.

To determine whether the reduced oxacillin MIC of CGK5mut was caused by increased expression of vraSR, and how this might be related to the fact that 2% NaCl could restore the oxacillin MIC of CGK5mut to wild-type levels, we examined whether the high salt concentration could inhibit the expression of vraSR in CGK5mut. A vraSR promoter activity assay showed that NaCl does indeed inhibit vraSR promoter activity (**Figure 6B**). The

FIGURE 2 | Growth curve of CGK5, CGK5mut, and CGK5mutR in Mueller Hinton II broth (MHB). Overnight fresh cultures of the bacteria were adjusted in 0.85% NaCl to 0.5 McFarland turbidity, then diluted 1:200 in MHB to obtain ∼5 × 10<sup>5</sup> CFU/ml starting inoculum. The inoculum was dispensed at 120 µl per well in triplicates to a 100-well plate and incubated at 37◦C in Bioscreen C MBR. The OD<sup>600</sup> of each well was read every 30 min for 24 h. The average OD of the blank wells was subtracted from the average of the triplicate test wells at each time point and plotted.

effect is dose-dependent and 2% NaCl is the critical concentration for inhibition.

#### DISCUSSION

Daptomycin is an important last-line agent against serious S. aureus infection, particularly in patients with persistent MRSA bacteremia and endocarditis (Boucher and Sakoulas, 2007; Marco et al., 2008). Daptomycin non-susceptibility in S. aureus, albeit still rare, can emerge during treatment and usually in

patients with severe deep-seated infections (Julian et al., 2007; Lee et al., 2010; Boyle-Vavra et al., 2011). The mechanism of daptomycin-non-susceptibility is still not fully understood but has been associated with multiple genetic changes on the bacterial chromosome, with single point mutations in the mprF gene being the most frequently identified alteration (Friedman et al., 2006; Julian et al., 2007; Mehta et al., 2012a; Mishra et al., 2012). Although a few hot spots have been associated with daptomycin resistance, such as L826F, S295L, and S337L, there is considerable variation in the locations of the mutations in the MprF polypeptide structure (Friedman et al., 2006; Julian et al., 2007; Yang et al., 2009a; Lee et al., 2010; Boyle-Vavra et al., 2011; Mehta et al., 2012a; Peleg et al., 2012). A study just published by Yang et al. (2018) investigated the impact of laboratory generated single or dual point mutations within the hot spot loci of mprF and found that extra point mutation in mprF resulted in diminished characteristics associated with DAP-NS, which may explain why no clinically derived DAP-NS strains contained multiple point mutations within the mprF gene.

Our previous study on sequential clinical MRSA isolates from a patient with persistent bacteremia revealed an L431F amino acid substitution in the 3 DAP-NS isolates we characterized (Lee et al., 2010). Since this mutation had not been seen in other DAP-NS isolates reported to date, we carried out the present study to verify its contribution to daptomycin resistance. Because some DAP-NS isolates are also vancomycin non-susceptible, and a few DAP-NS isolates (including our own; Mishra et al., 2009; Lee et al., 2010; Yang et al., 2010; Mehta et al., 2012a) have been reported to exhibit a reduced oxacillin MIC, we also investigated the effects of this L431F MprF change on vancomycin and oxacillin MICs. The approach we employed differed from those of other studies in that we used a single-base-substitution (from CTT to TTT) method to create the L431F mutant (CGK5mut) from CGK5, the wild-type progenitor of our DAP-NS strains. We also generated a reverse mutant from CGK5mut, CGK5mutR, to verify our findings for CGK5mut. This approach ensured that the strains we tested and compared were isogenic, except for the single nucleotide difference at the site corresponding to codon 431 of MprF (and a mutated silent EcoRV site in CGKmutR which has no effect on the sequence of the polypeptide).

The increased daptomycin MIC observed in CGK5mut confirmed that the L431F missense mutation contributes to daptomycin resistance. We also showed that CGK5mut has reduced binding of cytochrome c. Since the L431 amino acid is located in the C-terminal LPG synthase domain of MprF (Ernst et al., 2009), the daptomycin-non-susceptibility in CGK5mut might be an effect of a reduced PG:LPG ratio (Ernst et al., 2009; Yang et al., 2009b). A lower PG:LPG ratio has been shown to reduce the negative charge of the cell membrane (Baltz, 2009; Rubio et al., 2012), thus diminishing the access of calciumdependent daptomycin to its cellular target.

Another factor that is commonly associated with daptomycinnon-susceptibility is VraSR (Camargo et al., 2008; Mehta et al., 2012a), a two-component system that positively regulates the cell wall biosynthesis pathway and is involved in the cell wall stimulon (Kuroda et al., 2003; Gardete et al., 2006). The VraSR system was originally identified as a key regulator of vancomycin resistance in VISA and hVISA isolates (Kuroda et al., 2000; Cui et al., 2005). We found that the expression levels of vraSR and several vraSR-regulated genes were significantly

increased in CGK5mut compared to CGK5 and CGK5mutR. In addition to the increased daptomycin MICs, the vancomycin MIC of CGK5mut was also increased, and population analysis showed that more CGK5mut isolates were able to grow in higher concentrations of vancomycin compared to CGK5 and CGK5mutR. CGK5mut cells also have a thicker cell wall and grow slower than CGK5 and CGK5mutR cells. These cellular features have been associated with both DAP-NS S. aureus and hVISA/VISA (Camargo et al., 2008; Song et al., 2013). The changes in CGK5mut vancomycin MIC and cell wall thickness likely arise from the upregulated vraSR-dependent system. But how the L431F mutant protein brings about the increased expression of vraSR and the cell wall-related genes is unknown. However, CGK5mut had significant increased vraSR promoter activity compared to CGK5 and CGK5mutR using vraSR promoter-reporter assay, indicating that the upexpression of vraSR in CGK5mut was mediated by its promoter. A previous study observed the mutual presence of both mprF point mutations and increased expression levels of vraSR in DAP-NS strains compared to their DAP-susceptible counterparts, and hypothesized that both genes were mechanistically linked to the DAP-NS phenotype (Mehta et al., 2012a). However, our results suggest a causal relationship between mprF point mutation and increased expression of vraSR, which may explain why the daptomycin resistance is often concomitant with vancomycin resistance in clinical isolates. There are a plethora of pathways to heterogeneous and intermediate resistance to vancomycin in S. aureus (Howden et al., 2010). In addition to VraSR, two other two-component systems, GraSR and WalKR, have also been associated with reduced susceptibility of S. aureus to both daptomycin and vancomycin. It has been reported that GraSR also plays an important role in S. aureus resistance to cationic antimicrobial peptides via altered expression of mprF and dltABCD resulting in increased electrostatic repulsion of cationic antimicrobial peptides. GraSR has also been shown to interact with the WalKR system and have significant regulatory overlap (Li et al., 2007; Meehl et al., 2007; Dubrac et al., 2008; Falord et al., 2011). However, we found no significant differences in the expression of mprF and mprF promoter activity among the tested strains. Therefore, it appears unlikely that reduced susceptibility of CGK5mut to both daptomycin and vancomycin are associated with these two two-component systems.

Although expression of PBP2a (encoded by the mecA gene) is thought to play a major role in oxacillin-resistance, several studies have reported that the amount of PBP2a expressed does not correlate with the level of methicillin-resistance (Chambers and Hackbarth, 1987; Murakami and Tomasz, 1989; De Lencastre et al., 1994), and other auxiliary genes have been reported to be essential for the optimal expression of methicillin-resistance (De Lencastre and Tomasz, 1994). When we determined the production level of PBP2a in CGK5mut and found it to be comparable to that of CGK5 and CGK5mutR in the normal condition, we thought its "seesaw-like" effect may be caused by other factors. This indicated that the expression of mecA gene in CGK5mut was not altered by the mprF point mutation. Since the oxacillin MIC of CGK5mut was reduced in medium lacking 2% NaCl, we measured the levels of PBP2a in these strains in the presence of oxacillin with and without 2% NaCl and found that the level of PBP2a production in CGK5mut reduced after exposure to oxacillin and 2% NaCl could abolish this effect. According to the CLSI guideline, 2% NaCl should be added to the medium in determining the oxacillin MIC for staphylococci (CLSI, 2013). Interestingly, our results also showed that the expression of vraSR decreased in the presence of

2% NaCl, with a concurrent restoration of oxacillin resistance in CGK5mut, indicating that the effect of upregulation of vraSR in CGK5mut could be abolished by adding 2% NaCl. Since the "seesaw" effect has been observed in only a few DAP-NS MRSA isolates (Mehta et al., 2012b), the factors associated with this phenomenon warrant further investigation. However, mprF point mutation in CGK5mut only exhibits the "seesaw-like" effect (meaning the seesaw effect was only observed in the absence of NaCl), we speculate that additional mutations at other genes are required to achieve truly "seesaw" effect observed in clinical isolates. In addition, how 2% NaCl counteracts the up-regulation effect of vraSR caused by MprF L431F mutation requires further research.

In fact, Mehta et al. (2012a) previously observed the phenotype in DAP-non-susceptible strains with upregulation of the twocomponent regulatory system vraSR; however, they did not prove this was caused by the MprF mutation. By contrast, they introduced an overexpression of vraSR in DAP-susceptible strains to show a reduction in oxacillin resistance. In their latest study, they found that introducing an overexpression of vraSR in DAP-susceptible strains triggering MprF mutation and result in impairment of PrsA chaperone functions, both events are required for β-lactam resistance via PBP2a maturation (Renzoni et al., 2017). It has long been recommended that 2% NaCl (or KCl) be added to the culture medium to reliably detect oxacillin resistance in staphylococci (Thornsberry and McDougal, 1983; Huang et al., 1993). When salt is placed in solvent, the solid salt dissolves into its component ions, sodium (Na+), potassium (K+), chloride (Cl−), which are the primary ions of electrolytes in physiology. It is plausible to speculate that these ions maintain the phosphorylation state of VraS in respond to mprF point mutation and/or β-lactam stimulon to precisely regulate its downstream genes for the generation of resistant phenotype.

#### CONCLUSION

The present study demonstrated that a single amino acid substitution (L431F) in the MprF protein contributes to

#### REFERENCES


both daptomycin and vancomycin resistance as well as increased oxacillin susceptibility in an isogenic MRSA strain. This point mutation in CGK5 also causes upexpression of vraSR. Further studies on the mechanisms contributing to these phenomena may lead to discovery of potential therapeutics against multidrug-resistant staphylococci.

#### AUTHOR CONTRIBUTIONS

F-JC contributed to the study design, data analysis, and manuscript preparation. T-LL performed the data analysis and prepared the manuscript. C-HL, Y-CH, I-WH, P-CH, and C-SY performed the work and data analysis.

#### FUNDING

This project was supported by intramural grants from the National Health Research Institutes to the National Institute of Infectious Diseases and Vaccinology (IV-101-SP-10, IV-102-SP-10, and IV-101-SP-11), and grants from the Ministry of Science and Technology in Taiwan (MOST 104-2320-B-400-013-). The transmission electron microscopy analysis was supported by an NHRI intramural grant to the Center for Nanomedicine Research (NM-101-PP-04).

#### ACKNOWLEDGMENTS

We wish to thank Ms. Yu-Ching Chen, Hsin-Ying Huang (Research Assistant, Center for Nanomedicine Research), and Yu-Lin Chen (Research Assistant, NIIDV) for their technical assistance. We are also grateful to Prof. Paul Williams (University of Nottingham, United Kingdom) for kindly providing many of important bacterial strains and plasmids used in this study.

daptomycin treatment failure. Antimicrob. Agents Chemother. 55, 2018–2025. doi: 10.1128/AAC.01593-10



strains: resistance mechanisms, laboratory detection, and clinical implications. Clin. Microbiol. Rev. 23, 99–139. doi: 10.1128/CMR.00042-09



**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 © 2018 Chen, Lauderdale, Lee, Hsu, Huang, Hsu and Yang. 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) and the copyright owner 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.

# Genetic Diversity of norA, Coding for a Main Efflux Pump of Staphylococcus aureus

Sofia Santos Costa<sup>1</sup> , Benjamin Sobkowiak<sup>2</sup> , Ricardo Parreira<sup>1</sup> , Jonathan D. Edgeworth<sup>3</sup> , Miguel Viveiros<sup>1</sup> , Taane G. Clark2,4 \* and Isabel Couto<sup>1</sup> \*

<sup>1</sup> Global Health and Tropical Medicine, Instituto de Higiene e Medicina Tropical, Universidade NOVA de Lisboa, Lisbon, Portugal, <sup>2</sup> Faculty of Infectious and Tropical Diseases, London School of Hygiene and Tropical Medicine, London, United Kingdom, <sup>3</sup> Department of Infectious Diseases, Centre for Clinical Infection and Diagnostics Research, Guy's and St Thomas' NHS Foundation Trust, King's College London, London, United Kingdom, <sup>4</sup> Faculty of Epidemiology and Population Health, London School of Hygiene and Tropical Medicine, London, United Kingdom

#### Edited by:

Silvia Buroni, University of Pavia, Italy

#### Reviewed by:

Maria José Saavedra, Universidade de Trás os Montes e Alto Douro, Portugal Teruo Kuroda, Hiroshima University, Japan

#### \*Correspondence:

Taane G. Clark taane.clark@lshtm.ac.uk Isabel Couto icouto@ihmt.unl.pt

#### Specialty section:

This article was submitted to Evolutionary and Genomic Microbiology, a section of the journal Frontiers in Genetics

Received: 28 August 2018 Accepted: 18 December 2018 Published: 09 January 2019

#### Citation:

Costa SS, Sobkowiak B, Parreira R, Edgeworth JD, Viveiros M, Clark TG and Couto I (2019) Genetic Diversity of norA, Coding for a Main Efflux Pump of Staphylococcus aureus. Front. Genet. 9:710. doi: 10.3389/fgene.2018.00710 NorA is the best studied efflux system of Staphylococcus aureus and therefore frequently used as a model for investigating efflux-mediated resistance in this pathogen. NorA activity is associated with resistance to fluoroquinolones, several antiseptics and disinfectants and several reports have pointed out the role of efflux systems, including NorA, as a first-line response to antimicrobials in S. aureus. Genetic diversity studies of the gene norA have described three alleles; norAI, norAII and norAIII. However, the epidemiology of these alleles and their impact on NorA activity remains unclear. Additionally, increasing studies do not account for norA variability when establishing relations between resistance phenotypes and norA presence or reported absence, which actually corresponds, as we now demonstrate, to different norA alleles. In the present study we assessed the variability of the norA gene present in the genome of over 1,000 S. aureus isolates, corresponding to 112 S. aureus strains with whole genome sequences publicly available; 917 MRSA strains sourced from a London-based study and nine MRSA isolates collected in a major Hospital in Lisbon, Portugal. Our analyses show that norA is part of the core genome of S. aureus. It also suggests that occurrence of norA variants reflects the population structure of this major pathogen. Overall, this work highlights the ubiquitous nature of norA in S. aureus which must be taken into account when studying the role played by this important determinant on S. aureus resistance to antimicrobials.

Keywords: Staphylococcus aureus, norA, alleles, variability, efflux

## INTRODUCTION

Staphylococcus aureus is one of the major human pathogens in the hospital and community settings, causing a wide array of clinical manifestations, from mild skin infections to lifethreatening systemic infections (Chambers and DeLeo, 2009; Tong et al., 2015). Development and acquisition of resistance to antibiotics and other antimicrobials is of paramount importance in S. aureus, as exemplified by the common occurrence and dissemination of strains displaying a phenotype of multidrug resistance (MDR), including methicillin-resistant

Staphylococcus aureus (MRSA) strains (Chambers and DeLeo, 2009). MRSA are spread worldwide and are a public health threat, ranked amongst the major nosocomial pathogens (Lee et al., 2018; Tacconelli et al., 2018). MRSA epidemiology has revealed the local or global dissemination of a few number of clones, namely the lineages in clonal complexes CC1, CC5, CC8, CC22, CC30, CC45, CC59, and CC80 (Lee et al., 2018).

In recent years, several studies have supported drug efflux as a player in the emergence of resistance toward antibiotics and other antimicrobials in S. aureus (DeMarco et al., 2007; Furi et al., 2013; Kwak et al., 2013; Costa et al., 2015). Of particular interest are multidrug efflux pumps (MDR EPs), which extrude a wide range of chemically dissimilar antimicrobials, being frequently associated with MDR phenotypes in bacteria (Piddock, 2006; Poole, 2007). In S. aureus, more than twenty putative MDR EPs are encoded in the chromosome (Schindler et al., 2015), of which several have already been characterized (Costa et al., 2013b). Among these, NorA is the most well studied, being frequently used as a model for studying efflux-mediated resistance in this pathogen.

NorA is a 388 aminoacid protein with 12 transmembrane segments (TMS) that belongs to the Major Facilitator Superfamily (MFS) of secondary transporters. This MDR EP uses the proton motive force to extrude from the cell fluoroquinolones, ethidium bromide, quaternary ammonium compounds, and other antimicrobials (Yoshida et al., 1990; Kaatz et al., 1993; Neyfakh et al., 1993). Several reports have associated NorA activity to low-level resistance to fluoroquinolones and reduced susceptibility to biocides (DeMarco et al., 2007; Huet et al., 2008; Kosmidis et al., 2010; Costa et al., 2011, 2015; Furi et al., 2013). Other studies support a broader substrate range for NorA, including siderophores (Deng et al., 2012) and fusaric acid (Marchi et al., 2015).

The NorA encoding gene, norA, was first identified in the chromosome of a fluoroquinolone-resistant S. aureus isolate, collected in 1986 at a Japanese hospital (Ubukata et al., 1989). Early studies have shown that genetic diversity of this gene can be captured by three alleles, differing up to 10% at the level of the nucleotide sequence and 5% in the polypeptide sequence; norAI (Yoshida et al., 1990), norAII (also described as norA23) (Noguchi et al., 2004) and norAIII (also described as norA1199) (Kaatz et al., 1993). The occurrence of norA variants is also strengthened by a recent study by Brooks et al. (2018) that refers to the variability of this gene in a set of over 150 S. aureus strains. Although some studies have been conducted to ascertain the impact of this genetic diversity on the efflux activity of NorA (Schmitz et al., 1998; Sierra et al., 2000; Noguchi et al., 2004), this effect remains unclear.

Despite this early characterization, there are still contradictory reports in literature on the role of NorA in S. aureus effluxmediated antimicrobial resistance. Several studies have reported on the putative absence of norA, most probably due to failure to amplify this gene with primers directed to only one of the possible norA alleles. The present study aims at clarifying some of these aspects, by demonstrating that the norA gene is part of the core genome of S. aureus, reflecting on the genetic variability of the gene, its distribution amongst S. aureus clonal lineages, including both methicillin-resistant and -susceptible strains and possible impact on NorA function.

## MATERIALS AND METHODS

#### Study Datasets

Four sets of nucleotide sequences of the norA structural gene and the corresponding polypeptide sequences were used in this study, comprising (i) the sequences of the three norA alleles described to date in literature; (ii) the norA sequences from 112 S. aureus whole genome sequences retrieved from the GenBank database; (iii) the norA sequences from 917 MRSA strains sourced from a London-based study (Auguet et al., 2018) (ENA accession PRJEB11177); (iv) the norA sequences of nine MRSA isolates collected in a major Hospital in Lisbon, Portugal, representative of the circulating norA alleles in that hospital at the time (Costa et al., 2011, 2016), which were deposited in GenBank.

A detailed list of all the sequences comprised in this study and respective accession numbers can be found in **Supplementary Information S1**. The Lisbon MRSA isolates were previously characterized for their susceptibility toward fluoroquinolones and biocides and presence of mutations in the quinoloneresistance determining region of grlA/gyrA genes (Costa et al., 2011, 2013a) – **Supplementary Information S1**. Information on the remaining S. aureus strains was gathered from the GenBank database and relevant published papers.

Whenever sequence types (ST) were not provided, in silico MLST was performed using the MLST 1.8 (Larsen et al., 2012) and SRST2 (Inouye et al., 2014) softwares. The datasets were used to construct a pan-genome (set of all genes within a given species), and a core genome (set of genes found in all strains of that species) for S. aureus (Tettelin et al., 2005; van Tonder et al., 2014).

#### Sequence Analysis of norA and MLST Alleles

De novo assemblies were performed for all London samples using Velvet (Zerbino and Birney, 2008) and VelvetOptimiser (Zerbino, 2010). Resulting contig FASTA files and FASTA files of the S. aureus samples obtained from GenBank were annotated using Prokka (Seemann, 2014). The pan-genome of these samples was then constructed with Roary (Page et al., 2015), and the sequences of the norA and MLST genes (arcC, aroE, glpF, gmk, pta, tpi, and yqiL) isolated using custom R scripts (R Core Team, 2016).

For the set of the nine Portuguese MRSA clinical isolates, norA was amplified by PCR using three pairs of primers (**Table 1**). PCR reaction mixtures were prepared in 0.05 mL containing 2.5 U Taq Polymerase (Thermo Scientific, Waltham, MA, United States); 1X Taq buffer (Thermo Scientific); 30 pmol of each primer (Invitrogen, Carlsbad, CA, United States); 0.2 mM dNTPs (GE Healthcare, Chicago, IL, United States); 1.75 mM MgCl<sup>2</sup> (Thermo Scientific). The amplification conditions were the following: initial DNA denaturation step at 95◦C for 3 min, followed by 35 cycles of denaturation at 94◦C for 1 min, annealing at 52◦C [norA(a)], 45◦C [norA(b)] or 50◦C [norA(c)] for 1 min,

TABLE 1 | Primers used to sequence the norA promoter and coding region of Staphylococcus aureus clinical strains.


<sup>a</sup>Position of the primers in the genome of S. aureus DSM20231 (accession no. CP011526.1). <sup>b</sup>The primer was designed with the aid of the free software Primer3 (http://primer3.sourceforge.net/) and tested in silico (http://insilico.ehu.es/PCR/).

and extension at 72◦C for 1 min, followed by a final extension step at 72◦C for 5 min. The PCR products were sequenced using the same set of primers and the sequences analyzed and assembled to make up the entire fragment using the software MAFFT v. 6<sup>1</sup> and BioEdit v. 7.0.9.0<sup>2</sup> .

The norA and MLST gene sequences from all four datasets were aligned using custom R scripts. Samples that shared identical norA and MLST sequences were grouped together and representative sequences used for phylogenetic reconstruction. Maximum-likelihood (ML) tree construction was performed on the aligned norA and concatenated MLST gene sequences separately using RAxML (Stamatakis, 2014) with a GTRGAMMA evolutionary model. The resulting phylogenetic trees were displayed and annotated using FigTree<sup>3</sup> .

#### Analysis of the Impact of norA Variability on NorA Activity

The tridimensional structure of the NorA efflux pump was predicted via the in silico platform PHYRE2 (Protein Homology/analogY Recognition Engine v2.0<sup>4</sup> ) (Kelley et al., 2015) based on the nucleotide sequences of the three norA variants described in literature. A mutational analysis based on the predicted tridimensional structure was conducted with the Phyre2 Investigator tools, using the SuSPect algorithm (Yates et al., 2014). This algorithm produces a table of scores from 0 to 100 according to predicted deleteriousness (0 = neutral to 100 = deleterious). A score of 50 is recommended as a cut-off between neutral and deleterious variants, with extreme scores being more confident predictions (Yates et al., 2014). In this work, a score of ≥75 was used as cut-off value.

## RESULTS AND DISCUSSION

#### norA Is Part of the Core Genome of S. aureus

Our study sought to establish if norA is part of the core genome of S. aureus. This is a relevant question since a significant part of the studies on efflux-mediated resistance in this pathogen focus mainly on the activity of NorA and some studies have reported the absence of the norA gene in several S. aureus strains (Monecke and Enricht, 2005; Vali et al., 2008; Conceição et al., 2015; Hasanvand et al., 2015; Liu et al., 2015; Ammar et al., 2016; Taheri et al., 2016; Antiabong et al., 2017; Hassanzadeh et al., 2017; Hadadi et al., 2018; Kernberger-Fischer et al., 2018). This reported absence could be explained by the genetic diversity

FIGURE 1 | (A) The distribution of genes in the pan-genome of the London and GenBank S. aureus isolates (n = 1029). A total of 5,489 genes were identified with 1,551 (28.25%) genes found in more than 99% of strains. The majority of genes (55.3%) were identified as 'cloud' genes, found only in a small number of samples (<15%), demonstrating the diversity of the S. aureus genome. (B) Rarefaction curves for total (red line) and core (blue line) genes identified when increasing the sample size to reconstruct the pan-genome. Random sampling of samples was conducted 100 times and the mean size of the core genome and number of total genes [and standard deviation (dotted lines)] is shown as the sample size is increased by 1 to the total sample size (n = 1029). 90% of the total genes are identified after 514 samples (49.5% of total samples), and there is a plateau in size of the core genome (±5% of the final genome size) at sample 256 (24.9% of total samples). The rarefaction curve confirms that the gene diversity of the strains has been adequately captured with the number of samples included in the analysis.

<sup>1</sup>mafft.cbrc.jp/alignment/software/

<sup>2</sup>https://bioedit.software.informer.com/

<sup>3</sup>http://tree.bio.ed.ac.uk/software/figtree/

<sup>4</sup>http://www.sbg.bio.ic.ac.uk/phyre2

of the gene and a consequent primer failure during norA PCR screening. To test if norA is part of the core genome of S. aureus, we constructed the pan-genome from the 1029 S. aureus isolates whole genome sequence datasets (112 GenBank; 917 London S. aureus isolates). The core genome of S. aureus consists of 1551 genes (found in ≥99% of individuals with 95% BLAST identity; **Figure 1A**) with the genome of each individual isolate containing a median of 2196 genes (with a range of 1955 to 2469), marginally lower than the number of genes found in S. aureus reference strains (Li et al., 2011). Rarefaction curves were calculated and determined that the number of samples used was adequate to accurately reconstruct the pan-genome of the population; 90% of the total gene diversity was present when considering just 49.5% (514/1029) of the samples and there was a stabilization of the core genome size after 24.9% (256/1029) of the total samples (**Figure 1B**). The norA gene was found to be present in all London and GenBank samples and thus is part of the core genome, with each sample containing a single gene variant. Additionally, we performed a Blastn search using as query the nucleotide sequence of the norAI allele (GenBank accession no. D90119.1) against all S. aureus genome sequences (complete, scaffold or contig formats) deposited in GenBank<sup>5</sup> , corresponding to more than 8.000 (circa 8.150) sequences. All the Blastn searched sequences retrieved hits with ≥90% identity to norAI allele. This result supports the notion that norA is a S. aureus core gene, occurring in all S. aureus genomes. The finding that norA is part of S. aureus core genome and thus is present in all S. aureus strains, implies that reporting the detection of this gene is not sufficient to make a direct association with a particular resistance phenotype. To make such a correlation, one must carry out expression analysis of the norA gene as well as of other efflux pump genes, as it has been shown that S. aureus strains can display different efflux pump gene expression patterns (DeMarco et al., 2007; Kosmidis et al., 2010, 2012), even under pressure of the same antimicrobial (Huet et al., 2008; Costa et al., 2011, 2015).

## Genetic Diversity of the norA Gene

One of the goals of this study was to determine the distribution of the norA variants amongst the several S. aureus clonal lineages. A phylogenetic analysis was performed with the three

<sup>5</sup>www.ncbi.nlm.nih.gov/genome/genomes/154

norA alleles and norA nucleotide sequences retrieved from complete S. aureus genomes available in GenBank and from the Portuguese and London-based S. aureus collections (**Figure 2**). As expected, a certain degree of genetic variability was revealed by the maximum-likelihood tree obtained, despite this gene being relatively conserved. **Figure 3** shows the distribution of the pairwise genetic distance within norA across the samples used to construct the phylogenetic tree, with a maximum of 143 SNP differences between samples (12.3% of the total gene length of 1164 bp).

A global analysis showed a clear correlation between occurrence of norA alleles and specific S. aureus clonal lineages. The norAI allele was mainly associated with genetic backgrounds of the clonal complexes CC1, CC5, CC8, and CC80 or closely related lineages. On the other hand, norAII and close variants appear to be restricted to strains belonging to CC30, CC398, and CC22 (**Figure 2**). The norAIII allele is only associated with the S. aureus lineage CC45 and novel but related STs.

Among the sequences retrieved from GenBank, the norAI allele is the most frequent (76 out of 112 strains) and the norAII allele the second most frequent allele (21 out of 112 strains), whereas norAIII occurs in just two CC45 strains. We also detected 13 more divergent variants, most closely related to the norAI allele, that were associated with other lineages, such as CC59 and CC121 (**Figures 2**, **3**). Although these findings may be biased as the dataset analyzed reflects the current epidemiologically S. aureus relevant clones, with a frequency of ∼70% MRSA strains (78 out of 112 genomes), it clearly shows that norAI and related variants are the most prevalent alleles across lineages, as suggested by earlier studies (Schmitz et al., 1998; Sierra et al., 2000; Noguchi et al., 2004). Interestingly, a sample isolated in Brazil in 2010 (FCFHV36) carried a norAI type variant of the norA gene that included a frameshift mutation in codon 129. This mutation caused a premature stop codon at codon number 147, resulting in potential pseudogenization of norA in this isolate.

Regarding the London isolates, we found the norAII allele to be the most common allele, identified in 671/917 samples, due to the high number of CC22 strains in the study samples (539/917). As with the GenBank samples, the norAIII allele was the least common variant found in the London isolates, with only the three CC45 and four novel ST strains carrying closely related variants. The norAI allele was found in 220/917 strains consisting of the broadest range of lineages, comprising 21 different STs. Additionally, one ST22 sample possessed a norA variant with a frameshift mutation at codon 365 that results in a premature stop codon at codon number 367 and, again, possible pseudogenization.

Nine MRSA isolates, representative of a collection of 53 S. aureus clinical isolates isolated at a major Portuguese Hospital and previously characterized for efflux activity and clonal lineage (Costa et al., 2011, 2013a, 2015, 2016) were added to the study to ascertain their norA allele. The full sequence of the norA gene was determined for these isolates and compared with norA alleles from major clonal lineages (**Figure 4**). Both the norAI and norAII alleles were found amongst the Portuguese isolates and their distribution reflects the findings for the other datasets analyzed. Eight isolates from

norAI group, norAII group, norAIII group, and CC121/CC59 group.

FIGURE 4 | Multiple alignment of norA sequences of the three norA alleles, from selected Portuguese MRSA clinical isolates (SM) and from strains of representative clonal lineages. The MRSA strains displayed were selected to represent the main clones found, namely SM10 – CC5; SM3 – CC22; MW2 – CC1; Newman – CC8; MRSA252 – CC30. The Shine-Dalgarno (SD) sequence is underlined in black. The shaded regions correspond to the putative TMS of NorA as suggested by Paulsen et al. (1996). The aminoacid regions inside the boxes are relative to the conserved motifs described for MFS transporters with 12 TMS. The black residues shown above the main polypeptide sequence represent the variations present in the various NorA polypeptide sequences relatively to the NorAI variant. The red residues represent alterations found among the SM collection of Portuguese MRSA clinical isolates. The red boxes indicate the primers described in Table 2 for distinction of norA alleles.

clonal lineages CC5, CC8, and CC88 carried norAI and a single isolate from clonal lineage CC22 harboured the norAII allele (**Figure 4**). Combining information on norA allele, clonal lineage and previous characterization of these isolates (Costa et al., 2011) showed no correlation between norA allele, norA expression levels, efflux capacity, or antimicrobial resistance levels (**Supplementary Information S1**).

A wet lab PCR-based strategic approach is proposed for detection and differentiation of the several norA alleles occurring in S. aureus. The norAI and norAII alleles can be easily distinguished using two sets of specific primers (**Table 2** and **Figure 4**). To amplify and differentiate norAIII from the norA allele associated with CC59/CC121, an approach consisting in PCR amplification followed by restriction of the PCR product with HindIII can be applied (**Table 2** and **Figure 4**).

#### Impact of Genetic Variability on NorA Function

A mutational analysis was performed in silico via the PHYRE2 platform to construct a NorA structural model, using the SuSPect algorithm to predict the effect of each residue substitution on

TABLE 2 | List of primers proposed to amplify and differentiate the norA alleles by conventional PCR.


<sup>a</sup>The primers were designed with the aid of the free software Primer3 (http://primer3.sourceforge.net/). The primers and the HindIII restriction were tested in silico (http://insilico.ehu.eus/) against several S. aureus genome sequences of representative clonal lineages.

NorA function. This analysis is limited because of the lack of available structural data for bacterial MFS pumps, mostly limited to substrate-specific transporters. Nevertheless, the polypeptide sequences encoded by the three norAI alleles were used to predict a tri-dimensional structure of NorA. All three NorA variants showed the highest identity with the putative MFS transporter YajR of Escherichia coli (Jiang et al., 2013). A second NorA in silico model, based on the structure of glycerol-3 phosphate MFS transporter from E. coli, described by Bhaskar et al. (2016) was taken into account for comparison purposes. The SuSPect-derived mutational analysis revealed 42 residues for which particular mutations could potentially impair NorA activity (**Supplementary Information S2**). Of these, residues Pro110, Pro158, Pro311, and Gly326 were particularly susceptible to mutations.

Analysis of the polypeptide sequence corresponding to the three norA variants revealed a total of 43 amino acid alterations. **Figure 5** summarizes these alterations, their location in NorA and their distribution amongst S. aureus lineages. There was no overlap between these alterations and the residues identified by SuSPect. We also compared the S. aureus NorAI variant with the NorA efflux pump encoded in the Staphylococcus epidermidis ATCC 12228 chromosome, which presents the same substrate profile (Costa et al., 2018; **Supplementary Information S2**). The differences encountered between the two polypeptide sequences correspond to residues that were not highlighted as pivotal for the protein activity by SuSPect. These results suggest conservation of NorA function in these two main staphylococcal species.

One alteration, Gly147Ser located within putative TMS 5, may affect NorA function (**Figure 4**). This alteration was detected in all four strains belonging to ST105 (three of which are of Portuguese origin) and in one of the strains belonging to ST225 (Portuguese origin) (**Supplementary Information S2**), but was absent from strains from other CC5 lineages, such as ST5. Residue 147 is predicted to be located within the conserved motif C (gxxxGPxiGGxl) (Paulsen et al., 1996) on the TMS 5, facing a water-filled channel. This motif is conserved among MFS drug-H<sup>+</sup> antiporters and has been implicated in the binding of the substrate, with the glycines being essential for that function (Ginn et al., 2000; Tamura et al., 2001). In fact, in a mutagenesis study in S. aureus efflux pump TetA(K), also a 12-TMS member of the MFS family, the mutation Gly147Ser was responsible for the loss of 80% of the activity of that efflux pump (Ginn et al., 2000). Although Tet transporters and NorA are distinct in their substrate specificity, these pumps share abundance of glycines in TMS 5, a trait that has been postulated to confer conformational plasticity to EPs (Ginn et al., 2000). Thus, the

found for the strains with genomes publicly available and for the Portuguese strains (SM10 – CC5; SM39 – CC88). Sequences are grouped according to the respective norA allele (orange – norAI; green – norAII; blue - norAIII; yellow – norA-CC59/121). The lines correspond to the transmembrane segments (TMS) of NorA, as predicted by Paulsen et al. (1996) (green), by the HMMTOP method as presented in Bhaskar et al. (2016) (red) and as predicted by the Phyre2-based NorA model (purple). The aminoacid regions inside the boxes are relative to the conserved motifs described for MFS transporters with 12 TMS (Paulsen et al., 1996). The red box corresponds to the amino acid substitution likely to affect NorA function.

mutation Gly147Ser may have a deleterious effect on NorA activity. Additional studies are necessary to fully ascertain the impact of this alteration in the efflux capacity of NorA. The same applies to the full understanding of possible correlations between a given norA allele and corresponding NorA efflux capacity, probably requiring carefully controlled genetic systems because of overlapping substrates between many staphylococcal multidrug efflux systems and the redundancy in their response to antimicrobials (Costa et al., 2013b; Schindler and Kaatz, 2016).

#### CONCLUSION

In this study we provide evidence that the efflux pump gene norA is part of core genome of S. aureus. This gene presents some genetic variability which may impair its detection by conventional laboratorial techniques, such as PCR, due to potential primer mismatch. Strikingly, a correlation was observed between the several norA alleles and specific S. aureus lineages, suggesting that the occurrence of norA variants reflects the population structure of this major pathogen.

#### DATA AVAILABILITY

The sequences analyzed in this study are available in GenBank and their accession numbers detailed in **Supplementary Information S1**, except for WGS data for the London-based strains that are available from the European Nucleotide Archive database under accession number PRJEB11177.

#### AUTHOR CONTRIBUTIONS

IC, SC, and TC conceived and designed the study. SC performed the wet lab norA analysis and the in silico mutational analysis. BS

#### REFERENCES


conducted the pan-genome analysis. SC and BS performed the phylogenetic analysis. JE provided genomic data on the Londonbased MRSA strains. RP, JE, MV, TC, and IC contributed to the analysis of these data. SC, BS, TC, and IC wrote the first draft of the manuscript. All authors have reviewed the final version of the manuscript.

#### FUNDING

This work was partially supported by Fundação para a Ciência e a Tecnologia (FCT, Portugal), through funds to GHTM – UID/Multi/04413/2013. SC was supported by grant SFRH/BPD/97508/2013 from FCT, Portugal. TC was funded by the Medical Research Council United Kingdom (Grant Nos. MR/K000551/1, MR/M01360X/1, MR/N010469/1, and MR/R020973/1) and BBSRC United Kingdom (BB/R013063/1). BS was funded by the Medical Research Council United Kingdom (Grant No. MR/N010469/1).

#### ACKNOWLEDGMENTS

The MRC eMedLab computing resource was used for bioinformatics and statistical analysis. The authors deeply acknowledge Professor José Melo-Cristino for providing the Portuguese S. aureus strains analyzed in this study.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fgene. 2018.00710/full#supplementary-material


fluoroquinolones and biocides in a major opportunistic pathogen. J. Antimicrob. Chemother. 73, 320–324. doi: 10.1093/jac/dkx400



Bayesian decision model. PLoS Comput. Biol. 10:e1003788. doi: 10.1371/journal. pcbi.1003788


**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 © 2019 Costa, Sobkowiak, Parreira, Edgeworth, Viveiros, Clark and Couto. 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) and the copyright owner(s) 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 and Distribution of Tetracycline Antibiotics and Resistance Genes in Longshore Sediments of the Three Gorges Reservoir, China

Lunhui Lu<sup>1</sup>† , Jie Liu<sup>2</sup>† , Zhe Li1,2 \*, Zhiping Liu<sup>2</sup> \*, Jinsong Guo<sup>2</sup> , Yan Xiao<sup>1</sup> and Jixiang Yang<sup>1</sup>

<sup>1</sup> CAS Key Laboratory of Reservoir Aquatic Environment, Chongqing Institute of Green and Intelligent Technology, Chinese Academy of Sciences, Chongqing, China, <sup>2</sup> Key Laboratory of the Three Gorges Reservoir Region's Eco-Environment, Ministry of Education, Chongqing University, Chongqing, China

#### Edited by:

Gian Maria Rossolini, Università degli Studi di Firenze, Italy

#### Reviewed by:

Andrea Di Cesare, Università di Genova, Italy Luciana Migliore, Università degli Studi di Roma Tor Vergata, Italy

#### \*Correspondence:

Zhe Li lizhe@cigit.ac.cn Zhiping Liu liulqs@163.com †These authors have contributed equally to this work and share first

Specialty section:

authorship

This article was submitted to Evolutionary and Genomic Microbiology, a section of the journal Frontiers in Microbiology

Received: 16 April 2018 Accepted: 30 July 2018 Published: 17 August 2018

#### Citation:

Lu L, Liu J, Li Z, Liu Z, Guo J, Xiao Y and Yang J (2018) Occurrence and Distribution of Tetracycline Antibiotics and Resistance Genes in Longshore Sediments of the Three Gorges Reservoir, China. Front. Microbiol. 9:1911. doi: 10.3389/fmicb.2018.01911 The widespread use of antibiotics and the induced antibiotic resistance genes have attracted much attention in recent years. The longshore sediments in the waterlevel-fluctuating zone of the Three Gorges Reservoir were selected to investigate the spatial-temporal distribution of antibiotics and antibiotic resistance genes in two different operation stages (low-water level in summer and high-water level in winter). Three kinds of tetracycline antibiotics (tetracycline, oxytetracycline, and chlortetracycline) and three kinds of tetracycline resistance genes [tet(A), tet(C), and tet(M)] were analyzed and quantified. The results showed that the distribution of tetracyclines and resistance genes in riverine, transition and lacustrine zones showed a certain regularity, and the tetracycline antibiotics concentration and the total abundance of the tetracycline resistance genes were highest in the transition zone, and then the riverine zone. Meanwhile, there were significant seasonal variations of tetracycline and the resistance genes. The concentrations of the tetracycline and resistance genes were higher in summer than those in winter, while the relative abundance of resistance genes was higher in winter. It was suggested that the different seasonal distribution of antibiotics and resistance genes may be correlated with the reservoir operation in the Three Gorges Reservoir and the higher use of antibiotics in winter. In addition, Pearson correlation analysis showed that the concentrations of the tetracycline, class 1 integron and 16S rRNA were positively correlated with the abundance of the tetracycline resistance genes.

Keywords: Three Gorges Reservoir, longshore sediments, tetracycline antibiotics, antibiotic resistance genes, spatial-temporal distribution

## INTRODUCTION

In recent years, with the widespread abuse of antibiotics in human medicine and animal husbandry, exogenous antibiotics and antibiotic resistance genes (ARGs) in the environment have become an increasingly global problem (Sarmah et al., 2006). China leads the world in antibiotic production capacity, with a significant percentage of antibiotics utilized in animal agriculture and medicine

(Luo et al., 2010). In general, after the consumption, only a small portion of antibiotics can be bio-degraded, most antibiotics are discharged into the environment without being metabolized (Xiong et al., 2018). As the most commonly applied antibiotics, tetracyclines (TCs), such as tetracycline (TC), oxytetracycline (OTC), and chlortetracycline (CTC), are broad-spectrum and affordable (Patterson et al., 2007). Accordingly, TCs are used extensively in medical treatment, as well as for veterinary drugs and growth promoter in poultry farming and aquaculture, particularly in developing countries (Chopra and Roberts, 2001; Roberts, 2003). Remarkably, compared with other antibiotics, TCs are more persistent, highly adsorptive and hardly degradable in soils and sediments (Chee-Sanford et al., 2009). Thus, they are also more easily accumulated (Rabolle and Spliid, 2000; Sarmah et al., 2006; Jiang et al., 2011), posing a potential threat to ecosystems and human health.

TCs may trigger the production of tetracycline resistance genes, which directly cause environmental pollution (Kim, 2004). A typical feature of ARGs is that they can be transferred horizontally between microorganisms using mobile genetic elements as carriers, such as plasmids and transposons (Zhou et al., 2014; Lin et al., 2015). As a type of genetic assembly platform, integrons associated with these mobile genetic elements are considered as the key players in horizontal gene transfer, particularly class 1 integrons (Gaze et al., 2011; Berglund et al., 2014). Through this horizontal gene transfer, ARGs can migrate and transform from different environmental media, and are likely to be introduced into the food chain along with the resistant plasmid and finally enter the human body to increase human antibiotic resistance (Ghosh et al., 2009; Stoll et al., 2012; Sengupta et al., 2013). The aquatic environment is an important pool for ARGs because many pollutants from wastewater treatment plants, industries, hospitals and swine farms finally circulate in water environments and drive the propagation of ARGs (Zhang et al., 2009; Lupo et al., 2012; Yan et al., 2014a,b; Yang et al., 2017). Sediments may act as a sink but also as a secondary source of various contaminants including antibiotics and ARGs (Su et al., 2014), constituting a great potential danger for aquatic organisms (Siedlewicz et al., 2017). Hence, the diversity of antibiotics and ARGs in sediments of the watershed is important for us to understand the diffusion of antibiotic resistance at the catchment scale. To date, reports on the pollution of antibiotics and ARGs in sediments are mainly concentrated in the coastal areas, some rivers and lakes (Pei et al., 2006; Luo et al., 2010; Su et al., 2014; Yang et al., 2016; Calero-Caceres et al., 2017; Siedlewicz et al., 2017). However, there are relatively few studies on longshore sediments in the large reservoir.

As the largest reservoir in China, the Three Gorges Reservoir (TGR) performs several functions, such as flood control, power generation, navigation, irrigation and aquaculture, and its ecological functions are crucial. Since the completion of the Three Gorges Project, the operation of the reservoir has been implemented as an anti-seasonal artificial water level regulation and management method for "winter storage and summer drainage." The water level rises to 175 m in October and begins to decline in May of the following year. From June to September, the reservoir water level is maintained at the lowest level of 145 m. Periodic changes in reservoir water level affect the sediment deposition in the riparian zone (Wang et al., 2016), which inevitably leads to a change in the transport characteristics of pollutants adsorbed on sediment. Hence, the purpose of this study was to focus on detecting and discussing the spatialtemporal distribution of TCs and tetracycline resistance genes in the longshore sediments of the TGR.

In this study, eight sampling sites were selected in the sediments of the water-level-fluctuating zone of the TGR. Three tetracycline antibiotics (TC, OTC, and CTC) were determined by liquid chromatography-mass spectrometry (LC-MS). Three tetracycline resistance genes [tet(A), tet(C), and tet(M)], one genetic element (class 1 integron, intI1) and 16S rRNA gene abundance were assessed by real-time quantitative PCR to reveal the presence of the tetracycline resistance genes in the TGR. This study would give insight into further research on assessment of antibiotics and antibiotic resistance genes pollution in the TGR.

## MATERIALS AND METHODS

#### Site Selection and Sampling

The TGR is formed after the completion of the Three Gorges hydropower station, with a water surface area of 1084 km<sup>2</sup> . This reservoir is from Chongqing (west) to Yichang of Hubei province (east), and the distance is approximately 662.9 km (Chang et al., 2010). According to Thornton et al. (1990), a pure river-like reservoir can be divided into three longitudinal zones: riverine zone, transition zone and lacustrine zone. A riverine zone is located at the upper domain that has a narrow and channelized basin with relatively high flow rates; the lower part immediately upstream of the dam is often described as the lacustrine zone, which has a broad, deep and lakelike basin with low flow rates; a transitional zone is defined between the riverine and lacustrine zones that has a relatively broader and deeper basin with modest flow rates (Kimmel and Groeger, 1984; Thornton et al., 1990). Longitudinal zonation in a reservoir changes in the zone extension due to the difference in hydrodynamic conditions. TGR is periodically regulated by dam operations, and thus exhibits the characteristics of both rivers and lakes, showing the same zonation characteristics. The TGR is a typical reservoir containing three typical zones (riverine zone, transition zone and lacustrine zone) according to the differences in hydrological characteristics, and these three zones are distinguishable along the longitudinal axis (Li et al., 2017). The average water velocity in the sites Zhutuo, Mudong and Fuling (about 0.6∼1.05 m/s in August and 0.30 in December), Fengjie and Zigui (about 0.2 m/s in August and <0.1 m/s in December) (**Supplementary Figure S1**). Moreover, the hydraulic retention time in the TGR varied from 6∼19 days between the sites ZT and FL, from 11∼36 days between Zhongxian and Wanzhou, from 35∼110 days between the sites FJ and ZG. According to the hydraulic retention time and average water velocity calculated, the sampling sites were classified qualitatively. For this reason, eight sampling sites were established along the reservoir: three sites belong to the riverine zone (ZT, MD, and

FL), two sites belong to the transition zone (ZX and WZ), two sites belong to the lacustrine zone (FJ and ZG), and one site YC after the dam (**Figure 1**). Detailed information of these sampling sites was summarized in **Table 1**. According to its regulation plan, the water level in the TGR is impounded to 175 m for power generation during the winter and discharged to 145 m for flood control during the summer (Li et al., 2010). Water residence time in August and December represented two periods of the reservoir, i.e., the lotic and lentic periods throughout the year. In this case, sediment samples were collected in the TGR during periods of low-water level (August 2015) and high-water level (December 2015).

Sampling sites were located along the riparian zone within a distance of 10 m from the water edge, which were subjected to alternate inundation and exposure due to the fluctuation of water level. At each sampling site, three surface sediment subsamples (0–20 cm) were randomly collected by a small shovel. After mixing and splitting into about 1 kg by quartering technique, samples were taken to the laboratory in sterile containers and stored at −80◦C for further treatment and analysis (Zhang et al., 2012).

## Physicochemical Properties of the Samples

Sediment samples were air-dried and sieved through a mesh screen (<0.149 mm) and then stored at 4◦C before use. Nitrate (NO<sup>3</sup> <sup>−</sup>-N), nitrite (NO<sup>2</sup> <sup>−</sup>-N) and ammonium (NH<sup>4</sup> <sup>+</sup>-N) were extracted from the sediments using 2 M KCl. The extracted NO<sup>3</sup> <sup>−</sup>-N was then determined by a UV spectrophotometry method (Song et al., 2007); the extraction liquid containing NO<sup>2</sup> <sup>−</sup>-N and NH<sup>4</sup> <sup>+</sup>-N was subsequently analyzed according to previously published methods (International Organization for Standardization [ISO], 2003). Water-soluble organic carbon (WSOC) was measured using a Vario TOC Cube analyzer (Elementar, Hanau, Germany). All the above determinations were carried out in triplicate, and the average results were calculated.

## Extraction and Determination of Tetracyclines

Sediment samples were extracted using the most common solidphase extraction method (Kim and Carlson, 2007a). First, the freeze-dried samples were broken into 0.3 mm particles, and 20 mL McIlvaine buffer was added to 4 g samples. The upper extract of the mixed sample after 30 min of shaking was filtered through a 0.45 µm filter. Then, the filtrate was extracted by an OASIS hydrophilic-lipophilic balance solid-phase extraction column (Waters, Milford, MA, United States), and 5 mL of methanol was added to elute the antibiotics. The eluent was dried by a nitrogen blower, then sealed with 1 mL of methanol and stored at −20◦C for LC-MS.

The target compounds TC, OTC and CTC in the extracts were determined by LC-MS. First, analyses were carried out using a Waters ACQUITY UPLCTM system (Waters, Milford, MA, United States). The chromatographic conditions were as follows: chromatographic column, Waters ACQUITY UPLC BEH C18 column (1.7 µm; 2.1 mm × 100 mm); mobile phase, acetonitrile (A) and 0.1% formic acid (formic acid/ultrapure water, V/V) (B); flow rate, 0.2 mL/min; injection volume, 2 µL; column temperature, 30◦C. Then, mass spectrometry was performed. The mass spectrometry system consisted of a Waters Micromass Quattro Premier XE (triple-quadrupole) detector and an ion source. The mass spectrometry conditions were as follows: ion source, ESI (+); source temperature, 110◦C; desolvation temperature, 400◦C; capillary voltage, 2.80 kV; desolvation gas flow, 600 L/h; cone gas flow, 50 L/h. The detailed mass spectrometric operating conditions are found in reference (Jia et al., 2009). For each sample, three replicate assays were performed for LC-MS analysis. The tetracyclines concentrations were all determined based on a dry weight.

## DNA Extraction and PCR

fmicb-09-01911 August 17, 2018 Time: 12:39 # 4

The DNA in the longshore sediment samples was extracted in triplicate with the PowerWater DNA Extraction Kit (Mo Bio, CA, United States) according to the manufacturer's protocol. The DNA extracts were pooled for each sample to reduce sample variability for the following research. The content and purity of the extracted DNA were detected by 1% (weight/volume) agarose gel and Nanodrop ND-1000 (Nanodrop, United States). The A260/A280 values ranged from 1.8 to 2.0, and A260/230 ranged from 2.2 to 2.5, indicating that the purity of the DNA obtained was high.

Three tet genes [tet(A), tet(C), and tet(M)], one genetic element (class 1 integron, intI1) and the 16S rRNA gene were investigated. Information about the primers and the annealing temperature is listed in **Table 2**. PCR mixtures contained 25 µL of 2× Power Taq PCR MasterMix (TIANGEN, Beijing, China), 1 µL of each primer (25 µM), and 10 ng of the DNA extracts in a total volume of 50 µL. PCR amplification was run on a MyCycler (Bio-Rad, Hercules, CA, United States) with an initial cycle of denaturation (3 min at 95◦C) followed by 30 cycles (30 s at 94◦C, 30 s at the annealing temperature (**Table 2**) and 60 s at 72◦C), and a final elongation step (5 min at 72◦C).

The above PCR amplification products were analyzed using electrophoresis with a Bio-Rad Gel Doc2000 on 2% agarose gel electrophoresis. The universal DNA purification and recovery kit (centrifugal column) was then used to purify the sediment DNA in the PCR reaction solution.

## Gene Quantification With Real-time Quantitative PCR

Quantitative PCR (qPCR) was used to quantify the target genes. The 16S rRNA gene was included to quantify the total bacterial load and to normalize the abundance of ARGs (ARG copies/16S rRNA gene copies, defined as relative abundance) in the collected samples. Real-time QPCR was performed on an iCycler iQ5 thermocycler (Bio-Rad, Hercules, CA, United States) to determine the abundance of the target genes in the samples. Standard plasmids carrying target genes were obtained by TA clones and extracted using a TIANpure Mini Plasmid kit (TIANGEN, Beijing, China) (Yu et al., 2005). Concentrations of the standard plasmids (ng/µL) were determined with the Nanodrop ND-1000 spectrophotometer (Nanodrop, United States). Copy concentrations (copies/µL) were then calculated by the following formula (Pei et al., 2006).

copy concentration (copies/g)=

DNA mass concentration (ng/g) DNA molecular weight (g/mol) .

The 25 µL qPCR mix contained 2 ng of sediment microbial DNA, 0.75 µL of each primer (25 µM), 10 µL of 2.5× RealMasterMix (TIANGEN, Beijing, China) and 1.5 µL of 20× SYBR solution (TIANGEN, Beijing, China). The protocol was as follows: 15 min at 95◦C; 40 cycles consisting of 10 s at 95◦C, 20 s at the annealing temperature and 30 s at 72◦C; followed by a final


TABLE 2 | Information of the PCR primers and amplification conditions.


extension of 30 s at 72◦C. The fluorescent signal was measured at the end of each extension step (Bustin et al., 2009). The melting process was automatically generated by the iCycler iQ5 system.

For the generation of the standards for all qPCR assays, a plasmid construct was used containing the inset of each ARG obtained by conventional PCR. The following requirements were satisfied to obtain reliable quantification (**Supplementary Table S1**): R<sup>2</sup> was higher than 0.99 for all standard curves over five orders of magnitude. Amplification efficiencies based on slopes were between 95 and 110%. The specificity was assured by the melting curves and gel electrophoresis. The minimum quantification limits for all target genes were within the range of 1.0 × 10<sup>2</sup> to 1.0 × 10<sup>3</sup> copies per µL DNA. The qPCR performed included two sets of standards and no-template negative (sterile water) controls on all samples and standards. The presence of PCR inhibitors in the DNA extracted from sediment samples was examined by diluting the DNA extract and mixing a known amount of standard DNA to a DNA extract before qPCR. In none of these cases was inhibition detected. The target genes concentrations were all determined based on a dry weight.

#### Statistical Analysis

Pearson correlation analysis was conducted to detect significant differences between physicochemical parameters and the target genes. Differences with p < 0.05 were considered statistically significant. The distribution map of antibiotics and resistance genes was completed by Origin 8.0. To understand the distribution characteristics of resistance genes at different sites, principal component analysis (PCA) was performed with Canoco 4.5. Linear-regression analysis was used to assess the association between TCs, intI1, 16S rRNA and ARGs. Statistical analyses were performed using SPSS 19.0 (IBM, United States).

## RESULTS

### Physicochemical Parameters of Sediment Samples

The variation of the physicochemical parameters of longshore sediments at the eight sampling sites is presented in **Table 3**. There were significant differences between the parameters at different sites. The mean values of all physicochemical descriptors at site WZ were higher in winter than those in summer. For the other sampling sites, most parameters at the upstream locations of ZT, MD, FL, and ZX appeared analogous distribution characteristics, while only NH<sup>4</sup> <sup>+</sup>-N at the downstream locations of FJ, ZG, and YC were higher in winter than those in summer. In terms of the physicochemical parameters, the mean values of organic matter were higher in summer than in winter. In contrast, WSOC, NH<sup>4</sup> <sup>+</sup>-N, NO<sup>2</sup> <sup>−</sup>-N, and NO<sup>3</sup> <sup>−</sup>-N were higher in winter. Briefly, seasonal differences were noticed between these parameters (p > 0.05).

Pearson correlation analysis of these indicators at each site showed that the conductivity and WSOC were significantly correlated with NO<sup>3</sup> <sup>−</sup>-N, with p values of 0.023 and 0.040, respectively, while there was no obvious correlation between other physicochemical indicators (**Supplementary Table S2**).

#### Spatial-Temporal Distributions of Tetracyclines

Of the three tetracycline antibiotic compounds, CTC was only detected at Wanzhou site in the summer. TC and OTC were


Abbreviations: a, standard deviation; b, conductivity; c, organic matter; d, water-soluble organic carbon.

detected at all sampling sites in both sampling seasons. The results of the concentrations of all detected tetracyclines are shown in **Figure 2**. The TC concentration ranged from 12.11 to 1142.67 ng/kg in the summer, 11.74 to 347.21 ng/kg in the winter. For OTC, the concentration ranged from 0.33 to 216.93 ng/kg in the summer and 0.12 to 156.44 ng/kg in the winter. In addition, the mean concentrations of TC and OTC were 263.60 and 54.73 ng/kg, respectively. Clearly, TC was the dominant tetracycline antibiotic in the TGR, contributing 20.45–99.74% to the total tetracycline antibiotic detected.Except for ZT and TC concentrations at all sites were noticeably higher in the summer. For OTC, only the concentrations at ZT, ZX, WZ, and FJ were slightly higher in summer, indicating that the distribution of OTC did not have a clear seasonal variation. On the other hand, the concentrations of TC and OTC in WZ were the highest, followed by MD and FL. The concentrations of TC and OTC were added together as the sum of the concentrations of tetracyclines at each site, and recorded as TCs (the gray bars in **Figure 2**). It was found that the concentration of TCs in the transition zone (ZX and WZ) was the highest, followed by the riverine zone (ZT, MD, and FL), and the lacustrine zone (FJ and ZG) was the lowest.

#### Distribution of Tetracycline Resistance Genes

**Figure 3** shows the occurrences and quantities of ARGs and intI1 in the sediments of the TGR. Of tetracycline resistance genes analyzed, tet(C) and tet(M) had 100% detection frequency in all eight sediment samples of the TGR, followed by tet(A) (87.5%). The concentrations of these three genes varied greatly by five orders of magnitude, ranging from 2.11 × 10<sup>2</sup> copies/g [tet(M) of site MD in the winter] to 3.23 × 10<sup>7</sup> copies/g [tet(A) of site MD in the summer]. Additionally, among the three investigated genes, the absolute and relative abundance of tet(A) was the highest, while the tet(M) gene was found with the lowest average absolute abundance and relative abundance. IntI1 was also detected in sediment samples; however, the detection frequency (75%) was lower than that of tet genes, with mean absolute abundance of 1.91 × 10<sup>5</sup> copies/g and mean relative abundance of 7.14 × 10<sup>5</sup> copies/16S rRNA. On the other hand, from the time perspective, the absolute concentration of ARGs and intI1 in summer was higher than that in winter, while the relative abundance of ARGs and intI1 was the opposite.

The principle component analysis (PCA) was carried out to reveal the distribution characteristics of the target genes at different sites (**Figure 4**). As shown in the figure, ZT was located at the end of the PCA graphs. Principal components 1 and 2 accounted for 87.2% of the total sample variability (64.2 and 23.0% for PC1 and PC2, respectively). PCA coordinates showed comparatively high discrepancies between the different sites and periods in the TGR. In summer, WZ and MD had a highly similar distribution of resistance genes, and it was probably due to the sites affected by human activities seriously. Sites of the lacustrine zone (FJ and ZG) clustered together in the PCA plot, showing a certain similarity. However, the riverine and transition zone did not show similarity between the sites in the same zone. In winter, the two sites of the riverine zone (MD and FL) and the two sites of the transition zone (ZX and WZ) were clustered together.

#### Distribution of TCs and ARGs in Relation to Environmental Variables Within the TGR

To explore the factors influencing the distribution characteristics of TCs and ARGs, some environmental factors (physicochemical indicators detected in this study) and social factors (population density, GDP per capita, GDP, land area, and permanent population<sup>1</sup> ) were selected for Pearson correlation analysis. In summer, no correlation was found between any factors (**Supplementary Table S3**). But in winter, there was a significant correlation between intI1 and organic matter, NH<sup>4</sup> <sup>+</sup>-N, NO<sup>2</sup> <sup>−</sup>-N, and NO<sup>3</sup> <sup>−</sup>-N (**Supplementary Table S4**). Besides, an obvious correlation was found between intI1 and permanent population, land area, population density, organic matter, which indicating that intI1 was greatly affected by these factors in winter. When all data in both seasons were grouped together, as shown in **Supplementary Table S5**, correlation analysis demonstrated that only tet(A) had negative correlation relationship with NO<sup>2</sup> <sup>−</sup>-N (p = 0.047).

#### Correlation Between ARGs Abundances and Antibiotic Concentrations

As shown in **Figure 5**, significant correlations existed between TC concentration and the three tet genes abundance at the p < 0.01 level. The correlation between OTC concentration and tetracycline resistance genes was different. It was noteworthy that no correlation was found between tet(A), tet(C) abundance and OTC concentration, but there was a significant correlation between tet(M) abundance and OTC concentration. In addition, Pearson correlation analysis also showed that intI1 and 16S rRNA abundance had positive correlation relationships with the three tet genes abundance at the p < 0.01 level.

<sup>1</sup>http://www.cqtj.gov.cn/

## DISCUSSION

#### Relationships Between TCs, intI1, 16S rRNA, and ARGs

Tet genes can be classified into four categories according to their different resistance mechanisms: efflux genes, ribosomal protection genes, enzymatic genes and other genes (Roberts, 2005). The three tet genes detected in this study cover two of them: the efflux genes [tet(A) and tet(C)] and the ribosome protection genes [tet(M)]. Previous studies had shown that efflux genes were more common in the environment compared with ribosomal protection genes (Borjesson et al., 2010). This may be the reason why the abundance (absolute and relative) of tet(M) was quite different from that of tet(A) and tet(C). Chopra and Roberts (2001) pointed out that ribosomal protection protein conferred a wider spectrum of resistance to TCs than those carrying tetracycline efflux proteins, with the exception of tet(B). Correlation analysis with antibiotics in this study showed that tet(A) and tet(C) had a relatively consistent correlation, while tet(M) was opposite. This difference may be caused by different mechanisms of the genes. However, due to the limited number of genes selected in this study, the above phenomenon may be accidental, and further research is needed to verify it.

Apart from the possible internal factors of the resistance mechanism, many external factors were also the key factors affecting the distribution of antibiotics and resistance genes, like environmental conditions (Siedlewicz et al., 2017) and human activities (Pruden et al., 2006; Zhou et al., 2011; Su et al., 2014; Yang et al., 2016). However, the sampling location and

FIGURE 4 | Principle component analysis (PCA) of absolute concentrations of three tet genes, 16S rRNA and intI1 in the TGR. The circle in the graph represented the sampling site. Red circle and blue circle represented the samples in summer and winter, respectively. The PC1 and PC2 explained 64.2 and 23.0% of the total variance, respectively. Seven sampling sites along the longitudinal axis in the TGR were selected from upstream to downstream: riverine zone consisted of sites ZT, MD, and FL; the transition zone consisted of sites ZX and WZ; the lacustrine zone consisted of FJ and ZG. YC belonged to the site after the dam.

sampling strategy selected by different research institutes were inconsistent, which may be the reason for the lack of correlation between antibiotics, ARGs and other factors chosen in this study.

Except for the above elements, other factors also affected the abundance of resistance genes, such as antibiotics (Luo et al., 2010), intI1 (Zhu et al., 2013) and 16S rRNA (Chen and Zhang, 2013). In theory, as the direct selection pressure of ARGs, the abundance of ARGs in the environment was consistent with the corresponding concentration of induced antibiotics (Wang et al., 2014; Niu et al., 2016). However, some articles suggested that there was no correlation between some antibiotics and their corresponding ARGs (Niu et al., 2016). Because in addition to the above factors, light (Pei et al., 2007), temperature (Engemann et al., 2008), heavy metals (Baker-Austin et al., 2006; Gillings and Stokes, 2012) and other factors also affect the abundance of ARGs, especially antibiotic resistance can be co-selected by heavy metals (Zhu et al., 2013). Nevertheless, antibiotic residues are still considered to be the direct and primary selection pressures for acceleration of the expansion of the bacterial resistance (Caucci et al., 2016). In generally, such a significant difference existed in the association of antibiotics with ARGs, one of which may be due to the different habits of antibiotic use in different regions and the different chemical properties of different ARGs. IntI1 was known to play a key role in horizontal gene transfer, leading to widespread antibiotic resistance (Zhu et al., 2013). Meanwhile, intI1 was also regarded as an environmental marker of anthropogenic pollution (Gillings et al., 2015). The high correlation between the tet genes and intI1 revealed the potential horizontal gene transfer in these samples. The characteristics of the bacterial abundance were analyzed based on 16S rRNA. As the potential hosts of ARGs, bacteria would amplify or attenuate under different conditions, so the persistence and proliferation of ARGs were also closely associated with the bacterial abundance (Su et al., 2017). The above results indicated that the higher abundance of bacteria in sediments in summer than that in winter may play an important role in the horizontal spread of tet(A), tet(C) and tet(M) genes in the TGR.

## Spatial-Temporal Variation of TCs and ARGs in the Sediments of the TGR

The TGR is different from other basins, with a low-water level in summer and a high-water level in winter. The seasonal distribution of TGR antibiotics and resistance genes was also different from other basins that was, the concentration of TC and the abundance of ARGs were higher in summer than those in winter. In general, many factors would cause antibiotics to present different distribution characteristics in different seasons, such as temperature and water flow rate (Kim and Carlson, 2007b; Loftin et al., 2008). For example, photo degradation and biodegradation were less effective at low temperature in winter (Cheng et al., 2014). At the same time, TCs were commonly administered to livestock for the prevention and treatment of most respiratory infection and diarrhea (Matsui et al., 2008) and their usage tended to increase in winter (Cheng et al., 2014). Under the influence of these factors, most studies found that the concentration of antibiotics in the water body was higher in the dry season (Yan et al., 2013;

Chen et al., 2018). The seasonal variation in antibiotics would produce a selective pressure, so the abundance of ARG in winter would be higher (Caucci et al., 2016; Sui et al., 2017). But the opposite phenomenon were also found in other literatures (Arikan et al., 2008; Jiang et al., 2011; Knapp et al., 2012; Calero-Caceres et al., 2017), which indicating that the seasonal distribution of antibiotics and ARGs was controversial. It was suggested that the hydrological conditions of different basins may be inconsistent, therefore we speculate that the seasonal distribution in this study may be caused by unique hydrological conditions of the TGR.

It had been reported that the sediment could act as a significant secondary source of antibiotics and could release pollutants into the water if the aquatic environment changed (Cheng et al., 2014), with the antibiotics strongly bonding in the sediments. This tended to protect them from potential degradation processes, which resulted in a greater persistence and a longer-term fate of the antibiotics (Chen and Zhou, 2014). In summer, the TGR is no longer at the high-water level, but operates stably at the low-water level of 145 m. The aquatic environment was changed, sediments in the riparian zone were also changed, and fluctuation of the water level turned the original bottom sediment in winter into the longshore sediment in summer. Previously, the antibiotics in water bodies were adsorbed and accumulated in sediments by natural sedimentation, which resulting in an increase in the concentration of TCs in these sediments. On the other hand, heavy rainfall occurred in the summer, and rainfall runoff carried a large amount of antibiotics into the water, leading to an increase in external input. As the main selection pressure of ARGs, the concentration of TCs in the water during the summer was higher and thus the absolute abundance of ARGs was also higher. In addition, as the rainfall increased, the velocity of the flowing water in the reservoir also increased. ARGs could migrate among different environmental compartments (Chen et al., 2013). Hence, it was reasonable to assume that aqueous-phase mechanical agitation should be enhanced, possibly resulting in the increased exchange of ARGs between the longshore sediments and the water phase (Fang et al., 2017), thereby resulting in a higher concentration of ARGs in the longshore sediments. Meanwhile, we also hypothesized that more frequent exchanges of ARGs may likely occur in the summer than in the winter, bringing about higher concentrations of ARGs in the sediments. This was maybe because the warmer temperatures may promote the survival of bacteria in sediments, and thus lead to a greater number of copies of ARGs (Luo et al., 2010). This assumption was supported by our findings that the abundance of 16S rRNA in sediments was higher in the summer than in the winter and that massive microbial carriers of the ARGs provided the fundamental conditions for the exchange of ARGs (Fang et al., 2017). In December, the reservoir operated at a highwater level and the water flow rate slowed down, which in turn accelerated the sedimentation of antibiotics in the water, leading to a lower concentration of TCs and ARGs in the longshore sediments.

The distribution of TCs and ARGs exhibited clear spatial heterogeneity. Reservoirs are created ecosystems that behave as "hybrids" of river and lake ecosystems (Scott et al., 2009). The riverine, transition and lacustrine zones are three typical zones in the TGR. These longitudinal gradients influence the physical, chemical and biological processes that affect sedimentation (Liu et al., 2012). Along the longitudinal axis, sites MD, FL, and WZ were near a large metropolitan area with a large population and sizeable industries (Li et al., 2017), which resulted in more external antibiotics and ARGs flowing into the TGR. The sediments in the riverine zone were mainly exogenous (Cooke et al., 2011), so the concentrations of antibiotics and resistance genes in the riverine zone were relatively high. However, the riverine zone was close to the inlet of the reservoir, where the fastest water flow rate led to a degree of dilution of the antibiotics and resistance genes. The sediments in the transition zone were exogenous and endogenous (Filstrup and Lind, 2010). Due to the large input of WZ, the content of antibiotics and ARGs in the transition zone were also highest among the eight sites in this study. Moreover, and the decrease in the reservoir flow rate in the transition zone resulted in the residence of the antibiotics and ARGs. Besides, it was worth noting that, in addition to ZT, the concentrations of antibiotics and ARGs at upstream sites (ZT, MD, FL, ZX, and WZ) were higher than those at downstream sites (ZG and FJ), in agreement with previous studies on the Liao River (Dong et al., 2016). Our previous work indicated that anthropogenic activities might lead to the different characteristics of the upstream and downstream sites (Li et al., 2017). Hence, fewer exogenous inputs at FJ and ZG may also result in lower concentrations of antibiotics and ARGs at the downstream sites. In addition, the relatively low antibiotic concentration at the downstream sites could be mainly attributed to dilution of the water from the TGR (Loftin et al., 2008).

#### CONCLUSION

This study demonstrated the occurrence and distribution of antibiotic and ARGs in the longshore sediments of the TGR, China. The differences in hydrological conditions and surrounding environmental conditions also probably led to different distribution characteristics of antibiotic and resistance genes in three typical zones of the TGR (riverine, transition, and lacustrine zones). Statistical analyses revealed that both antibiotics and 16S rRNA played a role in the distribution of ARGs within the TGR. Besides, the concentration of tet genes was significantly correlated with intI1 (p < 0.05) implying horizontal transfer of ARGs, which posed a great threat to public health. It is necessary to conduct further researches on the environmental effects and migration transformation mechanisms of the ARGs.

## AUTHOR CONTRIBUTIONS

LL, ZL, ZPL, and JG conceived and designed the study. JL, YX, and JY performed the experiments. LL, ZL, JL, ZPL, and JG analyzed the data and wrote the paper. All authors contributed to the editing of the manuscript.

#### FUNDING

fmicb-09-01911 August 17, 2018 Time: 12:39 # 10

This study was supported by the National Key Research and Development Plan of China (No. 2017YFC0404705), the National Natural Science Foundation of China (No. 51509233, 51679226, 51779240, and 51861125204), Chongqing Science and Technology Commission (No. cstc2017jcyjAX0243) and Key Research Program of Frontier Sciences, CAS (QYZDJSSWDQC008). All the data in the manuscript are

#### REFERENCES


accessible and can be requested from the corresponding author.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2018.01911/full#supplementary-material


largest municipal wastewater treatment plant in Southwest China: mass balance analysis and consumption back-calculated model. Chemosphere 99, 160–170. doi: 10.1016/j.chemosphere.2013.10.062


**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 © 2018 Lu, Liu, Li, Liu, Guo, Xiao and Yang. 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) and the copyright owner(s) 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.

# Genetic Differentiation, Diversity, and Drug Susceptibility of *Candida krusei*

Jie Gong1†, Meng Xiao2,3†, He Wang2,3, Timothy Kudinha4,5, Yu Wang<sup>6</sup> , Fei Zhao<sup>1</sup> , Weiwei Wu<sup>7</sup> , Lihua He<sup>1</sup> , Ying-Chun Xu2,3 \* and Jianzhong Zhang<sup>1</sup> \*

*<sup>1</sup> State Key Laboratory of Infectious Disease Prevention and Control, Collaborative Innovation Center for Diagnosis and Treatment of Infectious Diseases, National Institute for Communicable Disease Control and Prevention, Chinese Center for Disease Control and Prevention, Beijing, China, <sup>2</sup> Department of Clinical Laboratory, Peking Union Medical College Hospital, Chinese Academy of Medical Sciences, Beijing, China, <sup>3</sup> Beijing Key Laboratory for Mechanisms Research and Precision Diagnosis of Invasive Fungal Diseases, Beijing, China, <sup>4</sup> School of Biomedical Science, Charles Sturt University, Orange, NSW, Australia, <sup>5</sup> Central West Pathology Laboratory, Orange, NSW, Australia, <sup>6</sup> Key Laboratory of Wildlife Biotechnology, Conservation and Utilization of Zhejiang Province, Zhejiang Normal University, Jinhua, China, <sup>7</sup> Department of Dermatology, Hainan Provincial Center for Skin Disease and STI Control, Haikou, China*

#### *Edited by:*

*Silvia Buroni, University of Pavia, Italy*

#### *Reviewed by:*

*Paula Sampaio, University of Minho, Portugal Avi Peretz, The Baruch Padeh Medical Center, Poriya, Israel Sona Kucharikova, University of Trnava, Slovakia*

#### *\*Correspondence:*

*Ying-Chun Xu xycpumch@139.com Jianzhong Zhang Zhangjianzhong@icdc.cn*

*†These authors have contributed equally to this work*

#### *Specialty section:*

*This article was submitted to Evolutionary and Genomic Microbiology, a section of the journal Frontiers in Microbiology*

*Received: 19 July 2018 Accepted: 24 October 2018 Published: 20 November 2018*

#### *Citation:*

*Gong J, Xiao M, Wang H, Kudinha T, Wang Y, Zhao F, Wu W, He L, Xu Y-C and Zhang J (2018) Genetic Differentiation, Diversity, and Drug Susceptibility of Candida krusei. Front. Microbiol. 9:2717. doi: 10.3389/fmicb.2018.02717* *Candida krusei* is a notable pathogenic fungus that causes invasive candidiasis, mainly due to its natural resistance to fluconazole. However, to date, there is limited research on the genetic population features of *C. krusei*. We developed a set of microsatellite markers for this organism, with a cumulative discriminatory power of 1,000. Using these microsatellite loci, 48 independent *C. krusei* strains of clearly known the sources, were analyzed. Furthermore, susceptibility to 9 antifungal agents was determined for each strain, by the Clinical and Laboratory Standards Institute broth microdilution method. Population structure analyses revealed that *C. krusei* could be separated into two clusters. The cluster with the higher genetic diversity had wider MIC ranges for six antifungal agents. Furthermore, the highest MIC values of the six antifungal agents belonged to the cluster with higher genetic diversity. The higher genetic diversity cluster might have a better adaptive capacity when *C. krusei* is under selection pressure from antifungal agents, and thus is more likely to develop drug resistance.

Keywords: *Candida krusei*, invasive candidiasis, genetic differentiation, genetic diversity, microsatellites, drug susceptibility

#### INTRODUCTION

Invasive candidiasis is the most common fungal disease among hospitalized patients, and affects more than 250,000 people worldwide annually, with more than 50,000 deaths reported (Kullberg and Arendrup, 2015). In the Candida genus, Candida krusei attracts much medical attention because it is intrinsically resistant to fluconazole (Akova et al., 1991; Schuster et al., 2013). In addition, C. krusei exhibits resistance to other antifungal drugs such as voriconazole, echinocandins, and amphotericin B (Fukuoka et al., 2003; Hakki et al., 2006; Pfaller et al., 2008). It has been known for some time that mutations in ERG11 and FKS 1 genes are the major mechanisms responsible for azole- and echinocandin-resistance in Candida species, including C. krusei (Jensen et al., 2014; Forastiero et al., 2015; Feng et al., 2016; Perlin et al., 2017). In addition, antifungal resistance can be acquired by over-expression of efflux pump e.g., Abc1p (Lamping et al., 2009; Ricardo et al., 2014). However, there have been some C. krusei antifungal resistant phenotypes, including resistance to azoles other than fluconazole and to enchinocandins e.g., caspofungin, that cannot be explained by currently known mechanisms of resistance (Hakki et al., 2006; Whaley et al., 2017).

From an evolutionary perspective, drug resistance in a microorganism is part of the adaptive evolutionary response of a species to environmental pressures (Salmond and Welch, 2008). Nowadays, the environmental pressure of antifungal drugs comes not only from the use of clinical drugs, but also from the use of agricultural drugs (Sanglard, 2016).

The adaptive capacity is usually related to the level of genetic diversity. From the points of molecular ecology, genetic diversity can allow species or populations to adapt quickly to changing environment conditions and different habitats (Freeland et al., 2011). Similarly, from the perspective of conservation genetics, genetic diversity allows species or populations to tolerate a wider range of environmental changes, including bacteria, fungi and so on. Also, genetic diversity is helpful to maintain the evolutionary vigor (Frankham et al., 2010). In general, a higher genetic diversity enables the organism to respond better to new selection pressures (McDonald and Linde, 2002). When there was a selection pressure for exogenous antifungal agents, the more genetic diversity the fungal populations had, the higher the probability of survival. In other words, antifungal agents were the directional selection factors from Darwin's theory of Evolution. If the genetic diversity of the fungal population was high, there might be some individual death under the pressure of drug selection, but some individuals carrying different genes would survive. Therefore, it is reasonable to hypothesize that microbial populations with higher genetic diversity are more likely to develop antimicrobial drug resistance.

However, to our best knowledge, no research targeting the genetic population features of C. krusei has been carried out to date. This is partly due to the lack of a flexible molecular typing method. Therefore, in this study, we (1) developed a novel set of microsatellite markers for molecular typing and population genetic analysis of C. krusei; (2) used the developed assay to type 48 multicenter collected C. krusei clinical strains; and (3) analyzed the correlation between genetic diversity and drug susceptibility among the studied strains.

#### MATERIALS AND METHODS

#### Ethics Statement

This study was reviewed and approved by the ethics committee of the National Institute for Communicable Disease Control and Prevention, Chinese CDC. Written informed consent was obtained from patients for use of the samples in research.

#### Isolate Collection and Identification

A total of 48 C. krusei isolates, each from a single patient, were collected from 15 hospitals distributed in 10 cities across China during the period 2009–2012, as part of the national surveillance program for invasive fungal infections (the

#### TABLE 1 | Characterization of *Candida krusei* microsatellite loci.


*(Continued)*


#### TABLE 1 | Continued

*<sup>a</sup>Annealing temperature; <sup>b</sup>Polymorphic information content; <sup>c</sup>discriminatory power.*

CHIF-NET study, **Figure 1** and **Supplemental A**). The isolates were stored at −80◦C until use at Peking Union Medical College Hospital, Beijing, China (PUMCH). Before testing, the isolates were inoculated on CHROMagarTM Candida medium (Difco Laboratories, Detroit, MI, USA) and incubated at 37◦C for 24 h. Species identification of the isolates was confirmed by matrix-assisted laser desorption/ionization-time of flight mass spectrometry (MALDI-TOF MS, Vitek MS, bioMérieux, Marcyl'E'toile, France) as per manufacturer's instructions, and by sequence analysis of their rDNA internal transcribed spacer (ITS) regions (Wang et al., 2012). The identities of all the isolates was confirmed by sequencing.

#### DNA Extraction, Microsatellite Development, and Genotyping

All the fungal isolates were grown on potato dextrose agar at 37◦C for 24 h. DNA extraction was performed using a QIAamp DNA Mini Kit (Qiagen, Hilden, Germany).

The software SciRoKo was used to identify microsatellites in the C. krusei genome (GenBank assembly accession: GCA\_001983325.1, Kofler et al., 2007; Cuomo et al., 2017). Primers were designed using Primer premier 5.0 (PREMIER Biosoft International, Palo Alto, CA, USA) in regions flanking microsatellite loci, and annealing temperatures were optimized with a gradient PCR. Polymorphic microsatellite loci were selected for molecular typing and population genetic analysis of C. krusei. There were two criteria used for selection of microsatellite loci: first, the locus had to have a relatively high genetic polymorphism (the number of alleles was >3); second, the locus could be amplified relatively stable. The microsatellite loci would be abandoned if the loci was not amplified in more than two strains.

For the 33 selected microsatellite loci (**Table 1**), PCR was performed on 48 clinical isolates. Amplification was carried out using a Taq polymerase kit (Takara, Dalian, China). Each of the amplification reactions was composed of 1 × PCR buffer, 0.2µM dNTP, 0.5 U Taq polymerase, 0.2µM each primer, and 2 µl genomic DNA (20–50 ng/µl). The thermocycler conditions were as follows: initial denaturation at 95◦C for 5 min, followed by 35 cycles of denaturation at 94◦C for 30 s, annealing at an optimized primer-specific annealing temperature for 30 s (**Table 1**), extension at 72◦C for 30 s and final extension at 72◦C for 10 min. The primers for these selected loci were fluorescently labeled with 6-carboxy-fluorescein (6-FAM). Allele length was determined by migration of PCR products on an ABI 3,700 automated capillary DNA sequencer (Applied Biosystems). Allele sizes were assigned with GeneMapper software (version 3.7) according to an internal size standard (LIZ 500, Applied Biosystems).

#### Antifungal Susceptibility Testing

The in vitro susceptibility to nine antifungal drugs- fluconazole, voriconazole, itraconazole, posaconazole, caspofungin, micafungin, anidulafungin, amphotericin B, and 5-flucytosine, was determinedfor 48 isolates using the Clinical and Laboratory Standards Institute (CLSI) broth microdilution method<sup>22</sup> . Minimum inhibitory concentration (MIC) results for fluconazole, voriconazole, caspofungin, micafungin, and anidulafungin, were interpreted using clinical breakpoints in accordance with the CLSI guidelines (CLSI, 2017), and those for itraconazole, posaconazole, amphotericin B, and 5-flucytosine, were interpreted using epidemiological cut-off values (Xiao et al., 2014). The quality control strains used were C.krusei ATCC 6,258 and Candida parapsilosis ATCC 22,019.

#### Data Analysis

Deviation from Hardy-Weinberg equilibrium was computed using GENEPOP version 4 (Rousset, 2008). Hardy-Weinberg equilibrium was tested using the score test for heterozygote deficiency and the significance was addressed by a Markov Chain algorithm (Markov chain parameters: dememorization number = 2,000, number of batches = 250, number of iterations per batch = 2,000).

The discriminatory power of markers was calculated according to the method of Hunter and Gaston (1988). Number of alleles (nA), effective number of alleles (ne), Shannon's Information Index (I), and Nei's unbiased gene diversity (HS), were calculated using GENALEX 6.5 (Peakall and Smouse, 2012). Allelic Richness (AR) was calculated by FSTAT 2.9.3 (Goudet, 1995). Ne, I, HS, and AR were used to measure genetic variability of populations.

Population composition was inferred for the C. krusei isolates using the program Structure 2.3 (Pritchard et al., 2000), which estimates the log probability of the data for each value of K (number of clusters or populations). A series of independent runs were performed by using K from 1 to 12 populations, a burnin of 100,000 Markov chain Monte Carlo (MCMC) iterations, and a data collection period of 100,000 MCMC iterations. Each simulation of K was replicated 10 times. The method of Evanno et al. (2005) was used to estimate the most likely K given the data with Structure Harvester (Earl, 2012). The level of genetic differentiation at microsatellite loci among clusters was estimated as FST, which is simply a measure of how genetically similar populations are to one another. FST was calculated using Arlequin 3.5 (Excoffier and Lischer, 2010). Principal coordinates analysis (PCoA) of FST value among clusters was calculated using GenAlEx (Peakall and Smouse, 2012).

For antifungal susceptibility results, MIC50, MIC90, and geometric mean (GM) MIC values were calculated using WHONET software (version 5.6, WHO Collaborating Center for Surveillance of Antimicrobial Resistance, Boston, USA).

#### RESULTS

### Microsatellite Loci of *C. krusei*

Based on the genome of C. krusei, a total of 200 microsatellite loci were identified, and primers were designed (data not shown). Of these microsatellite loci, 33 polymorphic microsatellite loci (Cakr 001–Cakr 033) could be stably amplified in all C. krusei isolates (**Table 1**). The cumulative discriminatory power of the 33 loci was 1.000. If only 8 polymorphic sites with the highest polymorphism were selected (Cakr004, Cakr005, Cakr011, Cakr019, Cakr025, Cakr026, Cakr029, Cakr031), it was found that the cumulative discriminatory power would still be 1.000. This might mean that the molecular typing of strains

which corresponded to the 2 clusters divided by STRUCTURE software.

could be achieved effectively by using only these 8 microsatellite loci. All loci showed significant deviation from Hardy-Weinberg equilibrium (P < 0.05).

In addition, it must be noted that many isolates were heterozygote (**Supplemental A**).

#### Genetic Differentiation and Diversity

When performing STRUCTURE analyses, the clustering level K = 2 yielded the largest delta-K value (**Figure 2A**). At K = 2, individual isolates could be assigned to two clusters (**Figure 2B**). Cluster A included 17 strains and cluster B 31 strains. There was no clear relationship between cluster patterns and geographical source of the isolates, and between cluster patterns and disease

clinical manifestations. The FST between the two clusters was 0.188 (P < 0.01). The principal coordinate analysis (PCoA) supported the result of the STRUCTURE analyses (**Figure 3**), suggesting that the population of C. krusei was divided into two lineages. The PCoA also suggested that two lineages would both consist of strains from different geographical origins and clinical manifestations of disease. However, if patients from whom the isolates were obtained is considered, there appears to be some differences between the two clusters. The hosts of Cluster A strains were mainly children younger than 10 years old or aged people older than 50 years old. In contrast, the hosts of Cluster B covered almost all age groups (**Figure 4**). Furthermore, the specimen types of Cluster B are also more complex than Cluster A (**Figure 5**).

The genetic diversity of C. krusei was assessed by Shannon's Information Index, Nei's unbiased gene diversity and allelic richness. These indices are shown in **Table 2**. Four indices (Mean number of effective alleles, Shannon's Information Index, Nei's unbiased gene diversity, allelic richness), all showed that there was a higher genetic diversity in cluster B.

#### *In vitro* Susceptibilities

All isolates were intrinsically resistant to fluconazole (MICs ≥ 16 mg/L; **Figure 6**, **Supplemental A**). Of the other eight antifungal agents tested, all isolates were susceptible or of wild-type phenotype to voriconazole, itraconazole, posaconazole, anidulafungin, micafungin, 5-flucytosine, and amphotericin B. Only two of 48 isolates (4.2%) were interpreted as intermediate to caspofungin, both of which belonged to microsatellite cluster B, while the rest 95.8% (46/48) isolates remained susceptible to caspofungin (**Figure 6**, **Supplemental A**).

The Geometric mean (GM) MIC, MIC50, and MIC<sup>90</sup> of the two clusters were generally similar, while the MIC range differed between the two clusters (**Table 3**, **Figure 6**). For most antifungal agents (6/9) (including caspofungin, posaconazole, voriconazole, itraconazole, fluconazole, amphotericin B), cluster B had a wider MIC range. It is worth noting that the highest MIC values of all 6 antifungal agents were confined to cluster B.

### DISCUSSION

The correlation between genetic diversity and adaptive capacity of the population has long been studied in the field of molecular ecology. For example, it was found that the Arabidopsis thaliana population with higher genetic diversity had better colonization success (Crawford and Whitney, 2010). In principle, the process of fungal infection and clinical manifestation of disease is also considered a colonization success. Unfortunately, very few studies have been done on fungal infections and drug-resistance from the perspective of population genetics. In this study, we carried out a comprehensive analysis of the population genetic features and drug resistance, and attempted to elucidate the drug resistance of C. krusei from an evolutionary perspective.

Although an important pathogenic fungal species, the population genetic parameters of C. krusei have remained largely unknown. In this study, a novel array of microsatellite markers was developed for molecular typing and population genetic analysis of the species. The discriminatory index of the new

TABLE 2 | Genetic diversity of *Candida krusei* subgroups.


*<sup>a</sup>Mean number of alleles; <sup>b</sup>Mean number of effective alleles; <sup>c</sup>Shannon's Information Index; <sup>d</sup>Nei's unbiased gene diversity; <sup>e</sup>Allelic richness.*



*<sup>a</sup>Geometric mean Minimum inhibitory concentration; <sup>b</sup>Mean minimal inhibitory concentrations against 50 percent of strains; <sup>c</sup>Mean minimal inhibitory concentrations against 90 percent of strains.*

method (1.000) was slightly higher than MLST, which exhibited a discriminatory index of 0.998 (Jacobsen et al., 2007). It has been shown by Cuomo et al. (2017) that the genome of C. krusei is highly heterozygous, and this was also confirmed in the present study. For all microsatellite loci, there were some heterozygous individual isolates. As previously described in Candida albicans, a significant departure from Hardy-Weinberg equilibrium expectations was found (Sampaio et al., 2003). This finding supports the previous conclusions that reproduction in C. krusei is mainly clonal.

Based on the STRUCTURE software analysis, the C. krusei population in China appears to be divided into two clusters, a result which was also supported by PCoA. The two clusters showed no obvious differences with respect to geographical distribution of the isolates. These findings are very similar to those of another pathogenic fungus, Trichophyton rubrum, which was also divided into two clusters with similar geographical distributions of the clusters (Gong et al., 2016). For pathogenic fungi, it might be a common phenomenon that different clusters of the same organism co-exist in the same geographical locale. Dispersal of pathogenic fungi is generally affected by host activity. It is highly possible that C. krusei strains of different clusters existed in different geographical areas and were carried to the same geographical area by hosts including humans. When the survival capacity of different clusters is similar, it indicates that the different clusters co-existed in the same area. However, there were significant differences in host age between the two clusters, which might suggest that cluster B strains have a higher pathogenicity. Moreover, the type of specimens in Cluster B were also more complicated and seemed to confirm this. However, this is only a hypothesis, and more work needs to be done to demonstrate these findings in animal infection model experiments.

Cluster B had a higher genetic diversity which suggests a better adaptive capacity for survival in challenging conditions. As previously mentioned, the population with a higher genetic diversity is more likely to develop antimicrobial drug resistance. In our study, cluster B had a wider MIC range for 6 antifungal drugs, although there was no obvious difference in GM MIC, MIC50, and MIC<sup>90</sup> between the two clusters. Specifically, the strains with the highest MIC value were either in the current group B (including 6 drugs: caspofungin, posaconazole, voriconazole, itraconazole, fluconazole, and amphotericin B), or both in group A and group B (including 3 drugs: anidulafungin, micafungin, and 5-flucytosine). Meanwhile, there was no strain with the highest MIC value was observed in Cluster A. These findings suggest that the population with higher genetic diversity may have more diverse phenotypes, including drug resistance. When subjected to selective pressure from antifungal drugs, cluster B might have a better adaptive capacity, and thus would be more likely to develop drug resistance. This suggests that the C. krusei population or lineage with higher genetic diversity needs more attention in terms of fungal drug resistance.

In conclusion, C. krusei was divided into two clusters by novel high-resolution microsatellite markers. The cluster with higher

#### REFERENCES


genetic diversity had wider MIC ranges for six antifungal agents, and the highest MIC values of the six antifungal agents belonged to the cluster of higher genetic diversity. It is plausible that the C. krusei cluster with higher genetic diversity might have better adaptive capacity when under the selection pressure of antifungal agents.

#### AUTHOR CONTRIBUTIONS

JZ and Y-CX designed the experiments. JG, MX, HW, FZ, LH, and WW collected the samples and performed the experiments. JG, MX, TK, and YW analyzed data. JG, MX, and TK wrote the manuscript. All authors read and approved the final manuscript.

#### ACKNOWLEDGMENTS

This work was supported by a Peking Union Medical College Hospital Out-standing Young Talents Program (JQ201703), a Chinese Academy of Medical Sciences Innovation Fund for Medical Sciences (2016-I2M-1-014), a Beijing Innovation Base Cultivation and Development Special Fund (Z171100002217068), and the Major Infectious Diseases Such as AIDS and Viral Hepatitis Prevention and Control Technology Major Projects (2018ZX10712001).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2018.02717/full#supplementary-material

echinocandin resistance during caspofungin therapy. Antimicrob. Agents. Chemother. 59, 6975–6982. doi: 10.1128/AAC.01005-15


**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 © 2018 Gong, Xiao, Wang, Kudinha, Wang, Zhao, Wu, He, Xu and Zhang. 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) and the copyright owner(s) 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 NDM-Encoding Plasmids From *Enterobacteriaceae* Recovered From Czech Hospitals

Veronika Paskova1,2, Matej Medvecky 3,4, Anna Skalova1,2, Katerina Chudejova1,2 , Ibrahim Bitar 1,2, Vladislav Jakubu<sup>5</sup> , Tamara Bergerova1,2, Helena Zemlickova5,6 , Costas C. Papagiannitsis 1,2 \* and Jaroslav Hrabak 1,2

<sup>1</sup> Department of Microbiology, Faculty of Medicine and University Hospital in Plzen, Charles University, Plzen, Czechia, <sup>2</sup> Faculty of Medicine, Biomedical Center, Charles University, Plzen, Czechia, <sup>3</sup> Veterinary Research Institute, Brno, Czechia, <sup>4</sup> Faculty of Science, National Centre for Biomolecular Research, Masaryk University, Brno, Czechia, <sup>5</sup> National Reference Laboratory for Antibiotics, National Institute of Public Health, Prague, Czechia, <sup>6</sup> Department of Clinical Microbiology, University Hospital and Faculty of Medicine in Hradec Kralove, Charles University, Hradec Kralove, Czechia

#### *Edited by:*

Simona Pollini, Università degli Studi di Firenze, Italy

#### *Reviewed by:*

Remy A. Bonnin, Université Paris-Saclay, France Angela Novais, Universidade do Porto, Portugal

#### *\*Correspondence:*

Costas C. Papagiannitsis c.papagiannitsis@gmail.com

#### *Specialty section:*

This article was submitted to Evolutionary and Genomic Microbiology, a section of the journal Frontiers in Microbiology

*Received:* 17 February 2018 *Accepted:* 21 June 2018 *Published:* 10 July 2018

#### *Citation:*

Paskova V, Medvecky M, Skalova A, Chudejova K, Bitar I, Jakubu V, Bergerova T, Zemlickova H, Papagiannitsis CC and Hrabak J (2018) Characterization of NDM-Encoding Plasmids From Enterobacteriaceae Recovered From Czech Hospitals. Front. Microbiol. 9:1549. doi: 10.3389/fmicb.2018.01549 The aim of the present study was to characterize sporadic cases and an outbreak of NDM-like-producing Enterobacteriaceae recovered from hospital settings, in Czechia. During 2016, 18 Entrobacteriaceae isolates including 10 Enterobacter cloacae complex (9 E. xiangfangensis and 1 E. asburiae), 4 Escherichia coli, 1 Kluyvera intermedia, 1 Klebsiella pneumoniae, 1 Klebsiella oxytoca, and 1 Raoultella ornithinolytica that produced NDM-like carbapenemases were isolated from 15 patients. Three of the patients were colonized or infected by two different NDM-like producers. Moreover, an NDM-4-producing isolate of E. cloacae complex, isolated in 2012, was studied for comparative purposes. All isolates of E. cloacae complex, except the E. asburiae, recovered from the same hospital, were assigned to ST182. Additionally, two E. coli belonged to ST167, while the remaining isolates were not clonally related. Thirteen isolates carried blaNDM−4, while six isolates carried blaNDM−<sup>1</sup> (n = 3) or blaNDM−<sup>5</sup> (n = 3). Almost all isolates carried blaNDM-like-carrying plasmids being positive for the IncX3 allele, except ST58 E. coli and ST14 K. pneumoniae isolates producing NDM-1. Analysis of plasmid sequences revealed that all IncX3 blaNDM-like-carrying plasmids exhibited a high similarity to each other and to previously described plasmids, like pNDM-QD28, reported from worldwide. However, NDM-4-encoding plasmids differed from other IncX3 plasmids by the insertion of a Tn3-like transposon. On the other hand, the ST58 E. coli and ST14 K. pneumoniae isolates carried two novel NDM-1-encoding plasmids, pKpn-35963cz, and pEsco-36073cz. Plasmid pKpn-35963cz that was an IncFIB(K) molecule contained an acquired sequence, encoding NDM-1 metallo-β-lactamase (MβL), which exhibited high similarity to the mosaic region of pS-3002cz from an ST11 K. pneumoniae from Czechia. Finally, pEsco-36073cz was a multireplicon A/C2+R NDM-1-encoding plasmid. Similar to other type 1 A/C<sup>2</sup> plasmids, the blaNDM−<sup>1</sup> gene was located within the ARI-A resistance island. These findings underlined that IncX3 plasmids have played a major role in the dissemination of blaNDM-like genes in Czech hospitals. In combination with further evolvement of NDM-like-encoding MDR plasmids through reshuffling, NDM-like producers pose an important public threat.

Keywords: NDM, metallo-β-lactamases, *Enterobacter xiangfangensis*, ST182, IncX3

## INTRODUCTION

Acquired carbapenem-hydrolyzing β-lactamases are resistance determinants of increasing clinical importance in Gram-negative pathogens. Of these, NDM-1 metallo-β-lactamase (MβL) was first described in Klebsiella pneumoniae and Escherichia coli isolated in Sweden in 2008 from an Indian patient transferred from a New Delhi hospital (Yong et al., 2009). Since then, NDM-1-producing bacteria, including clinical isolates of Enterobacteriaceae and Acinetobacter baumannii, have been reported from the Indian subcontinent but also worldwide (Nordmann et al., 2011).

In Czechia, the occurrence of NDM-producing bacteria was rare, with only three sporadic cases being detected during 2011– 2013. These cases included an NDM-1-producing A. baumanni isolated from a patient repatriated from Egypt (Hrabák et al., 2012), an NDM-4-producing strain of Enterobacter cloacae complex from a patient previously hospitalized in Sri Lanka (Papagiannitsis et al., 2013b) and a ST11 K. pneumoniae isolate carrying two NDM-1-encoding plasmids, from Slovakia (Studentova et al., 2015). However, an increase in the isolation frequency of NDM-like-producing Enterobacteriaceae from Czech hospitals was observed, during 2016.

Thus, the aim of the present study was to characterize the NDM-like producers detected in Czech hospitals, during 2016. Also, we describe the complete nucleotide sequences of representative blaNDM-like-carrying plasmids harbored by the studied isolates.

## MATERIALS AND METHODS

#### Bacterial Isolates and Confirmation of Carbapenemase Production

In 2016, Czech hospitals referred a total of 410 Enterobacteriaceae isolates with a meropenem MIC of >0.125µg/ml (EUCAST, 2012) to the National Reference Laboratory for Antibiotics. Species identification was confirmed by matrix-assisted laser desorption ionization-time of flight mass spectrometry (MALDI-TOF MS) using MALDI Biotyper software (Bruker Daltonics, Bremen, Germany). All isolates were tested for carbapenemase production by the MALDI-TOF MS meropenem hydrolysis assay (Rotova et al., 2017). Isolates that were positive by the MALDI-TOF MS meropenem hydrolysis assay were subjected to metallo-β-lactamase, KPC, and OXA-48 detection using the double-disc synergy test with EDTA, the phenylboronic acid disc test, and the temocillin disc test (Lee et al., 2003; Doi et al., 2008; Glupczynski et al., 2012), respectively. Additionally, carbapenemase genes (blaKPC, blaVIM, blaIMP, blaNDM, and blaOXA−48-like) were detected by PCR amplification (Poirel et al., 2004; Ellington et al., 2007; Naas et al., 2008; Yong et al., 2009). PCR products were sequenced as described below. Isolates positive for blaNDM-like genes were further studied. Moreover, the NDM-4-producing isolate of E. cloacae complex, recovered at the University Hospital Pilsen (Pilsen, Czechia) during 2012 (Papagiannitsis et al., 2013b), was included in this study for comparative purposes.

## Susceptibility Testing

The MICs of piperacillin, piperacillin-tazobactam, cefotaxime, ceftazidime, cefepime, aztreonam, meropenem, ertapenem, gentamicin, amikacin, chloramphenicol, tetracycline, trimethoprim-sulfamethoxazole, ciprofloxacin, colistin, and tigecycline were determined by the broth dilution method (EUCAST, 2003). Data were interpreted according to the criteria of the European Committee on Antimicrobial Susceptibility Testing (EUCAST; www.eucast.org).

## Typing

All isolates were typed by multilocus sequence typing (MLST) (Diancourt et al., 2005; Wirth et al., 2006; Miyoshi-Akiyama et al., 2013; Herzog et al., 2014). The databases at https:// pubmlst.org/ecloacae/, http://mlst.warwick.ac.uk/mlst/dbs/Ecoli, http://bigsdb.web.pasteur.fr/klebsiella and https://pubmlst.org/ koxytoca/ were used to assign STs.

## Detection of β-Lactamases

The β-lactamase content of all blaNDM-like-positive isolates was determined by isoelectric focusing (IEF). Bacterial extracts were obtained by sonication of bacterial cells suspended in 1% glycine buffer and clarified by centrifugation. Sonicated cell extracts were analyzed by IEF in polyacrylamide gels containing ampholytes (pH 3.5–9.5; AP Biotech, Piscataway, NJ). The separated β-lactamases were visualized by covering the gel with the chromogenic cephalosporin nitrocefin (0.2 mg/ml; Oxoid Ltd., Basingstoke, United Kingdom; Papagiannitsis et al., 2015).

On the basis of the IEF data, PCR detection of various bla genes was performed by the use of primers specific for blaTEM−1, blaOXA−1, blaSHV, blaCTX−M, and blaCMY, as reported previously (Pałucha et al., 1999; Pérez-Pérez and Hanson, 2002; Woodford et al., 2006; Coque et al., 2008). Both strands of the PCR products were sequenced using an ABI 377 sequencer (Applied Biosystems, Foster City, CA).

## Transfer of *bla*NDM-Like Genes

Conjugal transfer of blaNDM-like genes from the clinical strains was carried out in mixed broth cultures (Vatopoulos et al., 1990), using the rifampin-resistant E. coli A15 laboratory strain as a recipient. Transconjugants were selected on MacConkey agar plates supplemented with rifampin (150 mg/l) and ampicillin (50 mg/l). Plasmid DNA from clinical isolates, which failed to transfer blaNDM-like by conjugation, was extracted using a Qiagen Maxi kit (Qiagen, Hilden, Germany) and used to transform E. coli DH5α cells. The preparation and transformation of competent E. coli cells were done using calcium chloride (Cohen et al., 1972). Transformants were selected on Luria-Bertani agar plates with ampicillin (50 mg/l). Transconjugants or transformants were confirmed to be NDMlike producers by PCR (Yong et al., 2009) and the MALDI-TOF MS meropenem hydrolysis assay (Rotova et al., 2017).

## Plasmid Analysis

To define the genetic units of the blaNDM-like genes, the plasmid contents of all NDM-producing clinical and recombinant strains were analyzed by pulsed-field gel electrophoresis (PFGE) of total DNA digested with S1 nuclease (Promega, Madison, WI, USA; Barton et al., 1995). Following PFGE, the DNA was transferred to a BrightStar-Plus positively charged nylon membrane (Applied Biosystems, Foster City, CA) and hybridized with digoxigeninlabeled blaNDM-like probe.

Plasmid incompatibility (Inc) groups were determined by the PCR-based replicon typing (PBRT) method (Carattoli et al., 2005; Johnson et al., 2012), using total DNA from transconjugants or transformants. Additionally, the IncR replicon was detected as described previously (García-Fernández et al., 2009).

#### Detection of Characteristic Regions

Based on the results from Illumina sequencing (see below), six PCRs targeting characteristic regions of NDM-4-encoding IncX3 plasmids and genomes of ST182 isolates of E. cloacae complex sequenced during this study were designed. The selected regions included: (i) a Tn3-like transposon found in NDM-4-encoding IncX3 plasmids, and (ii) four insertions identified in the genome of Encl-922 (see section Comparative Analysis of Enterobacter Isolates). All NDM-producing clinical or recombinant strains were screened for the presence of the regions described above by the use of specific primers (see Table S1).

#### Plasmid and Chromosome Sequencing

Ten plasmids were selected for complete sequencing. These plasmids were selected as representatives of different origins, plasmid sizes and hospitals. Additionally, clinical isolates Encl-922 and Encl-44578 were also selected for whole genome sequencing. The selected isolates were isolated 4-year apart (2012 and 2016).

Plasmid DNAs from transconjugants or transformants were extracted using a Qiagen Large-Construct kit (Qiagen, Hilden, Germany). Additionally, the genomic DNAs of clinical Encl-922 and Encl-44578 isolates were extracted using a DNA-Sorb-B kit (Sacace Biotechnologies S.r.l., Como, Italy). Multiplexed DNA libraries were prepared, using the Nextera XT Library Preparation kit, and 300-bp paired-end sequencing was performed on the Illumina MiSeq platform (Illumina Inc., San Diego, CA, USA) using the MiSeq v3 600-cycle Reagent kit. Initial paired-end reads were quality trimmed using the Trimmomatic tool v0.33 (Bolger et al., 2014) with the sliding window size of 4 bp, required average base quality ≥17 and minimum read length of 48 bases. Genomic DNA reads of clinical isolates of E. cloacae complex were consequently assembled using the de Bruijn graph-based de novo assembler SPAdes v3.9.1 (Bankevich et al., 2012), using k-mer sizes 21, 33, 55, 77, 99, and 127. For assembly of the plasmids, reads were mapped to the reference E. coli K-12 substrain MG 1655 genome (GenBank accession no. U00096) using the BWA-MEM algorithm (Li, 2013), in order to filter out the chromosomal DNA. Then, all the unmapped reads were assembled in the same way as described above. The sequence gaps were filled by a PCR-based strategy and Sanger sequencing. For sequence analysis and annotation, the BLAST algorithm (www.ncbi.nlm.nih.gov/BLAST), the ISfinder database (www-is.biotoul.fr/), and the open reading frame (ORF) finder tool (www.bioinformatics.org/sms/) were utilized. Comparative genome alignments were performed using the Mauve v2.3.1 program (Darling et al., 2010).

Antibiotic resistance genes were identified using the ResFinder 2.1 tool (https://cge.cbs.dtu.dk/services/ResFinder/) with an identity threshold of >90% (Zankari et al., 2012).

#### Comparative Analysis of Clinical Isolates of *E. cloacae* Complex

Comparative genomic analysis of clinical isolates of E. cloacae complex was based on statistics calculated by QUAST v4.5 (Gurevich et al., 2013) and VarScan v2.3.9 (Koboldt et al., 2012) tools. All quality trimmed Illumina reads of Encl-922 were mapped to contigs of Encl-44578, employing BWA-MEM algorithm v0.7.12 (Li, 2013) and SAMtools v1.3 (Li et al., 2009), for the format conversions and analysis of the results. Then, single nucleotide polymorphisms (SNPs) and indels were detected employing VarScan with parameters set as follows: minimum read depth at a position = 6, minimum base quality at a position = 20 and minimum variant allele frequency threshold of 0.45. Moreover, SNPs and indels located in a region within 127 bp from any edge of a contig, as well as SNPs and indels harbored by contigs smaller than 2 kb were excluded from further analysis. Remaining SNPs and indels were also manually checked and refined by visualization of mapped data via Tablet v1.14.04.10 (Milne et al., 2013). Differences in assembly of E. cloacae complex genomes were inspected using QUAST's Icarus viewer (Mikheenko et al., 2016). In order to examine whether SNPs and indels were located in intergenic or coding regions, as well as to find out what are the differences in genetic information between studied isolates, contigs of clinical strains were annotated using Prokka v1.10 (Seemann, 2014). Genes harboring SNPs were compared against NCBI's conserved domain database (Marchler-Bauer et al., 2017) via CD-Search (Marchler-Bauer and Bryant, 2004) to identify conserved domain hits. Finally, sequencing data of clinical strains were examined for the presence of prophage sequences using PHAST web server (Zhou et al., 2011).

#### Nucleotide Sequence Accession Numbers

The nucleotide sequences of the pEsco-5256cz, pEncl-922cz, pRor-30818cz, pKpn-35963cz, pEsco-36073cz, pEncl-44578cz, pEnas-80654cz, pEnin-51781cz, pEsco-4382cz, and pKlox-45574cz plasmids have been deposited in GenBank under accession numbers MG252891, MG252892, MG252893, MG252894, MG252895, MG833402, MG833403, MG833404, MG833405, and MG833406, respectively. Whole genome assemblies of isolates of E. cloacae complex were deposited in NCBI under accession number PRJNA432167.

#### RESULTS

#### Carbapenemase-Producing *Enterobacteriaceae*

A total of 40 Enterobacteriaceae isolates showing carbapenemase activity on MALDI-TOF MS meropenem hydrolysis assay were recovered from Czech hospitals during 2016. PCR screening showed that 18 of the isolates were positive for blaNDM, 14 isolates were positive for blaOXA−48, while the remaining 8 isolates were positive for blaKPC.

### NDM-Like-Producing Isolates

Altogether, 18 nonrepetitive isolates producing NDM-like carbapenemases were isolated from 15 patients in 2016. Among them, 10 were presumptively identified as belonging to E. cloacae complex, 4 were identified to be E. coli, while the remaining isolates belonged to unique species (Enterobacter intermedius, K. pneumoniae, Klebsiella oxytoca, and Raoultella ornithinolytica). A previous study showed that 16S rRNA gene sequence of E. intermedius was included within the cluster of the genus Kluyvera, and therefore, the transfer of E. intermedius to the genus Kluyvera as Kluyvera intermedia was proposed (Pavan et al., 2005). Three of the patients were colonized or infected by two different NDM-like producers (**Table 1**).

NDM-like producers were collected from five Czech hospitals located in three different Czech cities. In hospital B1, an outbreak that included ten patients diagnosed with NDM-like-producing Enterobacteriaceae lasted the studied period. Additionally, two patients colonized or infected with NDM-like producers were reported in hospital B2. The three remaining cases were identified in three different hospitals. None of the patients, treated in hospital B1, had recently traveled abroad or had been previously hospitalized. The patient treated in hospital C was directly repatriated from a hospital in China, while clinical data weren't available for the remaining patients.

Additionally, the NDM-4-producing isolate of E. cloacae complex identified in 2012 (Papagiannitsis et al., 2013b), was studied.

All 19 NDM-like producers exhibited resistance to piperacillin, piperacillin-tazobactam, cephalosporins, and ertapenem (Table S2), while the observed variations in the MICs of aztreonam might reflect the presence of additional resistance mechanisms in some of the isolates. Seventeen of the NDM-like producers also exhibited resistance to ciprofloxacin; 15 were resistant to gentamicin, 13 were resistant to trimethoprimsulfamethoxazole, 1 was resistant to amikacin, and 1 was resistant to colistin, whereas all isolates were susceptible to tigecycline.

The population structure of NDM-like-producing isolates studied by MLST is shown in **Table 1**. All isolates of E. cloacae complex, except the E. asburiae strain, which were recovered from hospital B1, belonged to ST182. Of note was that the NDM-4-producing isolate of E. cloacae complex that was isolated, in 2012, from the patient previously hospitalized in Sri Lanka (Papagiannitsis et al., 2013b) was also assigned to ST182. Two of E. coli, both of which were from hospital B2, belonged to ST167. E. coli ST167 was recently found among NDM-5-producing isolates from different healthcare institutions in China (Yang et al., 2014; Zhang et al., 2016). The two remaining E. coli isolates were not clonally related and belonged to different STs (ST58 and ST69). The K. pneumoniae isolate was assigned to the high risk clone ST14 (Woodford et al., 2011), while the K. oxytoca isolate was classified into ST2 that belongs to a growing international clonal complex (CC2) (Izdebski et al., 2015).

Sequencing of the PCR products revealed three blaNDM-type genes encoding the NDM-1, NDM-4, and NDM-5 enzymes (**Table 1**; Yong et al., 2009; Hornsey et al., 2011; Nordmann et al., 2012). NDM-5 is an NDM-1-related MβL variant that differs from NDM-1 by two amino-acid substitutions, Val88Leu and Met154Leu, the former one being its only change with NDM-4. Thirteen of the isolates, all of which were from hospital B1, were found to produce the NDM-4 MβL (**Table 1**). The three isolates from hospital B2 produced the NDM-5 enzyme, while the three remaining isolates that were recovered from sporadic cases in three different hospitals expressed NDM-1 carbapenemase. Additionally, most of blaNDM-like-positive isolates were confirmed to coproduce the extended-spectrum βlactamase CTX-M-15 (n = 18) either alone or along with TEM-1 (n = 13) and/or OXA-1 (n = 13), whereas the K. pneumoniae and R. ornithinolytica isolates also expressed the SHV-12 enzyme. The ST58 NDM-1-producing E. coli isolate coproduced CMY-16, CTX-M-15, OXA-10, and TEM-1 β-lactamases.

## *bla*NDM-Like-Carrying Plasmids

The blaNDM-like genes from all clinical strains were transferred by conjugation (n = 14) or transformation (n = 5) (**Table 1**). All blaNDM-like-positive recombinants exhibited resistance to piperacillin, piperacillin-tazobactam, cephalosporins, and ertapenem, while they remained susceptible to meropenem (Table S2). The three NDM-1-producing recombinants also exhibited resistance to aztreonam. Additionally, most of blaNDMlike-positive recombinants (n = 18) were susceptible to non-βlactam antibiotics.

Plasmid analysis of NDM-4-producing donor and transconjugant strains revealed the transfer of plasmids, all of which were ∼55 kb in size (**Table 1**). The three NDM-5 producing transformants harbored plasmids of ∼45 kb, whereas the three remaining recombinants carried blaNDM−1-positive plasmids of different sizes (∼55, ∼150, and ∼300 kb). Replicon typing showed seventeen of the plasmids, including those sizing ∼45 and ∼55 kb, were positive for the IncX3 allele. The blaNDM−1-positive plasmid of ∼300 kb was positive for replicons R and A/C, whereas the one remaining blaNDM−1-carrying plasmid was non-typeable by the PBRT method (Carattoli et al., 2005; Johnson et al., 2012).

#### Structure of *bla*NDM-Like-Carrying Plasmids

The complete sequence of blaNDM-like-carrying plasmids representative of different plasmid sizes, replicons, and resistance genes (n = 10) was determined (**Table 1**). Sequence analysis revealed that all IncX3 blaNDM-like-carrying plasmids exhibited a high similarity to each other and to previously described NDMlike-encoding plasmids, belonging to IncX3 group, reported from worldwide (Krishnaraju et al., 2015; Zhu et al., 2016; Pál et al., 2017). The blaNDM−5-positive plasmids, pEsco-5256cz and pKlox-45574cz, were almost identical to NDM-5 encoding plasmid pNDM-QD28 (100% coverage, 99% identity) (GenBank accession no. KU167608) that was characterized from a ST167 E. coli in China (Zhu et al., 2016). Differences among these plasmids consisted in few SNPs (n = 5), almost all

#### TABLE 1 | Characteristics of NDM-like-producing Enterobacteriaceae.


NA, not applicable.

<sup>a</sup>White circles, black squares, and black triangles each indicate the NDM-like-producing isolates recovered from the same patient.

<sup>b</sup>Data for plasmids found in transconjugants are shown in bold; data for plasmids observed in transformants are underlined.

located in mobile elements. Similar to pNDM-QD28, no other resistance genes were detected in these plasmids. Compared to other IncX3 NDM-encoding plasmids, all blaNDM−4-encoding plasmids differed by the insertion of a Tn3-like transposon (nt 7108-14624 in pEncl-44578cz) downstream topB gene (**Figure 1**). The Tn3-like sequence was composed by the 38-bp inverted

repeats (IR) of the transposon, tnpA, tnpR, and two ORFs encoding hypothetical proteins. Target site duplications of 5 bp (GTACC) at the boundaries of the Tn3-like element indicated insertion by transposition. Of note was that the sequence of pEncl-922cz, isolated in 2012 (Papagiannitsis et al., 2013b), was identical to the respective sequences of NDM-4-encoding plasmids recovered in the same hospital, during 2016. PCR screening confirmed the presence of the Tn3-like transposon in all NDM-4-encoding IncX3 plasmids, isolated in hospital B1, while Tn3-like wasn't detected in the remaining blaNDM-like-positive plasmids that belonged to IncX3 group. Furthermore, the blaNDM−1-positive plasmid, pRor-30818cz, harbored an additional 7875-bp sequence (nt 40617-48491 in pRor-30818cz) encoding the extended-spectrum β-lactamase SHV-12 (**Figure 1**). A similar SHV-12-encoding region was found in the IncX3 blaNDM−1-positive plasmid pKP04NDM (100% coverage, 99% identity) (GenBank accession no. KU314941) described from a K. pneumoniae isolate in China.

The NDM-1-encoding plasmid pKpn-35963cz that was nontypeable by the PBRT method (Carattoli et al., 2005) was 161,324 bp in size. Plasmid pKpn-35963cz was composed of two distinct parts: a contiguous plasmid backbone of 115,998 bp (nt 1–58,655 and 103,982–161,324) and an acquired sequence of 45,326 bp (nt 58,656–103,981). The plasmid backbone, which shared similarities with the respective regions of plasmid p1605752FIB (GenBank accession no. CP022125) recovered from a pan-resistant isolate of K. pneumoniae from the United States, harbored regions responsible for replication [repB gene; IncFIB(K) replicon], conjugative transfer (tra and trb genes) and plasmid maintenance (vagCD, psiAB, umuCD and parAB operons, and ssb gene; **Figure 2**). The acquired sequence of pKpn-35963cz contained a 17,836-bp segment (nt 77,360–95,195) encoding NDM-1, which was similar to the mosaic region of pS-3002cz (99% identity) (Studentova et al., 2015). The acquired sequence of pKpn-35963cz contained two additional segments that have also been described in pS-3002cz. The first segment (nt 65,518–72,935) included genes encoding an EcoRII methylase and EcoRII endonuclease, and the class 1 integron In191 carrying the dfrA14 resistance gene. The second segment (nt 101,342– 103,981) contained fragments of transposons Tn1000 (1Tn1000) and Tn1331 (1Tn1331). 1Tn1331 comprised tnpR and aacA4 resistance gene. Furthermore the acquired sequence of pKpn-35963cz carried a macrolide resistance operon [mph(A)], and regions encoding OXA-1 and CTX-M-15 β-lactamases (**Figure 2**). In the acquired sequence of pKpn-35963cz, intact and truncated copies of several mobile elements that may have been implicated in the formation of this region were found.

The plasmid pEsco-36073cz, which encoded the NDM-1 carbapenemase, is 300,958 bp in size. The plasmid showed a complex structure, being composed of sequences of diverse origin (**Figure 3**). A 170,314-bp sequence (nt 232,204–300,958 and 1–101,559) resembled the type 1 A/C<sup>2</sup> plasmid pRH-1238 (94% coverage, 99% identity; **Figure 3**), characterized from a Salmonella enterica serovar Corvallis strain isolated from a migratory wild bird in Germany (Villa et al., 2015). Analysis of A/C2-associated sequence by the core gene PMLST (cgPMLST) scheme (Hancock et al., 2017) indicated that it belonged to cgST3.4. The A/C<sup>2</sup> backbone was composed of regions responsible for replication (repA gene), conjugative

red. IS elements and transposases are shown in yellow and green, respectively. intI1 genes are shaded blue. The remaining genes are shown in white. Thin lines below the map correspond to highly similar sequences from other plasmids.

transfer (Tra1 and Tra2 regions), and plasmid maintenance (higBA and parAB operons and xerD- and kfrA-like genes). Apart from the backbone, pEsco-36073cz carried the blaCMY−2 like-containing region, and the ARI-B and ARI-A resistance islands, as previously described in other type 1 A/C<sup>2</sup> MDR plasmids (Harmer and Hall, 2015; Papagiannitsis et al., 2016). The blaNDM−<sup>1</sup> gene was located within ARI-A, in a genetic environment similar to those previously identified in pRH-1238 (Villa et al., 2015). However, unlike in pRH-1238, the ARI-A of pEsco-36073cz lacked the macrolide resistance determinant mphA-mel-repAciN. Furthermore, a class 1 integron with aacA4 and aphA1 gene cassettes was located between resI and resII sites of the Tn1696 module. The ARI-A of pEsco-36073cz also carried a new integron, In1459, whose variable region comprised the dfrA14, arr-2, cmlA1, blaOXA−10, aadA1 cassettes. Additionally, pEsco-36073cz included fragments resembling the backbone of the recently described IncR plasmid pKP1780 (Papagiannitsis et al., 2013a), and sequences previously found in the plasmid pPSP-a3e (Conlan et al., 2014) and in the chromosomes of several Gram-negative rods. Genes encoding for resistance to arsenate, cooper and mercury were identified in the three remaining acquired regions of pEsco-36073cz.

## Comparative Analysis of *Enterobacter* Isolates

"In silico" hsp60 typing of the genome sequences (Hoffmann and Roggenkamp, 2003) showed that both isolates belonged to the recently recognized E. xiangfangensis species (Gu et al., 2014).

Since all isolates of E. cloacae complex, except the E. asburiae isolate, belonged to the same ST and carried the same IncX3 blaNDM−4-carrying plasmid, the WGS data of clinical strains Encl-922 and Encl-44578 were compared, using QUAST and VarScan tools, in order to examine the phylogenetic relationship of the isolates recovered in 2012 and 2016.

Comparative analysis of clinical isolates revealed that the genome of Encl-922 exhibited extensive similarity (99.87% identity) to the genome of Encl-44578. Sixteen SNPs were identified in the genome of Encl-922, compared to that of Encl-44578, five of which were located within prophage regions (**Table 2**). Interestingly, Encl-922 harbored three large insertions of 8,933 bp (nt 439,392–448,324 in node 2), of 17,903 bp (nt 17,786–35,688 in node 32) and of 13,165 bp (nt 1–13,165 in node 27; prophage sequence PHAGE\_Salmon\_SPN3UB \_NC\_019545). Additionally, Encl-922 harbored an insertion of 33-bp sequence (AACCCTCTCCCCAAAGGGGAGAGG GGACGATTA) located in an intergenic region. Moreover,

rectangles (arrows within rectangles indicate the direction of transcription). blaNDM-like genes are shaded purple, while other resistance genes are shown in red. IS elements and transposases are shown in yellow and green, respectively. intI1 genes are shaded blue; teal blue arrow indicates the group II intron. The remaining genes are shown in white. Thin lines below the map correspond to highly similar sequences from other plasmids.

Encl-922 showed a single nucleotide (G) deletion leading to CDS annotation change of general stress protein 39 to putative oxidoreductase YghA. Analysis of whole genome sequencing (WGS) data by PHAST web server found five intact prophage sequences (PHAGE\_Haemop\_HP2\_NC\_003315, PHAGE \_Salmon\_SPN3UB\_NC\_019545, PHAGE\_Entero\_mEp390 \_NC\_019721, PHAGE\_Pseudo\_PPpW\_3\_NC\_023006, and PHAGE\_Salmon\_SP\_004\_NC\_021774) and one questionable prophage region (PHAGE\_Entero\_SfI\_NC\_027339), in both E. xiangfangensis isolates. However, Encl-922 included one additional incomplete prophage region (PHAGE\_Salmon \_SPN3UB\_NC\_019545), which was absent from the Encl-44578 genome.

Screening by PCR and sequencing identified that all E. xiangfangensis isolates, recovered during 2016, didn't harbor any of the four mentioned insertions. Thus, this finding indicated that Enterobacter isolates from 2016 differed from Encl-922.

## DISCUSSION

The present study investigated sporadic cases and an outbreak of NDM-like-producing Enterobacteriaceae recovered from Czech hospitals, during 2016. Specifically, 12 NDM-4-producing isolates, which belonged to E. xiangfangensis (n = 9), E. asburiae (n = 1), K. intermedius (n = 1), and E. coli (n = 1) species, 3 NDM-5 producers of E. coli (n = 2) and K. oxytoca (n = 1) species, and one E. coli, one K. pneumoniae and one R. ornithinolytica producing NDM-1 MβL were characterized.

The setting that was most affected was hospital B1, in which an outbreak of NDM-4-producing ST182 E. xiangfangensis isolates took place. ST182 isolates of E. cloacae complex were previously identified in Mexico and were associated with the production of NDM-1 enzyme (Torres-González et al., 2015; Bocanegra-Ibarias et al., 2017). Of note was that the isolate of E. cloacae complex, isolated in 2012 from a TABLE 2 | Summary table of 16 SNPs found between the genomes of E. xiangfangensis isolates Encl-44578 (reference) and Encl-922 (query).


<sup>a</sup>The first two SNPs are located in intergenic regions.

patient who had been previously hospitalized in Sri Lanka (Papagiannitsis et al., 2013b), also belonged to ST182 and harbored an IncX3 blaNDM−4-positive plasmid being identical to respective plasmids characterized from isolates recovered from patients treated in hospital B1 (**Table 1**), during 2016. However, comparative genome analysis revealed the presence of four insertions in the genome of E. xiangfangensis Encl-922 isolate. These insertions were not found in the genomic DNA of E. xiangfangensis isolates from 2016, suggesting a second insertion event of NDM-4-producing E. xiangfangensis isolates in Czech hospitals.

In three of the patients, two different NDM-like producers were identified during their hospitalization, supposing the in vivo horizontal transfer of blaNDM-like-carrying plasmids. Sequencing and PCR screening data revealed the presence of the same blaNDM−4- or blaNDM−5-carrying plasmid in these isolates (**Table 1**). These results confirmed the hypothesis of the in vivo horizontal transfer of blaNDM-like-carrying plasmids.

Results from Illumina sequencing showed that IncX3 plasmids have played a major role in the dissemination of blaNDMlike genes in Czech hospitals, which is in agreement with the findings from previous studies from worldwide (Krishnaraju et al., 2015; Zhu et al., 2016; Pál et al., 2017). In the current study, three blaNDM-type genes, encoding the NDM-1, NDM-4, and NDM-5 enzymes, were associated with IncX3 plasmids exhibiting high similarity to each other. Considering also the fact that NDM-1, NDM-4, and NDM-5 differ by one or two aminoacid substitutions may indicate the possibility that blaNDMlike genes encoding NDM-1-related variants have evolved in the same plasmid type. Additionally, Illumina data showed the presence of a unique sequence, a Tn3-like transposon, in sequenced blaNDM−4-carrying plasmids. PCR confirmed the presence of the Tn3-like sequence in all transconjugants, carrying blaNDM−4-positive plasmids. Thus, the PCR targeting of the Tn3-like sequence was able to distinguish blaNDM−4-positive plasmids from other IncX3 plasmids carrying blaNDM−<sup>1</sup> or blaNDM−5. On the other hand, two of the sporadic isolates carried novel NDM-1-encoding plasmids. Plasmid pKpn-35963cz that was an IncFIB(K) molecule contained an acquired sequence, encoding NDM-1 MβL, which exhibited high similarity to the mosaic region of pS-3002cz. pS-3002cs was characterized from an ST11 K. pneumoniae isolate, producing NDM-1 carbapenemase, identified in Czechia (Studentova et al., 2015). Whereas plasmid pEsco-36073cz was a multireplicon A/C2+R NDM-1-encoding plasmid, being a fusion derivative of sequences of diverse origin. Similar to other type 1 A/C<sup>2</sup> plasmids (Harmer and Hall, 2015; Villa et al., 2015), the blaNDM−<sup>1</sup> gene was located within the ARI-A resistance island.

In conclusion, the data presented here contribute to the current knowledge of NDM-like-producing Enterobacteriaceae. In agreement with previous studies, our findings punctuate that NDM-like producers constitute an important public threat, mainly due to the rapid horizontal transfer of IncX3 blaNDMcarrying plasmids but, also, due to further evolvement of NDMlike-encoding MDR plasmids via reshuffling.

#### AUTHOR CONTRIBUTIONS

CP and JH played an important role in interpreting the results and in writing the manuscript. VJ, TB, and HZ helped to acquired data. VP, MM, AS, KC, and IB carried out experimental work. CP supervised the experiments and revised the final manuscript, which was approved by all authors.

### REFERENCES


#### FUNDING

This work was supported by the Medical Research Foundation of the Czech Republic (grant numbers 15-28663A and 17- 29239A); by the National Sustainability Program I (NPU I; grant number LO1503) provided by the Ministry of Education Youth and Sports of the Czech Republic; and the Charles University Research Fund- PROGRES (grant number Q39).

#### ACKNOWLEDGMENTS

We are very thankful to Dana Kralova for technical assistance during the present study. The authors also thank Thomas Jové for curating integron sequencing data.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2018.01549/full#supplementary-material


patients with antibiotic-associated hemorrhagic colitis. J. Clin. Microbiol. 52, 1607–1616. doi: 10.1128/JCM.03373-13


coli isolates in the Arabian Peninsula. Microb. Drug Resist. 23, 871–878. doi: 10.1089/mdr.2016.0216


dissemination of antibiotic resistance. FEMS Microbiol. Rev. 35, 736–755. doi: 10.1111/j.1574-6976.2011.00268.x


Zhou, Y., Liang, Y., Lynch, K., Dennis, J. J., and Wishart, D. S. (2011). PHAST: a fast phage search tool. Nucleic Acids Res. 39, W347–W352. doi: 10.1093/nar/gkr485

Zhu, Y. Q., Zhao, J. Y., Xu, C., Zhao, H., Jia, N., and Li, Y. N. (2016). Identification of an NDM-5-producing Escherichia coli sequence type 167 in a neonatal patient in China. Sci. Rep. 6:29934. doi: 10.1038/srep29934

**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 © 2018 Paskova, Medvecky, Skalova, Chudejova, Bitar, Jakubu, Bergerova, Zemlickova, Papagiannitsis and Hrabak. 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) and the copyright owner(s) 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.

# Identification of a Novel Plasmid Lineage Associated With the Dissemination of Metallo-β-Lactamase Genes Among Pseudomonads

Vincenzo Di Pilato<sup>1</sup> , Alberto Antonelli<sup>1</sup> , Tommaso Giani1,2, Lucia Henrici De Angelis<sup>3</sup> , Gian Maria Rossolini1,2 and Simona Pollini1,2 \*

#### Edited by:

Raffaele Zarrilli, University of Naples Federico II, Italy

#### Reviewed by:

Remy A. Bonnin, Université Paris-Saclay, France Stefano Gaiarsa, Fondazione IRCCS Policlinico San Matteo, Italy

> \*Correspondence: Simona Pollini simona.pollini@unifi.it

#### Specialty section:

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

Received: 08 April 2019 Accepted: 14 June 2019 Published: 02 July 2019

#### Citation:

Di Pilato V, Antonelli A, Giani T, Henrici De Angelis L, Rossolini GM and Pollini S (2019) Identification of a Novel Plasmid Lineage Associated With the Dissemination of Metallo-β-Lactamase Genes Among Pseudomonads. Front. Microbiol. 10:1504. doi: 10.3389/fmicb.2019.01504 <sup>1</sup> Department of Experimental and Clinical Medicine, University of Florence, Florence, Italy, <sup>2</sup> Microbiology and Virology Unit, Careggi University Hospital, Florence, Italy, <sup>3</sup> Department of Medical Biotechnologies, University of Siena, Siena, Italy

Acquisition of metallo-β-lactamases (MBLs) represents one of most relevant resistance mechanisms to all β-lactams, including carbapenems, ceftolozane and available β-lactamase inhibitors, in Pseudomonas spp. VIM-type enzymes are the most common acquired MBLs in Pseudomonas aeruginosa and, to a lesser extent, in other Pseudomonas species. Little is known about the acquisition dynamics of these determinants, that are usually carried on integrons embedded into chromosomal mobile genetic elements. To date, few MBL-encoding plasmids have been described in Pseudomonas spp., and their diversity and role in the dissemination of these MBLs remains largely unknown. Here we report on the genetic features of the VIM-1 encoding plasmid pMOS94 from P. mosselii AM/94, the earliest known VIM-1-producing strain, and of related elements involved in dissemination of MBL. Results of plasmid DNA sequencing showed that pMOS94 had a modular organization, consisting of backbone modules associated with replication, transfer and antibiotic resistance. Plasmid pMOS94, although not typable according to the PBRT scheme, was classifiable either in MOBF11 or MPF<sup>T</sup> plasmid families. The resistance region included the class I integron In70, carrying blaVIM-1, in turn embedded in a defective Tn402-like transposon. Comparison with pMOS94-like elements led to the identification of a defined plasmid lineage circulating in different Pseudomonas spp. of clinical and environmental origin and spreading different MBL-encoding genes, including blaIMP-63, blaBIM, and blaVIMtype determinants. Genetic analysis revealed that this plasmid lineage likely shared a common ancestor and had evolved through the acquisition and recombination of different mobile elements, including the MBL-encoding transposons. Our findings provide new insights about the genetic diversity of MBL-encoding plasmids circulating

**86**

among Pseudomonas spp., potentially useful for molecular epidemiology purposes, and revealed the existence and persistence of a successful plasmid lineage over a wide spatio-temporal interval, spanning over five different countries among two continents and over 20-years.

Keywords: metallo-β-lactamase, mobile genetic elements, integron, antibiotic resistance, plasmid

## INTRODUCTION

fmicb-10-01504 June 28, 2019 Time: 15:13 # 2

Pseudomonas is a genetically and metabolically diverse genus of bacteria, which inhabit a wide variety of environments and can act as pathogens of humans, animals and plants. Among the wide variety of Pseudomonas species, only few have been recognized as human pathogens, with Pseudomonas aeruginosa being the most common cause of infections (Driscoll et al., 2007; Høiby et al., 2015; Peix et al., 2018).

The acquisition by pathogenic Pseudomonas species of β-lactamases able to degrade most anti-pseudomonas β-lactams and resistant to currently available β-lactamase inhibitors, such as the metallo-β-lactamases (MBLs), can provide a remarkable contribution to the emergence of strains with difficult-to-treat resistance (DTR) phenotypes (Hawkey et al., 2018). VIMtype enzymes are among the most widespread and prevalent acquired MBLs in P. aeruginosa (Hong et al., 2015), and have occasionally been reported also in other Pseudomonas spp. of clinical interest including Pseudomonas putida, Pseudomonas mosselii, Pseudomonas monteilii, Pseudomonas stutzeri, and Pseudomonas mendocina (Yan et al., 2001; Lombardi et al., 2002; Giani et al., 2012; Ocampo-Sosa et al., 2015; Almuzara et al., 2018).

Metallo-β-lactamase genes, and in particular blaVIM- and blaIMP-type genes, are typically carried on mobile gene cassettes inserted into integron platforms (Partridge et al., 2018). While among Enterobacterales MBL-encoding integrons are mostly associated with plasmid lineages of IncN-, IncI1- and IncHI2 type replicons (Carattoli, 2009; Villa et al., 2014), less is known about the genetic support of these elements among Pseudomonas spp. In these species, MBL-encoding integrons have been described both as chromosomal- (Lauretti et al., 1999; Perez et al., 2014; Di Pilato et al., 2015; Roy Chowdhury et al., 2016) and plasmid-borne, but only few plasmids have been completely sequenced (Li et al., 2008; Bonnin et al., 2013; Vilacoba et al., 2014; San Millan et al., 2015; Botelho et al., 2017, 2018; Partridge et al., 2018; Liapis et al., 2019), and information about prevalence and diversity of MBL-carrying plasmid lineages remains largely unknown.

In this study we have investigated the genetic support of the blaVIM-<sup>1</sup> gene from a non-aeruginosa Pseudomonas strains of clinical origin, namely Pseudomonas mosselii AM/94, which represents the earliest known VIM-1-producing strain (Giani et al., 2012). In that strain, blaVIM-<sup>1</sup> was previously shown to be carried on a gene cassette inserted into integron In70, but its genetic support was not further investigated. Here we report that in this strain the integron was plasmid-borne and describe the structure of the blaVIM-1-carrying plasmid from AM/94, named pMOS94. The results of a comparative genetic analysis with pMOS94, allowed the definition of a novel plasmid lineage involved in dissemination of MBL genes among Pseudomonas spp.

### MATERIALS AND METHODS

### Bacterial Strains and Antimicrobial Susceptibility Testing

Pseudomonas mosselii AM/94 was isolated in 1994 from the lower respiratory tract of an inpatient in Genoa, Italy (Giani et al., 2012). P. aeruginosa C/53 and C/57 were isolated in 2014 from bloodstream infections from intensive care patients from Milan, Italy, and the complete genome has been previously determined (Giani et al., 2018). Identification to the species level and Sequence Type (ST) of C/57 strain were carried out in silico on whole genome sequence data. Antimicrobial susceptibility testing was carried out by reference brothmicrodilution method and results were interpreted according to EUCAST clinical breakpoints (EUCAST breakpoint tables version 9, 2019<sup>1</sup> ).

## Plasmid Transfer and Typing

Electrotransformation experiments were performed with electrocompetent E. coli DH10B and P. aeruginosa PAO-1 using plasmid DNA preparations, obtained as described previously (Docquier et al., 2003). The broad-host range vector plasmid pME6001 (Blumer et al., 1999) was used as positive control in electrotransformation experiments. Transformants were selected on LB agar containing ceftazidime (10 mg/L for E. coli and 20 mg/L for P. aeruginosa) or gentamicin (30 mg/L) for the pME6001 vector. Conjugation experiments were performed in solid medium as previously described (Docquier et al., 2003), using E. coli J53 Azi<sup>R</sup> as the recipient strain. Ceftazidime (10 mg/L) was used for the selection of transconjugants and sodium azide (150 mg/L) for counterselection of the donor. Plasmid incompatibility groups were determined following the previously described PCR-based replicon typing (PBRT) schemes (Carattoli et al., 2005; García-Fernández et al., 2009; Villa et al., 2010; Johnson et al., 2012). In silico typing using plasmid relaxases (MOB) regions (Garcillán-Barcia et al., 2009) and the mating pair formation (MPF) apparatus (Smillie et al., 2010) was performed according to the previously proposed classification scheme (Guglielmini et al., 2014; Orlek et al., 2017).

<sup>1</sup>www.eucast.org/clinical\_breakpoints/

## DNA Sequencing and Bioinformatic Analysis

Genomic DNA was extracted as previously described (Johnson, 1994). Plasmid DNA was extracted using Wizard <sup>R</sup> Plus SV Minipreps DNA purification system (Promega, Madison, United States) according to manufacturer's instructions. Plasmid DNAs from strain AM/94 was subjected to complete sequencing with the Illumina MiSeq platform (Illumina Inc., San Diego, United States) with a 2 × 250 or 2 × 300 bp paired-end approach. A total of 171,952 high quality reads were generated and assembled using SPAdes 3.11 (Bankevich et al., 2012). Sequence annotation was performed using the RAST web-server (Aziz et al., 2008). Finishing of pMOS94 and pAER57 plasmid sequences was achieved through Sanger sequencing of PCR products spanning contigs' gaps. Sequence comparisons were performed using BLAST<sup>2</sup> and Mauve<sup>3</sup> Core genome alignments were performed using Roary, using the non-paralog splitting method (Page et al., 2015; Rozwandowicz et al., 2017). Plasmid circular and linear maps were generated using the CGView Server and EasyFig tools, respectively (Grant and Stothard, 2008; Sullivan et al., 2011). Comparator plasmid sequences were selected from the NCBI-NIH database (non-redundant) to include all circular molecules with 100% sequence identity at nucleotide level of the repA-oriV region of plasmid pMOS94. A tentative plasmid classification based on replicase (rep) homology was inferred through nucleotide sequence alignments of rep genes from representative members of the IncN (García-Fernández et al., 2011), IncW (Suzuki et al., 2010; Aoki et al., 2018), and IncP9 replicons (Sevastsyanovich et al., 2008), collectively clustering into the MOBF11 relaxase family, to which pMOS94-like plasmids belong. Plasmids used in this analysis are reported in **Supplementary Table S1**. Sequences were aligned using the MAFFT v.7 server<sup>4</sup> with the G-INS-1 progressive method, and the rep phylogeny was inferred using the Neighborjoining method implemented by MAFFT. Phylogenetic trees were visualized and annotated through the Evolview software (He et al., 2016).

### RESULTS

#### Mapping of blaVIM-<sup>1</sup> in a Plasmid From P. mosselii AM/94

Pseudomonas mosselii AM/94 was previously shown to carry the blaVIM-<sup>1</sup> gene cassette as part of an In70 integron platform (50CS – blaVIM-<sup>1</sup> – aacA4 – aphA15 – aadA15 – 30CS) (Giani et al., 2012).

To investigate the genetic support of the blaVIM-<sup>1</sup> cassette and of its cognate integron in P. mosselii AM/94, an S1 nuclease assay and hybridization with a blaVIM-specific probe were performed. Results mapped blaVIM to a 50 kb plasmid (data not shown), hereafter referred to as pMOS94.

Gene transfer experiments failed in yielding transconjugants using E. coli J53 as recipient in conjugation experiments, and transformants using E. coli DH10B and P. aeruginosa PAO-1 as recipients in electrotransformation experiments.

The PBRT assay (Carattoli et al., 2005; García-Fernández et al., 2009; Villa et al., 2010; Johnson et al., 2012), performed on total DNA purified from AM/94, could not identify any plasmid replicon.

#### Sequence Analysis of pMOS94: General Features and Plasmid Backbone

In order to determine the genetic features of pMOS94, plasmid DNA was purified from AM/94 and completely sequenced. De novo sequence assembly yielded a single contig of 51,660 kb in length, which resulted in a 52,002 kb circular molecule, with a raw coverage of 380× and a mean G+C content of 59.6%, following sequence completion.

A total of 71 coding sequences (CDS) were identified by the RAST annotation pipeline, of which 37 had a predicted function consisting in: replication and stability (n = 7), transfer (n = 14), mobile genetic elements (n = 9) and antibiotic resistance (n = 7) (**Figure 1**).

Plasmid pMOS94 showed a modular organization associated with the following functions: (i) plasmid replication; (ii) conjugal transfer; (iii) antibiotic resistance.

The replication module included genes essential for plasmid replication (repA, res) and stability (stbABC), as well as a type II toxin-antitoxin system (vapBC) for stable maintenance. A putative replication origin (oriV) was identified immediately adjacent to repA, characterized by four 19-bp (TTCGTCACTCCAGGGACCG) and seven 9-bp (TTCGTCACT) iteron repeats, suggesting a theta-type replication mechanism for pMOS94 (Chattoraj, 2000).

The transfer module comprised a trw-type gene cluster encoding a T4SS machinery, composed by three structurally different components within the same segmented operon (Llosa and Alkorta, 2017) encoding: (i) the inner membrane complex (TrwMKIGB); (ii) the outer membrane core complex (TrwNHFE); (iii) the external pilus (TrwLJ). In silico analyses according to the MOB and MPF typing schemes, exploiting T4SS components for classification (Garcillán-Barcia et al., 2009; Smillie et al., 2010; Guglielmini et al., 2014; Orlek et al., 2017), revealed that pMOS94 was classifiable either in the MOBF11 or in the MPF<sup>T</sup> families.

The resistance region was organized in two non-contiguous segments, of which one was represented by the In70 integron platform carried on a defective Tn402-like transposon, whereas the second included Tn5393c, a Tn3-family transposon carrying the strA–strB genes, coding for phosphotransferases able to confer high level resistance to streptomycin (Chiou and Jones, 1993; Sunde and Norström, 2005). These segments were located downstream of and within the transfer module, respectively. The interruption of the trw operon by insertion of Tn5393c, apparently related with a transposition event generating 5-bp direct repeats (DRs) (5<sup>0</sup> -GAATA-3<sup>0</sup> ) flanking the element, could explain the unability of pMOS94 for conjugational transfer.

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

<sup>3</sup>http://darlinglab.org/mauve/mauve.html

<sup>4</sup>https://mafft.cbrc.jp/alignment/server/

Conversely, the Tn402-like transposon carrying In70 was inserted downstream of the transfer module and upstream of the locus coding for a DNA invertase, a core component of the plasmid backbone associated with replication and stability functions.

## Identification and Comparative Analysis of pMOS94-Like Plasmids From Other Pseudomonas Strains

In order to identify relatives of pMOS94 and to evaluate the possible host range and dissemination of a similar plasmid lineage, and its role in spreading clinically relevant antibiotic resistance determinants, an BLAST screening using the repAoriV region was performed against publicly available databases in NCBI (non-redundant and wgs) (last accessed on April 2, 2019).

Altogether, seven records were identified showing 100% identity with the repA-oriV region, of which five were circular plasmids ranging in size from 24 to 80 kb, while two corresponded to putative plasmid sequences from a recently characterized collection of MBL-producing P. aeruginosa from Italy (Giani et al., 2018). Strains harboring the latter putative plasmids, namely P. aeruginosa C/53 and C/57, were available for further analysis, and gaps closure by PCR resulted in two circular elements showing a size of 102 kb. Since C/53 and C/57 were isolated from the same surveillance center in the same

period (2014), and showed the same clonality (ST 532) and resistome (data not shown), C/57 was selected as representative for plasmid closure.

Interestingly, the analyzed plasmids were uniquely detected within members of the Pseudomonas genus, including strains of P. aeruginosa, P. putida, and P. mosselii. All of them carried an MBL-encoding gene (i.e., blaVIM, blaIMP, or blaBIM) in association with additional antibiotic resistance determinants (**Table 1**). In most of cases (6/7) the Pseudomonas strains were from clinical sources, with the exception of P. putida IEC33019, which was from an environmental source. Overall these strains were from five different countries, belonging to two continents, and had been isolated over a prolonged time period spanning over 20-years (**Table 1**).

Sequence analysis and comparison of pMOS94 with the other plasmids revealed an overall conserved backbone structure, containing genes involved in stability, maintenance and, in some cases, encoding the T4SS or parts thereof (**Figure 2**). Except for pAMBL2, where the transfer operon and the cognate relaxase were fully lacking, all plasmids were found to belong to the MOBF11/MPF<sup>T</sup> families and were untypeable by the replicon typing scheme, as pMOS94.

Compared to pMOS94, pIEC33019 exhibited the closest conserved architecture, with major differences due to the acquisition of mobile genetic elements targeting a different backbone region (**Figure 2**). Conversely, the most diverging elements were pAMBL2, pPC9 and pAER57. Plasmid pAMBL2 was a small (24.1 kb) deletion derivative of pMOS94, where the entire trw locus was lacking, carrying only genes essential for replication processes and a multiresistance region (**Figure 2**; San Millan et al., 2015). Plasmid pPC9 was characterized by a mosaic architecture, carrying mobile genetic elements and multiple genes coding for metal (e.g., merEA) and antibiotic resistance (e.g., In860) (**Figure 2** and **Table 1**; Molina et al., 2014). Plasmid pAER57 was the largest identified element of this lineage (>100 kb), and was characterized by a more complex chimeric arrangement, including a 43 kb segment sharing 87% nucleotide identity with the archetypal IncP-9 pWW0 catabolic plasmid from P. putida (GenBank Acc. No. AJ344068) (Greated et al., 2002; Sevastsyanovich et al., 2008). The 43 kb segment, likely acquired through recombination with a pWW0-like plasmid, included genes coding for replication (e.g., res-oriV-rep), partitioning (e.g., parABC, korA), transfer (e.g., mpfABCDEFGHIJ) and unknown functions. Therefore, pAER57 was classifiable as a multi-replicon plasmid, being characterized by an additional replication locus (i.e., res-oriV-rep) 91% identical to that of pWW0.

Analysis of SNPs evaluated on core orthologous genes of pMOS94-like plasmids showed an overall limited sequence diversity. In detail, all plasmids differed only by 1–2 SNPs from each other, with the exception of pAER57 that was the more divergent, showing 13–14 SNPs differences.

Further sequence analysis revealed that members of this plasmid lineage were also able to integrate into the host chromosome as linear elements. Evidences of integration came from a recent study reporting on the structure of a chromosomal genomic island carrying an MBL determinant from a P. aeruginosa strain of clinical origin isolated in Prague in 2015 (GenBank acc. no. KY860572). In this case, the integrated element was apparently a deletion-derivative of pMOS94 (carrying only backbone essential regions coding for replication functions) inserted within a Tn2 family transposon (Tn6162), in turn embedded into a genomic island (PACS171b) previously described in high risk clones of P. aeruginosa (i.e., ST 235) as chromosomal support of MBL-encoding genes (Martinez et al., 2012; Di Pilato et al., 2015).

Noteworthy, a plasmid lineage closely related (84–87% identity of the repA-oriV locus and belonging to the MOBF11 and MPF<sup>T</sup> families) to the one identified in this study was detected in the NCBI non-redundant database and included elements preferentially circulating within the genus Pseudomonas. These elements included: (i) plasmid pCT14 encoding mercury resistance from Pseudomonas veronii (Bramucci et al., 2006); (ii) two unnamed plasmids from P. aeruginosa AR\_0356 and 163940 producing a KPC-2 carbapenemase (GenBank acc. nos. CP027168.1 and CP029092.1); (iii) plasmid pTROUS1 from



<sup>a</sup>Uncharacterized gene variants are named according to the closest homolog plus "v".

P. aeruginosa 163940 producing an IMP-63 carbapenemase (GenBank acc. no MK047610.1), and (iv) a non-circular molecule resembling a plasmid element (GenBank acc. no. LC350084.1) coding for a VIM-2 carbapenemase. Notably, most of these elements were associated with clinically relevant resistance determinants and, according to available metadata, were of clinical (pTROUS1 and LC350084.1) or environmental (pCT14) origin. Using the repA-oriV locus as probe, at least 96 records were found among Pseudomonas spp. draft genome sequences in NCBI wgs database (i.e., showing 100–90% identity), highlighting a potential wide dissemination of pMOS94-related elements. Nevertheless, the plasmidic nature of such records could not be verified due to their fragmentation and unfinished status. In order to infer possible phylogenetic relationships between pMOS94, pMOS94-related plasmids (i.e., those showing 84–87% identity of the repA-oriV locus) and known plasmid replicons belonging to the MOBF1relaxase family, to which pMOS94-like plasmids belong, a classification according to the rep locus was attempted. The analysis showed that pMOS94 and related plasmids clustered in a separate branch, that appeared more closely related to IncW than IncN and IncP9 replicons (**Figure 3**).

## Genetic Environment of the MBL-Encoding Genes in the pMOS94-Like Plasmids

All seven plasmids of the pMOS94-like lineage described above carried an MBL-encoding gene, embedded in a class

I integron platform (**Figure 4**), including blaVIM-<sup>1</sup> in four plasmids, and the blaVIM-2, blaBIM, and blaIMP-<sup>63</sup> in the remining ones. The blaVIM-<sup>1</sup> gene was found embedded in the In70 integron (in pMOS94) or in its derivatives (e.g., In110, In860, In1167) carrying a modified version of the typical gene cassette array of this element (i.e., in pAMBL2, pAER57, and pPC9). On the other hand, the other MBL genes were carried in integrons of different structure (e.g., In58, In1326, and In1297; **Figure 4**). Despite their heterogeneity, the MBL-

encoding integrons always carried aminoglycosides resistance determinants, with aacA4 variants being the most frequent (**Figure 4**). Most of the integrons carried typical 50CS and 30CS regions, except for In110 in pAER57, where the intI1 gene was partially deleted (**Figure 4**).

The MBL-encoding integrons were always associated with defective Tn402-like transposons. In most cases (namely pAMBL2, pPC9, pMOS94, pAER57, and pCB58), the Tn402 25 bp inverted repeats (IRi and IRt) were present at the boundaries of the defective transposons, while in the remaining plasmids (i.e., pTROUS2 and pIEC33019) IRi was present adjacent to the integrase gene, suggesting recombination events which deleted the IRt end. The Tn402 tni transposition module was present downstream the 30CS region in a partially deleted form (in five plasmids), or completely absent (in the others) (**Figure 4**).

In some cases, additional mobile elements were found within the Tn402 elements, targeting the cassette array or other regions (**Figure 4**). The ISPsp7-like and IS110-like insertion sequences, respectively targeting the cassette arrays of pTROUS2 and pIEC33019 were flanked by DRs, suggesting acquisition by transposition.

### Evolutionary History of the Tn402-Like Transposons Carried by pMOS94-Like Plasmids

In all cases the Tn402-like elements carried by the pMOS94 like plasmids were not flanked by 5 bp DRs, that represent the molecular markers of the transposition of this kind of mobile elements, suggesting that acquisition had involved different recombination events, or that recombination events involving one transposon moiety had occurred after transposition.

In all plasmids but pIEC33019, the Tn402-like element was inserted into an A-T rich region located upstream of the locus coding for a DNA invertase. Within this region, all the Tn402 like elements carrying blaVIM-<sup>1</sup> were inserted at the same site (i.e., 258 bp upstream of the DNA invertase gene), while the remaining ones were inserted slightly upstream (i.e., 281 bp upstream of the DNA invertase gene). On the other hand, in plasmid pIEC33019 the Tn402-like element was inserted in a different backbone region located downstream of repA (**Figure 4**).

Interestingly, all Tn402-like transposons carrying blaVIM-<sup>1</sup> were flanked by a copy of the ISPa17 insertion sequence (Haines et al., 2005), located upstream of the IRi end (**Figure 4**). Conversely, a different ISPa17/Tn402-like transposon assembly was found in plasmid pTROUS2, where the IS element was inserted downstream of the class I integron 30CS region and disrupted the tniA gene (**Figure 4**).This IS was already described as associated to MBL-encoding Tn402-like transposons aboard plasmids from GNNFs, where it likely contributed to the MBL gene platform acquisition through a novel transposition mechanism exploiting the similarity between the Tn402 IRs and their counterparts in ISPa17 (Di Pilato et al., 2014; Botelho et al., 2017; Liapis et al., 2019). Nevertheless, in pMOS94-like plasmids no evidence of an ISPa17-mediated mobilization was found, since no DRs bounding the ISPa17 IR-L and the Tn402 IRt end were detected (**Figure 4**).

## DISCUSSION

In this work we characterized the blaVIM-1-carrying plasmid pMOS94 from P. mosselii AM/94 isolated in Italy in 1994,

100% identity with the repA-oriV region.

which represents the earliest known VIM-producing strain (Giani et al., 2012). Analysis of the complete sequence of pMOS94 led to the identification of a novel plasmid lineage characterized by an original replication origin (i.e., repAoriV) and included in the MOBF11 and MPF<sup>T</sup> families, according to the typing schemes based on plasmid relaxases and on the MPF system, respectively (Guglielmini et al., 2014; Orlek et al., 2017). Interestingly, the MOBF11 family includes relaxases of plasmids belonging to IncW and IncN replicons, responsible for the spreading of antibiotic resistance determinants among enterobacteria, and to the IncP9 group, including metal-resistance and xenobiotic-biodegradation plasmids from Pseudomonas spp. (Garcillán-Barcia et al., 2009;

striped blocks that are connected across the different structures.

Shintani et al., 2015). Based on replicase sequence homology, pMOS94-like plasmids showed a closer relatedness with the IncW replicon family. Further transfer experiments will be needed to evaluate the incompatibility behavior of these plasmids.

Sequence analyses revealed that members of this novel plasmid lineage were involved in spread of MBL-encoding genes, and accounted for a broad dissemination of such resistance determinants among pseudomonads of clinical and environmental origin, spanning over a time period of at least 20 years, among at least five different countries from two continents. However, the association of pMOS94 and pMOS94 like plasmids with resistance genes could be somewhat biased by

25-bp IRs are indicated by a black vertical line. When present, DRs flanking the different mobile elements are shown as black dots. The diverse insertion point of Tn402-like transposons in pTROUS2 and pCB58 is indicated by an asterisk. Homologous segments (>99%) generated by a BLASTn comparison are shown as gray

the fact that the main purpose of bacterial sequencing focuses on antibiotic resistance.

The comparison of pMOS94-like plasmids highlighted a high phylogenetic relatedness within the lineage, which had likely evolved from a common ancestor. This point was also supported by the low sequence divergence observed among core backbone genes with the exception of pAER57, which was slightly more divergent. Nonetheless, the different insertion points of Tn402-like transposons and cognate integrons within the accessory region might represent a molecular marker of different evolutionary pathways for pMOS94 and pMOS94 like plasmids. In this scenario, different hypotheses could be made on the evolution of MGEs associated with this plasmid lineage: (i) the Tn402-like transposons could have been acquired following different insertion events that, in some cases (e.g., plasmids harboring blaVIM-1), occurred at the same site; (ii) at least three insertion events (occurring at different sites) led to the acquisition of the Tn402-like transposons, that subsequently evolved independently by genetic rearrangements and gene cassettes loss/acquisition. The number of the analyzed plasmids and sequence data publicly available for comparison, however, are limited and the chronological order of these events (i.e., the evolution of the backbone core genes and the acquisition of the accessory resistance regions) remains to be clarified.

Recently, a number of studies have described novel plasmid supports for different MBL genes among pseudomonads (Li et al., 2008; Bonnin et al., 2013; Vilacoba et al., 2014; San Millan et al., 2015; Botelho et al., 2017, 2018; Partridge et al., 2018; Liapis et al., 2019), reporting on the structure of single plasmids or on structures circulating in a restricted geographical area or within a relatively short time period.

The present study contributed to extend current knowledge about the diversity and phylogeny of plasmids circulating in Pseudomonas spp., reporting on a successful plasmid lineage carrying MBL genes. Although mobilization experiments were not successful under the experimental conditions used in this study, the heterogeneous range of species carrying pMOS94-like plasmids suggests that these elements are able to be transferred. Therefore, further experiments are needed to evaluate their mobilization potential among GNNFs and eventually Enterobacterales, in order to better understand the possible contribution

#### REFERENCES


in disseminating antibiotic resistance traits by this novel plasmid lineage.

#### NUCLEOTIDE SEQUENCE ACCESSION NUMBERS

The complete nucleotide sequence of pMOS94 from P. mosselii AM/94 and pAER57 from P. aeruginosa C/57 have been registered in GenBank under accession numbers MK671725 and MK671726, respectively.

### DATA AVAILABILITY

The nucleotide sequences described for the first time in this study, namely pMOS94 and pAER57 sequences, have been deposited in GenBank database and will be made available to other researchers following publication. The remaining plasmid nucleotide sequences analyzed in this study are publicly available in the GenBank database under the accession numbers cited in the manuscript.

#### AUTHOR CONTRIBUTIONS

VDP and AA did plasmid sequencing, analyzed the data, and drafted the manuscript. TG and LHDA produced the phenotypic data and handled the samples. GR and SP coordinated the experiments, drafted and edited the manuscript.

#### FUNDING

This work was supported in part by EvoTAR (No. HEALTH-F3- 2011-2011-123282004) to GR.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2019.01504/full#supplementary-material



systems involved in bacterial conjugation or protein secretion. Nucleic Acids Res. 42, 5715–5727. doi: 10.1093/nar/gku194



**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 © 2019 Di Pilato, Antonelli, Giani, Henrici De Angelis, Rossolini and Pollini. 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) and the copyright owner(s) 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 a Multiresistance Plasmid Carrying the optrA and cfr Resistance Genes From an Enterococcus faecium Clinical Isolate

Gianluca Morroni<sup>1</sup> , Andrea Brenciani<sup>2</sup> \*, Alberto Antonelli<sup>3</sup> , Marco Maria D'Andrea3,4 Vincenzo Di Pilato<sup>3</sup> , Simona Fioriti<sup>2</sup> , Marina Mingoia<sup>2</sup> , Carla Vignaroli<sup>5</sup> , Oscar Cirioni<sup>1</sup> Francesca Biavasco<sup>5</sup> , Pietro E. Varaldo<sup>2</sup> , Gian Maria Rossolini3,6 and Eleonora Giovanetti<sup>5</sup>

1 Infectious Diseases Clinic, Department of Biomedical Sciences and Public Health, Polytechnic University of Marche Medical School, Ancona, Italy, <sup>2</sup> Unit of Microbiology, Department of Biomedical Sciences and Public Health, Polytechnic University of Marche Medical School, Ancona, Italy, <sup>3</sup> Department of Experimental and Clinical Medicine, University of Florence, Florence, Italy, <sup>4</sup> Department of Medical Biotechnologies, University of Siena, Siena, Italy, <sup>5</sup> Unit of Microbiology, Department of Life and Environmental Sciences, Polytechnic University of Marche, Ancona, Italy, <sup>6</sup> Microbiology and Virology Unit, Florence Careggi University Hospital, Florence, Italy

Enterococcus faecium E35048, a bloodstream isolate from Italy, was the first strain where the oxazolidinone resistance gene optrA was detected outside China. The strain was also positive for the oxazolidinone resistance gene cfr. WGS analysis revealed that the two genes were linked (23.1 kb apart), being co-carried by a 41,816-bp plasmid that was named pE35048-oc. This plasmid also carried the macrolide resistance gene erm(B) and a backbone related to that of the well-known Enterococcus faecalis plasmid pRE25 (identity 96%, coverage 65%). The optrA gene context was original, optrA being part of a composite transposon, named Tn6628, which was integrated into the gene encoding for the ζ toxin protein (orf19 of pRE25). The cfr gene was flanked by two ISEnfa5 insertion sequences and the element was inserted into an lnu(E) gene. Both optrA and cfr contexts were excisable. pE35048-oc could not be transferred to enterococcal recipients by conjugation or transformation. A plasmid-cured derivative of E. faecium E35048 was obtained following growth at 42◦C, and the complete loss of pE35048-oc was confirmed by WGS. pE35048-oc exhibited some similarity but also notable differences from pEF12-0805, a recently described enterococcal plasmid from human E. faecium also co-carrying optrA and cfr; conversely it was completely unrelated to other optrA- and cfr-carrying plasmids from Staphylococcus sciuri. The optrA-cfr linkage is a matter of concern since it could herald the possibility of a co-spread of the two genes, both involved in resistance to last resort agents such as the oxazolidinones.

Keywords: multiresistance plasmid, optrA gene, cfr gene, oxazolidinone resistance, Enterococcus faecium

#### Edited by:

Teresa M. Coque, Instituto Ramón y Cajal de Investigación Sanitaria, Spain

#### Reviewed by:

Yoshikazu Ishii, Toho University, Japan Ana R. Freitas, Universidade do Porto, Portugal

#### \*Correspondence:

Andrea Brenciani a.brenciani@univpm.it; andreabrenciani@yahoo.it

#### Specialty section:

This article was submitted to Evolutionary and Genomic Microbiology, a section of the journal Frontiers in Microbiology

Received: 30 April 2018 Accepted: 27 August 2018 Published: 11 September 2018

#### Citation:

Morroni G, Brenciani A, Antonelli A, D'Andrea MM, Di Pilato V, Fioriti S, Mingoia M, Vignaroli C, Cirioni O, Biavasco F, Varaldo PE, Rossolini GM and Giovanetti E (2018) Characterization of a Multiresistance Plasmid Carrying the optrA and cfr Resistance Genes From an Enterococcus faecium Clinical Isolate. Front. Microbiol. 9:2189. doi: 10.3389/fmicb.2018.02189

## INTRODUCTION

fmicb-09-02189 September 11, 2018 Time: 14:14 # 2

Enterococci are members of the gut microbiota of humans and many animals, and are widespread in the environment. They are also major opportunistic pathogens, mostly causing healthcarerelated infections. Among the reasons of their increasing role as nosocomial pathogens, the primary factor is their inherent ability to express and acquire resistance to several antimicrobial agents, with Enterococcus faecium emerging as the most therapeutically challenging species (Arias and Murray, 2012).

Oxazolidinones are among the few agents that retain activity against multiresistant strains of enterococci (Shaw and Barbachyn, 2011; Patel and Gallagher, 2015), and the emergence of resistance to these drugs is an issue of notable clinical relevance. Particularly worrisome, due to their potential for horizontal dissemination, are the oxazolidinone resistances caused by cfr, encoding a ribosome-modifying enzyme (Kehrenberg et al., 2005; Deshpande et al., 2015; Munita et al., 2015), and optrA, encoding a ribosome protection mechanism (Wang et al., 2015; Wilson, 2016; Sharkey et al., 2016). Both these genes were found to be associated with a number of different mobile genetic elements.

The optrA gene, in particular, was discovered in China in enterococci of human and animal origin isolated in 2005- 2014 (Wang et al., 2015) where it was detected in different genetic contexts (He et al., 2016). Since then, optrA-positive enterococci have been reported worldwide (Mendes et al., 2016; Cavaco et al., 2017; Freitas et al., 2017; Pfaller et al., 2017a,b), including Italy, where optrA was found — the first report outside China — in two bloodstream isolates of E. faecium which were also positive for the cfr gene, which was not expressed (Brenciani et al., 2016b). By further investigating one of those isolates (strain E35048), we noticed that both optrA and cfr were capable of undergoing excision as minicircles (Brenciani et al., 2016b). It is worth noting that among the reported optrA protein variants (Morroni et al., 2017), the one detected in E. faecium E35048, named optrAE35048, is the most divergent, differing by 21 amino acid substitutions from the firstly described optrA variant (Wang et al., 2015).

The goal of the present work was to investigate the locations, genetic environments, and transferability of the optrA and cfr resistance genes detected in E. faecium E35048. We characterized the genetic contexts and location of optrA and cfr in E. faecium E35048, and found that both genes were co-carried on a plasmid of original structure, named pE35048-oc. This plasmid, which also carried the macrolide resistance gene erm(B), shared regions of homology with the well-characterized (Schwarz et al., 2001) and widely distributed (Rosvoll et al., 2010; Freitas et al., 2016) conjugative multiresistance enterococcal plasmid pRE25, but was unable to transfer. In pE35048-oc, the genetic context of optrA was different from those so far described in other optrAcarrying plasmids, underscoring the plasticity of these resistance regions.

## MATERIALS AND METHODS

### Bacterial Strain

optrA- and cfr-positive E. faecium E35048 (linezolid MIC, 4 µg/ml; tedizolid MIC, 2 µg/ml) was isolated in Italy in 2015 from a blood culture (Brenciani et al., 2016b).

## WGS and Sequence Analysis

Genomic DNA was extracted using a commercial kit (Sigma-Aldrich, St. Louis, MO). WGS was carried out with the Illumina MiSeq platform (Illumina Inc., San Diego, CA, United States) by using a 2 × 300 paired end approach and a DNA library prepared using Nextera XT DNA Sample Prep Kit (Illumina, San Diego, CA, United States). De novo assembly was performed with SPAdes V 3.10.0 (Bankevich et al., 2012) using default parameters. Scaffolds characterized by a length ≤ 300 bp were filtered out. Raw reads were mapped to the filtered scaffolds by using bwa (Li and Durbin, 2009) to check the quality of the assembly. Tentative ordering of selected scaffolds of plasmid origin was performed by BLASTN comparisons of data from WGS to homologous plasmids, and eventually confirmed by PCR approach followed by Sanger sequencing. The ST was determined through the Center for Genomic Epidemiology<sup>1</sup> . Analysis of insertion sequences was carried out using ISFinder online database<sup>2</sup> (Siguier et al., 2006).

## PCR Mapping Experiments

PCR mapping with outward-directed primers topo-FW (50 -GAAGCGACAAGAGCAAGTAT-3<sup>0</sup> ) and optrA-RV (5<sup>0</sup> - TCTTGAACTACTGATTCTCGG-3<sup>0</sup> ), and Sanger sequencing were used to close the pE35048-oc plasmid sequence.

To investigate the excision of the optrA and cfr genetic contexts, PCR mapping and sequencing assays were performed using: (i) primer pairs targeting the regions flanking their insertion sites [orf7-FW (5<sup>0</sup> - ATTTTTCTTTTGATTTGGTA-3<sup>0</sup> ) and orf14up-RV (50 -AAGTAATCTTTTTTTGTTTT-3<sup>0</sup> ) for the cfr genetic context; and orf33-FW (5<sup>0</sup> -CTTGTTTTGGTGTTGCCCTGG-3 0 ) and orf33-RV (5<sup>0</sup> -CCACCAAGTAAAAAAGCGG-3<sup>0</sup> ) for the optrA genetic context]; (ii) outward-directed primer pairs designed from cfr and optrA genes [cfr-INV (50 -TTGATGACCTAATAAATGGAAGTA-3<sup>0</sup> ) and cfr-FW (5<sup>0</sup> -ACCTGAGATGTATGGAGAAG-3<sup>0</sup> ); optrA-INV (50 -TTTTTCCACATCCATTTCTACC-3<sup>0</sup> ) and optrA-FW (50 -GAAAAATAACACAGTAAAAGGC-3<sup>0</sup> )] (**Figure 1**).

## S1-PFGE, Southern Blotting and Hybridization

Total DNA in agarose gel plugs was digested with S1 nuclease (Thermo Fisher Scientific, Milan, Italy) and separated by PFGE as previously described (Barton et al., 1995). After S1-PFGE, DNA was blotted onto positively charged nylon membrane (Ambion-Celbio, Milan, Italy) and hybridized with specific

<sup>1</sup>https://cge.cbs.dtu.dk/services/MLST/

<sup>2</sup>https://isfinder.biotoul.fr/

probes (Brenciani et al., 2007). cfr and optrA probes were obtained by PCR as described elsewhere (Wang et al., 2015; Brenciani et al., 2016a).

are indicated by thin arrows below pE35048-oc. Gray areas between ORF maps denote > 90% DNA identity.

## Transformation and Conjugation Experiments

Purified plasmids extracted from E. faecium E35048 were transformed into the E. faecalis JH2-2 recipient by electrotransformation as described previously (Brenciani et al., 2016a). The transformants were selected on plates supplemented with florfenicol (10 µg/ml) or erythromycin (10 µg/ml).

In mating experiments, E. faecium E35048 was used as the donor. Two florfenicol-susceptible laboratory strains were used as recipients: E. faecium 64/3 (Werner et al., 1997), and E. faecalis JH2-2, both resistant to fusidic acid. Conjugal transfer was performed on a membrane filter. Transconjugants were selected on plates supplemented with florfenicol (10 µg/ml) or erythromycin (10 µg/ml) plus fusidic acid (25 µg/ml).

## Curing Assays

Enterococcus faecium E35048 was grown overnight in brain heart agar (BHA) at 42◦C for some passages. After each passage a few colonies were picked up, and their DNA was extracted and screened for the presence of the optrA and cfr genes by PCR with specific primers (Brenciani et al., 2016b). In case of negative testing, the strain was regarded as possibly cured and subjected to WGS for confirmation.

## Nucleotide Sequence Accession Numbers

The complete nucleotide sequence of plasmid pE35048-oc has been assigned to GenBank accession no. MF580438, available under the BioProject ID PRJNA481862.

## RESULTS AND DISCUSSION

#### Genome General Features and Resistome of E. faecium E35048

Assembly of the raw WGS data followed by filtering of low length contigs gave a total of 172 scaffolds (range: 310-137, 266 bp; N50: 41,930 bp; L50: 18; mean coverage: 92X). E. faecium E35048 was assigned to ST117, a globally disseminated hospital-adapted clone (Hegstad et al., 2014; Tedim et al., 2017). Resistome analysis revealed the presence of six acquired resistance genes in addition to the previously described optrA and cfr genes: erm(B) (resistance to macrolides, lincosamides and group B streptogramins), msr(C) (resistance to macrolides and group B streptogramins), tet(M) (resistance to tetracycline), aphA and aadE (resistance to aminoglycosides), and sat4 (resistance to streptothricin).

## Characterization of the optrA- and cfr-Carrying Plasmid pE35048-oc

The optrA and cfr genes were found to be linked, 23.1 kb apart in the linearized form, on the same contig, which also contained regions of high similarity (96% nucleotide identity) to the E. faecalis plasmid pRE25 (50 kb) (65% coverage) (Schwarz et al., 2001) (GenBank accession no. NC\_008445).

PCR and Sanger sequencing using outward-directed primers targeting orf1 and optrA demonstrated that the region containing optrA and cfr was part of a plasmid which was designated pE35048-oc (**Figure 1**). The plasmid was 41,816 bp in size, contained 42 open reading frames (ORFs), and had a G + C content of 35%.

S1-PFGE analysis of genomic DNA extracted from E. faecium E35048 showed four plasmids, ranging in size from ∼10 to

∼250 kb (data not shown). Both optrA and cfr probes hybridized with a plasmid of ∼45 kb, in agreement with sequencing data.

The characteristics of the plasmid ORFs and of their products are detailed in **Table 1**. In particular, pE35048-oc carried (i) a repS gene (orf6, corresponding to orf6 of pRE25), encoding a theta mechanism replication protein responsible for the plasmid replication; (ii) a putative origin of replication downstream of orf6; and (iii) a region containing the putative minimal conjugative unit of pRE25, consisting of 15 ORFs (orf28 to orf14, corresponding to orf24 to orf39 of pRE25) and the origin of transfer (oriT) found upstream of orf28 (Schwarz et al., 2001). BLASTN analysis showed that the oriT nucleotide sequence was shorter in pE35048-oc (only 15 bp vs. 38 bp in pRE25). Compared to pRE25, pE35048-oc lacked (i) the region spanning from orf41 to orf5 (two IS1216 elements probably involved in the rearrangement occurred during plasmid evolution); (ii) orf10, i.e., the chloramphenicol resistance cat gene; and (iii) orf11, another replication gene encoding a rolling-circle replication protein. In addition, compared to pRE25, pE35048-oc carried the optrA and cfr genes and their respective genetic environments.

The optrA context (5,850 bp) consisted of the optrAE35048 gene followed by a novel insertion sequence of the IS21 family, named ISEfa15. Consistently with other members of this family (Berger and Haas, 2001), ISEfa15 included two CDS encoding a transposase and a helper protein, and was bounded by 11 bp imperfect inverted repeats (IRL 5<sup>0</sup> -TGTTTATGATA-3<sup>0</sup> and IRR 5<sup>0</sup> -TGTATTTGTCA-3<sup>0</sup> ). A truncated copy of ISEfa15, named ISEfa15<sup>∗</sup> , was present also upstream of optrA gene. This optrA context was flanked by 5-bp target site duplications (5<sup>0</sup> -CTAAT-3 0 ) suggesting its mobilization as a composite transposon, named Tn6628 (**Figure 1**). This transposon was previously shown to form circular intermediate (3,350 bp) including optrA and the truncated copy of ISEfa15 (Brenciani et al., 2016b).

The proposed role of IS1216 in the dissemination of optrA among different types of enterococcal plasmids (He et al., 2016) is likely to be true also for other transposase genes. The optrA context was located downstream of the erm(B) gene (orf15 of pRE25) and was integrated into orf33 (orf19 of pRE25, which encodes the ζ toxin protein of the ω-ε-ζ toxin/antitoxin system). This integration inactivates ζ toxin encoded by orf33, a condition that could prevent the correct partitioning of pE35048-oc and lead to the appearance of plasmid-free segregants (Magnuson, 2007).

The cfr context (6,098 bp) was located between orf7 and orf14 (orf39 and orf40 of pRE25) and consisted of the cfr gene flanked by two ISEnfa5 elements, inserted in turn into the lnu(E) gene. The same genetic context of cfr [including the direct repeats and the lnu(E) gene] has been reported in China in a plasmid from a Streptococcus suis isolate from an apparently healthy pig (Wang et al., 2013) and in Italy in an MRSA isolated from a patient with cystic fibrosis (Antonelli et al., 2016), with cfr being untransferable in both instances. Very recently, the cfr gene, flanked by only one ISEnfa5, inserted upstream, has been described in a chromosomal fragment shared by three pig isolates of Staphylococcus sciuri (Fan et al., 2017).

PCR assays, using primer pairs targeting regions flanking the optrA and the cfr contexts (**Figure 1**), and sequencing experiments confirmed that both genes could be excised leaving one of the two flanking genes (ISEfa15 or ISEnfa5, respectively) at the excision sites.

#### Transferability of the optrA and cfr Genes and Curing of E. faecium E35048 From pE35048-oc

Repeated attempts of conjugation and transformation assays failed to demonstrate any optrA or cfr transfer from E. faecium E35048 to enterococcal recipients. The partial deletion of oriT and the lack of the rolling-circle replication protein might be responsible for the non-conjugative behavior of pE35048-oc compared to pRE25 (Schwarz et al., 2001).

An optrA- and cfr-negative isogenic strain of E. faecium E35048 was obtained after three passages on BHA at 42◦C. It was subjected to WGS. Compared to the wild type, it disclosed complete loss of pE35048-oc.

### pE35048-oc vs. Other Plasmids Sharing Co-carriage of optrA and cfr

Since this study was started, co-location of optrA and cfr has been reported in a few additional plasmids, some from pig isolates of S. sciuri (Li et al., 2016; Fan et al., 2017) and one, pEF12-0805, from a human isolate of E. faecium (Lazaris et al., 2017). Comparison of pE35048 oc with the S. sciuri plasmids revealed completely unrelated backbones and optrA and cfr contexts. On the other hand, pE35048-oc was related with pEF12-0805 (accession no. KY579372.1) although with significant differences (**Figure 1**). In particular:

(i) pE35048-oc and pEF12-0805 share a pRE25-related backbone (Schwarz et al., 2001), but pEF12-0805 is much larger (72,924 bp vs. 41,816 bp) due to the presence of a larger amount of pRE25-related regions, including the pRE25 region spanning from orf51 to orf5 (∼12,5 kb) and a rearranged region of pRE25 containing antibiotic resistance genes aphA, aadE, and lnu(B) (∼13 kb). (ii) A ∼4-kb remnant of the ermA-carrying transposon Tn554 (Murphy et al., 1985) is found only in pEF12-0805. (iii) The optrA contexts of the two plasmids are completely different, only the optrA gene of pE35048-oc being part of a composite transposon. The absence of insertion sequences makes it unlikely that the optrA gene of pEF12-0805 is excisable. Moreover, whereas in pE35048-oc the optrA context is found downstream of erm(B), the optrA gene of pEF12-0805 is associated with the ermA-carrying Tn554 remnant. (iv) Interestingly, the cfr contexts of the two plasmids are the same, including some plasmid backbone flanking regions on either side (**Figure 1**), suggesting that the two plasmids might be derived from a pRE25 related common ancestor that had initially acquired the mobile cfr element. (v) Repeated transfer assays were unsuccessful with both plasmids. Finally, (vi) whereas we obtained a pE35048-occured derivative of our E. faecium isolate, curing assays were unsuccessful with E. faecium strain F120805 (Lazaris et al., 2017).

#### TABLE 1 | Amino acid sequence identities/similarities of putative proteins encoded by the pE35048-oc (GenBank accession no. MF580438).


(Continued)

#### Table 1 | Continued

fmicb-09-02189 September 11, 2018 Time: 14:14 # 6


<sup>a</sup>For each ORF, only the most significant identity detected is listed. <sup>b</sup>1 represents a truncated ORF.

The E. faecium hosts of the two plasmids belonged to different sequence types and were isolated from different sources. Strain E35048 was recovered in 2015 in Italy from a blood culture, belonged to ST117, exhibited no mutational mechanisms of oxazolidinone resistance, and was vancomycin susceptible. Strain F120805, recovered in 2013 in Ireland from feces and reported to have a linezolid MIC of 8 µg/ml, belonged to ST80, exhibited also mutational mechanisms of oxazolidinone resistance (involving both 23S rRNA and ribosomal protein L3), and was vancomycin resistant (vanA genotype). Although belonging to different sequence types, ST80 and ST117 were part of the same clonal group, ST78.

## CONCLUSION

Distinctive findings of the optrA- and cfr-carrying plasmid pE35048-oc are its relation to the well-known enterococcal plasmid pRE25, shared with plasmid pEF12-0805

(Lazaris et al., 2017); a unique optrA context, that has never been described before; and the fact that both the optrA and cfr contexts are capable of excising to form minicircles. This, in addition to the belonging of E. faecium E35048 to ST117, a globally disseminated clone recovered in many European health institutions (Hegstad et al., 2014; Tedim et al., 2017), might favor the spread of optrA and cfr in the hospital setting. Under this respect, the in vitro non-transferability of pE35048-oc is someway reassuring, although transfer in vivo cannot be ruled out. Moreover, at the hospital level, it cannot be excluded that co-carriage of optrA and cfr by the same plasmid ends up turning into co-spread, as already highlighted with pheromone-responsiveness plasmids (Francia and Clewell, 2002), and also in consideration of the

#### REFERENCES


very recent finding that, in enterococci, non-conjugative plasmids can be mobilized by co-resident, conjugative plasmids (Di Sante et al., 2017). Co-spread would be a cause for special concern, considering that both optrA and cfr encode resistance, through diverse mechanisms, to different antibiotics, including last resort agents such as oxazolidinones.

#### AUTHOR CONTRIBUTIONS

AB, PV, and EG designed the study and wrote the paper. FB, OC, and GR have contributed to critical reading of the manuscript. GM, AA, MD, VD, SF, MM, CV, and SF did the laboratory work.



**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 © 2018 Morroni, Brenciani, Antonelli, D'Andrea, Di Pilato, Fioriti, Mingoia, Vignaroli, Cirioni, Biavasco, Varaldo, Rossolini and Giovanetti. 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) and the copyright owner(s) 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.

# Excision and Circularization of Integrative Conjugative Element Tn5253 of Streptococcus pneumoniae

#### Francesco Santoro, Alessandra Romeo, Gianni Pozzi and Francesco Iannelli\*

Laboratory of Molecular Microbiology and Biotechnology, Department of Medical Biotechnologies, University of Siena, Siena, Italy

The integrative conjugative element (ICE) Tn5253 of Streptococcus pneumoniae, conferring resistance to tetracycline and chloramphenicol, was found integrated at a 83-bp specific target site (attB) located in the rbgA gene of the pneumococcal chromosome. PCR analysis of Tn5253-carrying strains showed evidence of precise excision of Tn5253 from the pneumococcal chromosome with production of (i) circular forms of the ICE in which the ends were joined by a 84-bp sequence (attTn), and (ii) reconstituted chromosomal attB. When integrated into the chromosome, Tn5253 was flanked by attL, identical to attB, and attR, identical to attTn. Circular forms of Tn5253 were present at a concentration of 3.8 × 10−<sup>4</sup> copies per chromosome, whereas reconstituted attB sites were at 3.0 × 10−<sup>4</sup> copies per chromosome. Deletion of int-xis of Tn5253 abolished production of circular forms (<7.1 × 10−<sup>6</sup> copies per chromosome) and was associated to the lack of Tn5253 conjugal transfer suggesting, as expected, that Tn5253 circular form acts as a conjugation intermediate.

Keywords: integrative conjugative element (ICE), circular form, attachment site, conjugative transposon, Tn5253,

## INTRODUCTION

conjugation, mobile genetic elements

Horizontal gene transfer, mediated by MGEs, significantly drives bacterial genome evolution including the acquisition and dissemination of new patterns of antibiotic resistance (Burrus and Waldor, 2004). Functional characterization of MGEs is essential to understand the evolution and spread of antibiotic resistance within a given bacterial species and also among different species (Frost et al., 2005). ICEs, which include CTs, are MGEs that integrate into the bacterial genome and are capable of intracellular transposition to a new genomic location or intercellular transposition to a new genome host upon conjugative transfer (Mullany et al., 2002). ICEs account for up to 25% of the genetic material in a bacterial genome (Paulsen et al., 2003) and are the major promoters of genetic diversity in bacteria (Burrus and Waldor, 2004; Johnson and Grossman, 2015).

The CT Tn916, carrying the tet(M) gene, is the prototype of the Tn916–Tn1545 family of ICEs, and one of the most studied ICEs of gram positive bacteria (Santoro et al., 2014). Tn916 was shown to excise from the bacterial chromosome producing a covalently closed circular form of the element which was called "CI." Production of CIs of Tn916 was demonstrated to be essential for conjugative transposition of the element (Scott et al., 1988). Recombination processes of ICEs are catalyzed by

#### Edited by:

Simona Pollini, Università degli Studi di Firenze, Italy

#### Reviewed by:

Marina Mingoia, Università Politecnica delle Marche, Italy Filipa Grosso, Universidade do Porto, Portugal

#### \*Correspondence:

Francesco Iannelli francesco.iannelli@unisi.it

#### Specialty section:

This article was submitted to Evolutionary and Genomic Microbiology, a section of the journal Frontiers in Microbiology

> Received: 10 April 2018 Accepted: 16 July 2018 Published: 31 July 2018

#### Citation:

Santoro F, Romeo A, Pozzi G and Iannelli F (2018) Excision and Circularization of Integrative Conjugative Element Tn5253 of Streptococcus pneumoniae. Front. Microbiol. 9:1779. doi: 10.3389/fmicb.2018.01779

**Abbreviations:** CI: circular intermediate; CDS: coding sequence; CT: conjugative transposon; HGT: horizontal gene transfer; ICE: integrative conjugative element; MGE: mobile genetic element.

site specific recombinases (serine or tyrosine) or by DDE transposases (Ambroset et al., 2016). The Tn916 element carries the int and xis genes which code for a tyrosine site specific recombinase and an excisionase, respectively (Lu and Churchward, 1995). Excision and circularization require both Xis and Int, whereas Int alone is sufficient for integration (Storrs et al., 1991). Dosage of Tn916 CIs demonstrated that their number correlates with conjugation frequency and is variable among different strains (Manganelli et al., 1995).

Tn5253 is a 64,528-bp composite ICE of Streptococcus pneumoniae which contains integrated two distinct genetic elements: Tn5251, belonging to the Tn916–Tn1545 family of ICEs, and cat(pC194) which carry tet(M) and cat resistance genes, respectively (Ayoubi et al., 1991; Provvedi et al., 1996; Santoro et al., 2010; Iannelli et al., 2014). Tn5253 contains two pairs of xis/int recombinase genes one of which is part of Tn5251 (Kiliç et al., 1994; Iannelli et al., 2014). Genomic sequence analysis and PCR genotyping studies demonstrated that Tn5253-like elements are very common in multidrug-resistant pneumococcal strains including pandemic isolates (Croucher et al., 2009; Henderson-Begg et al., 2009; Mingoia et al., 2011). A study on 240 different pneumococcal isolates of the multidrugresistance 23F Spanish strain lineage, carrying the Tn5253-like element ICESpn23FST81, showed that the element is maintained among all derivative strains (Croucher et al., 2011). In this work, we investigated excision and circularization of the composite ICE Tn5253, including the respective contribution of each xis/int recombinase pair to the conjugal transfer of the genetic element.

## MATERIALS AND METHODS

TABLE 1 | Streptococcus pneumoniae strains.

### Bacterial Strains, Growth, and Mating Conditions

The bacterial strains used in this study and their relevant properties are described in **Table 1**. Bacterial growth and plate mating conjugation experiments were performed as reported (Santoro et al., 2010).

## Pneumococcal Lysate Preparation

Pneumococcal cultures (1 ml) were harvested in exponential phase (OD<sup>590</sup> about 0.2, roughly corresponding to 5 × 10<sup>8</sup> CFU/ml) and centrifuged at 11,000 × g for 2 min. Bacterial pellets were resuspended in 30 µl of lysis solution (DOC 0.1%, SDS 0.008%) and incubated at 37◦C until clarification (about 10 min). Two hundred and seventy micro liters of TE 1×, pH 8.0 were then added to the lysate.

### PCR, Sequencing, and Sequence Analysis

PCR and direct PCR sequencing were carried out following an already described protocol (Iannelli et al., 1998; Santoro et al., 2010) and DNA sequence analysis was obtained with standard softwares. DNA sequence alignments were performed using Clustal Omega<sup>1</sup> and Lalign<sup>2</sup> . Oligonucleotide primers and their characteristics are reported in **Table 2**.

#### PCR Mutagenesis

Isogenic deletion mutant strains were constructed transforming FR24 with linear PCR mutagenic constructs assembled by gene splicing by overlap extension as already described (Pearce et al., 2002; Iannelli and Pozzi, 2004). Deletion of Tn5251 int and xis CDS (orf21 and orf22 of Tn5253) was obtained with a mutagenic construct containing the ami/aad9 spectinomycin resistance cassette flanked at the left by a 496-bp DNA fragment and at the right by a 569-bp fragment corresponding to nucleotides 16,810–17,305 and 18,780–19,348 of Tn5253 (GenBank EU351020), respectively. The primer pair IF100/IF101 was used to amplify the spectinomycin-resistance cassette from plasmid pR412 (Bergé et al., 2002), whereas IF517/IF681 and IF520/IF682 were used to amplify the flanking fragments from FR24.

The xis and int CDSs of Tn5253 (orf78 and orf79) were deleted with a mutagenic construct containing the ami/aphIII kanamycin resistance cassette, flanked at the left by a 471-bp DNA fragment and at the right by a 603-bp corresponding

<sup>1</sup>https://www.ebi.ac.uk/Tools/msa/clustalo/ <sup>2</sup>http://vega.igh.cnrs.fr/bin/lalign-guess.cgi


a str-41 indicates a point mutation conferring resistance to streptomycin, while nov-1 indicates a point mutation conferring resistance to novobiocin. Cm, chloramphenicol; Km, kanamycin; Nov, novobiocin; Sm, streptomycin; Tc, tetracycline; Spe, spectinomycin; R, resistance.

FR51 Tn5253; cat, tet(M), 1xis-int of Tn5253; CmR, TcR, SmR; KmR; recombinases (orf78-orf79) deletion mutant of FR24 This study FR82 Tn5253; cat, tet(M), 1int-xis of Tn5251; CmR, TcR, SmR; SpeR; recombinases (orf21-orf22) deletion mutant of FR24 This study

#### TABLE 2 | Oligonucleotide primers.

fmicb-09-01779 July 29, 2018 Time: 15:52 # 3


<sup>a</sup>Nucleotides complementary to resistance cassettes primers are reported in italics.

to nucleotides 62,969–63,439 of Tn5253 and 65,260–65,862 of Tn5253, respectively. The primer pair IF149/IF210 was used to amplify the kanamycin-resistance cassette from plasmid pR410 (Bergé et al., 2002), while IF373/IF521 and IF522/IF523 were used to amplify the left and right fragments from FR24. Linear PCR constructs were used directly as donor DNA in transformation experiments. Mutant strains were selected for acquisition of spectinomycin or kanamycin resistance and the correct integration of constructs was confirmed by PCR and sequencing (Iannelli and Pozzi, 2004).

#### Real-Time PCR Quantification

Real-time PCR experiments were carried out with the KAPA SYBR FAST qPCR kit Master Mix Universal (2X) (Kapa Biosystems) on a LightCycler 1.5 apparatus (Roche). Real-time PCR mixture contained, in a final volume of 20 µl, 1× KAPA SYBR FAST qPCR reaction mix, 5 pmol of each primer and 1 µl of bacterial lysate as starting template. Thermal profile was an initial 3 min denaturation step at 95◦C followed by 40 cycles of repeated denaturation (0 s at 95◦C), annealing (20 s at 50◦C), and polymerisation (10 s at 72◦C). The temperature transition rate was 20◦C/s in the denaturation and annealing step and 5 ◦C/s in the polymerisation step. The primer pair IF327/IF328 amplified a 411 bp fragment used for CIs quantification, while IF496/IF356 amplified a 215 bp fragment used for free locus quantification, a 171 bp fragment of chromosomal gyrB gene obtained with primers IF138/IF139 was used to standardize results (**Table 2**). A standard curve for the gyrB gene was built plotting the threshold cycle against the number of chromosome copies using serial dilutions of chromosomal DNA with known concentration. This external standard curve was used to quantify in each sample the number of (i) chromosome copies, (ii) CIs, and (iii) reconstituted attB. Lower limit of detection of the assay was 10 copies/reaction. The quantification was corrected for the primer efficiency. Melting curve analysis was performed to differentiate the amplified products from primer dimers.

## RESULTS AND DISCUSSION

#### Excision of Tn5253 From the Pneumococcal Chromosome

PCR analysis of cell lysates of Tn5253-carrying pneumococcal strains showed evidence of precise excision of Tn5253 from its specific attachment site (attB) in the pneumococcal chromosome. This excision was investigated in liquid cultures of BM6001, the clinical isolate in which Tn5253 was originally found, and four other Tn5253-carrying laboratory strains all deriving from classic type 2 D39 (**Table 1**). Using divergent primers (IF327, IF328; **Table 2**) designed on the ends of the element, PCR analysis showed the presence of junctions between the left and right ends of Tn5253 (attTn), whereas with convergent primers (IF496, IF356; **Table 2**) designed on the regions flanking the insertion site it was possible to show the presence of chromosomes with reconstituted target sites for integration of Tn5253 (attB) (**Figure 1**). In all Tn5253-carrying pneumococci, DNA sequence analysis of PCR fragments indicated that: (i) attTn was 84 bp in size and was identical to attR, the 84-bp direct repeat present at the right end of the integrated element, whereas (ii) attB was 83 bp in size and was identical to attL, the direct repeat present

at the left end of the integrated element (**Figure 2**). The attRattTn repeat contained 4 nucleotide changes and 1 insertion compared to attL-attB (**Figure 2**). These results suggest that in Tn5253-carrying strains, recombination occurs between the two imperfect direct repeats attL and attR leading to precise excision of the element from the chromosome, with production of circular forms of Tn5253 in which the ends are joined by attR/attTn, while the attL/attB repeat remains in the bacterial chromosome (**Figure 1**).

#### Quantification of Circular Forms and Reconstituted attB Sites

To obtain a quantitative estimate of Tn5253 excision from the S. pneumoniae chromosome, Real-time PCR was used to quantify concentration of circular forms and reconstituted attB sites in liquid bacterial cultures. Different Tn5253-carrying laboratory strains of D39 ancestry, and BM6001 showed very homogeneous quantitative results (**Table 3**). In the laboratory strain DP1322 circular forms of Tn5253 were present at a concentration of 5.1 × 10−<sup>4</sup> (±2.7 × 10−<sup>4</sup> ) copies per chromosome, whereas reconstituted attB sites were at 2.1 × 10−<sup>4</sup> (±3.0 × 10−<sup>5</sup> ) copies per chromosome. These values were comparable to those obtained in the Tn5253-carrying laboratory strains (**Table 3**). Autonomous plasmid-like replication is common in ICEs and contributes to the stability and maintenance of these elements (Lee et al., 2010; Carraro et al., 2015; Johnson and Grossman, 2015). The hypothesis that also Tn5253 circular forms undergo few cycles of autonomous replication can explain why the copy


TABLE 3 | Real-time PCR quantification of Tn5253 circular form and reconstituted attB a .

<sup>a</sup>Concentration was expressed as copies per chromosome. <sup>b</sup>Frequency refers to mating experiments where S. pneumoniae FP11 was the recipient.

number of circular forms is higher than the copy number of the reconstituted attB site.

### Site-Specific Integration of Tn5253 Into the rbgA Gene

In S. pneumoniae D39 and in its derivatives used as conjugation recipients (**Table 1**) DNA sequence analysis showed that attB of Tn5253 was 83 bp in size, and was always located within the rbgA gene (nucleotides 1,043,779 to 1,043,861, GenBank CP000410) (**Figure 1**). The ribosomal biogenesis GTPase A encoded by rbgA is a conserved, essential bacterial protein involved in the 50S ribosome subunit assembly (Uicker et al., 2006). The attB site contained the first 20 nucleotides of the rbgA CDS together with the 63 nucleotides upstream of the start codon (**Figures 1**, **2**). The junction fragments of Tn5253 with the bacterial chromosome were investigated in a total of 12 transconjugants obtained in independent matings in which Tn5253 was transferred by conjugation from 2 pneumococcal donors (BM6001 and DP1322; **Table 1**) to 3 pneumococcal recipients (FP58, FP10, FP11; **Table 1**). Left and right junction fragments were amplified by PCR (using primer pairs IF496/IF327 and IF328/IF356) and sequenced in all transconjugants. DNA sequence analysis showed that in all cases Tn5253 integration occurred at the same site within rbgA, between the two direct repeats: (i) attL, corresponding to the attB of the recipient and (ii) attR, corresponding to the attTn of the circular forms of Tn5253 (**Figure 2**). These results indicated that attR, one of the two repeats flanking Tn5253 in the donor chromosome, was always transferred by conjugation to the recipients. Since the integrated form of Tn5253 was invariably flanked by attB at the left end (attL) and attTn at the right end (attR), we hypothesize a polarization in the DNA integration process. Site specific integration of MGEs often occurs at one end of essential and highly conserved genes, such as the 3<sup>0</sup> end of tRNA genes and the 3<sup>0</sup> or 5<sup>0</sup> end of genes coding for ribosomal proteins (Ambroset et al., 2016). Also for Tn5253 integration occurs at the 5<sup>0</sup> end of an essential gene, with target site duplication allowing restoration of an intact CDS. The use of essential and conserved genes as target sites guarantees the presence and conservation of attB in bacterial genomes favoring the spread of ICEs such as Tn5253, which can overpass the border of a single species and thus favor the dissemination of multiple antibiotic resistance genes. In fact, in other bacterial species such as Streptococcus pyogenes and Streptococcus mitis, Tn5253-like elements are found integrated at the 5<sup>0</sup> end of rbgA orthologous genes (Mingoia et al., 2014; Petrosyan et al., 2016).

#### Recombinase Genes Involved in Tn5253 Excision

Two sets of xis/int recombinase genes are present in the sequence of Tn5253, one set (orf78/orf79) is at the right end of the element,

transcription direction are indicated as arrows, sequences corresponding to Tn5251 and cat(pC194) are indicated by solid bars.

while the other (orf21/orf22) belongs to Tn5251 (**Figure 3**). Excisionase Xis and tyrosine integrase Int are known to work in synergy, for this reason we decided to construct mutants where the xis and int genes were both deleted. For each set of xis/int recombinase genes, we constructed an isogenic deletion mutant in the Tn5253-carrying strain FR24. In FR51 a 1,820-bp DNA fragment (position 63,440–65,259, GenBank No. EU351020) encompassing orf78/orf79 CDSs was deleted and replaced with the 876-bp ami/aphIII cassette. In FR82 a 1,474-bp DNA fragment (position 17,306–18,779, GenBank No. EU351020) encompassing orf21/orf22 CDS was deleted and replaced with the 894-bp ami/aad9 cassette (**Table 2**). The deletion of xis/int of Tn5251 abolished the production of circular forms and the conjugal transfer of Tn5251, but did not affect the frequencies of Tn5253 circular forms, of attB site reconstitution, and of Tn5253 conjugal transfer (data not shown). Deletion of xis/int of Tn5253 in FR51 abolished the circular forms generation and the reconstitution of attB site (<7.1 × 10−<sup>6</sup> copies per chromosome for both genetic structures, **Table 3**). The absence of circular forms in FR51 was associated to the lack of Tn5253 conjugal transfer suggesting, that the circular form of Tn5253 acts as a conjugation intermediate as proposed for other characterized ICEs including Tn916. Data obtained using xis/int deletion mutants showed that the two recombinase pairs act independently and do not complement each other. This finding suggests that the association between the two elements is physical but not functional.

## CONCLUSION

In this work we have shown that: (i) Tn5253 is capable of precise excision from the chromosome, producing circular forms of the

#### REFERENCES


element, and leaving chromosomes with reconstituted attB sites; (ii) in the circular forms, the two ends of Tn5253 are joined by attTn, an 84-bp DNA fragment identical to the attR junction fragment flanking the element in its integrated form; (iii) attR is always transferred to the recipient strain during conjugation; (iv) production of Tn5253 circular forms and their conjugal transfer were abolished when xis/int of Tn5253 were deleted. Even if the importance of ICEs in shaping bacterial genomes is widely recognized and nucleotide sequences of ICEs are increasingly available, a functional characterization is available only for a few of these genetic elements. This work on Tn5253 contributes to elucidating the transfer functions of one of the prototypes of ICEs of gram positive bacteria.

#### AUTHOR CONTRIBUTIONS

FS, FI, and GP designed the experiments. FS and AR performed the experimental work. All authors analyzed and interpreted the data. FI, FS, and GP wrote the paper.

## FUNDING

The research leading to these results has received funding from the European Union Seventh Framework Programme (FP7/2007-2013) under grant agreement no. 241446 (project ANTIRESDEV).

#### ACKNOWLEDGMENTS

We are grateful to Marc Prudhomme and Jean-Pierre Claverys for kindly providing us plasmids pr410 and pr412.

multidrug-resistant pandemic clone Streptococcus pneumoniae Spain23F ST81. J. Bacteriol. 191, 1480–1489. doi: 10.1128/JB.01343-08



mitis strain SVGS\_061 isolated from a neutropenic patient with viridans group streptococcal shock syndrome. Genome Announc 4:e00259-16. doi: 10.1128/ genomeA.00259-16


**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 © 2018 Santoro, Romeo, Pozzi and Iannelli. 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) and the copyright owner(s) 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 Versatile Mutational Resistome of Pseudomonas aeruginosa

#### Carla López-Causapé, Gabriel Cabot, Ester del Barrio-Tofiño and Antonio Oliver\*

Servicio de Microbiología y Unidad de Investigación, Hospital Universitari Son Espases, Institut d'Investigació Sanitaria Illes Balears, Palma de Mallorca, Spain

One of the most striking features of Pseudomonas aeruginosa is its outstanding capacity for developing antimicrobial resistance to nearly all available antipseudomonal agents through the selection of chromosomal mutations, leading to the failure of the treatment of severe hospital-acquired or chronic infections. Recent whole-genome sequencing (WGS) data obtained from in vitro assays on the evolution of antibiotic resistance, in vivo monitoring of antimicrobial resistance development, analysis of sequential cystic fibrosis isolates, and characterization of widespread epidemic high-risk clones have provided new insights into the evolutionary dynamics and mechanisms of P. aeruginosa antibiotic resistance, thus motivating this review. Indeed, the analysis of the WGS mutational resistome has proven to be useful for understanding the evolutionary dynamics of classical resistance pathways and to describe new mechanisms for the majority of antipseudomonal classes, including β-lactams, aminoglycosides, fluoroquinolones, or polymixins. Beyond addressing a relevant scientific question, the analysis of the P. aeruginosa mutational resistome is expected to be useful, together with the analysis of the horizontally-acquired resistance determinants, for establishing the antibiotic resistance genotype, which should correlate with the antibiotic resistance phenotype and as such, it should be useful for the design of therapeutic strategies and for monitoring the efficacy of administered antibiotic treatments. However, further experimental research and new bioinformatics tools are still needed to overcome the interpretation limitations imposed by the complex interactions (including those leading to collateral resistance or susceptibility) between the 100s of genes involved in the mutational resistome, as well as the frequent difficulties for differentiating relevant mutations from simple natural polymorphisms.

Keywords: antibiotic resistance, resistome, Pseudomonas aeruginosa, multidrug resistance, evolution, resistance development, mutation

## INTRODUCTION

Pseudomonas aeruginosa is one of the most frequent and severe causes of hospital-acquired infections, particularly affecting immunocompromised (especially neutropenic) and Intensive Care Unit (ICU) patients. Indeed, P. aeruginosa is the first cause of ventilator associated pneumonia (VAP) and burn wound infections, both associated with a very high mortality rate (Vincent, 2003; Bassetti et al., 2012). Likewise, P. aeruginosa is the most frequent driver of chronic respiratory infections in cystic fibrosis (CF) patients or other chronic underlying diseases (Döring et al., 2011).

Edited by:

Gian Maria Rossolini, Università degli Studi di Firenze, Italy

#### Reviewed by:

Daniel Pletzer, The University of British Columbia, Canada Max Maurin, Université Grenoble Alpes, France

> \*Correspondence: Antonio Oliver antonio.oliver@ssib.es

#### Specialty section:

This article was submitted to Evolutionary and Genomic Microbiology, a section of the journal Frontiers in Microbiology

Received: 24 January 2018 Accepted: 23 March 2018 Published: 06 April 2018

#### Citation:

López-Causapé C, Cabot G, del Barrio-Tofiño E and Oliver A (2018) The Versatile Mutational Resistome of Pseudomonas aeruginosa. Front. Microbiol. 9:685. doi: 10.3389/fmicb.2018.00685

One of the most striking features of P. aeruginosa is its outstanding capacity for developing antimicrobial resistance to nearly all available antipseudomonal agents through the selection of chromosomal mutations. Indeed, treatment failure caused by the development of antimicrobial resistance is a too frequent outcome of P. aeruginosa infections. The problem of mutationmediated antibiotic resistance is further amplified in the chronic infection setting, due to the very high prevalence of hypermutable strains, showing greatly enhanced spontaneous mutation rates caused by defective DNA repair or error avoidance systems (Oliver et al., 2000; Maciá et al., 2005).

Beyond the obvious negative impact of resistance development for the treated patient, the accumulation of several of these chromosomal mutations leads to the emergence of multidrug resistant (MDR), extensively drug-resistant (XDR) or even pan-antibiotic-resistant (PDR) strains, which can be responsible for notable epidemics in the hospital setting (Deplano et al., 2005; Suarez et al., 2011). Moreover, recent studies have evidenced the existence of MDR/XDR global clones disseminated in different hospitals worldwide, and for that reason they have been denominated high-risk clones (Woodford et al., 2011; Oliver et al., 2015). Although high-risk clones are frequently associated with transferable antimicrobial resistance determinants, they also typically show a wide range of chromosomal mutations playing a major role in the resistance phenotype (Oliver et al., 2015). Likewise, recent reports have evidenced the interpatient spread of antimicrobial resistance mutations linked to the transmission of epidemic CF strains (López-Causapé et al., 2017).

Along with growing information from mechanistic studies on chromosomal resistance mechanisms and their complex regulatory pathways, involved in adaptive resistance (Lister et al., 2009; Muller et al., 2011; Skiada et al., 2011; Juan et al., 2017), the introduction of whole-genome sequencing (WGS) approaches is shaping up a new dimension for the mutational resistance landscape. The term resistome was first used to account for the set of primary antibiotic resistance genes that could be eventually transferred to human pathogens (D'Costa et al., 2006). Soon after the concept of intrinsic resistome was introduced to include all chromosomal genes that are involved in intrinsic resistance, and whose presence in strains of a bacterial species is independent of previous antibiotic exposure and is not due to horizontal gene transfer (HGT) (Fajardo et al., 2008). Finally, the term mutational resistome was more recently implemented to account for the set of mutations involved in the modulation of antibiotic resistance levels in the absence of HGT (Cabot et al., 2016b; López-Causapé et al., 2017). Recent WGS data obtained from in vitro assays on the evolution of antibiotic resistance, in vivo monitoring of antimicrobial resistance development, analysis of sequential CF isolates, and characterization of wide spread epidemic high-risk clones provide new insights into the evolutionary dynamics and mechanisms of P. aeruginosa antibiotic resistance (Cabot et al., 2016a; Feng et al., 2016; Del Barrio-Tofiño et al., 2017; Jaillard et al., 2017; López-Causapé et al., 2017). Indeed, the analysis of WGS mutational resistomes has proven to be useful for understanding the evolutionary dynamics of classical resistance mechanisms and to depict new ones for the majority of antimicrobial classes, including β-lactams, aminoglycosides, fluoroquinolones, polymixins and others, as reviewed in the following sections. **Table 1** summarizes the main genes and mutations known to increase resistance levels and therefore shaping up the P. aeruginosa mutational resistome.

### β-LACTAM MUTATIONAL RESISTOME

The most frequent mutation-driven β-lactam resistance mechanism is likely the overproduction of the chromosomal cephalosporinase AmpC, involving a wide range of genes belonging to complex regulatory pathways of cell-wall recycling (Juan et al., 2017). Among them, the mutational inactivation of dacB, encoding the non-essential penicillin-binding protein (PBP) PBP4, and ampD, encoding a N-acetyl-muramyl-L-alanine amidase have been found to be the most frequent cause of ampC derepression and β-lactam resistance (Juan et al., 2005; Moya et al., 2009). The inactivation of PBP4 has also been shown to activate the BlrAB/CreBC regulatory system, further increasing resistance levels (Moya et al., 2009). Additionally, specific point mutations leading to a conformation change in the transcriptional regulator AmpR, causing ampC upregulation and β-lactam resistance, have been noted among clinical strains. They include the D135N mutation, described in several species besides P. aeruginosa, including Stenotrophomonas maltophilia, Citrobacter freundii, or Enterobacter cloacae (Juan et al., 2017) or the R154H mutation, linked to the widespread MDR/XDR ST175 high-risk clone. Mutation of many other genes, including those encoding other amidases (AmpDh2 and AmpDh3), other PBPs (such as PBP5 and PBP7), lytic transglycosylases (such as SltB1 and mltB), MPL (UDP-N-acetylmuramate:Lalanyl-γ-D-glutamyl-meso-diaminopimelate ligase), or NuoN (NADH dehydrogenase I chain N) have been shown to enhance ampC expression, either alone or combined with other mutations, although their impact on β-lactam resistance among clinical strains still needs to be further analyzed (Juan et al., 2017).

In addition to ampC overexpression, recent studies have revealed that β-lactam resistance development, including the novel combinations of β-lactam-β-lactamase inhibitors ceftolozane/tazobactam and ceftazidime/avibactam, may result from mutations leading to the structural modification of AmpC (Cabot et al., 2014; Lahiri et al., 2014; Fraile-Ribot et al., 2017a; Haidar et al., 2017). Likewise, recent studies identified diverse AmpC variants associated with high-level cephalosporin resistance, including ceftolozane/tazobactam and ceftazidime/avibactam, in a small proportion (around 1%) of clinical P. aeruginosa isolates (Berrazeg et al., 2015). Over 200 Pseudomonas Derived Cephalosporinase (PDC) variants have been described so far, including those associated with enhanced ceftolozane/tazobactam and ceftazidime/avibactam resistance (**Table 1**). An update database of PDC variants is maintained in our laboratory and is freely available at https://arpbigidisba.com. Additionally, resistance development to ceftolozane/tazobactam and/or ceftazidime/avibactam may involve mutations leading to the structural modification of narrow spectrum OXA-2 and OXA-10 acquired oxacillinases (Fraile-Ribot et al., 2017a,b).

#### TABLE 1 | Main genes and mutations known to be involved in increased P. aeruginosa antibiotic resistance.


(Continued)

#### TABLE 1 | Continued


<sup>a</sup>FQs, fluoroquinolones; COL, colistin; AMGs, aminoglycosides; CAZ, ceftazidime; CEF, cefepime; PPT, piperacillin-tazobactam; IMP, imipenem; MER, meropenem; CAZ/AVI, ceftazidime/avibactam; C/T, ceftolozane/tazobactam; FOS, fosfomycin; RIF, rifampicin.

Besides β-lactamases, there is increasing evidence on the role of target modification in P. aeruginosa β-lactam resistance. Particularly noteworthy are the mutations in ftsI, encoding PBP3, an essential high molecular class B PBP with transpeptidase activity (Chen et al., 2016). Indeed, data from CF patients (Diaz Caballero et al., 2015; López-Causapé et al., 2017), epidemic high-risk clones (Cabot et al., 2016b; Del Barrio-Tofiño et al., 2017) as well as from in vitro studies (Cabot et al., 2016a) have recently shown that PBP3 is under strong mutational pressure, and that specific mutations in this PBP contribute to β-lactam resistance development. R504C/H and F533L mutations are likely those most commonly reported, and are located within the protein domains implicated in the formation and stabilization of the inactivating complex β-lactam-PBP3 (Han et al., 2010). Moreover, these specific mutations have been documented to emerge in vivo during chronic respiratory infection in CF patients (Diaz Caballero et al., 2015; López-Causapé et al., 2017) and upon meropenem (Cabot et al., 2016a) and aztreonam (Jorth et al., 2017) exposure in vitro. However, the precise contribution of PBP3 mutations to β-lactam resistance phenotypes needs to be further investigated using isogenic strains. Likewise, despite unique polymorphisms have also been detected in some clinical strains for other PBPs, their role in β-lactam resistance, if any, still needs to be experimentally addressed.

Other relevant components of the β-lactam mutational resistome are the porins and RND efflux pumps. Mutational inactivation of OprD is well-known to be the primary carbapenem resistance mechanisms in P. aeruginosa (Lister et al., 2009; Castanheira et al., 2014). OprD inactivation typically results from indels or nonsense mutations, including the Q142X mutation, characteristic of the widespread ST175 high-risk clone (Cabot et al., 2016b). Additionally, some amino acid substitutions have also been recently associated with OprD-driven resistance, particularly in the CF setting (Richardot et al., 2015). Finally, carbapenem resistance may also result from oprD repression caused by mutations in the MexEF-OprN efflux pump regulators

(mexS/T) or the ParRS two-component system (Li et al., 2015). Overexpression of MexAB-OprM, caused by mutation of several genes involved in its regulation (mexR, nalC, or nalD) increases MICs of most β-lactams except imipenem, whereas overexpression of MexXY (mexZ, parSR, amgS mutations) is particularly involved in cefepime resistance (Li et al., 2015). Additionally, sequence variations in unique residues are detected in the genes coding for the efflux pump (Del Barrio-Tofiño et al., 2017; López-Causapé et al., 2017); however, their contribution to resistance profiles, if any, still needs to be further explored.

Finally, another potentially relevant mutational β-lactam resistance mechanism is the selection of large (>200 kb) deletions affecting specific parts of the chromosome (Cabot et al., 2016a). Although the basis of the conferred resistance phenotype still needs to be further clarified, these mutants can be recognized by the characteristic brown pigment (pyomelanine) caused by the deletion of one of the affected genes, hmgA, coding for a homogentisate-1,2- dioxygenase. This type of deletion has been documented in both, in vitro evolved β-lactam-resistant mutants and CF isolates (Cabot et al., 2016a; Hocquet et al., 2016). However, the deletion of hmgA is not responsible for the resistance phenotype, which could be linked to the deletion of another of the affected genes, galU. This gene codes for a UDP-glucose pyrophosphorylase involved in the synthesis of the lipopolysaccharide (LPS) core. Indeed, analysis of transposon mutant libraries has shown that inactivation of galU increases ceftazidime and meropenem MICs (Dötsch et al., 2009; Alvarez-Ortega et al., 2010).

#### AMINOGLYCOSIDE MUTATIONAL RESISTOME

In the absence of horizontally-acquired aminoglycoside modifying enzymes, resistance to this antibiotic class has been particularly linked to the mutational overexpression of MexXY-OprM. Indeed, mutational overexpression of this pump, mainly caused by mexZ, amgS, or parRS mutations (**Table 1**), is very frequent among clinical isolates, from both, CF patients and nosocomial infections (Guénard et al., 2014; Prickett et al., 2017). Moreover, recent studies show that the epidemic high-risk clone ST175 overexpresses MexXY due to a specific mutation in mexZ (G195E) (Cabot et al., 2016b). However, recent studies have revealed that the aminoglycoside mutational resistome extends far beyond MexXY overexpression, and that high-level resistance may result from the accumulation of multiple mutations, and the involvement of several novel resistance determinants has been recently documented (El'Garch et al., 2007; Schurek et al., 2008; Feng et al., 2016). Among them is particularly noteworthy fusA1, coding for the elongation factor G. Indeed, specific FusA1 mutations have been associated with aminoglycoside resistance in vitro (Feng et al., 2016; López-Causapé et al., 2018) and among clinical, particularly CF, strains (Chung et al., 2012; Markussen et al., 2014; Greipel et al., 2016; López-Causapé et al., 2017, 2018). Moreover, the implication of fusA1 mutations in aminoglycoside resistance has been recently confirmed through site-directed mutagenesis (Bolard et al., 2017).

## FLUOROQUINOLONE MUTATIONAL RESISTOME

The fluoroquinolone mutational resistome generally includes specific missense mutations in DNA gyrase (gyrA and/or gyrB) and topisomerase IV (parC and/or parE) Quinolone Resistance-Determining Regions (QRDRs) (**Table 1**) (Bruchmann et al., 2013; Kos et al., 2015). High-level fluoroquinolone resistance in P. aeruginosa high-risk clones is nearly universal, and typically involves combinations of mutations in GyrA-T83 and ParC-S87 (Del Barrio-Tofiño et al., 2017). QRDR mutations involved in fluoroquinolone resistance in CF might be more variable (López-Causapé et al., 2017). It is also well-known that the mutational overexpression of efflux pumps modulate fluoroquinolone resistance (**Table 1**). While the overexpression of MexAB-OprM and MexXY-OprM is globally more frequent among clinical strains, its contribution to clinical fluoroquinolone resistance is likely more modest (Bruchmann et al., 2013). On the other hand, the mutational overexpression of MexEF-OprN or MexCD-OprJ is associated with high-level (clinical) fluoroquinolone resistance, and although their prevalence is considered low except in the CF chronic infection setting, recent data show that it might be higher than expected (Richardot et al., 2015).

#### POLYMIXIN MUTATIONAL RESISTOME

Whereas the prevalence of polymyxin (colistin and polymyxin B) resistance is still globally low (<5%), it has increased in the last years because of the frequent use of these last-resource antibiotics for the treatment of MDR/XDR nosocomial and CF strains. Polymyxin resistance results most frequently from the modification of the LPS caused by the addition of a 4-amino-4-deoxy-L-arabinose moiety in the lipid A structure (Olaitan et al., 2014; Jeannot et al., 2017). The involved mutations are frequently located in the PmrAB or PhoPQ two-component regulators, which lead to the activation of the arnBCADTEF operon (Barrow and Kwon, 2009). More recent studies have revealed that mutations in the ParRS twocomponent regulator, not only produce polymyxin resistance due to the activation of the arnBCADTEF operon, but also lead to a MDR phenotype determined by the overprexpression of MexXY and the repression of OprD (Muller et al., 2011). Moreover, two additional two-component regulators, ColRS and CprRS, have been recently found to be involved in polymyxin resistance (Gutu et al., 2013). The analysis of colistin resistance mechanisms among clinical strains is not always straight forward, since the presence of mutations in these two-component regulators is not always associated with clinical colistin resistance, probably denoting partial complementation between the different regulators (Moskowitz et al., 2012; Gutu et al., 2013; López-Causapé et al., 2017). Moreover, recent in vitro evolution assays have revealed the implication of additional mutations in high level colistin resistance, facilitated by the emergence of mutator (mutS deficient) phenotypes (Dößelmann et al., 2017). Particularly noteworthy among them are those occurring in LptD, LpxC, or MigA.

## OTHER ANTIBIOTICS

fmicb-09-00685 April 4, 2018 Time: 16:14 # 6

Even if not considered a classical antipseudomonal agent, fosfomycin has emerged in the last decade as a potentially useful antibiotic in urinary tract infections and combined therapy for MDR/XDR P. aeruginosa (Michalopoulos et al., 2011). However, fosfomycin resistance spontaneous mutation rates are high and the mechanism involved is typically the mutational inactivation of glpT, coding for a glycerol-3-phospate permease required for fosfomycin uptake (Castañeda-García et al., 2009; Rodríguez-Rojas et al., 2010). glpT mutations, conferring high-level fosfomycin resistance are also frequently found among MDR/XDR high-risk clones (Del Barrio-Tofiño et al., 2017), and some specific mutations, such as T211P, have been fixed in some widespread lineages as described for ST175 (Cabot et al., 2016b). Another potentially useful antimicrobial for combined therapy against MDR/XDR P. aeruginosa is rifampicin (Cai et al., 2017). However, rifampicin resistance may emerge at high frequency due to the selection of specific missense mutations in rpoB, coding for the beta subunit of the RNA polymerase (Jatsenko et al., 2010). Another example of newer antibiotic families with antipseudomonal activity are the pacidamycins, uridyl peptide antibiotics, targeting translocase I, an essential enzyme in peptidoglycan biosynthesis (Mistry et al., 2013). Emergence of high-level resistance to this antibiotic class has been shown to involve the selection of mutations in the Opp transporter, a binding proteindependent ABC transporter used for oligopeptide import. Finally, the P. aeruginosa mutational resistome, particularly in the CF setting, may also include resistance to other used antibiotics such as the frequent mutations of domain V of 23S rRNA, conferring macrolide resistance (Mustafa et al., 2017).

#### CONCLUDING REMARKS AND FUTURE PERSPECTIVES

The analysis of the P. aeruginosa mutational resistome, together with the analysis of the horizontally-acquired resistance determinants, should be useful for establishing the antibiotic resistance genotype, which should correlate with the antibiotic resistance phenotype and as such, it should permit the design of therapeutic strategies and for monitoring the efficacy of administered antibiotic treatments. However, the current applicability of the analysis of the mutational resistome is still limited by the large number of genes involved and the complexity of the resistance phenotypes generated, and, particularly, by the difficulties, in many cases, for differentiating relevant mutations from simple natural polymorphisms. Obviously, the evolution of the mutational resistome is a direct consequence of antimicrobial exposure and as such, it is not surprising that exposure to one antibiotic drives evolution of the mutational resistome for that antibiotic. However, the complexity of the actual resistance profile is further increased by the specificity and interactions among different resistance mechanisms. Indeed, a resistance mutation selected by one antibiotic may have a variable effect among the different agents within the same antibiotic class or family. Likewise, cross resistance (or collateral resistance) implies that exposure to one antibiotic drives also the development of resistance to a different one from the same or other classes. Typically, this is caused by the developed resistance mechanism (such as efflux pump overexpression) affecting simultaneously different antibiotics. Indeed, potential development of cross resistance is a major issue to consider when using antibiotic combinations (Vestergaard et al., 2016). Moreover, cross resistance between antibiotics and antiseptics and other biocides may also occur (Li et al., 2015). Perhaps less obvious is collateral susceptibility, which implies that exposure to one antibiotic increases the susceptibility to a different one (Pál et al., 2015; Imamovic et al., 2017). This might be achieved through two mechanisms. One possible mechanism is that exposure to one antibiotic directly causes increased susceptibility to a different one, for example, mutations in the β-lactamase AmpC increases cephalosporin hydrolysis while reducing that of penicillins or carbapenems (Cabot et al., 2014). The second possibility is that the development of a resistance mechanism impairs the activity of another existing resistance mechanism. An example is competition between the different efflux pumps, since the overexpression of one may impair the expression of another (Mulet et al., 2011). Thus, the evolution of the mutational resistome for a given antibiotic is not only dependent on the exposure to this antibiotic, but it is also conditioned by the simultaneous or even previous exposures to other antibiotics. An illustrative example is provided in a recent in vitro study that demonstrated, for a broad range of antibiotic classes, that the history of exposure and resistance development to a given antibiotic, conditions the dynamics and mechanisms of resistance development when exposed to a second one (Yen and Papin, 2017). In summary, the comprehensive analysis of the mutational resistome of P. aeruginosa in CF and nosocomial infections is expected to become a useful tool for optimizing therapeutic strategies and monitoring the efficacy of administered antibiotic treatments in the near future.

## AUTHOR CONTRIBUTIONS

CL-C and AO wrote the manuscript. CL-C, GC, EdB-T, and AO reviewed the literature.

## FUNDING

The authors were supported by Plan Nacional de I+D+i 2013– 2016 and Instituto de Salud Carlos III, Subdirección General de Redes y Centros de Investigación Cooperativa, Ministerio de Economía, Industria y Competitividad, Spanish Network for Research in Infectious Diseases (REIPI RD16/0016) and grant PI15/00088 (PI AO) and co-financed by European Development Regional Fund ERDF "A way to achieve Europe," Operative program Intelligent Growth 2014–2020. AO was also supported by the European Union through the 11th Call of the Innovative Medicines Initiative (grant COMBACTE-MAGNET).

## REFERENCES

fmicb-09-00685 April 4, 2018 Time: 16:14 # 7



the cystic fibrosis setting. Antimicrob. Agents Chemother. doi: 10.1128/AAC. 02583-17 [Epub ahead of print].


in Pseudomonas aeruginosa. Antimicrob. Agents Chemother. 52, 4213–4219. doi: 10.1128/AAC.00507-08


resistance. FEMS Microbiol. Rev. 35, 736–755. doi: 10.1111/j.1574-6976.2011. 00268.x

Yen, P., and Papin, J. A. (2017). History of antibiotic adaptation influences microbial evolutionary dynamics during subsequent treatment. PLoS Biol. 15:e2001586. doi: 10.1371/journal.pbio.2001586

**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 © 2018 López-Causapé, Cabot, del Barrio-Tofiño and Oliver. 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) and the copyright owner 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.

# Genome Sequencing and Comparative Analysis of Stenotrophomonas acidaminiphila Reveal Evolutionary Insights Into Sulfamethoxazole Resistance

Yao-Ting Huang<sup>1</sup> , Jia-Min Chen<sup>1</sup> , Bing-Ching Ho<sup>2</sup> , Zong-Yen Wu3,4, Rita C. Kuo<sup>3</sup> and Po-Yu Liu5,6,7 \*

<sup>1</sup> Department of Computer Science and Information Engineering, National Chung Cheng University, Chiayi, Taiwan, <sup>2</sup> Department of Clinical Laboratory Sciences and Medical Biotechnology, National Taiwan University Hospital, Taipei, Taiwan, <sup>3</sup> DOE Joint Genome Institute, Walnut Creek, CA, United States, <sup>4</sup> Department of Veterinary Medicine, National Chung Hsing University, Taichung, Taiwan, <sup>5</sup> The Department of Nursing, Shu-Zen Junior College of Medicine and Management, Kaohsiung, Taiwan, <sup>6</sup> Rong Hsing Research Center for Translational Medicine, College of Life Sciences, National Chung Hsing University, Taichung, Taiwan, <sup>7</sup> Division of Infectious Diseases, Department of Internal Medicine, Taichung Veterans General Hospital, Taichung, Taiwan

#### Edited by:

Silvia Buroni, University of Pavia, Italy

#### Reviewed by:

Prabhu B. Patil, Institute of Microbial Technology (CSIR), India Haitham Sghaier, Centre National des Sciences et Technologies Nucléaires, Tunisia

\*Correspondence:

Po-Yu Liu pyliu@vghtc.gov.tw; liupoyu@gmail.com

#### Specialty section:

This article was submitted to Evolutionary and Genomic Microbiology, a section of the journal Frontiers in Microbiology

Received: 28 February 2018 Accepted: 30 April 2018 Published: 17 May 2018

#### Citation:

Huang Y-T, Chen J-M, Ho B-C, Wu Z-Y, Kuo RC and Liu P-Y (2018) Genome Sequencing and Comparative Analysis of Stenotrophomonas acidaminiphila Reveal Evolutionary Insights Into Sulfamethoxazole Resistance. Front. Microbiol. 9:1013. doi: 10.3389/fmicb.2018.01013 Stenotrophomonas acidaminiphila is an aerobic, glucose non-fermentative, Gramnegative bacterium that been isolated from various environmental sources, particularly aquatic ecosystems. Although resistance to multiple antimicrobial agents has been reported in S. acidaminiphila, the mechanisms are largely unknown. Here, for the first time, we report the complete genome and antimicrobial resistome analysis of a clinical isolate S. acidaminiphila SUNEO which is resistant to sulfamethoxazole. Comparative analysis among closely related strains identified common and strain-specific genes. In particular, comparison with a sulfamethoxazole-sensitive strain identified a mutation within the sulfonamide-binding site of folP in SUNEO, which may reduce the binding affinity of sulfamethoxazole. Selection pressure analysis indicated folP in SUNEO is under purifying selection, which may be owing to long-term administration of sulfonamide against Stenotrophomonas.

Keywords: genome sequencing, Stenotrophomonas acidaminiphila, sulfamethoxazole resistance, Stenotrophomonas, comparative genomics, dihydropteroate synthase

## INTRODUCTION

Bacteria within the genus Stenotrophomonas species are aerobic, glucose non-fermentative, Gram-negative bacilli which inhabit diverse marine and terrestrial environments. The genus Stenotrophomonas currently comprises of 14 species<sup>1</sup> . Stenotrophomonas acidaminiphila was identified in 2002 (Assih et al., 2002). Initially isolated from sewage sludge from wastewater treatment, it isolated mostly from aquatic environments. Reports of S. acidaminiphila human isolates are limited. To our best knowledge, no case of S. acidaminiphila infections has ever been reported to date. However, studies of environmental isolates revealed highly resistant to multiple antibiotics (Assih et al., 2002; Vinuesa and Ochoa-Sanchez, 2015).

<sup>1</sup>http://www.bacterio.net/stenotrophomonas.html

#### TABLE 1 | Overview of the S. acidaminiphila strains in this study.


The antimicrobial options for Stenotrophomonas infections are limited because of its inherent resistance to most antibiotics, where trimethoprim-sulfamethoxazole (trimethoprim and sulfonamide combination in a 1:5 ratio) has long been regarded as the agent of choice (Sanchez, 2015). The main component, sulfamethoxazole interrupts the biosynthesis of tetrahydrofolic acid in both bacteria and primitive eukaryotes by targeting the dihydropteroate synthase (DHPS) catalyses, which catalyzes the condensation of 6-hydroxymethyl-7,8-dihydropterin monophosphate (DHPP) with p-aminobenzoic acid (PABA) (Skold, 2000). However, the resistance to sulfamethoxazole is increasing and is mainly caused by single amino acid mutations in the chromosomal gene encoding DHPS or by the acquisition of sul genes encoding alternative drug-resistance variants of the DHP via mobile genetic elements (Toleman et al., 2007).

To date, only two environmental strains of S. acidaminiphila genomes have been sequenced, all of which were isolated from river sediments (Assih et al., 2002; Vinuesa and Ochoa-Sanchez, 2015). However, the genome, pathogenome, and antimicrobial resistome of clinical isolate can still differ a lot in comparison with those of environmental strains, owing to the adaptation to host immune system and antibiotic pressure. Therefore, a complete genome from clinical isolates is valuable for designing effective treatment strategies.

Here, we sequenced the genome of the S. acidaminiphila strain SUNEO, a first clinical isolate that possessed trimethoprimsulfamethoxazole resistance. We propose a scenario for the origin and evolution of S. acidaminiphila SUNEO, based on its genomic features. Gene annotation and comparative analysis further revealed a unique profile of folP mutation that could play a role in drug resistance.

#### MATERIALS AND METHODS

#### Bacterial Strain Isolation, Identification, and Antimicrobial Susceptibility Testing

Strain SUNEO was isolated from the bile of a cholangiocarcinoma patient with obstructive jaundice and cholangitis. The bile

cellular functions. The Circles display from the outermost to the innermost: (1) DNA coordinates. (2,3) Function-based color coded mapping of the CDSs predicted on the forward and reverse strands, respectively. (4,5) tRNA and rRNA genes, respectively. (6) GC plot with regions above and below average in green and violet, respectively. (window size: 10,000 bp). (7) GC skew showing regions above and below average in yellow and light blue, respectively. (window size: 10,000 bp).


sample was inoculated on trypticase soy agar with 5% sheep blood (Becton–Dickinson, Franklin Lakes, NJ, United States) and incubated aerobically at 37 ◦C overnight. The isolate was identified through 16S rRNA gene sequencing as previously described (Assih et al., 2002 ; Mangwani et al., 2014). Antibiotic susceptibility tests for the strain SUNEO was performed by the automated Vitek 2 system (bioMérieux, Inc., Durham, NC, United States) according to the manufacturer's instructions.

#### Library Preparation and Whole-Genome Sequencing

Overnight cultures were grown in Luria-Bertani broth overnight at 37 ◦C. Genomic DNA was extracted using DNeasy blood and tissue kit (Qiagen, Valencia, CA, United States) as per the manufacturer's instructions. High-molecular-weight gDNA was sheared to 10-kb lengths using g-TUBES (Covaris, Woburn, MA, United States). Sheared DNA was processed into PacBio sequencing library. DNA damage repair, end repair, and ligation of SMRT adapters were performed using PacBio SMRTbell Template Prep Kit (Pacific Biosciences). Whole genome sequencing was performed using PacBio sequencing platform (Pacific Biosciences, Menlo Park, CA, United States). Sequence runs of three single-molecule real-time (SMRT) cells were performed on the PacBio RS II sequencer with a 120-min movie time/SMRT cell. SMRT Analysis portal version 2.1 was used for read filtering and adapter trimming, with default parameters, and post-filtered data of 1.479 Gb ( ∼404X coverage) with an average read length of ∼6.2 kb was used for subsequent assembly (Supplementary Table 1).

#### Genome Assembly and Gene Annotation

The post-filtered genome reads were de novo assembled by Canu (v1.4) (Koren et al., 2017), which produced one single large contig ( ∼3.6 Mb). Circlator was used to circularize this contig into a complete circular genome (Hunt et al., 2015). Protein-coding and non-coding genes in the SUNEO genome were annotated using National Center for Biotechnology Information (NCBI) Prokaryotic Genomes Automatic Annotation Pipeline (PGAAP). Functional classification of these annotated genes was carried out by RPSBLAST version 2.2.15 in conjunction with Clusters of Orthologous Groups (COGs) of proteins databases ( E-values < 0.001).

#### Comparative Genomics Analysis and Classification of Pan-Genomic Core Genes and Strain-Specific Genes

To study the comparative genomics of S. acidaminiphila , three whole genome sequences of S. acidaminiphila strains; S. acidaminiphila SUNEO, JCM 13310 (Assih et al., 2002), and ZAC14D2\_NAIMI4\_2 (Vinuesa and Ochoa-Sanchez, 2015) were downloaded from the NCBI database (**Table 1**). The protein sequences of all three strains were BLAST-aligned against each other ( E-value < 0.001). However, BLAST may identify false homologs due to repeat sequences commonly shared by multiple genes. Thus, a gene is considered to be shared by both strains if the alignment coverage of both genes is at least 60%. The cutoff

TABLE 2 | Genomicfeatures of S. acidaminiphila strain SUNEO, ZAC14D2\_NAIMI4\_2, and JCM

 13310.

Strain SUNEO

ZAC14D2\_NAIMI4\_2

JCM 13310 ∗1Total genes = Total CDS + rRNA + tRNA +

Complete Complete

Draft

ncRNA. ∗2Total CDS = Total Proteins +

3,660,864 4,138,297 3,942,520

69.8 68.5 68.8 Pseudo genes.

3,247 3,709 3,636

3,573

3,635

3,173

292 65 131

2,881

3,570

3,442

 3

 56

 11 April, 2017

 9

 61

 11 April, 2017

 9

 61

16 February, 2017

Genome assembly status

Genome size (bp)

 GC content (%)

 Total Genes∗1

Total CDS∗2

Pseudo genes

 Total proteins

 rRNA

 tRNA

 Modify date


size: 10,000 bp). (9) GC skew showing regions above and below average in yellow and light blue, respectively. (window size: 10,000 bp).

was determined by the statistics of alignment coverage of all gene-pairs. We observed that 60% act as a good cutoff for balancing sensitivity and specificity. We consider each gene to be strain-specific if it is presented in only one strain and lost in all other strains. On the other hand, the genes shared by all strains are considered to be pan-genomic core genes.

## 16S rRNA Phylogenetic Analysis

The publicly available 16S rRNA sequences of type strains of Stenotrophomonas spp. (including the S. acidaminiphila strain SUNEO) were retrieved from the National Center for Biotechnology Information (NCBI) nucleotide database (Supplementary Table 2) (Pruitt et al., 2007; Alavi et al., 2014; Davenport et al., 2014; Patil et al., 2016). In particular, two S. acidaminiphila strains (i.e., JCM 13310 and ZAC14D2 NAIMI4) were included in order to confirm the phylogenetic status of SUNEO. At first, multiple sequence alignment of the 16S rRNA gene sequences of all strains was first performed by MEGA (v7). Specifically, ClustalW was used for multiple sequence alignment. Evolution history was reconstructed using the built-in maximum-likelihood method with 1,000 bootstraps.

## Multi-Locus Sequence Typing Using Multiple Housekeeping Genes

To further validate these clade assignments, multilocus sequence analysis (MLSA) was performed by concatenation of housekeeping genes: atpD, guaA, mutM, nuoD, ppsA, and recA (Kaiser et al., 2009). Multiple sequence alignment of these housekeeping genes in 15 Stenotrophomonas genomes was performed using MEGA in order to infer their phylogeny (Alavi et al., 2014; Davenport et al., 2014; Patil et al., 2016; Sanchez-Castro et al., 2017). Two conventional gyrB and gapA were not included because gyrB is completely absent in S. acidaminiphila JCM 13310 and only very short piece of gapA is found in S. ginsengisoli DSM 24757. Additional MLSA using the six housekeeping genes plus 16S rRNA is also performed using MEGA (v7) to confirm the phylogenetic position of SUNEO.

## Whole-Genome Average Nucleotide Identity Analysis

To measure the nucleotide-level genomic similarity between SUNEO and related Stenotrophomonas genomes

(Alavi et al., 2014; Davenport et al., 2014; Patil et al., 2016), the Average Nucleotide Identity (ANI) (Konstantinidis and Tiedje, 2005) was calculated by the USEARCH program (Yoon et al., 2017) based on modified OrthoANI algorithm (Lee et al., 2016). A radial phylogram was constructed using distance matrix computations using the Integrated Microbial Genomes pipeline (Chen et al., 2017).

## Annotation of Antibiotic-Resistance Genes

The SUNEO resistome is annotated by using both the Resistance Gene Identifier from the Comprehensive Antibiotic Resistance Database (McArthur et al., 2013) and the IMG database (Markowitz et al., 2012). RGI prediction of resistome is based on homology and SNP models, where the strict criteria were chosen for prediction. In homolog models, BLAST is used to detect functional homologs of antimicrobial resistance genes. In contrast, SNP models identify candidate genes which acquire mutations conferring antimicrobial resistance genes based on curated SNP matrices. The SUNEO resistome is predicted through alignment against the IMG database using BLASTN with a 95% sequence identity threshold.

## Sequence, Structural, and Selection Analysis of folP Gene

The phylogeny of the DHPS protein was constructed by MEGA7. The publicly available folP homolog gene sequences of 18 Stenotrophomonas strains (including S. acidaminiphila strain SUNEO) were retrieved from the National Center for Biotechnology Information (NCBI) nucleotide database

(Supplementary Table 3). The amino acid sequences of DHPS from SUNEO and JCM 13310 were BLAST-aligned against each other in order to identify the mutation loci and annotate the conservative loop regions. To probe its topological structure, the 3D structure of DHPS proteins were predicted by Robetta<sup>2</sup> , and subsequently visualized by NOC 3.01<sup>3</sup> . In conjunction with the annotated loop regions in the sequence level, we were able to compare the local substructure of Loop2 between the two strains. The ratio of non-synonymous to synonymous substitutions (i.e., Ka/Ks) is used to estimate positive and purifying selection at each amino-acid site in folP for SUNEO, JCM 13310, and

<sup>2</sup>http://robetta.bakerlab.org/

<sup>3</sup>http://noch.sourceforge.net

ZAC14D2\_NAIMI4\_2. The M8 site model (with ω allowed to > 1) with intermediate precision level (ε = 0.1, which defines when two likelihood values converged) is used to compute the Ka/Ks ratio<sup>4</sup> .

#### RESULTS

#### Genome Overview and Annotation

The total size of the genome is 3,660,864 bp with a G+C content of 69.8%. An illustration of the genomic contents in the genome of SUNEO is shown in **Figure 1**. A total of 3,173 Coding Sequences (CDSs) were predicted (**Table 2**). In

<sup>4</sup>http://selecton.tau.ac.il

TABLE 3 | Antimicrobial susceptibility profiles of S. acidaminiphila SUNEO and JCM 13310.


<sup>∗</sup>R, resistant; S, susceptible; I, intermediate.

addition, 70 RNAs including rRNA and tRNA were identified. No extrachromosomal elements were found in SUNEO.

#### Comparative Genomic Study and Identification of Core and Strain-Specific Genes of S. acidaminiphila Genomes

General genomic features of S. acidaminiphila SUNEO were compared to the S. acidaminiphila JCM 13310 and ZAC14D2\_NAIMI4\_2 (**Table 2**). The genome size of the S. acidaminiphila strain ZAC14D2\_NAIMI4\_2 was the largest (4,138,297 bp) amongst all genomes (ranging from 3,660,864 to 3,942,520 bp). The highest genomic G+C content (69.8%) was from the S. acidaminiphila strain SUNEO followed by the S. acidaminiphila strain JCM 13310 (68.8%), and the S. acidaminiphila strain ZAC14D2\_NAIMI4\_2 (68.5%).

The protein-coding genes of SUNEO were compared with those of S. acidaminiphila JCM 13310 and ZAC14D2\_NAIMI4\_2, in order to identify orthologous core genes which are shared across all strains and strain-specific genes. **Figure 2** depicts both the positions and the color-coded functions of S. acidaminiphila SUNEO genes in comparison with all other strains, whereas the number of orthologous and strain-specific genes is shown in **Figure 3**. In summary, the core genome of S. acidaminiphila consisted of 2,736 core genes shared across all strains, whereas 807 genes are specific only to S. acidaminiphila SUNEO (**Figure 3A**). Functional analysis of SUENO-specific genes revealed that, in addition to hypothetical proteins, a relative abundance of the gene is involved in carbohydrate transport, along with metabolism and cell wall/membrane/envelop biogenesis (**Figure 3B**). DHPS encoding folP homologs genes are presented in all strains.

#### Phylogenetic Analysis of S. acidaminiphila

A maximum-likelihood tree of the three S. acidaminiphila genomes and 12 reported Stenotrophomonas strains (comprising of S. daejeonensis, S. humi, S. pictorum, S. terrae, S. nitritireducens, S. ginsengisoli, S. koreensis, S. maltophilia, S. pavanii, S. chelatiphaga, S. panacihumi, and S. rhizophila) was created based on 16S rRNA gene sequences (Supplementary Figure 1). This phylogenetic tree shows S. acidaminiphila SUNEO, JCM 13310, and ZAC14D2\_NAIMI4\_2 grouped together. The result was further supported by Alignment Fraction analysis, which showed SUNEO was included in the S. acidaminiphila JCM 13310 and ZAC14D2\_NAIMI4\_2 phylogenetic subgroup (Supplementary Figure 2).

The MLSA using conventional housekeeping genes (with or without 16S RNA) both revealed high phylogenetic similarities higher than 99% among SUNEO, JCM 13310, and ZAC14D2\_NAIMI4\_2 (**Figure 4** and Supplementary Figure 3), which was the accepted species threshold (Kaiser et al., 2009). Genomic-wide relatedness comparison was calculated with the OrthoANI program using publicly available genomes from type strains of Stenotrophomonas species (Alavi et al., 2014; Davenport et al., 2014; Patil et al., 2016; Sanchez-Castro et al., 2017). As the ANI value of SUNEO to S. acidaminiphila strain is 92.94–92.83% (Supplementary Table 4), indicated a taxonomic

TABLE 4 | Summary of the antibiotic resistance genes among the three strains of S. acidaminiphila and their related locus tag.


outlier (Gomila et al., 2015). Together, all these analysis (from single gene, multiple genes, to entire genome) concordantly concluded that the phylogenetic position of SUNEO is indeed belonging to S. acidaminiphila.

#### Comparative Analysis of Antibiotic Resistance Genes in SUNEO

Antimicrobial susceptibility test showed that SUNEO is resistant to both imipenem and trimethoprim/sulfamethoxazole (**Table 3**). Comparative analysis was performed on the three S. acidaminiphila genomes, among which JCM 13310 is trimethoprim/sulfamethoxazole susceptible (**Table 3**). All three strains harbor similar resistance genes. There are Ambler class B β-lactamase L1 and Ambler class A β-lactamase L2 in all of the three strains. The resistance-nodulation-division (RND) family efflux pump, consisting of the smeDEF and smeOP genes, along with the efflux pumps genes that are homologous to efflux pumps encoded in S. maltophilia and Escherichia coli were also identified in all of the tested strains (**Table 4**).

#### Phylogeny of folP Homologs in Stenotrophomonas Strains

Comparative genomic analysis has revealed the gene folP is commonly presented among S. maltophilia strains. folP encodes DHPS and is the target of sulfonamides, to which SUNEO is resistant. A BLASTP search in Stenotrophomonas strains with folP as the primer sequence returned a collection of related homologs with the annotation of DHPS (Baker and Sali, 2001). We retrieved 18 protein sequence candidates with available MIC data for each strain. The phylogeny of the DHPS protein constructed by MEGA7 (Kumar et al., 2016) clearly presented two distinct groups: one denotes a family of folP homologs from S. maltophilia whereas the other comprises a series of folP homologs from non-maltophilia Stenotrophomonas strains (**Figure 5**). Of particular note, folP homologs between

SUNEO and JCM 13310 are highly similar in comparison with other distant-related strains. This implies that the resistance of SUNEO to sulfonamides is due to few/key mutations acquired occasionally instead of continual accumulation of resistant alleles after speciation.

## Sulfonamide-Binding Site Mutation Revealed by Structural Analysis

In order to identify key mutation in folP leading to resistance of SUNEO, sequence composition of DHPS between JCM 13310 and SUNEO was compared, which exhibited an amino acid change (Gly<sup>72</sup> → Glu72) in one highly conservative region (termed Loop2) (**Figure 6**). Structural modeling allows us to visualize the difference in DHPS architecture between these two strains. This conservative region stabilizes the binding of PABA and variation at this region has been shown to contribute to resistance to sulfonamide (Yun et al., 2012). We further reconstruct their protein 3D structures to verify the difference of DHPS architecture between these two strains. The protein substructure at Loop2 in SUNEO appears disordered in comparison with that of JCM 13310 (**Figure 7**), which may reduce the binding stability of sulfonamide and lead to resistance.

The selection pressure (Ka/Ks) was measured along the entire folP (**Figure 8**). The results indicated that strong signals of purifying selection (Ka/Ks < 0.08) are widely spread in folP. This implies evolution of folP in SUNEO is probably constrained by high selection pressure from long-term exposure of sulfonamide.

(sulfonamide) resistance mutations. The last line in the data means the consensus amino acid symbol of the residue in the corresponding position for all strains: an asterisk indicates positions which have a single, fully conserved residue; a colon indicates conservation between groups of strongly similar properties; a period indicates conservation between groups of weakly similar properties.

FIGURE 7 | Structure predication for the folP gene products (dihydropteroate synthase) of S. acidaminiphila SUNEO and JCM 13310. (A) Glu72 in the SUNEO dihydropteroate synthase. (B) Gly72 in the JCM 13310 dihydropteroate synthase.

#### DISCUSSION

The data we present represents a first glimpse into the evolution and mechanism for sulfamethoxazole resistance in S. acidaminiphila. Our discovery of novel mutations in DHPS offers new insight into the newly emerging field of Stenotrophomonas infections, furthering our understanding of the diversity in the dissemination of sulfamethoxazole resistance. Sulfamethoxazole interrupt the essential folate pathway in bacteria by targeting the enzyme DHPS, which in turn catalyzes the condensation of DHPP with PABA in the production of the folate intermediate. The locus of the mutation is consistent with earlier observations that many sulfamethoxazole resistance mutations are located within two conserved loops (called Loop1 and Loop2) of DHPS, which creates a specific binding pocket for PABA (Estrada et al., 2016).

To gain further structural insight of the DHPS homologs in SUNEO and JCM 13319, structural modeling was performed, where the ribbon structures of their DHPS were generated. The 3D structural DHPS proteins in SUNEO illustrated that the substructure at Loop2 is disordered. Further structural comparisons of DHPS also indicated that the resistant (SUNEO) and sensitive (JCM 13310) strains display a different substructure around the PABA-binding pockets. In agreement with the earlier proposal by Yun et al. (2012), sulfonamide resistance is associated with the Loop2 mutation and subsequent DHPS substructure disorder affected the binding of PABA. It was also reported that mutations in DHPS were associated with sulfamethoxazole resistance in both prokaryotes (Huovinen, 2001) and primitive eukaryotes (Triglia et al., 1997), where associated structural changes in Loop1–Loop2 PABA binding sites occurred (Capasso and Supuran, 2014). As sulfa drugs interrupt the folate pathway by competing with PABA as DHPS substrates, mutation at both the sequence and the structure of DHPS in SUNEO could be attributed to its resistance to sulfamethoxazole.

The selection pressure measured along the DHPS-encoding gene folP reveals strong signals of purifying selection, implying the evolution of folP is constrained by high selection pressure. Sulfamethoxazole has been the first-line antibiotic agent against Stenotrophomonas for decades and is widely use in agriculture. The usage of specific antibiotic agent has been shown to result in

purifying selection in certain genes of resistant strains (Mortimer et al., 2018), because purifying selection on a subset of genes can be intensified in the setting of resistance. After sufficiently long-term usage of the same antibiotic agent, resistance may even reach a point of stabilizing evolution, completely reducing or invaliding the efficiency of the drug (Cornick et al., 2014). As folP in S. acidaminiphila already exhibited purifying selection, the administration of sulfamethoxazole for S. acidaminiphila infections in the future should be taken with caution.

Stenotrophomonas acidaminiphila SUNEO is also resistant to various antibiotics, suggesting multiple resistance mechanisms. Current understanding of the resistance determinants of S. acidaminiphila is limited. Vinuesa and Ochoa-Sanchez (2015) reported on predicted antibiotic resistant genes (without phenotypic resistance) in S. acidaminiphila ZAC14D2\_NAIMI4\_2 isolated from river sediment in Mexico, while Assih et al. (2002) reported on phenotypic resistance (without genotypic changes) in S. acidaminiphila JCM 13310. To address this concern, we conducted resistome analysis of S. acidaminiphila SUNEO and predicted multiple efflux pumps, which have also been detected in S. maltophilia strains (Crossman et al., 2008). In particular, we identified the RND efflux pump genes smeDEF, which was associated with trimethoprim/sulfamethoxazole resistance (Sanchez and Martinez, 2015). The resistance of trimethoprim/sulfamethoxazole in SUNEO could be a combination effect related to DHPS mutation and efflux pumps.

Stenotrophomonas acidaminiphila is able to degrade a number of organic pollutants, including Fomesafen [5-(2chloro-4-[trifluoromethyl]phenoxy)-N-methylsulfonyl-2-nitrobenzamide] (Huang et al., 2017), Diuron [3-(3,4 dichlorophenyl)-1,1-dimethylurea] (Egea et al., 2017), and azo dye crystal violet (Kim et al., 2002). Our comparative analysis of resistome in S. acidaminiphila revealed that the efflux pumps genes presented in all examined S. acidaminiphila strains. Efflux pumps play a major role in both solvent tolerance and bioremediation (Fernandes et al., 2003), which is consistent with recent observations of the biodegradation of sulfonamide (Liao et al., 2016) and aminoglycoside (Selvaraj et al., 2018) in S. acidaminiphila. The mechanistic insight we gained further raises the possibility of cross-resistance to both environmental toxic compounds and antibiotics which would have a major impact on the use of disinfectants and disinfecting procedures.

Currently, most of the reported Stenotrophomonas infections are caused by S. maltophilia which is frequently recovered from clinical samples and is an emerging opportunistic pathogen associated with substantial morbidity and mortality, particularly in immunocompromised patients (Falagas et al., 2009). Incidences of human infection appear to have increased recently, where a variety of clinical syndromes have been described, including pneumonia, bacteremia, and peritonitis (Denton and Kerr, 1998; Sattler et al., 2000). However, biliary tract infections remain uncommon. Our report outlines the first human biliary infection caused by S. acidaminiphila. The virulence factors of S. acidaminiphila are largely unknown. Our data has revealed RND pump homologs of the acrAB in SUNEO. acrAB encodes a bile-induced efflux system and is expressed in both animal models and infected patients (Gunn, 2000; Piddock, 2006). Additional studies are required in order to clarify its role in S. acidaminiphila pathogenesis.

#### CONCLUSION

fmicb-09-01013 May 15, 2018 Time: 16:56 # 12

Our analysis reveals a possible core genome of S. acidaminiphila, along with accessory genomes specific to each strain, providing insights into the resistant potential of the clinical isolate. We propose a scenario for the origin and evolution of S. acidaminiphila, based on its genomic features. Gene annotation and comparative analysis further revealed a unique profile of folP mutation. The mechanism for sulfonamide resistance in S. acidaminiphila SUNEO appears to involve the mutation of the Loop2 region of DHPS, thereby leading to alterations in the structural conformation of the site and the multi-drug efflux pumps.

#### DATA AVAILABILITY

This genome project, which includes the raw read data, assembly, and annotation, has been deposited at NCBI/GenBank as BioProject PRJNA374779. The assembly is available under accession CP019797; the version described in this paper is the first version.

## REFERENCES


## AUTHOR CONTRIBUTIONS

Y-TH, J-MC, and P-YL designed and coordinated the study and carried out the data analysis. Y-TH, J-MC, and P-YL performed the bioinformatics analysis. Z-YW, B-CH, and P-YL carried out the experiments and interpreted data for the work. Y-TH, Z-YW, RK, and P-YL wrote the manuscript. Y-TH, RK, and P-YL checked and edited the manuscript. All authors have read and approved the manuscript.

## FUNDING

Y-TH was supported in part by the Ministry of Science and Technology (MOST) with 106-2221-E-194-056-MY3. P-YL was supported by the Taichung Veterans General Hospital with TCVGH-1073901B and TCVGH-NK10 79003.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2018.01013/full#supplementary-material



**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 © 2018 Huang, Chen, Ho, Wu, Kuo 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) and the copyright owner 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.

# Mutational Evolution of Pseudomonas aeruginosa Resistance to Ribosome-Targeting Antibiotics

#### Fernando Sanz-García, Sara Hernando-Amado\* and José L. Martínez\*

Centro Nacional de Biotecnología, Consejo Superior de Investigaciones Científicas, Madrid, Spain

The present work examines the evolutionary trajectories of replicate Pseudomonas aeruginosa cultures in presence of the ribosome-targeting antibiotics tobramycin and tigecycline. It is known that large number of mutations across different genes – and therefore a large number of potential pathways – may be involved in resistance to any single antibiotic. Thus, evolution toward resistance might, to a large degree, rely on stochasticity, which might preclude the use of predictive strategies for fighting antibiotic resistance. However, the present results show that P. aeruginosa populations evolving in parallel in the presence of antibiotics (either tobramycin or tigecycline) follow a set of trajectories that present common elements. In addition, the pattern of resistance mutations involved include common elements for these two ribosome-targeting antimicrobials. This indicates that mutational evolution toward resistance (and perhaps other properties) is to a certain degree deterministic and, consequently, predictable. These findings are of interest, not just for P. aeruginosa, but in understanding the general rules involved in the evolution of antibiotic resistance also. In addition, the results indicate that bacteria can evolve toward higher levels of resistance to antibiotics against which they are considered to be intrinsically resistant, as tigecycline in the case of P. aeruginosa and that this may confer cross-resistance to other antibiotics of therapeutic value. Our results are particularly relevant in the case of patients under empiric treatment with tigecycline, which frequently suffer P. aeruginosa superinfections.

#### Edited by:

Elena Perrin, Università degli Studi di Firenze, Italy

#### Reviewed by:

Antonio Oliver, Hospital Universitario Son Dureta, Spain Michael L. Vasil, University of Colorado, United States

#### \*Correspondence:

Sara Hernando-Amado shernando@cnb.csic.es José L. Martínez jlmtnez@cnb.csic.es

#### Specialty section:

This article was submitted to Evolutionary and Genomic Microbiology, a section of the journal Frontiers in Genetics

Received: 26 July 2018 Accepted: 18 September 2018 Published: 18 October 2018

#### Citation:

Sanz-García F, Hernando-Amado S and Martínez JL (2018) Mutational Evolution of Pseudomonas aeruginosa Resistance to Ribosome-Targeting Antibiotics. Front. Genet. 9:451. doi: 10.3389/fgene.2018.00451 Keywords: antibiotic resistance, Pseudomonas aeruginosa, tobramycin, tigecycline, mutation, evolution

## INTRODUCTION

Antibiotic resistance has been a major public health concern since the dawn of the antibiotic era, but in recent decades there has been an alarming increase in the number and type of antibiotic-resistant bacteria (Appelbaum, 2012), posing a threat to health worldwide (Roca et al., 2015). Predicting the mechanisms by which bacteria may acquire resistance is therefore important in the prevention and treatment of infections (Martinez et al., 2007; Martinez et al., 2011).

Resistance can be acquired via horizontal transfer of antibiotic resistance genes (HGT), or through mutation (Hernando-Amado et al., 2017). While exhaustive information is available on the mechanisms of antibiotic resistance at the basic science and epidemiological levels, the evolutionary trajectories leading to high level antimicrobial resistance, as well as the reproducibility of these trajectories among populations evolving concurrently, have been studied in less detail. It is worth mentioning, however, that the reconstruction of mutants that are selected in patients

**134**

under treatment have shown that fitness costs and the selection of compensatory mutations are critical for the success of some specific antibiotic resistance mutations (Shcherbakov et al., 2010; Brandis et al., 2015; Meftahi et al., 2016; Huseby et al., 2017). Nevertheless, this type of retrospective analyses is useful just for studying already known mechanisms of resistance, not for predicting new ones (Pietsch et al., 2017).

Strategies to predict the emergence of resistance mutations (Martinez et al., 2007) were implemented soon after the discovery of antibiotics (Bryson and Szybalski, 1952), one of the most useful of which is experimental evolution. Since the seminal work of Richard Lenski, experimental evolution has been used to analyze different bacterial traits, including the development of resistance to antibiotics (Bryson and Szybalski, 1952; Toprak et al., 2011; Turrientes et al., 2013; Feng et al., 2016; Ibacache-Quiroga et al., 2018). Recent research has shown experimental evolution able to predict the emergence of resistance against different antimicrobial agents, including colistin (Jochumsen et al., 2016), beta-lactams, quinolones and aminoglycosides (Cabot et al., 2016; Feng et al., 2016; Ibacache-Quiroga et al., 2018; Lopez-Causape et al., 2018b). In recent years, the potential of experimental evolution has been further boosted by the development of technologies that allow the fast and affordable sequencing of whole bacterial genomes.

Pseudomonas aeruginosa, an opportunistic pathogen widely distributed in nature (Silby et al., 2011), commonly causes lung, airway and other infections in hospitalized patients. It is the main cause of chronic infections in patients with cystic fibrosis (CF) and chronic obstructive pulmonary disease (Martinez-Solano et al., 2008; Tummler et al., 2014). These infections are usually fought using aminoglycosides, β-lactams and polymyxins (Palmer and Whiteley, 2015). Unfortunately, P. aeruginosa intrinsically shows low-level susceptibility to a number of drugs, even against the recently developed glycylcycline tigecycline (Pankey, 2005), which works via tightly binding to the ribosome and thus evading the most common tetracycline resistance mechanisms. In addition, mutants presenting increased levels of antibiotic resistance are selected along chronic infections. Indeed, while resistance to aminoglycosides in P. aeruginosa isolates from acute infections has been largely attributed to the acquisition of antibiotic resistance genes, mutation plays a major role for the acquisition of resistance by P. aeruginosa causing chronic infections (Vogne et al., 2004; Guenard et al., 2014; Bolard et al., 2018).

In the present work, experimental evolution and wholegenome sequencing (WGS) were used to examine evolutionary trajectories of P. aeruginosa toward resistance against two ribosome-targeting antimicrobials: tobramycin and tigecycline. The aim was to determine whether the mechanisms that impair the actions of different drugs targeting the same cell machinery are shared (at least in part), or whether resistance to each drug is specific via a different mechanism. Tigecycline binds to the 30S ribosomal subunit, thereby blocking the interaction of aminoacyl-tRNA with the A site of the ribosome, whereas tobramycin prevents the formation of the 70S complex (Kotra et al., 2000). While tobramycin forms part of usual therapy regimens against P. aeruginosa (Cheer et al., 2003), the pathogen is intrinsically resistant to tigecycline (following the clinical definition of antibiotic resistance) (Martinez et al., 2015). One cause of this phenotype is the capability of the multidrug efflux pump MexXY, also involved in intrinsic P. aeruginosa tobramycin resistance (Westbrock-Wadman et al., 1999), to extrude tigecycline (Dean et al., 2003). Nonetheless, tigecycline was used in the present study since it provides an opportunity to examine whether or not bacterial pathogens can acquire clinically relevant characteristics when challenged with the antibiotics to which they are considered to be intrinsically resistant. In this regard, it is worth mentioning that P. aeruginosa has emerged as a major cause of superinfection in nosocomial patients treated with tigecycline (Garcia-Cabrera et al., 2010; Ulu-Kilic et al., 2015; Katsiari et al., 2016). Knowing whether or not the empirical use of tigecycline for treating Gramnegative hospital infections might challenge P. aeruginosa, affecting its susceptibility to other antibiotics commonly used for treating P. aeruginosa infections, is of relevance for developing a rational approach for treating such superinfections. The results showed that even for microorganisms dubbed intrinsically resistant to an antibiotic, the challenge with this antibiotic selects mutants presenting reduced susceptibility to antibiotics of clinical value.

Experimental evolution studies allow one to determine whether evolutionary trajectories are reproducible, i.e., whether the process of evolution is mainly deterministic and hence predictable (Martinez et al., 2007), or whether it is largely stochastic. The present work provides a predictive analysis of the potential mutational causes of resistance in P. aeruginosa against two ribosome-targeting antibiotics belonging to different structural families, as well as the different evolutionary trajectories taken toward this resistance. This information may allow new strategies to be designed for predicting, managing, and eventually reducing resistance in this important nosocomial pathogen. In addition, the results throw light on whether bacterial evolution is largely stochastic or presents some deterministic features, and thus whether the emergence and spread of antibiotic resistance can be predicted to a certain extent (Martinez et al., 2007, 2011).

## MATERIALS AND METHODS

#### Growth Conditions and Determination of Susceptibility to Antibiotics

Unless otherwise stated, bacteria were grown in Mueller Hinton Broth (MHB, Pronadisa) at 37◦C with agitation at 250 rpm. The initial concentrations of tigecycline (Pfizer) and tobramycin (Normon, S. L) that inhibit the growth of P. aeruginosa PA14 under the culture conditions used in the evolution experiments were determined at 37◦C.

General susceptibility to a wide range of antibiotics tigecycline, tetracycline, aztreonam, ceftazidime, imipenem, ciprofloxacin, levofloxacin, norfloxacin, tobramycin, streptomycin, amikacin, colistin, polymyxin B, chloramphenicol, fosfomycin and erythromycin – was examined by disk diffusion in Mueller Hinton Agar (MHA) (Sigma) at 37◦C.

The MICs of different antibiotics were determined for the bacterial populations over the evolution period at 37◦C in MHA using E-test strips (MIC Test Strip, Liofilchem <sup>R</sup> ). MICs of colistin and polymyxin B were determined in MHB II by double dilution in microtiter plates. The MICs to the antibiotics of selection and to fosfomycin were repeated twice and in all cases, the results were the same in the replicated assays.

#### Experimental Evolution

fgene-09-00451 October 16, 2018 Time: 19:31 # 3

Twelve independent bacterial populations (four controls without antibiotics, four populations challenged with tigecycline, and four populations challenged with tobramycin) were grown in parallel in MHB for 35 consecutive days. All replicates were established from a stock culture of the P. aeruginosa PA14 strain. Each day, the cultures were diluted (1/250) with fresh MHB: 8 µl of bacterial culture in 2 ml of medium. The concentrations of tigecycline and tobramycin used for selection increased over the evolution period from the initial MIC up to 32MIC, doubling them every 5 days. Every 5 days, samples from each culture were taken and preserved at −80◦C for future investigation.

### Whole-Genome Sequencing

Genomic DNA was extracted at the end of the evolution assays from all 12 populations using the Gnome <sup>R</sup> DNA kit (MP Biomedicals). Whole-genome sequencing was performed by Sistemas Genómicos S.L. Libraries were obtained without amplification following Illumina protocols and recommendations. The quality of the extracted material was analyzed via a 4200 Tapestation High Sensitivity assay, and the DNA concentration determined by real-time PCR using a LightCycler 480 device (Roche). The pool of libraries was pairend sequenced (100 × 2) in an Illumina HiSeq 2500 sequencer. The average number of reads per sample was 8646177, which represents a coverage of 200x on average. Short reads used in this publication are deposited in SRA database<sup>1</sup> with accession PRJNA490803.

#### Bioinformatic Analysis

Mutations in the evolved bacteria were detected using CLC Genomics Workbench 9.0 (QIAGEN) software. WGS data were trimmed and the reads aligned with the P. aeruginosa UCBPP-PA14 reference chromosome (NC\_008463.1). The single nucleotide polymorphisms (SNPs) present in the populations kept under selective pressure were identified and filtered against those present in the populations maintained in the absence of such pressure. The cut-off threshold of a mutation to be included in the analysis was ≥15%.

#### Confirmation of SNPs

Sanger sequencing was used to verify the mutations found via WGS (**Supplementary Table S1**) and to ascertain the order of appearance of these modifications. Twenty-four pairs of primers, which amplified 100–200 base pair regions containing each putative mutation, were designed (**Supplementary Table S2**). After PCR amplification, the corresponding amplicons were purified using the QIAquick PCR Purification Kit (QIAGEN) and sequenced at GATC Biotech.

### RNA Extraction and Real-Time RT-PCR

One flask containing 20 ml of MHB was inoculated with an overnight culture of the selected clones and P. aeruginosa PA14 to a final O.D.<sup>600</sup> = 0.01, and they were incubated until exponential phase was reached (O.D.<sup>600</sup> = 0.6). 10 ml of each culture were centrifuged at 7000 rpm for 15 min and at 4◦C. This process was performed with three independent biological replicates.

Then, RNeasy mini Kit (QIAGEN) extraction protocol was followed: 570 µl of TE buffer (10 mM Tris-HCl, 1 mM EDTA [pH 8.0]) and 30 µl of lysozyme (Sigma), for a final concentration of 1 mg/ml of the latter, were added to each sample. Afterward, the samples were mixed by vortexing for 10 s and were incubated at room temperature for 10 min with regular vortexing. A volume of 2100 µl of buffer RLT (QIAGEN) was added, and samples were sonicated at 0.45 Hz for 20 s. Next, 1410 µl of ethanol (Merck) was added and the protocol continued according to the manufacturer's instructions. In order to remove any residual DNA, two DNase treatments were carried out, with DNase I (QIAGEN) and TURBO DNase (Ambion). Finally, a PCR with rplU primers (**Supplementary Table S2**) was performed to check that no residual DNA was present in the RNA samples. High Capacity cDNA Reverse Transcription Kit (Applied Biosystems) was used to synthesize cDNA from 5 µg of RNA. Then, realtime RT-PCR was performed using 50 ng of cDNA and the Power SYBR green PCR Master Mix (Applied Biosystems) in the ABI PRISM 7500 real-time PCR system (Applied Biosystems). Gene expression data were normalized using the gene rplU (**Supplementary Table S2**). Differences in the relative amounts of mRNA were obtained following the 2−11Ct method (Livak and Schmittgen, 2001).

## RESULTS AND DISCUSSION

Different publications have shown that the genes that contribute to antibiotic resistance (intrinsic resistome) comprise around 3% of the genome of a given bacterial species (Breidenstein et al., 2008; Fajardo et al., 2008; Tamae et al., 2008; Alvarez-Ortega et al., 2010; Blake and O'Neill, 2013; Fernandez et al., 2013). In the case of P. aeruginosa the search of a comprehensive transposon-tagged library has shown that the inactivation of 135 genes renders lowlevel tobramycin resistance (Schurek et al., 2008). Even if we do not take into consideration gain of function mutations (which are not considered in transposon insertion libraries studies), neither that different mutations might be selectable in each of these genes (Martinez and Baquero, 2000), hence increasing the number of potential antibiotic resistance mutants (Lopez-Causape et al., 2018a), the number of possible combinations of these 135 genes is 2.7E230. It can be argued that the number of mutations that a single bacterium can accumulate is likely low. However, even if only five mutations are accumulated as reported in the present work (see below), the number of combinations of these 135 potential mutations, taken five by five, is 4.1E10 if the order of

<sup>1</sup>https://www.ncbi.nlm.nih.gov/sra

selection is taken into consideration and 3.4E8 if the order of selection of each of the mutations along evolution is not taken into consideration. If all possible combinations were equivalent in terms of antibiotic resistance, these numbers would mean that mutation-driven antibiotic resistance should be un-predictable.

## Stepwise Evolution of P. aeruginosa Toward Antibiotic Resistance

To determine whether similar potential evolutionary trajectories are followed by different populations, four biological replicates were allowed to evolve in parallel under selective pressure from tobramycin (populations 1–4), tigecycline (populations 5–8) and in the absence of any selective pressure (populations 9–12). In populations 1–8 the antibiotic concentration was doubled every 5 days, from 0.5 µg/ml for tobramycin and 4 µg/ml for tigecycline, up to 32MIC. At 64MIC no growth was seen after 5 days incubation, suggesting this concentration to be beyond limits of P. aeruginosa when evolving toward tobramycin and tigecycline resistance in these experimental conditions.

To monitor the evolution of resistance over the selection period, the susceptibility of each population to the selecting antibiotic was determined every 5 days by E-test. When bacteria are confronted to increased concentrations of antibiotics, two different phenotypic trajectories can be foreseen. Sudden selection of high-level resistance at the first steps of evolution or stepwise selection of mutants presenting increasing resistance levels. As shown in **Figure 1**, a stepwise evolutionary trajectory was observed for both antibiotics, suggesting either accumulation of sequential mutations after each evolution step (i.e., each change in antibiotic concentration), or the displacement of lowlevel resistance mutants by higher-level resistance ones as the selection pressure increased. It is important to note that the evolutionary trajectory was similar in all replicated experiments (**Figure 1**). This suggests the number of evolutionary trajectories possible (at least at phenotypic level) is limited, although the amount of genotypic evolutionary trajectories can be larger (Lassig et al., 2017).

An increase in the MIC of an antibiotic after experimental evolution does not, however, necessarily mean that antibioticresistant mutants have been selected for: resistance may be due to a phenotypic adaptation to the presence of an antibiotic rather than to mutations (Levin and Rozen, 2006; Martinez et al., 2009; Martinez and Rojo, 2011). To address this possibility, the evolved populations were sub-cultured in the absence of selection pressure (three sequential passages in MHB) and the MICs again determined. These were found not to change, indicating that the observed modifications were mainly due to the selection of stable mutants.

## Cross-Resistance and Collateral Sensitivity of Evolved Populations

To determine whether the development of resistance was specific to the selecting antibiotic or also affected susceptibility to other antimicrobials, a range of representative antibiotics was tested (beta-lactams, quinolones, tetracyclines, macrolides, aminoglycosides, polymyxins, and chloramphenicol) by disk diffusion (**Supplementary Table S4**). Despite it has been shown macrolides as azithromycin can select mexCD-oprJ overpressing mutants in P. aeruginosa biofilms (Mulet et al., 2009), no differences in susceptibility to imipenem and erythromycin were detected between the wild-type parental P. aeruginosa PA14 strain and the evolved populations. Whether these differences can be due to the different experimental model (biofilm or planktonic cells) or the different genetic background where mutants are selected (PAO1 or PA14) remains to be established. Nevertheless, almost every evolved population developed resistance against other antibiotics belonging to the different structural families, implying that at least some resistance mutations are not tigecycline- or tobramycin-specific (**Table 1**). Notably, in all cases the evolved populations were more susceptible to fosfomycin than the wild-type strain, and it is important to know whether this is related to the development of resistance to ribosometargeting agents. In this regard, it is worth mentioning that Listeria monocytogenes is more susceptible to fosfomycin when growing intracellularly than when growing extracellularly. The reason for this lies in the overexpression of a hexose-phosphate transporter when the bacteria are grown intracellularly, which provides an entry route for this antibiotic (Scortti et al., 2006). These results indicate that the resistance to ribosome-targeting drugs may correlate with increasing susceptibility to fosfomycin, although the underlying mechanisms remain obscure. Given that these types of drug are widely used clinically, it is important to determine whether this trade-off occurs commonly, and whether the combined or sequential use of both types of antibiotic offers a better alternative to current therapies.

## Mutations Selected in the Presence of Antibiotics

To gain insight into the genetic events associated with the development of resistance in the evolved populations, the genomes of each, as well as that of the original population (for which a frozen sample was available), were sequenced on the last day of the experiment (when the antibiotic concentration was 32MIC). Different mutations can appear by chance or because of the P. aeruginosa adaptation for growing in MHB. Consequently, only those mutations present in the populations evolving under antibiotic selective pressure and not in the control populations evolving in absence of selection were taken into consideration. In other words, only those mutations that are enriched (and hence have been selected) under antibiotic selective pressure were taken into consideration. Indeed, all mutant alleles selected in the presence of antibiotics and not present in the populations grown without antibiotics present always a coverage >50% and typically >90% (**Supplementary Table S1**) in the whole population, indicating that they are under positive selection when P. aeruginosa grows in presence of antibiotics. **Supplementary Table S1** shows the locations of all 35 confirmed genetic changes potentially associated with the development of resistance. A total of 31 single-nucleotide variants (SNVs) and 4 multi-nucleotide variants (MNVs; deletions and substitutions of various nucleotides and one transposition) were found, 31 located in genes and 4 in intergenic regions. The majority of the

concentration every 5 days). The detection limit of the tigecycline E-test was 256 µg/ml, limiting the assessment of resistance levels from day 15 to the end of the experiment. MIC values for each replicate are provided in Supplementary Table S3.

mutations located in genes resulted in amino acid alterations, frameshifts or stop codons. In addition, the coverage of the obtained reads was mapped to the P. aeruginosa genome to search for gene amplifications and deletions. One 377 bp deletion was detected in the tigecycline-treated population 8, comprising part of the gene coding for the transcriptional repressor of the multidrug efflux pump mexCD-oprJ, nfxB, plus a small region of the adjacent gene morA.

To further verify the presence in the evolved populations of the mutations identified by WGS (**Supplementary Table S1**), the regions containing these mutations were amplified and the amplicons Sanger-sequenced.

#### Common Aspects of the Evolution of P. aeruginosa Toward Resistance to Tobramycin or Tigecycline

The present results shed light on the development of resistance to ribosome-targeting antibiotics in P. aeruginosa. Although this opportunistic pathogen is already intrinsically resistant to


TABLE 1 | MICs (µg/ml) of antibiotics of different structural families in the populations evolved at 32MIC tobramycin and tigecycline.

Tgc, tigecycline; tet, tetracycline; atm, aztreonam; caz, ceftazidime; cip, ciprofloxacin; tob, tobramycin; s, streptomycin; ak, amikacin; c, chloramphenicol; f, fosfomycin; cs, colistin; pb, polymyxin B. All MICs were obtained by E-test, excepting for polymyxin B and colistin, which were analyzed via double dilution in a microtiter plate.

TABLE 2 | Previously described role on antibiotic resistance of genes presenting mutations in the evolved populations.


tigecycline, highly resistant populations with MICs several times higher than that seen for the wild-type strain were selected for in the experimental evolution assays (**Figure 1**).

Overall, the results indicate that increased resistance to tigecycline and tobramycin comes about via distinguishable evolutionary trajectories but which show some similarities (**Table 2**). In particular, mutations in orfN (selected in all eight replicates) and pmrB (selected in 6 out of 8 replicates) were selected under pressure from either antibiotic (**Figure 2**). It should be noted that all populations evolved in the presence of antibiotic, and none of those evolved in unexposed control populations show mutants in orfN during the first step in evolution (i.e., after the first change in antibiotic concentration). This gene codes for a putative glycosyl transferase needed for the glycosylation of type A flagellins (Schirm et al., 2004). Six of the studied mutants carried a single base pair deletion in a poly-G repeat in orfN, and two contained a single base pair insertion in the same region (**Supplementary Table S1**), all of them leading to a Val50fs mutation. Similar mutations have been found in P. aeruginosa when exposed to ciprofloxacin under experimental evolution conditions (Wong et al., 2012). The fact that these mutations are located in a poly-G repeat region supports that this gene might have a specific high mutation rate as the consequence of polymerase slippage (Moxon et al., 2006). We hypothesize that this might be the reason why orfN mutants are detected in all replicates.

pmrAB is a two-component system involved in polymyxin resistance (Moskowitz et al., 2012). The system regulates the

expression of operons involved in the biosynthesis of lipid A with 4-aminoarabinose, which produces a more positively charged lipopolysaccharide, thus reducing the binding and the activity of cationic peptides (Gunn et al., 2000). Gain-of-function mutants with lipid A modifications can be selected for and are resistant to colistin (Fernández et al., 2010). Moreover, a recent evolution study of P. aeruginosa involving colistin gave rise to the selection of mutations in pmrB (Jochumsen et al., 2016), and it has been proposed that mutations in such regulatory elements may potentiate the effect of other mutations (Lind et al., 2015). Indeed, pmrB is involved in a wide range of antibiotic resistances, including those to quinolones, tigecycline, tobramycin and cationic peptides (Muller et al., 2011; Lopez-Causape et al., 2018b). In agreement with these data, the populations selected in our evolution experiment present a reduced susceptibility to colistin and polymixin B (**Table 1**).

#### Evolutionary Pathways Toward Tobramycin Resistance in P. aeruginosa

Having shown the common features of the evolutionary trajectories toward resistance to both test antibiotics, the specific pathways toward resistance to each were sought. Mutations in fusA and ptsP were found to be involved in evolution toward tobramycin resistance (Schurek et al., 2008; Wang et al., 2015; Feng et al., 2016). Together with the above mentioned orfN mutants, mutations in fusA, which codes for elongation factor G, may be envisaged as a first response to aminoglycosides (Wang et al., 2015; Feng et al., 2016) since they were observed in all evolved populations (**Figure 2**). Further, recent works have shown this type of mutations to be present in clinical isolates as well as in in vitro selected aminoglycoside resistant mutants (Bolard et al., 2018; Ibacache-Quiroga et al., 2018; Lopez-Causape et al., 2018b), reinforcing its role in the acquisition of aminoglycosides resistance. Although mexXY overexpression is considered a hallmark of aminoglycoside resistance development of CF chronic infections (Guenard et al., 2014), our data using PA14, as well as a recent experimental study using PAO1 (Lopez-Causape et al., 2018b), show that this does not necessarily always occurs, at least in vitro.

Consistent with a proposed role of ptsP in low-level tobramycin resistance (Schurek et al., 2008), the present results indicate that its mutation might also be important, because of its presence in 2 out of 4 tobramycin replicates. In fact, the MIC clearly increased when this mutation appeared alone (population 2, 8–24 µg/ml; **Figure 3** and **Supplementary Table S3**). Whether or not this type of mutation is selected for in clinical settings deserves further investigation. Finally, FleQ is a major flagellar regulator that has been found responsive to c-di-GMP and plays a role in biofilm formation (Hickman and Harwood, 2008; Jimenez et al., 2012; Matsuyama et al., 2016), although it has not been previously described to be involved in antibiotic resistance in planktonic cells.

It has been previously shown that mutations in at least 135 genes can render low-level aminoglycoside resistance in P. aeruginosa (Schurek et al., 2008). Despite the large number of possible combinations that might be selected along evolution under antibiotic selective pressure, we found that all four evolved populations contained mutants in orfN, fusA and pmrB and two of these evolved populations also presented mutants in ptsP. Further, recent work performed independently in another laboratory has also shown that fusA and pmrB mutants are

frequently selected in presence of aminoglycosides (Lopez-Causape et al., 2018b). Taking into consideration that the number of possible combinations of five mutations from the above mentioned 135 genes exceeds 3.4E8, our results indicate that P. aeruginosa mutation-driven resistance to tobramycin displays a certain degree of predictability.

## Evolutionary Pathways Toward Tigecycline Resistance in P. aeruginosa

Pseudomonas aeruginosa PA14 evolving under selective pressure from tigecycline accumulated resistance mutations, indicating that, even in intrinsically resistant microorganisms, resistance may increase. Among the selected mutated genes, three – rpsJ, parRS and nfxB - have already been described involved in antibiotic resistance (a result that validates the present experimental strategy). The Val57 change in the 30S ribosomal protein S10 (rpsJ), which was present in two out of four replicates, is commonly seen in tetracycline-resistant clinical isolates of pathogens such as Neisseria gonorrhoeae, (Hu et al., 2005) while in Klebsiella pneumoniae different mutations in this gene have been related to tigecycline resistance (Ahn et al., 2016). Further, the two-component parRS system has been extensively examined in association with resistance to various drugs. In our experimental evolution assay, three out of the four tigecycline evolved populations presented mutations in this two-component system (**Figure 3**), which may be indicative of its importance in resistance. ParRS may promote multidrug resistance via three mechanisms: by inducing the expression of efflux pumps (such as MexXY-OprM), by repressing the expression of porins, and via LPS alteration (Muller et al., 2011).

The transcriptional repressor of the multidrug efflux pump MexCD-OprJ (Purssell and Poole, 2013), one of the most clinically important of all efflux pumps in P. aeruginosa, is coded for by nfxB. Loss-of-function mutations in this gene result in the pump's overexpression. Different mutations in nfxB were selected in all four replicates of the evolving populations, and most were fully inactivating mutations (a 377 bp deletion leading to a truncated protein; a transposition; and an 11 bp deletion). Although it has been suggested that the main efflux pump contributing to tigecycline resistance in P. aeruginosa is MexXY-OprM (Masuda et al., 2000a; Dean et al., 2003), the first step along the path to resistance in the present work was seen to be the selection of mutants able to extrude tigecycline via MexCD-OprJ. To ascertain whether or not nfxB selected mutations inactivate the repressor, hence allowing MexCD overexpression, one individual clone from each 5 days tigecycline evolved population was isolated in MHA. The presence of their corresponding nfxB mutations and the absence of orfN modification were confirmed by Sanger-sequencing. Expression

of mexC was measured in comparison with the one of the wildtype strain. As shown in **Figure 4**, all clones carrying nfxB mutations overexpressed mexC, confirming that these mutations inactivate, in different degree, the NfxB repressor. Two further mutations in each of the genes coding for the subunits of this efflux pump (mexC and mexD) were then selected for in populations 5 and 6 respectively. It may be that these mutations alter the specificity of MexCD-OprJ improving the capacity of the pump to extrude tigecycline. The MexD SNV (Phe608Cys) is located in one of the two large periplasmic loops known to be involved in substrate specificity (Elkins and Nikaido, 2002). In agreement with this, it has been described that mutations in AcrB, the Enterobacteriaceae ortholog of MexD, alter the pump's substrate profile (Blair et al., 2015). While a role for these mutations in reducing the susceptibility to tigecycline is a compelling hypothesis, it cannot be ruled out that mutations in mexD or mexC may compensate for any increased nonphysiological extrusion of some important cellular metabolite by those mutants that overexpress this pump.

The present results indicate that tigecycline resistance might also come about through changes in the Sec system (SecA and SecG), likely via the modification of the permeability of the membrane and the impairment of tigecycline uptake. It has also been suggested that the Sec pathway could be involved in the translocation of components of multidrug efflux pumps across the bacterial membranes (Yoneyama et al., 2010; Akiba et al., 2013; Chaudhary et al., 2015), but it is unclear how changes in this translocation could increase resistance in the selected mutants. Tigecycline increased resistance may also come about through modifications of the σ<sup>54</sup> RpoN factor that has been found to modulate P. aeruginosa virulence (Kazmierczak et al., 2005), bacterial tolerance to carbapenems (Viducic et al., 2016), susceptibility to quinolones (Viducic et al., 2007) and survival in the presence of tobramycin (Viducic et al., 2017). Both the Sec pathway and RpoN have been proposed as excellent targets in the search of novel antibiotics (Jin et al., 2016; Lloyd et al., 2017). Our findings suggest that tigecyline might select mutants presenting cross-resistance to these potential inhibitors, still under development. Mutations in the gene encoding the ribosome recycling factor frr are likely related to modifications in the tigecycline target (ribosome). Indeed, this particular mutation is located in the start codon (**Supplementary Table S1**), so it may affect the level of frr, impairing the steady state amount of active ribosomes. A similar situation might arise with PA14\_00180, which codes for a putative rRNA small subunit methyltransferase, hence likely able to modify the ribosome.

All the changes seen during the evolution of the populations subjected to selective pressure from tigecycline indicate that evolutionary pathways toward tigecycline resistance present some common features. All four replicates present mutations in orfN and nfxB and mutations in parRS were selected in three out of the four populations. Mutations at secAG, PA14\_00180, rpsJ or mexCD were selected in half (two) of the populations. This indicates that P. aeruginosa can develop resistance to tigecycline by following a limited number of different evolutionary trajectories, which share some specific type of mutations, suggesting a certain degree of determinism.

In addition of conferring resistance to the selecting antibiotics, the respective mutants also showed resistance to antibiotics belonging to other structural categories and with different targets. This might be explained, at least in part, via the important role that efflux pumps seem to play in the development of resistance. Indeed, all mutants selected in the presence of tigecycline showed mutations in nfxB, which would lead to the overexpression of MexCD-OprJ. In turn this might reduce susceptibility to quinolones, tetracyclines, beta-lactams and chloramphenicol (De Kievit et al., 2001) – phenotypes in agreement with the present data (**Table 1**). Moreover, all four population replicates analyzed showed reduced susceptibility to aminoglycosides. This phenotype may be explained in three populations as a result of the mutations selected for in the ParRS system, which is involved in modification of the lipopolysaccharide and in the regulation of the expression of MexXY-OprM, an efflux pump that contributes to aminoglycoside resistance in P. aeruginosa (Masuda et al., 2000a), and which is described as being regulated by the two-component system ParRS (Muller et al., 2011).

## CONCLUSION

Experimental evolution approaches may allow determining basic aspects of evolution, among which knowing to what extent evolution (in our case of asexual organisms) can be predictable and hence deterministic or is basically stochastic (non-predictable) (Lassig et al., 2017), is notably interesting. This is particularly relevant in the case of antibiotic resistance, a field in which the implementation of novel therapeutic approaches is based in the assumption that evolution of antibiotic resistance can be largely predictable (Martinez et al., 2007), a feature that goes against most common views on evolution.

The present work provides information on the evolutionary trajectories leading to resistance to antibiotics belonging to different structural families but targeting the ribosome in P. aeruginosa. The results suggest that, although mutations in several genes may contribute to the development of antibiotic resistance (Fajardo et al., 2008; Tamae et al., 2008; Alvarez-Ortega et al., 2011; Martinez, 2012; Vestergaard et al., 2016; Lopez-Causape et al., 2018a), which may imply a large degree of stochasticity in the evolutionary trajectories, mutation-driven evolution toward resistance is partially deterministic, at least when bacteria grow in the same conditions. This opens up the possibility of predicting the appearance of antibiotic resistance (Martinez et al., 2007).

The selection of specific mutations depends primarily on the fitness of each mutant before/after selection and the rate of mutation supply for each of the mutations (Huseby et al., 2017; Hughes and Andersson, 2017). Recent work has shown that a high generalized mutation supply does not alter the types of mutations selected upon experimental evolution (Ibacache-Quiroga et al., 2018), in which case fitness costs, together with population bottlenecks (Vogwill et al., 2016), will be the main constrains for selecting some mutants over other mutations able of providing the same resistance phenotype. However, the situation might be different in the case of gene-specific high mutation rates, a situation that might have happened in the case of orfN. To note here that all mutations in orfN have been selected in a poly-G region, present in the P. aeruginosa PA14 strain, a situation that might increase the mutation rate as the consequence of polymerase slippage (Moxon et al., 2006). We hypothesize that this gene-specific high mutational supply might be in the basis of the presence of orfN mutants in all replicates. However, OrfN sequence is highly polymorphic in P. aeruginosa (Arora et al., 2004) becoming doubtful if these results can be extrapolated to strains with different orfN alleles.

The present results clearly show that challenging P. aeruginosa with tigecycline has clinically relevant consequences, despite this bacterial pathogen is considered to be intrinsically resistant to this antibiotic. Tigecycline is used for empiric treatment of Gram-negative infections, and in this type of patients, the main agent causing superinfection is P. aeruginosa (Garcia-Cabrera et al., 2010; Ulu-Kilic et al., 2015; Katsiari et al., 2016), which is most likely under tigecycline selection during treatment. Our results indicate that tigecycline selects mutants presenting reduced susceptibility to antibiotics of clinical value as aztreonam, ceftazidime, ciprofloxacin or aminoglycosides. Notably, these mutants also display an increased susceptibility to fosfomycin. Our results suggest that fosfomycin might be an antibiotic of choice for treating superinfections by P. aeruginosa in tigecycline treated patients. A recent work has shown that

#### REFERENCES

Ahn, C., Yoon, S. S., Yong, T. S., Jeong, S. H., and Lee, K. (2016). The resistance mechanism and clonal distribution of tigecycline-nonsusceptible Klebsiella pneumoniae isolates in Korea. Yonsei Med. J. 57, 641–646. doi: 10.3349/ymj. 2016.57.3.641

most antibiotic resistance mutations display strain-independent phenotypes (Knopp and Andersson, 2018). Nevertheless, it is also true that the current work has been performed with a single, model strain and, because of the diversity of P. aeruginosa, it would be important to examine additional strains to establish whether this phenotype depends on the genomic context or can be extrapolated to other P. aeruginosa clinical isolates.

Antibiotic resistance can be acquired by the acquisition of antibiotic resistance genes, which usually confer resistance to members of the same family of antibiotics. However, particularly relevant in the case of chronic infections is the selection of antibiotic resistant mutants. Our results indicate that antibiotic resistance mutations frequently have a pleiotropic effect, altering susceptibility to other drugs. In agreement with this is the fact that some of the mutations selected for by exposure to tobramycin or tigecycline have previously been described to confer resistance to other antibiotics belonging to different structural families.

#### AUTHOR CONTRIBUTIONS

FS-G and SH-A performed the experiments. JM and SH-A designed the work. All the authors participated in the interpretation of the results and in writing the article.

## FUNDING

Work in our laboratory is supported by grants from the Instituto de Salud Carlos III (Spanish Network for Research on Infectious Diseases [RD16/0016/0011]), from the Spanish Ministry of Economy and Competitiveness (BIO2017-83128-R and BIO2014-54507-R) CNB, and from the Autonomous Community of Madrid (B2017/BMD-3691). FS-G is the recipient of an FPU fellowship from the Spanish Ministry of Education. The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

## ACKNOWLEDGMENTS

We thank Adrian Burton (www.physicalevidence.es) for editing the manuscript.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fgene. 2018.00451/full#supplementary-material

Akiba, K., Ando, T., Isogai, E., Nakae, T., and Yoneyama, H. (2013). Tat pathwaymediated translocation of the Sec pathway substrate OprM, an outer membrane subunit of the resistance nodulation division xenobiotic extrusion pumps, in Pseudomonas Aeruginosa. Chemotherapy 59, 129–137. doi: 10.1159/000353894

Alvarez-Ortega, C., Wiegand, I., Olivares, J., Hancock, R. E., and Martinez, J. L. (2010). Genetic determinants involved in the susceptibility of Pseudomonas

aeruginosa to beta-lactam antibiotics. Antimicrob. Agents Chemother. 54, 4159– 4167. doi: 10.1128/AAC.00257-10




in Pseudomonas aeruginosa. Antimicrob. Agents Chemother. 54, 1492–1497. doi: 10.1128/aac.01495-09

**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 © 2018 Sanz-García, Hernando-Amado and Martínez. 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) and the copyright owner(s) 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.

# Insights Into the Evolution of Staphylococcus aureus Daptomycin Resistance From an in vitro Bioreactor Model

Erica Lasek-Nesselquist<sup>1</sup> \*, Jackson Lu<sup>2</sup> , Ryan Schneider<sup>3</sup> , Zhuo Ma<sup>2</sup> , Vincenzo Russo<sup>2</sup> , Smruti Mishra<sup>2</sup> , Manjunath P. Pai<sup>4</sup> , Janice D. Pata1,3, Kathleen A. McDonough1,3 and Meenakshi Malik<sup>2</sup>

<sup>1</sup> Wadsworth Center, New York State Department of Health, Albany, NY, United States, <sup>2</sup> Albany College of Pharmacy and Health Sciences, Albany, NY, United States, <sup>3</sup> Department of Biomedical Sciences, University at Albany, School of Public Health, Albany, NY, United States, <sup>4</sup> Department of Clinical Pharmacy, University of Michigan, Ann Arbor, MI, United States

#### Edited by:

Simona Pollini, Università degli Studi di Firenze, Italy

#### Reviewed by:

Soojin Yang, Chung-Ang University, South Korea Juan M. Pericas, University Hospital Arnau de Vilanova, Spain

#### \*Correspondence:

Erica Lasek-Nesselquist erica.lasek-nesselquist@health.ny.gov

#### Specialty section:

This article was submitted to Evolutionary and Genomic Microbiology, a section of the journal Frontiers in Microbiology

Received: 24 July 2018 Accepted: 08 February 2019 Published: 28 February 2019

#### Citation:

Lasek-Nesselquist E, Lu J, Schneider R, Ma Z, Russo V, Mishra S, Pai MP, Pata JD, McDonough KA and Malik M (2019) Insights Into the Evolution of Staphylococcus aureus Daptomycin Resistance From an in vitro Bioreactor Model. Front. Microbiol. 10:345. doi: 10.3389/fmicb.2019.00345 The extensive use of daptomycin for treating complex methicillin-resistant Staphylococcus aureus infections has led to the emergence of daptomycin-resistant strains. Although genomic studies have identified mutations associated with daptomycin resistance, they have not necessarily provided insight into the evolution and hierarchy of genetic changes that confer resistance, particularly as antibiotic concentrations are increased. Additionally, plate-dependent in vitro analyses that passage bacteria in the presence of antibiotics can induce selective pressures unrelated to antibiotic exposure. We established a continuous culture bioreactor model that exposes S. aureus strain N315 to increasing concentrations of daptomycin without the confounding effects of nutritional depletion to further understand the evolution of drug resistance and validate the bioreactor as a method that produces clinically relevant results. Samples were collected every 24 h for a period of 14 days and minimum inhibitory concentrations were determined to monitor the acquisition of daptomycin resistance. The collected samples were then subjected to whole genome sequencing. The development of daptomycin resistance in N315 was associated with previously identified mutations in genes coding for proteins that alter cell membrane charge and composition. Although genes involved in metabolic functions were also targets of mutation, the common route to resistance relied on a combination of mutations at a few key loci. Tracking the frequency of each mutation throughout the experiment revealed that mutations need not arise progressively in response to increasing antibiotic concentrations and that most mutations were present at low levels within populations earlier than would be recorded based on single-nucleotide polymorphism (SNP) filtering criteria. In contrast, a serial-passaged population showed only one mutation in a gene associated with resistance and provided limited detail on the changes that occur upon exposure to higher drug dosages. To conclude, this study demonstrates the successful in vitro modeling of antibiotic resistance in a bioreactor and highlights the evolutionary paths associated with the acquisition of daptomycin non-susceptibility.

Keywords: Staphylococcus aureus, bioreactor culture, daptomycin, whole-genome sequencing analysis, evolution of resistance

## INTRODUCTION

fmicb-10-00345 February 26, 2019 Time: 15:3 # 2

Staphylococcus aureus is one of the pathogens most commonly associated with both community and hospital acquired infections (Chambers and DeLeo, 2009; Leclercq, 2009; Patel, 2009). Antibiotic management of infections caused by S. aureus is complicated by the emergence of methicillin resistant S. aureus (MRSA), vancomycin intermediate S. aureus (VISA), and vancomycin resistant S. aureus (VRSA) (Chambers and DeLeo, 2009; Howden et al., 2010; Bayer et al., 2013). More recently, daptomycin nonsusceptible (according to Clinical Laboratory Standards Institute 2016 guidelines) strains of S. aureus (herein referred to as DAP resistant or DAP-R) have emerged as daptomycin increasingly becomes the drug of choice for infections caused by MRSA and VISA strains (Marty et al., 2006; Skiest, 2006; Fowler et al., 2007; Jones et al., 2008; Stryjewski and Corey, 2014).

Previous studies have described the acquisition of DAP-R by S. aureus as a complex multifactorial process (Chen et al., 2015; Tran et al., 2015; Kang et al., 2017), with contributions from expression changes and mutations in genes involved in cell wall and cell membrane homeostasis (Friedman et al., 2006; Camargo et al., 2008; Yang et al., 2009; Fischer et al., 2011; Patel et al., 2011; Bayer et al., 2014, 2015; Capone et al., 2016; Kang et al., 2017; Ma et al., 2018; Sabat et al., 2018). The majority of these studies characterized transcriptional changes and/or genetic mutations in clinical DAP-R isolates of S. aureus that emerged as a result of prolonged treatment with (1) daptomycin alone, (2) other antibiotics (typically vancomycin) followed by the administration of daptomycin, or (3) other antibiotics with no daptomycin treatment (Friedman et al., 2006; Murthy et al., 2008; Mishra et al., 2009, 2011, 2014; Yang et al., 2009; Howden et al., 2010; Boyle-Vavra et al., 2011; Fischer et al., 2011; Cameron et al., 2012; Hafer et al., 2012; Peleg et al., 2012; Bayer et al., 2014, 2015; Cafiso et al., 2014; Berti et al., 2015; Gaupp et al., 2015; Capone et al., 2016; Kang et al., 2017; Sabat et al., 2018). Serial passaging of S. aureus on plates or liquid culture by exposing the bacteria to fixed concentrations of antibiotic and identifying the associated molecular and/ or phenotypic changes represents another strategy employed to elucidate the mechanisms of daptomycin resistance (Silverman et al., 2001; Friedman et al., 2006; Mishra et al., 2009, 2012; Patel et al., 2011; Peleg et al., 2012; Berti et al., 2015; Müller et al., 2018). While these studies identify targets of mutation associated with killing by antibiotics, they only provide a snapshot of the evolutionary process after all mutations have been acquired. In the case of clinical isolates there is no knowledge of which mutations accumulate in response to specific levels of treatment while in vitro studies involving serial passage often characterize the response to only a single dose of antibiotic or describe the final DAP-R strain. Similarly, mutations selected in response to vancomycin treatment or other antibiotics do not necessarily represent predominant mutations that would emerge after daptomycin treatment alone. It has been reported that S. aureus strains can develop vancomycin heteroresistance which makes them DAP-R despite never having been exposed to daptomycin (Pillai et al., 2007). Therefore, isolates of S. aureus that were resistant to other antibiotics prior to acquisition of DAP-R do not provide the details of DAP-R evolution in a DAP-S population. Better understanding of the evolution of DAP-R in S. aureus would derive from a system, such as a bioreactor, where a DAP-S bacterial population grown at a steady state is exposed to increasing concentrations of daptomycin, with mutations responsible for daptomycin resistance identified at each step (Toprak et al., 2011, 2013).

The goals of our study were to establish the bioreactor as a tool for realistically modeling the emergence of drug resistance and to use the bioreactor to further elucidate the evolution of daptomycin resistance in susceptible S. aureus populations by overcoming some of the limitations of alternative methods. Our results demonstrate that a continuous culture bioreactor maintains bacterial growth at a steady state while also maintaining the concentration of daptomycin at desired levels. We recovered single nucleotide polymorphisms (SNPs) in genes previously implicated in DAP-R as well as mutations in possible new targets of resistance. In comparison, the serial passaged culture showed only one mutation potentially associated with daptomycin resistance and did not offer the level of insight into mutation frequency fluctuations that the bioreactor populations provided. Our bioreactor results support the hypothesis that resistance at higher daptomycin concentrations requires only a few key changes and that these mutations often emerge much earlier in the history of the population, underscoring the need to examine the path to high minimum inhibitory concentrations (MICs), not just the endpoint.

## MATERIALS AND METHODS

### Bacteria Parental Strain

Staphylococcus aureus strain N315 (ATCC 29213) obtained from BEI Resources (Catalog No. NR-45898), Manassas, VA, United States (Biodefense and Emerging Infections Research Resources)<sup>1</sup> served as the parental DAP-S strain and was propagated in Mueller Hinton Broth (MHB) (BD BBLTM) and frozen in 15% glycerol in 1 ml aliquots for further use. Daptomycin resistant isolates were derived from the parent N315 strain which was consistently exposed to increasing doses of daptomycin in MHB in a bioreactor or in serial passaged flask culture. For every experiment, blood agar plates (Trypticase Soy Agar II with 5% sheep blood, BD Biosciences) were streaked with bacteria from frozen glycerol stocks. After overnight incubation at 37◦C with 5% CO2, a single colony was used to seed the bioreactor or flask culture populations. Three bioreactor replicates (populations A–C) and one serial passaged flask culture experiment (population P) were performed with populations B and C being exposed to 0, 6, 10, and 14 µg/ml daptomycin and populations A and P being exposed to 6 and 10 µg/ml of daptomycin.

#### Daptomycin

Clinical grade Cubicin <sup>R</sup> (injectable daptomycin) from a single lot was purchased from Albany Medical Center Outpatient

<sup>1</sup>https://www.beiresources.org/

Pharmacy, Albany NY, United States. The daptomycin was supplied as 500 mg vials and was reconstituted in sterile water to achieve desired concentrations to be used in various experiments.

#### Serial Passage Flask Culture

Bacterial flask cultures were grown overnight in MHB supplemented with 50 mg/L calcium and 12.5 mg/L magnesium and adjusted to 10<sup>7</sup> CFU/ml concentration. This served as the 0-h time point. After two hours, the concentration of bacteria was roughly 10<sup>8</sup> CFU/ml at which time 6 µg/ml daptomycin was added. Both treated and untreated flasks were grown overnight and subsequently plated on 6 µg/ml daptomycin drug plates. Colonies were isolated, grown overnight, and frozen for further analysis. This process was repeated at 10 µg/ml daptomycin with the colonies from 6 µg/ml drug exposure serving as the seed populations.

### Bioreactor Culture and Adaptation Conditions

The bioreactor system is a continuous culture system in which the rate of inflow (fresh medium) and outflow (spent medium containing culture) is equal, resulting in a constant volume of 400 ml culture in the vessel. The system maintains the bacterial population in exponential growth at a constant density (except after initial additions of antibiotic), as confirmed by CFU counts and optical densities (ODs) of cultures sampled at different time points. MHB was inoculated with a single colony of S. aureus N315 strain and maintained for a period of 14 days at a constant OD<sup>600</sup> of 0.045 (equivalent to 1 × 10<sup>7</sup> CFU/ml) in a 400 ml volume by automated dilution of the culture from feed reservoirs in a Sartorius Stedim Biostat A plus bioreactor. Briefly, thawed glycerol stock of N315 was streaked out on blood agar plates. A single colony was picked to inoculate 20 ml of MHB, which was incubated overnight in an orbital shaker at 37◦C. The next day, the inoculum was diluted 1:1 to make a 40 ml starting culture of 0.4 OD600. The starting culture was added to 360 ml in the bioreactor and then maintained at an OD<sup>600</sup> of 0.45. Daptomycin was added in step-wise increments at a concentration of 6 µg/ml after 24 h, 10 µg/ml after 120 h, and 14 µg/ml after 220/320 h of growth.

#### Isolation of DAP-S and DAP-R Isolates

Samples were collected from the bioreactor at 24, 120, and 220/320 h following exposure to 0, 6, 10, and 14 µg/ml of daptomycin. Samples from population P were collected after exposure to 0, 6, and 10 µg/ml of daptomycin after 24 h of growth. Stocks of both population-based and single colony-based resistant isolates were made in 15% glycerol and stored at −80◦C for further analysis. Colonies were isolated by streaking Mueller-Hinton agar plates supplemented with 50 mg/ml calcium and 6 µg/ml daptomycin.

#### Measurement of Bacterial Growth Rate

Isolates were grown in a shaking incubator at 37◦C. Aliquots were collected at 2 h intervals for a period of 6–8 h for populations A–C. Samples from population P were collected at 0, 2, 4, 6, and 24 h starting from the 0-h time point. The aliquots were serially diluted and plated on sheep blood agar plates to quantify bacterial numbers. The plates were incubated at 37◦C overnight. The colonies were counted and the results were expressed as log<sup>10</sup> CFU/ml.

### Antimicrobial Susceptibility Studies

The MIC values were determined for isolates using either Epsilometer test (E test) or broth microdilution method. The MIC was considered to be the lowest concentration of daptomycin required to inhibit the growth of S. aureus N315. All susceptibility testing was performed using samples in triplicate and each experiment was repeated at least twice. For E tests, bacterial cultures of isolates were grown to OD<sup>600</sup> of 0.08–0.09 in 5 ml cation-adjusted MHB (CAMHB) containing 25 mg/ml calcium and 12.5 mg/ml magnesium. Sterile cotton swabs were used to uniformly streak a lawn of each sample on trypticase soy agar plates containing 5% sheep blood. The cultures were allowed to be completely adsorbed. Using sterile tweezers, the E test strips containing varying concentrations of daptomycin (BioMérieux Inc.) were placed on the streaked agar plates. The plates were incubated at 37◦C for 18–20 h. The elliptical zones of inhibitions were read to determine the MIC values.

The broth dilution method was performed according to the Clinical Laboratory Standards Institute (CLSI) guidelines. Blood agar plates streaked with bioreactor-derived isolates and a control N315 strain were grown overnight. Single colonies were picked, inoculated in 5 ml CAMHB and grown to achieve an OD<sup>600</sup> of 0.08–0.09, which corresponds to 1 × 10<sup>8</sup> CFU/ml. Daptomycin with a starting concentration of 32 µg/ml was serially diluted two-fold in a sterile 96-well plate to achieve a final concentration of 16, 8, 4, 2, 1, 0.5, 0.25, and 0 µg/ml. In accordance with the CLSI standards, bacterial suspension diluted to a final concentration of 5.5 × 10<sup>5</sup> CFU/ml was added to each well of the plate containing varying concentrations of daptomycin. The 96-well plate was incubated for 24 h after which the results were recorded.

#### DNA Extraction and Whole Genome Sequencing

Bacterial pellets from overnight cultures of the reference S. aureus N315 strain, bioreactor-derived isolates and colonies, and serial passaged isolates and colonies were treated with lysostaphin and lysozyme to disrupt the peptidoglycan layer of the cell wall. Genomic DNA was purified using PureLink Genomic DNA Mini Kit (Invitrogen) and samples were sent for whole genome DNA sequencing at Wadsworth Center, New York State Department of Health, Albany, NY, United States. Whole genome sequencing libraries were prepared with the Nextera DNA library preparation kit (Illumina) and sequenced using the standard 500 cycle V2 protocol on an Illumina MiSeq. Whole genome sequences were required to have an average depth of coverage of at least 60 × before analysis.

## Identification of Mutations Associated With Daptomycin Resistance

The subroutine BBDuk from BBtools v36.38<sup>2</sup> quality trimmed raw reads and removed any remaining adaptors with the following parameters: qin = 33, ktrim = r, mink = 11, trimq = 20, minlength = 100, tbp = t, and tpe = t. Sequence libraries for our parent N315 genome and all subsequent populations and clones were reference aligned to the N315 genome from NCBI (Refseq accession number NC\_002745.2) with BWA mem v.0.7.5 (Li and Durbin, 2009). SAMtools and BCFtools 0.1.19 (Li and Durbin, 2009; Li, 2011) detected single nucleotide polymorphisms (SNPs) from each reference alignment – considering only base positions with a Phred score > = 20 and reads with a minimum mapping score of 20. We further filtered the list of SNPs for each population by only examining mutations at positions covered by a depth of 20 or more reads and where 5% or more of the reads supported the alternative nucleotide. We considered a SNP to be present in colonies if positions were covered by 20 or more reads and 95% or more of the reads supported the alternative nucleotide. The presence of each potential SNP and its quality were confirmed in IGV v.2.3.78 (Robinson et al., 2011; Thorvaldsdóttir et al., 2013) and mutations that occurred in highly variable regions (such as phage insertions) were discarded. Custom designed Python scripts associated mutations with coding or non-coding regions to identify changes potentially involved in daptomycin resistance and to translate codons into amino acids.

Accession numbers for sequence data generated from this project have been deposited in GenBank under the BioProject PRJNA446060 http://www.ncbi.nlm.nih.gov/bioproject/446060 and SRA accession SRP136646.

## RESULTS

#### Staphylococcus aureus N315 Acquires Resistance to Daptomycin in a Bioreactor

A 35-fold reduction in bacterial viability was observed following the addition of 6 µg/ml concentration of daptomycin that was corroborated by a sudden drop in the OD<sup>600</sup> values (**Figures 1A,B**). Bacterial numbers returned to the starting concentration of 1 × 10<sup>8</sup> CFUs within the next 24 h of growth (**Figure 1B**) at which time the second bolus of daptomycin at 10 µg/ml was added. The second treatment resulted in a moderate 10-fold drop in bacterial viability, which remained at the same level for next 96–120 h of growth (**Figure 1B**). An identical pattern for OD<sup>600</sup> values was also observed (**Figure 1A**). The addition of the third bolus of 14 µg/ml of daptomycin after 220 h of growth did not reduce bacterial viability. On the contrary, a 5–10-fold increase in bacterial numbers was observed (**Figure 1B**). Corresponding increases in the MIC values were observed in association with enhanced bacterial viability in response to increasing doses of daptomycin (**Figure 1C**). In summary, the parent S. aureus was initially sensitive to the smallest dose (6 µg/ml) of daptomycin but DAP resistant S. aureus emerged, which progressively exhibited an enhanced resistance to increasing exposures of daptomycin.

### The Bioreactor Recovers Clinically Relevant Mutations Associated With Daptomycin Resistance

Our parental N315 genome (sequenced after acquisition from BEI but before introduction to the bioreactor or serial passage) deviated from the NCBI N315 strain (NC\_002745.2) by eight SNPs, which were present in all bioreactor passaged populations (A–C) and clonal lines (**Supplementary Table S1**) and the serial passaged population P and its clonal lines (with the exception of the absence of the substitution at position 939485 in population C; **Supplementary Table S1**).

All three bioreactor populations (A–C) evolved mutations at the phosphatidylglycerol lysyl-transferase mprF gene (SA1193) after exposure to daptomycin. Populations B and C also exhibited mutations at the cardiolipin synthase cls2 locus (SA1891; **Table 1**). These mutations, as well as the SNPs retrieved from the CDP-diacyclycerol-glycerol-3 phosphate 3-phosphatidyltransferase pgsA gene (SA0828) and the PAS sensor histidine kinase walK (previously yycG) gene (SA0018), have been recorded in clinical or laboratory DAP-R strains (Friedman et al., 2006; Howden et al., 2011; Peleg et al., 2012; Bayer et al., 2015; Chen et al., 2015; Kang et al., 2017). We also detected mutations in the clpX gene (SA1498) and the ABC transporter permease gene vraG gene (SA0617), which have been shown to vary in response to daptomycin exposure (Cameron et al., 2012; Peleg et al., 2012; Song et al., 2013).

The bioreactor also recovered new mutations that might be candidates for contributing to daptomycin resistance but require further exploration, including the nonsynonymous substitutions in the genes coding for an ATP-dependent helicase and the pur operon repressor and an intergenic mutation near the DNA-3-methladenine glycosylase locus (**Table 1**). Serial passaged population P showed only three mutations after exposure to 10 µg/ml daptomycin and only the SNP in the pgsA gene occurred at a locus previously associated with daptomycin resistance in laboratory strains (**Table 1**; Peleg et al., 2012).

## Frequency Profiles Reveal Early Appearance of Mutations in Bioreactor Experiments

Tracking the frequency of mutations throughout the course of the bioreactor experiments revealed that many SNPs appeared earlier than were identified by our filtering criteria (See example: **Figure 2**). Mutations present at very low frequencies at no or low daptomycin concentrations were often observed reaching high frequencies at 10 or 14 µg/mL daptomycin in population B. This suggests that variants observed in less than 5% of the population, which did not pass our reporting threshold, actually represent real mutants present in earlier populations (**Supplementary Table S2**.

<sup>2</sup>https://sourceforge.net/projects/bbmap



<sup>a</sup>Pop, Bioreactor and serial passaged populations; <sup>b</sup>mutations detected in > 5% but < 20% of population as assessed by read depth; <sup>c</sup>Dap exposure, the concentration of daptomycin at which the mutation was recorded to have occurred in the population – 0, 0 µg/ml daptomycin; 6, 6 µg/ml daptomycin; 10, 10 µg/ml daptomycin; 14, 14 µg/ml daptomycin; \*dchange to a stop codon; <sup>e</sup>Gene abbreviations included with the gene names highlight genes with previously determined associations to DAP resistance in clinical and/or laboratory strains of S. aureus; <sup>f</sup>Protein product descriptions for intergenic mutations represent the closest gene to the intergenic SNP.

For example, the SNP at position 2143178 (in cls2) in population B was present in 3% of the reads at 6 µg/ml daptomycin (**Figure 2**), but was not recorded at this time point because it did not pass our 5% reporting threshold. However, this SNP increased to 15% at 10 µg/ml and almost reached fixation at 14 µg/ml (represented in 93% of the population; **Supplementary Table S2** and **Figure 2**). Similarly, the SNPs at positions 26957, 261590, 1364495, and 2399255 in population B (in genes walK, SA0219, mprF, and the noncoding region near SA2134, respectively) reached intermediate frequencies at 10 µg/ml daptomycin but were actually present earlier in < = 5% of population (**Supplementary Table S2**). Although population C appeared relatively stable until 14 µg/ml, some SNPs were already present at low frequencies at 0, 6, and 10 µg/ml daptomycin concentrations (**Supplementary Table S2** and **Figure 2**). In contrast, serial passaged population P showed no indication of mutations until after the highest drug exposure of 10 µg/ml (**Supplementary Table S2**. Thus, sampling throughout drug exposures in the bioreactor revealed that mutations relevant to resistance at high

profiles of SA0610, SA2134, SA1228, SA\_RS07380, SA2023, and SA2062.

antibiotic concentrations often appeared much earlier in the history of the population.

#### Colonies Confirm Presence of Low Frequency Variants

The three colonies sequenced from population A at 6 µg/ml and populations B and P 10 µg/ml daptomycin exposure helped to confirm the presence of several SNPs in the populations from which they derived (**Supplementary Table S3**). Mutations present in the two populations but absent in their colonies most likely reflect the heterogeneous composition of the populations, but could also indicate poor recovery of mutant cells on plates or represent false-positive SNPs. For example, the nonsynonymous mutation in the pgsA gene was detected in 68% of the reads in population A at 6 µg/ml DAP (**Supplementary Table S2**) but did not appear in any of the three colonies derived from this population (**Supplementary Tables S2**, **S3**). The relatively high number of reads with this SNP, the fact that the same mutation occurred in population A at 10 µg/ml DAP exposure, and the fact that this SNP has been documented in other DAP resistant S. aureus strains (Peleg et al., 2012) all suggest that this mutation is real but inhibited growth on plates or was not represented by the small sampling size in the colonies. The SNP at position 1706314 was present in only 8% of population B but occurred in 100% of the reads for two of the three colonies (**Supplementary Table S2**). Similarly, a mutation present at position 352545 in the ABC transporter permease gene SA0296 occurred in only 5% of

the reads for population P (not reaching our cutoff threshold) but was present in 100% of the reads for colony 3 from this population (**Supplementary Table S2**).

### DISCUSSION

We describe a novel approach to studying the accumulation of genetic mutations that leads to enhanced daptomycin resistance in S. aureus N315, a highly DAP-S strain. We employed a continuous culture model using a bioreactor which eliminates the conditions of nutrient limitation and allows bacterial replication to occur unhindered except for antibiotic exposure. Thus, growth of DAP-S strains in a bioreactor continuously exposes the bacteria to a known increasing concentration of daptomycin that constantly challenges the bacterial population and selects for changes related to antibiotic pressure (Toprak et al., 2011, 2013). Additionally, the bioreactor model allows us to continuously monitor bacterial growth in a way that is not possible with hollow fiber or plate dependent methods. The consistently increasing concentration of daptomycin in a bioreactor simulates clinical situations where the bacteria are present in a nutritionally rich environment in pockets where the drug concentration may not reach lethal bactericidal levels. This differential distribution of antibiotic allows some bacteria to persist and selects for DAP-R S. aureus strains that are refractory to killing when exposed to higher concentrations of antibiotics.

S. aureus populations subjected to increasing concentrations of daptomycin in the bioreactor displayed mutations at loci associated with resistance in clinically and laboratory-derived isolates (Friedman et al., 2006; Howden et al., 2011; Peleg et al., 2012; Bayer et al., 2015; Chen et al., 2015; Kang et al., 2017; Sabat et al., 2018), thereby demonstrating the utility of the bioreactor for studying the in vivo and in vitro evolution of daptomycin resistance. The three experimental replicates (populations A–C) revealed that the bioreactor yields consistent, reproducible results as evidenced by the repeated evolution of several genetic mutations in more than one population. This supports the hypothesis that daptomycin resistance can occur through multiple routes but evolution converges on and targets a few main pathways.

We also detected SNPs at or near new loci that might contribute to daptomycin resistance, such as a nonsynonymous change in a DNA helicase and a mutation in the noncoding region near DNA-3-methladenine glycosylase (**Table 1**). These two genes code for proteins that have DNA repair functions and could either assist in mediating mutational load or facilitate mutation depending on gain or loss of function. Mutations in genes that code for proteins involved in oxidative phosphorylation or electron transport might contribute to the altered metabolic activities that accompany the acquisition of antibiotic resistance in S. aureus (Somerville and Proctor, 2009; Proctor et al., 2014). In population C, amino acid changes to NADH dehydrogenase (coded for by SA0802), which transfers electrons to menaquinone in the electron transport chain, and MenH (coded for by SA0897), which functions in menaquinone synthesis, could reflect the presence of small-colony variants that we isolated from the bioreactor. These small-colony variants display deficient electron transport chains and are associated with persistent, antibiotic resistant infections (Somerville and Proctor, 2009; Proctor et al., 2014). Metabolic shifts in DAP-R strains were associated with a decrease in the TCA cycle and an increased production of pyrimidines and purines that diverted carbon to the synthesis of cell wall components such as teichoic acid and peptidoglycan (Gaupp et al., 2015). Transcriptomic profiling revealed that the pur operon – responsible for purine synthesis – was strongly regulated in DAP-R strains (Müller et al., 2018). Correspondingly, we observed mutations in a transcription factor that controls purine biosynthesis (SA0454) as well as changes in metabolic genes involved in glucose metabolism (SA0219), amino acid synthesis (SA0756), and steroid biosynthesis and degradation (SA0613; **Table 1**). While serial passaged population P evolved resistance to daptomycin, the SNP in pgsA – the only mutation associated with this condition – arose after exposure to the highest concentration of daptomycin. This suggests that transcriptional changes mediated resistance at lower concentrations and illustrates the different results produced by the two culture methods.

Although populations A–C independently acquired many of the same SNPs or mutations at the same loci, the evolutionary landscapes explored were varied and the mutation profiles were distinct (e.g., **Figure 2**). However, all three populations converged upon a combination of changes – either at pgsA and mprF or mprF and cls2 – supporting our hypothesis that the presence of only a few key adaptations drives daptomycin resistance in S. aureus N315. Both cardiolipin synthase (coded for by cls2) and CDP-diacylglycerol-glycerol-3-phosphatidyltransferase (coded for by pgsA) are involved in the production of membrane phospholipids (Peleg et al., 2012; Bayer et al., 2013). Thus, modifications to these enzymes could influence daptomycin binding and/or translocation across the membrane via compositional changes or changes to membrane fluidity or charge (Peleg et al., 2012; Bayer et al., 2013). The MprF enzyme lysinylates PG – a major cell membrane component for S. aureus – and translocates it to the outer membrane, which influences the charge of the membrane and could ultimately cause repulsion of the DAP-Ca2<sup>+</sup> complex (Ernst et al., 2009). The presence of mutations in phospholipid biosynthesis genes also represented the unifying characteristic across 21 clinical and 12 laboratory DAP-R isolates, underscoring the importance of a few key genes in providing initial resistance as well as the genomic background capable of surviving increased antibiotic exposure (Peleg et al., 2012). While mutations at mprF, rpoB/C, and yycG (walK) loci consistently arose in (varying) sequential order from serial passage experiments (Friedman et al., 2006) and mutations also accumulated sequentially in vancomycin resistant clinical isolates (Mwangi et al., 2007), our results highlight the fact that the evolution of daptomycin resistance does not necessarily rely on a linear addition of genomic mutations. Furthermore, the bioreactor revealed that most mutations arose in the population at an earlier time point than might be recorded from sequencing only the final DAP-R population, a level of detail that our serial passage experiment could not provide. For example, some SNPs

present in less than 5% of the population, reached fixation by the end of the experiment. Additionally, the presence of low frequency variants in colonies plated from the DAP-R populations supports the hypothesis that most low-frequency variants were likely real and could be selected for in subsequent generations. The discrepancies between our results and past experiments could reflect strain specific responses or the effects of bioreactor versus plate-dependent growth. Additional bioreactor experiments with different S. aureus strains would help to clarify whether the trends we observed are more generalizable. The early appearance of low frequency variants suggests that there could be pre-existing mutations within the population that are selected for at higher concentrations of daptomycin. However, the appearance of additional low frequency variants at later time points, such as in Population C, presents the possibility that mutagenesis induced by drug exposure also plays a role in generating variation, a hypothesis that requires further experimentation as well.

While bioreactor populations consistently evolved pairs of mutations affiliated with daptomycin resistance, not all mutation combinations appear to be created equally as demonstrated by the different fates of mutations at mprF and cls2 loci in populations B and C. At 10 µg/ml, population B contained two SNPs at the mprF locus (herein referred to as mprF S295L and L341F) represented in approximately equal proportions and two SNPs at the cls2 locus (herein referred to as cls2 A23V and L52F) also represented in equal proportions (**Figure 2**). However, after increasing the antibiotic concentration, mprF S295L and cls2 A23V were represented in almost 100% of the population while mprF L341F and cls2 L52F were essentially extinct (**Figure 2**). The simultaneous rise of mprF S295L and cls2 A23V with the corresponding extinction of mprF L341F and cls2 L52F suggests that two DAP-R subpopulations existed with the first conferring a selective advantage (**Figure 2**). Population C displayed the same cls2 L52F mutation as Population B and an S337L mprF mutation (which resides in the same transmembrane domain as position 341). Interestingly, it required an additional 24 h for population C to rebound to the same CFU levels reached by population B at 14 µg/ml. A fourth bioreactor experiment produced a DAP-R population with the same mprF and cls2 mutations as Population C with a nearly identical mutational profile (i.e., SNPs appearing at the same time points with changes in SNP frequencies occurring at the same rates at the same dosages; **Supplementary Figure S1**) and confirmed the delayed rebound in growth after 14 µg/ml daptomycin exposure. Thus, it is tempting to speculate that the behavior of population C would represent the behavior of the mprF L341F/cls2 L52F subpopulation in population B in the absence of competition. However, validation of fitness differences requires additional experiments that directly test the competition among populations with varying mutation combinations.

#### CONCLUSION

In conclusion, the bioreactor model represents a novel, effective method for recapitulating in vivo conditions and inducing DAP-R mutations typically observed in clinical isolates. The bioreactor model reveals that certain key mutations are most likely required to be present to confer daptomycin resistance but the combination of key mutations can vary and mutations do not necessarily arise in a progressive fashion. Furthermore, the competitive advantages of mutation combinations do not appear to be equal even if the mutations arise in the same gene combinations (such as mprF and cls2). To our knowledge, this study is the first to unveil a possible hierarchical dominance to mutation combinations, although additional experimental confirmation is required. The bioreactor model illustrates that the evolution of DAP-R is a dynamic process, a characteristic often not reflected when examining static cultures or isogenic strains. Tracking the collective evolutionary fates of genomic mutations over the course of antibiotic treatment will assist in identifying early signs of resistance and might provide knowledge of which mutations will be more responsive to treatment.

## AUTHOR CONTRIBUTIONS

KM, JP, MP, and MM conceived of the study. MM, JL, ZM, VR, and SM maintained the bioreactor and all cultures. ZM performed DNA extractions while RS submitted DNA for sequencing and contributed to data interpretations. MP generated **Figure 1**. EL-N and MM wrote the manuscript and EL-N performed all whole-genome sequencing analyses. All authors read and approved the final manuscript.

## FUNDING

This work was supported by funding from an "Investment in Collaborative Research at the Wadsworth Center and Albany College of Pharmacy and Health Sciences Grant." JP was supported by NIH R01-GM-080573.

## ACKNOWLEDGMENTS

The following reagent was provided by the Network on Antimicrobial Resistance in Staphylococcus aureus (NARSA) for distribution by BEI Resources, NIAID, NIH: Staphylococcus aureus, Strain N315, NR-45898. The authors acknowledge the assistance of Wadsworth Center Applied Genomic Technologies and Bioinformatics and Statistics Cores.

## SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb.2019. 00345/full#supplementary-material

FIGURE S1 | Mutation profiles of population C and a fourth bioreactor experiment (population D). The frequency of mutations as a function of antibiotic concentration for population C (top) and population D (bottom). Mutations highlighted are SNPs that lead to amino acid changes in MprF and Cls2 and lower frequency variants NADH (NADH dehydrogenase, SA0802), menH (2-succinyl-6-hydroxy-2,

4-cyclohexadiene-1-carboxylate synthase, SA0897), and rpoA (DNA-directed RNA polymerase subunit alpha, SA2023). SNPs are identical between the populations and display the same dynamics (i.e., SNPs appear at the same time and change in frequency at the same rate and at the same dosages).

TABLE S1 | SNPs present in the parental DAP-S isolate of N315 in comparison to the NCBI Refseq N315 isolate. These changes were detected in our BEI parental N315 strain and all descendant populations, except for the mutation at position 939485, which is not present in Population C. Annotations associated with noncoding SNPs are for the nearest coding sequence.

### REFERENCES


TABLE S2 | Allele Frequencies of SNPs for populations 6–8, population P and colonies 1–3 derived from population 6 at 6 µg/ml and populations 7 and P at 10 µg/ml daptomycin. Alt, Alternative (SNP) nucleotide frequency; Reads, total number of reads at a position. All concentrations are of daptomycin. No colonies were sequenced from Population C.

TABLE S3 | SNPs present in colonies from Population A taken at 6 µg/ml daptomycin and populations B and P at 10 µg/ml daptomycin. Three colonies were sequenced from each population. SNPs present in the ancestral population are not shown.


methicillin-resistant Staphylococcus aureus strains: relative roles of mprF and dlt operons. PLoS One 9:e107426. doi: 10.1371/journal.pone.0107426


**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 © 2019 Lasek-Nesselquist, Lu, Schneider, Ma, Russo, Mishra, Pai, Pata, McDonough and Malik. 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) and the copyright owner(s) 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.

# Construction and Characterization of Synthetic Bacterial Community for Experimental Ecology and Evolution

Johannes Cairns<sup>1</sup>† , Roosa Jokela<sup>1</sup>‡ , Jenni Hultman1§, Manu Tamminen2||, Marko Virta1¶ and Teppo Hiltunen1,2 \*

<sup>1</sup> Department of Microbiology, University of Helsinki, Helsinki, Finland, <sup>2</sup> Department of Biology, University of Turku, Turku, Finland

#### Edited by:

Elena Perrin, Università degli Studi di Firenze, Italy

#### Reviewed by:

Henning Seedorf, Temasek Life Sciences Laboratory, Singapore Vasvi Chaudhry, Institute of Microbial Technology (CSIR), India

#### \*Correspondence:

Teppo Hiltunen teppo.hiltunen@helsinki.fi orcid.org/0000-0001-7206-2399

†orcid.org/0000-0003-1329-2025 ‡orcid.org/0000-0002-4144-6573 §orcid.org/0000-0002-3431-1785 ||orcid.org/0000-0001-5891-7653 ¶orcid.org/0000-0001-5981-7566

#### Specialty section:

This article was submitted to Evolutionary and Genomic Microbiology, a section of the journal Frontiers in Genetics

Received: 27 April 2018 Accepted: 23 July 2018 Published: 14 August 2018

#### Citation:

Cairns J, Jokela R, Hultman J, Tamminen M, Virta M and Hiltunen T (2018) Construction and Characterization of Synthetic Bacterial Community for Experimental Ecology and Evolution. Front. Genet. 9:312. doi: 10.3389/fgene.2018.00312 Experimental microbial ecology and evolution have yielded foundational insights into ecological and evolutionary processes using simple microcosm setups and phenotypic assays with one- or two-species model systems. The fields are now increasingly incorporating more complex systems and exploration of the molecular basis of observations. For this purpose, simplified, manageable and well-defined multispecies model systems are required that can be easily investigated using culturing and highthroughput sequencing approaches, bridging the gap between simpler and more complex synthetic or natural systems. Here we address this need by constructing a completely synthetic 33 bacterial strain community that can be cultured in simple laboratory conditions. We provide whole-genome data for all the strains as well as metadata about genomic features and phenotypic traits that allow resolving individual strains by amplicon sequencing and facilitate a variety of envisioned mechanistic studies. We further show that a large proportion of the strains exhibit coexistence in co-culture over serial transfer for 48 days in the absence of any experimental manipulation to maintain diversity. The constructed bacterial community can be a valuable resource in future experimental work.

Keywords: microbial community, model system, synthetic ecology, experimental evolution, whole-genome sequencing

#### INTRODUCTION

Testing ecological and evolutionary theory in a highly controlled manner using simple laboratory setups with one or two microbial species (Fraser and Keddy, 1997; Buckling et al., 2009) has produced important insights into ecological interactions–e.g., competition, cooperation, and crossfeeding interactions (Helling et al., 1987; Treves et al., 1998; Rozen and Lenski, 2000; Shou et al., 2007; Harcombe, 2010); the role of cheaters (MacLean and Gudelj, 2006); predator–prey interactions (Shertzer et al., 2002); and host–parasite interactions (Morgan et al., 2005)–and evolutionary processes–e.g., the evolution of coexistence (Good et al., 2017), coevolution between species (Hall et al., 2011; Brockhurst and Koskella, 2013), and eco-evolutionary feedback dynamics (Yoshida et al., 2003; Hiltunen and Becks, 2014). However, there is an increasing awareness that ecological and evolutionary processes can be fundamentally altered in more complex multispecies communities owing to several features such as altered competitive interactions and multiple selection pressures (Dunham, 2007). Recent empirical findings show, for example, that pairwise

**158**

interactions can be strongly altered in the presence of other species (Kastman et al., 2016) and the rate of adaptation of species can differ between monocultures and communities (Lawrence et al., 2012). Even a basic understanding of certain characteristics of microbial life such as horizontal gene transfer (Smillie et al., 2011), metabolic interactions and spatial heterogeneity (Elias and Banin, 2012; van Gestel et al., 2014) requires investigation of multispecies settings integral to them. Furthermore, several key ecological features are specific to multispecies communities, such as diversity, stability, succession and high-order (e.g., fourway) species interactions (Bairey et al., 2016). There is therefore a profound need to expand the biotic complexity of study systems used in the fields of experimental microbial ecology and evolution.

The design of multispecies model communities in experimental ecology and evolution is part of the emerging field of synthetic ecology where synthetic communities are used for a plethora of basic and applied purposes. Several research attempts have sought mechanistic understanding of specific natural systems, such as methane consuming communities (Yu et al., 2017), plant root colonizing bacteria (Lebeis et al., 2015), the human gut microbiota (Goodman et al., 2011), and cheese rind communities (Wolfe et al., 2014), by complementing observational findings with findings from controlled in vitro or in vivo studies using synthetic communities. These studies focus on designing synthetic communities that capture the essential features of the natural system being investigated. A typical approach is to determine the prevalent taxa in the natural system or in the core microbiome (Shade and Handelsman, 2012) common to similar systems, to construct a synthetic community of taxonomically representative strains isolated from the natural system, and to culture the community in conditions mimicking the natural system. The limitations of this approach include the potentially important role in ecological functions or evolutionary processes of low-abundance taxa (Liu et al., 2017), microdiversity (Chase et al., 2017), or interactions between bacteria and members of other taxonomic groups such as viruses, unicellular eukaryotic predators or fungi, as well as technical difficulties in mimicking natural conditions in the laboratory (Wolfe, 2018). Likely owing to such limitations, among these studies, cases have been observed both where simple synthetic communities representing predominant taxa in natural systems recapitulate the dynamics in natural systems (Goodman et al., 2011; Wolfe et al., 2014; Lebeis et al., 2015) and where major differences are observed between synthetic and natural systems (Yu et al., 2016).

Compared with studies employing synthetic communities to understand specific natural systems, more applied studies focusing, among others, on medical therapeutics (Petrof et al., 2013; Sheth et al., 2016), bioremediation (Dejonghe et al., 2003; Zomorrodi and Segre, 2016), or biofuel production (Wang et al., 2015), rely even more heavily on a detailed understanding of the characteristics and functions of specific bacterial taxa in natural systems to engineer communities that can successfully perform desired functions. In contrast, studies attempting to investigate highly general ecological or evolutionary processes, similar to traditional experimental microbial ecology and evolution using one- or two-species model systems, do not necessarily seek to, or prefer simple culture conditions over the ability to, accurately represent a particular natural community. For instance, a synthetic community of 72 bacterial strains isolated from tree-hole bacterial communities–but limited to aerobic heterotrophs cultivable in simple laboratory conditions–has been used to investigate several key ecological questions, including the relationship between diversity and ecosystem productivity (Bell et al., 2005) and the success of multispecies invasions during different stages of ecological succession (Rivett et al., 2018). Celiker and Gore (2014), in turn, used a completely synthetic model community comprising six apparently random soil bacterial strains from culture collections to examine the repeatability of change in community composition over time.

In simplified synthetic communities, verisimilitude is sacrificed to obtain relative ease of analysis and modeling, and control of species interactions, non-linear effects from added traits and strains, and evolutionary change (Widder et al., 2016). There is ongoing debate about the utility of simple microcosm setups to understand ecological and evolutionary phenomena (Carpenter, 1996, 1999; Fraser and Keddy, 1997; Drenner and Mazumder, 1999; Benton et al., 2007; Buckling et al., 2009), yet the approach continues to produce major scientific discoveries (van Houte et al., 2016; Good et al., 2017; Betts et al., 2018; Frickel et al., 2018). Similarly, the definition of, need for, and necessary level of representativeness of synthetic communities remain matters of debate (Dolinsek et al., 2016; Widder et al., 2016; Zomorrodi and Segre, 2016; Wolfe, 2018), and are likely strongly dependent on the research questions. Although debated, completely synthetic communities composed of strains isolated from different habitats can also be used to study general questions as well as having special use in studying questions such as community assembly and the evolution of coexistence in newly formed communities.

In this context, the detailed, mechanistic understanding of simpler, less representative synthetic communities can be thought to inform, or even be a prerequisite to, understanding more complex synthetic systems, and ultimately, natural systems. It has been argued that such efforts should focus on understanding a limited set of well-defined model synthetic communities, which would make results comparable between studies and allow collaborative efforts toward mechanistic understanding (Widder et al., 2016). However, not many such systems exist to our knowledge despite the general boom in synthetic ecology. Furthermore, the highly general level studies that exist primarily focus on simple phenotypic analyses (Bell et al., 2005; Foster and Bell, 2012; Celiker and Gore, 2014; Rivett et al., 2016, 2018; Rivett and Bell, 2018), although high-throughput molecular methods, such as amplicon sequencing, (meta)genomics, (meta) transcriptomics, proteomics and metabolomics, which have been promisingly utilized in studies focusing on synthetic systems mimicking natural systems (Goodman et al., 2011; Wolfe et al., 2014; Lebeis et al., 2015; Stopnisek et al., 2016; Yu et al., 2017), could provide valuable insights into the mechanisms behind observed phenotypic and community level features.

To address these needs, we here constructed a simplified experimental system comprising a completely synthetic

community of 33 bacterial strains representing two phyla and six classes that can be cultured individually and in co-culture in highly simple laboratory conditions, and that can be individually resolved based on amplicon sequencing. Furthermore, we present draft-level whole genome sequence data as well as information regarding genomic features and phenotypic traits for all strains in the community, facilitating further mechanistic studies. We also present proof of concept for coexistence of a large proportion (14/33) of the strains in co-culture after serial passage of cultures for 48 days. Such a community can be a highly useful resource for experimental microbial ecology and evolution. For instance, we recently used a closely related model system to track the mobility of antibiotic resistance genes in a complex bacterial community (Cairns et al., 2018). For future studies, we envision, for example, exploration of the trajectories of ecosystem composition and genetic structure in response to environmental perturbations or variability in functional trait space. We are also considering using the community as an internal control for improving high-throughput microbial single cell genome sequencing techniques such as epicPCR (Spencer et al., 2016) or metagenome assemblies.

#### MATERIALS AND METHODS

#### Constructing Model Community

Strains from the University of Helsinki culture collection (HAMBI) representing diverse taxa were initially screened for the ability to grow individually at 28◦C in two complex liquid media: the nutrient-rich proteose peptone yeast extract (PPY: 20 g proteose peptone and 2.5 g yeast extract in 1 l deionized H2O) medium or a custom lower-nutrient-level medium containing M9 salt solution and King's B (KB) nutrients at a 1% concentration compared with full-strength medium (concentrations used: 0.2 g peptone number 3 and 0.1 ml of 85% glycerol in 1 l of dH2O), and 0.2 g l−<sup>1</sup> protozoan pellets (Carolina Biological Supply Co., Burlington, United States). Protozoan pellets were prepared by dissolving in dH2O, bringing to boil and filtering through 40 µm to remove particulate matter. Notably, the strains were not selected based on being representative of any particular natural system but rather as a collection of strains, each representing a different taxonomic group, that can grow in simple, uniform laboratory conditions and therefore be easily used to test general ecological and evolutionary theory or techniques (Wolfe, 2018).

Previously, we performed a serial transfer experiment with an initial version of the model community consisting of 62 strains, including the Escherichia coli K-12 strain JE2571(RP4) harboring the multidrug resistance plasmid RP4 (Bradley, 1980; Cairns et al., 2018). The study demonstrates the ability of dozens of the strains to coexist in culture over 40 days, as well as a method for manipulating the level of spatial heterogeneity (biofilm mass) in community cultures. For tracking strain abundance over time in the study, we Sanger sequenced the near-full-length 16S rRNA gene sequences of the strains for use as a reference database for mapping high-throughput sequencing amplicons from experimental samples (Cairns et al., 2018). Based on these previous data, we refined the model community, removing strains with > 97% identity of 16S rRNA gene sequence with another strain in the community or uncertain strain identity. This resulted in a set of 33 gram-negative strains with confirmed culture collection identity that can be tracked at strain-level resolution through amplicon sequencing (**Supplementary Tables S1, S2**). The community was given the acronym HMC 33.1 (standing for HAMBI Mock Community containing 33 strains, version 1). The strains represent three classes in the phylum Proteobacteria (Alpha-, Beta- and Gammaproteobacteria) and three classes in the phylum Bacteroidetes (Chitinophagia, Flavobacteriia and Sphingobacteriia) (**Figure 1**). Each class contains a minimum of two strains, with the most representatives (13 strains) for the class Gammaproteobacteria. The strains have been isolated from diverse environments across the globe (**Supplementary Table S2**). Most of the strains originate from the soil, and the second highest group is human associated bacteria. Other sources include animals, aquatic environments, plants and the industry.

We further refined the co-culture medium to contain 1 g l <sup>−</sup><sup>1</sup> R2A solution and 0.5 g l−<sup>1</sup> of cereal grass medium (Ward's Science, St. Catharines, ON, Canada) in M9 salt solution. Cereal grass medium stock was prepared by autoclaving in dH2O and removing particulate matter by filtering through 5 µm.

To provide proof of concept for the coexistence of the strains in the medium, we performed a 48-day serial transfer experiment. The strains were grown separately for 4 days at 28◦C with constant rotation at 50 rpm in the new medium and pooled together at equal volumes for the serial transfer experiment. The growth curves of strains when growth separately for 4 days are shown in **Supplementary Figure S1**. The stock community was stored in 30% glycerol at –80◦C. The experiment was started by adding 10 µl of 100-fold diluted freeze-stored stock community to wells in a ABgeneTM 96 Well 2.2 ml Polypropylene Deepwell Storage Plate (Thermo Fisher Scientific, Waltham, MA, United States) containing 550 µl of medium. The community was grown at 28◦C with constant rotation at 100 rpm. The experiment was maintained every 96 h by transferring 10% to fresh medium. To determine bacterial density, optical density values at 600 nm were obtained for undiluted grown cultures after each growth cycle (Tecan Infinite M200 well-plate reader), and samples were frozen in glycerol for further analysis.

#### Phenotypic Analyses

Minimum inhibitory concentration (MIC) was determined for each strain for six antibiotics representing different antibiotic classes: ampicillin (class: penam), erythromycin (macrolide), kanamycin (aminoglycoside), nalidixic acid (fluoroquinolone), rifampicin (rifamycin), and streptomycin (aminoglycoside). For MIC determination, bacterial inoculum from an individual colony was suspended in M9 salt solution to 0.5 McFarland standard turbidity. Subsequently, 100 µl was spread-plated on 50% PPY agar medium containing a Liofilchem <sup>R</sup> MIC test strip (Liofilchem, Italy) for a specific antibiotic. The MIC was interpreted according to manufacturer's instructions after culturing at 28◦C for 2 days, or for slow-growing strains, 6 days. Therefore, MICs were estimated in customized standard

conditions rather than strain-specific optimal growth conditions or standard conditions recommended by the European Committee on Antimicrobial Susceptibility Testing (EUCAST) for clinical bacterial isolates (Kahlmeter et al., 2006), and the data might therefore not be comparable with clinical MIC data.

The ability of the strains to utilize 31 different carbon sources was assessed using the Biolog EcoPlate system (Biolog, Inc., Hayward, CA, United States), with the exception of HAMBI 2948 which grew poorly in experimental conditions. The experimental procedure was modified from MacLean et al. (2004). Strains were precultured in liquid PPY medium for 48 or 96 h, depending on strain growth ability, at 28◦C with constant rotation. Subsequently, cells were spinned down, resuspended in M9 salt solution and nutrient-starved for 48–72 h. Starved cells were spinned down to remove any carryover nutrients in the supernatant, and resuspended in fresh M9 salt solution. Following 100-fold dilution into M9 salt solution, 150 µl of the culture was pipetted to wells in an EcoPlate (one EcoPlate per strain) containing three technical replicate wells for each carbon source and a negative control. The ability of a strain to utilize a particular carbon source was interpreted as a significantly higher optical density, based on a t-test, at 590 nm (measured with Tecan Infinite M200 well-plate reader) compared to the negative control after culturing for 7 days at 28◦C.

## DNA Extraction and Sequencing for 16S rRNA Amplicon Analysis

DNA from three technical replicates from the original pooled bacterial community and three replicate communities from days 16, 32, and 48 in the serial transfer experiment was extracted using DNeasy 96 Blood & Tissue Kit (Qiagen, Hilden, Germany). DNA extraction was performed according to the manufacturer's instructions using 400–600 µl of sample. DNA concentrations were measured using the Qubit <sup>R</sup> 3.0 fluorometer (Thermo Fisher Scientific, Waltham, MA, United States).

Paired-end sequencing was performed using the Illumina MiSeq platform at the Institute for Molecular Medicine Finland (FIMM) amplifying the V3 and V4 regions of ribosomal RNA with Phusion High Fidelity PCR Master Mix (Thermo Fisher Scientific, Waltham, MA, United States). Reactions were done with a 2-step PCR method with primers carrying Illumina adapter tails (PCR1: forward 5<sup>0</sup> -ACACTCTTTCCCT ACACGACGCTCTTCCGATCTCTCCTACGGGAGGCAGCAG-3 0 , reverse 5-0AGACGTGTGCTCTTCCGATCTTCTRCGMATT YCACYKCTACAC-3<sup>0</sup> ; PCR2: forward 5<sup>0</sup> -ACACTCTTTCCCT ACACGACGCTCTTCCGATCTGACTACHVGGGTATCTAATC C-3<sup>0</sup> , reverse 5<sup>0</sup> -AGACGTGTGCTCTTCCGATCTCCTACGGGN GGCWGCAG-3<sup>0</sup> ). For PCR2, the primers are the same but the locus-specific part and adapter part have been paired vice versa

to create more diversity in the final sequencing library. Products from PCR1 and PCR2 were pooled together and indexed with two Illumina P5/P7 index primers (every sample had their own unique combination).

PCR amplification was performed in a volume of 20 µl containing approx. 20 ng of sample DNA, 1 µl (5 µM) of each locus-specific primer (final concentration 0.25 µM), and 10 µl of 2 × Phusion High-Fidelity PCR Master Mix, and the reaction mix was brought to a final volume with laboratory grade water. The cycling conditions were as follows: 98◦C for 30 s, 27 cycles of 98◦C for 10 s, 62◦C for 30 s, and 72◦C for 15 s, with a final extension at 72◦C for 10 min, followed by hold at 10◦C.

Sample indexing was performed in a volume of 20 µl containing 1.5 µl (5 µM) of each index primer (final concentration 0.375 µM), 10 µl of 2 × Phusion High-Fidelity PCR Master Mix (Thermo Fisher Scientific, Waltham, MA, United States) and 1 µl of pooled PCR product. The reaction mix was brought to a final volume with laboratory grade water. The cycling conditions were as follows: 98◦C for 30 s, 8 cycles of 98◦C for 10 s, 65◦C for 30 s, and 72◦C for 20 s, with a final extension at 72◦C for 5 min, followed by hold at 10◦C.

After PCR, random samples were measured with LabChip GX Touch HT DNA High Sensitivity Reagent Kit (Perkin Elmer, Waltham, MA, United States) to check that the PCR was successful with the correct product size. The same run was repeated after indexing. Samples were pooled together in equal volumes and purified with Agencourt <sup>R</sup> AMPure <sup>R</sup> XP beads (Beckman Coulter, Brea, CA, United States) twice using 0.8 × volume of beads compared to the sample pool volume (40 µl). The ready amplicon library was diluted to 1:10 and 1:20 and quantified with the Agilent 2100 Bioanalyzer High Sensitivity DNA Analysis Kit (Agilent Genomics, Santa Clara, CA, United States). The 16S rRNA gene amplicon pool was sequenced with Illumina MiSeq System using the Illumina MiSeq Reagent Kit v3 600 cycles kit (Illumina, San Diego, CA, United States). The read length for the paired-end run was 2 × 300 bp.

### DNA Extraction and Sequencing for Whole-Genome Analysis

For this study, 16 of the 33 model community strains were wholegenome sequenced (**Table 1**). For the remaining 17 strains, 16 assembled genomes were obtained from the NCBI Reference Sequence Database (RefSeq) and 1 raw sequence dataset was obtained from the Joint Genome Institute Genomes OnLine Database (JGI GOLD; **Supplementary Table S1**). For the strains sequenced, DNA was extracted from 1 ml of liquid culture using the DNeasy 96 Blood & Tissue Kit (Qiagen, Hilden, Germany) according to manufacturer's instructions. DNA concentrations were measured using the Qubit <sup>R</sup> 3.0 fluorometer (Thermo Fisher Scientific, Waltham, MA, United States). High-throughput sequencing was performed by the Next Generation Sequencing Services, Institute for Molecular Medicine Finland (FIMM). For this, 2.5 ng of dsDNA was prepared according to the Nextera XT DNA Library Prep Kit Reference Guide (Illumina, San Diego, GA, United States) with the following modifications: All reactions were performed in half of the normal volume, and normalization was done according to the concentration measured on LabChip GX Touch HT (PerkinElmer, United States). 530–670 bp fragments were size selected from the pool using BluePippin (Sage Science, United States). Sequencing was performed with the Illumina HiSeq2500 system in the HiSeq high output mode using v4 kits (Illumina, San Diego, CA, United States). The read length for the paired-end run was 2 × 101 bp.

#### 16S rRNA Amplicon Analysis

Sequencing adapters were removed from unpaired sequence data using Cutadapt 1.12 (Martin, 2011), with the parameter –minimum-length 100. Sequence pairing was done with Pear 0.9.11 (Zhang et al., 2014). Quality was assessed before and after Cutadapt and Pear with FASTQC<sup>1</sup> , and further trimming was done with PRINSEQ (Schmieder and Edwards, 2011) using the parameters -trim\_left 5 and -trim\_right 40 to obtain better quality. Quality filtering was done with the USEARCH 10 (Edgar, 2013) command –fastq\_filter with fastq\_maxee setting 1.0 (to discard all reads with > 1.0 total expected errors). The command –usearch\_global was used to align the sequences to the 16S rRNA fragment database of the strains with > 97% identity. The Shannon diversity index was computed using the USEARCH 10 command -alpha\_div. Difference in community composition between days 16–48 (disregarding initial stock at time point 0 that might not represent viable cell counts for each strain) was assessed in R 3.4.0 (R Core Team, 2017) using permutational multivariate analysis of variance (PERMANOVA) as implemented in the adonis function in the vegan package (Anderson, 2001; Oksanen et al., 2016).

#### Genome Assembly and Annotation

Prior to genome assembly, Cutadapt 1.12 (Martin, 2011) was used to remove sequencing adapters and quality trim sequence data, with the parameters -O 10 (minimum overlap for an adapter match) and -q 28 (quality cutoff for the 3<sup>0</sup> end of each read). Sequence data quality before and after Cutadapt was assessed using FASTQC<sup>1</sup> and multiQC (Ewels et al., 2016). Genome assembly was performed using SPAdes 3.11.1 (Bankevich et al., 2012) with default settings (i.e., no specified parameters since SPAdes performs estimation of e.g., k-mer sizes, coverage cutoff value and PHRED quality offset for input reads). Following assembly, contigs were filtered by minimum coverage 30 and minimum length 1000 bp. Genome assembly quality was assessed using QUAST 4.0 (Gurevich et al., 2013), and by mapping reads back to the assembly using Bowtie 2 with default settings (Langmead and Salzberg, 2012) combined with the Picard<sup>2</sup> command CollectWgsMetrics performed for alignment files after sorting with SAMtools (Li et al., 2009) and marking duplicates with Picard (**Table 1**). Subsequently, all assemblies, including those obtained from databases, were annotated using Prokka 1.12 (Seemann, 2014), providing genus and species names, with the parameters –centre X, –compliant, and –usegenus.

<sup>1</sup>http://www.bioinformatics.babraham.ac.uk/projects/fastqc <sup>2</sup>http://broadinstitute.github.io/picard

TABLE 1 | Genome sequencing and assembly metrics.

fgene-09-00312 August 11, 2018 Time: 19:23 # 6


<sup>∗</sup>HAMBI 1874 was assembled but not sequenced in this study.

FIGURE 2 | Genomic features of model community strains. E. coli K-12, HAMBI 216, HAMBI 1923 and HAMBI 2659 are closed genomes composed of a single chromosome. Other genomes are composed of blocks of adjacent contigs whose order may not reflect genomic position. Contigs are ordered by size for genomes assembled in this study and by order of appearance in assembly file for genomes obtained from databases. Predicted features are depicted by colored blocks along the genome. These include plasmid-derived sequences, genomic islands, prophages, integrons, insertion sequences, antibiotic resistance genes, virulence factors, and CRISPR loci.

#### Phylogenetic Analysis

We identified and aligned genes conserved among all community members using Roary 3.8.0-0 (Page et al., 2015) with minimum BLASTp identity set to 65% and protein group limit set to 200,000 to account for high diversity in the community. The core gene alignment, consisting of four ribosomal proteins (rplN, rpmA, rpsL, and rpsS) and the ATP synthase subunit beta (atpD), was used to create a phylogenetic tree with PhyML (Guindon et al., 2010), visualized in iTol (Letunic and Bork, 2016).

#### Characterization of Genomic Features

To generate comprehensive genomic metadata for model community strains, the genome assemblies were scanned for genomic elements using several tools, with default parameters unless otherwise specified. The Metabolic And Physiological potentiaL Evaluator (MAPLE) 2.3.0 (Takami et al., 2016) was used to map genes in assemblies to functional modules as defined by the Kyoto Encyclopedia of Genes and Genomes (KEGG) and for calculating module completion ratios (MCR).

Genomic islands, i.e., large genomic regions assumed to have horizontal origins, were predicted using IslandViewer 4 (Bertelli et al., 2017). For strains other than E. coli K-12, HAMBI 216, HAMBI 1923 and HAMBI 2659 which have closed genomes, genomic island predictions were obtained for incomplete genomes with contigs ordered against closely related reference genomes (**Supplementary Table S3**) using the Mauve contig orderer (Rissman et al., 2009; Bertelli et al., 2017).

Plasmid-derived sequences were predicted using cBar 1.2 (Zhou and Xu, 2010) and PlasmidFinder 1.3 (Carattoli et al., 2014), which was implemented through ABRicate<sup>3</sup> using databases dated 2018-01-02. In addition, for genomes assembled for this study, plasmidSPAdes (Antipov et al., 2016) implemented in SPAdes 3.11.1 (Bankevich et al., 2012) was used, with assignment of plasmid annotation to > 1000 bp regions in genome assembly contigs displaying 100% BLASTn (Altschul et al., 1990) identity with plasmid assemblies generated by plasmidSPAdes.

Antibiotic resistance and virulence genes were predicted with ABRicate (https://github.com/tseemann/abricate, databases dated 2018-01-02) using ResFinder 3.0 (Zankari et al., 2012), the comprehensive antibiotic resistance database (CARD) (Jia et al., 2017), and the virulence factors database (VFDB) (Chen et al., 2016), discarding hits with proportion of gene covered < 50%. Prophages, integrons, insertion sequences, and clustered regularly interspaced short palindromic repeats (CRISPR) were predicted using the PHAge Search Tool – Enhanced Release (PHASTER; accessed 2018-01-08) (Arndt et al., 2016), Integron Finder 1.5.1 (Cury et al., 2016), ISEScan 1.5.4.3 (Xie and Tang, 2017) and CRISPRFinder (accessed 2018-02-01) (Grissa et al., 2007), respectively.

<sup>3</sup>https://github.com/tseemann/abricate

Notably, in the absence of further validation steps, plasmid predictions obtained with cBar or plasmidSPAdes have low accuracy (Arredondo-Alonso et al., 2017), and genomic island predictions for draft level genomes may contain errors (Bertelli et al., 2017), and the predicted loci must therefore be considered as regions of interest rather than high-confidence predictions.

#### RESULTS

#### Genomic Composition

The strains display high variability in genome size and genomic content, including the resistome, mobilome and functionome (**Figure 2**). Antibiotic resistance and virulence genes are particularly prevalent among strains in the Gammaproteobacteria class (**Figures 1**, **2**). Genomic elements associated with horizontal gene transfer, including plasmids, genomic islands, prophages, integrons, and insertion sequences, occur frequently among the strains, as do CRISPR arrays that might be associated with CRISPR/Cas systems conferring adaptive immunity against mobile genetic elements.

### Metabolic and Physiological Characteristics

The strains display diverse antibiotic susceptibility and carbon source utilization phenotypes (**Figure 3**). Antibiotic resistance phenotypes are variably associated with the presence of bioinformatically predicted resistance factors to the antibiotic class. The number of carbon sources on which the strains can grow displays high variability, with a maximum of 25 out of 31 carbon sources (HAMBI 2159 and 2494, which both belong to the genus Paraburkholderia), indicating large differences between strains in metabolic capacity and fastidiousness. Furthermore, the presence of genomic functional modules for benzoic acid, galactonic acid, galacturonic acid or lactose metabolism is frequently associated with the ability of a strain to grow with a corresponding compound as the sole carbon source (**Figure 3**).

The metabolic characteristics predicted from the genome sequences by MAPLE display hierarchical clustering by

phylogenetic classes of the isolates as well as by their isolation origin (**Figure 4**). Soil isolates typically cluster separately from isolates from humans and animals. Most of this clustering is driven by metabolic functions related to environmental information processing. Core metabolic pathways such as the tricarboxylic acid cycle are present across all isolates while specialized metabolic features, such as utilization of less common carbohydrates, exhibit differences in their distribution across the isolates.

#### Coexistence of Strains in Co-culture

Serially transferring the 33 strain community for 48 days seemingly led to a loss of over half of the strains within the first 16 days, although the initial number of viable strains is uncertain owing to different growth dynamics of individual strains prior to initiating the experiment (**Supplementary Figure S1** and **Figure 5**). After these transient dynamics, there was only a marginally significant change in community composition over time between days 16 and 48 (PERMANOVA: F1,<sup>7</sup> = 3.018, r <sup>2</sup> = 0.301, p = 0.074). Furthermore, the apparent initial loss in richness (strain count) was accompanied by an increase in evenness, seen as a relatively small temporal change in Shannon diversity, which accounts for both factors. There were three dominant strains: Elizabethkingia meningoseptica HAMBI 1875, Aeromonas caviae HAMBI 1972 and Pseudomonas chlororaphis HAMBI 1977. However, the degree to which the dominance of these strains is caused by their relative abundance or viability in the input community (strain mix) or competitive ability during co-culture is uncertain.

## DISCUSSION AND CONCLUSION

We developed a multispecies synthetic bacterial community, which was characterized at the genomic and phenotypic levels, revealing high diversity in the resistome, mobilome and functionome of the community. Furthermore, we demonstrated the utility of the community by showing coexistence of a large proportion (14/33) of the strains in co-culture over serial transfer for 48 days. These observations indicate that the community is suitable for use as a cultivated model community to ask a wide range of biological questions.

Along with characterizing the community, we produced comprehensive genomic and phenotypic metadata for the community members. This facilitates future mechanistic work with the community. For instance, our pipeline allows the use of amplicon sequencing to track community composition over time at strain-level resolution. Moreover, the genome assemblies provide a reference database for (meta)genomic and (meta)transcriptomic studies. Isolation of individual colonies combined with colony PCR, in turn, allows the rapid identification of substrains possessing mutations or horizontal gene transfer events of interest.

Our model community is composed of diverse bacterial strains isolated from different environments and hence does not mimic any specific natural community. Therefore, the community is more suited to study general questions about ecology and evolution such as community assembly and response to environmental perturbations rather than to explain patterns in any particular natural microbial community (Wolfe, 2018).

The ability of the community to answer general questions may be limited by the potential absence of focal taxonomic groups or high-order interactions that occur in more complex communities. The community is composed of bacteria alone, while interactions between bacteria and fungi, protozoa or bacteriophages often play a key role in natural microbial communities. In a separate study, however, we have introduced a method for incorporating protozoa in the community (Cairns et al., 2018).

The utility of the 33-strain model community and its predecessor is demonstrated in the current and previous work by us (Cairns et al., 2018). Here we present the community and a collection of methods and metadata to the scientific community. A previous version of the community has already been successfully used for tracking the mobility of antibiotic resistance genes (Cairns et al., 2018), and as prominent cases of future use we envision, for instance, replicated ecosystem microcosms to explore the trajectories of ecosystem composition and genetic structure in response to various environmental perturbations. We also envision the community as an efficient spike-in control to increase the statistical rigor of highthroughput microbial single cell assays such as epicPCR (Spencer et al., 2016) or as a test system for validating metagenome assemblers.

#### DATA AVAILABILITY

Raw fastq files and genome assemblies for the genomes sequenced or assembled in this study (N = 17) have been deposited in the National Center for Biotechnology Information (NCBI) Sequence Read Archive (SRA) and GenBank, respectively, under the BioProject Accession Number PRJNA476209. Accession numbers for raw fastq files or genome assemblies obtained for this study from databases are listed in **Supplementary Table S1**. Previously Sanger sequenced near-full-length 16S rRNA gene sequences for all strains are available in the European Nucleotide Archive (ENA) under the Accession Number PRJEB21728 (https://www.ebi.ac.uk/ena/data/view/ PRJEB21728). The complete sequence for the plasmid RP4 harbored by E. coli K-12 JE2571(RP4) is available from GenBank under the accession BN000925 (Pansegrau et al., 1994). Genome annotations, including those obtained in this study for previously sequenced or assembled genomes, are available upon request.

The following community metadata is available via Dryad (doi: 10.5061/dryad.53b6n5f): Genomic predictions for functional modules obtained using MAPLE 2.3.0 (Takami et al.,

#### REFERENCES


2016); genomic feature data collated into a single data frame, using custom R scripts (R Core Team, 2017), displaying strain, contig, locus, and genomic feature information; antibiotic MICs; and carbon source utilization phenotypes. Raw output files for each strain from all genomic feature prediction tools are available upon request.

The model community strains, except for E. coli K-12 strain JE2571(RP4), are available individually from the HAMBI culture collection at the University of Helsinki, Finland (http://www. helsinki.fi/hambi/). For further requests concerning the use of the model community, please contact the corresponding author.

## AUTHOR CONTRIBUTIONS

JC and TH designed the experiments. RJ contributed to phenotypic measurements, proof of concept experiments, and performed sequence analyses for 16S rRNA amplicon data. TH, JH, and MV provided sequencing resources. JC and JH designed the sequence analyses. JC performed sequence analyses for whole-genome data and drafted the manuscript. JC, MT, and RJ contributed to data visualization. RJ, JH, MT, MV, and TH edited and commented on the manuscript. All authors gave final approval for publication and accept accountability for the content and work performed.

### FUNDING

This work was supported by Academy of Finland projects (Grant Nos. 308128 and 314114 to JH and 307213 to TH), University of Helsinki (3-year Grant to JH), and Doctoral Programme in Microbiology and Biotechnology, University of Helsinki (funding to JC).

#### ACKNOWLEDGMENTS

We thank Anna Hartikainen and Liisa Ruusulehto for technical assistance, and acknowledge CSC – IT Center for Science Ltd., for the allocation of computational resources.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fgene. 2018.00312/full#supplementary-material

Arndt, D., Grant, J. R., Marcu, A., Sajed, T., Pon, A., Liang, Y. J., et al. (2016). PHASTER: a better, faster version of the PHAST phage search tool. Nucleic Acids Res. 44, W16–W21. doi: 10.1093/nar/gkw387

Arredondo-Alonso, S., Willems, R. J., van Schaik, W., and Schürch, A. C. (2017). On the (im)possibility of reconstructing plasmids from whole-genome shortread sequencing data. Microb. Genom. 3:e000128. doi: 10.1099/mgen.0.00 0128

Bairey, E., Kelsic, E. D., and Kishony, R. (2016). High-order species interactions shape ecosystem diversity. Nat. Commun. 7:12285. doi: 10.1038/ncomms 12285



**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 © 2018 Cairns, Jokela, Hultman, Tamminen, Virta and Hiltunen. 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) and the copyright owner(s) 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.

digital media

of impactful research

article's readership