# NTM – THE NEW UBER-BUGS

EDITED BY : Veronique Dartois, Christine Sizemore and Thomas Dick PUBLISHED IN : Frontiers in Microbiology

#### Frontiers Copyright Statement

© Copyright 2007-2019 Frontiers Media SA. All rights reserved. All content included on this site, such as text, graphics, logos, button icons, images, video/audio clips, downloads, data compilations and software, is the property of or is licensed to Frontiers Media SA ("Frontiers") or its licensees and/or subcontractors. The copyright in the text of individual articles is the property of their respective authors, subject to a license granted to Frontiers.

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

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

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

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

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

The above represents a summary only. For the full conditions see the Conditions for Authors and the Conditions for Website Use. ISSN 1664-8714 ISBN 978-2-88963-012-7 DOI 10.3389/978-2-88963-012-7

#### 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

# NTM – THE NEW UBER-BUGS

Topic Editors:

Veronique Dartois, Center for Discovery and Innovation, Hackensack Meridian Health, United States

Christine Sizemore, National Institute of Allergy and Infectious Diseases, United States

Thomas Dick, Center for Discovery and Innovation, Hackensack Meridian Health, United States

"Mycobacterium colonies" by Michelle Yee

Incidence of pulmonary disease caused by nontuberculous mycobacteria (NTM), such as Mycobacterium abscessus and M. avium, is increasing at an alarming rate, surpassing tuberculosis (TB) in many countries. Treatment durations are extremely long. With low sputum conversion rates, the outcomes are often disappointing. Thus, NTM disease resembles extremely drug resistant TB. There is an urgent need for increased research on this dreadful disease. For this book we brought together experts from a wide range of disciplines addressing current status, gaps and needs, and new developments in several NTM-relevant research areas. We start with the discussion of NTM disease presentations and clinical research. This is followed by contributions on epidemiology, antibiotic resistance mechanisms, drug discovery and development, bacteriology and targets, and new diagnostic tools. We would like to thank the 130 authors for sharing their work and insights. We hope that this collection of articles will stimulate discussions and research activities on the challenging lung disease caused by the diverse group of pathogens called NTM.

Citation: Dartois, V., Sizemore, C., Dick, T., eds. (2019). NTM – The New Uber-Bugs. Lausanne: Frontiers Media. doi: 10.3389/978-2-88963-012-7

# Table of Contents

*06 Editorial: NTM—The New Uber-Bugs* Veronique Dartois, Christine Sizemore and Thomas Dick

#### DISEASE PRESENTATIONS AND CLINICAL RESEARCH


Christopher Vinnard, Alyssa Mezochow, Hannah Oakland, Ross Klingsberg, John Hansen-Flaschen and Keith Hamilton

#### EPIDEMIOLOGY

*28 A Closer Look at the Genomic Variation of Geographically Diverse*  Mycobacterium abscessus *Clones That Cause Human Infection and Disease*

Rebecca M. Davidson


#### RESISTANCE MECHANISMS


#### DRUG DISCOVERY AND DEVELOPMENT


Jimin Zhang, Franziska Leifer, Sasha Rose, Dung Yu Chun, Jill Thaisz, Tracey Herr, Mary Nashed, Jayanthi Joseph, Walter R. Perkins and Keith DiPetrillo

*119 Optimization and Lead Selection of Benzothiazole Amide Analogs Toward a Novel Antimycobacterial Agent*

Mary A. De Groote, Thale C. Jarvis, Christina Wong, James Graham, Teresa Hoang, Casey L. Young, Wendy Ribble, Joshua Day, Wei Li, Mary Jackson, Mercedes Gonzalez-Juarrero, Xicheng Sun and Urs A. Ochsner

*129* Mycobacterium abscessus *and ß-Lactams: Emerging Insights and Potential Opportunities* Elizabeth Story-Roller, Emily C. Maggioncalda, Keira A. Cohen and

Gyanu Lamichhane

*139 Novel Screen to Assess Bactericidal Activity of Compounds Against Non-replicating* Mycobacterium abscessus

Bryan J. Berube, Lina Castro, Dara Russell, Yulia Ovechkina and Tanya Parish


### BACTERIOLOGY AND TARGETS

*179 MmpL3 as a Target for the Treatment of Drug-Resistant Nontuberculous Mycobacterial Infections*

Wei Li, Amira Yazidi, Amitkumar N. Pandya, Pooja Hegde, Weiwei Tong, Vinicius Calado Nogueira de Moura, E. Jeffrey North, Jurgen Sygusch and Mary Jackson

*188 Lsr2 is an Important Determinant of Intracellular Growth and Virulence in*  Mycobacterium abscessus

Vincent Le Moigne, Audrey Bernut, Mélanie Cortès, Albertus Viljoen, Christian Dupont, Alexandre Pawlik, Jean-Louis Gaillard, Fabienne Misguich, Frédéric Crémazy, Laurent Kremer and Jean-Louis Herrmann

*199 Glycopeptidolipids, a Double-Edged Sword of the* Mycobacterium abscessus *Complex*

Ana Victoria Gutiérrez, Albertus Viljoen, Eric Ghigo, Jean-Louis Herrmann and Laurent Kremer

*207* Mycobacterium abscessus *subsp.* massiliense mycma\_0076 *and*  mycma\_0077 *Genes Code for Ferritins That are Modulated by Iron Concentration*

Fábio M. Oliveira, Adeliane C. Da Costa, Victor O. Procopio, Wanius Garcia, Juscemácia N. Araújo, Roosevelt A. Da Silva, Ana Paula Junqueira-Kipnis and André Kipnis

#### DIAGNOSTIC TOOLS

*223 Evaluation of a Novel MALDI Biotyper Algorithm to Distinguish*  Mycobacterium intracellulare *From* Mycobacterium chimaera

L. Elaine Epperson, Markus Timke, Nabeeh A. Hasan, Paul Godo, David Durbin, Niels K. Helstrom, Gongyi Shi, Markus Kostrzewa, Michael Strong and Max Salfinger

#### *229 MALDI Spectra Database for Rapid Discrimination and Subtyping of*  Mycobacterium kansasii

Jayaseelan Murugaiyan, Astrid Lewin, Elisabeth Kamal, Zofia Bakuła, Jakko van Ingen, Vit Ulmann, Miren J. Unzaga Barañano, Joanna Humięcka, Aleksandra Safianowska, Uwe H. Roesler and Tomasz Jagielski

## Editorial: NTM—The New Uber-Bugs

#### Veronique Dartois <sup>1</sup> , Christine Sizemore<sup>2</sup> and Thomas Dick <sup>1</sup> \*

<sup>1</sup> Center for Discovery and Innovation, Hackensack Meridian Health, Nutley, NJ, United States, <sup>2</sup> Division of Microbiology and Infectious Diseases, National Institute of Allergy and Infectious Diseases, National Institutes of Health, Rockville, MD, United States

Keywords: non-tuberculous mycobacteria, environmental mycobacteria, lung disease, repositioning, from biology to solutions

**Editorial on the Research Topic**

#### **NTM—The New Uber-Bugs**

Pulmonary disease caused by non-tuberculous mycobacteria (NTM), relatives of Mycobacterium tuberculosis, is increasing at an alarming rate, surpassing tuberculosis (TB) in many countries. Patients suffering from chronic pulmonary diseases, including Cystic Fibrosis and Chronic Obstructive Pulmonary Disease, are particularly susceptible to NTM infections. These infections are difficult to diagnose and present significant challenges for treatment. The environmental reservoir of the infectious agents and an incomplete understanding of what makes individuals susceptible to infection complicates preventive approaches. Considering the rising incidence and likely underreporting of these infections, a clear understanding of the barriers to diagnosis, treatment and prevention, and focused research on the pathobiology and microbiology of NTMs and new methods to combat them is urgently needed.

#### Edited by:

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

#### Reviewed by:

Philippe Lanotte, Université de Tours, France

\*Correspondence: Thomas Dick thomas.dick@HMH-CDI.org

#### Specialty section:

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

> Received: 23 April 2019 Accepted: 24 May 2019 Published: 12 June 2019

#### Citation:

Dartois V, Sizemore C and Dick T (2019) Editorial: NTM—The New Uber-Bugs. Front. Microbiol. 10:1299. doi: 10.3389/fmicb.2019.01299

Recognizing this emerging new health threat and the complex nature of these infections, the National Institute of Allergy and Infectious Diseases organized the first NTM-focused workshop "Advancing Translational Science for Pulmonary NTM Infections" in Rockville, MD, in September 2017. This workshop, bringing together scientists and physicians from the United States and abroad, identified gaps in fundamental biological and epidemiological research and mapped out possible approaches for targeted translational science to expedite improvements in detection and care (Daniel-Wayman et al., 2018). The gathering not only generated increased awareness of this new health issue, it also brought together research disciplines and investigators from across the world who previously had few opportunities to interact, and kick-started new collaborative research activities. Momentum created during this workshop also resulted in the Research Topic "NTM—The New Uber-Bugs."

NTM basic and translational research are still in their infancy with many questions remaining to understand the biological diversity of the infecting pathogens, their similarities and differences and how these affect pathobiology, transmission, and response to therapy. Furthermore, depending on where NTM infections occur globally, different species of NTM predominate, making rapid diagnostics that can speciate mycobacteria a necessity. This acute need to reduce the biological uncertainties around NTM pulmonary disease by increasing research efforts reminds of the state of knowledge TB research was facing two decades ago. Despite an initial research void in the TB field during the last quarter of the twentieth century, in part driven by the absence of molecular tools and animal models to study this difficult pathogen, significant progress has been achieved since the early 2000s. Knowledge and tools developed for the study of TB disease and approaches and technologies and platforms that have been applied to TB product development can be leveraged for NTM research.

**6**

For instance, chemical entities with activity against Mycobacterium tuberculosis that have been generated in TB screening campaigns can quickly also be tested against NTM and may facilitate repositioning and repurposing of chemical leads and create a complementary approach to de novo drug discovery. By exploiting strategies and tools developed for TB, together with increasing research and product development focused on high-priority NTM species, we believe that significant medical advances for NTM diseases will be achieved in the medium term.

The present Research Topic, "NTM—The New Uber-Bugs" compiles contemporary results and insights from a group of leading researchers into the high-priority research area of NTM infections. It also highlights that in order to truly understand the diverse nature of these infections and their impact on human health, more thorough trans-national collaborations and data sharing will be important. We would like to thank the reviewers for their many thoughtful and insightful comments, and the authors for their excellent contributions. We hope that this collection of manuscripts will stimulate much-needed discussions and research activities on the challenging lung disease caused by the diverse group of pathogens called nontuberculous mycobacteria.

#### REFERENCES

Daniel-Wayman, S., Abate, G., Barber, D. L., Bermudez, L. E., Coler, R. N., Cynamon, M. H., et al. (2018). Advancing translational science for pulmonary nontuberculous mycobacterial infections. a road map for research. Am. J. Respir. Crit. Care Med. 199, 947–951. doi: 10.1164/rccm.201807-1273PP

**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.

#### AUTHOR CONTRIBUTIONS

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

#### FUNDING

TD holds a Toh Chin Chye Visiting Professorship at the Department of Microbiology and Immunology, Yong Loo Lin School of Medicine, National University of Singapore. His work is supported by the National Institute of Allergy and Infectious Diseases of the National Institutes of Health under Award Numbers R01AI132374 and 2R01AI106398-05 and by the Cystic Fibrosis Foundation under Award Number DICK17XX00. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health or the Cystic Fibrosis Foundation.

#### ACKNOWLEDGMENTS

We are grateful to all the authors and reviewers for their contributions to this Research Topic.

Copyright © 2019 Dartois, Sizemore and Dick. 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.

# Host Variability in NTM Disease: Implications for Research Needs

Colin Swenson<sup>1</sup> \*, Christa S. Zerbe<sup>2</sup> and Kevin Fennelly<sup>3</sup>

<sup>1</sup> Division of Pulmonary, Allergy, Critical Care, and Sleep Medicine, Emory University, Atlanta, GA, United States, <sup>2</sup> Laboratory of Clinical Immunology and Microbiology, National Institute of Allergy and Infectious Diseases, National Institutes of Health, Bethesda, MD, United States, <sup>3</sup> Laboratory of Chronic Airway Infection, Pulmonary Branch, Division of Intramural Research, National Heart, Lung, and Blood Institute, National Institutes of Health, Bethesda, MD, United States

Non-tuberculous mycobacteria (NTM) are ubiquitous environmental organisms that may cause opportunistic infections in susceptible hosts. Lung infections in immunocompetent persons with structural lung disease are most common, while disseminated disease occurs primarily in immunocompromised individuals. Human disease caused by certain species, such as Mycobacterium avium complex, Mycobacterium abscessus, and Mycobacterium kansasii, is increasing in incidence and varies by geographic distribution. The spectrum of NTM disease varies widely in presentation and clinical outcome, but certain patterns can be organized into clinical phenotypes. Treatment options are limited, lengthy, and often toxic. The purpose of this case-based review is to provide non-clinician scientists with a better understanding of human NTM disease with an aim to stimulate more research and development.

#### Edited by:

Thomas Dick, Rutgers, The State University of New Jersey, United States

#### Reviewed by:

Jennifer R. Honda, National Jewish Health, United States Mary Ann Degroote, SomaLogic, United States

#### \*Correspondence:

Colin Swenson colin.swenson@emory.edu

#### Specialty section:

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

Received: 06 August 2018 Accepted: 12 November 2018 Published: 03 December 2018

#### Citation:

Swenson C, Zerbe CS and Fennelly K (2018) Host Variability in NTM Disease: Implications for Research Needs. Front. Microbiol. 9:2901. doi: 10.3389/fmicb.2018.02901 Keywords: non-tuberculous mycobacteria, pulmonary, bronchiectasis, disseminated, COPD, cystic fibrosis, osteomyelitis

"It is more important to know what sort of person has a disease than to know what sort of disease a person has."

– Hippocrates, Greek Physician, 460–370 B.C.

### INTRODUCTION

Non-tuberculous mycobacteria (NTM) are ubiquitous environmental organisms that are frequently isolated from soil and water, as well as other environmental sources. NTM are acid-fast bacilli (AFB) that are clinically categorized as either rapid growers (RG), taking fewer than 7 days to grow on culture media, or slow growers (SG), taking more than 7 days (**Table 1**). Certain species are well known causes of human disease, most commonly Mycobacterium avium complex (MAC), which includes the species M. avium, M. intracellulare, and M. chimaera, and the M. abscessus complex (MABSC), which includes the subspecies abscessus, massiliense, and bolletii. Considered opportunistic pathogens, NTM organisms may occasionally be isolated from the sputa samples of healthy individuals, but there is a higher prevalence of NTM infection among patients with cystic fibrosis and non-cystic fibrosis bronchiectasis, as well as chronic obstructive pulmonary disease (COPD) (Adjemian et al., 2012; Fleshner et al., 2016). While all MAC species may be isolated from water and soil sources, M. avium is more commonly isolated from fresh water sources, and M. intracellulare is associated with garden soils (Falkinham, 2015). This predilection has been proposed as an explanation for the epidemiologic clusters observed in regions with abundant surface water and loamy soil.

TABLE 1 | Clinically important mycobacterial species organized by growth time.


Slow growing organisms take greater than 7 days to culture. Rapid growing organisms take fewer than 7 days to culture. <sup>∗</sup> Includes M. avium, M. intracellulare, and M. chimaera. ∗∗Includes M. abscessus, M. massiliense, and M. bolletii.

While NTM species may rarely cause pulmonary infections in normal hosts, those with structural lung disease, such as bronchiectasis and emphysema, are at increased risk. The hallmark of the disease is granulomatous inflammation of the airways, which, over time, may lead to worsening bronchiectasis and cavitary lung disease (Fujita et al., 1999). Among patients with pulmonary NTM infections, distinct phenotypes have been described based primarily on the underlying lung disease. These phenotypes have characteristic presentations on lung imaging (**Table 2**).

Although once considered rare diseases, human infections with NTM appear to be increasing in North America and Europe, and they are increasingly recognized in areas of the world considered endemic for tuberculosis (TB), including Africa, South America and India (Adjemian et al., 2012). Human lung disease caused by NTM was first recognized in the 1950s as the incidence of TB was decreasing (Timpe and Runyon, 1954). Given the similarities of clinical microbiology and of the drugs used to treat TB, the care of patients with NTM disease has often been provided by clinicians and hospitals initially established to manage TB. The HIV/AIDS epidemic introduced many physicians and scientists to disseminated infections with MAC, which has waned in incidence along with other opportunistic infections due to the development and widespread use of highly active antiretroviral therapy. Most disseminated NTM infections occur in immunocompromised patients. Similar to TB, pulmonary NTM disease accounts for approximately 85% of cases and disseminated NTM disease for about 15% (Jones et al., 2018). The major types of human NTM disease are summarized in **Table 3**.

The primary purpose of this review is to introduce microbiologists and other scientists who are not clinicians to the wide variety of clinical presentations among humans infected with NTM. Given the variable responses observed to infections with the same pathogen among various inbred strains of mice and other experimental animal models, it should not be surprising that the 'human host' is not just one host, but rather many. However, in spite of the enormous genetic variability among humans, there are similarities in presentations of NTM disease among various human 'phenotypes.' We will present representative cases, most of which are actual cases that we have seen, followed by a discussion of that category of disease and figures of imaging studies. We hope that a better understanding of the various human diseases caused by NTM can help inspire our

FIGURE 1 | Axial CT chest image demonstrating bronchiectasis and mucoid impaction in the right middle lobe. The tree-in-bud nodularity in the left lung (yellow arrow).

non-clinical colleagues to develop better diagnostic tests, drug therapies and vaccines and other preventive interventions.

#### PROTECTION OF RESEARCH PARTICIPANTS

Written consent was obtained from each patient presented in this article. He or she explicitly consented to the inclusion of background and clinical data, as well as radiographic images.

#### NON-CYSTIC FIBROSIS BRONCHIECTASIS (NCFB)

#### Case

A 67-year-old female lifelong non-smoker with a history of scoliosis and early osteoporosis developed a non-productive cough for 6 months, an unintentional weight loss, and daytime fatigue. She had experienced a seven-pound weight loss over 6 months, to a body mass index (BMI) of 17.5 kg/m<sup>2</sup> (normal: 19–24 kg/m<sup>2</sup> ). An avid swimmer and jogger, she had also noticed a gradual decline in her stamina, with an increase in shortness of breath with moderate exercise. Her primary care provider ordered a chest radiograph, which showed possible right middle lobe pneumonia. She was prescribed 10 days of levofloxacin, which did not improve her respiratory symptoms. She developed severe diarrhea, and on the seventh day of levofloxacin, her husband took her to the emergency department, where she was admitted for dehydration secondary to Clostridium difficile colitis. Because the cough had not improved with antibiotics, a CT of the chest was performed, showing right middle lobe and lingula bronchiectasis, nodularity, and mucus impaction (**Figure 1**). She was placed on airborne isolation for possible pulmonary tuberculosis. MTB PCR probe was negative on an TABLE 2 | Radiographic presentations of pulmonary NTM diseases.

fmicb-09-02901 November 29, 2018 Time: 13:6 # 3


induced sputum sample, so she was taken for bronchoscopy with lavage. These samples were smear positive for AFB, and 10 days later, grew M. avium. She was discharged from the hospital on a combination of ethambutol, rifampin, and azithromycin three times weekly, and after 3 months of treatment, her cough resolved and she began to gain weight. Repeat CT imaging at 6 months demonstrated fewer lung nodules, with less mucus impaction.

#### Discussion

The nodular-bronchiectatic phenotype of pulmonary NTM disease is also known as "Lady Windermere Syndrome," in reference to the character from Oscar Wilde's 1892 play entitled "Lady Windermere's Fan" (Reich and Johnson, 1992). While the term has had wide adoption in the medical literature, it is, in fact, a malapropism, since the character for which the syndrome is named is a young woman without respiratory symptoms of any kind. Further, the authors who originally named this syndrome believed that the infection resulted from women chronically suppressing their cough, leading to mucus inspissation in the longer airways of the right middle lobe and lingula. This hypothesis has more recently been challenged, since a hallmark of the nodular-bronchiectatic phenotype is chronic cough and sputum production (Olivier, 2016; Tsai et al., 2017). Reflex cough responses in elderly women with stable pulmonary NTM disease appear similar to age-matched controls, except for a mildly decreased urge-to-cough, arguing against cough suppression (Tsai et al., 2017).

Individuals with this phenotype are typically female, elderly, are tall with thin body habitus, have pectus excavatum, scoliosis, and a non-smoking history (Kim et al., 2008; Park et al., 2009; Lee et al., 2013). Radiographically, the right middle lobe and lingula are most often affected, though any area of the lung may be affected (**Figure 1**). Typically, the radiographic pattern is described as "tree-in-bud," indicating bronchiolar inflammation and mucus impaction. Coinfection with more than one species of Mycobacterium, or with Gram-negative bacteria, characterizes later, more advanced disease.

Symptomatically, patients with the nodular-bronchiectatic phenotype typically complain of chronic cough, which may or may not be productive, and may or may not be associated with hemoptysis, persistent fatigue, night sweats, and unintentional weight loss. The infection may be mistaken for active pulmonary tuberculosis, leading to airborne isolation and prolonged hospitalization, as happened with our patient. Treatment is typically a prolonged combination regimen of a rifamycin, most often rifampin, ethambutol, and a macrolide, typically azithromycin or clarithromycin. The frequency of dosing


of cystic bronchiectasis (yellow arrow).

depends on the extent of radiographic disease: mild nodularbronchiectatic disease may be treated with a thrice-weekly regimen, while more severe bronchiectasis requires a daily regimen. In the most severe cases, particularly those with cystic bronchiectasis and cavitary lung disease, a concurrent thriceweekly parenteral aminoglycoside, such as amikacin, may be necessary (Griffith et al., 2007). As with the M. abscessus group, surgical resection in localized or cavitary disease is an option where medical treatment has failed to eradicate infection. Since individuals with the nodular-bronchiectatic phenotype are typically elderly and frail, the risks of surgical resection may outweigh potential benefits, in which case a chronic suppressive strategy with antibiotics is a reasonable alternative.

Complete clearance of MAC infection is variable, and often depends on the severity of disease. Further limiting the successful clearance of MAC infection is the low rate of clinician adherence to the 2007 ATS/IDSA treatment guidelines (Adjemian et al., 2014a,b). With the guidelines-based treatment approach, more than 60% of patients with nodular-bronchiectatic disease will have treatment success (Wallace et al., 2014; Diel et al., 2018). These patients continue to be susceptible to reinfection, however, often with entirely new strains or subspecies. Overall, the recurrence of pulmonary MAC infection approaches 40%, with the highest rates of recurrence among the nodular-bronchiectatic cohort (Lee et al., 2015; Boyle et al., 2016). The rate of relapsed disease, where the same clinical isolate is again isolated within 12 months of supposed cure, is unknown. While environmental exposure has been proposed as a source of infection, it is unclear what role remediation and exposure avoidance have on preventing reinfection.

#### CYSTIC FIBROSIS LUNG DISEASE (CF)

#### Case

A 21-year-old man with CF presented to the outpatient clinic with worsening dyspnea, weight loss, fever, and hemoptysis over approximately 2 months. He had last been in clinic 3 months prior when he had reported feeling well. Induced sputum sample during that clinic visit grew scant mucoid fluoroquinoloneresistant Pseudomonas aeruginosa. Mycobacterial cultures were negative at that time, and his forced expiratory volume (FEV1: typically the best overall measure of lung function) was stable at 2.5 L (normal for patient's age: 4.5 L). His BMI was chronically low at 17.8 kg/m2 (normal: 19–24 kg/m<sup>2</sup> ). At the current clinic visit, his FEV<sup>1</sup> had declined to 2.1 L, and his sputum sample was a dark gray streaked with blood. Because his resting pulse oximetry reading was consistently in the upper-80% range, he was directly admitted to hospital, where sputum sample was smear positive for AFB. M. tuberculosis PCR probe was negative. Chest radiograph on admission demonstrated upper and midlung zone cystic bronchiectasis with a new possible left upper lobe infiltrate (**Figure 2**, left). CT of chest confirmed the development of a left upper lobe infiltrate, with new tree-in-bud attenuation (**Figure 2**, right). The sputum culture grew M. abscessus 5 days after admission.

#### Discussion

Patients with CF have a higher prevalence of NTM disease than the general population. Pathogenic NTM species are isolated from approximately 1 in 5 patients with CF, and the incidence appears to be increasing by 5% per year, depending on geographic location (Adjemian et al., 2014a, 2018). Among the NTM species affecting this patient population, approximately 60% of isolates are MAC, while almost 40% are MABSC. MABSC is comprised of three subspecies: M. abscessus, M. bolletii, and M. massiliense, all of which are rapidly growing organisms. MABSC is a notoriously difficult infection to treat, owing in part to the erm(41) gene, which confers inducible macrolide resistance (Adjemian et al., 2018). The subspecies M. massiliense is an exception, however, since it lacks a functional erm(41) gene, and is therefore more easily eradicated (Kim et al., 2010; Roux et al., 2015). Patients with CF who develop MABSC infection typically have a lower BMI and, paradoxically, a higher FEV<sup>1</sup> than those without MABSC infection (Adjemian et al., 2018).

Infection with MABSC is considered a relative contraindication to lung transplantation, since MABSC infection may prove fatal in the post-transplant period, when recipients are on maximal immunosuppressive therapy (Chalermskulrat et al., 2006; Gilljam et al., 2010). Because subjects with CF have progressive lung disease, lung transplantation is often a consideration at the later stages of disease. Our patient was concurrently followed by the lung transplant clinic, though he had never been listed. His sputum isolate was ultimately identified as subspecies M. massiliense, and he was successfully treated with a regimen consisting of azithromycin, amikacin, and cefoxitin. Sputum samples at 3, 6, and 9 months were negative for NTM species. His hemoptysis resolved, and he did not require supplemental oxygen at his three-month follow-up appointment.

#### CHRONIC OBSTRUCTIVE LUNG DISEASE (COPD)

#### Case

A 55-year-old man with a history of COPD associated with smoking developed a chronic cough productive of sputum, malaise, and night sweats. The social history was remarkable

FIGURE 3 | Left: chest radiograph demonstrating a right upper lobe cavitary lesion on a background of emphysema (green arrow). Right: axial CT image demonstrating the same cavitary lesion in the right lung (yellow arrow). Reprinted with permission of the American Thoracic Society. Copyright© 2018 American Thoracic Society. Fennelly et al. (2016) is an official journal of the American Thoracic Society.

the inner wall of the lung cavity, with many AFB seen at higher power using Ziehl Nissen stain (right, yellow arrows). Reprinted with permission of the American Thoracic Society. Copyright© 2018 American Thoracic Society. Fennelly et al. (2016) is an official journal of the American Thoracic Society.

for use of a hot tub at his home. Sputum microbiology was notable for 3+ AFB on smear and three cultures grew M. abscessus, susceptible to amikacin and linezolid, and intermediate to clarithromycin (without erm gene testing) and cefoxitin. Pulmonary function testing showed airflow obstruction and a low diffusing capacity, and imaging demonstrated a right upper lobe cavity and emphysematous changes, consistent with COPD (**Figure 3**). He and his wife had just retired and sold their home in northern Florida. They were en route to vacation in the American West in their recreational vehicle (RV), hoping to receive some oral antibiotics and be on their way. His symptoms responded well to treatment with intravenous amikacin once daily and intravenous imipenem-cilastatin twice daily and azithromycin (home health nursing visited him in a state forest campground where they lived in their RV during his treatment). However, 4 months after treatment he was admitted to the hospital with new shortness of breath, fevers, and with ground-glass opacities, consistent with parenchymal inflammation, in the posterior aspects of both upper lobes. Bronchoalveolar lavage (BAL) grew only M. abscessus, and the BAL cell count was markedly lymphocytic. This pattern was consistent with an allergic alveolitis due to a spill of the cavity, so

he was treated with oral corticosteroids and his antimycobacterial regimen intensified by the addition of linezolid once daily. As the cavity was the major site of disease and his lung function was only mildly decreased, he was referred for surgical resection of the right upper lobe. The histopathology of the cavity was remarkable for aggregates of AFB along the inner wall of the cavity (**Figure 4**). M. abscessus grew from a culture of the resected lung cavity within 3 days. Scanning electron microscopy of the inner wall of the resected cavity demonstrated bacilli embedded within a matrix. There were 3.83 × 10<sup>7</sup> colony forming units (CFU) of total bacteria in a 0.5 g sample from the lung cavity; 7.17 × 10<sup>5</sup> CFU were in biofilm. The bacterial composition was 97.7% mycobacterial using 16s ribosomal RNA sequencing (Fennelly et al., 2016). The patient's symptoms and imaging improved post-operatively, and all sputa specimens were negative on culture in the subsequent 10 months of his tour of the West.

#### Discussion

Patients with COPD often have emphysema that is typically more prominent in the upper lobes. The 'blebs' or holes within the lung parenchyma are presumably infected by NTM. The mode of infection is not well understood, but we suspect that it was likely due to inhalation in this case. We see a similar pattern of Aspergillus fumigatus infecting the upper lobes of COPD patients, and Aspergillus in the most ubiquitous airborne mold. As TB often presents with upper lobe cavities, COPD patients with AFB in their sputum are often initially placed in respiratory isolation as TB suspects unless a rapidly done molecular diagnostic can confirm that the AFB are not M. tuberculosis.

Pulmonary NTM infections in COPD patients have a low cure rate, generally considered to be about 50% compared to about 75% in patients with idiopathic bronchiectasis (Jones et al., 2018). Whether or not this lower cure rate is due to the presence of cavities is unknown. Surgical resection of a cavity can be curative, as in this case. Unfortunately, many patients with cavitary disease have lung function that it so diminished that they cannot tolerate surgery. Surgical resection may also be contraindicated if there is extensive bilateral disease. lower lobes.

fmicb-09-02901 November 29, 2018 Time: 13:6 # 6

However, medical treatment can have a role in preventing further progression of disease, possibly by suppressing bacterial growth within cavities or by protecting healthy lung tissue by eradicating small amounts of bacteria that are spread from diseased areas. **Figure 5** shows the chest radiographs of a woman with COPD who had cavitary pulmonary MAC disease (**Figure 5**, Left). Although her sputum remained AFB smear- and culture-positive during her course of treatment, her symptoms had improved and there was no further radiographic evidence of disease progression. Unfortunately, she stopped all her medications when she was given a diagnosis cervical cancer and became depressed, and she was lost to follow-up. Approximately 6 months later she was admitted to the intensive care unit, and a repeat chest radiograph showed dramatic progression of disease (**Figure 5**, Right). The right upper lobe cavity appeared to have extended and spilled, destroying her entire right lung and significantly involving the left. She died 2 weeks later. COPD appears to be the most important underlying condition leading to mortality in patients with pulmonary NTM infections (Hayashi et al., 2012; Ito et al., 2012; Mirsaeidi et al., 2014; Gommans et al., 2015; Jones et al., 2018), although fibrocavitary disease without COPD is also associated with increased mortality (Fleshner et al., 2016).

### PRIMARY CILIARY DYSKINESIA (PCD)

#### Case

A 53 year old woman is followed for bronchiectasis, sinusitis and chronic lung infections with MAC and P. aeruginosa. She had no respiratory distress as an infant, but she began to develop recurrent sinus infection and a productive cough around age four. At the age of 18 she was diagnosed with bronchiectasis. She had been evaluated for CF with sweat chloride tests were negative on three occasions. She has two children, and no other family members have respiratory disease. On exam now, she is thin with a body-mass index of 17.9 (normal: 19–24 kg/m<sup>2</sup> ). On lung exam she has scattered wheezes and rales in the bases. She inhales nebulized hypertonic saline twice daily and engages in aerobic exercise daily for airway clearance.

Imaging of her sinuses shows chronic sinusitis, with airfluid levels in both maxillary sinuses and opacified ethmoid and frontal sinuses, in spite of evidence of prior surgeries. Computed tomography (CT) of the chest is remarkable for severe bronchiectasis, predominantly in both lower lobes (**Figure 6**).

At the age of 37, she was first diagnosed with pulmonary MAC infection after presenting with worsening cough. She improved on treatment with clarithromycin, ethambutol and rifampin for 9 months. However, her cough returned several months after treatment was stopped, and she was restarted on the same treatment except with azithromycin instead of clarithromycin. Unfortunately, she noticed difficulties discriminating red and green colors and was diagnosed with optic neuritis secondary to ethambutol but this resolved after discontinuation of the drug. She was treated with inhaled amikacin but it was stopped due to severe hoarseness. Linezolid was recently added but the discontinued after she developed numbness and tingling in her fingers and toes consistent with peripheral neuropathy.

At the age of 47, the diagnosis of PCD was confirmed by low nasal nitric oxide of 15 nL/min and genetic analysis demonstrating two disease-causing mutations in DNAH11. Electron microscopy showed no structural changes of the cilia as seen in classic PCD.

P. aeruginosa was first cultured from her sputum in early 2017, but she was feeling relatively well. She received three courses of oral ciprofloxacin without eradication. In the fall of 2017, she complained of more cough and sputum production. She was treated with cefepime and tobramycin by the intravenous route for 2 weeks to treat a bronchiectasis exacerbation due to P. aeruginosa. She had baseline mild tinnitus that did not change. A follow-up audiological evaluation 1 month later showed the development of mild to moderate high frequency sensorineural hearing loss bilaterally, although the patient did not perceive hearing loss. Follow-up audiological evaluations after another month and then 6 months later showed no improvement in the hearing acuity, but it remained stable. Unfortunately, her Pseudomonas infection now appears chronic. She is feeling relatively well on suppressive therapy with inhaled aztreonam.

#### Discussion

PCD, or immotile cilia syndrome, is an autosomal recessive disorder of motile cilia dysfunction resulting in impairment of mucociliary clearance. It is heterogenous, with over 30 genetic variants described to date and a wide range of clinical manifestations. Unlike the patient above, approximately 80% of newborns have neonatal respiratory distress. About one-half of PCD patients have 'laterality defects' such as situs inversus, in which the organs are on the opposite side of the body due to defects in the embryonic nodal cilia. In children, CF and PCD may both present with respiratory infections and distress. In CF, the problem is due to alterations of the airway surface fluid,

whereas in PCD the problem is with poor movement of the fluid due to the dysfunctional cilia. Both result in bronchiectasis (dilated, chronically inflamed bronchi), but the location tends to be in the upper lobes in CF and in the lower lobes in PCD. Nearly all PCD patients have rhinosinusitis. Infertility is a possibility, but not universal as in this case. The median age of diagnosis in childhood is about 5 years of age, but the age of diagnosis in adults is widely variable. Diagnosis has traditionally been difficult due to variable availability and accuracy of tests such as transmission emission microscopy to evaluate 'classic' ciliary ultrastructural defects and nasal nitric oxide measurements. However, the increased availability of genetic testing has been a significant advance. A recent clinical practice guideline provides a very helpful diagnostic algorithm (Shapiro et al., 2018).

Most PCD patients live an active life and have a normal lifespan, although lung infections can interfere with their normal social functions. The rate of decline in their lung function is slower than in CF. The rates of NTM infections among patients with PCD is not known. In a recent registry study, PCD patients were less likely to have NTM infections than those with bronchiectasis associated with gastroesophageal reflux disease (GERD).

This case highlights problems of drug toxicity experienced by most patients with NTM lung diseases, who are also frequently co-infected with Gram-negative bacteria or fungal pathogens. In review, this patient has had adverse reactions to ethambutol (optic neuritis), inhaled amikacin (laryngitis), linezolid (peripheral neuropathy), and intravenous tobramycin (ototoxicity). Due to the multiple adverse reactions, her chronic MAC infection now can be treated only with azithromycin and rifampin, a suboptimal regimen for cure, which is probably unlikely to occur due to the severity of her bronchiectasis. The goal of therapy for both the MAC and the Pseudomonas is now suppression of bacterial growth to prevent exacerbations and further lung destruction. Such adverse reactions to drugs limit the armamentarium available to physicians to treat patients, and contribute to the urgent need for additional antibacterial drugs.

#### HYPERSENSITIVITY PNEUMONITIS (HP)

#### Case

One of us (KF) was consulted on the case of a 52 year-old pulp mill worker for evaluation of possible underlying occupational lung disease. He had been referred to the infectious disease service after a surgical lung biopsy at an outside hospital showed non-caseating granulomas and AFB. The lung tissue culture had grown MAC. The patient had noted progressive shortness of breath. On physical examination, the patient had a very rapid breathing rate in spite of being on supplemental oxygen by nasal cannula with a reservoir device at 15 l/min. There were inspiratory crackles on lung exam. The lung imaging was remarkable for diffuse 'mosaic attenuation' and fine nodules (**Figure 7**) which was not consistent with what is usually seen in MAC infections. It was originally suspected that he might have HP due to airborne molds within the pulp mill. However, the history revealed that he had not worked inside the pulp mill for several years due to a severe back injury. He worked in a control room in another building at a distance from the plant. To manage his back pain, the patient had the habit of soaking in a hot tub for 30 min in the morning after awakening and then again in the evening when he returned from work. The hot tub was in a room just outside his bedroom. He was too short of breath to be able to do pulmonary function testing. We considered performing

a bronchoscopy to obtain BAL fluid for cell counts, but we realized that both HP and mycobacterial infection would be predominantly lymphocytic. Given his tenuous respiratory status and his tissue diagnosis, the risk of the procedure far outweighed any benefit. He was treated with both systemic corticosteroids for the HP and with azithromycin, rifampin and ethambutol for the MAC infection. He began to improve within a few days in the hospital, and was discharged home with the advice to drain the hot tub after obtaining a water sample for culture. He sent the water to the local state public health laboratory and received a report that it grew MAC. Unfortunately, when we called to request the isolate to determine if it was identical to the clinical one, the laboratory in a southeastern state had already discarded it because they were overwhelmed with such cultures. When we saw him 18 months after his removal from exposure and initiation of treatment, his symptoms, exam, imaging and pulmonary function tests were completely normal.

#### Discussion

This was a classic case of 'hot tub lung' due to inhalation of NTM. This form of HP was first described in case reports in 1997, and multiple cases have now been reported in the literature. In a retrospective case series of 21 patients at a referral center, all patients presented with shortness of breath and cough and about half were hypoxemic. MAC was isolated from both the hot tub water and the respiratory secretions or tissue in all cases (Hanak et al., 2006). All improved with avoidance of exposure, and there was a mix of additional corticosteroid and antimycobacterial therapy, but five (24%) received neither. Thus, avoidance of exposure appears to be the most important element. However, in patients with severe illness like our case, more aggressive treatment is warranted. Some of the more interesting case reports have involved multiple immunocompetent family members developing HP due to MAC and possibly other NTM (Mangione et al., 2001; Kitahara et al., 2016).

An occupational respiratory illness associated with NTM that emerged in the last decade is HP associated with metalworking fluids in both the United States and the United Kingdom (Dawkins et al., 2006). These fluids were found to be contaminated with a rapidly growing Mycobacterium now known as M. immunogenum (Wallace et al., 2002).

Mycobacterium chelonae was isolated from a bassoon played by a professional musician diagnosed with HP, but molds were also isolated and no NTM or molds were isolated from the BAL from the patient (Moller et al., 2017). Thus, the association may not be causal. NTM contamination in a water-damaged building was recently associated with asthma symptoms (Park et al., 2017). Environmental sampling for NTM is not commonly done in indoor environments, suggesting a need for further research about environmental exposures to airborne NTM.

#### DISSEMINATED DISEASE

#### Case

A 49-year-old denied any infections or hospitalizations in his childhood. As a young adult he had histoplasmosis lymphadenitis and was treated successfully. He did well until 47 years of age when he presented to a hospital complaining of diarrhea after having several year history of GERD, intermittent nausea and emesis. Colonoscopy was performed and pathology showed Periodic Acid Schiff stain (PAS) positivity, so empiric treatment for Whipple's disease, a rare systemic infection caused by the bacterium Tropheryma whipplei, was initiated. One month later, he presented to a large tertiary hospital with mycobacteremia, hypotension and sepsis. He was diagnosed with idiopathic CD4 lymphopenia and disseminated MAC. He suffered multiple cerebral embolic events, had a tracheostomy placed, underwent PEG placement and suffered extensive end-organ damage from long term aminoglycoside therapy. His GI disease was indicative of his overall burden of disease. He was referred to the NIH after the diagnoses of two variants (IL12Rβ1 c.512 A > C, p.Q171P; and IL12Rβ1 c. 1442A > G, p.Y481C) were found in IL-12 receptor 1. He required years of intensive multidrug therapy in addition to immune therapies including exogenous interferon gamma.

#### Discussion

Genetic susceptibility to disseminated mycobacterial infection is well described and should be considered in anyone who presents with disseminated disease without HIV infection (Wu and Holland, 2015). Even diseases that are often diagnosed in childhood can present with disseminated infection in adulthood when their mutation results in a less severe phenotype (Hsu et al., 2018). Host control depends on the interleukin 12-interferon pathway that connects myeloid and lymphoid cells. The genetic susceptibilities, if inherited, can be elucidated by a careful family medical history. Autosomal dominant patterns of inheritance include: IFNR1/R2, STAT1, IRF8, GATA2, IL12Rβ1/R2 X-linked: NEMO. Autosomal recessive patterns of inheritance typically present early in life and would not have been considered in this age patient. Non-HIV Acquired immunodeficiency such as autoantibodies to interferon gamma (IFN) would have been another consideration in this patient and one that is important to consider in adults with disseminated infection. However, in the United States, women have been identified more frequently than men, and have mainly been described in the East Asian population.

IL-12Rβ1 encodes the 1 chain of IL-12 and IL-23 receptors and activation of pSTAT4 and is important in the subsequent signaling of IFN and activation of macrophage intracellular killing. IL-12Rβ1 has been reported in more than 200 patients worldwide (de Beaucoudrey et al., 2010). Typically, they will present after BCG vaccination in countries where TB is endemic. However, in the United States, these patients may present later in life, as our patient presented. This case of a primary immune deficiency that phenotypically presented in adulthood highlights the importance of evaluating genetic disorders that may predispose individuals to disseminated infection with a low virulent organism in the absence of known risk factors.

#### Case

A 32-year-old Filipina woman presented to her primary care provider with fevers, neck and back pain, and cervical

lymphadenopathy. She was diagnosed with communityacquired pneumonia, and treated with azithromycin and oral corticosteroids as well as a subsequent course of doxycycline with more oral corticosteroids. While her pain improved on the corticosteroids, she started having night sweats off of prednisone. An eventual CT of her neck showed destruction of her C3 vertebra, and lymphadenopathy. A PET/CT showed uptake in activity in lymph nodes, femurs, left acetabulum, left pubic ramus, sacrum, right ischium, both iliac bones, ribs, and clavicles. She underwent C3 corpectomy with fusion; followed by right femoral curettage 1 month later. When AFB+ organisms were seen, an interferon gamma release assay (IGRA) was indeterminate. Cultures grew MAC. Three months after the first surgery, she continued to have pain and underwent debridement of an epidural abscess in T8, involving a laminectomy, foraminectomy and hardware placement of screws at T6, T7, T10, and T11. Her disease continued to worsen despite optimal multi-drug therapy. Subsequent testing at the National Institutes of Health (NIH) showed very high titer autoantibodies to Interferon gamma. She was referred to the NIH and continued on multi-drug antibiotic therapy and given rituximab with improvement in her disease. Unfortunately, when her B-cells returned after 2 years, her disease recurred and she remains on therapy for her infection and is getting repeat rituximab infusions for the IFN autoantibodies.

#### Discussion

Autoantibodies to IFN were first reported by Doffinger et al. (2004). An adult-onset immunodeficiency due to these autoantibodies in Thailand and Taiwan was described in 2012 in a population of HIV negative patients presenting with opportunistic infections (Browne et al., 2012). They have since been described in patients with thymic neoplasia (Burbelo et al., 2010). An indeterminate IGRA test is characteristic in these patients, as sera from them will block positive controls for this commercial test and will therefore be reported as indeterminate. In vitro, the pSTAT1 activation is negative when patient sera is tested in the presence of normal PBMCs. Especially in Asians, one should consider autoantibodies to IFN when encountering an adult with disseminated NTM, particularly when there is no family history of disease.

#### VERTEBRAL OSTEOMYELITIS

#### Case

A 16-year-old girl was evaluated for progressive mid-thoracic back pain. She reported the onset after abatement of symptoms of a flu-like illness with fever, nasal congestion, sore throat and non-productive cough and muscle aches. She was a competitive roller-skater, and she had experienced multiple falls in the prior year. She recalled multiple ecchymoses and abrasions due to these falls. She had no history of unusual infections earlier in her life, and there was no suggestion of immunodeficiency in her laboratory data. The C-reactive protein (a non-specific marker of systemic inflammation) was

minimally elevated at 1.3 mg/dL. Her examination was normal with no neurological deficits. A CT scan of her lungs was normal. A radionuclide bone scan demonstrated increased uptake in the vertebral body of T9, and magnetic resonance imaging (MRI) showed a soft tissue mass anterior to the vertebral bodies of T8, T9, and T10. Thoracoscopic biopsy of the T9 vertebral body demonstrated necrotizing granulomas of the bone; special stains for bacteria, mycobacteria and fungi were negative (**Figure 8**). She was treated initially for both M. tuberculosis and NTM with isoniazid, rifampin, ethambutol and clarithromycin. After the culture grew only M. abscessus, she was treated with only clarithromycin; the other drugs for tuberculosis were discontinued. A follow-up MRI showed resolution of the edema and overall improvement at 4 months. The clarithromycin was discontinued after 6 months, and she was well with a normal C-reactive protein 3 months later. She was advised to discontinue roller-skating and other potentially traumatic activities and to be closely monitored for relapse.

#### Discussion

Although most cases of disseminated NTM disease suggest the presence of an immunodeficiency, NTM infections can occasionally occur at distant sites from the lung in special situations. Obviously these organisms can gain access to bone or other internal structures during penetrating trauma. However, this case was associated with blunt trauma and so illustrates the principle of 'locus minoris resistentiae,' i.e., a place of less resistance, in this case to microbes. This unusual case was previously published along with two similar cases in 2001 (Chan et al., 2001). In retrospect, such a case now would probably be treated not only with a macrolide like clarithromycin, but also with at least one intravenous medication such as amikacin. The isolate in this case was most likely a M. massiliense subspecies that was susceptible to clarithromycin. Interestingly, shorter courses of therapy than are used for pulmonary disease are acceptable for bone infections.

### NOSOCOMIAL INFECTIONS WITH Mycobacterium chimaera AFTER CARDIAC SURGERY

fmicb-09-02901 November 29, 2018 Time: 13:6 # 10

#### Case

A 64-year-old man had cardiac surgery for a mitral valve porcine mitral valve replacement and a new prosthetic aortic valve in March of 2012. His postoperative course was uneventful until 5 months after the surgery when he developed persistent fevers between 103 and 104 degrees Fahrenheit. Blood cultures were negative, and transthoracic and transesophageal echocardiograms were normal. He was treated empirically with both oral and intravenous antibiotics for culture negative endocarditis in the setting of his recent surgery. In December 2012 (9 months after surgery) blood cultures drawn 30 days prior grew M. chimaera (a species in the M. avium complex). Despite optimal medical management with intravenous and oral multi-drug therapy, he remained persistently bacteremic for 2 years. Unfortunately, despite multi-drug therapy and close medical follow-up, the patient died of his persistent bacteremia.

#### Discussion

A prolonged outbreak of M. chimaera infections after openchest cardiac surgery was first reported by Swiss investigators in 2015. After the recognition of a cluster of cases of disseminated M. chimaera after cardiac surgery at the same hospital as our patient's, the US Centers for Disease Control and Prevention (CDC) and the Pennsylvania Department of Health investigated the outbreak. They found that the heatercooler devices (HCDs) used to maintain the patients' body temperature during extracorporeal circulation (also known as cardiac bypass) were contaminated with M. chimaera, confirming similar findings by the Swiss investigators (Perkins et al., 2016). The isolates obtained from HCDs from three different European countries were nearly identical to those from the manufacturing site in another study. The results of these investigations strongly suggested a point-source contamination in the manufacturing of a specific HCD. There is a growing body of literature on this problem, and investigations are on-going.

Whole exome sequencing revealed that our patient had a mutation of interferon regulatory factor 8 (IRF8), which regulates expression of interferon-alpha and interferon beta genes. IRF8 is critically important in immune responses to intracellular pathogens, especially mycobacteria, involving the production of interleukin-12 in response to interferon-gamma (Hambleton et al., 2011). This mutation likely contributed to his susceptibility and death due to this infection.

### CONCLUSION

The cases presented in this paper demonstrate that NTM infections vary widely in presentation, treatment, and clinical outcome. While ubiquitous in the environment, NTM species tend to infect certain vulnerable patient populations, and these infections can often be organized into clinical phenotypes. While many NTM species have low pathogenicity, some opportunistic pathogens are important causes of human disease. Infection with these organisms, including MAC, MABSC, M. kansasii, M. xenopi, and others, are increasing in incidence and geographic distribution. Most human disease is pulmonary, primarily affecting immunocompetent patients with underlying structural lung disease such as bronchiectasis or COPD. Disseminated disease is usually more life-threatening, and primarily occurs in immunocompromised individuals. Among those most severely affected by disseminated disease are children with specific primary immunodeficiencies, and those with acquired T cell immunodeficiency due to anti-cytokine antibodies or iatrogenic administration of immunosuppressive medications, particularly tumor-necrosis factor alpha antagonists.

We hope that the host variability in NTM disease presented in this paper serves to stimulate and inspire basic and translational scientists to consider a similar variety of innovative interventions. For example, the poor response to treatment and increased mortality seen in fibrocavitary lung disease might be improved by drugs targeting biofilm dispersion (COPD Case). Inhaled liposomal amikacin was recently approved for pulmonary MAC disease in the United States, and translational studies on its role in biofilm infections may be a next step. Elucidation of the pathogenesis of non-CF bronchiectasis in tall, thin women may enable us to 1 day prevent the development of the disease. Candidate areas of research for the nodular-bronchiectasis MAC phenotype include leptin and ghrelin hormone influences on the immune response, cellular senescence, and the aging lung. In CF, research is needed on potential airborne transmission of M. abscessus, biofilm formation in the airways by M. abscessus, the effect of new CFTR modulator drugs on mycobacterial infection, among others. In PCD, it is possible that drugs improving the function of cilia may have a role in preventing NTM infections, and perhaps gene mutations could be corrected in the future. HP could be prevented by improved water disinfection techniques, and disseminated disease acquired during cardiac bypass could also be prevented by improved disinfection of heart-lung machines. Disseminated diseases associated with immunodeficiencies could be prevented by better immunomodulatory drugs that could correct such defects. Obviously, most patients would benefit from simpler, shorter and more effective drug treatment regimens against the mycobacteria, and the role of inhaled liposomal amikacin in such regimens continues to be defined.

The cases that we have presented illustrate the broad spectrum of NTM disease. We hope that these clinical vignettes have provided non-clinical scientists with a better understanding of human NTM diseases.

### AUTHOR CONTRIBUTIONS

All authors contributed sections to the manuscript, and all reviewed and contributed to revisions.

### REFERENCES

fmicb-09-02901 November 29, 2018 Time: 13:6 # 11


diffuse panbronchiolitis. J. Korean Med. Sci. 24, 427–432. doi: 10.3346/jkms. 2009.24.3.427


mycobacterial lung infections. J. Appl. Physiol. 122, 1262–1266. doi: 10.1152/ japplphysiol.00939.2016


**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 Swenson, Zerbe and Fennelly. 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.

# Assessing Response to Therapy for Nontuberculous Mycobacterial Lung Disease: Quo Vadis?

Christopher Vinnard<sup>1</sup> \*, Alyssa Mezochow<sup>2</sup> , Hannah Oakland<sup>3</sup> , Ross Klingsberg<sup>3</sup> , John Hansen-Flaschen<sup>2</sup> and Keith Hamilton<sup>2</sup>

<sup>1</sup> Public Health Research Institute, New Jersey Medical School, Newark, NJ, United States, <sup>2</sup> Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA, United States, <sup>3</sup> Department of Medicine, Tulane University School of Medicine, New Orleans, LA, United States

Assessing progression of disease or response to treatment remains a major challenge in the clinical management of nontuberculous mycobacterial (NTM) infections of the lungs. Serial assessments of validated measures of treatment response address whether the current therapeutic approach is on track toward clinical cure, which remains a fundamental question for clinicians and patients during the course of NTM disease treatment. The 2015 NTM Research Consortium Workshop, which included a patient advisory panel, identified treatment response biomarkers as a priority area for investigation. Limited progress in addressing this challenge also hampers drug development efforts. The Biomarker Qualification Program at the FDA supports the use of a validated treatment response biomarker across multiple drug development programs. Current approaches in clinical practice include microbiologic and radiographic monitoring, along with symptomatic and quality-of-life assessments. Blood-based monitoring, including assessments of humoral and cell-mediated NTMdriven immune responses, remain under investigation. Alignment of data collection schemes in prospective multicenter studies, including the support of biosample repositories, will support identification of treatment response biomarkers under standard-of-care and investigational therapeutic strategies. In this review, we outline the role of treatment monitoring biomarkers in both clinical practice and drug development frameworks.

Keywords: nontuberculous mycobactena, biomarker (development), response, therapeutics, clinical trial, radiography, quantitative culture

Nontuberculous mycobacteria (NTM) are ubiquitous environmental organisms capable of causing significant morbidity (Yeung et al., 2016) and mortality (Vinnard et al., 2016; Diel et al., 2018), predominantly in the form of chronic lung disease (Wassilew et al., 2016). Profound knowledge gaps regarding the diagnosis of NTM lung infections (Griffith et al., 2007) include distinguishing NTM colonization from infection (van Ingen, 2015), identification of patients most likely to benefit from treatment (Mirsaeidi et al., 2014), selection of an optimal initial drug regimen (Philley and Griffith, 2015), and classification of treatment endpoints. As an explanation for these knowledge gaps, the formal study of NTM lung disease is relatively recent compared to tuberculosis, and basic understanding of NTM lung disease pathophysiology has only recently emerged.

In this review, we address uncertainties regarding the clinical "waypoints" of NTM treatment. Intermediate and definitive markers of treatment response provide crucial feedback to NTM patients, their clinicians, and developers of novel therapies. We outline the role of

#### Edited by:

Rustam Aminov, University of Aberdeen, United Kingdom

#### Reviewed by:

David E. Griffith, The University of Texas Health Science Center at Tyler, United States Debra Hanna, Bill & Melinda Gates Foundation, United States

\*Correspondence:

Christopher Vinnard christopher.vinnard@njms.rutgers.edu

#### Specialty section:

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

Received: 02 July 2018 Accepted: 01 November 2018 Published: 20 November 2018

#### Citation:

Vinnard C, Mezochow A, Oakland H, Klingsberg R, Hansen-Flaschen J and Hamilton K (2018) Assessing Response to Therapy for Nontuberculous Mycobacterial Lung Disease: Quo Vadis?. Front. Microbiol. 9:2813. doi: 10.3389/fmicb.2018.02813

**20**

treatment monitoring biomarkers in both clinical practice and drug development. We summarize the existing knowledge base regarding biomarkers of treatment response for NTM disease and we identify areas for ongoing and potential future investigation.

### THE UTILITY OF A TREATMENT RESPONSE BIOMARKERS IN CLINICAL CARE AND DRUG DEVELOPMENT

In his seminal framework for defining surrogate endpoints in clinical trials, Ross Prentice proposed that a valid surrogate endpoint must "capture" the underlying relationship between the intervention and the clinical endpoint, allowing for comparison of intervention groups based on the surrogate rather than the definitive clinical endpoint at any time during the followup period (Prentice, 1989). By comparison, the concept of a treatment monitoring biomarker is less restrictive than a surrogate endpoint. The Institute of Medicine defines a biomarker as "a characteristic that is objectively measured and evaluated as an indicator of normal biologic processes, pathogenic processes, or biologic responses to a therapeutic intervention" (Institute of Medicine (US) Committee on Qualification of Biomarkers and Surrogate Endpoints in Chronic Disease, 2010).

The BEST (Biomarkers, EndpointS, and other Tools) Resource, published by the FDA-NIH Biomarker Working Group, provides descriptions and examples of disease monitoring biomarkers, which are "assessed serially over time to measure presence, status, or extent of disease or medical condition, or to provide evidence of an intervention effect or exposure, including exposure to a medical product or environmental agent" (Cagney et al., 2017).

Thus, the interpretation of a monitoring biomarker is not based on a snapshot level, as with a baseline clinical predictor of response to therapy, but rather on changes in values during serial measurements. As examples, quantification of HIV RNA in blood serves as a monitoring biomarker for the clinical efficacy of antiretroviral therapy, and the amount of air that an individual can forcefully exhale in one second (FEV1) reflects response to chronic obstructive pulmonary disease management. Ideally, a treatment monitoring biomarker could be readily adopted into existing procedures that support routine use in clinical settings (Nahid et al., 2011).

The therapeutic approach to NTM lung disease lacks the evidence base of tuberculosis, where standard-of-care combination drug regimens are based on randomized clinical trials that enrolled tens of thousands of tuberculosis patients. Unlike tuberculosis, the NTM lung disease research agenda has not been propelled by public health pressure or urgency. Consequently, much of the clinical approach to NTM disease rests on a foundation of observational data and expert opinion, supported by only a few randomized clinical trials (Research Committee of the British Thoracic Society, 2001). In general, this evidence base has demonstrated that antimycobacterial therapies are less effective for NTM lung disease than tuberculosis (Griffith et al., 2007). The 2015 NTM Research Consortium Workshop, which included a patient advisory panel, identified treatment response biomarkers as a priority area for investigation (Henkle et al., 2016). One or more treatment response biomarkers would help to answer a fundamental question for clinicians and patients during the course of NTM disease treatment: is the current therapeutic approach on track toward a clinical cure?

This question has a particular significance for NTM disease given the complexity of treatment and the frequency of drugrelated adverse events. Interval monitoring of treatment response would support clinical decision-making with regards to the intensification or de-escalation of therapies. When an initial period of intravenous therapy is used, for example with disease caused by M. abscessus complex, the optimal time point for transitioning to an all-oral regimen is uncertain. Surgical interventions may be the most appropriate approach for a subset of patients with severe M. abscessus complex lung disease, but potentially could be deferred in a subset of patients that are responsive to medical management (Jarand et al., 2011). Such decision nodes provide a potential point of application for a treatment monitoring biomarker. Monitoring biomarker response early in treatment could also inform clinical decisionmaking in advance of drug susceptibility testing (DST) results, which have a delay of several weeks and lack correlation between in vitro susceptibility and clinical outcomes for many antibiotics.

Aside from their value in clinical management decisions, treatment monitoring biomarkers support drug development efforts. According to guidelines established by the Food and Drug Administration (FDA), the use of a biomarker in the context of drug development relies on "qualification" that specifies its contextual use and interpretation. The Biomarker Qualification Program at the FDA provides a pathway for a particular biomarker to become accepted into the regulatory framework independent of approval of a specific drug (Amur et al., 2015). This program was recently updated in response to the 21st Century Cures Act, establishing formal qualification plans for biomarkers and other drug development tools. As an example, the Critical Path Institute (Tuscon, AZ, United States) is evaluating sputum levels of lipoarabinomannan, a component of the mycobacterial cell wall, as a non-culture based measure of tuberculosis treatment response, which could be incorporated into adaptive clinical trial designs (FDA, 2018a). Similarly, a qualified biomarker of treatment response for NTM disease would support drug development efforts across multiple programs, a notable advantage given the likelihood that combination antimicrobial therapies will remain essential to future drug regimens.

#### CURRENT APPROACHES TO MONITORING TREATMENT RESPONSE TO NTM DISEASE

#### Microbiologic Monitoring

Presently, culture of serial sputum samples serves as the primary biomarker for monitoring the response to treatment

of mycobacterial lung infections. In 2016, an expert panel composed of NTM clinicians reached consensus definitions for several endpoints related to the treatment of NTM lung disease, including cure, relapse, and re-infection (van Ingen et al., 2018). Serial culturing of sputum and deep respiratory samples during treatment of NTM lung disease provides the basis for definition of microbiologic cure, which is defined as 3 negative sputum cultures collected at least 4 weeks apart.

Interval microbiologic data may also provide insight into the effectiveness of therapy when the microbiologic cure endpoint has not yet been reached. Mycobacterial cultures from respiratory specimens can be classified semiquantitatively according to growth characteristics in liquid and solid media (Griffith et al., 2015; **Table 1**). Using this scoring system, Griffith et al. (2015) analyzed semiquantitative sputum culture scores on 180 patients with lung disease caused by M. avium complex (MAC), grouping patients into converters and nonconverters based on sputum culture status after 12 months of treatment. Although baseline semi quantitative culture scores were not significantly different between converters and nonconverters, the change in semi quantitative culture scores after 2 months of MAC-directed treatment was highly predictive of sputum culture conversion, with each 1-point decline in the scoring scale associated with a 20% increase in the likelihood of achieving conversion at 12 months. Importantly, early microbiologic response also correlated with symptomatic and radiographic improvements, suggesting that semi quantitative culture data can be explored as a surrogate endpoint in drug development. One limitation of this approach may be that some NTM lung disease patients do not expectorate sputum (Huang et al., 1999), although proper application of sputum induction technique considerably improves microbiologic yield (Guiot et al., 2017).

Interval treatment response monitoring based on microbiologic data is supported by recently published NTM lung disease treatment guidelines. The British Thoracic Society recommends obtaining interval sputum cultures every 4– 12 weeks during treatment to assess microbiologic response (Haworth et al., 2017). Similarly, consensus recommendations provided by the US Cystic Fibrosis Foundation and the European Cystic Fibrosis Society for the management of NTM lung disease include interval monitoring of sputum culture every 4–8 weeks during treatment (Floto et al., 2016). An update from the American Thoracic Society/Infectious Disease Society of America regarding management of NTM lung disease is currently in preparation. Unlike tuberculosis, there are no data supporting non-culture based microbiologic monitoring during NTM lung disease treatment.

### Radiographic Monitoring

Radiographic imaging studies have long served an essential role, not only in diagnosing NTM lung infection and determining the extent of disease, but also in assessing response to treatment (Griffith et al., 2007). A simple chest radiograph scoring system with 3 tiers (improved, no change, or worsened in comparison with the preceding examination) was predictive of sputum culture conversion among MAC lung disease patients,



with repeated routine chest radiography performed a mean of 59 days after the baseline (pre-treatment) examination (Griffith et al., 2015). Computed tomographic (CT) imaging considerably improves assessment and comparison of NTM disease extent compared to chest radiographs. Indeed, several scoring criteria for CT imaging studies have been developed and evaluated for this purpose (Kim et al., 2012; Lee et al., 2013).

Untreated patients with NTM lung disease demonstrate progression of CT findings on serial studies that inform decisions to initiate treatment, particularly with the nodular bronchiectatic form of MAC lung disease (Park et al., 2017). Similar to microbiologic endpoints, radiographic improvement at the conclusion of therapy can characterize successful treatment of NTM lung disease. However, less is known regarding the contribution of early radiographic changes as a marker of treatment response. Moreover, interval radiographic findings during treatment of NTM lung disease may be confounded by the extent of underlying lung disease and by the presence of concomitant infection or colonization of airways (Park et al., 2016; Cohen-Cymberknoh et al., 2017). Other limitations of serial CT imaging for monitoring treatment response include the difficulty with classification of NTM lung disease patients with fibrocavitary disease (an alternate form of MAC lung disease), or structural lung changes due to additional treatment modalities (such as surgical resection or radiation fibrosis) (Lee et al., 2013).

Patients with active chronic lung diseases are exposed to significant radiation when monitored by chest CT imaging. Because proton magnetic resonance imaging (MRI) does not use ionizing radiation this imaging modality has been studied for assessment of cystic fibrosis and sarcoidosis, performing comparably to chest CT in identifying various pathologic patterns of the disease in lung parenchyma. Chung and associates have proposed that MRI may emerge as a viable alternative to CT imaging for assessing the distribution and extent of NTM lung infection (Chung et al., 2016). In an exploratory study of 25 patients known to have lung infection with MAC, they found excellent agreement between noncontrast MRI and CT imaging for detection of nodules and cavities, but only limited agreement for detection of bronchial mucous plugging (Chung et al., 2016; Chung et al. (2016).

fmicb-09-02813 November 17, 2018 Time: 16:32 # 4

**Figure 1**). The authors postulated that further improvement in special discrimination of NTM lung infection may be achievable by application of maximum intensity projection reconstruction and contrast enhancement. To our knowledge, no published studies have reported on the usefulness of MRI for serial assessment of NTM lung infection.

Interval examination with <sup>18</sup>F-fluorodeoxyglucose positronemission tomography [18F-FDG PET] is a routine component of monitoring response to chemotherapy for lung cancer, based on the detection of increased glucose metabolism that is characteristic of malignant lesions. More recently, the use of PET modalities has been examined for diagnosis and treatment monitoring of tuberculosis, detecting the increase in glucose metabolism associated with the host inflammatory response (Vorster et al., 2014). Nodular MAC lung disease may show a similar pattern of increased glucose metabolism on PET/CT imaging, raising the possibility of interval monitoring during treatment. Among 6 patients with MAC lung disease assessed with <sup>18</sup>F-FDG PET studies before and after MAC-directed treatment, all patients demonstrated a decline in <sup>18</sup>F-FDG uptake after treatment, even in the absence of significant improvement in high-resolution CT images (Demura et al., 2009; **Figure 2**).

#### Symptomatic Improvement and Quality-of-Life Monitoring

The concept of clinical cure of NTM disease is tied to symptomatic improvement at the conclusion of treatment, which may not reach complete resolution of symptoms due to underlying structural lung disease (van Ingen et al., 2018).

A number of studies have measured baseline quality-oflife indicators among NTM lung disease patients using various assessment tools, including the St. George Respiratory Questionnaire (SGRQ) (Mehta and Marras, 2011), the EuroQOL Five Dimensions (Hong et al., 2014), and the Medical Outcomes Short-Form-36 Questionnaire (SF-36) (Mehta and Marras, 2011; Asakura et al., 2015). However, there are a paucity of data regarding the formal assessment of quality-of-life during NTM treatment as an interval marker of clinical response, a notable knowledge deficit given the importance of patient reported outcomes in the context of drug development efforts. Importantly, under the 21st Century Cures Act, the FDA has updated the regulatory framework for evaluating these types of patient-reported outcomes in the Clinical Outcome Assessment Qualification Program, which operates in parallel with the Biomarker Qualification Program to support the evaluation of drug development tools for a disease area (FDA, 2018b).

Czaja et al. (2016) studied 47 patients with MAC lung disease, assessing quality-of-life at regular intervals during treatment with the SGRQ, and observing a significant improvement as early as 3 months from the treatment initiation. Notably, improvements in quality-of-life assessments were not related to improvements in CT imaging, perhaps a reflection of the limitations of CT imaging as discussed above, along with the greater degree of overall disease status and complexity that can be captured by quality-of-life assessments. A crosssectional study of MAC complex lung disease patients, based on the SF-36, found that quality-of-life scores were lower among currently treated patients compared to previously- or nevertreated patients, but repeated assessments during treatment were not performed (Asakura et al., 2015). While one interpretation of these findings would be that current MAC treatment modalities negatively affect quality-of-life, an alternate explanation would be that patients are selected for MAC treatment based on current symptoms, as reflected in quality-oflife scores.

#### Blood-Based Monitoring

fmicb-09-02813 November 17, 2018 Time: 16:32 # 5

Both humoral and cell-mediated NTM-driven immune responses have been examined as potential sources of treatment monitoring biomarkers. As an alternative to sputum-based biologic sampling, blood-based monitoring may offer several advantages. Unlike sputum, blood is readily available throughout the treatment course, and continuously distributed biomarkers in blood would support precise comparisons with baseline levels without relying on hierarchical scoring systems (as used for radiographic findings or microbiology results). Furthermore, patients undergoing treatment for extra-pulmonary forms of NTM disease could potentially benefit from a validated blood-based biomarker of treatment response, for example during treatment of deep tissue infections such as osteomyelitis, where clinical response is less apparent on physical examination. Immune-based biomarkers of treatment response may also shed insight into the immunologic correlates of protection against disease in at-risk patient populations.

The Capilia MAC Ab ELISA kit (TAUNS, Shizuoka, Japan) developed for the diagnosis of MAC lung disease, measures serum levels of IgA antibodies against glycopeptidolipid (GPL), a major component of the MAC cell wall (Shibata et al., 2016). Serum anti-GPL IgA levels have been proposed as potential biomarkers of treatment response (Jhun et al., 2017; Kitada et al., 2017), although clinical evidence is limited to studies where serum anti-GPL IgA levels were compared before and after treatment, rather than at earlier timepoints during treatment.

Kitada et al. (2005) followed 27 patients with pulmonary MAC disease, obtaining serum samples before and after anti-MAC therapy. Among patients achieving clinical cure, a reduction in anti-GPL IgA levels was observed, while no difference was observed among 13 patients with treatment failure. Interestingly, a single patient with medical treatment failure then underwent surgical lobectomy, and subsequently demonstrated a fall in serum anti-GPL IgA levels, corresponding to conversion of sputum cultures. In a subsequent study that included 34 patients with MAC lung disease, serum anti-GPL IgA post-treatment declined from pre-treatment values among 19 patients who achieved sputum culture conversion, but not among 7 patients who either had disease recurrence or 8 patients who failed to achieve conversion (Kitada et al., 2017).

Jhun et al. (2017) evaluated serial changes in serum anti-GPL IgA levels during treatment of MAC lung disease at baseline, 3 months, and 6 months of treatment. Across all 57 patients, a decrease in serum anti-GPL IgA levels was observed. In contrast to the earlier findings discussed above (Czaja et al., 2016), patients with unfavorable treatment responses also had significant declines in anti-GPL IgA levels, although baseline levels were higher among patients who eventually developed treatment failure, perhaps reflecting a greater initial disease burden.

Cell-mediated immune responses, rather than antibody responses, could provide an alternate approach for immunologic monitoring of treatment response. Kim et al. (2014) enrolled 42 patients with MAC lung disease and measured serum concentrations of 36 different type I cytokine-associated molecules before and after 12 months of treatment. Declining levels of IL-17 and IL-23 during follow-up were associated with treatment failure, suggesting that continued impairment of the IL-17 pathway reflects ongoing disease processes. These findings are consistent with earlier work demonstrating that low IL-17 levels are a risk factor for developing MAC lung disease (Lim et al., 2010). Depending on the temporal dynamics of serum cytokines in the IL-17 pathway earlier in treatment, the measurement of one or more molecules on this pathway could be of interest as a potential monitoring biomarker.

Carbohydrate antigen 19-9 (CA 19-9) is a treatment monitoring biomarker for several forms of cancer. Elevated blood levels of CA 19-9 are also seen in various forms of chronic lung disease, including bronchiectasis. Recently, CA 19- 9 was examined as biomarker of treatment response among 24 patients with NTM lung disease in South Korea (Hong et al., 2016). Among 17 patients with treatment response (defined as symptomatic improvement and sputum culture conversion within 12 months of therapy), CA 19-9 levels significant declined, while no difference was observed among the 7 patients without clinical response. Larger cohort studies with serial measurements of CA 19-9 earlier in treatment may be of interest.

#### Pulmonary Function Monitoring

Measurement of forced expiratory volume is a prognostic indicator for several chronic respiratory diseases, including COPD and cystic fibrosis, and the FEV1 is a qualified biomarker for COPD drug development efforts. The utility of pulmonary function monitoring as a general biomarker of treatment response for NTM lung disease will likely be limited by the tremendous heterogeneity in underlying chronic lung diseases and NTM disease presentations (for example, nodular bronchiectatic vs. fibrocavitary disease). In a single center study of 37 cystic fibrosis patients with pulmonary M. abscessus infection, early improvements in FEV1 after treatment initiation (either 30 or 60 days) were not associated with longterm microbiologic outcomes or sustained improvements in pulmonary function (DaCosta et al., 2017).

### FUTURE DIRECTIONS

Clinical research centers specializing in the treatment of NTM lung disease will benefit from alignment of prospective data collection schemes that include specimen repositories for blood (both plasma and peripheral blood mononuclear cells), sputum,

and urine. The Bronchiectasis and NTM Research Registry, originally designed to enroll patients with bronchiectasis without underlying cystic fibrosis, supports serial banking of blood and deep respiratory specimens (Aksamit et al., 2017). This framework could provide the basis for investigating some of the potential microbiologic and serum and biomarkers discussed above. Future advances in MRI and PET imaging and in quantitative comparison of serial imaging studies may accelerate future research on changes in anatomic extent and metabolic activity of NTM infections in response to treatment.

Complementary prospective, observational studies would support comprehensive phenotyping of NTM disease patients with serial assessments of novel biomarkers of disease status and treatment response, and potentially could compare treatment monitoring biomarkers between patients treated for pulmonary and extra-pulmonary NTM disease. Prospective studies of novel monitoring biomarkers should also consider the potential for ongoing environmental re-exposure to NTM pathogens leading to re-infection, as can occur with contaminated household water sources (Lande et al., 2018) or nosocomial infections (Baker et al., 2017). Finally, much of the treatment response work thus far has focused on NTM lung disease has centered on MAC, and multicenter enrollment strategies may be essential to study response biomarkers for non-MAC lung disease, in particular lung disease caused by M. abscessus complex. These types of large scale, multicenter studies will also be essential to the prospective evaluation of novel treatment regimens for NTM lung disease. Multiple enrollment sites will be required to provide the statistical power for comparing alternative therapeutic strategies, given the smaller number of NTM lung disease patients compared with tuberculosis on a global scale. Furthermore, clinical studies supported by the National Institutes of Health (Sigal et al., 2017) and the Bill & Melinda Gates Foundation (Chen et al., 2017) have demonstrated the

#### REFERENCES


feasibility of integrating biomarker assessment into clinical trial design.

### SUMMARY

"Quo vadis" ("whither goest thou?") remains a central question for NTM patients and their physicians. It is uncertain whether a single biomarker of clinical response will sufficiently capture the complexity of NTM disease presentations in a heterogeneous patient population, or whether the current "all of the above" strategy can be improved with refinement of microbiologic, radiographic, and quality-of-life assessments. Should a composite biomarker be developed under the FDA Biomarker Qualification program, it would require a "Context of Use" defining the specific algorithm for combining these individual components into a single treatment response measure. Ongoing research efforts for NTM disease should support the evaluation of biomarkers of response to treatment in parallel with studies of novel therapeutics and mechanisms of disease. Patients, clinicians, and investigators will benefit from the knowledge that the road they travel leads toward clinical cure.

### AUTHOR CONTRIBUTIONS

AM, CV, JH-F, and KH performed the literature review. CV wrote the initial draft. AM, HO, RK, JH-F, and KH critically appraised and revised.

### FUNDING

CV was supported by NIH (K23AI10263).


response in patients with pulmonary mycobacteriosis. Eur. J. Nucl. Med. Mol. Imaging 36, 632–639. doi: 10.1007/s00259-008-1009-5



**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 Vinnard, Mezochow, Oakland, Klingsberg, Hansen-Flaschen and Hamilton. 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.

# A Closer Look at the Genomic Variation of Geographically Diverse Mycobacterium abscessus Clones That Cause Human Infection and Disease

#### Rebecca M. Davidson\*

Center for Genes, Environment and Health, National Jewish Health, Denver, CO, United States

#### Edited by:

Veronique Anne Dartois, Rutgers, The State University of New Jersey, United States

#### Reviewed by:

Joseph Oliver Falkinham, Virginia Tech, United States Rachel Thomson, The University of Queensland, Australia

> \*Correspondence: Rebecca M. Davidson DavidsonR@NJHealth.org

#### Specialty section:

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

Received: 15 August 2018 Accepted: 19 November 2018 Published: 05 December 2018

#### Citation:

Davidson RM (2018) A Closer Look at the Genomic Variation of Geographically Diverse Mycobacterium abscessus Clones That Cause Human Infection and Disease. Front. Microbiol. 9:2988. doi: 10.3389/fmicb.2018.02988 Mycobacterium abscessus is a multidrug resistant bacterium that causes pulmonary and extrapulmonary disease. The reported prevalence of pulmonary M. abscessus infections appears to be increasing in the United States (US) and around the world. In the last five years, multiple studies have utilized whole genome sequencing to investigate the genetic epidemiology of two clinically relevant subspecies, M. abscessus subsp. abscessus (MAB) and M. abscessus subsp. massiliense (MMAS). Phylogenomic comparisons of clinical isolates revealed that substantial proportions of patients have MAB and MMAS isolates that belong to genetically similar clusters also known as 'dominant clones'. Unlike the genetic lineages of Mycobacterium tuberculosis that tend to be geographically clustered, the MAB and MMAS clones have been found in clinical populations from the US, Europe, Australia and South America. Moreover, the clones have been associated with worse clinical outcomes and show increased pathogenicity in macrophage and mouse models. While some have suggested that they may have spread locally and then globally through 'indirect transmission' within cystic fibrosis (CF) clinics, isolates of these clones have also been associated with sporadic pulmonary infections in non-CF patients and unrelated hospital-acquired soft tissue infections. M. abscessus has long been thought to be acquired from the environment, but the prevalence, exposure risk and environmental reservoirs of the dominant clones are currently not known. This review summarizes the genomic studies of M. abscessus and synthesizes the current knowledge surrounding the geographically diverse dominant clones identified from patient samples. Furthermore, it discusses the limitations of core genome comparisons for studying these genetically similar isolates and explores the breadth of accessory genome variation that has been observed to date. The combination of both core and accessory genome variation among these isolates may be the key to elucidating the origin, spread and evolution of these frequent genotypes.

Keywords: nontuberculous mycobacteria (NTM), genome evolution, pulmonary infection, core genome, accessory genome, Mycobacterium abscessus

### INTRODUCTION

fmicb-09-02988 December 4, 2018 Time: 15:45 # 2

Nontuberculous mycobacteria (NTM) include slowly and rapidly growing species in the genus Mycobacterium that are opportunistic pathogens in humans. A common clinical presentation is pulmonary infection, which occurs in susceptible individuals such as those with cystic fibrosis (CF), COPD, and adults with certain morphological phenotypes (Cassidy et al., 2009; Daley and Griffith, 2010; Winthrop et al., 2010; Martiniano et al., 2017). Another presentation is extrapulmonary NTM infection that can result from environmental exposures or healthcare-associated outbreaks (Lai et al., 1998; Duarte et al., 2009; Li et al., 2017). One of the most clinically challenging NTM that causes human disease is the rapidly growing species, Mycobacterium abscessus (M. abscessus), which is classified into three subspecies: M. abscessus subsp. abscessus (MAB), M. abscessus subsp. massiliense (MMAS) and M. abscessus subsp. bolletii (MBOL). MAB is the most frequently observed subspecies in clinical populations followed by MMAS, and MBOL is relatively rare (Bryant et al., 2016). Epidemiological studies in the United States (US) show an increase in pulmonary NTM infections (including M. abscessus) in CF populations (Olivier et al., 2003; Adjemian et al., 2018) and patients over 60 years old (Prevots et al., 2010). M. abscessus is also one of the most prevalent species associated with NTM infections in Europe (Roux et al., 2009; Hoefsloot et al., 2013). M. abscessus infections are of particular concern due to the bacteria's innate resistance to a range of antibiotics often resulting in lengthy treatment courses and poor clinical outcomes (Jarand et al., 2011; Koh et al., 2011).

An interesting aspect of M. abscessus infections is that little is known about exposure risks and modes of transmission associated with pulmonary infections. The prevailing wisdom is that NTMare acquired from the environment, as they inhabit plumbing and water systems (Feazel et al., 2009; Falkinham, 2011; Ovrutsky et al., 2013; Honda et al., 2016; Zhao et al., 2017), as well as aquatic environments and soil (Primm et al., 2004; De Groote et al., 2006). These surveys have predominantly detected slowly growing NTM species, however, underscoring the potential for M. abscessus to occupy different and possibly yet to be identified niches within the environment. Studies of M. abscessus outbreaks and pseudo-outbreaks have found the offending bacteria in contaminated laboratories (Lai et al., 1998), clinic disinfectants (Tiwari et al., 2003) and hospital water sources (Baker et al., 2017), and some epidemic strains show disinfectant resistance (Duarte et al., 2009; Leao et al., 2010). Environmental surveys of NTM in household environments pinpoint the ecological niche of M. abscessus as indoor water and plumbing biofilms (Thomson et al., 2013a; Honda et al., 2016), M. abscessus also exhibits long-term survival on fomite particles (Malcolm et al., 2017). Epidemiological studies, however, have not found clear links between household exposure risks and acquisition of NTM (Prevots et al., 2014). Only two studies from Australia have demonstrated genetic links between pulmonary patient isolates of M. abscessus and environmental isolates from household water sources (Thomson et al., 2013a,b), though the directionality of infection has not been confirmed. Finally, attempts to isolate M. abscessus from environmental sources in the context of suspected outbreaks within CF centers have been unsuccessful.

The first case of person-to-person transmission of any NTM species was proposed among five CF patients with MMAS cared for at a US CF Center, raising concerns about adequacy of infection control and risks associated with patient exposure during CF care (Aitken et al., 2012). This coincided with using whole genome sequencing (WGS) to characterize genomic diversity and perform genetic strain matching. Since then, several studies have used WGS to examine genetic diversity of pulmonary M. abscessus isolates from patients in a US referral hospital (Davidson et al., 2014), suspected outbreaks of pulmonary infections in CF centers (Bryant et al., 2013; Tettelin et al., 2014; Harris et al., 2015; Tortoli et al., 2017); a nationwide epidemic of soft tissue infections in Brazil (Davidson et al., 2013a; Everall et al., 2017) and a global population study of CF-associated pulmonary isolates from multiple CF centers in three continents (Bryant et al., 2016). One of the most intriguing observations from these studies was that a large proportion of MAB and MMAS clinical isolates grouped into "dominant clones", which have high genetic similarity, but are not identical. Moreover, these clones have been found in nearly every WGS study regardless of worldwide location. This finding led to speculation that the clones may have both spread locally within CF clinics, likely through 'indirect transmission' of contaminated clinical environments, and then globally between CF centers (Bryant et al., 2016). However, detection of these clones sporadically from non-CF patients (Davidson et al., 2014) and in cases of extrapulmonary infections (Ripoll et al., 2009; Everall et al., 2017), combined with their unknown prevalence in the environment, challenge this hypothesis.

This review will take a closer look at the genomic studies of M. abscessus (subspecies MAB and MMAS) and summarize observations of the dominant clones (**Table 1**). It will address the limitations of core genome phylogenomic analysis for genetic matching, and explore how accessory genome variation may be the key to revealing the relationships among clonal isolates from disparate geographic regions and infection types. Genomic studies of M. abscessus will undoubtedly be needed in the future to further study the evolution and spread of the dominant clones in both environmental and clinical settings.

#### GENOTYPING METHODS FOR M. abscessus

Several molecular techniques have been used to identify and classify M. abscessus isolates. Early studies relied on single gene sequencing or multi-locus sequence typing (MLST) using genes such as rpoB (Adekambi et al., 2006), hsp65, secA and the internal transcribed spacer (Adekambi et al., 2003; Adekambi and Drancourt, 2004; Zelazny et al., 2009), to classify isolates to the subspecies level. To further evaluate whether strains were "genetically matched" in the context of potential outbreaks, pulsed field gel electrophoresis (PFGE), randomly amplified polymorphic DNA polymerase chain reaction (RAPD-PCR) and repetitive sequence polymerase chain reaction (rep-PCR)


(Healy et al., 2005) have been used. The highest resolution method for genetic matching is whole genome sequencing (WGS) as millions of genomic positions and thousands of loci can be compared, vs. a limited number of loci (usually less than 100) with the previous methods.

In 2009, a complete reference genome at 5.03 megabases (Mb) was published for the MAB type strain, ATCC19977<sup>T</sup> (Ripoll et al., 2009) followed by genomes for the type strains of MMAS at 4.8 Mb (CCUG48898<sup>T</sup> ) (Tettelin et al., 2012) and MBOL at 5.0 Mb (BD<sup>T</sup> ) (Choi et al., 2012). This ushered in a new era in population genomic studies. Since 2013, several groups have utilized WGS to examine M. abscessus subspecies at the population level, comparing patient isolates within a single clinic (Bryant et al., 2013; Davidson et al., 2014; Harris et al., 2015) and on a larger scale, between CF clinics in one or more countries (Bryant et al., 2016; Tortoli et al., 2017).

#### GENOME STRUCTURE OF M. abscessus

To understand the nuances in interpreting comparative genomics analyses, we should consider the structure of bacterial genomes, which are classified into two components. The **core genome** is defined as the genomic regions or genes that are shared by all isolates within a population (at the genus, species or subspecies level) (Medini et al., 2005). These include "housekeeping genes" that are necessary for survival and which evolve at relatively predictable rates. Variation in the core genome is measured by the single nucleotide polymorphisms (SNP) found in the shared regions. Depending on the species, as much as 65% or more of the genomic space can be analyzed in this way. The **accessory genome** includes genetic material that is shared by one or more, but not all, isolates in a population. This includes regions that are acquired by horizontal gene transfer (HGT) such as genomic islands, transposases, mycobacteriophages, and plasmids, which have the potential to change bacterial phenotypes through acquisition of antimicrobial resistance genes and entire gene cassettes that code for novel metabolic processes (Juhas et al., 2009; Sassi and Drancourt, 2014; Garcia et al., 2015; Gray and Derbyshire, 2018). Accessory genome features, such as genes, groups of genes and plasmids, are measured in an isolate population as presence or absence.

The sum of the core genes plus accessory genes in an isolate population is called the **pan genome**. A pan genome study of 40 M. abscessus genomes from all three subspecies revealed a core genome of 3,345 genes, which is about 68% of a typical MAB or MBOL genome and 71% of a typical MMAS genome (Choo et al., 2014). This means that approximately 30% of the M. abscessus genome is made up of accessory genes that are present in only a subset of isolates or are strain-specific. The accessory genome of M. abscessus is known to include prophages, plasmids and genomic islands transferred by HGT from other bacterial genera (Ripoll et al., 2009; Leao et al., 2013; Davidson et al., 2014; Gray and Derbyshire, 2018) consistent with an environmental bacterium exposed to a diverse microbiome. Population genomic studies can use both core and accessory genome data in their genetic matching analyses, but it is common to find studies using only core genome comparisons as these

have the most well-defined and reproducible computational methods. The accessory genome, however, can provide additional discrimination among closely related isolates.

#### GENOMICS STUDIES OF M. abscessus FROM CF POPULATIONS

Following the first published case of inter-patient transmission of MMAS isolates among CF patients in the US (Aitken et al., 2012), multiple European studies used core genome analysis to look for evidence of person-to-person transmission in their CF clinics (reviewed in detail in Martiniano et al., 2017). A retrospective analysis of 168 M. abscessus isolates from 31 adult CF patients at the Papworth hospital in the United Kingdom (UK) (Bryant et al., 2013) revealed two genetic clusters of MMAS with high genetic similarity. This study was the first to define a quantitative SNP threshold indicating probable transmission between patient isolates as less than 25 SNPs. The use of this SNP threshold for isolates from different patients coupled with social network analysis showed that while the patients lived in different geographic areas, they had multiple opportunities for cross-infection through overlapping clinic visits. Moreover, the potentially transmissible isolates shared the same SNP in the 16S rRNA that confers amikacin resistance, including isolates from patients that had not previously used nebulized aminoglycosides. The authors were unable to identify a local environmental source of MMAS inoculum, and thus surmised that person-to-person transmission did occur within their hospital. The absence of isolation in the environment, however, could also reflect insufficient culturing methods or nonexhaustive sampling. Another study at the Great Ormond Street Hospital in the United Kingdom used a similar experimental design to assess the potential for cross-infection among 20 pediatric CF patients (Harris et al., 2015). Core genome analysis of 27 M. abscessus isolates revealed genetic clusters of both MAB and MMAS. The authors used the suggested threshold of less than 25 SNPs between patient isolates as defined in the Papworth study (Bryant et al., 2013) and identified four patients involved in suspected transmission events, but they did not find intersecting clinic visits between any of the patients. Finally, an Italian study analyzed 162 M. abscessus isolates from 48 patients attending CF centers in "four geographically distinct regions in Italy" (Tortoli et al., 2017). Using a conservative threshold of less than 30 SNPs between patient isolates, they found isolate clusters of MAB, MMAS and MBOL, and identified and seven "possible transmission episodes" among patients. In three potential episodes including two clusters of MMAS and one cluster of MBOL, two patients were found to have attended the same clinic within the same timeframe. The authors concluded that the lack of major outbreaks over the 12-year study period signified minimal risk of inter-human transmission in their CF centers. In summary, the occurrence and frequency of crossinfection between CF patients in the clinic remains controversial and further studies will be needed. It would also be useful to compare and contrast infection control strategies in this context.

In 2016, a large-scale study utilized WGS and core genome analyses to examine the global population structure of 1,080 M. abscessus isolates from 510 CF patients sampled at clinics across multiple European countries, the US and Australia (Bryant et al., 2016). The study identified predominant clusters of genotypes that they coined "dominant circulating clones" as well as several other genetically diverse genotypes. The dominant clones are defined as phylogenetically clustered isolates found in a high proportion of patients. The study identified a primary dominant clone of MAB (known as Abscessus Cluster 1) and a dominant clone of MMAS (known as Massiliense Cluster 1) that corresponds to the "transmissible clone" from the previous Papworth study (Bryant et al., 2013). Collectively, clustered isolates had a higher proportion of drug resistance mutations to amikacin and clarithromycin and were more likely to be associated with chronic infections compared with unclustered isolates. Moreover, the clones had increased phagocytic uptake and intracellular survival in macrophages, and significantly greater bacterial burden and granulomatous inflammation in SCID mice suggesting differences in virulence compared with unclustered isolates.

### DOMINANT CLONE OF M. abscessus subsp. massiliense

Three "transmissible" MMAS isolates from the CF clinic outbreak in Seattle, Washington (Aitken et al., 2012) were sequenced in 2014, and the genomes were compared with other publically available genomes (Tettelin et al., 2014). Intriguingly, the Seattle MMAS isolates were highly similar to the transmissible MMAS clones from the Papworth, United Kingdom CF clinic study (Bryant et al., 2013) and an epidemic MMAS strain from Rio de Janiero, Brazil (CRM-0020) (Davidson et al., 2013a,b). Core genome diversity among the geographically disparate isolates was only 11 to 86 SNPs, and the accessory genome variation included a 11.5 kb genomic island that was unique to the Papworth strains and three genomic regions (totaling 95 kb) and a 44 kb IncP-1β plasmids that were unique to the Brazilian strain. Then, two years later, it was revealed that this potentially "transmissible clone" was the most prevalent MMAS (Massiliense Cluster 1) in the global population study of CF-NTM isolates and was present in all the countries sampled including the United States, United Kingdom, Ireland, Sweden, Netherlands, Denmark, Ireland and Australia (Bryant et al., 2016).

The Brazilian clone (also called BRA100) with high genetic similarity to the pulmonary CF strains was responsible for an epidemic of soft tissue infections from 2004–2009 (Duarte et al., 2009; Leao et al., 2010). The series of infections that spread through multiple states were attributed to contaminated surgical equipment that that had been cleaned with glutaraldehyde, which ultimately selected for a disinfectant resistant clone. A follow up population genomics study of 188 epidemic strains from nine Brazilian states revealed that the isolates had one or two different plasmids (pMAB01 and pMAB02) that were absent from

the corresponding pulmonary CF isolates (Everall et al., 2017). Divergence dating analysis estimated that the Brazilian lineage emerged in 2003 suggesting a recent introduction, however, it is unknown if BRA100 exists locally in the environment of Brazil. The two plasmids from the Brazilian isolates have been fully sequenced (Leao et al., 2013; Everall et al., 2017), but the mechanism of glutaraldehyde resistance is still not known.

#### DOMINANT CLONE OF M. abscessus subsp. abscessus

The dominant clone of MAB identified in the global NTM population study (Abscessus Cluster 1) was found in CF patients from all of the countries and continents sampled (Bryant et al., 2016). Interestingly, this clone is highly genetically similar to the MAB type strain, ATCC19977<sup>T</sup> , that was initially isolated in the early 1950's from a midwestern US patient with "an acid-fast infection of the knee" and "subcutaneous abscess-like lesions of the gluteal region" (Moore and Frerichs, 1953). This clone was present in pulmonary isolate clusters found in the CF study in Great Ormand Hospital in the United Kingdom (Harris et al., 2015), the M. abscessusstudy from four Italian CF centers (Tortoli et al., 2017), and in a study of pulmonary isolates from CF and non-CF patients in a US referral hospital in Colorado (Davidson et al., 2014).

The MAB dominant clone isolates appear highly similar by core genome SNP analyses, but they also show significant variation in their accessory genomes. For example, the type strain (ATCC19977<sup>T</sup> ) reference genome includes a 23.3 Kb plasmid with a mercury resistance operon (Ripoll et al., 2009). In the Colorado WGS study with nine clustered pulmonary isolates of the MAB dominant clone, only one (1/9 = 11%) contained the full plasmid, and up to 8.3% of the reference ATCC19977<sup>T</sup> genome was absent in these isolates including a prophage region and multiple transposase-related genomic islands (Davidson et al., 2014). This study also showed that the nine isolates clustered geographically when comparing the presence or absence of genomic islands suggesting that the

#### REFERENCES


accessory genome may be subject to local, environmental-specific adaptations.

#### CONCLUSION AND PERSPECTIVES

The dominant clones of M. abscessus (including MMAS and MAB) have been observed several times over the years associated with a range of disease etiologies (**Table 1**), Recently, WGS and population genomics has revealed the extent of genomic variation in the core genome with less than 100 variable SNP positions, while the accessory genome can include presence or absence of large plasmids and genomic islands. The widespread geographic diversity of the dominant clones is intriguing, and it is currently not known if they have spread globally via inter-continental transmission or if they inhabitant local environments and just happen to be the most effective genotypes for human infection. A large scale genomic study of randomly sampled environmental NTM isolates would reveal the niches of M. abscessus dominant clones in nature and the true risk of environmental exposure to susceptible patient populations (Strong and Davidson, 2017). Such environmental studies are time consuming and costly, and appropriate funding mechanisms will need to be identified. Future genomic studies of M. abscessus will benefit from analyzing diversity in both the core and accessory genes to reveal local adaptations, identify potential mechanisms of virulence and monitor the evolution and mutation rates of these clinically important genotypes.

#### AUTHOR CONTRIBUTIONS

RD conceived of topic and wrote the manuscript.

#### FUNDING

RD acknowledges the NIH/NIAID (award no. 5K01AI125 726-02).




**Conflict of Interest Statement:** The author declares 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 Davidson. 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.

# Global Environmental Nontuberculous Mycobacteria and Their Contemporaneous Man-Made and Natural Niches

Jennifer R. Honda<sup>1</sup> \*, Ravleen Virdi<sup>1</sup> and Edward D. Chan2,3,4

<sup>1</sup> Department of Biomedical Research and the Center for Genes, Environment, and Health, National Jewish Health, Denver, CO, United States, <sup>2</sup> Medicine and Academic Affairs, National Jewish Health, Denver, CO, United States, <sup>3</sup> Division of Pulmonary Sciences and Critical Care Medicine, University of Colorado Denver, Aurora, CO, United States, <sup>4</sup> Department of Medicine, Denver Veterans Affairs Medical Center, Denver, CO, United States

#### Edited by:

Thomas Dick, Rutgers, The State University of New Jersey, Newark, United States

#### Reviewed by:

Joseph Oliver Falkinham, Virginia Tech, United States Giovanni Delogu, Università Cattolica del Sacro Cuore, Italy

> \*Correspondence: Jennifer R. Honda hondaj@njhealth.org

#### Specialty section:

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

Received: 26 June 2018 Accepted: 10 August 2018 Published: 30 August 2018

#### Citation:

Honda JR, Virdi R and Chan ED (2018) Global Environmental Nontuberculous Mycobacteria and Their Contemporaneous Man-Made and Natural Niches. Front. Microbiol. 9:2029. doi: 10.3389/fmicb.2018.02029 Seminal microbiological work of environmental nontuberculous mycobacteria (NTM) includes the discovery that NTM inhabit water distribution systems and soil, and that the species of NTM found are geographically diverse. It is likely that patients acquire their infections from repeated exposures to their environments, based on the well-accepted paradigm that water and soil bioaerosols – enriched for NTM – can be inhaled into the lungs. Support comes from reports demonstrating NTM isolated from the lungs of patients are genetically identical to NTM found in their environment. Well documented sources of NTM include peat-rich soils, natural waters, drinking water, hot water heaters, refrigerator taps, catheters, and environmental amoeba. However, NTM have also been recovered in biofilms from ice machines, heated nebulizers, and heater-cooler units, as well as seat dust from theaters, vacuum cleaners, and cobwebs. New studies on the horizon aim to significantly expand the current knowledge of environmental NTM niches in order to improve our current understanding of the specific ecological factors driving the emergence of NTM lung disease. Specifically, the Hawaiian Island environment is currently being studied as a model to identify other point sources of exposure as it is the U.S. state with the highest number of NTM lung disease cases. Because of its geographic isolation and unique ecosystem, the Hawaiian environment is being probed for correlative factors that may promote environmental NTM colonization.

Keywords: nontuberculous mycobacteria, environments, man-made, natural, Hawaii

### BACKGROUND

Nontuberculous mycobacteria (NTM) and their environments are intricately bound. NTM share these environments with humans and domesticated animals and repeated exposure is a wellaccepted mode of acquiring these infections. While most casual NTM exposures do not results in disease, those with anatomic lung abnormalities of bronchiectasis and emphysema are particularly predisposed to develop NTM lung disease (NTM-LD). However, in individuals without obvious pre-existing risk factors, it is likely that multiple risk factors – some genetic or other acquired factors – may collude to increase their vulnerability. While there have been case reports and small case series linking genetically identical NTM in patients to their home environment, the

specific factors facilitating their acquisition remains poorly characterized. These factors encompass (i) varied sources of infection, (ii) modes of aquisition, and (iii) other physical aspects of the environment such as temperature, humidity, air exchange, surface types, and turbulence created by wind and natural disasters, and (iv) human behaviors, the combination of which are likely to be relatively unique among affected individuals (Honda et al., 2015; Nishiuchi et al., 2017).

Herein, we discuss the traditional environmental niches associated with NTM organisms, but also review the lesser-recognized environmental locales that NTM colonize including non-traditional niches and non-human hosts. We also summarize previous studies that link ecological factors with risk for infection and epidemiological information. Finally, we introduce new, ongoing work to study the particular environmental drivers of NTM emergence in Hawai'i, a geographic location deemed a major hot spot for NTM-LD.

### NTM ORIGINS?

The interaction of humans with their natural and built environments along with changes humans make to their environment (e.g., installation of mechanical devices that change environmental temperature and humidity, or others), and differences in the robustness of human health may impact the emergence of infectious diseases. The origins of NTM-LD remain a mystery, but may be intrinsically linked to human interaction with their environments. As soon as early humans learned to prepare small, controlled fires, the light and heat produced gathered people together for social interactions, vastly improved food preparation, and helped communities defend against invasion – activities that significantly lengthened human survival (Chisholm et al., 2016). Because NTM are found in the environment, it is plausible that these bacteria were historically aerosolized from soil with increased fire-making, infecting human lungs already impacted by long-term exposure to campfire smoke. Nonetheless, there is no evidence to show that the number of NTM-LD cases increase after camping or large scale environmental fires. For now, the origins of lung disease caused by NTM organisms remains an area of investigation.

### GLOBAL GEOGRAPHY AND NTM

The ability of NTM to cause LD and its clinical relevance varies globally. Geographic distribution of NTM species provides information regarding geographic-specific drivers of exposure such as climate, environment, and host factors associated with NTM-LD that would be specific to a global region (Falkinham, 1996; Griffith et al., 2007).

Mycobacterium avium complex (MAC) is the most frequently isolated group of NTM species worldwide and the most common organism associated with NTM-LD. MAC consists of various species of slow-growing mycobacteria (SGM) including M. avium, M. intracellulare, M. chimaera, M. colombiense, M. marseillense, M. arosiense, M. timonense, M. bouchedurhonense, and M. ituriense. Subspecies of M. avium include avium, silvaticum, hominissuis, and paratuberculosis (Hoefsloot et al., 2013; Johnson and Odell, 2014). Currently, the lowest number of MAC isolates in the world are seen in South America, but this may be because there is little NTM information available from this region (Hoefsloot et al., 2013; Halstrom et al., 2015; Stout et al., 2016). High incidence of NTM-LD due to MAC, M. kansasii, M. gordonae, and M. malmoense are observed in Europe, North America and Australia.

Rapid-growing mycobacteria (RGM) including the Mycobacterium abscessus and Mycobacterium fortuitum groups also contribute to a large proportion of NTM-LD cases globally. The Mycobacterium abscessus group is comprised of three subspecies: M. abscessus subsp. abscessus, M. abscessus subsp. massiliense, and M. abscessus subsp. bolletii. In the United States (U.S.), M. abscessus complex infections are secondary only to MAC infections, comprising 3-13% of all NTM-LD cases (Lee et al., 2015). The M. abscessus group is also commonly observed in patients in East Asia; in Taiwan 17.2% of all clinical NTM isolates belong to the M. abscessus group (Lai et al., 2010). The M. fortuitum group comprises M. fortuitum, M. peregrinum, M. senegalense, M. alvei, M. houstonense, M. neworleansense, M. boenickei, M. septicum, and M. porcinum (Adekambi and Drancourt, 2004; Schinsky et al., 2004). Besides LD, M. fortuitum cause soft tissue, skeletal, catheter-related, and disseminated infections in immunocompromised patients (Brown-Elliott and Wallace, 2002).

### United States

The most commonly occurring NTM species in many parts of U.S. are MAC and M. kansasii (Hoefsloot et al., 2013). Nearly 80% of all NTM-LD in the U.S. is due to a species of MAC, followed by M, kansasii - the second most common NTM associated with LD (Falkinham, 1996; Griffith et al., 2007). The U.S. states with the highest overall risk for NTM-LD include Hawai'i, California, New York, Louisiana, Pennsylvania, Florida, Oklahoma and Wisconsin (Adjemian et al., 2012a). M. abscessus is the most commonly recovered RGM from southeastern parts of U.S. (Halstrom et al., 2015; Stout et al., 2016). For unknown reasons, NTM-LD cases are lowest in North Dakota, South Dakota, Minnesota, Michigan, New Mexico, and West Virgina (Adjemian et al., 2012a).

### Europe and United Kingdom

Higher isolation rates of MAC (44%) are seen in northern Europe as compared to southern Europe (31%) with M. avium as the most prevalent species. M. kansasii is the predominant NTM species to cause LD in London, United Kingdom. and M. lentiflavum is most frequently isolated from clinical samples in Crete, Greece (Neonakis et al., 2007; Wassilew et al., 2016). Slovakia, Poland, and the United Kingdom have the highest amount of M. kansasii isolates in Europe (Hoefsloot et al., 2013; Wassilew et al., 2016). In contrast, higher isolation rates for M. xenopi are observed in southern Europe (21%) as compared to northern Europe (6%), but are most commonly isolated in Hungary (46%) (Hoefsloot et al., 2013; Wassilew et al., 2016). RGM are more commonly seen in the United Kingdom and

Greece as compared to the rest of Europe (Hoefsloot et al., 2013). While not considered to be a pathogen, M. gordonae is most commonly recovered from environmental sources in Canada and Europe (Hoefsloot et al., 2013; Halstrom et al., 2015; Stout et al., 2016).

## Asia

A recent study in China observed that an increase in latitude was associated with higher isolation rates of MAC species (predominantly M. intracellulare) whereas the number of RGM (most commonly M. chelonae) increased with a decrease in latitude (Yu et al., 2016). A similar trend is observed in Taiwan, with higher cases of NTM due to MAC recovered in the north and RGM like M. abscessus in the south (Huang et al., 2017). M. scrofulaceum and M. szulgai are also intermittently found in respiratory specimens from Asia (Simons et al., 2011). Overall, elderly women are disproportionately affected by NTM-LD, but in Saudi Arabia and most of the Persian Gulf countries, elderly men are found to be more affected (perhaps due to a lifetime of extended outdoor exposure) with MAC and M. abscessus being the main causative agents (Al-Ghafli and Al-Hajoj, 2017). MAC, M. simiae, and M. marinum are most commonly observed in NTM-LD individuals in Oman (Al-Ghafli and Al-Hajoj, 2017). In Western Asia, M. fortuitum and M. flavescens are the most prevalent RGM and SGM organisms, respectively with much higher frequency of NTM in northern Iran (73.2%) (Khaledi et al., 2016). In India, M. abscessus, M. fortuitum, and M. intracellulare are most commonly isolated from clinical samples (Desikan et al., 2017). M. abscessus are more widespread in Singapore and Okinawa (Hoefsloot et al., 2013).

### Australia

Similar to the aforementioned geographic areas, LD caused by NTM are also increasingly observed in Australia (Thomson, 2010). MAC is the most commonly isolated NTM in Queensland with M. intracellulare comprising nearly 80% of the MAC isolates (Hoefsloot et al., 2013). RGM are the second most common cause of NTM-LD with M. abscessus the most commonly recovered from southern Australia (Hoefsloot et al., 2013; Wassilew et al., 2016). Unlike many other regions of the world, NTM are notifiable infections under the Queensland Public Health Act, 2005 which has facilitated the surveillance of potentially highly virulent and transmissible NTM strains (Thomson et al., 2017).

### Africa

Information regarding the causative agents of NTM-LD in Africa are limited and is likely due to the overwhelming burden of tuberculosis in the regions. However, M. abscessus, M. avium, M. fortuitum, and M. nebraskense are recognized as the most frequently isolated NTM species from clinical samples in Zambia (Monde et al., 2018). M. gordonae has been recently found to be highly prevalent in water reservoirs like borehole wells, rivers dams, and tap water (Monde et al., 2018).

## FACTORS THAT HELP SUSTAIN NTM IN THE ENVIRONMENT

NTM are slow-growing compared to other types of bacteria, with the ability to form biofilms, resist high temperatures, and grow in marginal environments with low nutrient and oxygen content (Kirschner et al., 1992; Falkinham, 2009). However, cell surface hydrophobicity is the major driver sustaining NTM in the biofilms of both natural waters and man-made drinking water distribution systems, hospitals, and household plumbing. Due to their repulsion to water, NTM are found in aerosolized particles present above natural water bodies, showerheads, humidifiers, hot tubs and spas as well as in biofilms that form in these places (Falkinham, 2013). Environmental factors such as high humidity levels and high evapotranspiration (movement of water from land to the atmosphere) rates are known to be associated with an increased risk of NTM infection in susceptible individuals (Adjemian et al., 2012a). This is particularly translatable to the recovery of NTM during different seasons. For example, the species of NTM isolated from municipal water distribution systems in Brisbane, Australia differed in the samples collected in summer as compared to those collected in the winter months with higher numbers of M. gordonae, M. kansasii, M. abscessus, M. mucogenicum, and MAC isolated in the winter (Thomson R. M. et al., 2013).

Other environmental factors have a profound impact of NTM viability. For example, NTM have been isolated from water bodies with moderate salinity (1–2% NaCl) like estuaries (Chesapeake Bay) (Kirschner et al., 1992; Falkinham, 2009). But in a separate study, reduced numbers of NTM isolates were observed when water salinity exceeded 2% (Gruft et al., 1981). NTM also favor environments with acidic pH. Humic and fulvic acids and acidic brown water swamps along the southeastern coast of the U.S. support high numbers of MAC (Falkinham, 2009, 2013). Pine forest (boreal rich) and peat rich soils, brackish marshes, and drainage water are also rich in NTM (Kirschner et al., 1992; Falkinham, 2009). Minerals widely found in clay soils such as kaolin and dust have also been demonstrated to facilitate the growth of M. abscessus (Malcolm et al., 2017).

## TYPICAL ENVIRONMENTAL HABITATS OF NTM

M. avium, M. fortuitum, M. chelonae, M. kansasii, M. gordonae, and M. xenopi are the NTM species most commonly found in water distribution systems, water bodies including lakes, rivers and streams as well as soil and dust (Falkinham, 2009; Ulmann et al., 2015). Falkinham et al. (2008) was the first to demonstrate that the M. avium isolated from a patient with NTM-LD had a clonal relationship with the M. avium isolated from her home showerhead biofilm. Falkinham also first reported identical NTM DNA fingerprints from patients' sputa and matched shower water isolates, shedding light on the paradigm that inhalation of aerosols while showering is a likely mode of NTM acquisition (Falkinham, 2011). Similar

reports are observed from Japan where MAC isolates recovered from bathtub inlets and showerheads showed identical pulse-field gel electrophoresis profiles when compared to their respective clinical isolates (Nishiuchi et al., 2009; Ichijo et al., 2014). Thus, showerhead biofilms remain one of the most frequently sampled environmental sources used to describe the presence of NTM globally. In an interesting turn of events, early methods such as hybridization probes and multiplex 16S rRNA gene PCR methodologies identified M. intracellulare in households and potable water (Tichenor et al., 2012; Whiley et al., 2012). However, Falkinham et al. (2008) and Wallace et al. (2013) reanalyzed environmental household water and biofilm isolates originally identified as M. intracellulare by sequencing the 280 bp 16S to 23S internal transcribed spacer region and discovered that these isolates were instead M. chimaera (Thomson, 2010; Falkinham, 2011; Koh W. J. et al., 2012; Tichenor et al., 2012). Using the same sequencing method, isolates originally called M. avium were confirmed as M. avium. Thus, M. chimaera is now widely recognized in water biofilms, while M. intracellulare is found to be absent from them. A clue to their sources may come from a study conducted in American Samoa where M. intracellulare was identified in roofharvested rainwater (Kirs et al., 2017) and soil (Honda et al., 2016).

NTM found in households are likely piped into home plumbing systems from public utility sources where biofilms are commonly formed. However, NTM are also known inhabitants of natural freshwater ecosystems. Of two recreational lakes, RGM were the dominant NTM, but a diversity of other mycobacteria were found in high density in the water column, air-water interface, sediment, and in association with benthic algae growing on plants and fine sediment using quantitative real-time PCR and the MiSeq Illumina platform (Roguet et al., 2016). Yet, NTM remain seldomly recovered from well and groundwaters (Martin et al., 1987).

Nosocomial NTM lung infections have been reported in the literature. MAC species have been detected in hospital potable hot water distribution systems, hospital tap water used for dialysis, and in water used to prepare medical solutions, highlighting their propensity to stick to piped surfaces (Bolan et al., 1985; Safranek et al., 1987; Hector et al., 1992). After a significant increase in NTM-positive sputa was observed from patients referred to respiratory wards in Rome, an infection control investigation revealed a massive presence of NTM in the hospital water network (D'Antonio et al., 2016). In another study, 83% of U.S. dialyses centers examined showed NTM in municipal water supplies (Carson et al., 1988). More recently, global outbreaks of M. chimaera associated with heater-cooler units used during open-heart surgery have provided unique challenges for the medical community (Sax et al., 2015; Schreiber et al., 2016; Marra et al., 2017). Investigations point to model-specific designs in air flow direction, location of cooling ventilators, and the continuous cooling of unit water tanks significantly increased the risk of disseminating colonized M. chimaera into the air of operating rooms (Kuehl et al., 2018). Of the first thirty cases affected by this outbreak in the United Kingdom, 60% (18/30) died at a median of 30 months after initial surgery (Scriven et al., 2018). Poor disinfection and resistance to glutaraldehydes have been highlighted in pseudo-outbreaks of M. abscessus subsp. bolletii in bronchoscopes, endoscopes, and disinfection units (Guimaraes et al., 2016). In other areas of the hospital, patient accessible ice machines have been shown to be laden with NTM, particularly M. paraffinicum and M. fortuitum (Gebo et al., 2002; Wang et al., 2009).

M. fortuitum skin infections have been associated with pedicure-associated whirlpool footbaths in California and Georgia nail salons (Winthrop et al., 2004). Skin infections due to other NTM such as M. chelonae have been linked to contaminated water used for diluting tattoo ink and to unsterilized instrumentation (Mudedla et al., 2015). Dental unit water lines have also been shown to harbor a variety of environmental NTM species (Schulze-Robbecke et al., 1995), but remain as unproven sources of NTM lung infection. MACassociated hypersensitivity pneumonitis (HP) has been linked to exposure to warm, bubbly water found in rarely cleaned hot tubs and spa baths (Embil et al., 1997; Sugita et al., 2000; Rickman et al., 2002). However, MAC has also been isolated from cold water sources including swimming pool water. In fact, lifeguards with long-term exposure to indoor swimming pool aerosols are susceptible to work-associated exposures and are at increased risk for MAC-associated HP (Rose et al., 1998; Koschel et al., 2006). Besides tap water, a Dutch group found NTM in swimming pools and whirlpool water (Havelaar et al., 1985). NTM-associated HP has also been linked to occupational exposures to aerosols produced through manipulation of metalworking fluids (Bernstein et al., 1995; Wilson et al., 2001).

Soil is also a widely recognized environmental niche for NTM. NTM patients' potting soils yielded NTM that were identical by DNA fingerprinting to the NTM isolates from the same patients' lungs (Ichiyama et al., 1988; De Groote et al., 2006). Moreover, M. avium subsp. hominissuis (MAH) was predominant in soil and dust, but not identified in German water and biofilm samples by culture (Lahiri et al., 2014). A study in Iran found that 6- 15% of soil samples compared to 10–27% of water samples collected from the suburbs of Tehran had NTM isolated by culture with M. farcinogens and M. fortuitum being the most common species (Velayati et al., 2014). Furthermore, the risk for acquiring NTM is significantly higher in communities engaged in occupations that generate aerosols and are exposed to soil for a longer time (e.g., agriculture, mining, landscaping, and tunnel work) as compared to communities that have a limited exposure to soil (Hamada et al., 2016). In West Harima, Japan, NTM-LD was associated with natural resource activities, construction, mining, and soil exposure (Hamada et al., 2016). In a separate study, M. chelonae, M. fortuitum, and M. kansasii were identified in 85% of the alpine and subalpine soil, peat, humus, porous ricks, mosses, and wood examined suggesting NTM thrive in mountain ranges and elevations (Thorel et al., 2004). Rare and unique NTM species have also been described in polluted soils of Hawai'i where they functioned as polycyclic aromatic hydrocarbon pollutant-degrading organisms (Hennessee et al., 2009).

#### NTM IN THE KITCHEN

fmicb-09-02029 August 29, 2018 Time: 10:35 # 5

NTM have been reported in kitchen sink biofilms as well as household refrigerator taps and home ice machines (Ichijo et al., 2014). MAC organisms have been found to colonize point-of-use filters used to filter tap water including carbon filers impregnated with silver (Rodgers et al., 1999; Hamilton et al., 2017). Due to the appearance of disseminated M. avium infections during the height of the HIV-AIDS epidemic, two studies tested for the presence of NTM in foods consumed by HIV-infected patients. PCR typing revealed 29 different mycobacterial isolates in 21% (25/121) of food samples tested; 41% of the 29 samples (n = 12) were M. avium (Yoder et al., 1999). One of the clinical M. avium isolates was identical to a food isolate, suggesting food as a potential source of M. avium infections. In the second study, water, food and soil samples from 290 homes of HIV-infected patients were tested for mycobacteria using DNA probes, serotyping, and multilocus enzyme electrophoresis and compared to clinical isolates (Yajko et al., 1995). Soil, rather than water sources, were found to harbor more M. avium. While not considered a food-borne illness, M. avium subspecies DNA was also identified in raw meats (Lorencova et al., 2014). In contrast, smoked fish products did not show NTM, but 12% of samples collected from pond fish (4%), retail sold fish (61%), and frozen fish (91%) contained NTM DNA (Lorencova et al., 2013). Although unproven, the acid-resistant NTM organisms may remain viable in the stomach where food is consumed and digested. Evidence suggests patients with the nodular bronchiectatic form of NTM-LD have a high prevalence of increased esophageal acid exposure and gastroesophageal reflux disease was found to be significantly associated with RGM organisms (Winthrop et al., 2010; Koh E. et al., 2012).

#### NTM IN NON-CANONICAL ENVIRONMENTS

Besides their well-known habitation in the numerous sources detailed above, rare NTM species (e.g., M. algericum, M. arabiense, M. heraklionense) have been cultured and identified from primary sludge samples of water treatment plants even after decontamination (Makovcova et al., 2015). M. avium, M. gordonae, and M. flavescens have been also been identified in non-traditional water sources including untreated, drinking well water in rural areas of Montana as well as in treated municipal wastewater from arid regions (Richards et al., 2015; Amha et al., 2017). Qualitative assessments were used to determine the risk of MAC exposure in Queensland, Australia, an area that utilizes rainwater catchment systems. Untreated rainwater is commonly used for showering, car washing, toilet flushing, and food preparation (Hamilton et al., 2017). But in this study, rainwater used for drinking presented the greatest risk for MAC infection; yet the species of MAC responsible was not reported. In most cases, disinfection methods to remove potential pathogens from these water sources is a decision left up to the household and can range from no disinfection methods to point-of-use filters, UV irradiation, solar disinfection, chlorine or a combination of them. More work is needed in these areas to determine the impact of using different disinfection methods to reduce exposures in NTM patients who use these water source types.

NTM have also been detected, albeit, rarely in cobwebs above hen nests, soil fertilized with chicken droppings, and moss but more commonly in dust from vacuum cleaners, and air conditioners (Nishiuchi et al., 2009; Kaevska et al., 2011). Of particular significance to smokers is the recovery and identification of M. avium from cigarettes (Eaton et al., 1995). While not directly shown to cause LD, MAC has been identified from clothing washed during laundry cycles suggesting laundry water maybe an unintentional source of household NTM (Yajko et al., 1993). Finally, M. avium was detected in samples of condensation water formed from the coagualation of steam in three different rooms inside the Russian space station, Mir (Kawamura et al., 2001).

#### THE "ENVIRONMENTAL MACROPHAGE" AND NTM

Amoeba are free-living, freshwater associated protozoans that are ubiquitously found in water systems often cohabited by NTM. Many species of amoeba phagocytose free-living bacteria and feed on them; however, some NTM are able to resist and evade degradation. In particular, most species of MAC, including M. avium, M. intracellulare, M. chimaera, M. colombiense, M. arosiense, M. marseillense, M. timonense, and M. bouchedurhonense reside within free-living Acanthamoeba polyphaga and their exocysts as well as A. castellanii found in potable water (Cirillo et al., 1997; Taylor et al., 2009; Ben Salah and Drancourt, 2010). Delafont et al. (2014) demonstrated amoeba-mycobacteria associations in drinking water networks in a year-long sampling study. Nearly 88% of amoeba including the genera Acanthamoeba, Vermamoeba, Echinamoeba, and Protacanthamoeba recovered from drinking water were found to contain M. llatzerense and M. chelonae (Delafont et al., 2014). NTM cultured in amoeba also show increased resistance to antibiotics and enhanced virulence compared to NTM grown in chickens and mice (Cirillo et al., 1997; Falkinham et al., 2001).

#### EVIDENCE FOR ANIMAL AND INSECT RESERVOIRS FOR NTM ASSOCIATED WITH EXTRAPULMONARY INFECTIONS

NTM infections in mammals occur sporadically and are rarely transmissible between animals and seldomly considered bone fide zoonotic diseases. Except under particular scenarios, NTM infections are also generally not transmissible from human to human (Bryant et al., 2016). Nonetheless, NTM have been known to infect animals such as chickens and quails (Meissner and Anz, 1977; Morita et al., 1999). Drug-susceptible M. fortuitum and M. abscessus were identified in a cutaneous lesion on the snout and nostrils of a captive Trichechus inunguis (manatee) in the Amazon (Reisfeld et al., 2018). In captive non-domesticated hoofed animals and in immunosuppressed dogs and cats,

M. avium has been reported as disseminated disease (Thorel et al., 2004). With the potential for cross species infection or transmission in the Serengeti, tissues from wildlife species and indigenous cattle were probed for mycobacteria, revealing M. intracellulare as the most frequently isolated species, followed by M. lentiflavum, M. fortuitum, and M. chelonae/abscessus. MAC organisms were also detected in animal feces and huts from pastoral Uganda (Kankya et al., 2011). Because M. avium isolates from pigs showed shared genetic characteristics to M. avium isolated from humans, pigs have been theorized as potential sources of infection (Bono et al., 1995). Among 1,249 mandibular lymph node samples collected from the wild boar, Sus scrofa, between 2007 and 2011 in Spain, M. chelonae and M. avium represented 61 and 11% of the NTM isolates (Garcia-Jimenez et al., 2015). In a separate study, a low degree of similarity between MAH isolates from Japanese patients and local pigs was found, while there was a high degree of similarity between European patient and pig isolates, suggesting geographic distinctions (Iwamoto et al., 2012).

NTM infections are among the most common chronic disease of aquatic animals (Gcebe et al., 2018). After their original discovery in 1897, NTM organisms continue to cause disease in sea bass, mullets, and amberjacks that live in temperate zones (Lescenko et al., 2003). The most common NTM pathogens of fish include M. chelonae (sea horses), M. avium, and M. fortuitum. M. marinum remains the most typical NTM found in aquatic environments and often coinhabit on shrimp, frogs, eels, oysters, and shellfish (Piersimoni and Scarparo, 2009). M. marinum is one of the only known species of NTM that grow in waters with high salt concentrations (>3% NaCl) (Kirschner et al., 1992; Falkinham, 2009). In some cases, M. marinum cause extrapulmonary infections and cutaneous "fish tank granulomas" in humans. Controlling NTM infections in the aquaculture setting is difficult, relying only on destruction of the infected stocks in the absence of effective treatments (Gcebe et al., 2018). Carassius auratus (goldfish) have been evaluated as a novel in vivo model to study the pathogenesis of M. marinum. After intraperitoneal administration of 10<sup>2</sup> and 10<sup>9</sup> CFU of M. marinum organisms, an acute and chronic infection was respectively observed with high recovery of NTM from inoculated animals (Talaat et al., 1999).

Although seldom reported, reptiles and insects can potentially carry pathogenic NTM. M. chelonae was originally isolated from the lungs of sea turtles in 1903 (Männikkö, 2011), M. intracellulare has been identified in a rusty monitor reptile with lung nodules (Friend and Russell, 1979), and M. szulgai has been isolated from a crocodile showing tuberculosis-like lung lesions (Gcebe et al., 2018). In a separate case, M. gordonae, M. avium, and M. kansasii were isolated off cockroaches from a South Taiwan hospital (Pai et al., 2003).

Mycobacterium ulcerans can infect skin and subcutaneous tissues developing into non-ulcerated nodules or lesions. Although M. ulcerans has not yet been cultured from the environment, its DNA has been detected in low levels in suspended solids/water residues and soil. High concentrations of M. ulcerans DNA are observed in the feces of Australian ringtails and brushtails possums residing in locations where human cases were reported. These findings suggest possums are naturally infected and are potential environmental reservoirs (Fyfe et al., 2010).

### STUDIES LINKING NTM ECOLOGY WITH AVAILABLE EPIDEMIOLOGICAL DATA

To understand the environmental risk factors for infection, most studies described above have either probed for the occurrence of NTM in various environmental sources or in pulmonary samples from various geographic areas. Yet, the number of studies that have overlaid and integrated ecological and geographic information with epidemiological information of NTM-LD risk are scant. However, an exciting recent study has enhanced our understanding of how ecology relates to risk for NTM infection. In Colorado, soil acidity, low manganese concentrations, and silt were significantly associated with increased disease risk for NTM (Lipner et al., 2017). This same study also identified high-risk clusters of NTM-LD and high-risk watershed locations in the same geographic region using spatial scanning methods. Another study from Queensland, Australia (2001–2010) investigated the associations between climate, soil characteristics, land use, demographic, and socio-economic variables with spatial patterns of NTM infection (Chou et al., 2014). High risk clusters of M. kansasii, M. intracellulare, and M. abscesses infections were found to be associated with areas of high agricultural, mining, and tourism activity.

Higher rates of NTM lung infections have been associated with oceanic coastlines, accounting for 70% of annual NTM cases in the U.S. (Strollo et al., 2015). Statistically significant increases in NTM identification have been reported in the U.S. Affiliated Pacific Islands (USAPI) including American Samoa, Guam, Northern Marina Islands, Palau, Marshall Islands, and the Federated States of Micronesia (Lin et al., 2018). After analyzing respiratory cultures submitted for species identification between 2007 and 2011 at a clinical reference laboratory, the overall period prevalence of NTM isolation in this study was 106 cases/100,000 persons. The authors acknowledge that the prevalence of NTM isolation varied by island nation and may be related to urbanization. The lowest period prevalence of NTM isolation (22 cases/100,000 persons) was reported in American Samoa with 87.2% of patients living in high urban areas. The highest period prevalence (164/100,000 cases) was observed in respiratory samples from the Federated States of Micronesia which also showed the lowest percentage of urban dwellers (22.4%).

Of interest, the first nationwide population-based analysis on the prevalence of NTM-LD found that Hawai'i had the highest period prevalence (1997–2007) of any U.S. state with 396 cases/100,000 persons among persons > 65 years-old (Adjemian et al., 2012b), a rate that is four times greater than the national average. Using national Medicare claims, U.S. Census data, as well as U.S. Geological and Forest Services environmental and climatic data, high and low risk U.S. counties were identified (Adjemian et al., 2012a). High-risk regions for NTM were noted in particular geographic areas of

Hawai'i. Higher daily evapotranspiration levels, higher number of surfaces covered by water, higher soil copper and sodium, and lower manganese levels were characteristics of high risk areas. More ominous, Hawai'i also shows the highest, national age-adjusted mortality rates from NTM-LD (Mirsaeidi et al., 2014).

In a recent large-scale study, researchers partnered with a network of citizen scientists to collect showerhead biofilm samples from 606 households across 49 of 50 U.S. states to simultaneously superimpose the identities of showerhead microbes over geographic areas where NTM-LD is most prevalent (Gebert et al., 2018). Multiple concluding findings were reported that reinforce known paradigms. First, Mycobacteria was the most abundant group of bacteria identified in showerhead biofilms by 16S rRNA gene sequencing, corroborating prior studies (Feazel et al., 2009; Thomson R. et al., 2013). Next, the most frequently identified NTM species by hsp65 gene sequencing were MAC, M. abscessus, and the M. fortuitum complex which are also the most problematic NTM for patients. As already observed (Garcia et al., 2013; Ovrutsky et al., 2013), mycobacteria often co-occurred with free-living amoeba. Second, similar to other studies (Feazel et al., 2009), homes using municipal treated water showed twice more mycobacteria than homes using water from wells and the former contained higher chlorine and iron concentrations. Third, by overlapping these data with previously reported epidemiological information, this study is the first to show showerhead-associated mycobacteria with the potential to cause pathogenic LD are indeed found in U.S. households in regions with the highest number of NTM-LD cases including Florida, southern California, and northeast states. It is clear mycobacterial lineages showed distinct geographic distributions. Finally, supporting prior work, this study confirms Hawai'i as the state with the most abundant mycobacterial showerhead biofilms and show these are predominated by MAC and M. abscessus organisms. Undoubtedly, Hawai'i is a U.S. hot spot for NTM-LD (**Figure 1A**) and the islands' unique characteristics may contribute to greater exposures.

#### MOVEMENT TO COLLECTIVELY UNDERSTAND NTM ECOLOGY, EPIDEMIOLOGY, AND ORGANISM VIRULENCE IN THE HAWAIIAN ISLAND ENVIRONMENT

We have previously shown increased prevalence of NTM infection in Hawai'i using patient laboratory data from a representative population of in-state residents enrolled in a closed healthcare system (Adjemian et al., 2017). Indeed, the most frequently isolated species was MAC, followed by M. fortuitum that was more common among Vietnamese and Korean patients, while M. abscessus was more associated with Japanese and Filipino patients. In this study, Native Hawaiian/Other Pacific Islander populations were less likely to have NTM infection than other racial/ethnic groups examined. A logical hypothesis is that native populations inhabited the islands for a longer period of time, evolving and developing resistance to infection.

We have also previously demonstrated not only the overwhelming presence of MAC in both clinical samples in respiratory specimens from Hawai'i, but also in water biofilm samples collected from households in Hawai'i (Honda et al., 2016). Using partial rpoB gene sequencing, M. avium was not identified in any of the samples examined, providing a contrast to previous studies from the continental U.S. M. intracellulare was found in 27% (4/15) SGM respiratory specimens examined and a soil sample. While absent from soil, M. chelonae was significantly more common in kitchen sink biofilms (9/35, 35%) compared to bathroom sink biofilms (3/34, 9%) and M. abscessus was found equally in kitchen (5/34, 15%) and bathroom sink (4/30, 13%) biofilms. M. porcinum was also frequently recovered

from showerhead biofilm samples, a species normally associated with swine, but also a cause of human infections (Wallace et al., 2004; Brown-Elliott et al., 2011; Perez-Martinez et al., 2013). This observation may be particularly important in Hawai'i, which has a sizeable feral pig population (Hess et al., 2006). Importantly, we identified M. chimaera as the predominant species of MAC in these islands; in fact, M. chimaera was recovered from 56% of the 75 environmental samples tested and 67% of ten clinical isolates tested (Honda et al., 2016). By comparison, of 8,800 isolates analyzed using rpoB gene sequencing in 26 months at National Jewish Health, only 6% were M. chimaera (Dr. Max Salfinger, National Jewish Health, personal communication). Taken together, M. chimaera is emerging as a major NTM species of interest in Hawai'i, underscoring the need for further studies to define the drivers for NTM emergence there and in other Pacific Islands.

In new and on-going work to understand the Hawai'i specific environmental, host, and NTM factors that contribute to NTM-LD emergence (**Figure 1B**), we are synergizing environmental and human behavioral/genetic findings with microbiological and NTM genomic data in dynamic statistical, spatial/temporal models to uncover not only disease drivers, but also potential points of intervention to prevent future infections. Isolates collected through this work will also be applied in future studies of NTM virulence. To accomplish this, a complementary team has been formed including a NTM microbiologist born and raised in Hawai'i with significant ties to the local community, a mycobacterial pulmonologist, epidemiologist, earth geochemist, climatologist, microbial ecologist, volcanic scientists, ecological modelers, local clinicians, and a team of microbial genomic and computational scientists as well as Hawai'i residents, high school/college students, their mentors, and local pig hunters.

#### REFERENCES


### FUTURE DIRECTIONS

Future studies should investigate the: (i) factors that drive the relative absence of NTM from seawater, but association with higher humidity and associated biofilms collected from different areas of the world, (ii) NTM prevalence and diversity in areas of the world with endemic tuberculosis, (iii) role of animal and protozoal reservoirs in the maintenance and spread of environmental NTM, and (iv) contributions of temperature, evapotranspiration, and air pollution including climatic determinants of disease. In the meantime, we spotlight the NTM crisis in Hawai'i as a useful model system to understand NTM transmission and disease dynamics. New information gathered from this work will then be used and applied to study NTM-LD in other areas of the world where NTM-LD is prevalent and emergent.

### AUTHOR CONTRIBUTIONS

JH conceived the project. JH, RV, and EC wrote and edited the manuscript.

#### ACKNOWLEDGMENTS

JH would like to thank the Shoot for the Cure and Padosi Foundations. The authors also acknowledge Drs. Michael Strong, Stephen Nelson, James Crooks, Krishna Pacifici and Stacey Honda as co-PI and collaborators on our work supported by the National Science Foundation, Ecology, Evolution, and Infectious Disease Program, Award # 1743587. Assistance to create the U.S. map shown in this manuscript provided by James Crooks.



and prevention of nontuberculous mycobacterial diseases. Am. J. Respir. Crit. Care Med. 175, 367–416.


ecosystems of Uganda: public health significance. BMC Public Health 11:320. doi: 10.1186/1471-2458-11-320



**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 Honda, Virdi and Chan. 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.

# Mycobacterium abscessus: Shapeshifter of the Mycobacterial World

#### Keenan Ryan<sup>1</sup> and Thomas F. Byrd<sup>2</sup> \*

<sup>1</sup> Department of Pharmacy, University of New Mexico Hospital, Albuquerque, NM, United States, <sup>2</sup> Department of Medicine, The University of New Mexico School of Medicine, Albuquerque, NM, United States

In this review we will focus on unique aspects of Mycobacterium abscessus (MABS) which we feel earn it the designation of "shapeshifter of the mycobacterial world." We will review its emergence as a distinct species, the recognition and description of MABS subspecies which are only now being clearly defined in terms of pathogenicity, its ability to exist in different forms favoring a saprophytic lifestyle or one more suitable to invasion of mammalian hosts, as well as current challenges in terms of antimicrobial therapy and future directions for research. One can see in the various phases of MABS, a species transitioning from a free living saprophyte to a host-adapted pathogen.

Keywords: glycopeptidolipid, toll-like receptor 2, cystic fibrosis, bronchiectasis, serpentine cording, fibroblasts

#### Edited by:

Thomas Dick, Rutgers, The State University of New Jersey, United States

#### Reviewed by:

Michael Henry Cynamon, Syracuse VA Medical Center, United States Joon Liang Tan, Multimedia University, Malaysia Albertus Johannes Viljoen, IRIM, France

> \*Correspondence: Thomas F. Byrd tbyrd@salud.unm.edu

#### Specialty section:

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

Received: 11 July 2018 Accepted: 16 October 2018 Published: 01 November 2018

#### Citation:

Ryan K and Byrd TF (2018) Mycobacterium abscessus: Shapeshifter of the Mycobacterial World. Front. Microbiol. 9:2642. doi: 10.3389/fmicb.2018.02642

### EVOLVING NOMENCLATURE

Mycobacterium abscessus (MABS), a rapidly growing non-tuberculous mycobacterium (NTM) (Howard and Byrd, 2000) is an emerging pathogen worldwide. Perhaps the earliest case of MABS was reported in 1951 and described infection which occurred in the setting of traumatic knee injury. This infection was characterized by "subcutaneous, abscess-like lesions with a peripherally tuberculoid structure." The unique characteristics of this mycobacterium prompted the investigators to propose it as a new species, Mycobacterium abscessus (Moore and Frerichs, 1953). In more recent times, MABS was considered a subspecies of Mycobacterium chelonae until 1992 when genetic analysis demonstrated that it was a distinct species and it was elevated to species status as MABS (Kusunoki and Ezaki, 1992). This resulted in the realization that most prior reports of M. chelonae lung infection were likely a mischaracterization of actual MABS lung infection as it is now recognized that pulmonary infection is a common clinical manifestation of MABS whereas it is rarely caused by M. chelonae. In recent years, MABS has been further characterized into three distinct subspecies; MABS subspecies abscessus, MABS subspecies bolletii and MABS subspecies massiliense (Adekambi et al., 2017). They differ in terms of drug susceptibility, and may have differences related to transmissibility as well which will be discussed (Tan et al., 2017). Genomic analysis of MABS reveals evidence of horizontal gene transfer (Howard et al., 2002). In addition to genes associated with mycobacterial virulence, genes with similar function to those found in Pseudomonas aeruginosa and Burkholderia cepacia, pathogens commonly found in the lungs of cystic fibrosis (CF) patients, have also been identified (Ripoll et al., 2009). Thus MABS can be thought of as a pathogen uniquely adapted to different niches within the host lung environment.

### EPIDEMIOLOGY AND EMERGENCE AS A PATHOGEN

The incidence of NTM infection is increasing in the United States (Prevots et al., 2010) and worldwide (Prevots and Marras, 2015). Pulmonary infection is the most common clinical

presentation, however, extrapulmonary infection either due to direct inoculation into the skin or due to disseminated disease, often in association with disease modifying anti-TNFα therapy, is also being recognized with increased frequency (Winthrop et al., 2009). In the United States, Mycobacterium avium complex (MAC) is the most common NTM clinical pulmonary isolate followed by MABS (Prevots and Marras, 2015). It has also been reported that there are geographic differences in the types of NTM isolates in the United States. MAC is the most frequent and predominant isolate in the South and Northeast with proportionally higher rates of isolation of MABS, M. chelonae, M. fortuitum, and M. kansasii in the Western United States (Spaulding et al., 2017). It is noteworthy that in parts of Asia MABS is the predominant pulmonary NTM pathogen (Umrao et al., 2016; Nagano et al., 2017; Lim et al., 2018). Furthermore, limited epidemiologic data suggests that there may be a relative genetic susceptibility to MABS infection in certain Asian populations (Adjemian et al., 2017).

Perhaps what is most disconcerting about MABS from an epidemiologic standpoint is its unique ability among NTM to cause outbreaks of infection over wide geographic regions without linkage to a single or specific point source. The best described instance of this relates to a study of postsurgical wound infections with MABS subspecies massiliense in Brazil. Using genomic analysis it was found that a recently emergent single clone caused multiple outbreaks of post-surgical wound infections. The organism spread throughout Brazil and persisted in hospital environments. Evidence was also provided that there is a loss of genetic material from this lineage raising the possibility that it is undergoing reductive evolution as it adapts to its new niche in the hospital environment (Everall et al., 2017). It is noteworthy that this isolate is genetically related to another MABS subspecies massillense clone that has been reported to be circulating among CF treatment centers throughout the world. There is evidence that acquisition of this strain by CF patients may be associated with worse clinical outcomes. Data from cell culture and animal infection experiments provide evidence that this clone may be more virulent in comparison to unclustered MABS subspecies abscessus isolates. Review of this data shows that although the observed differences are statistically significant, they are small (Bryant et al., 2016), and whether they have clinical relevance in terms of virulence remains unclear. Nonetheless these studies raise the disconcerting possibility that evolutionary changes affecting transmissibility and adaptation to mammalian hosts could lead to further spread of MABS into the susceptible general population of patients with abnormal lung airways and/or states of immunosuppression (Winthrop et al., 2009). The possible mechanisms responsible for spread of MABS include fomites, aerosolized airway secretions and contaminated hospital and municipal tap water (Baker et al., 2017; Donohue, 2018). One study has reported an association between warm humid climates and high atmospheric vapor pressure with the prevalence of NTM infection (Prevots and Marras, 2015). This raises the question of whether climate change is contributing to the rising incidence of NTM infection.

## CLINICAL INFECTION

The clinical spectrum of MABS and other rapidly growing mycobacteria has been well-described and broadly categorized as pulmonary and extrapulmonary disease (Howard and Byrd, 2000). MABS extrapulmonary infection can involve a variety of sites, most commonly the skin. It can occur via dissemination in immunosuppressed patients (Scholze et al., 2005), often from an occult source, or from direct inoculation, either iatrogenically [for example, contaminated acupuncture needles, injection solutions, and surgical procedures (Ryu et al., 2005; Yuan et al., 2009; Schnabel et al., 2016)], or as a result of wound contamination in the setting of trauma (Petrini et al., 2006). Infection of many anatomic sites has been reported including the eye, bone, joint, and central nervous system (Chu et al., 2015; Fukui et al., 2015; Baidya et al., 2016; Jeong et al., 2017). Extrapulmonary infection is generally responsive to treatment with antibiotics to which the isolate is susceptible with adjunctive surgical debridement where indicated.

MABS was first recognized as a cause of chronic pulmonary infection in 1993. The majority of pulmonary infections occur in older adults often with no history of cigarette smoking who are otherwise healthy, but who have underlying lung airway abnormalities (Griffith et al., 1993). The clinical presentation is indistinguishable from lung infection caused by MAC. Unlike MAC, however, effective antibiotic treatment options are limited with many patients requiring a combination of medical and surgical intervention for cure (Jarand et al., 2011). In patients with CF, MABS pulmonary infection is an important cause of morbidity and mortality, and is a relative contraindication to lung transplantation (Dorgan and Hadjiliadis, 2014). It is important to note that immune dysregulation as a result of mutations in the CF Transmembrane Conductance Regulator gene (CFTR) contributes to the inflammatory phenotype in CF lung disease, and may result in a pathogenic process that differs from that seen in otherwise healthy individuals who have MABS infection and abnormal lung airways (Rieber et al., 2014). One clinically apparent difference is that MABS chronic colonization of the lung by smooth variants as well as by rough invasive variants is a cause of morbidity in CF patients (Bryant et al., 2016). In otherwise healthy patients with abnormal lung airways MABS colonization is more likely to be transient.

A common characteristic of patients with MABS pulmonary infection is bronchiectasis. An obvious question is whether bronchiectasis is the end result of infection with MABS or a predisposing factor for MABS infection? An experiment of nature provides evidence that bronchiectasis per se predisposes to NTM infection in the absence of preceding inflammation. Patients with the genetic disorder known as primary ciliary dyskinesia have a loss of normal ciliary structure and function leading to development of bronchiectasis which may occur in the absence of antecedent inflammation. These individuals are prone to recurrent oto-sino-pulmonary infections. In addition to lung infection with pyogenic bacteria such as P. aeruginosa and Staphylococcus aureus, these patients are also susceptible to colonization and invasive infection with MABS and MAC (Noone et al., 2004). The other lung airway disease which

predisposes to colonization and infection with NTM such as MABS and MAC is chronic obstructive pulmonary disease (COPD) (Billinger, 2009). Large upper lobe bullous cavities in the lung are a predisposing factor to NTM infection in these patients. It is also noteworthy that bronchiectasis is also present in a high percentage of patients diagnosed with COPD (Patel et al., 2004; Martinez-Garcia et al., 2013) suggesting that the presence of bronchiectasis may be an important factor in MABS and MAC disease pathogenesis in these patients as well.

#### PATHOGENESIS: THE IMPORTANCE OF ROUGH AND SMOOTH COLONY PHENOTYPES

Microdroplet nuclei aerosolized after an individual with pulmonary tuberculosis coughs, with subsequent inhalation by an uninfected host, is the mode of person to person spread of M. tuberculosis. In contrast, epidemiologic studies have not demonstrated this to be an efficient mechanism for acquisition of pulmonary MABS infection although there is experimental evidence that long lived infected aerosols generated by a coughing patient may contain respiratable bacteria (Bryant et al., 2016). Respiratable bacteria could also result from other sources of aerosols contaminated with MABS, for example colonized showerheads. Fomites have also been implicated in transmission of an outbreak strain (Bryant et al., 2016). It would seem unlikely that bacteria adherent to a fomite could be easily re-aerosolized. This mechanism of transmission could involve contact transfer from the fomite to the mouth with subsequent colonization of the oropharynx. Microaspiration of bacteria into the lung would lead to colonization of the lung airways and/or invasive lung infection (Thomson et al., 2007).

We first described rough and smooth colony phenotypes of MABS which arose from a single parental strain, and demonstrated that the rough variant persists in the lungs of SCID mice, replicates in macrophages, and forms corded, invasive microcolonies in fibroblast monolayers. The smooth variant demonstrated none of these characteristics (Byrd and Lyons, 1999). We then identified glycopeptidolipid (GPL) in the cell wall of the smooth variant as being responsible for the smooth colony phenotype and for the ability of the smooth strain to exhibit sliding motility and biofilm formation (Howard et al., 2006; Nessar et al., 2011). Importantly, we demonstrated that a mutant which lacks GPL and exhibits the rough phenotype can spontaneously arise from the smooth phenotype and regain virulence characteristics (Howard et al., 2006). We showed by light and electron microscopy (Byrd and Lyons, 1999; Howard et al., 2006), and in a later publication using scanning electron microscopy (Sanchez-Chardi et al., 2011), that the rough phenotype exhibits cording, a property associated with virulence in M. tuberculosis. These observations were replicated by others, and the ability of spontaneous rough mutants to arise in infected mice from initial infecting MABS smooth variants was subsequently demonstrated (Catherinot et al., 2007; De Groote et al., 2014). The clinical relevance of these observations is supported by studies indicating that the

majority of clinical isolates from individuals with chronic lung disease exhibit the rough phenotype whereas environmental contaminants and isolates from wounds exhibit the smooth phenotype (Jonsson et al., 2007). A study which longitudinally characterized MABS isolates from ten CF and three non-CF patients over a 10 year period found that a switch from smooth to rough colony morphology was observed in 6 of the patients during the course of long-term infection and was associated with increased severity of clinical symptoms (Kreutzfeldt et al., 2013). In addition, it has been documented that deterioration of lung function in a MABS-infected CF patient was associated with conversion of a sputum isolates from the smooth phenotype to an isogenic rough phenotype (Catherinot et al., 2009). These observations are consistent with our hypothesis that MABS initially gains entry to the lung as a GPL-expressing smooth variant which colonizes abnormal lung airways, and that spontaneous loss of GPL expression leads to a virulent phenotype capable of causing inflammation and invasive lung disease (Howard et al., 2006; Rhoades et al., 2009). In support of this pathogenic mechanism, we have provided evidence that MABS GPL is immunologically inert and that loss of GPL unmasks underlying MABS cell wall molecules such phosphatidyl-myo-inositol mannosides (PIMs) which are recognized by toll like receptor 2 (TLR2) on macrophages and epithelial cells thereby initiating an inflammatory response. Establishment of a smooth strain in the lungs of individuals with normal lung airways is likely prevented by the normal lung mucociliary clearance mechanism. This prevents establishment of lung airway biofilm from which invasive rough variants which have lost GPL can emerge. In patients with abnormal lung airways and diminished mucociliary clearance, MABS smooth variants have been demonstrated to establish biofilm in lung airways (Qvist et al., 2015). In CF patients, heavy colonization of lung airways with smooth variants in the absence of lung invasion is also likely to contribute to morbidity. Evidence indicates that genetic mutation involving a gene coding for one of the MmpL proteins involved in the biosynthesis of the MABS cell envelope is responsible for the smooth to rough transition in one MABS strain (Bernut et al., 2016b). However, we have presented evidence that in some strains of MABS, expression of GPL is temperature dependent and reversible, with expression at lower temperatures and loss of expression at higher temperatures (Rhoades et al., 2009). This suggests that alternative mechanisms regulating GPL production in the smooth to rough transition are operating. In addition, temperature-dependent transitioning of the smooth to rough phenotype could be a mechanism whereby a GPL expressing environmental strain gains access to abnormal lung airways where higher internal core temperature of the lungs would result in loss of GPL and emergence of the rough virulent phenotype. Our current model of MABS pathogenesis is summarized in **Figure 1**.

With the exception of MABS rough strains from chronically infected CF patients contaminating the hospital/surgical or CF clinic environment, the majority of environmental strains have the smooth colony phenotype (Jonsson et al., 2007; Baker et al., 2017). Direct acquisition of a MABS rough variant into abnormal lung airways as might occur in a contaminated clinical

environment would likely result in a less indolent infection due to the greater virulence of rough variants and stimulation of a robust innate immune response.

### COMPARISON TO OTHER CLINICALLY SIGNIFICANT NTM

Other clinically significant NTM may express some form of GPL or exhibit cording, but there are few examples of other NTM which can transition between both phenotypes with such a clear correlation between colonization and invasion. For example, MAC expresses serovar-specific GPLs that differ from the nonspecific core GPL found in MABS through modification by addition of oligosaccharides (Brennan et al., 1981). In contrast to MABS GPL, these molecules are immunogenic and are associated with MAC virulence (Barrow et al., 1995; Sweet and Schorey, 2006). M. kansasii variants may have a smooth colony phenotype associated with expression of characteristic lipids distinct from GPLs (Nataraj et al., 2015). Both MAC and M. kansasii form microscopic bacterial aggregates in broth culture, but do not exhibit the serpentine cording found in M. tuberculosis and MABS that is associated with mycobacterial virulence (Tu et al., 2003). In contrast, M. marinum does not express GPLs, but does exhibit serpentine cording (Hall-Stoodley et al., 2006). Thus, MABS may be viewed as a pathogen which has a unique ability to shift from a ubiquitous environmental saprophyte to an invasive human pathogen – this transformation is apparent upon inspection of the shape of its bacterial colonies growing on nutrient agar.

## THE PARADOX OF LOW VIRULENCE AND INEFFECTIVE HOST RESPONSE

Perhaps the most puzzling aspect of MABS pulmonary infection is that it rarely occurs in immunocompetent individuals with normal lung airways in spite of the ability of this pathogen to invade and replicate in mononuclear phagocytes and nonprofessional phagocytes (Byrd and Lyons, 1999). Determining the reason(s) why has been impeded by the lack of a suitable mouse model that mimics human lung bronchiectasis. For example, mice with in which the CFTR gene has been disrupted have organ specific pathology which is mild in the lung (Clarke et al., 1994).

A non-CF mouse model utilizes GM-CSF knock out mice in which a chronic lung infection can be established. In this model, colony forming units in mouse lung persist and gradually decline over a month but then begin to increase at 2 months – at that time the mice have begun to develop bronchiectasis. The limitations of this model are that the mice are immunocompromised, and that bronchiectasis develops late in the course of infection (De Groote et al., 2014). In spite of the current lack of an ideal model, a pattern of bacterial – host interaction is beginning to emerge based on clinical and experimental evidence to date.

At the cellular level, important differences have been reported regarding the intracellular life cycle of MABS smooth and rough variants. Since both variants replicate in tissue culture medium assessing growth in macrophages requires extensive washing after infection, incubation with amikacin to kill extracellular bacteria and then removal of amikacin followed by lysis of macrophages and plating lysates for bacterial CFU. Using this model, rough variants replicate intracellularly resulting in lysis of cell monolayers whereas smooth variants persist and/or decline within cells, but do not destroy cell monolayers in the time frames studied in these experiments (Byrd and Lyons, 1999; Howard et al., 2006; Greendyke and Byrd, 2008; Nessar et al., 2011). Recent studies have demonstrated differences in intracellular behavior comparing smooth and rough variants which may account for these findings. Smooth variants have been found to reside in phagosomes in which they are surrounded by electron translucent zone representing cell wall GPL. They are typically present as a single organism which is likely due to the fact that they are easily dispersed into single cell suspensions prior to infecting cells. In contrast, rough variants clump in a manner similar to M. tuberculosis and are difficult to get into single cells suspension, thus they are usually present as two or more bacteria per phagosome (Byrd and Lyons, 1999; Roux et al., 2016). Consistent with the lack of destruction of macrophage monolayers infected with smooth variants (Nessar et al., 2011) is the finding that smooth variants are poor inducers of apoptosis and autophagy in contrast to rough variants which end up in phagolysosomes but nonetheless replicate intracellularly and induce apoptosis (Roux et al., 2016). In fact, uptake of large clumps of MABS rough variants leads to rapid dissolution of macrophages and emergence of MABS cords (Brambilla et al., 2016). There is evidence that smooth variants cause disruption of the phagosomal membrane allowing for direct cytosolic contact (Roux et al., 2016). Thus, as a mechanism of apoptosis inhibition, polar GPLs found on the surface of MABS smooth variants (Lopez-Marin et al., 1994; Howard et al., 2006) have access to cytosolic contents and have been demonstrated to interact with mitochondrial cyclophilin D, a component of the mitochondrial permeability transition pore (MPTP) to stabilize the pore and inhibit apoptosis (Whang et al., 2017).

MABS has an ESX-4 type VII secretion system encoded for by five genes. A gene in this locus, eccB4 was found to be necessary for low level replication (approximately 0.5 log over 5 days), but not persistence, of the smooth MABS subspecies massiliense 43S strain in Acanthamoeba castellani and J774.2 mouse macrophages. This gene was also found to be necessary for prevention of phagosome acidification and for causing phagosome rupture (Laencina et al., 2018). Other studies have found that under stringent conditions to prevent extracellular replication in tissue culture media, colony forming units of MABS smooth variants in human monocyte-derived macrophage monolayers have limited replicative capacity over a 5 day period and do not persist in the lungs of infected SCID mice (Byrd and Lyons, 1999; Howard et al., 2006). These latter findings raise the question of the significance of the slight loss of replicative ability in the eccB4 deletion mutant. The question of the effect of expression of the genes of the ESX-4 type VII type secretion system by the more virulent rough MABS phenotype is unexplored although as noted, another study did not report phagosome disruption but rather trafficking of rough variants through the autophagic pathway ending up in phagolysosomes where they replicate and rapidly induce apoptosis (Roux et al., 2016). It remains unclear why rough variants, which are able to arise from the smooth variant used in this study (Ripoll et al., 2009), would not also rupture the phagosome. Finally, as would be expected for an intracellular pathogen establishing cytosolic contact and activating the inflammasome, this study found that IL1β release was increased by cells infected with wild type bacteria. There was a significant decrease in IL1β release from cells infected with the eccB4 deletion mutant. Since inflammasome activation results in cell death via pyroptosis (Sharma and Kanneganti, 2016), and human macrophage monolayers infected with MABS smooth variants typically remain intact and viable throughout the course of infection, this discrepancy remains unexplained. It may be that GPL-mediated inhibition of macrophage apoptosis counteracts the effect of inflammasome activation (Whang et al., 2017).

One important function of GPL is that it prevents bacteria to bacteria interaction of underlying surface molecules such as trehalose polyphleates which may play a role in the cording exhibited by rough variants (Llorens-Fons et al., 2017). With loss of GPL, rough variants taken up by macrophages rapidly induce apoptosis (Roux et al., 2016) and demonstrate rapid cell–cell spread via serpentine cord formation (Byrd and Lyons, 1999). The fact that serpentine cording is an important virulence determinant of rough MABS strains was demonstrated using a rough deletion mutant lacking a gene coding for a dehydratase necessary for cord formation. This mutant was found to be markedly attenuated for virulence in the zebrafish model (Halloum et al., 2016). A comparison of the behavior of MABS rough and smooth variants to the virulent M. tuberculosis strains H37Rv/Erdman, and the avirulent strain H37Ra shows similar behavior in a fibroblast microcolony assay which we described (Byrd et al., 1998; Byrd and Lyons, 1999). The differences relate to the addition of extracellular acting aminoglycoside antibiotics. Both the MABS rough variant and H37Rv/Erdman demonstrate elongated, corded microcolonies within the plane of the agar-overlaid monolayers, while the MABS smooth variant and H37Ra demonstrate significantly smaller, rounded microcolony morphology. Importantly, the addition of streptomycin does not prevent formation of H37Rv/Erdman microcolonies, presumably due to a mechanism of direct cell– cell spread wherein M. tuberculosis avoids exposure to the extracellular environment. In contrast, addition of amikacin to

fibroblast monolayers infected with the MABS rough variant prevents the formation of corded microcolonies. This suggests that M. tuberculosis is host adapted to favor replication within the intracellular environment while MABS may be viewed as an environmental saprophyte that has not quite made the transition to an intracellular lifestyle.

In spite of the ability to replicate extracellularly and form microabscesses in the zebrafish model of infection, MABS pulmonary infection does not occur in immunocompetent humans (or mice) with normal lung airways. Why this is so is a central question in terms of MABS pathogenesis. TLR2 has been found to be important in innate immune recognition and signaling in response to MABS. PIMs exposed on the surface of rough variants interact directly with TLR2 on human macrophages and epithelial cells to promote release of TNFα and IL-8, respectively, whereas PIMs are "masked" by GPL on the surface of smooth variants which are not recognized by TLR2 (Rhoades et al., 2009; Davidson et al., 2011). Both TNFα and IL-8 have both been found to be important for control of infection by MABS rough variants in the zebrafish model of infection (Bernut et al., 2016a). On the other hand, human neutrophils have been found to be less effective at killing MABS than killing S. aureus, and dead and dying neutrophils have been found to enhance biofilm formation by smooth variants (Malcolm et al., 2013). Thus in abnormal lung airways with impaired mucociliary clearance and MABS smooth variant biofilm formation, ongoing inflammation with IL-8 mediated neutrophil recruitment may promote biofilm persistence. This may have particular relevance in patients with CF who are often co-infected with multiple pulmonary pathogens and in whom chronic lung airway inflammation dominated by the presence of neutrophils is felt to be central to disease pathogenesis (Cantin et al., 2015). Another aspect of the innate immune response relates to single nucleotide polymorphisms (SNPs) that alter host responses to infectious agents. TLR2 signaling in response to mycobacterial lipopeptides depends upon formation of TLR2/TLR1 heterodimers at the cell surface. SNPs that alter the function of either TLR2 and/or TLR1 may thus affect innate immune responses to mycobacteria. We have reported that a well described SNP, TLR1 SNP I602S, is present in the respiratory epithelial cell line CFBE41o-, which was derived from the bronchus of a patient with CF and immortalized with SV40. This cell line is hyporesponsive to TLR2/TLR1 receptor agonist and the rough MABS 390R strain. This SNP is likely to be present in the CF patient population (Kempaiah et al., 2013). Paradoxically the presence of this SNP is protective against infection with Mycobacterium leprae (Johnson et al., 2007). An unanswered question is how the presence of this SNP affects susceptibility to NTM infection in both CF and non-CF patients.

In terms of the cell-mediated immune response, as would be predicted, anti-TNFα inhibitor therapy has been associated with disseminated NTM infection, including MABS (Winthrop et al., 2009). Th1 CD4+ T cell responses are also important for control of infection with NTM. It is established that patients with advanced HIV and low CD4+ T cell counts are susceptible to infection with MAC, and cases of disseminated MABS infection have been reported in these patients as well (Tan et al., 2010). The importance of IFNγ in control of MABS infection is highlighted in a recent case-control study from Thailand in which MABS was found to be the most common NTM clinical isolate, and the presence of anti-IFNγ autoantibody was strongly associated with disseminated infection (Phoompoung et al., 2017). Since defects in cell-mediated immunity have not been identified in the majority of non-CF patients with chronic MABS lung infection, it remains unclear why patients with underlying lung airway abnormalities such as bronchiectasis are uniquely susceptible to infection, and why an effective cell-mediated immune response does not develop in response to pulmonary infection.

### A LIMITED ANTIMICROBIAL ARMAMENTARIUM

In spite of its low virulence compared to pathogens such as M. tuberculosis, MABS infection of the lung is the most difficult NTM to treat, resembling multi-drug resistant tuberculosis. Numerous resistance mechanisms in MABS have been identified that limit the number of available antibiotics in comparison to infections caused by other NTM (Brown-Elliott et al., 2012; Nessar et al., 2012; Luthra et al., 2018). Of the limited antibiotics used to treat MABS infection most are bacteriostatic and not bactericidal for both intra- and extracellular bacteria in vitro (Greendyke and Byrd, 2008; Maurer et al., 2014). MABS biofilm formation has been described for GPL-expressing smooth variants, and biofilm consisting primarily of smooth variants has been found in the lung airways of explanted lungs from CF patients prior to lung transplantation (Qvist et al., 2015). Under certain in vitro culture conditions MABS rough variants grow as biofilms as well (Clary et al., 2018). Biofilms formed by both smooth and rough MABS variants are relatively resistant to antibiotics when compared to planktonic bacteria (Greendyke and Byrd, 2008; Clary et al., 2018). The bacteriostatic activity of currently used antibiotics against intracellular MABS, and the relative antibiotic resistance of MABS biofilms in abnormal lung airways make eradication of MABS from the lung extremely difficult.

Of the antibiotics used to treat MABS infection, macrolides are felt to be the cornerstone of therapy. Importantly, there are differences in macrolide susceptibility among the different MABS subspecies based on the presence and functional status of the erythromycin ribosomal methylation gene 41 (erm41). The erm41 gene encodes for an enzyme that confers intrinsic, inducible resistance in MABS (Nash et al., 2009). In MABS strains with a functional erm41 gene, clarithromycin may initially appear to be active; however, resistance may develop in the time frame of 3–14 days, the Clinical and Laboratory Standard Institute (CLSI) recommendation for length of incubation (Nash et al., 2009; Koh et al., 2011; CLSI, 2011). PCR is now being used to identify MABS subspecies and erm41 gene status (Shallom et al., 2015).

Broth microdilution with determination of minimum inhibitory concentration (MIC) is recommended as the gold standard for NTM antibiotic susceptibility testing. Due to the relative rarity of MABS infections in the past, correlating in vitro modeling and susceptibility testing with effective clinical

response has been limited. In fact, tigecycline, a glycine antibiotic commonly used in treatment of MABS infection, currently lacks MIC interpretation from CLSI (Brown-Elliott et al., 2012). When MABS infection involves sites such as the CNS and bone, it is important to consider pharmacokinetic properties to maximize antibiotics exposure at the target site (Landersdorfer et al., 2009; Nau et al., 2010). Dosing regimens should be individualized to maximize drug exposure at the target site while limiting potential side-effects. The use of inhaled aminoglycosides for pulmonary MABS infection is one strategy to maximize the concentration at the active site while decreasing the likelihood of side-effects. Several small trials of inhaled amikacin for the treatment of NTM, specifically MABS and MAC, have had positive results including negative culture conversion and improvement seen on lung imaging (Olivier et al., 2014; Yagi et al., 2017). Even though serum concentrations are lower with inhaled amikacin it is not completely without risk. In one study, ototoxicity occurred at a relatively high rate (Olivier et al., 2014). Inhaled liposomal amikacin may be a further advancement in localized therapy for MABS. The liposome capsule penetrates biofilms and is also taken up by macrophages within the lung, thus delivering amikacin directly to the site of infection (Zhang et al., 2018). It should be noted that a Phase 2 trial of inhaled liposomal amikacin in addition to standard therapy failed to meet its primary endpoint for both MABS and MAC but did show improvement in culture clearance and 6-minute walk test (Olivier et al., 2017). Inhaled liposomal amikacin has recently been approved in the United States for the treatment of refractory MAC following a Phase 3 trial that demonstrated an increase in sputum clearance at 6 months (Griffith et al., 2018). More data about the use specifically in MABS is still needed. It is apparent that development of new drugs specifically targeting MABS infection should be a high priority and drug discovery efforts are underway utilizing novel high throughput screening platforms (Gupta et al., 2017). There are currently two new antibiotics designed and approved for alternative infectious indications which may add to the MABS antimicrobial armamentarium and are worth mentioning. The first antibiotic is an FDA-approved agent named tedizolid. Compared to the related antibiotic linezolid which is generally active against MABS, tedizolid has a higher ribosomal binding affinity that allows for lower effective serum concentration and once daily dosing (Burdette and Trotman, 2015). In vitro testing of NTM isolates shows promise as MIC values are 1–8 times lower than that of linezolid (Brown-Elliott

#### REFERENCES


and Wallace, 2017). Importantly, tedizolid has a lower incidence of side-effects and drug interactions as compared to linezolid (Burdette and Trotman, 2015). There is little data on longterm use of tedizolid that has been reported, but overall it appears to be well tolerated (Nigo et al., 2018). The second antibiotic is avibactam, a non-β-lactam β-lactamase inhibitor commercially available as a coformulation with ceftazidime, which can inactivate the MABS β-lactmase Blamab. When avibactam is combined with β-lactams, this combination may have enhanced activity for the treatment of MABS infection (Kaushik et al., 2017; Le Run et al., 2018).

#### FUTURE DIRECTIONS

Mycobacterium abscessus (MABS) has emerged as a significant infectious disease threat and warrants the designation of "shapeshifter of the mycobacterial world." Its ability to exist as a GPL-expressing environmental saprophyte forming biofilms along with the ability to "unmask" itself within the human host and display virulence properties such as serpentine cording is unprecedented for a bacterial pathogen. The ability of evolving clones to spread among clinical environments foreshadows an increasing incidence of nosocomial infections. The fact that many strains are multidrug resistant and that the MABS subspecies differ in terms of their inherent antimicrobial susceptibility creates an enormous challenge, particularly since patients often require months of therapy to achieve cure. There is an urgent need for better models that mimic human pulmonary infection to better understand disease pathogenesis and test new compounds for antimicrobial activity.

#### AUTHOR CONTRIBUTIONS

KR contributed ideas and expertise, and drafted the section related to antimicrobial therapy. TFB conceived and wrote the final manuscript.

### FUNDING

Ongoing work in the TFB laboratory is supported by the National Institutes of Health (NIH) grant AI137633.


bolletii unravels a functional Tyr residue conserved in all mycobacterial MmpL family members. Mol. Microbiol. 99, 866–883. doi: 10.1111/mmi.13283



of Mycobacterium abscessus but is absent from Mycobacterium chelonae. Antimicrob. Agents Chemother. 53, 1367–1376. doi: 10.1128/AAC.01275-08


abscessus variants inside macrophages. Open Biol. 6:160185. doi: 10.1098/rsob. 160185


**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 Ryan and Byrd. 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 Role of Antibiotic-Target-Modifying and Antibiotic-Modifying Enzymes in Mycobacterium abscessus Drug Resistance

#### Sakshi Luthra<sup>1</sup> , Anna Rominski<sup>1</sup> and Peter Sander1,2 \*

1 Institute of Medical Microbiology, University of Zurich, Zurich, Switzerland, <sup>2</sup> National Center for Mycobacteria, Zurich, Switzerland

#### Edited by:

Thomas Dick, Rutgers, The State University of New Jersey, Newark, United States

#### Reviewed by:

Uday Suryan Ganapathy, Rutgers, The State University of New Jersey, Newark, United States Juan Manuel Belardinelli, Colorado State University, United States

> \*Correspondence: Peter Sander psander@imm.uzh.ch

#### Specialty section:

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

Received: 02 July 2018 Accepted: 24 August 2018 Published: 12 September 2018

#### Citation:

Luthra S, Rominski A and Sander P (2018) The Role of Antibiotic-Target-Modifying and Antibiotic-Modifying Enzymes in Mycobacterium abscessus Drug Resistance. Front. Microbiol. 9:2179. doi: 10.3389/fmicb.2018.02179 The incidence and prevalence of non-tuberculous mycobacterial (NTM) infections have been increasing worldwide and lately led to an emerging public health problem. Among rapidly growing NTM, Mycobacterium abscessus is the most pathogenic and drug resistant opportunistic germ, responsible for disease manifestations ranging from "curable" skin infections to only "manageable" pulmonary disease. Challenges in M. abscessus treatment stem from the bacteria's high-level innate resistance and comprise long, costly and non-standardized administration of antimicrobial agents, poor treatment outcomes often related to adverse effects and drug toxicities, and high relapse rates. Drug resistance in M. abscessus is conferred by an assortment of mechanisms. Clinically acquired drug resistance is normally conferred by mutations in the target genes. Intrinsic resistance is attributed to low permeability of M. abscessus cell envelope as well as to (multi)drug export systems. However, expression of numerous enzymes by M. abscessus, which can modify either the drug-target or the drug itself, is the key factor for the pathogen's phenomenal resistance to most classes of antibiotics used for treatment of other moderate to severe infectious diseases, like macrolides, aminoglycosides, rifamycins, β-lactams and tetracyclines. In 2009, when M. abscessus genome sequence became available, several research groups worldwide started studying M. abscessus antibiotic resistance mechanisms. At first, lack of tools for M. abscessus genetic manipulation severely delayed research endeavors. Nevertheless, the last 5 years, significant progress has been made towards the development of conditional expression and homologous recombination systems for M. abscessus. As a result of recent research efforts, an erythromycin ribosome methyltransferase, two aminoglycoside acetyltransferases, an aminoglycoside phosphotransferase, a rifamycin ADP-ribosyltransferase, a β-lactamase and a monooxygenase were identified to frame the complex and multifaceted intrinsic resistome of M. abscessus, which clearly contributes to complications in treatment of this highly resistant pathogen. Better knowledge of the underlying mechanisms of drug resistance in M. abscessus could improve selection of more effective chemotherapeutic regimen and promote

**56**

development of novel antimicrobials which can overwhelm the existing resistance mechanisms. This article reviews the currently elucidated molecular mechanisms of antibiotic resistance in M. abscessus, with a focus on its drug-target-modifying and drug-modifying enzymes.

Keywords: non-tuberculous mycobacteria, Mycobacterium abscessus, antibiotic, drug resistance, resistance genes, antibiotic-target-modifying enzymes, antibiotic-modifying enzymes

#### INTRODUCTION

fmicb-09-02179 September 10, 2018 Time: 17:34 # 2

Non-tuberculous mycobacteria (NTM) encompass all species of mycobacteria that do not cause tuberculosis (TB) or leprosy (Lee et al., 2015). NTM are ubiquitous in the environment and occasionally infect humans with various predisposing conditions like cystic fibrosis, bronchiectasis or immunosuppression, causing a variety of pathological conditions including pulmonary, skin and soft tissue infections and disseminated diseases (Chan and Iseman, 2013). The last few decades saw an alarming increase in the incidence and prevalence of NTM-pulmonary disease throughout the globe. Importantly, more than 90% of all reported NTM-pulmonary disease cases involved infections with the Mycobacterium avium complex (comprising of M. avium, Mycobacterium chimaera and Mycobacterium intracellulare) or Mycobacterium abscessus, accentuating the medical importance of these emerging pathogens (Hoefsloot et al., 2013; Ryu et al., 2016; Wu et al., 2018). For a long while, it was believed that infections with genetically diverse NTM strains were exclusively acquired upon exposure to the environment. However, a recent whole genome analysis of more than 1000 M. abscessus clinical isolates from different geographical locations, revealed the presence of genetically clustered strains in patients. This suggests a human-to-human transmission of M. abscessus, presumably through indirect mechanisms like cough aerosols or surface contamination (Bryant et al., 2016).

Mycobacterium abscessus is an opportunistic pathogen, ubiquitous in the environment, that often causes infections in humans with compromised natural defenses such as patients with cystic fibrosis or other chronic lung diseases (Sanguinetti et al., 2001; Brown-Elliott and Wallace, 2002; Howard, 2006; Griffith et al., 2007). A current taxonomic classification suggests separation of M. abscessus into three distinct subspecies: M. abscessus subsp. abscessus, M. abscessus subsp. bolletii, and M. abscessus subsp. massiliense (Tortoli et al., 2016). Although a saprophyte in water and soil, following lung infection M. abscessus can swiftly grow and survive intra-cellularly within macrophages as well as in extra-cellular caseous lesions and airway mucus (Wu et al., 2018). Several factors contribute to the success of this rapidly growing mycobacterium. A plethora of intrinsic resistance mechanisms renders almost all clinically used antibiotics ineffective against M. abscessus (Nessar et al., 2012). In addition, the presence of a highly dynamic open pan-genome in M. abscessus might explain the ease with which the bacterium evolves and adapts to a wide-spectrum of stressful environmental conditions encountered in diverse habitats (Choo et al., 2014; Wu et al., 2018). Importantly, the respiratory habitat of M. abscessus brings it in close proximity to highly virulent pathogens (for example, Pseudomonas aeruginosa in cystic fibrosis lung) which can serve as donors of novel drug resistance or virulence genes (Ripoll et al., 2009).

Treatment of an M. abscessus pulmonary infection is very difficult owing to the bacterium's high-level innate resistance towards most antibiotics commonly used for Gram-negative and Gram-positive bacterial infections including majority of β-lactams, tetracyclines, aminoglycosides and macrolides. In addition, anti-TB drugs including first-line drugs (such as rifampicin and isoniazid) as well as some second-line agents (such as capreomycin) are ineffective against this pathogen (**Table 1**). Thus, the term "incurable nightmare" is often used to describe M. abscessus (Brown-Elliott et al., 2012; Nessar et al., 2012). So far, no reliable antibiotic regimen has been established for M. abscessus pulmonary disease. Antibiotic administration is largely empirical and relies on in vitro antibiotic susceptibility testing and definitive subspecies identification (Griffith et al., 2007; Ryu et al., 2016). Treatment of M. abscessus pulmonary disease as recommended by the British Thoracic Society involves administration of a multi-drug regimen encompassing intravenous antibiotics - amikacin, tigecycline, and imipenem along with an oral macrolide (clarithromycin or azithromycin) for clinical isolates susceptible to macrolides, during the initial treatment phase. For the continuation phase of treatment, nebulised amikacin and an oral macrolide combined with one to three of the following oral antibiotics: linezolid, clofazimine, minocycline, co-trimoxazole, and moxifloxacin are generally recommended (Haworth et al., 2017). Standard of care calls for continuation of antibiotic therapy for a minimum of 12 months after culture conversion. However, microbiological eradication of M. abscessus bacilli from lung tissues, using recommended antibiotic regimens, is rare and recurrence is often after completion of a treatment course and successful symptom management (Jarand et al., 2011; Stout et al., 2016).

As a result of extensive, repeated or inappropriate use of macrolides and aminoglycosides, which inhibit protein biosynthesis by binding to the large and small ribosomal subunits, respectively, M. abscessus strains with clinically acquired panmacrolide and pan-aminoglycoside resistance have emerged, due to mutation(s) in the corresponding 23S (rrl) and 16S (rrs) rRNA genes (Wallace et al., 1996; Prammananan et al., 1998; Maurer et al., 2012; Nessar et al., 2012). Acquired resistance to macrolides and aminoglycosides severely limits the remaining treatment options and highlights the urgent need for new antimicrobial agents.

Mechanisms underpinning intrinsic drug resistance of M. abscessus are multi-fold and fall into two main groups: first, the presence of a highly impermeable cell envelope


TABLE 1 | Drug susceptibility of Mycobacterium abscessus wild-type and isogenic mutants in comparison to screening concentrations for M. tuberculosis with decreased susceptibility.

Epidemiologic cutoff values (ECOFF; MGIT960) as defined by Cambau et al. (2015); MIC, minimal inhibitory concentration; N.A., Not available; N.D., Not determined.

and/or multi-drug efflux pumps might reduce the effective concentration of antibiotics within the bacterial cells; second, the genome of M. abscessus encodes several putative enzymes which can inactivate antibiotics by modification and/or degradation or lower the affinity of the drug for its target by modifying the target (Ripoll et al., 2009; Nessar et al., 2012). For long, molecular investigations aimed at elucidating antibiotic resistance mechanisms of M. abscessus were limited, however, significant progress has been made in recent years owing to the development of efficient tools for genetic manipulation of this bacterium (**Figure 1**). In this article, we provide an up-to-date overview of the main molecular mechanisms of antibiotic resistance in M. abscessus with a focus on the drugtarget-modifying or drug-modifying enzymes and discuss the potential impact on clinical treatment. We also discuss how this knowledge could facilitate the discovery and development of new improved antimicrobial agents or help rescue the function of currently available antibiotics against M. abscessus and provide an update on the latest research underway to combat multi-drug resistant M. abscessus infections.

#### ANTIBIOTIC-TARGET-MODIFYING ENZYMES IN M. abscessus

#### Macrolide Resistance: MAB\_2297 [erm(41)] Encoded Erythromycin Ribosome Methylase

Macrolides are among the most successful antibiotics in the world and are highly prescribed for infections caused by nontuberculous mycobacteria. A multidrug regimen recommended for M. abscessus infections almost always includes macrolides, particularly clarithromycin or azithromycin. Macrolides target the nascent polypeptide exit tunnel within the 50S ribosomal subunit, which accommodates the newly synthesized polypeptide chain as it emerges from the ribosome. Following macrolide binding, the elongation of polypeptide chain continues up to a few amino acids before the peptidyl-tRNA dissociates from the ribosome thereby arresting protein synthesis (Wilson, 2014).

In bacteria, innate resistance towards macrolides is most often the result of increased efflux, ribosomal methyltransferases and GTP-dependent macrolide kinases (Pawlowski et al., 2018). A well-studied example of macrolide resistance by target modification in bacteria is the erythromycin ribosome methylase (Erm) enzyme, which specifically methylates A2058 nucleotide of 23S rRNA and lowers the macrolide affinity for ribosome exit tunnel (Blair et al., 2015). An Erm methyltransferase is also responsible for inducible macrolide resistance in M. abscessus (**Figure 1**). Interestingly, the erm(41) gene, first described by Nash et al. (2009), is only functional in two out of three subspecies of the M. abscessus complex. The subspecies M. massiliense harbors deletions in its erm(41) gene copy which render the Erm(41) enzyme inactive and the bacterium susceptible to macrolides. This may explain why patients with M. massiliense infections have better treatment outcomes than patients with M. abscessus or M. bolletii infections. To further complicate matters, a T/C polymorphism at position 28 of the erm(41) sequence determines the appearance of inducible macrolide resistance in M. bolletii and M. abscessus subspecies. Only those isolates that harbor a T28 erm(41) sequevar develop inducible macrolide resistance. Phenotypic detection of inducible macrolide resistance in the M. abscessus complex by drug susceptibility testing requires extended incubation (14 days rather than 3 days) in the presence of the drug (Nash et al., 2009; Kim et al., 2010).

A functional role of erm(41) in inducible resistance to macrolides has been confirmed by genetic studies. Deletion of erm(41) in M. abscessus lowered the minimal inhibitory concentration (MIC) of clarithromycin by 128-fold (**Table 1**), while introduction of a functional M. abscessus erm(41) gene into M. massiliense led to a 64-fold increase in the clarithromycin MIC on day 14 (Choi et al., 2012). Inducible Erm(41)-dependent rRNA methylation severely affects the efficacy of macrolides against M. abscessus. The long duration of partially effective macrolide therapy, eventually favors the emergence of isolates with highlevel constitutive macrolide resistance. Mutations that alter the

23S rRNA nucleotides 2058 and 2059 within the peptidyltransferase center (PTC), are a frequent cause of constitutive macrolide resistance in M. abscessus (Wallace et al., 1996; Bastian et al., 2011; Brown-Elliott et al., 2015).

#### ANTIBIOTIC-MODIFYING ENZYMES IN M. abscessus

#### Aminoglycoside Resistance: MAB\_4395 [aac(2<sup>0</sup> )] & MAB\_4532c (eis2) Encoded Acetyltransferases; MAB\_2385 Encoded 3 00 -O-Phosphotransferase

Aminoglycosides are among the broadest classes of antibiotics that interfere with specific steps in bacterial protein synthesis (Kohanski et al., 2010). All aminoglycosides primarily target the 16S rRNA component of 30S ribosomal subunit, however, their exact binding site and mode of action may differ slightly depending on their chemical structure. For example, 2 deoxystreptamine (2-DOS) aminoglycosides (such as kanamycin, amikacin, apramycin, arbekacin), which are chemically distinct from the non-DOS aminoglycosides (for example, streptomycin), specifically bind to the A-site region of small ribosomal subunit at the decoding center, encompassing conserved nucleotides A1492 and A1493 of 16S rRNA. 2-DOS aminoglycosides function by inhibiting the translocation of tRNA-mRNA complex through the ribosome. Additionally, several members of this class also induce protein mistranslation by promoting and stabilizing the interaction of non-cognate aminoacylated tRNAs with mRNA. In contrast, the non-DOS aminoglycoside streptomycin targets a different region of 16S rRNA in proximity of the decoding center

and mainly interferes with the delivery of aminoacylated tRNA to the A-site (Wilson, 2014).

Bacterial resistance to aminoglycosides may occur due to low cell permeability/efflux, mutations or modification of 16S rRNA, mutations in ribosomal protein genes and enzymatic drug modification. Drug and target modification being the most extensively studied and clinically relevant aminoglycoside resistance mechanisms in pathogenic bacteria. Covalent modification of key amino or hydroxyl groups within the aminoglycoside molecule by bacterial enzymes markedly decreases the binding affinity for the ribosomal target and confers aminoglycoside resistance. Aminoglycosidemodifying enzymes in bacteria can be grouped into three main classes namely, acetyltransferases, nucleotidyltransferases, and phosphotransferases (D'Costa et al., 2011; Blair et al., 2015). A worrying discovery that the genome of emerging pathogen M. abscessus encodes several putative aminoglycoside-modifying enzymes which include members of all three classes, came in 2009 when Ripoll et al first sequenced the complete genome of M. abscessus (Ripoll et al., 2009). Some predictions were later substantiated by drug susceptibility testing and biochemical analysis (Maurer et al., 2015). More recently, direct genetic evidence for a functional role of some of these putative aminoglycoside-modifying enzymes in M. abscessus intrinsic resistance to this antibiotic class, was obtained.

Similar to all other mycobacteria that have been studied so far (Aínsa et al., 1997), M. abscessus also harbors a 2<sup>0</sup> -Nacetyltransferase [AAC(2<sup>0</sup> )], encoded by MAB\_4395 gene. This enzyme is capable of acetylating several aminoglycosides bearing a 2<sup>0</sup> amino group including gentamicin C, dibekacin, tobramycin and kanamycin B (**Figure 1**). This was evident upon deletion of MAB\_4395 in M. abscessus ATCC 19977 strain which led to a 4-64-fold reduction in the MICs of these aminoglycosides (**Table 1**) (Rominski et al., 2017c). Interestingly, another N-acetyltransferase capable of conferring aminoglycoside resistance in M. abscessus is a homolog of Anabaena variabilis enhanced intracellular survival (Eis) protein (**Figure 1**). Eis proteins are widespread in mycobacteria as well as other Gram-positive bacteria. Eis protein in Mycobacterium tuberculosis has been shown to enhance intracellular survival within macrophages by acetylating and activating dual-specificity protein phosphatase 16/mitogen-activated protein kinase phosphatase-7 (DUSP16/MPK-7), a JNK-specific phosphatase that inhibits inflammation, autophagy and subsequent death of infected macrophages (Shin et al., 2010; Kim et al., 2012). Furthermore, upregulation of eis gene expression owing to mutations in its promoter was associated with kanamycin resistance in one-third of clinical isolates encompassing a large collection of M. tuberculosis strains from diverse geographical locations (Zaunbrecher et al., 2009). More recently, structural and biochemical characterization of M. tuberculosis Eis revealed its unprecedented ability to acetylate multiple amines of several aminoglycosides (Chen et al., 2011).

Analysis of M. abscessus genome revealed two genes namely, MAB\_4124 (eis1) and MAB\_4532c (eis2), which show homology to eis from M. tuberculosis (Rv2416c) and A. variabilis (Ava\_4977), respectively. Drug susceptibility testing could confirm a functional role for Eis2 but not Eis1, in intrinsic aminoglycoside resistance. Of note, deletion of MAB\_4532c enhanced M. abscessus susceptibility towards a heterogenous group of aminoglycosides including the cornerstone drug amikacin as well as a non-aminoglycoside antibiotic capreomycin (**Table 1**) (Rominski et al., 2017c). Furthermore, heterologous expression of eis2 in Mycobacterium smegmatis decreased amikacin susceptibility (Hurst-Hess et al., 2017). Capreomycin is a cyclic peptide antibiotic that inhibits protein synthesis by blocking peptidyl-tRNA translocation. The drug is particularly active against M. tuberculosis and commonly used as a secondline agent for the treatment of TB (Stanley et al., 2010; Wilson, 2014). Although remarkable, the discovery that M. abscessus Eis2 exhibits a broad-spectrum acetyltransferase activity is not a real surprise. In fact, recent structural and functional studies on Eis enzymes from M. tuberculosis (Eis\_Mtb), M. smegmatis (Eis\_Msm), and A. variabilis (Eis\_Ava) revealed the presence of an unusually large and complex active site unlike in other aminoglycoside acetyltransferases. The unique structure equips Eis with a regio-versatile multi-acetylating acetyltransferase activity towards a broad range of substrates (Chen et al., 2012; Pricer et al., 2012; Jennings et al., 2013). However, there seems to be a notable difference between the acetylation activities of M. abscessus Eis2 and M. tuberculosis Eis especially towards capreomycin as the MIC of capreomycin in M. abscessus is about 100-fold higher than that in M. tuberculosis and this difference is almost completely abolished by the deletion of eis2 gene (**Table 1**) (Rominski et al., 2017c).

Aminoglycoside acetyltransferases Eis2 and AAC(2<sup>0</sup> ) do not appear to modify streptomycin, an aminoglycoside which exhibits moderate in vitro activity against M. abscessus (MIC: 32 mg/L) (Rominski et al., 2017c). Instead, a MAB\_2385 encoded putative aminoglycoside 3<sup>00</sup> -O-phosphotransferase is the main determinant for intrinsic streptomycin resistance in M. abscessus (**Figure 1**). Twelve putative aminoglycoside phosphotransferases are encoded in the genome of M. abscessus (Ripoll et al., 2009). Among them, MAB\_2385 was found to be a close homolog of aph(300)-Ic, an aminoglycoside 3<sup>00</sup> -Ophosphotransferase gene which was previously shown to confer streptomycin resistance in Mycobacterium fortuitum (Ramón-García et al., 2006). Deletion of MAB\_2385 enhanced the susceptibility of M. abscessus towards streptomycin by 16 fold (**Table 1**). Furthermore, MAB\_2385 was able to confer streptomycin resistance when heterologously expressed in M. smegmatis (Dal Molin et al., 2017). To the best of our knowledge other annotated aminoglycoside phosphotransferases (such as MAB\_0163c, MAB\_0313c, MAB\_0327, MAB\_0951, and MAB\_1020) and aminoglycoside acetyltransferases (for example, MAB\_0247c, MAB\_0404c, MAB\_0745, MAB\_4235c, and MAB\_4324c) have not been addressed experimentally, however, interesting regulatory features of drug resistance genes have been elucidated.

A WhiB7 like protein encoded by MAB\_3508c was recently uncovered as a multi-drug inducible transcriptional regulator that modulates the expression of genes conferring aminoglycoside and macrolide resistance in M. abscessus. WhiB7, an autoregulatory transcriptional activator, is a determinant of

innate antibiotic resistance in mycobacteria. In M. tuberculosis, genes conferring antibiotic resistance including ermMT (Rv1988), eis (Rv2416c) and tap (Rv1258c) are induced in a whiB7-dependent manner, upon exposure to antibiotics. Similarly, the erm(41) and eis2 genes in M. abscessus are included in the whiB7 regulon (Hurst-Hess et al., 2017). A recent study by Pryjma et al. (2017) demonstrated that exposure to subinhibitory levels of clarithromycin, enhanced amikacin as well as clarithromycin resistance in M. abscessus in a whiB7-dependent manner. Of note, deletion of MAB\_3508c rendered M. abscessus more susceptible to amikacin (fourfold) and clarithromycin (eightfold), while complementation with an intact gene copy restored resistance to these antibiotics, suggesting a role of whiB7 in M. abscessus amikacin and clarithromycin resistance. Interestingly, pre-exposure to clarithromycin, which is a potent inducer of M. abscessus whiB7, enhanced the resistance of M. abscessus wild-type strain towards amikacin by fourfold but had no effect on amikacin susceptibility of a 1whiB7 mutant. In addition, pre-treatment with clarithromycin markedly increased resistance to itself in the wild-type strain but did not alter clarithromycin sensitivity of a 1whiB7 mutant. Furthermore, a quantitative reverse transcription-PCR assay revealed a whiB7 dependent upregulation of eis2 and erm(41) genes following pre-exposure to clarithromycin. These observations support the following conclusions: (i) in M. abscessus, whiB7 (MAB\_3508c) activates the expression of genes which confer amikacin and clarithromycin resistance, i.e., eis2 and erm(41), (ii) strong induction of whiB7 following exposure to clarithromycin confers cross-resistance to amikacin in addition to activating inducible macrolide resistance (Pryjma et al., 2017). Thus, in vitro findings by Pryjma et al. (2017) suggest that front-line antibiotics, amikacin and clarithromycin may exhibit antagonistic effects when used in combination for the treatment of M. abscessus infections. This may at least partially explain the limited clinical efficacy of currently recommended multi-drug regimen for M. abscessus which almost always includes these two antibiotics.

### Tetracycline Resistance: MAB\_1496c Encoded Flavin Monooxygenase

Tetracyclines are a class of broad-spectrum natural product antibiotics that interfere with bacterial protein synthesis. These drugs bind with high affinity to the small ribosomal subunit and specifically interfere with the delivery of aminoacylated tRNA to the A-site (Kohanski et al., 2010; Wilson, 2014). While tetracyclines are an important class of ribosome-targeting antibiotics, their anthropogenic and prolific use in the clinic and agriculture has led to emergence of resistance among benign as well as pathogenic bacteria, thereby limiting their efficacy. Despite growing resistance, tetracyclines continue to remain one of the most successful and widely used chemotherapeutics against bacterial infections (Allen et al., 2010). Furthermore, a renaissance for tetracyclines is being fuelled by the development of next-generation derivatives such as tigecycline, approved for clinical use in 2005 and omadacycline and eravacycline, undergoing late-phase clinical trials (Kinch and Patridge, 2014). In fact, tigecycline is the only new antibiotic and the sole member of its class, which has been introduced in the treatment of recalcitrant M. abscessus lung disease with clinical evidence (Wallace et al., 2014).

Previously, resistance towards tetracyclines was thought to be exclusively mediated by two mechanisms namely, drug efflux and ribosome protection. However, evidence for enzymatic inactivation of tetracyclines by a family of flavin adenine dinucleotide (FAD)-dependent monooxygenases in both benign and pathogenic bacteria has been recently documented. Few well characterized examples of such tetracycline-modifying enzymes include TetX from the obligate anaerobe Bacteroides fragilis and Tet(56) found in human pathogen Legionella longbeachae, the causative agent of Legionnaires' disease and Pontiac fever (Yang et al., 2004; Park et al., 2017). Whether such a mechanism of tetracycline resistance existed in mycobacteria was not known until recently, when Rudra and colleagues demonstrated through elegant experiments, that high levels of intrinsic resistance towards tetracycline and second generation derivative doxycycline in M. abscessus was mediated by a flavin monooxygenase, MabTetX (**Figure 1**). Deletion of MAB\_1496c (encoding a putative FAD binding monooxygenase) enhanced the susceptibility of M. abscessus towards tetracycline and doxycycline by 15–20-fold (**Table 1**), while a complemented 1MAB\_1496c strain exhibited even higher levels of resistance towards both antibiotics compared to the wild-type strain, supporting the notion that MAB\_1496c is a primary determinant for M. abscessus intrinsic tetracycline resistance. Furthermore, UV-Visible and mass spectrometry assays with purified MAB\_1496c protein provided evidence for hydroxylation and subsequent degradation of tetracycline and doxycycline by MabTetX (Rudra et al., 2018). Interestingly, MAB\_1496c is not a member of the whiB7 regulon, even though tetracycline is a strong inducer of whiB7 in M. abscessus (Hurst-Hess et al., 2017). In fact, MabTetX expression is regulated by a TetR family repressor, MabTetR<sup>x</sup> (encoded by the upstream gene MAB\_1497c) which represses the MAB\_1497c-MAB\_1496c operon by binding to a 35bp operator sequence tetO. This repression is relieved in the presence of tetracycline and doxycycline which can bind the repressor protein thereby preventing its interaction with the tetO operator (Rudra et al., 2018).

The glycylcycline antibiotic tigecycline, specifically designed to evade resistance by ribosome protection and efflux, is not invulnerable to modification by Bacteroides TetX, which was recently also identified in numerous clinically relevant pathogens (Volkers et al., 2011; Leski et al., 2013). Although tigecycline was found to be a poor substrate of TetX, alarmingly enzymatic activity of TetX was substantially improved upon acquisition of single amino acid substitutions to confer resistance at clinically relevant tigecycline concentrations when overexpressed in Escherichia coli (Linkevicius et al., 2016). Surprisingly, tigecycline resisted MabTetX-dependent degradation and did not induce its expression, which might contribute to its excellent in vitro activity and moderate clinical efficacy against M. abscessus (Rudra et al., 2018). However, selective pressure exerted by tigecycline will eventually favor

the emergence of M. abscessus isolates carrying extendedspectrum FAD-monooxygenases with the ability to modify glycylcyclines.

Another interesting feature is the ability of anhydrotetracycline (ATc), a degradation product of tetracycline with poor antibiotic activity which was recently shown to be a competitive inhibitor of TetX and Tet(56) flavin monooxygenases, to rescue antibiotic activities of tetracycline and doxycycline against M. abscessus expressing MabTetX (Park et al., 2017; Rudra et al., 2018). However, ATc was also found to induce MabTetX expression by binding the MabTetR<sup>x</sup> repressor, a property at odds with its MabTetX inhibition activity. While this finding together with the severe side effects associated with ATc use, makes it an unattractive drug candidate, it nevertheless represents a flexible starting point for developing MabTetX inhibitors with improved activity, better tolerability and minimal binding affinity towards MabTetRx. Co-administration with a potent MabTetX monooxygenase inhibitor can potentially rescue the clinical efficacies of tetracycline and doxycycline which are currently inactive against the M. abscessus complex (Burgos et al., 2011; Rudra et al., 2018).

#### β-Lactam Resistance: MAB\_2875 Encoded β-Lactamase

β-lactams constitute an important class of antibiotics that function by inhibiting cell wall synthesis in bacteria. The β-lactam structure closely resembles the terminal D-alanyl-Dalanine dipeptide of peptidoglycan, a substrate of penicillin binding proteins (also known as D, D-transpeptidases) that catalyze the final cross-linking step of peptidoglycan synthesis. Penicilloylation of the D, D-transpeptidase active site following treatment with β-lactams, disables the enzyme and prevents the formation of peptidoglycan cross-links, thereby interfering with cell wall biosynthesis (Kohanski et al., 2010). Despite being the cornerstones of antimicrobial chemotherapy and constituting one of the largest groups of antibiotics available today, only two members of the β-lactam class namely, cefoxitin (a cephalosporin) and imipenem (a carbapenem) form a part of the antibiotic arsenal for M. abscessus. Furthermore, cefoxitin and imipenem have been shown to exhibit only moderate in vitro activity against M. abscessus with MICs of 32 and 4 mg/L, respectively (Lavollay et al., 2014). Majority of the other β-lactam antibiotics are not effective against this bacterium.

Several possible mechanisms may contribute to β-lactam resistance in M. abscessus. Low mycomembrane permeability and β-lactamase production may reduce the effective concentration of β-lactams at the site of action. Alternatively, modification of transpeptidase profile of the cell by replacement of penicillin binding proteins (D, D-transpeptidases) with L, D-transpeptidases for instance, may reduce the activity of penicillins and cephalosporins that have lower affinity for L, D-transpeptidase (Wivagg et al., 2014). However, most of the abovementioned resistance mechanisms are relatively unexplored in this organism except for one. Recently it was shown that the genome of M. abscessus encodes a strong, constitutive class A β-lactamase (Bla\_Mab) which renders it highly resistant towards most β-lactams (**Figure 1**). The M. abscessus β-lactamase Bla\_Mab, encoded by MAB\_2875 is endowed with an exceptional broadspectrum activity and can effectively hydrolyse several members of first- and second-generation cephalosporins, carbapenems, and penams (Soroka et al., 2014). The currently recommended β-lactams namely imipenem and cefoxitin are substrates of Bla\_Mab, however, they are hydrolysed at a very slow rate, which may contribute to their clinical efficacy. The instability of imipenem complicates drug susceptibility testing when generation times of test bacteria require incubations for several days as in the case of M. abscessus (Rominski et al., 2017b).

One strategy to overcome the hurdle of chromosomally encoded β-lactamase in M. abscessus is to co-administer a β-lactamase inhibitor with the failing β-lactam. Soroka and colleagues evaluated the in vitro efficacy of nitrocefin and meropenem in combination with approved β-lactamase inhibitors clavulanate, tazobactam, and sulbactam against Bla\_Mab. Surprisingly, Bla\_Mab inhibition was not detected for all the three inhibitor drugs (Soroka et al., 2014). This remarkable ability of Bla\_Mab to resist inhibition even surpasses that of BlaC, a class A β-lactamase encoded by its relative M. tuberculosis. Although BlaC is able to reverse inhibition caused by tazobactam and sulbactam, and return to its native functional form, clavulanate can slowly but irreversibly inhibit this enzyme (Sagar et al., 2017). In fact, a recent demonstration of in vitro synergy between meropenem and clavulanate in 13 extensively drug resistant (XDR) M. tuberculosis strains raised a renewed interest in β-lactams for use in TB therapy (Hugonnet et al., 2009). Fortunately, a surprising finding that M. abscessus Bla\_Mab is effectively disabled by the newly approved β-lactamase inhibitor avibactam provides a new hope for old antibiotics. A study by Dubée et al. (2015) showed that co-administering a small amount of avibactam (4 mg/L) with each of several representative antibiotics from three important β-lactam sub-classes (carbapenems, penams, and cephalosporins) lowered their MICs in M. abscessus CIP104536 strain to levels comparable to those observed in a β-lactamase deficient mutant. The authors also reported synergy between avibactam and amoxicillin within macrophages as well as in a zebrafish model of M. abscessus infection (Dubée et al., 2015). More recently, an in vitro synergy screen of 110 β-lactam - β-lactamase inhibitor combinations against an M. abscessus clinical isolate identified six potential hits. Five selected β-lactamase inhibitors encompassing both β-lactam-based (clavulanate, tazobactam, and sulbactam) as well as non-β-lactam-based inhibitors (avibactam and vaborbactam) were evaluated for synergistic effects with a large panel of β-lactams. Importantly, all observed synergy combinations exclusively involved non-β-lactam-based inhibitors, either avibactam or vaborbactam. Combinations of avibactam with each of the following drugs including ampicillin, amoxicillin, tebipenem, and panipenem displayed synergistic effects against M. abscessus. Furthermore, the carbapenems, tebipenem, and panipenem also exhibited synergy with a novel boronic acid β-lactamase inhibitor, vaborbactam (Aziz et al., 2018). Thus, non-β-lactam-based β-lactamase inhibitors such as avibactam and vaborbactam can potentially extend the spectrum of β-lactams employed for the treatment of largely incurable, chronic M. abscessus pulmonary infections.

### Rifamycin Resistance: MAB\_0591 Encoded ADP-Ribosyltransferase

fmicb-09-02179 September 10, 2018 Time: 17:34 # 8

Rifampicin, a rifamycin antibiotic commonly employed as a firstline agent in the treatment of M. tuberculosis infections, hardly exhibits any activity towards the M. abscessus complex (MIC: 128 mg/L) (**Table 1**). Since rifamycins inhibit transcription by binding with high affinity to rpoB encoded β-subunit of bacterial RNA polymerase, resistance towards this group of antibiotics is most commonly conferred by point mutations within the target gene rpoB, especially in M. tuberculosis (Campbell et al., 2001). However, additional rifamycin resistance mechanisms that involve enzymatic drug inactivation are also prevalent in other mycobacterial species. The presence of a rifamycin ADPribosyltransferase in M. smegmatis was known for quite a while (Quan et al., 1997), however, only recently genetic studies involving gene knockout and heterologous expression unveiled the existence of a MAB\_0591 encoded ADP-ribosyltransferase as the major determinant for high levels of innate rifamycin resistance in M. abscessus (**Figure 1**). Deletion of MAB\_0591 enhanced the susceptibility of M. abscessus towards three tested rifamycins including rifaximin, rifapentine and rifampicin by 64–512-fold, respectively (Rominski et al., 2017a).

Surprisingly, a rifampicin derivative, rifabutin showed up as an attractive hit (MIC: 2.5 mg/L), when a set of 2,700 FDA-approved drugs were screened against a clinical isolate of M. abscessus. Rifabutin exhibited approximately 10-fold higher potency in vitro relative to rifampicin and rifapentine, against reference strains representative of all three subspecies of the M. abscessus complex as well as a collection of clinical isolates and its bactericidal activity was comparable to or better than that of clarithromycin, one of the cornerstone drugs for M. abscessus infections (Aziz et al., 2017). The higher efficacy of rifabutin relative to rifampicin in M. abscessus as well as in M. tuberculosis (with no known rifamycin inactivating enzymes) suggests its increased accumulation within the bacterial cells. This may be due to differences in membrane penetration or specificity of efflux systems towards different rifamycins (**Figure 1**). Based on the findings reported by Aziz et al. (2017), rifampicin analog rifabutin, appears to be a potential candidate for repurposing and its clinical efficacy in patients with M. abscessus pulmonary disease should be further explored.

### CLINICAL RELEVANCE

Although the inherent ability of M. abscessus to exhibit resistance to a wide array of antibiotics has long been recognized, our knowledge of the prodigious diversity of mechanisms involved has improved immensely in recent years, owing to the development of tools for genetic manipulation of M. abscessus. A better understanding of the genetic basis of innate antibiotic resistance is a prerequisite for the discovery and development of synergistic drug combinations, where one agent serves to salvage the antibacterial activity of a failing antibiotic by inhibiting an intrinsic resistance mechanism. This is exemplified by the recent discovery that avibactam is a potent inhibitor of M. abscessus β-lactamase Bla\_Mab. When used in combination, avibactam lowered the MICs of several β-lactams in M. abscessus by 4–32-fold (Dubée et al., 2015; Kaushik et al., 2017). Despite the success of β-lactamase inhibitors, the concept of designing drugs which target intrinsic resistance mechanisms in bacteria, has not gained substantial clinical exploitation beyond this antibiotic class (Brown, 2015). Recently, a repertoire of drug-modifying and target-modifying enzymes conferring innate resistance to aminoglycosides, tetracyclines, rifamycins, and macrolides in M. abscessus were delineated (Nash et al., 2009; Dal Molin et al., 2017; Rominski et al., 2017a,c; Rudra et al., 2018). Likewise, attempts to identify chemical entities which could potentially rescue the function of one or more key members from each aforementioned antibiotic group, should be pursued in the near future.

Since the discovery that mutations within eis promoter are associated with kanamycin resistance in M. tuberculosis, vigorous attempts to identify Eis inhibitors that can potentially rescue kanamycin activity in this pathogen have been pursued with notable success. Recently, several structurally diverse competitive Eis inhibitors with 1,2,4-triazino[5,6b]indole-3-thioether-based, sulfonamide-based, pyrrolo[1,5-a]pyrazine-based and isothiazole S,S-dioxide heterocyclic scaffolds were identified by highthroughput screenings of large compound libraries. Importantly, few promising inhibitors could completely restore kanamycin susceptibility in a kanamycin resistant M. tuberculosis strain in vitro at concentrations that were non-cytotoxic to mammalian cells (Green et al., 2012; Garzan et al., 2016, 2017; Willby et al., 2016; Ngo et al., 2018). Remarkably, inhibitors designed for M. tuberculosis Eis, were also found to be active against Eis homologs in M. smegmatis and A. variabilis (Chen et al., 2012; Pricer et al., 2012). These findings give rise to an intriguing possibility that Eis\_Mtb inhibitors may be able to overcome Eis2 mediated resistance towards aminoglycoside and/or nonaminoglycoside antibiotics like capreomycin and restore their effectiveness against M. abscessus. Hence efforts to explore the activity of Eis\_Mtb inhibitors against M. abscessus Eis2 enzyme should be undertaken. Alternatively, extending drug discovery efforts towards targeting WhiB7, could be a possible strategy to tackle intrinsic macrolide and aminoglycoside resistance in M. abscessus. WhiB7 is a conserved transcriptional regulator in mycobacteria that co-ordinates intrinsic resistance to a wide range of ribosome-targeting drugs (Burian et al., 2012). Recent evidence suggests that genes which contribute to intrinsic aminoglycoside and macrolide resistance in M. abscessus, i.e., eis2 and erm(41), are induced in a whiB7-dependent manner upon exposure to ribosomal antibiotics (Hurst-Hess et al., 2017). Thus, with the aim to kill two birds with one stone, compounds that prevent the induction of eis2 and erm(41) resistance genes in M. abscessus by modulating the synthesis and/or activity of WhiB7, should be sought.

Since the discovery and development of compounds that can circumvent existing resistance mechanisms in M. abscessus is a daunting and time-consuming task, an immediate and relatively simple solution to improve treatment efficacy may

involve repurposing and repositioning of currently available antibiotics. A number of candidate drugs suited for this purpose have been reviewed in detail elsewhere (Wu et al., 2018) and are therefore only briefly mentioned here. For example, a systematic screening of a collection of FDA-approved drugs revealed rifabutin, an analog of rifampicin, exhibiting potent in vitro activity against M. abscessus (MIC: 2.5 mg/L) (Aziz et al., 2017). Similarly, clofazimine, a leprosy drug currently repurposed for treatment of TB, was also found to be active against M. abscessus. In addition to promising in vitro activity (MIC ≤ 1 mg/L), clofazimine displayed adequate clinical efficacy and safety when evaluated in patients with M. abscessus pulmonary infections (Yang et al., 2017). Likewise, a novel oxazolidinone LCB01-0371, presently in Phase II clinical development for TB, exhibited high potency against M. abscessus compared to linezolid, an FDA-approved drug of the same class (Kim et al., 2017). Moreover, some aminoglycosides such as kanamycin A, apramycin, isepamicin, and arbekacin were also found to exhibit potent in vitro activity against M. abscessus (MICs ≤ 1 mg/L) (Rominski et al., 2017c). Besides these, other compounds that have been extensively studied for their inhibitory effects against M. abscessus encompass anti-TB drug bedaquiline, some TB actives like indole-2-carboxamides and piperidinolbased compound 1 (PIPD1), as well as second generation thiacetazone derivatives like D6, D15, and D17 (Dupont et al., 2016, 2017; Franz et al., 2017; Halloum et al., 2017; Kozikowski et al., 2017).

In addition, several recent studies have reported in vitro synergies between novel combinations of existing antibiotic classes. For example, vancomycin, an inhibitor of cell wall synthesis, exhibited synergy with clarithromycin, an inhibitor of protein synthesis, in M. abscessus (Mukherjee et al., 2017). Likewise, a combination of tigecycline (which inhibits protein synthesis) with teicoplanin (which disrupts peptidoglycan synthesis) displayed a strong synergistic effect against M. abscessus (Aziz et al., 2018). Since the mycobacterial cell envelope forms a major permeability barrier, drugs that perturb cell wall integrity, are likely to synergize with drugs having intracellular targets (Le Run et al., 2018). Such combinations based on mechanistic drug action should be further explored for synergistic effects against M. abscessus. For example, PIPD1 and indole-2-carboxamides, which disrupt mycolic acid synthesis by targeting MmpL3 transporter in mycobacteria (Dupont et al., 2016; Franz et al., 2017), may be able to potentiate the effects of other drugs modulating intracellular targets.

Taken together, a huge reservoir of intrinsic resistance genes combined with an unusually high evolutionary potential of its genome to acquire resistance renders almost all available antibiotics ineffective against M. abscessus. Therefore, in the long run, a better understanding of natural resistance mechanisms in M. abscessus will help jump start drug discovery projects for (i) compounds that can rescue current antibiotics by neutralizing intrinsic resistance (termed antibiotic resistance breakers), (ii) new drugs with one or more improved properties like ability to resist enzymatic modification and/or degradation, high affinity for modified or mutated targets, better solubility/uptake and low propensity for efflux (Brown, 2015). In the short run, available knowledge from studies seeking synergy between antibiotics together with those investigating potential candidates for repurposing, will aid in the development of dosing regimens that display adequate clinical efficacy and resistance to which will only emerge slowly.

### OUTLOOK AND PERSPECTIVES

Molecular genetic tools that allow investigators to systematically probe gene function in mycobacteria by generation of unmarked gene deletions and functional complementation, have proven extremely useful in dissecting the intrinsic resistome of M. abscessus while genetic investigations in other NTM species are still lagging behind. Using the available genetic 'toolkit,' 7 orthologs of known resistance gene families have been discovered so far in M. abscessus that provide resistance or 'immunity' against antibiotics targeting major cellular pathways including cell wall synthesis (β-lactams), RNA synthesis (rifamycins) and protein synthesis (macrolides, aminoglycosides, and tetracyclines). The impressive resistance diversity shown by M. abscessus is expected given its natural habitat is shared by many antibiotic producers, primarily the soil dwelling actinomycetes which are also reservoirs of extensively diverse resistance elements (Pawlowski et al., 2016; Crofts et al., 2017; Koteva et al., 2018). Exposure to a variety of noxious antimicrobial molecules in the soil might have favored selection of specialized and diverse resistance mechanisms in M. abscessus as well as in other NTM species.

Till now, the approach that researchers used to systematically probe the intrinsic resistome of M. abscessus involved analysis of few selected genes identified based on homology to known resistance determinants from other bacteria. As a result, this approach mainly revealed genes within M. abscessus that were involved in previously known resistance mechanisms and disguised the presence of potential determinants of entirely new resistance mechanisms that lack characterized homologs in other bacteria. To capture the remarkable resistance diversity of M. abscessus as well as other environmental NTM with unprecedented depth, future studies in this field should incorporate next-generation approaches that provide a comprehensive genome-wide definition of loci required for antibiotic resistance, using cutting-edge technologies. One such technique which can efficiently mine resistance determinants with novel sequences as well as assign resistance functions to known genes that were not previously shown to be involved in drug resistance, is the most recent incarnation of insertional mutagenesis, called transposon insertion sequencing (TIS) (Chao et al., 2016).

The utility of TIS in elucidating novel antibiotic resistance functions is highlighted by its recent application to the discovery of a repertoire of previously unknown intrinsic factors impacting resistance to antibiotics of different classes including oxazolidinones, fluoroquinolones, and aminoglycosides in pathogens of high clinical concern such as Staphylococcus aureus, P. aeruginosa and E. coli (Gallagher et al., 2011; Shan et al., 2015; Rajagopal et al., 2016). Screening large collections

of transposon insertion mutants for antibiotic susceptibility using TIS approach, has identified promising novel targets for drugs that can restore or enhance susceptibility to existing antibiotics. For example, analysis of insertion mutants spanning all non-essential S. aureus genes for fitness defects resulting from exposure to antibiotics, identified pathways/genes including graRS and vraFG (graRS/vraFG), fmtA, mprF, SAOUHSC\_01025, and SAOUHSC\_01050 which if inhibited, can greatly improve the efficacy of existing antibiotics including daptomycin, vancomycin, gentamicin, ciprofloxacin, oxacillin, and linezolid and extend their utility for treating S. aureus infections (Rajagopal et al., 2016).

In the coming years, high-throughput next-generation strategies like TIS will undoubtedly rise to prominence in the NTM drug resistance field and radically advance our understanding of resistance mechanisms in M. abscessus and other emerging NTM pathogens. Such breakthrough technologies will greatly impact the kinds of questions which can be addressed in this field, for example, which resistance elements unique to M. abscessus render it unusually more drug resistant than other NTM species? or does M. abscessus already

#### REFERENCES


carry genes which confer resistance to novel NTM drugs in development?

#### AUTHOR CONTRIBUTIONS

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

#### FUNDING

Research in the group of PS was supported by the Swiss National Science Foundation (#31003A\_153349), Lungenliga Schweiz, the Institute of Medical Microbiology and the University of Zurich.

#### ACKNOWLEDGMENTS

We acknowledge the members of the research group for stimulating discussions and support.

Clin. Microbiol. Rev. 15, 716–746. doi: 10.1128/CMR.15.4.716-746. 2002



glycopeptidolipid and reacquisition of an invasive phenotype. Microbiology 152, 1581–1590. doi: 10.1099/mic.0.28625-0



pathogen Mycobacterium abscessus. PLoS One 4:e5660. doi: 10.1371/journal. pone.0005660



Zaunbrecher, M. A., Sikes, R. D., Metchock, B., Shinnick, T. M., and Posey, J. E. (2009). Overexpression of the chromosomally encoded aminoglycoside acetyltransferase Eis confers kanamycin resistance in Mycobacterium tuberculosis. Proc. Natl. Acad. Sci. U.S.A. 106, 20004–20009. doi: 10.1073/pnas.0907925106

**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.

The reviewer UG and handling Editor declared their shared affiliation at time of review.

Copyright © 2018 Luthra, Rominski and Sander. 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.

# Mechanistic and Structural Insights Into the Unique TetR-Dependent Regulation of a Drug Efflux Pump in Mycobacterium abscessus

Matthias Richard<sup>1</sup> , Ana Victoria Gutiérrez1,2, Albertus J. Viljoen<sup>1</sup> , Eric Ghigo<sup>3</sup> , Mickael Blaise<sup>1</sup> \* and Laurent Kremer1,4 \*

<sup>1</sup> CNRS UMR 9004, Institut de Recherche en Infectiologie de Montpellier, Université de Montpellier, Montpellier, France, <sup>2</sup> Unité de Recherche, Microbes, Evolution, Phylogeny and Infection, Institut Hospitalier Universitaire Méditerranée Infection, Marseille, France, <sup>3</sup> Centre National de la Recherche Scientifique, Campus Joseph Aiguier, Marseille, France, <sup>4</sup> Institut National de la Santé et de la Recherche Médicale, Institut de Recherche en Infectiologie de Montpellier, Montpellier, France

#### Edited by:

Thomas Dick, Rutgers University, United States

#### Reviewed by:

Peter Sander, Universität Zürich, Switzerland Kyle Rohde, University of Central Florida, United States

#### \*Correspondence:

Mickael Blaise mickael.blaise@irim.cnrs.fr Laurent Kremer laurent.kremer@irim.cnrs.fr

#### Specialty section:

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

Received: 13 February 2018 Accepted: 20 March 2018 Published: 05 April 2018

#### Citation:

Richard M, Gutiérrez AV, Viljoen AJ, Ghigo E, Blaise M and Kremer L (2018) Mechanistic and Structural Insights Into the Unique TetR-Dependent Regulation of a Drug Efflux Pump in Mycobacterium abscessus. Front. Microbiol. 9:649. doi: 10.3389/fmicb.2018.00649 Mycobacterium abscessus is an emerging human pathogen causing severe pulmonary infections and is refractory to standard antibiotherapy, yet few drug resistance mechanisms have been reported in this organism. Recently, mutations in MAB\_4384 leading to up-regulation of the MmpS5/MmpL5 efflux pump were linked to increased resistance to thiacetazone derivatives. Herein, the DNA-binding activity of MAB\_4384 was investigated by electrophoretic mobility shift assays using the palindromic sequence IRS5/L5 located upstream of mmpS5/mmpL5. Introduction of point mutations within IRS5/L5 identified the sequence requirements for optimal binding of the regulator. Moreover, formation of the protein/IRS5/L5 complex was severely impaired for MAB\_4384 harboring D14N or F57L substitutions. IRS5/L5/lacZ reporter fusions in M. abscessus demonstrated increased β-galactosidase activity either in strains lacking a functional MAB\_4384 or in cultures treated with the TAC analogs. In addition, X-ray crystallography confirmed a typical TetR homodimeric structure of MAB\_4384 and unraveled a putative ligand binding site in which the analogs could be docked. Overall, these results support drug recognition of the MAB\_4384 TetR regulator, alleviating its binding to IRS5/L5 and steering up-regulation of MmpS5/MmpL5. This study provides new mechanistic and structural details of TetR-dependent regulatory mechanisms of efflux pumps and drug resistance in mycobacteria.

Keywords: Mycobacterium abscessus, TetR regulator, MmpL, efflux pump, structure, thiacetazone analogs, EMSA

### INTRODUCTION

Mycobacterium abscessus is a rapid growing mycobacterium (RGM) that has recently become an important health problem (Mougari et al., 2016). This non-tuberculous mycobacterial (NTM) pathogen can cause serious cutaneous, disseminated or pulmonary infections, particularly in cystic fibrosis (CF) patients. In CF patients, infection with M. abscessus is correlated with a decline in lung function and poses important challenges during last-resort lung transplantation (Esther et al., 2010; Smibert et al., 2016). An epidemiological study has recently documented the prevalence

and transmission of M. abscessus between hospital settings throughout the world, presumably via fomites and aerosols and uncovered the emergence of dominant circulating clones that have spread globally (Bryant et al., 2016). In addition, M. abscessus exhibits innate resistance to many different classes of antimicrobial agents, rendering infections with this microorganism extremely difficult to treat (Nessar et al., 2012; van Dorn, 2017). Recent studies have started to unveil the basis of the multi-drug resistance characterizing M. abscessus, uncovering a wide diversity of mechanisms or regulatory networks. These involve, for example, the induction of the erm(41) encoded 23S rRNA methyltransferase and mutations in the 23S rRNA that lead to clarithromycin resistance (Nash et al., 2009), the presence of a broad spectrum β-lactamase that limits the use of imipenem (Dubée et al., 2015; Lefebvre et al., 2016, 2017) or the presence of eis2, encoding an acetyltransferase that modifies aminoglycosides, specifically induced by whiB7, which contributes to the intrinsic resistance to amikacin (Hurst-Hess et al., 2017; Rominski et al., 2017b). Other studies reported the role of the ADP-ribosyltransferase MAB\_0591 as a major contributor to rifamycin resistance (Rominski et al., 2017a) whereas MAB\_2385 was identified as an important determinant in innate resistance to streptomycin (Dal Molin et al., 2018).

Recently, we reported the activity of a library of thiacetazone (TAC) derivatives against M. abscessus and identified several compounds exhibiting potent activity against a vast panel of clinical strains isolated from CF and non-CF patients (Halloum et al., 2017). High resistance levels to these compounds were linked to mutations in a putative transcriptional repressor MAB\_4384, together with a strong up-regulation of the divergently oriented adjacent locus encoding a putative MmpS5/MmpL5 transporter system. That ectopic overexpression of MmpS5/MmpL5 in M. abscessus also increased the minimal inhibitory concentration (MIC) to analogs of TAC further suggested that these two proteins may act as an active efflux pump which was sufficient to confer drug resistance (Halloum et al., 2017). In addition to uncovering new leads for future drug developments, this study also highlighted a novel mechanism of drug resistance in M. abscessus. Unexpectedly, an important difference relies on the fact that, in M. tuberculosis, MmpS5/MmpL5 acts as a multi-substrate efflux pump causing low resistance levels to antitubercular compounds such as clofazimine, bedaquiline, and azoles (Milano et al., 2009; Andries et al., 2014; Hartkoorn et al., 2014) whereas the M. abscessus strains overexpressing MmpS5/MmpL5 are very resistant to the TAC analogs but fail to show cross-resistance against clofazimine or bedaquiline (Halloum et al., 2017). This implies that, despite their high primary sequence identity, the MmpS5/MmpL5 orthologs from M. tuberculosis and M. abscessus do not share the same substrate specificity. Moreover, whereas Rv0678, the cognate regulator of MmpS5/MmpL5 in M. tuberculosis belongs to the MarR family (Radhakrishnan et al., 2014), MAB\_4384 is part of the TetR family of regulators and the change in the transcriptional level of mmpS5/mmpL5 was much more pronounced in the M. abscessus mutants than in the

M. tuberculosis mutants (Milano et al., 2009; Hartkoorn et al., 2014).

The TetR transcriptional regulatory factors are common single component signal transduction systems found in bacteria. These proteins possess a conserved helix-turn-helix (HTH) signature at the N-terminal of the DNA-binding domain as well as a ligand binding domain (LBD) located at the C-terminal part (Cuthbertson and Nodwell, 2013). They often act as repressors and interact with a specific DNA target to prevent or abolish transcription in the absence of an effector. In contrast, the binding of a specific ligand to the LBD induces structural changes, conducting the dissociation of the repressor from the target DNA, and the subsequent transcription of the TetR-regulated genes. Being largely associated with resistance to antibiotics and regulation of genes coding for small molecule exporters, TetR regulators also govern expression of antibiotic biosynthesis genes, quorum sensing and in distinct aspects in bacterial physiology/virulence (Cuthbertson and Nodwell, 2013). A recent global analysis indicated that the TetR regulators represent the most abundant class of regulators in mycobacteria, the vast majority remaining uncharacterized (Balhana et al., 2015). In order to provide new insight into the mechanism of gene regulation by TetR regulators in mycobacteria and to describe a new and specific drug resistance mechanism in M. abscessus, we focused our efforts on the molecular and structural characterization of the MAB\_4384-dependent regulation of MmpS5/MmpL5.

In this study, a combination of genetic and biochemical analyses was applied to determine the specificity of this regulatory system in M. abscessus and to describe the contribution of key residues that are important in driving the DNA-binding of MAB\_4384 to its operator. We report also the crystal structure of the MAB\_4384 TetR regulator. Overall, the results provide new insights into the regulation of members of the MmpL family and on a novel mechanism of drug resistance in M. abscessus.

## MATERIALS AND METHODS

#### Plasmids, Strains, Growth Conditions, and Reagents

The Mycobacterium abscessus subsp. abscessus CIP104536<sup>T</sup> reference strain and all derived mutant strains are listed in Supplementary Table S1. Strains were grown in Middlebrook 7H9 broth (BD Difco) supplemented with 0.05% Tween 80 (Sigma-Aldrich) and 10% oleic acid, albumin, dextrose, catalase (OADC enrichment; BD Difco) (7H9T/OADC) at 30◦C or in Sauton's medium in the presence of antibiotics, when required. On plates, colonies were selected either on Middlebrook 7H10 agar (BD Difco) supplemented with 10% OADC enrichment (7H10OADC) or on LB agar. For drug susceptibility testing, cultures were grown in Cation-Adjusted Mueller-Hinton Broth (CaMHB; Sigma-Aldrich). The TAC analogs D6, D15, and D17 were synthesized as reported previously (Coxon et al., 2013) and dissolved in DMSO. Other antibiotics were purchased from Sigma-Aldrich.

### Cloning of Wild-Type and Mutated MAB\_4384 and Site-Directed Mutagenesis

MAB\_4384 was PCR-amplified from M. abscessus CIP104536<sup>T</sup> purified genomic DNA using the MAB\_4384\_full primers (Supplementary Table S2) and Phusion polymerase (Thermo Fisher Scientific). The amplicon was cloned into pET32a restricted with KpnI and HindIII (New England Biolabs), enabling the introduction of MAB\_4384 in frame with the thioredoxin and poly-histidine tags as well as a Tobacco Etch Virus protease (TEV) cleavage site between the N-terminus of MAB\_4384 and the tags. The MAB\_4384 alleles harboring the g40a (D14N) and t169c (F57L) mutations were PCR-amplified using the primers described above and using the purified genomic DNA of the two spontaneous resistant M. abscessus strains to TAC analogs reported previously (Halloum et al., 2017). The double mutant carrying both g40a and t169c mutations was obtained from the MAB\_4384 (g40a) allele using the PCR-driven primer overlap extension method (Aiyar et al., 1996). Briefly, two separate PCR reactions were set up using Phusion polymerase. The first was generated using the forward primer MAB\_4384\_full Fw and a reverse internal primer MAB\_4384\_DM Rev harboring the nucleotide substitution responsible for the mutation. The second PCR was set up with an internal forward primer MAB\_4384\_DM Fw overlapping the internal reverse primer MAB\_4384\_DM Rev and with the original reverse primer MAB\_4384\_full Rev. PCR products were purified, annealed and amplified by a last PCR amplification with the MAB\_4384\_full primers. All mutated genes were cloned into pET32a, as described for wild-type MAB\_4384.

### Expression and Purification of MAB\_4384 Variants

The various pET32a-derived constructs containing either the wild-type or the mutated MAB\_4384 gene alleles were used to transform Escherichia coli strain BL21 Rosetta 2 (DE3) (Novagen). Cultures were grown in Luria-Bertani (LB) medium containing 200 µg/mL ampicillin and 30 µg/mL chloramphenicol until an optical density at 600 nm (OD600) of between 0.6 and 1.0 was reached. Liquid cultures were then placed on icy water for 30 min prior to the addition of 1 mM isopropyl β-D-1-thiogalactopyranoside (IPTG) and incubation for an additional 20 h at 16◦C. Bacteria were then collected by centrifugation (6,000 × g, 4◦C, 60 min) and the pellets were resuspended in lysis buffer (50 mM Tris-HCl pH 8, 200 mM NaCl, 20 mM imidazole, 5 mM β-mercaptoethanol, 1 mM benzamidine). Cells were opened by sonication and the lysate clarified by centrifugation (28,000 × g, 4◦C, 45 min) and subjected to a first step of nickel affinity chromatography (IMAC) (Ni-NTA Sepharose, GE Healthcare Life Sciences). After elution, the protein was dialyzed overnight at 4◦C in a buffer containing 50 mM Tris-HCl pH 8, 200 mM NaCl, 5 mM β-mercaptoethanol and TEV protease (1 mg of protease/50 mg of total protein) to cleave the thioredoxin and histidine tags from the recombinant proteins. The dialyzed preparations were purified again by IMAC, followed by an anion exchange chromatography step (HiTrap Q Fast Flow, GE Healthcare Life Sciences) as well as a final polishing step using size exclusion chromatography (SEC) (SephadexTM 75 10/300 GL, GE Healthcare Life Sciences) and a buffer containing 50 mM Tris-HCl pH 8, 200 mM NaCl and 5 mM β-mercaptoethanol.

The selenomethionine-substituted protein was expressed in the methionine auxotroph E. coli strain B834 (DE3) (Novagen). A 1L culture was grown very densely in LB medium containing 200 µg/mL ampicillin for 36 h at 37◦C. Bacteria were harvested by centrifugation and the pellets resuspended in minimal medium A without antibiotic and methionine traces (M9 medium, trace elements, 0.4% glucose, 1 µM MgSO4, 0.3 mM CaCl2, biotin and thiamine at 1 µg/mL). After an additional wash in medium A, the bacterial pellet was resuspended in 6L of medium A containing 200 µg/mL ampicillin and incubated for 2 h at 37◦C. Finally, S/L selenomethionine was added at a final concentration of 100 µg/mL. After 30 min of incubation, expression of the protein was induced with 1 mM IPTG for 5 h at 37◦C. The protein was purified using a protocol similar to the one used for the proteins expressed in the E. coli strain BL21 Rosetta 2(DE3).

#### Determination of Oligomeric States of MAB\_4384 by Size Exclusion Chromatography

The oligomeric state of MAB\_4384 and MAB\_4384:DNA complex in solution were assessed on an ENrichTM SEC 650 size exclusion column (Bio-Rad) run on an ÄKTA pure 25M chromatography system (GE Healthcare Life Science). The protein, DNA or protein:DNA complex were eluted with 50 mM Tris-HCl pH 8, 200 mM NaCl and 5 mM β-mercaptoethanol at a flow rate of 0.4 mL/min at 4◦C. MAB\_4384 (dimer) was concentrated to 3.9 mg/mL, while DNA was at 2.8 mg/mL and complexes were formed at different protein(dimer)/DNA molar ratios of 1:1, 2:1, 3:1. The molecular weights were determined based on a calibration curve generated using the Gel Filtration Markers Kit (Sigma-Aldrich) for proteins ranging from 12,400 to 200,000 Da. The column void volume was assessed with the elution peak of dextran blue. The apparent mass was obtained by plotting the partition coefficient Kav against the log values of the molecular weights of the standard proteins.

## Disruption of MAB\_4384 and mmpL5 in M. abscessus

To generate MAB\_4384 and mmpL5 knock-out mutants, internal fragments of the genes were PCR-amplified using Phusion polymerase and the specific oligonucleotide sets: MAB\_4384::pUX1 Fw with MAB\_4384::pUX1 Rev and mmpL5::pUX1 Fw with mmpL5::pUX1 Rev, respectively, digested with NheI and BamHI and ligated to NheI-BamHI-linearized pUX1 (Supplementary Table S1), a suicide vector specifically designed to perform gene inactivation in M. abscessus (Viljoen et al., 2018). Electrocompetent M. abscessus was transformed with the plasmids pUX1-MAB\_4384 and pUX1-mmpL5 and plated on 250 µg/mL kanamycin LB plates. After 5 days of incubation at 37◦C, red fluorescent colonies were selected and gene disruption resulting from homologous recombination

between the plasmid DNA and the target genes was confirmed by PCR and sequencing with appropriate primers (Supplementary Table S2).

### Quantitative Real-Time PCR

Isolation of RNA, reverse transcription and qRT-PCR were done as reported earlier (Halloum et al., 2017) using the primers listed in Supplementary Table S2.

### Drug Susceptibility Assessment

The MICs were determined according to the CLSI guidelines (Woods et al., 2011), as reported earlier (Halloum et al., 2017).

#### Electrophoretic Mobility Shift Assays

First, a typical DNA binding motif recognized by the TetR regulator, often composed of palindromic sequences or inverted repeats, was identified in silico within the intergenic region located between MAB\_4384 and the MAB\_4383c (mmpS5Mabs)/MAB\_4382c (mmpL5Mabs) gene cluster, hereafter referred to as IRS5/L5, using the MEME Suite 4.20.0 online tool<sup>1</sup> (Bailey et al., 2009). A 45 bp double stranded DNA fragment (Probe 1) containing the 27 bp palindromic sequence was labeled with fluorescein at their 5<sup>0</sup> ends (Sigma-Aldrich). Increasing amounts of purified MAB\_4384 protein were co-incubated with 280 nM of the fluorescein-labeled probes in 1X Tris Base/acetic acid/EDTA (TAE) buffer for 1 h at room temperature. The samples were then subjected to 6% native polyacrylamide gel electrophoresis for 30 min at 100 V in 1X TAE buffer. Gel shifts were visualized by fluorescence using an Amersham Imager 600 (GE Healthcare Life Sciences). All additional modified probes listed Supplementary Table S2 and used in this study were synthetized and used in electrophoretic mobility shift assay (EMSA) assays, as described above.

#### Construction of β-Galactosidase Reporter Strains and β-Gal Assays

The lacZ reporter gene encoding the β-galactosidase was amplified from the E. coli HB101 using primers listed in Supplementary Table S2. The amplicon was cloned into pMV261 cut with BamHI and HindIII, thus yielding pMV261\_Phsp60\_lacZ. The 208 bp intergenic region IRS5/L5 was amplified by PCR using M. abscessus CIP104536<sup>T</sup> genomic DNA and the MAB\_4384\_PS5/L5 primers (Supplementary Table S2) and subsequently cut with XbaI and BamHI. The hsp60 promoter was removed from pMV261\_Phsp60\_lacZ construct by restriction using XbaI and BamHI and replaced with IRS5/L5, thus creating pMV261\_PS5/L5\_lacZ. A promoterless pMV261\_lacZ construct was also generated by removing the hsp60 promoter from the pMV261\_Phsp60\_lacZ with BamHI and XbaI, blunting the overhang extremities using the T4 DNA Polymerase and selfreligation.

The β-galactosidase activity of the M. abscessus strains carrying either wild-type or mutated MAB\_4384 alleles and the various β-gal reporter constructs was monitored streaking the strains directly on 7H10OADC agar plates supplemented with 100 µg/mL kanamycin and 50 µg/mL X-gal (Sigma-Aldrich). The quantification of the β-gal activity was also assayed in liquid medium using a protocol adapted from Miller's method. Briefly, a 10 mL culture in Sauton's medium supplemented with 0.025% tyloxapol was grown until the OD<sup>600</sup> reached 0.6–1. Cultures were collected by centrifugation (4,000 × g for 10 min at 4◦C) and the bacterial pellets were resuspended in 700 µL 1X phosphate-buffered saline (PBS) prior to mechanical lysis by bead beating (3 min treatment, 30 Hz). Lysates were finally centrifuged at 16,000 × g for 10 min at 4◦C. 10 µL of clarified lysate were co-incubated 30 min at 37◦C with 100 µL of reaction buffer (60 mM Na2HPO4, 40 mM NaH2PO4, 10 mM KCl, 1 mM MgSO4, 50 mM β-mercaptoethanol) in 96-well plates. Enzymatic reactions were initiated by adding 35 µL of 2-Nitrophenyl β-D-galactopyranoside (ONPG, Sigma-Aldrich) at 4 mg/mL and absorbance was recorded at 420 nm at 34◦C using a Multimode Microplate Reader POLARstar Omega (BMG Labtech). The β-galactosidase specific activity (SAβ−Gal) was calculated using the following formula: SAβ−Gal = (Absorbance420nm × min−<sup>1</sup> )/(OD280nm × liter of culture). To test the β-gal-induction by the TAC analogs, the drugs were added directly to the cultures grown in Sauton's medium (OD<sup>600</sup> = 0.6–1) and incubated with slow shaking for 96 h at 37◦C. The β-gal activity was determined as described above.

### Crystallization, Data Collection, and Refinement

The MAB\_4384 crystals were grown in sitting drops in MR Crystallization Plates (Hampton Research) at 18◦C by mixing 1.5 µl of protein solution concentrated to 4.7 mg/mL with 1.5 µl of reservoir solution made of 100 mM sodium cacodylate pH 6.5, 200 mM MgCl2, 16% PEG 8000 and 5% DMSO. Crystals were briefly soaked in 100 mM Cacodylate buffer pH 6.5, 200 mM MgCl2, 16% PEG 8000, 5% DMSO and 10% PEG 400 prior to being cryo-cooled in liquid nitrogen. The selenomethioninesubstituted MAB\_4384 crystals were obtained in sitting drops in 96-well SWISSCI MRC plates (Molecular Dimension) at 18◦C by mixing 0.8 µl of protein solution concentrated to 2.5 mg/mL with 0.8 µl of reservoir solution consisting of 35% (v/v) 1,4 dioxane. Crystals were cryo-cooled without any additional cryoprotection. Data were processed with XDS and scaled and merged with XSCALE (Kabsch, 2010). Data collection statistics are presented in **Table 1**. The MAB\_4384 structure was solved by the single wavelength anomalous dispersion method. AutoSol from the Phenix package was used to solve the structure (Adams et al., 2010). Twelve of the fourteen potential selenium sites in the asymmetric unit were found using a resolution cutoff of 3.4 Å for the search of the Se atoms. After density modification, a clear electron density map for the two TetR monomers allowed initial model building. The resulting partial model was used to perform molecular replacement with the 1.9 Å native dataset using Phaser (McCoy et al., 2007) from the Phenix package (Adams et al., 2010). Coot (Emsley et al., 2010) was used for manual rebuilding while structure refinement and validation were performed with

<sup>1</sup>http://meme-suite.org

#### TABLE 1 | Data collection and refinement statistics.

fmicb-09-00649 April 2, 2018 Time: 16:9 # 5


The values in parenthesis are for the last resolution shell.

the Phenix package (Adams et al., 2010). The statistics for data collection and structure refinement are displayed in **Table 1**. Figures were prepared with PyMOL<sup>2</sup> . The atomic coordinates and the structure factors for the reported MAB\_4384 crystal structure has been deposited at the Protein Data bank (accession number 5OVY).

#### Docking of TAC Analogs Into the Ligand Binding Site

Docking studies was performed with PyRx (Dallakyan and Olson, 2015) running AutoDock Vina (Trott and Olson, 2010). Search was done with grid dimensions of 39.45, 39.05, 29.25 Å and origin coordinates at x = −17.8, y = 6.9, z = 0.63. The search was performed on chain B of the crystal structure of MAB\_4384 without any additional model modification.

#### RESULTS

#### MAB\_4384 Specifically Regulates Susceptibility to TAC Analogs in M. abscessus

We recently showed that mutations in the MAB\_4384 regulator were associated with the transcriptional induction of the divergently oriented adjacent genes coding for an MmpS5/MmpL5 efflux pump and accounting for high resistance levels toward various TAC analogs (Halloum et al., 2017). To gain more insight into this drug resistance mechanism in M. abscessus, detailed genetic, functional and structural characterizations of the MAB\_4384 regulator were undertaken. First, the expression profile of 19 mmpL genes was analyzed by qRT-PCR using the M. abscessus D15\_S4 strain which possesses an early stop codon in MAB\_4384 resulting in high resistance levels to the TAC analogs D6, D15 and D17 (MIC > 200 µg/mL), presumably due to derepression of the MmpS5/MmpL5 efflux pump machinery (Halloum et al., 2017). The results clearly showed a pronounced increase in the expression level of MAB\_4382c (mmpL5) mRNA in D15\_S4 in comparison to the wild-type strain as reported previously, while no marked effect on the remaining mmpL genes was detected **(Figure 1A)**. The expression levels of tgs1, encoding the primary triacylglycerol synthase responsible for the accumulation of triglycerides in M. abscessus (Viljoen et al., 2016) was included as unrelated gene control. As expected, expression of tgs1 stayed unchanged **(Figure 1A)**. To further confirm these results, the MAB\_4384 and MAB\_4382c genes were inactivated by homologous recombination using the recently developed genetic tool dedicated to facilitate gene disruption in M. abscessus (Viljoen et al., 2018), as illustrated in Supplementary Figures S1A,B. The mutant strain, designated MAB\_4384::pUX1, failed to show morphological changes (Supplementary Figure S1C) and grew similarly to its parental strain and the MAB\_4382c::pUX1 mutant (Supplementary Figure S1D), suggesting that MAB\_4384 does not play a significant role under normal in vitro conditions. However, this mutant exhibited high resistance levels to D6, D15, and D17 (MIC > 200 µg/mL, corresponding to >8-, >32-, and >16 fold-increases in MIC levels, respectively) (**Table 2**) similarly to our previous results for D15\_S4 (Halloum et al., 2017), thus validating the expected phenotype of the strain. Analysis of the transcriptional profile of all 19 mmpL genes in MAB\_4384::pUX1 confirmed the results obtained in the D15\_S4 strain **(Figure 1B)**. Interestingly, expression of MAB\_4384 itself was significantly induced in MAB\_4384::pUX1, **(Figure 1C)**, albeit lower than the expression level of mmpL5, thus indicating that MAB\_4384 is self-regulated.

Overall, these results suggest that MAB\_4384 is a unique and highly specific regulator controlling expression of mmpL5 in M. abscessus, which was strongly up-regulated in the absence of MAB\_4384.

<sup>2</sup>www.pymol.org

TABLE 2 | Drug susceptibility profile of M. abscessus S strains inactivated in either MAB\_4384 (tetR gene) or MAB\_4382c (mmpL5 gene).


The MIC (µg/mL) was determined in CaMH medium. Data are representative of three independent experiments. CFZ, clofazimine; BDQ, bedaquiline.

### MAB\_4384 Negatively Regulates Expression of mmpS5/mmpL5

The pMV261\_PS5/L5\_lacZ plasmid was constructed, containing the β-galactosidase gene as a reporter in M. abscessus to further confirm the negative regulation of MAB\_4384 on the target gene (mmpS5/mmL5 locus) expression. To do this, the 208 bp intergenic region located between MAB\_4384 and mmpS5/mmpL5 **(Figure 2A)**, designated IRS5/L5, was cloned upstream of lacZ. pMV261\_PS5/L5\_lacZ was subsequently introduced in parental smooth (S) and rough (R) variants of M. abscessus as well as in three different strains carrying single point mutations in MAB\_4384 (M1A, F57L, and D14N), previously selected for their high resistance phenotype to TAC analogs (Halloum et al., 2017). In addition, pMV261\_Phsp60\_lacZ allowing constitutive expression of lacZ under the control of the strong hsp60 promoter was produced. As expected, pMV261\_Phsp60\_lacZ led to high expression of lacZ in the wildtype strain and in the mutants compared to the promoter-less plasmid, as evidenced by the production of intense blue colonies and a strong β-Gal activity (**Figure 2B**). In contrast, whereas IRS5/L5 resulted in very low expression of LacZ in the wildtype strains, characterized by a pale blue color on plates and low β-Gal activity in liquid-grown cultures, a pronounced lacZ induction was detected in all three mutant strains (**Figure 2B**). Strikingly, the lacZ expression levels in these strains was almost comparable to the one observed in the pMV261\_Phsp60\_lacZcontaining strains.

Overall, these results indicate that expression of lacZ is strongly repressed in the presence of an intact MAB\_4384 regulator and that under derepressed conditions, the promoter driving expression of mmpS5/mmpL5 appears almost as strong as the hsp60 promoter.

## MAB\_4384 Binds to a Palindromic Sequence Within IRS5/L5

Motif-based sequence analysis using MEME (Bailey et al., 2009) revealed the presence of a 27 bp segment within the divergently oriented IRS5/L5 intergenic region and harboring a palindromic sequence **(Figure 3A)**. To test whether this motif represents a DNA binding site for MAB\_4384, EMSA was first performed using increasing concentrations of purified MAB\_4384 expressed in E. coli in the presence of a 45 bp fragment of IRS5/L5 (Probe 1; **Figure 3B**) carrying extra nucleotides flanking the palindromic sequence. Under these conditions, a DNA–protein

FIGURE 2 | Identification of a MAB\_4384-dependent regulatory promoter region in M. abscessus. (A) Schematic representation of the genome organization of MAB\_4384 and the mmpS5/mmpL5 gene cluster in M. abscessus with the purple rectangle indicating the IRS5/L5 intergenic region cloned upstream of the lacZ reporter construct. (B) The effect of MAB\_4384 on mmpS5/mmpL5 expression was assayed by constructing a plasmid with the lacZ reporter under control of IRS5/L5. A positive control plasmid consisting of the constitutive expression of lacZ under the control of the hsp60 promoter and a negative control consisting of a promoter-less lacZ were also generated. All constructs were introduced into wild-type smooth (S) and rough (R) variants as well as into three different strains harboring single point mutations in MAB\_4384 (M1A, D14N and F57L replacements). Exponentially growing M. abscessus cultures were streaked onto 7H10 plates containing 100 µg/mL kanamycin and 50 µg/mL X-gal. The plates were subsequently incubated for 3–4 days at 37◦C and visualized for their appearance with respect to white-to-blue coloration. The β-galactosidase specific activity (SAβ−Gal) was quantified in liquid cultures using ONPG as a substrate. Results were obtained from three independent experiments and the error bars represent standard deviation. The capped lines indicate the groups compared. For statistical analysis, the Student's t-test was applied with ns, ∗∗ , ∗∗∗ , ∗∗∗∗ indicating non-significant, p < 0.01, p < 0.001, and p < 0.0001, respectively.

complex was seen **(Figure 3C)**. To confirm the specificity of the binding, a competition assay with increasing concentrations of the corresponding unlabeled probe (cold probe) was carried out, leading to a dose-dependent decrease of the DNA–protein complex **(Figure 3C)**. In addition, in the presence of an excess of a non-related labeled probe, the shift was maintained, thus indicating that a specific protein–DNA complex was seen only when MAB\_4384 was incubated with DNA containing the specific inverted repeat sequence. To better define the

FIGURE 3 | Binding activity of MAB\_4384 to a palindromic region within IRS5/L5. (A) DNA sequence of IRS5/L5 and representation of the operator composed of a 27 bp region containing two degenerated inverted sequences of 13 nucleotides each (black arrows) and separated by a one nucleotide spacer. The probe used to perform the EMSA is delimited by dotted lines (Probe 1). (B) DNA sequences of all the various 5<sup>0</sup> fluorescein-labeled probes used in this study. (C) EMSA and competition assay using probe 1. Protein and DNA concentrations are expressed in µM. In competition assays, the concentration of Probe 1 was fixed at 280 nM. Gel shifts were revealed by fluorescence emission. (D–H) EMSA using Probes 2 to 8, each time compared to the shift profile obtained with Probe 1. Experiments were reproduced three times with similar results.

minimal motif and the importance of the nucleotides involved in recognition and binding of the protein, a large set of fluoresceinlabeled probes differing in their size and/or sequence were next assayed **(Figure 3B).** In the presence of Probe 2, in which only the right inverted sequence was conserved, a delay in the DNA shift was observed (partial in the presence of 1.75 µM of protein as compared to the reaction in the presence of Probe 1). Shifts using Probe 3, where the extra nucleotides surrounding the palindromic sequence were changed randomly were comparable to those obtained with Probe 1 **(Figure 3D)**, indicating that the extra-palindromic sequence does not influence protein binding. With Probe 4, where the inverted repeats were shortened by six nucleotides, the formation of the DNA–protein complex was severely impeded even with the highest concentration of protein tested (3.5 µM) **(Figure 3E)**. Increasing the spacer between the two inverted repeats by 10 nucleotides (Probe 5) negatively impacted the shift **(Figure 3F)**. Substitutions of two nucleotides at the extremities in each repeat sequence (Probe 6) was accompanied by a pronounced shift alteration **(Figure 3G)** and similar results were obtained when di-nucleotide substitutions occurred at other positions within the conserved palindromic sequence (Probes 7 and 8) **(Figure 3H)**.

Overall, these results confirm that a strict preservation of this inverted sequence and space separating these two motifs are crucial for binding of MAB\_4384 to its target.

#### Asp14 and Phe57 Are Critical for Optimal DNA-Binding Activity of MAB\_4384

Multiple primary sequence alignments of the MAB\_4384 N-terminus with other TetR regulators with known threedimensional structures indicate that the N-terminus Asp14 residue is well conserved in several other Tet regulators **(Figure 4A)**. Similarly, Phe57 was also found to be part of a highly conserved stretch of amino acids in these proteins, although, in some instances, Phe was replaced by bulky/hydrophobic residues **(Figure 4A)**. The importance of the conservation of these two residues for the function of MAB\_4384 and presumably also for that of the other TetR regulators, was next assessed by EMSA using the purified MAB\_4384 mutated variants. As compared to the shift profile with wild-type MAB\_4384, the production of the DNA–protein complex was severely impaired in the presence of either MAB\_4384 (D14N) **(Figure 4C)** or MAB\_4384 (F57L) **(Figure 4D)** and fully abrogated in the presence of the double mutant (D14N/F57L) **(Figure 4E)**.

FIGURE 4 | D14N and F57L mutations abrogate binding of MAB\_4384 to its palindromic DNA target. (A) Multiple primary sequence alignments of MAB\_4384 N-terminus with other TetR family members whose crystal structures are known (PDB code indicated) from different microorganisms showing the conservation of the Asp14 and Phe57 residues (indicated with black arrows). (B) Mapping of the mutations on the crystal structure of MAB\_4384, the degree of residue conservation (obtained from the sequence alignment in A) is represented by the coloration, from white (low conservation) to red (high conservation). (A,B) Were prepared using the ENDscript server (http://endscript.ibcp.fr/ESPript/ENDscript/). EMSA were performed using increasing concentrations (in µM) of the purified MAB\_4384 (D14N) (C), MAB\_4384 (F57L) (D) or MAB\_4384 (D14N/F57L) (E) in the presence of Probe 1. Wild-type MAB\_4384 was included as a positive control in each assay. The concentration of Probe 1 was fixed at 280 nM. Gel shifts were revealed by fluorescence emission. Three independent experiments were performed with similar results.

Overall, these results support the importance of both residues in the DNA-binding capacity of MAB\_4384 and the impaired ability of the mutants to bind to the operator is in agreement with the derepression of lacZ transcription in the M. abscessus strains carrying the D14N or F57L mutations **(Figure 2B)**.

### Oligomeric States of MAB\_4384 and MAB\_4384:DNA in Solution

To further characterize the MAB\_4384:DNA complex formation, we next assessed its stability in solution by SEC **(Figure 5)**. The oligomeric state of MAB\_4384 in solution has an apparent 42.6 kDa molecular weight as compared to its 24.7 kDa theoretical molecular mass calculated from its primary sequence, thus highlighting the dimeric state of MAB\_4384 in solution. MAB\_4384 (dimer) was next incubated with the non-fluorescent DNA Probe 1 **(Figure 3A)** in a 1:1 molar ratio. A stable complex elution peak at 12.6 mL clearly shifted from the elution peak of MAB\_4384 and DNA alone **(Figure 5)**, allowing deduction of the molecular mass of the protein:DNA complex at 102 kDa. As the DNA alone in solution appeared on SEC as a 24.5 kDa molecule, these results strongly suggest the existence of two MAB\_4384 dimers bound to one DNA molecule as such a complex would possess a molecular mass of 109.7 kDa (2 × 42.6 kDa + 24.5 kDa). This 2-to-1 binding mode was further corroborated by the fact that an elution peak corresponding to free DNA can be seen at 14.4 mL when we mixed the MAB\_4384 dimer and DNA in a 1:1 ratio. Moreover, increasing the molar ratio of the MAB\_4384 dimer:DNA complex (2:1 and 3:1) did not yield larger protein:DNA complexes (data not shown), suggesting that, at a 1:1 molar ratio, the operator is already saturated by the protein. This observation is not unique as two TetR dimers have been shown to bind their DNA targets in other microorganisms, such as in Staphylococcus aureus (Grkovic et al., 2001) or in Thermus thermophilus (Agari et al., 2012).

#### Crystal Structure of MAB\_4384

To understand, at a molecular basis, how the D14N or F57L mutations generate resistance to TAC analogs, we first crystallized and determined the X-ray structure of MAB\_4384. Although the structure of the protein could not be solved by molecular replacement, the phase problem was overcome with the SAD method using crystals of selenomethionine-substituted MAB\_4384 **(Table 1)**. The crystal structure of the native protein was subsequently solved with the partial model obtained from the SAD data and refined to a resolution of 1.9 Å. The asymmetric unit contains two subunits. Chain A was modeled from residues Asp14 to Thr213, indicating that the first thirteen residues, one residual Gly residue from the tag and the last eight residues in the C-terminus were not visible in the electron density. Chain B showed also disordered regions as the first ten residues, one Gly residue from the tag in N-terminus as well as the last nine residues in the C-terminus, could not be modeled. Analysis of the crystal packing using the PISA server (Krissinel and Henrick, 2007) predicted the existence of a stable homodimer formed within the crystal, consistent with other TetR regulators (Cuthbertson

FIGURE 5 | Oligomerization of MAB\_4384 and the MAB\_4384:DNA complex in solution. The oligomeric states of MAB\_4384 alone or complexed to its DNA target were determined by size exclusion chromatography. The elution profile of the proteins, used for the calibration is displayed as a dashed black line. Calibration was established using β-amylase (200 kDa) (1), bovine serum albumin (66 kDa) (2), carbonic anhydrase (29 kDa) (3), and cytochrome C (12.4 kDa) (4), eluting with estimated volumes of 11.7, 13.2, 14.9, 15.9 mL, respectively. The void volume was determined with the elution volume (8.8 mL) of dextran blue. MAB\_4384 (dimer) was at a concentration of 3.9 mg/mL. The elution profiles of MAB\_4384 dimer, DNA target and MAB\_4384 bound to the DNA target are shown in orange, blue, and pink and their respective elution peaks were at 13.9, 14.7, and 12.6 mL. Kav indicates the partition coefficient.

and Nodwell, 2013) and with the SEC profile of MAB\_4384 in solution **(Figure 5)**.

The two subunits are very similar as their superposition leads to a root mean square deviation (r.m.s.d.) of 0.53 Å over 198 aligned residues. The N-terminus comprises the DNA binding domain (DBD), followed by the LBD. The DBD is composed of three α-helices α1: residues 12–33, α2: 39–46, and α3: 50–56, where helices 2 and 3 form a helix-turn-helix (HTH) motif. The LBD is made of seven α-helices, α4: 60–82, α5: 88–104, α6: 107–114, α7: 121–143, α8: 155–170, α9: 178–188, and α10: 205–211 **(Figure 6A)**. The surface of dimerization of about 1,700 Å 2 is mediated by 31 residues mainly from helices α8 and α9 of each subunit and involves numerous interactions notably five salt bridges, fourteen hydrogen bonds and van der Waals interactions.

Interestingly, we noticed the presence of extra electron density in the LBD of chain B in a rather hydrophobic pocket **(Figure 6B)**. Although we could not interpret this density, we hypothesize that it may correspond to a compound present in the crystallization solution, such as PEG. Search for structural homologs in the PDBeFold server indicated that the closest structure to MAB\_4384 corresponds to the LfrR TetR transcriptional regulator from Mycobacterium smegmatis bound to proflavin (PDB id : 2V57) (Bellinzoni et al., 2009) with an r.m.s.d. of 2.6 Å and sharing 16% primary sequence identity with MAB\_4384. However, only one subunit of each structure could

be superposed as the overall dimers differed largely **(Figure 6C)**. LrfR represses the expression of the LfrA efflux pump (Buroni et al., 2006) and mediates resistance to ethidium bromide, acriflavine, and fluoroquinolones (Takiff et al., 1996).

Due to the occurrence of an extra electron density within the LBD of MAB\_4384 and that the closest structure of MAB\_4384 is LfrR in its ligand bound form, it is very likely that MAB\_4384 was crystallized in its open conformation, i.e., its derepressed form that is not able to interact with DNA. This was assessed by determining the distance between two residues from the DBD susceptible to interact with DNA. Residues Arg55 from chains A and B are about 56 Å apart **(Figure 6D)**. In comparison, the distances between the equivalent residues in various TetR:DNA complexes are largely reduced. In the TetR:DNA complex (PDB id: 4PXI) from Streptomyces coelicolor this distance is 45 Å (Bhukya et al., 2014), in the TetR:DNA complex from M. smegmatis (PDB id: 4JL3) (Yang et al., 2013) **(Figure 6E)**, E. coli (PDB id: 1QPI) (Orth et al., 2000), Corynebacterium glutamicum (PDB id: 2YVH) (Itou et al., 2010) or Staphylococcus aureus (PDB id: 1JT0) (Schumacher et al., 2002) the distances are 38 Å, 30 Å, 42 Å, and 37 Å, respectively. From these results it can be inferred that the DBDs of MAB\_4384 are too far from each other to bind to the DNA groove. These observations combined with the presence of an unidentified ligand in the LBD strongly suggest that the MAB\_4384 structure is in an open conformation.

### Structural Basis of the Resistant Phenotype of the Mutants

To determine the impact of the mutations in the spontaneous resistant M. abscessus mutants, the D14N and F57L residues were mapped on the crystal structure of MAB\_4384. Asp14 is located at the beginning of helix α1 and is conserved in several TetR protein members **(Figures 4A,B)**. Residues from helix α1 are often found in contact with DNA as seen in several TetR:DNA crystal structures. Nonetheless, due to the acidic nature of Asp, it is more likely that this residue repulses DNA. We, therefore, hypothesize that it may instead contribute to the correct positioning of other residues located in its close vicinity. Alternatively, repulsion may promote important interactions of DNA with other residues. Indeed, in other collected datasets at lower resolution (data not shown), Asp14 was found to establish a salt bridge interaction, thereby stabilizing the side chain of Arg17 that could interact with DNA. In the absence of a crystal structure of MAB\_4384 bound to DNA it is, however, difficult to convincingly affirm the impact of the D14N substitution. However, neither the repulsion of DNA nor the establishment of a salt bridge would be possible if Asn is present instead of Asp, presumably explaining the loss of DNA binding activity of the TetR D14N mutant.

The role of Phe57 situated on helix α3 is more obvious as this position appears always occupied by bulky residues (Phe, Tyr, or Trp) in numerous TetR proteins **(Figure 4A)**. The side chain of Phe57 contacts the side chains of Val22 from helix α1 and Leu63 from helix α4 **(Figure 4B)**. Phe57 is very likely to perform an important structural role in stabilizing the DBD. Replacement with a less bulky side chain such as Leu would abolish these contacts with helices α1 and α4 residues, thus perturbing the overall structural fold of this domain and suppressing the DNAbinding capacity of MAB\_4384.

### Drug Recognition of MAB\_4384 Induces Expression of MmpS5/MmpL5

TetR regulators can respond to small molecules and the best characterized member of this family of regulators is E. coli TetR itself. It confers resistance to tetracycline by regulating the expression of the tetracycline TetA efflux pump (Hillen and Berens, 1994). When tetracycline binds to TetR, the regulator loses affinity for the operators, conducting derepression of tetA and extrusion of tetracycline out of the bacteria (Lederer et al., 1995). To investigate whether TAC derivatives could bind to the LBD of MAB\_4384, in silico docking was performed. Despite using a large grid box covering the entire LBD, all three compounds seem to be accommodated by the same binding pocket **(Figure 7A)**. Interestingly, this pocket positioned exactly where the extra electron density was seen in the LBD **(Figure 6B)**. All the compounds bind with similar energies in the aforementioned hydrophobic binding pocket. A slightly stronger interaction for the most hydrophobic derivative D17 was nonetheless observed. D17 and D6 that seem to bind stronger are more hydrophobic and in their best docking poses their thiosemicarbazide group is differently oriented as compared to D15.

Next, we determined whether expression of mmpS5/mmpL5 can be conditionally induced by the substrates that are extruded by the efflux pump system. This was achieved by determining the effect of the D6, D15, and D17 analogs on LacZ production using the pMV261\_PS5/L5\_lacZ reporter strain in M. abscessus incubated in Sauton's medium with various drug concentrations consisting of 1X, 2.5X, and 5X the MIC for D15 and of 0.5X, 1X, and 2.5X for D6 and D17. Kinetic studies indicated that optimal expression was obtained after 96 h of treatment (data not shown). The LacZ assay showed that transcription was induced by the TAC analogs in a dose-dependent manner whereas non-related drugs such as amikacin or the DMSO control had no effects **(Figure 7B)**. In comparison with the basal transcriptional level in Sauton's medium (no drug control), the addition of TAC derivatives in the cultures resulted in a reproducible 2.5- to 5-fold increase in the detection of β-Gal activity with D17 being the most potent inducer at 2.5X MIC. However, ethionamide, an antitubercular drug that, like the TAC and TAC analogs, requires to be activated by the EthA monooxygenase (Baulard et al., 2000; DeBarber et al., 2000; Dover et al., 2007; Halloum et al., 2017) failed to induce lacZ at 5X MIC (previously determined at 16 µg/ml). Induction of lacZ by D17 treatment was further confirmed at a transcriptional level from the pMV261\_PS5/L5\_lacZ cultures treated with 2.5× MIC of D17 for 8 h **(Figure 7C**, left). This effect was specific to D17 as no gene induction was observed in the DMSO-treated cultures. Consistently, transcription profiling of mmpS5 and mmpL5 in the D17-exposed cultures clearly showed a marked induction level as compared to the DMSO-treated cultures and no effect on tgs1 expression **(Figure 7C**, right).

involved in van der Waals, hydrophobic bonds or hydrogen bonds (in black dashes) are displayed as sticks. D15 in salmon has a binding energy of 1G = –6.6 kcal/mol, D6 in magenta has a 1G = –7.3 kcal/mol, and D17 seems to bind slightly stronger with a 1G = –8.3kcal/mol. (B) Conditional induction of lacZ by structural analogs of TAC in M. abscessus. Induction of β-Gal activity in wild-type M. abscessus S carrying pMV261\_PS5/L5\_lacZ was assayed using mid-log phase cultures incubated with increasing drug concentrations varying from 1X to 5X the MIC for D15 and varying from 0.5X to 2.5X the MIC for D6 and D17. Inductions were performed for 96 h at 37◦C. The β-galactosidase specific activity (SAβ−Gal) was quantified in liquid cultures using ONPG as a substrate. Amikacin (AMK) and ethionamide (ETH) were included as unrelated drug controls. (C) Transcriptional profile of lacZ in the M. abscessus pMV261\_PS5/L5\_lacZ reporter strain exposed to 2.5X the MIC of D17 for 8 h (left). tgs1 was included as a non-relevant control. Replacing D17 by an equal volume of DMSO had no effect on lacZ transcription. Transcriptional induction of mmpS5Mabs and mmpL5Mabs following exposure to 2.5X the MIC of D17 for 8 h (right). Results were obtained from three independent experiments and the error bars represent standard deviation. For statistical analysis the Student's t-test was applied with ns, <sup>∗</sup> , ∗∗ , ∗∗∗ , ∗∗∗∗ indicating non-significant, p < 0.05, p < 0.01, p < 0.001, and p < 0.0001, respectively.

Together, these results support the view that TAC analogs, which are substrates of MmpS5/mmpL5, are also effectors of MAB\_4384-induced transcription of mmpS5/mmpL5.

### DISCUSSION

Herein, a combination of genetic, biochemical and structural studies was used to demonstrate that MAB\_4384 is part of the TetR family of regulators, which represses the transcriptional expression of the MmpS5/MmpL5 efflux pump. MAB\_4384 belongs to the type I class TetR family of regulators, characterized by a divergent orientation to one of the adjacent target genes (Cuthbertson and Nodwell, 2013). In M. tuberculosis, MmpS5/MmpL5 is under the control of the MarR repressor Rv0678 (Radhakrishnan et al., 2014) and mutations in this regulator leads to drug resistance (Andries et al., 2014; Hartkoorn et al., 2014; Zhang et al., 2015). EMSA indicated a direct binding of Rv0678 to the intergenic region located between mmpS5 and Rv0678. However, shifts were also found using the promoter regions of mmpS2-mmpL2, mmpS4-mmpL4, and Rv0991-Rv0992 (Radhakrishnan et al., 2014), suggesting that a single regulator can control expression of several mmpS/mmpL loci. Our analysis indicates that, despite the fact that M. abscessus possesses the highest number of mmpL genes among all mycobacterial species studied (Viljoen et al., 2017), the MAB\_4384 regulator is highly specific to the mmpS5/mmpL5 pair as demonstrated by the lack of transcriptional regulation of a large set of mmpL genes and the presence of a unique inverted DNA sequence target that was not found elsewhere in the chromosome. This unique trait might also be reflected by the modest structural homology of MAB\_4384 with other TetR crystal structures. The tight regulation and the high specificity of interaction with its target DNA, however, cannot be solely explained on the basis of the MAB\_4384 crystal structure and the structure of the MAB\_4384:DNA complex is, therefore, greatly warranted to dissect these underlying mechanisms. Nevertheless, our structural analysis underscores the strategy employed by M. abscessus to acquire mutations

impacting the DNA-binding capacity or the folding/stability of the DBD of MAB\_4384 to become resistant.

EMSA and lacZ reporter fusions confirmed that D14N and F57L mutations, alleviating the DNA-binding activity of MAB\_4384, cause a strong up-regulation of mmpS5/mmpL5 gene expression, in agreement with our previous qRT-PCR analyses (Halloum et al., 2017). This leads to extrusion of the TAC derivatives out of the cells, contributing to the high MIC values for TAC derivatives against these mutants, as illustrated in **Figure 8**. Expression of multi-drug resistant efflux pumps can also be conditionally induced using structurally diverse substrates (Kaatz and Seo, 1995; Rosenberg et al., 2003; Buroni et al., 2006). This induction is caused by the direct interaction of these substrates with the repressors, interfering with binding of the repressors to their target operators and resulting in increased expression of the pumps. Here, we show inducible β-galactosidase activity following treatment with D6, D15, or D17, a mechanism that is very likely to be meditated by MAB\_4384. This view is reinforced by the fact that docking studies highlighted the possibility that all three analogs could be accommodated in the LBD of the protein, which perfectly coincided with the extra electron density observed. Since, the LBD are remote from the DBD, the derepression of TetR family regulators involves allosteric mechanisms that include conformational changes transmitted largely within the same subunit (Ramos et al., 2005). The interaction of ligands with the LBD captures a conformational state where the DBD is repositioned relative to the LBD in a way that the dimer is prevented from binding to its target DNA. However, definitive proof of this mechanism awaits the elucidation of the crystal structure of the D17-bound form of MAB\_4384, as reported for instance with the hexadecyl octanoate-bound EthR repressor (Frénois et al., 2004) or the LfrR regulator complexed with proflavine (Bellinzoni et al., 2009). Lack of inducible lacZ expression in M. abscessus cultures exposed to amikacin, for which mutations in 16S rRNA represent a major mechanism of resistance (Prammananan et al., 1998), indicates that MmpL5 mediated efflux cannot mediate resistance toward this antibiotic in line with the lack of cross-resistance toward amikacin observed for TAC derivative-spontaneous resistant mutants (Halloum et al., 2017). The specificity of the MAB\_4384-driven resistance mechanism described herein is further supported by the lack of lacZ induction during exposure to ETH, that similarly to TAC and TAC analogs, requires bio-transformation by EthA, whose expression is also dependent on the EthR regulator belonging to the TetR family (Baulard et al., 2000; Engohang-Ndong et al., 2004; Halloum et al., 2017). Together, these findings strongly suggest that when TAC analogs bind to MAB\_4384, the regulator loses affinity for its DNA target, resulting in up-regulation of mmpS5/mmpL5 and export of the drugs from the cells **(Figure 8)**. These results also point out the selectivity of this efflux-based mechanism. Indeed, no change in the MIC of clofazimine or bedaquiline were noticed in a MAB\_4384-disrupted strain, which appears intriguing as MmpL5 has been reported as a multisubstrate efflux pump responsible for low-level resistance to both of these drugs in M. tuberculosis (Hartkoorn et al., 2014). The LBD of MAB\_4384 potentially can accommodate bulky molecules and might thus indicate that MAB\_4384 is involved in efflux of other types of compounds in addition to TAC analogs. However, we could neither dock clofazimine nor bedaquiline in the LBD of MAB\_4384 (not shown). Several reasons can be put forth to explain these species-specific variations. In M. tuberculosis, expression of MmpL5 is under the control of a MarR regulator rather than a TetR regulator. Alternatively, we have previously reported the occurrence in M. abscessus of three mmpS5/mmpL5 paralogs (Halloum et al., 2017), thus, it remains possible that either of the two remaining genes may participate in co-resistance to these drugs in M. abscessus.

The highly pronounced expression of lacZ under derepressed conditions found in the M1A, F57L, or D14N mutant strains, almost at levels similar to those driven by the strong and constitutive hsp60 promoter, confirmed the very high expression levels of mmpS5 and mmpL5 detected by qRT-PCR and probably explains the very high level of resistance of the mutants (MIC > 200 µg/mL). This contrasts also with findings where MmpL5 mediates only low-levels of resistance in M. tuberculosis (Andries et al., 2014; Hartkoorn et al., 2014), presumably because expression of mmpL5 is driven by a weaker promoter than in M. abscessus. That mmpS5/mmpL5 expression is tightly controlled suggests that the MmpS5/MmpL5 machinery may exert an important function in the assembly and/or maintenance of the cell wall by exporting a yet unidentified lipid, as already reported for several MmpL transporters in M. tuberculosis (Chalut, 2016; Viljoen et al., 2017). Alternatively, they may participate in adaptation during the infection process. However, the growth curves of the wild-type or the strain constitutively expressing high levels of MmpL5 (due to the M1A mutation in MAB\_4384) were comparable in vitro. In addition, microinjections of the different strains were done in the zebrafish embryo, an animal model previously developed to study the early events of the M. abscessus infection (Bernut et al., 2014, 2015). No differences in virulence were noticed between the wild-type and MAB\_4384 (M1A) strains (Supplementary Figure S2). Interestingly, the mmpS5/mmpL5 locus was found to be induced when M. abscessus was exposed to a defined, synthetic medium that mimics the composition of CF sputum (Miranda-CasoLuengo et al., 2016). This may be part of a complex adaptive transcriptional response to the mucus layer of the CF airways that leads to the chronic infections of M. abscessus.

In summary, this study provides new functional and structural insights into TetR-dependent regulation of MmpL efflux pumps in mycobacteria. Considering the exceptionally high abundance of TetR transcriptional regulators (more than 130) as well as the important MmpL repertoire (around 30) in M. abscessus, one can anticipate that mechanisms similar to the one described here are exploited by this pathogen to express its intrinsic resistance level to other antibiotics.

#### ETHICS STATEMENT

Zebrafish experiments were done at IRIM, according to European Union guidelines for handling of laboratory animals

(http://ec.europa.eu/environment/chemicals/lab\_animals/home\_ en.htm) and approved by the Direction Sanitaire et Vétérinaire de l'Hérault and Comité d'Ethique pour l'Expérimentation Animale de la Région Languedoc Roussillon (CEEA-LR) under the reference CEEA-LR-1145.

#### AUTHOR CONTRIBUTIONS

MR, AVG, AV, MB, and LK acquired and analyzed the data. EG, MB, and LK wrote the manuscript. LK conceived and designed the study.

#### FUNDING

This work was supported by the Fondation pour la Recherche Médicale (FRM) (grant number DEQ20150331719 to LK; grant

### REFERENCES


number ECO20160736031 to MR) and by the InfectioPôle Sud for funding the Ph.D. Fellowship of AVG.

#### ACKNOWLEDGMENTS

The authors wish to thank G. S. Coxon for the generous gift of the TAC analogs. The crystallographic data were collected on beamline ID30B at the European Synchrotron Radiation Facility (ESRF), Grenoble, France. The authors are grateful to Local Contact at the ESRF for providing assistance in using beamline ID30B.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2018.00649/full#supplementary-material

multidrug-resistant nontuberculous mycobacterium. Science 354, 751–757. doi: 10.1126/science.aaf8156


in cystic fibrosis. J. Cyst. Fibros. 9, 117–123. doi: 10.1016/j.jcf.2009.12. 0001



**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 Richard, Gutiérrez, Viljoen, Ghigo, Blaise and Kremer. 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.

# GlnR Activation Induces Peroxide Resistance in Mycobacterial Biofilms

Yong Yang<sup>1</sup> , Jacob P. Richards1,2, Jennifer Gundrum<sup>1</sup> and Anil K. Ojha1,3 \*

<sup>1</sup> Division of Genetics, Wadsworth Center, New York State Department of Health, Albany, NY, United States, <sup>2</sup> Department of Infectious Diseases and Microbiology, University of Pittsburgh, Pittsburgh, PA, United States, <sup>3</sup> Department of Biomedical Sciences, University at Albany, Albany, NY, United States

Mycobacteria spontaneously form surface-associated multicellular communities, called biofilms, which display resistance to a wide range of exogenous stresses. A causal relationship between biofilm formation and emergence of stress resistance is not known. Here, we report that activation of a nitrogen starvation response regulator, GlnR, during the development of Mycobacterium smegmatis biofilms leads to peroxide resistance. The resistance arises from induction of a GlnR-dependent peroxide resistance (gpr) gene cluster comprising of 8 ORFs (MSMEG\_0565-0572). Expression of gpr increases the NADPH to NADP ratio, suggesting that a reduced cytosolic environment of nitrogenstarved cells in biofilms contributes to peroxide resistance. Increased NADPH levels from gpr activity likely support the activity of enzymes involved in nitrogen assimilation, as suggested by a higher threshold of nitrogen supplement required by a gpr mutant to form biofilms. Together, our study uniquely interlinks a nutrient sensing mechanism with emergence of stress resistance during mycobacterial biofilm development. The gpr gene cluster is conserved in several mycobacteria that can cause nosocomial infections, offering a possible explanation for their resistance to peroxide-based sterilization of medical equipment.

Keywords: mycobacteria, biofilms, peroxide, starvation, GlnR

## INTRODUCTION

Under most detergent-free in vitro conditions, mycobacterial species grow as surface-associated, three-dimensionally organized multicellular communities, called biofilms, which develop through dedicated genetic programs (Ojha et al., 2005, 2008; Weiss and Stallings, 2013; Gupta et al., 2015; Chuang et al., 2016; Yang et al., 2017; Clary et al., 2018). Biofilm-like multicellular aggregates of non-tuberculous mycobacteria (NTMs) have also been reported from clinical and environmental specimens (Feazel et al., 2009; Bosio et al., 2012; Mullis and Falkinham, 2013; Fennelly et al., 2016). Biofilms of Mycobacterium avium, a prominent member of NTMs, have been implicated in pathogenesis (Rose and Bermudez, 2014), although biofilms of other pathogenic mycobacterial species including M. tuberculosis in the context of their host environments remain to be evaluated. Further clinical significance of mycobacterial biofilms is highlighted by at least two unique phenotypes, which are not associated with their single-cell planktonic counterparts. First, mycobacterial biofilms harbor a sizable subpopulation of bacilli that can survive extreme conditions including exposure to antibiotics and antiseptics (Falkinham, 2007; Ojha et al., 2008; Rose et al., 2015; Yang et al., 2017; Clary et al., 2018). Second, biofilm growth of some mycobacterial species, including the pathogenic species M. canettii, fosters horizontal gene transfer

#### Edited by:

Thomas Dick, Rutgers, The State University of New Jersey, Newark, United States

#### Reviewed by:

John T. Belisle, Colorado State University, United States Tanya Parish, Infectious Disease Research Institute, United States

#### \*Correspondence:

Anil K. Ojha anil.ojha@health.ny.gov

#### Specialty section:

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

Received: 27 April 2018 Accepted: 11 June 2018 Published: 04 July 2018

#### Citation:

Yang Y, Richards JP, Gundrum J and Ojha AK (2018) GlnR Activation Induces Peroxide Resistance in Mycobacterial Biofilms. Front. Microbiol. 9:1428. doi: 10.3389/fmicb.2018.01428

(Nguyen et al., 2010; Boritsch et al., 2016), which possibly accelerates the propagation of drug resistance mutations in these species. Although mycobacterial biofilms are increasingly being recognized as potential targets for effective anti-mycobacterial strategies, mechanisms underlying the emergence of stress tolerance in biofilms remain unknown.

Biofilm development in the model mycobacterial species, M. smegmatis, is a genetically programmed process that appears to occur in distinct stages, each demarcated by its specific genetic requirements (Ojha et al., 2005; Ojha and Hatfull, 2007; Yang et al., 2017). While the cell surface glycopeptidolipid (GPL) is necessary for optimum substratum attachment, a nucleoidassociated protein, Lsr2, is required for cell–cell aggregation (Yang et al., 2017). Moreover, gene expression analysis of an extragenic suppressor of an lsr2 mutant revealed that cell–cell aggregation is a critical checkpoint in the developmental process (Yang et al., 2017). Expression levels of 83 genes are dependent on intercellular aggregation and aggregated growth (Yang et al., 2017), suggesting that the physicochemical interactions among cells induce transcriptional reprogramming for further maturation of architecture and physiological adaptation of resident cells.

A large number of 83 aggregation-dependent genes are under the control of GlnR, a conserved OmpR-like transcription factor that regulates nitrogen assimilation in response to its limited availability (Amon et al., 2008; Jenkins et al., 2013; Yang et al., 2017). GlnR-dependent upregulation of three ammonium transporters (Amt), glutamine/glutamate synthases (GlnA) and nitrite/nitrate reductases facilitate efficient assimilation of environmental nitrogen in a cell (Amon et al., 2008; Yang et al., 2017). In addition, GlnR also induces urecase, amidase, xanthin permeases, which likely maximize the total intracellular nitrogen pool (Jenkins et al., 2013). However, the fact that GlnR induces over 100 genes in mycobacteria opens up questions about its wider influence in mycobacterial growth and adaptation. Studies in other species support a global role of GlnR, extending beyond nitrogen assimilation. In Saccharopolyspora erythraea, GlnR controls the expression of carbohydrate ATPbinding cassette (ABC) transporters, thereby facilitating carbon uptake in response to nitrogen starvation (Liao et al., 2015). Similarly, GlnR also appears to control the expression of a key phosphate-sensing regulator, PhoP, in S. erythraea, implying a possible role of GlnR in phosphorous homeostasis (Yao and Ye, 2016). In addition to controlling nutrient balance in bacteria, GlnR also influences secondary metabolism in actinomycetes. In Streptomyces coelicolor and Streptomyces avermitilis, GlnR modulates the synthesis of antibiotics by directly regulating the transcription of pathway-specific genes (He et al., 2016; Urem et al., 2016). Lastly, GlnR also controls osmolyte levels in S. coelicolor (Shao et al., 2015), and pH homeostasis in Streptococcus salivarius (Huang and Chen, 2016).

Given a global role of GlnR in other species, we asked whether its activation during biofilm development in M. smegmatis has any significance beyond nitrogen assimilation. We report here that GlnR activation for nitrogen assimilation during biofilm growth also increases resistance to peroxide. The phenotype is caused by induction of a GlnR-dependent peroxide resistance (gpr) cluster of genes. The gpr cluster is comprised of 8 open reading frames (ORFs) – MSMEG\_0565-0572– encoding genes of diverse functions. The upstream region of this uncharacterized operon has binding sites for both GlnR and SoxR, which is a MarR-family transcription factor that responds to oxidative stress to maintain redox homeostasis (Dietrich et al., 2008; Jenkins et al., 2013). However, gpr induction responds differently to GlnR and SoxR activities. While GlnR is a strong positive regulator of gpr, inducing it by ∼100-fold under limiting nitrogen, SoxR is a modest negative regulator causing twofold de-repression under peroxide stress. Emergence of peroxide resistance through a nutrient sensing mechanism provides a direct link between form and function of mycobacterial biofilms.

### MATERIALS AND METHODS

### Bacterial Strains and Growth Conditions

All plasmids and strains used in this study are listed in Supplementary Tables S1 and S2, respectively. Unless indicated, M. smegmatis, mc<sup>2</sup> 155 (wild-type), was maintained at 37◦C in 7H9ADC (Becton Dickinson) with 0.05% (v/v) Tween-80 for planktonic cultures. 7H10ADC agar (Becton Dickinson) was used for plate cultures. When necessary, hygromycin, kanamycin, and zeocin were added at 150, 20, or 25 µg/mL, respectively, to culture recombinant strains. Escherichia coli (DH5α) was grown at 37◦C in LB broth or LB agar. Pellicle biofilms of M. smegmatis strains were grown as described earlier (Yang et al., 2017). Briefly, 10 µL of saturated planktonic cultures were inoculated into 10 mL of detergent-free Sauton's medium or modified M63 medium in either 60 mm polystyrene dishes or 12-well polystyrene plates, and incubated stationary at 30◦C untill indicated time. The N<sup>0</sup> version of Sauton's medium was prepared by omitting asparagine, and by replacing the ferric ammonium citrate with ferric citrate. The N1/<sup>2</sup> version was prepared by reducing the initial concentrations of the above mentioned nitrogen sources by half.

### Construction of Mutants and Plasmids

Mutations in M. smegmatis mc<sup>2</sup> 155 were constructed using recombineering as described previously (Yang et al., 2017). Briefly, allelic exchange substrates for a given target gene were generated by SOEing-PCR on either side of a loxP flanked zeocin-resistant cassette using the respective primers listed in Supplementary Table S3. The purified PCR-products were electroporated into an electrocompetent recombineering strain, mc<sup>2</sup> 155-pJV53-SacB, and plated on 7H10ADC with 25 µg/mL zeocin. Mutant genotypes of zeo<sup>r</sup> colonies were confirmed by PCR. The recombineering plasmid, pJV53-SacB, was rescued from mutants by plating them on 7H10ADC with 15% sucrose, and screening sucrose resistant colonies for kanamycin sensitivity. The zeo<sup>r</sup> marker was removed by excision using a Cre recombinase expressed from pCre-SacB, which was electroporated into the rescued mutant cells and transformants were screened for loss of zeo<sup>r</sup> . The zeo<sup>s</sup> colonies were screened on 7H10ADC with 15% sucrose to obtain clones

without the pCre-SacB plasmid. The rescued unmarked mutants were complemented as indicated.

#### RNA-seq

Mycobacterium smegmatis mc<sup>2</sup> 155 and 1glnR were grown in detergent-free Sauton's medium to form matured pellicle biofilms. Total RNA was extracted using a Qiagen RNeasy kit and contaminating DNA was removed with the turbo DNAfree kit (Thermo Fisher Scientific). For each sample, 5 µg of total RNA was processed for rRNA removal using the Ribo Zero kit (Illumina). Strand-specific DNA libraries were then prepared with 100 ng of mRNA using the Scriptseq Complete Kit- Bacteria (Illumina). Libraries were sequenced on the NextSeq500 platform (Illumina) and analyzed by Rockhopper (McClure et al., 2013) at default settings using the reference genome of M. smegmatis mc<sup>2</sup> 155 (NC\_008596).

### RT-qPCR

All oligonucleotides used for RT-qPCR are listed in Supplementary Table S3. DNA-free RNA for RT-qPCR was extracted as described for RNA-seq. For each sample, 200 ng of RNA was used for reverse transcription using the Maxima First Strand cDNA Synthesis Kit (Thermo Fisher Scientific). RT-qPCR was performed on an Applied Biosystems 7000 fast RT-qPCR System (Applied Biosystems) with SYBR Green Master Mix following the manufacturer's instructions. Relative expression of target gene is calculated either as 2 <sup>−</sup>1Ct(gene−SigA) or 2−1Ct(target gene1−SigA)−1Ct(target gene2−SigA) , in which SigA transcript was the endogenous control.

Antibiotics and Peroxide Sensitivity Assays

Exponential phase culture (OD<sup>600</sup> = 0.3) of each strain grown in Sauton's medium with 0.05% (v/v) Tween80 was harvested and washed once with phosphate buffered saline with 0.05% Tween-80 (PBST). Approximately 2 × 10<sup>7</sup> CFU/mL of each strain was resuspended in nitrogen-free (N0) Sauton's medium with 0.05% (v/v) Tween80 at 37◦C for 3 h. 400 µg/mL rifampicin, 1 µg/mL streptomycin or 20 mM H2O<sup>2</sup> were added and incubated at

37◦C for the indicated period; unexposed cultures were used as control. At the indicated time point, the exposed and unexposed cultures were diluted and plated on 7H10ADC for bacterial viability.

### Measurement of Intracellular NADPH/NADP Ratio

Average intracellular NADPH/NADP ratios of planktonic culture and biofilms were measured according to previous publication with minor modifications (Vilcheze et al., 2005). Briefly, exponential phase planktonic and biofilm cultures from normal and N<sup>0</sup> Sauton's medium were harvested and washed once with PBS and then resuspended in PBS. Large aggregates from biofilms were broken up by 8–10 repeated passaging through 18-G needles. 1.5 mL single cell suspensions at density of 108– 10<sup>9</sup> CFU/mL of each strain were pelleted and resuspended in 0.75 mL 0.2 M HCl (for NADP extraction) or 0.75 mL of 0.2 M NaOH (for NADPH extraction). After 10 min at 55◦C, the suspensions were cooled to 0◦C and neutralized by adding either 0.75 mL of 0.1 M NaOH for NADP extraction or 0.75 mL of 0.1 M HCl for NADPH extraction, while vortexing at high speed. After incubation for 10 min on ice, the suspensions were centrifuged at 3000 rpm for 10 min at 4◦C. The supernatants were filtered and transferred to a new tube and used immediately. The concentrations of NADP and NADPH in the suspensions were determined by spectrophotometric measurement of the rate of 3-[4,5-dimethylthiazol-2-yl]-2,5 diphenyltetrazolium bromide (Sigma # M2128) reduction by the glucose-6-phosphate dehydrogenase (Sigma # G6378) in the presence of phenazine ethosulfate (Sigma # P4544) at 570 nm. The rate of 3-[4,5-dimethylthiazol-2-yl]-2,5-diphenyltetrazolium bromide reduction is proportional to the concentration of the nucleotides. Purified NADP (Sigma # 10128031001) and NADPH (Sigma # 10107824001) were used for standard curves, which were used for determination of the nucleotides in each sample. Serial dilutions of samples were tested to ensure the values were in the linear range of the NADP/NADPH standard curve.

#### Peroxide Sensitivity Assay for Microfluidic Biofilms

Biofilms as microcolonies were grown in a CellASIC ONIX (Cat # EV262) microfluidic platform, using CellASIC microfluidic plates (M04S-03) with headspace of 150 µm. Approximately 10<sup>6</sup> CFU/mL of bacteria in 10 µL media were perfused at a pressure of 1.7 kPa (0.25 psi) into each culture chamber of the plate for 6 s, followed by no perfusion for 30 min. This allowed optimum attachment of single cells to the culture chamber surface at a density that then grew into separate colony biofilms. For microfluidic culturing, detergent-free Sauton's media was perfused across each culture chamber at the manufacturer's recommendation of a dual pressure of 3.4 kPa (0.5 psi) at 37◦C for 4 days to provide adequate nourishment and minimal stress to biofilm-like colonies. Incubation of 1glnR was extended by an additional day to allow colony biofilms to grow to the same size as wild type. Colony biofilms were exposed to 20 mM H2O<sup>2</sup> in Sauton's media via microfluidic perfusion for 3 h followed by overnight perfusion of Sauton's media with 1 µg/mL Calcein AM to stain survivors. Images of colony biofilms were collected by confocal laser scanning microscopy (CLSM) at 20× magnification under green channel (excitation 488 nm). Corresponding DIC image of each colony was also captured for overlaying the fluorescence signal. The images were analyzed by ImageJ. To compare the numbers of surviving cells among strains after H2O<sup>2</sup> exposure, the number of green CalceinAM-stained cells was calculated from the maximum intensity projection of the z-stacks of each biofilmlike colony, normalized to colony surface area. For each strain, at least three biofilm-like colonies over two fields of view were analyzed.

## RESULTS

### GlnR and Biofilm Formation in M. smegmatis

In our earlier study we identified 61 genes induced during maturation of M. smegmatis biofilms to be GlnR-dependent (Jenkins et al., 2013; Yang et al., 2017). We therefore tested the effect of glnR mutation on development of M. smegmatis biofilms. Deletion of glnR produced no apparent phenotype in modified M63 medium, which was used in our earlier study (Yang et al., 2017) (Supplementary Figure S1). However, the mutation caused delayed planktonic growth and retarded biofilm development in Sauton's medium (**Figures 1A,B**), which has poorer nitrogen source relative to the modified M63 medium. Lack of adequate nitrogen source in Sauton's medium appeared to be the primary cause for 1glnR phenotype, because the mutant growth could be substantially rescued by addition of casamino acid as supplemental nitrogen source (**Figure 1A**). Moreover, similar to the observation in modified M63 medium (Yang et al., 2017), late-stage (6-day) biofilms of wild-type (mc<sup>2</sup> 155) M. smegmatis in Sauton's medium also exhibited > 50 fold induction of a GlnR-dependent ammonium transporter (MSMEG\_2425; Amt1) (**Figure 1C**). The induction was not observed in planktonic culture of the strain (**Figure 1C**), indicating that nitrogen availability in the medium is sufficiently high to prevent peak level of GlnR activation in planktonic cells, but not high enough to do so in biofilms. Based on these findings, and to maintain consistent growth conditions used in previous studies of GlnR mutant of M. smegmatis (Jenkins et al., 2013), we chose to use Sauton's medium for this study.

Growth retardation of 1glnR in planktonic cultures (**Figure 1A**) suggests that a basal activity of the regulator is necessary for optimum uptake of nitrogen sources. Delayed but matured biofilm development by 1glnR raised the possibility of either an alternative mechanism of induction of GlnRdependent genes, or existence of alternative pathways for nitrogen assimilation. To investigate these possibilities we compared the transcriptomes of 6-day biofilms of wild-type and 1glnR strains. Expression of GlnR-dependent genes was significantly retarded in biofilms of 1glnR mutant, compared to the wild-type (Supplementary Table S4), suggesting that

secondary mechanisms of nitrogen assimilation are triggered in the mutant. This was further substantiated by upregulation in 1glnR mutant biofilms of acetamidase (amiE or MSMEG\_5335) and D-amino acid dehydrogenase (Supplementary Table S4).

#### GlnR Activation and Peroxide Resistance

A global effect of GlnR on gene expression patterns (Jenkins et al., 2013; Jessberger et al., 2013; Yang et al., 2017) raised a possibility that its activation during adaptation of M. smegmatis to low nitrogen may also impact other functions of cells. We investigated the effect of GlnR on persistence under stress exposure by comparing the sensitivity of nitrogen-starved cells of wild-type, 1glnR and 1glnR complemented strains to two commonly used anti-TB antibiotics: rifampicin (Rif) and streptomycin (Str) at 10X MIC. We also included hydrogen peroxide – a routinely used sterilizing agent for control of biofilm-related contaminants of surgical equipment in nosocomial settings (Falagas et al., 2011). We excluded isoniazid (INH) due to its selective activity on growing cells. We chose to test GlnRactivated planktonic cells by exposing them to N0-Sauton's medium. GlnR was activated within 3 h of exposure to N0- Sauton's medium (Supplementary Figure S2). Because viability of 1glnR remains unaltered during the first 48 h of exposure to the N0-Sauton's medium (**Figure 2A**), the exposure periods to the stressors were kept within this time limit. All three strains appeared equally sensitive to high concentrations of Rif and Str (**Figure 2B**). Interestingly, 1glnR mutant showed significantly greater sensitivity to peroxide exposure in a 60-min period (**Figure 2B**), and the phenotype was substantially reversed in the complemented strain (**Figure 2B**).

#### Peroxide Resistance Arises From GlnR-Dependent Induction of gpr

To determine the basis of GlnR-dependent peroxide resistance we analyzed the nucleotide sequence of GlnR-dependent genes. Upstream region of one of the GlnR-dependent operons comprising of 8 ORFs (MSMEG\_0565 to MSMEG\_0572) contained binding sites for both GlnR and SoxR (Dietrich et al., 2008; Jenkins et al., 2013) (**Figure 3A**). Since SoxR plays important role in redox homeostasis in many bacterial species (Storz and Imlay, 1999), we speculated that this locus could be under dual regulation of SoxR and GlnR, and that its activation by either of the two regulators possibly confers peroxide resistance. Biofilm-specific expression of all 8 ORFs in Sauton's medium was verified by RT-qPCR (**Figure 3B**). We next tested the role of SoxR and GlnR in activation of the operon using MSMEG\_0572 as a representative member. As expected from earlier studies (Jenkins et al., 2013; Jessberger et al., 2013; Yang et al., 2017), expression of the operon in wild-type cells was highly (>500 fold) induced upon 3-h exposure to N<sup>0</sup> Sauton's medium in a GlnRdependent manner (**Figure 3C**). However, SoxR activity appears to have very modest (∼2-fold) negative effect on the induction of the operon (**Figure 3D**). Interestingly, exposure to peroxide did not induce the operon (**Figure 3D**). Together, these expression profiles indicate that regulation of the operon exclusively depends on GlnR under the tested conditions.

We next asked if expression of MSMEG\_0565-72 operon is necessary and sufficient to exhibit GlnR-dependent peroxide resistance. A deletion mutant of the operon exhibited similar level of peroxide sensitivity as 1glnR under N0-Sauton's medium (**Figure 3E**), and the phenotype was rescued by plasmid-borne expression of the operon by a constitutive (Phsp60) promoter. Importantly, constitutive expression of the operon by Phsp<sup>60</sup> promoter was also able to substantially rescue peroxide sensitivity of 1glnR, indicating that peroxide resistance in M. smegmatis is primarily contributed by GlnRdependent activation of MSMEG\_0565-72 operon. We therefore call this operon as GlnR-dependent peroxide resistance or gpr.

To obtain a direct evidence for roles of glnR and gpr in peroxide resistance of M. smegmatis biofilms, we employed a microfluidic-based growth model to visualize surviving cells in peroxide exposed biofilms by confocal microscopy. To calibrate the growth model we first determined the timing of activation of GlnR by using a reporter strain of M. smegmatis, which

dinucleotide phosphate in biofilms of wild-type, 1gpr and the complemented strains cultured in normal Sauton's medium. (C) Ratio of NADPH to NADP calculated from (A,B). Data represent mean of three biologically independent experiments <sup>∗</sup> , ∗∗, and ∗∗∗ indicate p (t-test) < 0.05, 0.01, and 0.001, respectively.

harbored constitutively expressing mCherry and Dendra-2 fused to the promoter of Amt<sup>1</sup> (MSMEG\_2425). Expression of Dendra-2 in biofilms could be visualized after 4 days of growth in Sauton's medium (Supplementary Figure S3). Subsequent incubation led to bacterial growth in the flow channels, leading to increased backflow pressure. We therefore used 4-day stage of wild type biofilms for our analysis, although biofilms of 1glnR were cultured for an additional day to allow them to achieve similar size as wild-type. Following peroxide exposure, live cells in biofilms were probed by calcein AM, which remains non-fluorescent until its passive diffusion to the cytosol and hydrolysis by intracellular hydrolases produces fluorescent calcein (Rego et al., 2017). Compared to wild-type biofilms, the number of viable cells in peroxide exposed 1glnR biofilms was significantly reduced (**Figures 4A,B**). The mutant phenotype could be complemented by plasmid-borne expression of either glnR from its native promoter or a constitutive expression of gpr from the hsp60 promoter. We thus conclude that induced expression of gpr upon activation of GlnR during maturation of M. smegmatis biofilms directly contributes to their peroxide resistance.

### Physiological Role of gpr in Biofilm Development

Nitrogen assimilation in majority of bacterial species occurs at the expense of the redox currency, NADPH (van Heeswijk et al., 2013), which serves as a co-factor for several enzymes, including glutamate synthase, involved in synthesis of ammonia and amino acids. Two subunits of NADPH-dependent glutamate

Sauton's medium and modified Sauton's medium with half the normal nitrogen source (N1/2). In contrast to the wild type, which formed thick-textured biofilms in both medium, 1amt<sup>1</sup> and 1gpr mutants formed untextured biofilms in N1/<sup>2</sup> Sauton's medium, and the phenotype was partially rescued by nitrogen restoration in normal Sauton's medium. (B) Biomass of the pellicles described in (A) from three independent experiments. The order of columns in the plot corresponds to the order in which the strains are indicated in (A). Data represent mean of three biologically independent experiments. ∗∗ , ∗∗∗, and ∗∗∗∗ denote p < 0.01, 0.001, and 0.0001, respectively (t-test). (C) Planktonic growth of the indicated strains in N1/<sup>2</sup> Sauton's medium.

synthase, encoded by MSMEG\_3225 and MSMEG\_3226, are induced in biofilms by ∼30-fold (Yang et al., 2017), and their induction is GlnR-dependent (Supplementary Table S4). This suggests a greater demand of NADPH for nitrogen-starved cells in biofilms. This is consistent with ∼5-fold increase NADPH/NADP ratio after 3 h of incubation of wild-type cells in N0-Sauton's medium, relative to normal Sauton's medium (Supplementary Figure S4). The increase in the ratio appears to be contributed by a modest decrease (<2-fold) in NADPH, relative to NADP (∼7-fold), suggesting that NADPH is likely regenerated from existing NADP by reductases induced in nitrogen-starved cells. Regeneration of NADPH, as opposed to new synthesis, is preferred perhaps due to lack of nutrients in N0-Sauton's medium. We therefore hypothesized that induction of the putative reductases encoded by genes in gpr cluster likely regenerate NADPH pool to meet the metabolic demand of nitrogen-starved cells. Nitrogen-starved 1gpr mutant indeed produced a lower NADPH/NADP ratio than wild-type and complemented cells (Supplementary Figure S4). The decreased ratio in the mutant was due to reciprocal change in NADP and NADPH levels, consistent with the idea that existing NADP are reduced to regenerate NADPH (Supplementary Figure S4).

To test gpr-dependent redox homeostasis during biofilm development, we first compared the NADPH and NADP levels between planktonic and wild-type cells. Interestingly,

the average NADPH/NADP ratio in biofilm cells of wildtype increased by ∼18%, relative to its planktonic counterpart, indicating that biofilm cells have a more reduced cytosolic environment (**Figure 5C**). The increase in the ratio resulted from a disproportionate increase in NADPH (∼34%), compared to NADP, which was maintained at a steady level (**Figures 5A,B**). This suggests increase in both new synthesis of NADP and its reduction. This was in contrast to the scenario observed in planktonic cells in N0-Sauton's medium (Supplementary Figure S4). Expectedly, the average NADPH/NADP ratio in biofilms of 1gpr mutant declined by nearly 50%, relative to wild-type, and the phenotype was substantially restored in the complemented strain (**Figures 5A–C**). Decline in the ratio in the mutant was due to accumulation of NADP, with concomitant decrease in NADPH (**Figures 5A–C**). Together, we infer that GlnR-dependent induction of gpr in a subpopulation of biofilm cells that experience nitrogen starvation is critical for maintenance of higher NADPH pool to meet their metabolic demand.

A corollary to gpr-dependent redox homeostasis is that 1gpr mutant has impaired ability to assimilate nitrogen that impacts its growth in biofilms. Relative to wild-type, biofilm growth of 1gpr cells in normal Sauton's medium was moderately retarded, but the effect was more pronounced in nitrogen-depleted (N1/2) Sauton's medium, which has been shown to increase GlnRdependency of M. smegmatis in biofilms (Yang et al., 2017) (**Figures 6A,B**). A mutant lacking a GlnR-dependent operon (MSMEG\_2425-27), which was previously shown to display nitrogen-responsive biofilm defect (Yang et al., 2017), served as a reference in our analysis of 1gpr phenotype (**Figures 6A,B**). Phenotype of 1gpr could be complemented by plasmid-borne expression of gpr (**Figures 6A,B**). The defect was specific to biofilm culture, as no difference in planktonic growth of 1gpr mutant was observed (**Figure 6C**).

Our preliminary attempts to identify specific gene(s) of gpr that could complement 1gpr phenotype were unsuccessful (data not shown), suggesting that interaction between multiple genes of the cluster give rise to its function.

#### DISCUSSION

Mycobacteria express and utilize dedicated genes to build stress resistant biofilms (Richards and Ojha, 2014), raising a possibility that genetic programs involved in adaptation of resident cells within the architecture, and those in stress resistance overlap with each other. In this study, we provide evidence supporting this hypothesis by demonstrating a causal relationship between nitrogen-starvation response exhibited by biofilm cells and emergence of peroxide resistance in these cells. Peroxide resistance is likely a result of recalibration of NADPH/NADP ratio, skewed toward a more reduced cytosolic environment, to meet the increasing demand of NADPH for nitrogen assimilation in starved cells. These cells likely reside in interiors of biofilms, as suggested by localization of GlnRactivated cells in these regions of biofilms (Supplementary Figure S3). It is noteworthy that E. coli also acquire greater resistance to peroxide upon exposure to nitrogen starvation (Jenkins et al., 1988), although underlying mechanism remains unknown in this species.

The contribution of gpr cluster in maintenance of NADPH pool suggests specific function of encoded reductases in the process, although identity of these enzymes and their mechanisms remain open to further investigation. MSMEG\_0572 appears to represent DsrE-family reductases, which are conserved in a wide range of environmental species of bacteria. Moreover, similar to MSMEG\_0572, its orthologs in these species exhibit a syntenic arrangement with respect to the remaining seven genes of gpr cluster, suggesting that the entire cluster has migrated across genomes during evolution. This also raises the possibility that the cluster could function as a unit, rendering a possible explanation to our inability in identifying individual genes responsible for the phenotype associated with 1gpr strain. Interestingly, one of the genes in the cluster, MSMEG\_0567, is homologous to selenophosphate synthetase, which is directly involved in the synthesis of selenocysteine (Sec) (Turanov et al., 2011). Sec in prokaryotes is incorporated in polypeptide by a set of specialized accessory factors, which facilitate Sec-tRNA to decode the UGA codon (Krol, 2002). The UGA codon in mRNA decoded as Sec must have a downstream Sec insertion sequence (SECIS), which forms a unique stem-loop structure recognized by the accessory factors that recruit Sec-tRNA during translation (Krol, 2002). Bioinformatics search using bSECIS program (Zhang and Gladyshev, 2005) of all gpr ORFs identified a SECIS element comprising of 39 bp downstream of UGA codon of MSMEG\_0571, suggesting that the codon in this ORF is decoded as Sec, thereby giving rise to a selenoprotein. The resulting selenoprotein is extended by 518 amino acids (aa) to terminate at the originally annotated stop codon (UAG) of MSMEG\_0569 (Supplementary Figure S5). BLAST search of thus encoded 818aa long selenoprotein encompassing MSMEG\_0567 to MSMEG\_0569 reveals a conserved nitrilase-like domain in the N-terminal region and a flavin-dependent oxidoreductase domain in the C-terminal region (Supplementary Figure S5). This domain architecture is consistent with the fact that majority of the selenopoteins known so far are oxidoreductase (Hatfield et al., 2014), supporting the function of gpr cluster in redox homeostasis. The gpr cluster also includes a putative aliphatic amidase (MSMEG\_0566), which could directly contribute to nitrogen assimilation.

A homology search for gpr cluster in 29 mycobacterial species reveals its presence in only a few members, which are evolutionarily linked based on 16S rRNA phylogeny (**Figure 7**). Most of these species are classified as rapidly growing mycobacteria (RGM) (Runyon, 1959). Interestingly, gpr is absent in a clinically important RGM, M. abscessus, suggesting that its transfer to RGM from a common ancestor is a relatively recent evolutionary event. Moreover, evidence of horizontal gene transfer is supported by the presence of gpr in the only slowgrowing non-tuberculosis mycobacteria, M. simiae (**Figure 7**). Presence of gpr in M. mucogenicum, M. goodii, and M. simiae, which are emerging pathogens in nosocomial infections (van Ingen et al., 2008; Adekambi, 2009; Parikh and Grant, 2017; Salas and Klein, 2017), raises clinical significance of our findings and

possibly offers insight into resistance of these species to peroxidemediated sterilization of surgical and medical equipment.

Lack of gpr in a majority of mycobacteria suggests a different mechanism of GlnR-dependent nitrogen assimilation in these species, consistent with the differences between GlnR activities in M. smegmatis and M. tuberculosis as described previously (Williams et al., 2015). Further understanding of molecular underpinnings of these differences is likely to identify a role of GlnR in biofilm development and associated stress tolerance in these species.

#### AUTHOR CONTRIBUTIONS

AO and YY conceived this study. YY, JR, JG, and AO designed, performed, and analyzed the experiments. YY and AO wrote the manuscript.

#### REFERENCES


#### FUNDING

AO received funding from the NIH (AI107595).

#### ACKNOWLEDGMENTS

The authors acknowledged technical help from Advanced Light Microscopy and Image Analysis, Applied Genomic Technologies, and Media core laboratories of the Wadsworth Center.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2018.01428/full#supplementary-material


conditions for biofilm formation but not for planktonic growth. Mol. Microbiol. 66, 468–483. doi: 10.1111/j.1365-2958.2007.05935.x


**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 Yang, Richards, Gundrum and Ojha. 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.

# Challenges of NTM Drug Development

#### Joseph O. Falkinham III\*

Department of Biological Sciences, Virginia Tech, Blacksburg, VA, United States

Discovery and development of antibiotics active against the environmental opportunistic non-tuberculous mycobacteria (NTM) have been retarded by innate antibiotic-resistance of NTM cells and methodological challenges in the laboratory. The basis for the innate resistance of NTM cells is its lipid rich outer membrane that results in hydrophobic cells and the outer membrane's impermeability, and the residence of NTM cells in phagocytic cells, and the slow growth and dormancy of NTM. Laboratory challenges include: the choice of species and strains for screening and measurement of anti-NTM activity, the high frequency colony switching between antibiotic-susceptible and resistant variants, the preference of NTM to adhere to surfaces and form biofilms, and the aggregation of NTM cells. Understanding these challenges can guide and inform our approaches to discovery and development of antibiotics with activity against NTM.

#### Edited by:

Thomas Dick, Rutgers, The State University of New Jersey, United States

#### Reviewed by:

David E. Griffith, The University of Texas Health Science Center at Tyler, United States César de la Fuente, Massachusetts Institute of Technology, United States

> \*Correspondence: Joseph O. Falkinham III jofiii@vt.edu

#### Specialty section:

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

Received: 11 April 2018 Accepted: 28 June 2018 Published: 18 July 2018

#### Citation:

Falkinham JO III (2018) Challenges of NTM Drug Development. Front. Microbiol. 9:1613. doi: 10.3389/fmicb.2018.01613 Keywords: slow growth, rapid metabolism, adaptation, impermeable, hydrophobic, dormancy, biofilms, intracellular

### INTRODUCTION TO THE NON-TUBERCULOUS MYCOBACTERIA (NTM)

The term non-tuberculous mycobacteria (NTM) encompass a large number (>150) Mycobacterium species that are environmental opportunistic pathogens causing pulmonary disease and skin infections in adults and cervical lymphadenitis in children (Marras and Daley, 2002). There is growing awareness of the emergence of NTM as causative agents of nosocomial infections (Marras and Daley, 2002). In the United States the species most commonly reported as causing disease are Mycobacterium avium subsp. hominissuis, Mycobacterium intracellulare, Mycobacterium chimaera, Mycobacterium kansasii, and Mycobacterium abscessus complex. Proven sources of infection include NTM from drinking water, potting soils, and medical equipment. Routes of infection include aerosols from water, dusts from soil, and water and soil contact with abraded or injured skin (Falkinham, 2015).

Humans are surrounded by NTM. NTM have been shown to be normal residents of drinking water distribution systems and premise plumbing; in fact, they grow in distribution systems between the treatment plant and buildings (Falkinham et al., 2001). Premise plumbing is an ideal habitat for the NTM. The water is heated and the low concentration of organic carbon is sufficient to support NTM growth. Although NTM can be grown on rich laboratory medium, they are oligotrophs able to grow on the small amount of organic matter in drinking water (George et al., 1980; Norton et al., 2004). Periods of water stagnation brought about by increased water age and oxygen consumption do not limit NTM growth, as they can grow at 12% oxygen as well as in air (21% oxygen) and can even grow, albeit at half the rate, at 6% oxygen (Lewis and Falkinham, 2015). Finally, NTM cells are surrounded by a lipid-rich outer membrane (Brennan and Nikaido, 1995) that makes NTM cells hydrophobic (van Oss et al., 1975) and impermeable

(Jarlier and Nikaido, 1994). Hydrophobicity leads to the preferential attachment of NTM cells to the walls of any container or pipe, so the slowly growing NTM cells cannot be washed out. Surveys of household plumbing of M. avium patients showed that in 50% of the households the NTM isolates from the houses were of the same species and shared the same rep-PCR DNA fingerprint patterns (Falkinham, 2011). It is important to note that later work has shown that VNTR-RFLP is more discriminatory than rep-PCR (Iakhiaeva et al., 2016). Further, M. avium was common in plumbing of homes with a water heater temperatures of 125◦ F (50◦ C) or less, but rare if the water heater temperature was 130◦ F (55◦ C) or higher. That has led to a current study of the role of water heaters in increasing numbers of M. avium in premise plumbing. All the factors listed above undoubtedly contributed to the colonization and growth of M. chimaera in heater-coolers that have been shown to generate M. chimaera-laden aerosols and infected patients undergoing cardiac surgery (Sax et al., 2015).

### CHALLENGES TO DEVELOPMENT OF ANTI-NTM ANTIBIOTICS

#### Introduction

The foregoing brief introduction to the NTM is necessary before discussion of how NTM physiologic and genetic features of NTM are challenges to antibiotic development. The challenges fall into two categories: those involving difficulties in overcoming innate genetic and physiologic traits of NTM (**Table 1**) and difficulties in measuring anti-NTM antibiotic activity in the laboratory (**Table 2**). The approach that follows focuses first on laboratory challenges of susceptibility measurements and then turns to challenges due to innate NTM resistance to antibiotics. In a number of instances, it will be clear that a single NTM feature drives antibiotic resistance and laboratory measurement problems. Finally, a table is included to suggest possible solutions to the challenges (**Table 3**).

## Laboratory Challenges

#### NTM Strain Selection

When searching among libraries of synthetic or nature compounds for anti-mycobacterial activity the choice of Mycobacterium spp. strains is of paramount importance. First, do not use strains of Mycobacterium smegmatis as indicators of potential anti-NTM activity by novel compounds in spite of their rapid rate of growth and ease of handling. Current

TABLE 1 | Non-tuberculous mycobacteria (NTM) antibiotic development challenges for NTM due to laboratory measurement of antibiotic susceptibility.

Mycobacterium species strain selection

Colonial variation

Preference for surface adherence and biofilm formation Cell aggregation Dormancy Residence in phagocytic cells

TABLE 2 | Non-tuberculous mycobacteria antibiotic development challenges due to innate traits of NTM.


TABLE 3 | Possible solutions to challenges impeding antibiotic development for NTM.


isolates of M. smegmatis have been linearly cultivated for so long, that they do not resemble the original ATCC deposit. Most of the M. smegmatis isolates being used are derivatives of ATCC strain 607, and those lab-adapted strains have lost many unique characteristics – slow growth, hydrophobicity, and impermeability – that directly increase antibiotic susceptibility. Thus, measurements of minimal inhibitory concentrations (MIC) are not useful to guide compound selection as the M. smegmatis strains in use are unusually susceptible to a wide range of antibiotics. Prolonged, linear cultivation of NTM isolates selects for variants that can grow faster, have lost hydrophobicity and impermeability. Even employing M. smegmatis as a primary screen can be misleading, as some compounds that do not display anti-M. smegmatis inhibition has activity against M. avium, M. intracellulare, or M. abscessus. It is best to identify the ultimate target NTM species (e.g., M. avium) and measure MICs against representatives of that species (**Table 3**).

Another problem concerns selection of isolates from patients for measurement of MIC to guide and select antibiotics for therapy. More than a single isolate needs to be selected and its susceptibility measured (CSLI, 2011), as it has been shown that colonies of M. avium from AIDS patients having identical morphology, yielded different patterns of susceptibility to antibiotics (von Reyn et al., 1995). Further, as NTM-infected patients may be infected by more than a single NTM species as detected directly (Wallace et al., 1998) or evidenced by relapse (Boyle et al., 2016), selection of single colonies may lead to an incorrect assessment of the antibiotic of choice.

#### NTM Colonial Variation

fmicb-09-01613 July 16, 2018 Time: 16:26 # 3

Another problem associated with strain choice is that NTM strains can display alternative colonial variants. For M. avium and M. intracellulare, isolates exhibit either a transparent or opaque colony form (McCarthy, 1970; Stormer and Falkinham, 1989). Likewise, M. abscessus complex isolates can exhibit either a rough and smooth colonial morphology (Howard et al., 2006). In M. avium and M. intracellulare the transparent (T) "Mexican Hat" colony morphology is found upon primary culture from patients. The transparent (T) variant is slow-growing, hydrophobic, virulent, and antibiotic-resistant compared to the opaque (O) variant that emerges during laboratory medium cultivation (Schaefer et al., 1970). The opaque (O) variants are faster growing, less hydrophobic, and antibiotic-sensitive compared to the isogenic transparent (T) variants. In M. avium, single colony isolates switch between the T and O forms; colony variation is fully reversible (McCarthy, 1970). The frequency of colony variants of the opposite type amongst isolates of M. avium is approximately 1 in 1,000 (Stormer and Falkinham, 1989). Thus, in a barely turbid culture of one-hundred million (10<sup>8</sup> ) M. avium (T) variant cells, there are approximately one-hundred thousand (10<sup>5</sup> O) cells. Thus, special care must be taken to sample every culture to ensure that the frequency of colonies of the opaque (O) type are always less than or equal to 1 in 1,000 (**Table 3**).

As the alternative colonial variants have different susceptibilities to antibiotics (McCarthy, 1970; Stormer and Falkinham, 1989), measurement of MIC values may not provide the guidance to direct either chemotherapy or selection of promising novel anti-mycobacterial compounds. For example, as transparent (T) colony variants predominate in patients, selection of an isogenic opaque (O) variant for MIC measurement will lead to choice of an antibiotic that may lack activity against the patient's transparent (T) variant. This may be the basis, in part, for the lack of any correlation between MIC measurements and prediction of therapeutic success (Sison et al., 1996; Van Engen et al., 2010; Schön and Chryssanthou, 2017). It might also result in the advancement of a compound for further testing based on MIC values from an opaque (O) variant, that will be ineffective in patients as the transparent variant (T) is resistant to the compound.

#### NTM Hydrophobicity and Measurement of Antibiotic Susceptibility

Non-tuberculous mycobacteria hydrophobicity also creates problems in measuring MICs to guide compound selection or patient therapy. As NTM are hydrophobic, they prefer surface attachment and growth rather than replicating in aqueous suspension. That means that the NTM cells are not in suspension, but rather adhere to the walls of the individual wells in 96-well plates used to measure MICs as recommended by Clinical and Laboratory Standards Institute (CSLI, 2011). Measurements of NTM cells adhering to the walls of water heaters and pipes in households and hospitals, in medical equipment (e.g., heatercoolers), and in test tubes in the laboratory shows that the majority of NTM cells (>99%) are surface-attached (Falkinham et al., 2001; Falkinham, 2011). This observation alone provides a reason why measurement of antibiotic susceptibility in tubes or 96-well plates has not provided guidance for therapeutic efficacy in patients (Sison et al., 1996; Van Engen et al., 2010; Schön and Chryssanthou, 2017). Simply, measurement of turbidity of cell suspensions does not accurately reflect the number of total cells; most are not in suspension but adhering to the wells or tubes. The solution is to use methods to measure MIC of biofilm-grown cells in biofilms (**Table 3**).

Non-tuberculous mycobacteria preference for adherence and biofilm-formation (Falkinham et al., 2001; Falkinham, 2011) also explain why disinfection of hospital pipes and medical equipment generally fails. The NTM cells in the biofilms survive disinfection and can re-inoculate the water.

A second hydrophobicity-driven problem reducing the utility of standard methods of MIC measurement (CSLI, 2011) is the spontaneous aggregation of NTM cells. Typically, early growth of NTM cells in laboratory medium (e.g., M7H9 broth) can be followed by increases in turbidity, but upon reaching the mid to late exponential phase of growth turbidity disappears and visible aggregates of various dimensions appear. Thus, for many NTM isolates, antibiotic susceptibility cannot be measured with accuracy. Aggregation between NTM cells is driven by hydrophobicity, just as hydrophobicity drives the adherence of NTM to the walls of tubes and wells (van Oss et al., 1975). The best solution to this problem is to measure MIC of isolates during the early phases of cell growth, before aggregation occurs (**Table 3**). This may require amplification, such as PCR or dyes, to increase the signal from the low cell numbers.

There are no other current useful approaches to reduce the aggregation of NTM cells, as the use of detergent reduces cell-surface hydrophobicity and abolishes aggregation, but increases NTM permeability and growth rate, and thereby the susceptibility to antibiotics. Selection of non-aggregating variants of NTM isolates is not a solution, as the non-aggregating variants have reduced cell hydrophobicity and altered outer membrane composition. Thus, the cells are not representative of those recovered from the patient. This is represented by the laboratory-culture-induced selection of opaque (O) variants of the transparent (T) variants that are isolated from patients.

#### NTM Adherence and Biofilm Formation

Non-tuberculous mycobacteria cells prefer surface adherence to residence in suspension, whether in humans, in water, or in laboratory growth medium. Surface attachment allows NTM cells to reduce the interaction of their hydrophobic surfaces with the positive and negative charges in aqueous suspension. Once the NTM cells adhere, they can grow to form a biofilm that consists of NTM and other microbial cells within a polymeric matrix consisting of polysaccharide, lipid, DNA, and protein (Mullis and Falkinham, 2013). The layers of cells within the matrix are relatively impermeable to antibiotics and disinfectants. There are significant differences in the susceptibility to chlorine (Falkinham, 2003; Steed and Falkinham, 2006) or antibiotics (Falkinham, 2007) of M. avium cells grown and exposed in biofilms compared to suspension-grown and exposed cells. The increased resistance of M. avium cells in biofilms to antimicrobial agents, compared to those grown and exposed in suspension, prevents obtaining an accurate measurement of disinfectant-

or antibiotic-susceptibility for guidance in water treatment or patient care. Again, the solution to biofilm formation is to measure MICs of NTM cells grown in biofilms (**Table 3**).

That is not the only problem of NTM biofilm formation. Surprisingly, cells of M. avium grown in biofilms, yet harvested and exposed as single cell suspensions were significantly more tolerant of chlorine or antibiotics than cells grown and exposed in suspension (Steed and Falkinham, 2006; Falkinham, 2007). That tolerance was transient, as cells lost the increased resistance to chlorine or antibiotics exhibited by suspensiongrown cells following overnight growth in fresh medium (Steed and Falkinham, 2006; Falkinham, 2007). I speculate that such a physiologic adaptation (adaptation, because cells are not permanently altered in susceptibility) is a consequence of growth of NTM-cells in a biofilm.

#### NTM Dormancy

NTM-dormancy has been demonstrated in a variety of NTM species, particularly in M. avium (Archuleta et al., 2005). M. avium cells subjected to nutrient starvation entered a dormant stage in which cells are effectively non-growing and thereby resistant to any antibiotic challenge (Archuleta et al., 2005). Whether such NTM-dormant cells are in patient tissue or in the wells of an antibiotic-containing medium, such non-growing cells would be expected to be resistant to most antibiotics, as the antibiotics do not actively destroy cell components.

As demonstrated by the studies of Larry Wayne, the slow decrease in oxygen levels as M. tuberculosis triggers the formation of lung tubercles, points out that dormancy of M. tuberculosis is triggered by the absence of oxygen. Those dormant cells can, under certain conditions, be triggered to grow again. The discovery of proteins that can induce growth in such dormant cells, suggests an approach to measurement of MIC of cells suspected to be dormant; namely resuscitate the dormant cells (Wivagg and Hung, 2012) and measure the MIC following the subsequent growth (**Table 3**).

#### NTM Residence in Phagocytic Cells

Non-tuberculous mycobacteria cells are not only found within phagocytic cells in infected individuals, but also in granulomas in infected organs like the lungs and liver. NTM-cells in granulomas have not only the NTM outer membrane to protect them, but they also have the layers of host, human cells that make up the granuloma. When considering that it is likely that the majority of NTM-cells are located within mammalian phagocytic cells in infected individuals, a problem of permeability re-appears. Here, it is not the NTM cell's impermeability that is the only factor, but rather the impermeability of the phagocytic cell's membrane now plays a role. Recognition of that fact led David Yajko and his colleagues to measure antibiotic susceptibility of NTM cells in human macrophages (Yajko et al., 1991). However, as discussed above, impermeability can be possibly overcome using combinations of agents; namely, one agent to increase the permeability of the phagocytic cell's membrane to achieve bacteriostatic or bactericidal concentrations within those cells and a second for killing or inhibiting the growth of the NTM cells (**Table 3**). Documentation that the MICs of the intracellular NTM cells were higher than those of the NTM cells alone, forces us to consider this approach.

#### Innate Resistance of NTM to Antibiotics NTM Hydrophobicity and Impermeability and NTM Susceptibility to Antibiotics

Both the hydrophobic and impermeable NTM outer membrane are major and independent contributors to the lack of susceptibility of NTM to commonly used antibiotics (Rastogi et al., 1981; Jarlier and Nikaido, 1994; Brennan and Nikaido, 1995). NTM are the most hydrophobic of bacteria (van Oss et al., 1975), due to the fact that cells are surrounded by the lipid-rich outer membrane that comprises 30% of the cell mass (Brennan and Nikaido, 1995). Hydrophobicity, measured as water droplet contacts angles on filters covered with NTM cells are high (≥75◦ ) almost resembling water droplets on a freshly waxed automobile. The non-polar cell surface prevents the adherence or binding of antibiotics that carry positive or negative charges.

The lipid-rich outer membrane is also impermeable. Measurement of rates of uptake of compounds showed the uptake of charged antibiotics, nutrients, metals, oxyanions, and disinfectants was only 1% of the rate measured in rapidly growing bacteria such as Escherichia coli (Jarlier and Nikaido, 1994). Thus, not only is adherence of charged compounds reduced, but the rate of transport though the outer membrane is greatly reduced. Therefore, it is not surprising that NTM exhibit resistance to most commonly used antibiotics. That having been stated, it is important to point out that the targets for antibiotic action, for example the 23S rRNA for the erythromycin-family antibiotic clarithromycin, are intact and resemble those in other bacteria. That can be shown by exposing NTM cells to antibiotics or other anti-microbial compounds in the presence of detergent. Detergent disrupts the outer membrane, reducing hydrophobicity and increasing permeation, artificially creating phenocopy cells that are antibiotic-susceptible.

One successful route to development of anti-NTM antibiotics is to synthesize hydrophobic derivatives of existing antibiotics. That approach was taught in a paper describing the synthesis of effective anti-NTM drugs through the addition of hydrophobic groups, starting with an ineffective and hydrophilic drug (Rastogi et al., 1988). Although those drugs are not in use, there is the example of synthesizing hydrophobic derivatives of erythromycin, a common antibiotic with little activity against NTM. Hydrophobic derivatives of erythromycin, namely clarithromycin and azithromycin have been shown effective in treating NTM-infected patients whether suffering from NTN pulmonary disease or NTM bacteremia. Following that rational for anti-NTM antibiotic development, a family of Amphipathic dendritic amphiphiles were synthesized and their anti-NTM activity measured (**Table 3**). These compounds consist of a fatty acid tail linked to (Williams et al., 2007; Falkinham et al., 2012). A number of these compounds displayed strong antibiotic activity against M. avium, M. intracellulare, and M. abscessus with MICs all below 10 µg/mL (Williams et al., 2007; Falkinham et al., 2012).

Another successful approach to overcoming the hydrophobic barrier of the NTM outer membrane is through combinations

of antibiotics (**Table 3**). For example, the antibiotic ethambutol is an inhibitor of formation of the arabinogalactan polymer that links the NTM outer membrane to the peptidoglycan cell wall. Although ethambutol is not a first line drug, it acts synergistically with other antibiotics with targets within cells. Ethambutol reduces the hydrophobic and impermeable outer membrane barrier resulting in increased transport of the second drug (Yajko et al., 1988). That demonstrated example of rational synergism, suggests that the search for anti-NTM antibiotics includes screening methods to identify those that target and disrupt the outer membrane and then measure MICs of those compounds in combination with drugs whose targets are intracellular. Adoption of this view strongly suggests that screening natural and synthetic compounds for anti-NTM activity include all possible combinations and screening in combination with known, albeit low, anti-NTM active antibiotics such as ethambutol.

#### NTM Slow Growth and Adaptation

The unique structural feature of the mycobacteria; namely the presence of a long chain lipid and wax-rich outer membrane, is the major determinant of their physiology, growth, ecology, and epidemiology (Brennan and Nikaido, 1995). As the outer membrane makes up 30% of the total mycobacterial cell weight, its synthesis reduces energy available for production of new cells. Consequently, NTM grow slowly with a generation time of 1 doubling per day under nutrient-rich conditions at 37◦C. However, it is important to point out that the NTM do not replicate slowly because of a slow rate of metabolism. NTM metabolism, as reflected by oxygen uptake and ATP production, is as active as that of rapidly growing bacteria; the NTM simply expend a great deal of energy making the outer membrane. That rapid rate of metabolism means that NTM cells can induce gene products that protect against environmental stress before cells divide.

In M. avium, growth rate modulates antibiotic susceptibility. The susceptibility of M. avium cells to antibiotics or chlorine is greater in medium that supports more rapid growth compared to a minimal, nutrient-poor medium (Falkinham, 2003). In measuring antibiotic MICs of NTM isolates, it is valuable to examine and identify MICs on a daily basis (**Table 3**). For example, inspection of MICs for the antibiotic clarithromycin against strains of Mycobacterium abscessus identified some whose MIC values increased over time. Significantly, the MIC measured after 3 days incubation was significantly lower than that measured at 7 days (Nash et al., 2009). Further examination of that behavior led to the discovery of an erythromycin methylase that protected cells from inhibition of protein synthesis due to its induction and synthesis leading to methylation of the 23S rRNA target of clarithromycin (Nash et al., 2009).

The studies of Maaløe and Kjeldgaard (1966) taught microbiologists the concept of "balanced" and "unbalanced" growth and the relationship of growth rate to the efficacy of antibiotics. Growth of microorganisms in the absence of antibiotics leads to "balanced" conditions where the increases in the amount of DNA, RNA, protein, and cell walls are proportionate with the growth rate. Exposure of cells growing under "balanced" conditions to an antibiotic unbalances growth. For example, as antibiotics have a single target (e.g., DNA polymerase), it follows that exposure to an antibiotic that inhibits the activity of DNA polymerase would lead to an "unbalanced" condition where increases in of DNA, are not proportionate with increases in cell mass. "Unbalanced" growth of microbial cells leads to death, particularly in rapidly dividing cells (e.g., E. coli). However, unlike E. coli, NTM cells can react to antibiotic stress by synthesizing proteins and other cell constituents able to protect the cells before they are forced to divide (24 h/generation).

To explain NTM-adaptation further, what follows are the results of a study of M. avium heat-adaptation. My students and I have found heat-tolerance in M. avium isolates recovered from plumbing biofilm samples collected from homes of M. aviuminfected patients. The isolates are able to survive exposure to 60◦C for 3 h without any loss in colony-forming units; they are only killed at 65◦C. This data is in contrast to published data on heat-susceptibility of NTM species (Schulze-Röbbecke and Buchholtz, 1992). As the patient, plumbing, and waterprovider M. avium isolates shared an identical VNTR-RFLP DNA fingerprint patterns (Iakhiaeva et al., 2016) and heat-tolerance was an adaptation, we hypothesized that passage through the home's water heater may have triggered an adaptive response leading to increased heat-tolerance. Accordingly, we grew the household M. avium strains at 25◦ , 30◦ , 35◦ , and 42◦ (the latter the highest temperature of growth for M. avium) and measured their susceptibilities to 55◦ , 60◦ , and 65◦C. Further, we measured the cell concentration of trehalose that has been shown to be a modulator of temperature-susceptibility. Compared to cells grown at 25◦C, cells grown at 42◦C had 10-times more trehalose/cell protein. Further, cells grown at 42◦C were significantly more tolerant to survival at 60◦C compared to those grown at 25◦C.

#### CONCLUDING COMMENTS

The foregoing information provides strong support for the notion that the search for novel anti-NTM antibiotics will continue to be difficult. It is already well-established that laboratory measures of antibiotic MICs of suspension-grown NTM cells do not provide guidance for therapy (Sison et al., 1996; Van Engen et al., 2010; Schön and Chryssanthou, 2017). The challenges involve both methods of measuring antibiotic susceptibility of NTM strains in the laboratory (**Table 1**) as well as developing novel anti-NTM antibiotics that can overcome innate characters of these waterborne opportunistic pathogens (**Table 2**). I have suggested several solutions to these problems (**Table 3**), but further approaches are needed.

First, identification of suitable novel anti-NTM antibiotics requires development of accurate laboratory measures of NTM antibiotic susceptibility. That requires choice of species and strains coupled with the development of accurate measures of suspension-, biofilm-, and intracellular- (e.g., macrophage) NTM susceptibility (**Table 1**). Those in vitro measures must be supplemented with ways to prevent NTM cell aggregation, without altering NTM susceptibility.

The second challenge facing development of anti-NTM antibiotics involves efforts to overcome NTM cell surface hydrophobicity and outer membrane impermeability (**Table 2**). A starting point might be the realization that successful anti-NTM drug therapy involves synergistic combinations; one antibiotic to disrupt or inhibit the outer membrane, and a second antibiotic to inhibit a critical cellular process (i.e., DNA, RNA, or protein and outer membrane synthesis) and hence

#### REFERENCES


unbalance growth. For such an approach, high throughput screening involving pairs or compounds seems the likely tool.

#### AUTHOR CONTRIBUTIONS

JF researched, practiced, and wrote the article.


in antimicrobial susceptibilities of different strains of M. avium isolated from the same patient. J. Clin. Microbiol. 33, 1008–1010.


virus infection. Antimicrob. Agents Chemother. 35, 1621–1625. doi: 10.1128/ AAC.35.8.1621

Yajko, D. M., Kirihara, J., Sanders, C., Nassos, P., and Hadley, W. K. (1988). Antimicrobial synergism against Mycobacterium avium complex strains Isolated from patients with acquired immune deficiency syndrome. Antimicrob. Agents Chemother. 32, 1392–1395. doi: 10.1128/AAC.32.9.1392

**Conflict of Interest Statement:** The author declares 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 Falkinham. 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.

# Amikacin Liposome Inhalation Suspension (ALIS) Penetrates Non-tuberculous Mycobacterial Biofilms and Enhances Amikacin Uptake Into Macrophages

Jimin Zhang<sup>1</sup> \* † , Franziska Leifer<sup>1</sup>† , Sasha Rose1,2, Dung Yu Chun<sup>1</sup> , Jill Thaisz<sup>1</sup> , Tracey Herr<sup>1</sup> , Mary Nashed<sup>1</sup> , Jayanthi Joseph<sup>2</sup> , Walter R. Perkins<sup>1</sup> and Keith DiPetrillo<sup>1</sup>

1 Insmed Incorporated, Bridgewater, NJ, United States, <sup>2</sup> Department of Biomedical Sciences, College of Veterinary Medicine, Oregon State University, Corvallis, OR, United States

#### Edited by:

Thomas Dick, Rutgers, The State University of New Jersey, Newark, United States

#### Reviewed by:

Murugesan V. S. Rajaram, The Ohio State University, United States Nagendran Tharmalingam, Alpert Medical School, Brown University, United States

#### \*Correspondence:

Jimin Zhang jimin.zhang@insmed.com

†These authors have contributed equally to this work.

#### Specialty section:

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

Received: 28 February 2018 Accepted: 20 April 2018 Published: 16 May 2018

#### Citation:

Zhang J, Leifer F, Rose S, Chun DY, Thaisz J, Herr T, Nashed M, Joseph J, Perkins WR and DiPetrillo K (2018) Amikacin Liposome Inhalation Suspension (ALIS) Penetrates Non-tuberculous Mycobacterial Biofilms and Enhances Amikacin Uptake Into Macrophages. Front. Microbiol. 9:915. doi: 10.3389/fmicb.2018.00915 Non-tuberculous mycobacteria (NTM) cause pulmonary infections in patients with structural lung damage, impaired immunity, or other risk factors. Delivering antibiotics to the sites of these infections is a major hurdle of therapy because pulmonary NTM infections can persist in biofilms or as intracellular infections within macrophages. Inhaled treatments can improve antibiotic delivery into the lungs, but efficient nebulization delivery, distribution throughout the lungs, and penetration into biofilms and macrophages are considerable challenges for this approach. Therefore, we developed amikacin liposome inhalation suspension (ALIS) to overcome these challenges. Nebulization of ALIS has been shown to provide particles within the respirable size range that distribute to both central and peripheral lung compartments in humans. The in vitro and in vivo efficacy of ALIS against NTM has been demonstrated previously. The key mechanistic questions are whether ALIS penetrates NTM biofilms and enhances amikacin uptake into macrophages. We found that ALIS effectively penetrated throughout NTM biofilms and concentration-dependently reduced the number of viable mycobacteria. Additionally, we found that ALIS improved amikacin uptake by ∼4 fold into cultured macrophages compared with free amikacin. In rats, inhaled ALIS increased amikacin concentrations in pulmonary macrophages by 5- to 8-fold at 2, 6, and 24 h post-dose and retained more amikacin at 24 h in airways and lung tissue relative to inhaled free amikacin. Compared to intravenous free amikacin, a standardof-care therapy for refractory and severe NTM lung disease, ALIS increased the mean area under the concentration-time curve in lung tissue, airways, and macrophages by 42-, 69-, and 274-fold. These data demonstrate that ALIS effectively penetrates NTM biofilms, enhances amikacin uptake into macrophages, both in vitro and in vivo, and retains amikacin within airways and lung tissue. An ongoing Phase III trial, adding ALIS to guideline based therapy, met its primary endpoint of culture conversion by month 6. ALIS represents a promising new treatment approach for patients with refractory NTM lung disease.

Keywords: non-tuberculous mycobacteria, amikacin, biofilm, macrophage uptake, ALIS, LAI, Arikayce, Arikace

## INTRODUCTION

fmicb-09-00915 May 14, 2018 Time: 15:46 # 2

Non-tuberculous mycobacteria (NTM) are commonly found throughout the environment and cause pulmonary infections in patients with structural lung damage, impaired immunity, or other risk factors (Johnson and Odell, 2014; van Ingen and Kuijper, 2014). Lung disease is the most common manifestation of NTM infections (NTM lung disease; NTM-LD) and patients typically present with either nodular bronchiectatic disease or fibrocavitary disease (Johnson and Odell, 2014). Various mycobacterial species can cause NTM-LD, but species from the Mycobacterium avium complex (MAC; including avium, intracellulare, chimaera, and others) are the most common (Johnson and Odell, 2014).

Non-tuberculous mycobacteria can persist extracellularly in biofilms or intracellularly within macrophages and other cells in infected hosts. NTM biofilms are evident in both expectorated sputum samples and in alveolar walls of explanted lungs from infected patients (Qvist et al., 2015). Several lines of evidence demonstrate that MAC species can also live as intracellular infections within monocytes and macrophages (Appelberg, 2006; Awuh and Flo, 2017). In vitro, several MAC strains grew well (9- to 43-fold increases) inside isolated human peripheral blood monocytes cultured with human serum, but failed to grow extracellularly in the presence of human serum (Nozawa et al., 1984). In mice systemically infected with M. avium, the mycobacteria proliferated in spleen, liver, and lung tissues over the course of several months (Frehel et al., 1991). Electron microscopy localized the mycobacteria exclusively within cells, with no extracellular mycobacteria detected in thin tissue sections of livers and spleens (Frehel et al., 1991). Clinical case reports have also described intracellular MAC infections in humans, with M. avium detected inside peripheral blood leukocytes and bone marrow aspirate (Graham et al., 1984; Moffie et al., 1989). These findings indicate that MAC species can live and replicate within macrophages. Therefore, delivering effective antibiotic concentrations into NTM biofilms and cells infected with NTM are important components of treatment.

Mycobacterium avium complex species are generally sensitive in vitro to aminoglycoside antibiotics, such as amikacin (Brown-Elliott et al., 2013; Cowman et al., 2016), that exhibit concentration-dependent bactericidal effects. However, amikacin and other aminoglycoside antibiotics accumulate poorly in cells, which can limit their effectiveness against intracellular infections (Kesavalu et al., 1990). One way to increase intracellular amikacin delivery is to package the antibiotic into liposomes, which are nanometer sized vesicles composed of a phospholipid bilayer membrane surrounding an aqueous interior compartment (Gregoriadis, 1976a). The physical characteristics of liposomes (small size, low toxicity, tissue/cell targeting) and their capability for delayed or triggered release of cargo molecules make them highly useful drug carriers (Gregoriadis, 1976b). In fact, liposome encapsulation has been shown to improve the ability of amikacin to kill intracellular M. avium infections in macrophages (Kesavalu et al., 1990). Moreover, intravenous injection of liposome-encapsulated amikacin improved treatment of systemic M. avium infections compared to free amikacin (Cynamon et al., 1989; Ehlers et al., 1996; Petersen et al., 1996), reducing mycobacterial counts in liver and spleen at doses 100-fold lower than free amikacin (Ehlers et al., 1996).

Despite the effectiveness of liposome-encapsulated amikacin against disseminated M. avium infections, intravenous administration of liposomal amikacin failed to provide longterm benefit against pulmonary infections in mice (Ehlers et al., 1996). NTM-LD is particularly difficult to treat because it requires delivery of high amounts of antibiotics to the lung while keeping systemic concentrations low to avoid toxicities (Olivier et al., 2014). Inhalation delivery of liposomal amikacin directly into the lungs may address this problem, but this approach faces three major delivery hurdles: (1) efficient nebulization delivery of liposomes to the lungs; (2) distribution of the liposomes throughout the lungs; and, (3) penetration into biofilms and macrophages to reach the sites of infection. Therefore, we developed amikacin liposome inhalation suspension (ALIS), also referred to in previous publications as Arikayce, Arikace or liposomal amikacin for inhalation (LAI), to overcome these challenges and improve the treatment of NTM-LD.

To provide efficient nebulization delivery to the lungs, ALIS was designed to have a high drug to lipid ratio and to retain a consistent fraction of amikacin within the liposome during the nebulization process. ALIS liposomes are composed of dipalmitoylphosphatidylcholine (DPPC) and cholesterol at a 2:1 weight ratio, with 70 mg/mL of amikacin and 47 mg/mL of lipid (DPPC and cholesterol combined). The high drug loading is achieved through a proprietary infusion process that mixes a stream of lipids dissolved in ethanol with a stream of amikacin sulfate dissolved in water at a specific flow rate ratio (Li et al., 2008), causing amikacin to be in a semi-soluble, coacervated state during liposome formation. The high drug load within each liposome is critical to deliver an effective dose of amikacin to NTM-LD patients in about 14 min using a product-specific eFlow <sup>R</sup> nebulizer (LamiraTM Nebulizer System; PARI Respiratory Equipment, Inc.). The liposomes are relatively small (<300 nm in diameter) and release about 30% of the initial amikacin load during nebulization. Approximately 67% of the aerosol droplets produced are within the respirable range (<5.0 µm), which enables 43% of the nominal liposome dose to be deposited in the lungs of NTM-LD patients after nebulization (Olivier et al., 2016).

Once inhaled into the lungs, ALIS must disperse throughout the lungs and penetrate through mucus to reach infected areas. In healthy animals dosed with ALIS by inhalation, amikacin was distributed evenly throughout the lungs after single and multiple doses, with equal amikacin concentrations in all lobes of both lungs (Malinin et al., 2016). Similarly, inhaled ALIS distributed throughout the lungs in both healthy volunteers (Weers et al., 2009) and patients with NTM-LD (Olivier et al., 2016); the ratio of distribution to central and peripheral lung regions was approximately 1.6–2.0 (Weers et al., 2009; Olivier et al., 2016) and more than 50% of the deposited dose was detectable in the lung 24 h post-dose (Olivier et al., 2016). Mucus represents a further barrier to distribution of inhaled particles, but ALIS

liposomes effectively diffuse through 500–1000 µm thick human mucus in 30 min (Meers et al., 2008). These studies show that inhaled ALIS can distribute across lung regions and through mucus.

The final delivery challenge for ALIS is reaching sites of infection where NTM reside. Previous work demonstrated in vitro efficacy with ALIS against multiple M. avium and M. abscessus strains in a THP-1 macrophage model of infection (Rose et al., 2014). Furthermore, inhaled, nebulized ALIS reduced the NTM burden in mice with respiratory M. avium infections (Rose et al., 2014). However, the mechanistic data demonstrating ALIS penetration into NTM biofilms and macrophages are limited. Therefore, we evaluated a range of ALIS concentrations against M. avium biofilms and we also quantified ALIS uptake into macrophages, both in vitro and in rats given inhaled ALIS.

### MATERIALS AND METHODS

### Preparation and Characterization of ALIS Containing TAMRA-Amikacin

Amikacin liposome inhalation suspension was prepared using a microscale flow focusing method. DPPC (NOF America; cat # COATSOME <sup>R</sup> MC-6060) and cholesterol (Avanti Polar Lipids; cat # 700000) were dissolved at a 2:1 weight ratio in 100% ethanol at 20 mg/mL. Amikacin sulfate was dissolved in sterile, deionized water at 80 mg/mL (53.4 mg/mL amikacin base; pH adjusted to 6.7 using 5 M NaOH). For studies requiring fluorescent liposomes, 0.01% AF647-labeled dipalmitoylphosphatidylethanolamine (DPPE; Avanti Polar Lipids; cat # 850705) was included in the lipid solution and 3% tetramethylrhodamine (TAMRA)-amikacin (AAT Bioquest) was included in the amikacin stock. The lipid and amikacin solutions were infused (17 and 34 mL/min flow rates, respectively) through a Y connector, mixed with 2/3 volumes of 275 mM sodium citrate buffer at pH 8, and the resulting liposome suspension was diluted 1:1 in the collection vial with 1.5% NaCl. Samples were subsequently washed via tangential flow filtration using 2 EMD Millipore Pellicon XL cassettes (500 kDa polyethersulfone membrane; cat # PXB500C50).

Total amikacin concentrations in the liposome suspensions were measured by high-performance liquid chromatography (HPLC) using a Hypersil GOLD C18 column (175 Å, 3 µm, 150 mm × 4.6 mm; Thermo Fisher; cat. # 25003-154630) with a mobile phase of 65% methanol, 35% water, and 0.3% pentafluoropropionic acid (PFPA). An aliquot of each liposome suspension was centrifuged in an Amicon Ultra-0.5 centrifugal filter unit with Ultracel-30 KDa membrane to separate free amikacin, and the amount of free amikacin was then determined by HPLC. The lipid concentrations were measured by HPLC using an XBridge C8 column (130 Å, 3.5 µm, 150 mm × 4.6 mm; Waters; cat. # 186003055) with a mobile phase A consisting of 49.9% acetonitrile, 49.9% water, 0.1% acetic acid, and 0.1% triethylamine and mobile phase B consisting of 44.9% acetonitrile, 45% isopropyl alcohol, 10% water, 0.1% acetic acid, and 0.1% triethylamine. Liposome sizes were determined using a Mobius dynamic light scattering instrument (MOB-131, Wyatt Technology).

To confirm the TAMRA-amikacin concentration in the final sample of fluorescent ALIS, liposomes were dissolved with octylthioglucoside detergent to eliminate any minor-self quenching effect and TAMRA fluorescence was measured on a plate reader (Biotek Synergy NEO; BioTek Instruments, Inc., Winooski, VT, United States). Standards with decreasing percentages of TAMRA-amikacin mixed with amikacin (total amikacin concentration fixed at 80 mg/mL) were measured simultaneously and the final TAMRA-amikacin concentration of 0.44% in the ALIS sample was calculated from the standard curve. The same amount (0.44%) of TAMRA-amikacin was used for the free amikacin sample.

#### Mycobacterium avium Biofilm Assays

The biofilm studies were performed with the A5 strain of M. avium subspecies hominissuis that was originally isolated from an AIDS patient with a pulmonary infection. For imaging studies, mycobacteria were cultured for 7 days in 2-well chamber slides to establish biofilms. The biofilms were treated with 512 µg/mL of fluorescently labeled ALIS (AF647) for 4 h, fixed with 4% paraformaldehyde for 15 min, stained with Syto9 for 15 min, and sealed with a coverslip. Biofilms were imaged with a Zeiss LSM 780 confocal scanning microscope (630× magnification) and Zen 2.3 software. For efficacy studies, mycobacteria were cultured on plates containing Middlebrook 7H10 agar supplemented with 10% oleic acid-albumin-dextrose-catalase (OADC; Hardy Diagnostics, Santa Maria, CA, United States) at 37◦C for 7–10 days and then resuspended into Hank's Balanced Salt Solution (HBSS; Corning, Corning, NY, United States) at 3 × 10<sup>8</sup> CFU/mL using optical density. Input suspensions were serially diluted 10-fold and plated to determine the inoculum for each experiment. The input suspension was added to flat-bottom, 96-well polystyrene plates (150 µL/well) and biofilms were formed statically at room temperature for 7 days. After 7 days, supernatants were removed and replaced with HBSS containing 60 mM lactose ± increasing concentrations of either 100% ALIS or 70% ALIS/30% free amikacin (total amikacin concentrations tested were 16, 32, 64, 128, 256, 512, or 1024 µg/mL). Biofilms were treated for 4 days and then disrupted by pipetting the supernatant 50 times. The mixture was serially diluted 10-fold to a final dilution of 10<sup>4</sup> and the 102–10<sup>4</sup> dilutions were spot plated in duplicate to quantify CFU. Three independent experiments were performed on different days, with three technical replicates for each condition tested.

### Macrophage Uptake Assay

THP-1 human peripheral blood monocytes (ATCC; cat. # TIB-202) were seeded in 12-well plates at 1.2 × 10<sup>6</sup> cells/well in RPMI-1640 media (ATCC; cat. # 30-2001) containing 10% fetal bovine serum (FBS; HyClone; cat. # SH30070.03) and 50 ng/mL phorbol myristate acetate to induce attachment and differentiation into macrophages. After 24 h, the media was replaced with fresh RPMI media containing 5% FBS. After another 24 h, the media was replaced with 1 mL of fresh RPMI media with 5% FBS,

and 250 µL of ALIS (0.44% TAMRA-amikacin), free amikacin (0.44% TAMRA-amikacin), or control solution were added to respective wells to achieve the final desired concentrations (8, 16, 32, 64, or 128 µg/mL total amikacin). ALIS and free amikacin stocks were dissolved in 300 mM lactose solution. Following 4 or 24-h incubation periods, the media was removed, and then the cells were washed once with phosphate-buffered saline (PBS) and harvested into a 96-well plate. The cells were centrifuged at 300 × g, washed twice with PBS, and suspended in 200 µL of PBS. A small aliquot of the cell suspension was removed for imaging and the remaining cell suspension was used to quantify TAMRA-amikacin fluorescence using a Guava easyCyte 6HT flow cytometer (Millipore) with Easy Check beads as an internal control. The mean fluorescence intensity (MFI) was measured for 5,000 cells and normalized to the TAMRA fluorescence added to each well to account for any slight differences. The assay was repeated in three independent experiments, and then averages and standard errors of the mean were calculated for each treatment condition.

#### Macrophage Imaging

The aliquot removed prior to flow cytometry analysis was transferred to another 96-well plate and fixed in 4% formaldehyde. Fixed cells were mounted to glass slides with ProLongTM Diamond Antifade Mountant with 4<sup>0</sup> ,6-diamidino-2-phenylindole (DAPI; Thermo Fisher; cat. # P36962) and coverslips were sealed with clear nail polish. Samples were imaged with a Zeiss Axio fluorescence microscope (400× magnification) and Zen 2.3 software. Settings were kept constant while imaging all samples from the same timepoint.

#### Animals and Housing Conditions

Male, Han-Wistar rats were purchased from Charles River Laboratories, group-housed in plastic cages, and maintained with an average room temperature of 69◦F (range 68–78◦F), an average room humidity of 49% (range 30–70%), and a 12 h light cycle (light on 6:00–18:00). Animals received standard chow (Purina diet; cat. #5053) and water ad libitum throughout the study, except during the nebulization period. This study was carried out in accordance with the recommendations of the Guide for the Care and Use of Laboratory Animals. The protocol was approved by the animal care and use committee at Rutgers University.

#### Evaluation of Macrophage Content in Cells Collected by Bronchoalveolar Lavage

Four naïve rats were anesthetized under isoflurane anesthesia and bronchoalveolar lavage (BAL) was performed with 2 mL of PBS. The BAL fluid was centrifuged at 2,000 × g, the resulting cell pellet was resuspended in FACS staining buffer, and Fc receptors were blocked for 30 min at 4◦C. Cells were stained for 30 min with an Alexa Fluor 700 anti-rat CD45 antibody (BioLegend, San Diego, CA, United States), an Alexa Fluor 488 mouse antirat CD11b antibody (Bio-Rad, Hercules, CA, United States), a mouse anti-rat CD11c antibody (Bio-Rad) pre-conjugated with Cy5, and a biotin-conjugated mouse anti-rat CD68 antibody (Bio-Rad), followed by a phycoerythrin-conjugated streptavidin (BioLegend). Finally, all cell samples were stained with DAPI. Cell populations were analyzed using a BD Influx Cell Sorter (BD Biosciences, San Jose, CA, United States), and flow cytometry compensation setting was performed with UltraComp eBeadsTM (Thermo Fisher Scientific, Waltham, MA, United States) and with cell samples. The NR8383 rat alveolar macrophage cell line was used as a positive control for staining and flow cytometry.

### Administration of ALIS and Free Amikacin

For inhalation dosing, rats were placed into plastic tube restrainers and loaded into an ONARES nose-only inhalation delivery system (CH Technologies, Westwood, NJ, United States). Drug suspension/solution (40 mL of ALIS or free amikacin) was split between two Aeroneb Pro nebulizers (Aerogen, Chicago, IL, United States) placed in series along the air flow (3 L/min) into the ONARES chamber. The aerosol concentration in the ONARES chamber was sampled 5 min after the start of nebulization by drawing aerosol through a Pall #61631 glass-fiber filter (MilliporeSigma, St. Louis, MO, United States) at 0.25 L/min for 5 min. The filters were stored at 4◦C in a glass vial for subsequent amikacin measurements. Nebulization continued until 40 mL of drug solution/suspension was nebulized and the nebulizers ran clear for one additional minute (total nebulization time was 60–90 min).

The study consisted of three experimental groups: ALIS (40 mL at 53.4 mg/mL amikacin base); low-dose free amikacin (40 mL at 20.0 mg/mL amikacin base); and, high-dose free amikacin (40 mL at 53.4 mg/mL amikacin base). Each timepoint for each experimental group required a separate nebulization group consisting of 10 rats; lungs were collected from 2 rats immediately post-dose at the end of the nebulization period to determine the deposited dose of amikacin, and the remaining 8 rats were euthanized together at one timepoint, either 2, 6, or 24 h post-dose. Therefore, each experimental group consisted of 30 rats total, with n = 6 euthanized immediately post-dose to measure deposited doses (2 from each of 3 nebulization groups) and n = 8 euthanized at each timepoint of 2, 6, and 24 h.

The distribution of intravenous amikacin into pulmonary macrophages, airways, and lung tissue was also assessed using a similar study design. Ten rats in each group were injected with a 20 mg/mL amikacin solution in 0.9% NaCl at 5 mL/kg to achieve a dose of 100 mg/kg. From each group, 2 animals were euthanized 0.25 h post-dose (similar to the immediately postdose group for the nebulization studies) and the remaining 8 were euthanized together at either 2, 6, or 24 h post-dose. BAL fluid and cell pellets, lungs, and plasma were collected from each rat as described below.

### Blood, Bronchoalveolar Lavage Fluid, and Lung Tissue Collection

At the appropriate timepoint after dosing, rats were weighed and deeply anesthetized with isoflurane, blood was collected

by retroorbital bleeding into ethylenediaminetetraacetic acid (EDTA) coated BD Microtainer <sup>R</sup> tubes (cat. # 02 669-33; Becton Dickinson, Franklin Lakes, NJ, United States), and then rats were euthanized by exsanguination. The lungs were lavaged with 2 mL of PBS and the BAL fluid was collected. Lungs were excised, weighed, rinsed in PBS, placed in 5 mL Eppendorf tubes, and flash frozen in liquid nitrogen. Blood samples were kept on ice prior to centrifugation at 20,817 × g for 10 min, and then the plasma was collected and stored at −80◦C until analysis. BAL fluid was centrifuged at 2,000 × g for 3 min and the supernatant was removed and stored at −80◦C. The cell pellet was resuspended in 1 mL of fresh PBS and centrifuged at 2,000 × g for 3 min. The supernatant was discarded and the cell pellet was stored at −80◦C until analysis.

### Quantification of Sample Amikacin Concentrations

Amikacin standards were generated in each sample matrix: cultured NR8383 rat alveolar macrophage cells were used for BAL cell pellet standards; PBS was utilized for BAL fluid standards; lung homogenate from naïve rats was employed for lung homogenate standards; and, plasma from naïve rats was used for plasma standards. For each standard curve, a 1.0 mg/mL intermediate in the appropriate matrix was first prepared from a primary amikacin stock of 20.0 mg/mL in 20% methanol and 80% water, then further diluted to 100 µg/mL in the appropriate matrix and serially diluted to make calibration standards. Quality control samples were also prepared from the highest standard. Experimental samples were mixed with an equal volume of internal standard working solution (gentamicin C1 in water) and 10% trichloroacetic acid, vortexed for 5 s, and centrifuged at 21,130 × g at 4◦C for 15 min. Depending on the analytical range, the supernatant was mixed with either equal volume or 10× volume of mobile phase A (0.2% formic acid and 0.1% heptafluorobutyric acid in water) and 10 µL was injected into a Phenomenex Kinetex Phenyl-Hexyl column (4.6 mm × 50 mm; 2.6 µm; 50◦C; 1 mL/min) with an elution gradient consisting of mobile phase A and mobile phase B (100% acetonitrile). Liquid chromatography, tandem mass spectrometry (LC-MS/MS) was performed using a Sciex API4500 mass spectrometer (AB SCIEX, Framingham, MA, United States).

#### Quantification of Protein Concentrations

A portion of rat BAL cell pellet lysate and lung homogenate samples were retained for protein quantification using a PierceTM BCA Protein Assay Kit (Thermo Fisher) according to the manufacturer's instructions.

#### Statistical Analyses

fmicb-09-00915 May 14, 2018 Time: 15:46 # 6

Statistical analyses were performed with GraphPad Prism (La Jolla, CA, United States). Because the inoculation levels were different between the three independent biofilm experiments (P = 0.0007 by one-way ANOVA with Tukey post-test), the CFU count for each concentration was normalized as percent of the untreated CFU count within each experiment and then group means and standard errors were calculated. Significant changes from the untreated control group were identified by one-way ANOVA with Dunnett's multiple comparisons test. For the in vitro macrophage uptake experiments, the data were analyzed by two-way ANOVA with Tukey post-test. For the in vivo studies, differences between the dose groups at each timepoint were determined by one-way analysis of variance (ANOVA) with Tukey post-test. P < 0.05 was considered significant.

#### RESULTS

#### ALIS Penetrates and Kills Mycobacterium avium Biofilms

During respiratory NTM infection, one of the niches the bacteria can reside within are biofilms. ALIS is effective at treating NTM infections, but it is unknown if it can penetrate and kill NTM biofilms. We treated M. avium biofilms for 4 h with AF647-labeled ALIS to evaluate whether the liposomes can penetrate into the biofilm. Mycobacteria formed a dense biofilm on the slide surface, with more diffuse bacteria and extracellular biofilm components above (**Figure 1**). Liposomes were visible throughout the biofilm, indicating that the liposomes penetrated through the extracellular components and reached the cell dense region (**Figure 1**). We also tested whether increasing concentrations of ALIS can effectively kill the mycobacteria within the biofilms, using both 100% ALIS and a mix of 70% ALIS/30% free amikacin that represents the ratio of liposomal/free amikacin after nebulization. ALIS concentration-dependently reduced viable cell counts in biofilms at concentrations greater than 16 µg/mL, and the 100% ALIS and 70% ALIS/30% free amikacin treatment groups were equally effective (**Figure 2**).

#### ALIS Enhanced in Vitro Amikacin Uptake Into Macrophages Compared to Free Amikacin

Non-tuberculous mycobacteria are intracellular pathogens that mainly reside within macrophages during infection,

creating an additional challenge to deliver effective antibiotic concentrations inside the cell. Taking this important point into consideration, we used fluorescently labeled amikacin to investigate macrophage uptake of ALIS. The fluorescent ALIS formulation produced for uptake studies contained 0.44% TAMRA-amikacin; therefore, we used the same 0.44% TAMRA-amikacin concentration in the free amikacin sample for all experiments to ensure that the total fluorescence added was equal for each comparison between ALIS and free amikacin. The batch of fluorescent ALIS had characteristics similar to a previous engineering batch (ENG1505) of ALIS that represents a typical batch produced by the manufacturing process (**Table 1**).

To confirm that TAMRA conjugation did not alter macrophage uptake of amikacin, we performed a pilot study using mass spectrometry to measure the ratio of TAMRA-labeled to unlabeled amikacin in the cell pellet after 24-h incubation with



<sup>1</sup>Batch ENG1505 represents a typical batch produced by the ALIS manufacturing process.

labeled with TAMRA) for 4 h (gray symbols) or 24 h (black symbols) and, cell monolayers were resuspended, then cellular TAMRA fluorescence was quantified by flow cytometry. Panel (A) shows normalized mean fluorescence intensity (MFI) at each concentration averaged from three independent experiments. Panel (B) shows representative flow cytometry histograms with gating for the MFI after 24 h treatment from one experiment. Green represents control cells, red represents cells treated with ALIS or free amikacin, and yellow indicates overlap between the two histograms. <sup>∗</sup>P < 0.05 vs. free amikacin at the same concentration and timepoint.

free amikacin or ALIS. Cells treated with either free amikacin or ALIS had the same ratio of unlabeled/TAMRA-labeled amikacin (data not shown).

We quantified the uptake of various concentrations of free or liposomal TAMRA-amikacin into human macrophages after incubation for 4 or 24 h. Amikacin uptake was low after 4-h incubation, but cells treated with 128 µg/mL ALIS contained significantly more amikacin than cells treated with the same concentration of free amikacin (**Figure 3A**). After 24-h incubation, macrophages treated with ALIS generally contained ∼4-fold more amikacin than cells exposed to the same concentrations of free amikacin, with significant differences between ALIS and free amikacin groups exposed to 64 and 128 µg/mL concentrations (**Figure 3A**). The flow cytometry histograms demonstrated unimodal distributions for both free and liposomal amikacin at all concentrations tested (**Figure 3B**).

Fluorescence microscopy images taken after 24-h incubation were consistent with the flow cytometry measurements. Macrophages exposed to 32, 64, or 128 µg/mL of ALIS clearly exhibited TAMRA fluorescence, whereas TAMRA fluorescence was sparse and dim in cells incubated with the same concentrations of free amikacin (**Figure 4**).

#### ALIS Enhanced in Vivo Amikacin Uptake Into Macrophages Compared to Free Amikacin

To establish a method to collect pulmonary macrophages, we used flow cytometry to determine the percentage of

macrophages among cells collected by BAL from naïve rats. Side and forward scatter differentiated single cells from other debris and cell aggregates in the sample, and high DAPI fluorescence identified dead cells for removal from the analysis (**Figure 5**). CD45<sup>+</sup> staining identified leukocytes and CD11chigh/CD11blow staining differentiated alveolar macrophages within the leukocyte population (**Figure 5**). High CD68<sup>+</sup> staining confirmed the identity of macrophages. Overall, 82.0 ± 3.3% of live cells isolated by BAL were alveolar macrophages.

Because the percentage of macrophages collected by BAL was high, we used total BAL cells for subsequent studies evaluating in vivo amikacin uptake into macrophages from rats given either ALIS or free amikacin by nebulization. From each nebulization group, we euthanized two rats immediately post-dose and measured the amikacin dose deposited in the lungs. The amikacin doses delivered to the nose or deposited in the lungs were not significantly different between the ALIS and low-dose amikacin groups (**Table 2**), but high-dose amikacin achieved significantly higher delivered and deposited doses than the other groups (**Table 2**).

We collected cells in the BAL fluid after inhalation administration of ALIS or free amikacin and measured amikacin concentrations in the cell pellets. The amount of protein in the cell pellets was not different between dosing groups at any timepoint (**Table 3**), indicating consistent collection of cells across the different dose groups. Cells harvested from the BAL fluid from rats dosed with ALIS exhibited 5- to 8-fold higher amikacin concentrations than BAL cells collected from rats dosed with low-dose free amikacin (**Figure 6A**). Additionally, the ALIS group contained 4- to 6-fold higher amikacin concentrations in the BAL cell pellet compared to the high-dose free amikacin group (**Figure 6A**), despite a twofold lower deposited dose in the ALIS group.

We also quantified amikacin levels in BAL fluid (representing amikacin collected from airways) and found that samples from rats in the ALIS group had significantly higher amikacin concentrations at 2 h post-dose than BAL fluid samples from rats in the low-dose amikacin group (**Figure 6B**). Despite a higher dose deposited in the lungs, high-dose amikacin did not achieve significantly higher BAL fluid concentrations than ALIS (**Figure 6B**). At 24 h post-dose, the ALIS group exhibited significantly higher BAL fluid amikacin concentrations than both groups administered free amikacin (**Figure 6B**).

Following each BAL procedure, we excised and froze the lungs for measurement of tissue amikacin concentrations. Lung concentrations at 2 and 6 h post-dose in the ALIS group were significantly higher than those in the low-dose amikacin group (although the two groups achieved similar deposited doses) and equivalent to amikacin concentrations in the high-dose amikacin group (despite a ∼2-fold lower deposited dose with ALIS; **Figure 6C**). At 24 h post-dose, lung concentrations in the ALIS group were significantly higher than both low- and high-dose amikacin (**Figure 6C**).

Plasma amikacin concentrations were not significantly different between the ALIS and low dose amikacin groups at 2 and 6 h post-dose (**Figure 6D**). Plasma concentrations were significantly higher in the ALIS group relative to the low-dose amikacin group, but not the high-dose group, at 24 h post-dose (**Figure 6D**).

Because intravenous amikacin is a common treatment for refractory NTM infections, we also evaluated the distribution of amikacin into pulmonary macrophages, airways, and lung tissue after an intravenous bolus dose. The 100 mg/kg dose

TABLE 2 | Delivered and deposited amikacin doses in rats administered either ALIS or free amikacin by nebulization.


Delivered dose was calculated as the aerosol concentration (calculated from filter samples captured during nebulization) × respiratory minute volume × duration of nebulization/body weight. The deposited dose was measured as the lung amikacin concentration immediately post-dose in two rats from each nebulization group. Values represent mean ± SEM. <sup>∗</sup>P < 0.05 vs. low-dose amikacin; n = 3 filters per group. ∗∗P < 0.05 vs. both low-dose amikacin and ALIS; n = 6 rats per group.

TABLE 3 | Protein concentrations of cell pellets collected by bronchoalveolar lavage (BAL).


BAL cell pellet protein concentrations (µg/mL) were measured by BCA protein assay. n = 8 per timepoint in each group.

given intravenously was comparable to the 96 mg/kg amikacin dose delivered by ALIS nebulization. The mean peak plasma concentration at 0.25 h post-dose was 337 µg/mL and the mean lung concentration at that timepoint was 94 µg/g, approximately 10-fold lower than the 977 µg/g deposited dosed achieved with nebulized ALIS (**Table 2**). In pulmonary macrophages, amikacin concentrations were less than 0.004 µg/µg protein throughout the 24-h time course and the area under the curve (AUC) was 274-fold lower than the macrophage AUC following ALIS inhalation (**Table 4**). Similarly, AUCs in BAL fluid and lung tissue were 69- and 42-fold lower, respectively, after intravenous dosing compared with inhalation dosing of ALIS (**Table 4**).

#### DISCUSSION

Amikacin liposome inhalation suspension was designed to provide efficient nebulization delivery of liposomes to the lungs, distribute to both central and peripheral regions of the lungs, and penetrate into biofilms and macrophages to reach the sites of infection. Li et al. (2008) showed that ALIS nebulization provides a consistent combination of encapsulated and free amikacin and forms aerosol droplets within the respirable range, and Meers et al. (2008) and Malinin et al. (2016) demonstrated that ALIS distributes well throughout different lobes of the lungs and into peripheral airways. The data presented herein clearly establish that ALIS can penetrate M. avium biofilms and into macrophages.

Amikacin liposome inhalation suspension effectively penetrated M. avium biofilms to reduce the number of viable bacteria; this was true for 100% ALIS as well as 70% ALIS/30% free amikacin (representing nebulized ALIS). In vivo, ALIS retained more amikacin at 24 h in airways and lung tissue compared to free amikacin, which would increase the duration of amikacin exposure at the sites of biofilm infections within mucus and alveolar walls.

Additionally, ALIS significantly enhanced amikacin uptake into macrophages compared with free amikacin. Using flow cytometry to quantify TAMRA fluorescence in macrophages, we found that the liposomal formulation enhanced amikacin uptake following 4- and 24-h incubations. Enhanced uptake with ALIS relative to free amikacin was also evident by fluorescence microscopy. These in vitro studies relied on the use of fluorescent amikacin, but we confirmed in a pilot study that TAMRA-conjugation did not alter uptake of free amikacin into macrophages. These data demonstrate that ALIS significantly increased amikacin uptake into human macrophages by ∼4-fold relative to free amikacin over a range of concentrations. The finding of improved uptake into macrophages in vitro is consistent with published data showing that ALIS kills intracellular M. avium infections in macrophages better than free amikacin (Rose et al., 2014).

Pulmonary macrophages from rats dosed with ALIS exhibited 5- to 8-fold higher amikacin levels compared with cells from animals given low-dose free amikacin, although the deposited dose immediately after nebulization was not significantly different between the groups. Additionally, the ALIS group exhibited significantly higher amikacin concentrations in BAL fluid and lung tissue relative to the low-dose amikacin group. These results show that the ALIS formulation retains amikacin within macrophages, airways, and lung tissue significantly better than free amikacin. Furthermore, high-dose amikacin deposited twofold more amikacin in the lungs immediately postdose compared to ALIS, but failed to produce macrophage amikacin concentrations equaling those in the ALIS treatment group (4- to 6-fold lower with high-dose amikacin vs. ALIS). Amikacin levels in the bronchoalveolar space and lung tissue declined more over 24 h after administration of high-dose free amikacin relative to ALIS, indicating that free amikacin passes out of the lungs faster than ALIS. Taken together, these data provide strong evidence that inhaled ALIS can deliver more amikacin into macrophages in vivo than free drug. Importantly, simply depositing more free amikacin into the lungs failed to match the macrophage amikacin concentrations achieved with ALIS.

Intravenous amikacin is commonly used in patients with severe or refractory NTM-LD, but we found that free amikacin penetrated very poorly into lung tissue, airways, and pulmonary macrophages after intravenous administration. Compared to intravenous amikacin, ALIS achieved ∼42-, ∼69-, and ∼274 fold higher amikacin exposures in lung tissue, airways, and pulmonary macrophages, respectively, with ∼5-fold lower plasma exposure. Although intravenous amikacin is the current standard of care, ALIS delivered more antibiotic to sites relevant for NTM lung infections with substantially lower systemic exposure.

Amikacin liposome inhalation suspension has been shown to effectively treat NTM-LD in both preclinical studies and

FIGURE 6 | Amikacin liposome inhalation suspension (ALIS) enhanced amikacin uptake into macrophages and retained more amikacin in airways and lung tissue. Rats were administered either ALIS or free amikacin by nebulization and BAL cells (representing >80% macrophages) (A), BAL fluid (B), lungs (C), and plasma (D) were collected at 2, 6, or 24 h after the end of nebulization. Amikacin concentrations were quantified by LC-MS/MS. The delivered and deposited doses for each group are presented in Table 2. <sup>∗</sup>P < 0.05 vs. both high-dose amikacin and ALIS groups. ∗∗P < 0.05 vs. both low and high-dose amikacin groups. #P < 0.05 vs. both ALIS and low-dose amikacin groups. ˆP < 0.05 vs. the low-dose amikacin group. n = 8 rats per group.


TABLE 4 | Summary of macrophage, airway, lung tissue, and plasma exposures.

Values represent area under the curve (AUC) calculated for each group using the mean concentrations at each timepoint. AUC2−<sup>24</sup> was calculated for macrophage, airway, and plasma exposures, whereas AUC0−<sup>24</sup> was calculated for lung exposure.

clinical trials. In mice with pulmonary M. avium infections, inhalation administration of ALIS lowered viable mycobacteria in the lungs by more than 2 log units (Rose et al., 2014). A Phase II trial of patients with refractory NTM-LD demonstrated that ALIS increased the proportion of patients who achieved negative sputum cultures compared with placebo, and the time to first negative sputum culture was shorter with ALIS treatment versus placebo (Olivier et al., 2017). Based on top-line results, an ongoing Phase III study met the primary endpoint by demonstrating that the addition of ALIS to guideline based therapy eliminated evidence of NTM-LD caused by MAC in sputum by month 6 in a greater proportion of patients than guideline based therapy alone (Insmed, 2018).

Overall, the data presented herein demonstrate that ALIS delivers amikacin to pulmonary macrophages, airways, and lung tissue better than free amikacin given by either inhalation or intravenous administration. Simply delivering more free amikacin to the lungs by nebulization failed to match the macrophage uptake or duration of exposure achieved by ALIS, indicating the benefit of the liposomal formulation. This mechanism of improved delivery into pulmonary macrophages and retention within airways and lung tissue has been shown to effectively treat refractory NTM-LD in clinical trials and represents a promising new therapy for patients.

#### AUTHOR CONTRIBUTIONS

fmicb-09-00915 May 14, 2018 Time: 15:46 # 12

FL, JZ, SR, WP, and KD contributed conception and design of the studies. KD performed the statistical analyses. FL and KD wrote the manuscript. JZ, DC, SR, MN, JJ, JT, and TH performed the experiments. All authors contributed to manuscript revision, and then read and approved the submitted version.

#### REFERENCES


#### FUNDING

This work was funded by Insmed Incorporated.

#### ACKNOWLEDGMENTS

Special thanks to Daniel Martin in the High Resolution Microscopy Facility, Department of Biomedical Engineering at Rutgers University for assistance with biofilm imaging.

liposomal amikacin in chronic Pseudomonas aeruginosa lung infections. J. Antimicrob. Chemother. 61, 859–868. doi: 10.1093/jac/dkn059


**Conflict of Interest Statement:** Authors JZ, FL, SR, DC, JT, TH, MN, WP, and KD were employed by Insmed Incorporated. The other author JJ declares no competing interests. The funder, Insmed Incorporated, had no role in study design, data collection and analysis, but was involved in the decision to publish the manuscript.

Copyright © 2018 Zhang, Leifer, Rose, Chun, Thaisz, Herr, Nashed, Joseph, Perkins and DiPetrillo. 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.

# Optimization and Lead Selection of Benzothiazole Amide Analogs Toward a Novel Antimycobacterial Agent

Mary A. De Groote1†, Thale C. Jarvis 2†, Christina Wong<sup>2</sup> , James Graham<sup>2</sup> , Teresa Hoang<sup>2</sup> , Casey L. Young<sup>2</sup> , Wendy Ribble<sup>2</sup> , Joshua Day <sup>2</sup> , Wei Li <sup>1</sup> , Mary Jackson<sup>1</sup> , Mercedes Gonzalez-Juarrero<sup>1</sup> , Xicheng Sun<sup>2</sup> and Urs A. Ochsner <sup>2</sup> \*

*<sup>1</sup> Mycobacteria Research Laboratories, Department of Microbiology, Immunology and Pathology, Colorado State University, Fort Collins, CO, United States, <sup>2</sup> Crestone, Inc., Boulder, CO, United States*

#### *Edited by:*

*Veronique Anne Dartois, Rutgers, The State University of New Jersey, Newark, United States*

#### *Reviewed by:*

*Philip Hipskind, Lgenia Inc, United States Courtney Cortez Aldrich, University of Minnesota Twin Cities, United States*

#### *\*Correspondence:*

*Urs A. Ochsner uochsner@crestonepharma.com*

*†These authors have contributed equally to this work*

#### *Specialty section:*

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

*Received: 30 May 2018 Accepted: 31 August 2018 Published: 20 September 2018*

#### *Citation:*

*De Groote MA, Jarvis TC, Wong C, Graham J, Hoang T, Young CL, Ribble W, Day J, Li W, Jackson M, Gonzalez-Juarrero M, Sun X and Ochsner UA (2018) Optimization and Lead Selection of Benzothiazole Amide Analogs Toward a Novel Antimycobacterial Agent. Front. Microbiol. 9:2231. doi: 10.3389/fmicb.2018.02231* Mycobacteria remain an important problem worldwide, especially drug resistant human pathogens. Novel therapeutics are urgently needed to tackle both drug-resistant tuberculosis (TB) and difficult-to-treat infections with nontuberculous mycobacteria (NTM). Benzothiazole adamantyl amide had previously emerged as a high throughput screening hit against *M. tuberculosis* (*Mtb*) and was subsequently found to be active against NTM as well. For lead optimization, we applied an iterative process of design, synthesis and screening of several 100 analogs to improve antibacterial potency as well as physicochemical and pharmacological properties to ultimately achieve efficacy. Replacement of the adamantyl group with cyclohexyl derivatives, including bicyclic moieties, resulted in advanced lead compounds that showed excellent potency and a mycobacteria-specific spectrum of activity. MIC values ranged from 0.03 to 0.12 µg/mL against *M. abscessus* (*Mabs*) and other rapid- growing NTM, 1–2µg/mL against *M. avium* complex (MAC), and 0.12–0.5µg/mL against *Mtb.* No pre-existing resistance was found in a collection of *n* = 54 clinical isolates of rapid-growing NTM. Unlike many antibacterial agents commonly used to treat mycobacterial infections, benzothiazole amides demonstrated bactericidal effects against both *Mtb* and *Mabs*. Metabolic labeling provided evidence that the compounds affect the transfer of mycolic acids to their cell envelope acceptors in mycobacteria. Mapping of resistance mutations pointed to the trehalose monomycolate transporter (MmpL3) as the most likely target. *In vivo* efficacy and tolerability of a benzothiazole amide was demonstrated in a mouse model of chronic NTM lung infection with *Mabs*. Once daily dosing over 4 weeks by intrapulmonary microspray administration as 5% corn oil/saline emulsion achieved statistically significant CFU reductions compared to vehicle control and non-inferiority compared to azithromycin. The benzothiazole amides hold promise for development of a novel therapeutic agent with broad antimycobacterial activity, though further work is needed to develop drug formulations for direct intrapulmonary delivery via aerosol.

Keywords: NTM, tuberculosis, antibacterial therapeutics, benzothiazole amide, MmpL3, efficacy, tolerability, aerosol

## INTRODUCTION

Infections with non-tuberculous mycobacteria (NTM) are increasing in incidence (Falkinham, 2013; Prevots and Marras, 2015; Strollo et al., 2015; Vinnard et al., 2016; Prevots et al., 2017; Spaulding et al., 2017) and are notoriously difficult to treat (Henkle et al., 2017; Lande et al., 2018). Resistance of NTM to disinfectants may contribute to hospital acquired infections (Caskey et al., 2018) especially in cystic fibrosis disease where infection is common as patients age (Nick et al., 2016; Martiniano et al., 2017). Increasing evidence suggests that there is a protective effectiveness of the Bacillus Calmette-Guerin (BCG) vaccination against NTM disease (Zimmermann et al., 2018). Tuberculosis may give some cross-protection to NTM infection. With the declining incidence of TB in the US, NTM may fill this immunity void and we will see rising numbers of cases and will be challenged with treating NTM infection as the tuberculosis prevalence is at an all-time low in the US. Given that the US has never undertaken a BCG vaccination strategy like most other nations there may be a niche for NTM infections. NTM are ubiquitous in the environment and disease due to these organisms are heterogeneous disorders with pulmonary manifestations being the most common presentation. Lung disease manifests as nodular bronchiectasis and/or fibrocavitary disease. Some individuals will remain culture positive but clinically stable, but those with significant respiratory symptoms and radiographic abnormalities including destructive cavitary manifestations and microbiological evidence of an NTM will require treatment as disease progression frequently occurs and mortality can be high (Fleshner et al., 2016). Increasing prevalence of NTM will continue to occur as the population ages (Falkinham, 2003). Immune dysfunction has been seen in patients with pulmonary NTM (Cowman et al., 2018). Since many patients have underlying host predispositions they are susceptible to re-infection and may need intermittent treatment when symptomatic infection recurs (Lake et al., 2016). Treatment of non-tuberculous mycobacterial lung disease (NTM-LD) is challenging for several reasons including the relative resistance of NTM to currently available drugs and the difficulty in tolerating prolonged treatment with multiple drugs (Philley et al., 2016).

There are virtually no antimicrobial drug discovery programs specifically targeting NTM. Although therapeutic agents developed to treat Mtb infections often lack activity against NTM, it remains an attractive approach to initiate new NTM drug discovery projects via screening a library of TB active compounds against NTM, which has indeed resulted in high hit rates (Low et al., 2017). In fact, our development candidate, benzothiazole adamantyl amide, was discovered via that approach in a screening for Mabs whole cell activity of a library composed of hits from a previous Mtb screening (Franzblau et al., 2012). Another strategy for NTM drug discovery is to revisit or repurpose older drugs. A potential role for clofazimine in NTM treatment regimens has been suggested, since this agent showed bactericidal activity and synergy with amikacin or clarithromycin, both of which are commonly used antibiotics to treat NTM infections (Ferro et al., 2016). Semisynthetic spectinamides derived from the old spectinomycin discovered over half a century ago have shown potent activity against MDR and XDR tuberculosis (Liu et al., 2017).

NTM treatment regimens differ by species (Daley and Glassroth, 2014; Kasperbauer and De Groote, 2015), particularly between rapid growers (RGM) comprised of M. abscessus (Mabs) complex (M. abscessus, ssp. abscessus, bolletti, and massiliensis), M. chelonae, M. fortuitum, and others; and slow growers represented by M. avium, M. intracellulare, and M. chimaera (M. avium complex, or MAC). M. abscessus (Mabs) is particularly difficult to treat (Haworth et al., 2017). This organism is capable of forming more virulent rough phenotypes and smooth forms that tend to form biofilms (Claeys and Robinson, 2018) associated with antibiotic drug resistance (Clary et al., 2018). Resistance to macrolides, a cornerstone class of therapeutic agents, is innate in certain Mabs subspecies that carry an inducible methylase genetic element and the presence of this gene affects the outcomes (Koh, 2017; Choi et al., 2018). This resistance is detected clinically as well as in the laboratory (Carvalho et al., 2018) and is an important development limiting effective therapy (Kasperbauer and De Groote, 2015). Therapy involves multiple agents and recalcitrant or severe infections are optimally managed with the addition of an injectable agent (De Groote and Huitt, 2006; Philley et al., 2016). Therapeutic intolerances and side effects limit adherence to treatment regimens and affect outcomes. In addition, the potency of available agents is relatively low. Thus, there is a compelling need for novel antibiotics that are more potent and better tolerated.

The most common form of NTM disease is lung disease. NTM can result in chronic progressive lung disease especially for those with known risk factors, the most common of which is aging. Patients with cystic fibrosis, chest structural abnormalities and pre-existing lung disease such as bronchiectasis and autoimmune diseases and their treatments are all risk factors for infection and disease. For pulmonary disease, direct delivery to the airway would allow greater penetration and less systemic exposure. An inhaled route of delivery would be advantageous for intermittent therapy, in particular to suppress systemic side effects. Advances have been made in the area of aerosol delivery carriers and devices. The success of inhalational therapies with liposomal amikacin (Olivier et al., 2014; Caimmi et al., 2018) and ArikayceTM (Olivier et al., 2017) has paved the way for new therapies in this category. The development of a multicenter clinical trial network will allow more rapid enrollment in new drug treatment trials (Kevin Winthrop, personal communication).

Much remains to be done in the field of NTM as this has been a neglected area of drug development. As more treatment response surrogate markers are discovered, the future holds great promise for utilization of biomarkers to match therapeutic regimens to appropriate patient subpopulations and to monitor treatment response (Asakura et al., 2015) similar to biomarker work in TB to enhance drug development (Sigal et al., 2017). This will require clinical studies involving multi-center cooperative studies to ensure an adequate number of enrollees. In addition, diseases like Buruli ulcer caused by M. ulcerans are emerging infections in need of novel treatment approaches (Tai et al., 2018; Zingue et al., 2018). Wound healing is prolonged and problematic (O'Brien

et al., 2018) and topical therapy for Buruli ulcer or M. marinum cutaneous infections may have some utility depending on drug penetration into deeper tissues (Simoes et al., 2016). Mycobacteria live in lipid-rich host environments (Aguilar-Ayala et al., 2018; Ayyappan et al., 2018) so a lipophilic drug might be attractive for a variety of disease manifestations if lipophilicity can be managed. There is always the need to balance good drug properties such as hydrophilicity with the impenetrable lipid rich cell walls of mycobacteria. In addition to delivery improvements, future work to optimize potency against members of the slowly growing mycobacteria, particularly (MAC), will be a priority.

## MATERIALS AND METHODS

#### Synthesis of Benzothiazole Amide Analogs

Unsubstituted benzothiazole adamantyl amide had emerged as a hit during screening of a focused library of TB-actives (Franzblau et al., 2012) for compounds that also possessed activity against M. abscessus. This scaffold represented the starting point for a campaign to elucidate the structure-activity relationship in greater detail. The initial strategy involved trimming of the adamantyl group to a minimum of lipophilic structure required for activity. The general synthetic route started from substituted 2-amino-benzothiazole intermediates and variably substituted cycloalkyl carboxylic acids under standard amide coupling conditions using 1-[Bis(dimethylamino)methylene]-1H-1,2,3 triazolo[4,5-b]pyridinium 3-oxide hexafluoro-phosphate (HATU) in the presence of N,N-diisopropylethylamine (DIEA) in dichloroethane (DCE). A detailed description of chemical synthesis, purification and SAR of substituted benzothiazole cyclohexyl amides is presented elsewhere (Graham et al., 2018). The products were purified by column chromatography using ethyl acetate and hexanes as eluents and fully characterized by NMR and LC-MS. Purity of the lead analogs described here was >95%.

#### Microbiological Evaluation of Benzothiazole Amide Compounds

NTM clinical isolates were from University of Colorado Hospital (De Groote et al., 2014a). M. chimaera strains were provided by the CDC (van Ingen et al., 2017), Reference strain Mabs 19977 and other bacterial strains, including anaerobes, were from ATCC, BEI Resources, or from the biorepository at Crestone, Inc. The avirulent Mtb strain H37Rv mc<sup>2</sup> 6206 is a leuC leuD mutant derivative of strain mc<sup>2</sup> 6020 (1lysA 1panCD) and was provided by Dr. Bill Jacobs (Sambandamurthy et al., 2005). Compounds were tested for antimicrobial activity against mycobacteria and other pathogens following guidelines published by the Clinical Laboratory Standards Institute (CLSI). Muller-Hinton broth (MHB, 3 days) was used for rapid-growing NTM and Middlebrook medium (7H9 + OADC, 7 days) for slow-growing NTM and for Mtb (CLSI, 2003). Accurate endpoint minimum inhibitory concentration (MIC) values were obtained after addition of resazurin for 24 h to monitor viable cells colorimetrically (Khalifa et al., 2013). Protein and serum binding of compounds was tested by MIC determination in the presence of human albumin (45 mg/mL) or 50% complement-inactivated human serum. Potential non-specific effects on membranes were monitored in a hemolysis assay, where equine erythrocytes were exposed to compounds over a concentration range of up to 128µg/mL in 10 mM Tris-HCl buffered 0.9% saline (pH 7.5) for 10 min at room temperature, centrifuged and evaluated for absence of erythrocyte pellets and buffer discoloration due to hemolysis. Spontaneous resistant mutants were isolated by plating 10<sup>9</sup> -10<sup>10</sup> Mabs ATCC 19977 cells onto 7H11-ADC agar plates containing benzothiazole amide compounds at multiples of their MIC, followed by colony purification on selective agar, preparation of genomic DNA, PCR amplification and sequencing of the mmpL3 gene. For time-kill assays, cultures of Mabs and Mtb were diluted with broth to cell densities of 10<sup>5</sup> -10<sup>6</sup> CFU/mL, compounds and control agents were added at 10 times their MICs and samples were removed for colony enumeration over a period of 5 days (Mabs) and 21 days (Mtb). Synergy studies applied the MIC checkerboard method of CRS400393 with 10 commercially available antibiotics, and the fractional inhibitory concentration index (FIC-I) was determined (Li et al., 2017). Metabolic labeling assays using [14C]acetate of Mabs ATCC19977 cultures either untreated or treated with benzothiazole amides at 2 and 10 times their MICs and lipid extraction and analysis were performed as described (Li et al., 2014).

### Preliminary Determination of Pharmacological Properties

ADMET assays were performed through NIAID Preclinical Services and Eurofins. Potential for cytotoxicity was monitored in HepG2 cells after 72 h growth in the presence of test compounds, either prepared as 10-point 3-fold serial dilutions or at a fixed 10µM concentration. Cell viability was measured using the CellTiter-Glo <sup>R</sup> Luminescent Cell Viability Assay (Promega) or via fluorescent image analysis. Cytotoxicity was expressed as IC<sup>50</sup> or as percent reduction of viable cells relative to a reference compound. Metabolic stability of compounds was evaluated via incubation with human liver microsomes at 37◦C and quantitative analysis of parent compound by LC-MS/MS in samples removed at 0, 5, 15, 30, and 45 min (Obach, 1999). For cytochrome P450 (CYP) inhibition assays, compounds were prepared as a 7-point dilution series and incubated with human liver microsomes in buffer containing 2 mM NADPH and probe substrate. After incubation at 37◦C for the optimal time (10– 60 min) the reactions were processed for quantitative LC-MS/MS analysis of probe substrate metabolites in order to calculate IC<sup>50</sup> values (Walsky and Obach, 2004). Thermodynamic solubility of compounds in aqueous media was determined by incubation in 50 mM potassium phosphate buffer (pH 7.4) for 24 h, followed by HPLC-UV detection of compound in the filtrate (Analiza, Cleveland, OH, United States).

#### Efficacy and Tolerability Assessment

In vivo assessment of efficacy and tolerability of benzothiazole amide CRS400226 was assessed in a granulocyte macrophage colony stimulated factor-knockout (GM-CSF KO) mouse model of Mabs infection (Gonzalez-Juarrero et al., 2012; De Groote et al., 2014b). This study was carried out at Colorado State University which maintains a centralized IACUC registered by the USDA and accredited by the Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC) International. The protocols were reviewed and approved by the CSU Animal Care and Usage Committee (ACUC) prior to development of the experimental infections. Male GM/CSF-KO mice were bred from the Gonzalez-Juarrero Lab colony at Colorado State University in the Painter Center for Laboratory Animal Resources. Mice at least 2 months old and at least 20 grams of weight were selected and acclimated for 1 week in the BSL3 laboratory prior to infection. Inoculum was prepared from a stock vial of Mabs strain #21, a clinical isolate, in 4 mL ddH2O to a targeted dose of 10<sup>6</sup> CFU/50 µL. The actual CFU in the inoculum used for infection was determined by plating of serial dilutions onto 7H11 agar plates, followed by colony enumeration after 4 days. The mice were infected via intratracheal aerosol delivery using a Penn-Century microsprayer. Actual bacterial deposition in the lungs was determined in three mice that were sacrificed to remove the lungs for bead homogenization in 500 uL of PBS, and CFU determination in serial dilutions of the homogenate. Established infections were confirmed after 10 days, again by determination of CFU in the lungs of three mice. Post infection, animals were monitored daily and weights recorded weekly, to ensure <20% weight loss and to adjust drug dosing to the actual weight. Four groups of n = 5 mice were included in the study, and treatment started 10 days post infection. The benzothiazole amide CRS400226 was formulated in saline containing 5% corn oil and 0.05% Tween-80. CRS400226 and vehicle were administered via intrapulmonary liquid aerosol delivery (50 µL/dose) via liquid Penn-Century MicroSprayer (liquid) 5 days per week over a 4-week treatment course. Azithromycin (TOCRIS #1A/203325) was prepared as a solution in 0.6% acetic acid and was administered via oral gavage of 0.1 mL to achieve a dose of 100 mg/kg. One group of mice served as untreated controls. After 4 weeks of treatment, the mice were sacrificed and the lung lobes separated. The left lobes of the lungs were homogenized in PBS and plated to determine bacterial load. The middle right lobes were placed in 4% paraformaldehyde and processed for histology.

### RESULTS

#### Improved Potency of Advanced Lead Compounds and *in vitro* Profiling

Following previous screening of compound collections against Mtb (Falzari et al., 2005; Franzblau et al., 2012), initial hits with modest activity against Mabs ATCC 19977 emerged from a whole cell activity screen of small molecule compound libraries. This led to the identification of scaffolds with broad antimycobacterial spectrum, such as the adamantyl amides. We applied an iterative process of design-synthesis-screen and increased potency by 1–2 orders of magnitude compared to the initial library hits (Graham et al., 2018). Optimization of the adamantyl amide series resulted in compound CRS400226, an adamantyl benzothiazole amide. Due to its high lipophilicity and consequent potential for nonspecific binding, the adamantyl group was eventually replaced with cyclohexyl derivatives, such as in CRS400153. Some loss of activity, particularly against slow-growing NTM, was noted when comparing the cyclohexyl derivatives to adamantyl compounds. A breakthrough in structure-activity relationship was achieved with bicyclic moieties, such as in CRS400359 and CRS400393, which resulted in improved potency against both rapid- and slow-growing NTM as well against Mtb. The properties of these advanced leads with regard to in vitro activity and ADMET profiles are shown in **Table 1**. The lead compounds demonstrated a very narrow, mycobacteria-specific spectrum of activity, without appreciable activity against other aerobic or anaerobic bacteria. In vitro activity and MIC distributions were further investigated for the top compounds against rapid-growing NTM clinical isolates. MIC ranges and MIC<sup>90</sup> values, defined as the minimum concentration required to inhibit growth of at least 90% of all strains tested, are shown in **Table 2**. Advanced lead compounds had MIC<sup>90</sup> of ≤1µg/mL against all mycobacteria tested, which included n = 20 Mabs, n = 11 M. chelonae, n = 11 M. fortuitum, and n = 12 other rapid-growing NTM. Outlier strains with elevated MIC values for benzothiazole amides were not found and the MIC ranges were tight, while commonly used antibiotics such as amikacin, linezolid, and azithromycin showed bimodal MIC distributions (**Supplementary Figure S1**). Reduced susceptibility (MIC ≥ 8µg/mL) to at least one of these three antibiotics was observed for 45 of 54 (83%) of the clinical isolates. The benzothiazole amides were equally active against drug-resistant strains, which is consistent with the absence of preexisting resistance to an agent directed against a novel target. Lead compound CRS400393 demonstrated MIC values ranging from ≤0.03 to 0.5µg/mL against rapidgrowing NTM, MIC = 1–2µg/mL against MAC, and MIC ≤ 0.12µg/mL against Mtb. Notably, all mycobacterial species were susceptible, including isoniazid, rifampin, or fluoroquinolone resistant strains of Mtb, and activity was maintained under low oxygen conditions and against intracellular Mtb, as shown for CRS400226 (**Supplementary Table S2**). In addition, lead compounds have so far shown favorable in vitro ADMET properties with regard to cytotoxicity, hemolysis, CYP inhibition, and metabolic stability, although in vivo safety and toxicity data have not yet been generated for the more potent analogs. Aqueous solubility was low (<3µM) and protein binding was high (>95%) for all compounds of this series, likely attributable to the highly hydrophobic moieties of the compounds.

#### Mode of Action and Target Identification

Benzothiazole amides showed bactericidal activity in time-kill assays. CRS400226 at concentrations of 25µM (10µg/mL) caused a reduction in Mtb of 3 Log<sup>10</sup> CFU within 7 days. CRS400393 was bacteriostatic against Mabs, for 3 days, followed by a drop of 2 Log<sup>10</sup> CFU after 5 days. Importantly, no regrowth was observed following a single addition of compound due to emergence of resistance (**Figure 1**). In vitro checkerboard synergy assays of the benzothiazole amides showed additive effects with many commonly used antimycobacterial agents (amikacin, ciprofloxacin, azithromycin, tobramycin, clofazimine, linezolid, and cefoxitin), and indifference when combined with doxycycline or bedaquiline (**Supplementary Table S3**). TABLE 1 | *In vitro* activity of advanced benzothiazole amides.


*<sup>a</sup>Efflux mutant strain of E. coli.*

*<sup>b</sup>Permeabilized strain of E. coli; PMBN, polymyxin B nonapeptide.*

*c Including genus Clostridium, Bifidobacterium, Fusobacterium, Peptostreptococcus, Bacteroides, Lactobacillus, Veillonella, Blautia, Parvimonas, Finegoldia, Coprococcus, Actinomyces, Eubacterium, and Prevotella.*

*<sup>d</sup>Clearance measured in human liver microsome assay.*

*NT, not tested.*

Mode-of-action studies were performed with CRS400153 and CRS400226, which represent the cyclohexyl and the adamantyl subseries of benzothiazole amides, respectively. Evidence was obtained that compounds from both subseries at 2x MIC and 10x MIC affect the transfer of mycolic acids to their cell envelope acceptors in both Mabs and Mtb, most likely through the inhibition of the trehalose monomycolate transporter, MmpL3. Indeed, metabolic labeling with [14C]acetate of Mabs ATCC19977 cultures either untreated or treated with benzothiazole amides at 2x and 10x MIC resulted in a concentration-dependent inhibition of mycolic acid transfer to both arabinogalactan and trehalose dimycolates, a hallmark of MmpL3 inhibitors (**Supplementary Figure S4**). Moreover, spontaneous Mabs ATCC19977 mutants resistant to CRS400153 were isolated with a frequency of 6 × 10−<sup>9</sup> at 4x MIC and were found to harbor non-synonymous mutations in


CRS400226 0.25–2 0.5 0.5–0.5 0.5 0.12–0.5 0.5 0.12–1 0.5 CRS400153 0.5–2 1 0.5–1 1 0.5–0.5 0.5 0.25–1 1 CRS400359 0.06–2 0.25 0.25–0.5 0.25 0.06–1 1 0.12–2 0.5 CRS400393 ≤0.03– 0.25 0.25 0.12–0.12 0.12 ≤0.03–0.5 0.25 0.12–0.5 0.5 Amikacin 1–>16 16 16–>16 >16 0.5–4 2 0.5–16 16 Linezolid 0.5–>8 >8 1–8 8 0.5–>8 >8 0.5–>8 >8 Azithromycin 0.12–>8 >8 0.12–0.5 0.5 0.12–>8 >8 0.12–>8 >8

TABLE 2 | *In vitro* Activity (MIC, µg/mL) of benzothiazole amides and control agents against rapid-growing NTM clinical Isolates.

*<sup>a</sup>Other rapid-growing NTM included two strains each of M. peregrinum, M. mucogenicum, M. massiliense, M. bolletii, M. phocaicum, and M. porcinum.*

*<sup>b</sup>MIC*<sup>90</sup> *defined as the minimum drug concentration that inhibited growth of* ≥*90% of the strains tested.*

MmpL3 (L551S, I306S, or A309P). These mutations were sufficient to confer reduced susceptibility (4 to >32-fold MIC increase) to CRS400153 and CRS400359 when introduced via recombineering into the isogenic background of Mabs ATCC19977. Moreover, testing of 15 additional analogs from the same compound subclass indicated that all of them displayed 8–64-fold reduced activity against CRS400153-resistant mutants harboring mutations in MmpL3. Interestingly, the same A309P and L551S mutations were previously reported to decrease 16- to 64-fold the susceptibility of M. abscessus to MmpL3 inhibitors of the indole-2-carboxamide and piperidinol series (Dupont et al., 2016; Franz et al., 2017) (unpublished data). The I306 residue has not previously been associated with resistance to MmpL3 inhibitors but maps in the fifth transmembrane region of the transporter, very close to A309 and to a conserved serine residue (S302 in M. abscessus, S288 in Mtb) shown to confer resistance to a variety of MmpL3 inhibitors in Mtb (Belardinelli et al., 2016). These data strongly suggest that compounds from the structural class of benzothiazole cyclohexyl amides such as CRS400153 kill M. abscessus through the direct or indirect inhibition of MmpL3. The fact that CRS400153 showed no effect on the membrane potential and electrochemical pH gradient of Mtb intact cells and inverted membrane vesicles (data not shown) supports a direct mechanism of inhibition of the transporter (Li et al., 2018).

#### *In vivo* Efficacy and Tolerability

Several compounds were screened in mouse PK and tolerability studies to evaluate various doses and routes of administration. Although essentially insoluble in water, we found CRS400226 was soluble in corn oil at concentrations up to 250 mg/mL. Oral dosing of this compound at 100 mg/kg in female CD-1 mice resulted in very low plasma levels of 2.3µg/mL at 1 h and 2.0µg/mL at 8 h. Although highly tolerated, it became clear that the lipophilic nature and high protein binding of the compounds precluded oral routes of administration for efficacy. Instead, intrapulmonary delivery methods for treatment of lung infections such as NTM and Mtb were explored. CRS400226 was synthesized in multi-gram quantities to support in vivo studies and was further developed for this delivery pathway. These initial attempts to formulate CRS400226 as a corn oil/saline emulsion afforded the opportunity to obtain proofof-concept in vivo efficacy. We assessed CRS400226 against Mabs 21 in a 28-day mouse model of chronic NTM lung infection in granulocyte macrophage colony stimulated factorknockout (GM-CSF KO) mice, which represent a clinically relevant model of human infection (De Groote et al., 2014b). Actual dose deposited in the lungs was 4.53 Log<sup>10</sup> CFU, which increased to 4.98 Log<sup>10</sup> CFU after 10 days of infection, as determined in groups of three mice. Intratracheal administration of CRS400226 at 25 mg/kg once daily, 5 days per week for 4 weeks resulted in a statistically significant (p = 0.0005) reduction of 0.64 Log<sup>10</sup> CFU compared to the vehicle control (**Figure 2A**). Azithromycin at 100 mg/kg achieved 0.38 Log<sup>10</sup> CFU reduction when compared to the untreated group (p = 0.05), though the data points in the azithromycin group were more scattered than in the other groups. There was no statistical difference (p = 0.17) in the bacterial burden between the CRS400226 and azithromycin treated groups. No animals died during the 4 week course of the efficacy model, though all treatment groups demonstrated a slight decrease in weight ranging from 1.6 to 7.8%. CRS400226 was well tolerated up to 4 weeks of treatment, without any observed adverse side effects (**Figure 2B**). Histology showed that the azithromycin treated animals had only minimal pathologic evidence of inflammatory infiltrates with one animal displaying normal appearing lung tissue. CRS400226 treated animals had only very mild areas of airway-centric inflammation compared to the peribronchial inflammatory infiltrate seen in the vehicle treated group. The untreated and vehicle treated lung tissue demonstrated scattered nodules and more significant inflammatory infiltrates. In many of the animal lungs, evidence of bronchiectasis was observed and inflamed peri-bronchial lymphoid tissue (**Supplementary Figure S5**).

#### DISCUSSION

We have developed a lead series that exhibits broad antimycobacterial activity, which could revolutionize the treatment of NTM, while simultaneously adding much-needed novel agents to the Mtb treatment armamentaria. The strength of the benzothiazole amide series is that advanced lead compounds are very potent in vitro and mid-stage compounds showed in vivo efficacy. Good potency of the benzothiazole amides was demonstrated against a collection of rapid growing NTM clinical isolates which included many strains that were resistant to amikacin, linezolid, or azithromycin illustrating the high prevalence of non-susceptibility of NTM to many antibiotics commonly used to treat NTM infections and the need for novel agents whose activity is not affected by existing resistance mechanisms. Apparent MIC values against slow-growing NTM and Mtb were 3higher compared to the MIC values seen against rapid-growing NTM. This has been attributed to differences in the media used to test the different mycobacterial species, specifically to the presence of albumin (5 mg/mL) in the OADC supplement used for slow- growing NTMs and for Mtb, causing an MIC shift of the relatively highly (>95%) protein-bound compounds (**Supplementary Table S6**). CRS400393 showed additivity and no antagonism when combined with other antimycobacterial agents, an essential property for treating chronic NTM infections that often require combination drug regimens. Another favorable property of the benzothiazole amides is their narrow spectrum, since such a mycobacterialspecific agent may prove more tolerable for prolonged duration of therapy, i.e., by not affecting the beneficial gut microbiome that protects the host from gastroenteric side effects such as Clostridium difficile infections that often arise from exposure to broad-spectrum antibiotics.

The apparent target of benzothiazole amides, MmpL3, is a mycolate transporter essential for assembly of the outer membrane and has high potential to improve and shorten current

FIGURE 2 | GM-CSF KO mouse model of *Mabs* lung infection demonstrating efficacy (A) and tolerability (B) of CRS400226. The Penn Century microsprayer was used for delivery of 5 × 10<sup>5</sup> to 1 × 10<sup>6</sup> bacteria to the lungs. Starting 10 days later, the test compound was dosed once daily over 4 weeks by intratracheal microspray administration as 5% corn oil/saline emulsion. CRS400226 at 25 mg/kg daily for 28 days resulted in a statistically significant (*p* = 0.0005) reduction of 0.64 Log10 CFU compared to the vehicle control.

drug-susceptible and drug-resistant tuberculosis chemotherapies (Li et al., 2016, 2018). Several MmpL3 inhibitors with different chemical scaffolds have been reported (Poce et al., 2013, 2016; Li et al., 2014; Rayasam, 2014; Stec et al., 2016). One compound, SQ109, which is unrelated to our compound series, has been in clinical development through Phase II (Sacksteder et al., 2012; Tahlan et al., 2012). The nature of the membrane-bound MmpL3 target favors lipophilic and therefore poorly watersoluble inhibitors. Efforts to optimize oral drug-like properties such as improving compound solubility have largely resulted in loss of activity. This potential limitation had been discovered through detailed SAR of the early lead compound CRS400153, by exploring hydrophilic substitutions at the 4-, 5-, 6-, and 7-position of the benzothiazole moiety, such as a hydroxyl, tertiary amine, and ether or ester groups. In all cases, these hydrophilic substitutions resulted in MIC > 64µg/mL against M. abscessus compared to MIC = 0.5µg/mL for the 4,6 diF-benzothiazole analog CRS400153, except for the 7-OH benzothiazole analog which retained some activity with an MIC range of 2–16µg/mL against M. abscessus (compound #37) (Graham et al., 2018). Our lead compounds are very potent, but have very low solubility (≤3µM), high protein binding (MIC ≥ 16µg/mL in 50% serum) and low oral bioavailability due to their lipophilic nature, precluding their utility as oral agents. Therefore, further development will focus on inhalational therapy for NTM first and possibly development of a topical formulation later. Similar approaches have been reported for aerosolized liposomal amikacin (Rose et al., 2014) and for inhaled dry powder capreomycin with the potential of rendering patients non-infectious (Dharmadhikari et al., 2013). Our proofof-concept efficacy study used corn oil for formulation, and we noted an increased bacterial burden in the vehicle treated group compared to the untreated group, which is likely attributable to corn oil being a carbon source for mycobacteria, increasing their replication; hence, developing a more sophisticated formulation will be paramount.

The desired target product profile for this drug includes (1) broad-spectrum coverage of the clinically-relevant mycobacteria Mabs, MAC and Mtb, (2) bactericidal activity, (3) antibacterial potency equal or better compared to antibiotics currently used in the standard of care for NTM infections, (4) suitability for inhaled therapy via aerosol, (5) dosing frequency preferably once or twice a day, (6) excellent safety profile suitable for treating chronic infections and for pediatric use, (7) low propensity for resistance development, (8) compatibility with other antibacterial drug cocktails, and (9) shelf-life of at least 2 years. Our data indicate that the benzothiazole amides already meet the criteria regarding the in vitro antimyobacterial activity and low resistance frequencies. The next steps will be the evaluation of the compounds regarding efficacy, safety, tolerability, and chemical stability. Cytotoxicity studies, however, have been limited by the low solubility of the compounds, which made testing at high concentrations impossible, and potential effect may be masked due to the high protein binding of the compounds.

In conclusion, the limitations of the scaffold are high lipophilicity, high protein binding, and poor PK with oral dosing. Such challenges in drug development can often be overcome by exquisite potency of the agents and improved formulation and delivery methods. A recent example of an antimycobacterial drug with some undesirable physicochemical properties is bedaquiline, which nonetheless was developed successfully and became the first FDA-approved TB drug in 40 years (Palomino and Martin, 2013).

In vivo efficacy and tolerability of mid-stage lead compound CRS400226 provides exciting evidence of the potential for this series; new analogs have already shown further potency increases, providing a solid foundation for further optimization. Next steps will include testing of pharmacokinetic properties, metabolic stability, toxicity and efficacy of the most recent potent compounds such as CRS400393 against NTM and Mtb as inhaled therapy, hopefully leading to IND-enabling studies.

#### AUTHOR CONTRIBUTIONS

MD, UO, XS, JD, and TJ conceived the project and designed the strategy. Experiments were carried out by CW, JG, JD, and XS (chemical synthesis and SAR), TH, CY, and UO (in vitro susceptibility and other microbiological assays), WL and MJ (mode-of-action studies) and MG-J (in vivo efficacy). WR was Project Manager and assisted with data analysis. MD and UO wrote the manuscript.

### FUNDING

Initial funding was obtained from the American Lung Association and NTM Info & Research (NTMir) and the Cystic Fibrosis Foundation through grants awarded to MD (CSU), followed by funding from NIH via SBIR Phase I and II grants GM110848 to Crestone, Inc. (PIs TJ, MD, and XS) and AI116525 to MJ and MG-J. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.

#### ACKNOWLEDGMENTS

We thank Scott Franzblau (Univ. Illinois at Chicago) for providing screening hits. We are grateful to Jim Boyce at NIAID preclinical services (Task Order A19) for performing in vitro activity and ADMET assays. NTM strains were provided by Denver Health, and Mtb H37Rv mc<sup>2</sup> 6206 was provided by Dr. Bill Jacobs (Albert Einstein College of Medicine, NY).

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2018.02231/full#supplementary-material

#### REFERENCES


age in Mycobacterium avium complex lung disease: a cross-sectional analysis of 235 patients. Respir. Res. 16:145. doi: 10.1186/s12931-015-0304-5


transporter from Mycobacterium tuberculosis. ACS Infect. Dis. 2, 702–713. doi: 10.1021/acsinfecdis.6b00095


screen of 70 host markers in a randomized clinical trial. EBioMed. 25, 112–121. doi: 10.1016/j.ebiom.2017.10.018


**Conflict of Interest Statement:** UO, XS, JD, TJ, CW, JG, JD, XS, TH, and CY are employees of Crestone, Inc.

The remaining 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 De Groote, Jarvis, Wong, Graham, Hoang, Young, Ribble, Day, Li, Jackson, Gonzalez-Juarrero, Sun and Ochsner. 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.

# *Mycobacterium abscessus* and β-Lactams: Emerging Insights and Potential Opportunities

Elizabeth Story-Roller <sup>1</sup> , Emily C. Maggioncalda<sup>1</sup> , Keira A. Cohen<sup>2</sup> and Gyanu Lamichhane<sup>1</sup> \*

*<sup>1</sup> Division of Infectious Diseases, School of Medicine, Johns Hopkins University, Baltimore, MD, United States, <sup>2</sup> Division of Pulmonary and Critical Care Medicine, School of Medicine, Johns Hopkins University, Baltimore, MD, United States*

β-lactams, the most widely used class of antibiotics, are well-tolerated, and their molecular mechanisms of action against many bacteria are well-documented. *Mycobacterium abscessus* (*Mab*) is a highly drug-resistant rapidly-growing nontuberculous mycobacteria (NTM). Only in recent years have we started to gain insight into the unique relationship between β-lactams and their targets in *Mab*. In this mini-review, we summarize recent findings that have begun to unravel the molecular basis for overall efficacy of β-lactams against *Mab* and discuss emerging evidence that indicates that we have yet to harness the full potential of this antibiotic class to treat *Mab* infections.

#### *Edited by:*

*Thomas Dick, Rutgers, The State University of New Jersey, Newark, United States*

#### *Reviewed by:*

*Marcel Behr, McGill University, Canada Bavesh Davandra Kana, University of the Witwatersrand, South Africa*

> *\*Correspondence: Gyanu Lamichhane lamichhane@jhu.edu*

#### *Specialty section:*

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

*Received: 30 May 2018 Accepted: 05 September 2018 Published: 25 September 2018*

#### *Citation:*

*Story-Roller E, Maggioncalda EC, Cohen KA and Lamichhane G (2018) Mycobacterium abscessus and* β*-Lactams: Emerging Insights and Potential Opportunities. Front. Microbiol. 9:2273. doi: 10.3389/fmicb.2018.02273* Keywords: *Mycobacterium abscessus*, β-lactams, peptidoglycan, LD-transpeptidase, β-lactamase inhibitor

### INTRODUCTION

Although Mycobacterium abscessus (Mab) was first discovered in 1953 (Moore and Frerichs, 1953), it was only recently that genomic sequencing differentiated the Mab complex into three subspecies: M. abscessus sensu stricto, M. abscessus subsp. bolletii, and M. abscessus subsp. massiliense (Adekambi et al., 2004, 2006; Viana-Niero et al., 2008). These subspecies exhibit differential susceptibilities to certain antibiotics and differential clinical outcomes.

Mab can cause pulmonary disease in addition to skin and soft tissue infections, lymphadenitis, and disseminated disease. Mab is sometimes considered a respiratory colonizer; however, in the setting of immunosuppression or structural lung disease, such as cystic fibrosis (CF) and bronchiectasis, Mab can cause chronic pulmonary disease. In CF patients, Mab infections are often incurable and associated with rapid lung function decline (Griffith et al., 2007; Esther et al., 2010; Benwill and Wallace, 2014). The cure rate for Mab lung disease is only 30–50% (Jarand et al., 2011), with a recent review reporting sputum culture conversion rates as low as 25% with antibiotic treatment alone (Diel et al., 2017).

Poor treatment outcomes of Mab infection have been ascribed to both innate and acquired drug resistance. Mab is intrinsically resistant to multiple antibiotic classes which has been attributed to various factors (Brown-Elliott and Wallace, 2002; Nessar et al., 2012; van Ingen et al., 2012). Acquired resistance has further limited therapeutic options (Flume, 2016). Current treatment regimens are suboptimal, as they require several months of intravenous multidrug therapy with potentially cytotoxic antibiotics and produce poor outcomes (Wallace et al., 1985; Floto et al., 2016).

In this review, we will briefly summarize Mab treatment recommendations, discuss unique molecular targets of β-lactams in Mab, and highlight emerging insights into how β-lactams may be leveraged to treat individuals infected with Mab.

**129**

### CURRENT *Mab* TREATMENT RECOMMENDATIONS

The US Cystic Fibrosis Foundation and European Cystic Fibrosis Society recently developed consensus guidelines for management of Mab lung disease in CF patients (Floto et al., 2016). Similar to tuberculosis, Mab infection is treated with multidrug regimens divided into an intensive phase, followed by a continuation phase. Per recent guidelines, the intensive phase of Mab therapy should consist of an oral macrolide, combined with 3–12 weeks of intravenous amikacin, plus at least one of the following: intravenous cefoxitin, imipenem, or tigecycline (Floto et al., 2016). Guidelines for the continuation phase include a daily oral macrolide, inhaled amikacin, and two to three additional oral antibiotics, including minocycline, clofazimine, moxifloxacin, and linezolid.

Macrolides have historically been considered the backbone of treatment against Mab. They have relatively low toxicity, are orally bioavailable (Griffith et al., 2007; Floto et al., 2016), and exhibit consistent activity against Mab in vitro (Griffith et al., 2007). However, subspecies abscessus and bolletii harbor a functional erm(41) gene, which confers inducible macrolide resistance and can limit the effectiveness of this drug class. In contrast, subspecies massiliense carries a non-functional erm(41) gene (Nash et al., 2009), thus cannot exhibit inducible macrolide resistance and is associated with improved outcomes on macrolide-based regimens (Koh et al., 2011). Consequently, the CF guidelines recommend subspeciation of Mab complex, which many clinical laboratories are not equipped to perform routinely. Therefore, some CF centers prescribe initial treatment regimens comprised of intravenous amikacin plus either cefoxitin or imipenem, rather than a macrolide (Philley et al., 2016).

Cefoxitin and imipenem are currently the only two β-lactams included in the guidelines for treatment of Mab infections. This antibiotic class has been largely understudied against Mab and may be a potential untapped resource in combating this highlyresistant microbe.

### MECHANISM OF ACTION OF β-LACTAMS AGAINST *Mab*

β-lactams are the most widely-used antibiotic class to treat bacterial infections (Hamad, 2010) and their safety and efficacy profiles have been well-established. There are five subclasses of β-lactams currently available in the clinical setting: penicillins, cephalosporins, monobactams, carbapenems, and penems. βlactams have been studied extensively for treatment of drugresistant Mycobacterium tuberculosis (Mtb) infections, which is summarized elsewhere (Story-Roller and Lamichhane, 2018). Certain β-lactam subclasses also exhibit activity against Mab (Lavollay et al., 2014; Kaushik et al., 2015; Lefebvre et al., 2016). While initial insights into the molecular mechanism of action of β-lactams against mycobacteria were gleaned largely from Mtb, recent studies have begun to elucidate the relationship between Mab and β-lactams (Lavollay et al., 2014; Lefebvre et al., 2016; Kumar et al., 2017a).

β-lactams exert their activity by inhibiting synthesis of an essential component of the bacterial cell wall, the peptidoglycan (PG) (Hartmann et al., 1972). The building block of PG is a disaccharide with a stem peptide comprised of four or five amino acids; specifically N-acetyl-glucosamine-N-acetyl-muramic acid-L-alanyl-D-glutaminyl-meso-diaminopimelyl-D-alanyl-D-alanine in Mab (Lavollay et al., 2011). Polymerization

of disaccharides by transglycosylases and stem peptides by transpeptidases produces a three-dimensional macromolecule, the PG (**Figure 1**).

The dominant model of PG architecture was largely established by studies using model organisms, such as E. coli. According to this historical model, the final step of PG synthesis is catalyzed by D,D-transpeptidases (DDT), also known as penicillin binding proteins, which link the 4th amino acid of one stem peptide to the 3rd amino acid of the adjacent stem peptide, thereby generating a 4 →3-linked peptide network. However, as early as 1974, it became clear that the chemical architecture of mycobacterial PG, and therefore the enzymes necessary for its synthesis, were distinct from those described in the historical model. This study reported that stem peptides in the PG of M. smegmatis, Mtb, and M. bovis BCG were predominantly crosslinked with non-canonical linkages between the 3rd amino acid of one peptide and the 3rd amino acid of another (Wietzerbin et al., 1974). That same year, this group also demonstrated that the enzyme, L,D-transpeptidase (LDT), generated these 3 →3 linkages in Streptococcus faecalis (Coyette et al., 1974).

The first direct evidence demonstrating that stem peptides in Mab PG are predominantly cross-linked by 3 →3 linkages was reported in 2011 (Lavollay et al., 2011). Subsequently, five putative LDTs, LdtMab1−<sup>5</sup> , were identified in Mab (Mattoo et al., 2017) and the first crystal structure of one of these enzymes, LdtMab2, was described (Kumar et al., 2017a). These studies confirmed that Mab utilizes both LDTs and DDTs to generate 3 →3 and 4 →3 linkages between stem peptides, respectively (**Figure 1**). The majority of linkages in Mab are 3 →3, which suggests that LDTs are at least as important as DDTs for synthesis of its PG. Several studies report that genes involved in PG synthesis and remodeling are largely conserved across mycobacteria, implying a similar PG chemical composition, architecture, and metabolism (Sanders et al., 2014; Mattoo et al., 2017). A review of PG biosynthesis in Mtb by Pavelka et al. is recommended for further insight into mycobacterial PG biology (Pavelka et al., 2014).

### LDTs ARE PREFERENTIALLY INHIBITED BY CARBAPENEMS AND CEPHALOSPORINS

β-lactams mimic the C-terminal end of the native stem peptide of PG, bind to the active site of transpeptidases, and irreversibly inhibit their enzymatic activity (Park and Strominger, 1957). As the historical model considered DDTs to be the only enzymes that synthesized PG, they were assumed to be the sole targets of βlactams. The discovery of LDTs (Mainardi et al., 2005) prompted inquiry into whether β-lactams also interacted with this enzyme

class. Subsequent studies have demonstrated that LDTs and DDTs of mycobacteria differ in their binding affinities to β-lactam subclasses and are consequently inhibited by different subclasses to varying degrees (Dubee et al., 2012; Kumar et al., 2017b). DDTs are effectively bound and inhibited by all β-lactam subclasses, whereas Mab LDTs are preferentially bound and inhibited by carbapenems and to a lesser extent by cephalosporins (Kumar et al., 2017a,b).

Although the crystal structures of LDTs of Mab bound to β-lactams are not yet available, several groups have reported crystal structures of LDTs of Mtb bound to carbapenems and penems (Kim et al., 2013; Li et al., 2013; Bianchet et al., 2017; Kumar et al., 2017b; Steiner et al., 2017). As LDTs and DDTs of Mab are differentially inhibited by β-lactam subclasses, comprehensive inhibition of PG synthesis will likely require simultaneous administration of multiple β-lactams belonging to different subclasses to optimally inhibit the two enzyme classes.

### FACTORS THAT DETERMINE POTENCY OF β-LACTAMS AGAINST *Mab*

The major molecular factors limiting effectiveness of β-lactams against Mab are β-lactamase activity and the bacterial cell wall. Factors that commonly affect other antibiotic classes, including poor permeability of the cellular envelope, low affinity of antibiotic targets, drug efflux pumps, and chromosomallyencoded neutralizing enzymes, have been elegantly summarized elsewhere (Nessar et al., 2012; van Ingen et al., 2012).

### β-Lactamases

The potent activity of the chromosomally-encoded β-lactamase, BlaMab, is primarily responsible for poor efficacy of β-lactams against Mab (Soroka et al., 2014). β-lactamases hydrolyze the β-lactam ring, thereby inactivating these antibiotics (Kasik et al., 1971). Not only does BlaMab degrade several β-lactams with significantly higher efficiency than BlaC of Mtb, BlaMab

is not effectively inhibited by common β-lactamase inhibitors (BLI) clavulanate, tazobactam, and sulbactam (Soroka et al., 2017); agents that inhibit BlaC of Mtb (Wang et al., 2006). The observation that these BLIs do not reduce the minimum inhibitory concentration (MIC) of β-lactams against Mab in a whole-cell assay (Kaushik et al., 2017) is additional confirmation that β-lactamase activity in Mab is more robust than in Mtb. Subspecies massiliense harbors an additional β-lactamase, BlaMmas (Ramirez et al., 2017).

BlaMab is inactivated by avibactam (Dubee et al., 2015a), a recently-developed BLI whose core chemical composition differs from older BLIs and lacks a β-lactam ring (Coleman, 2011). Observations that avibactam reduces the MIC of several βlactams against Mab provides further validation of its efficacy against both the BlaMab protein and whole-cell Mab (Dubee et al., 2015a; Kaushik et al., 2017; Lefebvre et al., 2017). A recent study showed avibactam not only inhibits β-lactamases but also inhibits LDTs (Edoo et al., 2018). A recombinant Mab strain lacking blaMab exhibited increased sensitivity to β-lactams and was rendered susceptible to amoxicillin and ceftaroline (Lefebvre et al., 2016). This study also observed that β-lactams plus avibactam exhibited similar efficacy against the parental Mab strain as compared to each drug alone against 1blaMab, suggesting that avibactam fully inhibits BlaMab. While BlaMab and BlaMmas hydrolyze penicillins and cephalosporins with similar efficacy, BlaMmas also exhibits mild carbapenemase activity, a potential concern as it suggests an evolutionary movement toward β-lactamases with extended spectra (Ramirez et al., 2017). This study also noted that BlaMmas is structurally similar to other acquired carbapenemases normally found in gram negative bacteria, such as KPC-2 and SFC-1. Avibactam activity against BlaMmas has not yet been assessed and further study is warranted.

## Cell Wall

Mycobacteria possess an unusually thick cell wall composed of layers of complex hydrophobic molecules including fatty acids, mycolic acids, lipoproteins, glycopeptidolipids (GPL), and largely insoluble PG and arabinogalactan layers. Although poorlyunderstood in Mab, epigenetic factors generating differential levels of these molecules, especially GPLs, are associated with two distinct colony morphotypes—rough and smooth—within a clonal population. The rough morphotype tends to be associated with higher rates of antimicrobial resistance, including against β-lactams (Cangelosi et al., 1999; Greendyke and Byrd, 2008; Lavollay et al., 2014). Additionally, glycosylation of lipoproteins limits permeability of the cell wall to antibiotics that inhibit PG synthesis (Becker et al., 2017). Cell wall porins are also partially responsible for β-lactam resistance, as they allow transport of small hydrophilic molecules across the membrane, which interact with targets within the cytoplasm to potentially activate expression of drug resistance genes (Nguyen and Thompson, 2006; Nessar et al., 2012).

### ACTIVITY OF β-LACTAMS AGAINST *Mab*

We identified thirty-five studies with documented MIC ranges of β-lactams against clinical isolates of Mab globally (**Table 1**). These data serve to highlight the high degree of variability in observed MIC ranges among clinical isolates, even within each study, and this variability is partially why standardized treatment regimens against Mab are often not practical in the clinical setting. Imipenem and cefoxitin were the most commonlytested β-lactams and nearly all studies included Mab strains that were resistant to these agents based on established MIC breakpoints (Woods et al., 2011). Only in four studies were all strains susceptible or intermediate to cefoxitin (Lee et al., 2012; Lavollay et al., 2014; Singh et al., 2014; Jeong et al., 2017). Two studies performed subspeciation and observed that all strains of subspecies massiliense and/or bolletii were either susceptible or intermediate to imipenem, whereas subspecies abscessus exhibited higher MICs to this drug (Lavollay et al., 2014; Singh et al., 2014). The reason for this is not currently known. Although seventeen studies also evaluated additional β-lactams, it is evident that this antibiotic class is largely understudied against Mab.

#### FURTHER POTENTIATION OF β-LACTAMS AGAINST *Mab* BY BLIs

Several studies have investigated the ability of BLIs to potentiate β-lactams against Mab, both in vitro and in vivo. The combination of amoxicillin and avibactam effectively reduced abscess formation and prolonged survival of zebrafish infected with Mab reference strain ATCC 19977 compared to amoxicillin alone (Dubee et al., 2015a). A subsequent study found that a combination of imipenem and avibactam also prolonged zebrafish survival compared to imipenem alone (Lefebvre et al., 2017). Avibactam also decreases the MIC of ceftaroline against Mab (Dubee et al., 2015b). Combinations of carbapenems and avibactam against clinical isolates of Mab showed that avibactam reduced MICs to therapeutically-achievable levels (Kaushik et al., 2017). The greatest MIC reductions were noted with tebipenem, ertapenem, and panipenem; demonstrating that avibactam can successfully overcome β-lactamase activity and further suggests that carbapenems, especially those developed after imipenem, such as doripenem, biapenem and tebipenem, have untapped potential for use against Mab (Kaushik et al., 2017).

### SYNERGY STUDIES WITH β-LACTAMS AND OTHER DRUGS

As combination regimens are essential for clinical management of Mab infections, several studies have evaluated antibiotic synergy against Mab with mixed results (Cremades et al., 2009; Shen et al., 2010; Bastian et al., 2011; Choi et al., 2012; van Ingen et al., 2012; Oh et al., 2014; Singh et al., 2014; Ferro et al., 2016; Mukherjee et al., 2017; Aziz et al., 2018; Pryjma et al., 2018; Schwartz et al., 2018; Zhang et al., 2018). In vitro studies have shown variable synergy of β-lactams in combination with other drugs. One study found no evidence of synergy among combinations of either imipenem or ertapenem with various other antibiotics (Cremades et al., 2009). However, another study reported high levels of synergy against Mab clinical isolates when clofazimine and amikacin were combined with several β-lactam subclasses (Schwartz et al., 2018). In a final study, rifampin combined with either doripenem or biapenem significantly reduced the MICs of both drugs to within therapeutic levels, compared with each carbapenem alone (Kaushik et al., 2015).

## DUAL β-LACTAMS FOR *Mab*

Given that different subclasses of β-lactams target distinct aspects of mycobacterial cell wall biosynthesis, Mab regimens that contain two β-lactams from different subclasses may have high efficacy in Mab. As mentioned above, mycobacterial DDTs are inhibited by all β-lactams, whereas LDTs are preferentially inhibited by carbapenems and cephalosporins (Kumar et al., 2017a,b). A combination of cefdinir and doripenem was observed to be synergistic against Mab 19977 (Kumar et al., 2017a), demonstrating that dual β-lactams have therapeutic potential against Mab. This promising finding warrants further investigation into the effects of dual β-lactams against clinical isolates of Mab, further potentiation with BLIs, and additional in vivo studies.

### PRECLINICAL MODELS AND CLINICAL TRIALS

At least two groups have taken initiatives to develop animal models of Mab infection (Lerat et al., 2014; Obregon-Henao et al., 2015). Two studies have assessed efficacy of antibiotic treatment of mice infected with Mab, one of which included a β-lactam, cefoxitin. Lerat et al. assessed regimens containing clarithromycin, amikacin, or cefoxitin monotherapy vs. a threedrug combination in nude mice infected with Mab ATCC 19977. Cefoxitin monotherapy was equally effective as triple therapy, resulting in prolonged survival and reduced splenic bacillary


*(Continued)*

September 2018 | Volume 9 | Article 2273


TABLE

1


Continued

loads compared to untreated controls (Lerat et al., 2014). Several clinical trials assessing efficacy of non-β-lactam antibiotics against NTMs have been undertaken (clinicaltrials.gov). To date, there are no published clinical trials that have specifically investigated β-lactams for the treatment of Mab; however, we are hopeful that an increasing awareness of β-lactams as viable treatment options may lead to clinical trials with this class in the future.

#### FUTURE DIRECTIONS AND CONCLUSIONS

There is a dearth of research exploring β-lactams as potential treatments for Mab. Given the increasing prevalence of highly drug-resistant Mab isolates leading to poor clinical outcomes, new therapeutic approaches are needed to adequately treat these infections. Given our understanding of the differential mechanisms of β-lactam subclasses, and the ability of certain BLIs to overcome β-lactamase activity, currently-available β-lactams are a largely untapped resource for Mab treatment. Of the βlactam subclasses, carbapenems/penems have the greatest activity against Mab, followed by cephalosporins, then penicillins. As noted above (Kumar et al., 2017a), it is likely that combinations of different β-lactam subclasses are required to fully inhibit PG synthesis in Mab. This insight may partially explain why prior studies evaluating β-lactams individually have not shown significant efficacy against this microbe. Further investigation may identify novel treatments utilizing combinations of βlactams that optimally inhibit the distinct enzymatic targets present in Mab.

Appropriate selection of companion BLIs is another area in which β-lactams can be potentiated for use against Mab. Several studies have demonstrated efficacy of the BLI avibactam in inhibiting BlaMab activity, which is a major factor contributing to the high MIC of most β-lactams against Mab. However, avibactam is currently only available as a coformulated combination with ceftazidime, which itself does not appear to

#### REFERENCES


have activity against Mab (Dubee et al., 2015a; Kaushik et al., 2017). If avibactam were to be made available as an individual formulation, this would significantly increase its clinical usefulness, as regimens could be tailored to combine it with any β-lactam shown to be effective against a particular microbe or strain. Recently, two novel carbapenem-BLI combinations have been developed. These are meropenemvaborbactam, which was recently FDA-approved for use against gram-negative organisms, and imipenem-relebactam, which is currently in phase II clinical trials (Zhanel et al., 2018). There are no published studies assessing efficacy of these BLIs against Mab, but their coformulation with carbapenems may confer greater potential for clinical use and further studies with these drugs are certainly warranted. It is possible that β-lactam-BLI combinations will become integral to effective treatment of drugresistant Mab in the future. Additional animal studies as well as clinical trials with this drug class will be essential for the development of novel treatment regimens with improved clinical outcomes. Furthermore, repurposing already FDA-approved βlactams for use against Mab may allow for expedited clinical implementation of regimens that show promise in preclinical models.

#### AUTHOR CONTRIBUTIONS

ES-R, EM, KC, and GL discussed relevant literature. ES-R and KC focused on clinical aspects of the literature and EM focused on the basic biology. ES-R, EM, KC, and GL wrote the manuscript.

#### ACKNOWLEDGMENTS

This work was supported by the Cystic Fibrosis Foundation grant LAMICH17GO and NIH grant R21 AI121805. ES-R was supported by the NIH T32 AI007291. The content is solely the responsibility of the authors and does not necessarily represent the official views of the Cystic Fibrosis Foundation or the National Institutes of Health.

antibiotic resistance of Mycobacterium abscessus. Front. Microbiol. 8:2123. doi: 10.3389/fmicb.2017.02123


**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 Story-Roller, Maggioncalda, Cohen and Lamichhane. 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.

# Novel Screen to Assess Bactericidal Activity of Compounds Against Non-replicating Mycobacterium abscessus

Bryan J. Berube, Lina Castro, Dara Russell, Yulia Ovechkina and Tanya Parish\*

TB Discovery Research, Infectious Disease Research Institute, Seattle, WA, United States

Mycobacterium abscessus infections are increasing worldwide. Current drug regimens are largely ineffective, yet the current development pipeline for M. abscessus is alarmingly sparse. Traditional discovery efforts for M. abscessus assess the capability of a new drug to inhibit bacterial growth under nutrient-rich growth conditions, but this does not predict the impact when used in the clinic. The disconnect between in vitro and in vivo activity is likely due to the genetic and physiological adaptation of the bacteria to the environmental conditions encountered during infection; these include low oxygen tension and nutrient starvation. We sought to fill a gap in the drug discovery pipeline by establishing an assay to identify novel compounds with bactericidal activity against M. abscessus under non-replicating conditions. We developed and validated a novel screen using nutrient starvation to generate a non-replicating state. We used alamarBlue <sup>R</sup> to measure metabolic activity and demonstrated this correlates with bacterial viability under these conditions. We optimized key parameters and demonstrated reproducibility. Using this assay, we determined that niclosamide was bactericidal against non-replicating bacilli, highlighting its potential to be included in M. abscessus regimens. In contrast, most other drugs currently used in the clinic for M. abscessus infections, were completely inactive, potentially explaining their poor efficacy. Thus, our assay allows for rapid identification of bactericidal compounds in a model using conditions that are more relevant in vivo. This screen can be used in a highthroughput way to identify novel agents with properties that promise an increase in efficacy, while also shortening treatment times.

Keywords: Mycobacterium abscessus, nutrient starvation, high-throughput screen, drug discovery, nonreplicating, niclosamide

#### INTRODUCTION

Infections by nontubercular mycobacteria (NTMs), including Mycobacterium avium complex and Mycobacterium abscessus complex, are a growing public health concern (Nessar et al., 2012). Traditionally considered an opportunistic pathogen, M. abscessus predominantly causes pulmonary disease in patients with cystic fibrosis (CF), chronic obstructive pulmonary disease (COPD), or bronchiectasis. However, recent trends suggest M. abscessus infection among otherwise healthy

#### Edited by:

Thomas Dick, Rutgers, The State University of New Jersey, United States

#### Reviewed by:

Giovanna Poce, Università degli Studi di Roma "La Sapienza", Italy Umayal Lakshmanan, Agency for Science, Technology and Research (A∗STAR), Singapore

> \*Correspondence: Tanya Parish Tanya.Parish@idri.org

#### Specialty section:

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

Received: 18 May 2018 Accepted: 20 September 2018 Published: 10 October 2018

#### Citation:

Berube BJ, Castro L, Russell D, Ovechkina Y and Parish T (2018) Novel Screen to Assess Bactericidal Activity of Compounds Against Non-replicating Mycobacterium abscessus. Front. Microbiol. 9:2417. doi: 10.3389/fmicb.2018.02417

individuals is also on the rise with recent studies showing a 3% increase in prevalence per year (Prevots et al., 2010). Additionally, many cases likely go misdiagnosed as Mycobacterium tuberculosis infections due to their similarity in disease presentation and colony morphology upon microscopic examination of sputum smears. This not only underestimates the number of M. abscessus infections worldwide but also leads to improper antibacterial treatment (Brown-Elliott et al., 2012).

The increase in NTM infections is troubling in part due to NTMs being particularly refractory to antibiotic treatment with M. abscessus resistant to all front-line anti-tuberculosis drugs (Brown-Elliott and Philley, 2017). The American Thoracic Society/Infectious Disease Society of America guidelines state that no current drug regimens are proven efficacious against pulmonary M. abscessus infection (Griffith et al., 2007). Drug regimens can be tailored to individuals based on laboratory drug susceptibility testing and individual tolerance to drugs, but generally include a macrolide-based antimicrobial in combination with intravenously administered agents (CF Foundation Clinical Care Guidelines; Griffith et al., 2007). Current guidelines from the Center for Disease Control and Prevention list amikacin, cefoxitin, and clarithromycin as the leading drugs for treatment of M. abscessus infection (Lee et al., 2015). These regimens are not only ineffective (29–58% cure rate) but also lead to considerable toxic side-effects due to a typical course of treatment lasting 18–24 months (Jeon et al., 2009; Jarand et al., 2011; Koh et al., 2017; Wu et al., 2018). Antibiotic treatment alone also leads to higher relapse rates than highly invasive surgical procedures involving resection of infected tissue (Griffith et al., 1993; Jarand et al., 2011). These highly invasive procedures should be seen as a last resort, but due to the failure of current drug regimens, these procedures are often the only realistic path to a relapse-free cure (Griffith et al., 2007).

Failure of current drug regimens against M. abscessus is likely multi-factorial. One factor is intrinsic antibiotic resistance (Nessar et al., 2012). M. abscessus has a waxy cell wall composed of peptidoglycan, arabinogalactan, and long-chain mycolic acids, which is relatively impermeable to antibacterial compounds (Jarlier and Nikaido, 1990; Brennan and Nikaido, 1995). M. abscessus also possesses efflux pumps capable of exporting drug, as well as enzymes capable of modifying antibiotics or their targets (Nasiri et al., 2017). These factors make it difficult for drugs not only to enter the cell but also to remain effective if they get past the cell wall. In the few small-scale screens for compounds active against M. abscessus, screens suffered from low hit rates compared to other mycobacterial species (Chopra et al., 2011; Gupta et al., 2017; Low et al., 2017).

A major factor in the failure of current drug regimens is the diversity of niches and physiological states encountered by the bacterium during infection, which are not adequately captured with current in vitro model systems. The disease pathology of M. abscessus infection is similar to that of M. tuberculosis (Griffith et al., 2007; Koh, 2017). M. abscessus is phagocytosed by macrophages in the lung, but prevents phagosomal acidification and can survive for an extended time (Laencina et al., 2018). During this time physiological and transcriptional changes can lead to phenotypic resistance (tolerance) to drugs (Adams et al., 2011; Larsson et al., 2012). As the infection progresses, NTMs are contained in necrotic granulomas and eventually survive extracellularly in caseum, where they likely experience both low oxygen tensions and/or nutrient starvation (Wu et al., 2018). In M. tuberculosis, these conditions lead to a non-replicating state controlled by the DosR regulon, which is characterized by low metabolic activity and enhanced resistance to antibiotics (Wayne and Hayes, 1996; Xie et al., 2005; Leistikow et al., 2010; Sarathy et al., 2018). In this state, M. tuberculosis maintains a basal level of metabolic activity to ensure bacterial viability, and the electron transport chain (ETC) is required to maintain the proton-motive force and generate ATP (Rao et al., 2008; Cook et al., 2014).

The ability of M. abscessus to survive nutrient starvation or low oxygen tensions has not been established. However, M. abscessus does contain a DosR regulon (Gerasimova et al., 2011) and is known to survive in biofilms in the lung of CF patients (Qvist et al., 2015). Additionally, other NTMs have been shown to establish latent infections in patients with subsequent reactivation to active disease (Rosenmeier et al., 1991; Tsilimparis et al., 2014). Together with the similarity in disease progression to TB disease, we believe it is highly likely M. abscessus can survive in patients in a dormant state. Previous screens against M. abscessus specifically addressed drug activity against actively growing, metabolically active bacteria (Chopra et al., 2011; Gupta et al., 2017; Low et al., 2017). Such efforts will miss compounds which are active against non-replicating or metabolically inactive cells.

The current state of the drug pipeline for NTMs was recently reviewed (Wu et al., 2018), highlighting the need for novel screens to address different physiological conditions encountered by NTMs during infection, including nutrient starvation. Our study aimed to fill part of that gap by developing and validating a novel assay to identify compounds with bactericidal activity against non-replicating M. abscessus generated by nutrient starvation. This will allow screening of compound libraries for compounds with sterilizing capabilities under a physiological state likely to be relevant to in vivo infection. We demonstrate that a number of drugs currently employed to treat M. abscessus infection are inactive in this assay, highlighting the need for drug discovery and development campaigns specifically targeted against M. abscessus.

#### MATERIALS AND METHODS

#### Mycobacterium abscessus Culture

Mycobacterium abscessus strain # 103 was purchased directly from BEI Resources (NIAID, NIH) and was cultured in Middlebrook 7H9 medium containing 10% v/v OADC (Oleic Acid, Albumin, Dextrose, Catalase; Becton Dickinson) supplement and 0.05% v/v Tween 80 (7H9-OADC-Tw). Cultures were maintained standing at 37◦C. Viable bacteria were determined by colony forming units by plating 10-fold serial dilutions onto Middlebrook 7H10 agar plus 10% v/v OADC supplement. For starvation, mid-log cultures were harvested and resuspended in phosphate-buffered saline (PBS) + 0.05% v/v Tyloxapol (PBS-Tyl) at an OD<sup>590</sup> of 0.4. Cultures were incubated shaking at 100 rpm for 96 h at 37◦C before use. Cultures were harvested and resuspended in PBS-Tyl to an OD<sup>590</sup> of 0.5 and 50 µL was dispensed into 96-well plates using a Multidrop Combi.

#### Preparation of Assay Plates

fmicb-09-02417 October 8, 2018 Time: 15:42 # 3

Compound plates were prepared as described (Ollinger et al., 2013). In brief, 10 mM stock solutions in DMSO were prepared as a 10-point twofold serial dilution in 96-well plates using a Biomek 3000. For the final assay plate, 50 µL of PBS-Tyl was dispensed into each well of a 96-well plate (µCLEAR plates Greiner). Compounds or controls (2 µL of each) were added to each well using a Biomek 4000. The final assay plate contained DMSO as a negative inhibition control and 25 µM niclosamide as a positive control for bacterial killing (**Figure 1**). The final concentration of DMSO in all wells was 2%. Plates were inoculated with 50 µL of bacterial culture; 50 µL of PBS-Tyl was dispensed into column 12 (contamination control). Assay plates were incubated in a humidified incubator for 48 h at 37◦C after which 50 µL of 20% v/v alamarBlue <sup>R</sup> in PBS-Tyl (Bio-Rad) was added using a Multidrop Combi. Plates were incubated at 37◦C for 24 h and relative fluorescent units (RFUs) measured at Ex560/Em590 using a BioTek Synergy 4.

#### Data Analysis and Statistics

The mean value from the buffer-only wells (background) was subtracted. The coefficient of variance (CV) was calculated as:

$$CV = \frac{\frac{\text{SD}}{\sqrt{n}}}{AVG}$$

Z <sup>0</sup> was calculated by the following formula:

$$\text{Z}' = \frac{\left(\text{AVG}\_{\text{max}} - \frac{\text{3SD}\_{\text{max}}}{\sqrt{n}}\right) - \left(\text{AVG}\_{\text{min}} + \frac{\text{3SD}\_{\text{min}}}{\sqrt{n}}\right)}{\text{AVG}\_{\text{max}} - \text{AVG}\_{\text{min}}}$$

where AVG is the average RFU, SD is standard deviation, and n is the number of controls per plate. The signal-to-background (S/B) ratio is calculated as the mean of the maximum signal wells divided by the mean of the background (no bacteria) wells.

### RESULTS

#### Rationale and Establishment of Assay Parameters

We sought to develop a screen to identify compounds with bactericidal activity against non-replicating M. abscessus. We used nutrient starvation to generate a non-replicating state, since this is relevant in vivo and is technically feasible. We first established the ability of M. abscessus to survive under nutrient starvation. We grew M. abscessus to mid-log phase and resuspended in buffer to induce complete nutrient starvation. After 96 h, we adjusted the cultures to an OD<sup>590</sup> of 0.25 (∼2.5 × 10<sup>7</sup> CFU) and monitored survival. After a further 72 h, bacterial counts were 2.6 ± 0.4 × 10<sup>7</sup> CFU/mL (n = 6, three biological replicates), indicating M. abscessus is capable of surviving under non-replicating, nutrient starvation conditions. Our data confirmed that there was no increase in viable bacteria during this time and thus the bacilli are in a non-replicating state.

We needed a rapid measure of bacterial viability which was amenable to high throughput and required minimal manipulation. Since the ETC is essential under nutrientstarvation, we predicted that respiration would be a good measure of bacterial viability. We tested alamarBlue <sup>R</sup> , which uses the reducing power of the cell to convert resazurin to resorufin, since we could use fluorescence as the readout. We tested the correlation between bacterial viability, measured by optical density, and respiratory output measured by alamarBlue <sup>R</sup> turnover. We generated nutrient-starved bacteria after incubation in PBS-Tyl and then measured OD<sup>590</sup> and RFU for twofold serial dilutions, starting at an OD<sup>590</sup> of 1.0. Optical density correlated in a linear fashion to respiratory output up to an OD<sup>590</sup> of 0.25 (**Figure 2A** and **Supplementary Figure 1**). Above this level, the RFU signal flattened, likely due to a complete turnover of any alamarBlue <sup>R</sup> present in the wells or having reached the upper limit of detection of the microplate reader. We confirmed a linear correlation between fluorescence and colony forming units, which determined that a reduction of alamarBlue <sup>R</sup> signal by 90% is equivalent to ≥1-log reduction in CFU (**Figure 2B** and **Supplementary Figure 2**). We varied a number of parameters (data not shown) and selected the key parameters for the inoculum as 50 µL of a starting culture with a final OD<sup>590</sup> in the assay plate of 0.25 for assay validation. We established the remaining parameters of the assay as starvation for 96 h, exposure to compounds for 48 h, and incubation with alamarBlue <sup>R</sup> for 24 h. These parameters gave us sufficient signal to background and signal to noise ratios.

### Reference Compound Activity Under Nutrient–Starvation Conditions

We next sought a control compound which would have rapid bactericidal activity against M. abscessus under starvation conditions. We selected a number of compounds based on current treatment guidelines for M. abscessus infection, as well as compounds known to be active against non-replicating bacteria. Surprisingly, almost all the compounds we tested were inactive, including most of those currently used as part of M. abscessus drug regimens (**Table 1**). Only three of the tested compounds inhibited alamarBlue <sup>R</sup> turnover by at least 90%, i.e., MBC<sup>90</sup> (**Table 1**). Kanamycin and amikacin had MBC<sup>90</sup> values of 39 and 16 µM, respectively, while niclosamide was the most effective compound with a MBC<sup>90</sup> of 8.4 µM (**Table 1**) and maximal activity at 25 µM (**Figure 3**).

#### Confirmation of Assay Serving as a Measure of Bactericidal Activity

To confirm our screen's ability to serve as a measure of bacterial killing, we assayed M. abscessus survival over time in the presence of niclosamide. Concentrations of niclosamide below the MBC<sup>90</sup> (3.1and 6.3 µM) had minimal activity (**Figure 4A**), while concentrations above the MBC<sup>90</sup> had >1-log reduction in CFU

within three days; 25 µM reduced bacterial viability by 1.5-log CFU (**Figure 4A**). This data corresponds to the concentration– response curve seen in **Figure 3**, where 25 µM niclosamide caused the greatest reduction in alamarBlue <sup>R</sup> turnover. At higher concentrations (50 and 100 µM), niclosamide had reduced activity due to solubility issues. We measured alamarBlue <sup>R</sup> turnover each day. **Figure 4B** shows bacterial viability correlates in a linear fashion with RFU in our assay (R <sup>2</sup> = 0.59). Over time, niclosamide killed M. abscessus, and there was a concomitant decrease in alamarBlue <sup>R</sup> turnover (**Figure 4B**). These data confirm that alamarBlue <sup>R</sup> can be used as a readout for bacterial viability.

#### Validation of High-Throughput Screen

We tested the reproducibility of our screen by testing for intraexperiment and inter-experiment variability (Iversen et al., 2004). We ran three separate experiments each containing duplicate plates with full 96-well plates of the maximum signal (2% DMSO), minimum signal (25 µM niclosamide) and dose– response curves (niclosamide 10-point, twofold dilutions). Interand intra-assay variability were low (**Figure 5A**). The average CV across all DMSO control plates was 11%, while average CV for niclosamide plates was only 5% (**Table 2**), both well below the minimum standard for CV of 20% specified in the assay guidance manual (Iversen et al., 2004). The curves for niclosamide were consistent within and across runs (**Figure 5B**) with an average MBC<sup>90</sup> of 9.1 ± 3.2 µM and MBC<sup>50</sup> of 5.0 ± 0.9 µM across 48 replicates and an average Z<sup>0</sup> score for these plates of 0.84 ± 0.16 (n = 6; **Table 2**). Thus, the assay is reproducible and suitable for high throughput screening.

#### DISCUSSION

The drug development pipeline for M. abscessus is extremely sparse at a time when infections worldwide are on the rise


Nutrient-starved bacteria were exposed to compounds for 48 h, alamarBlue was added, and RFU measured after 24 h. MBC<sup>90</sup> is the concentration required to reduce alamarBlue turnover by 90% as determined by the Gompertz equation. Data are mean ± SD of at least two independent biological replicates.

and current drug regimens are highly ineffective (Griffith et al., 2007). Recently, a call was issued for the development of novel assays to assess the efficacy of compounds against M. abscessus under various physiological states with the hope these assays might lead to hits that are more physiologically relevant, and thus more likely to be successful in the clinic. In this work, we describe one such assay that specifically identifies compounds that are bactericidal to nutrient-starved M. abscessus.

Nearly all tested compounds were completely inactive against non-replicating M. abscessus (**Table 1**). This could be due to an inability of the compounds to penetrate the M. abscessus cell wall or to their inactivity under nutrientstarvation conditions. Surprisingly, we found only one of the drugs recommended by the Center for Disease Control and Prevention to treat M. abscessus was active in this assay. Amikacin, an aminoglycoside inhibitor of the 30S ribosomal subunit, had a MBC<sup>90</sup> of 16 µM, indicating this drug actively targets non-replicating M. abscessus. On the other hand, cefoxitin and clarithromycin were completely inactive (**Table 1**). Inactivity of cefoxitin is potentially explained by its mechanism of action. As a beta-lactam antibiotic, cefoxitin targets the cell wall of actively dividing cells, a potential liability against nonreplicating M. abscessus. Clarithromycin was also inactive in this assay, despite being a protein synthesis inhibitor like amikacin. One possible explanation is the known induction of clarithromycin resistance in some M. abscessus isolates (Benwill and Wallace, 2014; Koh, 2017). Whatever the mechanism, it is troubling two of the three recommended treatments for M. abscessus are inactive. It is tempting to speculate the general inactivity of the tested compounds against non-replicating M. abscessus may play a role in the failure of these drugs to adequately treat patients. More work needs to be done to assess their efficacy under other physiological states likely encountered by M. abscessus during infection.

We did identify two other compounds capable of killing nonreplicating M. abscessus, namely kanamycin and niclosamide. Kanamycin, like amikacin, is an aminoglycoside inhibitor of ribosomal translocation (Misumi and Tanaka, 1980) and is commonly used as a second-line drug in regimens to treat multidrug-resistant tuberculosis (MDR-TB). However, kanamycin has considerable negative side effects shared with amikacin (Lane et al., 1977; Jiang et al., 2017). First, kanamycin is not orally bioavailable; it is given as an injectable drug, making its delivery more difficult than drugs available as pills. Second, patients undergoing prolonged aminoglycoside treatment can experience nephrotoxicity, as well as significant hearing loss, and thus must undergo repeated testing to determine whether treatment can continue (Ariano et al., 2008; Al-Malky et al., 2015; Garinis et al., 2017). With current drug regimens for M. abscessus lasting 18–24 months, kanamycin is not an ideal drug to be included in a new regimen.

Niclosamide was the most effective compound in this assay, causing 1–2 log-CFU kill in just three days (**Figure 4**). Niclosamide, an anti-tapeworm drug, is an inhibitor of oxidative phosphorylation, potentially working by inhibiting production of adenosine triphosphate (ATP) (Frayha et al., 1997; Chen et al., 2018). Niclosamide is an essential medicine according to the World Health Organization and is available as a chewable tablet (World Health Organization [WHO], 2017), although there are issues with solubility, absorption, and distribution to tissues (Chen et al., 2018). Recent work has been done to improve pharmacokinetic (PK) and pharmacodynamic (PD) properties of niclosamide, which could lead to its use to treat various bacterial infections (Mook et al., 2015). Our work suggests niclosamide could be used as part of a drug regimen to treat M. abscessus. If PK/PD properties were suitable to provide adequate exposure in the

FIGURE 4 | Metabolic activity correlates with bactericidal activity. Nutrient-starved bacteria were exposed to compounds for 24–72 h, alamarBlue <sup>R</sup> was added and RFU measured after 24 h. Aliquots from the same samples were used to determine: (A) Bacterial viability (CFU; data are the average ± SD of 2 biological replicates); (B) metabolic activity.


TABLE 2 | Assay reproducibility and robustness testing.

The final assay was run in three independent runs containing two plates each of minimum signal (2% DMSO), maximum signal (25 µM niclosamide), and niclosamide CRCs with positive and negative controls. Nutrient-starved bacteria were exposed to compounds for 48 h, alamarBlue <sup>R</sup> was added, and RFU measured after 24 h. Z<sup>0</sup> , % CV, and signal/background (S/B) ratio were calculated for each experimental run.

lungs, niclosamide might be able to target non-replicating bacilli residing in granulomas or caseum and help clear the infection.

#### CONCLUSION

In conclusion, we have developed a new assay to find antimycobacterial agents with activity against non-replicating bacilli, where positive hits can be followed up for bactericidal activity by plating for CFU. This assay allows us to screen for compounds that are bactericidal within a short period (3 days), increasing the possibility that we can identify rapidly bactericidal agents that could have an impact on the treatment of M. abscessus infections. We used this assay to determine that niclosamide is a rapidly bactericidal agent against non-replicating M. abscessus.

#### AUTHOR CONTRIBUTIONS

BB and TP conceived the experiments. BB, LC, YO, and TP designed all the experiments. BB, LC, DR performed the

experiments. BB, LC, and TP wrote the manuscript. All authors edited and approved the final manuscript.

#### FUNDING

This work was funded in part by NIAID of the National Institutes of Health under award number R01AI099188 and by the Bill and Melinda Gates Foundation, under grant OPP1024038. The content is solely the

#### REFERENCES


responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2018.02417/full#supplementary-material


chemotype: identification of derivatives with improved drug exposure. Bioorg. Med. Chem. 23, 5829–5838. doi: 10.1016/j.bmc.2015.07.001


caseum. Antimicrob. Agents Chemother. 62, e02266-17. doi: 10.1128/AAC. 02266-17


**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 Berube, Castro, Russell, Ovechkina and Parish. 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.

# Tedizolid Activity Against Clinical *Mycobacterium abscessus* Complex Isolates—An *in vitro* Characterization Study

Ying Wei Tang<sup>1</sup> , Bernadette Cheng<sup>2</sup> , Siang Fei Yeoh<sup>3</sup> , Raymond T. P. Lin2,4 and Jeanette W. P. Teo<sup>2</sup> \*

*<sup>1</sup> Department of Biological Sciences, National University Singapore, Singapore, Singapore, <sup>2</sup> Department of Laboratory Medicine, National University Hospital, Singapore, Singapore, <sup>3</sup> Pharmacy, National University Hospital, Singapore, Singapore, <sup>4</sup> National Public Health Laboratory, Ministry of Health, Singapore, Singapore*

#### *Edited by:*

*Thomas Dick, Rutgers, The State University of New Jersey, Newark, United States*

#### *Reviewed by:*

*Timothy Barkham, Tan Tock Seng Hospital, Singapore Avi Peretz, The Baruch Padeh Medical Center, Poriya, Israel*

> *\*Correspondence: Jeanette W. P. Teo jeanette\_teo@nuhs.edu.sg*

#### *Specialty section:*

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

*Received: 18 July 2018 Accepted: 16 August 2018 Published: 07 September 2018*

#### *Citation:*

*Tang YW, Cheng B, Yeoh SF, Lin RTP and Teo JWP (2018) Tedizolid Activity Against Clinical Mycobacterium abscessus Complex Isolates—An in vitro Characterization Study. Front. Microbiol. 9:2095. doi: 10.3389/fmicb.2018.02095* *Mycobacterium abscessus* complex consist of three rapidly growing subspecies: *M. abscessus*, *M. massiliense,* and *M. bolletii*. They are clinically important human pathogens responsible for opportunistic pulmonary and skin and soft tissue infections. Treatment of *M. abscessus* infections is difficult due to *in vitro* resistance to most antimicrobial agents. Tedizolid (TZD) is a next-generation oxazolidinone antimicrobial with a wide spectrum of activity even against multidrug resistant Gram-positive bacteria. In this study, the *in vitro* activity of TZD against the *M. abscessus* complex (*n* = 130) was investigated. Susceptibility testing by broth microdilution showed lower TZD minimum inhibitory concentrations (MICs) when compared to linezolid. The MIC<sup>50</sup> and MIC<sup>90</sup> was 1 mg/L and 4 mg/L, respectively across all *M. abscessus* complex members, reflecting no difference in subspecies response to TZD. Pre-exposure of *M. abscessus* complex to subinhibitory concentrations of TZD did not trigger any inducible drug resistance. Single-drug time kill assays and bactericidal activity assays demonstrated bacteriostatic activity of TZD in all three *M. abscessus* subspecies, even at high drug concentrations of 4 to 8x MIC. Combination testing of TZD with clarithromycin, doxycycline and amikacin using the checkerboard approach showed no antagonistic interactions. TZD may be an effective therapeutic antimicrobial agent for the treatment of *M. abscessus* infections.

Keywords: multidrug resistant (MDR), inducible resistance, oxazolidinone, time-kill, repurposable drugs

### INTRODUCTION

Mycobacterium abscessus complex consists of three rapidly-growing mycobacteria (RGM) subspecies: M. abscessus subspecies abscessus, M. abscessus subspecies massiliense and M. abscessus subspecies bolletii (Lee et al., 2015). They have emerged as clinically important multi-drug resistant (MDR) human pathogens responsible for a wide spectrum of skin and soft tissue infections (SSTIs), opportunistic infections in immunocompromised patients and pulmonary infections in patients with chronic pulmonary disease or cystic fibrosis (Nessar et al., 2012). Nosocomial outbreaks of M. abscessus have been reported worldwide, highlighting its clinical significance (Nessar et al., 2012). M. abscessus complex accounts for approximately 65–80% of pulmonary infections caused by RGM (Koh et al., 2011). In Singapore, M. abscessus complex is the most prevalent RGM isolated in hospitals and accounts for approximately 35% of all nontuberculous mycobacteria (NTM) infections (Tang et al., 2015).

M. abscessus pulmonary infections are infamously difficult to treat, with low cure rates ranging from 30 to 50%. This is attributed to natural resistance to most antimicrobial agents (Van Ingen et al., 2012). Existing treatment regimens are combinationbased therapies usually consisting of a macrolide antibiotic such as clarithromycin (CLR), amikacin (AMK) and either cefoxitin (FOX), imipenem (IPM), or tigecycline (TGC) (Van Ingen et al., 2012). The administration of combination therapy (usually CLR and AMK) is lengthy, lasting for periods of between 2 and 4 months before clinical and microbiological improvements are noticeable (Huang et al., 2010). And the lack of alternative antimicrobial options further complicates the treatment of NTM infections (Benwill and Wallace, 2014).

Tedizolid (TZD) is a next-generation oxazolidinone antibiotic approved by the Food and Drug Administration (FDA) in 2014 for the treatment of acute bacterial skin and skin structure infections (ABSSSI) caused by certain Streptococcus spp. and methicillin-resistant Staphylococcus aureus (MRSA). Phase three clinical trials demonstrated non-inferiority of TZD to the first-in-class oxazolidinone LZD for the treatment of ABSSI, with improved clinical efficacy against MRSA and slightly improved safety profile (Moran et al., 2014). Oxazolidinones are protein synthesis inhibitors (Rybak et al., 2014) whose action is primarily bacteriostatic (Rybak et al., 2014). In vitro, TZD has demonstrated activity against acid-fast bacilli such as slowgrowing Mycobacterium tuberculosis and the rapidly-growing Mycobacterium fortuitum (Kisgen et al., 2014). TZD MIC values against NTM were equivalent or 1- to 8-fold lower than those of LZD, indicating improved in vitro potency (Brown-Elliott and Wallace, 2017). Another study showed that TZD exhibited good bacteriostatic activity against M. abscessus, with MICs twoto 16-fold lower as compared to LZD (Compain et al., 2018). The combination of in vitro activity against MDR Gram-positive bacteria, an oral dosage formulation and once-daily dosing makes TZD a promising investigational antimicrobial therapeutic agent (Kisgen et al., 2014).

In this study, we explored the potential use of TZD for anti-mycobacterial therapy by characterizing the in vitro activity of TZD against 130 clinical isolates of M. abscessus complex members.

#### MATERIALS AND METHODS

#### Mycobacterial Isolates and Genetic Characterization

A total of 130 retrospective non-duplicate clinical M. abscessus complex isolates were evaluated. This collection consisted of 43 M. abscessus isolates, 82 M. massiliense isolates and five M. bolletii isolates. The subspecies of the M. abscessus complex isolates was determined by multi-locus sequencing employing the rpoB and hsp65 genes (Macheras et al., 2011). CLR resistance was analyzed by full-length sequencing of the erm(41) and rrl genes (Aziz et al., 2017). For erm(41), the full-length 673 bp gene sequence was examined for T/C polymorphism at the 28th nucleotide position as well as for gene deletions. erm(41) T28 sequevars have wild-type inducible CLR resistance whilst C28 sequevars are phenotypically CLR susceptible (Choi et al., 2012). For the rrl gene, the nucleotides 2058–2059, associated with CLR resistance were examined.

### MIC Determination

Antibiotic powders of TZD, CLR, and LZD were purchased from MedChem Express (NJ, USA). Antimicrobial susceptibility testing of TZD, CLR and LZD were performed using the microdilution method according to the Clinical & Laboratory Standards Institute (CLSI) guidelines (CLSI, 2015). The working range for all tested antimicrobials was 0.125–64 mg/L. For TZD and LZD, the inoculated microdilution plates were incubated at 30◦C for 3–5 days before growth was assessed by visual inspection. The MIC was determined as the concentration of antibiotic at which there was no visible growth. Staphylococcus aureus ATCC (American type culture collection) 6538 and Enterococcus faecalis ATCC 29212 were used as susceptibility testing quality control strains. The MIC for the control strains fell within the acceptable MIC range of 0.25–1 mg/L for both TZD and LZD (Woods et al., 2011; Brown-Elliott and Wallace, 2017). For TZD, there are currently no interpretative criteria for RGM. For LZD, RGM with MICs of ≤8 were classified as sensitive and ≥32 as resistant (Woods et al., 2011).

#### Bactericidal/Static Activity Determination

For the bactericidal/static activity determination assay, M. abscessus isolates (n = 7), M. massiliense isolates (n = 15) and M. bolletii (n = 5) were tested. After three days of TZD incubation at 30◦C, the entire 96-well microtiter plate well contents corresponding to the two-fold diluted TZD concentrations (64–0.0625 mg/L) were plated and the CFU determined. The Minimum Bactericidal Concentration (MBC) of TZD against the tested isolates was defined as the lowest drug concentration required to induce ≥99.9% cell death as compared to the untreated control at the 0 h time point. For bactericidal antibiotics, the MBC is classified as ≤4 times the MIC while the MBC is usually >4 times the MIC for bacteriostatic antibiotics.

#### TZD Time Kill Assay for the *M. abscessus* Complex

Time-kill assays were performed according to CLSI guidelines (CLSI, 1999) and were setup for a single isolate each of M. abscessus, M. massiliense and M. bolletii using a 10<sup>6</sup> CFU/mL inoculum exponential growth phase bacterial suspension. Twofold increasing concentrations of TZD (from 0.25 to 8x MIC) and a drug-free growth control was used. At time intervals of 0, 4, 8, 12, 24, 36, 48, 72, 96, and 120 h CFU enumerations were made. Bactericidal activity was defined as a ≥3-log<sup>10</sup> decrease in CFU/mL at 120 h when compared to the 0 h time point. All timekill experiments were performed in duplicate and the mean CFU counts plotted.

#### TZD Pre-exposure Assay

erm(41) confers inducible macrolide resistance in the M. abscessus complex, observable phenotypically at day 14 of


TABLE 1 | MICs of tedizolid, clarithromycin and linezolid for 130 clinical isolates of *Mycobacterium abscessus* complex.

\**For LZD, isolates with MICs of* ≤*8 were classified as sensitive and* ≥*32 were resistant (Woods et al., 2011).*

*For CLR, isolates with MICs of* ≤*2 were classified as sensitive and* ≥*8 were resistant.*

incubation (Rubio et al., 2015). To examine if a similar inducible phenomenon existed for TZD, M. abscessus complex isolates were pre-exposed to sub-inhibitory concentrations of TZD prior to MIC determination as previously described (Aziz et al., 2017). TZD pre-exposure assays were performed for three isolates each of M. abscessus, M. massiliense and M. bolletii. Briefly, 10<sup>6</sup> CFU/mL bacterial suspension was treated with TZD at a subinhibitory concentration of 0.25 mg/L for M. abscessus and M. massiliense isolates, and at 1 mg/L for M. bolletii isolates, fourfold lower than their MIC<sup>50</sup> values. An untreated, drug-free culture was setup as a growth control. The MICs were determined at day 3 and at day 14.

### Synergy Studies Using Checkerboard Titration Assay

The in vitro interactions of TZD and CLR, TZD and DOX, as well as TZD and AMK were investigated by the checkerboard approach using the broth microdilution method as previously described (Kaushik et al., 2015). Five isolates of M. abscessus, four M. massiliense and five M. bolletii isolates were used for evaluation. The fractional inhibitory concentration index (PFIC) for each isolate was calculated as follows: PFIC = MIC of antibiotic 1 in combination MIC of antibiotic <sup>1</sup> only + MIC of antibiotic 2 in combination MIC of antibiotic <sup>2</sup> only . Synergy was defined as a FIC index of ≤0.5, indifference by a FIC index of >0.5 to ≤4 and antagonism when the FIC index was >4.

#### RESULTS

#### Susceptibility of *M. abscessus* Complex Isolates to Tedizolid, Clarithromycin, and Linezolid

For TZD, the MIC range was 0.0625–8 mg/L, compared to 0.0625–>32 mg/L for LZD. The MIC<sup>50</sup> and MIC<sup>90</sup> for TZD was 1 and 4 mg/L, consistent across all three subspecies suggesting that the 3 subspecies were similarly responsive to TZD. In general, the TZD MICs were 2- to 16-fold lower than those of LZD. Due to the lack of interpretive criteria for TZD for RGM, susceptibility rates were not assigned (**Table 1**). For LZD, 52.3% of all isolates were susceptible (MIC ≤ 8 mg/L); 53.5% of M. abscessus, 20% of M. bolletii and 53.7% of M. massiliense (**Table 1**).

For CLR, the MIC<sup>50</sup> and MIC<sup>90</sup> were both >16 mg/L for M. abscessus and M. bolletii, as compared to 1 and 12 mg/L for M. massiliense. According to CLSI interpretive criteria for susceptibility (CLSI, 2017), 55.4% (72/130) of all isolates were susceptible to CLR (MIC < 2 mg/L). M. massiliense isolates showed susceptibility rates of 76.8%. This is consistent with the observation that M. massiliense usually possesses a truncated non-functional erm(41) (Chew et al., 2017). In contrast, M. abscessus (20.9% susceptible) and M. bolletii isolates (0% susceptible) showed higher rates of resistance to CLR. The susceptible M. abscessus isolates were erm(41) C28 sequevar.

#### TZD Does Not Exhibit Bactericidal Activity Against the *M. abscessus* Complex

Time kill assays were performed using TZD for one isolate each of M. abscessus (MIC = 2 mg/L), M. bolletii (MIC = 8 mg/L) and M. massiliense (MIC = 0.25 mg/L). TZD did not exhibit bactericidal activity in all three subspecies, even at concentrations of 4- and 8-fold higher than the MIC determined by the microdilution method (**Figure 1**). There was a general decline in CFU count over time for M. bolletii and M. massiliense. Bacterial regrowth (0.2 log10CFU/ml greater than the starting inoculum) was observed at time point 72 h for M. abscessus for all drug concentrations (0.5–8x MIC), following which a reduction in CFU was only observed at concentrations of 4x and 8x MIC. For M. bolletii, regrowth was noted at time point 120 h for concentrations of 0.25x and 1x MIC. Regrowth was observed at time point 12 h for 0.25x MIC and 72 h for 0.5x MIC for M. massiliense (**Figure 1**).

### TZD Exhibits Bacteriostatic Activity Against *M. abscessus* Complex

TZD exhibited bacteriostatic activity against all tested isolates of the M. abscessus complex (**Table 2**). The MBC of all three subspecies was greater than four times of MIC, which is characteristic of a bacteriostatic antimicrobial agent.

#### TZD Pre-exposure Does Not Induce Resistance

Pre-treatment of three M. abscessus, M. bolletii and M. massiliense isolates to sub-inhibitory concentrations of TZD did not affect MICs (**Table 3**). MICs after pre-exposure to TZD were similar to those without pre-exposure, suggesting that M. abscessus did not harbor inducible TZD resistance mechanisms.

### Checkerboard Testing of TZD in Combination With Clarithromycin, Doxycycline and Amikacin Suggests Interactions That Are Largely Indifferent

Amikacin (AMK) and clarithromycin (CLR) are currently the only two antimicrobial agents with reliable in vitro activity against M. abscessus (Tang et al., 2015). TZD in combination with CLR, DOX and AMK were evaluated for antimicrobial activity against five isolates of M. abscessus, four isolates of M. bolletii and five isolates of M. massiliense by checkerboard synergy approach (**Table 4**). No instances of antagonism were observed in any antimicrobial combination tested. Indifference was the primary interaction accounting for 90.5% of all interactions. For the combination of TZD and CLR, all interactions were indifferent. One instance of synergistic interaction was observed in the erm(41) C28 sequevar M. bolletii isolate for the combination of TZD and DOX. In two erm(41) T28 sequevar M. abscessus isolates and one M. massiliense isolate, synergistic interactions were observed for the combination of TZD and AMK (**Table 4**). Overall, the findings suggest that TZD has no interaction when used in combination with CLR, DOX and AMK against M. abscessus.

### DISCUSSION

M. abscessus pulmonary infections are notoriously difficult to treat with low cure rates of 30–50% (5). This has



TABLE 3 | MICs of *Mycobacterium abscessus* complex after exposure to sub-inhibitory concentrations of tedizolid.


\**Three unique isolates of each subspecies were used for testing.*

spurred drug repurposing, defined as the "off-label" usage of existing antimicrobials (Palomino and Martin, 2014). LZD was initially developed for the treatment of infections caused by β-lactam-resistant Gram-positive bacteria, but it is now a recommended second-line drug for the treatment of MDR and extensively drug-resistant tuberculosis (Dheda et al., 2017). Furthermore, TZD has demonstrated in vitro activity against mycobacterial pathogens such as Mycobacterium tuberculosis and Mycobacterium fortuitum (Kisgen et al., 2014).

In our study, the potential of TZD for the treatment of M. abscessus complex infections was investigated in vitro. In comparison to two recent studies, our TZD MIC range of 0.0625– 8 mg/L (MIC<sup>50</sup> = 1 mg/L, MIC<sup>90</sup> = 4 mg/L) for 43 M. abscessus isolates was lower than the MIC range of 0.12–>32 µg/mL (MIC<sup>50</sup> = 4µg/mL, MIC<sup>90</sup> = 8µg/mL) reported by Brown-Elliott and Wallace (2017) and the MIC range of 1–16µg/mL (MIC<sup>50</sup> = 2µg/mL, MIC<sup>90</sup> = 8µg/mL) reported by Compain et al. (2018). For the 82 M. massiliense isolates a TZD MIC range of 0.0625–8 mg/L (MIC<sup>50</sup> = 1 mg/mL, MIC<sup>90</sup> = 4 mg/mL) was obtained. Brown-Elliott & Wallace Jr. reported a TZD MIC range of 0.12–>32µg/mL (MIC<sup>50</sup> = 2µg/mL, MIC<sup>90</sup> = 4µg/mL) for a smaller set of 12 isolates whilst Compain et al. reported a MIC range of 1–8µg/mL (MIC<sup>50</sup> = 4µg/mL, MIC<sup>90</sup> = 8µg/mL) for 14 M. massiliense isolates (n=14). The TZD MIC range of 1–8 mg/L (MIC<sup>50</sup> and MIC<sup>90</sup> = 4 mg/L) for 5 M. bolletii isolates were comparable to the MIC range of 1–4 mg/L (MIC<sup>50</sup> = 2µg/mL, MIC<sup>90</sup> = 4µg/mL) as determined by Compain et al. (2018).

There are currently no CLSI recommended TZD breakpoints for Mycobacteria but LZD is considered a reliable surrogate antimicrobial agent for TZD susceptibility, with the European Committee on Antimicrobial Susceptibility Testing (EUCAST) recommending the reporting of isolates susceptible to LZD as also susceptible to TZD (EUCAST, 2016). LZD susceptibility was found to be highly predictive of TZD susceptibility, with high categorical agreement between MIC values of LZD and TZD, and low rates of very major errors for Gram-positive bacteria (e.g., Staphylococcus spp. and Enterococcus spp.) (Zurenko et al., 2014). All 130 M. abscessus isolates were susceptible to TZD when a breakpoint ≤8 mg/L was applied. Although the suitability of LZD as a surrogate for TZD susceptibility has only been recommended for Gram-positive bacteria, these findings suggest that M. abscessus may be more susceptible to TZD than LZD.

In M. tuberculosis oxazolidinone resistance is associated with point mutations in the 23S rRNA gene (rrl) and in the 50S ribosomal protein L3 (Klitgaard et al., 2015; McNeil et al., 2017). rrl and L3 mutant strains were resistant to LZD and crossresistant to sutezolid, a next-generation oxazolidinone currently in clinical development with improved potency against M. tuberculosis (McNeil et al., 2017). Furthermore, Gram-positive bacteria with 23S rRNA gene mutations were found to have high LZD (16 mg/L) and TZD MICs (>1 mg/L) as reported in a 2011–2012 surveillance report of TZD activity (Bensaci and Sahm, 2017). In our study, 5 isolates with high TZD MIC (8 mg/L) had their full–length rrl gene sequenced. All the sequenced isolates possessed a wild type rrl gene. This suggests alternative resistance mechanisms, such as efflux pumps (Gupta et al., 2006). We acknowledge our study limitation where mutations in bacterial 50S ribosomal protein L3, which are associated with oxazolidinone resistance, were not investigated.

This is currently the first study (to the best of our knowledge) to perform a time-kill assay for all three subspecies of the M. abscessus complex. TZD exhibits little concentration-dependent killing and no significant bactericidal activity against the three subspecies at all tested drug concentrations (0.5x−8x MIC). Compain et al. reported similar time-kill kinetics for M. abscessus ATCC 19977/CIP 104536, with no bactericidal activity at TZD concentrations of 4 and 8 mg/L. Bacterial regrowth was observed in M. abscessus in the logarithmic phase of growth for TZD concentrations tested. In comparison, regrowth was only observed at lower TZD concentrations of 0.25x and 1x MIC for M. bolletii and 0.25x and 0.5x MIC for M. massiliense. These findings were consistent with the findings by Ferro et al. (2015)


\$*T28 sequevar are CLR resistant, C28 sequevars are CLR susceptible. Deleted, refers to 274 bp erm(41) gene deletion characteristic in the M. massiliense subspecies.* #*FIC index was calculated as [(MIC of tedizolid in combination/MIC of tedizolid alone)* + *(MIC of second antibiotic in combination/MIC of second antibiotic alone)].*

*Only FIC index* <*0.5 was considered as a synergistic interaction.*

where regrowth was also observed for AMK and CLR after 72 hours even at concentrations of 2x to 8x MIC. The findings suggest that a TZD concentration of ≥4x MIC may be required to induce significant killing activity against M. abscessus, whereas a TZD concentration of 1x/2x MIC sufficiently reduces bacterial counts in M. bolletii and M. massiliense over time. Similar to other active antimicrobials against M. abscessus complex, TZD exhibits a bacteriostatic effect that is more pronounced in M. bolletii and M. massiliense than M. abscessus.

Synergy studies of TZD with CLR, DOX, AMK demonstrated that all combinations primarily showed indifferent interactions with no instances of antagonism. A similar study performed by Compain et al. reported indifferent interactions of TZD

#### REFERENCES


with TGC, AMK and ciprofloxacin, with the combination of CLR and TZD showing one synergistic interaction out of 6 tested isolates. The findings suggest that TZD could be used in the existing combination regime of CLR and AMK with no antagnostic interactions.

#### AUTHOR CONTRIBUTIONS

YT drafted the manuscript and performed the experiments. BC performed the experiments. SY provided clinical feedback and scientific review. RL provided clinical feedback and scientific review. JT drafted the manuscript and oversaw the project execution.


Antimicrob. Agents Chemother. 61:e01296–e01217. doi: 10.1128/AAC. 01296-17


**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.

The handling editor declared a past co-authorship with one of the authors JT.

Copyright © 2018 Tang, Cheng, Yeoh, Lin and Teo. 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.

# Therapy for Mycobacterium kansasii Infection: Beyond 2018

#### Michelle S. DeStefano<sup>1</sup> , Carolyn M. Shoen<sup>1</sup> and Michael H. Cynamon1,2 \*

<sup>1</sup> Central New York Research Corporation, Syracuse, NY, United States, <sup>2</sup> Veterans Affairs Medical Center, Syracuse, NY, United States

The current standard of care therapy for pulmonary Mycobacterium kansasii infection is isoniazid (300 mg/day), rifampin (600 mg/day), and ethambutol (15 mg/kg/day) for 12 months after achieving sputum culture negativity. Rifampin is the key drug in this regimen. The contribution of isoniazid is unclear since its in vitro MICs against M. kansasii are near the peak achievable serum levels and more than 100-fold greater than the MICs for Mycobacterium tuberculosis. Ethambutol likely decreases the emergence of rifampin resistant organisms. There are several new drug classes (e.g., quinolones, macrolides, nitroimidazoles, diarylquinolines, and clofazimine) that exhibit antimycobacterial activities against M. tuberculosis but have not yet been adequately studied against M. kansasii infections. The evaluation of in vitro activities of these agents as well as their study in new regimens in comparison to the standard of care regimen in mouse infection models should be undertaken. This knowledge will inform development of human clinical trials of new regimens in comparison to the current standard of care regimen. It is likely that shorter and more effective therapy is achievable with currently available drugs.

Keywords: M. kansasii, antimycobacterials, mouse models, non-tuberculous mycobacteria, pulmonary infection

### INTRODUCTION

Mycobacterium kansasii (M. kansasii) is a group I non-tuberculous mycobacterium (NTM) (Buhler and Pollack, 1953). These organisms are ubiquitous in the environment and are often found in aquatic settings (Amha et al., 2017). Seven genotypes, or subtypes have been identified, along with an intermediate (I/II) and atypical (IIb) subtype (Bakuła et al., 2018). Types I and II are the most common clinical subtypes found while types III-VII have generally only been recovered from environmental samples (Taillard et al., 2003).

Mycobacterium kansasii causes infection in both immunocompetent and immunosuppressed individuals. Additionally, it is the non-tuberculous mycobacterium most frequently found in immunocompetent patients. It is also the second most frequent mycobacterium found in HIVinfected patients and is only surpassed by Mycobacterium avium complex (Canueto-Quintero et al., 2003). The number of NTM infections is increasing and are geographically wide-spread. M. kansasii is the second most prevalent cause of NTM disease in the United States, China, South American countries, and some European countries such as Poland, Slovakia, and the United Kingdom (Hoefsloot et al., 2013).

Mycobacterium kansasii causes lung disease that clinically and radiologically resembles tuberculosis. The American Thoracic Society (ATS) and the Infectious Diseases Society of America (IDSA) recommend a combination of three anti-tuberculosis drugs, isoniazid (INH) at 300 mg/day, rifampin (RIF) at 600 mg/day, and ethambutol (EMB) at 15 mg/kg/day for treatment of M. kansasii

#### Edited by:

Thomas Dick, Rutgers, The State University of New Jersey, Newark, United States

#### Reviewed by:

Amanda S. MacLeod, Duke University, United States Tianyu Zhang, Guangzhou Institutes of Biomedicine and Health (CAS), China Eric Nuermberger, Johns Hopkins University, United States

#### \*Correspondence:

Michael H. Cynamon Michael.Cynamon@va.gov

#### Specialty section:

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

Received: 28 June 2018 Accepted: 05 September 2018 Published: 24 September 2018

#### Citation:

DeStefano MS, Shoen CM and Cynamon MH (2018) Therapy for Mycobacterium kansasii Infection: Beyond 2018. Front. Microbiol. 9:2271. doi: 10.3389/fmicb.2018.02271

**154**

pulmonary disease in HIV negative individuals. Treatment should be continued for 1 year after culture negativity (Griffith et al., 2007).

Mycobacterium kansasii infection is challenging to treat since the course of therapy requires multiple drugs and treatment periods of up to 2 years. Long periods of treatment often give rise to additional problems including patient non-adherence to the treatment plan, expense, and potential drug interactions and/or adverse events (Larsson et al., 2017). There are several antimycobacterial drugs that are effective against tuberculosis and other NTMs that might be better alternatives to the current three drug combination. The British Thoracic Society recommends the inclusion of a macrolide, such as azithromycin (AZI) at 250 mg/day or 500 mg of clarithromycin (CLA) twice a day as an alternative to INH (Haworth et al., 2017).

The focus of this review is to discuss in vitro susceptibility testing, in vivo evaluation in animal models, and clinical reports or studies utilizing these antimycobacterial agents. We will also discuss future efforts that should be undertaken to improve the clinical outcome of M. kansasii pulmonary infection.

#### IN VITRO SUSCEPTIBILITY TESTING

In vitro susceptibility testing of M. kansasii to antimycobacterial agents has been accomplished using several different methodologies, including agar dilution, broth macrodilution or microdilution method, or BACTEC. Investigations to determine the minimal inhibitory concentration (MIC) of antimicrobials against M. kansasii have been performed on multiple clinical isolates. **Table 1** contains the MIC<sup>50</sup> and MIC<sup>90</sup> (defined as the lowest concentration at which 50 and 90% of the clinical isolates tested were inhibited, respectively) of drugs currently used to treat M. kansasii infection, as well as other drugs used to treat tuberculosis and some NTM infections. The list includes oxazolidinones, quinolones, macrolides, riminophenazines, diarylquinolines, nitroimidazoles, glycylcyclines, and rifamycins.

Throughout the literature there is a divergence in the MIC values for CLA and AZI. The reported MIC values of susceptible strains of M. kansasii for CLA are generally consistent with those listed in **Table 1** with the MIC<sup>50</sup> 0.12 and MIC<sup>90</sup> 0.5 µg/mL. Other MIC values reported are: 0.125–0.25 µg/mL (Griffith et al., 2003), MIC<sup>50</sup> ≤ 0.06 MIC<sup>90</sup> 0.125 µg/mL (Guna et al., 2005), MIC<sup>50</sup> 0.25 and MIC<sup>90</sup> 0.5 µg/mL (Witzig and Fransblau, 1993), and MIC<sup>50</sup> 0.125, MIC<sup>90</sup> 0.25 µg/mL (Brown et al., 1992). Although a more recent study (Li et al., 2016) suggests that CLA resistance is related to the Subtype I phenotype of M. kansasii and reports much higher MIC<sup>90</sup> values (MIC<sup>50</sup> 0.25, MIC<sup>90</sup> 128 µg/mL). Subtype 1 is the most prevalent type of M. kansasii worldwide. In the same study the MIC values of AZI (MIC<sup>50</sup> 16 MIC<sup>90</sup> 32 µg/mL) are similar to those reported in **Table 1** (MIC<sup>50</sup> 8 MIC<sup>90</sup> 8 µg/mL). It is possible that these higher values are due to cross resistance with CLA as there are CLA resistant isolates identified in the same study. Other studies report MIC values of >8 µg/mL (Yew et al., 1994). Conversely, there are also reports of MIC values of AZI being similar to those seen with clarithromycin (Brown et al., 1992; Klemens and Cynamon, 1994).

The clinical MIC cutoff for EMB to define susceptibility/ resistance against M. kansasii is 4 µg/ml, however, there is growing evidence that many M. kansasii isolates are resistant to EMB by this definition (**Table 1**). Patients diagnosed with this infection are not routinely evaluated for EMB susceptibility, therefore, resistance may be more prevalent than previously thought. It has been suggested that resistance to EMB (and INH) in M. kansasii occurs when a patient develops RIF resistance while on the standard regimen (Hjelm et al., 1992; Griffith et al., 2003). In a 2016 study, isolates belonging to M. kansasii Subtype I were found to exhibit greater resistance to EMB than the other subtypes (Li et al., 2016) although the differences were not statistically significant perhaps due to their small sample size. In a study looking at isolates from Poland, Germany, and Netherlands, 97.9% of the isolates were EMB resistant (all subtypes with the exception of Subtype V), however, published results from other countries indicates a resistance range for EMB of 0–94% as well as variable methods used to determine susceptibility (Bakuła et al., 2018). This raises the question whether EMB and INH should be used as companion drugs to rifampin when newer, potentially more efficacious drugs are currently available.

A model of intracellular M. kansasii infection has been designed to mimic the pharmacodynamics and pharmacokinetics between infected macrophages and therapeutic agents (Srivastava et al., 2015; Srivastava and Gumbo, 2018). The hollow fiber model of M. kansasii has been adapted from similar models for Mycobacterium tuberculosis and M. avium complex (Deshpande et al., 2010; Srivastava and Gumbo, 2011). It has been reported that this system evaluates compound efficacy, rank orders kill rates compared to standard regimens, and assesses antibiotic tolerance by the use of efflux pump inhibitors. In the case where samples were available from moxifloxacin (MOX) treated patients, analysis of bronchial secretions, alveolar macrophages and lung epithelium lining fluid were used to simulate optimal doses and breakpoints for resistance with the aid of computer software (Srivastava et al., 2015). This preclinical disease model will provide valuable additional information to help select the appropriate agents for in vivo testing and perhaps subsequent evaluation in clinical studies.

The immune status of a patient determines the course of infection and the selection of an appropriate regimen. RIF is currently the cornerstone for the therapy of M. kansasii infection, but in patients co-infected with HIV, RIF presents a problem since it increases the hepatic metabolism of protease inhibitors, often used for the treatment of HIV infection (Brown-Elliott et al., 2012). Rifabutin (RBT) has less effect on the hepatic metabolism, therefore, it is often used as an alternative to RIF in HIV-infected patients. Rifapentine (RPT) is an alternative to RIF or RBT. All three rifamycins have demonstrated good in vitro activity against M. kansasii (Cynamon and Sklaney, 2013).

Linezolid (LZD), an oxazalidinone, is currently an option for multi-drug resistant tuberculosis infections and has been shown to have good in vitro activity against M. kansasii (**Table 1**), but there are significant toxicity issues with this drug. These


toxicity issues make the use of this drug for M. kansasii therapy problematic. The newer oxazolidinones, tedizolid, and contezolid, have fewer side effects than LZD, and have good in vitro activity against M. kansasii (Shoen et al., 2018).

Four other drugs that are used to treat drug resistant tuberculosis have also shown good in vitro activity against M. kansasii (**Table 1**). They are clofazimine (CFZ) (riminophenazine), bedaquiline (diarylquinoline), delamanid (nitroimidazole), and pretomanid (nitroimidazopyran) although the MIC<sup>50</sup> of pretomanid is significantly high and therefore may not be a viable candidate compound. Attempting to correlate in vitro efficacy with achievable serum levels is not always useful when selecting compounds for in vivo testing and treatment.

Several factors affecting serum levels and compound availability have to be considered. These include tissue/cell accumulation, interaction with other drugs, and length of treatment (single dosing versus multiple doses). CFZ illustrates the complexity of this issue. A single 200 mg dose in humans results in peak plasma levels of 0.41 µg/ml, which is lower than reported MIC values (**Table 1**), however, CFZ concentrates in macrophages and tissues long-term due to its highly lipophilic character (Cholo et al., 2012). INH treatment with CFZ increases the plasma concentration and reduces tissue concentration of CFZ (Haworth et al., 2017). Several studies with other antimycobacterial agents allude to a synergistic relationship with CFZ when evaluated in combination therapy against a variety of NTM species (McGuffin et al., 2017).

Since treatment for M. kansasii and other NTM infections consist of a multi-drug regimen, a comprehensive review of each compound's known characteristics need to be considered to better predict potential efficacy. The established method for evaluation of in vitro combination therapy, "the checkerboard" has been used to define synergy or antagonism between two compounds. It is not clear whether results from this method correlate with those from in vivo modeling. New methods are being developed for prioritizing drug combinations, however, they have not yet been utilized for modeling regimens of M. kansasii therapy. Due to the lack of resources available for M. kansasii research it continues to be important to utilize available information on candidate compounds from in vitro combination studies focused on M. tuberculosis to advise our prioritization of regimens for M. kansasii.

#### ANIMAL STUDIES

Relatively little research has been done utilizing animal models to develop new therapies for NTM generally (Soni et al., 2016) and M. kansasii infection specifically. Research into the efficacy of compounds has largely been limited to drugs currently in use to treat tuberculosis or licensed drugs such as quinolones and macrolides. Due to the emerging importance of this infection, and the challenges of treatment for patients with resistant infections, more research into the development of an appropriate animal model to test new therapeutics is needed.

A review of the available literature indicates that several different strains of mice, levels of inocula, and routes of infection have been used, all with the ability to sustain a measurable bioburden. Presently there is no standard model and the strains of mice and routes of infection used are varied. Described below is a summary of the available literature on animal studies evaluating antimycobacterial compounds against M. kansasii.

BALB/c athymic (nude) mice infected with 10<sup>7</sup> CFU M. kansasii intravenously resulted in disseminated infection (Graybill and Bocanegra, 2001). Untreated control mice succumbed to the infection after about 2 months. Therapy given daily for 21 days was started 1-week post-infection. Increasing doses of RPT (0.15, 0.3, and 0.6 mg/kg), AZI (10, 15, 30, 50, and TABLE 2 | In vivo results of M. kansasii intranasal infection in mice.

fmicb-09-02271 September 24, 2018 Time: 12:19 # 4


<sup>∗</sup>95% confidence interval.

150 mg/kg) and EMB (10, 25, and 100 mg/kg) were evaluated in a survival study. RPT dosed at 0.6 mg/kg, AZI at 15 mg/kg, and EMB at 10 mg/kg or higher doses of each prolonged survival. There was no difference in survival between RPT and AZI. The investigators also evaluated bioburden reduction in the spleens and livers of mice treated with RPT and AZI alone and in combination. The results indicated that the combination was no more effective than RPT alone.

In another study, beige mice (C57BL/6J background) were infected intravenously with 10<sup>7</sup> CFU of M. kansasii and treated 1-week post-infection, 5 days/week for 4 weeks. AZI 200 mg/kg, CLA 200 mg/kg, EMB 125 mg/kg, RIF 20 mg/kg, and CFZ 20 mg/kg were evaluated for activity compared to untreated controls. Viable cell counts were enumerated from the spleens and lungs. RIF reduced lung CFUs by about 1 log compared to the early controls and yielded the greatest reduction in organisms in the lungs. AZI and CLA had similar, but modest activity (Klemens and Cynamon, 1994).

In a study using beige mice infected intranasally with approximately 10<sup>6</sup> CFU of M. kansasii, treatment with gatifloxacin (GAT) 100 mg/kg, CLA 200 mg/kg, LZD 100 mg/kg, RIF 20 mg/kg, GAT + CLA, CLA + LZD, GAT + LZD, and RIF + CLA was given for 4 weeks, 5 days/week. All of the treatment groups were significantly better than the early controls (evaluated at the initiation of therapy, 1-week post-infection). CLA was more effective than any other monotherapy. CLA combined with LZD, GAT, or RIF was better than any single drug therapy (Cynamon et al., 2003; **Table 2**).

More recent studies with immunocompetent C57BL/6J mice demonstrate that M. kansasii can also survive and replicate in these mice. Two studies were carried out where mice were infected with approximately 5 × 10<sup>5</sup> CFU/mouse via the intranasal route. In the first experiment, RIF 10 mg/kg, rifalazil (RZL) 10 mg/kg, AZI 100 mg/kg and combinations of RIF + AZI and RZL + AZI were evaluated. All treatment groups were significantly better than the early controls. The combination of RZL + AZI was the most effective treatment group (Cynamon and Buswell, 2002; **Table 2**). In the second experiment, 4 weeks (5 days/week) of GAT 100 mg/kg and RIF 20 mg/kg were compared to increasing doses of RZL (1, 5,

and 10 mg/kg). The individual drugs were also compared to a combination of GAT + RZL10 mg/kg. All single agents and the combination regimen had significant activity compared to the early controls (evaluated 1-week post-infection). There was a dose response with RZL and the combination of RZL + GAT reduced lung CFUs to an undetectable level after 4 weeks of therapy (Cynamon and Monica, 2003; **Table 2**).

Although the published literature is limited regarding the use of M. kansasii infection models as a screening tool to evaluate recently licensed or new experimental compounds, the above studies suggest that these models should be used to determine if more effective therapy is achievable and whether their use may lead to a shorter duration of therapy compared to the current standard regimen.

#### CLINICAL RESEARCH

Rifampin, INH, and EMB (ATS/IDSA) or RIF, CLA (or AZI), EMB (British Thoracic Society) are suggested regimens for treatment of M. kansasii infection in humans taking between 12– 18 months depending on the sputum culture time of conversion. The standard treatment regimen is effective in most instances; however, therapy is lengthy and it does not provide a cure for all patients (Santin et al., 2009). Treatment failure is almost always associated with resistance to RIF and/or CLA (Brown-Elliott et al., 2012). The Clinical and Laboratory Standards Institute (CLSI) currently recommends testing all initial patient isolates of M. kansasii for RIF and CLA in vitro susceptibility. Broth dilution is the suggested methodology utilizing cation-adjusted Mueller Hinton broth and 5% OADC (oleic acid, albumin, dextrose, catalase) enrichment as the media (Griffith et al., 2007). When isolates are RIF resistant (MIC > 1 µg/ml), the suggestion is to test secondary agents such as AMK, EMB, LZD, MOX, RBT, and trimethoprim-sulfamethoxazole for in vitro activity (Brown-Elliott et al., 2012).

The current method for testing and interpretation of the MIC for INH in the clinical laboratory utilizes the critical concentrations (0.2 µg/ml, 1 µg/ml) for M. tuberculosis susceptibility testing. This results in a false interpretation of INH resistance because the MIC values for M. kansasii usually range from 0.5 µg/ml to 5 µg/ml (Brown-Elliott et al., 2012). Heifets' work illustrated this issue stating that of more than 100 M. kansasii isolates evaluated by agar dilution methodology almost all were completely resistant to 0.2 µg/ml and susceptible, or partially resistant to 1 µg/ml of INH (Heifets, 1991). MIC testing of INH is not recommended.

Although infection can occur in other organs or become disseminated, the focus of our literature review of clinical outcomes is pulmonary infection caused by M. kansasii. A systematic review of microbiological and clinical treatment outcomes for M. kansasii pulmonary infections highlighted the lack of randomized controlled clinical trials for this infection (Diel et al., 2017; Haworth et al., 2017). A small cohort of prospective and retrospective observational studies exist and a few representative studies are discussed below.

A prospective study by the British Thoracic Society evaluated 173 patients with pulmonary M. kansasii infection. Prior to species identification, 149 of these patients were treated with regimens of INH + RIF + EMB, INH + RIF + EMB + PZA, or INH + RIF + PZA + streptomycin. Once M. kansasii was identified patients were given RIF + EMB for an additional 9 months (Jenkins et al., 1994). The remaining 24 patients received 9 months of RIF + EMB. Their results suggested that the RIF + EMB was acceptable treatment for non-immune compromised patients. Only 11% of their patients had positive sputum cultures after 3 months of therapy. Fifteen patients (9.7%) of the 154 patients who entered the post-chemotherapy follow up period relapsed between 6 and 50 (median 23) months.

In a retrospective study of 111 patients diagnosed with M. kansasii lung disease and treated with RIF + EMB + INH supplemented with streptomycin during the initial 2–3 months, 75 patients completed 12 months of therapy (Santin et al., 2009). After a 41.5-month median follow-up, five (6.6%) patients relapsed. The authors concluded that a 12-month course is adequate for most, but not all patients.

A prospective trial by Griffith et al. (2003) evaluated a thriceweekly regimen of CLA 1000 mg (two female patients who weighed <50 kg received 500 mg), EMB 25 mg/kg, and RIF 600 mg in 18 patients with M. kansasii lung disease. This regimen yielded a durable cure with no relapse after a follow-up of 46 ± 8 months (mean time ± standard deviation) for the 14 patients that completed therapy. Four patients were lost to followup. The mean time to sputum conversion was 1 ± 0.9 months and the mean duration of therapy was 13.4 ± 0.9 months (Griffith et al., 2003).

#### FUTURE EFFORTS

The efficacy of INH when added to RIF + EMB should be measured in an appropriate mouse model to evaluate the contribution of INH to the standard regimen. New drug regimens should be tested in a mouse model of M. kansasii infection and compared to RIF + EMB ± INH to determine comparative efficacy. Similar to in vitro evaluation, multiple clinical isolates of M. kansasii should be studied in a mouse model to ensure that isolate differences do not affect efficacy in the mouse model. It is preferable to deliver the inoculum by the aerosol or intranasal route since this method most closely parallels the human pulmonary route of infection.

Initially, 4 weeks of treatment should be used to determine which regimens to evaluate in more detail (i.e., longer duration of therapy followed by a non-treatment observation period). It is yet to be determined whether one particular mouse strain is preferable to another. It would be of interest to evaluate C3HeB/FeJ mice as a potential model system for M. kansasii infection since they develop lesions similar to humans when infected with M. tuberculosis (Driver et al., 2012; Harper et al., 2012).

Several agents included in **Table 1** (bedaquiline, delamanid, MOX, tedizolid, contezolid, CFZ, and CLA) should be studied

in a mouse model of M. kansasii infection to determine whether they can improve outcomes with regard to efficacy and duration of therapy. CFZ added to INH + RIF + PZA in mice infected with tuberculosis was shown to decrease the time needed to achieve a durable cure (Tyagi et al., 2015; Ammerman et al., 2018). It should be determined whether CFZ has a similar effect with RIF + EMB or in the case of possible RIF resistance, a combination of CFZ with 2 other drugs in M. kansasii infections.

Several tuberculosis clinical trials with RIF suggest that doses of 20 and 35 mg/kg are both well tolerated and more efficacious than the standard 10 mg/kg dose (600 mg daily) (Tiberi et al., 2018). Higher doses of RIF should be evaluated to assess whether this treatment is also beneficial in reducing the lung bioburden of M. kansasii infected mice.

In order to accomplish randomized controlled clinical trials evaluating new regimens for therapy of pulmonary M. kansasii infection it is necessary to develop a clinical trials consortium, similar to the Mycoses Study Group<sup>1</sup> . This would enable multiple clinical trial sites to recruit and enter patients into clinical trials. In 2009 the European based MTB clinical trial

1 http://www.msgerc.org

#### REFERENCES


consortium TBnet, formed the Non-tuberculous Mycobacteria Network European Trials Group (NTM-NET). NTM-NET is an international consortium of clinical NTM researchers that is primarily focused in Europe<sup>2</sup> and is able to provide limited funding for NTM research. Perhaps now is an appropriate time for a Non-tuberculous Mycobacteria Study Group funded by NIAID to advance clinical trials for NTM infections by providing opportunities and resources for this important work.

#### AUTHOR CONTRIBUTIONS

MD, CS, and MC have equally contributed to the research of the content, writing, and editing of this article. All authors have approved the final version of the article.

### ACKNOWLEDGMENTS

This manuscript is dedicated to Dr. Leonid B. Heifets for his many contributions to expanding our understanding of susceptibility testing of mycobacteria.

2 http://www.ntm-net.org



kansasii lung disease. Eur. Respir. J. 33, 148–152. doi: 10.1183/09031936.000 24008


**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 DeStefano, Shoen and Cynamon. 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.

# Teicoplanin – Tigecycline Combination Shows Synergy Against Mycobacterium abscessus

Dinah B. Aziz1,2, Jeanette W. P. Teo<sup>3</sup> , Véronique Dartois<sup>4</sup> and Thomas Dick4,5 \*

<sup>1</sup> Department of Medicine, Yong Loo Lin School of Medicine, National University of Singapore, Singapore, Singapore, <sup>2</sup> Department of Pharmacy, Faculty of Science, National University of Singapore, Singapore, Singapore, <sup>3</sup> Department of Laboratory Medicine, National University Hospital, Singapore, Singapore, <sup>4</sup> The Public Health Research Institute, Rutgers, New Jersey Medical School, The State University of New Jersey, Newark, NJ, United States, <sup>5</sup> Department of Microbiology and Immunology, Yong Loo Lin School of Medicine, National University of Singapore, Singapore, Singapore

Lung disease caused by non-tuberculous mycobacteria (NTM), relatives of Mycobacterium tuberculosis, is increasing. M. abscessus is the most prevalent rapid growing NTM. This environmental pathogen is intrinsically resistant to most commonly used antibiotics, including anti-tuberculosis drugs. Current therapies take years to achieve cure, if cure if achieved. Thus, there is an urgent medical need to identify new, more efficacious treatments. Here, we explore the possibility of repurposing antibiotics developed for other indications. We asked whether novel two-drug combinations of clinically used antibiotics can be identified that show synergistic activity against this mycobacterium. An in vitro checkerboard titration assay was employed to test 180 dual combinations of 41 drugs against the clinical isolate M. abscessus Bamboo. The most attractive novel combination was further profiled against reference strains representing three sub-species (M. abscessus subsp. abscessus, massiliense and bolletii) and a collection of clinical isolates. This resulted in the identification of a novel synergistic antibiotic pair active against the M. abscessus complex: the glycopeptide teicoplanin with the glycylcycline tigecycline showed inhibitory activity at 2–3 µM (teicoplanin) and 1–2 µM (tigecycline). This novel combination can now be tested in M. abscessus animal models of infection and/or patients.

Keywords: Mycobacterium abscessus, teicoplanin, tigecycline, synergy, repurposing

### INTRODUCTION

Among the rapid growing non-tuberculous mycobacteria (NTM), M. abscessus is the most common cause of lung disease (Griffith et al., 2007; Medjahed et al., 2010; Hoefsloot et al., 2013). A poor rate of successful chemotherapeutic treatment makes M. abscessus disease a chronic incurable infection (Griffith et al., 2007). The bacterium is intrinsically drug resistant to most antibiotics (Brown-Elliott et al., 2012; Nessar et al., 2012). Currently, M. abscessus infections are treated by a multi-drug regimen consisting of a macrolide (clarithromycin), amikacin and either cefoxitin or imipenem (Benwill and Wallace, 2014; Ryu et al., 2016). Different clinics may choose to add on additional antibiotics and recently, tigecycline has been used (Wallace et al., 2014; Floto et al., 2016). The treatment issues are further complicated by the ability of two out of three sub-species of M. abscessus to develop macrolide resistance upon exposure to subinhibitory concentrations of the drug (Nash et al., 2009; Bastian et al., 2011; Maurer et al., 2014).

#### Edited by:

Farhat Afrin, Taibah University, Saudi Arabia

#### Reviewed by:

Yusuf Akhter, Babasaheb Bhimrao Ambedkar University, India Anna D. Tischler, University of Minnesota, United States Gyanu Lamichhane, Johns Hopkins Medicine, United States

> \*Correspondence: Thomas Dick td367@njms.rutgers.edu

#### Specialty section:

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

Received: 27 February 2018 Accepted: 23 April 2018 Published: 11 May 2018

#### Citation:

Aziz DB, Teo JWP, Dartois V and Dick T (2018) Teicoplanin – Tigecycline Combination Shows Synergy Against Mycobacterium abscessus. Front. Microbiol. 9:932. doi: 10.3389/fmicb.2018.00932

Indeed, a recent study conducted in a hollow fiber model showed that the standard regimen of clarithromycin, amikacin, and cefoxitin exerted low sterilizing activity within the first 14 days of treatment, and re-growth of the bacteria was seen after this period due to inducible macrolide resistance (Ferro et al., 2016). Demonstrated transmission of M. abscessus between cystic fibrosis patients (Bryant et al., 2016) has increased the urgency to identify novel treatments for this NTM pathogen.

Screening for synergy interactions of approved drugs is an approach to new medicines that allows rapid benchto-bedside translation (Hill and Cowen, 2015). A series of synergy studies have been conducted for M. abscessus and among the combinations that have been identified so far are imipenem + clarithromycin, imipenem + levofloxacin, clarithromycin + linezolid, clarithromycin + vancomycin, clofazimine + amikacin, tigecycline + clarithromycin, tigecycline + clofazimine, tigecycline + linezolid, clavulanate + meropenem, doripenem + rifampicin, biapenem + rifampicin, avibactam + ertapenem, avibactam + tebipenem, and avibactam + panipenem (Miyasaka et al., 2007; Cremades et al., 2009; Shen et al., 2010; van Ingen et al., 2012; Huang et al., 2013; Oh et al., 2014; Singh et al., 2014; Kaushik et al., 2015, 2017; Mukherjee et al., 2017).

To identify novel synergistic combinations, we carried out a large scale study using the checkerboard assay employing two different strategies. The first strategy was to screen combinations of β-lactams with β-lactamase inhibitors (Livermore, 1995; Bebrone et al., 2010). M. abscessus harbors the blamab gene encoding an Ambler class A β-lactamase (Soroka et al., 2014) and an inhibitor might restore activity of β-lactams against M. abscessus. The second strategy was to screen combinations of cell wall-targeting antibiotics with antibiotics that engage intracellular targets. This approach is based on our previous findings that vancomycin displayed (moderate) activity against M. abscessus (Aziz et al., 2017) and showed synergy with clarithromycin (Mukherjee et al., 2017). We screened a total of 180 dual drug combinations against a clinical isolate of M. abscessus and found that the combination of teicoplanin and tigecycline displayed synergistic activity. We characterized the in vitro activity of this novel combination against M. abscessus reference strains and diverse clinical isolates.

## MATERIALS AND METHODS

#### Compounds

The 36 antibiotics and 5 β-lactamase inhibitors used in this study were obtained from commercial sources and dissolved according to the manufacturer's recommendations. Teicoplanin was obtained from Sigma-Aldrich, while tigecycline was obtained from Adooq BioScience. Both antibiotics were dissolved in 90% dimethyl sulfoxide (DMSO).

#### Bacterial Strains and Culture Media

Mycobacterium abscessus Bamboo (Yee et al., 2017) was used for screening of combinations and the subsequent confirmation of synergy hit combinations. For the checkerboard titration assay determination of the activity of the teicoplanin + tigecycline hit against the various M. abscessus subspecies within the M. abscessus complex, M. abscessus subsp. abscessus (ATCC 19977), M. abscessus subsp. bolletii (CCUG 50184-T) and M. abscessus subsp. massiliense (CCUG 48898-T) were used. Reference strains were obtained from the American Type Culture Collection (ATCC) and the Culture Collection University of Goteborg (CCUG), respectively. For further characterization of the teicoplanin + tigecycline combination in the macrolide resistance induction assay, M. abscessus subsp. abscessus (ATCC 19977) harboring the T28 sequevar of erm41 gene, conferring inducible resistance upon exposure to sub-inhibitory concentrations of macrolides (Nash et al., 2009; Bastian et al., 2011) was used. For determination of synergy of teicoplanin + tigecycline against a variety of clinical isolates, strains were obtained from the strain collection of the clinical microbiology laboratory at the National University Hospital, Singapore. The strains were characterized by the lab as previously described (Aziz et al., 2017). For the evaluation of the bactericidal activity of the synergy combination M. abscessus subsp. abscessus (ATCC 19977) was used.

Liquid cultures were grown in standard mycobacterium medium, Middlebrook 7H9 broth (BD Difco) supplemented with 0.5% albumin, 0.2% glucose, 0.085% sodium chloride, 0.0003% catalase, 0.2% glycerol and 0.05% Tween 80. Solid cultures were grown on Middlebrook 7H10 agar (BD Difco) supplemented with 0.5% albumin, 0.2% glucose, 0.085% sodium chloride, 0.5% glycerol, 0.0003% catalase and 0.006% oleic acid.

Mycobacterium abscessus bacterial work was carried out under BSL-2 conditions according to approved biosafety protocols.

### Checkerboard Titration Assay

This assay was carried out in 96-well microtiter plates as previously described (Hsieh et al., 1993; Kaushik et al., 2015), with some modifications. Drugs were tested within the range of concentrations of either 0–25 µM or 0–50 µM, at twofold serial dilutions. For each combination, 8 concentrations of a drug were tested for synergy against 11 concentrations of another drug. Hence, for each two-drug combination screened for synergy, 88 different combination concentrations are tested. A total of 180 different two-drug combinations were tested in this study. For the screening of combinations, this assay was carried out using the Tecan D300e Digital Dispenser for dispensing of drugs. For confirmation of the teicoplanin + tigecycline hit, as well as its subsequent characterization against sub-species, clinical isolates and induced cultures, drugs were dispensed manually. Results were reproducible between the two methods of dispensing drugs for this assay. Briefly, this assay was carried out in 96-well flat bottom plates, with two different compounds, with a starting inoculum of an optical density at 600 nm (OD600) of 0.05 (10<sup>7</sup> colony forming units or cfu/mL) in a final volume of 200 µL. The culture for the starting inoculum was diluted from a pre-culture at mid-log phase (OD<sup>600</sup> = 0.4 to 0.6). The plates were sealed using parafilm, put in an airtight container with moist tissue and incubated for 3 days at 37◦C on an orbital shaker at 110 rpm. Each plate had a media-only control, a drug free control as well as a positive control of clarithromycin at 20 µM. After 3 days of

incubation, the cultures in the wells were manually re-suspended before OD<sup>600</sup> was read in the plate reader (Tecan Infinite 200 Pro) and used to calculate growth inhibition percentage of each well. The Fractional Inhibitory Concentration Index (FICI) was used to analyze the results from the checkerboard assay. FICI was calculated by using the concentrations at which at least 90% inhibition of the culture in the well as compared to the drug free culture was observed. It was computed as FICI = [(concentration of drug A in combination/concentration of drug A when used alone) + (concentration of drug B in combination/concentration of drug B when used alone)] (Hsieh et al., 1993). Synergy is defined as FICI ≤ 0.5, indifference is defined as 0.5 < FICI ≤ 4, and antagonism is defined as FICI > 4 (Hsieh et al., 1993).

#### Macrolide Resistance Induction Assay

M. abscessus subsp. abscessus (ATCC 19977) mid-log phase culture was diluted to OD<sup>600</sup> = 0.05 and treated with clarithromycin at a sub-inhibitory concentration of 0.075 µM (fourfold lower than clarithromycin MIC<sup>50</sup> (concentration that causes 50% growth inhibition). An untreated culture was set up as a control. Cultures were grown to mid-log phase overnight and then subjected to the checkerboard titration assay as described above.

#### Bactericidal Assay

Bactericidal activity determinations were carried out in 14 mL round bottom tubes with the compounds added at set concentrations, with a starting inoculum of OD<sup>600</sup> 0.05 (10<sup>7</sup> cfu/mL) in a final volume of 1 mL. The culture for the starting inoculum was diluted from a pre-culture at mid-log phase (OD<sup>600</sup> = 0.4 to 0.6). Tubes were incubated for 3 days at 37◦C with shaking at 160 rpm. After 3 days of drug exposure, 10 µL of the cultures were plated at different dilutions in 12 well plates containing 2 mL of 7H10 agar in each well. The plates were sealed with parafilm and incubated at 37◦C for 4 days and then colonies were counted. We report fold-kill, which is the

TABLE 1 | Outcome of screening 110 combinations of β-lactams and β-lactamase inhibitors against Mycobacterium abscessus Bamboo: 6 two-drug hits.


S, synergistic; I, indifferent; N.D, not determined. Synergy is defined as FICI ≤ 0.5. However, FICI values cannot be calculated for combinations involving β-lactamase inhibitors or other drugs that do not have MIC values when used individually. Hence, in the tables, synergistic interactions are defined as either combinations with FICI ≤ 0.5 or combinations that exhibit potentiation where a fourfold or more reduction in concentration of both drugs when used together is observed to inhibit 90% of growth as compared to when they are each used alone. Indifferent is defined as 0.5 < FICI ≤ 4. Where FICI cannot be calculated, 'indifferent' is defined as a less than fourfold in concentration of each antibiotic needed to achieve inhibition of 90% of growth when used together as compared to when the antibiotic is used alone. AVI, avibactam; VAB, vaborbactam; CLA, clavulanate; SUL, sulbactam; TZB, tazobactam.

reduction in cfu/mL of the treated culture compared to the time zero untreated control.

#### RESULTS

#### Screening of 180 Two-Drug Combinations for Synergy Against M. abscessus Identifies 11 Hits

We screened a total of 180 two-drug combinations of approved antibiotics for their growth inhibition potency against the clinical isolate M. abscessus Bamboo using the checkerboard assay. Hits were defined as combinations that showed at least a fourfold decrease in concentration of each drug that was needed to achieve 90% inhibition as compared to the concentration needed to achieve that same level of inhibition when either drug was used alone. Screening of β-lactam – β-lactamase inhibitor combinations, identified 6 hits out of 110 combinations (**Table 1**). Screening of combinations of cell envelope targeting drugs with antibiotics that inhibit intracellular targets, identified 5 hits out of 70 combinations (**Table 2**). Taken together, the screen identified 11 primary hits (6.1% hit rate) which were re-confirmed with fresh solids (**Table 3**).

Three of our combination hits, panipenem + avibactam, tebipenem + avibactam, and amoxicillin + avibactam were reported previously (Dubee et al., 2015a; Kaushik et al., 2017).

Out of our eight novel hits, the potencies of ampicillin + avibactam, panipenem + vaborbactam, tebipenem + vaborbactam, and ceftobiprole + linezolid were only modest, with MIC<sup>90</sup> (concentrations that inhibit 90% of growth) of 25 + 19 µM, 50 + 2 µM, 10 + 10 µM, and 38 + 19 µM (**Table 3**).

One of our novel hits involved the glycopeptide ramoplanin in combination with clarithromycin, however, this is not unexpected since we had previously reported synergy between the glycopeptide vancomycin with clarithromycin (Mukherjee et al., 2017).

Three novel hits showed encouraging synergy effects: Ramoplanin + tigecycline, vancomycin + tigecycline and teicoplanin + tigecycline inhibited growth at 5 + 0.8 µM, 2 + 1 µM and 3 + 1 µM, respectively (**Table 3**). Ramoplanin, vancomycin, and teicoplanin are all glycopeptides. Ramoplanin is not well absorbed and unstable in the bloodstream due to hydrolysis of the lactone bond (Farver et al., 2005). This makes ramoplanin unsuitable to repurpose for use in treatment of M. abscessus lung infections. As teicoplanin shows systemic exposure upon intravenous or intramuscular administration and has been reported to have a better safety and efficacy profile compared to vancomycin (Svetitsky et al., 2009), we characterized the activity of teicoplanin in combination with the glycylcycline tigecycline in more detail.

#### Teicoplanin + Tigecycline Displays Activity Against Reference Strains Representing the Three Subspecies of the M. abscessus Complex

To determine whether the teicoplanin + tigecycline combination shows similar attractive potency across the three subspecies of the M. abscessus complex, we carried out the checkerboard titration assay to determine the FICI value of the combination against the reference strains M. abscessus subsp. abscessus ATCC 19977, M. abscessus subsp. bolletii CCUG 50184-T and M. abscessus subsp. massiliense CCUG 48898-T. The teicoplanin + tigecycline combination was synergistic against all three subspecies (**Table 4**). These results suggest that this novel combination is active across the phylogenetically divergent M. abscessus complex.

TABLE 2 | Outcome of screening 70 combinations of cell envelope-targeting antibiotics with antibiotics targeting intracellular targets against M. abscessus Bamboo: 5 synergistic two-drug hits.


S, synergistic; I, indifferent; N.D, not determined. For definition of terms see legend of Table 1. LZD, linezolid; LVX, levofloxacin; MXF, moxifloxacin; AZM, azithromycin; CLR, clarithromycin; TGC, tigecycline; RFB, rifabutin. <sup>a</sup>Reported previously in Mukherjee et al. (2017).

### Teicoplanin + Tigecycline Retains Its Activity Against M. abscessus subsp. abscessus ATCC 19977 Cultures Displaying Induced Macrolide Resistance

The checkerboard titration assay was performed using M. abscessus subsp. abscessus ATCC 19977 cultures that had been exposed to a sub-inhibitory concentration of clarithromycin to induce macrolide resistance to determine whether the teicoplanin + tigecycline combination retains its activity under these conditions. The combination still exhibited synergy against the culture with induced macrolide resistance as seen by its FICI value of 0.22 (**Table 4**).

#### Teicoplanin + Tigecycline Shows Potent Activity Against M. abscessus Clinical Isolates

The teicoplanin + tigecycline combination showed potent growth inhibition activity against the screening strain as well as the reference strains representing the three subspecies of M. abscessus. This suggests that most clinical M. abscessus strains may be susceptible to this combination.

TABLE 3 | Reconfirmation of 11 two-drug hits identified from screening of 180 combinations against M. abscessus Bamboo.


The synergy or potentiation concentration of the 11 drug pair hits identified in the screens shown in Tables 1, 2 are reported. Results shown are the mean of two replicates. Standard deviations were ±50% of shown values.

To provide evidence for a widespread susceptibility of M. abscessus to the teicoplanin + tigecycline combination, we tested its activity against a collection of clinical isolates covering various subspecies of M. abscessus, including clarithromycin resistant as well as clarithromycin sensitive strains. The combination displayed synergy against 70.4% of the isolates with FICI values ranging from 0.32 to 0.48 (**Table 5**). This result indicates that this combination is active against a large number of M. abscessus isolates.

#### Teicoplanin + Tigecycline Is Not Bactericidal Against M. abscessus subsp. abscessus ATCC 19977

To determine whether the teicoplanin + tigecycline combination shows bactericidal activity against M. abscessus, cultures were treated with the drug combination and the effect on viability was determined by cfu enumeration as described in Section "Materials and Methods." The teicoplanin + tigecycline combination showed no bactericidal activity.

## DISCUSSION

A synergy screen of 180 dual antibiotic combinations against M. abscessus yielded a total of 11 hits. Six hits were obtained from combinations of β-lactams with β-lactamase inhibitors, and five hits from combinations of cell walltargeting antibiotics with antibiotics that have intracellular targets.

From the analyses of combinations of β-lactams with β-lactamase inhibitors, the most striking observation is that all six hits involved a non-β-lactam-based β-lactamase inhibitor. 4 out of the 6 hits involved avibactam. This is consistent with previous reports showing that this inhibitor is effective against M. abscessusβ-lactamases (Ehmann et al., 2012; Soroka et al., 2014; Dubee et al., 2015a; Kaushik et al., 2017). The remaining 2 hits involved the novel non-β-lactam-based β-lactamase inhibitor vaborbactam, which has not been previously studied for activity against M. abscessus β-lactamases. Activity of avibactam and now vaborbactam suggests that it may be worthwhile to characterize other types of non-β-lactam β-lactamase inhibitors like phosphonates, hydroxamates, or vanadate-catechol complexes, in combination with β-lactams for any potentiation effect of the combinations against M. abscessus (Bebrone et al., 2010).

It is to note that out of the three sub-classes of β-lactams we tested, avibactam appears not to improve the activity of cephalosporins, in contrast to a previous report describing potentiation between ceftaroline and avibactam against M. abscessus (Dubee et al., 2015b). A recent study by Kaushik et al. (2017) which focused on combinations of avibactam, sulbactam and tazobactam with carbapenems against M. abscessus reported 3 hits, with potentiation observed for combinations of avibactam with ertapenem, tebipenem, or panipenem (Kaushik et al., 2017). In this study, we could confirm potentiation for combinations of

avibactam with tebipenem or panipenem, however, we did not observe any potentiation between avibactam and ertapenem. Another study found potentiation between clavulanate and meropenem, which we also did not observe (Kaushik et al., 2015). The reasons for these discrepancies remain to be determined (Fisher et al., 2005). One possible explanation may be the use of different M. abscessus strains. This study used the clinical isolate M. abscessus Bamboo in the initial screening, while other studies used other strains including M. abscessus subsp. abscessus ATCC19977. In comparison to avibactam, vaborbactam improved the activity of selected compounds only from the carbapenems but not from the penicillins.

Vaborbactam is a new β-lactamase inhibitor and is the first one to contain a cyclic boronic acid structure (Lomovskaya et al., 2017). Despite its difference in structure from avibactam, both β-lactamase inhibitors were able to potentiate the activity of the same two carbapenems, panipenem and tebipenem. Panipenem is an earlier carbapenem and the drug needs to be administered together with betamipron to block its deactivation by dehydropeptidase I (Papp-Wallace et al., 2011). Tebipenem is a more recently discovered carbapenem and is the first oral drug of this class (Papp-Wallace et al., 2011).

From the analyses of combinations of cell wall-targeting antibiotics with drugs that have intracellular targets, we obtained five novel hits, with the teicoplanin + tigecycline combination being most attractive. Teicoplanin + tigecycline combination displayed synergy at a similar range across reference strains representing the three subspecies of M. abscessus with growth inhibitory combination concentrations of 2–3 µM teicoplanin + 1–2 µM tigecycline. The combination also retained

TABLE 4 | Synergy concentrations and FICI values of teicoplanin + tigecycline combination against M. abscessus screening strain and three reference strains of the M. abscessus sub-species.


'Pre-treated with clarithromycin' shows the respective data for a strain displaying induced macrolide resistance due to pre-treatment with a sub-inhibitory concentration of clarithromycin (see section "Materials and Methods"). Results shown are the mean of two replicates. Standard deviation was ±50% of shown values. Synergy is defined as FICI ≤ 0.5. Checkerboard assay was conducted with 7H9 broth.

TABLE 5 | Synergy concentrations and FICI values of the teicoplanin + tigecycline combination against 14 clinical M. abscessus isolates.


Results shown are the mean of two replicates. Standard deviations were ±50% of shown values. Synergy is defined as FICI ≤ 0.5. Indifference is defined as 0.5 < FICI ≤ 4, and antagonism is defined as FICI > 4. N.D, not determined.

activity against most clinical isolates. A limitation of the combinations tested in this category is, that we only tested combinations of tigecycline with glycopeptides and not with other classes of cell wall-targeting antibiotics such as β-lactams, and this should be explored in future studies.

Teicoplanin is a glycopeptide and acts by interacting with the D-ala-D-ala terminal of the muramyl-pentapeptide which results in inhibition of the cell wall peptidoglycan synthesis (Parenti, 1986). The drug is reported to have good tissue and cellular penetration (Parenti, 1986). Teicoplanin was found to have lower adverse event rates compared to the glycopeptide vancomycin (Svetitsky et al., 2009). Tigecycline is a glycylcycline acting via inhibiting protein synthesis (Olson et al., 2006). Both teicoplanin and tigecycline are administered intravenously, which may limit their application. However, it is noteworthy that despite this limitation tigecycline is used to treat M. abscessus infections (Wallace et al., 2014). The exact molecular mechanism by which the synergistic combination of teicoplanin + tigecycline exerts its activity remains to be determined. Tigecycline may have limited ability to penetrate the bacterium to gain access to its intracellular target. One may speculate that with the administration of teicoplanin together with tigecycline, teicoplanin is able to 'weaken' the bacterial cell wall and allow greater penetration of tigecycline into the bacterium.

#### CONCLUSION

This study has identified teicoplanin + tigecycline as a novel synergistic combination against M. abscessus in vitro. The drug pair can now be tested in M. abscessus animal models of infection and/or in patients.

#### REFERENCES


#### AUTHOR CONTRIBUTIONS

DA, VD, and TD conceived the idea, developed the strategy, and wrote the manuscript. DA carried out the experiments. JT provided and characterized the clinical isolates.

#### FUNDING

Research reported in this publication was supported by the National Institute of Allergy and Infectious Diseases of the National Institutes of Health under Award Number R01AI132374 and by the Cystic Fibrosis Foundation under Award Number DICK17XX00 to TD. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health or the Cystic Fibrosis Foundation. This research was also supported by the Singapore Ministry of Health National Medical Research Council under its TCR Flagship grant NMRC/TCR/011- NUHS/2014 as part of the Singapore Programme of Research Investigating New Approaches to Treatment of Tuberculosis (SPRINT-TB; www.sprinttb.org). TD holds a Toh Chin Chye Visiting Professorship at the Department of Microbiology and Immunology, Yong Loo Lin School of Medicine, National University of Singapore.

#### ACKNOWLEDGMENTS

We thank Wei Chang Huang, Taichung Veterans General Hospital, Taichung, Taiwan for the M. abscessus Bamboo strain.


and prevention of nontuberculous mycobacterial diseases. Am. J. Respir. Crit. Care Med. 175, 367–416. doi: 10.1164/rccm.200604-571ST


**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 Aziz, Teo, Dartois and Dick. 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.

fmicb-09-00932 May 9, 2018 Time: 16:41 # 8

# Mycobacterium avium Infection in a C3HeB/FeJ Mouse Model

Deepshikha Verma<sup>1</sup> , Megan Stapleton<sup>1</sup> , Jake Gadwa<sup>1</sup> , Kridakorn Vongtongsalee<sup>1</sup> , Alan R. Schenkel<sup>1</sup> , Edward D. Chan2,3,4 and Diane Ordway<sup>1</sup> \*

<sup>1</sup> Mycobacteria Research Laboratories, Department of Microbiology, Immunology and Pathology, Colorado State University, Fort Collins, CO, United States, <sup>2</sup> Department of Medicine, Denver Veterans Affairs Medical Center, Denver, CO, United States, <sup>3</sup> Departments of Medicine and Academic Affairs, National Jewish Health, Denver, CO, United States, <sup>4</sup> Division of Pulmonary Sciences and Critical Care Medicine, University of Colorado Anschutz Medical Campus, Aurora, CO, United States

Infections caused by Mycobacterium avium complex (MAC) species are increasing worldwide, resulting in a serious public health problem. Patients with MAC lung disease face an arduous journey of a prolonged multidrug regimen that is often poorly tolerated and associated with relatively poor outcome. Identification of new animal models that demonstrate a similar pulmonary pathology as humans infected with MAC has the potential to significantly advance our understanding of nontuberculosis mycobacteria (NTM) pathogenesis as well as provide a tractable model for screening candidate compounds for therapy. One new mouse model is the C3HeB/FeJ which is similar to MAC patients in that these mice can form foci of necrosis in granulomas. In this study, we evaluated the ability of C3HeB/FeJ mice exposure to an aerosol infection of a rough strain of MAC 2285 to produce a progressive infection resulting in small necrotic foci during granuloma formation. C3HeB/FeJ mice were infected with MAC and demonstrated a progressive lung infection resulting in an increase in bacterial burden peaking around day 40, developed micronecrosis in granulomas and was associated with increased influx of CD4<sup>+</sup> Th1, Th17, and Treg lymphocytes into the lungs. However, during chronic infection around day 50, the bacterial burden plateaued and was associated with the reduced influx of CD4<sup>+</sup> Th1, Th17 cells, and increased numbers of Treg lymphocytes and necrotic foci during granuloma formation. These results suggest the C3HeB/FeJ MAC infection mouse model will be an important model to evaluate immune pathogenesis and compound efficacy.

Keywords: Mycobacterium avium, C3HeB/FeJ mouse model, immunity, pathology, nontuberculosis mycobacteria

### INTRODUCTION

Infections due to nontuberculosis mycobacteria (NTM) are increasing worldwide, resulting in a **serious** public health problem (Prevots and Marras, 2015). NTM are mycobacterial species other than Mycobacterium tuberculosis complex and Mycobacterium leprae, that can cause pulmonary and extrapulmonary disease in vulnerable individuals, and are reported throughout the world (Prevots and Marras, 2015; Bryant et al., 2016). Given the ubiquitous nature of NTM in the environment, it is likely that repeated exposures from multiple sources- such as shower heads, swimming pools, and Jacuzzi baths and soil- increases the likelihood of established disease in susceptible individuals (Chan et al., 2010; Bryant et al., 2016). Other reports demonstrate the

#### Edited by:

Veronique Anne Dartois, Rutgers, The State University of New Jersey, United States

#### Reviewed by:

Eric Nuermberger, Johns Hopkins University, United States Michael Henry Cynamon, Syracuse VA Medical Center, United States

> \*Correspondence: Diane Ordway d.ordway@colostate.edu

#### Specialty section:

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

Received: 28 June 2018 Accepted: 19 March 2019 Published: 03 April 2019

#### Citation:

Verma D, Stapleton M, Gadwa J, Vongtongsalee K, Schenkel AR, Chan ED and Ordway D (2019) Mycobacterium avium Infection in a C3HeB/FeJ Mouse Model. Front. Microbiol. 10:693. doi: 10.3389/fmicb.2019.00693

**169**

possibility that cystic fibrosis patients acquire NTM infection by contact to fomites or from patient to patient transmission (Bryant et al., 2016). Recent population-based studies have demonstrated this worldwide increase in NTM began in 2000 and currently in the United States, the prevalence of NTM lung disease exceeds that of tuberculosis by ∼10-fold (Epson et al., 2012; Prevots and Marras, 2015).

The most common NTM causing disease and outbreaks in the United States are species within the MAC—comprised of at least 9 species including M. avium, M. intracellulare, and M. chimaera, followed by Mycobacterium abscessus complex (including M. abscessus sensu stricto, M. massiliense and M. bolletii), Mycobacterium chelonae and Mycobacterium kansasii (De Groote and Huitt, 2006). The two most common preexisting conditions for NTM lung disease are emphysema and bronchiectasis, both of which may be acquired or heritable in origin (Chan et al., 2010; Prevots and Marras, 2015). Other host factors and phenotypes that are associated with NTM lung disease include advanced age, thin body habitus often with thoracic cage abnormalities such as scoliosis and pectus excavatum, gastroesophageal reflux, and use of inhaled corticosteroids and anti-tumor necrosis factor-alpha (anti-TNF-α) therapies (Shang et al., 2011; Bryant et al., 2016; Honda et al., 2018).

The two major clinical forms of MAC lung disease are manifested by two major radiographic patterns—the nodular bronchiectatic and fibrocavitary forms—which have differing lung pathology and bacterial burdens (Kartalija et al., 2013). MAC lung disease associated with upper lobe fibrocavitary pattern occurs dominantly in men with chronic obstructive pulmonary disease (COPD) (Glassroth, 2008). The fibrocavitary pattern is characterized by the formation of granulomas and increased bacterial burden (Gadkowski and Stout, 2008). The nodularbronchiectasis form is often associated with immunocompetent women with granuloma formation in the airway walls and lower bacterial burdens (Glassroth, 2008). However, each type is not exclusively seen in one gender, and both types can be evident in a single patient (Okumura et al., 2008). The number of patients with MAC lung disease has increased worldwide, emphasizing the need for improved MAC modeling systems (Kartalija et al., 2013).

Treatment recommendations for MAC lung disease rely on a paucity of small clinical drug trials and mostly on expert opinion based on combined years of clinical experience (Griffith et al., 2007). Clarithromycin or azithromycin – used as part of a multidrug regimen – is currently the most important antibiotic in the treatment of MAC and many of the other NTM infections (Griffith et al., 2007; van Ingen et al., 2013). The present treatment guidelines for MAC lung disease recommend the combination of a macrolide (clarithromycin or azithromycin) with ethambutol and a rifamycin (rifampin), given for a minimum of 12 months after sputum culture conversion, with or without an aminoglycoside (streptomycin or amikacin given for the first several months) (Griffith et al., 2007; van Ingen et al., 2013). However, this recommended treatment regimen is only considered successful in roughly 50 to 60% of patients, with some recrudescence due to relapse and others due to a new infection from the environment (Griffith et al., 2007; van Ingen et al., 2013). The lack of success of the recommended treatment regimen (macrolide, ethambutol, rifamycin, aminoglycoside combination) is largely due to drug toxicities resulting in poor tolerability.

Four main animal models are routinely used for preclinical MAC experiments: C57BL/6, Balb/c, nude and beige mice infected by aerosol, intranasal, and intravenous exposure with MAC (Gangadharam et al., 1989; Pedrosa et al., 1994; Roger and Bermudez, 2001; Haug et al., 2013; Andrejak et al., 2015; Moraski et al., 2016; Blanchard et al., 2018). The majority of compound screening studies have been carried out with these mouse models due to their low cost and abundance of immunological reagents albeit a major drawback with them is a lack of necrotic granuloma formation (Abendano et al., 2014; Andrejak et al., 2015; Shin et al., 2015). In contrast, C3HeB/FeJ mice, develop highly organized encapsulated necrotic, hypoxic lesions following a M. tuberculosis infection (Kramnik et al., 2000; Obregon-Henao et al., 2013; Henao-Tamayo et al., 2015). Using a forward genetics approach, a region was identified as responsible for the 54.0-cM location of chromosome 1, termed the "super-susceptibility to tuberculosis-1" (sst1) locus (Kramnik et al., 2000). This locus was responsible for a reduced ability to control M. tuberculosis replication in the lungs. The susceptible sst1 allele was also shown to be responsible for the formation of caseous necrosis in the lungs (Kramnik et al., 2000). The C3HeB/FeJ mouse model has been principally used to study compound screening, host immune response and vaccine efficacy with M. tuberculosis infection (Obregon-Henao et al., 2013; Henao-Tamayo et al., 2015) because they are capable of forming necrotic, hypoxic tubercle granulomas (Pichugin et al., 2009; Obregon-Henao et al., 2013; Henao-Tamayo et al., 2015). A MAC infection mouse model with the ability to develop necrotic granulomas is required to understand if a compound or vaccineinduced immune response can enter the site of necrosis to eradicate the bacilli.

Hence, we undertook a study to infect C3HeB/FeJ mice with an aerosol of a rough strain of Mycobacterium avium 2285 to determine if they develop a progressive infection and develop necrotic foci as well as characterize the cellular immune response. To better characterize the immune and pathologic responses induced by MAC infection in the C3HeB/FeJ mice would greatly improve the usefulness of this animal model for the testing of urgently needed new antimicrobial compounds and vaccines.

### MATERIALS AND METHODS

#### Mice

Specific pathogen-free female C3Heb/FeJ, 6 to 8 weeks old, were purchased from the Jackson Laboratories (Bar Harbor, ME). Mice were maintained in the biosafety level 3 facilities at Colorado State University and were given sterile water, chow, bedding, and enrichment for the duration of the experiments. The specificpathogen-free nature of the mouse colonies was demonstrated by testing sentinel animals. All experimental protocols were approved by the Animal Care and Use Committee of Colorado State University.

#### MAC Model

fmicb-10-00693 April 1, 2019 Time: 18:5 # 3

The M. avium 2285 strain with a rough colony morphology and positive for biofilm formation was obtained from a pulmonary MAC patient with a fibrocavitary form of disease (gift from Drs. Stephen Holland and Kenneth Olivier, National Institute of Allergy and Infectious Diseases). The M. avium 2285 strain was found to be of high virulence in Beige mouse model infected with a high dose aerosol (1.0 × 10<sup>9</sup> CFU) (unpublished data). The inoculum was prepared by thawing the bacterial vial, sonicating for 10 to 15 s, and vortexing to remove any clumps that formed during freezing. Thereafter, the mycobacterial suspension was obtained from the vial with a 1-ml tuberculin syringe fitted with a 26.5-gauge needle and expelled back into the vial. This procedure was repeated into the vial 20 times without removing the needle to mix the suspension and break up any small clumps of bacilli (Ordway et al., 2008b; Obregon-Henao et al., 2015).

C3HeB/FeJ mice were infected with MAC using a Glas-Col aerosol generator (Glas-Col, Terre Haute, IN, United States), using the clinical isolate M. avium 2285 rough strain (henceforth referred to as MAC 2285) calibrated to deliver 2,000 bacteria into the lungs per mouse (Obregon-Henao et al., 2013; Henao-Tamayo et al., 2015). The following day, five mice were euthanized and their whole lungs, spleens, and livers were harvested to determine the baseline bacterial burden. The organs were homogenized in phosphate-buffered saline (PBS), and serial dilutions were plated on nutrient 7H11 agar and tryptic soy agar (TSA) for 3 weeks at 37◦C, then CFU were enumerated (Bryant et al., 2016).

The dose of MAC 2285 resulted in progressive infection without showing signs of mortality at the 2,000 CFU. The increased susceptibility of the C3HeB/FeJ mice to mycobacteria justified the use of lower infective doses of MAC 2285 compared to the other more resistant mice (C57BL/6, Balb/c, nude or beige) which require an infective dose of 1 × 108–1 × 10<sup>11</sup> CFUs (Ji et al., 1994; Andrejak et al., 2015).

#### Animal Infection

Using a Glas-Col aerosol generator, C3HeB/FeJ mice were infected by an aerosol of the clinical isolate MAC 2285 rough strain calibrated to deliver 2,000 bacteria into the lung, spleen and liver per mouse (Obregon-Henao et al., 2013; Henao-Tamayo et al., 2015). The whole lung, spleen, and liver from the mice for each condition at each time point (n = 5) were harvested to quantify bacterial burden (lung, spleen and liver, n = 5), histological analysis (lungs, n = 5), and flow cytometric analyses (lung and spleen, n = 5) at 20, 30, 40, 50, and 60 days post-challenge. In brief, bacterial loads were determined by plating whole organ serial dilutions of organ homogenates onto nutrient 7H11 and TSA agar plates. The rationale for using both 7H11 and TSA plates are any additional environmental NTM contamination of our CFUs can be identified on the TSA plates. Colony forming units (CFU) were quantified after incubation for 3 weeks at 37◦C. The tissues of additional groups of mice were analyzed for pathological and immune response analysis. The results shown in this study are representative of two independent experiments using five animals per time point.

### Histological Analysis

The whole lung lobe from each mouse was fixed with 10% formalin in PBS. Sections from these tissues were stained with haematoxylin-eosin and with Ziehl-Neelsen acid-fast stains (Obregon-Henao et al., 2013; Henao-Tamayo et al., 2015).

#### Lung and Spleen Cell Digestion

Briefly, single cell suspensions were prepared as described previously (Ordway et al., 2008a; Shang et al., 2011; Obregon-Henao et al., 2013). The lungs and spleens were aseptically removed, teased apart and treated with a solution of deoxyribonuclease IV (DNAse) (Sigma Chemical, 30 µg/ml) and collagenase XI (Sigma Chemical, 0.7 mg/ml) for 45 min at 37◦C. To obtain a single-cell suspension, the organs were gently passed through cell strainers (Becton Dickinson, Lincoln Park, NJ, United States). Any remaining erythrocytes were lysed with Gey's solution (0.15 M NH4Cl, 10 mM KHCO3) and the cells were washed with Dulbecco's modified Eagle's minimal essential medium. Cell suspensions from each individual mouse were incubated with monoclonal antibodies to various cytokines and cell surface markers labeled with fluorescein isothiocyanate (FITC), phycoerythrin (PE), peridinin chlorophyll-a protein (PerCP), or allophycocyanin (APC) at 4◦C for 30 min in the dark as described previously (Ordway et al., 2008a; Shang et al., 2011; Obregon-Henao et al., 2013). Total cell numbers were determined by flow cytometry using BDTM Liquid Counting Beads, as described by the manufacturer (BD PharMingen, San Jose, CA, United States). All analyses were performed with an acquisition of at least 100,000 total events for T cells and 200,000 total events for antigen-presenting cells.

#### Cell Surface and Intracytoplasmic Cytokine Staining

Cells were first stained for cell surface markers as indicated above and thereafter the same cell suspensions were prepared for intracellular staining as described previously (Ordway et al., 2007, 2008a; Obregon-Henao et al., 2013). For flow cytometry analysis, single-cell suspensions prepared from the lungs of naïve and MAC 2285 infected mice were re-suspended in PBS containing 0.1% of sodium azide. Cells were incubated in the dark for 25 min at 37◦C with pre-determined optimal titrations of specific antibody (directly conjugated to FITC, PE, PerCP, APC, Pacific Blue, or Alexa 700); or after biotin antibody incubations washed and incubated for 25 min more with streptavidin Qdot800 (Invitrogen), followed by two washes in PBS containing 4% sodium azide. Measurement of intracellular cytokines was conducted by pre-incubating lung cells with monensin (3 µM) (Golgi Stop, BD PharMingen), anti-CD3 and anti-CD28 (both at 0.2 µg/10<sup>6</sup> cells) for 4 h at 37◦C, 5% CO2. The cells were then surface stained, incubated for 30 min at 37◦C, washed, fixed and permeabilized with Perm Fix/Perm Wash (BD Pharmingen). Finally, the cells were stained with fluorescentlylabeled antibodies directed against intracellular Foxp3 (FJK-16s), IL-17 (clone N49-653) and IFN- γ many of the IFN-γ (clone B27) and separately with their respective isotype controls (BD Pharmingen) for a further 30 min. All the samples were run on a Becton Dickinson LSR-II and data were analyzed using FACSDiva v5.0.1 software. Cells were gated on lymphocytes based on characteristic forward and side scatter profiles. Individual cell populations were identified according to their presence of specific fluorescent-labeled antibodies. All the analyses were performed with a minimum acquisition of 100,000 events for T cells and 200,000 events for macrophages and dendritic cells.

#### Statistical Analysis

fmicb-10-00693 April 1, 2019 Time: 18:5 # 4

Data are presented using the mean values from 5 mice per group performed in duplicate experiments. The Student t-test was used to assess statistical significance between groups of mice. In addition, bacterial burdens for each experimental condition were analyzed with GraphPad Prism version 4 (GraphPad Software, San Diego, CA), using analysis of variance (ANOVA) comparison test. Data are presented using the mean values (n = 5) plus or minus the standard error of the mean (SEM). Significance was considered with a P value of < 0.05 (Obregon-Henao et al., 2015).

#### RESULTS

#### Course of MAC Infection in C3HeB/FeJ Mice

Our goal was to assess the course of an aerosol infection during acute and chronic infection with MAC 2285 in mice. We evaluated aerosol exposure of a clinical, drug- resistant strain of MAC 2285 (∼2,000 bacilli per mouse) to determine if there is a progressive infection in the C3HeB/FeJ during the acute and relative chronic phases after infection.

Mice infected with MAC were evaluated for bacterial loads in the lungs (**Figure 1A**), spleens (**Figure 1B**), and livers (**Figure 1C**) after 1, 20, 30, 40, 50, and 60 days of infection.

C3HeB/FeJ mice showed an increase in bacterial burden 40 days after infection, peaking at ∼5.2 log<sup>10</sup> CFU in the lungs (**Figure 1A**), followed by bacilli persisting in the lungs and plateauing at ∼5.1 log<sup>10</sup> CFU during the chronic phase of disease. C3HeB/FeJ mice showed a delayed dissemination to the spleens (**Figure 1B**), and livers (**Figure 1C**) over the first 30 days of infection, followed by an increase of ∼4.0 log<sup>10</sup> CFU in the spleens peaking at 50 days and ∼3.0. log<sup>10</sup> CFU in the livers during the chronic phase of disease.

These results demonstrate that aerosol exposure of MAC can progress slowly over the acute and subacute phases of disease, reaching a peak in bacterial burden in the lungs ay day 40. Furthermore, by the chronic phase of the infection, the bacterial burden remained persistently elevated in all the organs examined (Kitahara et al., 1997; Boyle et al., 2015).

#### Development of Pathology in MAC Infection in C3HeB/FeJ Mice

The lung histopathology and acid-fast staining over the course of infection are shown in **Figure 2** (left / middle and right columns, respectively). As early as day 40 after infection, C3HeB/FeJ mice developed small pulmonary necrotic lesions (**Figure 2**, left / middle columns), and multiple extracellular clusters of bacilli were observed within these lesions (**Figure 2**, arrows). During the chronic stage, the lungs of MAC -infected C3HeB/FeJ mice developed increased numbers of granulomas, greater numbers of intracellular and extracellular acid fast bacilli and progressive development of lesions, with notable lung tissue inflammation and small areas of necrosis (**Figure 2** day 40–60). During the later stages of chronic disease much of the lung tissue was grossly consolidated, with inflammation and necrosis was evident. Overall, MAC infection of C3HeB/FeJ mice demonstrate disease progression resulting in pulmonary inflammation, lung consolidation and small foci of necrosis in pulmonary lesions.

#### Kinetic Influx of the T Cell Response in M. avium Infected C3HeB/FeJ Mice

T cells were gated with a primary gate on viable FSClow vs. SSClow lymphocytes and then on CD3<sup>+</sup> T cells and compared to the isotype controls (**Figures 3A–C**), and analyzed for changes in the total mean cell number of CD3+CD4<sup>+</sup> and CD3+CD8<sup>+</sup> cells over the course of infection.

T cells that migrated to the lungs of naive control and MAC-infected C3HeB/FeJ mice were harvested and analyzed by flow cytometry. Specifically, naive CD3+CD4<sup>+</sup> and CD3+CD8<sup>+</sup> T cells, as well as the kinetic influx of activated T effector cells (CD3+CD4+CD44hi and CD3+CD8+CD44hi) and effector memory T cells (CD3+CD4+CD44hiCD127<sup>+</sup> and CD3+CD8+CD44hiCD127+) were determined in naive mice and MAC-infected C3HeB/FeJ mice (Henao-Tamayo et al., 2014). As expected, at days 30 and 40, when bacterial burden increased (**Figure 1A**) and pathology began to develop (**Figure 2**), increased numbers of (CD3+CD4<sup>+</sup> and CD3+CD8<sup>+</sup> T cells) were already found in MAC-infected C3HeB/FeJ mice compared to naive mice (**Figures 3D,G**). In addition, in MAC infected C3HeB/FeJ mice the (CD3+CD4+CD44hi and CD3+CD8+CD44hi) activated T effector cells showed a delay in trafficking into the lungs peaking on day 50 compared to the overall CD4<sup>+</sup> and CD8<sup>+</sup> populations (**Figures 3E,H**). Similarly, in the MAC infected C3HeB/FeJ mice, the effector memory T cells (CD3+CD4+CD44hiCD127<sup>+</sup> and CD3+CD8+CD44hiCD127+) cells demonstrated a lag in reaching the lungs, peaking on day 50 compared to the overall CD4<sup>+</sup> and CD8<sup>+</sup> populations, thereafter, these memory T cells declined (**Figures 3F,I**).

#### Kinetic Influx of CD4 and CD8 T Cells Expressing IFN-γ, IL-17, and Foxp3 in MAC-Infected C3HeB/FeJ Mice

The kinetic influx of IFN- γ -producing CD3+CD4<sup>+</sup> and CD3+CD8<sup>+</sup> T cells, IL-17 producing CD3+CD4<sup>+</sup> and CD3+CD8<sup>+</sup> T cells, and suppressive T regulatory cells (CD3+CD4<sup>+</sup> and CD3+CD8<sup>+</sup> CD25+Foxp3+), were determined in the lungs of the MAC infected C3HeB/FeJ mice compared to naive mice. Interestingly, after day 40–50, when bacterial burden (**Figure 1A**) and pathology (**Figure 2**) were increasing, higher numbers of CD4<sup>+</sup> IFN-γ-producing T cells were found in the lungs which declined at 60 days (**Figures 4A,D**). However, with respect to CD8<sup>+</sup> IFNγ-producing T cells in C3HeB/FeJ mice, what was observed

FIGURE 1 | Mycobacterium avium infection in C3HeB/FeJ mice. Bacterial counts in the lungs (A), spleens (B), and livers (C) of C3HeB/FeJ mice infected with 2,000 MAC 2285 rough strain per mouse are shown. CFU were determined at 1, 20, 30, 40, 50, and 60 days after infection by plating serial dilutions of organ homogenates on nutrient 7H11 and TSA agar and quantifying CFU after 3 weeks incubation at 37◦C. The C3HeB/FeJ mice after 40 days of infection showed increased bacterial burden ∼4.5 to 5.0 log<sup>10</sup> in the lungs followed by bacterial burden plateauing during chronic phase of infection ∼50–60 days (A). The C3HeB/FeJ mice showed a delay in bacterial dissemination in both the spleens (B) and liver (C) peaking after ∼50–60 days to ∼4.0 log<sup>10</sup> CFU in the spleen and ∼3.0 log<sup>10</sup> CFU in the liver. Results represent the average of two experiments (n = 5 mice per time point) and are expressed as log<sup>10</sup> CFU (±SEM).

increased clusters of acid-fast-staining bacilli, accumulating in areas of necrosis (N) (arrows). Magnifications, 1X (left), 20X (middle) and 100X (right).

to be overall lower were observed and increased slightly between 40 and 50 days over the course of the infection compared to naive mice (**Figure 4B**). MAC-infected C3HeB/FeJ mice demonstrated increased numbers of CD4<sup>+</sup> and CD8<sup>+</sup> T cells producing IL-17 (**Figures 4B,E**), and regulatory T cell markers (CD25hi and Foxp3) (**Figures 4C,F**), that were associated with increased bacterial burden and organ pathology.

#### Kinetics of Macrophages and Dendritic Cells in M. avium-Infected C3HeB/FeJ Mice

Flow cytometric analysis was performed on MAC- infected C3HeB/FeJ mice and naïve mice in order to analyze the influx of CD11b<sup>+</sup> macrophages and CD11c<sup>+</sup> dendritic cells into the lungs. Between 40 and 50 days, a time at which the infection progressed, we observed a relative increase in cells staining positive for CD11b<sup>+</sup> and CD11c<sup>+</sup> MHC class II expression compared to the naïve mice (**Figures 5A,D**). This was consistent with the presence of increasing numbers of CD11b<sup>+</sup> and CD11c<sup>+</sup> cells expressing IL-27 (**Figures 5B,E**) and programmed deathligand 1 (PD-L1) markers although both plateaued at 50 days of infection (**Figures 5C,F**) related with T cell immune suppression (Shu et al., 2017).

The slow increase of MHC II<sup>+</sup> macrophages and dendritic cells between 40 and 50 days of infection, was associated with an increase number of CD4<sup>+</sup> T cells producing IFN-γ and IL-17 increased arriving in the lungs at this later time range, whereas

cells were gated with a primary gate on viable FSClow vs. SSClow lymphocytes compared to the isotype controls (A–C) and then on CD3<sup>+</sup> T cells, and analyzed for changes in the total mean cell number of CD3+CD4<sup>+</sup> and CD3+CD8<sup>+</sup> (D,G) cells over the course of infection. Shown are the numbers of activated effector T cells (CD4+CD44hi and CD8+CD44hi) (E,F) and memory T cells (CD4+CD44hiCD127<sup>+</sup> and CD8+CD44hiCD127+) (F,I) migrating to the lungs of C3HeB/FeJ mice peaking after 40–50 days of infection. During the later stages on day 60 of infection, C3HeB/FeJ mice expressed diminished numbers of activated effector and memory T cells. Results represent the mean number of cells of five mice for each condition from two independent experiments (±SEM).

FIGURE 4 | Kinetic influx of CD4 and CD8 T cells expressing IFN-γ, IL-17, and Foxp3 in M. avium-infected C3HeB/FeJ mice. Lung cells obtained from MAC-infected C3HeB/FeJ and naive control mice were analyzed by flow cytometry. (A–C) show CD4<sup>+</sup> effector cells expressing IFN-γ, IL-17 and CD25hiFoxp3. MAC-infected C3HeB/FeJ mice showed increased numbers of CD4+IFN-γ <sup>+</sup> and CD4+IL-17<sup>+</sup> producing cells peaking between 40 and 50 days after infection with concomitant increased numbers of CD4+CD25hiFoxp3<sup>+</sup> cell compared to the naïve mice. (D–F) show CD8<sup>+</sup> effector cells expressing IFN-γ, IL-17, and CD25hiFoxp3. Interestingly, CD8+IFN-γ <sup>+</sup> effector cells demonstrated decreased migration to the lungs with a concomitant increased number of CD8+IL-17<sup>+</sup> and CD8+CD25hiFoxp3<sup>+</sup> cells compared to the naïve mice. The data are expressed as the mean number of pulmonary cells in each organ ± SEM (n = 5 mice for each condition from two independent experiments (±SEM).

there was only a modest increase in the number of CD8<sup>+</sup> T cells that produced IFN-γ (**Figure 4D**).

#### DISCUSSION

While MAC is less virulent than M. tuberculosis, these NTM are capable of causing chronic lung disease. Thus, development of a sustained progressive MAC infection in a mouse strain is an important endeavor to not only better characterize the immunopathogenesis in an in vivo model, but to also test candidate antimicrobials. Previous work using immunocompetent mouse strains demonstrated rapid clearance with lower amounts of MAC (Gangadharam, 1995), making such models suboptimal for more chronic human lung infection. In a MAC aerosol model using high doses of bacteria (1 × 109– 1 × 10<sup>11</sup> CFU per mouse) in comparing BALB/c, C57BL/6, Nude and Beige mice, the Nude mice were identified as the most sensitive mouse strain to MAC while treatment efficacy was most noticeable in BALB/c mice (Torrelles et al., 2002; van Ingen et al., 2013; Andrejak et al., 2015; Cha et al., 2015). Beige mice are not uncommonly utilized as an in vivo mouse model for MAC infection and display many immune deficiencies similar to those occuring in AIDS patients such as a dominant Th2 response resulting in enhanced susceptibility to infection with NTM following either intravenous or aerosol infection (Caverly et al., 2015). Indeed, dissemination of MAC from gut of the Beige mice was shown to be more rapid than that seen in wildtype C57BL/6 mice (Torrelles et al., 2002). The Beige mice also had a defect in the influx of neutrophils to the site of infection and transfusing exogenous neutrophils mitigated their susceptibility; neutrophil depletion studies with wildtype C56BL/6 mice demonstrated increased susceptibility (Torrelles et al., 2002). Subsequent studies in another mouse strain revealed that a defect in CXCR2 chemokine signaling impaired the early and rapid recruitment of neutrophils in response to MAC infection (Torrelles et al., 2002; van Ingen et al., 2013). Nevertheless, the initial discovery and further confirmation of sustained MAC disease in the Beige mouse have encouraged many investigators to use this model to screen potential chemotherapeutic compounds for the treatment of MAC disease (Obregon-Henao et al., 2015; van Ingen et al., 2013).

Finding demonstrate that C3HeB/FeJ mice infected with MAC (∼2,000 organisms per mouse) – as compared to an inoculum dose of 1 × 108–1 × 10<sup>9</sup> organisms per mouse typically utilized with infection of mouse models (Torrelles et al., 2002; van Ingen et al., 2013; Andrejak et al., 2015; Cha et al., 2015) resulted in a progressive infection, distinct temporal immune responses, and histopathologic changes resulting in small foci of necrosis in the lungs.

More specifically, after 20 days of infection, the C3HeB/FeJ mice showed a burden of 3.5 log<sup>10</sup> CFU in the lungs followed by an increase of 1.8 log<sup>10</sup> CFU during the chronic phase of disease. C3HeB/FeJ mice showed a delayed dissemination to the

spleen, and liver, over the first 30 days of infection, followed by an increase of ∼4.0 log<sup>10</sup> CFU in the spleen and ∼3.0 log<sup>10</sup> CFU in the liver during the chronic phase of disease. An advantage of the C3HeB/FeJ model over the currently utilized MAC mouse models that use 1 × 108–1 × 10<sup>9</sup> CFU (Torrelles et al., 2002; van Ingen et al., 2013; Andrejak et al., 2015; Cha et al., 2015) to infect mice is that much fewer bacteria were used to infect the mice, mimicking the relative paucibacillary load seen in non-HIV infected subjects infected with NTM (Farhi et al., 1986; Hibiya et al., 2013; Kobashi et al., 2013).

Our prior studies with M. tuberculosis infection of C3HeB/FeJ mice demonstrated necrotic granulomas in the lungs during the chronic phase of the disease, allowing for bacterial replication, while fewer tubercular granulomas were present in the spleen and liver, leading to better bacterial clearance (Henao-Tamayo et al., 2015). However, presumably due to the reduced virulence of MAC compared to M. tuberculosis, MAC demonstrated a slower disease progression prior to 20 days where thereafter the development of small foci of necrosis appeared in MAC-infected granulomas. An additional advantage of the C3HeB/FeJ model over currently utilized MAC mouse models is the presence of small foci of necrosis in the granulomas - absent in the latter mouse strains (Torrelles et al., 2002; van Ingen et al., 2013; Andrejak et al., 2015; Cha et al., 2015) - since formation of these characteristic lesions is important in the pathogenesis of human NTM lung infections (Farhi et al., 1986; Hibiya et al., 2013; Kobashi et al., 2013).

Development of small foci of necrosis is an important aspect of NTM pathogenesis. Bacteria residing in necrotic granulomas have access to an abundant source of carbon in the form of cholesterol and triglycerides (Obregon-Henao et al., 2013), and this may be the reason for the relatively high NTM bacilli growth in these lesions. Furthermore, antibiotic penetration is significantly reduced in necrotic granulomas (Farhi et al., 1986), which could help explain the protracted antibiotic therapy required to eradicate NTMs. Thus, it is important to have these aspects in our mouse models that are utilized for compound and vaccine screening.

C3HeB/FeJ mice infected with MAC showed increased numbers of CD11b<sup>+</sup> macrophages and CD11c<sup>+</sup> dendritic cells expressing MHC class II during acute phase of the infection, however, this waned during the chronic phase of the disease when expression of IL-27 and PD-1L increased. Studies investigating macrophage deactivation by examining the expression of a panel of IFN-γ-inducible genes and activation of Janus kinase (JAK)- STAT pathway in MAC-infected macrophages showed reduced expression of IFN-γ-inducible genes—MHC class II gene Eβ; MHC class II transactivator; IFN regulatory factor-1; and Mg21, a gene coding for a GTP-binding protein in MAC-infected macrophages (Hussain et al., 1999). These studies support our results showing a reduction of MHC class II-expressing CD11b<sup>+</sup> macrophages and CD11c<sup>+</sup> dendritic cells during chronic disease (Taylor-Robinson et al., 1994).

C3HeB/FeJ mice infected with MAC was associated with increased numbers of CD3+CD4+CD44hi and CD3+CD8+CD44hi activated T effector cells and CD3+CD4+CD44hiCD127<sup>+</sup> and CD3+CD8+CD44hiCD127<sup>+</sup> memory T cells between 40 and 50 days after infection, followed by diminished numbers. In addition, C3HeB/FeJ mice were infected with MAC also demonstrated increased numbers of IFN- γ and IL-17 CD4<sup>+</sup> cells as well as increased Foxp3+CD4<sup>+</sup> T regulatory cells during late chronic infection (60 days). Interestingly, MAC-infected C3HeB/FeJ mice displayed increasing numbers of CD8<sup>+</sup> T cells expressing IL-17 and Foxp3 over the course of the infection period but a modest increase of CD8+IFNγ <sup>+</sup> T cells. CD4<sup>+</sup> T cell depleted C57BL/6 mice clearly indicate a role for CD4<sup>+</sup> T cell needed for control of an intranasal infection of MAC (Florido et al., 1999). In this previous study, IFN-γ <sup>+</sup> depletion before and during MAC infection led to increased bacterial burden in then lung, spleen and liver, suggesting a protective role for IFN-γ against this pathogen.

In previous studies using C57BL/6 mice and C3HeB/FeJ mice, we reported that protection waned with specific alterations of the adaptive immune response, such as increased numbers of regulatory T cells (Henao-Tamayo et al., 2015), concomitant with reduced numbers of protective CD8<sup>+</sup> T cells. However, our previous studies infecting C3HeB/FeJ mice with M. tuberculosis (Henao-Tamayo et al., 2015) showed a reduction of both IFN-γ producing CD4<sup>+</sup> and CD8<sup>+</sup> T cell responses with a concomitant increase in the influx of CD4<sup>+</sup> and CD8<sup>+</sup> Foxp3<sup>+</sup> T regulatory cells. Due to the lower dose of MAC used to infect C3HeB/FeJ mice, it is plausible that MAC disease requires additional immune deficits to display progression or a higher infection dose.

A MAC-infected C3HeB/FeJ mouse model is supported by other murine studies showing early in vivo expression of IFN-γ during MAC-infection correlated with resistance to the infection (Gangadharam, 1995; Young and Bermudez, 2001; Appelberg, 2006). The use of specific neutralizing antibodies in vivo led to the identification of IFN-γ and TNF-α as protective cytokines acting at the effector level of resistance to MAC (Gangadharam, 1995; Young and Bermudez, 2001; Appelberg, 2006).

### CONCLUSION

In conclusion, MAC infections are becoming an emerging problem worldwide. To deal with the increasing number of infected NTM patients and the resultant morbidity and mortality caused by these pathogens, multiple laboratories are focused on developing new preclinical models to screen new or repurposed compounds to fight the emergence of these pathogens. Our studies support the use of a MAC infected C3HeB/FeJ mouse model for testing candidate compounds against MAC.

#### ETHICS STATEMENT

This study was carried out in accordance with the recommendations of NRC Guide for the Care and Use of Laboratory Animals (National Research Council, 2010), the

requirements of the Public Health Service (PHS) Grants Administration Manual, and The Animal Welfare Act as amended. CSU files assurances with the DHHS Office of Extramural Research, Office of Laboratory Animal Welfare (OLAW), the Public Health Service, and adheres to NIH standards and practices for grantees. The protocol was approved by the Colorado State Universities Animal Care and Usage Committee.

#### AUTHOR CONTRIBUTIONS

DO, EC, AS, DV, and MS conceived and designed the study, acquired, analyzed, and interpreted the data, and drafted the manuscript. JG, KV, and AT analyzed and interpreted the data, and drafted and revised the manuscript. DO, EC, AS, and DV conceived and designed the study, analyzed

### REFERENCES


and interpreted the data, and drafted, revised, and approved the manuscript.

### DEDICATION

In memory of Dr. Ian Orme.

#### FUNDING

This work was supported by funding from the National Institutes of Health [Grant Nos. R21AIO99534-02 and AIO99534-02, and NIH/NIAID task order HHSN272201000009I/HHSN27200001, principal investigator (PI), Anne J. Lenaerts, co-PI, DO] (program officers, Christine Sizemore, Jim Boyce, and Andre McBride).



(MAC) isolates in mice. Clin. Exp. Immunol. 98, 210–216. doi: 10.1111/j.1365- 2249.1994.tb06127.x


**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 Verma, Stapleton, Gadwa, Vongtongsalee, Schenkel, Chan and Ordway. 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.

# MmpL3 as a Target for the Treatment of Drug-Resistant Nontuberculous Mycobacterial Infections

Wei Li 1†, Amira Yazidi 2,3†, Amitkumar N. Pandya4†, Pooja Hegde<sup>4</sup> , Weiwei Tong<sup>1</sup> , Vinicius Calado Nogueira de Moura<sup>1</sup> , E. Jeffrey North<sup>4</sup> \*, Jurgen Sygusch2,3 \* and Mary Jackson<sup>1</sup> \*

<sup>1</sup> Mycobacteria Research Laboratories, Department of Microbiology, Immunology and Pathology, Colorado State University, Fort Collins, CO, United States, <sup>2</sup> Biochimie et Médecine Moléculaire, Université de Montréal, Montréal, QC, Canada, <sup>3</sup> Groupe d'Étude des Protéines Membranaires, Université de Montréal, Montréal, QC, Canada, <sup>4</sup> Department of Pharmacy Sciences, School of Pharmacy and Health Professions, Creighton University, Omaha, NE, United States

#### *Edited by:*

Thomas Dick, Rutgers, The State University of New Jersey, Newark, United States

#### *Reviewed by:*

Shashank Gupta, Brown University, United States Kapil Tahlan, Memorial University of Newfoundland, Canada

#### *\*Correspondence:*

E. Jeffrey North jeffreynorth@creighton.edu Jurgen Sygusch jurgen.sygusch@umontreal.ca Mary Jackson mary.jackson@colostate.edu

†Co-first authors.

#### *Specialty section:*

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

> *Received:* 18 May 2018 *Accepted:* 21 June 2018 *Published:* 10 July 2018

#### *Citation:*

Li W, Yazidi A, Pandya AN, Hegde P, Tong W, Calado Nogueira de Moura V, North EJ, Sygusch J and Jackson M (2018) MmpL3 as a Target for the Treatment of Drug-Resistant Nontuberculous Mycobacterial Infections. Front. Microbiol. 9:1547. doi: 10.3389/fmicb.2018.01547 Nontuberculous mycobacterial (NTM) pulmonary infections are emerging as a global health problem and pose a threat to susceptible individuals with structural or functional lung conditions such as cystic fibrosis, chronic obstructive pulmonary disease and bronchiectasis. Mycobacterium avium complex (MAC) and Mycobacterium abscessus complex (MABSC) species account for 70–95% of the pulmonary NTM infections worldwide. Treatment options for these pathogens are limited, involve lengthy multidrug regimens of 12–18 months with parenteral and oral drugs, and their outcome is often suboptimal. Development of new drugs and improved regimens to treat NTM infections are thus greatly needed. In the last 2 years, the screening of compound libraries against M. abscessus in culture has led to the discovery of a number of different chemotypes that target MmpL3, an essential inner membrane transporter involved in the export of the building blocks of the outer membrane of all mycobacteria known as the mycolic acids. This perspective reflects on the therapeutic potential of MmpL3 in Mycobacterium tuberculosis and NTM and the possible reasons underlying the outstanding promiscuity of this target. It further analyzes the physiological and structural factors that may account for the apparent looser structure-activity relationship of some of these compound series against M. tuberculosis compared to NTM.

Keywords: *Mycobacterium abscessus*, nontuberculous mycobacteria, MmpL3, mycolic acids, drug development

## INTRODUCTION

The prevalence of pulmonary nontuberculous mycobacterial (NTM) infections caused by Mycobacterium avium complex (MAC) and Mycobacterium abscessus complex (MABSC) species is increasing worldwide and poses a particular threat to susceptible individuals with structural or functional lung conditions such as cystic fibrosis (CF), chronic obstructive pulmonary disease, and bronchiectasis (Park and Olivier, 2015; Parkins and Floto, 2015; Bryant et al., 2016; Floto et al., 2016; Martiniano et al., 2016). Treatment options for NTM pulmonary infections involve lengthy (12–18 months) combination regimens with antibiotics that lack bactericidal activity and are associated with significant toxicity. For pulmonary MAC, the recommended treatment includes a macrolide, rifamycin, and ethambutol to which intravenous amikacin may be added. Treatment of pulmonary MABSC typically consists of an oral macrolide in conjunction with intravenous or inhaled amikacin, and one or more of the following drugs: intravenous cefoxitin, imipenem, or tigecycline, in addition to oral antibiotics (minocycline, clofazimine, moxifloxacin, linezolid; Floto et al., 2016). The impermeability of the cell envelopes of NTM to drugs and the high number of efflux systems and antibiotic inactivation mechanisms with which NTM are typically endowed confer upon these microorganisms high intrinsic protection against antibiotics (Brown-Elliott et al., 2012). There is clearly an urgent need for more active and bettertolerated drugs to improve therapeutic outcome (Jarand et al., 2011; Maurer et al., 2014; Park and Olivier, 2015; Martiniano et al., 2016).

In last 3 years, the phenotypic screening of compound libraries against NTM has yielded a number of hits with activity against MABSC, MAC, or both complexes. Interestingly, several of these compounds appear to kill NTM through the inhibition of MmpL3, an essential mycolic acid transporter present in all mycobacteria whose therapeutic potential in the treatment of M. tuberculosis infections was highlighted in a number of recent studies (Sacksteder et al., 2012; Kondreddi et al., 2013; Lun et al., 2013; Rao et al., 2013; Remuinan et al., 2013; Yokokawa et al., 2013; Li et al., 2016, 2017; Poce et al., 2016, 2018; Stec et al., 2016; Degiacomi et al., 2017). The availability of cidal inhibitors against this new target, some of which have already demonstrated activity in in vivo models of MABSC infection (Dupont et al., 2016; De Groote et al., in revision; Pandya et al., in revision), provides much-needed novel opportunities for the treatment of pulmonary NTM infections.

This perspective reflects on the therapeutic potential and promiscuity of MmpL3 in NTM, and discusses recent findings from our laboratories toward understanding the basis for the better activity and looser structure-activity relationship of MmpL3 inhibitors against M. tuberculosis compared to NTM.

### THE PHENOTYPIC SCREENING OF COMPOUND LIBRARIES AGAINST NTM IDENTIFIES INHIBITORS OF MmpL3

In the last 3 years, the screening of compound libraries, including libraries of TB actives, against MABSC and MAC, has yielded a number of potent hits that appear to target the mycolic acid transporter MmpL3. These include indole-2-carboxamides (ICs) (Franz et al., 2017; Kozikowski et al., 2017; Low et al., 2017), benzothiazole amides (De Groote et al., in revision), and a piperidinol derivative (PIPD1) (Dupont et al., 2016; Low et al., 2017). Earlier work on analogs of the M. tuberculosis MmpL3 inhibitor BM212 had further highlighted the activity of pyrole derivatives against a variety of NTM including M. avium, M. gordonae, M. smegmatis, and M. marinum (Biava et al., 1999, 2007; Biava, 2002). A subset of these hits and their MIC against M. tuberculosis, M. avium and MABSC isolates (including M. abscessus subsp. abscessus, M. abscessus subsp. Massiliense, and M. abscessus subsp. bolletti) is presented in **Table 1** along with that of other chemotypes reported to inhibit MmpL3 activity in M. tuberculosis (i.e., the 1,2-ethylene diamine SQ109, the tetrahydropyrazolopyrimide carboxamide THPP1, and the adamantyl urea AU1235) (Grzegorzewicz et al., 2012; La Rosa et al., 2012; Tahlan et al., 2012; Remuinan et al., 2013).

ICs have previously been identified as a novel chemical scaffold showing promise in the treatment of tuberculosis (Kondreddi et al., 2013; Lun et al., 2013; Rao et al., 2013; Stec et al., 2016). Based on their high anti-MABSC potency, bactericidal activity on extracellularly- and intracellularly-grown bacilli and promising safe pharmacological profile (Franz et al., 2017; Kozikowski et al., 2017; Pandya et al., in revision), two lead molecules were advanced for efficacy studies in a mouse model of MABSC infection. Oral administration of the lead compounds showed a statistically significant reduction in bacterial load in the lungs, spleen and liver of MABSCinfected mice compared to an untreated control group, with one of the two compounds (compound # IC25; see **Table 1**) showing similar efficacy to amikacin (Pandya et al., in revision). The intrapulmonary delivery of a lead benzothiazole amide compound also demonstrated in vivo efficacy in a mouse model of chronic MABSC lung infection (De Groote et al., in revision), whereas the piperidinol-based compound PIPD1 was reported to restrict bacterial growth in a zebrafish model of MABSC infection (Dupont et al., 2016). An interesting property of MmpL3 inhibitors first revealed in M. tuberculosis is their ability to synergize with a number of other antimycobacterial drugs or drug candidates including rifampicin, bedaquiline, clofazimine, and β-lactams (Li et al., 2017). Our preliminary studies with IC25 (see **Table 1**) in MABSC similarly point to the existence of a synergistic interaction between this MmpL3 inhibitor and clofazimine in MABSC (Table S1).

Collectively, these results highlight the therapeutic potential of MmpL3 inhibitors in the treatment of NTM infections and provide a strong incentive to develop these compounds into new generation antimycobacterial drugs as their inclusion in anti-NTM drug regimens has the potential to lead to the faster and more efficient clearance of NTM from infected tissues.

### MmpL3: A PROMISCUOUS MYCOBACTERIAL TARGET

The reason why so many chemical scaffolds kill M. tuberculosis and NTM through the inhibition of MmpL3 remains incompletely understood. MmpL3 belongs to the Resistance, Nodulation and Division (RND) superfamily of transporters that requires the transmembrane electrochemical proton gradient for activity. The observation that the most common resistance mutations identified in both M. tuberculosis and MABSC tend to map to a transmembrane region of MmpL3 overlapping with functional residues required for proton translocation or proton-driven conformational changes in the transporter has led to the hypothesis that inhibitors might target the proton relay site of MmpL3 (Belardinelli et al., 2016). MmpL3 inhibitors are typically lipophilic (logP ∼2.6–7.0) and many suffer from poor aqueous solubility which likely favors their concentration in

TABLE 1 | MICs of MmpL3 inhibitors against M. tuberculosis [Mtb], M. abscessus complex species (M. abscessus subsp. abscessus ATCC 19977 [Mabs]; M. abscessus subsp. massiliense CIP 108297 [Mmas]; M. abscessus subsp. bolletii ATCC 14472 [Mbol]), M. avium 104 [Mav], and M. smegmatis recombinant strains expressing different mmpL3 orthologs.


#### TABLE 1 | Continued


(Continued)

#### TABLE 1 | Continued


MICs (in µg/mL) were determined were determined in 96-well microtiter plates at 37◦C in 7H9-ADC-0.05% Tween 80 medium (M. smegmatis), 7H9-OADC-0.05% tyloxapol supplemented with 0.2% casaminoacids, 48µg/ml pantothenate, and 50µg/ml L-leucine (M. tuberculosis mc26206), cation-adjusted Mueller-Hinton broth (M. abscessus complex) or in cation-adjusted Mueller-Hinton broth supplemented with 5% OADC (M. avium) using the resazurin blue test (Martin et al., 2003) and by visually scanning for growth. The orthologs of mmpL3 from M. abscessus (mmpL3abs), M. tuberculosis (mmpL3tb), and M. smegmatis (mmpL3smg) are expressed from the replicative plasmid pMVGH1 under control of the hsp60 promoter in the background of a M. smegmatis null mutant (Msmg1mmpL3) of which the entire mmpL3 ORF was deleted and replaced with a kanamycin-resistance cassette. Control drugs: APRA, apramycin; BDQ, bedaquiline; CFZ, clofazimine; CLA, clarithromycin. nd, not determined. <sup>a</sup>MIC value against M. tuberculosis H37Rv ATCC 27294 (Low et al., 2017); <sup>b</sup>The precise M. avium strain used by Dupont et al. (2016) in the MIC determination of PIPD1 was not indicated and may be different from M. avium 104.

the inner membrane where MmpL3 is located. This property and the extreme vulnerability of MmpL3 (Li et al., 2016) that may allow inhibitors with relatively weak binding affinity to the transporter to still inhibit enough of its activity to cause growth arrest, could explain the bias of phenotypic screens toward small hydrophobic inhibitors of MmpL3. The exquisite vulnerability of MmpL3 may further mask potential secondary targets of the inhibitors as illustrated by THPP derivatives that were found to target another essential mycolic acid-related protein in M. tuberculosis (Cox et al., 2016) and compounds such as SQ109, BM212, and some THPPs that show activity against non-replicating M. tuberculosis bacilli, a property typically not shared by other MmpL3 inhibitors Li W. et al., 2014). The fact that the hydrophobicity of ICs, THPPs, SQ109 analogs and urea derivatives is a key driver of their efficacy provides further support to the notion that the concentration of MmpL3 inhibitors in the phospholipid bilayer plays a key role in their activity (Biava et al., 2005, 2007; Onajole et al., 2010, 2013; Brown et al., 2011; Scherman et al., 2012; Kondreddi et al., 2013; North et al., 2013; Li K. et al., 2014; Poce et al., 2016; Stec et al., 2016; Franz et al., 2017; Kozikowski et al., 2017).

A second mechanism through which high rates of apparent MmpL3 inhibitors may arise from phenotypic screens was proposed after it was found that unspecific uncouplers such as carbonyl cyanide m-chlorophenyl hydrazone (CCCP) or the K<sup>+</sup> ionophore, valinomycin, both abolished MmpL3 activity in M. tuberculosis and M. smegmatis (Li W. et al., 2014). This finding indicated that any compound with the ability to dissipate the proton motive force may indirectly inhibit MmpL3 activity with immediate consequences on mycobacterial growth and viability. Accordingly, and most likely explaining the relatively broad spectrum of activity of some of these compounds including against bacteria devoid of MmpL3 homolog, inhibitors such as the 1,2-ethylenediamine SQ109, the adamantyl urea AU1235 and the 1,5-diarylpyrrole derivative BM212 were found to impact to some degree the membrane potential, the electrochemical proton gradient or both components of the proton motive force of mycobacterial cells (Li K. et al., 2014; Li W. et al., 2014; Feng et al., 2015; Foss et al., 2016). This unspecific activity, however, was later disputed in the case of BM212 and this compound proposed to directly inhibit MmpL3 on the basis of its demonstrated binding to the purified MmpL3 protein from M. smegmatis (Xu et al., 2017).

In conclusion, both direct and indirect mechanisms can lead to MmpL3 inhibition in treated mycobacterial cells and contribute to the promiscuity of the target. While not mutually exclusive, a precise understanding of how these two mechanism(s) play out for each inhibitor to eventually abolish mycolic acid export will require a detailed analysis of how each of them interacts with the transporter and affects the energy metabolism of the bacterium.

#### NTM VS. *M. TUBERCULOSIS* EFFICACY

From the MIC data presented in **Table 1** and previous studies (Biava et al., 1999, 2007; Biava, 2002; Li K. et al., 2014; Franz et al., 2017; Kozikowski et al., 2017; Low et al., 2017; De Groote et al., in revision), it is obvious that the overall activity of MmpL3 inhibitors against NTM, particularly MAC, is less than that observed against M. tuberculosis. While the structural diversity of the chemotypes found to inhibit MmpL3 is very broad, spanning from compounds such as BM212 and THPP1, which are large (for MmpL3 inhibitors) multicyclic compounds, to SQ109 which is an ethylene diamine originally designed as an ethambutol analog, the majority of MmpL3 inhibitors reported to date have come from two other classes that contain the same pharmacophore which are the ureas (e.g., AU1235) and indole-2-carboxamides. The pharmacophore for these classes of MmpL3 inhibitors are two hydrogen bond donors and one hydrogen bond acceptor in the center of the molecule and one bulky lipophilic aliphatic ring (adamantyl, cyclooctyl, cycloheptyl, or substituted cyclohexyl groups) and one aromatic ring on either side of the core. For structure activity relationships, generally, as lipophilicity is increased on either the bulky aliphatic ring (typically through ring expansion or addition of methylene or methyl groups) or aromatic ring (typically through addition of halogens or methyl groups), anti-NTM activity is improved.

Since a number of factors could account for the overall better activity of MmpL3 inhibitors against M. tuberculosis than NTM, including species-specific variations in the structure of MmpL3 orthologs, increased drug efflux/degradation/modification in NTM relative to M. tuberculosis, or reduced compound penetration in NTM, we first sought to compare the MIC of the inhibitors presented in **Table 1** against M. smegmatis recombinant strains expressing different MmpL3 orthologs. To this end, mmpL3 from M. tuberculosis, M. smegmatis and M. abscessus subsp. abscessus were expressed from the same expression plasmid in the background of a M. smegmatis mutant whose endogenous mmpL3 gene was deleted by allelic replacement (Msmg1mmpL3) (Belardinelli et al., 2016). Expressing all orthologs from the same promoter in the same M. smegmatis strain abolished any potential differences in compound uptake, modification and efflux allowing for a direct comparison of the effect of the inhibitors against the three MmpL3 proteins. Importantly, all three mmpL3 orthologs were expressed at comparable levels in this recombinant system (Figure S1). The results of these comparative MIC studies clearly indicated that the MICs of the inhibitors against the different M. smegmatis strains generally reflected their MICs against the Mycobacterium species from which the rescue mmpL3 ortholog originated (**Table 1**). The structure of MmpL3 thus appears to be the main driver of the susceptibility of each Mycobacterium species to these inhibitors.

(gray), TMS-8 (violet purple), TMS-10 (pink), TMS-12 (light orange), TMS-11 (violet), TMS-5 (pale cyan), TMS-4 (marine), TMS-6 (green), TMS-2 (red), TMS-3 (wheat), and TMS-1 (cyan). The TM helices encompass residues whose mutations resulted in significant reduction in transport activity, shown in green, and residues 251, 288, 640, 641, 710, and 715 whose mutation abolished transport activity, colored yellow (Belardinelli et al., 2016). The positions of frequently encountered resistance mutations to one or more MmpL3 inhibitor series are shown as red and salmon spheres centered on the Cα atom of the native MmpL3 residue (Belardinelli et al., 2016). These residues map to TM helices. Resistance mutations also producing a significant reduction in growth are shown in red and those that slightly attenuated growth are colored salmon. (B) Stereo model as in (A) showing regions of the TM helices where the majority of residues are not conserved between MmpL3 orthologs (dark gray). Most of the dissimilar residues represent semi-conservative and non-conservative mutations (see Figure S2). Several of these regions map vicinal to the functional residues and mutations that induce resistance.

To investigate the structural basis for the different susceptibilities of MmpL3 orthologs to various classes of inhibitors, the I-TASSER server (Yang et al., 2015) was used for automated full-length 3D structure prediction of MmpL3 transporters from M. tuberculosis H37Rv, M. abscessus ATCC 19977, M. smegmatis mc<sup>2</sup> 155, and M. avium 104. The top predicted structure for each MmpL3 transporter corresponded to a Cscore of >-1.5 suggesting a correct fold. All MmpL3 orthologs resemble each other (root mean square differences based Calpha atoms < 0.4Å) and had as closest target the crystal structure of the Burkholderia multivorans hopanoid transporter HpnN (PDB 5khnB). Comparison of the predicted structures with that of HpnN yielded a high TM-score >0.8 and low RMSD <2.0 between residues that were structurally aligned by TM-align (Zhang, 2008). The superposition of all three NTM MmpL3 orthologs onto MmpL3 from M. tuberculosis H37Rv shows very similar spatial overlap of the C-alpha positions of essential residues identified in the reference MmpL3 transporter (Belardinelli et al., 2016) that can be seen in **Figure 1A**. Each of the predicted structures contained 12 TM helices.

We next aligned the amino acid sequences of the MmpL3 orthologs among each other using PSI/TM-Coffee (Chang et al., 2012; Figure S2). Essential functional residues identified in M. tuberculosis MmpL3 (namely, residues: 251, 288, 640, 641, 710, and 715; boxed in green in Figure S2; Belardinelli et al., 2016) are conserved and are all located in the central regions of the 12 TM helical bundle that is thought to be involved in proton translocation.

We then searched for sequence dissimilarities among the TM helices given the lipophilicity of the MmpL3 inhibitors. The stereo model in **Figure 1B** shows regions of the transmembrane helices in dark gray where the majority of amino acid residues are not conserved between MmpL3 orthologs and which span the central regions of the 12 TM helical bundle. Most of the dissimilar residues represent semi-conservative and non-conservative mutations that are shown as red boxes on the sequence alignment shown in Figure S2. Several of these regions map vicinal to the functional residues and mutations that induce resistance. The shape differences in the geometries of the hydrophobic and polar side chains alters the packing of the 12 TM helices and is likely to concomitantly modify their dynamical behavior important for transport activity and inhibitor binding. Given that the inhibitors are partitioned into the lipid bilayer, from a thermodynamic perspective, their propensity to interact with the TM helices will further depend on two factors: their ability to interact preferentially with the hydrophobic side chains of the TM helices and their ability to form polar interactions with either backbone or polar side chains. It follows that both the differential helical packing modifying the binding loci of the inhibitors and the nature of the side chains of the TM helices probably account for the ortholog-dependent activity of MmpL3 inhibitors.

## FUTURE DIRECTIONS

There is an unmet medical need for the development of new bactericidal agents to treat pulmonary NTM infections. The novel classes of bactericidal MmpL3 inhibitors that have been reported in the last few years, some of which have demonstrated activity against M. tuberculosis and MABSC in vivo, highlight the therapeutic potential of this transporter in tuberculous and nontuberculous mycobacteria and provide much needed translational opportunities for the treatment of NTM infections. Future research is expected to gain further insight into the structure of MmpL3 and its variations across Mycobacterium species in order to leverage the emerging structure-activity relationship information now available for some of these compound series (Brown et al., 2011; Scherman et al., 2012; North et al., 2013; Li K. et al., 2014; Poce et al., 2016, 2018; Stec et al., 2016; Franz et al., 2017; Kozikowski et al., 2017). Also, critical to the further development of these inhibitors will be the availability of a simple, nonradioactive, and relatively high-throughput assay to screen optimized analogs with increased activity against MmpL3. Currently available cell-free and whole cell-based assays (e.g., Grzegorzewicz et al., 2012; Li W. et al., 2014; Li et al., 2016; Xu et al., 2017) indeed lack the simplicity of use and/or specificity required to rapidly screen such analogs. The development of such assays is currently the object of intense efforts in our laboratories.

### AUTHOR CONTRIBUTIONS

EN, JS, and MJ conceived the project, analyzed the data, and wrote the manuscript. WL, PH, WT, and VC generated and characterized the M. smegmatis recombinant strains, and carried out the MIC determinations and checkerboard assays. AP synthesized inhibitors, contributed to the preparation of **Table 1** and analyzed SAR data. AY performed the MmpL3 modeling studies.

## ACKNOWLEDGMENTS

This work was supported by a grant from the National Institutes of Health/National Institute of Allergy and Infectious Diseases (AI116525) (to MJ and EN) and a grant from the Bill and Melinda Gates Foundation (OPP1181207) (to MJ and JS). The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH. We are grateful to the Global Alliance for TB Drug Development for the provision of NITD-304 and NITD-349.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2018.01547/full#supplementary-material

#### REFERENCES


monomycolate involved in mycolic acid donation to the cell wall core of Mycobacterium tuberculosis. Antimicrob. Agents Chemother. 56, 1797–1809. doi: 10.1128/AAC.05708-11


**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 Li, Yazidi, Pandya, Hegde, Tong, Calado Nogueira de Moura, North, Sygusch and Jackson. 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.

# Lsr2 Is an Important Determinant of Intracellular Growth and Virulence in Mycobacterium abscessus

Vincent Le Moigne<sup>1</sup> , Audrey Bernut<sup>2</sup>† , Mélanie Cortès<sup>3</sup> , Albertus Viljoen<sup>2</sup> , Christian Dupont<sup>2</sup> , Alexandre Pawlik<sup>4</sup> , Jean-Louis Gaillard1,5, Fabienne Misguich<sup>1</sup> , Frédéric Crémazy<sup>1</sup> \*, Laurent Kremer2,6 and Jean-Louis Herrmann1,5 \*

#### Edited by:

Thomas Dick, Center for Discovery and Innovation, Hackensack Meridian Health, United States

#### Reviewed by:

Thomas F. Byrd, The University of New Mexico, United States Joseph Oliver Falkinham, Virginia Tech, United States

#### \*Correspondence:

Frédéric Crémazy frederic.cremazy@uvsq.fr Jean-Louis Herrmann jean-louis.herrmann@aphp.fr

#### †Present address:

Audrey Bernut, Department of Infection, Immunity & Cardiovascular Disease, The Bateson Centre, The University of Sheffield, Sheffield, United Kingdom

#### Specialty section:

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

Received: 05 January 2019 Accepted: 09 April 2019 Published: 30 April 2019

#### Citation:

Le Moigne V, Bernut A, Cortès M, Viljoen A, Dupont C, Pawlik A, Gaillard J-L, Misguich F, Crémazy F, Kremer L and Herrmann J-L (2019) Lsr2 Is an Important Determinant of Intracellular Growth and Virulence in Mycobacterium abscessus. Front. Microbiol. 10:905. doi: 10.3389/fmicb.2019.00905 <sup>1</sup> 2I, UVSQ, INSERM, Université Paris-Saclay, Versailles, France, <sup>2</sup> UMR 9004, Centre National de la Recherche Scientifique, Institut de Recherche en Infectiologie de Montpellier, Université de Montpellier, Montpellier, France, <sup>3</sup> VitamFero, Tours, France, <sup>4</sup> Unité de Pathogénomique Mycobactérienne, Institut Pasteur, Paris, France, <sup>5</sup> APHP, GHU PIFO, Hôpital Raymond-Poincaré – Hôpital Ambroise-Paré, Boulogne-Billancourt, France, <sup>6</sup> INSERM, Institut de Recherche en Infectiologie de Montpellier, Montpellier, France

Mycobacterium abscessus, a pathogen responsible for severe lung infections in cystic fibrosis patients, exhibits either smooth (S) or rough (R) morphotypes. The S-to-R transition correlates with inhibition of the synthesis and/or transport of glycopeptidolipids (GPLs) and is associated with an increase of pathogenicity in animal and human hosts. Lsr2 is a small nucleoid-associated protein highly conserved in mycobacteria, including M. abscessus, and is a functional homolog of the heat-stable nucleoidstructuring protein (H-NS). It is essential in Mycobacterium tuberculosis but not in the non-pathogenic model organism Mycobacterium smegmatis. It acts as a master transcriptional regulator of multiple genes involved in virulence and immunogenicity through binding to AT-rich genomic regions. Previous transcriptomic studies, confirmed here by quantitative PCR, showed increased expression of lsr2 (MAB\_0545) in R morphotypes when compared to their S counterparts, suggesting a possible role of this protein in the virulence of the R form. This was addressed by generating lsr2 knockout mutants in both S (1lsr2-S) and R (1lsr2-R) variants, demonstrating that this gene is dispensable for M. abscessus growth. We show that the wild-type S variant, 1lsr2- S and 1lsr2-R strains were more sensitive to H2O<sup>2</sup> as compared to the wild-type R variant of M. abscessus. Importantly, virulence of the Lsr2 mutants was considerably diminished in cellular models (macrophage and amoeba) as well as in infected animals (mouse and zebrafish). Collectively, these results emphasize the importance of Lsr2 in M. abscessus virulence.

Keywords: non-tuberculous mycobacteria, Mycobacterium abscessus, Lsr2, virulence, pathogenesis, zebrafish, mouse

#### INTRODUCTION

Mycobacterium abscessus is a rapidly growing mycobacterium (RGM) increasingly acknowledged as a serious non-tuberculous mycobacterial (NTM) pathogen (Mougari et al., 2016; Diel et al., 2017). Although it causes extrapulmonary infections (Jeong et al., 2017) and disseminated pulmonary diseases among otherwise healthy individuals (Varghese et al., 2012), it has become

**188**

notorious for the serious threat it poses to patients with cystic fibrosis (CF) and with other underlying lung disorders (Brown-Elliott and Wallace, 2002; Medjahed et al., 2010). M. abscessus infection correlates with a decline in pulmonary function and is an appreciable concern for CF patients requiring lung transplantation (Esther et al., 2010; Qvist et al., 2016; Smibert et al., 2016). M. abscessus has high intrinsic levels of resistance to many antibiotics, making infections with this mycobacterium hard to treat and eradicate (Brown-Elliott and Wallace, 2002; van Dorn, 2017).

Mycobacterium abscessus presents either smooth (S) or rough (R) colony morphotypes leading to different clinical outcomes (Howard et al., 2006; Catherinot et al., 2009; Medjahed and Reyrat, 2009; Ripoll et al., 2009; Roux et al., 2011). They also exhibit different morphological aspects (Howard et al., 2006; Sánchez-Chardi et al., 2011) and virulence phenotypes (Byrd and Lyons, 1999; Catherinot et al., 2007; Bernut et al., 2014a). These differences rely mainly on the presence (in S) or absence (in R) of surface-associated glycopeptidolipids (GPL) (Gutiérrez et al., 2018). The S variant is thought to be the colonizing form, capable of producing mature biofilms and exhibiting a significant motility on soft agar (Howard et al., 2006). The R variant, on the other hand, is impaired in these abilities and forms pronounced serpentine cords, a feature associated with hypervirulence (Byrd and Lyons, 1999; Catherinot et al., 2007; Bernut et al., 2014a; Halloum et al., 2016). Being the most pathogenic RGM, it is not surprising that both variants of M. abscessus resist phagocytosis by immune cells, a trait shared with pathogenic slow-growing mycobacterial (SGM) species, such as M. tuberculosis (Oberley-Deegan et al., 2010; Nessar et al., 2011; Roux et al., 2016). Therefore, the S-to-R transition, which occurs essentially in vivo and is responsible for a significant increase in pathogenicity (Jönsson et al., 2007; Rottman et al., 2007; Bernut et al., 2014b) likely represents an evolutionary adaptation mechanism to the host immune response (Pawlik et al., 2013; Roux et al., 2016).

Genome comparison of three distinct isogenic S/R couples of M. abscessus revealed the presence of genetic lesions in the R variant, such as single nucleotide polymorphisms and/or insertions/deletions, in genes belonging to the GPL biosynthesis and transport locus (Pawlik et al., 2013) and causing the S-to-R transition. Although these mutations can account for the change in colony morphology of M. abscessus, it is unlikely that they fully explain the virulence of the R variant observed in multiple cellular and animal models (Catherinot et al., 2007; Bernut et al., 2014b, 2017). This view is supported by transcriptomic studies performed in the same isogenic S/R pairs, which revealed differential expression for a large set of genes, including lsr2 (Pawlik et al., 2013).

Lsr2 is a nucleoid-associated protein (NAP) and a functional analog of the heat-stable nucleoid-structuring proteins or H-NS (Chen et al., 2008; Gordon et al., 2008; Qu et al., 2013) that is conserved in all actinomycetes and mycobacteria (Chen et al., 2008; Gordon et al., 2008; Qu et al., 2013). Lsr2 was originally shown to be an immunodominant antigen of Mycobacterium leprae (Laal et al., 1991) and has since been reported to regulate a broad range of processes (Chen et al., 2008; Colangeli et al., 2009; Gordon et al., 2010). In Mycobacterium smegmatis, lsr2 mutants show a different colony morphology and are defective in biofilm formation, a phenotype presumably resulting from an altered expression of key surface lipids, such as mycolyl-diacylglycerols (Chen et al., 2006; Arora et al., 2008; Kocíncová et al., 2008; Yang et al., 2017). Subsequent studies also demonstrated that Lsr2 is necessary for mycobacterial conjugal DNA transfer (Nguyen et al., 2010). In contrast to M. smegmatis, in which it is dispensable for planktonic growth, Lsr2 is considered essential in M. tuberculosis (Chen et al., 2006; Colangeli et al., 2007; Arora et al., 2008; Kocíncová et al., 2008; Yang et al., 2017). In addition, it has also been shown to contribute to antibiotic resistance (Colangeli et al., 2007) and to control the expression of a large panel of genes acquired by horizontal gene transfer in M. tuberculosis (Gordon et al., 2010). Like H-NS, Lsr2 preferentially binds to AT-rich sequences and can oligomerize through its protein-protein interaction domain to form filaments along the M. tuberculosis chromosome (Gordon et al., 2010, 2011). Moreover, it has the ability to bridge distant DNA segments, suggesting a role in the organization and compaction of the nucleoid (Chen et al., 2008).

Herein, we generated lsr2 knock-out mutants in both S and R variants of M. abscessus to define the contribution of Lsr2 in major physiological processes that are relevant to the context of increased pathogenicity of the R variant.

### MATERIALS AND METHODS

#### Strains and Culture Media

Mycobacterium abscessus subsp. abscessus S and R 19977-IP strains were used in this study and designated Mabs-S and Mabs-R, respectively. Mycobacterial strains were grown aerobically at 37◦C in Middlebrook 7H9 broth or on Middlebrook 7H11 agar, supplemented with 0.2% glycerol and 1% glucose. When necessary; kanamycin, hygromycin, and zeocin were added to medium at 250 µg/ml, 500 µg/ml (or 1000 µg/ml for selection) and 25 µg/ml, respectively. After infection experiments, bacterial CFU were counted on agar plates (BioMérieux, France), either Columbia blood agar plates after infections in macrophages and amoeba, and H2O<sup>2</sup> exposition, either VCAT (Vancomycin, Colistin sulfate, Amphotericin B, and Trimethoprim) chocolate agar plates after mice infections. The Escherichia coli TOP10 strain (Thermo Fisher) was grown in Lysogeny Broth (LB) medium with or without kanamycin (25 µg/ml).

### Construction of lsr2 Mutants in M. abscessus

The lsr2 downstream and upstream regions were amplified by PCR using the primer pairs MC75/MC76 and MC80/MC81, respectively, then cloned with the zeocin resistance cassette yielding pMC34. Replacement of the endogenous lsr2 gene by the zeocin resistance cassette was performed as described previously (van Kessel and Hatfull, 2007; Medjahed and Reyrat, 2009; Bakala N'Goma et al., 2015). lsr2 knock-out mutants, designated Mabs-S-1lsr2 and Mabs-R-1lsr2 were confirmed by PCR (primer pairs lsr2-5/lsr2V and MC75/ZeoR) (**Supplementary Figure S1**) and Southern blotting (**Figure 2B**).

Complementation was achieved by PCR amplification of lsr2 under the control of its endogenous promoter region (961 bp) using the primers Comp-MAB\_0545+reg-AclI/Comp-MAB\_0545-HpaI and insertion into the integrative vector pMV361. The resulting construct was then introduced into the Mabs-S-1lsr2 strain. Mabs-R-1lsr2 complementation was achieved by PCR amplification (primer pair Comp-MAB\_0545- NdeI/ Comp-MAB\_0545-HindIII) and cloning of lsr2 under the control of the hsp60 promoter in the replicative plasmid pVV16 (Grzegorzewicz et al., 2012) Expression of lsr2 mRNA transcripts was checked by quantitative RT-PCR analysis. To generate the Mabs-S-1lsr2-Clsr2-FLAG strain used for the ChIPqPCR experiments, lsr2 was fused to the FLAG tag by PCR amplification using the primer pair Comp-MAB\_0545-reg-AclI/ Comp-MAB\_0545-FLAG-HpaI and cloning with its endogenous promoter into the integrative plasmid pMVH361. All the primers used in this study are listed in **Supplementary Table S1**.

#### Quantitative RT-PCR Analysis

mRNA was reverse transcribed using the "iScript reverse transcription" kit (Bio-Rad). Quantitative PCR was performed with a MasterMix qPCR (Eurogentec) in a Chromo4 instrument (Bio-Rad), as described previously (Pawlik et al., 2013; Le Moigne et al., 2016).

#### ChIP-qPCR Analysis

For each ChIP library, 50 ml of Mabs-S-1lsr2-Clsr2-FLAG was grown until an OD<sup>600</sup> of 0.5 and fixed in 1% formaldehyde (Euromedex) and lysed using a Precellys grinder (3 cycles: 6,700 rpm −3 × 20 s ON/60 s OFF, Bertin Technologies) and VK05 beads. The bacterial chromatin was sheared in 100–300 bp fragments using a Bioruptor Pico (Diagenode). Chromatin Immunoprecipitation was performed as described previously (Grainger et al., 2006) using IgG M2 anti-FLAG (Sigma, F1804) and Anti-GFP (G6539, Sigma) mouse monoclonal antibodies attached to proteins A/G coupled to magnetic beads (Thermo fisher). The immunoprecipitated DNA and 1% of the total input were reverse-crosslinked and eluted using the iPure v2 kit (Diagenode). The quantitative PCR tests were performed using the SsoFast evergreen supermix (Bio-Rad). Enrichment of Lsr2 binding at the GPL operon operator was calculated using the "Percent Input" method using the primers MAB\_4100c\_qPCR1 and MAB\_4100c\_qPCR3 (**Supplementary Table S1**).

### GPL Analysis

Apolar and polar lipid fractions were obtained from exponentially growing mycobacteria cultured as reported previously (Besra, 1998). The polar lipid fraction containing GPL was separated by TLC using silica gel 60 coated aluminum TLC plates (Merck) and chloroform/methanol (90:10; v/v) as solvent system, as previously described (Sondén et al., 2005; Pawlik et al., 2013). GPL were revealed by treating TLC plates with a mist of 0.2% (w/v) of Anthrone diluted in sulfuric acid, followed by charring.

### Susceptibility Profile to H2O<sup>2</sup>

Exponential growth phase cultures (OD<sup>600</sup> between 0.6 and 0.8) of M. abscessus were diluted to obtain an OD<sup>600</sup> of 0.1 and transferred in two new tubes. One of the two cultures was exposed to 20 mM H2O<sup>2</sup> and the other one to an equivalent volume of sterile water. CFUs were determined by plating aliquots at various time points after the addition of H2O<sup>2</sup> (2, 4, 6, and 8 h) on blood agar plates. For each morphotype and mutant, three independent experiments were performed, and one experiment is shown.

### Intracellular Survival in Macrophages and Amoeba After Infection

Murine J774.2 macrophages (M8) were grown at 37◦C under 5% CO<sup>2</sup> in DMEM medium supplemented with 5% heat-inactivated Fetal Calf Serum (FBS), penicillin (100 IU/ml) and streptomycin (100 µg/ml). M8 infections with the M. abscessus strains were carried out, as described previously (Bakala N'Goma et al., 2015; Roux et al., 2016). CFU were counted after 3 to 5 days of incubation at 37◦C. Amoeba infections were done using Acanthamoeba castellanii (ATCC30010), as described earlier (Bakala N'Goma et al., 2015; Dubois et al., 2018a). At 1, 2, and 3 days of co-culture, the A. castellanii monolayer was disrupted with 0.1% SDS for 30 min at 32◦C and CFU were counted.

### Systemic Infection in Mice

Six to eight-week-old BALB/c mice were infected with M. abscessus strains, as described earlier (Catherinot et al., 2007; Rottman et al., 2007; Bakala N'Goma et al., 2015; Bernut et al., 2015). A Student t-test was carried out to test significance of differences between groups. All procedures were performed according to the institutional and national ethical guidelines and approved by the Comité d'éthique en experimentation animale N◦ 047 with agreement A783223 under the reference APAFIS#11465.

### Zebrafish Infections

All zebrafish experiments were done according to European Union guidelines for handling of laboratory animals<sup>1</sup> and approved by the Comité d'Ethique pour l'Expérimentation Animale de la région Languedoc Roussillon under the reference CEEALR36-1145. Experiments were performed using the golden mutant (Lamason et al., 2005). Zebrafish embryos were obtained, maintained and microinjected in the caudal vein of 30 hpf dechorionated embryos as described (Bernut et al., 2014a, 2015). Survival curves and statistics were carried out as described (Bernut et al., 2014a).

## RESULTS

#### Transcription Levels of lsr2 in M. abscessus R and S Strains

Our initial comparative genomic analyses performed on three isogenic M. abscessus S/R pairs identified several mutations in the GPL biosynthesis/transport locus that are responsible for the S-to-R transition (Pawlik et al., 2013). Microarray data obtained from RNAseq highlighted the differential expression

<sup>1</sup>http://ec.europa.eu/environment/chemicals/lab\_animals/home\_en.htm

pattern of several genes, including MAB\_0545 (Pawlik et al., 2013; **Supplementary Table S2**). MAB\_0545, encoding the Lsr2 protein, was found to be moderately more expressed in the R than in the S variants of the CIP104536 and ATCC19977 strains as well as in one clinical isolate, referred to as the CF R strain (Catherinot et al., 2009). To confirm these results, quantitative RT-PCR analysis was performed on lsr2 using mRNA isolated from both the S and R variants of all three strains. **Figure 1** shows that transcription of lsr2 is increased in the R strains relative to their S counterparts, particularly in the two type strains, thus validating the microarray data. Because Lsr2 appears to be involved in the virulence and the antibiotic resistance of M. tuberculosis (Colangeli et al., 2009; Gordon et al., 2010), we reasoned that induction of Lsr2 expression in the R strain may explain, at least in part, the increase of virulence in the R over the S variant.

### Lsr2 Null Mutants Are Not Affected in Their Glycopeptidolipid Profile

To address the impact of Lsr2 on the M. abscessus morphotype, null mutants of lsr2 were generated in both variants, designated Mabs-S-1lsr2 and Mabs-R-1lsr2, using the recombineering method (van Kessel and Hatfull, 2007) that was adapted to M. abscessus (Bakala N'Goma et al., 2015; Halloum et al., 2016; Dubois et al., 2018b; Laencina et al., 2018). This was achieved by replacing lsr2 with a zeocin resistance cassette using homologous recombination (**Figure 2A**). Proper allelic exchange was subsequently confirmed by PCR (**Supplementary Figure S1**) and Southern blotting (**Figure 2B**). qRT-PCR revealed that complementation with a functional lsr2 gene was partial (**Supplementary Figure S2**).

The viability of the mutant indicates that lsr2 is dispensable in M. abscessus, as previously reported in M. smegmatis (Chen et al., 2006; Arora et al., 2008; Cortes et al., 2011). However, Mabs-S-1lsr2 and Mabs-R-1lsr2 colonies were heterogenous in size with a tendency toward small colonies when compared to their wild-type progenitors (**Supplementary Figure S3A**). All strains exhibited a similar growth rate, suggesting that the reduced colony size did not affect in vitro growth in this liquid medium (**Supplementary Figures S3B,C**). Lsr2 has been reported to be a negative regulator of GPL expression

FIGURE 2 | Generation of lsr2 null mutants. (A) Schematic representation of the genomic regions surrounding the lsr2 gene in the WT and lsr2 deletion mutant. (B) Southern blot of Mabs-S (S), Mabs-R (R), Mabs-S-1lsr2 (S 1lsr2) and Mabs-R-1lsr2 (R 1lsr2) using Probe 1 and Probe 2 depicted in panel (A). Genomic DNA was digested with the restriction enzyme PvuI, migrated on an agarose gel, transferred to a nitrocellulose membrane and hybridized with the two probes. Bands of the corresponding DNA fragments observed are of the expected size.

through binding to an AT-rich element upstream of the coding region of mbtH and mps1, both known to be involved in GPL production in M. smegmatis ATCC607 (Kocíncová et al., 2008). However, two other independent studies reported an unaltered GPL profile in lsr2 mutants generated from the M. smegmatis mc<sup>2</sup> 155 strain (Chen et al., 2006; Arora et al., 2008), which is the rough counterpart of the ATCC607 strain. We therefore addressed whether the inactivation of lsr2 may affect the GPL profile in both S and R M. abscessus variants. First, chromatin immunoprecipitation was carried out on lysates of Mabs-S-1lsr2 in which a FLAG-tagged Lsr2 was introduced and Lsr2 enrichment in the AT-rich element within the promoter region of mbtH/mps1 was then measured (**Figure 3A**). Binding of Lsr2 was found to be significantly enriched in this region (**Figure 3B**). However, GPL analysis by thin layer chromatography failed to show differences in the GPL profiles in Mabs-S-1lsr2 and in its parental S strain (**Figure 3C**), suggesting that Lsr2 inactivation alone is not sufficient to alter the morphotype of the mutant. As expected, Mabs-R-1lsr2 also failed to produce GPL, which can be explained by the presence of severe genetic lesions that impair the transcription of genes such as mps1 and mmpL4b (Pawlik et al., 2013).

Collectively, these data showed Lsr2 ability to bind the promoter region upstream of mbtH/msp1. However, loss of Lsr2 expression has no effect on production or composition of GPL in the S M. abscessus variant.

### Lsr2 Promotes Resistance to Oxidative Species in the R Variant

An important function of Lsr2 resides in protecting the integrity of genomic DNA from the deleterious action of reactive oxygen species (ROS) during macrophage infection (Colangeli et al., 2009). We aimed at addressing the potential protective effect of the higher expression of lsr2 in M. abscessus R cultures under H2O<sup>2</sup> exposure. Exponential phase cultures, diluted 10 times, were exposed to 20 mM H2O<sup>2</sup> and CFU were counted at various time points during treatment. When comparing WT M. abscessus S and R variants, a high proportion of the R population survived in the presence of H2O<sup>2</sup> (p < 0.001) (**Figure 4A**). Whether

this advantage is conferred by a higher lsr2 expression was next investigated. The growth defects of Mabs-S and Mabs-S-1lsr2 in the presence of H2O<sup>2</sup> were similar (**Supplementary Figure S4**). In contrast, Mabs-R-1lsr2 was more sensitive than its WT R progenitor, while genetic complementation in the R lsr2 mutant restored the WT R H2O<sup>2</sup> resistance phenotype (**Figure 4B**). These results show a difference in the response of the M. abscessus S/R variants to H2O2, the R variant being less sensitive to oxidative stress than the S variant. However, the intrinsic resistance of the R variant was partially lost upon deletion of the lsr2 gene.

#### The Intracellular Survival of lsr2 Mutants Is Reduced in Amoebae and Macrophages

The role of Lsr2 in the resistance of M. abscessus to oxidative species suggests that it may also protect the bacilli from the microbicidal activity of macrophages (Roux et al., 2016) and/or amoebae (Bakala N'Goma et al., 2015). This prompted us to investigate and compare the impact of lsr2 deletion on intracellular growth of the S and R mutants with their corresponding complemented strains (Mabs-S-1lsr2-Clsr2 and Mabs-R-1lsr2-Clsr2) and wild-type progenitors. Mabs-S-1lsr2 (**Figure 5A**) and Mabs-R-1lsr2 (**Figure 5B**) exhibited a pronounced defect inside murine M8 as compared to the WT and complemented strains. This survival defect was already significant at 1 dpi for the S strain (p < 0.0001) and at 5 dpi for the R strain (p < 0.0001). Inside A. castellani, a similar decrease in bacterial viability was observed for both Mabs-S-1lsr2 and Mabs-R-1lsr2 mutants (**Figures 5C,D**). At 3 dpi, both mutants showed a strong difference (p < 0.0001) in intra-amoebal multiplication compared to their respective WT and complemented strains.

Together, these results indicate that Lsr2 is required for intracellular survival of M. abscessus S/R forms in both macrophages and amoebae.

#### Lsr2 Plays a Key Role in M. abscessus Virulence in vivo

The impact of Lsr2 on mycobacterial virulence was originally suggested by demonstrating the preferential binding of Lsr2 to

AT-rich regions in the M. tuberculosis genome, including several genomic islands acquired by horizontal gene transfer, encoding major virulence factors (Gordon et al., 2010). The behavior of the lsr2 mutants was first addressed by infecting zebrafish larvae (Bernut et al., 2014a) with the wild-type R (Mabs-R) and Mabs-R-1lsr2 strains. A significant reduction in the virulence of Mabs-R-1lsr2 was observed when compared with Mabs-R (**Figure 6A**). Whereas only 10% of zebrafish were killed when infected with Mabs-R-1lsr2 at 12 dpi, around 40% of the larvae died at 6 dpi with Mabs-R while nearly 70% died at 10 dpi (**Figure 6A**). Complementing with lsr2 partially reversed the attenuated phenotype of Mabs-R-1lsr2.

To further evaluate the contribution of Lsr2 in pathogenesis of M. abscessus in a more complex animal host, BALB/c mice were intravenously infected with the wild-type S/R variants or Mabs-S/R-1lsr2 strains (Catherinot et al., 2007; Rottman et al., 2007). Bacterial loads were monitored by measuring the CFU counts in the lung, spleen and liver homogenates from mice sacrificed at 1, 15, and 45 dpi. A significant reduction in the number of Mabs-S-1lsr2 bacilli was observed in the lungs at 15 dpi when compared to the wild-type S progenitor (p < 0.01). This difference was further increased at 45 dpi, along with a complete clearance of Mabs-S-1lsr2 (p < 0.05) (**Figure 6B**, upper panels). However, no significant differences in CFU counts in the spleen and the liver were observed between Mabs-S and Mabs-S-1lsr2. Monitoring the bacterial burden of Mabs-R-1lsr2 revealed a significant reduction in all three organs at 45 dpi as compared to the WT parental strain (p < 0.05 in lungs and spleen; p < 0.01 in liver) (**Figure 6C**, lower panels).

Overall, these results indicate that Lsr2 plays a critical role in M. abscessus virulence in different animal hosts and highlight the requirement of Lsr2 for persistence in mice, particularly in the lungs, which represents the major targeted organ during infection in patients, especially in those with already underlying pulmonary diseases.

#### DISCUSSION

Lsr2, a unique nucleoid associated protein in the Actinomyces and Mycobacterium genera, has been extensively studied in M. tuberculosis (Colangeli et al., 2009; Gordon et al., 2010;

Bartek et al., 2014). Like its ortholog H-NS, homodimers of Lsr2 can bind DNA cooperatively to form long oligomers on AT-rich sequences that can further interact to bridge distant DNA regions and contribute to loop formation (Chen et al., 2008; Liu and Gordon, 2012). At the genomic level, Lsr2 filaments are found at the promoters of 401 and 272 genes in M. tuberculosis and M. smegmatis, respectively, often extending into their coding regions (Gordon et al., 2010). This observation, as well as its known role as a negative regulator in several cellular functions, strongly suggest that like H-NS, Lsr2 represses transcription through promoter occlusion of RNA polymerase targets or by interfering with transcription elongation (Chen et al., 2008; Liu and Gordon, 2012; Landick et al., 2015). Lsr2 has been shown to be involved in M. tuberculosis virulence by binding to important genes, such as genes involved in the ESX secretion systems, the biosynthesis of the PDIM and PGL cell wall lipids or encoding antigenic proteins of the PE/PPE family (Gordon et al., 2010). Previous studies also demonstrated the implication of Lsr2 in multi-drug tolerance of M. tuberculosis (Colangeli et al., 2007) and in protection against reactive oxygen intermediates (Colangeli et al., 2009). All these phenotypes were investigated and confirmed in M. abscessus, except for the increased resistance to antibiotics (not shown), thus involving Lsr2 in increased resistance to oxidative species as well as in intracellular multiplication and survival of M. abscessus in zebrafish and mice.

The S-to-R transition of M. abscessus has only been found to occur during infection either in mice (Rottman et al., 2007) or in humans (Jönsson et al., 2007; Catherinot et al., 2009; Roux et al., 2009). In addition, the underlying genetic causes for this transition pointed to irreversible mechanisms, due to selective mutations such as indels that prevent the R form to reverse its morphotype to a S form (Pawlik et al., 2013). Thus, the S-to-R transition represents a clear advantage for M. abscessus during the course of the infection. When comparing pairs of isogenic S/R variants, it was found that lsr2 was expressed at a higher level in the R morphotype (Pawlik et al., 2013). This was confirmed in the present study using qRT-PCR analysis. By inducing a higher expression of lsr2, the R variant becomes more resistant to reactive oxygen species. Interestingly, these conditions look similar to those prevailing in the CF airways, characterized by polymicrobial and chronic infections, which maintain a very harsh and oxidative environment in the lungs. Similarly, CF patients receive frequent cures involving large-spectrum antibiotics, such as inhaled tobramycin therapy. Herein, we show that the R form is more resistant to H20<sup>2</sup> than the S form. But we were unable to confirm an increased resistance to antibiotics (not shown). This resistant phenotype was lost in the lsr2 mutant but restored to wild-type levels following complementation with a functional copy of lsr2.

A major outcome of this work is to point out the role of Lsr2 in the intracellular behavior of M. abscessus and on its requirement for persistence in infected mice and zebrafish. It was previously shown that a deletion of lsr2 in M. smegmatis has an impact on its survival in murine macrophages (Colangeli et al., 2009). In vivo studies of the role of lsr2 in virulence in mice were also performed on M. tuberculosis, using an attenuated strain that was complemented by a replicative plasmid expressing Lsr2 under the control of an inducible promoter (Ehrt et al., 2005) to circumvent the absence of viability of the lsr2 M. tuberculosis deletion mutant (Colangeli et al., 2007). Using this strategy, it was shown that lsr2 is important to protect M. tuberculosis against ROI during macrophage infection. The present study shows that despite being dispensable in M. abscessus, lsr2 remains essential for optimal intracellular growth and virulence in different vertebrates.

Previous work in M. smegmatis emphasized the contribution of Lsr2 in defining the lipid profile, colony morphology, bacterial motility and biofilm formation by specifically repressing transcription at the GPL locus in the rough morphotypes (Chen et al., 2006; Arora et al., 2008; Kocíncová et al., 2008; Yang et al., 2017). However, this study clearly shows that lsr2 mRNAs are detected in both M. abscessus S and R strains. Thanks to ChIP-qPCR, we also observed that Lsr2 is able to bind the promoter region of mbtH in M. abscessus S, despite the equivalent GPL profiles in the WT and the mutant strain. This suggests that binding of Lsr2 to this particular genomic region is not sufficient to regulate the biosynthetic GPL locus and that this process may involve other partners that remain to be identified. An alternative explanation is that the higher expression level of Lsr2 in M. abscessus R is needed to successfully reduce GPL production, eventually by increasing the formation of a DNA loop that can occlude RNA polymerase binding or inhibit transcription elongation, as proposed for H-NS in Gram-negative bacteria (Landick et al., 2015). One would have anticipated that increasing expression of Lsr2 in M. abscessus R would enhance the R phenotype by further repressing the genes of the GPL locus. However, deletion of lsr2 in M. abscessus R did not restore GPL production as it was previously observed in M. smegmatis (Kocíncová et al., 2008). As mentioned above, the GPL locus in the M. abscessus R variant suffers from irreversible genetic lesions, adding a layer of complexity to the molecular mechanisms involved in the regulation of this locus. The pleiotropic nature of Lsr2 implies that its deletion has a broader impact on the fitness of M. abscessus as well as on the general compaction of its nucleoid. Thus, additional analysis using genome-wide methods will help to unravel the details of the molecular role of Lsr2 during S-to-R transition and its impact on pathogenicity.

Taken together, this work showed the impact of Lsr2 as a key virulence factor on the intracellular and in vivo survival of M. abscessus. These results also strongly suggest that Lsr2, like other NAPs, might represent a target of choice for the development of new antimicrobials. Indeed, compounds that would either silence expression of lsr2 or inhibit its function could therefore be considered as a new anti-virulence approach against M. abscessus (Liu and Gordon, 2012).

#### ETHICS STATEMENT

All procedures involving mice were performed according to the institutional and national ethical guidelines and approved by the comité d'éthique en experimentation animale N◦ 047

with agreement A783223 under the reference APAFIS#11465. All zebrafish experiments were done according to European Union guidelines for handling of laboratory animals (http://ec. europa.eu/environment/chemicals/lab\_animals/home\_en.htm) and approved by the Comité d'Ethique pour l'Expérimentation Animale de la région Languedoc Roussillon under the reference CEEALR36-1145.

### AUTHOR CONTRIBUTIONS

J-LH, LK, and FC designed the project and experiments. VLM, AB, MC, AV, CD, and AP performed the experiments. J-LG, FM, FC, LK, and J-LH wrote and corrected the manuscript.

#### REFERENCES


#### FUNDING

This work was supported by the Vaincre la Mucoviscidose association (Post-doctoral fellowship to VLM #RF20110600446/1/ 3/130) and the ANR DIMYVIR Grant (ANR-13-BSV3- 0007-01 to J-LH and LK) from the French National Research Agency Program.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2019.00905/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 © 2019 Le Moigne, Bernut, Cortès, Viljoen, Dupont, Pawlik, Gaillard, Misguich, Crémazy, Kremer and Herrmann. 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.

# Glycopeptidolipids, a Double-Edged Sword of the Mycobacterium abscessus Complex

Ana Victoria Gutiérrez1,2, Albertus Viljoen<sup>1</sup> , Eric Ghigo<sup>3</sup> , Jean-Louis Herrmann<sup>4</sup> and Laurent Kremer1,5 \*

<sup>1</sup> Centre National de la Recherche Scientifique, Institut de Recherche en Infectiologie de Montpellier, UMR 9004, Université de Montpellier, Montpellier, France, <sup>2</sup> CNRS, IRD 198, INSERM U1095, APHM, Institut Hospitalo-Universitaire Méditerranée Infection, UMR 7278, Aix-Marseille Université, Marseille, France, <sup>3</sup> CNRS, Campus Joseph Aiguier, Marseille, France, <sup>4</sup> 2I, UVSQ, INSERM UMR 1173, Université Paris-Saclay, Versailles, France, <sup>5</sup> INSERM, IRIM, Montpellier, France

Mycobacterium abscessus is a rapidly-growing species causing a diverse panel of clinical manifestations, ranging from cutaneous infections to severe respiratory disease. Its unique cell wall, contributing largely to drug resistance and to pathogenicity, comprises a vast panoply of complex lipids, among which the glycopeptidolipids (GPLs) have been the focus of intense research. These lipids fulfill various important functions, from sliding motility or biofilm formation to interaction with host cells and intramacrophage trafficking. Being highly immunogenic, the induction of a strong humoral response is likely to select for rough low-GPL producers. These, in contrast to the smooth high-GPL producers, display aggregative properties, which strongly impacts upon intracellular survival. A propensity to grow as extracellular cords allows these low-GPL producing bacilli to escape the innate immune defenses. Transitioning from high-GPL to low-GPL producers implicates mutations within genes involved in biosynthesis or transport of GPL. This leads to induction of an intense pro-inflammatory response and robust and lethal infections in animal models, explaining the presence of rough isolates in patients with decreased pulmonary functions. Herein, we will discuss how, thanks to the generation of defined GPL mutants and the development of appropriate cellular and animal models to study pathogenesis, GPL contribute to M. abscessus biology and physiopathology.

#### Edited by:

Thomas Dick, Rutgers, The State University of New Jersey, Newark, United States

#### Reviewed by:

Anil Ojha, Wadsworth Center, United States Olivier Neyrolles, Centre National de la Recherche Scientifique (CNRS), France

> \*Correspondence: Laurent Kremer laurent.kremer@irim.cnrs.fr

#### Specialty section:

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

Received: 21 March 2018 Accepted: 14 May 2018 Published: 05 June 2018

#### Citation:

Gutiérrez AV, Viljoen A, Ghigo E, Herrmann J-L and Kremer L (2018) Glycopeptidolipids, a Double-Edged Sword of the Mycobacterium abscessus Complex. Front. Microbiol. 9:1145. doi: 10.3389/fmicb.2018.01145 Keywords: Mycobacterium abscessus, glycopeptidolipid, cell wall, pathogenesis, host/pathogen interactions

#### INTRODUCTION

Mycobacterium abscessus is a fast-growing non-tuberculous mycobacterium (NTM) and an emerging human pathogen that causes nosocomial skin and soft tissue infections (Brown-Elliott et al., 2012) but also pulmonary infections, especially in patients with cystic fibrosis (CF) and other lung disorders (Sermet-Gaudelus et al., 2003; Esther et al., 2010). Recent investigations reported mechanisms of virulence and physiopathological processes characterizing M. abscessus infection thanks to (i) genetic tools that allowed generation of defined mutants and transposon libraries, particularly useful to seek out genetic determinants of intracellular survival (Medjahed and Reyrat, 2009; Cortes et al., 2011; Gregoire et al., 2017; Laencina et al., 2018) and (ii) the development of various complementary cellular and animal models, which have allowed delineation of the early

stages of the infection and the role of important cell types participating in controlling the infection and/or in the formation of granulomas (Ordway et al., 2008; Bernut et al., 2014a, 2017; Laencina et al., 2018). Evidence exists that granulomas harbor persistent M. abscessus for extended periods of time (Tomashefski et al., 1996; Medjahed et al., 2010). Additionally, these models have been used successfully to test the in vivo therapeutic efficacy of compounds against M. abscessus, considered as one of the most drug-resistant mycobacterial species (Bernut et al., 2014b; Dubée et al., 2015; Obregón-Henao et al., 2015; Dupont et al., 2016).

Like other NTMs, M. abscessus displays smooth (S) or rough (R) colony morphotypes, associated with distinct in vitro and in vivo phenotypes. This colony-based distinction is dependent on the presence (in S) or absence (in R) of surface-associated glycopeptidolipids (GPLs) (Howard et al., 2006; Medjahed et al., 2010). The presence or lack of GPL considerably influences important physiological and physiopathological aspects, including sliding motility or biofilm formation, interaction with host cells, intracellular trafficking in macrophages and virulence, ultimately conditioning the clinical outcome of the infection. This review gathers some of the most recent findings related to biosynthesis and transport of GPL in M. abscessus, the mechanisms driving the S-to-R switch and how this transition influences the surface properties of the bacilli, interaction with host cells, virulence and potentially the mode of transmission of M. abscessus.

### GENOMICS AND STRUCTURAL ASPECTS OF GPL IN M. abscessus

The mycobacterial envelope comprises three layers: a typical plasma membrane, a complex cell wall partly resembling a Gram-positive wall and an outer layer (Daffé and Draper, 1998). Particularly unusual, the cell wall consists of a thick peptidoglycan layer covalently-linked to arabinogalactan, itself esterified by mycolic acids, forming the inner leaflet of the mycomembrane. In addition, a large variety of extractible lipids form the outer leaflet of the mycomembrane. Among these are the GPL, found in many NTM (**Figure 1A**). GPL are subdivided into alkali-stable C-type GPL and alkalilabile serine-containing GPL. The C-type GPL are found in saprophytic mycobacteria such as Mycobacterium smegmatis or in opportunistic pathogens like Mycobacterium avium, Mycobacterium chelonae, or Mycobacterium abscessus (Schorey and Sweet, 2008), whereas the alkali-labile serine-containing GPL were found in Mycobacterium xenopi (Besra et al., 1993). C-type GPL share a common lipopeptidyl core consisting of a mixture of 3-hydroxy and 3-methoxy C28-30 fatty acids amidated by a tripeptide-amino-alcohol core of D-Phe-D-allo-Thr-D-Ala-L-alaninol. This lipopeptide core is glycosylated with the allo-Thr linked to a 6-deoxy-α-L-talose and the alaninol linked to an α-L-rhamnose. These di-glycosylated GPL make up the less polar species (**Figure 1A**). In the case of M. avium, the 6-deoxytalose is non-methylated or 3-O-methylated, and the rhamnose is either 3-O-methylated or 3,4-di-O-methylated. M. avium GPL can also be O-acetylated at various locations, depending on the strain. In contrast, M. smegmatis, M. chelonae, and M. abscessus produce di-glycosylated GPL that contain a 3,4-di-O-acetylated 6-deoxytalose and a 3,4-di-O-methylated or 2,3,4-tri-O-methylated rhamnose (Villeneuve et al., 2003; Ripoll et al., 2007; Whang et al., 2017). These species also produce more polar GPL by the addition of a 2,3,4-tri-hydroxylated rhamnose to the alaninol-linked 3,4-di-O-methyl rhamnose. Although being structurally identical, triglycosylated GPL are more abundant in M. abscessus than in M. smegmatis (Ripoll et al., 2007). GPL are heterogenous in structure and vary according to the fatty acyl chain length and the degree of hydroxylation or O-methylation of the glycosidic moieties (**Figure 1B**).

The gpl locus is highly conserved in M. smegmatis, M. abscessus, and M. avium (**Figure 1C**) but differences exist, like the presence of an IS1096 in M. smegmatis. The tripeptideaminoalcohol moiety of GPL is assembled by the products of mps1 and mps2 (Billman-Jacobe et al., 1999). The genes gtf1 and gtf2 catalyze glycosylation of the lipopeptide core whereas gtf3 adds the extra rhamnose defining triglycosylated GPL. The genes rmt2, rmt3, and rmt4 participate in O-methylation of the rhamnose and fmt, absent in M. avium, in O-methylation of the lipid moiety. In contrast to M. smegmatis which possesses a single atf gene involved in acetylation of the two positions of the deoxytalose, two genes, atf1 and atf2, transfer the acetyl residues in a sequential manner in M. abscessus (Ripoll et al., 2007). Separated by gtf3, rmlA and rmlB are responsible for monosaccharide activation and epimerization. On the proximal end of the gpl locus is found mmpS4, mmpL4a and mmpL4b in an operon and encoding membrane proteins required for the transport of GPL across the plasma membrane (Medjahed and Reyrat, 2009; Deshayes et al., 2010; Bernut et al., 2016b). MmpS4 has been proposed to mediate formation of the GPL biosynthesis/transport machinery megacomplex located at the bacterial pole (Deshayes et al., 2010). GPL transport requires also the integral membrane protein Gap in M. smegmatis (Sondén et al., 2005) (**Figure 1A**). How GPL are translocated from the periplasmic space to the outer membrane, however, remains unknown. Additionally, a block of eight genes [MSMEG\_0406 (fadE5) to MSMEG\_0413 (gap-like)] in M. smegmatis, originally proposed to catalyze the lipid synthesis and attachment to the tripeptide-aminoalcohol moiety of GPL (Ripoll et al., 2007) was recently reattributed to the synthesis of trehalose polyphleates (TPP) (Burbaud et al., 2016). In M. abscessus, this cluster is far away from the main gpl locus with Rv0926 and fadE5 being scattered in the M. abscessus chromosome (**Figure 1C**).

### MOLECULAR MECHANISMS OF THE SMOOTH-TO-ROUGH TRANSITION AND ASSOCIATED PHENOTYPES

Comparative genomics to understand the molecular basis of the S and R phenotypes using isogenic S and R pairs revealed multiple indels or single nucleotide polymorphisms within the gpl locus (Pawlik et al., 2013). A single nucleotide deletion in mmpL4b and nucleotide insertions in mps1 were identified in the R variants

orientation of the different genes are drawn to scale. A color code has been used to specify the participation of these genes in synthesis of the lipopeptide core, glycosylation, methylation, acetylation, transport and other biological functions, as indicated. Genes in the tpp locus are displayed in pink.

when compared to the S variants from the three different isogenic S/R couples. Moreover, RNA sequencing demonstrated that S and R isogenic strains differed considerably at the transcriptomic level, with the transcriptional extinction of mps1, mps2, and gap in the R strain caused by an insertion in the 5<sup>0</sup> -end of mps1 (Pawlik et al., 2013). Additional mutations in mps2, mmpL4a, and mmpS4 were subsequently identified in R strains isolated from later disease stages (Park et al., 2015). Disruption of mmpL4b in

M. abscessus S was initially reported to abrogate GPL production, leading to a rough colonial morphotype (Medjahed and Reyrat, 2009; Nessar et al., 2011) (**Figure 2A**). Point mutations in MmpL4a at Tyr842 or MmpL4b at Tyr854, corresponding to two critical residues presumably involved in the proton-motive force of the MmpL proteins, were also associated with loss of GPL production (Bernut et al., 2016b) (**Figure 2A**), suggesting that no functional redundancy exists between MmpL4a and MmpL4b.

Hydrophilic and hydrophobic properties of bacteria can influence surface adhesion and biofilm formation (Krasowska and Sigler, 2014). As shown in M. smegmatis (Recht and Kolter, 2001), M. abscessus (Howard et al., 2006) and M. bolletii (Bernut et al., 2016b) the presence of GPL in the S variants facilitates sliding across the surface of motility agar and biofilm formation on the liquid medium/air interface whereas lack of GPL promotes bacterial aggregation (Brambilla et al., 2016) and cording (Howard et al., 2006; Bernut et al., 2014a, 2016b). However, whether M. abscessus R fails at producing biofilms was recently readdressed and proposed that it can grow in biofilm-like structures, which, like S biofilms, are significantly more tolerant than planktonic cultures to acidic pH, hydrogen peroxide, and drugs (Clary et al., 2018). In M. smegmatis, the nucleoid-associated protein Lsr2 negatively regulates GPL production (Kocíncová et al., 2008) and plays a role during the initial stages of biofilm development (Yang et al., 2017). Despite the presence of an lsr2 gene in M. abscessus, which is up-regulated in the R variant (Pawlik et al., 2013), the contribution of Lsr2 in regulating GPL expression, sliding motility and biofilm formation remains to be established.

External factors, such as sub-inhibitory antibiotic concentrations, can promote a transient S-to-R change with more aggregated cultures and a higher resistance to phagocytosis (Tsai et al., 2015). Strangely, these phenotypes were neither linked to a loss of GPL production nor to the differential expression of genes within the gpl cluster, but were rather mediated by MAB\_3508c, homologous to whiB7 and conferring extreme resistance to antibiotics in M. abscessus (Hurst-Hess et al., 2017). In contrast, another study reported that sub-inhibitory amikacin treatment, also leading to a S-to-R transition, was associated with decreased GPL, resulting from down-regulation of several gpl biosynthetic genes (Lee et al., 2017). Overall, these results suggest that exposure to sub-inhibitory amikacin doses may induce alterations in GPL content, increase virulence and influence the outcome of the infection as well as the therapeutic efficacy of drugs.

#### PRESENCE OR LOSS OF GPL CONDITIONS BACTERIAL SURFACE PROPERTIES AND INTERACTIONS WITH HOST CELLS

The S and R variants interact differently with host cells and exhibit different intracellular behaviors, as first reported in human monocytes with the R variant persisting longer in these cells (Byrd and Lyons, 1999) and then confirmed in other phagocytes (Howard et al., 2006; Roux et al., 2016). Macrophages encountering the aggregative R strain, are incapable of engulfing clumps of R bacteria, which remain embedded in phagocytic cups on the exterior of the cells (Roux et al., 2016). However, smaller clumps are phagocytosed, resulting in social phagosomes (containing numerous bacilli) that rapidly fuse with lysosomes. In addition, R variant-containing THP1 cells were more acidified, more autophagic and more apoptotic than those infected with the S variant. In contrast, in the majority of S variant-containing phagosomes, a continuous tight apposition is maintained between the phagosome membrane and the mycobacterial cell envelope, leading to phagosome maturation blockage and the absence of acidification in S-infected macrophages. Another conspicuous trait of the S-containing phagosomes is the occurrence of a large electron translucent zone (ETZ) enclosing the bacilli (**Figure 2B**). This ETZ is barely detected in R-containing phagosomes or in phagosomes containing S strains mutated in either mmpL4a or mmpL4b, indicating that the ETZ relies on the presence of cell surface GPL (Bernut et al., 2016b; Roux et al., 2016). Interestingly, infection with M. abscessus S, but not R, leads to phagosome membrane lesions, suggesting that, similarly to pathogenic slow-growing mycobacteria, this variant has the capacity to induce phagosome–cytosol communications (Simeone et al., 2012), through a mechanism that likely involves the type VII secretion system ESX-4 (Laencina et al., 2018).

The distinct mechanisms responsible for the S- and R-induced responses in macrophages are being delineated, highlighting a model whereby the loss of GPL at the surface unmasks underlying phosphatidyl-myo-inositol dimannoside (PIM2) (Rhoades et al., 2009) and lipoproteins (Roux et al., 2011). These TLR-2 agonists stimulate the expression of TNF and intense inflammation. Exacerbation of this response can also lead to tissue lesions associated with R strains. In S strains, by covering underlying immunostimulatory cell wall components, GPL may delay the activation of the immune response during early infection stages and facilitate colonization by preventing TLR-2 signaling in the respiratory epithelial cells (Davidson et al., 2011).

Apoptosis represents an innate response of cells to restrict multiplication of intracellular pathogens (Lamkanfi and Dixit, 2010). M. abscessus R was found to be more apoptotic than M. abscessus S in different types of macrophages (Roux et al., 2016; Whang et al., 2017). Supporting these findings, purified GPL from M. abscessus S inhibits macrophage apoptosis, presumably by suppressing the production of radical oxygen species (ROS), the release of cytochrome c and by preserving the mitochondrial transmembrane potential (Whang et al., 2017) (**Figure 2B**). This mechanism appears to be mediated by targeting of acetylated GPL to the mitochondria where they interact with cyclophilin D, a component of the mitochondrial permeability transition pore (MPTP), inhibiting the MPTP that results in a block on cell death in similar fashion to cyclosporin A. That exogenous GPL-dependent apoptosis inhibition restricts intracellular growth and spreading of M. abscessus R, suggests also that GPL may limit M. abscesssus virulence (Whang et al., 2017).

Collectively, these observations emphasize the diversity of the infection programs orchestrated by S and R variants. While the S variant is promptly phagocytosed by macrophages without immediately affecting cell survival, phagocytosis of the R variant is more harmful to macrophage viability and, following apoptosis, the released bacteria replicate extracellularly in the form of serpentine cords (**Figure 2A**).

### IMPACT OF THE GPL CONTENT ON VIRULENCE

Epidemiological surveys document the prominence of the R strain in patients with severe pulmonary infections (Catherinot et al., 2009) and with chronic colonization of the airways in CF patients (Jönsson et al., 2007), but the exact proportions of S and R forms in these populations remain largely unknown. This distinction is, however, of crucial importance considering that the cord-forming R variant causes much more aggressive and invasive pulmonary disease that ends in severe respiratory failure. Further alarm is raised by several studies using various cellular and animal models, confirming the increased virulence of the R over the S form (Bernut et al., 2017). Among these models, the zebrafish (Danio rerio) has been proposed as a relevant and genetically tractable host–pathogen conjugate for dissecting M. abscessus interactions with host cells (Bernut et al., 2015). This led to important breakthroughs regarding mechanisms of M. abscessus pathogenesis, involving cording and granuloma formation (Bernut et al., 2014a) or the importance of the TNF response in controlling the infection and establishment of protective granulomas (Bernut et al., 2016a). S-to-R transitioning is associated with exacerbation of the bacterial burden, the formation of massive serpentine cords, abscess formation, notably in the central nervous system, and increased larval killing (Bernut et al., 2014a). Deletion of MAB\_4780, encoding a dehydratase required for cording, resulted in extreme attenuation in wild-type and immunocompromized larvae (Halloum et al., 2016), further incriminating cording as a major virulence determinant in strains lacking GPL. Replacing the endogenous mmpS4-mmpL4a-mmpL4b promoter with the leaky acetamidase promoter from M. smegmatis in M. abscessus S resulted in a strain with low-GPL levels, but still aggregating in culture with a rough appearance, similar to R strains. In zebrafish, this mutant exhibited an intermediate virulence phenotype with a delay in killing compared to the R strain. Moreover, the number and size of the abscesses in larvae infected with this low-GPL producing strain were significantly reduced compared to the R strain (Viljoen et al., 2018). This indicates that low-GPL levels impede the induction of the physiopathological signs and virulence of M. abscessus R, confirming the opposite relationship between the amount of GPL and virulence. In addition, the S variant is more hydrophilic than the R variant or the rough low-GPL producing strain. This suggests that lack of the hydrophilic GPL components is responsible for their increased hydrophobicity over the S strain and confirm a positive correlation between GPL production and hydrophilicity.

Supporting the theory of transmission via aerosols from the environment, M. abscessus, along with other NTM, was isolated from household water and shower aerosols in the homes of patients with pulmonary disease (Thomson et al., 2013). Importantly, a direct person-to-person transmission of M. abscessus by aerosol inhalation has been asserted in recent world-wide surveys, although the link between morphology/GPL profile and mode of transmission remains to be investigated (Bryant et al., 2013, 2016).

### CONCLUSION AND PERSPECTIVES

Fundamental aspects of the M. abscessus lifecycle rely on the beneficial effects of GPL in promoting and facilitating the early stages of colonization of the S variant, presumably the major form existing in the environment. By covering the bacilli, the highly immunogenic GPL induce a strong humoral response in infected individuals and it is possible that this strong immune pressure leads to selection of GPL-deficient strains, allowing M. abscessus to escape the anti-GPL response and the emergence of R bacilli. The lack of GPL, in turn, leads to increased apoptosis, promoting extracellular replication and cording, acute infection and the most severe forms of the disease. This has also detrimental consequences for the host since, by unmasking other pro-inflammatory cell-surface components, the loss of GPL translates to severe inflammation and lung damage. Therefore, the double edged sword effect of GPL allows M. abscessus to efficiently transition between a colonizing environmental micro-organism to an invasive human pathogen. Given the importance of the GPL content in driving the interaction with host cells and in conditioning the issue of the infection, it appears important to pay more attention to the variant (S or R) selected for experimental infections and to systematically report the morphotype of the strains isolated in clinical studies.

Important unsolved questions remain on GPL in M. abscessus. Future investigations should describe the complete GPL export machinery since, while transfer of GPL across the plasma membrane has been addressed to some extent, additional unidentified outer-membrane proteins are likely to participate in this important physiological process. So far, the literature only reports the effects of near total loss of GPL in M. abscessus, portraying an incomplete picture of the functions of these lipids in the physiology of this pathogen. Polar GPL species in M. smegmatis are only produced under carbon starvation and induce smooth-colony formation (Ojha et al., 2002) opening up the possibility that in M. abscessus GPL composition is modulated in response to changing environments. Therefore, studies are required to address whether GPL composition affects M. abscessus persistence and/or host inflammation, as well as how the dynamics of GPL production, potentially mediated by yet unidentified factors, influences adaptation of M. abscessus to its host.

### AUTHOR CONTRIBUTIONS

AG designed the figures. All authors contributed to writing the manuscript.

## FUNDING

LK acknowledges the support by the Fondation pour la Recherche Médicale (FRM) (DEQ20150331719) and the Infectiopôle Sud Méditerranée for funding the Ph.D. Fellowship of AG.

#### REFERENCES

fmicb-09-01145 June 5, 2018 Time: 14:34 # 7


antimycobacterial drugs in preclinical models. Antimicrob. Agents Chemother. 59, 6904–6912. doi: 10.1128/AAC.00459-15


toxicity and host cell death. PLoS Pathog. 8:e1002507. doi: 10.1371/journal.ppat. 1002507


**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 Gutiérrez, Viljoen, Ghigo, Herrmann and Kremer. 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.

# Mycobacterium abscessus subsp. massiliense mycma\_0076 and mycma\_0077 Genes Code for Ferritins That Are Modulated by Iron Concentration

Fábio M. Oliveira<sup>1</sup> , Adeliane C. Da Costa<sup>1</sup> , Victor O. Procopio<sup>1</sup> , Wanius Garcia<sup>2</sup> , Juscemácia N. Araújo<sup>2</sup> , Roosevelt A. Da Silva<sup>3</sup> , Ana Paula Junqueira-Kipnis<sup>1</sup> and André Kipnis<sup>1</sup> \*

#### Edited by:

Thomas Dick, Rutgers, The State University of New Jersey, Newark, United States

#### Reviewed by:

Divakar Sharma, Aligarh Muslim University, India Zeeshan Fatima, Amity University, India

> \*Correspondence: André Kipnis andre.kipnis@gmail.com

#### Specialty section:

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

Received: 27 February 2018 Accepted: 04 May 2018 Published: 01 June 2018

#### Citation:

Oliveira FM, Da Costa AC, Procopio VO, Garcia W, Araújo JN, Da Silva RA, Junqueira-Kipnis AP and Kipnis A (2018) Mycobacterium abscessus subsp. massiliense mycma\_0076 and mycma\_0077 Genes Code for Ferritins That Are Modulated by Iron Concentration. Front. Microbiol. 9:1072. doi: 10.3389/fmicb.2018.01072 <sup>1</sup> Tropical Institute of Pathology and Public Health, Department of Microbiology, Immunology, Parasitology and Pathology, Federal University of Goiás, Goiânia, Brazil, <sup>2</sup> Centro de Ciências Naturais e Humanas, Federal University of ABC (UFABC), Santo André, Brazil, <sup>3</sup> Collaborative Center of Biosystems, Regional Jataí, Federal University of Goiás, Goiânia, Brazil

Mycobacterium abscessus complex has been characterized in the last decade as part of a cluster of mycobacteria that evolved from an opportunistic to true human pathogen; however, the factors responsible for pathogenicity are still undefined. It appears that the success of mycobacterial infection is intrinsically related with the capacity of the bacteria to regulate intracellular iron levels, mostly using iron storage proteins. This study evaluated two potential M. abscessus subsp. massiliense genes involved in iron storage. Unlike other opportunist or pathogenic mycobacteria studied, M. abscessus complex has two genes similar to ferritins from M. tuberculosis (Rv3841), and in M. abscessus subsp. massiliense, those genes are annotated as mycma\_0076 and mycma\_0077. Molecular dynamic analysis of the predicted expressed proteins showed that they have a ferroxidase center. The expressions of mycma\_0076 and mycma\_0077 genes were modulated by the iron levels in both in vitro cultures as well as infected macrophages. Structural studies using size-exclusion chromatography, circular dichroism spectroscopy and dynamic light scattering showed that r0076 protein has a structure similar to those observed in the ferritin family. The r0076 forms oligomers in solution most likely composed of 24 subunits. Functional studies with recombinant proteins, obtained from heterologous expression of mycma\_0076 and mycma\_0077 genes in Escherichia coli, showed that both proteins were capable of oxidizing Fe2<sup>+</sup> into Fe3+, demonstrating that these proteins have a functional ferroxidase center. In conclusion, two ferritins proteins were shown, for the first time, to be involved in iron storage in M. abscessus subsp. massiliense and their expressions were modulated by the iron levels.

Keywords: rapid growing mycobacteria, iron storage protein, pathogenic, ferritin, ferroxidase center, iron homeostasis

### INTRODUCTION

fmicb-09-01072 June 1, 2018 Time: 14:47 # 2

The Mycobacterium abscessus complex, composed of M. abscessus subsp. abscessus, M. abscessus subsp. massiliense, and M. abscessus subsp. bolletii has emerged as human pathogens in the last few years (Petrini, 2006; Medjahed et al., 2010; Lee et al., 2015; Tortoli et al., 2016). Mycobacteria belonging to this complex cause several diseases in humans. These include severe lung, skin, post-traumatic, and post-surgical infections, especially in patients that have cystic fibrosis as well as in immunocompetent individuals (Medjahed et al., 2010). Due to its capacity to adapt and persist in the environment as well as to resist disinfection procedures, the infections caused by the M. abscessus group are usually due to cross contamination, through surgical equipment or other contaminated procedures (Cardoso et al., 2008). The transmission of M. abscessus has been already documented among cystic fibrosis individuals. Therefore, infections in humans can occur by both direct and indirect transmission (Bryant et al., 2013; Lee et al., 2015; Bryant et al., 2016).

Its capacity to infect and multiply within phagocytic cells indicates that M. abscessus can evade the defense mechanisms imposed by the host, resulting in successful infection (Martins de Sousa et al., 2010; Shang et al., 2011; Bernut et al., 2014; Abdalla et al., 2015a; Bakala et al., 2015; Caverly et al., 2015). One mechanism used by this bacillus to multiply within macrophages is the induction of Heme-Oxygenase-1 (HO-1), which reduces the toxic oxidative stress effects produced by the cell (Abdalla et al., 2015a). Part of the HO-1 action is accomplished by the reduction of free ferrous ion Fe2<sup>+</sup> inside macrophages through the storage of this metal by ferritins, thus preventing the formation of free radicals by the Fenton reaction (Imlay et al., 1988; Vile and Tyrrell, 1993). Hence, mechanisms of extracellular iron level regulation are crucial for the bacilli to survive within macrophages. However, studies have shown that the intracellular levels of iron are also important for bacilli to establish infection, because both absence and excess of iron are deleterious (De Voss et al., 2000; Pandey and Rodriguez, 2012). Consequently, in order to survive within macrophages, the bacilli must be able to obtain, store, and regulate the iron levels during entire infection (Gold et al., 2001; Pandey and Rodriguez, 2014; Pandey et al., 2014).

The main protein family involved in regulating intracellular iron levels and reducing its toxic effects are the ferritins. The proteins within this family may be divided into three sub-classes: ferritin (non-heme binding), bacterioferritin (heme bound) and Dps (DNA-binding protein from starved cells) (Andrews et al., 2003). Bacterial ferritins and bacterioferritins have similar structures, and they are composed of 24 identical subunits arranged in an octahedral form, with a ferroxidase catalytic site at its center. At this catalytic site, Fe2<sup>+</sup> is oxidized to Fe3<sup>+</sup> and stored within its hollow interior, where it can store up to 4,500 atoms of this metal ion (Andrews, 1998; Bou-Abdallah, 2010). Consequently, iron is stored in its non-reactive form (Fe3+), avoiding its toxic effects on the cell, and can be released when needed (Andrews, 1998).

Despite similar structure between baterial ferritins and bacterioferritins, their amino acid sequences present low identity and bacterioferritins have a heme group suggesting different origins for these proteins (Andrews, 1998; Carrondo, 2003). M. tuberculosis (Mtb) has two types of ferritin-like molecules, bacterioferritin (BfrA) and ferritin (BfrB), coded by the genes Rv1876 and Rv3841, respectively (Cole et al., 1998). Crystallographic studies showed that these proteins are organized similar to the ferritin superfamily, which is an oligomer in the form of a shell with a catalytic center of ferroxidase (Gupta et al., 2009; Khare et al., 2011). Studies using Mtb mutants, which had their bfrA and bfrB genes deleted solely or together, demonstrated the importance of both in iron homeostasis, as well as in the virulence and pathogenicity of this bacillus in different infection models (Pandey and Rodriguez, 2012, 2014; Reddy et al., 2012; Khare et al., 2017).

In addition, ferritins appear to be involved in drug resistance of Mtb, because it was shown that the lack of ferritin in this bacillus increased the susceptibility to antimicrobials used to treat tuberculosis (Pandey and Rodriguez, 2012). Proteomic analysis indicated that both BfrA and BfrB were overexpressed in aminoglycosides resistant as compared to sensitive clinical isolates of Mtb (Kumar et al., 2013; Sharma et al., 2015). Additionally, overexpression of Rv3841 (bfrB) by recombinant Escherichia coli resulted in increased kanamycin and amikacin resistance (Sharma et al., 2016). Taken together, BfrA and BfrB proteins, could be promising drug targets against mycobacteria infection.

Recent studies with drugs that act in the iron metabolism of M. abscessus confirm that this metal is crucial for the development of this bacillus (Abdalla et al., 2015b). However, the genes and proteins involved in the iron homeostasis and their importance for establishing infection remain unclear. The present study demonstrates for the first time that M. abscessus subsp. massiliense has two ferritin (non-heme binding) proteins involved in iron storage and related in the iron homeostasis both in vitro and in infected macrophages.

#### MATERIALS AND METHODS

#### M. abscessus subsp. massiliense GO06 Genome Annotation

The complete genome sequences of M. abscessus subsp. massiliense GO06 (Mycma GO06, taxid: 1198627) and the pathogen reference strain of M. tuberculosis H37Rv (taxid: 83332) used in this study were obtained from NCBI<sup>1</sup> . The BLAST tool from NCBI<sup>2</sup> was used for genome and proteome annotations of the Mycma GO06 strain as well as other mycobacteria species genomes.

#### Bacterial Strains and Growth Conditions

Escherichia coli XL1-Blue and BL21 (DE3) pLysS were used for cloning and expression of the recombinant proteins, respectively. E. coli strains were cultured in Luria Bertani (LB) broth (Himedia) and M. abscessus subsp. massiliense

<sup>1</sup>http://www.ncbi.nlm.nih.gov/genome/

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

(Cardoso et al., 2008) was cultured in Mueller Hinton broth or agar at 37◦C under 180 rpm shaking. For growth in different iron concentrations, the minimal medium contained 3.6 mM of KH2PO4, 2.0 mM of MgSO<sup>4</sup> 7H2O, 6% (v/v) of glycerol, 30 mM of L-asparagine, 0.006 mM of ZnSO4, and 0.05% (v/v) of Tween 80, pH 6.8 in iron free conditions. For low iron conditions, media was supplemented with 1.25 mM of deferoxamine mesylate (DFO). In high iron conditions, media was supplemented with 50, 150, 300, or 450 µM of FeCl3. Minimal media without supplementation (iron or DFO) contains enough iron concentration to sustain M. abscessus subsp. massiliense growth, and thus this condition was used as normalizer. E. coli transformants were selected in medium supplemented with the antibiotics kanamycin (KAN) and chloramphenicol (CAM) at 20 µg/ml.

#### Homology Models

The predicted 0076 and 0077 amino acid sequences were initially submitted to I-TASSER (Zhang, 2008) and an initial model was obtained for each sequence. I-TASSER strategy starts from the structure templates identified by LOMETS (Wu and Zhang, 2007) in the PDB library. I-TASSER only uses the templates of the highest significance in the threading alignments, which are measured by the Z-score. The C-score for each model was verified to evaluate the quality of the predictions from I-TASSER. C-score values are related to the expected TM-score (Zhang and Skolnick, 2004, 2005) between the model and native structure (structural similarity of two proteins).

#### Molecular Dynamics Simulations

To explore the stability and conformational variability of the initial models in its native environment, molecular dynamics (MD) simulations were performed with Gromacs 5.1.3 (Berendsen et al., 1995; Yu, 2012) using force field AMBER99SB-ILDN (Yildirim et al., 2010). The proteins were solvated with a box cubic wall distance of 10 Å using water model TIP3P (Mahoney and Jorgensen, 2000). The system was neutralized by adding the required number of counter ions according of each protein. The system was initially minimized using the steepest descent energy. The simulations were complete when the tolerance no longer exceeded 1000 kJ/mol. In the next three steps consisted of 50 picoseconds MD simulations in NVT and NPT ensemble at 300 K with a restraint of 50 kcal/mol/Å on the protein atoms and 0.5 ns without restraint in NPT ensemble at 300 K. Finally, the simulations were performed for 100 ns for all proteins with a constant temperature of 300 K, 1 atm pressure, time-set of 2 fs, and without any restriction of protein conformations. All information concerning the trajectory of these times were collected every 50 ps. The equilibration of the trajectory was evaluated by monitoring the equilibration of quantities, such as the root-mean-square deviation (RMSD) (Coutsias et al., 2004) of the non-hydrogen atoms, total energy, potential energy, and kinetic energy. The conformations that best represented the structures of the entire trajectory were selected following the algorithm described by Daura et al. (1999). A cutoff of 0.2 nm for the clusters was used considering the profile of the RMSD observed for each protein. The clusters were determined using the non-hydrogen atom RMSD values. The quality of the predicted structure was assessed using the MolProbity server (Chen et al., 2010).

### Gene Expression Evaluation of M. abscessus subsp. massiliense

Mycobacterium abscessus subsp. massiliense cultures grown in minimal media with different iron concentrations were harvested by centrifugation at 3,200 × g for 10 min. The pellet was suspended with 1 ml of nuclease-free water (Ambio Life Technologies) and 0.25 ml of glass beads (0.1 mm in diameter, Glass Glass Disruptor Beads; USA Scientific, Inc.) was added. The suspension was maintained in ice and vortexed five times for 2 min with 30 s intervals. The Lysate was centrifuged at 12,000 × g for 10 min at 4◦C, and the aqueous phase was transferred to a new tube for addition of 215 µl of ethanol (J.T. Baker) for each 400 µl of recovered solution. The solution was then applied to an RNA purification column (Phenol-Free Total RNA Purification, Amresco) for purification according manufacture's instructions. The obtained RNA was treated with RNAse free DNAse (Sigma-Aldrich) and stored at −80◦C until further use.

The Reverse Transcriptase M-MLV kit (Sigma-Aldrich) was used for cDNA synthesis. The reaction consisted of 200 ηg of total RNA, 0.5 mM of dNTPS and 0.63 µM of random hexamer primers (Gibco/Thermo Fisher Scientific) and was incubated for 5 min at 65◦C. Next, the system was transferred to ice and reverse transcriptase buffer, 200 U of M-MLV reverse transcriptase, and 40 U of RNAse OUT (Invitrogen) were added. The reaction was incubated at 37◦C for 1.5 h. Then the synthesized cDNA was stored at −20◦C until its use for Real Time PCR (RT-PCR). RT-PCR was set up in a 0.2 ml tube, using 10 µl of SYBR Green mix (Bio-Rad), 0.5 µM of each primer (**Supplementary Table S2**) and 5 µl of cDNA in a final volume of 20 µl. The reaction was run on a IQ5 thermocycler (Bio-Rad). RT-PCR conditions were as follows: 95◦C (5 min), 40 cycles of 95◦C (15 s), 58◦C (30 s), and 72◦C (1 min), and at the end a melting temperature curve ranging from 70 to 99◦C (ramp rate of 0.5◦C per cycle and 30 s in each temperature) was performed and the detected fluorescence emission recorded. Positive samples were considered when the fluorescence surpassed the threshold baseline. The cycle of threshold crossing corresponded to the Ct value. Ct values greater than 35 were considered negative. Ct values were tabulated on an Excel 2011 spreadsheet, and the relative expression was determined with the Delta Delta Ct (2−11Ct) method using the expression of the 16s rRNA gene as normalizer. The calibrator condition in this study was bacteria grown in minimal media without DFO and FeCl3, as these conditions contain sufficient iron levels to support mycobacteria growth. The relative gene expression levels were analyzed on GraphPad Prism 7 (version Prism 7a, Graph Pad) for statistical analysis and graphic representations.

## Infection of Bone Marrow-Derived Macrophages (BMDM)

To evaluate the ex vivo gene expression of M. abscessus subsp. massiliense, bone marrow from C57BL/6 mice were collected

(Becker et al., 2012) and submitted to differentiation as previously described da Costa et al. (2017). BMDM (1 × 10<sup>6</sup> /ml) were infected with M. abscessus subsp. massiliense at a MOI of 10 in a 24 well plate with or without coverslip. Three hours after infection, extracellular bacteria were removed by washing the wells twice with RPMI with 10 µg/ml of kanamycin and then adding RPMI media supplemented with 10% fetal bovine serum (FBS). The CFU determination and expression profile of the genes mycma\_0076 and mycma\_0077 during infection was assessed by recovering the bacilli at three different times: 3, 24, 48, and 72 h post infection and additionally the wells with coverslip 24 and 72 h were randomly selected to stained with Instant-Prov (NewPRQV) according to the manufactured instruction's. At these times, the wells were randomly selected and from them the supernatant was removed and substituted with nuclease free water (Ambion) to lyse macrophages. The lysate was transferred to 1.5 ml nuclease free tubes, centrifuged at 16,000 × g for 10 min at 4◦C and the pellet was processed for RNA extraction. The relative gene expression was determined by the delta delta Ct (2−11Ct) method using the expression of the 16s rRNA gene as normalizer. The bacilli, obtained from culture supernatant after 3 h of macrophage infection, was used as calibrator.

#### Cloning and Expression of mycma\_0076 and mycma\_0077 Genes

The mycma\_0076 and mycma\_0077 genes were amplified by PCR using M. abscessus subsp. massiliense GO06 (Raiol et al., 2012) genomic DNA as template. The primers were designed using NCBI Primer designing tool<sup>3</sup> . The following primers were used to amplify the mycma\_0076 gene include: forward 5<sup>0</sup> CATATGACCGCGACCGACACCCCGA 3<sup>0</sup> that incorporates an NdeI restriction site (underlined) and reverse 5<sup>0</sup> GGATCCTCTTGTTGACGTGCTTAGAGCG 3<sup>0</sup> that incorporates a BamHI restriction site (underlined). Similarly, the primers for the mycma\_0077 gene amplification were: forward 5 <sup>0</sup> CATATGGTGGCTACCACCGATCTCCATG 3<sup>0</sup> and reverse 5<sup>0</sup> CTCGAGCGCAAAATTATCAGAGCGCGC 3<sup>0</sup> that incorporate an NdeI and an XhoI restriction sites, respectively. The PCR products of each gene were cloned into pET28a (Novagen) vector using their respective flanking sites. Recombinant plasmids were confirmed by sequencing. Recombinant protein expression was performed by transforming the recombinant plasmids into E. coli BL21 (DE3) pLysS cells.

### Recombinant Protein Purification

Escherichia coli BL21 (DE3) pLysS containing the recombinant plasmids were grown in LB containing kanamycin (20 µg/ml) and chloramphenicol (20 µg/ml) until OD550 nm reached 0.5. Then the culture was induced with 1 mM of isopropryl-1 thio-β-D-galactopyranoside (IPTG) at 37◦C, 180 rpm for 4 h. Cells were then harvested by centrifugation at 4,000 × g for 20 min at 4◦C. The pellet was used for protein extraction using the commercial protein purification QIAexpress-Ni\_NTA Fast Start kit (Qiagen) according to the manufacturer's instructions. Proteins eluted from the nickel column were further purified on a gel filtration Superdex 200 10/300 GL chromatographic column (GE Healthcare). The column was previously equilibrated with 50 mM NaH2PO<sup>4</sup> buffer adjusted at pH 8.0 containing 300 mM NaCl (pH 8.0) buffer. The column was calibrated with molecular weight standards (GE Healthcare), and chromatography was performed at a 1 ml/min rate with 5 MPa pressure and detection at 280 nm on an AKTA purifier system (GE Healthcare). Eight microliters from each collected fraction were analyzed on 12% SDS-PAGE. Protein concentration was determined by using Bradford's reagent with bovine serum albumin as the standard.

### Mouse Anti-r0076 Antibody Production

Three C57BL/6 mice were immunized by the subcutaneous route with purified r0076 protein. In the first immunization, a formulation consisting of 50 µg of r0076 protein and 50% (v/v) of complete Freund adjuvant was administered. Fourteen days later the same amount of protein was used mixed with incomplete Freund adjuvant. The third immunization was performed 14 days after the second with the same formulation as the second. Ten days after the last immunization, total blood was collected and incubated for 30 min at room temperature. The blood was centrifuged at 3,000 × g for 10 min and the sera was aliquoted in 50 µl volumes and stored at −20◦C until their use. The antiserum was titrated and used in western blotting experiments.

#### Culture Filtrate Proteins (CFP) and Cell Lysate Obtention

Mycobacteria cultures at logarithmic growth were harvested at 6,000 × g for 10 min. The supernatant and pellet were processed for CFP and cell lysate preparations, respectively. The supernatant was filtered through a 0.22 µm filter and then concentrated by centrifugation using a 10 kDa (Amicon) centricon filter at 7,000 × g for 30 min at 4◦C. Glycerol was added to the obtained CFP to a final concentration of 20% and CFP was stored at −20◦C until use. The culture pellet was resuspended in PBS buffer and sonicated in an ice bath twice for 1 min to obtain cell lysate. The cell lysate was adjusted to 20% glycerol and stored at −20◦C.

## Western Blotting

After electrophoresis of the proteins by PAGE, under denaturing or non-denaturing conditions, the separated proteins were electrotransferred to a nitrocellulose membrane. The membrane was blocked with an incubation of 2 h with PBS containing 5% skimmed milk at room temperature. Then the membranes were incubated overnight at 4◦C with the mouse serum against r0076 diluted 1:500 in PBS containing 2% skimmed milk. The membrane was then washed three times with PBS buffer and incubated with 4 µg of secondary anti-Mouse-F (ab') 2-xx-biotin (Molecular Probes) for 2 h at 37◦C. Then, horse anti-mouse antibody conjugated with avidin-peroxidase (Sigma-Aldrich) was added and incubated for 1 h. The reaction was developed by adding 0.05% diaminobenzidine (DAB, Roche) in 10 ml of H2O2. The image was acquired with the help of Gel documentation system (Bio-Rad) and analyzed with Quantity One 4.5.6 software (Bio-Rad).

<sup>3</sup>http://www.ncbi.nlm.nih.gov/tools/primer-blast

### Circular Dichroism (CD) Spectroscopy

Circular dichroism spectrum was collected using a Jasco-815 spectropolarimeter equipped with a temperature control device. The r0076 concentration was 5 µM in 50 mM NaH2PO<sup>4</sup> buffer adjusted at pH 8.0 containing 50 mM NaCl. All data were collected using 1 mm quartz cuvette. The spectrum was recorded over the wavelength range from 195 to 260 nm. A total of eight accumulations were averaged to form the CD spectrum, using a scanning speed of 100 nm/min, a spectral bandwidth of 1 nm, and a response time of 0.5 s. The buffer contribution was subtracted in each experiment. Thermal denaturation of r0076 at pH 8.0 was characterized by measuring the ellipticity changes at 222 nm induced by a temperature increase from 20 to 90◦C. The fraction of denatured protein (α) was calculated from the relationship: α = (θ<sup>n</sup> − θobs)/(θ<sup>n</sup> − θd) and α + β = 1, in which θobs is the ellipticity obtained at a particular temperature, and θ<sup>d</sup> and θ<sup>n</sup> are the values of the ellipticity characteristic of the denatured and native states, respectively.

### Dynamic Light Scattering (DLS)

The size of r0076 was examined by means of the Nano-ZS dynamic light scattering system (Malvern Instruments Ltd., Malvern, United Kingdom). This system employs a λ = 633 nm laser and a fixed scattering angle of 173◦ . The r0076 solution (1 mg/ml), in buffer 50 mM NaH2PO<sup>4</sup> buffer adjusted at pH 8.0 containing 50 mM NaCl, was centrifuged at 16,000 × g for 10 min at room temperature, and subsequently loaded into a quartz cuvette prior to measurement. The temperature was raised from 20 to 90◦C and the sample was allowed to equilibrate for 2 min in each temperature prior to DLS measurements. The hydrodynamic radius (RH) was determined from a second-order cumulant fit to the intensity auto-correlation function (size distribution by volume). The determined R<sup>H</sup> was converted to molecular mass (kDa) based on the assumption of a spherical particle and using the Zetasizer software.

#### Iron Oxidation Assays

Oxidation reactions were performed according to Khare et al. (2011) using a fresh solution of 0.1 M of HEPES, pH 6.5 containing 125 µM of ammonium ferrous sulfate. The recombinant protein was added to the buffer containing ferrous sulfate for a final concentration of 0.25 µM, and the optical density was monitored at 310 nm for 18 min at 37◦C. At this wavelength, the Fe3<sup>+</sup> is detected, and consequently, the amount of oxidation can be monitored. To determine the amount of oxidized iron, additional replicate reactions were performed, but ferrozine iron reagent was also added to the reaction. Ferrozine makes a complex with free ferrous iron in solution, resulting in a violet color solution that can be detected at 570 nm. Ferrozine was added at 3-min intervals to individual wells and the 570 nm was recorded. A ferrous iron concentration curve was generated by adding ferrozine to different Fe2<sup>+</sup> concentrations and recording the optical density (O.D.) at 570 nm. The experimental readings were converted to concentration based on the generated curve. The concentration at time zero was considered 100%, and the remaining concentrations were transformed in percentages relative to time zero. In all oxidation reactions, the 50 mM NaH2PO<sup>4</sup> buffer adjusted to pH 8.0 containing 50 mM NaCl was used as negative control.

### Ethical Committee

The study was approved by the Ethics Committee for Animal use (CEUA: Comite de Ética no uso de animais; #229/11) of the Universidade Federal de Goiás (UFG), Goiânia, Brazil.

### Statistical Analysis

Comparison between means was assayed for variance (ANOVA) and non-paired t-test using Prism software version 6.0c (GraphPad). Values of p < 0.05 were considered statistically significant.

## RESULTS

#### Mycobacteria Belonging to the Mycobacterium abscessus Complex Have Two Genes Possibly Coding for Ferritin Proteins

To identify possible genes coding for bacterioferritin and ferritin in the M. abscessus subsp. massiliense genome, a BLAST using the genes bfrA (Rv1876) and bfrB (Rv3841) from M. tuberculosis H37Rv performed against M. abscessus genomes and M. abscessus subsp. massiliense did not present any gene with significant similarity to the Rv1876 gene (**Supplementary Table S1**). However, M. abscessus subsp. massiliense has two genes with similarities higher than 70% to the Rv3841 gene (**Table 1**). Both mycma\_0076 and mycma\_0077 genes (**Figure 1A**) are located in tandem in the genome. Similar results were seen for other M. abscessus subspecies and other non-tuberculosis mycobacteria (NTM). A phylogenetic tree with the bacterioferritin and ferritin protein sequences from different mycobacteria species was constructed (**Figure 1B**). This shows that mycobacteria species closest to the group of M. abscessus may have two ferritins, while M. tuberculosis and other mycobacteria have one of both ferritin and bacterioferritin proteins. Thus, M. abscessus subsp. massiliense and other closely related genetic mycobacterial species do not have genes that are similar to the bacterioferritin gene bfrA (Rv1876) from M. tuberculosis, which suggests for the first time that mycobacteria from the M. abscessus complex and their closely related species have two genes possibly coding for ferritin.

#### Molecular Dynamics Evaluation of 0076 and 0077 Proteins Demonstrate a Ferritin Like Protein

To correlate the genes mycma\_0076 and mycma\_0077 with ferritin, their hypothetic structures were modeled and analyzed by molecular dynamics (MD). Initial models of 0076 and 0077 proteins were built from I-Tasser server using as

principal templates (PDB files) 3qd8A (Crystal structure of M. tuberculosis BfrB), 1vlgA (Crystal structure of Ferritin from Thermotoga maritima), 3unoA (M. tuberculosis ferritin homolog, BfrB), and 1z6oA (Crystal Structure of Trichoplusia ni secreted ferritin). The information from each template compared to 0076 and 0077 proteins are shown in (**Table 2**). The best model (model 1) for 0076 and 0077 structures had a C-score of 1.13 and 0.92, respectively. These values provide an estimate for TM-score above 0.84 and an RMSD below 3.5 Å for both models. These predicted values indicate that the models determined by ITASSER have a great chance of representing the expected native structures for the 0076 and 0077 proteins.

The MD simulations from theses initial models were performed to achieve stability and/or improve the structure quality of them. In **Figure 2**, the RMSD evolution from initial models is shown for 0076 and 0077. For both proteins, after 60 ns, a transient stability could be verified for them. Just one simulation from 0077 protein had high fluctuations after 60 ns, which is mainly associated to moves from the residues located at the N and C-terminal (**Figure 2**).

From the trajectory of each simulation, cluster analysis of the conformations with a cutoff of 2 Å helped identify multiple conformations that could represent their flexibility. We selected only the center structure of the cluster most common during the simulations to represent each protein. **Figure 3** shows the clusters obtained for structures over time. For 0076 simulations (**Figure 3A**), cluster number one (most frequent structure) appeared only after about 60 ns, which remained stable until the end of the simulations. The same feature was observed for 0077 simulations (**Figure 3B**). In MD1 simulation of 0077 protein, the number of clusters observed between 60 ns and 80 ns fell from 70 to less than 20 (**Figure 3B**). Outside this interval, more intense structural fluctuations occur around N terminal region. The quality of the selected structures was measured by molprobity score, which indicates a better quality for the structure when its value tends to zero. **Table 3** shows that the quality of the models (model 1) had comparable molprobity scores from high-resolution structures. The highest value was 1.65 for 0077 (DM2) model, which is still a very common value in high resolution structures. The structural alignment of the likely active site of the Helicobacter pylori ferritin structure (PDB id 3bvi – chain C) and proteins 0076 and 0077 is shown in **Figure 4**.

#### Evaluation of mycma\_0076 and mycma\_0077 Genes Expression

Bacteria require a mechanism for iron storage for efficient homeostasis of this ion and to avoid the deleterious effects of iron excess (Pandey and Rodriguez, 2012; Reddy et al., 2012). M. abscessus subsp. massiliense growth did not alter in different iron concentrations, ranging from minimal concentrations to excess conditions, such as 450 µM FeCl<sup>3</sup> (**Figure 5A**). However, when iron was completely removed from the media, the mycobacteria growth was seriously compromised (**Figure 5A**). Thus, M. abscessus subsp. massiliense has mechanisms for iron homeostasis that allows this bacterium to grow in conditions of iron overload.

As mycma\_0076 and mycma\_0077 genes possibly correspond to ferritin genes, their expression was evaluated during M. abscessus subsp. massiliense growth in different iron concentrations. Surprisingly, the mycma\_0076 gene had its expression up regulated 50 times in high iron concentrations, while mycma\_0077 gene was not induced under those conditions (**Figure 5B**). Thus, only mycma\_0076 gene seemed to have a positive correlation between iron levels and expression (**Figure 5B**).

#### The Expression of mycma\_0076 and mycma\_0077 Genes Is Modulated During Macrophage Infection

In order to understand if the differential expression observed in vitro was also used by M. abscessus subsp. massiliense to overcome the infection, the expression of mycma\_0076 and mycma\_0077 genes were evaluated during macrophage infection.

TABLE 1 | BLAST results from similarity search for Mycobacterium tuberculosis H37Rv Rv3841 gene.


Alignments were performed with Basic Local Alignment Search Tool (BLAST) program at http://blast.ncbi.nlm.nih.gov, using the nucleotide BLAST algorithm.

from alignments of ferritin and bacterioferritin protein sequences. (A) Genomic organization of the M. abscessus subsp. massiliense genes similar to M. tuberculosis H37Rv Rv3841. (B) Phylogenetic tree constructed from ferritin (BfrB) and bacterioferritin (BfrA) mycobacterial protein sequences performed on ClustalX. The protein sequences were obtained from NCBI BLAST searches for proteins with similarity to ferritin and bacterioferritin from M. tuberculosis H37Rv.

TABLE 2 | Accuracy of models and threading templates information.


∗ Identity is the percentage sequence identity of the whole template chains with query sequence. # Coverage represents the coverage of the threading alignment and is equal to the number of aligned residues divided by the length of query protein.

In contrast to the in vitro observations, expression of both genes was induced during macrophage infection, but these genes were differently modulated (**Figures 6A,B**). While the expression of mycma\_0076 gene was reduced 3 h after macrophage infection (**Figure 6A**), the mycma\_0077 had its expression up regulated 80 times (**Figure 6B**). At 24 h of infection both genes had similar levels of expression, however, after 48 h the expression of gene mycma\_0076 was reduced again, while the expression of mycma\_0077 remained highly expressed (**Figures 6A,B**). After 72 h of infection, both genes were expressed at lower levels compared to 48 h (**Figures 6A,B**). These results suggested that the expression of mycma\_0076 and mycma\_0077 genes could be related to the establishment of infection, but their possible role in this process is not redundant. Moreover, it was observed

FIGURE 3 | Cluster analysis of the 0076 (A) and 0077 (B) protein trajectories obtained over 100 ns. A cutoff of 0.2 nm was selected to include the main structures during the simulations. The cluster structures were determined using the non-hydrogen-atom RMSD values. Insert figures represent the cluster structures from MD1 (purple) and MD2 (green) independent simulations.


TABLE 3 | Molprobity score and ferritin active site key residues for proteins 0076 and 0077.

\$Clashscore measured from molprobity program. <sup>∗</sup>Molprobity scores after molecular dynamics simulations. &Percent scores observed in high resolution structures.

was analyzed by quantitative RT-PCR with SYBR green under high (150 µM FeCl3) and low (DFO) iron conditions. The relative gene expression was determined by the Delta Delta Ct (2−11Ct) method using the expression of the 16s rRNA gene as normalizer and the growth under non-supplemented (NS) condition as calibrator. ∗∗∗∗p < 0.0001 indicate statistically significant difference between groups.

that expression of mycma\_0076 and mycma\_0077 increased accompanied by the growth of bacilli inside of macrophages (24 h) (**Figures 6A–C**). However, the bacterial number within macrophage culture reduced after 72 h (**Figure 6C**), as did the expression both genes were (**Figures 6A,B**). It is of important notice that the observed decrease in intracellular bacteria was accompanied by increase in extracellular bacilli (**Figure 6D**) and macrophage death (data not shown). These data suggest that the expression of both genes are important for bacilli growth inside of macrophages, differently to extracellular growth as in RPMI when the mycma\_0076 gene was predominantly expressed (**Figures 6A,B**).

## Protein 0076 Cytolocalization in M. abscessus subsp. massiliense

Both mycma\_0076 and mycma\_0077 genes from M. abscessus subsp. massiliense were separately cloned in pET28a(+) plasmid, expressed in E. coli and the recombinant proteins were purified (**Figures 7A** and **Supplementary Figure S2**). While recombinant 0076 (r0076) protein was easily obtained in its soluble form, r0077 had very low solubility and yield. Consequently, some experiments for ferritin characterization was performed only for r0076. The protein 0076 was detected by specific polyclonal antibodies only in the cellular fraction of M. abscessus subsp. massiliense (**Figure 7B**).

FIGURE 6 | Relative gene expression of mycma\_0076 (A) and mycma\_0077 (B) genes and growth of M. abscessus subsp. massiliense during infection of BMDM. (A,B) Macrophages were infected at a multiplicity of infection of 1:10 with M. abscessus subsp. massiliense, and total RNA was extracted at 3, 24, 48, and 72 h after infection. Expression of mycma\_0076 and mycma\_0077 was analyzed by quantitative RT-PCR with SYBR green. The relative gene expression was determined by the delta delta Ct (2−11Ct) method using the expression of the 16s rRNA gene as normalizer and the bacilli obtained from supernatant of macrophage infection after 3 h as calibrator. (C) After 3, 24, 48, and 72 h of macrophage infection the macrophages were washed, lysed with water and plated on MH agar for CFU determination. (D) After 3, 24, 48, and 72 h of macrophage infection, the supernatant was collected for CFU quantification of extracellular bacilli. <sup>∗</sup>p-value < 0.05 indicate statistically significant difference between groups as compared with 3 h post infection. (E) After 24 and 72 h of incubation, uninfected and infected cells were stained and analyzed by light microscopy (Leica Application Suite v.4.4.0) to observe cell damage.

on a western blotting using polyclonal antibodies raised against r0076. The purified recombinant protein was used as control (r0076). M, prism protein marker (Amresco); CFP, culture filtrate protein.

## Recombinant 0076 Protein Complex Formation

The results presented above support the function of the protein 0076 as a ferritin. Several studies have shown that ferritin proteins require the formation of an oligomeric structure to perform iron oxidation and storage (Levi et al., 1988; Khare et al., 2011, 2013). We could show that this was the case for r0076 by performing western blotting of the protein under native polyacrylamide gel conditions. As shown in **Figure 8A**, r0076 forms a high molecular mass protein. Upon gel filtration analysis, r0076 elutes as single peak of apparent molecular mass of 480 kDa (**Figure 8B**) similar to ferritins.

### Recombinant Ferritin From M. abscessus subsp. massiliense (r0076) Forms Stable Oligomers in Solution

Circular dichroism spectroscopy was used to analyze the secondary structure of the r0076 in solution at pH 8.0 (**Figure 9A**). The CD spectrum of r0076 is characterized by two minima at 210 ± 1 nm and 222 ± 1 nm, a maximum near 200 ± 1 nm, and a negative to positive crossover at 201 ± 1 nm. The negative minimum was around 210 and 222 nm, which strongly indicate the presence of α helices, comparable to the secondary structure of other ferritins (Khare et al., 2011, 2013). As a next step, the quaternary structure of r0076 was analyzed in

solution at pH 8.0 by dynamic light scattering (DLS). When r0076 was analyzed by DLS the observed profile was characteristic of a monodisperse solution (**Figure 9B**). The value of hydrodynamic radius (RH) determined for r0076 was 8.0 ± 0.5 nm, certainly corresponding to an oligomeric form of the protein in solution. This result is consistent with the size-exclusion chromatography profile obtained for r0076 (**Figure 8B**).

#### Influence of Temperature on r0076 Stability and Compactness

The thermostability of the r0076 at pH 8.0 was monitored following changes in the ellipticity at 222 nm (**Figure 10A**). The spectrum remained constant at temperatures below 55◦C. However, the spectrum was progressively altered when the temperature was increased above 55◦C, which indicates loss of the regular secondary structure. The melting temperature (Tm), value determined by CD spectroscopy for r0076, was 57 ± 1 ◦C (**Figure 10A**). The structural alteration observed by CD spectroscopy was accompanied by DLS analyses. **Figure 10B** shows the variation of R<sup>H</sup> as a function of temperature for r0076 at pH 8.0. The R<sup>H</sup> of r0076 exhibited minimal temperature dependence between the ranges of 20 to 55◦C. However, when r0076 was incubated at temperature values above 55◦C, the R<sup>H</sup> increased significantly, suggesting the formation of aggregates as a consequence of the denaturation process. The thermal denaturation process was essentially irreversible in the conditions described in this study (data not shown).

### Recombinant r0076 and r0077 Proteins Promote Oxidation of Fe2<sup>+</sup> Into Fe3<sup>+</sup>

The capacity of the recombinant proteins to promote the oxidation of ferrous iron was evaluated by incubating them with Fe2<sup>+</sup> and observing the increase in optical density at 310 nm. Both r0076 and r0077 proteins were capable to oxidize ferrous iron (**Figure 11A**), but the activity of r0076 was much greater than r0077. The r0076 protein oxidized 25% of the available Fe2<sup>+</sup> (**Figure 11B**) after 3 min, while r0077 protein oxidized only 16% during the same time (**Figure 11B**). After 18 min, the r0076 protein oxidized more than 80% of the available Fe2<sup>+</sup> while, r0077 oxidized 55%. These results demonstrate that both

r0076 and r0077 proteins are capable of oxidizing Fe2<sup>+</sup> into Fe3+, evidencing their activities as ferritins.

#### DISCUSSION

In the last decades, the M. abscessus complex has emerged as a human pathogen (Petrini, 2006; Medjahed et al., 2010). Its ability to infect and persist inside phagocytic cells and in the extracellular environment indicates that this bacterium has evolved to adapt and establish infection in humans (Bernut et al., 2014; Helguera-Repetto et al., 2014; Brambilla et al., 2016). In this study, we showed that M. abscessus subsp. massiliense has two ferritins similar to the Mtb ferritin that might be important for intracellular iron homeostasis, which in turn is crucial for successful infection.

It has been shown that E. coli and Haemophilus influenza have more than one gene coding for ferritin, while M. tuberculosis and M. smegmatis have only one (Andrews, 1998; Bou-Abdallah et al., 2014). Here, we showed for the first time that bacteria from the M. abscessus complex have two genes coding for ferritins and none for bacterioferritin (**Figure 1**, **Table 1**, and **Supplementary Table S1**). It was observed that the proteinscoded by mycma\_0076 and mycma\_0077 genes do not have the methionine (Met) residue at position 52 as observed in bacterioferritin from M. tuberculosis (data not shown). The absence of Met-52 may render these proteins unable to bind heme (Gupta et al., 2009; Yasmin et al., 2011) as shown in Met-52 M. tuberculosis mutants (Gupta et al., 2009; Khare et al., 2017). Absence of the gene with similarity to Rv1867 gene (**Supplementary Table S1**) and lack of Met-52 raise the possibility that M. abscessus subsp. massiliense do not have bacterioferritin homolog.

To confirm if the proteins 0076 and 0077 might support ferritin functions, structural and molecular dynamic analyzes were performed. Considering model-1 structures from MD1 and MD2 simulations, the RMSD (CA atoms) between them is 7.03 Å and 6.24 Å for 0076 and 0077 proteins, respectively. The 0076 and 0077 proteins have about 60% of α helix secondary structures and low content of β-sheet secondary structures (see **Table 2**). When residues involved in segments other than the helices are removed from the RMSD estimates, RMSD values fall to 1.20 Å (113 CA atoms) and 1.23 Å (111 CA atoms). This becomes clear in the RMSD fluctuations for residues shown in **Figures 2**, **3** from all clusters. RMSD above 5.0 Å occurred mainly in the segments involved in the N and C terminus, except around residues 130– 135, where sensitive structural fluctuations were observed. This may have provided instability in this region and even partial loss of the helix structure, such as illustrated in **Supplementary Figure S1**. The slight fluctuation of the residues involved in segment 130–135 may be due to the templates used to construct the model (3d8A) whose irons were present in the structure and not included in the simulations. On the other hand, it may also be associated with the flexibility expected to assist in the conformational rearrangement of this region to accommodate iron ions and make them more stable. Above all, irrespective of the presence of iron, the conformations of the proteins were stable at their sites, as expected for the positions of the key residues of a ferritin (**Figure 4**). This reinforces the idea that these structures are not dependent on iron for their stability and formation (Stookey, 1970; Waldo and Theil, 1993; Theil et al., 2000).

Although the main crystal structure used to construct the models was not solved with the presence of iron in this region (3qd8A), the MD simulations showed the importance of these ions for the stability of this site. **Table 2** shows that the main residues from Helicobacter pylori ferritin structure (GLU17, HIS53, GLU50 GLU94, GLN 127, GLU 130) are conserved in 0076 and 0077 proteins (Cho et al., 2009). Additionally, we observed that the residues involved in the self-assembly and stability are the same as those recently reported by Khare et al. (2013). The 3D position of theses residues for both structures are highly correlated with that observed for Helicobacter pylori ferritin structure (**Figure 4**), which strongly supports the ferritin activity and the same iron binding mechanism of these two proteins.

The expression of the mycma\_0076 and mycma\_0077 genes, evaluated in vitro with different iron concentrations was found to be differently regulated, suggesting different roles in iron homeostasis for this Mycobacterium. Furthermore, M. abscessus subsp. massiliense was able to grow in highly toxic iron concentrations (450 µM FeCl3), which indicates the presence of a homeostasis mechanism (**Figure 5A**). A recent transcriptomic

analysis study of M. abscessus subsp. abscessus grown in the presence of cystic fibrosis patient sputum listed the different expression of the genes similar to mycma\_0076 (MAB\_0126c) and mycma\_0077 (MAB\_0127c), although that study did not investigate ferritins specifically (Miranda-CasoLuengo et al., 2016). In this previous study, the gene MAB\_0126c was induced when grown with patient sputum as the stress condition. We found that the gene mycma\_0076 was also induced under in vitro conditions with high stressing concentration of iron (**Figure 5B**). Similarly, the requirement of ferritin expression to reduce the effects of oxidative damage was observed in Mtb (Pandey and Rodriguez, 2012, 2014). Thus, among other functions, ferritins play an important role in the resistances against stress conditions (Khare et al., 2017).

The different expression profiles of both mycma\_0076 and mycma\_0077 genes under different iron concentrations (**Figure 5B**), suggests that the proteins coded by theses genes have different roles in iron homeostasis. Recent studies have shown that Mtb ferritin and bacterioferritin have different roles in the cellular homeostasis, suggesting that the presence of two classes of ferritins is non-redundant and important for virulence (Khare et al., 2017). The interesting question is why does M. abscessus subsp. massiliense have two similar proteins of the same ferritin group? The overexpression of 0076 in high iron concentrations suggests that this protein may be involved in the storage of the ion providing protection to the bacilli from iron-mediated toxicity (**Figure 5B**).

Nonetheless, both mycma\_0076 and mycma\_0077 genes were expressed during macrophage infection, but they were differentially regulated according to the time of infection, which indicates that inside of macrophages, M. abscessus subsp. massiliense find a different microenvironment as compared with the medium, requiring different expression of those genes. It was observed during macrophage infection that the mycma\_0077 gene was expressed at higher levels, when compared with the levels of expression the mycma\_0076 (**Figures 6A,B**). That difference suggested that the expressions of mycma\_0076 and mycma\_0077 genes and their respective involvement in iron homeostasis are largely dictated by the microenvironment surrounding the cell, and they may play different or redundant but independent roles in iron management. Additionally, during macrophage infection, the M. abscessus find a more oxidizing environment compared to an uninfected cell, but the bacilli grow better in this condition (Oberley-Deegan et al., 2010). Moreover, it was demonstrated that an enhanced oxidative stress happens at 24 h post infection of macrophages when the bacilli growth increase (Oberley-Deegan et al., 2010). Our results demonstrated that the mycma\_0076 and mycma\_0077 genes were induced at the same time during macrophage infection (**Figure 6**). Furthermore, we have previously shown that the infection of M. abscessus subsp. massiliense induced high levels production of NO by spleen and liver cells (Martins de Sousa et al., 2010). Our findings raise the possibility that induction of the mycma\_0076 and mycma\_0077 genes expression during macrophage infection could be related to the resistance of these mycobacteria from oxidative stress caused by the macrophage. Besides, it was demonstrated in M. tuberculosis that the ferritin provide protection against oxidative stress and the deletion of this protein increased the sensibility to oxidative damage (Pandey and Rodriguez, 2012; Khare et al., 2017).

After 72 h post infection the burden of bacilli inside macrophages reduced significantly as compared with 48 h (**Figure 6C**) and in the same time of infection, the expression of the mycma\_0076 and mycma\_0077 genes also were reduced (**Figures 6A,B**). An initial interpretation of this observation could be related to the control of infection by macrophages, however, it was observed that the mycobacteria was extravasating to the extracellular milieu (**Figure 6D**). It has been reported that the M. abscessus can induce apoptosis of macrophages as a mode of mycobacterial escape for release and growth at the extracellular milieu (Sohn et al., 2009; Bernut et al., 2014; Brambilla et al., 2016; Whang et al., 2017). To confirm that this was the case, infected and uninfected cells were stained and analyzed 24 and 72 h post incubation. Damaged macrophages were observed after 24 h of M. abscessus subsp. massiliense infection and increased as the infection progressed (**Figure 6E**) concomitant to the increase of extracellular bacteria (**Figure 6D**). This result indicates that M. abscessus subsp. massiliense can induce damage on macrophages and consequently be released to the extracellular environment when the expression of the

mycma\_0076 and mycma\_0077 genes would not be as much necessary as within the intracellular environment.

To confirm the iron storage characteristics of the protein encoded by the mycma\_0076 and mycma\_0077 genes, the recombinant proteins were expressed in E. coli. While r0076 was obtained in the soluble form, r0077 remained mostly in insoluble form despite different attempts to obtain it in a soluble form. The CD spectrum obtained for r0076 is typical of proteins with high content of α-helical secondary structure (**Figure 9A**). Moreover, the formation of oligomers in solution (**Figure 9B**), together with the presence of bound iron ions, indicates that r0076 folded correctly. As seen in **Figure 8**, r0076 eluted as a major peak (**Figure 8B**) with an apparent molecular mass of 480 kDa. This result is broadly consistent with an oligomer composed of 24 subunits, whose theoretical expected mass for each subunit is 20 kDa. When r0076 was analyzed by DLS, a R<sup>H</sup> of 8.0 ± 0.5 was determined. Assuming a spherical particle, the value determined for R<sup>H</sup> correspond to a molecular mass of 437 ± 63 kDa, which would also be consistent with an oligomer composed of 24 subunits. The quaternary structures of several ferritins were found to be strikingly similar (Khare et al., 2011, 2013). The ferritin from M. tuberculosis exhibits a quaternary structure where 24 subunits assemble in octahedral 432-symmetric arrangements to form a roughly spherical protein shell (Khare et al., 2011).

The CD spectrum of r0076 was constant below 55◦C; however, above this temperature there was a progressive loss of regular secondary structure (**Figure 10A**). Concomitantly, the R<sup>H</sup> of r0076 exhibited minimal temperature dependence in the range of 20–55◦C, while the R<sup>H</sup> increased significantly when the protein was incubated at temperatures above 55◦C (**Figure 10B**). Taken together, these results indicate that the increase in temperature does not induce r0076 dissociation prior to denaturation.

The main function of ferritin is to detoxify and store free cellular iron, which is accomplished by oxidation reaction at the ferroxidase center (Andrews, 1998). Molecular dynamics of both models obtained from the amino acid sequences of 0076 and 0077 proteins found conserved amino acid residues involved in the ferroxidase active site as shown for H. pylori crystallographic resolved ferritin protein (**Figure 4**) (Cho et al., 2009). We showed here that both proteins r0076 and r0077 are capable of oxidizing Fe2<sup>+</sup> into Fe3+, confirming their ferritin function similarly to Mtb ferritin (**Figure 11**) (Khare et al., 2011, 2013).

Although the essentiality of the ferritin genes reported here can only be demonstrated by silencing their expression (knockdown them out for example), we have clearly shown for the first time that mycma\_0076 and mycma\_0077 genes codes for a ferritin.

#### CONCLUSION

The genes mycma\_0076 and mycma\_0077 from M. abscessus subsp. massiliense code for bacterial ferritins, homologous to Mtb ferritin gene (Rv3841), that are differently modulated by iron concentration both in vitro and in vivo. Additionally, both proteins r0076 and r0077 were capable to oxidize Fe2<sup>+</sup> into Fe3<sup>+</sup> supporting an active ferroxidase center. The implications that M. abscessus complex has only ferritins and no bacterioferritins should be further explored.

### AUTHOR CONTRIBUTIONS

FO, ADC, and VP carried out most of the experiments. RS performed the MD experiments and wrote the pertinent data of these experiments. WG and JA performed the CD and DLS experiments and wrote the results and discussion regarding this data. AJ-K and AK designed the experiments and supervised all work. FO, AJ-K, and AK wrote the manuscript. All authors revised the manuscript and approved the final version.

#### FUNDING

This study had financial support from CNPq (307186/2013-0 and 303675/2015-2) and FAPEG (2012/0267000-48 and 2013/10267000-46).

### ACKNOWLEDGMENTS

The authors are thankful to Dr. Cirano Jose Ulhoa and Dr. Fabrícia Paula de Faria from Universidade Federal de Goiás for their laboratory support with the experimental procedures for the purification of the recombinant proteins. This manuscript is part of the Ph.D. thesis of FO.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2018.01072/full#supplementary-material

FIGURE S1 | Secondary fluctuations structures over 100 ns from 0076 to 0077 proteins. (A) Secondary structures of 0076 MD1. (B) Secondary structures of 0076 MD2. (C) Secondary structures of 0077 MD1. (D) Secondary structures of 0077 MD2.

FIGURE S2 | Solubilization of r0077 protein from inclusion bodies using sarkosyl and purification. (A) SDS-PAGE gel showing lysis of bacterial cells induced or not with 1 mM IPTG for 4 h. (B,C) SDS-PAGE gel showing solubilization of r0077 protein with various concentration of sarkosyl. (D) Purification of r0077 using His tag from soluble fraction obtained with 2.5% sarkosyl. (E) SDS-PAGE gel showing r0077 protein purified concentrated.

TABLE S1 | BLAST results from similarity search for the Rv1876 gene from M. tuberculosis H37Rv.

TABLE S2 | Primer sequences used to evaluate the expression of M. abscessus subsp. massiliense genes.

#### REFERENCES

fmicb-09-01072 June 1, 2018 Time: 14:47 # 15



Andrews, S. C., Robinson, A. K., and Rodriguez-Quinones, F. (2003). Bacterial iron homeostasis. FEMS Microbiol. Rev. 27, 215–237.

Mycobacterium abscessus. Free Radic. Biol. Med. 49, 1666–1673. doi: 10.1016/ j.freeradbiomed.2010.08.026


abscessus subsp. abscessus and Mycobacterium abscessus subsp. bolletii and designation of Mycobacterium abscessus subsp. massiliense comb. nov. Int. J. Syst. Evol. Microbiol. 66, 4471–4479. doi: 10.1099/ijsem.0.001376


**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 Oliveira, Da Costa, Procopio, Garcia, Araújo, Da Silva, Junqueira-Kipnis and Kipnis. 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.

# Evaluation of a Novel MALDI Biotyper Algorithm to Distinguish Mycobacterium intracellulare From Mycobacterium chimaera

L. Elaine Epperson<sup>1</sup> , Markus Timke<sup>2</sup> , Nabeeh A. Hasan<sup>1</sup> , Paul Godo<sup>3</sup> , David Durbin<sup>3</sup> , Niels K. Helstrom<sup>3</sup> , Gongyi Shi<sup>4</sup> , Markus Kostrzewa<sup>2</sup> , Michael Strong<sup>1</sup> and Max Salfinger3,5,6 \*

<sup>1</sup> Center for Genes, Environment and Health, National Jewish Health, Denver, CO, United States, <sup>2</sup> Bruker Daltonik GmbH, Bremen, Germany, <sup>3</sup> Mycobacteriology Laboratory, National Jewish Health, Denver, CO, United States, <sup>4</sup> Bruker Daltonics, Billerica, MA, United States, <sup>5</sup> Department of Medicine, National Jewish Health, Denver, CO, United States, <sup>6</sup> College of Public Health, University of South Florida, Tampa, FL, United States

#### Edited by:

Thomas Dick, Rutgers, The State University of New Jersey, United States

#### Reviewed by:

Jeanette Teo, National University Hospital, Singapore Maurizio Sanguinetti, Catholic University of the Sacred Heart, Italy

> \*Correspondence: Max Salfinger salfingerm@njhealth.org

#### Specialty section:

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

Received: 31 August 2018 Accepted: 04 December 2018 Published: 18 December 2018

#### Citation:

Epperson LE, Timke M, Hasan NA, Godo P, Durbin D, Helstrom NK, Shi G, Kostrzewa M, Strong M and Salfinger M (2018) Evaluation of a Novel MALDI Biotyper Algorithm to Distinguish Mycobacterium intracellulare From Mycobacterium chimaera. Front. Microbiol. 9:3140. doi: 10.3389/fmicb.2018.03140 Accurate and timely mycobacterial species identification is imperative for successful diagnosis, treatment, and management of disease caused by nontuberculous mycobacteria (NTM). The current most widely utilized method for NTM species identification is Sanger sequencing of one or more genomic loci, followed by BLAST sequence analysis. MALDI-TOF MS offers a less expensive and increasingly accurate alternative to sequencing, but the commercially available assays used in clinical mycobacteriology cannot differentiate between Mycobacterium intracellulare and Mycobacterium chimaera, two closely related potentially pathogenic species of NTM that are members of the Mycobacterium avium complex (MAC). Because this differentiation of MAC species is challenging in a diagnostic setting, Bruker has developed an improved spectral interpretation algorithm to differentiate M. chimaera and M. intracellulare based on differential spectral peak signatures. Here, we utilize a set of 185 MAC isolates that have been characterized using rpoB locus sequencing followed by whole genome sequencing in some cases, to test the accuracy of the Bruker subtyper software to identify M. chimaera (n = 49) and M. intracellulare (n = 55). 100% of the M. intracellulare and 82% of the M. chimaera isolates were accurately identified using the MALDI Biotyper algorithm. This subtyper module is available with the MALDI Biotyper Compass software and offers a promising mechanism for rapid and inexpensive species determination for M. chimaera and M. intracellulare.

Keywords: nontuberculous mycobacteria, diagnostics, MALDI, Mycobacterium chimaera, Mycobacterium intracellulare, species identification, MAC, NTM

### INTRODUCTION

The use of Matrix Assisted Laser Desorption Ionization-Time of Flight Mass Spectrometry (MALDI-TOF MS) for identification of microbial specimens has scaled up dramatically over the past decade due to the continual improvement of available tools, including expanded spectral peak databases [reviewed in Doern and Butler-Wu (2016); Rahi et al. (2016); and Alcaide et al. (2018)].

Species identification of bacteria is now honed to strain-specific and subspecies distinctions in many cases (Neville et al., 2011; Huang et al., 2018) as well as the characterization of nonbacterial organisms such as fungi (Chalupova et al., 2014; Levesque et al., 2015). These approaches are relatively rapid, inexpensive, and accurate for use in clinical diagnostics (Neville et al., 2011), environmental sample analysis, food safety, and other applications (Carbonnelle et al., 2011; Florio et al., 2018). MALDI-TOF MS is increasingly utilized to distinguish clinically relevant nontuberculous mycobacteria (NTM) (Levesque et al., 2015; Costa-Alcalde et al., 2018).

The Mycobacterium avium complex (MAC) comprises several clinically important mycobacterial species including M. avium (subsp. hominissuis), M. intracellulare, and M. chimaera (Diel et al., 2018; Forbes et al., 2018), among others [e.g., M. colombiense (Murcia et al., 2006) and M. yongonense (Castejon et al., 2018)]. Although M. avium and M. intracellulare species are more frequently observed in clinical cases, a recent series of M. chimaera infections originating in cardiac surgery suites (van Ingen et al., 2017) illuminates the importance of identifying members of the MAC complex at the species level in a rapid and accurate manner. While the spectrum of M. avium is relatively distinct from the other two species, M. intracellulare and M. chimaera are highly similar to each other (van Ingen et al., 2012), and not readily distinguished using the standard approaches of most MALDI-TOF peak interpretation algorithms (Boyle et al., 2015a; Leyer et al., 2017). Pulmonary infections caused by the different MAC species are distinct in their virulence and clinical features, indicating that species-level identification is important for effective prognostics (Boyle et al., 2015b; Kim et al., 2017). Bruker Daltonik (Bremen, Germany) developed a commercially available algorithm to differentiate M. chimaera and M. intracellulare from each other using only MALDI-TOF peak data. This algorithm was found to perform well against 59 bacterial isolates of European origin using the internal transcribed spacer (ITS) sequence to identify species (Pranada et al., 2017). Here we set out to evaluate the same MALDI algorithm against 111 isolates of United States origin. Sequence from the rpoB locus was used to determine species, but in 16 discrepant cases, the isolates were whole genome sequenced and species identity was determined using phylogenomics.

### MATERIALS AND METHODS

#### NTM Species Determination Using Sanger Sequencing

All samples were of clinical origin within the United States, either bronchoalveolar lavage, sputum, or tissue. Isolates were sent to National Jewish Health for NTM species identification and/or antimicrobial susceptibility testing. DNA from these isolated cultures of 300 acid-fast bacilli were analyzed using a targeted 711 base pair region of the rpoB locus that was Sanger sequenced using ABI 3730xL and queried against the National Center for Biotechnology Information (NCBI) Basic Local Alignment Search Tool (BLAST) database. The rpoB target is one of several recommended for NTM species determination according to current national clinical microbiology guidelines (Forbes et al., 2018); other targets that are utilized for this purpose are 16S rRNA, secA, ITS, and hsp65 (Lecorche et al., 2018). From this collection, a subset (185) were identified as one of the MAC species according to rpoB sequence, i.e., M. avium (n = 74), M. intracellulare (n = 55), or M. chimaera (n = 56).

### MALDI-TOF MS

MAC strains were subcultured onto Middlebrook 7H11 agar media and incubated for 7–10 days to obtain sufficient biomass for MALDI-TOF MS identification. Colonies from the 7H11 plates were harvested and heat-killed in high performance liquid chromatography (HPLC) grade water, followed by acetonitrile/formic acid extraction procedure according to the manufacturer protocol. The MALDI-TOF target was spotted with 1 µL of extract supernatant and overlaid with 1 µL of α-Cyano-4-hydroxycinnamic acid (HCCA) matrix prior to data capture using the Bruker MALDI-TOF Biotyper system (software v4.0). A minimum score of 1.8 was required. A complete MALDI spectrum dataset was sent to Germany for MALDI spectrum analysis in parallel at Bruker Daltonik in Bremen, Germany.

For samples identified by MALDI-TOF MS as M. chimaeraintracellulare group, a second-tier analysis was performed using a recently developed research subtyping software from Bruker Daltonik [(Pranada et al., 2017) herein referred to as MBT Subtyping Module]. One sample, NTM-184, was eliminated from the sensitivity and specificity calculations because the MALDI results placed it in the M. chimaera-intracellulare group, but the rpoB result was M. avium, and this sample did not have whole genome sequence (WGS) information (see below.) An additional seven samples, six of which were identified as M. chimaera using rpoB, were also eliminated from these calculations because the MBT Subtyping Module did not yield a result. These strains are listed in **Table 1** under the MBT Subtyper as "M. chim/M. int." It should be noted that the MBT Subtyping Module returned a high score on these seven samples but because it was unable to distinguish the two species, it does not make a call in these cases.

### Phylogenomic Analysis of Ambiguous NTM Strains

For almost all samples, the rpoB species identification was considered to be the actual species of that strain. In a few cases, samples yielded discrepant results, and for these samples, WGS was used to assign a species identification to a given mycobacterial strain. Genomic DNA was isolated according to a protocol adapted from Käser et al. (2010), employing a column DNA clean in lieu of a phenol chloroform extraction and alcohol precipitation. Genomic libraries were constructed using Nextera XT and sequenced using Illumina chemistry (Illumina, Inc., San Diego, CA).

Illumina reads were trimmed of adapters and low quality bases (< Q20) using Skewer (Jiang et al., 2014). Trimmed reads were assembled into scaffolds using Unicycler (Wick et al., 2017), and genome assemblies were compared against a selection of reference genomes to calculate average nucleotide identities (ANI, chjp/ANI on GitHub.com, **Supplementary Table S1**)

and to assign a species call to each isolate (Goris et al., 2007; Richter and Rossello-Mora, 2009). Trimmed reads and background reference genomes were mapped to the M. chimaera CDC 2015-22-71 genome (Hasan et al., 2017) using Bowtie2 software (Langmead and Salzberg, 2012). Single nucleotide polymorphisms (SNPs) were called using SAMtools mpileup program (Li et al., 2009), and a multi-fasta sequence alignment was created from concatenated basecalls from all strains (Davidson et al., 2014; Page et al., 2016). Resulting sequences were used to make a maximum likelihood (ML) phylogenetic tree using RAxML-NG with a GTR+G substitution model (Kozlov et al., 2018) and 1,000 bootstrap replicates. The tree was annotated and visualized with ggtree (Yu et al., 2017). Reference genomes for the tree are, for M. chimaera, listing genome name, accession number: AH16, CP012885.2; CDC 2015-22- 71,CP019221.1; DSM 44623, CP015278.1; SJ42, CP022223.1; ZUERICH-2, CP015267.1; LJHL01, LJHL00000000.1; LJHM01, LJHM00000000.1; LJHN01, LJHN00000000.1; 1956, CP009499.1; ATCC 13950, NC\_016946.1; MOTT-02, NC\_016947.1; MOTT-64, NC\_016948.1; and for M. intracellulare: 1956, CP009499.1; ATCC 13950, NC\_016946.1; MOTT-02, NC\_016947.1; MOTT-64, NC\_016948.1.

The whole genome sequence data are available at NCBI in Bioproject number PRJNA506060<sup>1</sup> with the following accession numbers: NTM-006, SAMN10441865; NTM-019, SAMN10 441945; NTM-035, SAMN10441946; NTM-054, SAMN1044 1947; NTM-105, SAMN10441948; NTM-107, SAMN10441949; NTM-168, SAMN10441950; NTM-178, SAMN10441951; NTM -203, SAMN10441952; NTM-204, SAMN10441953; NTM-206,

<sup>1</sup>https://www.ncbi.nlm.nih.gov/bioproject/PRJNA506060/

SAMN10441954; NTM-208, SAMN10441955; NTM-223, SAMN 10441956; NTM-224, SAMN10441957; NTM-230, SAMN10 441958; NTM-232, SAMN10441959.

#### Comparison of WGS to Single Locus Species Determination

Using primer sequences of known targets, three regions were extracted from the WGS data. These were previously targeted regions from the following loci: 16S rDNA (Springer et al., 1996; van Ingen et al., 2012), the 16S–23S internal transcribed spacer (ITS) (Schweickert et al., 2008), and rpoB (Adekambi et al., 2003; Ben Salah et al., 2008). These sequences were used to BLAST against the non-redundant NCBI database and the top species call or calls are reported in **Table 1**.

### RESULTS

From the initial strain collection, 185 isolates were identified by rpoB amplicon sequencing as one of the three prominent clinical MAC species, specifically 74 M. avium, 55 M. intracellulare, and 56 M. chimaera. These 185 strains were then analyzed using MALDI-TOF MS at National Jewish Health, in Denver, Colorado, United States and the data were re-analyzed in Bremen, Germany, yielding identical results according to the Biotyper 4.0 software. Of the 74 M. avium isolates, 73 were identified as M. avium. A single M. avium sample (as identified by rpoB) and the remaining 111 samples were all identified as being in the M. chimaera-intracellulare group.

These 112 strains were then subjected to a second tier of analysis wherein their MALDI-TOF MS spectra were evaluated

TABLE 1 | Comparison of single locus sequencing results to whole genome sequencing results for sixteen MAC strains for each sample that was whole genome sequenced, three loci were extracted and analyzed by BLAST to the non-redundant NCBI database giving the results listed.


Also listed are the species as determined using only the 403 SNP from 16S, assuming a strain was previously identified as belonging to the chimaera/intracellulare group. MBT Subtyper result and the WGS species call based on phylogenomic tree. M. chim = M. chimaera, M. int = M. intracellulare, M. yong = M. yongonense, M. mars = M. marseillense.

using the recently developed MBT Subtyping Module software; see Methods (Pranada et al., 2017). One sample, NTM-006, was identified as M. chimaera by rpoB but as M. intracellulare by WGS, thus the sample was categorized as M. intracellulare. This was the only sample for which WGS yielded a result that differed from rpoB. In this case, the WGS result was considered to be the actual species for calculations. After removing the eight samples that were unconfirmed (see Methods), the samples numbered 55 M. intracellulare and 49 M. chimaera that were suitable to evaluate the MBT Subtyping Module algorithm for its ability to distinguish these two species.

Sixteen isolates were selected for WGS. These strains were investigated in greater depth using extracted sequences of specific loci (**Table 1**) in addition to phylogenomic tree analysis using the WGS data (**Figure 1**). The specific loci were selected to reflect the regions that are most commonly used for mycobacterial species identification. For most Mycobacteria spp., these loci are useful for identifying to the NTM species level. But the close relationship of M. chimaera and M. intracellulare is apparent in these genes since the two species' sequences over these regions are largely interchangeable, consistent with the initial confounding observations that led to the naming of M. chimaera (Tortoli et al., 2004). Based on these extracted sequences, the rpoB fragment was most reflective of the WGS result (**Table 1**).

For the 49 samples identified by sequencing as M. chimaera, the MBT Subtyping Module identified 40 as M. chimaera and 9 as M. intracellulare for a sensitivity of 81.6%. However, because all of the Subtyper M. chimaera calls were in fact M. chimaera according to whole genome sequencing, the positive predictive value for a call of M. chimaera is 100%. The M. intracellulare (n = 55) data differed in that all 55 samples were identified as M. intracellulare for a sensitivity of 100%, but nine of the M. chimaera samples were called M. intracellulare by the Subtyping software, reducing the specificity, and the positive predictive value is 86% or 55/(55+9). This indicates that a minority (∼18%, or 9/49) of the true M. chimaera samples would be recognized as M. intracellulare using the MALDI Biotyper Compass software.

### DISCUSSION

The recent outbreak of M. chimaera-contaminated heater-cooler devices used in cardiac surgery has increased the clinical urgency for differentiating M. chimaera from M. intracellulare. The MALDI-TOF MS approach to species identification is less expensive and more rapid than sequencing methods, but until now has been unable to distinguish these two closely related MAC species. Our findings indicate that a call of M. chimaera by the Bruker subtyping algorithm tested here is correct every time for a positive predictive value of 100%. We also found that a call of M. intracellulare is correct 86% of the time, that is, 14% of our samples that were recognized as M. intracellulare by MALDI are actually more closely related to M. chimaera than to M. intracellulare. Given the regional diversity of NTM species, it is possible that these strains were not available in

the initial European sample set during test design. A number of these strains cluster more closely with the M. chimaera reference strains (samples 178, 019, 232, etc.), although they are arguably relatively distant from any available reference. The dynamic nature of bacterial phylogenomics requires us to integrate currently available references and whole genome approaches with an understanding that species nomenclature is always evolving as new branches and clades become apparent.

With increasing numbers of M. chimaera and M. intracellulare isolates requiring species identification in diagnostic reference laboratories, it is clear that improved tools are needed to quickly and accurately identify and distinguish these close mycobacterial species, in most cases inseparable using average nucleotide identity measures (**Supplementary Table S1**). We observe a qualitative difference in genome size for these two species and performed a statistical comparison of available complete reference genomes from M. chimaera (mean ± 1s.d., 6.24 ± 0.29Mb, n = 8) and M. intracellulare (5.52 ± 0.12Mb, n = 6), and found that M. chimaera was consistently larger by ∼700,000 bases (p < 0.001, Mann-Whitney U test.) This information may provide a path to discovering markers that can distinguish the two species, but the standard and reliable tools used to distinguish most NTM from one another often fail to discriminate between these very closely related strains. The discerning capability of the Bruker Subtyper module with MALDI spectral data

#### REFERENCES


offers an available and notable advance in MAC species identification.

#### AUTHOR CONTRIBUTIONS

MT, MK, GS, and MaS designed the study. LEE wrote the manuscript and performed whole genome sequencing. NKH, PG, and DD performed rpoB sequencing and MALDI-TOF analysis. MT, MK, and GS performed parallel MALDI-TOF spectral analysis and developed and implemented the subtyping software. NAH analyzed whole genome sequencing data, extracted loci of interest, and created visualizations. All authors contributed to writing, editing, and overall presentation of the manuscript.

#### ACKNOWLEDGMENTS

We thank Sara Kammlade and Rebecca Davidson for providing genomic ANIs. LEE, NAH, MiS, and MaS thank the Cystic Fibrosis Foundation, NICK15R0.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2018.03140/full#supplementary-material

Biotechnol. Adv. 32, 230–241. doi: 10.1016/j.biotechadv.2013. 11.002


whole-genome sequence similarities. Int. J. Syst. Evol. Microbiol. 57, 81–91. doi: 10.1099/ijs.0.64483-0


**Conflict of Interest Statement:** MT and MK are employed by Bruker Daltoniks and GS by Bruker Daltonics.

The remaining 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 Epperson, Timke, Hasan, Godo, Durbin, Helstrom, Shi, Kostrzewa, Strong and Salfinger. 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.

# MALDI Spectra Database for Rapid Discrimination and Subtyping of Mycobacterium kansasii

Jayaseelan Murugaiyan<sup>1</sup> \*, Astrid Lewin<sup>2</sup> , Elisabeth Kamal <sup>2</sup> , Zofia Bakuła<sup>3</sup> , Jakko van Ingen<sup>4</sup> , Vit Ulmann<sup>5</sup> , Miren J. Unzaga Barañano<sup>6</sup> , Joanna Humie˛cka<sup>7</sup> , Aleksandra Safianowska<sup>8</sup> , Uwe H. Roesler <sup>1</sup> and Tomasz Jagielski <sup>3</sup> \*

<sup>1</sup> Centre for Infectious Medicine, Institute of Animal Hygiene and Environmental Health, Freie Universität Berlin, Berlin, Germany, <sup>2</sup> Division 16, Mycotic and Parasitic Agents and Mycobacteria, Robert Koch Institute, Berlin, Germany, <sup>3</sup> Department of Applied Microbiology, Faculty of Biology, Institute of Microbiology, University of Warsaw, Warsaw, Poland, <sup>4</sup> Department of Medical Microbiology, Radboud University Medical Center, Nijmegen, Netherlands, <sup>5</sup> Institute of Public Health, Ostrava, Czechia, <sup>6</sup> Clinical Microbiology and Infection Control, Basurto Hospital, Bilbao, Spain, <sup>7</sup> Hospital for Infectious Diseases in Warsaw, Medical University of Warsaw, Warsaw, Poland, <sup>8</sup> Department of Internal Medicine, Pulmonology, and Allergology, Medical University of Warsaw, Warsaw, Poland

#### Edited by:

Veronique Anne Dartois, Rutgers University–Newark, United States

#### Reviewed by:

An Van Den Bossche, Institut Scientifique de Santé Publique, Belgium Haiqing Chu, Shanghai Pulmonary Hospital, China

#### \*Correspondence:

Jayaseelan Murugaiyan jayaseelan.murugaiyan@fu-berlin.de Tomasz Jagielski t.jagielski@biol.uw.edu.pl

#### Specialty section:

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

Received: 10 January 2018 Accepted: 14 March 2018 Published: 03 April 2018

#### Citation:

Murugaiyan J, Lewin A, Kamal E, Bakuła Z, van Ingen J, Ulmann V, Unzaga Barañano MJ, Humie˛cka J, Safianowska A, Roesler UH and Jagielski T (2018) MALDI Spectra Database for Rapid Discrimination and Subtyping of Mycobacterium kansasii. Front. Microbiol. 9:587. doi: 10.3389/fmicb.2018.00587 Mycobacterium kansasii is an emerging non-tuberculous mycobacterial (NTM) pathogen capable of causing severe lung disease. Of the seven currently recognized M. kansasii genotypes (I-VII), genotypes I and II are most prevalent and have been associated with human disease, whereas the other five (III-VII) genotypes are predominantly of environmental origin and are believed to be non-pathogenic. Subtyping of M. kansasii serves as a valuable tool to guide clinicians in pursuing diagnosis and to initiate the proper timely treatment. Most of the previous rapid diagnostic tests for mycobacteria employing the matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS) technology focused on species-level identification. The purpose of this study was to establish MALDI-TOF MS reference spectra database for discrimination of M. kansasii at the genotype level. A panel of 32 strains, representatives of M. kansasii genotypes I-VI were selected, whole cell proteins extracted and measured with MALDI-TOF MS. A unique main spectra (MSP) library was created using MALDI Biotyper Compass Explorer software. The spectra reproducibility was assessed by computing composite correlation index and MSPs cross-matching. One hundred clinical M. kansasii isolates used for testing of the database resulted in 90% identification at genus-level, 7% identification at species-level and 2% identification was below the threshold of log score value 1.7, of which all were correct at genotype level. One strain could not be identified. On the other hand, 37% of strains were identified at species level, 40% at genus level and 23% was not identified with the manufacturer's database. The MALDI-TOF MS was proven a rapid and robust tool to detect and differentiate between M. kansasii genotypes. It is concluded that MALDI-TOF MS has a potential to be incorporated into the routine diagnostic workflow of M. kansasii and possibly other NTM species.

Keywords: Biotyper, matrix-assisted laser desorption ionization—time of flight mass spectrometry, microflex, Mycobacterium kansasii, genotypes

## INTRODUCTION

Mycobacterium kansasii is one of the leading causative agents of pulmonary and extrapulmonary infections due to nontuberculous mycobacteria (NTM). It is also among the top six most frequently isolated NTM species across the world. The isolation rate of this pathogen is exceptionally high in Slovakia, Poland, and the UK, that is 36, 35, and 11% respectively, compared to a mean isolation rate of 5% in Europe and 4% globally (Hoefsloot et al., 2013).

The major clinical manifestations attributable to M. kansasii include fibro-cavitary or nodular-bronchiectatic lung disease, lymphadenitis, skin, and soft-tissue infections, and disseminated disease (van Ingen and Kuijper, 2014).

Pulmonary diseases caused by M. kansasii tend to cluster in specific geographical areas, such as central Europe or metropolitan centers of London, Brasilia, and Johannesburg (Hoefsloot et al., 2013). In most of the countries there is no obligation of registration of such cases, thus the true incidence of M. kansasii disease is largely unknown. The estimated incidence of infections due to this pathogen, reported in the general population, fall within the range of 0.3–1.5 cases per 100,000 (Santin et al., 2004; Moore et al., 2010; Namkoong et al., 2016).

As with other NTM, M. kansasii infections are believed to be acquired from environmental exposures rather than by personto-person transmission. Rarely has M. kansasii been isolated from soil, natural water systems or animals. Instead, the pathogen has commonly been recovered from municipal tap water, which is considered to be its major environmental reservoir (Thomson et al., 2014).

The existence of five (I–V) distinct M. kansasii types was initially evidenced by unique DNA profiles they produced upon hybridization with the major polymorphic tandem repeat (MPTR) probe, pulsed-field gel electrophoresis (PFGE), amplified fragment length polymorphism (AFLP) analysis, and PCR restriction enzyme analysis (PCR-REA) of the hsp65 gene (Alcaide et al., 1997; Picardeau et al., 1997). In somewhat later studies two novel types (VI and VII) have been described, with similar methods (Taillard et al., 2003). This intra-species polymorphism has important clinical and epidemiological implications. Genotypes I and II have been involved in human disease, with the former being the most prevalent, whereas the remaining genotypes have usually been non-pathogenic. Therefore, recovery of M. kansasii type I and II isolates from clinical specimens should raise a clinical suspicion of M. kansasii disease. Furthermore, M. kansasii type I has been found more likely to be recovered from immunocompetent individuals, whereas type II has been linked to patients with some form of immunosuppression (Taillard et al., 2003). Furthermore, despite a paucity of studies, there seems to be some genotype-specific differences in drug susceptibility profiles. For instance, clarithromycin resistance appears to be associated with M. kansasii type I (Li et al., 2016), while M. kansasii type V, contrary to all other types, shows susceptibility to ethambutol (Bakuła et al., 2018). Currently, identification of M. kansasii genotypes is performed by PCR-sequencing or PCR-REA of the hsp65, rpoB, and tuf genes (Telenti et al., 1993; Kim et al., 2001; Bakuła et al., 2016). These methods, though accurate and reproducible, require post-PCR manipulations, which are time-consuming and laborious. A fast, reliable and automated screening method for M. kansasii genotype determination is needed.

In the past two decades, MALDI-TOF MS-based species identification has been integrated into the routine diagnostic workflow in microbiology laboratories due to its rapidness, reliability, cost-effectiveness, and high throughput. A prerequisite for MALDI-TOF MS profiling is a robustly established reference spectra database. The growing accessibility of MALDI-TOF MS instruments along with spectra pattern matching software databases made that the interest in this technology and its applications across research and diagnosis has exponentially increased (Wattal et al., 2017).

In this study, a commercially available software package, MALDI Biotyper (Bruker Daltonics) was used for creation of genotype-specific reference database of M. kansasii.

#### MATERIALS AND METHODS

#### Mycobacteria Strains and Isolates

A total of 32 strains, representing I-VI genotypes of Mycobacterium kansasii were used to establish a MALDI-TOF MS spectra reference database (**Table 1**). In addition, 100 clinical M. kansasii isolates were used for evaluation of the database. The species identity of these isolates was confirmed with the high-pressure liquid chromatography (HPLC). Genomic DNA was extracted with the AMPLICOR Respiratory Specimen Preparation Kit (Roche, Basel, Switzerland). Genotyping was performed upon PCR-REA of hsp65 and tuf genes, as previously described (Telenti et al., 1993; Bakuła et al., 2016).

#### Culture and Protein Extraction

A single bacterial colony was picked from Löwenstein-Jensen (L-J) slant and used for inoculation of 10 mL of Middlebrook 7H9 medium supplemented with glycerol and OADC (10%) (Becton-Dickinson, Franklin Lakes, USA). The bacteria were cultured at 37◦C with shaking for 7 days. Each strain/isolate was cultured independently for three independent replicates. The cells were harvested by brief centrifugation (4,500 rpm, 2 min) and killed by suspending in 300µL of deionized water and 900µL of absolute ethanol. After centrifugation (12,000 rpm, 2 min) at room temperature (RT), excessive ethanol was discarded and the samples were air-dried completely. The cell pellet was dissolved in 50µL of 70% formic acid and 50µL acetonitril. The samples were then subjected to sonication on ice for 1 min (cycle, 1.0; amplitude, 100%) using a sonicator (UP100H, Hielscher Ultrasound Technology, Teltow, Germany). The suspension was centrifuged at 12,000 rpm for 2 min at RT, and 1.0µL of the clear supernatant was spotted in triplicate onto the MALDI target (MSP 96 target polished steel (MicroScout Target) plate, Bruker Daltonik, Bremen, Germany). Following

**Abbreviations:** MALDI-TOF MS, Matrix-assisted laser desorption ionizationtime-of-flight mass spectrometry; MSP, main spectra libraries.


<sup>T</sup> Type strain, <sup>E</sup> Environmental isolate.

air drying each sample was overlaid with 1.0µL of saturated αcyano-4-hydroxycinnamic acid matrix solution and allowed to dry completely prior to MALDI-TOF measurement.

#### MALDI Measurements

Bruker MALDI Microflex LT (Bruker Daltonics, Bremen, Germany) with a linear positive mode of spectra acquisition, at a laser frequency of 20 Hz was used for the measurements. A broad m/z range (2,000–20,000 kDa) of spectra was automatically acquired using the AutoXecute acquisition control software (Flex control 3.0, Bruker Daltonics, Leipzig, Germany) with the following instrument settings: ion source 1 at 20 kV, ion source 2 at 16.7 kV, lens at 6 kV, extraction delay time of 150 ns, 240 laser shots/spot in 40 shot steps (random walk movement), initial laser power of 45%, maximal laser power of 55%, laser attenuation Offset of 46% and at the range 30%). The bacterial test standard (BTS) mixture covering the mass range between 2,000 and 20,000 Da (BTS, Bruker Daltonics, Bremen, Germany), was used for external calibration.

#### MSP Library Construction

The spectra quality and reproducibility were assessed by using Flex analysis 3.0 software (Bruker Daltonics, Bremen, Germany). For every strain/isolate, 27 spectra representing three independent culturing and three technical replicates and 3 measurements per replicate were collected. The raw spectra were loaded in the MALDI Biotyper Compass Explorer 4.1 and matched against the commercial BDAL database—updated version 6.0.0.0 with 6,903 Main Spectra (MSP) entries (including 75 for mycobacteria with seven genotype undefined M. kansasii entries)—plus the in-house entries of Staphylococcus intermedius group of microorganisms (60 entries) (Murugaiyan et al., 2014) and microalgae Prototheca species (23 entries) (Murugaiyan et al., 2012). The spectra preprocessing parameters included mass adjustment (lower bound- 3,000, upper bound- 15,000; resolution-1 and spectra compressing factor- 10), smoothing (Savitzky-Golay algorithm with a frame size of 25 Da), baseline correction (multipolygon with search window 5 Da and number of runs 2), normalization (maximum norm), and peak detection (spectra differentiation with maximum peaks of 100, signal-to-noise ratio 3 and threshold of 0.001). Main spectrum profiles (MSPs) were created for each strain using the following parameters: maximum mass error for each single spectrum 2,000, mass error for the MSP 100, peak frequency minimum 70%, and maximum peak number 70. The details of the newly created reference spectra will be listed at the MALDI-UP, an online platform for the MALDI-TOF mass spectra (Rau et al., 2016) to facilitate exchange of the reference spectra as btmsp files with users from other laboratories.

#### Statistical Evaluations

To evaluate the spectral variation (similarity) between the spectra sets acquired from strains of all genotypes, the Composite Correlation Index (CCI) was calculated with MALDI Biotyper Compass Explorer 4.1. All measured spectra were loaded and CCI was computed using the following settings, mass lower bound 3,000, mass upper bound 12,000, resolution 4, and CCI parameter interval of 8. CCI value nearing 1.0 indicates the complete congruence and 0.0 – complete deviation within the spectra sets.

Following the creation of spectral reference database, the genotype discrimination limits were assessed by crosswise comparison by matching the newly created M. kansasii genotypespecific MSPs with an augmented BDAL commercial database described earlier. The resulted log score values of the first hits were imported into MS excel and conditional formatting, in accordance to the manufacturer's recommended cutoff log score values, was applied to generate a heat-map. The scoreoriented MSP dendrogram (distance level: correlation and linkage: average) was also calculated using the MALDI Biotyper Compass Explorer 4.1 software to explore the closeness of the genotypes in terms of arbitrary distance level.

### Blind Coded Sample Testing

For validation purposes, 100 clinical M. kansasii isolates and single reference strains of M. conspicuum, M. marinum, M. szulgai, and M. gastri were subjected to MALDI-TOF MS identification, by comparing with the database before and after addition of genotype-specific MSPs (**Supplementary Table 1**). The isolates used for validation, had previously been identified, as specified above and blind coded. The log score values for species discrimination, recommended by the manufacturer, were applied i.e., 0–1.699 indicating "no reliable identification", 1.7–1.999 indicating a "probable genus identification", 2.0– 2.299 indicating a "secure genus identification and probable species identification" (PSI), and 2.300–3.000 indicating a "highly probable species identification."

### RESULTS

A total of 32 strains representing genotypes I-VI of M. kansasii were used to establish M. kansasii genotype-specific reference spectra library. The choice of the strains was based on the availability of strains representing a given genotype and strains of genotype VII were not available in any of the culture collections. Each strain was cultured independently three times, each culturederived extract was spotted three times, and each spot was measured three times on MALDI-TOF MS to generate 27 spectra in total per strain. Altogether, 864 spectra within the mass range of 2,000–20,000 Da were acquired. The spectra quality in terms of peak presence and intensity were assessed using Flex analysis software. The flat line spectra and low-intensity spectra were replaced by new measurement of the same spots or fresh spotting of the same protein extract.

As shown in **Figure 1**, different genotypes displayed distinct peak patterns and the spectral sets of each strain were fully reproducible and comparable.

The spectra of the same strains showed a high level of similarity, with the CCI value ranging from 0.78 to 0.94 (**Figure 2**). Genotype I displayed CCI within the range of 0.78 to 0.99 while it was between 0.82 and 0.91 for genotype II, 0.80 to 0.94 for genotype III, 0.83 for a single genotype IV strain, 0.83 to 0.94 for genotype V, and 0.88 to 0.91 for genotype VI. The resulted CCI values reflected the similarities of the measured spectra sets. The CCIs between different genotypes displayed lower values indicating the possibility for spectra fingerprinting of genotype level discrimination.

The created M. kansasii genotype MSPs were then matched to the augmented BDAL database described earlier to generate the cross-matching result. Then, the first hits log score values were imported in MS excel and a heat map was generated using MS excel conditional color formatting option following the manufacturer's cut-off values (**Supplementary Figure 1**). The MSPs of any given strain matched 100% with themselves, as indicated by the log score of 3.0, and had high matching scores with the other strains within a genotype. The seven strains without any genotype information included in the original version of the manufacturer's database, displayed higher score values among each other, although the MSPs had been configured upon a much lower number of spectra. The manufacturer's MSPs displayed log score values of >2.0 with several strains of newly added genotype I and borderline or genus-level matching was observed with genotype III, IV, and V strains. One of the manufacturer's MSP (S.No. 37) did not match any of the newly introduced genotype-specific MSPs.

In addition, the individual spectra used for MSP creation were matched first with the manufacturer's database (BDAL databaseplus in-house entries of Staphylococcus intermedius group and microalgae Prototheca species) and then after introduction of our genotype-specific MSPs. This enhanced considerably the overall identification confidence and identification at the genotype level with first hits always being the same strain (**Figure 3**).

view of spectra acquired from a single strain (genotype I strain 1010001495). (C) Stacked view indicating the uniformity among spectra acquired from a genotype I strain 1010001495 and highlighted overlaid view as an inserted figure.

As shown in **Figure 4**, a score-oriented dendrogram demonstrated two major clusters. The first cluster contained all genotype I strains and two genotype VI strains. The second cluster accommodated all type II, III, IV, and V strains.

The augmented genotype-specific reference database was tested with 104 blind-coded samples, representing clinical isolates of M. kansasii (n = 100) and single strains of M. conspicuum, M. gastri, M. marimum, and M. szulgai, to find any cross-matching with M. kansasii database. The identification rates were significantly higher for the expanded MALDI Biotyper database when compared with the original library (**Table 2**). In **Figure 5**, a shift of identification confidence is shown for all 100 blind-coded M. kansasii samples (**Supplementary Table 1**). The augmented database resulted in an overall identification success rate of 97% isolates identified at the species/genotype level, among which 7 isolates, representatives of genotype I, were identified at the genus level. Two genotype I isolates could not be identified despite the first hit as the correct genotype but with log score value <1.7. One isolate (M. kansasii 17.14) did not display any improvement in the log score value despite the database augmentation. The results also indicated that, 89 strains belonged to genotype I (19-SI, 63-PSI, and 7-PGI), 4 – to genotype II, and 2 – to genotype III. Single isolates were recognized as representing genotypes IV and V. Whereas the MALDI Biotyper using the manufacture's BDAL database displayed "Not reliable identification" for 15% isolates of genotype I and six other genotype isolates. Two genotype II isolates were identified only as Mycobacterium sp.

## DISCUSSION

MALDI-TOF MS profiling has been successfully employed in various commercial software packages allowing the identification of a wide array of Mycobacterium species, including M. kansasii. The discrimination of different M. kansasii genotypes is useful for the clinical categorization of M. kansasii infection, however, it has never been attempted by MALDI-TOF analysis (Pignone et al., 2006; El Khechine et al., 2011; Saleeb et al., 2011; Balada-Llasat et al., 2013; Chen et al., 2013; Machen et al., 2013; Mather et al., 2014; Mediavilla-Gradolph et al., 2015; Rodriguez-Sanchez et al., 2015; Wilen et al., 2015; Ceyssens et al., 2017; Leyer et al., 2017). M. kansasii genotype level information is still not included in the commercial reference database package, MBT Mycobacteria library. Previously, identification of M. kansasii with low spectral scores was linked to the intraspecies heterogeneity, and supplementation of the database with additional strains of M. kansasii was proposed to remedy this problem (El Khechine et al., 2011; Saleeb et al., 2011). Responding to this necessity, this study established MSPs for all M. kansasii genotypes (I–VI), except for the genotype VII (Taillard et al., 2003), for which strains could not be retrieved in any of the culture collection, which could not be retrieved from any culture

collections. It has to be noted, however, that given the paucity of genotype II-VI strains, differentiation of these types may still pose certain difficulties, and much more strains would be needed to verify the genotype-specificity of the MALDI-TOF profiles established.

Inspection with Flex analysis software, the spectral profiles of the M. kansasii genotypes clearly showed their disparateness. At the same time, a high intra-genotype reproducibility of the spectra was demonstrated. The spectra sets acquired from the same strains and from strains representing the same genotypes were highly consistent, as evidenced with the computed CCI values.

The cross identification of the established MSPs together with seven MSPs originally deposited in the manufacturer's database resulted in a log score value of 3.0 indicating matching of the 70 most frequently encountered peaks of the species or genotypes. When compared against the manufacturer's database, with exception of genotypes I and II, whose spectra matched those of M. kansasii, all other genotypes produced spectra ascribable to merely Mycobacterium genus level. It clearly proved that the existing database is not sufficient for genotype-level discrimination of M. kansasii. Special mycobacteria library with lowered threshold log score values of 1.6 and 1.8 was implemented for genus and species identification. It was later reported that the log score values for M. kansasii were unchanged with the improved mycobacteria library (Rodríguez-Sánchez et al., 2016). In the present study, augmentation of genotype-specific MSPs resulted in an enhanced log score value, which indicates that the identification confidence improvement can be mainly achieved through inclusion of genotypically well-defined strains.

The grouping, as shown on score-oriented dendrogram, of genotypes I, II, and VI in three distinct clusters supports the potential of MALDI-TOF to easily differentiate between these three types. Whereas identification of genotypes III, IV, and V may pose certain difficulties, since they all belonged to a single cluster. Inclusion of additional strains of these genotypes may produce better identification scores.

The blind coded M. kansasii samples tested with augmented genotype specific MSPs resulted in an overall identification success rate of 97%. Eighty-eight (98%) out of 91 blinded coded samples of genotype I gave a genotype match, despite the fact that log score values for seven strains were within the genus level cutoff (1.7–1.99). Four strains of genotype II, two of genotype III, and single strains of genotype IV and V were used for blind coded study, which resulted in successful identification of all strains at the species level, otherwise, except one genotype II strain, all these strains resulted in log score value <1.7 before the MSP augmentation. Despite an exhaustive search through culture collections of M. kansasii strains, genotype II–IV strains were seriously underrepresented. According to previous studies, M. kansasii involved in human disease belong almost exclusively to genotypes I and II, with the former accounting for 42–100%

values were observed for the four non-M. kansasii species included in the database evaluation.

TABLE 2 | Identification results of 100 blind-coded M. kansasii and one each of M. conspicuum, M. marinum, M. szulgai, and M. gastri strains using MALDI Biotyper Explorer Compass 4.1 reference database and after augmentation of genotype-specific reference database.


of the clinical isolations (Alcaide et al., 1997; Bakuła et al., 2013). The other types (III-VII), only sporadically recovered, are predominantly of environmental origin (Alcaide et al., 1997; Santin et al., 2004). The reason why seven isolates could be identified at the genus level only, and three isolates unidentifiable at all can hardly be found. We propose that the addition of spectra from genotype II-VI from various geographical regions could enable accurate and rapid differentiation of these genotypes.

In conclusion, unique, reproducible spectra for six genotypes of M. kansasii were established. The expansion of the database with reference spectra resulted in successful species- and genotype-level identification despite the limited number of genotype II–VI strains. Overall, we believe that MALDI-TOF MS can easily be employed for efficient genotype-level discrimination of M. kansasii and that further augmentation of spectral patterns might enable fast and accurate diagnosis in future.

## AUTHOR CONTRIBUTIONS

JM, TJ conceived and designed the work, acquired the data and wrote the manuscript. AL, EK, VU, MU, JH, and AS collected the strains and performed species-level identification. ZB performed species confirmation and genotype-level identification. JI and UR critically revised the manuscript. All authors had read and approved the submission version of the manuscript.

#### ACKNOWLEDGMENTS

The study was in part financed by the National Centre for Research and Development LIDER Programme [LIDER/044/457/L-4/12/NCBR/2013]. We are grateful to Elvira Richter (Labor Limbach, Heidelberg, Germany) and Katharina Kranzer, and Doris Hillemann (NRC for Mycobacteria, Borstel, Germany) for patient isolates of M. kansasii. We would like to thank Michael Kühl (Freie Universitaet Berlin, Germany) for excellent technical assistance.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2018.00587/full#supplementary-material

#### REFERENCES


Supplementary Figure 1 | MSP matching matrix was created by matching with M. kansasii genotype-specific MSPs augmented BDAL database. Color coded as per the manufacturer's recommendation on log score cutoff values.

Supplementary Table 1 | Blind coded isolates testing. Manufacturers recommended interpretation of score values were followed: from 0 to 1.699 indicates "no reliable identification" (Red); 1.7 to 1.999 indicate a "probable genus identification" (yellow); 2.0 to 2.299 indicate a "secure genus identification and probable species identification" (light green), and values from 2.300 to 3.000 indicate a "highly probable species identification" (intensive green).

Mycobacteria from solid culture media by use of vitek matrix-assisted laser desorption ionization-time of flight mass spectrometry. J. Clin. Microbiol. 51, 4226–4229. doi: 10.1128/JCM.02612-13


defined geographical area in Catalonia, Spain. Epidemiol. Infect. 132, 425–432. doi: 10.1017/S095026880300150X


Wilen, C. B., McMullen, A. R., and Burnham, C. A. (2015). Comparison of sample preparation methods, instrumentation platforms, and contemporary commercial databases for identification of clinically relevant mycobacteria by matrix-assisted laser desorption ionization-time of flight mass spectrometry. J. Clin. Microbiol. 53, 2308–2315. doi: 10.1128/JCM.00567-15

**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.

The reviewer AVDB declared a past co-authorship with one of the authors JI to the handling Editor.

Copyright © 2018 Murugaiyan, Lewin, Kamal, Bakuła, van Ingen, Ulmann, Unzaga Barañano, Humi˛ecka, Safianowska, Roesler and Jagielski. 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.