# THE ROLE OF IRON IN BACTERIAL PATHOGENESIS

EDITED BY : Susu M. Zughaier and Pierre Cornelis PUBLISHED IN : Frontiers in Cellular and Infection Microbiology

#### Frontiers Copyright Statement

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

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

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

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

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

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

The above represents a summary only. For the full conditions see the Conditions for Authors and the Conditions for Website Use. ISSN 1664-8714 ISBN 978-2-88945-662-8 DOI 10.3389/978-2-88945-662-8

#### About Frontiers

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

#### Frontiers Journal Series

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

#### Dedication to Quality

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

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

#### What are Frontiers Research Topics?

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

## THE ROLE OF IRON IN BACTERIAL PATHOGENESIS

Topic Editors: Susu M. Zughaier, Qatar University, Qatar Pierre Cornelis, Vrije Universiteit Brussel, Belgium

The collection of articles published in this eBook represent different facets of the interactions between pathogens and their host concerning the battle for iron. Pathogens have developed different strategies to acquire iron from their host. These include the production of siderophores, heme acquisition and ferrous iron uptake.

Citation: Zughaier, S. M., Cornelis, P., eds. (2018). The Role of Iron in Bacterial Pathogenesis. Lausanne: Frontiers Media. doi: 10.3389/978-2-88945-662-8

## Table of Contents


Lauren O'Connor, Jacqueline D. Fetherston and Robert D. Perry


Aaron T. Butt and Mark S. Thomas


Sarah Hijazi, Paolo Visca and Emanuela Frangipani


## Editorial: Role of Iron in Bacterial Pathogenesis

#### Susu M. Zughaier <sup>1</sup> \* and Pierre Cornelis <sup>2</sup> \*

*<sup>1</sup> Department of Basic Medical Sciences, College of Medicine, Qatar University, Doha, Qatar, <sup>2</sup> Microbiology Unit, Department of Bioengineering Sciences, Vrije Universiteit Brussel and VIB Department of Structural Biology, Brussels, Belgium*

Keywords: iron depletion, virulence factors, host defense against pathogenic bacteria, iron-regulated genes, siderophore

#### **Editorial on the Research Topic**

#### **Role of Iron in Bacterial Pathogenesis**

#### Edited by:

*John S. Gunn, The Ohio State University, United States*

#### Reviewed by:

*Kevin Mason, The Ohio State University, United States*

\*Correspondence:

*Susu M. Zughaier szughaier@qu.edu.qa Pierre Cornelis pcornel@vub.ac.be*

#### Specialty section:

*This article was submitted to Molecular Bacterial Pathogenesis, a section of the journal Frontiers in Cellular and Infection Microbiology*

Received: *22 August 2018* Accepted: *11 September 2018* Published: *16 October 2018*

#### Citation:

*Zughaier SM and Cornelis P (2018) Editorial: Role of Iron in Bacterial Pathogenesis. Front. Cell. Infect. Microbiol. 8:344. doi: 10.3389/fcimb.2018.00344* Iron is the fourth-most abundant element on the earth, and it is needed by most organisms, including bacteria. It exists in two oxidation states, Fe2<sup>+</sup> and Fe3+, and is involved in many oxido-reduction reactions (Andrews et al., 2013). Ferric iron (Fe3+) is the dominant form in oxygenated environments and has a very low solubility, which presents a problem for microorganisms with an aerobic lifestyle (Andrews et al., 2013). Conversely, in anaerobic environments or in microaerobic conditions at low pH, the soluble ferrous iron (Fe2+) is the most abundant form (Andrews et al., 2003). Bacterial pathogens face a problem because free iron is not available since it is bound to heme or by circulating proteins such as transferrin or lactoferrin (Finkelstein et al., 1983; Cornelissen and Sparling, 1994). Pathogens use different strategies to obtain iron from the host, via the production of extracellular Fe3+-chelating molecules termed siderophores (either their own or produced by other microorganisms), the uptake of heme, and the uptake of Fe2<sup>+</sup> (Feo system) (Andrews et al., 2013). A single pathogen can adapt its iron-uptake strategy in response to the type of infection (acute or chronic) and the availability or lack of ferrous iron (Cornelis and Dingemans, 2013). In this issue, several authors present several facets around iron uptake in different bacterial pathogens. Yersinia pestis produces the yersiniabactin siderophore under aerobic conditions and the Feo Fe2<sup>+</sup> uptake system under microaerobic conditions (Fetherston et al., 2012). The feo operon of Y. pestis is peculiar since it is repressed by Fe via the Fur repressor only under microaerobic, but not under aerobic conditions, unless the promotor region is truncated. The other facet of the host-pathogen battle for iron is the host response to the bacterial pathogen. As shown, again for Y. pestis, a live vaccine induces an iron nutritional immunity via the production of hemopexin and transferrin iron-binding proteins.

Some pathogenic bacteria infect fish, such as Vibrio anguillarum and Photobacterium damselae, both belonging to the Vibrionaceae. Citrate is probably the simplest siderophore and is produced by the citrate synthase and excreted by P. damselae strains unable to produce the vibrioferrin siderophore, thus establishing a link between the cellular metabolism and iron uptake. In their review article, Li and Ma describe the ways by which different V. anguillarum pathogenic strains take up iron either via the production of siderophores (anguibactin, vanchrobactin), uptake of xenosiderophores enterobactin or ferrichrome, or uptake of heme or ferrous iron. Burkholderia represents a genus of β-proteobacteria with 90 species, including the B. cepacia complex (BCC), which cause infections in the lungs of patients with cystic fibrosis and B. pseudomallei, which causes meloidiosis. Butt and Thomas reviewed the different iron-uptake strategies of these highly adaptable bacteria, including the production of siderophores (ornibactins, cepaciachelin, pyochelin, malleobactin), the uptake of heme and of ferrous iron. Francisella tularensis is the causative agent of tularemia and able to replicate in macrophages. F. tularensis can take up the siderophore rhizoferrin, but relies on the Feo system for the uptake of ferrous iron. The uptake of rhizoferrin does, however, not need the TonB protein as in other bacteria, while the uptake of Fe2<sup>+</sup> involves an outer membrane protein termed FupA, which is also unusual. Inhibition of the uptake of iron by bacteria involves, among other approaches, the use of gallium–protoporphyrin IX as shown in the case of Pseudomonas aeruginosa. The GaPPX is a heme analog that can be taken up via outer membrane heme receptors, inhibiting the growth of P. aeruginosa under conditions of iron limitation. Once in the cell, GaPPX was shown to target the aerobic respiration.

Bacterial pathogens sense iron-limiting conditions and respond accordingly by upregulating iron-acquisition mechanisms and virulence genes (Zughaier et al., 2014). Mouméne et al. report that the intracellular bacteria Ehrlichia ruminantium upregulates the type 4 secretion system (T4SS) and virulence genes under iron depletion via the newly identified master regulatory protein ExrR. Ferric uptake regulator (Fur) is a transcription factor that upregulates virulence factors in bacteria during iron depletion. Guo et al. used an unmarked gene deletion system to investigate the role of Fur in the virulence of Riemerella anatipestifer, an avian pathogen. Using RNA-seq analysis, they determined fur-regulated genes and identified putative fur-binding sequences. Their study further demonstrated that deleting the fur gene led to a reduction of virulence in vivo. In response to infection, the host limits the bioavailability of iron by upregulating expression of hepcidin, the master iron-regulating hormone, which limits iron uptake from the gut and retains iron in macrophages. Nairz et al. investigated the role of dietary iron

#### REFERENCES


enrichment in host-pathogen interactions during Salmonella typhimurium infection in mice with hereditary hemochromatosis (genetic Hfe-deficiency) compared to wild type. They observed that Salmonella infection induced hepcidin and hypoferremia in an Hfe-independent manner. However, iron overload increased the bacterial load in mice. Further, Salmonella infection in mice responded to iron-depleting conditions in the host by upregulated iron-acquisition genes.

Malhotra et al. report that Mycobacterium tuberculosis (M. tb) utilizes its highly conserved glycolytic enzyme GAPDH to acquire iron from the host by binding to lactoferrin with high affinity. M. tb sequesters iron from lactoferrin bound to GAPDH on the surface of bacteria. Sharma and Bisht provide a perspective on the role of iron-storing proteins in the emergence of antibiotic resistance. Based on their previous observation that iron-storing proteins, bacterioferritin (Rv1876) and ferritin (Rv3841), were overexpressed in aminoglycosidesresistant isolates of M. tb, they used a computational approach (STRING analysis) to predict protein partners that interact with bacterioferritin and ferritin. Among the identified partners is the hypothetical transmembrane protein Rv1877, which is predicted to be involved in drug resistance; therefore, Rv1877 may be a potential drug discovery target.

#### AUTHOR CONTRIBUTIONS

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


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

Copyright © 2018 Zughaier and Cornelis. 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 *feoABC* Locus of *Yersinia pestis* Likely Has Two Promoters Causing Unique Iron Regulation

Lauren O'Connor † , Jacqueline D. Fetherston and Robert D. Perry \*

*Department of Microbiology, Immunology, and Molecular Genetics, University of Kentucky, Lexington, KY, United States*

The FeoABC ferrous transporter is a wide-spread bacterial system. While the *feoABC* locus is regulated by a number of factors in the bacteria studied, we have previously found that regulation of *feoABC* in *Yersinia pestis* appears to be unique. None of the non-iron responsive transcriptional regulators that control expression of *feoABC* in other bacteria do so in *Y. pestis.* Another unique factor is the iron and Fur regulation of the *Y. pestis feoABC* locus occurs during microaerobic but not aerobic growth. Here we show that this unique iron-regulation is not due to a unique aspect of the *Y. pestis* Fur protein but to DNA sequences that regulate transcription. We have used truncations, alterations, and deletions of the *feoA::lacZ* reporter to assess the mechanism behind the failure of iron to repress transcription under aerobic conditions. These studies plus EMSAs and DNA sequence analysis have led to our proposal that the *feoABC* locus has two promoters: an upstream P1 promoter whose expression is relatively iron-independent but repressed under microaerobic conditions and the known downstream Fur-regulated P2 promoter. In addition, we have identified two regions that bind *Y. pestis* protein(s), although we have not identified these protein(s) or their function. Finally we used iron uptake assays to demonstrate that both FeoABC and YfeABCD transport ferrous iron in an energy-dependent manner and also use ferric iron as a substrate for uptake.

#### *Edited by:*

*Pierre Cornelis, Vrije Universiteit Brussel, Belgium*

#### *Reviewed by:*

*Gregory Plano, University of Miami, United States Deborah Anderson, University of Missouri, United States Simon Colin Andrews, University of Reading, United Kingdom*

#### *\*Correspondence:*

*Robert D. Perry rperry@uky.edu*

#### *† Present Address:*

*Lauren O'Connor, Department of Human Genetics, University of Michigan Medical School, Ann Arbor, MI, United States*

> *Received: 04 April 2017 Accepted: 05 July 2017 Published: 21 July 2017*

#### *Citation:*

*O'Connor L, Fetherston JD and Perry RD (2017) The feoABC Locus of Yersinia pestis Likely Has Two Promoters Causing Unique Iron Regulation. Front. Cell. Infect. Microbiol. 7:331. doi: 10.3389/fcimb.2017.00331* Keywords: ferrous transport, plague, iron regulation, ferric uptake, promoter structure

### INTRODUCTION

Iron is an essential nutrient and plays an important role in the pathogenesis of many organisms. Mammals chelate inorganic iron and heme using ferritin and hemoglobin in intracellular environments while transferrin, lactoferrin, heme- and hemoglobin-binding proteins sequester it in extracellular environments. Successful pathogens are able to acquire iron from these sources using a number of mechanisms: (1) secreted siderophores (2) heme transporters (3) ferric and ferrous transporters, as well as (4) transferrin and lactoferrin binding proteins to acquire iron from the host (Skaar, 2010).

Iron acquisition is crucial for the pathogenesis of Yersinia pestis, the causative agent of bubonic, pneumonic, and septicemic plague (Perry and Fetherston, 1997). Although, Y. pestis has multiple uptake systems for acquiring iron, relatively few have been shown to play a role in virulence in mammals (Kirillina et al., 2004; Perry, 2004; Forman et al., 2010). The yersiniabactin (Ybt) siderophore system is one of the most important for obtaining ferric iron (Kirillina et al., 2006; Perry and Fetherston, 2011) and is critical for virulence during pneumonic and the early stages of bubonic plague (Bearden et al., 1997; Fetherston et al., 1999, 2010). Ybt is encoded on the chromosomal pgm locus, a 102 kb region that can be spontaneously deleted. Recently, Bobrov et al. showed that components of the Ybt system are also involved in zinc uptake (Bobrov et al., 2014). The functions of 4 loci within the pgm locus have been determined: (1) ybt which encodes genes for the biosynthesis, regulation, and transport of the Ybt siderophore, (2) fet-flp, which encodes a ferrous iron transport system, (3) hms, which encodes components for biofilm development and (4) rip, which is involved in intracellular survival in phagocytic cells (Jarrett et al., 2004; Kirillina et al., 2004; Pujol et al., 2009; Forman et al., 2010).

The Yfe, Yfu, Yiu, and Hmu systems are ABC transporters involved in iron or heme uptake (Perry et al., 1990; Darby et al., 2002; Kirillina et al., 2004; Forman et al., 2010). The Hmu system is functional in vitro and is essential for the use of heme, but it does not play a role in the progression of bubonic or pneumonic plague (Thompson et al., 1999; Forman et al., 2010). The Yfu and Yiu transporters are functional in vitro; however, they are not important for the disease progression of bubonic plague (Gong et al., 2001; Kirillina et al., 2006). The Yfe system is an ABC transporter that transports iron and manganese. YfeA is a periplasmic binding protein that can bind Fe, Mn, and Zn, however, only Fe and Mn can be transported into the cytoplasm through the inner membrane permeases, YfeC and YfeD, with the help of the ATPase, YfeB (Bearden and Perry, 1999; Perry et al., 2012). Bearden et al. demonstrated that in an Ybt<sup>−</sup> background, a ∆yfeAB mutant had ∼2.5- fold decrease in its ability to transport iron (Bearden and Perry, 1999). In 2012, Fetherston et al. showed that in a Ybt<sup>−</sup> background, a ∆yfeAB ∆feoB mutant was completely defective in transporting ferrous iron (Fetherston et al., 2012). The Yfe and Feo systems appear to play somewhat of a redundant role in vitro, and play a role in the virulence of bubonic, but not pneumonic plague (Bearden and Perry, 1999; Fetherston et al., 2012).

The FeoABC system is comprised of three proteins. FeoB is an 85 kDa permease, which transports iron from the periplasm into the cytoplasm. FeoA is a small, 8.5 kDa protein that is predicted to interact with FeoB and stimulate GTPase activity. FeoC is a small, 9 kDa protein of unknown function. Due to its winged-helix motif, it has been predicted to play a role in regulation in Escherichia coli (Kammler et al., 1993; Lau et al., 2013). FeoAYP and FeoBYP have 77 and 74% identity to FeoAEC and FeoBEC respectively, whereas FeoCYP only has 57% identity to FeoCEC (Altschul et al., 1997). Guo et al. provided indirect evidence that FeoC is a negative regulator of the feoABC operon in Y. pestis (Gao et al., 2008). However, in our hands, a ∆feoC mutant did not affect transcription from the feo promoter under aerobic or microaerobic conditions (Fetherston et al., 2012). Due to our data, and the lack of homology between FeoC in Y. pestis and E. coli, we do not believe FeoC is a transcriptional repressor of the feoABC operon in Y. pestis. FeoC has been shown to contain an oxygen sensitive 4Fe-4S cluster and to bind to FeoB preventing its proteolytic degration by FtsH in Klebsiella pneumoniae (Hung et al., 2012; Hsueh et al., 2013; Kim et al., 2013). However, a similar role in Y. pestis is uncertain since the Feo system is functional in a ∆feoC mutant (Fetherston et al., 2012).

Iron is an essential element for the growth and pathogenesis of Y. pestis. However, excess iron in the presence of oxygen can lead to damaging reactive oxygen radicals. Therefore, iron transport into the cell is a very tightly regulated process. All of the tested ferrous and ferric iron uptake systems in Y. pestis are repressed by Fe-Fur under aerobic conditions, except for the FeoABC system. However, the feoABC locus is repressed by Fe under microaerobic conditions (Forman et al., 2010; Fetherston et al., 2012).

The feoA promoter in Y. pestis contains putative Fnr- and ArcA-binding sites (Gao et al., 2008). Previously, we have shown that ∆fnr, ∆arcA, and ∆rstAB mutations do not affect feoB transcription (Fetherston et al., 2012), even though Fnr positively regulates the feo operon of E. coli and Shigella, and ArcA and RstA are positive regulators of the operons in Shigella and Salmonella, respectively (Kammler et al., 1993; Boulette and Payne, 2007; Jeon et al., 2008).

In this study, we examined the transcriptional regulation of the feoABC operon. Our results suggest that a unique feature of the feo promoter region in Y. pestis prevents Fe-Fur repression aerobically but not microaerobically. We also demonstrate that the Yfe and Feo systems function in iron uptake under both reducing and non-reducing conditions.

#### MATERIALS AND METHODS

#### Bacterial Strains and Growth Conditions

Y. pestis KIM6+ and derivatives of this strain (Table S1) were inoculated from −80◦C glycerol stocks onto Congo Red plates. Plates were incubated at 30◦C for ∼48 h. Individual colonies were inoculated onto Tryptose Blood Agar Base (TBA) slants with the appropriate antibiotic, and incubated at 30◦C overnight. E. coli strains were inoculated from −80◦C glycerol stocks onto TBA slants with appropriate antibiotics and incubated at 30◦C overnight. For transformants, E. coli cells were grown in Luria broth (LB) or on LB agar plates.

#### Plasmid Construction

#### 5 ′ -Truncations of the Y. pestis KIM6+ feo Promoter Region

Promoter truncations were constructed by PCR amplification using the primer pairs listed in Table S2. PCR products were digested with PmeI and AscI and and cloned into the same sites of pWSKfeoA::lacZ (Fetherston et al., 2012; Table S1). The resulting plasmids with the correct inserts were verified by sequencing and were electroporated into Y. pestis strains.

#### Construction of an E. coli feoA Promoter Reporter

Primer pair Ecfeo-5′ and Ecfeo-3′ was used to amplify an ∼329 bp fragment from DH5α. The 3′ end of this fragment corresponds to the equivalent of nucleotide 1 in **Figure 1** (just prior to the predicted ATG start at 0 to −2). The PCR fragment was digested with PmeI and AscI and cloned into the same sites of pWSKfeoA::lacZ. Clones containing the correct insert were identified using Ecfeo-5′ and the universal −40 primer, verified by sequencing, and one clone (pWSKfeoAE.coli::lacZ) was electroporated into KIM6+ and KIM6-2030.

promoter regions for all reporter constructions end with nt 1. The dotted box around TC residues of the Fur-repressible P2 promoter element indicate the −13 and

#### Construction of the Integration Vector pUC18R6K-lac-Gm

−14 residues indicative of a σ

To construct a pUC18R6K-mini-Tn7T-Gm integration vector containing lacZ and a multi-cloning site (MCS), we first moved the lacZ gene with the feo promoter from pBSfeoPlacZ1Rs (Fetherston et al., 2012) into pWSK129. The plasmid, pBSfeoPlacZ1Rs, was digested with EagI and filled in with Klenow. The ∼3.8 Kb fragment containing lacZ and the feo promoter was cloned into the EcoRV site of pWSK129 to generate pWSKfeoPlacZ1Rs. This plasmid was digested with PmeI and AscI to remove the feo promoter and a multiple cloning site (MCS) was inserted in its place. The MCS was isolated from a PCR reaction containing pNEB193 as a template with the universal −40 and reverse universal primers. The PCR product was digested with PmeI and AscI and cloned into the PmeI/AscI sites of pWSKfeoPlacZ1Rs to generate pWSKlacMCS. The presence of the MCS was confirmed by sequencing. An ∼3.5 Kb XhoI/XmaI fragment containing the lacZ gene with the MCS

<sup>S</sup>-dependent promoter.

was isolated from pWSKlacMCS and cloned into the XhoI and XmaI sites of pUC18R6K-mini-Tn7T-Gm (Choi et al., 2005) to generate the integration vector, pUC18R6K-lac-Gm. A clone containing the correct insertion was identified by PCR with the universal −40 primer and integvec-up.

#### Cloning feoA, feoA165, feoA208, and feoA447 Promoters into pUC18R6K-lac-Gm

A PmeI/XhoI fragment containing lacZ with the respective feo promoter was isolated from pWSKfeoPlacZ1Rs, pWSKfeoA165::lacZ as well as pWSKfeoA208::lacZ and ligated to the 3.4 Kb PmeI/XhoI fragment of pUC18R6Klac-Gm to yield pUCR6KfeoA-lac-Gm, pUCfeoA165-lac-Gm and pUCR6Kfeo208-lac-Gm. Plasmids with the correct configuration were identified by PCR using primers Feo-Pro2 and integvec-up as well as by restriction enzyme digests. Following electroporation into KIM6+ along with pTNS2, strains containing the appropriate sequences integrated at the Tn7 insertion site were identified by PCR with primer pair Feo-Pro2 and attTn7Yp-fwd and designated KIM6-2203+ (attTn7:: feoA::lacZ), KIM6-2204+ (attTn7:: feoA165::lacZ) and KIM6-2205+ (attTn7:: feoA208::lacZ).

The feoA447 promoter fragment (spanning nucleotides 396 to −50 in **Figure 1**) was isolated from a PCR reaction containing pWSKfeo (Fetherston et al., 2012) as a template with primers Feo-Pro1 and FeoP-4. The PCR products were eluted from Zymo columns, digested with PmeI and AscI, and ligated into the same sites in pNEB193. A plasmid containing the feoA447 fragment was identified by sequence analysis and termed pNEBfeoP447. The PmeI/AscI promoter fragment from pNEBfeoP447 was then ligated into the PmeI/AscI sites of pUCR6K-lac-Gm. The resulting clones were screened using primers FeoP-4 and integvec-up as well as by restriction enzyme digests to identify a pUCR6K derivative carrying lacZ with the feoA447 promoter. This plasmid (pUCR6KfeoP447-lac-Gm) was introduced into KIM6+ along with pTNS2 and screened for integration by PCR with Feo-Pro2 and attTn7Yp-fwd primers. A strain containing the proper integrated reported was designated KIM6-2206+ (attTn7:: feoA447::lacZ).

#### Deletion of Specific Sequences within the feo Promoter of Y. pestis

Twenty base pairs encompassing the putative upstream (P1) promoter region of the feo operon were deleted using primer pair dfeoP1-5 and dfeoP1-3 in an overlapping PCR reaction with pNEBfeoP DNA as a template and Phusion (NEB). The PCR products were cleaned through Zymo columns, digested with DpnI, and transformed into DH5α. Plasmids containing the deletion were detected by PCR using primers pfeo2 and pfeo-R6. The products were analyzed on 5% polyacrylamide gels to detect the 20 bp deletion which was confirmed by sequencing. A PmeI/AscI fragment was isolated from one of the clones (designated pNEB1feoP1) that contained the correct sequence and inserted into the corresponding sites of pUCR6Klac-Gm. The ligation reactions were transformed into DH5α (λ pir) and plated on LB agar with 100 µg of ampicillin/ml. Plasmids containing an insert were identified by PCR using primers Feo-Pro2 and integvec-up. One of these, designated pUCR6K-1feoP1-lac-Gm, was electroporated along with pTNS2 into KIM6+, incubated for 1 h at 30◦C and plated on TBA containing 10 µg of gentamicin/ml. A strain in which the feo lac construct was integrated into the Tn7 insertion site was identified by PCR using primers Feo-Pro2 and attTn7Yp-fwd and designated KIM6-2207+ (attTn7:: feoA1feoP1::lacZ).

The same basic procedure was used to delete the P2 promoter region [41 bps which include the Fur binding site (FBS), −10 and −35 regions; Gao et al., 2008] and three repeated sequences within the feo promoter region (R1-3; **Figure 1**). Primers dfeoFBS-3 and dfeoFBS-5 were used to delete the P2 promoter region. Initial clones (pNEB1feoFBS) were screened with primer pair Feo-Pro2 and feo120-5. Plasmids containing an insert in pUCR6K-lac-Gm were identified using integvec-up and the universal −40 primer and designated pUCR6K-1feoFBS-lac-Gm.

Primer pairs FeoR1d-5 and FeoR1d-3, FeoR2d-5 and FeoR2d-3, and Feo1R3-5 with Feo1R3-3 were used to delete 10 bp that included repeated sequences R1, R2, and R3 (**Figure 1**) in separate PCR reactions. The pNEB clones with deletions of R2 and R3, designated pNEBfeo1R2 and pNEBfeo1R3 respectively, were initially identified by PCR with primer pair pfeo2 and pfeo-R6. Clones with deletions of R1, termed pNEBfeo1R1, were screened with primers Feo-Pro1 and FeoR2d-5. The final constructs in the integration vector were named pUCR6Kfeo1R1, pUCR6Kfeo1R2, and pUCR6Kfeo1R3.

Suicide plasmids with the promoter deletions were separately electroporated along with pTNS2 into KIM6+, incubated for 1 h at 30◦C and plated on TBA containing 10 µg of gentamicin/ml. A strain in which the feo lacZ reporter deletion was integrated into the Tn7 insertion site was identified by PCR using primers Feo-Pro2 and attTn7Yp-fwd. These strains were designated KIM6-2208+ (attTn7:: feoA∆FBS::lacZ), KIM6-2209+ (attTn7:: feoA∆R1::lacZ), KIM6-2210+ (attTn7:: feoA∆R2::lacZ), and KIM6-2211+ (attTn7:: feoA∆R3::lacZ).

#### β-Galactosidase Assays

From TBA slants, Y. pestis cells were inoculated to an OD<sup>620</sup> of 0.1 in chelex 100-treated PMH2 (cPMH2) at a pH of 6.0, 6.5, or 7.5 with or without 10 µM FeCl<sup>3</sup> and grown at 37◦C at 250 rpm through 2 transfers (6-8 generations). Lysates were assayed for enzymatic activity using ONPG (4-nitrophenyl-β-Dgalactopyranoside) as a substrate as previously described (Miller, 1992). The statistical significances of differences were calculated using the 2-tailed student's t-test.

#### Iron Transport Studies

Y. pestis cells were grown in cPMH2 under aerobic conditions for ∼5 generations prior to use in iron transport assays under reducing conditions. Transport was initiated by adding <sup>55</sup>FeCl<sup>3</sup> to a final concentration of 0.2 µCi/ml to cultures containing 5 mM sodium ascorbate. Parallel cultures, preincubated for 10 min with 100 µM carbonyl cyanide m-chlorophenylhydrazone (CCCP), were used to demonstrate energy-independent binding. The <sup>55</sup>FeCl<sup>3</sup> used in the transport studies was diluted into PMH2 containing 100 mM sodium ascorbate, incubated for at least 30 min at room temperature, and filtered through a 0.45-µm filter before being added to the cultures. Duplicate samples (0.5 ml) were taken at various time points after the addition of <sup>55</sup>FeCl3, filtered through 0.45-µm GN-6 nitrocellulose membranes, and rinsed twice with PMH2 medium containing 20 µM FeCl3. The membranes were added to vials containing Bio-Safe II scintillation fluid (Research Products International) and counted in a Beckman LS3801 liquid scintillation counter. Unfiltered samples were used to determine the total amount of radioactivity in each culture. The results are expressed as percent uptake/0.4 OD<sup>620</sup> to compensate for slight increases in cell density during the course of the assay.

For transport under non-reducing conditions Y. pestis cells were grown in cPMH1 for ∼5 generations prior to use in iron transport assays. Transport was initiated by adding filtered <sup>55</sup>FeCl<sup>3</sup> to a final concentration of 0.2 µCi/ml to cultures containing cold FeCl3. Cultures were processed and analyzed as described above.

#### Electrophoretic Mobility Shift Assays

Electrophoretic mobility shift assays (EMSAs) were used to demonstrate binding to the feo promoter region. Full length and truncated feo promoter probes were amplified by PCR using primer pairs pfeoF4 and pfeoR4, pfeoF5 and pfeoR5, or pfeoF6 and pFeoR6 (Table S2). Y. pestis KIM6+ or KIM6 fur::kan-9 (KIM6-2030) were grown with or without added iron in cPMH2 media through 2 passages to an OD<sup>620</sup> of 0.3–0.4, cells centrifuged and frozen at −80◦C. Cell pellets were resuspended in 2 ml of B-PER II reagent, 2 µl of lysozyme (10 mg/µl), 2 µL of DNase I and 4 µL of DNase I buffer and incubated with shaking for 20 min at room temperature. Samples were centrifuged to remove debris and protein content analyzed by the Bradford assay and comparison by SDS-PAGE. Lysates were aliquoted and stored at 4◦C. Binding reactions [100 mM Tris, 750 mM KCl, 10 mM dithiothreitol (DTT), 50 mM MgCl2, 12.5% glycerol, 50 mM EDTA [pH 8], 20 fmol biotinylated DNA probe, and Y. pestis cell lysates] were incubated at room temperature for 20 min. 5X EMSA loading buffer (50 mM Tris, 250 mM KCl, 5 mM DTT, 37 µM bromophenol blue) was added and the reactions were loaded onto a 6% DNA retardation gel (Invitrogen) with prechilled 0.5X TBE buffer and run at a constant voltage of 120 V for 90 min. After electrophoresis, gels were transferred to a Biodyne B Nylon membrane (Thermo Fisher Scientific) with prechilled 0.5X TBE buffer at a constant voltatge of 120 V for 45 min. Following transfer, the membrane was crosslinked with a UV Stratalinker 2400 (Stratagene, La Jolla, CA) and incubated overnight with 20 mL blocking buffer (Thermo Fisher Scientific). The Chemiluminescent Nucleic Acid Detection Module (Thermo Fisher Scientific) was used according to the manufacturer's protocol. The membrane was visualized on CL-X Posure Film (Thermo Fisher Scientific).

#### RESULTS

#### Truncations of the *feoA* Promoter Affect Overall Expression and Repression by Fe-Fur

Although, all other tested promoters controlling expression of iron transport systems in Y. pestis, are repressed by iron under aerobic conditions, the feoA promoter is repressed only under static (microaerobic) conditions. The other Fe2<sup>+</sup> transporters (Yfe, Fet-flp, and Efe) are repressed by iron under borth aerobic and microaerobic conditionas and none of the 4 Y. pestis Fe2<sup>+</sup> transporters are repressed by aerobic conditions compared to microaerobic conditions (Perry and Fetherston, 2004; Fetherston et al., 2012; Perry et al., 2012). We hypothesized that some aspect of the feoA promoter is preventing Fe-Fur regulation under aerobic conditions. To test this hypothesis we made 5′ truncations of the feo promoter region and fused them to a lacZ transcriptional reporter (**Figure 1**). KIM6 (∆pgm Feo+) strains carrying the various reporter plasmids were grown in cPMH2 in the presence and absence of iron under aerobic

growth conditions at 37◦C. When the intergenic region between y3913 and feoA is truncated from 396 to 335 bp, transcriptional activity increased ∼2.5-fold in the presence and absence of iron (**Figure 2**). This activity was further increased to a 4-fold difference, both in the presence and absence of iron when the 396 bp promoter region was truncated to 271 bp. A modest but statistically significant 1.3-fold ron-mediated repression was not observed until the feoA promoter region was truncated to 208 bp. This construct still had overall increased transcriptional activity compared to the full-length 396 bp promoter region (**Figure 2**). A further truncation to 195 bp maintained repression in the presence of iron, while the overall transcriptional activity decreased. The 165 and 125 bp promoters, showed stronger ironmediated repression (2.6- and 3-fold, respectively), and overall transcriptional activity that returned to the 396bp promoter levels (**Figure 2**). Thus, the 396 to 165 bp region somewhat prevents Fe-Fur repression aerobically, while the 396 to 335 bp region somewhat lowers overall transcriptional activity from the feoA promoter region.

#### The *feoA* Promoter Is Iron-Repressed Aerobically under Acidic Conditions

In Salmonella, RstA activates transcription from the feo promoter (Jeon et al., 2008). Previous work showed that a ∆rstAB mutation does not affect transcription from the Y. pestis feoA promoter (Fetherston et al., 2012). Since the rstAB operon is activated by the PhoP/PhoQ two-component regulators, which are activated by acidic pH (Jeon et al., 2008; Choi et al., 2009), we examined whether acidic conditions affected transcription from the feoA promoter. Y. pestis cells with full-length or truncated promoters were cultured in cPMH2 at pHs of 6.0, 6.5, or 7.5. It should be noted that all of the strains grew very slowly at pH 6.

FIGURE 3 | The full-length *feoABC* promoter is repressed aerobically by iron under acidic conditions. KIM6 carrying the various reporter plasmids were grown in the presence or absence of 10 <sup>µ</sup>M FeCl<sup>3</sup> at 37◦C under aerobic conditions in cPMH2 at pH 6, pH 6.5, or pH 7.5. Samples were taken during mid-exponential phase for β-galactosidase assays. Activities are averages of multiple samples from at least two independent experiments expressed in Miller units. Error bars represent standard deviations. Statistically significant fold differences due to iron status are shown (\**P* ≤ 0.02, \*\**P* ≤ 0.002, \*\*\**P* ≤ 0.0002).

FIGURE 4 | Fur is necessary for the iron-repression of the 125 bp promoter clone. KIM6 (*fur*+) and KIM6-2030 (*fur::kan*), carrying either the 396 or 125 bp reporter plasmids were grown in the presence or absence of 10 µM FeCl3 at 37◦C under aerobic conditions in cPMH2. Samples were taken during mid-exponential phase for β-galactosidase assays. Activities are averages of multiple samples from at least two independent experiments expressed in Miller units. Error bars represent standard deviations. The statistically significant fold difference due to iron status is shown (\*\*\**P* < 0.0002).

The 396 bp promoter had a 2-fold repression in the presence of iron when the cultures were grown at pH 6. A slight, but not statistically significant repression was seen at pH 6.5. Expression levels at pH 7.5 (from **Figure 2**) are shown for all constructs for reference (**Figure 3**). However, RstAB is not responsible for this effect since KIM6+ (rstAB+) and KIM6-2189+ (∆rstAB) had similar levels of activity from the feoA::lacZ reporter (Fetherston et al., 2012).

When the promoter was truncated to 208 bp, there was no significant iron repression at any of the tested pH levels (**Figure 3**), perhaps due to the high standard deviations associated with this reporter. While the 165 and 125 bp promoter truncations, showed ∼3-fold iron-mediated repression at pH 7.5, iron repression increased to 5- and 6-fold at pH 6.5, respectively. When the pH was further lowered to 6, the iron repression became even more pronounced—12- and 8-fold respectively. At all pHs, the pattern of increased expression from the 208 bp reporter, compared to the 396, 165, and 125 bp reporters was maintained (**Figure 3**).

#### Iron Repression of the Truncated *feo* Promoters Is Dependent on Fur

The 396 bp reporter does not exhibit iron repression. However, truncation of the promoter region to 125 bp causes transcriptional repression of the reporter by iron (**Figure 2**). This experiment was repeated using KIM6-2030 (fur::kan mutant) carrying the 396 or 125 bp reporter. With this fur mutant the 125 bp promoter reporter no longer exhibited iron repression. This indicates that Fur is responsible for the observed repression of the 125 bp reporter (**Figure 4**).

#### The Lack of Iron-Repression of the *feo* Promoter Is due to the Promoter Sequence, Not a Property of the *Y. pestis* Fur Protein

To determine whether the feo promoter itself or a unique aspect of the Fur protein in Y. pestis was preventing aerobic iron repression of the feoA promoter, we used a lacZ transcriptional reporter fused to the E. coli feo promoter region. The E. coli feo::lacZ reporter showed iron repression under aerobic conditions in Y. pestis (Figure S1). This suggests that a unique aspect of the Y. pestis Fur protein is not responsible for the lack of iron repression of the Y. pestis feo promoter under aerobic conditions. Rather the sequence/structure of the Y. pestis feoA promoter region likely prevents iron repression under aerobic, but not microaerobic conditions.

#### A Protein(s) That Is Not Fur Is Binding to the *feo* Promoter Region

We hypothesized that a repressor binding to the feo promoter region between 396 and 208 bp is the reason for increased transcriptional activity in the truncated promoters. Additionally, a second protein binding to the feoA promoter may prevent iron-repression under aerobic conditions. To examine these possibilities, an electrophoretic mobility shift assay (EMSA) was performed using whole cell lysates from Y. pestis that had been grown aerobically in the presence or absence of iron. Biotinylated probes were generated by PCR using primers that spanned the 396 bp feoA promoter, a 251 bp 3′ -truncated feo promoter fragment, or a 146 bp 5′ -truncated feo promoter fragment (**Figure 5A**). A shift was observed using the 396 bp (**Figure 5B**), 251 bp (**Figure 5B**), and 126 bp (**Figure 5C**) probes regardless of growth conditions.

We would not expect to see Fur binding in the EMSAs since Fur needs to be bound to a divalent cation in order to bind DNA (de Lorenzo et al., 1988). Even in the cultures where iron was added, the intracellular iron bound to Fur would likely be converted in the ferric form after cell lysis. Typically MnCl<sup>2</sup> is added to cell lysates or purified Fur, when studying Fur binding, since the Mn2<sup>+</sup> serves as a stable divalent cation under aerobic conditions (Friedman and O'Brian, 2003). However, since there is a Fur-binding site in the feoA promoter it was necessary to ensure that Fur was not responsible for this shift. Cell lysates from the fur mutant grown with or without iron were also used, and also resulted in a shift with the 396 bp (**Figure 5B**) and 146 bp (**Figure 5C**) probes. The 251 bp probe was not tested. Therefore, this observed shift is not due to Fur binding, but some other Y. pestis protein. In order to further identify the region in which the protein(s) is binding, two additional biotinylated probes were generated by PCR, splitting the 251 bp probe into an upstream 126 bp probe and downstream 125 bp probe (**Figure 5A**). Regardless of growth condition, a shift was seen with the upstream 126 bp probe, but not with the 125 bp downstream probe (**Figure 5C**). These data suggest that there are (1) two proteins binding the feo promoter: one within the 126 bp probe (396 to 271 bp region in **Figure 1**) and one within the 146 bp probe (146 to 1 bp region in **Figure 1**) or (2) one protein binding to both regions.

#### The Lack of Aerobic Iron Regulation Is Not due to Titration of a Transcriptional Regulator

Although, the pWSK vectors are considered low-copy number (Wang and Kushner, 1991) which we have used in transcriptional analysis of transcriptional regulation of numerous promoters (Fetherston et al., 1999, 2012; Gong et al., 2001; Kirillina et al., 2006; Forman et al., 2010; Perry et al., 2012), we wanted to ensure that our results were not due to titration of some negative regulatory factor by the feoA reporter plasmids. Therefore, we constructed vectors where the reporter gene would integrate into the Tn7 insertion site of Y. pestis using the Tn7-based system developed by Choi et al. (2005). We tested integrated reporters with the feoA (396 bp), feoA165, and feoA208 lacZ reporters in KIM6+ grown under static and aerobic conditions in the presence or absence of iron. The integrated reporters behaved in the same fashion as their plasmid counterparts (compare Figure S2 and **Figure 2**) with regard to the overall transcriptional activity and iron regulation. Thus, the feoA construct was only repressed by iron in cultures grown under static conditions while feoA165 reporter was iron-regulated under both static and aerobic conditions. The activity of the integrated feoA208::lacZ construct was much higher than the other two feo constructs and exhibited the same regulation pattern as its plasmid counterpart (compare Figure S2 and **Figure 2**). The integrated lacZ gene without any promoter fragment exhibited very low activity under all conditions (Figure S2).

We also used the integrated reporter system to examine regulation of an extended fragment of the feoA promoter region. Subsequent analysis and annotation of the Y. pestis feoABC operon predicts the start codon for feoA is further downstream than the originally annotated initiating ATG (at 0 to −2) in **Figure 1**. To determine if the additional bp separating the 2 predicted ATG start codons would affect the regulation of the feo operon, we made another feoA::lacZ reporter (pUCR6KfeoP447 lac-Gm) that included these nucleotides and integrated it into the Tn7 insertion site of KIM6+. The reporter with the extended promoter fragment exhibited the same properties as our original feoA::lacZ construct (data not shown) indicating that the additional base pairs did not play a role in the regulation of the feoABC operon.

#### The *feoABC* Locus Is Likely Transcribed from Two Independent Promoters

The transcriptional and binding data caused us to re-examine the feoA promoter region. Another, putative RNA polymerase (RNAP)—binding site (labeled P1/-10/-35 in **Figure 1**; the putative P1 promoter spans the Fnr binding site) is located upstream of the FBS and the P2 promoter (**Figure 1**). To test this, we deleted 41 bp that encompass the FBS and includes the putative −35 and −10 regions for the downstream P2 promoter. This strain, designated KIM6-2208+ (attTn7:: feoA1FBS::lacZ) had higher β-galactosidase activity (1.5–2.5X) than the integrated reporter with the original feoA::lacZ construct (KIM6-2203+; attTn7:: feoA::lacZ) under all conditions (**Figure 6**). Absence of the downstream P1 promoter may allow better read through from transcription initiation at the upstream P2 promoter. Interestingly, the attTn7:: feoA1FBS::lacZ) reporter was slightly repressed (∼2-fold) by iron under static but not aerobic conditions. Examination of the feoA promoter revealed additional potential FBSs with weaker matches to the Y. pestis consensus FBS (Zhou et al., 2006), located at bp 262–280 (11/19 bp) and bp 204–222 (12/19). Whether, either of these regions is responsible for the iron repression observed with the attTn7:: feoA1FBS::lacZ reporter is unknown; however, the observed ∼2 fold iron repression was alleviated in a Y. pestis fur mutant (KIM6-2030+; **Figure 6**).

Deletion of 20 bp that would include the putative −10 region of the upstream P1 promoter resulted in an integrated construct (attTn7:: feoA1feoP1::lacZ in KIM6-2207+) with lower overall β-galactosidase activity than the original attTn7:: feoA::lacZ reporter in KIM6-2203+ under both static and aerobic conditions. However, like the attTn7:: feoA165::lacZ reporter in KIM6-2204+, the attTn7:: feoA1feoP1::lacZ reporter was repressed by iron under both conditions. This provides additional evidence that the Y. pestis feo operon contains 2 promoter regions with the downstream P2 promoter being repressed by iron when cells are grown either statically or aerobically and suggest that the upstream P1 promoter is more active and slightly repressed by iron only under static conditions.

In examining the Y. pestis feo promoter region, we noted the presence of 3 nearly identical 11 bp direct repeat sequences, two of them located before the upstream P1 promoter (R1 and R2)

and one (R3) present between the 2 promoters (**Figure 1** and Table S3). To determine if these regions regulated expression of the feoABC operon, we deleted each of them in turn and analyzed the effect on the β -galactosidase activity of strains carrying the respective integrated reporters. An R3 deletion (attTn7:: feoA1R3::lacZ in KIM6-2211+) had no significant effect on the expression of the feoA::lacZ reporter under any of the tested conditions. Similarly, deletion of R1, the most distally located repeat, had no effect on regulation in that the reporter construct (attTn7:: feoA1R1::lacZ in KIM6-2209+) was only repressed by iron in cells grown under static conditions. Under aerobic conditions, the attTn7:: feoA1R1::lacZ reporter was not repressed by iron but was less active (51–65%) than the attTn7:: feoA165::lacZ reporter. The attTn7:: feoA1R2::lacZ in KIM6- 2211+ reporter in KIM6-2210+ behaved essentially the same as the attTn7:: feoA1feoP1::lacZ reporter. Expression of the lacZ

gene was lower relative to the attTn7:: feoA::lacZ reporter but was repressed by iron under both static and aerobic growth conditions (Figure S3). The R2 deletion would remove a potential −35 box for the putative P1 promoter (**Figure 1**) which likely accounts for the results we obtained with this construct.

#### The *FeoABC* and *YfeABCD* Systems Transport Iron under Both Reducing and Non-reducing Conditions

To demonstrate that the Yfe and Feo systems are involved in active iron transport, transport assays were performed under reducing and non-reducing conditions. Sodium ascorbate was utilized to maintain <sup>55</sup>FeCl<sup>3</sup> in the reduced, ferrous state. An ∆irp2 mutant (unable to make Ybt), transported ∼40% of the ferrous iron under reducing conditions (**Figure 7A**) and ∼35% under non-reducing conditions (**Figure 7B**) by 30 min of incubation at 37◦C. The uptake was energy-dependent, since CCCP-poisoned cells showed very little cell-associated iron (**Figures 7A,B**). The ∆irp2 ∆yfe double mutant, transported iron at slightly reduced rates compared to the ∆irp2 mutant, under both reducing (∼3-fold) (**Figure 7A**) and non-reducing (∼2 fold) (**Figure 7B**). The ∆irp2 ∆feo double mutant had slightly reduced rates of uptake under reducing conditions compared to the ∆irp2 mutant (**Figure 7A**). However, the ∆irp2 ∆yfe ∆feo triple mutant showed very little cell-associated iron, a level comparable to the CCCP-poisoned cells, under both reducing (**Figure 7A**), and non-reducing (**Figure 7B**) conditions. This indicates, that Yfe and Feo have somewhat redundant in vitro ferrous uptake functions with one system partially compensating for loss of the other. However, together they are crucial for ferrous iron transport under the short-term in vitro uptake conditions used here. In addition, these assays indicate that the Yfe and Feo systems can use ferric iron as well as ferrous.

#### DISCUSSION

The FeoABC system is widespread in bacteria and transcriptional regulation by Fnr, ArcA, RstAB and Fur has been demonstrated, particularly in E. coli, Salmonella, and Shigella (Kammler et al., 1993; Boulette and Payne, 2007; Jeon et al., 2008). Although, the Y. pestis feoA promoter encodes putative putative Fnr- and ArcA-binding sites (Gao et al., 2008), ∆fnr and ∆arcA mutations do not affect feoB transcription. In addition, oxyR, rstA, or feoC mutations failed to affect feoB transcription. Finally, the feoABC operon is repressed by iron and Fur during microaerobic but not aerobic growth (Forman et al., 2010; Fetherston et al., 2012). Thus, the Y. pestis feoABC locus currently has unique regulatory properties.

To assess the mechanism behind the lack of iron and Fur repression during aerobic growth, we assessed the regulation of 5′ -truncations of feoA::lacZ reporter with a 396 bp intergenic/promoter region. Modest aerobic ironmediated repression occurred with the 208 bp feoA::lacZ and stronger aerobic iron- repression was observed with the 165 and 125 bp promoter truncations (**Figure 2**). This repression was increased when Y. pestis cells were grown at acidic pHs (6.0 and 6.5) and even the 396 bp feoA::lacZ was repressed by iron after growth at pH 6.0 (**Figure 3**). When the feoA promoter from E. coli linked to lacZ was expressed in Y. pestis, it was repressed by iron aerobically (Figure S1), confirming that it is a unique aspect of the Y. pestis feoA promoter region that is preventing iron-Fur repression aerobically and not an aspect of the Y. pestis Fur protein. Thus, we propose that the 396 to 165 bp region somewhat prevents Fe-Fur repression aerobically and that pH has some influence on this.

In addition, when the promoter was truncated from 396 to 335 bp or 271 bp, there was an overall increase in transcriptional activity (**Figure 2**), suggesting that a repressor binds to the 396 to 335 bp region, lowering overall transcription. A series of EMSA experiments found that a Y. pestis cytoplasmic protein(s) bound to two regions: 396 to 271 and 145 to 1 bp (**Figure 5**). Although, we have not identified this protein(s) it is not Fur (**Figure 5**).

These results caused us to reexamine the sequence of the region upstream of feoA. This led to our hypothesis that a second promoter region (designated P1 in **Figure 1**; 198 to 171 bp which overlaps the putative Fnr-binding site) is upstream of the originally annotated Fur-repressible promoter (69 to 29 bp including the FBS; P2 in **Figure 1**). Note that the defined binding region for the 396 to 271 bp region lies outside of the predicted second promoter region. Significant transcriptional activity from the putative P2 promoter was demonstrated in a construct that deleted the P1 promoter region (feoA∆FBS::lacZ). Iron addition showed a modest (∼2-fold) repression of the P1 promoter under static (microaerobic) but not aerobic conditions. Although two weaker potential FBS sites (262–280 and 204–222) are upstream of the P1 promoter, their role in iron regulation has not been examined. We also identified three direct repeats (R1- R3 in **Figure 1**) and tested their role(s) on transcription of the feoABC operon. Deletion of R3 (151 to 141 bp region in **Figure 1**) had no effect on expression compared to the feoA::lacZ reporter, while the R1 (366 to 356 bp region in **Figure 1**) deletion reduced expression under aerobic conditions (51–65%) without altering iron regulation under either aerobic or microaerobic conditions (Figure S3). In contrast, the feoA∆R2::lacZ had activity essentially the same as the feoA∆feoP1::lacZ reporter. Note that the R2 repeat (206 to 196 bp region in **Figure 1**) is not within regions to which protein(s) bind (396 to 271 and 145 to 1 bp; **Figure 5**). Thus, the transcriptional effect observed in the R2 deletion likely occurs because this repeat overlaps the P1 −35 box (**Figure 1** and Figure S3). Thus, there is no evidence that the R1-3 repeats play a direct role in transcriptional regulation from either P1 or P2.

In our current model, truncation of the 297 bp region to 125 or 165 bp, eliminates the upstream P1 promoter, forcing transcription from the second Fur-regulated P1 promoter (**Figure 1**). Although, the 195 and 208 bp promoter truncationss contained both RNAP-binding sites, binding would not be optimal since the RNAP typically spans −55 to +20. Thus, there may have been limited transcription from the upstream P1 promoter with the majority coming from the downstream iron-repressible P2 promoter. The upstream RNAP polymerase binding site may favor an alternative RNAP possibly with a σ S subunit. In E. coli, Becker and Hengge-Aronis found that σ S dependent promoters have a tendency to have a "C" at position −13 and either a "T" or "G" at position −14. Of the promoters they examined, 85% had a "C" at −13, 40% had a "T" at −14, and 40% had a "G" at −14 (Becker and Hengge-Aronis, 2001). The upstream RNA polymerase binding site contains a "C" at position −13 and a "T" at position −14 (**Figure 1**). Activity of σ <sup>S</sup>RNAP promoters is not limited to stationary phase—activity is also induced in response to stresses such as osmotic, heat shock, and low pH (Kazmierczak et al., 2005). Whether, the upstream promoter uses an alternative RNAP or simply responds to environmental stresses remains to be determined.

In our model, aerobic conditions, as well as acidic conditions are stresses that lead to transcription from the upstream putative P1 promoter instead of or in addition to the σ <sup>70</sup>-RNAP, Fe-Fur repressible P2 promoter making iron-regulation negligible under these conditions. However, under microaerobic conditions, transcription from the upstream promoter is repressed. Whether this repression is due to a requirement for a different sigma factor RNAP or a DNA looping mechanism related to the two unidientifed protein binding sites that occludes the upstream promoter is undetermined. However, this leaves the downstream Fur-regulated P2 promoter as the active promoter which can be repressed by iron and Fur. Unlike FeoA, which has high amino acid similarity to FeoA of E. coli and Shigella, the feoA promoter regions of Y. pestis and E. coli show no significant similarities when compared via a DNA BLAST search (Zhang et al., 2000). Thus, the avoidance of iron-repression aerobically may be unique to Y. pestis, at least until the feoA promoter regions of other bacteria are assessed under aerobic conditions. Whether this lack of regulation aerobically plays a positive role in Y. pestis remains to be determined. One possible role would be that Feo serves to acquire basal levels of iron aerobically even under iron-surplus conditions. Y. pestis also appears to be unique in that Fnr, ArcA, RstAB, and FeoC do not regulate the feoABC operon (Fetherston et al., 2012) although they are proven regulators of this operon in

#### REFERENCES


other enterics (Kammler et al., 1993; Boulette and Payne, 2007; Jeon et al., 2008; Kim et al., 2013).

As expected our transport studies showed that FeoABC and YfeABCD each function to take up iron in an energy-dependent manner under reducing conditions (ferrous) (**Figure 7A**) Together these two system are responsible for nearly all the iron uptake observed. However, the Fet-Flp system and likely the EfeUOB systems also function in iron uptake. Perhaps neither of these systems functioned significantly during the 30-min time frame or conditions of this experiment. Unexpectedly we found that both the FeoABC and YfeABCD systems could use ferric iron as a substrate for uptake over the 30 min period (**Figure 7B**). These experiments did not distinguish between the possibility that both systems transport ferric iron or that ferric iron is reduced to ferrous prior to transport. However, either alternative suggests that the FeoABC and YfeABCD transporters could function aerobically under non-reducing conditions.

#### AUTHOR CONTRIBUTIONS

LO is the first author. LO and JF contributed experimental work and planning of the research course. RP oversaw the research and contributed to the planning of the research course.

#### ACKNOWLEDGMENTS

This work was supported by Public Health Services Grant RO1 AI33481 from the US National Institutes of Health.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fcimb. 2017.00331/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 © 2017 O'Connor, Fetherston and Perry. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

## Host Iron Nutritional Immunity Induced by a Live *Yersinia pestis* Vaccine Strain Is Associated with Immediate Protection against Plague

Ayelet Zauberman\*, Yaron Vagima, Avital Tidhar, Moshe Aftalion, David Gur, Shahar Rotem, Theodor Chitlaru, Yinon Levy and Emanuelle Mamroud\*

*Department of Biochemistry and Molecular Genetics, Israel Institute for Biological Research, Ness-Ziona, Israel*

Prompt and effective elicitation of protective immunity is highly relevant for cases of rapidly deteriorating fatal diseases, such as plague, which is caused by *Yersinia pestis*. Here, we assessed the potential of a live vaccine to induce rapid protection against this infection. We demonstrated that the *Y. pestis* EV76 live vaccine protected mice against an immediate lethal challenge, limiting the multiplication of the virulent pathogen and its dissemination into circulation. *Ex vivo* analysis of *Y. pestis* growth in serum derived from EV76-immunized mice revealed that an antibacterial activity was produced rapidly. This activity was mediated by the host heme- and iron-binding proteins hemopexin and transferrin, and it occurred in strong correlation with the kinetics of hemopexin induction *in vivo*. We suggest a new concept in which a live vaccine is capable of rapidly inducing iron nutritional immunity, thus limiting the propagation of pathogens. This concept could

#### *Edited by:*

*Pierre Cornelis, Vrije Universiteit Brussel, Belgium*

#### *Reviewed by:*

*Girija Ramakrishnan, University of Virginia, United States Roger Derek Pechous, University of Arkansas for Medical Sciences, United States*

#### *\*Correspondence:*

*Ayelet Zauberman ayeletz@iibr.gov.il Emanuelle Mamroud emmym@iibr.gov.il*

*Received: 23 April 2017 Accepted: 06 June 2017 Published: 21 June 2017*

#### *Citation:*

*Zauberman A, Vagima Y, Tidhar A, Aftalion M, Gur D, Rotem S, Chitlaru T, Levy Y and Mamroud E (2017) Host Iron Nutritional Immunity Induced by a Live Yersinia pestis Vaccine Strain Is Associated with Immediate Protection against Plague. Front. Cell. Infect. Microbiol. 7:277.*

*doi: 10.3389/fcimb.2017.00277*

be exploited to design novel therapeutic interventions.

Keywords: *Yersinia pestis*, plague, iron nutritional immunity, hemopexin, live vaccine, EV76, protective immunity, innate immunity

#### INTRODUCTION

Live vaccines are able to elicit extensive humoral immunity against specific antigens as well as inflammatory and cellular protective immune responses. Furthermore, the presentation of antigens to the host immune system in the context of infection can potentiate the amplitude and longevity of the protective response.

While the ability of live vaccines to serve as platforms for the initiation of adaptive protective responses against infections has been well documented (Coward et al., 2014; Minor, 2015), their ability to evoke protective innate immune mechanisms remains largely unknown. The induction of rapid, non-specific protective responses by live vaccines could provide a means to delay the progression of infection during sudden outbreaks, thereby extending the therapeutic window and allowing for the elicitation of the adaptive response, the manifestation of which typically requires more time post-exposure. In addition, the identification of potent innate immune mechanisms could lead to the discovery of novel bacterial or host-borne therapy targets, which are essential countermeasures to antibiotic-resistant pathogens.

Plague, which is caused by Yersinia pestis, is a fatal disease that has caused millions of deaths in three world pandemics. Over the last few decades, plague has persisted in Africa, Asia and the Americas, and since the 1990s, it has been categorized as a re-emerging disease (Bertherat, 2016). At the same time, Y. pestis has been recognized as a Tier 1 select biological agent due to its potential use as a bioweapon (Inglesby et al., 2000). The severity of the disease in combination with concerns regarding the existence of antibiotic-resistant Y. pestis strains has prompted efforts to develop safe live vaccines (Titball and Williamson, 2004; Feodorova et al., 2007; Tidhar et al., 2009; Feodorova and Motin, 2012; Dentovskaya et al., 2015; Derbise et al., 2015; Tiner et al., 2016; Verma and Tuteja, 2016).

Protection against plague can be achieved by reiterated administrations of live EV76-based vaccines, which elicit a strong adaptive protective immune response. As of today, the only wellestablished molecular basis for the attenuation of these vaccines is attributed to the absence of the pigmentation locus (pgm)-which results in iron acquisition deficiency, a process essential for the survival and expansion of the pathogen in the host (Fetherston et al., 1992; Perry and Fetherston, 1997). Whether live vaccines such as EV76 are merely attenuated mutants or whether they manifest protection-inducing activities that are silent in their virulent parental strains remains an unanswered issue.

In the current report, we evaluated the efficacy of the Y. pestis EV76 live vaccine strain in providing immediate protection against both bubonic and pneumonic plague, and we characterized the mechanism involved in this innate immune response.

#### MATERIALS AND METHODS

#### Bacterial Strains

The Y. pestis strains used in this study included Kimberley53 (Kim53), a fully virulent Y. pestis strain (biovar orientalis), the live vaccine strain EV76 (pgm<sup>−</sup> [Girard's strain]) and the avirulent Kim53170110 strain (Ben-Gurion and Shafferman, 1981; Flashner et al., 2004; Zauberman et al., 2013). Kim53170110 strain was spontaneously cured of pCD1 (carrying the genes coding for elements of the Type III secretion system, essential for manifestation of virulence, Perry and Fetherston, 1997) and pPCP1 (carrying the gene coding for the virulence factor plasminogen-activating protease, Sebbane et al., 2006; Lathem et al., 2007).

To construct a bioluminescent Y. pestis, the plasmid pGENluxCDABE (a generous gift from Professor H. Mobley, Lane et al., 2007) was introduced into the Y. pestis Kim53 strain by electroporation, generating the Kim53-lux strain. Maintenance of the virulence-associated plasmids pMT1, pCD1, and pPCP1 and the pgm locus in the Kim53-lux strain was verified by PCR analysis.

#### Ethics Statement

This study was performed in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals from the National Institutes of Health. All of the animal experiments were performed in accordance with Israeli law and were approved by the Ethics Committee for animal experiments at the Israel Institute for Biological Research. (Protocols M-03-14, M-28-14, M-47-15, M-20-16, and M-37-16).

#### Animal Studies

Female 8 to 12-week-old C57BL/6 mice (Harlan, Israel) were used in this study. For s.c. infections in the mice, Y. pestis bacteria were grown on brain heart infusion agar (BHIA, Difco) for 48 h at 28◦C and were then suspended in saline solution (0.9% NaCl). Protection experiments were carried out by challenging vaccinated animals by two routes of infection-s.c. or i.n. (i) Mice were infected simultaneously with 1 × 10<sup>7</sup> CFU of EV76 in the upper back (s.c. vaccination) and with 100 CFU (100 LD50), of Kim53 in the lower back (s.c. challenge). (ii) Mice were infected with 1 × 10<sup>7</sup> CFU of EV76 in the upper back (s.c. vaccination) simultaneously or 2 days prior to i.n. infection (challenge) with 1 × 10<sup>4</sup> CFU (10 LD50) of Kim53. Mice infections were performed as described previously (Zauberman et al., 2009; Vagima et al., 2015). Bacteria were quantified by CFU counting on BHIA plates. For bacterial dissemination to internal organs and blood, groups of at least 4 mice were anesthetized; tail vein blood was collected; spleens, draining inguinal lymph nodes (I-LNs), draining axillary lymph nodes (Ax-LNs) and mediastinal lymph nodes were harvested; and then tissue homogenates were prepared in 1 ml of PBS/organ. Bacterial quantification in the tissue homogenates or in blood samples was performed by plating serial dilutions in PBS on BHIA and by calculating the CFU/organ or CFU/1 ml of blood. To differentiate between EV76 and Kim53 in the organs, the samples were plated on BHIA supplemented with 100 µg/ml streptomycin, conditions that allow for the growth of only the Kim53 streptomycin-resistant strain.

#### Whole-Animal Luminescent Imaging

Anesthetized mice previously infected s.c. with 100 CFU of Kim53-lux were analyzed by visualizing photon emission using an in vivo Imaging System-IVIS (Caliper Life Sciences, Hopkinton, MA, USA). Image acquisition was performed with the binning set to 2. The acquisition time ranged between 1 and 4 min. The luminescence signals for all of the images were normalized using Living image <sup>R</sup> 4 software (PerkinElmer, USA) and are reported as photons/second/cm<sup>2</sup> /sr.

#### *Ex vivo Y. pestis* Growth Assay

For ex vivo growth assays, mouse serum were mixed with 1 × 10<sup>3</sup> Y. pestis bacteria (total volume of 100 µl) and incubated in microtiter plates at 28◦C. Dilution of mouse serum was done in RPMI medium supplemented with 10% fetal calf serum. Test samples and control samples contained identical amounts of mouse serum. Kim53 or Kim53170110 bacterial growth was determined by serial dilution and viable cell counts. The inhibitory effect was similar in both strains. To assess bacterial growth in immunized mouse serum in the presence of iron, iron dextran (Sigma D8517) was added at concentrations between 1.3 µg/ml to 1 mg/ml in 3-fold serial dilutions.

#### Proteinase K Digestion

Proteolysis was performed by incubating the serum samples with proteinase K-conjugated beads (Sigma Aldrich P9290) for 2 h at 37◦C according to the manufacturer's recommendations. Proteinase K-treated serum samples were centrifuged to remove the beads, and the supernatants were collected for analysis.

#### SDS-Page, Western Blot Analysis and Anti-F1 Elisa

Serum proteins (10 µl) were separated by 4–12% gradient SDS-PAGE (Invitrogen) and were stained using a Silver Stain Plus kit (Bio-Rad). For Western blot analysis, electrophoresed samples were transferred to nitrocellulose membranes (Invitrogen). Membranes were developed with polyclonal antibodies against hemopexin (Abcam ab 90947), transferrin (Abcam ab 82411) and β-actin (Sigma A5060), followed by HRP-conjugated goat anti-rabbit IgG. Anti-F1 ELISA in the serum of immunized mice was performed as described previously (Levy et al., 2011, 2016). Briefly, microtiter plates were coated with 500 ng of purified rF1 [provided by the Biotechnology Department at IIBR, produced as described in Holtzman et al. (2006)]. Tested sera were serially diluted in 2-fold dilutions in a final volume of 50 µl and were incubated in the wells for 1 h at 37◦C. Alkaline phosphataseconjugated goat anti-rabbit IgG (1/2,000 dilution, Sigma) was used as the 2nd layer for rabbit anti-F1 IgG titer determination. Titers were defined as the reciprocal values of the endpoint serum dilutions that displayed OD<sup>405</sup> values 2-fold higher than the normal serum controls.

#### Ion-Exchange Chromatography

A pooled serum sample from EV76-infected mice (48 h postinfection) was first diluted and filtered through a 100 k-Da size exclusion membrane (VIVASPIN, Sartorius). The Y. pestis ex vivo growth inhibitory activity was identified to be in the upper fraction. The upper fraction was diluted 1:5 in buffer A (25 mM HEPES, pH 8.0) and was loaded onto a HiTrapQ HP 5-ml column from GE Healthcare equilibrated in the same buffer, using an AKTA Explorer system (GE-Healthcare) at 4◦C. After loading, the column was washed with 10 column volumes (cv) of 100% buffer A at 4 ml/min, and it was eluted using a composite gradient between buffer A and B (25 mM HEPES, pH 8.0, and 1 M NaCl) with 20 cv of 10–30% B, 2 cv of 30% B, and 3 cv of 30–100% B at 2 ml/min, collecting 2 ml fractions. Finally, the column was eluted with 3 cv of 100% B at 2 ml/min, collecting 4 ml fractions. The fractions were sampled and used for ex vivo bacterial growth assays.

The fraction containing the inhibitory growth activity was subjected to a second purification step using anion exchange chromatography. After loading the sample, the column was washed with 5 cv of 100% A at 2 ml/min, collecting 4 ml fractions, and was eluted using a composite gradient between buffer A and C (25 mM HEPES, pH 8.0, and 150 mM NaCl) with 20 cv of 0–15% C at 1 ml/min, collecting 2 ml fractions. Then, the elution was continued with 5 cv of 15% C, 4 cv of 20% C, 3 cv of 25% C, 3 cv of 25–100% C and 3 cv of 100% C at 2 ml/min, collecting 2 ml fractions. Finally, the column was eluted with 3 cv of 100% C at 2 ml/min, collecting 4 ml fractions. The fractions were sampled and used for ex vivo bacterial growth assays. Faction 26 from the second chromatographic purification step exhibited the highest growth inhibitory activity.

#### Mass Spectrometry

Fraction 26, which contained most of the growth inhibitory activity, was subjected to in-solution digestion and was partially dried in an Eppendorf SpeedVac, and the remaining sample was resuspended in 100 mM ammonium bicarbonate to a final volume of 50 µl. Prior to enzymatic digestion with trypsin, the sample was reduced with dithiothreitol (DTT) and alkylated with iodoacetamide. Briefly, reduction was conducted at a final concentration of 10 mM DTT at 56◦ for 1 h. Iodoacetamide was then added to a final concentration of 55 mM, and the reaction was incubated at room temperature for 30 min. Finally, the reaction was quenched with 10 mM DTT. Enzymatic digestion was conducted overnight at 37◦C with trypsin at an enzyme to protein ratio of 1:50 (wt/wt). The sample was partially dried in a SpeedVac. The resulting peptide mixtures were diluted in 80% formic acid and were then diluted 1:10 with Milli-Q water immediately before inline reverse-phase nano-LC (liquid chromatography)-electro spray ionization (ESI) tandem mass spectrometric analysis (MS/MS).

Samples were analyzed using an LTQ Orbitrap (Thermo Fisher Scientific, Bremen, Germany) operated in the positive ion mode. Nano-LC-ESI-MS/MS peptide mixtures were separated by inline reversed-phase nanoscale capillary LC and were analyzed by ESI-MS/MS. For LC-MS/MS, the samples were injected onto a 15 cm reversed-phase spraying fused-silica capillary column [constructed in-house; inner diameter 75 µm and packed with 3 µm of ReproSil-Pur C18A18 media (Dr. Maisch GmbH, Ammerbuch-Entringen, Germany)] using an UltiMate 3000 Capillary/Nano LC System (Dionex). The LC setup was connected to an LTQ Orbitrap mass spectrometer (Thermo Fisher Scientific, Bremen, Germany) equipped with a nanoelectrospray ion source (Thermo Fisher Scientific, Bremen, Germany). The flow rate through the column was 300 nl/min. An acetonitrile gradient was employed with a mobile phase containing 0.2% formic acid in Milli-Q water. The injection volume was 5 µl. The peptides were separated by 4 h gradients that ranged from 5 to 65% acetonitrile. The voltage applied to the union to produce an electrospray was 1.2 kV. Helium was introduced as a collision gas at a pressure of 3 psi. The mass spectrometer was operated in the data-dependent mode. Survey MS scans were acquired in the Orbitrap with the resolution set to a value of 60,000. For the analysis of tryptic peptides, survey scans were recorded in the FT mode, followed by data-dependent collision-induced dissociation (CID) of the 7 most intense ions in the linear ion trap (LTQ). Raw data were searched with MASCOT (Matrix Science, London, UK) against a UniProt mouse database. The search parameters included variable modifications of 57.02146 Da (carboxyamidomethylation) for Cys, 15.99491 Da (oxidation) for Met and 0.984016 Da (deamidation) for Asn and Gln. The search parameters were as follows: a maximum of 2 missed cleavages; initial precursor ion mass tolerance of 10 ppm; and fragment ion mass tolerance 0.6 Da.

#### Statistical Analysis

Statistical significance was measured using Student's unpaired t-test. Survival curves were compared using the log-rank test. In all of the analyses, p-values equal to 0.05 served as the limit of Zauberman et al. Vaccine-Mediated Iron Sequestration

significance. Calculations were performed using GraphPad Prism software.

#### RESULTS

#### *Y. pestis* EV76 Provides Immediate Robust Protection against Plague

To discover novel innate immune responses that could result in the rapid onset of protective immunity against infection, we evaluated the ability of the live attenuate Y. pestis EV76 strain to induce protection against subcutaneous (s.c.) lethal challenge with the fully virulent Y. pestis Kimberley53 strain (Kim53), when both strains were administered concomitantly to mice. Accordingly, the mice were infected by s.c. injection of a single dose (1 × 10<sup>7</sup> CFU) of the attenuated strain and were immediately challenged subcutaneously with a lethal dose of 100 LD<sup>50</sup> of the virulent strain. To avoid possible direct interaction between the two co-administered strains, the attenuated strain was injected in the upper right back close to the right axillary lymph node (RAx-LN), whereas the virulent strain was injected in the lower left back, close to the left inguinal lymph node (LI-LN). Control mice were injected with saline and were challenged with the virulent Y. pestis strain (see scheme in **Figure 1A**).

As shown in **Figure 1B**, all of the control mice died within 5 days of the challenge. In contrast, simultaneous coadministration of the attenuated and virulent strains provided a very high protection level of 91%. This result clearly indicated that the EV76 strain was able to elicit an unforeseen, rapid and potent protective response against plague, strongly suggesting that upon immunization, some unknown protective mechanisms manifested prior to the establishment of an adaptive protected state. Simultaneous co-administration of the virulent Y. pestis strain with 1 × 10<sup>7</sup> CFU of another attenuated Y. pestis strain (Kim53170110, which lacks the plasmids that carry major Y. pestis virulence determinants), could not protect the infected mice (**Figure 1B**). This observation indicated that the early protective response promoted by EV76 could not be attributed merely to the large amount of bacteria administered but to a particular feature of this strain, and/or to its relative ability to replicate and disseminate in the host.

To examine whether protection could also be achieved by post-challenge administration of the EV76 strain, the mice were injected with EV76 5 h post-challenge. Under this condition, a significantly increased survival rate of 34% was also observed (**Figure 1B**).

In light of these results, we addressed the question of whether EV76-administration might promote rapid protection against pneumonic plague, which is considered to be a more challenging and severe manifestation of the disease. Accordingly, the mice were challenged intra-nasally (i.n.) with a lethal dose of the virulent Y. pestis strain 2 days after s.c. administration of the EV76 strain. As seen in **Figure 1C**, all of the control mice died at 3 days post-challenge, whereas 60% of the immunized mice survived the challenge. When mice were concomitantly infected with both strains, the mean time to death was extended from 3 to 6.8 days (**Figure 1C**). These results indicated that

FIGURE 1 | The EV76 strain confers rapid protection against bubonic and pneumonic plague. (A) Schematic illustration of the injection sites of the EV76 vaccine strain and the virulent *Y. pestis* Kim53 strain. RAx-LN (right axillary lymph node), LAx-LN (left axillary lymph node), RI-LN (right inguinal lymph node), and LI-LN (left inguinal lymph node). (B) Kaplan-Meier survival profiles of mice challenged s.c. in the lower left side with 100 CFU of the virulent *Y. pestis* strain with concomitant immunization in the upper right side with 1 × 10<sup>7</sup> CFU of the EV76 strain (blue triangles), with 1 × 10<sup>7</sup> CFU of the Kim53170110 attenuated strain (green diamonds), or with saline (red circles). The black line represents the survival profiles of mice immunized with 1 × 10<sup>7</sup> CFU of the EV76 strain 5 h after s.c. challenge (p.c.) with the virulent Kim53 *Y. pestis* strain (100 CFU). (C) Mice were infected i.n. with 10<sup>4</sup> CFU of the virulent Kim53 *Y. pestis* strain and were then immediately infected s.c. with the EV76 strain (blue triangle) or with saline (red circle). The black line represents the survival of mice immunized s.c. with 1 × 10<sup>7</sup> CFU of EV76 2 days before i.n. challenge with 1 × 10<sup>4</sup> CFU of Kim53 (b.c.); *n* = 6 to *n* = 11 mice per group in at least 2 independent experiments. The *p*-values were calculated using Kaplan-Meier analysis with a log-rank (Mental-Cox) test compared to sham controls. \*\**p* < 0.01 and \*\*\**p* < 0.001.

exposure to the Y. pestis EV76 live vaccine strain induced a very potent protective mechanism within a time window of less than 4 days, which is much earlier than the time usually required for the manifestation of an adaptive protective response.

#### Co-administration of the Protective EV76 Strain Restricts the Growth and Dissemination of the Virulent *Y. pestis* Strain

To further characterize the early protective activity induced by the EV76 strain, the effects of the administration of the attenuated strain on the expansion and dissemination of the virulent bacteria in the host were inspected using whole-animal luminescent imaging of individual animals with an in vivo imaging system (IVIS). Subcutaneous co-administration of the bacterial strains was performed as described above; however, the virulent strain used for the challenge expressed a bioluminescent tag (Kim53 lux). The animals were monitored daily for bioluminescence emission from 1 day post-infection until death. As depicted in **Figure 2A** (upper panel), in the non-immunized group, a luminescent signal marking the presence of at least 5 × 10<sup>6</sup> bacteria (the detection limit) was visible 48 h after the challenge and was limited to the site of injection or to the inguinal lymph node proximal to the injection site. Dissemination of the virulent bacteria into internal organs was first observed 72 h post-challenge. Five days post-challenge, 4 of 6 mice died, and the two remaining mice exhibited strong systemically dispersed luminescent signals. In contrast, mice injected with the protective EV76 strain simultaneously with the virulent strain did not exhibit any luminescent signals throughout the experiment, apart from a single mouse in which a minor and transient signal was observed at 72 h post-challenge at the draining lymph node (**Figure 2A**, lower panel). No mortality was observed in the immunized group. These results suggested that the protective strain rapidly induced a bacterial inhibitory effect following its administration that blocked the replication and dissemination of the virulent strain early in the course of infection.

To determine the effects of EV76 administration during the early stages of infection when the quantity of virulent bacteria in the host is less than the threshold enabling IVIS-mediated detection, the bacterial loads in organs were quantified by direct CFU counts. The mice were injected s.c. with EV76, with the virulent strain or with both, as above. At different time points post-infection, the tissues of infected mice were harvested, and the bacterial loads in the draining lymph nodes, spleen and blood were determined.

In the mice infected s.c. with 1 × 10<sup>7</sup> CFU of EV76, bacteria reached the draining RAx-LN proximal to the injection site within 1 h. Then, the bacteria persisted in the lymph node for at least 4 days and typically did not disseminate to other internal organs (**Figure 2B**). In the mice injected with the virulent Kim53 strain (100 CFU), a small amount of bacteria was found in the draining LI-LN, close to the injection site, within 1 h. The bacterial load at the LI-LN grew rapidly to 6 × 10<sup>5</sup> CFU (geometric mean) after 72 h and was paralleled by dissemination to the spleen and blood (**Figure 2C**).

When the virulent strain was co-administered with the protective EV76 strain, it disseminated to the draining LI-LN within 1 h (**Figure 2D**). However, in contrast to the pattern of distribution observed for the virulent strain in the absence of EV76, by 24 h, the number of bacteria detected in the LI-LN was notably lower, and by 72 h post-challenge, in 4 of 5 mice, the virulent bacteria were cleared from the draining LN and were not detected in the spleen or blood (**Figure 2D**). These results suggested that EV76 has rapid and systemic effects that inhibit the growth and dissemination of the virulent strain from the draining lymph node to internal organs and indicated that direct interaction between the two strains of the bacteria was not involved in this process. Furthermore, the persistence of EV76 bacteria in the lymph node for at least 4 days post-immunization (**Figure 2B**) could enable continuous induction of this putative protective mechanism.

#### Serum Derived from Mice Following Exposure to the Protective EV76 Strain Exerts Potent Protein-Based Antimicrobial Activity against *Y. pestis*

The results of the dissemination experiments strongly suggested that an antibacterial component was produced upon immunization with the protective EV76 strain, and it possibly migrated through the circulation, affecting propagation of the virulent strain.

To probe a putative non-cellular antibacterial effector, an ex vivo growth assay was designed (**Figure 3A**) in which Y. pestis bacteria were subjected to mouse serum from mice 24 h post-immunization (p.i.) with EV76 or to saline-immunized mouse serum (control serum) (**Figure 3B**). During the first 6 h of incubation, Y. pestis bacteria incubated with serum derived from control mice or with EV76-immunized mouse serum exhibited similar growth curves. However, Y. pestis growth continued with the control serum for an additional 72 h, reaching a concentration of 2.8 × 10<sup>9</sup> CFU/ml. In sharp contrast, at 24 h post-incubation, Y. pestis growth in the EV76-immunized mouse serum ceased, resulting in significantly lower bacterial concentrations of only 4 × 10<sup>5</sup> CFU/ml at 72 h. Therefore, the time course of the Y. pestis growth inhibition ex vivo was correlated with the protection observed in vivo. Furthermore, serum collected from mice exposed to the attenuated Y. pestis strain Kim53170110, which did not confer protection to mice (**Figure 1B**), failed to demonstrate inhibition of Y. pestis growth ex vivo (**Figure 3C**). Notably, inspection of sera by anti-F1 ELISA clearly demonstrated that the induction of humoral immunity against the immunogenic F1 capsular antigen of Y. pestis, which is known to provide anti-plague protection, was observed only on day 8 p.i. with EV76 (anti-F1 IgG titer of 1:1250). Thus, the data suggested that at 24 h p.i. with EV76, the murine serum contained an innate-derived antibacterial component that inhibited bacterial growth.

To define the kinetics of the antibacterial agent, the mice were immunized with EV76, and serum was collected at different time points post-immunization for ex vivo Y. pestis growth assays. Between 1 and 6 h p.i., the serum did not exhibit any anti-growth inhibitory effects (**Figure 4A**). Inhibition of bacterial growth was observed beginning with serum collected 24 h p.i., and it persisted until day 13. The maximal growth

*(Continued)*

#### FIGURE 2 | Continued

(100 CFU) together with saline (upper panel) or after s.c. injection of *Y. pestis* Kim53-lux (100 CFU) together with 1 × 10<sup>7</sup> CFU of the EV76 strain (EV76-immunized, lower panel), as schematically described in Figure 1A. (B) Mice were infected s.c. with the protective EV76 strain, the virulent Kim53 strain (C) or both (D), as described above. At different time points post-infection, the bacterial loads in the draining lymph nodes, spleen and blood were determined; *n*= 3 to *n* = 11 mice per group. Values represent the total bacterial loads in organs (CFU/organ) or the bacterial concentrations in blood (CFU/ml). The limit of detection was 3 CFU. RAx-LN (right axillary lymph node), LAx-LN (left axillary lymph node), RI-LN (right inguinal lymph node), LI-LN (left inguinal lymph node), M-LN (mediastinal lymph node), spleen, and blood. Each dot indicates the value quantified in one animal; bars represent geometric means.

determined after 48 h of incubation. In all of the experiments, the data depict the mean and the standard error of the mean (SEM) of at least 3 individual sera of at least two independent experiments. Statistical significance was measured using Student's unpaired *t*-test with log-transformed values (\*\*\**p* < 0.001, and ns, not significant).

inhibitory effect was observed in serum collected at 48 h and 72 h p.i and the antibacterial activity was found to be highly potent and active even at high dilution (**Figure 4B**). As depicted in **Figure 4C**, proteinase K digestion completely eliminated the growth inhibitory effect of the serum, indicating that the antibacterial activity might be attributed to a proteinaceous component.

#### Identification of the Host Proteins Hemopexin and Transferrin in the Serum Fraction That Inhibits Bacterial Growth

To identify the antibacterial component in the circulation of EV76-immunized mice, serum samples exhibiting Y. pestis inhibitory activity were subjected to HiTrapQ HP anion

exchange chromatography fractionations (**Figure 5A**). The eluted fractions were collected, and their antimicrobial properties were assessed by ex vivo Y. pestis growth assays (**Figure 5B**). Fraction 26 exhibited the maximal growth inhibitory effect. This fraction contained two protein bands visible by silver-stained SDS-PAGE analysis (**Figure 5C**, left panel). Mass spectrometric tryptic-protein fingerprinting analysis established that the two proteins were host-borne hemopexin (a heme-sequestering protein) and transferrin (an iron-sequestering protein). Using specific antibodies, the identities of these proteins were further confirmed by Western blot analysis (**Figure 5C**, right panels). Thus, the proteins that seemed to be associated with the inhibitory activity present in the serum of the immunized animals exhibited biological functions involved in iron limitation.

To directly investigate the possibility that the antibacterial activity of serum derived from EV76-immunized mice resulted from iron starvation, the Y. pestis ex vivo serum-mediated growth inhibition assay was performed in the presence of increasing amounts of iron dextran. Growth inhibition was clearly reversed by iron dextran in a dose-dependent manner (**Figure 6A**).

#### Serum Hemopexin and Transferrin Levels Following EV76 Immunization

To monitor the expression levels of hemopexin and transferrin following the immunization of mice with EV76, serum samples were collected from EV76-inoculated mice at different time points post-immunization and were subjected to SDS-PAGE and Western blot analysis using specific anti-transferrin and anti-hemopexin antibodies. As shown in **Figure 6B**, 24 h after EV76 inoculation, hemopexin was dramatically upregulated as compare to control mice. A further increase in hemopexin was detected 48 h following immunization, and this expression level was sustained for at least 4 days. The hemopexin expression pattern strongly correlated with the appearance of antibacterial activity in EV76-immunized mouse serum ex vivo (**Figures 4A,B**). A basal level of transferrin was detected in the serum of non-immunized mice, and its expression level increased slightly upon immunization (**Figure 6B**). These results suggested that both transferrin and hemopexin were responsible for the antimicrobial activity, however, hemopexin was the major component induced in EV76-immunized mouse serum. The correlation between hemopexin protein levels and the rapid protective response against plague was further substantiated

by the observation that injection of the Kim53170110 strain (which did not protect mice from simultaneous challenge with the virulent strain; **Figure 1B**) did not induce the expression of hemopexin or transferrin in mouse serum (**Figure 6C**).

#### DISCUSSION

The live attenuated Y. pestis vaccine strain EV76 confers robust protection against plague due to its elicitation of strong acquired immune responses targeting Y. pestis key virulent factors and structural bacterial constituents, such as the F1 antigen (Walker et al., 1953; Tidhar et al., 2009). The establishment of an adaptive immune response is a relatively slow process, requiring several weeks to attain the amplitude needed for manifestation of its protective value. Since post-exposure prophylactic measures are highly relevant in the case of pathogens against which mass vaccination is not routinely conducted or in the case of bacterial pathogens potentially associated with intentional bioterror use, we inspected the effects of EV76 vaccination administered concomitantly with exposure of the experimental animals to a lethal dose of a fully virulent Y. pestis strain. Here, we provided evidence that upon s.c. immunization with a single dose of the EV76 live vaccine strain of Y. pestis, the mice acquired a protected state almost instantaneously that conferred high resistance to s.c. infection by a lethal Y. pestis strain. This effect was prevalent even when the live vaccine was administered several hours postinfection and was also effective when the mice were infected via pulmonary exposure only 2 days after immunization. Early induction of the protected state could not be attributed to an acquired response, which typically requires a much longer time to manifest (in the current study, a minimal 8-day period was measured for the detection of anti-F1 antibodies). To the best of our knowledge, this was the first time that such an immediate and effective stimulation of a protective response to plague by a live vaccine strain was shown. Furthermore, the effect appeared to be due to a particular ability of the EV76 strain because it could not be replicated by the administration of a commensurate dose of another attenuated Y. pestis strain. One may speculate that the specificity of the response could be related to the residual ability of EV76 to replicate and disseminate in the host.

Bacterial dissemination data obtained by IVIS-mediated visualization or by direct quantification of the bacterial load in the organs confirmed that the early innate protective response induced by the EV76 strain resulted in complete blockage of the spread of the virulent strain as well as profound limitation of its proliferation. Ex vivo assays showed that mouse sera collected 24 h after EV76 immunization (but not from control mice) exhibited a strong anti-Y. pestis growth inhibitory effect that was maximal at 48 h post-immunization and was maintained at least until the development of specific antibodies. To address the mechanistic basis for this early systemic antimicrobial activity induced by

from 0 to 4 days post-infection were inspected by Western blot analysis using anti-hemopexin, anti-transferrin or anti-β actin antibodies. (C) Pooled sera, collected

from mice 24 h post-infection with EV76 or Kim53170110, were examined by Western blot analysis using the indicated antibodies.

the EV76 strain, a protein anion exchange chromatographic fractionation of the sera collected from the immunized mice was implemented, enabling mass spectrometric identification of the two host proteins, hemopexin and transferrin, that co-eluted in the antibacterial active fractions. While both proteins appeared to be associated with the antibacterial activity, transferrin was present in control mice sera, and its level was only moderately increased following immunization with EV76. Hemopexin, in contrast, was barely detected in control mice sera, and its expression level was upregulated dramatically soon after EV76 immunization in strong correlation with the antimicrobial activity observed in the ex vivo assays with post-immunization sera.

The data therefore strongly suggested that the protective state induced by immunization could be attributed to the biological activities of the host iron and heme-binding proteins. Maintenance of low levels of free iron is a major host innate defense strategy that limits the growth of infectious bacteria and thus establishes nutritional immunity (Weinberg, 1975; Cassat and Skaar, 2013; Parrow et al., 2013; Elphinstone et al., 2016). Accordingly, pathogens have evolved mechanisms to overcome this limitation to gain access to the host iron supply by expressing siderophores and hemophores, which are able to exploit iron that is trapped by host proteins, such as transferrin and hemopexin (Cornelis, 2010; Cassat and Skaar, 2013; Parrow et al., 2013). In the case of Y. pestis, extensive studies have documented that deletion of the pigmentation (pgm) locus, which is composed of genes encoding for the siderophore-mediated iron acquisition system and hemin storage, is associated with significant virulence attenuation (Perry and Fetherston, 1997; Carniel, 2001). Notably, the EV76 strain itself owes its significantly attenuated phenotype to the absence of the pgm pathogenicity locus.

Transferrin was previously shown to participate in host nutritional immunity, which can prevent the availability of free iron in the organism (Sridhar et al., 2000). Induction of hemopexin in response to systemic infection was thought to protect infected host from heme-induced cell damage (Larsen et al., 2010; Medzhitov et al., 2012). Recently, hemopexin was also shown to limit the availability of heme in the context of an IL-22-induced response to systemic infection by enteropathogens (Sakamoto et al., 2017). Our observation that Y. pestis growth inhibition appeared to depend on the presence of transferrin and hemopexin and that these effects could be reversed by elevating the amounts of available iron is in good agreement with these studies. Moreover, while in previous studies, iron limitation was associated with the host response to bacterial challenge; our data indicated that this phenomenon, which may not be effectively activated in the case of plague infection, was exploited for the early manifestation of protection by a live attenuated vaccine.

The sequestration of iron promoted by hemopexin and transferrin could explain the early antibacterial effect induced by immunization with EV76. In fact, exogenously elevating the level of the serum iron-binding protein transferrin was suggested to represent a therapeutic approach to preventing iron acquisition by pathogens, such as Staphylococcus aureus, Acinetobacter, Candida and B. anthracis, consequently improving the survival of the infected host (Rooijakkers et al., 2010; Lin et al., 2014; Bruhn and Spellberg, 2015).

In this study, we showed for the first time that a Y. pestis live vaccine brings about a rapid and significant upregulation of the host hemopexin protein. Rapid elevation of hemopexin levels could, in turn, lead to effective heme withdrawal, which, together with preexisting transferrin, limits the availability of iron that is necessary for the growth and dissemination of the virulent strain. This rapid response was followed by the development of additional innate and adoptive immune responses in later stages of the infection. We speculate that the EV76-mediated rapid induction of hemopexin expression involves a bacterial factor that directly or indirectly interacts with host cells after immunization.

The data strongly suggest that the novel EV76-mediated mechanism described in this study may be relevant in the context of other bacterial infections. Yet, at this stage, it is early to speculate on the ability of other live attenuated vaccines to mount similar responses. Studies addressing these issues are currently being conducted in our laboratory.

This study provides the proof of principle for a new concept by which live attenuated vaccines have the potential to induce rapid, broad-range iron nutritional immunity, which could prove

#### REFERENCES


beneficial for post-exposure scenarios by enabling the extension of the time window necessary for the development of an acquired specific immune response or by providing additional countermeasures against antibiotic-resistant pathogens.

#### AUTHOR CONTRIBUTIONS

AZ and EM conceived and designed the study. AZ, YV, AT, MA, DG, YL, and EM conducted the experiments. AZ, EM, SR, and TC analyzed the data. AZ, EM, and TC wrote the paper.

#### FUNDING

Israel Institute for Biological Research funds, grant number SB-5112-81.

#### ACKNOWLEDGMENTS

We thank Professor H. Mobley for providing the plasmid used to generate the Kim53-lux derivative. We are deeply grateful to Dr. Mario Lebendiker from the Protein Purification Facility at the Hebrew University of Jerusalem for chromatographic fractionation of the active sera and to Dr. Tevie Mehlman from the Weizmann Institute of Science for the mass spectrometric analysis.


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

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

## Secreted Citrate Serves as Iron Carrier for the Marine Pathogen *Photobacterium damselae* subsp *damselae*

Miguel Balado<sup>1</sup> , Beatriz Puentes <sup>1</sup> , Lucía Couceiro<sup>1</sup> , Juan C. Fuentes-Monteverde<sup>2</sup> , Jaime Rodríguez <sup>2</sup> , Carlos R. Osorio<sup>1</sup> , Carlos Jiménez <sup>2</sup> and Manuel L. Lemos <sup>1</sup> \*

<sup>1</sup> Department of Microbiology and Parasitology, Institute of Aquaculture, University of Santiago de Compostela, Santiago de Compostela, Spain, <sup>2</sup> Department of Chemistry, Faculty of Sciences and Center for Advanced Scientific Research (CICA), University of A Coruña, A Coruña, Spain

Photobacterium damselae subsp damselae (Pdd) is a Vibrionaceae that has a wide pathogenic potential against many marine animals and also against humans. Some strains of this bacterium acquire iron through the siderophore vibrioferrin. However, there are virulent strains that do not produce vibrioferrin, but they still give a strong positive reaction in the CAS test for siderophore production. In an in silico search on the genome sequences of this type of strains we could not find any ORF which could be related to a siderophore system. To identify genes that could encode a siderophore-mediated iron acquisition system we used a mini-Tn10 transposon random mutagenesis approach. From more than 1,400 mutants examined, we could isolate a mutant (BP53) that showed a strong CAS reaction independently of the iron levels of the medium. In this mutant the transposon was inserted into the idh gene, which encodes an isocitrate dehydrogenase that participates in the tricarboxylic acid cycle. The mutant did not show any growth impairment in rich or minimal media, but it accumulated a noticeable amount of citrate (around 7 mM) in the culture medium, irrespective of the iron levels. The parental strain accumulated citrate, but in an iron-regulated fashion, being citrate levels 5–6 times higher under iron restricted conditions. In addition, a null mutant deficient in citrate synthase showed an impairment for growth at high concentrations of iron chelators, and showed almost no reaction in the CAS test. Chemical analysis by liquid chromatography of the iron-restricted culture supernatants resulted in a CAS-positive fraction with biological activity as siderophore. HPLC purification of that fraction yielded a pure compound which was identified as citrate from its MS and NMR spectral data. Although the production of another citrate-based compound with siderophore activity cannot be ruled out, our results suggest that Pdd secretes endogenous citrate and use it for iron scavenging from the cell environment.

Keywords: *Photobacterium damselae*, citrate, iron uptake, siderophores, vibrioferrin

#### *Edited by:*

Pierre Cornelis, Vrije Universiteit Brussel, Belgium

#### *Reviewed by:*

Iain Lamont, University of Otago, New Zealand Paul Edward Carlson, United States Food and Drug Administration, United States

> *\*Correspondence:* Manuel L. Lemos manuel.lemos@usc.es

*Received:* 31 May 2017 *Accepted:* 26 July 2017 *Published:* 08 August 2017

#### *Citation:*

Balado M, Puentes B, Couceiro L, Fuentes-Monteverde JC, Rodríguez J, Osorio CR, Jiménez C and Lemos ML (2017) Secreted Citrate Serves as Iron Carrier for the Marine Pathogen Photobacterium damselae subsp damselae. Front. Cell. Infect. Microbiol. 7:361.

doi: 10.3389/fcimb.2017.00361

#### INTRODUCTION

Iron restriction is an important host defense strategy, thus successful pathogens must possess mechanisms to acquire iron from host sources in order to cause disease. Although there are several strategies for iron acquisition, heme uptake and siderophore-based systems are certainly two of the most widespread among pathogenic bacteria and the most relevant to virulence (Hider and Kong, 2010; Cassat and Skaar, 2013; Saha et al., 2013; Ellermann and Arthur, 2017).

Photobacterium damselae is a member of Vibrionaceae that has been divided in two different subspecies: subsp. damselae (hereafter Pdd) and subsp. piscicida (Pdp). Pdd is an autochthonous marine bacterium resident in both natural and fish farms ecosystems that can behave as an opportunistic pathogen for fish and mammals (Labella et al., 2017). It is also a primary emerging pathogen causing haemorraghic septicemia in a variety of marine species including wild and cultured fish, sharks, crustaceans and marine mammals. In addition, Pdd can cause infections in humans due to exposure to marine fish, seawater or raw seafood (Rivas et al., 2013). In some of the reported human cases, the infection progressed into a severe necrotizing fasciitis leading to a fatal outcome (Yamane et al., 2004). In this bacterium it was demonstrated a positive correlation between iron availability in host fluids and degree of virulence (Fouz et al., 1997). Previous works proved that all Pdd strains are capable of utilizing hemoglobin and ferric ammonium citrate as the sole iron sources and that they also have the ability to produce siderophores (Fouz et al., 1994, 1997). It was postulated that the siderophore produced by Pdd could be a hydroxamate type different from aerobactin and desferal (Fouz et al., 1997), but the precise nature of the siderophores produced was not elucidated up to date.

In previous works, we described the presence of a functional heme uptake system in Pdd (Rio et al., 2005), but at present nothing is known about the siderophore-based systems used by Pdd to get iron from its hosts. In a recent work about Pdd proteome variations under diverse iron conditions (Puentes et al., 2017), we showed that some Pdd strains are likely able to produce the siderophore vibrioferrin, that was firstly described as the main siderophore of Vibrio parahaemolyticus and V. alginolyticus (Yamamoto et al., 1994). We showed that some Pdd strains express proteins with high similarity to the main proteins involved in the biosynthesis and transport of vibrioferrin in these vibrios (Puentes et al., 2017). However, many other pathogenic Pdd strains, including the reference strain CIP 102761 (ATCC 33539), do not express those proteins and lack the gene clusters involved in vibrioferrin production, but they still display a CAS reactivity, suggesting the existence of an alternative siderophore in these strains.

In this work we show that most pathogenic strains of Pdd secrete endogenous citrate and use it for iron scavenging from the cell environment.

### MATERIALS AND METHODS

#### Strains, Media, and Reagents

Pdd strains were routinely grown at 25◦C on Tryptone Soy Broth or Agar (Cultimed) supplemented with 1% NaCl (TSB-1 or TSA-1 respectively). Differential iron availability conditions were achieved by adding to CM9 minimal medium (Lemos et al., 1988) final concentrations of 10 µM Fe2(SO4)<sup>3</sup> (high iron conditions), or different concentrations of the non-assimilable and non-toxic iron chelator 2,2′ -dipyridyl (TCI): 40 µM to get low iron conditions or 75 µM for very low iron conditions. The addition of 2,2′ -dipyridyl to culture media mimics the low iron conditions found in the host environment. Strains and plasmids used are listed in **Table 1**.

#### Growth under Iron Limiting Conditions and Test of Siderophore Production

To test the ability of Pdd strains to grow under iron limiting conditions, overnight cultures in LB were adjusted to an OD<sup>600</sup> of 0.5 and diluted 1:100 in CM9 minimal medium containing the iron chelator 2,2′ -dipyridyl at 40 or 75 µM. 2,2′ -dipyridyl (TCI) was dissolved in distilled water to prepare a stock solution at 20 mM that was added to the sterile media at appropriated concentrations. When required, ferric-ammonium citrate (Panreac) was added to CM9 medium at 10 µM final concentration. Cultures were incubated at 25◦C with shaking at 150 rpm, and OD<sup>600</sup> was measured after 12 h of incubation. Siderophore production was measured using the colorimetric liquid assay of the Chrome-Azurol-S (CAS) dye, as previously described (Schwyn and Neilands, 1987; Balado et al., 2015, 2017b). For CAS-reactivity assays, strains were grown at 40 µM 2,2′ -dipyridyl to allow enough growth to make iron-chelating

TABLE 1 | Strains and plasmids used in this work.


activity detectable. A non-inoculated CM9 sample with the same 2,2′ -dipyridyl concentration was used as spectrophotometric blank and as negative control for CAS liquid assays. Growth curves and CAS assays were carried out in triplicate, and results shown are the means of three independent experiments.

#### Cross-Feeding Assays

Bioassays were designed to detect the production of vibrioferrin and to detect the production of other iron-chelating molecules. To detect vibrioferrin synthesis we used a mutant (AR13) of Vibrio alginolyticus deficient in pvsA, a gene involved in the synthesis of this siderophore (Osorio et al., 2015). For the detection of biological active supernatants we used strain RM71. The indicator strains were inoculated into CM9 plates containing the iron chelator 2,2′ -dipyridyl at 100 µM, a concentration at which they are unable to growth. Strains to be tested were cultured on LB agar plates, cells were harvested with a sterile loop, placed on top of the indicator strain plates and incubated at 25◦C for 48 h. The results were scored as positive when a growth halo of the indicator strains was visible around cells. A disc containing 10 µl of a solution of Fe2(SO4)310 mM was used as positive growth control.

#### DNA Purification and Analysis and PCR

Total genomic DNA from Pdd was purified with the Easy– DNA kit (Invitrogen). Plasmid DNA purification and extraction of DNA from agarose gels were carried out using kits from Fermentas (Thermo–Fisher). PCR reactions were routinely carried out in a T–Gradient Thermal Cycler (Biometra), with Taq polymerase KAPA Taq (Kapa Biosystems). Since all Pdd strains with genomic sequences available lack vibrioferrin genes, detection of pvsD gene by PCR was performed using degenerated oligonucleotides (**Table 2**) designed from conserved pvsD sequences from several Vibrio species aligned with the ClustalW Multiple Alignment tool of BioEdit program (http://www.mbio.ncsu.edu/BioEdit/bioedit.html). The choosen sequences were those of pvsD from V. parahaemolyticus (accession No: AB082123.1; GI: 23307114), V. alginolyticus (accession No: DQ201184.2; GI: 120564760), V. splendidus (accession No: NC\_011744.2; GI: 294510242), V. caribbenthicus

TABLE 2 | Oligonucleotides used in this work for mutant construction and detection of pvsD gene by PCR.


(accession No: EFP94847.1; GI: 309367283) and V. harveyi (accession No: EMR36193.1; GI: 471343428).

#### Mutagenesis by Mini-Tn*10* Transposon

A mini-Tn10 insertion library was constructed in Pdd strain RM71 using the conjugative suicide plasmid pLOF/Km, carrying a mini-Tn10 transposon with an isopropyl-beta-Dthiogalactopyranoside (IPTG)-inducible transposase located outside the mobile element, and a kanamycin-resistant gene flanked by transposable elements (Herrero et al., 1990; Rivas et al., 2015). Conjugation of pLOF/Km into Pdd RM71 was achieved my mixing equal amounts of the E. coli S17(pLOF/Km) (previously cultured in LB at 37◦C for 5 h) and RM71-Rif (rifampicin-resistant) (cultured in TSB-1 at 25◦C for 12 h). One mL of the mixture plus 10 µL IPTG was deposited in Marine Agar (Difco) plates and incubated at 25◦C for 72 h. Appropriate dilutions of the growing cells were then plated on LB agar containing kanamycin (50 µg/mL) and rifampicin (50 µg/mL) and incubated at 25◦C for 48 h to select for Pdd transformants. An Km<sup>R</sup> Rif<sup>R</sup> clone was considered a mini-Tn10 insertion mutant. Growth and CAS-reactivity for each mutant was tested in 96-wells microtiter plates containing 100 µL of CM9 with 25 µM 2,2′ -dipyridyl. Sequencing of DNA fragments flanking transposon sequences in the clones with expected phenotypes was achieved by partial digestion of genomic DNA with restriction enzyme BfuCI (New England Biolabs) for 35 s. The restriction products were ligated into plasmid pUC118 previously treated with BamHI and alkaline phosphatase. The ligation products were transformed into E. coli DH5α, and clones were selected in LB with Km 50 µg/mL. pUC118-cloned inserts containing the kanamycin resistance gene of mini-Tn10 plus flanking chromosomal DNA were purified using the GeneJET Plasmid Miniprep Kit (Thermo Fisher Scientific) and finally sequenced using a primer targeted to the extreme of the kanamycin cassette (TCCAGTTTACTT TGCAGGGC). DNA sequences were obtained using a capillary DNA Sequencer ABI 3730xl (Applied Biosystems), and then analyzed by BLAST (Altschul et al., 1990) to identify the insertion points.

#### Allelic Exchange Mutagenesis

In-frame deletion of gltA gene in Pdd RM71 was constructed by allelic exchange using the nucleotide sequence of locus A0J47\_03095, protein with accession No. ODA26781.1, from Pdd RM71 genome sequence. PCR amplifications of two fragments of the gene and flanking regions, when ligated together result in an in–frame (nonpolar) deletion. Primers used are listed in **Table 2**. The construction was ligated into the suicide vector pKNG101 (Kaniga et al., 1991). The resulting plasmids were mated from E. coli S17-1-λpir (Herrero et al., 1990) into Pdd RM71 wild type strain (Rif<sup>R</sup> ) and exconjugants with the plasmid (Streptomycin<sup>R</sup> ) integrated in the chromosome by homologous recombination were selected. A second recombination event was obtained by selecting for sucrose (15%) resistance and further checking for plasmid loss and for allelic exchange, generating the mutant RM711gltA. Deletion of the parental gene was checked by DNA sequencing of the region involved to ensure that mutation was in-frame.

#### Citrate Quantification

Bacterial cultures under high- and low-iron conditions in early exponential growth (OD<sup>600</sup> = 0.6) were centrifuged (5 min at 8,000 × g) and filtered through 0.22 µm pore size to obtain a cell-free supernatant. The citrate concentration in these supernatants was measured using a colorimetric Citrate Assay Kit (Sigma-Aldrich) according to the manufacturer protocol. To remove oxaloacetate and pyruvate background a blank sample was included for each sample, omitting the Citrate Enzyme Mix provided by the kit, according to the manufacturer instructions. The citrate concentration standard curve was made with increasing concentrations (ranging from 0.1 to 12.5 mM) of sodium citrate (Merck) in CM9 medium. Concentrations of citrate lower than 0.1 mM gave reactions that were indistinguishable from CM9 alone. Three biological replicates were measured in duplicate.

#### Isolation and Identification of Citrate from Culture Supernatants

Pdd strain RM71 was cultured in CM9 minimal medium supplemented with 35 µM 2,2′ -dipyridyl. Cultures were carried out in 2 L Erlenmeyer flasks containing 1 L of medium. These flasks were inoculated with 20 mL of a fresh culture in TSB-1. Flasks were incubated during 24 h at 25◦C with continuous shaking at 190 rpm. After incubation, bacterial cells were pelleted by centrifugation at 10,000 rpm in a Beckman J-21 high speed centrifuge. The supernatant was filtered through a continuous filtration cartridge with a 0.45 µm pore size membrane (Millipore) to remove any cell and organic debris from the supernatant. Siderophore activity present was evaluated by the CAS method as above. Equal volumes of supernatants and CAS solution were mixed and the absorbance at 630 nm was measured in a spectrophotometer after 20 min. Siderophore activity of fractions and purified compounds were monitored by bioassays as described above.

A 6-L batch of centrifuged cell-free culture broth was divided into five portions which were loaded onto a XAD-4 resin (Radius Column: 2.8 cm, resin mass: 120 g). After washing with distilled water (1.3 L) with a flow rate of 0.8 mL/min, the resin was eluted with a methanol/water (1:1) mixture followed by methanol. The siderophore and CAS activities remained in the fraction eluted with water which was then lyophilized to give 86.2 g of a solid. This solid was washed with MeOH (4 times × 290 mL) to yield, after removal of the solvent, 54.53 g of a siderophore/CAS active non-methanol-soluble white solid. The material was dissolved in H2O and chromatographed on a Sephadex LH-20 column which was eluted with a 9:1 mixture of H2O in MeOH at a flow rate of 1.8 mL/min. The collected fractions were submitted to CAS assay and the fractions of interest were concentrated under vacuum to afford a CAS-positive fraction (37.6 g). Part of that fraction (21.0 g) was divided in three batches and chromatographed on a Sephadex <sup>R</sup> G-25 Fine column which was eluted with 760 mL of H2O using a flow rate of 4.5 mL/min. The eluted fractions were submitted to CAS assay and a chloride test for the presence of salts. Fractions displaying an intense color change with the CAS reagent and with a relatively low amount of salts were pooled and lyophilized. The lyophilized material (400 mg) was dissolved in H2O and chromatographed on a Sephadex <sup>R</sup> G-10 column using H2O (87 mL) as eluent at a flow rate of 1.6 mL/min. Again, the eluted fractions were submitted to the same tests as before to afford 53.5 mg of a salt-free and CAS– positive fraction. Final purification of this fraction was achieved by HPLC using a Discovery <sup>R</sup> HS F5 (100 × 4.6 mm, 5 µm) column with a mobile phase consisting of an isocratic mixture of 0.3% of CH3CN in H2O (each containing 0.1% HCO2H) for 1.4 min and then, a gradient from 0.3 to 6.0% of CH3CN in H2O (each containing 0.1% HCO2H) over 15.8 min at a flow rate of 1.2 mL/min. The CAS-positive fractions, eluted with a retention time of 2.98 min, were pooled and concentrated in vacuo to provide 2.1 mg of compound **1** which was identified as citrate by its <sup>1</sup>H NMR and (–)-ESIMS data. Compound **1**: <sup>1</sup>H NMR (500 MHz in D2O) δ (ppm): 3.00 (2H, d, J = 15.7, 2H, 2-H), 2.84 (2H, d, J =15.7, 2H, 4-H); (–)-ESIMS m/z 191 [M – H]−.

### RESULTS

In order to identify the siderophore system in Pdd, two strategies were used simultaneously: (1) identification of genes that could be related to siderophore biosynthesis and uptake by screening a transposon-generated mutant collection; and (2) isolation of compounds with siderophore activity by HPLC fractionation of iron-deprived cell-free supernatants and further chemical characterization of the fractions that retained the biological activity.

#### Only Some *Pdd* Strains Harbor Vibrioferrin Biosynthetic Genes

In previous works we have shown that some strains of Pdd, when cultured under iron limitation, express several proteins likely involved in the synthesis of the siderophore vibrioferrin (Puentes et al., 2017). In order to confirm the presence of vibrioferrin related genes, we checked by PCR the presence of pvsD, one of the main genes involved in the biosynthesis of this siderophore in other vibrios (Tanabe et al., 2003; Wang et al., 2007), in a collection of Pdd strains. The results are shown in **Table 3**. Some strains isolated from different hosts and different origins harbor the pvsD gene and give a strong reaction in the CAS test. Vibrioferrin production by these strains was assessed by a growth promotion bioassay using the V. alginolyticus AR13 mutant defective in vibrioferrin biosynthesis (Osorio et al., 2015). However, as seen in **Table 3**, there are many other virulent strains from different origins that clearly lack pvsD and do not produce vibrioferrin, but they still give a positive reaction in the CAS test for siderophore production, although with lower intensity.

#### Search of Candidate Genes Encoding a Siderophore System in Vibrioferrin-Negative Strains

Pdd RM71 strain was chosen to decipher the putative siderophore-based iron uptake system present in Pdd strains lacking vibrioferrin system. This strain has been widely used


TABLE 3 | Results of the PCR assays to detect the presence of pvsD and results of the CAS test and bioassays to detect vibrioferrin production in a collection of Pdd strains.

<sup>a</sup>Reaction in CAS liquid test after growth in CM9 plus 40 µM 2,2′ -dypiridyl; +, A<sup>630</sup> < −0.6; ++, A<sup>630</sup> ranging from −0.6 to −0.8.

<sup>b</sup>Detected by growth promotion of Vibrio alginolyticus AR13 (Osorio et al., 2015).

in previous works of our research group to characterize diverse virulence factors (Rivas et al., 2011). To try to find candidate genes that could encode a siderophore-mediated iron acquisition system, we performed an in silico search on the genome sequence of strain RM71 (GenBank accession No. NZ\_LYBT01000000) and on the reference genome from strain CIP 102761 (GenBank accession No. NZ\_ADBS00000000). After this search, we were unable to find any ORF which could be potentially involved in siderophore synthesis or transport according to its homology to other siderophore-related genes.

Since in silico search gave negative results we performed a random mutagenesis approach in strain RM71, using the mini-Tn10 transposon, to isolate mutants with siderophore production inactivated or altered. Essentially, we sought mutants that, compared to parental strain, showed a growth deficiency or a negative CAS reaction when they were grown under iron deficiency conditions. We also looked at potentially deregulated mutants that were positive in CAS assay under iron excess conditions. Any of these phenotypes could be indicative of inactivation of a component of a siderophore-mediated iron uptake system.

From a total of 1,400 mutants screened with the above phenotypic criteria we could found one mutant (BP53) that showed a strong CAS-positive reaction under both irondeprivation and iron-excess conditions, which suggested an overproduction of iron-chelating compounds (**Figure 1**). Since the transposon library coverage was low (ca. 37%) we cannot discard the existence of other genes involved in iron uptake. Nucleotide sequence analysis revealed that mutant BP53 had the transposon inserted into idh gene (acc. No. ODA21722.1). This gene (named icg in Vibrio cholerae) encodes a NADPdependent isocitrate dehydrogenase that participates in the tricarboxylic acid (TCA) cycle and convert D-isocitrate to 2 oxoglutarate with the production of one molecule of CO2. Inactivation of icg in V. cholerae and other bacteria results in an accumulation of large amounts of citrate and a slight growth impairment compared to its parental strain, especially during the late exponential phase of growth (Lakshmi and Helling, 1976; Minato et al., 2013). However, it is noticeable that BP53 did not show any growth defect in TSA-1 or CM9 media under high, low or very low iron (see methods) conditions (**Figure 2**). Furthermore, a significant citrate accumulation in the supernatant (ranging from 6 to 8 mM) could be detected (**Figure 3**). Based in these findings we could hypothesize that BP53 mutant secretes endogenous citrate to the extracellular medium and that the CAS-positive reaction observed could be caused by citrate accumulation.

FIGURE 1 | CAS assay of cell-free supernatants obtained from RM71 and BP53 mutant cultures at OD<sup>600</sup> = 0.6 under high- and low-iron conditions. Each bar represents mean values from three replicates. Error bars indicate standard deviations. Color pictures show wells from a microtiter plate with representative CAS test results.

#### Extracellular Citrate Accumulation by *Pdd* Strains

To confirm presence of citrate in culture media of Pdd strains we measured citrate levels under different culture conditions using a colorimetric citrate assay kit. We showed that RM71 produces noticeable levels of citrate, around 5.5 mM, when cultured in

conditions. Vibrio anguillarum RV22 and E. coli K12 were used as controls. CAS reactivity of E. coli K12 is due to enterobactin production. Each bar represents mean values from three replicates. Error bars indicate standard deviations.

standard deviations.

CM9 plus 40 µM 2,2′ -dipyridyl, but under iron excess conditions the levels of extracellular citrate were under 1 mM (**Figure 3**). As CM9 does not include citrate in its formulation, the citrate present in the cell-free spent medium must come from bacterial cells synthesis. As comparative purposes we used E. coli K12 and Vibrio anguillarum RV22. The citrate amount present in these controls ranged from 0.8 to 1.3 mM and did not show any significant variation with iron availability. These data are in agreement with extracellular citrate amounts described for E. coli (Bennett et al., 2009). Thus, under iron-excess conditions the citrate amount found in RM71 supernatant (ca. 0.9 mM) is equivalent to the amount found in E. coli or V. anguillarum. Instead, citrate levels of RM71 increased 4- to 6-fold under low iron conditions. Hence, low levels of iron in the cell environment seem to trigger citrate synthesis and secretion in Pdd strains lacking the vibrioferrin system.

Interestingly, when citrate accumulation was measured in a vibrioferrrin-producing strain (strain RG91, see **Table 3**), the citrate levels were 5- to 6-fold lower (ca. 1.5 mM) and independent of the iron levels of the medium (**Figure 3**). This observation could be explained by the hypothesis that if cells get enough iron through the likely much more efficient vibrioferrin system, citrate levels are kept in the minimal amounts needed for TCA cycle functionality. Only when a high affinity siderophore is not present, citrate would be synthesized at high levels to act as iron carrier for the cell. To confirm this hypothesis extracellular citrate levels were measured in a vibrioferrindeficient mutant of RG91 lacking pvsD (RG911pvsD). The results showed that in this mutant citrate accumulation reached the levels of strain RM71 and were also iron-dependent. In this mutant, the CAS-reactivity levels were also equivalent to those of RM71. Further evaluation of citrate accumulation in the supernatants of other Pdd strains, including vibrioferrinproducing and non-producing strains (**Table 3**), showed that the described behavior for each type of strain is not restricted to RG91or RM71 strains, but it seems a common feature in Pdd strains (**Figure 4**): citrate accumulates extracellularly only when vibrioferrin is not produced. All this suggests that citrate is being secreted by Pdd strains to the medium in high amounts when no other siderophore-based iron acquisition system is available.

To further confirm that citrate is involved in iron uptake in Pdd, a mutant lacking citrate synthase (GltA) was constructed in strain RM71 (RM711gltA) and its phenotype was compared with parental strain RM71 and mutant BP53 (deficient in isocitrate deshidrogenase). All three strains showed identical growth curves under high iron conditions (**Figure 2**). When CM9 contained 40 µM 2,2′ -dipyridyl (low-iron conditions, 50% of 2,2′ -dipyridyl minimal inhibitory concentration, MIC) RM71, BP53, and RM711gltA also showed indistinguishable growth curves (**Figure 2**). However, under very low-iron conditions (75 µM 2,2′ -dipyridyl, 80% of MIC) RM711gltA mutant showed a growth which was 45% lower with respect to BP53 and RM71. Interestingly, when ferric-ammonium citrate was added to the medium, RM711gltA showed a growth curve identical to RM71 and BP53 strains (**Figure 2**). Therefore, RM711gltA showed a significant decrease in the growth ability under iron deprivation conditions when the concentration of 2,2′ -dipyridyl

is close to the MIC of this iron chelator for the parental strain RM71.

#### Citrate as Siderophore for *Pdd*

The RM71 supernatant produces a CAS-positive reaction under low-iron, being the absorbance between 5- and 7-fold lower than in iron-excess conditions, which denotes that, as expected, siderophore activity is stronger under iron-deprived conditions. The iron homeostasis is strictly controlled by Fur repression, therefore the genes involved in iron uptake, like siderophore synthesis, are expressed only under iron starvation (Fillat, 2014; Porcheron and Dozois, 2015). As previously mentioned, BP53 showed a strong CAS-positive reaction under both iron-restricted and iron-excess conditions which denotes that idh inactivation causes an overproduction of siderophore. Interestingly, the supernatants from the RM711gltA mutant showed a CAS-negative reaction in both high- and low-iron conditions (**Figure 3**) indicating that no detectable amount of siderophores was produced. Therefore, inactivation of citrate synthesis in RM71 abolishes siderophore production.

To further confirm that citrate serves as iron source for Pdd, we used the stretonigrin sensitivity method. Streptonigrin is a metal-dependent quinone-containing antibiotic which at high values of intracellular iron (Fe2<sup>+</sup> and Fe3+) increases its bactericidal action by the formation of reactive oxygen radicals (Yeowell and White, 1982). So that, the presence in the cell of an active iron-uptake mechanism implies a growth defect when streptonigrin is added to the medium. In our case, while under iron limitation streptonigrin 10 µM was sufficient to inhibit the growth of RM71 and BP53 strains, the growth of RM711gltA was inhibited only when streptonigrin was above 75 µM (**Figure 5**), indicating that no iron was entering into the cell. This result shows that blocking the endogenous citrate synthesis in Pdd RM71 prevents siderophore production, resulting in a decrease

of growth under extreme iron deprivation. Altogether our results clearly suggest that citrate or a citrate-like compound serves as an iron scavenging system in RM71.

#### Identification of CAS-Reactive Compounds in Culture Supernatants

To identify compounds with siderophore activity secreted by strain RM71, the cells were grown under iron restricted conditions (CM9 with 35 µM 2,2′ -dipyridyl) and the spent medium was subjected to chemical analysis using the CAS reactivity and biological activity as purification guides. The cell-free supernatants were fractionated by successive column chromatography (Amberlite XAD4 resin and size exclusion on Sephadex LH-20, G-25, and G-10) to give a CAS reactive and bioassay active fraction (**Figure 6**). From a final HPLC purification of this fraction we obtained a pure compound which spectral data (NMR, MS) were coincident to those of citrate (**Figure 7**). Hence, the CAS-reactive molecule produced by Pdd strain RM71 was unequivocally identified as citrate, and no other molecules with siderophore activity could be identified in the supernatants.

#### DISCUSSION

Citrate is produced by most microorganisms via the TCA cycle (Breusch, 1943) and, although it is not a powerful chelator, it binds Fe(III) forming a ferric-dicitrate complex (Silva et al., 2009). In this form, citrate can be used by many bacteria, including Pdd, as a source of iron (Pressler et al., 1988; Mazoy and Lemos, 1991; Fouz et al., 1994; Mazoy et al., 1997). Besides, citrate is a common moiety of many polycarboxilic siderophores (Hider and Kong, 2010).

Although, at neutral pH, citrate has lower affinity for iron than conventional siderophores, it is considered a highaffinity iron carrier, and it produces a CAS-positive reaction at concentrations above 0.1 mM (Frawley et al., 2013). Endogenous citrate secretion was described previously in other pathogenic bacteria and it is associated to intracellular iron homeostasis. However, while most bacteria can use externally supplied ferric

citrate to fulfill their nutritional requirement for iron, there are two examples of bacteria which secrete citrate in order to get iron: Pseudomonas syringae (Jones and Wildermuth, 2011) and Bradyrhizobium japonicum (Guerinot et al., 1990). This property depends on the existence of a secretion system that secretes citrate under appropriate conditions. In Salmonella typhimurium it was described a major facilitator superfamily pump (IceT) that regulates intracellular iron content by Fe-citrate secretion (Frawley et al., 2013). in silico searches on Pdd genomes did not detect IceT homologs. Whether citrate is host or endogenous derived, bacterial cells require an energy-dependent transporter to acquire it, with the likely participation of outer membrane transporters from the TonB family (Ogierman and Braun, 2003). In E. coli and Pseudomonas aeruginosa this role is played by outer membrane protein FecA (Pressler et al., 1988; Marshall

et al., 2009). Although up to five TonB-dependent transporters are present in Pdd genomes, a FecA homolog seems to be absent. However, a FecB homolog, a periplasmic transporter which is part of the ferric-citrate transport system in E. coli (Banerjee et al., 2016), has been previously identified as a protein expressed under low iron conditions in Pdd (Puentes et al., 2017). More work is needed to clarify the ferric-citrate transport in this bacterium.

Besides acting as iron chelator, citrate seems to play different roles. In S. typhimurium it was reported the existence of a stressinduced citrate secretion that would reduce the intracellular iron amount and arrest growth, increasing tolerance to antibiotics and to reactive oxygen species produced by the host defenses (Frawley et al., 2013).

It is known that TCA cycle is suppressed in anaerobiosis or in media containing glucose and that it is activated in presence of oxygen and acetate (Park et al., 1994). Fur is a negative regulator of iron acquisition systems that also exert a regulatory role in energy metabolism (Thompson et al., 2002). In addition, Fur regulates some TCA cycle steps. While citrate synthase is repressed by the Fe2+-Fur complex, the Aconitase A (that converts citrate into isocitrate) is induced by Fe2+- Fur, suggesting that some bacteria, like E. coli, respond to iron restriction by producing citrate (Thompson et al., 2002; McHugh et al., 2003).

Our results indicate that inactivation of TCA cycle observed in Pdd BP53 mutant results in a dramatic increase of citrate in the extracellular medium but it did not affect growth in glucose-containing media. These findings clearly suggest the necessary existence of a citrate secretion system that prevents intracellular citrate accumulation. Overproduction of citrate is deleterious to most bacteria. It was proved that cytoplasmic citrate accumulation causes a growth defect and promotes a strong selection pressure for mutations inactivating citrate synthase. Such mutations restore the growth potential in mutant strains that accumulate citrate. This was demonstrated for many microorganisms such as E. coli (Gruer et al., 1997), Corynebacterium glutamicum (Baumgart et al., 2011), Sinorhizobium meliloti (Koziol et al., 2009), Vibrio fischeri (Septer et al., 2015), and Bacillus subtilis (Pechter et al., 2013). Nevertheless, the accumulation of citrate seems to have no negative effect on Pdd growth, suggesting that most citrate excess must be secreted to the extracellular medium to avoid intracellular accumulation. In contrast, inactivation of citrate synthase in Pdd produces a significant decrease in growth capacity under extreme iron-restriction. Therefore, citrate secretion by Pdd has a positive effect on cell fitness and could be used for extracellular high-affinity iron scavenging, which suggests that Pdd likely synthesizes a citrate excess to use it for iron uptake from the cell surroundings. This is reinforced by the observation that a vibrioferrin-producing Pdd strain does not accumulate citrate, but when the vibrioferrin system is inactivacted, the mutant behaves like a strain lacking the vibrioferrin system, and accumulate extracellular citrate in an iron-dependent fashion (**Figure 3**). One plausible explanation of this result is that since citrate is a precursor for vibrioferrin (Tanabe et al., 2003), mutation of a vibrioferrin biosynthetic gene would result in the accumulation of precursors, and in this case citrate would be exported out of the cell to avoid toxicity. The cell would then use external citrate as an alternative iron carrier.

The advantages of using citrate or vibrioferrin for iron uptake are unclear. From the results described here we could deduce that vibrioferrin use would be a better strategy since all strains that produce this siderophore do not accumulate citrate. However, the vibrioferrin affinity constant for iron is not particularly high, with respect to citrate (Amin et al., 2009; Silva et al., 2009). Furthermore, and despite that the information available is scarce, seems that there are not clear differences in the virulence degree between the two types of strains (data not shown). In addition, it is likely that the use of citrate could be an ancestral mechanism to get iron since Pdd and Pdp subspecies share a common evolutionary origin (Balado et al., 2017a), and we have some evidence that Pdp seems to use also secreted citrate for iron uptake. Some strains of each subspecies could have later acquired by horizontal transfer additional siderophore-based mechanisms for iron uptake, like the plasmid-encoded piscibactin in Pdp (Osorio et al., 2015) or vibrioferrin in Pdd, resulting in the inactivation of the use of secreted citrate for iron uptake.

As conclusion, this work shows that although some strains of the marine pathogen Pdd produce vibrioferrin as siderophore,

#### REFERENCES


most pathogenic strains release endogenous citrate to the extracellular environment in response to iron deprivation, and that this trait have a positive effect in the cell fitness when it grows under extreme iron-restricted conditions. Although the production of another citrate-based or other compounds with siderophore activity cannot be completely discarded, our results suggest that endogenous citrate, besides being part of other siderophores, can be itself used for iron uptake by Pdd.

#### AUTHOR CONTRIBUTIONS

MB, CO, CJ, and ML conceived and designed the experiments. BP, LC, MB, and JF performed the laboratory experiments. CO and JR made substantial contributions to conception and design. MB, LC, CJ, and ML participated in experimental design and data analysis. MB, CJ, and ML drafted and wrote the manuscript. All authors read and approved the final manuscript.

#### ACKNOWLEDGMENTS

This work was supported by grants AGL2012-39274-C02-01/02 and AGL2015-63740-C2-1/2-R (AEI/FEDER, EU) from the State Agency for Research (AEI) of Spain, and co-funded by the FEDER Programme from the European Union. The support of Xunta de Galicia (Spain) with grant GRC-2014/007 is also acknowledged.


changes and novel potential virulence factors. Vet. Microbiol. 201, 257–264. doi: 10.1016/j.vetmic.2017.01.040


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

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

## Iron Acquisition Strategies of Vibrio anguillarum

Yingjie Li 1, 2 and Qingjun Ma1, 2 \*

*<sup>1</sup> Key Laboratory of Experimental Marine Biology, Institute of Oceanology, Chinese Academy of Sciences, Qingdao, China, <sup>2</sup> Laboratory for Marine Biology and Biotechnology, Qingdao National Laboratory for Marine Science and Technology, Qingdao, China*

The hemorrhagic septicemic disease vibriosis caused by *Vibrio anguillarum* shows noticeable similarities to invasive septicemia in humans, and in this case, the *V. anguillarum*–host system has the potential to serve as a model for understanding native eukaryotic host–pathogen interactions. Iron acquisition, as a fierce battle occurring between pathogenic *V. anguillarum* and the fish host, is a pivotal step for virulence. In this article, advances in defining the roles of iron uptake pathways in growth and virulence of *V. anguillarum* have been summarized, divided into five aspects, including siderophore biosynthesis and secretion, iron uptake, iron release, and regulation of iron uptake. Understanding the molecular mechanisms of iron acquisition will have important implications for the pathogenicity of this organism.

Keywords: Vibrio anguillarum, siderophore biosynthesis, siderophore secretion, iron uptake, iron release, iron acquisition mechanism

#### INTRODUCTION

#### Edited by:

*Pierre Cornelis, Vrije Universiteit Brussel, Belgium*

#### Reviewed by:

*Manuel L. Lemos, Universidade de Santiago de Compostela, Spain Isabelle Jeanne Schalk, Centre National de la Recherche Scientifique (CNRS), France*

\*Correspondence:

*Qingjun Ma qma@qdio.ac.cn*

Received: *03 May 2017* Accepted: *11 July 2017* Published: *25 July 2017*

#### Citation:

*Li Y and Ma Q (2017) Iron Acquisition Strategies of Vibrio anguillarum. Front. Cell. Infect. Microbiol. 7:342. doi: 10.3389/fcimb.2017.00342* The Gram-negative bacterium Vibrio anguillarum is a pathogen that causes vibriosis with lethal hemorrhagic septicemia in aquatic animals worldwide (Toranzo et al., 2005). Although up to 23 O serotypes of V. anguillarum are identified in the European serotyping system, with most serotypes encompassing free-living environmental strains (Pedersen et al., 1999), only serotypes O1, O2, and partial O3 are found to be implicated in vibriosis outbreaks (Toranzo et al., 2005). Many studies have been performed in an attempt to understand the virulence mechanism in V. anguillarum. Several main virulence factors have been recognized by using genetic approaches, including iron acquisition components (Naka and Crosa, 2011), hemolysins (Hirono et al., 1996; Rodkhum et al., 2005; Rock and Nelson, 2006; Li et al., 2008; Xu et al., 2011; Mou et al., 2013), metalloproteases (Milton et al., 1992; Yang et al., 2007; Varina et al., 2008; Mo et al., 2010), chemotaxis and motility (O'Toole et al., 1999; Ormonde et al., 2000), exopolysaccharides (Croxatto et al., 2007), and lipopolysaccharides (Welch and Crosa, 2005). Among them, iron uptake systems are a critical component for infection of the host fish leading to disease (Wolf and Crosa, 1986).

V. anguillarum, like most other organisms, has an absolute requirement for iron to synthesize a large number of crucial enzymes, which are involved in many fundamental cellular processes, such as cytochromes for cell respiration, ribonucleotide reductase for the biosynthesis of DNA precursors, and enzymes for the tricarboxylic acid (TCA) cycle (Crosa et al., 2004). However, due to the low solubility of iron (∼10−<sup>18</sup> M) at physiological pH in aerobic environments, ferric iron mainly forms insoluble hydroxides, whereas a cytoplasmic iron concentration of ∼10−<sup>6</sup> M is required for bacterial growth (Hantke, 1981). Therefore, iron is suggested to be the growthlimiting factor in ocean environments (Martin et al., 1991). To respond to this selective pressure, bacteria have evolved numerous mechanisms for iron acquisition, including transport of iron from the mammalian iron carriers, transferrin and heme, and synthesis of small ferric iron-binding molecules, known as siderophores. Some of these iron transport systems are conserved among all Vibrio species, reflecting their common ancestry, while other acquisition systems appear to have been developed by horizontal transfer, such as the anguibactin transport system that is mainly specific to V. anguillarum. It has been shown that V. anguillarum harbors a number of genes encoding for iron uptake and regulation, which are essential for its virulence beyond simple iron chelation (Lemos and Osorio, 2007; Naka et al., 2013b). In this article, we describe the developments in understanding the molecular mechanisms of iron acquisition systems in V. anguillarum, divided into the following aspects: siderophore biosynthesis and secretion, iron uptake, iron release, and regulation of iron uptake.

#### SIDEROPHORE BIOSYNTHESIS

Two different siderophore-dependent systems have been identified in V. anguillarum strains. One is mediated by a 65 kb pJM1 plasmid, which contains most of the genes encoding for biosynthesis and transport proteins of the siderophore anguibactin (Naka et al., 2013b). This anguibactin system is only found in pathogenic plasmid-bearing strains of serotype O1. The other system, existing in all serotype O2 strains tested thus far, and some plasmidless serotype O1 strains, synthesizes a catecholated-type siderophore, vanchrobactin (Alice et al., 2005; Balado et al., 2006).

#### Biosynthesis of Anguibactin

The structure of anguibactin is unique in containing both catechol and hydroxamate metal-chelating functional groups (Actis et al., 1986), derived from 2,3-dihydroxybenzoic acid (DHBA) and N-hydro-histamine, respectively. Utilizing the chorismate, plasmid-carrying V. anguillarum strain 775 generates a repertoire of molecules through a ribosomeindependent process and finally synthesizes anguibactin. This biosynthesis is controlled by a number of genetic determinants. To date, more than 10 different genes have been described and yield anguibactin-related phenotypes in plasmid-carrying V. anguillarum strains when mutated by genetic approaches (Table S1). Most are located on the plasmid pJM1 or pJM1-like plasmids while some are on the chromosomes (**Figure 1**). First, chorismate is catalyzed stepwise by a series of proteins, AngC/VabC (isochrorismate synthase), AngB/VabB (isochrorismatase; Du et al., 2017), and VabA (2,3-dihydro-2,3-dehydroxybenzoate dehydrogenase), to synthesize DHBA (**Figure 2**; Alice et al., 2005; Balado et al., 2008). In later steps, phosphopantetheine transferase AngD is required to transfer a phosphopantetheinyl moiety to a serine residue of AngB and AngM (Balado et al., 2008). AngE/VabE (2,3-dehydroxybenzoate-AMP ligase) activates DHBA to form acyl adenylate and further transfers it to the free thiol of the phosphopantetheine AngB (Liu et al., 2004; Alice et al., 2005). AngC, AngB, and AngE are pJM1-encoded proteins while VabC, VabB, VabA, and VabE are chromosomally encoded with VabC, VabB, and VabE showing functional redundancy with AngC, AngB, and AngE, respectively. AngN catalyzes this DHBA thioester to combine with cysteine, which is activated by AngR and tethered by AngM, thereby producing a dihydroxyphenylthiazoline-thioester. AngH (histidine decarboxylase; Tolmasky et al., 1995; Barancin et al., 1998) and possibly AngU (Naka et al., 2013b) modify the histidine to form N-hydro-histamine, which is then transferred to dihydroxyphenyl-thiazoline-thioester to yield anguibactin. It is interesting to note that homologs of these plasmid-located genes are all present on the chromosome of Vibrio harveyi, which is also able to produce anguibactin, suggesting that the plasmid-mediated anguibactin system might originate from V. harveyi or vice versa (Naka et al., 2013a,b) .

#### Biosynthesis of Vanchrobactin

Plasmidless O1 strains and those belonging to a number of other V. anguillarum serotypes synthesize a chromosomemediated siderophore, vanchrobactin (Lemos et al., 1988; Soengas et al., 2008). In a similar way to anguibactin synthesis, DHBA of these V. anguillarum strains is also produced from chorismate by the sequential activities of VabC, VabB, and VabA. However, how vanchrobactin is synthesized from the DHBA precursor remains obscure, even though some genes have been found to be indispensable for this process, including vabB, vabD, vabE, and vabF (Balado et al., 2006, 2008). According to well-studied pathways for synthesis of anguibactin and vibriobactin, a siderophore produced by Vibrio cholerae, late steps for vanchrobactin formation in V. anguillarum have been proposed during which DHBA is assembled (**Figure 3**; Balado et al., 2006). Specifically, VabD contributes to transfer of a phosphopantetheinyl moiety to the aryl carrier domain of VabB and the peptidyl carrier domain of VabF, respectively. Like anguibactin formation, VabE activates DHBA and arginine to yield acyl adenylates, and then delivers them to VabB and VabF, respectively. The DHBA-VabB is combined with arginine by the condensation domain of VabF to form (2,3 dihydroxybenzoyl)argininate. Finally, the condensation domain of VabF in (2,3-dihydroxybenzoyl)argininate-VabF may be loaded with VabE-activated serine, which adheres to the peptidyl carrier domain of VabF to produce vanchrobactin. Although similar roles of AngB and VabB are proposed in these processes, the aryl carrier protein (ArCP) domain of VabB is not able to complement the function of the AngB ArCP domain (Di Lorenzo et al., 2011), suggesting different but unknown roles may occur between AngB and VabB during anguibactin and vanchrobactin formation.

### SIDEROPHORE EXPORT

Siderophore secretion is an essential step in iron uptake, yet the mechanisms of this process remain largely unknown. Two siderophore export systems have been found so far, including the ATP-dependent efflux pump and the major facilitator superfamily protein (MFS)-mediated efflux pump. PvdRT-OpmQ is the first ATP-dependent export system to be identified in Pseudomonas aeruginosa (reviewed by Schalk and Guillon, 2013). Schalk and colleagues found that this system exports not only the newly synthesized mature siderophore

pyoverdine but also pyoverdine that has already delivered iron into the bacterium (Hannauer et al., 2010; Yeterian et al., 2010). In addition, PvdRT-OpmQ can secrete unwanted metalpyoverdine complexes into the periplasm of P. aeruginosa (Hannauer et al., 2012). By using the respective protein sequences from P. aeruginosa as a query in BLASTP analysis, only genes encoding for PvdR and PvdT are identified (Table S2) while the ompQ gene is absent in the genomes of the sequenced V. anguillarum strains, suggesting that the MFS system rather than PvdRT-OpmQ may participate in siderophore secretion in V. anguillarum.

The secretion of enterobactin in Escherichia coli is the beststudied paradigm of siderophore export via the MFS system (Horiyama and Nishino, 2014). Based on this, a proposed siderophore-export pathway is depicted in **Figure 4**, and putative genes involved in this process are listed in Table S2. First, siderophores are exported to the periplasm from the cytoplasm via the MSF VabS, a homolog of E. coli EntS that has been shown to transport enterobactin across the cytoplasmic membrane (Furrer et al., 2002). Subsequently, the resistance-nodulationcell division (RND) family proteins, which have been found to play a role in multidrug resistance in many microbes including V. cholerae strains (Rahman et al., 2007; Bina et al., 2008), capture the periplasmic siderophores and secrete them to the environment via the outer membrane channel TolC. In agreement with this model, a RND efflux system recently described in V. cholerae plays an essential role in maintenance of cellular homeostasis by secreting the siderophore vibriobactin (Kunkle et al., 2017), which again indicates that a MFS-mediated efflux pump system might be used for siderophore export in V. anguillarum.

### IRON UPTAKE

V. anguillarum strains contain several iron transport systems to sequester the different sources of iron, including anguibactin or vanchrobactin, heme, free Fe2+, free Fe3+, and ferrichrome, which are summarized in **Figure 5**.

#### Ferrisiderophore Import

Once siderophores are produced and exported to the environment, they capture iron to form a ferric-siderophore complex, which is recognized by a specific transporter on the surface of the outer membrane. In V. anguillarum, ferricanguibactin is translocated across the outer membrane via its specific transporter FatA, which is essential for ferric-anguibactin uptake (Walter et al., 1983; Actis et al., 1988; Lopez and Crosa, 2007; Lopez et al., 2007). This process requires the TonB2 system for energy transmission, which originates from the proton-motive force of the inner membrane (Stork et al., 2004).

The TonB2 complex is located across the inner membrane and comprises TonB2, ExbB2, ExbD2, and TtpC (TonB2 complexassociated transport protein C), all of which are indispensable for ferric-anguibactin import to the periplasmic space. Deletion of either of these genes completely abolishes ferric-anguibactin uptake, and they are thereby considered essential virulence factors for V. anguillarum since ferrisiderophore transport during iron uptake is a critical step for virulence (Stork et al., 2004, 2007). Notably, TtpC which shows homology to the TolR protein of E. coli, has been identified to be part of the TonB2 system in several Vibrio species, including V. anguillarum (Stork et al., 2007), V. cholerae (Stork et al., 2007), Vibrio alginolyticus (Wang et al., 2008), Vibrio parahaemolyticus (Kuehl and Crosa,

2010), and Vibrio vulnificus (Kuehl and Crosa, 2009). Despite a high similarity among all pathogenic vibrios studied thus far, the TtpC proteins likely play an important role in specific iron transport mediated by the TonB2 system, a conclusion drawn based on the observation that V. anguillarum 1ttpC cannot be complemented by the TtpC protein from V. cholerae (Stork et al., 2007; Kuehl and Crosa, 2009).

In addition to anguibactin transport, the TonB2 system is also involved in the uptake of vanchrobactin and the xenosiderophore (siderophore produced by other organisms) enterobactin. The transport of vanchrobactin and enterobactin is mediated by another outer membrane transporter, FvtA, which displays vanchrobactin-dependent expression (Balado et al., 2008, 2009; Naka et al., 2008). In the serotype O2 V. anguillarum strain RV22, loss of fvtA leads to impaired growth under iron-limiting conditions due to incapability of vanchrobactin transport in cells (Balado et al., 2009). Besides FvtA, during transport of the xenosiderophore enterobactin, it seems another transporter, FetA, is present that shows specific binding to enterobactin, and the expression of FetA is regulated by FetR (Naka and Crosa, 2012). This fetA-fetR cluster in V. anguillarum is speculated to derive from V. cholerae based on similar molecular organization and high similarity (Naka and Crosa, 2012).

Subsequently, when combined with the periplasmic lipoprotein FatB, ferric-anguibactin passes through the cytoplasmic membrane by using an ATP-binding cassette (ABC) transporter, which includes inner membrane permeases consisting of a heterodimer of FatC and FatD, and an ATPase FatE (Köster et al., 1991; Actis et al., 1995; Naka et al., 2010, 2013c). FatBCD are required for ferric-anguibactin transport while a fatE mutant is still capable of ferricanguibactin transport due to the presence of its homolog FvtE (Naka et al., 2013c). Double deletion of fatE and fvtE significantly impairs ferric-anguibactin uptake (Naka et al., 2013c). Similarly, the passage of vanchrobactin through the inner membrane is also achieved in a stepwise manner by a periplasmic protein, FvtB, and an ABC transporter, FvtCDE, which together are essential for ferric-vanchrobactin or ferricenterobactin import; deletion of either protein causes a defect in ferric-vanchrobactin and ferric-enterobactin transport (Naka et al., 2013c). Different from the specific involvement of FatE in the uptake of anguibactin, FvtE can transport both

anguibactin and vanchrobactin/enterobactin (Naka et al., 2013c).

It is noteworthy that although fatDCBA encoding for ferricanguibactin transport are located on the pJM1-type plasmid in serotype O1 strains, they are also present on the chromosomes of plasmidless vanchrobactin-synthesizing strains, albeit fatA is inactivated by an insertion sequence (Bay et al., 2007; Balado et al., 2009). In fact, gene clusters coding for vanchrobactin synthesis and uptake are ubiquitous in both anguibactinand vanchrobactin-producing V. anguillarum strains (Balado et al., 2009), while plasmid-harboring strains do not produce vanchrobactin due to an inactivated chromosomal vabF gene (**Figure 1**; Naka et al., 2008). This is caused by the presence of the insertion sequence RS1, observed in all serotype O1 strains carrying a pJM1-type plasmid, and when RS1 is removed, pJM1 containing strains can simultaneously produce the siderophore vanchrobactin (Naka et al., 2008). However, because anguibactin appears to have a higher affinity for iron than does vanchrobactin, it is suggested to be the primary siderophore for iron import even though vanchrobactin is produced in the cell (Naka et al., 2008). Nevertheless, the minimal inhibitory concentration (MIC) values for iron chelators in plasmid-bearing strains seem to be lower than those of plasmidless V. anguillarum strains, which might be caused by higher stoichiometry of vanchrobactin in the plasmidless strains (Conchas et al., 1991).

#### Heme Uptake

Heme usually exists abundantly but not freely as a chemical compound in host cells. Therefore, bacteria secrete exotoxins such as cytolysins, proteases, or hemolysins to release it for further uptake. In V. anguillarum, several hemolysin genes have been identified and some of them are involved in V. anguillarum virulence (Hirono et al., 1996; Rodkhum et al., 2005; Rock and Nelson, 2006; Li et al., 2008, 2013; Xu et al., 2011) . After its release, free heme is captured by a putative outer membrane transporter, HuvA, which has been shown to be essential for heme uptake because the huvA mutant does not grow when heme is used as the sole iron source (Mazoy et al., 2003). However, compared with the wild type, the huvA mutant still maintains heme-binding activity under the conditions tested (Mazoy et al., 2003), indicating that HuvA is not the only protein capable of binding heme. In line with this, two heme-binding proteins

protein; RND, resistance-nodulation-cell division.

with molecular masses of 39 and 37 kDa were isolated from V. anguillarum serotype O1 and O2 strains, respectively, which are completely different from the 79 kDa HuvA protein (Mazoy and Lemos, 1996b; Mazoy et al., 1996). Therefore, Mazoy et al. speculate that, besides HuvA, additional proteins are probably present in V. anguillarum and function in heme binding, but not in its transport (Mazoy et al., 2003; Lemos and Osorio, 2007). Like siderophore import, this process is also energy-dependent, where HuvA is energized by TonB systems. The difference is that not only is the TonB2 complex involved in heme utilization, but a TonB1 system composed of TonB1-ExbB1-ExbD1 is also available for energy supply (Stork et al., 2004). Deletion of all genes for these two TonB systems leads to a complete defect in heme uptake and thus avirulence to the host, suggesting a key role of TonB systems in V. anguillarum virulence (Stork et al., 2004). The recent structural studies of the Ton complex from E. coli provide a mechanistic insight into this complex (Celia et al., 2016). It is proposed that the functional unit of the Ton complex contains an ExbB pentamer, an ExbD dimer, and at least one TonB. Electrophysiology experiments suggest that the ExbB-ExbD forms pH-sensitive channels, by which the Ton complex likely harnesses the proton-motive force for energy production and transduction.

Bioinformatic studies indicate that genes of the two TonB systems exist ubiquitously among all Vibrio species, and moreover, a third TonB system is even observed in some vibrios and other marine organisms (Kuehl and Crosa, 2010; Kustusch et al., 2011). However, it remains unclear why so many TonB systems are present in the Vibrio species. The V. anguillarum strains lack HuvA but contain an alternative heme transporter, HuvS, which is able to restore heme transport in a huvA mutant (Mouriño et al., 2005). The observation that huvS and huvA possess similar flanking DNA sequences implies that horizontal transmission and recombination might have occurred and thus be responsible for this genetic diversity (Mouriño et al., 2005).

Furthermore, the huvA gene is located in a gene cluster coding for nine heme uptake-related proteins, including HuvA, HuvZ, HuvX, TonB1, ExbB1, ExbD1, HuvB, HuvC, and HuvD (Mouriño et al., 2004). The periplasmic heme-binding protein HuvB delivers periplasmic heme to an inner membrane complex consisting of a permease, HuvC, and an ATPase, HuvD, which subsequently transport heme into the cytosol (Mouriño et al., 2004). Therefore, HuvBCD are required for heme transport, and deletion of either gene results in heme transport deficiencies (Mouriño et al., 2004). HuvZ also plays an import role in heme uptake, and loss of huvZ severely affects the growth of cells when heme serves as the sole iron source (Mouriño et al., 2004). Little is known about the function of HuvZ in heme utilization, and in V. cholerae it is suggested to have a role in heme storage (Wyckoff et al., 2004). HuvX, a predicted intracellular heme delivery protein in V. cholerae (Sekine et al., 2016), is not required for heme uptake in the V. anguillarum 775 plasmidless avirulent strain because deletion of huvX does not cause obvious differences in growth and heme utilization compared with the wild type (Mouriño et al., 2004). However, it is still under debate whether heme utilization is indeed involved in V. anguillarum virulence in nature.

#### Other Iron Acquisition Systems

Besides the anguibactin/vanchrobactin and heme uptake systems, four operons encoding putative iron transport systems, including transport of unchelated ferrous (feoABC) and ferric iron (fbpABC1 and fbpABC2), and siderophore ferrichrome transport (fhuABCD), have been identified in the genomes of V. anguillarum strains (**Figures 1**, **5**). The presence of different iron transport systems in V. anguillarum probably results from differences in growth conditions because siderophores can only promote growth under limited iron conditions that must be insufficient for iron uptake under all environmental conditions. In line with this, all the vibrios examined have been shown to have additional iron acquisition systems (Table S3; Payne et al., 2016).

The ferrous iron transporter FeoABC is speculated to be the most ancient iron transport system and widely found among bacterial species including Vibrio species. In V. cholerae, all feoABC genes are required for ferrous iron uptake although their functions have not been fully characterized (Wyckoff et al., 2006; Weaver et al., 2013). However, it is still unknown how ferrous iron passes through the outer membrane for transport by the Feo system in the periplasm. FbpABC, a ferric iron transporter, is also found in vibrios and has been shown to promote better growth at alkaline pH in V. cholerae (Peng et al., 2016). FhuABCD, responsible for siderophore ferrichrome utilization, have been demonstrated to be required for ferrichrome utilization in V. parahaemolyticus and V. cholerae (Rogers et al., 2000; Funahashi et al., 2009).

Notably, although it has been demonstrated that V. anguillarum is capable of using ferric citrate as the only iron source by a siderophore-independent mechanism (Mazoy et al., 1997), we do not find any genes involved in ferric citrate

transport in the genome of V. anguillarum or other vibrios. In P. aeruginosa, iron delivered by citrate is suggested to enter the cell as Fe2+, and FeoB is further required for citrate-mediated Fe2<sup>+</sup> uptake (Marshall et al., 2009). Therefore, it is plausible that FeoB of V. anguillarum may also play a role in Fe2<sup>+</sup> uptake from ferric citrate to maintain iron homeostasis.

#### IRON RELEASE FROM FERRISIDEROPHORES

When iron–siderophore complexes are transported into the cell cytoplasm, bacteria can use different strategies to release iron:


Genes predicted for iron reduction and siderophore degradation occur in the genome of V. anguillarum, and proposed pathways are shown in **Figure 5**. VabH, a putative cytoplasmic esterase, exhibits homology to E. coli Fes, which can hydrolyze enterobactin during iron release (Brickman and McIntosh, 1992). Therefore, VabH may serve as a vanchrobactin degradation enzyme during ferric-vanchrobactin compound dissociation. Similarly, a putative thioesterase gene, angT, may be involved in iron release from ferric-anguibactin. However, deletion of vabH or angT does not completely block siderophore uptake (Wertheimer et al., 1999; Balado et al., 2006), indicating that additional pathways for iron release may occur. In accordance with this, a nicotinamide adenine dinucleotide phosphate (NADPH)-dependent ferric reductase has been identified in the genomes of V. anguillarum strains. It shows high homology to the reported siderophore-interacting proteins (SIP) YqjH of E. coli and FscN of Thermobifida fusca, which are suggested to participate in iron reduction during ferrisiderophore dissociation (Miethke et al., 2011; Li et al., 2015). Furthermore, Mazoy and Lemos observed that ferric reductase activities of cell fractions are significantly increased in the presence of NADPH compared to its absence (Mazoy and Lemos, 1996a), indicating a functional NADPHdependent ferric reductase may occur in V. anguillarum. In addition, they also found it is only in the cytoplasmic, but not in periplasmic or membrane faction, where ferric reductase activity is stimulated under iron-limiting conditions (Mazoy and Lemos, 1996a). This implies iron release from ferrisiderophore might happen in the cytoplasm. Further investigation of the role of SIP in the ferrisiderophore dissociation pathway will provide more detail to elucidate the strategies deployed by V. anguillarum to release iron for utilization.

#### REGULATION OF IRON TRANSPORT

As excess iron is lethal and may lead to oxidative damage to DNA when free Fe2<sup>+</sup> reacts with hydrogen peroxide via the Fenton reaction (Imlay, 2002), tight regulation of iron transport is a prerequisite to meet, but not exceed, the requirement for iron. To date, several regulators have been identified in V. anguillarum to control the uptake of the iron, including the negative regulators Fur and an antisense RNA (RNAα), and the positive regulators AngR, TAFr, and anguibactin.

In Gram-negative bacteria, Fur is the major global regulator of iron metabolism, which serves as a sensor of intracellular iron concentration. It can bind to a Fe2+-bound dimer at a specific site, termed the Fur box, in the promoter region and thereby negatively regulate the transcription of corresponding genes (Troxell and Hassan, 2013). Therefore, it is not surprising that the Fur protein of V. anguillarum blocks or depresses the expression of most genes involved in iron acquisition systems, such as those coding for anguibactin and vanchrobactin synthesis (Salinas and Crosa, 1995; Chen et al., 1996; Di Lorenzo et al., 2004; Alice et al., 2005; Balado et al., 2008), TonB systems (Mouriño et al., 2006), and iron transport systems (Waldbeser et al., 1993; Tolmasky et al., 1994; Chen and Crosa, 1996; Chai et al., 1998; Mouriño et al., 2006; Balado et al., 2008; Naka and Crosa, 2012). In addition to Fur, an antisense RNA, RNAα, is capable of negatively modulating the expression of the fatA and fatB genes involved in iron transport by specifically binding to fatA and fatB mRNAs and thus repressing their transcription under iron-rich conditions (Waldbeser et al., 1995; Chen and Crosa, 1996). Moreover, the Fur protein is also crucial for RNAα synthesis and regulates its transcription initiation, which is independent of the iron status of the cell (Chen and Crosa, 1996).

AngR is a bifunctional protein, involved not only in the formation of anguibactin but also in the positive regulation of transport and biosynthesis genes (Salinas et al., 1989; Singer et al., 1991). Although AngR consists of two helix-turn-helix (HTH) regulatory motifs, only the first HTH is essential for gene regulation, as demonstrated by the finding that modulation of an angR deletion mutant is restored by a construct containing a frame shift and leaving only the first HTH motif (Wertheimer et al., 1999). These data again suggest that AngR plays an important role in both anguibactin synthesis and regulation of gene expression. The transcription of genes including the iron transport biosynthesis operon (ITBO; **Figure 1**) and angN is tightly controlled by AngR, and they display the highest expression level when iron is limited (Actis et al., 1995; Chen and Crosa, 1996; Di Lorenzo et al., 2008). The TAF (trans-acting factor) region shown in **Figure 1** is essential for anguibactin biosynthesis and for maximal expression of the ITBO genes, which are achieved by two separate entities: one involved in anguibactin biosynthesis (TAFb) and the other in regulation (TAFr) (Tolmasky et al., 1988; Welch et al., 2000). Studies have demonstrated that TAFr and AngR work in a synergistic manner to modulate the level of anguibactin synthesis under ironlimiting conditions (Salinas et al., 1989; Salinas and Crosa, 1995). In addition, anguibactin by itself is also capable of increasing the transcriptional level of the ITBO, which reaches the highest level when AngR, TAFr, and anguibactin are all acting synergistically (Chen and Crosa, 1996).

Besides iron, many other environmental factors have been identified in vibrios, including V. anguillarum, that affect the expression of genes associated with iron transport, such as oxygen, temperature, carbon sources, and quorum sensing molecules (Mou et al., 2013; Payne et al., 2016). Therefore, an expanded search for sensors involved in other environmental signals may lead to a more complete picture of the strategies of iron regulation in V. anguillarum.

#### OUTLOOK

Despite the challenges of iron acquisition in various environments, such as competition with other organisms on host surfaces, or sequestration by the high-affinity ironbinding host proteins lactoferrin and transferrin, growth of V. anguillarum is proficient in both host and marine habitats, achieved by multiple iron transport systems. Moreover, there is increasing evidence for the position of iron uptake, especially siderophores, at the crux of the microbial infection process, thereby reducing the appeal of siderophores as antimicrobial targets. In this context, it is important to investigate the whole of the iron acquisition systems of V. anguillarum, which will not only provide new insights to explain the evolutionary origin of versatile iron transport systems but also supply more evidence to understand the pathogenicity of this organism.

However, despite the exciting developments in understanding siderophore synthesis and ferrisiderophore transport in some vibrios, such as V. anguillarum and V. cholerae, there are many unsolved aspects of the iron uptake systems. For example, the mechanisms that govern the synthesis of anguibactin or vanchrobactin are still not fully elucidated, and how proteins are regulated for anguibactin or vanchrobactin synthesis is not yet well understood. How is mature siderophore secreted to the environment, and how is the iron released to the periplasm or cytoplasm from the ferric-siderophore complex? To answer these questions, an effective way is required to monitor anguibactin or vanchrobactin in real time. Finally, the different iron uptake systems that are responsible for the utilization of diverse iron sources will need to be established at the genetic level in order to uncover the relationship between iron uptake and virulence.

#### AUTHOR CONTRIBUTIONS

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

#### FUNDING

YL was supported by the China Postdoctoral Science Foundation (no. 2016M602200). QM was supported by grants from the "1000 Talents Program" of China, "100 Talents Program" of the Chinese Academy of Sciences, and "AoShan Talents Program" of Qingdao National Laboratory for Marine Science and Technology (no. 2015ASTP).

#### ACKNOWLEDGMENTS

Wenteng Xu, Yellow Sea Fisheries Research Institute, Chinese Academy of Fishery Sciences, is greatly acknowledged for language proofing.

#### REFERENCES


#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fcimb. 2017.00342/full#supplementary-material

Vibrio anguillarum strains of different origins. Appl. Environ. Microbiol. 57, 2956–2962.


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

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

## Iron Acquisition Mechanisms and Their Role in the Virulence of *Burkholderia* Species

Aaron T. Butt and Mark S. Thomas\*

*Department of Infection, Immunity and Cardiovascular Disease, Faculty of Medicine, Dentistry and Health, University of Sheffield, Sheffield, United Kingdom*

*Burkholderia* is a genus within the β*-Proteobacteriaceae* that contains at least 90 validly named species which can be found in a diverse range of environments. A number of pathogenic species occur within the genus. These include *Burkholderia cenocepacia* and *Burkholderia multivorans,* opportunistic pathogens that can infect the lungs of patients with cystic fibrosis, and are members of the *Burkholderia cepacia* complex (Bcc). *Burkholderia pseudomallei* is also an opportunistic pathogen, but in contrast to Bcc species it causes the tropical human disease melioidosis, while its close relative *Burkholderia mallei* is the causative agent of glanders in horses. For these pathogens to survive within a host and cause disease they must be able to acquire iron. This chemical element is essential for nearly all living organisms due to its important role in many enzymes and metabolic processes. In the mammalian host, the amount of accessible free iron is negligible due to the low solubility of the metal ion in its higher oxidation state and the tight binding of this element by host proteins such as ferritin and lactoferrin. As with other pathogenic bacteria, *Burkholderia* species have evolved an array of iron acquisition mechanisms with which to capture iron from the host environment. These mechanisms include the production and utilization of siderophores and the possession of a haem uptake system. Here, we summarize the known mechanisms of iron acquisition in pathogenic *Burkholderia* species and discuss the evidence for their importance in the context of virulence and the establishment of infection in the host. We have also carried out an extensive bioinformatic analysis to identify which siderophores are produced by each *Burkholderia* species that is pathogenic to humans.

*Edited by:*

*Pierre Cornelis, Vrije Universiteit Brussel, Belgium*

#### *Reviewed by:*

*Vittorio Venturi, International Centre for Genetic Engineering and Biotechnology, India Jonathan Mark Warawa, University of Louisville, United States*

> *\*Correspondence: Mark S. Thomas m.s.thomas@shef.ac.uk*

*Received: 18 August 2017 Accepted: 18 October 2017 Published: 06 November 2017*

#### *Citation:*

*Butt AT and Thomas MS (2017) Iron Acquisition Mechanisms and Their Role in the Virulence of Burkholderia Species. Front. Cell. Infect. Microbiol. 7:460. doi: 10.3389/fcimb.2017.00460* Keywords: *Burkholderia,* iron, siderophores, haem uptake, cystic fibrosis, melioidosis

#### THE ROLE OF IRON IN BACTERIA AND SOURCES OF IRON WITHIN THE HOST

Iron mainly occurs in either of two oxidation states in biological systems, Fe2<sup>+</sup> and Fe3<sup>+</sup> (also referred to as Fe(II) and Fe(III) or ferrous and ferric, respectively), the latter being the oxidized form that also prevails in the earth's crust, whereas the former is favored by low pH and low oxygen concentrations (Sanchez et al., 2017). It is the ability of iron to be interconverted between these two states that is the basis of many redox reactions that occur in cells (Andrews et al., 2003). For almost all species of bacteria, iron is essential as it is an important component of many proteins. It may occur as part of the haem cofactor, as in cytochromes and haem-type catalases, or as an iron-sulfur center, as in ferredoxins, rubredoxins, nitrogenase, sulfite reductase, and other iron-sulfur proteins, or as mono- or dinuclear non-haem iron that occurs in Fe-dependent superoxide dismutase and in class Ia ribonucleotide reductases, respectively (Caza and Kronstad, 2013). However, despite the relatively high iron content within humans and animals, it is not freely available due to sequestration by proteins that include hemoglobin, transferrin, lactoferrin, and ferritin, and the fact that it is largely present in the intracellular compartment (Skaar, 2010). This presents a problem to pathogenic microbes that demands the possession of high affinity iron capturing systems if they are to cope with the otherwise bacteriostatic environment. The reader is referred to the following reviews for a more comprehensive discussion of this subject (Nairz et al., 2010; Skaar, 2010; Parrow et al., 2013; Runyen-Janecky, 2013).

#### THE GENUS *BURKHOLDERIA*

Burkholderia is a genus within the β-Proteobacteriaceae that contains at least 90 validly named species but will almost certainly include many more (Depoorter et al., 2016). Members of the genus are diverse and may be found as free-living species within soil or water, or in association with other hosts, including plants, fungi, animals and humans (Smith et al., 1995; Parke and Gurian-Sherman, 2001). They contain large genomes in the range of 7–9 Mb that are typically organized into two or three chromosomes. Based on 16S rRNA sequences, two major clades account for almost all of the currently described species (the endosymbionts B. rhizoxinica and B. endofungorum being the two exceptions; **Figure 1**). One clade consists of a large group of environmental and plant-associated species referred to as the Burkholderia xenovorans group, together with the deeper branching species B. caryophylli, B. soli, and B. symbiotica. The other clade consists of two large groups [the B. glathei group and the B. cepacia complex (or Bcc)] and two smaller groups (the Burkholderia pseudomallei group and a plant pathogenic group consisting of Burkholderia gladioli, B. glumae, and B. plantarii) (Depoorter et al., 2016). More recently, most of the non-pathogenic species (i.e., the B. xenovorans and B. glathei groups along with B. caryophylli, B. soli, and B. symbiotica) have been transferred to the new genus Paraburkholderia (Sawana et al., 2014; Oren and Garrity, 2015) and subsequently the B. glathei group has been transferred to the new genus Caballeronia (Dobritsa and Samadpour, 2016; **Figure 1**), although it is not clear whether the new classification schemes will be accepted by the scientific community (Depoorter et al., 2016). For the purposes of this review we will refer to all species as belonging to the genus Burkholderia (whichever classification scheme is adopted, the pathogenic species will remain within the genus Burkholderia). Of the pathogenic species, members of the Bcc and B. pseudomallei group can cause life-threatening infections in humans, and this feature will be discussed in this review in the context of their iron acquisition mechanisms. The phytopathogen B. gladioli also causes opportunistic infections in humans, but as little is known concerning its iron acquisition mechanisms it will not be discussed in detail.

#### BCC INFECTIONS AND CYSTIC FIBROSIS

The Bcc constitute a group of at least 20 closely related species within the genus (**Table 1**). Members of the Bcc are well-known for causing infections in the lungs of cystic fibrosis (CF) patients, although they are also associated with infections of patients with chronic granulomatous disease and in individuals who are compromised for other reasons (Song et al., 2011). Although almost all Bcc species have been recovered from CF patient sputum, the most prevalent species are B. cenocepacia and B. multivorans, and consequently they have been the subject of most studies on potential virulence mechanisms (Reik et al., 2005; Drevinek and Mahenthiralingam, 2010; Zlosnik et al., 2015). Infections with Bcc have variable outcomes and may include transient or chronic asymptomatic infections, or they may cause a rapid decline in lung function which in some cases is accompanied by bacteraemia leading to death of the patient ("Cepacia syndrome;" Isles et al., 1984; Mahenthiralingam et al., 2001; Courtney et al., 2004; Jones et al., 2004). Infections with Bcc species are extremely difficult to eradicate due to their high level of intrinsic resistance to many antibiotics and biocides (Rose et al., 2009; Rhodes and Schweizer, 2016) [Note: prior to 2005, bacteria described as B. cepacia (and as Pseudomonas cepacia prior to the proposal of the genus Burkholderia in 1992) largely included members of related species within the Bcc that were not recognized as such at the time (Yabuuchi et al., 1992; Lipuma, 2005). This needs to be borne in mind when considering the results from some of the earlier investigations described below, particularly those in which a large number of "P. cepacia" or "B. cepacia" isolates were analyzed].

Cystic fibrosis (CF) is the most common autosomal recessive disorder among Caucasians. It is caused by mutations to the gene encoding the CF transmembrane conductor regulator (CFTR), which primarily functions as a gated chloride ion transporter but also regulates other apical membrane ion transporters (Davies et al., 2007). This defect leads to a more viscous lung mucus that impairs the action of the mucociliary escalator (Matsui et al., 1998, 2005; Boucher, 2007). Moreover, the high viscosity of the mucosal secretions may hinder the access of secreted cationic antimicrobial peptides from submucosal glands to the epithelial surface and may also restrict migration of neutrophils (see Doring et al., 2011; Tang et al., 2014 for reviews). Additional effects of a defective CFTR are also likely to be at play in facilitating pathogen survival in the airway surface liquid (ASL) of the CF lung (reviewed in Doring and Gulbins, 2009; Tang et al., 2014; Elborn, 2016), including increased abundance of amino acids (Barth and Pitt, 1996; Thomas et al., 2000), lowered pH (Song et al., 2006; Yoon et al., 2006), increased neutrophilmediated oxidative stress (Kolpen et al., 2010), inflammation (Perez et al., 2007), hypoxic regions (Worlitzsch et al., 2002), and an altered iron status (see below). These phenomena conspire to make CF patients particularly susceptible to infection from a variety of bacterial, viral and fungal pathogens (Harrison, 2007).

Despite extensive research, the key virulence determinants of Bcc members that lead to establishment of an infection, persistence and morbidity in CF patients still remain to be established. Potential virulence determinants associated with the TABLE 1 | Siderophore biosynthesis in human pathogenic members of the genus *Burkholderia*<sup>a</sup> .


*<sup>a</sup>The potential of the listed species to produce each siderophore was deduced from a bioinformatic analysis of genome sequences using genes known to encode the biosynthesis of each siderophore as a search query (except for cepabactin where the biosynthetic genes remain to be identified). In some cases, the production (or not) of a siderophore by a specific strain has been demonstrated (see below and main text for details).*

*<sup>b</sup>Ornibactin production has been confirmed in B. ambifaria, B. cenocepacia, B. cepacia, and B. vietnamiensis (Stephan et al., 1993; Meyer et al., 1995; Barelmann et al., 1996; Darling et al., 1998; Agnoli et al., 2006).*

*<sup>c</sup>Presumed to be malleobactin E for B. pseudomallei and B. mallei based on the known structure of the B. thailandensis siderophore (Franke et al., 2015).*

*<sup>d</sup>Cepaciachelin production has only been confirmed in B. ambifaria (Meyer et al., 1989).*

*<sup>e</sup>B. cenocepacia, B. cepacia, B. paludis, and B. pseudomallei have been demonstrated to produce pyochelin whereas B. vietnamiensis isolates do not (Meyer et al., 1995; Darling et al., 1998; Alice et al., 2006; Kvitko et al., 2012; Ong et al., 2016).*

*<sup>f</sup> Cepabactin has been identified in culture supernatants from two B. cepacia environmental strains (ATCC 25416 and ATCC17759) but not in environmental or clinical isolates of B. vietnamiensis [including the type strain TVV75 (LMG 10929)] (Meyer et al., 1989, 1995). It was not detected in culture supernatants of B. cenocepacia clinical isolates K56-2 and 715j (Darling et al., 1998).*

*<sup>g</sup>Member species of the Bcc are as listed in Depoorter et al. (2016) with the addition of B. paludis (Ong et al., 2016) and a potential new member Bcc ATCC 31433 (Loveridge et al., 2017).*

*<sup>h</sup>B. ambifaria PHP7 (LMG 11351) and the type strain AMMD (LMG 19182) have been shown to produce cepaciachelin, whereas strains MC40-6, MEX-5 and IOP40-10 do not possess the required genes (this study; Barelmann et al., 1996; Esmaeel et al., 2016).*

*<sup>i</sup>Although, several B. multivorans strains encode the capacity to produce cepaciachelin, some (including ATCC 17616 and the type strain LMG 13010) do not (this study; Esmaeel et al., 2016).*

*<sup>j</sup>Cepaciachelin gene cluster is present in B. pseudomultivorans strain MSMB368 but not in other strains currently in the database.*

*<sup>k</sup>Bcc ATCC 31433 is closely related to B. ubonensis, but possibly constitutes a separate species (Loveridge et al., 2017).*

*<sup>l</sup>B. cepacia LK29 has the genetic capacity to produce cepaciachelin but other B. cepacia strains for which genome sequences are available do not (this study; Esmaeel et al., 2016). <sup>m</sup>B. cepacia ATCC 25416 (the type strain) and ATCC 17759 produce pyochelin whereas strain GG4 does not encode the capacity to produce this siderophore (Meyer et al., 1995; Deng et al., 2016; Esmaeel et al., 2016).*

*<sup>n</sup>B. paludis encodes the ferric ornibactin transport system but not the biosynthetic apparatus. The fact that it retains OrbE may suggest recycling of the siderophore.*

*<sup>o</sup>At the time of writing, three B. stabilis complete genome sequences had been deposited in the database. The type strain (ATCC BAA-67) and FERMP-21014 contain the pyochelin biosynthesis and utilization genes on chromosome 2, whereas strain LA20W lacks these genes and carries the cepaciachelin gene cluster.*

*<sup>p</sup>Siderophore status is unknown due to unavailability of genome sequence information.*

*<sup>q</sup>B. territorii A63 contains the pyochelin gene cluster. Other strains, including the type strain, do not have it.*

*<sup>r</sup>Pyochelin was not detected in culture supernatants of clinical and environmental isolates of B. vietnamiensis (Meyer et al., 1995).*

Bcc, including their ability to survive intracellularly (including within macrophages) (Lamothe and Valvano, 2008; Vergunst et al., 2010; Valvano, 2015; Mesureur et al., 2017), have been the subject of several comprehensive reviews and are not discussed here (Drevinek and Mahenthiralingam, 2010; Loutet and Valvano, 2010; Sousa et al., 2011). One of the problems in identifying the pathogenic mechanisms is that there are few candidate virulence determinants that are associated with all member species of the Bcc, and indeed even among different strains within the same species some these determinants may not be conserved. However, one trait that does appear to be required for virulence in Bcc species is the ability to acquire iron from iron depleted environments such as within the human host. A large number of studies have been carried out on the virulence strategies of the most commonly isolated bacterial pathogen from CF patients, P. aeruginosa, particularly in relation to its ability to colonize the CF lung and the role that iron acquisition mechanisms may play in this process. This knowledge may inform our understanding of the conditions prevailing within the CF lung and the iron acquisition mechanisms that are important for establishment of an infection by Bcc species. Where relevant, pertinent data obtained from studies on P. aeruginosa will be discussed in this review.

### THE ROLE OF IRON ACQUISITION MECHANISMS IN BCC INFECTIONS

#### Iron Availability in the CF Lung

In considering the potential role of iron acquisition systems in the CF lung it is worth reviewing what we know considering the iron content of CF sputum and its bioavailability. Based on the known iron limiting environment of the ASL of the healthy lung, where iron is sequestered by lactoferrin, transferrin, and ferritin, it was long assumed that the CF lung also generated an iron-deficient environment (Drevinek et al., 2008). Indeed, the results of some investigations into the regulation of iron acquisition genes in the major CF pathogen, Pseudomonas aeruginosa, appeared to lend support to this contention (see below). However, measurements of the iron content of the CF lung have led to a reappraisal of this environment. It is now clear that the lungs of CF patients have, on average, a higher iron content than that of a healthy individual. For example, it was shown that the abundance of the iron storage protein, ferritin, was on average nearly 20-fold higher in the lungs of CF patients compared to healthy individuals, while there was ∼50% less transferrin (Stites et al., 1998). Other investigators have confirmed the high ferritin concentrations in CF sputum (Reid et al., 2002, 2004). In a more recent investigation, the total iron content and the fractions that were in the Fe(II) and Fe(III) forms were measured in CF patients experiencing differing degrees of disease severity (Hunter et al., 2013). This showed a strong positive correlation between the iron content of CF sputum and the progression of the disease. Thus, patients with severe disease (as judged by their low FEV) had a high mean iron concentration in their sputum of 72µM compared to those with a milder condition where the mean was 18µM. Moreover, a substantial fraction of iron was present in the soluble ferrous form which increased in line with disease severity, such that in the most severe disease stage, ferrous iron constituted ∼40% of the total iron load (Hunter et al., 2013). The increased abundance of Fe(II) in severe or late-stage CF disease is likely to be due to the reduction of Fe(III) by neutrophil-generated superoxides and stabilization of the resultant ferrous form by the increased prevalence of hypoxic zones in parts of the lung and acidification of the ASL.

Another source of iron that is more abundant in the CF lung is haem. This molecule becomes available through its release from hemoglobin which can occur following oxidation of the coordinated ferrous iron atom or following proteolysis of hemoglobin by host- or pathogen-derived proteases (Balla et al., 1993; Cosgrove et al., 2011). Lung tissue may release iron, including sources of haem, through injury due to the ravages of chronic inflammation (Reid et al., 2004). Moreover, CF patients experience a high frequency of micro-bleeds in their lung tissue that results in hemoglobin entering their ASL (Cosgrove et al., 2011). The frequency of airway bleeding in CF patients increases during pulmonary exacerbations where symptoms become more severe (Reid et al., 2009). The availability of ferrous iron and haem, particularly in the later stages of the disease, has potential implications for the iron acquisition systems that may be deployed by a colonizing pathogen. This change in our understanding of the iron status of the CF lung has led some workers to propose that the CF lung environment actually facilitates the growth of organisms such as P. aeruginosa (Reid et al., 2007).

#### Iron Acquisition Mechanisms of the Bcc

Many bacteria synthesize and secrete low molecular weight, high affinity iron chelating compounds known as siderophores which they employ to capture iron from their local environment, particularly when this element is scarce (Chu et al., 2010). Due to its propensity to form poorly soluble hydroxides in solution, such as Fe(OH)<sup>+</sup> 2 , an important role of the siderophore is to solubilize the ferric form of iron (Chipperfield and Ratledge, 2000; Ratledge and Dover, 2000). The affinities of some siderophores for iron are sufficiently high to allow them to obtain iron from host iron transport proteins such as lactoferrin and transferrin, but not from haem (Skaar, 2010). These compounds contain one, two, or three bidentate ligands that allow them to coordinate to a single Fe(III) ion, the predominant form of iron in aerobic environments at physiological pH. As iron forms hexavalent coordination complexes with its ligands, a single siderophore molecule containing three bidentate ligands (i.e., a hexadentate siderophore) will form a 1:1 complex with one ferric ion giving rise to an overall octahedral geometry (Neilands, 1995; Ratledge and Dover, 2000). Under iron replete conditions, synthesis of these molecules (and expression of other iron acquisition systems) is, in most cases, strongly downregulated in order to prevent cytoplasmic iron overload that may generate high levels of toxic reactive oxygen intermediates via the Fenton reaction (for reviews see Andrews et al., 2003; Cornelis et al., 2011).

Members of the Bcc have been shown to produce one or more of four different siderophores with which they can acquire iron: ornibactin, cepaciachelin, pyochelin, and cepabactin (Meyer et al., 1989, 1995; Stephan et al., 1993; Barelmann et al., 1996; Darling et al., 1998). These siderophores include all three types (bidentate, tetradentate, and hexadentate) and all of the most common iron binding ligands are represented among them (hydroxamate, hydroxycarboxylate, catechol, and 2-hydroxyphenylthiazoline; **Figure 2**). The biosynthesis of pyochelin and ornibactin and the genetic regulation of their synthesis have been reviewed elsewhere, while the biosynthetic genes for cepaciachelin have been recently identified (Thomas, 2007; Esmaeel et al., 2016). For this review, we have surveyed the distribution of ornibactin, cepaciachelin, and pyochelin among the Bcc by carrying out a bioinformatic analysis of the genomes of 21 Bcc members using the corresponding biosynthetic genes as search queries (**Table 1**). The results accord with more limited surveys carried out previously (Deng et al., 2016; Esmaeel et al., 2016). We have also augmented the bioinformatics analysis by referencing those cases where production of a particular siderophore by specific Bcc species has actually been demonstrated. Currently, this is the only way of ascertaining which species specify cepabactin, as the biosynthetic genes remain to be identified.

Based on bioinformatic analysis of genome sequences, all Bcc species (apart from the recently described Burkholderia paludis) are predicted to produce the siderophore ornibactin, which is likely to act as the primary secreted iron chelator in these organisms based on its hexadenticity (**Table 1** and **Figure 3**). Although ornibactin and malleobactin E (the siderophore produced by members of the B. pseudomallei group) are very similar (**Figure 2**), and therefore require similar biosynthetic enzymes for their assembly (**Figure 3**), a key feature that distinguishes the type of siderophore produced by each species is the presence of a distinct amino acid activation (adenylation) domain at the N-terminus of the larger of the two non-ribosomal peptide synthetases (NRPSs) that assemble these tetrapeptide siderophores (OrbI in the case of ornibactin). This domain activates the derivatized ornithine that will be located at the N-terminus of the tetrapeptide (for further details the reader is referred to Thomas, 2007 and the legend to **Figure 3**). In addition, as the δ-amino group of the N-terminal ornithine residue of ornibactin is acylated with a β-hydroxycarboxylic acid (rather than formic acid as in malleobactin E), the ornibactin gene cluster is distinguished by the presence of at least one of two genes (orbK and orbL) that are predicted to encode an acylase that

of *B. xenovorans*, tentatively referred to here as "malleobactin X." (D) Cepaciachelin contains two 2,3-DHBA groups that form amide linkages with the two amino groups of lysine, which in turn is conjugated to a molecule of putrescine (1,4-diaminobutane) on its α-carboxyl group. (E) Pyochelin contains two less commonly occurring bidentate iron-chelating groups (2-hydroxyphenyl thiazoline and N-methylthiazolidine-4-carboxylate). (F) Cepabactin, a cyclic hydroxamate bidentate siderophore. Chemical groups that distinguish the ornibactins and malleobactins are indicated in red circles or ellipses.

catalyses this condensation reaction (**Figure 3**). Whereas orbL is always present, orbK may contain an internal deletion (as in B. ubonensis) or be absent from the cluster altogether (as in B. vietnamiensis). Although, the single reported B. paludis strain is an environmental isolate, it should be noted that in a survey of "B. cepacia" CF isolates carried out prior to the taxonomic reorganization of B. cepacia into separate Bcc species, two clinical strains were found not to produce detectable levels of ornibactin (Darling et al., 1998). It is not clear to which Bcc member species they belong or whether they are indeed members of the Bcc. One possibility is that these strains were capable of producing ornibactin prior to infection but this ability was lost through mutation during prolonged carriage as has been observed with respect to production of the major siderophore pyoverdine by some P. aeruginosa strains isolated from chronically infected CF patients (De Vos et al., 2001; Smith et al., 2006; Andersen et al., 2015).

In addition, most Bcc species produce one or more secondary siderophores that are likely to have lower affinity for iron than ornibactin. The gene clusters specifying the biosynthesis and utilization of two of these siderophores, cepaciachelin and pyochelin, are shown in **Figures 4A,B**. Based on our bioinformatics survey, at least 7 species of Bcc, including B. cenocepacia and B. lata, produce pyochelin as the secondary siderophore, a feature associated with some Pseudomonas species (Cornelis and Matthijs, 2002), whereas in 8 other species, including some strains of B. ambifaria and B. multivorans, the secondary siderophore is cepaciachelin (**Table 1**). We have not identified a species possessing the genetic information required to produce both cepaciachelin and pyochelin. Both compounds are tetradentate siderophores, although the former belongs to the 2-hydroxyphenylthiazoline family whereas cepaciachelin is a bis-catecholate siderophore (Barelmann et al., 1996; Thomas, 2007; Inahashi et al., 2017). Some species do not appear to produce either of these two compounds as a secondary siderophore (**Table 1**). Another siderophore, the bidentate cyclic hydroxamate, cepabactin, has been detected in culture supernatants of some environmental B. cepacia strains in addition to ornibactin and pyochelin (see **Table 1**; Meyer et al., 1989, 1995). The ability of some clinical Bcc isolates of unknown taxonomic status to produce cepabactin has also been observed (Darling et al., 1998). Currently, it is not possible to infer from bioinformatics how widespread the synthesis or utilization of this siderophore is likely to be among the Bcc, although its production has not been observed in B. cenocepacia and B. vietnamiensis strains (**Table 1**).

The uptake of ferric-siderophore complexes by Gramnegative bacteria such as the Burkholderia requires an outer

FIGURE 3 | Organization of ornibactin, malleobactin, and phymabactin biosynthesis and utilization genes in pathogenic *Burkholderia* and related species. Genes are represented by block arrows and are color coded as indicated in the figure [the precise role of the MbtH-like OrbH/MbaG/PhmF proteins is unknown but they are assumed to be required for biosynthesis of the siderophore based on the requirement for other MbtH-like proteins for NRPS-mediated biosynthesis of some peptides; (Wolpert et al., 2007; Baltz, 2011)]. The numbering of each gene cluster corresponds to the numbering system for the species listed at the bottom of the figure. The font color used for each species name corresponds to the siderophore produced as follows: black, ornibactin; red, malleobactins; green, phymabactin. NRPSpredictor2 (Rottig et al., 2011) was used to predict the siderophore product based on the substrates accepted by the four adenylation domains present in OrbI/OrbJ, MbaA/MbaB, and PhmA/PhmB for each species. For systems known to specify ornibactin, the first and last (N- and C-terminal) adenylation domains of the OrbI-OrbJ NRPS pair are both predicted to accept leucine with highest probability, reflecting the presence of *N* 5 -3-hydroxyacyl-*N* 5 -hydroxyornithine and *N* 5 -formyl-*N* 5 -hydroxyornithine, respectively, at these positions in the tetrapeptide product. For malleobactin, the predicted specificity of the N-terminal adenylation domain changes to β-hydroxytyrosine although the accepted substrate is *N* 5 -formyl-*N* 5 -hydroxyornithine. In some species, such as *B. thailandensis*, the N-terminal *N* 5 -formyl-*N* 5 -hydroxyornithine is formylated on the *N* 2 -amino group upon formation of the tetrapeptide to generate malleobactin E (Franke et al., 2015), whereas in *B. xenovorans* the N-terminal *N* 5 -formyl-*N* 5 -hydroxyornithine does not appear to undergo such a tailoring reaction (Vargas-Straube et al., 2016) and so we tentatively refer to this siderophore as "malleobactin X." The N- and C-terminal adenylation domains of PhmA-PhmB are predicted to accept aspartate and cysteine, respectively, but the structure of the product, phymabactin, is unknown, although it is predicted to have siderophore activity based on its genomic context (Esmaeel et al., 2016). In all cases, the second and third adenylation domains are predicted to accept aspartate and serine, respectively, which correspond to β-hydroxy-D-aspartate and L-serine in the final product. The nomenclature proposed for each gene is shown below the gene clusters, and the gene designations are color coded as follows: ornibactin, black (Agnoli et al., 2006); the two systems for malleobactin, red (upper, Alice et al., 2006; lower, Franke et al., 2013); the two systems for phymabactin, green (upper, Esmaeel et al., 2016; lower, this study). Genes indicated by a single letter have the same prefix as the gene name at the extreme left. Dashes indicate the absence of a gene. Question mark indicates where a gene name has not been proposed. The initial annotation of the phymabactin gene cluster (Esmaeel et al., 2016) did not include the third and last genes in the cluster or the fact that a TBDR gene occurs in other species bearing this gene cluster (upper annotation in green font). Therefore, we have proposed an alternative nomenclature based on the ornibactin gene cluster (lower annotation in green font). Scale bar refers to gene lengths and not intergenic regions, which in some cases have been exaggerated to permit alignment of each gene cluster. Species marked with an asterisk belong to a subclade within the *B. xenovorans* group and have been reassigned to the new genus *Paraburkholderia* (Sawana et al., 2014). *Collimonas* is a genus within the *Oxalobacteraceae*, a family belonging to the order *Burkholderiales*. Member species of the Bcc are enclosed in a box (*B. arboris* is not listed as its genome sequence is not currently available). Gene loci are shown in Supplementary Table 1.

FIGURE 4 | Organization of the cepaciachelin and pyochelin biosynthesis and utilization genes and the genes for the haem uptake system in *Burkholderia* species. (A) The cepaciachelin gene cluster in *Burkholderia* species encodes enzymes for the synthesis of the precursor 2,3-dihydroxybenzoic acid (DHBA) from chorismate and its assembly into the siderophore. Genes encoding possible cepaciachelin transport proteins are also present, including the TonB-dependent receptor, CpcG, and an MFS transporter, CpcH. Most of the genes have been previously annotated (Esmaeel et al., 2016) but the annotation has been extended here (genes labeled in red font) and includes components of a cytoplasmic membrane ABC transporter which may be involved in uptake of ferric cepaciachelin (CpcF, -I, and -J). EstA is homologous to the cytoplasmic enterobactin and salmochelin esterases Fes and IroD that are required for removal of iron from ferric-enterobactins following their uptake. However, it contains a putative Sec-dependent signal peptide and so may be periplasmically located like the *C. jejuni* enterobactin esterase, Cee (Zeng et al., 2013). This may suggest that the putative cepaciachelin receptor (CpcG) and the cytoplasmic membrane transporter also recognizes ferric-enterobactin. With the exception of *B. ambifaria* and *B. pseudomultivorans* the cepaciachelin gene cluster includes a gene encoding a DAHP synthase which catalyses the first step in the shikimate pathway that leads to the biosynthesis of chorismate from erythrose-4-phosphate and PEP. Note that in Esmaeel et al. (2016) *cphA-cphC* should be annotated as *cpcA-cpcC*, as shown here (V. Leclere, personal communication). -, gene not assigned a four letter name. Gene loci are shown in Supplementary Table 2. (B) The pyochelin gene cluster. Genes annotated with a single letter are designated with the prefix *pch*. Products of the biosynthetic genes generate the precursor salicylic acid from chorismate (PchAB), activate it (PchD) and assemble it into pyochelin along with two molecules of cysteine (PchCEFG). The *pchHI* and *fptABCX* genes encode membrane proteins, of which two (FptA and FptX) are involved in the transport of exogenous ferric-pyochelin across the outer and inner membranes, respectively. *fptBC* and *pchHI* appear not to be essential for export of pyochelin nor for uptake of iron-bound pyochelin (see Youard et al., 2011 for a review). Gene loci are shown in Supplementary Table 3. (C) Organization of the *Burkholderia* haem uptake genes, *bhuRSTUV* (Shalom et al., 2007; Thomas, 2007), also referred to as *hmuRSTUV*, *huvA-hmuSTUV* or *omr-hmuSTUV* (Yuhara et al., 2008; Kvitko et al., 2012; Tyrrell et al., 2015). Note that in *B. stagnalis* a VOC family protein is encoded between *bhuU* and *bhuV*, and in *B. gladioli* the *bhu* genes are organized into two operons present on separate chromosomes. Gene loci of representative species are given in Supplementary Table 4.

membrane receptor, a 75–85 kDa polypeptide that folds into a β-barrel containing a central plug domain. Binding of a ferricsiderophore complex to the external face of the receptor triggers a conformational change in the gated receptor that allows access of the complex to the periplasmic space. The energy required for this process is derived from the proton motive force through the action of the TonB system, a complex of three different cytoplasmic membrane-anchored protein subunits: TonB, ExbB, and ExbD (Noinaj et al., 2010; Celia et al., 2016). For this reason, ferric-siderophore receptors are referred to as TonBdependent receptors (TBDRs) or TonB-dependent transporters (TBDTs). As an example, OrbA is the TBDR for ferric-ornibactin (**Figure 5A**). Once the ferric-siderophore complex has entered the periplasmic space, the ferric ion is transported across the cytoplasmic membrane, either in complex with the siderophore or following release from the siderophore (depending on the system). The cytoplasmic membrane transporters are often ATPbinding cassette (ABC) transporters that consist of a periplasmic binding protein, an intrinsic membrane protein (the permease) and an ATPase located on the cytoplasmic face of the permease (Krewulak and Vogel, 2008). This type of system operates for the uptake of ferric ornibactin (Orb-B, -C, and -D) and possibly also for cepaciachelin (CpcF, -I, and -J) in the Bcc, as well as for the import of ferric malleobactin in B. pseudomallei and related bacteria (**Figures 5A,B**; Agnoli et al., 2006). In the case of ferricpyochelin, a single subunit permease, FptX, appears to serve as the cytoplasmic membrane transporter (**Figure 5C**; Cuiv et al., 2004; Cunrath et al., 2015). For ferric-siderophore complexes that enter the cytoplasm, the iron is removed from the siderophore through its reduction to Fe(II), which is presumed to occur for ornibactin (Agnoli et al., 2006), or through modification or hydrolysis of the siderophore (Brickman and McIntosh, 1992; Hannauer et al., 2010).

Like many other bacterial species, Burkholderia spp. encode additional TBDRs that may allow them to utilize siderophores that are produced by other bacteria and fungi ("xenosiderophores"), although the potential importance of these compounds for pathogenicity is only likely to be realized in the context of a polymicrobial infection. Based on an analysis of its translated genome, B. cenocepacia is predicted to encode at least 20 TBDRs, many of which are likely to be involved in utilization of xenosiderophores (our unpublished results). At present, little is known concerning the nature of the xenosiderophores that can be utilized by the Burkholderia. However, the presence of the ornibactin transport genes in B. paludis, but not the biosynthetic genes, strongly suggests that ornibactin is likely to be utilized as a xenosiderophore by this species.

Members of the Bcc also specify iron acquisition systems that are not siderophore-dependent. For example, it has been shown that B. cenocepacia can utilize haem as an iron source (Whitby et al., 2006; Mathew et al., 2014; Tyrrell et al., 2015). B. cenocepacia and B. multivorans contain a cluster of genes (bhuRSTUV) that are predicted to be required for uptake of this molecule (**Figure 4C**; Thomas, 2007; Yuhara et al., 2008). [Note that there is currently a lack of consistency regarding the genetic nomenclature for this system in the Burkholderia (see below and legend to **Figure 4**).] Our bioinformatic survey shows that the bhu cluster is present on chromosome 2 in nearly all Bcc species for which whole genome sequence information is available (Supplementary Table 4). The exception is B. vietnamiensis which completely lacks the bhuRSTUV gene cluster. Although, there

FIGURE 5 | Proposed iron uptake pathways in the *Burkholderia*. (A) The ornibactin/malleobactin uptake system. Ferric-ornibactin is recognized by the OrbA/MbaD TBDR and is translocated into the periplasmic space through a conformational change in the plug domain of the TBDR that requires energy transduction by the TonB complex (TonB-ExbB-ExbD). The iron-siderophore complex is then transported across the cytoplasmic membrane by a periplasmic binding protein-dependent ABC transporter (OrbBCD/MbaLIJ). Oncethe ferric-siderophore complex has been internalized, iron is released from ornibactin through its reduction to the ferrous form by OrbF/MbaK. (B) Ferric-cepaciachelin is proposed to require the CpcG TBDR. Genes neighboring *cpcG* encode a periplasmic binding protein-dependent ABC transporter (CpcFHI) that may be involved in transport of the bis-catecholate complex across the cytoplasmic membrane. (C) Ferric-pyochelin uptake requires the FptA TBDR and the single subunit cytoplasmic membrane transporter, FptX. (D) Uptake of haem via the Bhu system. Haem uptake is proposed to follow an analogous pathway to that of ornibactin/malleobactin and cepaciachelin. Cytoplasmic haem is bound by the BhuS protein which is proposed to play a role in haem trafficking and haemostasis. If required, iron can be released form haem by haem oxygenases (not shown). The FtrABCD system is not shown. OM, outer membrane; CM, cytoplasmic membrane.

have not been any major studies carried out to investigate the role of this system in haem acquisition in the Bcc, this function has been established for the bhuRSTUV system in B. pseudomallei (see below), and so one can confidently infer its role in haem uptake in the Bcc. Haem appears to be an important source of iron for P. aeruginosa during colonization of the CF lung, and therefore one might expect members of the Bcc to take advantage of this nutrient (Konings et al., 2013). B. cenocepacia can also obtain iron from ferritin in a protease-dependent process (Whitby et al., 2006; Mathew et al., 2014; Tyrrell et al., 2015). Interestingly, although iron is present in the ferric form when sequestered by ferritin, siderophores are not required for uptake of ferritin-derived iron in B. pseudomallei (Kvitko et al., 2012).

Ferrous iron is very soluble and can passage across the outer membrane of Gram-negative bacteria through porins in an energy-independent process. However, it requires specific transporters for translocation across the cytoplasmic membrane in all bacteria. The Feo system is one such ferrous iron-specific uptake system that is present in many Gram-positive and Gramnegative bacteria (Lau et al., 2016). This system consists of a cytoplasmic membrane transport protein (FeoB) and (in most cases) a cytoplasmic component of unknown function (FeoA). In some cases, a third component (FeoC) is also present. Ferrous iron is likely to be an important source of this essential nutrient for bacteria colonizing the CF lung based on its increased abundance in the ASL of more severely afflicted CF patients. Accordingly, it has been shown that the Feo system of P. aeruginosa is upregulated during colonization of the lungs of every individual in a cohort of 23 infected CF patients (Konings et al., 2013). For this reason, we conducted a survey of human pathogenic Burkholderia species for the presence of feoA and feoB. However, only a small a minority of B. multivorans and B. pseudomultivorans strains (not the type strains), as well as the single currently identified B. paludis strain (MSh1), were found to encode a FeoB-like protein, although these proteins lacked a region of ∼100 amino acids that is present in P. aeruginosa FeoB. There were no matches when FeoA was used as the query in a BLASTP search of the Bcc and B. pseudomallei. We conclude that the vast majority of pathogenic Burkholderia species lack this particular ferrous iron uptake system.

B. cenocepacia encodes a siderophore-independent mechanism for iron assimilation that is very similar to the FtrABCD systems reported in Bordetella and Brucella (Brickman and Armstrong, 2012; Elhassanny et al., 2013; Mathew et al., 2014). These systems share similarities with components of the Escherichia coli EfeUOB system that serves to import ferrous iron under aerobic conditions at low pH (Cao et al., 2007). Accordingly, in Bordetella spp. and Brucella abortus, ferrous iron is efficiently transported into these bacteria by the FtrABCD system over the pH range 6.0–7.5, in a process where the Fe(II) ion is oxidized to the ferric form using the FtrB cupredoxin component before translocation across the cytoplasmic membrane. In contrast, the B. cenocepacia FtrABCD system appears to utilize ferric iron as the substrate. Given the absence of a Feo system in B. cenocepacia, this begs the question as to the mechanism by which this species can uptake ferrous iron. B. multivorans also possesses an FtrABCD system, but in contrast to that of B. cenocepacia it does utilize ferrous iron as the substrate (S.C. Andrews, unpublished results). Therefore, it is possible that this system may play a role in ferrous iron acquisition in the CF lung for at least some Bcc species, particularly given the abundance of ferrous iron in combination with the relatively low pH of CF ASL.

Although, the substantial proportion of iron that is in the ferrous form and the increased abundance of haem in the ASL may implicate the siderophore-independent iron acquisition mechanisms of the Bcc in establishing CF lung infections, as yet, the role of these systems in the context of CF have not been investigated in any detail (see below).

#### Experimental Evidence for the Role of Iron Acquisition Systems in the Virulence of the Bcc

#### In Vivo Studies

The earliest study on the role of iron acquisition mechanisms in the virulence of members of the Bcc was carried out prior to the discovery of the primary siderophore, ornibactin, in this group of bacteria (Sokol, 1986). Here, it was observed that from a collection of 43 "P. cepacia" CF isolates, ∼50% produced detectable levels of pyochelin during iron limited growth in vitro (Pch<sup>+</sup> phenotype). However, 86% of the Pch<sup>+</sup> strains were associated with infections which had led to the death of the patient or were responsible for severe infections, whereas only 41% of Pch<sup>−</sup> strains were associated with such outcomes. Thus, while the ability to colonize a CF patient could not be linked to the ability to produce this siderophore, there was a link between pyochelin production and the morbidity/mortality of disease in CF patients. Similarly, while addition of exogenous pyochelin to two pyochelin-negative Bcc strains did not increase bacterial numbers or bacterial persistence in infected rat lungs, it did increase the severity of infection as assessed by lung pathology (Sokol and Woods, 1988). It was proposed that the observed enhancement of lung damage brought about by exogenous addition of pyochelin to these strains was most likely due to increased dissemination of the bacteria throughout the lungs. However, it is not clear from these studies whether it was the role of pyochelin in iron acquisition that was responsible for the more severe outcome or some other effect of the siderophore. In this regard, it is known that apart from binding iron and other metals (Cuppels et al., 1987; Visca et al., 1992; Baysse et al., 2000; Braud et al., 2009) pyochelin possesses an inherent chemical reactivity that may contribute to disease severity. For example, it can promote the degradation of organotin derivatives (Sun et al., 2006), and perhaps pertinently, it can catalyse the generation of ROS such as hydroxyl radicals that result in tissue damage (Coffman et al., 1990; Britigan et al., 1994, 1997; Adler et al., 2012). The latter property is proposed to contribute to the known antibiotic activity of this siderophore (Adler et al., 2012; Ong et al., 2016). It should be noted that one important Bcc pathogen of CF patients, B. multivorans, does not produce pyochelin (**Table 1**) and the clonally related B. cenocepacia CF epidemic strains K56-2 and J2315 produce very little of this siderophore (see below).

A direct genotypic-phenotypic link between iron acquisition and the virulence of B. cenocepacia was observed during an investigation of the virulence potential of ornibactin deficient mutants in rodent models of both chronic and acute respiratory infection (Sokol et al., 1999). In this study, mutants derived from the highly transmissible epidemic B. cenocepacia strain, K56-2, which contained an insertionally inactivated pvdA gene that is required for ornibactin synthesis (**Figure 3**), were generated by transposon mutagenesis (pvdA::Tn5-OT182) and allelic replacement (pvdA::tp). Both mutants were significantly attenuated in these models. Thus, in a rat lung chronic infection model, the number of pvdA::Tn5-OT182 bacterial cells recovered from the lung was 4 logs lower than that of the wild type K56- 2 strain at 28 days post infection. Furthermore, the pvdA::tp strain could not be recovered from the lungs after the same length of time, suggesting the infection had been cleared. The degree of pathology, as determined by the amount of inflammatory cell infiltration and exudate in the lungs, was also significantly reduced in the K56pvdA::tp strain compared to that of K56-2. Aerosol administration of K56-2 and the pvdA::Tn5-OT182 mutant into neutropenic mice as an acute respiratory infection model revealed that whereas the wild type strain was able to persist in the lung 7 days post infection, the pvdA mutant was cleared from most of the mice after 3 days. These experiments suggested the importance of ornibactinmediated iron acquisition by the bacteria for initial colonization, persistence and resulting pathological changes within the host. However, given the difference in the iron content of the lungs of a healthy individual compared to those of a CF patient, it is not clear to what extent the conclusions from this study, which did not involve CF mice or rats, can be extrapolated to the situation in the CF lung.

While the data indicated an important role for ornibactin in lung colonization in these models, the K56-2 strain (and other members of the highly transmissible ET12 epidemic lineage) produces very low amounts of the siderophore pyochelin compared to other B. cenocepacia strains due to a frameshift mutation in pchF (Darling et al., 1998; Holden et al., 2009) (Based on perusal of the genome sequence, we presume pyochelin biosynthesis in these strains occurs through independent initiation of translation from an internal in-frame GUG codon located upstream of the frameshift site that results in production of PchF as two separate components). Thus, the role of pyochelin in the lung infection model could not be established using ET12 strains. However, given that ET12 strains cause life-threatening infections in CF patients, this would again appear to rule out an important role for pyochelin in colonization of the CF lung by B. cenocepacia. Later work using a B. cenocepacia strain (Pc715jorbA::tp) that produced normal amounts of pyochelin but was unable to utilize ferric-ornibactin due to disruption of the gene encoding the ferric-ornibactin TBDR, OrbA, revealed that it was cleared from rat lungs much more quickly than the WT strain (Visser et al., 2004). A ferric-pyochelin receptor mutant (Pc715jfptA::tp) persisted with the same efficiency as that of the WT. These data suggested that while pyochelin may have a role in the severity of infection, it is unable to compensate for the loss of a functional ornibactin utilization system. Therefore, it is ornibactin which appears to be important in order to establish an infection in this system. This may be explained by the presumed lower affinity of pyochelin than ornibactin for iron (Cox and Graham, 1979; Visca et al., 1993).

The importance of the ornibactin system for the virulence of B. cenocepacia has also been assessed in other infection models, including invertebrates and plants. Both a K56-2 orbA mutant and a K56-2 pvdA mutant were attenuated in the Caenorhabditis elegans and Galleria mellonella invertebrate models. The pvdA mutant was also slightly attenuated in the plant alfalfa model (Uehlinger et al., 2009). An orbJ mutant of the B. cenocepacia CF strain, H111, that is also deficient in the production of ornibactin, was also attenuated in the G. mellonella system (Mathew et al., 2014). Consistent with the virulence of the K56-2 strain in the rat lung chronic infection model, an H111 1pchAB pyochelin deficient mutant was still virulent in G. mellonella.

Using a modified signature-tagged mutagenesis (STM) procedure to identify genes required for survival in the rat chronic lung infection model, one of the attenuated B. cenocepacia K56-2 mutants which could not survive for 10 days in this model contained a transposon inserted just upstream of an ORF that encoded a haem TBDR-like protein (Hunt et al., 2004). The authors of this study indicated that the gene was the first in a cluster of genes associated with haem uptake that were located on chromosome 2, and the encoded protein was very similar to the 79 kDa RS03722 gene product of Ralstonia solanacearum strain GMI1000 (now reannotated as RSp0244). As RSp0244 is highly similar to BhuR we conclude that the plasposon insertion exerted polar effects on expression of the bhuRSTUV operon that impaired or abolished haem uptake. In contrast, genes involved in the biosynthesis and transport of ornibactin and pyochelin were not implicated in this study (Hunt et al., 2004). These observations would suggest that haem acquisition, and not siderophore-mediated iron acquisition, is an essential trait for persistence in the rat lung. Notwithstanding the different time courses of the chronic lung infection models, it is not clear why the ornibactin system should be implicated in some studies (Sokol et al., 1999; Visser et al., 2004) but not in the STM study, particularly as the ornibactin gene cluster presents a large target for plasposon-mediated disruption. One possible explanation is the selection procedure employed to construct the B. cenocepacia transposon mutant library. Here, mutants were selected on a mineral salts based medium in order to exclude auxotrophs. A K56-2 mutant in which the plasposon has disrupted a gene required for ornibactin synthesis or utilization would effectively be unable to obtain iron in a siderophore-dependent manner. We have observed that B. cenocepacia siderophore deficient mutants grow more slowly that the parental wild type strain on mineral salts medium (see for example Asghar et al., 2011), and it is possible that such mutants were omitted from the library of Hunt et al. This is not a complete explanation, however, as the haem uptake deficient mutants in the STM study can still produce ornibactin and so they might be expected to retain virulence based on other studies. This may suggest that a combination of both iron acquisition systems (haem- and ornibactin-mediated) is required for efficient colonization and persistence in the rat lung model.

Finally, the B. cenocepacia FtrABCD system was also investigated for a potential role in virulence in the Galleria wax moth model. Whereas deletion of the ftr system in isolation did not result in reduced virulence, when deleted in a strain that was unable to biosynthesise ornibactin and pyochelin, the mutant was more attenuated in comparison to an ornibactin-negative strain. This observation suggests that while ornibactin is the more important iron acquisition system for virulence in this model, the FtrABCD system can play a role in iron acquisition during infection in the absence of siderophores (Mathew et al., 2014).

#### In Vitro Studies

Changes in the environmental iron concentration or availability can also trigger adaptive changes in expression of virulence traits that are not directly connected to iron acquisition but rather serve other roles that contribute to survival of the bacterium under the prevailing conditions. Such examples include regulating biofilm formation in P. aeruginosa and capsule production in Cryptococcus neoformans (Singh et al., 2002; Banin et al., 2005; Jung et al., 2006). For B. cenocepacia it has been shown that modulating the concentration of iron causes a switch from planktonic to sessile growth. Thus, supplementing liquid cultures with ferric iron concentrations ranging from 1 to 100µM resulted in increased levels of extracellular matrix production by strain PVI as the iron concentration was increased. Furthermore, biofilm formation induced by growth under high iron conditions resulted in more efficient invasion of A549 epithelial monolayers compared to cells grown in lower iron conditions that did not produce biofilm (Berlutti et al., 2005). In contrast, adherence of B. cenocepacia to A549 cells was more efficient under iron limiting conditions. The consequences of this for B. cenocepacia CF lung infections are not yet clear.

#### Gene Expression and Omics Studies

Transcriptomic and proteomic analyses can be used to identify sets of genes or proteins that may be required for survival under particular conditions by virtue of their differential expression. A few such studies have been carried out with B. cenocepacia that have suggested an important role for iron uptake mechanisms in bacterial persistence within CF patients. However, as we discuss below, the experimental set up may not necessarily be appropriate for addressing this particular question. The first notable study looked at global gene expression during growth of B. cenocepacia J2315 in a basal salts medium supplemented with CF sputum in comparison to growth in unsupplemented medium (Drevinek et al., 2008). The microarray revealed upregulation of 287 genes and downregulation of 437 other genes during growth in CF sputum medium. However, only two of the upregulated genes were associated with characterized iron acquisition systems in this bacterium. These two genes (pchR and pchD) encode the transcription activator of the pyochelin gene cluster and an enzyme required for biosynthesis of the pyochelin precursor, salicylic acid, respectively, but their expression increased only 2 fold. Transcription of the ornibactin genes was not upregulated, although the gene encoding the global iron repressor, Fur, that represses the ornibactin gene cluster, was found to be downregulated 2- to 3-fold (Agnoli et al., 2006; Drevinek et al., 2008).

The authors of this work measured the iron content of their sputum medium and found it to be ∼35µM, which is more than adequate to sustain growth of B. cenocepacia in standard laboratory medium without upregulating siderophore biosynthesis (Drevinek et al., 2008; Madeira et al., 2013; our unpublished results). Despite the presence of CF sputum in the medium, which has been argued by some investigators to sequester iron due to the presence of various iron binding components (Wang et al., 1996; Palmer et al., 2007), their results imply that sufficient iron is available to effect repression of the ornibactin system. Accordingly, the observed induction of the pch genes may not be related to iron depletion but is rather a response to the presence of another component in CF sputum. It is noteworthy that genes encoding other components of the pyochelin biosynthesis machinery (particularly those enzymes required to assemble pyochelin from salicylate and cysteine) and alternative iron uptake systems (FtrABCD and BhuRSTUV) were also not upregulated. The authors contend that the CF medium they employed restricted the available iron based on the observed gross upregulation of the BCAL0270 gene, which they considered to be involved in iron acquisition (described in the study as a "ferric reductase-like transmembrane component"). However, in a basal salts medium without iron supplementation, this gene was upregulated only 2-fold compared to medium containing the standard amount of ferrous sulfate (43µM). Moreover, this gene is currently annotated in the NCBI database as encoding a "sulfoxide reductase heme-binding subunit YedZ" and it is transcriptionally linked to BCAL0269, a gene that encodes a YedY-homologous protein. Our own bioinformatics analysis supports this annotation (results not shown). The E. coli YedYZ complex is a membrane anchored haem-molybdoenzyme that serves to reduce an as yet unknown S- or N-oxide (Iobbi-Nivol and Leimkuhler, 2013). Therefore, there is little evidence to support the suggestion that BCAL0270 is involved in iron acquisition. The possibility that the CF sputum medium of Drevinek et al. (2008) is iron sufficient accords with the high iron concentration included in the basal salts medium used to generate the medium. Notwithstanding the fact that the sputum is present at only 10% (w/v) in the medium, the inability of the added CF sputum to induce iron acquisition systems may suggest that its ability to sequester iron is somewhat limited and/or it is iron replete.

The results of Drevinek et al. (2008) contrast with those of Palmer et al. (2005), who monitored gene expression in the CF pathogen P. aeruginosa growing in a mineral salts based medium containing only CF sputum as the source of carbon and energy ("MOPS-sputum medium"), and noted that a large number of genes specifying the biosynthesis of the major siderophore, pyoverdine, as well as the entire cluster of genes specifying the biosynthesis and transport of the secondary siderophore, pyochelin, were considerably upregulated relative to their transcription in cells growing in sputum-free MOPS-glucose medium. However, although MOPS-sputum medium contained a similar amount of CF sputum to that used by Drevinek et al. (2008), the iron content (3.5µM) was approximately one tenth of that present in the medium used in the B. cenocepacia experiment (in both cases iron was added as ferrous sulfate). Thus, although it should be borne in mind that two different CF pathogens are being compared, each possessing a different primary siderophore system, the amount of iron added ranged from iron replete (in the B. cenocepacia experiment) to a concentration that will support bacterial growth but may require a degree of upregulation of the siderophore-mediated iron acquisition system (in the P. aeruginosa experiment), (pyoverdine synthesis is fully repressed at ∼4µM iron and is upregulated to a progressively greater degree as concentrations of iron are decreased below 4µM; Meyer and Abdallah, 1978). The reader may wish to consider which version of CF sputum medium more closely represents the true environment of the CF lung with respect to iron availability. The other variable at play that may influence the iron content is the source of the sputum. As discussed above, the iron content of sputum shows marked variation among CF patients, particularly in relation to the severity of the disease (Stites et al., 1998; Reid et al., 2002; Hunter et al., 2013). To illustrate the potential for a different outcome that may reflect variation in the iron content of CF sputum, in an IVET study conducted on P. aeruginosa growing in a mineral salts medium containing 10% CF mucus, and otherwise with no iron supplementation, only a single iron-regulated gene, fptA (encoding the ferric-pyochelin outer membrane receptor), was identified as being upregulated (Wang et al., 1996).

Global changes in B. cenocepacia gene expression have also been analyzed in a synthetic CF sputum medium (SCFM). SCFM is a defined (i.e., sputum-free) medium containing the average concentrations of ions, free amino acids, glucose and lactate as those found in the sputum of CF patients and has been shown to support similar growth rates and elicit similar changes in expression of some subsets of genes in P. aeruginosa to those observed during growth in MOPSsputum medium (Palmer et al., 2007). It also contains 3.6µM ferrous iron (i.e., similar to that of MOPS-sputum medium). Although the iron concentration of SCFM may be low enough to cause upregulation of P. aeruginosa siderophore gene expression relative to iron replete conditions, these genes were not (unsurprisingly) upregulated relative to cells growing in a MOPS-glucose based mineral salts medium containing an almost identical concentration of iron (Palmer et al., 2007). The contrasting high level of siderophore gene expression observed in P. aeruginosa growing in MOPS-sputum medium relative to cells growing in MOPS-glucose medium (Palmer et al., 2005) was rationalized on the basis that CF sputum also contains iron sequestering components which have been proposed to restrict the availability of iron (Palmer et al., 2007). Other analogous attempts to mimic CF sputum conditions using semi-synthetic media, such as ASMDM or Modified ASMDM, and comparing gene expression in P. aeruginosa to that in cells growing in standard laboratory media have likewise not suggested a requirement for the main siderophore-mediated iron acquisition systems in synthetic sputum-like media [apart from one case where a small (2-fold) increase in some pyochelin biosynthesis and transport genes was observed; Fung et al., 2010; Hare et al., 2012].

In apparent contrast to the observations with P. aeruginosa, the entire ornibactin gene cluster of B. cenocepacia, as well as genes encoding a number of TBDRs and a cluster of genes that encode a putative bacterioferritin-associated ferredoxin and a TonB system (BCAL2290-BCAL2293) were found to be strongly upregulated during growth of B. cenocepacia J2315 in SCFM in comparison to growth in soil extract medium, although the pyochelin biosynthesis genes were not upregulated in SCFM (Yoder-Himes et al., 2010). These results also contrast with those observed for B. cenocepacia growing in a basal salts medium supplemented with glucose, casamino acids and CF sputum (see above; Drevinek et al., 2008). The most likely reason for the latter difference is that there was a 10-fold higher concentration of iron in the CF sputum medium in comparison to SCFM. It might seem intriguing that SCFM stimulates ornibactin gene expression in B. cenocepacia but not expression of pyoverdine genes in P. aeruginosa. However, again one must proceed with caution in interpreting these data, as the fold induction of gene expression in P. aeruginosa cells growing in SCFM was expressed relative to medium containing the same iron concentration, whereas the comparator for the B. cenocepacia experiment were cells growing in soil extract medium, which has an indeterminate iron concentration. In fact, it is likely that relative to cells growing in an iron replete, nutrient rich laboratory medium, both the pyoverdine and ornibactin gene clusters may actually be upregulated in cells growing in SCFM.

To summarize the above, it is difficult to make informed judgements regarding the requirement or otherwise for various iron acquisition systems based on gene expression analysis in cells growing in defined or semi-defined media that seek to mimic CF conditions when the concentration of iron and its relative availability may not accurately reflect the situation in the patient. As an example, we note that in some more recent attempts to mimic CF conditions using synthetic media, a higher iron concentration has been used by including ferritin to better reflect the prevailing view that the iron content of the CF lung is relatively high (Hare et al., 2012). Moreover, experiments involving media which incorporate CF-derived sputa may be prone to a high degree of experimental variation according to the disease severity and consequential iron status of the sputa. Finally, obvious though this must appear, consideration of the comparator is fundamentally important in assessing whether or not iron acquisition genes are upregulated in such media.

The difficulty in reproducing CF conditions in vitro can be bypassed by measuring gene expression in bacterial pathogens that are present in sputum following collection of samples from CF patients. As an example, in one such study, a microarray experiment was performed using P. aeruginosa mRNA isolated from sputum obtained from a single patient (Son et al., 2007). In this investigation, genes specifying the biosynthesis of pyochelin were upregulated but not those encoding the biosynthesis and transport of the major siderophore pyoverdine. In a later study, an RT-qPCR analysis of gene expression in P. aeruginosa strains that were present in the lungs of a cohort of CF patients suggested that the siderophore pyoverdine was likely to contribute to iron acquisition in this context (Konings et al., 2013), and indeed the presence of the siderophore could be detected in CF sputa (Martin et al., 2011). In contrast, to our knowledge, studies involving direct sampling of RNA from Bcc bacteria colonizing the CF lung have not been carried out. However, sampling of B. cenocepacia mRNA directly from an animal model of a chronic lung infection has been carried out (O'Grady and Sokol, 2011). As discussed earlier, the ability to biosynthesise ornibactin plays an important role in chronic infections of the rat lung by B. cenocepacia (7 and 14 days post-infection; Visser et al., 2004). However, microarray data in which gene expression in B. cenocepacia K56-2 cells that were recovered from the rat lung model 3 days post-infection was compared to cells that were grown to stationary phase in a nutrient-rich broth (iron replete medium), showed no difference in ornibactin gene expression levels (O'Grady and Sokol, 2011). Therefore, the simplest interpretation of these data is that ornibactin is not required to establish an infection in this particular model system but it is required for persistence. Moreover, in contrast to the STM analysis of Hunt et al. (2004), the microarray analysis did not reveal a difference in expression of the bhuR and bhuS genes (referred to as huvA and hmuS by the authors) in the rat lung model compared to growth in iron replete medium (O'Grady and Sokol, 2011). As the STM study was conducted with animals that were infected for 10 days, one possible explanation is that haem utilization becomes important for longer term infections as has been observed in P. aeruginosa (see below).

Transcriptomics has been used to monitor B. cenocepacia adaptation to the host over time, although in this case RNA was isolated following in vitro culture of the bacteria. In one such study, gene expression was compared in two clonal variants that were isolated 3 years apart from a CF patient who died of cepacia syndrome. This study revealed that in the later clone, seven genes located within the ornibactin gene cluster were upregulated 1.9- to 5.9-fold compared to those in the earlier isolate when both strains were cultured on a nutrient-rich agar. Other genes potentially involved in iron uptake were also found to be more transcriptionally active in this isolate, including three genes that encode TBDRs that are not involved in ornibactin or pyochelin uptake, and two genes, bhuR and bhuS (referred to as huvA and hmuS by the authors), from the bhuRSTUV gene cluster that is proposed to be required for the uptake of haem (**Figure 4C**; Mira et al., 2011). A subsequent proteomic study employing the same pair of isolates, along with a third isolate collected just before the death of the patient, showed that the two later isolates exhibited an increased abundance of four proteins involved in siderophore-mediated iron uptake compared to the earliest clone (Madeira et al., 2013). These proteins included two TBDRs (one of which was FptA) that had increased in abundance by <2-fold, and one component of the ferric-ornibactin cytoplasmic membrane transporter (OrbC) which showed a relatively small increase in abundance (∼50%) in the third isolate compared to the first. However, in contrast to the transcriptomic study, a general increase in abundance of iron acquisition proteins was not observed. Moreover, based on a CAS assay, the latter two isolates were considered to be more tolerant to low iron concentrations (siderophore production was induced at 4 or 5µM iron, whereas in the primary isolate siderophore production was upregulated at 6µM iron).

In a later study carried out on the same sequential clonal isolates, the upregulation of ornibactin gene expression observed in response to the iron chelating activity of exogenously added pyoverdine was significantly less pronounced in the last isolate compared to the earlier isolates (Tyrrell et al., 2015). These observations are consistent with a scenario in which selection for genetic alterations has occurred upon long term colonization of the CF lung that lead to a decreased reliance on siderophoredependent iron acquisition by the bacterium. This may indicate a switchover to another means of iron acquisition or it may reflect a general downregulation of iron acquisition mechanisms due to increased inflammatory damage that occurs in the CF lung as the disease progresses and the consequent increased availability of iron (Cohen and Prince, 2012).

The results are analogous to those obtained from studies carried out on P. aeruginosa, in which it was observed that during infection of the CF lung mutations accrue in the bacterial population that result in a reduction or abolition of production of the major siderophore, pyoverdine (Marvig et al., 2014; Nguyen et al., 2014 and refs within). Evidence was also provided for increased haem usage as an iron source concomitant with a reduction in pyoverdine synthesis during the later stages of infection (Marvig et al., 2014; Nguyen et al., 2014). Consistent with this, in a separate microarray study carried out on mRNA isolated directly from a P. aeruginosa-infected CF patient, two genes from the pyochelin gene cluster (pchA and pchC) were observed to be upregulated 2- to 3-fold, but no pyoverdine genes were identified as being upregulated (Son et al., 2007). Although these results were obtained from work on a different CF pathogen, they are consistent with the idea that siderophoremediated iron acquisition in the CF lung may not be of major importance to some pathogenic bacteria, particularly later on in an infection.

While these types of analyses may provide useful pointers as to the iron acquisition mechanisms that may be employed by pathogenic bacteria colonizing the CF lung, particularly if the isolate has accrued mutations that have inactivated a particular uptake system, the fact that these systems are subject to genetic regulation means that a true understanding of the iron uptake mechanisms at play during an infection will also require analysis of bacterial gene expression in vivo, i.e., in the CF lung.

Studies on the effects of other environmental parameters on global gene expression in B. cenocepacia have also been conducted that have potential implications for iron acquisition in this organism. In one investigation, the effect of oxygen depletion on gene expression was assessed, as there is evidence to suggest that within the CF lung a steep oxygen gradient is generated due to increased activity of airway epithelial Na+-K+- ATPase pumps and excessive mucin secretion. This results in the deepest layers of mucus providing a hypoxic environment that can support high densities of micro-oxic and anaerobic microbes in the CF lung (Tunney et al., 2008). In this study it was shown that B. cenocepacia can grow in an atmosphere with oxygen concentrations as low as 0.1% (Pessi et al., 2013). The possibility that B. cenocepacia may therefore occupy a low oxygen niche when colonizing the lungs of CF patients prompted an RNA-seq and shotgun proteome analysis of B. cenocepacia H111 growing under micro-oxic conditions (0.5% oxygen) in comparison to aerobic growth (21% oxygen). RNA-seq showed strong down regulation of the pyochelin biosynthesis genes pchD, pchE, and pchF in addition to the gene encoding the ferric-pyochelin outer membrane receptor, FptA, in micro-oxic conditions. Overall production of siderophores was also reduced in micro-oxic conditions as assessed by the CAS agar assay (Pessi et al., 2013). This kind of environment promotes the increased stabilization of Fe(II) and may favor the use of ferrous iron transport systems, such as the FtrABCD system in B. multivorans, over siderophore based systems.

Another environmental parameter that appears to affect the expression of iron acquisition genes that may be pertinent in the context of CF lung infections is oxidative stress. Thus, it has been observed that exposure of B. cenocepacia biofilms to hydrogen peroxide causes increased expression of the first few genes of the ornibactin gene cluster (orbS, orbH, and orbG; Peeters et al., 2010). While the oxygen and oxidative stress status of the niche occupied by B. cenocepacia in the CF lung has not yet been established, these observations bring into focus the requirement to mimic, as close as possible, the conditions of the CF lung when assessing the potential role of iron acquisition mechanisms in the virulence of respiratory pathogens.

To summarize, there appears to be a role for ornibactinmediated iron acquisition in the virulence of B. cenocepacia in both vertebrate and invertebrate models of infection, including infections of the respiratory tract. However, in the context of a CF infection the role of ornibactin is less certain and there is a possibility that haem acquisition may come into play. Moreover, the severity of the disease and the particular niche that is occupied by Bcc bacteria in the lung—for example, whether it is oxygen rich or hypoxic—are also likely to dictate which iron acquisition mechanism(s) are primarily deployed. Currently, there is little evidence to support a role for the secondary siderophores in colonization and persistence by Bcc bacteria, although this is partly due to the fact that no studies on the possible roles of cepaciachelin and cepabactin have been reported. Further studies are required to establish the relative importance of the various iron acquisition mechanisms available to Bcc bacteria for colonization of the CF lung.

#### *BURKHOLDERIA PSEUDOMALLEI* AND *BURKHOLDERIA MALLEI*

B. pseudomallei is both an environmental saprophyte and the causative agent of the tropical disease melioidosis (Wiersinga et al., 2006). This disease is endemic in South East Asia and Northern Australia with sporadic cases increasingly reported in other topical regions (Perumal Samy et al., 2017). Contraction of the disease is via cuts and abrasions, inhalation or ingestion (Wiersinga et al., 2012). Although apparently healthy individuals can become infected, conditions such as diabetes and liver disease are highly associated risk factors of melioidosis (Perumal Samy et al., 2017). The disease manifests in a range of forms from acute infections, chronic reoccurring infections, fatal sepsis or even persistent asymptomatic infections lasting for up to 60 years (White, 2003; Ngauy et al., 2005). The lung is the most commonly infected organ with the liver, spleen, skeletal muscle and prostate other sites of infection (White, 2003).

B. mallei is a host restricted obligate pathogen with no known environmental reservoir (Whitlock et al., 2007). This bacterium causes the zoonotic disease glanders, which is spread directly or indirectly through secretions and excretions of infected animals. This disease is chronic in horses and an acute form of the disease occurs in donkeys and mules. Infection of humans is rare and is usually the result of occupational exposure (Verma et al., 2014). B. mallei is considered to be a clone of B. pseudomallei that has undergone a process of genome reduction during host adaptation (Godoy et al., 2003; Nierman et al., 2004).

#### Experimental Evidence for the Role of Iron Acquisition Systems in the Virulence of *B. pseudomallei*

The role of iron in B. pseudomallei and B. mallei virulence has been less well-studied than in B. cenocepacia mainly due to the increased hazard associated with handling these organisms which necessitates a more stringent level of containment and has also restricted the range of available selective genetic markers. Nevertheless, with the recent development of biosafety compliant tools for the genetic manipulation of these bacteria important progress has been made in our understanding of their iron acquisition systems and the role that these systems play in virulence. It is now well-established that access to iron is important for virulence by B. pseudomallei. For example, it has been demonstrated that the severity of B. pseudomallei infection of A549 macrophages and HeLa cells is increased if the cell lines are supplemented with iron. Thus, B. pseudomallei K96243 formed more plaques on iron-supplemented HeLa cells and invasion was significantly increased in iron-supplemented A549 cells. Furthermore, the intracellular survival of B. pseudomallei in A549 monolayers and the ability to induce MNGC formation, was greater when the A549 monolayers were supplemented with iron compared to non-iron supplemented controls (Amornrit et al., 2012). Interestingly, iron has also been shown to down regulate one of the type VI secretion systems (specifically T6SS-5) in B. pseudomallei and B. mallei (Burtnick and Brett, 2013). T6SS-5 is essential for virulence in hamsters and is required for multinucleated giant cell formation in infected tissue culture monolayers, a phenomenon that may facilitate cell-to-cell spread of the bacterium (Burtnick et al., 2011) [Note: T6SS-5, as designated by Shalom et al. (2007), is also referred to as T6SS-1 or the cluster 1 type VI secretion system by some authors (Schell et al., 2007; Burtnick et al., 2011)].

A limited number of studies have been carried out to determine the role of siderophores and other iron uptake systems on the virulence of B. pseudomallei and B. mallei. Both species produce the siderophore malleobactin (or more precisely, malleobactin E), that is structurally related to ornibactin (Yang et al., 1991; Alice et al., 2006; Franke et al., 2013, 2015; **Figure 2**). This siderophore was shown to be able to acquire iron from human transferrin and lactoferrin (Yang et al., 1993). B. pseudomallei, like several members of the Bcc, also produces pyochelin as a secondary siderophore (Alice et al., 2006; Kvitko et al., 2012), whereas in B. mallei, which has undergone extensive genome reduction, the gene cluster needed for pyochelin production is absent (Esmaeel et al., 2016). This is consistent with other evidence that pyochelin has a limited role in Burkholderia virulence, and loss of the ability to manufacture this siderophore in B. mallei may point to a more important role of pyochelin in environmental survival among other Burkholderia species. Moreover, B. mallei also produces reduced levels of malleobactin (as determined by the CAS assay) compared to B. pseudomallei and B. thailandensis which may also reflect the narrower range of niches in which it inhabits (Ong et al., 2004). B. pseudomallei may also secrete a third compound with iron chelating activity, although this compound has not yet been characterized (Kvitko et al., 2012).

The BPSS0240-BPSS0244 genes of B. pseudomallei K96243 were proposed to serve as a haem uptake system and were observed to be upregulated during growth under low iron conditions (Tuanyok et al., 2005). Based on an IVET screen, this system was shown to be induced during growth of B. pseudomallei NCTC 10274 within macrophages, suggesting that it may play an important role in iron acquisition during intracellular survival. In this study the authors referred to the uptake system as the Bhu (Burkholderia haem uptake) system based on the Pseudomonas Phu system (Shalom et al., 2007). By analogy with the Phu system and the related Shu system of Shigella, the Bhu system consists of the TBDR, BhuR (BPSS0244), a periplasmic binding protein-dependent type II ABC transporter (BhuT-BhuV) for translocation of haem across the cytoplasmic membrane, and a cytoplasmic haem binding protein (BhuS) that plays a role in haem trafficking and may also initiate haem degradation (**Figure 5D**; O'Neill and Wilks, 2013; Naoe et al., 2016; see Choby and Skaar, 2016) for a review. Based on a bioinformatic analysis, B. pseudomallei K96243 was predicted to specify two additional outer membrane receptors for haem (BPSL2724 and BPSS1742) with the former also associated with an ABC transporter system (Harland et al., 2007). However, neither of these systems shows a strong homology to characterized haem uptake systems. The BPSL2721-BPSL2724 and BPSS0240-BPSS0244 (Bhu) transport systems were later referred to as the Hem and Hmu systems, respectively, in a study conducted on strain 1710b (Kvitko et al., 2012). These authors observed that whereas deletion of the bhu/hmu locus compromised the ability of B. pseudomallei to utilize haem or hemoglobin as iron sources, deletion of the hem locus did not abrogate the ability to utilize haem, strongly indicating that the Bhu/Hmu system serves as the haem uptake system in this organism.

Kvitko and colleagues went on to explore the relative contributions of siderophore- and haem-mediated uptake systems to B. pseudomallei virulence in an acute murine melioidosis model following intranasal infection. They found that although utilization of lactoferrin-bound iron in vitro relied on malleobactin, inactivation of the malleobactin uptake system in strain 1710b did not cause attenuation in the murine melioidosis model (Kvitko et al., 2012). Interestingly, a strain that was defective for malleobactin, pyochelin, and haem uptake was also fully virulent in the melioidosis model, although the titres of bacteria recovered from some organs was significantly lower. As this mutant could still grow with ferritin as an iron source, it was suggested that another iron uptake system is present that could compensate for the loss of the other uptake pathways. There are at least two possibilities that could account for this. First, the CAS agar assay indicated that B. pseudomallei mutants lacking the ability to biosynthesise malleobactin and pyochelin specify an additional secreted ironchelating compound of unknown identity (Kvitko et al., 2012). Secondly, B. pseudomallei encodes the FtrABCD system which may play a role in iron acquisition during infection (Mathew et al., 2014). Perhaps surprisingly, although B. pseudomallei 708a (a strain containing a >130 kb genome deletion that removes the malleobactin synthesis genes) was virulent in the mouse model used by Kvitko and colleagues, it was attenuated in G. mellonella (Wand et al., 2011; Kvitko et al., 2012).

The outer membrane receptors required for uptake of ferricsiderophore complexes and haem require the action of the cytoplasmic membrane-anchored TonB-ExbB-ExbD complex to energize transport of these iron sources. A B. mallei tonB mutant, which is unable to internalize ferric-malleobactin or haem, was shown to be completely attenuated in mice at 10<sup>5</sup> CFU, in comparison to an LD<sup>50</sup> of 7.4 x 10<sup>4</sup> CFU for the wild type, and it also exhibited lower titres in target organs (lungs and spleen). Supplementation of the medium with ferrous iron, which is assimilated by a TonB-independent mechanism, partially restored virulence and led to higher titres of tonB mutant bacteria in the spleen, indicating that the inability to acquire iron is the main reason for the loss of virulence in the mutant (Mott et al., 2015). However, as TonB systems in other species have been shown to be involved in transport of other large molecules, including some enzyme cofactors (thiamine, cobalamin) and saccharides (sialic acid, sucrose, maltodextrins) the possibility exists that there are other transport requirements served by the B. mallei TonB system that must be met for full virulence (Schauer et al., 2008; Roy et al., 2010).

Another study in which potential novel therapeutics were screened for their ability to counteract B. pseudomallei killing of C. elegans suggested an important role for iron acquisition in virulence. Exposure of B. pseudomallei to the plant alkaloid curcumin (diferuloylmethane), prior to their administration to C. elegans, significantly enhanced the survival of the worms (Eng and Nathan, 2015). Curcumin is structurally related to bis-catecholates, which include bacterial siderophores such as azotochelin, cepaciachelin, and serratiochelin, and is itself a known iron chelator that may serve to cause a decrease in iron availability (Jiao et al., 2006). Accordingly, microarray data showed that B. pseudomallei treated with curcumin upregulated genes for iron transport, including those for malleobactin, pyochelin and haem uptake, as well as genes encoding the TonB system. Moreover, genes encoding the biosynthesis of malleobactin and pyochelin were upregulated and increased secretion of siderophores by treated B. pseudomallei was observed. The authors suggest that the anti-infective effects observed by curcumin are due to the bacteria diverting their metabolism away from virulence and toward iron uptake to ensure growth and maintenance. These observations provide suggestive evidence for the importance of iron acquisition for virulence in this organism.

#### *BURKHOLDERIA GLADIOLI*

B. gladioli was originally identified as a pathogen of Iridaceae, specifically irises and gladioli, and then later was recognized as the cause of infections in certain groups of immunocompromised patients, particularly among those with CF and CGD (Boyanton et al., 2005; Kennedy et al., 2007). The iron acquisition systems of this organism have yet to be determined. However, the genome sequence offers a few clues. We note that it does not contain orthologs of the genes for the biosynthesis of ornibactin, malleobactin, pyochelin, or cepaciachelin (**Table 1**). However, it does contain orthologs of the bhuRSTUV genes. Unusually, these genes are organized into two separate operons that are present on chromosome 1 (bhuRST) and chromosome 2 (bhuUV) (results not shown), but nevertheless suggest that B. gladioli may be able to utilize haem as an iron source during infection of human hosts. The B. gladioli genome also encodes a FtrABCD system. It is not known whether one or both of these systems is important for establishing an infection in humans.

#### CLOSING REMARKS

To conclude, we are still far from ascertaining the relative importance of the different iron acquisition mechanisms available to the Burkholderia that are brought to bear during infection of a human host. A number of studies have been carried out with B. cenocepacia to address this question which have provided what might at first appear to be conflicting results. However, as the model systems are so varied, and the iron content and/or availability in these systems is also markedly different, we believe that most of these apparent contradictory results can be rationalized. Nonetheless, in terms of whether high affinity iron acquisition systems are essential for successful colonization of the CF lung and which ones are deployed, further studies are required. Due to the difficulty of working safely with B. pseudomallei and B. mallei fewer studies have been conducted, but it would appear that neither the known siderophores of B. pseudomallei nor its haem uptake system are important for systemic melioidosis. There is great scope for further work into the role of iron acquisition in the virulence of this important group of bacteria.

#### AUTHOR CONTRIBUTIONS

AB and MT contributed equally to writing the manuscript. MT prepared the figures, table and Supplementary Material.

#### ACKNOWLEDGMENTS

We would like to thank the BBSRC who are supporting the current work of AB and MT on the regulation of iron acquisition in Burkholderia (research grant BB/M003531/1).

#### REFERENCES


#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fcimb. 2017.00460/full#supplementary-material


Cryptococcus neoformans. PLoS Biol. 4:e410. doi: 10.1371/journal.pbio. 0040410


is induced upon invasion of macrophages. Microbiology 153, 2689–2699. doi: 10.1099/mic.0.2007/006585-0


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

Copyright © 2017 Butt and Thomas. 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.

## Iron and Virulence in *Francisella tularensis*

#### Girija Ramakrishnan\*

*Department of Medicine/Division of Infectious Diseases, University of Virginia, Charlottesville, VA, USA*

*Francisella tularensis*, the causative agent of tularemia, is a Gram-negative bacterium that infects a variety of cell types including macrophages, and propagates with great efficiency in the cytoplasm. Iron, essential for key enzymatic and redox reactions, is among the nutrients required to support this pathogenic lifestyle and the bacterium relies on specialized mechanisms to acquire iron within the host environment. Two distinct pathways for iron acquisition are encoded by the *F. tularensis* genome- a siderophore-dependent ferric iron uptake system and a ferrous iron transport system. Genes of the Fur-regulated *fslABCDEF* operon direct the production and transport of the siderophore rhizoferrin. Siderophore biosynthesis involves enzymes FslA and FslC, while export across the inner membrane is mediated by FslB. Uptake of the rhizoferrin- ferric iron complex is effected by the siderophore receptor FslE in the outer membrane in a TonB-independent process, and FslD is responsible for uptake across the inner membrane. Ferrous iron uptake relies largely on high affinity transport by FupA in the outer membrane, while the Fur-regulated FeoB protein mediates transport across the inner membrane. FslE and FupA are paralogous proteins, sharing sequence similarity and possibly sharing structural features as well. This review summarizes current knowledge of iron acquisition in this organism and the critical role of these uptake systems in bacterial pathogenicity.

#### *Edited by:*

*Pierre Cornelis, Vrije Universiteit Brussel, Belgium*

#### *Reviewed by:*

*Charles Martin Dozois, Institut National de la Recherche Scientifique, Canada Erin R. Murphy, Ohio University, USA Paul Edward Carlson, US Food and Drug Administration, USA*

#### *\*Correspondence:*

*Girija Ramakrishnan girija@virginia.edu*

*Received: 30 January 2017 Accepted: 16 March 2017 Published: 04 April 2017*

#### *Citation:*

*Ramakrishnan G (2017) Iron and Virulence in Francisella tularensis. Front. Cell. Infect. Microbiol. 7:107. doi: 10.3389/fcimb.2017.00107* Keywords: Siderophore, FeoB, intracellular pathogen, FslE, FupA, TonB-independent, *Francisella tularensis*

### INTRODUCTION

Francisella tularensis, the etiological agent of the zoonosis tularemia, is a Gram-negative gamma-proteobacterium with a small genome of 1.89 Mb (Sjöstedt, 2007). The species is further differentiated into three subspecies, of which tularensis causes a more severe disease than holarctica, while mediasiatica is less well-studied. The closely related species F. novicida, considered to be of a more ancestral lineage (Svensson et al., 2005) and with a genome sequence identity of ∼98% (Larsson et al., 2009) is an opportunistic human pathogen, but can cause a virulent tularemia-like disease in mice.

In the laboratory, studies have largely focused on virulent strain Schu S4 of the tularensis subspecies, the attenuated live vaccine strain (LVS) derived from a holarctica isolate and the strain U112 of F. novicida. These three isolates share many biological attributes although their genetic and functional differences significantly impact virulence (Jones et al., 2012; Celli and Zahrt, 2013; Kingry et al., 2014). For this reason, all three strains are referred to in this review as F. tularensis unless specifically identified in order to highlight particular differences.

F. tularensis is a facultative intracellular pathogen infecting a wide variety of cells, including mammalian and arthropod cells (Ozanic et al., 2015). Following uptake into the macrophage, the bacteria at first reside within a phagosome, but then rapidly escape into the cytoplasm. Phagosomal escape is dependent on the igl operon and associated genes in the Francisella Pathogenicity Island (FPI) that encode components of a putative Type VI secretion system (Barker et al., 2009; de Bruin et al., 2011). The bacteria replicate to high numbers in the cytoplasm resulting finally in apoptotic death of the host cell. Adaptation to the specialized intracellular lifestyle is associated with evolutionary loss of genes for many metabolic pathways (Rohmer et al., 2007; Larsson et al., 2009), but F. tularensis has retained or evolved mechanisms to efficiently acquire essential nutrients within the intracellular niche of the different cell types that it infects (Meibom and Charbit, 2010).

Mice have been extensively used to model animal infection (Lyons and Wu, 2007). Phagocytic cells are thought to be the first infected (Hall et al., 2008); subsequently infection is disseminated to other tissues in the body. F. tularensis exercises several strategies to evade immune responses and is able to replicate to high levels in the liver, spleen and lungs before the immune system is provoked to respond with a destructive cytokine storm (Sharma et al., 2011; Jones et al., 2012).

#### Iron and Francisella

Francisella requires iron for essential cellular functions. Early studies reported that infection with F. tularensis induces an iron-withholding response typical of the innate nutritional immunity defense mechanism (Pekarek et al., 1969). However, the intracellular pathogen manipulates host cell iron metabolism to support growth; LVS induces infected macrophages to increase iron flow through the cell by enhanced expression of the transferrin receptor TfR1 for uptake of iron and in parallel, increased expression of Dmt1 that moves endosomal iron into the cytoplasm and a slight increase in ferroportin that promotes outflow of iron from the cell (Pan et al., 2010). A functional Nramp1protein that also transports endosomal iron into the cytoplasm restricts growth of endosome-resident bacteria but enhances Francisella growth (Kováˇrová et al., 2000, 2002), highlighting the importance of cytoplasmic iron availability for pathogenesis. However, the nature of the host iron sources accessed by the organism remains to be characterized. Ironlimitation restricts growth of bacteria in culture (Deng et al., 2006; Sullivan et al., 2006) as well as within the macrophage (Fortier et al., 1995).

As might be predicted, growth of F. tularensis is inhibited by gallium, which competes with ferric iron for uptake and also interferes with iron-dependent biological processes (Olakanmi et al., 2010; Lindgren and Sjöstedt, 2016). Inhibition of the ironassociated enzymes catalase and superoxide dismutase leads to increased susceptibility to oxidative stress (Bakshi et al., 2006; Lindgren et al., 2007; Olakanmi et al., 2010; Binesse et al., 2015).

Iron metabolism appears to differ among F. tularensis isolates. LVS expresses higher levels of bacterioferritin as compared to Schu S4 (Hubálek et al., 2003, 2004). Consistent with these findings, isolates of the holarctica subspecies have greater iron stores than tularensis isolates, and since iron is closely associated with generation of reactive oxygen species, holarctica strains are more susceptible to oxidative stress (Lindgren et al., 2011).

#### Fur and Iron Regulation of Genes

A fur ortholog is encoded in the F. tularensis genome and the predicted Fur protein contains elements known to be important for Fur function (Pérard et al., 2016). The fur gene is adjacent to the fsl operon encoding components of a siderophore- mediated iron uptake pathway, and a canonical Furbox is located upstream of the first gene of the operon, fslA (**Figure 1A**). Expression of the fsl operon is induced in iron-limiting media (Deng et al., 2006; Sullivan et al., 2006; Buchan et al., 2008). Loss of the fur gene results in deregulated transcription of the fsl operon and increased siderophore production (Buchan et al., 2008; Ramakrishnan et al., 2008). Expression of the inner membrane ferrous iron transporter feoB is also upregulated in a fur mutant (Pérez and Ramakrishnan, 2014).

Exposure to iron limitation was also shown to increase virulence of an F. tularensis isolate, suggesting that pathogenicity is influenced by iron levels (Bhatnagar et al., 1995). Microarray analysis of RNA indicated that besides the fsl operon, transcription of the igl operon was increased under iron limitation (Deng et al., 2006). Proteomic analysis also confirmed that iron limitation results in increased levels of the IglC protein (Lenco et al., 2007). However, although an fslB-lacZ reporter could be repressed by overexpression of Fur, an iglB-lacZ reporter was not similarly repressible (Buchan et al., 2008), suggesting that a mechanism besides Fur regulates the igl genes in response to iron levels in F. tularensis.

#### Siderophore-Mediated Iron Acquisition

A "growth inducing substance (GIS)" that promoted growth of F. tularensis bacteria from small inocula was reported in the 1960s (Halmann and Mager, 1967; Halmann et al., 1967); in all likelihood this substance was the siderophore now identified as the polycarboxylate rhizoferrin (Drechsel et al., 1991; Thieken and Winkelmann, 1992; Sullivan et al., 2006). Rhizoferrin is structurally simple, comprising 2 citrate moieties linked through amide bonds to a putrescine backbone (**Figure 1B**). Originally identified as a fungal siderophore (Drechsel et al., 1991; Thieken and Winkelmann, 1992), rhizoferrin was subsequently also isolated from a strain of the bacterium Ralstonia pickettii (Münzinger et al., 1999). The identification of the siderophore made by F. tularensis as rhizoferrin (Sullivan et al., 2006), and the subsequent identification of the Legionella pneumophila siderophore legiobactin as also rhizoferrin (Burnside et al., 2015) suggests that this siderophore may be more widely prevalent in bacteria than suspected. Structurally related siderophores made by bacteria include staphyloferrin A (Konetschny-Rapp et al., 1990) and corynebactin (Zajdowicz et al., 2012), where D-ornithine and lysine, respectively, constitute the siderophore backbone in place of the putrescine present in rhizoferrin. The R. pickettii rhizoferrin was shown by CD spectroscopy to be an S-S enantiomer in contrast to the R-R fungal molecule (Münzinger et al., 1999). Whether all bacterial rhizoferrins adopt the S-S conformation is not clear, but bacteria making rhizoferrin are capable of utilizing the fungal form for iron uptake (Münzinger et al., 1999; Kiss et al., 2008).

Genes for synthesis and transport of Francisella rhizoferrin are located on the siderophore operon fslABCDEF (also designated

figABCDEF; (Deng et al., 2006; Sullivan et al., 2006; Milne et al., 2007; Buchan et al., 2008; Ramakrishnan et al., 2008); **Figures 1A,C**). Analysis of individual mutants as well as complementation of a strain carrying a complete deletion of the fslA-F genes helped to determine the roles played by the different genes in siderophore-mediated iron acquisition, as detailed below.

fslA and fslC and share homology with genes found in siderophore biosynthetic loci of other bacteria. FslA is similar to the aerobactin synthetases IucA/IucC and a member of the nonribosomal peptide synthetase-independent siderophore (NIS) synthetases, enzymes that assemble non-peptide siderophores using dicarboxylic acids and diamines or amino-alcohols (Challis, 2005). FslC is predicted to be a member of the pyridoxal phosphate-dependent decarboxylases. Mutant analysis demonstrated that both fslA and fslC are required for Francisella rhizoferrin production (Deng et al., 2006; Sullivan et al., 2006; Lindgren et al., 2009; Thomas-Charles et al., 2013). Rhizoferrin biosynthesis in F. tularensis, involving just two dedicated enzymes, may be the simplest siderophore biosynthetic pathway identified thus far.

fslB encodes a transporter of the Major Facilitator superfamily (MFS) and deletion of this gene in F. novicida results in reduced levels of siderophore activity in the culture medium (Kiss et al., 2008). Additionally, detection of siderophore activity in culture supernatants of an LVS 1fslA-F mutant required complementation with the fslB gene in addition to the biosynthetic genes fslA and fslC (Pérez et al., 2016). These observations support a role for FslB in export of the siderophore across the cytoplasmic membrane. How the siderophore is channeled through the outer membrane into the extracellular space is currently not known.

fslD encodes an inner membrane MFS protein and deletion of this gene was found to have little effect on siderophore production (Kiss et al., 2008). The role of this protein in siderophore-mediated iron uptake across the cytoplasmic membrane was deduced on the basis of complementation studies: a Schu S4 1fslA-F mutant is able to transport <sup>55</sup>Fe3<sup>+</sup> complexed to siderophore only when complemented with the fslD gene in addition to the fslE receptor gene (see below; Pérez et al., 2016).

fslE, the fifth gene in the operon, encodes an outer membrane protein unique to the Francisella genus Larsson et al., 2005; Huntley et al., 2007). 1fslE mutants are impaired for growth in iron-limiting media and are unable to utilize exogenous siderophore for growth (Kiss et al., 2008; Ramakrishnan et al., 2008). In transport assays, fslE mutants proved incapable of siderophore-mediated <sup>55</sup>Fe3<sup>+</sup> uptake, establishing a role for FslE as receptor for the siderophore (Ramakrishnan et al., 2012). FslE can also transport the iron mimic gallium in complex with rhizoferrin as shown by the resistance of fslA and fslE mutants to gallium (Pérez et al., 2016).

The last gene of the fsl operon, fslF, varies structurally among the tularensis and the novicida species, being truncated in the tularensis isolates. Studies with a 1fslF mutant in Schu S4 indicate that the gene does not influence iron transport in F. tularensis (Pérez et al., 2016).

Siderophore-mediated iron transport by outer membrane receptors in Gram-negative bacteria is typically dependent on the proton motive force transduced by the TonB-ExbB-ExbD complex (Noinaj et al., 2010). The Francisella genome, however, does not encode orthologs of tonB, exbB, and exbD, implying that alternative mechanisms must facilitate siderophore -iron uptake.

#### Ferrous Iron Uptake

The F. tularensis genome encodes an inner membrane ferrous iron transport system comprising unlinked genes feoA and feoB. 1feoB mutants of LVS and Schu S4 are deficient for growth on iron-limiting media (Thomas-Charles et al., 2013; Pérez and Ramakrishnan, 2014; Pérez et al., 2016). <sup>55</sup>Fe uptake assays demonstrated that the F. tularensis 1feoB mutants are completely deficient in ferrous iron uptake (Pérez and Ramakrishnan, 2014; Pérez et al., 2016), implying that the Feo system is the sole ferrous iron transporter across the inner membrane (**Figure 2**). It is likely that FeoA supports FeoB function as seen in Salmonella Typhimurium (Kim et al., 2012) and Vibrio cholerae (Weaver et al., 2013; Stevenson et al., 2016).

Given the soluble nature of ferrous iron, the general assumption has been that it diffuses into the periplasmic space through non-specific porin proteins in the outer membrane. However, growth and <sup>55</sup>Fe transport assays indicate that F. tularensis is capable of high-affinity uptake of ferrous iron mediated by the specific outer membrane protein FupA (Ramakrishnan et al., 2012). FupA was initially characterized as

a virulence factor in Schu S4 (Twine et al, 2005) and found to influence bacterial intracellular replication (Twine et al, 2005; Asare et al., 2010). An involvement in iron acquisition was established by the finding that Schu S4 1fupA mutant grew poorly under iron limitation, had lowered internal iron levels and was deregulated for siderophore production (Lindgren et al., 2009). A link to ferrous iron uptake was indicated by the finding that ferrous and ferric iron supplements supported growth of a fupA mutant to different extents on agar plates (Ramakrishnan et al., 2012). <sup>55</sup>Fe transport assays clearly demonstrated that the fupA mutant was unable to transport ferrous iron at limiting concentrations (∼0.1 µM) although low affinity ferrous iron transport at high concentrations (∼3 µM) could still be observed, and siderophore-iron uptake was not perturbed (Ramakrishnan et al., 2012). These findings suggest that FupA serves as a F. tularensis adaptation to efficiently acquire ferrous iron even in low abundance settings when general diffusion-based transport across the outer membrane might prove inadequate (**Figure 2**).

### FslE and FupA: Related High-Affinity Iron Transport Proteins

FupA and FslE are paralogs belonging to a family of proteins unique to Francisella (Larsson et al., 2005). Both proteins have been localized to the outer membrane of F. tularensis (Huntley et al., 2007; Ramakrishnan and Sen, 2014) and share a global 54% identity and 69% similarity in amino acid sequence. FupA with 557 amino acid residues is larger than FslE (509 residues). The Hidden Markov Model-based PRED-TMBB program (Bagos et al., 2004a,b) predicts that both FupA and FslE fold as βbarrels in the outer membrane with amino-terminal periplasmic domains. FupA is predicted to form a 16-stranded barrel with a periplasmic domain of 201 residues while FslE could form a 14-stranded barrel with a 152 residue periplasmic domain. This structure is reminiscent of typical TonB-dependent transporters (Noinaj et al., 2010). The greatest similarity between FslE and FupA is in the predicted β-barrel domains.

The fupA gene is located adjacent to a paralog fupB on the chromosome. Infrequent recombinational deletion events have been observed leading to formation of fupA/B hybrid genes (Twine et al, 2005; Rohmer et al., 2006); such recombination accounts for a significant reduction in virulence as seen in LVS, which can be reversed by restoration of the full length fupA gene (Salomonsson et al., 2009). The FupA/B hybrid protein encoded by LVS is less efficient at high-affinity ferrous iron uptake than FupA, but gains siderophore-iron uptake capability (Sen et al., 2010; Ramakrishnan and Sen, 2014). The structural and functional overlap in protein function raises the intriguing possibility that a common mechanism may underlie transport by FslE and FupA.

fupA expression is independent of iron and fur regulation, suggesting that FupA may have functions in addition to iron transport. fupA mutants have increased resistance to copper, and transport assays indicated that copper competes with ferrous iron for transport (Pérez et al., 2016). A role for FupA in maintenance of outer membrane integrity has also been proposed (Nallaparaju et al., 2011). Alhough, the fsl operon is regulated in response to iron levels, it was reported recently that calcium and magnesium limitation also result in increased fslE transcription (Wu et al., 2016). These observations are consistent with the idea that the high affinity iron transport proteins in the outer membrane of F. tularensis may assume roles in transport of additional substrates under stress.

Interestingly, FslE appears structurally different from rhizoferrin receptors of other bacteria. The LbtU siderophore receptor of L. pneumophila is not predicted to have a distinct periplasmic domain (Chatfield et al., 2011). RumA, the rhizoferrin receptor in Morganella morganii is a TonBdependent transporter (Kühn et al., 1996). Mechanisms for rhizoferrin transport thus appear to have evolved independently in different bacteria.

#### Iron Uptake and Pathogenesis

Analysis of the transcriptome of Schu S4-infected macrophages demonstrated that the fsl genes are among the most highly induced genes in the intracellular niche (Wehrly et al., 2009). Similar results were obtained using LVS infected hepatocytes (Thomas-Charles et al., 2013), suggesting that the siderophore uptake system contributes to survival within different tissue types. fsl mutants were identified in a negative selection screen of a U112 transposon mutant library in mice (Weiss et al., 2007) and a signature-tagged mutant screen identified fslA and feoB to be important for pulmonary infection by LVS (Su et al., 2007). Nevertheless, individual mutants in the fsl, fupA, and feoB genes

#### REFERENCES


are capable of intracellular growth although the reduced growth of the feoB mutants in hepatocytes suggests that ferrous iron is likely the major iron source within these cells (Lindgren et al., 2009; Ramakrishnan et al., 2012; Thomas-Charles et al., 2013; Pérez and Ramakrishnan, 2014; Pérez et al., 2016). Both iron uptake pathways appear to contribute to utilization of iron from heme for growth (Lindgren et al., 2015), suggesting that iron needs to be released from the heme for use by the bacteria. Screens with U112 mutants have implicated additional genes in iron acquisition, but they have not been definitively characterized (Crosa et al., 2009).

A Schu 1fslE 1fupA mutant deficient for both siderophore and high affinity ferrous iron uptake grows slowly, is attenuated for growth in macrophages and completely avirulent in mice (Ramakrishnan et al., 2012). 1fslA 1feoB mutants of LVS and Schu S4 deficient for siderophore biosynthesis and for all ferrous iron uptake have an even more severe defect, with dependence on extraneous siderophore for growth, loss of all capacity for intracellular growth and complete loss of virulence (Pérez and Ramakrishnan, 2014; Pérez et al., 2016). These findings indicate that both subspecies of F. tularensis have a similar repertoire of iron uptake mechanisms, limited to just the fsl and feo-mediated mechanisms. Interestingly, although the 1fslE 1fupA and the 1fslA 1feoB mutants are avirulent in mice, exposure to these strains protects from subsequent challenge with the virulent strain (Ramakrishnan et al., 2012; Pérez et al., 2016), making them good candidates for further exploration as live vaccines.

#### CONCLUSIONS

With a reduced genome, F. tularensis has evolved to efficiently support its lifestyle as an intracellular pathogen with a minimal set of two iron acquisition pathways. Host iron sources utilized and mechanisms regulating the transport proteins FslE and FupA are interesting questions for future investigations.

#### AUTHOR CONTRIBUTIONS

The author confirms being the sole contributor of this work and approved it for publication.

#### FUNDING

Work in the author's lab has been supported by grants AI056227, AI067823, and AI119471 from the National Institute of Allergy and Infectious Diseases (NIH) and by intramural support from the School of Medicine, University of Virginia.

discriminating β barrel outer membrane proteins. BMC Bioinformatics 5:29. doi: 10.1186/1471-2105-5-29


of Francisella tularensis demonstrate hypersensitivity to oxidative stress and attenuated virulence. J. Bacteriol. 188, 6443–6448. doi: 10.1128/JB.00266-06


virulence factor of Francisella tularensis subsp. Novicida U112. J. Med. Microbiol. 56(Pt 10), 1268–1276. doi: 10.1099/jmm.0.47190-0


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

## Gallium-Protoporphyrin IX Inhibits *Pseudomonas aeruginosa* Growth by Targeting Cytochromes

Sarah Hijazi, Paolo Visca and Emanuela Frangipani\*

*Department of Science, Roma Tre University, Rome, Italy*

*Pseudomonas aeruginosa* is a challenging pathogen due to both innate and acquired resistance to antibiotics. It is capable of causing a variety of infections, including chronic lung infection in cystic fibrosis (CF) patients. Given the importance of iron in bacterial physiology and pathogenicity, iron-uptake and metabolism have become attractive targets for the development of new antibacterial compounds. *P. aeruginosa* can acquire iron from a variety of sources to fulfill its nutritional requirements both in the environment and in the infected host. The adaptation of *P. aeruginosa* to heme iron acquisition in the CF lung makes heme utilization pathways a promising target for the development of new anti-*Pseudomonas* drugs. Gallium [Ga(III)] is an iron mimetic metal which inhibits *P. aeruginosa* growth by interfering with iron-dependent metabolism. The Ga(III) complex of the heme precursor protoporphyrin IX (GaPPIX) showed enhanced antibacterial activity against several bacterial species, although no inhibitory effect has been reported on *P. aeruginosa.* Here, we demonstrate that GaPPIX is indeed capable of inhibiting the growth of clinical *P. aeruginosa* strains under iron-deplete conditions, as those encountered by bacteria during infection, and that GaPPIX inhibition is reversed by iron. Using *P. aeruginosa* PAO1 as model organism, we show that GaPPIX enters cells through both the heme-uptake systems *has* and *phu*, primarily *via* the PhuR receptor which plays a crucial role in *P. aeruginosa* adaptation to the CF lung. We also demonstrate that intracellular GaPPIX inhibits the aerobic growth of *P. aeruginosa* by targeting cytochromes, thus interfering with cellular respiration.

#### *Edited by:*

*Pierre Cornelis, Vrije Universiteit Brussel, Belgium*

#### *Reviewed by:*

*Isabelle Schalk, Ecole Superieure de Biotechnologie de Starsbourg, France Christine Baysse, University of Rennes 1, France*

#### *\*Correspondence:*

*Emanuela Frangipani emanuela.frangipani@uniroma3.it*

*Received: 15 November 2016 Accepted: 10 January 2017 Published: 26 January 2017*

#### *Citation:*

*Hijazi S, Visca P and Frangipani E (2017) Gallium-Protoporphyrin IX Inhibits Pseudomonas aeruginosa Growth by Targeting Cytochromes. Front. Cell. Infect. Microbiol. 7:12. doi: 10.3389/fcimb.2017.00012* Keywords: aerobic respiration, antibacterial, cystic fibrosis, gallium, heme, infection, iron-uptake, terminal oxidases

### INTRODUCTION

Pseudomonas aeruginosa is a challenging bacterial pathogen due to both innate and acquired resistance to several antibiotics (Moore and Flaws, 2011). This bacterium is capable of causing a variety of infections, including chronic lung infection, which represents the main cause of morbidity and mortality in patients suffering from cystic fibrosis (CF) (Murphy, 2006; Davies et al., 2007). The success of P. aeruginosa as an opportunistic pathogen relies, at least in part, on its metabolic versatility, including the ability to obtain energy from different sources under a variety of environmental conditions (Williams et al., 2007; Arai, 2011). P. aeruginosa possesses a branched respiratory chain terminated by oxygen or nitrogen oxides, to allow growth by aerobic respiration or by denitrification under anaerobic conditions, respectively (reviewed in Arai, 2011). Moreover, P. aeruginosa is able to ferment arginine and pyruvate anaerobically (Vander et al., 1984; Eschbach et al., 2004). Aerobic respiration in P. aeruginosa relies on five terminal oxidases (Matsushita et al., 1982, 1983; Fujiwara et al., 1992; Cunningham and Williams, 1995; Cunningham et al., 1997; Stover et al., 2000; Comolli and Donohue, 2002, 2004). Three of these enzymes, the aa<sup>3</sup> terminal oxidase (Cox), the cbb3-1 (Cco-1), and the cbb3-2 (Cco-2) are cytochrome c-type oxidases, while the other two, i.e., the cyanide-insensitive oxidase (Cio) and the bo<sup>3</sup> oxidase (Cyo), are quinol oxidases (**Figure 1**). All these terminal oxidases contain heme, and are differentially expressed depending on the growth conditions, likely as a consequence to their different affinity for oxygen (Alvarez-Ortega and Harwood, 2007; Kawakami et al., 2010). Denitrification is ensured by a set of enzymes which sequentially convert nitrate (NO<sup>−</sup> 3 ) to molecular nitrogen (N2). Among the denitrification enzymes, only nitrite reductase (Nir) and nitric oxide reductase (Nor) contain heme as a cofactor (**Figure 1**).

Like almost all pathogenic bacteria, P. aeruginosa has an absolute need for iron to cause infections and to persist within the host (Ratledge and Dover, 2000). Iron is required as a cofactor of many key enzymes involved in respiration, DNA synthesis and defense against reactive oxygen species (Andrews et al., 2003). However, in the human host, iron is poorly available to bacteria due to its incorporation into hemecontaining molecules (e.g., hemoglobin and myoglobin) and iron carrier proteins (e.g., transferrin and lactoferrin) (Weinberg, 2009). This iron-withholding capacity represents the first line of the host defense against invading pathogens, a phenomenon known as "nutritional immunity" (Skaar, 2010). To circumvent iron-limitation, P. aeruginosa possesses several systems that actively acquire this essential metal, such as (i) the production of the siderophores pyoverdine (Pvd, Meyer and Abdallah, 1978; Cox and Adams, 1985) and pyochelin (Pch, Cox et al., 1981; Heinrichs et al., 1991); (ii) the ability to utilize a wide range of siderophores synthesized by other organisms (Cornelis and Matthijs, 2002; Cornelis et al., 2009); (iii) the ability to acquire Fe(II) through the Feo system (Cartron et al., 2006). In addition, P. aeruginosa can utilize heme-iron, by expressing two distinct heme-uptake systems, namely phu and has (Ochsner et al., 2000). The phu system allows the direct acquisition of heme from hemoproteins, which bind to the outer membrane receptor PhuR (Ochsner et al., 2000). In the has system a secreted hemophore HasA withdraws heme from hemoproteins and delivers it to the outer membrane receptor HasR (Létoffé et al., 1998). Given the similarity with the well-known has system of Serratia marcescens (Rossi et al., 2003; Létoffé et al., 2004), it is likely that the has system of P. aeruginosa positively regulates its own expression, via the sigma factor HasI and anti-sigma HasS, upon interaction of heme-loaded HasA with the HasR receptor (Llamas et al., 2014). The expression of both has and phu heme-uptake systems is shut down in the presence of sufficient intracellular iron, due to the negative regulation exerted by the ferric-uptake regulator (Fur) protein (Ochsner et al., 2000).

It has been shown that P. aeruginosa aerobic respiration and iron-uptake capabilities play pivotal roles during chronic lung

FIGURE 1 | Branched respiratory chain of *P. aeruginosa.* Cio, Cyo, Cox, Cco-1, and Cco-2 represent the five terminal oxidases that reduce oxygen to water under aerobic conditions. Cio and Cyo are quinol oxidases while Cox, Cco-1, and Cco-2 are cytochrome *c* oxidases. Nar, Nir, Nor, and Nos are nitrate reductase, nitrite reductase, nitric oxide reductase, and nitrous oxide reductase, respectively. These enzymes transfer electron to nitrogen oxides under anaerobic conditions. Nar receives electrons directly from the quinone pool while the other three receive electrons *via* the cytochrome *c* or from the small blue-copper protein azurin. *a*, *b*, *c,* and *d* represent different types of low-spin heme while *a*3, *b*3, *d*1, and *o*3 indicate the high-spin ones (modified from Arai, 2011).

infection in CF patients. In particular, three terminal oxidases (Cco-1, Cco-2, and Cio) sustain bacterial growth in the CF lung, a particular environment where P. aeruginosa iron-uptake abilities are sought to evolve toward heme utilization (Alvarez-Ortega and Harwood, 2007; Marvig et al., 2014; Nguyen et al., 2014).

The paucity of effective antibiotics to treat P. aeruginosa infections have made bacterial respiration and/or iron metabolism promising targets for the development of new anti-Pseudomonas drugs (Ballouche et al., 2009; Foley and Simeonov, 2012; Imperi et al., 2013). The possibility of using iron mimetics as novel therapeutics to interfere with iron metabolism has been exploited (Kaneko et al., 2007; Banin et al., 2008; Minandri et al., 2014). Ga(NO3)3, the active component of the FDA-approved formulation Ganite <sup>R</sup> , has successfully been repurposed as an antimicrobial drug (Bonchi et al., 2014; Rangel-Vega et al., 2015). Interestingly, Ga(NO3)<sup>3</sup> has been shown to be very active against P. aeruginosa, by interfering with iron-dependent metabolic pathways (Kaneko et al., 2007; Bonchi et al., 2015). The antibacterial proprieties of Ga(III) reside in the fact that, different from Fe(III), Ga(III) cannot be reduced under physiological conditions. However, redox cycling is critical for many of iron-dependent biological functions, including respiration (Breidenstein et al., 2011). Moreover, the heme-mimetic GaPPIX [i.e., Ga(III) coupled with the heme precursor protoporphyrin IX] has been shown to possess a good antibacterial activity against several bacterial species, including Staphylococcus aureus and Acinetobacter baumannii (Stojiljkovic et al., 1999; Arivett et al., 2015; Chang et al., 2016). GaPPIX is likely to exploit heme-uptake routes to enter bacterial cells, where it could substitute for heme in heme-containing enzymes, including cytochromes, catalases, and peroxidases, resulting in the perturbation of vital cellular functions (Stojiljkovic et al., 1999). Due to the similarity between GaPPIX and heme, GaPPIX is predicted to interfere with heme-dependent b-type cytochromes, thus impairing their function and ultimately inhibiting bacterial respiration.

In this work, the in vitro effect of GaPPIX on P. aeruginosa was tested under iron-depleted conditions, as those encountered during infection. The entrance routes of GaPPIX into P. aeruginosa cells and possible targets of GaPPIX were investigated. We demonstrate that the sensitivity of P. aeruginosa to GaPPIX depends on both intracellular iron levels and the expression of heme-uptake systems. Furthermore, we show that GaPPIX enters P. aeruginosa cells mainly through the heme-uptake receptor PhuR. Evidence is also provided that intracellular GaPPIX inhibits the aerobic growth of P. aeruginosa by targeting heme-dependent b-type cytochromes.

#### MATERIALS AND METHODS

#### Bacterial Strains and Growth Conditions

Strains and plasmids used in this work are listed in **Table 1**. P. aeruginosa clinical isolates are listed in Table S1. P. aeruginosa strains from frozen cultures were maintained on Luria Bertani (LB) agar before being transferred to liquid culture media. Bacteria were cultured in iron-free Casamino Acids medium (DCAA, Visca et al., 1993) supplemented or not with 100 µM of FeCl<sup>3</sup> at 37◦C, with vigorous shaking. When required, antibiotics were added to the media at the following concentrations for Escherichia coli, with the concentrations used for P. aeruginosa shown in parentheses: Ampicillin 100 µg/ml; carbenicillin (300 µg/ml in LB and 200 µg/ml in DCAA); and tetracycline 12.5 µg/ml (100 µg/ml). DCAA agar plates were prepared by the addition of 15 g/l bacteriological agar (Acumedia, Neogen corporation). When GaPPIX was required, a 50 mM of stock solution of GaPPIX (Frontier Scientific) was prepared in dimethyl sulfoxide (DMSO) and stored at 4◦C in the dark. When Ga(NO3)<sup>3</sup> was required, a 100 mM of stock solution of Ga(NO3)<sup>3</sup> (Sigma-Aldrich), was prepared in double-distilled water and stored at −20◦C.

#### Susceptibility Testing

The activity of GaPPIX, Ga(NO3)<sup>3</sup> and Hemin (Hm) (Sigma-Aldrich) on P. aeruginosa was tested in 96-well microtiter plates (Falcon). Briefly, bacterial cells were grown over-night in DCAA supplemented with 100 µM FeCl<sup>3</sup> in order to obtain high cell densities, then washed in saline and diluted to an OD<sup>600</sup> of 0.01 in 200 µl of DCAA containing increasing concentrations (0–100 µM) of GaPPIX, Ga(NO3)<sup>3</sup> or Hm. Microtiter plates were incubated for 24 h at 37◦C with gentle shaking (120 rpm). Growth (OD600) was measured in a Wallac 1420 Victor3 V multilabel plate reader (PerkinElmer). The minimum inhibitory concentration (MIC) of gallium compounds was visually determined as the lowest concentration that completely inhibited P. aeruginosa growth. As a control experiment the same procedure was performed, except that 100 µM FeCl<sup>3</sup> was added in the medium containing the highest concentration of gallium compounds tested (100 µM).

The antibacterial activity of gallium compounds was also assessed by disk diffusion assays. Briefly, cells from an over-night culture in DCAA supplemented with 100 µM FeCl<sup>3</sup> were washed and diluted in saline to OD<sup>600</sup> = 0.1, then seeded on the surface of DCAA agar plates supplemented or not with FeCl3. Sterile 6-mm blank disks (ThermoFisher-Oxoid) soaked with 10 µl of a 15 mM solution of either GaPPIX or Ga(NO3)<sup>3</sup> were deposited on the agar surface and the Zone Of growth Inhibition (ZOI) was measured (in mm) after 16 h of incubation at 37◦C.

To observe the rescue effect of Hm and Hemoglobin (Hb), disks were soaked with 10 µl of a 7.5 mg/ml solution of bovine hemin chloride (Sigma-Aldrich) in 10 mM NaOH or bovine hemoglobin (Sigma-Aldrich) in phosphate buffered saline (PBS) and deposited on the plate surface nearby the disk soaked with GaPPIX. The appearance of a half-moon-shaped growth area around the disk soaked with Hm or Hb was detected after 16 h of incubation at 37◦C.

#### Construction of Plasmids for the Expression of Heme Receptors

Plasmid preparations and DNA cloning were performed according to standards methods (Sambrook et al., 1989). Restriction and DNA modifying enzymes were used following the instructions of the manufacturers. Oligonucleotide primers are listed in **Table 1**. To express hasR in the 1hasR1phuR

#### TABLE 1 | Bacterial strains and plasmids used in this study.


*TcR, tetracycline resistant; ApR, ampicillin resistant; CmR, chloramphenicol resistant; CbR, carbenicillin resistant; restriction sites in the oligonucleotides are underlined.*

mutant, a 2932 bp fragment containing the hasR gene with its own promoter region was amplified by PCR from the PAO1 genome using primers hasR compl FW and hasR compl RV (**Table 1**). The product was then digested with KpnI and BglII and directionally cloned into the corresponding sites of the shuttle vector pUCP18, giving plasmid pUCPhasR. To express phuR in 1hasR1phuR mutant, a 2575 bp fragment containing the phuR gene with its own promoter region was amplified by PCR from the PAO1 genome using primers phuR compl FW and phuR compl RV (**Table 1**). The product was then digested with EcoRI and KpnI and directionally cloned into the corresponding sites of the shuttle vector pUCP18, giving plasmid pUCPphuR. To express hasR and phuR in the 1hasR1phuR mutant strain, the pUCPhasRphuR plasmid previously described (Minandri et al., 2016) was used.

#### Generation of *P. aeruginosa* Mutants

For mutant construction, E. coli and P. aeruginosa strains were grown in LB, with or without antibiotics, at 37 and 42◦C, respectively, with vigorous aeration. Previously described suicide plasmids (**Table 1**) were used according to procedures detailed elsewhere (Milton et al., 1996; Frangipani et al., 2008).

#### Measurement of Cytochrome *c* Oxidase Activity in *P. aeruginosa* Intact Cells

Cytochrome c oxidase activity was assayed by using the artificial electron donor N,N,N',N'tetramethyl-p-phenylene diamine (TMPD) (Fluka). Briefly, bacteria were grown over-night in DCAA supplemented with 100 µM FeCl3, then washed in saline and inoculated in DCAA to a final OD<sup>600</sup> = 0.05. When the mid-exponential growth phase was reached (≈6 h post inoculum), cells were washed once in saline and adjusted to an OD<sup>600</sup> = 1 (corresponding to ≈10<sup>9</sup> CFU/ml).

Then, 10<sup>8</sup> bacterial cells (100 µl) were suspended in 1.4 ml of 33 mM potassium phosphate buffer (KPi, pH 7.0). The reaction was started by the addition of 5 µl of a 0.54 M TMPD solution to the sample cuvette. The rate of TMPD oxidation was recorded spectrophotometrically at 520 nm for 8 min at 25◦C. Results were expressed as µmol TMPD oxidized/min−<sup>1</sup> /10<sup>8</sup> cells using 6.1 as the millimolar extinction coefficient of TMPD (Matsushita et al., 1982).

#### Isolation of Outer Membrane Proteins (OMPs) and SDS-PAGE Analysis

OMPs were isolated following the sarcosyl solubilization method (Filip et al., 1973), with some modifications. Briefly, bacteria from over-night cultures in DCAA supplemented with 100 µM FeCl<sup>3</sup> and 200 µg/ml Cb were washed in saline, then diluted to OD<sup>600</sup> = 0.05 in 60 ml DCAA supplemented with 200 µg/ml Cb, and incubated over-night at 37◦C. Cells were collected by centrifugation (2500 × g, 20 min), washed with 5 ml of 30 mM Tris HCl (pH 8, Sigma-Aldrich) and suspended in 1 ml of the same buffer. Bacteria were lysed by sonication in an ice bath (8 × 20 s cycles in a Sonics Vibra-CellTM VCX 130 sonicator), punctuated by 20 s intervals (50% power). Phenyl methyl sulfonyl fluoride (PMSF, Sigma-Aldrich) was added to cell lysate at 1 mM final concentration. Unbroken cells were removed by centrifugation at 2400 × g for 20 min, and supernatants were transferred to fresh tubes. Sarcosyl (N-laurylsarcosinate sodium salt, Sigma) was added to the supernatant to a final concentration of 2%. After 1 h incubation at room temperature with gentle shaking, the mixture was centrifuged for 2 h at 55,000 × g at 4◦C. OMP pellets were suspended in 40 µl 2 x SDS-PAGE loading dye (Sambrook et al., 1989), boiled for 10 min, then separated by 8% SDS-PAGE and visualized by Coomassie brilliant blue staining.

#### Statistical Analysis

Statistical analysis was performed with the software GraphPad Instat (GraphPad Software, Inc., La Jolla, CA), using One-Way Analysis of Variance (ANOVA), followed by Tukey-Kramer Multiple Comparisons Test.

### RESULTS

#### *P. aeruginosa* is Inhibited by GaPPIX under Iron-Deplete Conditions

It has been previously reported that GaPPIX has no effect on P. aeruginosa (Stojiljkovic et al., 1999). This results is quite surprising given that P. aeruginosa is able to utilize heme as an iron source, by expressing two heme-uptake systems, i.e., has and phu (Ochsner et al., 2000). However, since the effect of GaPPIX has previously been investigated in iron-rich media (Stojiljkovic et al., 1999), we sought that under these conditions iron availability would have impaired Ga(III) activity. To verify this hypothesis, we preliminary tested the effect of GaPPIX on P. aeruginosa PAO1 growth using the iron-poor medium DCAA (Visca et al., 1993), supplemented with increasing concentrations of GaPPIX, the iron-binding porphyrin Hemin, or Ga(NO3)3, the latter resulting very active on P. aeruginosa in this medium (Bonchi et al., 2015). Ga(NO3)<sup>3</sup> completely inhibited P. aeruginosa growth at 12.5 µM, and its activity was abrogated by the addition of FeCl<sup>3</sup> (**Figure 2A**) consistent with previous findings (Kaneko et al., 2007; Frangipani et al., 2014). Although the minimal inhibitory concentration (MIC) could not be determined for up to 100 µM GaPPIX (**Figure 2A**), exposure of PAO1 to GaPPIX reduced bacterial growth by 50% (IC50) at 12.5 µM (**Figure 2A**). Also in the case of GaPPIX, growth inhibition was completely reversed by the addition of FeCl<sup>3</sup> (**Figure 2A**). As expected, exposure P. aeruginosa PAO1 to Hemin promoted bacterial growth at concentrations ranging between 1.55 and 25 µM, in line with the ability of P. aeruginosa to use Hemin as an iron source (Ochsner et al., 2000).

The GaPPIX susceptibility of P. aeruginosa PAO1 was also tested using the disk diffusion assays in DCAA agar plates supplemented or not with an excess of FeCl<sup>3</sup> (600 µM) (**Figure 2B**). In FeCl3-supplemented DCAA, both GaPPIX and Ga(NO3)<sup>3</sup> caused no inhibition of PAO1 growth. Conversely, in DCAA a clear ZOI was observed around the GaPPIX and Ga(NO3)<sup>3</sup> disks (**Figure 2B**). Different from the ZOI formed by Ga(NO3)3, the ZOI formed by GaPPIX was less transparent (**Figure 2B**), consistent with the evidence that no MIC (full inhibition) could be determined for GaPPIX in liquid DCAA (**Figure 2A**). Although more transparent, the ZOI caused by Ga(NO3)<sup>3</sup> was smaller than that of GaPPIX (**Figure 2B**). These preliminary data indicate that iron-deplete conditions render P. aeruginosa PAO1 susceptible to GaPPIX-mediated growth inhibition.

#### The Response of *P. aeruginosa* Cells to GaPPIX Depends on Intracellular Iron Carryover

The above results prompted us to investigate the effect of the intracellular iron content on GaPPIX-dependent growth inhibition. To this aim, the effect of GaPPIX was compared between P. aeruginosa PAO1 cells that had been pre-cultured in either DCAA containing 100 µM FeCl<sup>3</sup> (to increase the intracellular iron content) or DCAA without FeCl<sup>3</sup> (to lower the intracellular iron content). Iron-starved bacterial cells were significantly more susceptible to GaPPIX (P< 0.001) compared with those pre-cultured with FeCl<sup>3</sup> (**Figure 3A**). In particular, upon the addition of 0.38 µM GaPPIX, the growth of ironstarved PAO1 cells was reduced by 40% compared with cells pre-cultured in the presence of 100 µM FeCl<sup>3</sup> (**Figure 3A**).

To further investigate the correlation between the intracellular iron content and GaPPIX-dependent growth inhibition, GaPPIX

susceptibility was evaluated on P. aeruginosa mutants impaired in Fe(III)-siderophore uptake systems, i.e., mutants unable to synthesize pyoverdine (1pvdA), pyochelin (1pchD), or both siderophores (1pvdA1pchD) (**Figure 3B**). While GaPPIXdependent growth inhibition was similar in the wild type and the 1pchD mutant, both 1pvdA and 1pvdA1pchD mutants were extremely sensitive to GaPPIX (**Figure 3B**). In particular, 0.38 µM GaPPIX inhibited the growth of the 1pvdA and 1pvdA1pchD mutant strains by 75 and 78%, respectively, compared with the untreated cultures, while it reduced the growth of the wild-type strain and of the 1pchD mutant by only 40 and 30%, respectively (**Figure 3B**). Altogether, these data indicate that the response of P. aeruginosa PAO1 to GaPPIX also depends on the carryover of intracellular iron.

#### GaPPIX is Preferentially Uptaken *via* the *P. aeruginosa* PhuR Receptor

To investigate the hypothesis that GaPPIX may enter P. aeruginosa cells by exploiting the same routes as heme, P. aeruginosa mutants carrying a deletion of either of the known heme receptors (1hasR and 1phuR mutants; **Table 1**) were generated. The effect of GaPPIX on these mutants, as well as on a 1hasR1phuR double mutant lacking both heme receptors (Minandri et al., 2016), was investigated in DCAA in the presence of 12.5 µM GaPPIX (IC50; **Figure 4A**). While all strains showed the same growth profiles in the untreated medium, both 1phuR and 1hasR1phuR mutants grew better than the wild type or the 1hasR mutant in the presence of 12.5 µM GaPPIX, displaying ≈50% higher growth levels relative to the wild type or the 1hasR mutant (**Figure 4A**). These data suggest that, among the P. aeruginosa heme-uptake systems, phu has a more prominent role than has in the uptake of GaPPIX. Then, the effect of GaPPIX on heme-receptor mutants was evaluated in DCAA agar plates, by performing the disk diffusion assays (**Figure 4B**). Results showed a similar ZOI (27.6 ± 2.0 mm) for both the wild-type strain and the 1hasR mutant, while a smaller ZOI (24.5 ± 0.7 mm) was observed for the 1phuR mutant, indicating a less susceptible phenotype (**Figure 4B,** Table S2).

In addition, no ZOI was observed for the 1hasR1phuR double mutant, indicating a fully resistant phenotype (**Figure 4B**). These observations indicate that both has and phu systems are implicated in GaPPIX transport, although the phu system appears to be the preferential route for the entrance of GaPPIX in P. aeruginosa cells (**Figure 4B**).

#### The Sensitivity of *P. aeruginosa* to GaPPIX Depends on the Expression of the Heme-Uptake Receptors

To further investigate the contribution of the HasR and PhuR receptors to GaPPIX-uptake, we individually expressed multicopy hasR, phuR, or both hasR and phuR in the 1hasR1phuR mutant strain (using plasmids pUCPhasR, pUCPphuR, or pUCPhasRphuR, respectively) (**Figure 5A**). The effect of GaPPIX on these strains was initially tested by the disk diffusion assays (**Figure 5A**). While, the empty pUCP18 vector did not alter the susceptibility of 1hasR1phuR to GaPPIX **(cfr Figures 5A**, **4B**), the expression of hasR from the multicopy plasmid pUCPhasR made the 1hasR1phuR mutant more susceptible to GaPPIX (ZOI = 27.6 ± 2.0 mm) (**Figure 5A,** Table S2). The effect of GaPPIX was even more pronounced in the 1has1phuR mutant overexpressing either phuR (1has1phuR carrying the multicopy plasmid pUCPphuR; ZOI = 34.0 ± 1.0 mm) or both hasR and phuR (1has1phuR carrying the multicopy plasmid pUCPhasRphuR; ZOI = 33.3 ± 0.5 mm) (**Figure 5A,** Table S2). GaPPIX sensitivity of the 1has1phuR strain expressing hasR, phuR, or both genes, was also evaluated in DCAA liquid medium, in the presence of different concentrations of GaPPIX (**Figure 5B**). All strains grew equally in the untreated medium, and GaPPIX did not

affect the growth of 1hasR1phuR/pUCP18 up to 25 µM (**Figure 5B**). Conversely, strains 1hasR1phuR/pUCPhasR, 1hasR1phuR/pUCPphuR, and 1hasR1phuR/pUCPhasRphuR were very sensitive to GaPPIX. In particular, 0.38 µM GaPPIX reduced the growth of the 1hasR1phuR/pUCPhasR strain by 56%, and by >80% in both 1hasR1phuR/pUCPphuR and 1hasR1phuR/pUCPhasRphuR strains (**Figure 5B**). This effect was much more pronounced than that observed for the parental strain PAO1 (**Figure 2A)**. Of note, no further growth reduction was observed for both the 1hasR1phuR/pUCPphuR and 1hasR1phuR/pUCPhasRphuR mutant strains at > 0.38 µM GaPPIX. The increased sensitivity of the 1hasR1phuR strain expressing either hasR or phuR, relative to the wild type, can be explained by the overexpression of heme receptors from the multicopy plasmid pUCP18 (**Figure 5A**). To confirm this hypothesis, HasR and PhuR protein levels were visualized by SDS-PAGE analysis of OMPs purified from the different P. aeruginosa strains cultured in DCAA (**Figure 5C**). By comparing P. aeruginosa outer-membrane-proteins profiles of the wild type, the 1phuR or the 1hasR1phuR mutant strains, the lack of a ca. 75 kDa protein in the 1phuR or the 1hasR1phuR mutants, was observed. This was in good agreement with a predicted molecular mass of 82 kDa for the mature PhuR receptor. Moreover, a protein band at that position was evident in SDS-PAGE electropherograms of the 1hasR1phuR/pUCPphuR and the 1hasR1phuR/pUCPhasRphuR complemented mutants (**Figure 5C**). Similarly, a protein band corresponding to ca. 94 kDa, consistent with the HasR receptor mass, was absent in the 1hasR and 1hasR1phuR mutants, while it was clearly detectable in the 1hasR1phuR/pUCPhasR and 1hasR1phuR/pUCPhasRphuR complemented mutants (**Figure 5C**). In line with previous results (Ochsner et al., 2000), protein levels greatly differed between PhuR and HasR, the latter being poorly expressed in wild-type PAO1. These results confirm that both HasR and PhuR direct GaPPIX entrance in P. aeruginosa cells, and argue for a prominent role of PhuR as a consequence of its higher expression levels, compared with HasR.

To confirm the specificity of GaPPIX for both heme-uptake systems, we investigated whether the growth inhibitory effect of GaPPIX could be rescued by the presence of Hemin (Hm) or Hemoglobin (Hb), which are known to deliver iron via hemeuptake receptors (Ochsner et al., 2000). To this aim, the hemeuptake mutant 1hasR1phuR overexpressing either PhuR or HasR was tested in the GaPPIX disk diffusion assays in the presence of Hm and Hb (**Figure 5D**). Both Hm and Hb partly rescued the growth of the 1hasR1phuR mutant overexpressing either PhuR (from pUCPphuR) or HasR (from pUCPhasR) thus

FIGURE 5 | The sensitivity of *P. aeruginosa* to GaPPIX depends on the expression of heme-uptake systems. (A) Approximately 5 × 10<sup>6</sup> bacterial cells were seeded on DCAA agar plates containing 200 µg/ml Cb, then disks soaked with 10 µl of a 15 mM solution of GaPPIX were deposited on the agar surface. Plates were incubated for 16 h at 37◦C. (B) Growth of the *P. aeruginosa* 1*hasR*1*phuR* mutant strain carrying the empty vector pUCP18 (white bars), or overexpressing *hasR* from plasmid pUCP*hasR* (light gray bars) or *phuR* from plasmid pUCP*phuR* (dark gray bars), or both *hasR* and *phuR* from plasmid pUCP*hasRphuR* (black bars), in DCAA supplemented with increasing concentrations of GaPPIX for 24 h at 37◦C. (C) SDS-PAGE analysis of the outer membrane proteins of different *P. aeruginosa* PAO1 strains. Strains and plasmids are indicated on the top of each lane. M is the molecular mass marker (kDa), with band sizes on the left. The position of the putative 82 kDa PhuR and 94 kDa HasR outer membrane receptors is indicated by arrows on the right. (D) Rescue effect of hemin (Hm) or hemoglobin (Hb) from GaPPIX growth inhibition in the 1*hasR*1*phuR* mutant strain overexpressing either *phu* from plasmid pUCP*phuR* or *hasR* from plasmid pUCP*hasR*. Approximately 5×10<sup>6</sup> bacterial cells were seeded on DCAA agar plates containing 200 µM Cb. Blank discs were soaked with 10 µl of a 15 mM solution of GaPPIX or with 75 µg of either Hm or Hb. Plates were incubated at 37◦C for 16 h. (E) Growth of the *P. aeruginosa* 1*pvdA*1*pchD* mutant strain carrying the empty vector pUCP18 (black bars), or overexpressing both *hasR* and *phuR* from plasmid pUCP*hasRphuR* (white bars), in DCAA supplemented with increasing concentrations of GaPPIX for 24 h at 37◦C. Values are the mean of two independent experiments, each one performed in triplicate ± the standard deviation. Images in B and D are representative of two independent experiments yielding similar results.

confirming that (i) Hm, Hb and GaPPIX compete with heme receptors and (ii) GaPPIX enters P. aeruginosa cells through PhuR and HasR (**Figure 5D**).

It has been observed that P. aeruginosa isolates evolving during chronic lung infection in CF patients tend to accumulate mutations in siderophore loci, concomitant with preferential utilization of heme iron (Cornelis and Dingemans, 2013; Marvig et al., 2014; Andersen et al., 2015). To simulate this situation, we tested GaPPIX susceptibility of a siderphore-defective P. aeruginosa mutant overexpressing both PhuR and HasR receptors (1pvdA1pchD/pUCPhasRphuR). Whereas, exposure of the 1pvdA1pchD mutant to GaPPIX reduced bacterial growth by 82% at 0.38 µM, expression of both hasR and phuR from multicopy plasmid pUCPhasRphuR made the 1pvdA1pchD mutant extremely susceptible to GaPPIX, displaying 90% growth reduction (IC90) at 0.38 µM (**Figure 5E**). Notably, full inhibition of the 1pvdA1pchD/pUCPhasRphuR strain was observed upon challenge with 50 µM GaPPIX.

#### GaPPIX Targets the Aerobic Respiration of *P. aeruginosa*

GaPPIX has been proven effective against a wide range of pathogenic bacteria by targeting metabolic pathways that require heme as an enzymatic cofactor, such as cellular respiration (Stojiljkovic et al., 1999). Thus, we investigated whether GaPPIX could interfere with the activity of terminal oxidases implicated in P. aeruginosa aerobic respiration. In particular, we focused on Cco-1, Cco-2, and Cio, which have been shown to sustain P. aeruginosa growth under low oxygen conditions, as those encountered in the lung of CF patients (Alvarez-Ortega and Harwood, 2007; Kawakami et al., 2010). To this aim, we initially tested the sensitivity of cytochrome c oxidases (i.e., Cox, Cco-1, and Cco-2) to GaPPIX. Strains deleted of the whole operon encoding the terminal oxidase Cox (1cox) or both the Cco-1 and Cco-2 terminal oxidases (1cco) were generated in the same parental strain used to generate the heme-receptor mutants (**Table 1**). The effect of GaPPIX was then assayed on these cytochrome-defective mutants using the TMPD redox indicator, which is an artificial electron donor to the cytochrome c (Matsushita et al., 1982). Oxidation of TMPD to a blue indophenol compound indicates electron flow to the cytochrome c terminal oxidases. Thus, cytochrome c oxidase activity was measured on P. aeruginosa PAO1 and in the 1cox and 1cco mutants grown in DCAA supplemented or not with a subinhibitory concentration of GaPPIX (4 µM). In whole cells cultured in the untreated medium, no cytochrome c oxidase activity could be measured in the 1cco strain (**Figure 6A**), confirming that in our conditions the TMPD test mainly measures the activity of Cco. Indeed, the 1cox mutation does not affect the TMPD oxidase activity (**Figure 6A**), as previously reported (Frangipani and Haas, 2009). This is because Cox is known to be poorly expressed during P. aeruginosa exponential growth (Kawakami et al., 2010). Interestingly, 4 µM GaPPIX reduced the respiratory activity by more than 50% in the wildtype strain PAO1 and the 1cox mutant, compared with the untreated condition (**Figure 6A**). These observations suggest that Cco-1 and Cco-2 terminal oxidases are sensitive to GaPPIX. To confirm these preliminary results, the effect of GaPPIX was tested on a mutant expressing only Cco-1 and Cco-2. To this aim, a 1cyo1cio1cox triple mutant strain was generated. Disk diffusion assays showed that the 1cyo1cio1cox mutant was more sensitive to GaPPIX than the wild type (ZOI = 34.6 ± 1.24 vs 27.6 ± 2.0 mm, respectively) (**Figure 6B,** Table S2). Similar results were obtained in DCAA liquid cultures. PAO1 wild type and the 1cyo1cio1cox mutant showed a similar growth profile in the untreated medium (**Figure 6C**), whereas exposure to 0.38 µM GaPPIX reduced bacterial growth by 40% and 68%, respectively, relative to the untreated cultures (**Figure 6C)**. Interestingly, it was possible to determine an IC<sup>90</sup> at 82 µM for the 1cyo1cio1cox mutant strain (**Figure 6C**). These results confirm that Cco-1 and Cco-2 are targeted by GaPPIX.

Then, the effect of GaPPIX on the Cio terminal oxidase was assessed. To this purpose, sodium azide (NaN3) was used as a specific inhibitor of copper-dependent oxidases, i.e., all terminal oxidases except Cio (Cunningham and Williams, 1995). Preliminarily, we determined the minimal NaN<sup>3</sup> concentration inhibiting all terminal oxidases except Cio in DCAA, by comparing the growth of wild-type PAO1 and the 1cio mutant in the presence of increasing NaN<sup>3</sup> concentrations (250–1000 µM). We observed that 350 µM of NaN<sup>3</sup> completely inhibited the 1cio mutant without affecting PAO1 growth (data not shown). Then, the sensitivity of Cio to GaPPIX was tested by performing a GaPPIX disk diffusion assays with wild-type PAO1 in DCAA supplemented or not with 350 µM NaN3. It was observed that PAO1 remains sensitive to GaPPIX in the presence of 350 µM NaN3, displaying a ZOI even greater than that obtained for PAO1 without NaN<sup>3</sup> (36.6 ± 3.0 vs 27.6 ± 2.0 mm, respectively) (**Figure 7A,** Table S2). This result provides evidence that Cio is a target for GaPPIX. To strengthen this evidence, a P. aeruginosa 1cyo1cco1cox triple mutant, which expresses only Cio (**Table 1**) was constructed and assayed for GaPPIX susceptibility. Disk diffusion assay results showed that the 1cyo1cco1cox mutant was more sensitive to GaPPIX than the wild-type PAO1 (ZOI = 30.0 ± 0.7 vs 27.6 ± 2.0 mm, respectively) (**Figure 7A,** Table S2). Similar results were also obtained in DCAA liquid medium, showing that GaPPIX significantly reduced (P < 0.001) the growth of the 1cyo1cco1cox mutant relative to the wild type, at concentrations ranging between 0.38 and 6.25 µM (**Figure 7B**). Altogether, the above results indicate that P. aeruginosa Cco-1, Cco-2, and Cio terminal oxidases are targets for GaPPIX.

#### *P. aeruginosa* Clinical Isolates Are Sensitive to GaPPIX

The expression of P. aeruginosa genes encoding heme-uptake systems has recently been detected in sputum samples collected from CF patients (Konings et al., 2013), and an evolution toward preferential heme utilization has been documented in P. aeruginosa during the course of chronic lung infection in CF patients (Marvig et al., 2014; Nguyen et al., 2014). Given the importance of heme in sustaining P. aeruginosa growth during infection, we have comparatively assessed the response to Ga(NO3)<sup>3</sup> and GaPPIX in a collection of P. aeruginosa clinical

isolates from CF and non-CF patients (**Figure 8**, Table S1). Although GaPPIX (up to 100 µM) never abolished P. aeruginosa growth, the majority of clinical isolates (>70%) was sensitive to GaPPIX, displaying an IC<sup>50</sup> values in the range 0.1–15.2 µM (Table S1). Moreover, all but one P. aeruginosa clinical isolates were significantly more susceptible than the reference PAO1 strain (**Figure 8**). In line with previous reports (Bonchi et al., 2015), all clinical isolates except one (FM1, Table S1) were very sensitive to Ga(NO3)3, showing IC<sup>50</sup> values ranging from 0.2 to 9 µM (Table S1).

#### DISCUSSION

The ability of pathogenic bacteria to colonize the host and cause infections is dependent on their capability to acquire iron and generate energy to sustain in vivo growth (Ratledge and Dover, 2000; Alvarez-Ortega and Harwood, 2007; Hammer et al., 2013). The success of P. aeruginosa as a pathogen relies on the presence of several iron-uptake systems (reviewed in Llamas et al., 2014), as well as on a multiplicity of terminal oxidases which allow bacterial respiration in vivo. Both iron-uptake systems and respiratory cytochromes have been shown to contribute to P. aeruginosa fitness during chronic lung infection in CF patients (Alvarez-Ortega and Harwood, 2007; Konings et al., 2013). Recent observations have documented an adaptation of P. aeruginosa toward heme iron acquisition in the CF lung, where bacterial energy metabolism mainly relies on the three terminal oxidases Cco-1, Cco-2, and Cio, all of which have high affinity for oxygen (Alvarez-Ortega and Harwood, 2007). These data suggest that heme utilization pathways and respiratory cytochromes could represent candidate targets for the development of new anti-Pseudomonas drugs (Alvarez-Ortega and Harwood, 2007; Marvig et al., 2014; Nguyen et al., 2014). Indeed, targeting bacterial membrane functions such as cellular respiration, are considered promising therapeutic opportunities, especially in the case of persistent or chronic infections (Hurdle et al., 2011). Given that all terminal oxidases require heme as a cofactor, and that heme-uptake systems are expressed during chronic lung infection, in this work we have investigated the effect of the heme-mimetic GaPPIX against P. aeruginosa. We focused on

Cco-1, Cco-2, and Cio since P. aeruginosa uses any of these three terminal oxidases to support the microaerobic growth necessary to thrive in the lung of CF patients. Cox and Cyo are not expressed or strongly repressed under these conditions (Alvarez-Ortega and Harwood, 2007).

We have initially demonstrated that GaPPIX is able to reduce the growth of P. aeruginosa only under iron-limiting growth conditions. However, different from Ga(NO3)3, bacterial growth was never completely inhibited at GaPPIX concentrations up to 100 µM (**Figure 2A**), in line with the fact that the ZOI for GaPPIX was less transparent compared with that generated by Ga(NO3)<sup>3</sup> in the disk diffusion assays (**Figure 2B)**. This diverse response of P. aeruginosa upon exposure to GaPPIX or Ga(NO3)<sup>3</sup> (**Figures 2A,B)** could be explained by the fact that GaPPIX and Ga(NO3)<sup>3</sup> enter bacterial cells through different pathways. Ga(NO3)<sup>3</sup> may enter P. aeruginosa cells (i) by diffusion; (ii) through the HitAB iron transport proteins (García-Contreras et al., 2013); or (iii) via the siderophore Pch (Frangipani et al., 2014). On the other hand, we have demonstrated that GaPPIX can cross the P. aeruginosa outer membrane only through the heme-receptors HasR and PhuR, since a 1hasR1phuR mutant is fully resistant to GaPPIX (**Figure 4**). Indeed, overexpression of heme receptors in the 1hasR1phuR mutant makes this strain susceptible to GaPPIX, at even lower GaPPIX concentrations compared with wild-type PAO1 (**Figures 5A,B)**. However, it should also be taken into consideration that GaPPIX and Ga(NO3)<sup>3</sup> likely have different targets. In fact, while Ga(NO3)<sup>3</sup> is known to target a variety of essential iron-containing enzymes (Bernstein, 1998; Soo et al., 2016), less is known about GaPPIX targets. Several studies have demonstrated that the antibacterial activity of GaPPIX relies on the molecule as a whole, since GaPPIX cannot be cleaved by bacterial enzymes (Stojiljkovic et al., 1999; Hammer et al., 2013). In fact, we demonstrated that the homolog of GaPPIX (Hemin) did not affect P. aeruginosa PAO1 growth. Indeed, Hemin promoted bacterial growth at concentrations ranging between 1.55 and 25 µM (**Figure 2A**), likely as a consequence of iron delivery to the cell, combined with positive regulation of the has system (Llamas et al., 2014). Hence, GaPPIX might be erroneously incorporated in hemecontaining proteins such as cytochromes. However, due to the multiplicity of pathways involving cytochromes, exposure to GaPPIX never results in a complete growth inhibition. This

hypothesis is supported by the observation that GaPPIX is more active against P. aeruginosa mutants deleted in some of the cytochrome-dependent terminal oxidases (**Figures 6**, **7**). In fact, a P. aeruginosa mutant that only expresses the terminal oxidases Cco-1 and Cco-2 (1cyo1cio1cox) is much more sensitive to GaPPIX than the wild-type strain. In addition, the 1cyo1cio1cox mutant showed a 68% growth reduction in liquid DCAA at 0.38 µM GaPPIX, compared to the untreated cultures, and an IC<sup>90</sup> of 82 µM (**Figures 6B,C**). Along the same lines, a P. aeruginosa strain that only relies on the terminal oxidase Cio to respire oxygen, is more sensitive to GaPPIX than the wildtype strain (**Figure 7**). Taken together, our results demonstrate that GaPPIX targets P. aeruginosa respiratory cytochromes Cco-1, Cco-2, and Cio, which are exclusively found in bacteria (Cunningham and Williams, 1995; Pitcher and Watmough, 2004), although we cannot discriminate which of the Cco cytochromes is preferentially targeted by GaPPIX (the 1cco-1,2 strain is mutated in both). Moreover, it is tempting to speculate that GaPPIX may also inhibit the other terminal oxidases Cyo and Cox (**Figure 1**), as well as some of the enzymes involved in denitrification, such as the heme-containing protein complexes Nir and Nor (**Figure 1**). Moreover, GaPPIX could also be incorporated into heme-containing enzymes involved in the protection from oxidative stress, increasing the susceptibility of P. aeruginosa to reactive oxygen species.

Although it was not possible to determine the MIC of GaPPIX for wild-type PAO1, it is worth to point out that GaPPIX was extremely active against a P. aeruginosa mutant impaired in siderophore production (1pvdA1pchD) and overexpressing both HasR and PhuR heme receptors from plasmid pUCPhasRphuR (**Figure 5**). Ninety percent growth reduction and full inhibition were observed upon exposure of this mutant to 0.38 and 50.0 µM GaPPIX, respectively. It is tempting to speculate that such strong inhibition could also occur in the CF lung, where siderophore-defective P. aeruginosa variants emerge during chronic infection, and heme represents the principal iron source (Marvig et al., 2014; Nguyen et al., 2014). Inhibition could further be enhanced under the microaerobic conditions encountered by P. aeruginosa in the CF airways (Hogardt and Heesemann, 2010), where the three high affinity terminal oxidases targeted by GaPPIX (Cco-1, Cco-2, and Cio) are essential for bacterial growth (Alvarez-Ortega and Harwood, 2007). Irrespective of the Ga(III) delivery system and of the energy metabolism adopted by P. aeruginosa, the balance between Fe(III) and Ga(III) availability in vivo will be the main determinant of Ga(III) efficacy. The inhibitory activity of GaPPIX was not limited to the prototypic strain PAO1, as it was also exerted on a representative collection of P. aeruginosa clinical isolates (Table S1). The great majority of clinical isolates (>70%) was sensitive to GaPPIX, irrespective of their origin, and all but one were significantly more susceptible than PAO1 (IC<sup>50</sup> ≤ 3.2 µM, Table S1).

Interestingly, studies on several human cell lines report that GaPPIX does not show cytotoxicity at concentrations ≤ 128 µM (Stojiljkovic et al., 1999; Chang et al., 2016), far above the concentrations that we found active on P. aeruginosa clinical isolates. Moreover, GaPPIX did not show to affect the health and behavior of mice, when administered by intraperitoneal injections (25–30 mg/kg) followed by four daily doses of (10–12 mg/kg) (Stojiljkovic et al., 1999), though it reduced the survival of Galleria mellonella larvae by 50% (LC50) when injected at 25 mM (Arivett et al., 2015).

Although further studies are needed to assess the effect of GaPPIX against P. aeruginosa infection in vivo, our work should encourage future research directed to the development of hememimetic drugs targeting cellular respiration for the treatment of P. aeruginosa chronic lung infection.

#### AUTHOR CONTRIBUTIONS

PV and EF designed research; SH performed research; SH, EF, and PV analyzed data; SH, EF, and PV wrote the paper.

#### ACKNOWLEDGMENTS

This work was supported by grants from the Italian Ministry of University and Research PRIN-2012 (prot. 2012WJSX8K) and from the Italian Cystic Fibrosis Foundation (grant FFC#21/2015) to PV.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fcimb. 2017.00012/full#supplementary-material

#### REFERENCES


related to the cytochrome bd quinol oxidases. Mol. Microbiol. 24, 579–591. doi: 10.1046/j.1365-2958.1997.3561728.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 © 2017 Hijazi, Visca and Frangipani. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

## Iron Starvation Conditions Upregulate *Ehrlichia ruminantium* Type IV Secretion System, *tr1* Transcription Factor and *map1* Genes Family through the Master Regulatory Protein ErxR

Amal Moumène1,2,3, Silvina Gonzalez-Rizzo4,5, Thierry Lefrançois 1,2, Nathalie Vachiéry 1,2 and Damien F. Meyer 1,2 \*

<sup>1</sup> Centre de Coopération Internationale en Recherche Agronomique Pour le Développement, UMR ASTRE, Petit-Bourg, France, <sup>2</sup> ASTRE, Univ Montpellier, Centre de Coopération Internationale en Recherche Agronomique Pour le Développement, Institut National de la Recherche Agronomique, Montpellier, France, <sup>3</sup> UFR Sciences Exactes et Naturelles, Université des Antilles, Pointe-à-Pitre, France, <sup>4</sup> Institut de Biologie Paris Seine (EPS - IBPS), Sorbonne Universités, UPMC Univ Paris 06, Univ Antilles, Univ Nice Sophia Antipolis, Centre National de la Recherche Scientifique Evolution Paris Seine, Paris, France, <sup>5</sup> Equipe Biologie de la Mangrove, UFR Sciences Exactes et Naturelles, Université des Antilles, Pointe-à-Pitre, France

#### *Edited by:*

Susu M. Zughaier, Emory University, United States

#### *Reviewed by:*

Jere W. McBride, University of Texas Medical Branch, United States David O'Callaghan, Université de Montpellier, France

> *\*Correspondence:* Damien F. Meyer damien.meyer@cirad.fr

*Received:* 19 June 2017 *Accepted:* 22 December 2017 *Published:* 19 January 2018

#### *Citation:*

Moumène A, Gonzalez-Rizzo S, Lefrançois T, Vachiéry N and Meyer DF (2018) Iron Starvation Conditions Upregulate Ehrlichia ruminantium Type IV Secretion System, tr1 Transcription Factor and map1 Genes Family through the Master Regulatory Protein ErxR. Front. Cell. Infect. Microbiol. 7:535. doi: 10.3389/fcimb.2017.00535 Ehrlichia ruminantium is an obligatory intracellular bacterium that causes heartwater, a fatal disease in ruminants. Due to its intracellular nature, E. ruminantium requires a set of specific virulence factors, such as the type IV secretion system (T4SS), and outer membrane proteins (Map proteins) in order to avoid and subvert the host's immune response. Several studies have been conducted to understand the regulation of the T4SS or outer membrane proteins, in Ehrlichia, but no integrated approach has been used to understand the regulation of Ehrlichia pathogenicity determinants in response to environmental cues. Iron is known to be a key nutrient for bacterial growth both in the environment and within hosts. In this study, we experimentally demonstrated the regulation of virB, map1, and tr1 genes by the newly identified master regulator ErxR (for Ehrlichia ruminantium expression regulator). We also analyzed the effect of iron depletion on the expression of erxR gene, tr1 transcription factor, T4SS and map1 genes clusters in E. ruminantium. We show that exposure of E. ruminantium to iron starvation induces erxR and subsequently tr1, virB, and map1 genes. Our results reveal tight co-regulation of T4SS and map1 genes via the ErxR regulatory protein at the transcriptional level, and, for the first time link map genes to the virulence function sensu stricto, thereby advancing our understanding of Ehrlichia's infection process. These results suggest that Ehrlichia is able to sense changes in iron concentrations in the environment and to regulate the expression of virulence factors accordingly.

Keywords: *Ehrlichia ruminantium*, master regulator, iron regulation, T4SS, *map* genes, *tr1* transcription factor, environmental cues

### INTRODUCTION

Ehrlichia ruminantium, the causal agent of heartwater, a fatal disease of ruminants in sub-Saharan African and other tropical regions, belongs to the Anaplasmataceae family and is transmitted by ticks of the genus Amblyomma (Dumler et al., 2001; Allsopp, 2010). In the mammalian host, E. ruminantium mainly infects brain capillary endothelial cells and replicates inside membrane-bound vacuoles (Zweygarth and Josemans, 2001). E. ruminantium has a biphasic developmental cycle in which elementary bodies (EB), the infectious form of the bacterium, first adhere and enter host cells. After internalization, EB differentiate into reticulate bodies (RB), the vegetative and non-infectious form, which divide by binary fission. Within 4–5 days, RB reorganize into EB, which are released from the vacuole by the lysis of the host cell to initiate a new infectious cycle (Moumène and Meyer, 2015).

Intracellular pathogenic bacteria belonging to the Anaplasmataceae family such as Ehrlichia, use a dedicated system, the type IV secretion system (T4SS), to inject some bacterial proteins, named effectors, to evade the host's immune responses and to hijack host cell processes in order to survive and proliferate in a safe replicative niche (Moumène and Meyer, 2015). T4SS is well conserved in the Anaplasmataceae family and several T4SS effectors (T4Es) have been shown to be crucial for the pathogenicity of Anaplasma phagocytophilum, Anaplasma marginale, and Ehrlichia chaffeensis (as reviewed by Rikihisa, 2010). Little is known about T4SS in E. ruminantium and no T4Es have yet been characterized (Collins et al., 2005; Frutos et al., 2007).

In several bacteria, it has been shown that the expression of T4SS is tightly regulated by transcription factors (Li and Carlow, 2012; Martín-Martín et al., 2012). Cheng et al. (2008) showed that the five virB/D4 genetic loci of E. chaffeensis T4SS are co-regulated by the transcription factor EcxR to allow specific expression, depending on the developmental stage of the bacteria (Cheng et al., 2008). Previous studies have also demonstrated developmental regulation of the expression of T4SS components during the intracellular life cycle of A. phagocytophilum in human peripheral blood neutrophils (Niu et al., 2006). A ecxR ortholog, apxR, has been found in several A. phagocytophilum strains (Wang et al., 2007a,b). ApxR regulates the expression of the transcription factor tr1 (Wang et al., 2007b) and of the downstream p44E locus (Wang et al., 2007a). Interestingly, the virB/D4 loci of A. phagocytophilum, E. chaffensis, and E. canis are flanked by genes encoding outer membrane proteins (OMP) belonging to the p44/msp2 family, which are paralogs of Ehrlichia ruminantium map1 genes (Dunning Hotopp et al., 2006; Rikihisa, 2011). The exact environmental cues, which stimulate the expression of apxR and ecxR, are not known.

Considering all these observations, we analyzed the genome of E. ruminantium and found an ortholog of ecxR and apxR in E. ruminantium, hereafter termed erxR, for the Ehrlichia ruminantium expression regulator. Moreover, as we observed in A. phagocytophilum, tr1 is present in E. ruminantium upstream of the cluster of map1 OMPs (Wang et al., 2007b). In E.ruminantium, the 16 paralogs of the map1 multigene family are expressed in bovine endothelial cells and some are preferentially transcribed in the tick or in the mammalian host (van Heerden et al., 2004). However, whether or not ErxR is a regulatory protein, which drives the expression of tr1 gene, the map1 genes family, and the virB/D4 loci in E. ruminantium and the triggering stimuli are currently unknown.

Microorganisms have evolved sensory mechanisms to regulate their cellular activities in response to environmental changes. This is particularly true for bacterial pathogens whose expression of virulence factors is tightly regulated in response to host and non-host environments (Hyytiäinen et al., 2003). Thus, the regulation of T4SS in response to host cues enables the efficient use of bacterial resources and facilitates colonization, leading to full infection (Abromaitis et al., 2013). One such environmental signal is iron, which is an essential cofactor in various enzymatic reactions like respiration, DNA replication, oxygen transport, response to oxidative stress, but can be toxic at excessive intracellular concentrations (Andrews et al., 2003). Therefore, iron scavenging from the limited sources of free iron available in the host is a crucial determinant of bacterial pathogenicity (Ratledge and Dover, 2000). Pathogens have evolved several ways to scavenge iron from the host, including the expression of iron acquisition genes in response to low iron concentrations (Brickman et al., 2011; Portier et al., 2014). In Gram negative bacteria, TonB-dependent outer membrane receptors (TBDR) are required to transfer iron chelates and heme into the periplasm under poor iron conditions, with subsequent transport to the cytoplasm (Shultis et al., 2006). Whether Map1 proteins play a similar role in E. ruminantium's iron uptake or sensing is currently not known.

In this study, we consequently investigated the effect of the newly identified ErxR protein and iron conditions on the regulation of the virB/D4, tr1 and map genes expression in E. ruminantium. For the first time in Ehrlichia, we identified (i) a master regulatory protein responsible for the coordinated regulation of the expression of virB/D4, tr1, and map genes, and (ii) one triggering environmental signal i.e., iron depletion, which activates this regulation cascade. Our work also demonstrates that map genes share at least one regulatory pathway with genes encoding the T4SS and may therefore be important pathogenicity determinants for iron acquisition or host cell infection.

#### MATERIALS AND METHODS

#### Culture Conditions

E. ruminantium Gardel strain (passages 30 to 52) was routinely propagated in bovine aortic endothelial (BAE) cells as previously described (Marcelino et al., 2005). To evaluate the growth characteristics of E. ruminantium under iron depletion conditions, the strain was grown in TC25 cm<sup>2</sup> flasks in BHK-21 cell medium supplemented with 2 mM glutamine, 10% heat inactivated fotal bovine serum (FBS), penicillin (100 IU/ml), streptomycin (100 mg/ml). For iron response experiments, the medium was supplemented with iron (100µM FeSO4) or an iron chelator (100µM 2,2′ -bipyridyl; BPD) as described in Breuer et al. (1995) and Romeo et al. (2001). The cells were kept in a humidified atmosphere supplemented with 5% CO<sup>2</sup> at 37◦C. FeSO<sup>4</sup> or 2,2′ -bipyridyl was added when 80% cell lysis was observed, at 120 h post-inoculation (hpi). The E. ruminantium infected cell monolayer (1 TC25 cm<sup>2</sup> flask) was harvested by trypsinisation 24 h after chemical inoculation. From the 6 ml of infected cell supernatant, a 600 µl sample was collected by centrifugation at 20,000 × g for 10 min. The pellet was stored at −80◦C until DNA extraction. The remaining 5,400 µl of infected cells were centrifuged at 20,000 × g for 10 min. The pellet was resuspended in TRIzol reagent (Invitrogen) and stored at −80◦C until RNA extraction (Pruneau et al., 2012).

#### Quantitative Detection of *E. ruminantium*

Genomic DNA was extracted from the 600 µl samples described above using the QIAamp DNA Mini Kit (Qiagen, France). The number of bacteria per sample was quantified by q-PCR, targeting the single copy of map1 gene encoding a major antigenic protein. The primer sequences are shown in **Supplementary Table 1** (Pruneau et al., 2012). A standard curve was established using gDNA of Gardel serially diluted from 7 × 10<sup>6</sup> to 7 × 10<sup>1</sup> copies µL <sup>−</sup>1, to determine the number of bacteria per microliter (Pruneau et al., 2012). Four microliters were added to Taqman master mix (Applied Biosystems, France), following the manufacturer's instructions. PCR conditions were as follows: 2 min at 50◦C, 10 min at 95◦C, and 40 cycles with 15 s at 95◦C and 1 min at 60◦C.

#### Relative Gene Expression: RNA Preparation and qRT-PCR

Total RNA was extracted using TRIzol reagent. RNA pellets were dissolved in 100 µl of DEPC water and treated with turboDNAse (Ambion, France). The purity and concentration of the isolated RNA were assessed using a NanoDrop 2000c (Thermo Scientific, France). RNA samples were diluted in RNase-free water to obtain a final concentration of 0.5 µg/µL. RNA samples were reverse-transcribed with the SuperScript VILO cDNA Synthesis Kit (Invitrogen, France), according to the manufacturer's instructions. Quantitative PCR was performed in a 7500 Real-Time PCR System (Applied Biosystems, France) using a Power SYBR Green PCR Master Mix (Applied Biosystems, France) and the primers listed in **Supplementary Table 2**. Reactions were performed in 25 µl volume with 5 ng template cDNA and 5µM of each primer. The amplification conditions were as follows: 2 min at 50◦C, 10 min at 95◦C and 40 PCR cycles (30 s at 95◦C and 1 min at 60◦C). An additional dissociation step of 15 s at 95◦C, 20 s at 60◦C and 15 s at 95◦C was added to assess nonspecific amplification. A negative control without cDNA template was included for each primer combination. Amplifications were performed in technical replicates consisting of independent cDNA syntheses derived from the same RNA sample and in three independent biological replicates. The relative expression of erxR was calculated by dividing the number of transcripts by the total number of bacteria at each time point. Fold change was then calculated by comparing the relative expression at each time point and the relative expression at 96 hpi, the stationary phase. Ratios were calculated from the number of transcripts and normalized to recA as described in Gonzalez-Rizzo et al. (2006).

#### Identification of Orthologs of EcxR in the *E. ruminantium* Genome

To identify orthologs of E. chaffeensis ecxR in the genome of E. ruminantium, we used the same strategy as that described in Li and Carlow (2012). The protein sequence of ECH\_0795 (YP\_507593) was used as a query to search the genome of E. ruminantium. Multiple sequence alignment was performed using ClustalW (http://www.ebi.ac.uk/Tools/ msa/clustalw2/) (Larkin et al., 2007). Sequence identity values between the two sequences were generated using BlastP. A Helix-Turn-Helix motif was determined using Pfam (https://www. ebi.ac.uk/Tools/hmmer/search/hmmscan) to find domains and motifs present in the protein sequence.

#### Cloning and Expression of *erxR*

Full-length erxR was PCR amplified using the primers listed in **Supplementary Table 2**, and ligated into the NdeI and XhoI sites of the pET29a(+) vector (Novagen). The resulting plasmid, defined herein as pErxR, was cloned into E. coli DH5α (Invitrogen) for amplification, purified using the QIAGEN Plasmid Maxi Kit (Qiagen, France), and cloned into E. coli BL21 (DE3) (Invitrogen) for protein expression. Protein expression was induced with 4 mM isopro-pyl-β-D-thiogalactopyranoside (IPTG) in 250 ml terrific broth. The protein then was purified using the Ni-NTA Fast Start Kit (Qiagen, France). ErxR expression was determined by Western blot analysis using anti-His tag antibody (Qiagen, France).

#### Construction of pUA66-Derived Promoter Plasmids

The pUA66 plasmid was used to analyse promoter activity based on the expression of a green fluorescent protein (GFP). The promoter regions were PCR amplified from the genomic DNA of Ehrlichia ruminantium Gardel strain, using the primers listed in **Supplementary Table 2**. Forward and reverse primers contained HindIII and BamHI restriction sites for cloning into the pET29a(+) plasmid. After cloning in pET29a(+), the promoters were digested with XhoI and BamHI for directed cloning into the pUA66 plasmid (Castaño-Cerezo et al., 2011). BL21 (DE3) cells were co-transformed with pErxR and each of the GFP reporter constructs, individually. The pET29a(+) vector alone was used as a negative control. Cotransformants were grown in LB medium supplemented with 50µg/ml kanamycin at 37◦C for 2 h, followed by induction with 1 mM IPTG for 4 h. Induced bacteria were visualized as described below.

#### Microscopy

A 6-µl drop of E. coli BL21 (DE3) co-transformed with the pUA66 promoter containing one of the different promoters and the pET29a-erxR plasmid (**Supplementary Table 3**), resuspended in LB growth medium, was spotted onto a Superfrost Plus slide (Fisher Scientific Ltd, UK) and visualized using a Nikon Eclipse 80i epifluorescence microscope (Nikon, France). Fluorescent images were acquired with a Nikon DXM1200F digital camera (Nikon, France), using Nikon ACT-1 software (Nikon, France). Fluorescence intensity was calculated by measuring the area, integrated intensity and mean gray

value of the fluorescent bacteria and the background with ImageJ (National Institute of Health, USA). Corrected total cell fluorescence (CTCF) was calculated using the following formula: integrated density—(area of the cell × mean background readings). The average and statistical differences between the bacteria containing the plasmids with the different promoters and controls were calculated using the CTCF values of all the bacteria in four different fields of view. Images were processed to size, and brightness and contrast were adjusted after the measurements, using Adobe Photoshop cs5 (Adobe Systems Inc., California, USA).

#### Statistical Analyses

Statistical analyses used Student's t-test and a P < 0.05 was considered significant.

#### RESULTS

#### Identification of One *ecxR* Ortholog in the *E. ruminantium* Genome

We used the ecxR sequence (YP\_507593.1) from E. chaffeensis to search NCBI databases using the BLAST tool. With this approach, we identified ERGA\_CDS\_03000 (YP\_196226.1) in the genome of E. ruminantium as the closest ortholog to ecxR. The results of the BLAST search revealed a putative conserved domain belonging to the HXT\_XRE superfamily of DNA binding proteins (cl17200). We identified a helix-turn-helix structure by comparing it with the structure found in Wolbachia (Li and Carlow, 2012) and further confirmed by Pfam (https://www.ebi. ac.uk/Tools/hmmer/search/hmmscan). This protein comprising 124 amino acids has a predicted molecular mass of 14.25 kDa. Alignment of the deduced amino acid sequences of the various orthologs is shown in **Figure 1A**. By homology with ApxR and EcxR, we named this protein E. ruminantium expression regulator, ErxR. Comparison of the sequence identities of these proteins revealed a high degree of conservation (82% identity) between EcxR and ErxR. ApxR and ErxR showed 40% identity at the amino acid level. Structural analyses indicated that all orthologs shared a conserved helix-turn-helix domain that may function as a sequence specific DNA binding domain, such as in transcription regulators (Aravind et al., 2005).

#### Architecture of T4SS and *map1* Genes Clusters of *E. ruminantium*

We compared the genetic arrangement of E. ruminantium to that of E. chaffeensis. The five virBD loci are represented in **Figures 1B,C**. In E. ruminantium, the genome sequence revealed the presence of two major operons. virD4, virB11, virB10, virB9a, and virB8a were located in operon 1. Operon 2 was seen to be located in the negative strand and contained four copies of virB6, along with one copy of virB4a and virB3, all located downstream of sodB. Four duplicated versions of virB2 were also located upstream of virB4b, while copies of virB8 and virB9 (namely virBb8 and virB9b) were scattered along the genome. We also present the arrangement of the map1 gene family previously reported by Postigo et al. (2007) (**Figure 1C**).

#### Analysis of *erxR* Expression during the Life Cycle of *E. ruminantium*

To determine the relative expression of erxR throughout the developmental cycle of E. ruminantium, erxR mRNA expression was analyzed by qRT-PCR. Compared to 96 hpi, the expression of erxR decreased at 24, 48, and 72 hpi, and peaked at 120 hpi, having increased 4 fold (**Figure 2A**), which corresponds to the time of lysis (early time point in the following round of infection). Data from previous independent microarray experiments (Pruneau et al., 2012) also demonstrated that the maximal expression of erxR in the post-exponential growth phase is significant and reproducible (data not shown). These results suggest that the up-regulation expression of erxR correlates with early stages in the development cycle in vitro before the bacteria enter the host cell.

#### rErxR Activates *gfp* Reporter Fusions

The promoter regions (hatched boxes in **Figure 1**) for virBD genes, map1, tr1, and erxR were cloned into a gfp reporter plasmid and transformed into E. coli BL21 (DE3) carrying pErxR or an empty pET29a(+) vector to investigate if ErxR regulatory protein drove their expression. The virB3-gfp reporter constructs presented a significant increase in fluorescence intensity after IPTG induction (270,000 units of fluorescence intensity) compared to samples lacking IPTG (10,000 units of fluorescence) or compared to the control (50,000 units of fluorescence) (**Figure 2B**). Similarly, the reporter construct showed an induction of virB2a by the recombinant ErxR protein. Activation was also observed for tr1 (200,000 units), erxR promoter (250,000), and map1 promoter (∼45,000) (**Figure 2B**). Western blot experiments confirmed that the expression of ErxR (14 kDa band) was only detected following induction with IPTG (**Figure 3**).

#### The Expression of T4SS, *erxR*, *tr1,* and *map* Genes Is Induced by Iron Depletion in *E. ruminantium*

As iron uptake mechanisms are closely associated with bacterial pathogenesis and may be connected to the expression of certain virulence determinants in E. ruminantium, we investigated the expression of T4SS, tr1 and map genes clusters in response to iron starvation. We incubated the bacteria in media containing iron or an iron chelator (BPD) for 24 h after lysis. The five T4SS transcription units of E. ruminantium were significantly up-regulated under iron depletion (25 to 190 fold increase as shown for virB3, virB4a, virB4b, virB8a, virB8b, and virB9b) (**Figure 4A**). virB2a was the only gene that did not present a significant change in expression (**Figure 4A**). Moreover, the erxR gene was up-regulated 25 fold during iron starvation (**Figure 4B**). We tested the expression of the map1 gene cluster under iron starvation in the two genes at the border of the cluster (tr1 and map1+1) as well as in the archetypal map1 and map1-6, which are located in a central position in the cluster. Culture under iron limitation strongly increased the expression of all genes (**Figure 4C**), however map1+1 showed the highest change in expression with 400-fold up-regulation (**Figure 4C**),

starvation condition was tested and are represented by gray arrows. The names and length of the genes are indicated above and below the arrows. The upstream

suggesting that this gene may play a key role during iron starvation.

regions amplified by PCR for regulation assays (hatched boxes) are indicated.

#### DISCUSSION

The preferential expression of virulence factors with diverse functions in response to host and environmental cues has already been characterized in several bacteria (Oogai et al., 2011; Weber et al., 2014). Thus, several Ehrlichia and other members of the Anaplasmataceae family have been shown to differentially express certain genes in response to the host cells (Singu et al., 2006; Nelson et al., 2008).

These genes include genes encoding components of the T4SS, which have been shown to play an essential role in pathogenicity (Rikihisa, 2010; Moumène and Meyer, 2015). Many intracellular bacterial pathogens use T4SS to deliver effector molecules, which subvert the eukaryotic host cell defenses and other cellular processes to their own advantage (Trokter et al., 2014). The genetic arrangement of the 18 genes encoding the T4SS apparatus in E. ruminantium T4SS genes resembles that of E. chaffeensis (**Figure 1**).

Likewise, OMPs are often regulated by environmental signals and play an important role in bacterial pathogenesis by enhancing the ability of the bacteria to adapt to different environments (Blanvillain et al., 2007). Interestingly, T4SS affects outer membrane properties, which might be important for the adaptation of Brucella to growth both in vitro and in vivo (Wang et al., 2010). Herein, we have described the effects of one environmental cue, iron concentration, on the expression of the transcriptional regulator erxR, the components of the T4SS apparatus and members of the Map1 family during E. ruminantium infection cycle of BAE cells.

Regulatory proteins are known to play an important role in the survival and persistence of intracellular pathogens in their host (Cheng et al., 2008). EcxR (ECH\_0795) is the only transcriptional regulator to be associated with the expression

FIGURE 2 | time point to 96 hpi (the stationary phase). Data were obtained from triplicate samples and are expressed as means + standard deviation. (B) ErxR activates the transcription of virBD, tr1 and map genes. Fluorescence intensity was used to measure the transcriptional activities of gfp reporter constructs. The values are means + standard deviations for three specimens, and measurements were taken from four different view fields. An asterisk indicates that values differ significantly (P < 0.001) from the controls. Western blot analyses were performed of samples from the fluorescence assays using an anti-His antibody to verify the expression of rErxR. Arrowheads indicate the position of rErxR.

of T4SS components in E. chaffeensis (Cheng et al., 2008). In E. ruminantium, we identified erxR, an orthologous gene of ecxR (**Figure 1A**), by sequence homology. ErxR (Ehrlichia ruminantium expression regulator) binds and regulates its own promoter and the promoters of some virBD genes. These results show that ErxR is the regulatory protein of the T4SS of E. ruminantium. Using qRT-PCR at different time points, we also show that erxR is strongly expressed at an elementary body stage (120 hpi), like that demonstrated in E. chaffeensis (Cheng et al., 2008).

ErxR orthologs in other Anaplasmataceae also appear to be associated with the expression of important antigenic OMPs. Thus, ApxR regulates the transcription of p44 transcription by binding to the tr1 promoter during A. phagocytophilum infection of mammalian host cells (Wang et al., 2007a,b). p44E encodes the immunodominant pleomorphic 44-kDa major surface protein, which shows homology with the Map1 family in E. ruminantium as well as the P30 family in E. canis and the P28 family in E. chaffeensis (Dunning Hotopp et al., 2006). The major antigenic protein Map1 is part of a multigene family containing 16 paralogs tandemly organized in a head to tail arrangement that are located downstream of a hypothetical transcriptional regulator gene (tr1) (Postigo et al., 2007; **Figure 1C**), a similar arrangement to that reported for p44, p30, and p28 in the other Anaplasmataceae (Dunning Hotopp et al., 2006). tr1 is one of the three promoters identified in the p44 expression locus and shown to be the strongest promoter driving the expression of a polycistronic mRNA containing OMP1, p44ESup, and p44 (Barbet et al., 2005). tr1 harbors a winged helix-turn-helix and a DNA binding motif and its part of the xenobiotic response element family of transcriptional regulators. However, the function of tr1 remains unclear (Nelson et al., 2008), and whether or not tr1 drives the expression of a polycistronic tandem mRNA containing several map homologs is still not known. According to our results, ErxR binds to tr1 and to map1 promoters (**Figure 3**). It is thus possible that it regulates the expression of the map1 members in a similar way to tr1 in A. phagocytophilum for p44 expression.

Taken together, our results show that ErxR binds to tr1, map1, and virBD promoters, suggesting for the first time coordinated regulation of T4SS and OMP in Anaplasmataceae. Interestingly, in the endosymbiotic bacterium Wolbachia (wBm) wBmxR1 and wBmxR2, which are orthologs of ErxR, have been shown to coregulate genes of the T4SS and riboflavin biosynthesis pathway (Li and Carlow, 2012). Riboflavin is an important co-factor for the survival of the endosymbiont's host, the filarial parasite Brugia malayi (Li and Carlow, 2012).

Next, we investigated whether the regulation of genes encoding the T4SS and members of the Map1 family by ErxR could be triggered by environmental and nutritional cues in the host cell. Our transcriptional analysis showed that erxR, virBD, tr1, and map1 genes were upregulated in response to iron starvation (**Figure 4**). Interestingly, the fact that virB2a is not expressed under iron depletion could be due to a functional redundancy depending on the environment (e.g. mammalian host, vector cell, etc.) as previously shown for PopF1 and PopF2 proteins of the T3SS of Ralstonia solanacearum (Meyer et al., 2006). Similarly, A. phagocytophilum appears to express specific virB2 paralogs in a host cell dependent manner as well as differential expression of virB2 gene in tick cells or in human cells (Nelson et al., 2008). Many bacterial pathogens sense iron depletion as a signal indicating that they are within a vertebrate host (Skaar, 2010).

Although the sensing mechanisms of changes in iron concentration are not known in E. ruminantium, three pathways are possible. One possibility is that the regulator ErxR is activated by an unidentified sensor kinase which responds to this environmental signal and up-regulates expression of the T4SS genes and certain map1 genes in response to iron starvation (**Figure 5**). Sensor kinases are part of two component systems (TCS), which regulate the differential expression of genes in bacteria in response to environmental cues. Three TCS composed of three response regulators and three sensor kinases have been identified in A. phagocytophilum and E. chaffeensis (Cheng et al., 2006). All three sets of TCS have orthologs in E. ruminantium and one of these TCS could thus be involved in the regulation of pathogenic factors in response to changes in iron abundance.

Another possibility is the presence of an unknown Fur repressor in the Anaplasmataceae family. This sensing involves transcriptional control mediated by the transcriptional repressor,

Fur (Escolar et al., 1998). But many organisms, including E. coli, Campylobacter jejuni, and Vibrio cholera, have been shown to use Fur to negatively regulate gene expression with increasing concentrations of iron (O'Sullivan et al., 1994; Palyada et al., 2004; Mey et al., 2005). For example, Fur activates sodB, an iron superoxide dismutase, under iron repletion. SODs are metalloproteins, which play an important role in protection against oxidative stress by catalyzing dismutation of the superoxide radical (O<sup>−</sup> 2 ). Interestingly, in E. ruminantium, sodB is located upstream of operon 2 of the T4SS and is co-transcribed along these genes in Ehrlichia and Anaplasma, suggesting an effect of iron in the expression of virBD genes. Moreover, it has been demonstrated that the regulator ApxR binds to the promoter regions upstream of sodB (Wang et al., 2007b). We searched for putative Fur boxes in the genome of E. ruminantium and found one located upstream of erxR (**Supplementary Figure 1**). The 19 bp sequence consisted in two repeated hexamers (nATWAT) flanking a 7 nt sequence, which is commonly found upstream of iron regulated enzymes such as succinate dehydrogenase iron-sulfur subunits, major ferric iron binding protein precursors, and adenosine tRNA methylthiotransferase (Escolar et al., 1998; Grifantini et al., 2003). Thus, it is possible that E. ruminantium is capable of sensing low iron concentrations in the environment and of regulating the expression of ErxR through this putative Fur box.

Finally, it is possible that some Map1 proteins play a role like that of TBDRs in the perception of environmental cues and in iron uptake (Blanvillain et al., 2007). As mentioned above, Map1 proteins are orthologs of members of the p28 family in E. chaffeensis, which have been shown to function as porins and possibly act in nutrient uptake during intracellular infection (Kumagai et al., 2008). It has been shown that porins, such as OmpA and OmpC, bind to transferrin and act in iron uptake in

the enteropathogenic strains of E. coli, Salmonella typhimurium, and several Shigella species (Sandrini et al., 2013). Similarly, Mycobacterium smegmatis is able to acquire ferric ions through members of the Msp family of porins (Jones and Niederweis, 2010). The up-regulation of map1, map1+1, and map1-6 showed under iron starvation suggests that these three porins play a role in iron acquisition and raises the possibility that Map1 proteins may also act as sensors, although more evidence is needed before concluding. These results suggest that the Map1 proteins may fulfil several functions during E. ruminantium infection. Characterizing these functions could advance our understanding of the adaptation of E. ruminantium to its host, as done by Blanvillain et al. (2007).

In conclusion, we have demonstrated that exposure of E. ruminantium to iron limitation induces ErxR-dependent expression of the T4SS apparatus and map1 genes. These findings reveal an important degree of coordination between T4SS and map1 genes at the transcriptional level and raise the possibility of the involvement of Map proteins in environmental sensing and in the infection process. The data presented herein enables us to propose a model for the regulation of E. ruminantium T4SS and map1 genes in which ErxR acts as a global regulator integrating iron as an important triggering environmental signal (**Figure 5**). Understanding Ehrlichia gene regulation in response to environmental signals provides valuable cues for the development of alternative treatments.

#### AUTHOR CONTRIBUTIONS

DM, Conceived and designed the experiments. AM, SG-R, and DM, Performed the experiments. AM, SG-R, TL, NV, and DM, Analyzed the data. AM and DM, Wrote the paper.

#### ACKNOWLEDGMENTS

We are grateful to. S. Castaño-Cerezo at the Universidad de Murcia for providing pUA66 plasmid. We thank E. Albina for critical reading of the manuscript and A.O. Chavez for technical assistance in the regulation experiments. AM acknowledges financial support from FEDER grant FED 1/1.4-30305, 2007– 2013, "Risque en santé animale et végétale" (PhD grant to AM), and from EPIGENESIS RegPot European project (n◦ 315988).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fcimb. 2017.00535/full#supplementary-material

#### REFERENCES


Supplementary Figure 1 (A) | Nucleotide sequence of the predicted erxR promoter. The initiation codon of erxR is shown in capital letters at the end of the sequence. The sequence in red represents the nATWAT motif. The−35 and−10 boxes of the erxR promoter are capitalized. (B) The potential erxR Fur box is compared with the consensus E. coli Fur box. Conserved residues are shown in red.

Supplementary Table 1 | Oligonucleotide primers used for qPCR and qRT-PCR.

Supplementary Table 2 | Oligonucleotide primers used to construct plasmids.

Supplementary Table 3 | Strains and plasmids used in this work.


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

Copyright © 2018 Moumène, Gonzalez-Rizzo, Lefrançois, Vachiéry and Meyer. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

## The Role of the Regulator Fur in Gene Regulation and Virulence of *Riemerella anatipestifer* Assessed Using an Unmarked Gene Deletion System

Yunqing Guo1, 2, Di Hu1, 2, Jie Guo1, 2, Xiaowen Li 1, 2, Jinyue Guo1, 2, Xiliang Wang1, 2 , Yuncai Xiao1, 2, Hui Jin1, 3, Mei Liu1, 2, Zili Li 1, 2, Dingren Bi 1, 2, 3 \* and Zutao Zhou1, 2 \*

<sup>1</sup> College of Veterinary Medicine, Huazhong Agricultural University, Wuhan, China, <sup>2</sup> Key Lab of Preventive Veterinary Medicine of Hubei Province, Huazhong Agricultural University, Wuhan, China, <sup>3</sup> State Key Laboratory of Agricultural Microbiology, Huazhong Agricultural University, Wuhan, China

#### *Edited by:*

Susu M. Zughaier, Emory University, United States

#### *Reviewed by:*

Teresa Olczak, University of Wrocław, Poland Michael Marceau, Université Lille 2 Droit et Santé, France

#### *\*Correspondence:*

Dingren Bi bidingren@mail.hzau.edu.cn Zutao Zhou ztzhou@mail.hzau.edu.cn

*Received:* 15 May 2017 *Accepted:* 09 August 2017 *Published:* 25 August 2017

#### *Citation:*

Guo Y, Hu D, Guo J, Li X, Guo J, Wang X, Xiao Y, Jin H, Liu M, Li Z, Bi D and Zhou Z (2017) The Role of the Regulator Fur in Gene Regulation and Virulence of Riemerella anatipestifer Assessed Using an Unmarked Gene Deletion System. Front. Cell. Infect. Microbiol. 7:382. doi: 10.3389/fcimb.2017.00382 Riemerella anatipestifer, an avian pathogen, has resulted in enormous economic losses to the duck industry globally. Notwithstanding, little is known regarding the physiological, pathogenic and virulence mechanisms of Riemerella anatipestifer (RA) infection. However, the role of Ferric uptake regulator (Fur) in the virulence of R. anatipestifer has not, to date, been demonstrated. Using a genetic approach, unmarked gene deletion system, we evaluated the function of fur gene in the virulence of R. anatipestifer. For this purpose, we constructed a suicide vector containing pheS as a counter selectable marker for unmarked deletion of fur gene to investigate its role in the virulence. After successful transformation of the newly constructed vector, a mutant strain was characterized for genes regulated by iron and Fur using RNA-sequencing and a comparison was made between wild type and mutant strains in both iron restricted and enriched conditions. RNA-seq analysis of the mutant strain in a restricted iron environment showed the downregulation and upregulation of genes which were involved in either important metabolic pathways, transport processes, growth or cell membrane synthesis. Electrophoretic mobility shift assay was performed to identify the putative sequences recognized by Fur. The putative Fur-box sequence was 5 ′ -GATAATGATAATCATTATC-3′ . Lastly, the median lethal dose and histopathological investigations of animal tissues also illustrated mild pathological lesions produced by the mutant strain as compared to the wild type RA strain, hence showing declined virulence. Conclusively, an unmarked gene deletion system was successfully developed for RA and the role of the fur gene in virulence was explored comprehensively.

Keywords: *Riemerella anatipestifer, fur, pheS,* unmarked gene deletion system, virulence, Fur-box, RNA-seq

#### INTRODUCTION

Riemerella anatipestifer (R. anatipestifer, RA) is non-spore forming, non-motile, Gram-negative, rod-shaped bacterium belonging to the family Flavobacteriaceae. Other than ducks, RA can also affect the majority of poultry including turkeys, geese, which has resulted in significant economic losses to the poultry industry worldwide. Infection leads to polyserositis and septicaemia often with neurological symptoms. At least 21 serotypes have been identified in different countries, with serotypes 1, 2, and 10 most prevalent in China (Loh et al., 1992; Cheng et al., 2003). Due to extensive genomic divergences, even within a given serotype, there is often limited cross protection and variation of virulence (Higgins et al., 2000). Presently, little is known regarding RA pathogenesis, although a number of attempts have been made to explore the molecular mechanisms underlying virulence. In previous studies, the role of outer membrane protein A (OmpA), TonB dependent receptor 1 (TbdR1), TonB family protein (TbfA), siderophore interacting protein (Sip) and CAMP cohemolysin have all been proposed as virulence associated factors (Crasta et al., 2002; Hu et al., 2011; Lu et al., 2013; Tu et al., 2014; Liu et al., 2016). All of these studies were based on gene knockout, which result in modified expression of downstream genes, known as the polar effect. As a genetic analysis tool, unmarked gene deletion system is advantageous over gene knockout strategy, as well as able to provide a more accurate estimation of gene expression and has a limited polar effect. Indeed, in many species of bacteria, such as Enterococcus faecalis, Burkholderia family, Streptococcus mutans, Bacillus amyloliquefaciens, unmarked gene deletion system has been established to elucidate molecular mechanisms of pathogenesis and virulence (Kristich et al., 2007; Barrett et al., 2008; Xie et al., 2011; Zhou et al., 2016). To the best of our knowledge, no data is available on unmarked gene deletion system in RA. The development and application of such strategies will accelerate our understanding of the mechanism of pathogenesis, virulence and antibiotic resistance in RA.

Earlier studies have established the role of Ferric uptake regulator (Fur) proteins in virulence in a variety of bacterial species (Ernst et al., 2005a; Haraszthy et al., 2006; Yuhara et al., 2008; Porcheron and Dozois, 2015; Pi et al., 2016). Fur is a regulator of transcription in bacteria, involved in iron homeostasis, acid resistance, oxidative stress and virulence (Bijlsma et al., 2002; Ernst et al., 2005b; Mathieu et al., 2016). Iron is an essential element in various metabolic pathways of bacteria and eukaryotic host (Holmes et al., 2005). To date, the function of fur gene in virulence of RA has not been demonstrated in any previous study. In this novel study, the role of fur gene in virulence of RA has been examined by adopting unmarked gene deletion system. Having observed the limitations of other counter-selectable markers, pheS is an appropriate non-antibiotic resistance counter-selectable marker. Previously, the applications of the mutant E. coli pheS gene (A294G), the mutant E. faecalis pheS gene (A312G), the mutant Burkholderia pheS gene (A294G), and the mutant S. mutans pheS gene (A314G) were successful for allelic replacement in those organisms (Kast and Hennecke, 1991; Ibba et al., 1994; Kristich et al., 2007; Barrett et al., 2008; Xie et al., 2011). Therefore, we postulated the role of pheS gene for this purpose.

In summary, in this study, we engineered a suicide vector pRE-lacZ-mpheS-spc, using mutated pheS as a counter-selectable marker and lacZ to select a fur gene deletion mutant RA-YM 1fur. This is the first successful attempt to construct mutant RA using an unmarked gene deletion system. The RA-YM 1fur complemented strain was constructed to confirm virulence of the wild type strain, compared with the mutant. Lastly, using whole genome transcriptional sequencing, genes regulated by the fur gene were screened out in mutant and wild types. Moreover, the predictive sequence of Fur-box of RA was analyzed.

#### MATERIALS AND METHODS

#### Bacterial Strains, Plasmids, Media, and Growth Conditions

The bacterial strains and plasmids used in this study, and their relevant characteristics are described in **Table 1**. R. anatipestifer strains were grown at 37◦C in tryptic soy broth (TSB) (Difco, Detroit, USA) in an atmosphere of 5% CO2, E. coli strains were cultured at 37◦C in Luria Bertani broth (Sigma-Aldrich, St. Louis, USA). Both R. anatipestifer strains and E. coli strains included in this study were obtained from laboratory stocks of the Department of Veterinary Microbiology and Immunology of Huazhong Agricultural University, China. Where necessary, the following antibiotics were added in to the selection media: ampicillin (Amp), 100 mg/mL; spectinomycin (Spc), 100 mg/mL; kanamycin (Kan), 100 mg/mL; and medium was supplemented with 2, 6-diaminopimelic acid (DAP), 100 mg/mL; 5-bromo-4-chloro-3-indolyl β-D-galactopyranoside (X-gal), 20 mg/mL; Isopropyl β-D-1-thiogalactopyranoside (IPTG), 20 mg/mL; 4 chloro-DL-phenylalanine (cPhe), 0.2% (w/v).

#### Construction of Suicide Vector pRE-lacZ-mpheS-spc and Complemented Shuttle Vector pRES-JXrep-spc

For the construction of suicide vector pRE-lacZ-mpheS-spc, a 3.9 kb fragment of pRE was amplified from suicide vector pRE112 using primer S1L (**Table 1**) (introducing BamHI, SalI, and NdeI site) and S2R (**Table 1**) (introducing XhoI and BamHI site); then digested with BamHI enzyme and ligated to generate circular pRE which contained the essential components of the conjugational transfer. As RA cannot catabolize sucrose, the selected marker SacB was removed. The mutated pheS gene (mpheS) and the sequence of the multiple cloning sites were engineered into pUC57 to generate pUC57-mpheS and pUC57- MCS by the GenScript Corporation (Nanjing, China). This pUC57-mpheS vector contained the mutated R. anatipestifer pheS gene with altered DNA sequences (Supplemental Figure 1), which was driven by an upstream PS12 promoter of the R. anatipestifer rpsL gene. The 1.1 kb spc cassette was amplified from plasmid pIC333 using primers spcL and spcR (**Table 1**), mpheS was amplified from plasmid pUC57-mpheS using primers pheS1and pheS2 (**Table 1**). The spc cassette was then fused with the mpheS fragment using overlap PCR (introducing

#### TABLE 1 | Strains, plasmids and primers.


(Continued)

TABLE 1 | Continued


KpnI and XbaI site). The fragment of mpheS-spc was inserted into pMD18T to generate pMD18T-mpheS-spc. The pMD18TmpheS-spc was digested with KpnI and XbaI enzymes, and 2.2 kb PS12-mpheS-spc fragment was ligated into pUC57-MCS and digested with KpnI and XbaI to obtain pUC57-MCS-mpheS-spc. The pRE and pUC57-MCS-mpheS-spc were digested with XhoI and XbaI and ligated to generate pRE-mpheS-spc. The next step was to amplify a 3.3 kb fragment of lacZ from E. coli BL21 genome using primer lacZR and lacZL (**Table 1**) (introducing a BamHI site). A 135 bp fragment of PS12 promoter was amplified from RA-YM using primer rpsL-LacZ (introducing a XhoI site) and rpsR-LacZ (**Table 1**) and fused to lacZ fragment by overlap PCR using rpsL-LacZ and lacZR primers. pRE-mpheS-spc and lacZ fragments were digested with BamHI and XhoI and ligated to obtain 9 kb pRE-lacZ-mpheS-spc.

The shuttle plasmid pRES-JXrep-spc was constructed in several steps including the amplification of 2.5 kb fragment of replicon region and replicase gene (Genebank: KY806579) with primers rep1 and rep2 (**Table 1**) (introducing a XhoI site). The wild type plasmid of R. anatipestifer strain RA-JX was used as a template. A 1.1 kb spc cassette was amplified from plasmid pIC333 using primers spcL1 and spcR1 (**Table 1**), thereby introducing a BamHI site. The replicon region of RA-JX was joined with the spc cassette using overlap PCR. The JXrep-spc fragment was inserted into pMD18T to obtain pMD18T-JXrepspc. In the next step, plasmid pRE and pMD18T-JXrep-spc were digested with XhoI and BamHI and ligated to generate the shuttle plasmid pRES-JXrep-spc.

#### Construction of Unmarked Deletion *R. anatipestifer* 1*fur* and Complemented Mutant Strains

To obtain the suicide vector pRE-lacZ-mpheS-spc-fur for the deletion of whole fur gene from R. anatipestifer RA-YM strain, upstream (738 bp) and downstream (802 bp) DNA fragments were amplified using primers Fur-L1 and Fur-L2 (**Table 1**) (introducing SphI site), Fur-R1 and Fur-R2 (**Table 1**) (introducing KpnI site), respectively. The two fragments were joined together by overlap PCR. The LR fragment and pRElacZ-mpheS-spc were digested with SphI and KpnI; 1.5 kb LR fragment was inserted into pRE-lacZ-mpheS-spc to generate the suicide vector pRE-lacZ-mpheS-spc-fur. E. coli strain x7213 was used as a donor in conjugation step to introduce the suicide vector pRE-lacZ-mpheS-spc-fur into RA-YM strain as described previously (Hu et al., 2011). For phenotypic detection of mutant strains, conjugation filters were plated on tryptic soya agar (TSA) containing 100 µg/mL Spc. Colonies were then grown on TSA containing cPhe (0.2%) and X-gal (40 µg/mL). Appearance of white colonies confirmed successful construction of deletion mutant strains. For identification of recombinants carrying the chromosomal fur gene deletion, colonies were analyzed using PCR primers FurL1 and FurR2 to determine presence of wildtype or mutant allele at the target locus. The wild-type and deleted alleles could be differentiated on the basis of size of amplicon by agarose gel electrophoresis.

Similarly, for generation of complemented mutant strain, shuttle vector pRES-JXrep-spc-fur was constructed by amplification of the promoter sequence (171 bp) and the coding sequence (486 bp) of fur gene. The promoter sequence and the coding sequence were amplified using primers Promoterfur1 and Promoter-fur2 (introducing KpnI site, **Table 1**), primers Fur-inL and Fur-inR (introducing SphI site, **Table 1**). The two fragments were joined together by overlap PCR. The plasmid pRES-JXrep-spc and fur gene fragment were digested with SphI and KpnI, then the fragment of 657 bp was inserted into plasmid pRES-JXrep-spc to obtain shuttle vector pRES-JXrep-spc-fur. The E. coli strain x7213 was used as donor in conjugation transfer of shuttle vector into the RA-YM 1fur strain (Hu et al., 2011). The phenotypic identification of complemented mutant strain was conducted on TSA plates containing Spc 100 µg/mL. Furthermore, PCR reaction was performed using primers Fur inL and Fur inR to ensure that recombinant strains were harboring shuttle vectors.

#### RNA-Sequencing of Wild-Type and RA-YM 1*fur* Deletion Mutant in Iron-Restricted and Enriched Conditions

The colonies of the wild-type and the RA-YM 1fur mutant were suspended into tryptic soya broth (TSB) and incubated overnight with shaking at 37◦C to an OD<sup>600</sup> of 0.2. FeCl<sup>3</sup> (Sigma-Aldrich) or 2, 2-Dipyridyl (2, 2-DP, Sigma-Aldrich) was added to the bacterial suspension to produce a final concentration of 200 and 30 µM, as iron restricted and iron rich conditions, respectively and incubated at 37◦C until the OD<sup>600</sup> reached 0.8. Total RNA was extracted from bacteria solution using Bacterial RNA Kit (OMEGA, Norcross, USA) following the guidelines. Extracted RNA was purified with RNase-free DNase (Promega, Wisconsin, USA) at 37◦C for 30 min to remove impurities of DNA, the DNA-free purified RNA was examined by 1% agarose gel electrophoresis. Purified RNA (23S rRNA and 16S rRNA) was sent to Huada Gene Center (Shenzhen, China) for RNA sequencing. All RNA samples were performed in two independent biological replicates (BioProject: SRP106941).

#### Quantitative Reverse Transcription PCR (RT-qPCR)

RT-qPCR was performed to quantify the expression of genes regulated by Fur. Primers were designed with Primer 5.0 software. RNA was extracted from wild-type and RA-YM 1fur strains grown in iron-restricted and iron-rich medium. RNA was reverse transcribed to cDNA using PrimerScript RT regent Kit with gDNA Eraser (Takara, Dalian, China). Real-time PCR reaction was performed using SYBR Premix (Takara, Dalian, China). Each reaction was performed in triplicate. Relative quantification of gene expression was calculated according to 2 <sup>−</sup>11Ct method, RA-YM 16S rRNA was used as reference gene for normalized expression for each RNA sample.

#### The Expression of Fur Protein and Electrophoretic Mobility-Shift Assay (EMSA) of the Putative Fur-Box Sequence

The fragment of fur gene was amplified using the primers Fur1 and Fur2 (introducing the BamHI and XhoI sites); the fragment and vector pET-28a were digested with BamHI and XhoI and restricted fragment was ligated into the expression vector pET-28a to generate the expression vector pET-28a-fur. The expression plasmid was then transformed into competent cells of E. coli BL21 (DE3). Then the Fur protein was purified with an ÄKTA Purifier (GE His Trap FF, USA).

EMSA was performed with the Lightshift Chemiluminescent EMSA Kit (Thermo fisher scientific, Waltham, USA). The reaction was incubated at 30◦C for 1 h, then loaded into 6% nondenaturing polyacrylamide gel electrophoretic and exposured. The reaction mixture (20 µL) contained 1 µg biotin labeled DNA fragment, 2 µL binding buffer, 1 µL KCl, 1 µL MgCl2, 1 µL glycerol, 1 µL NP-40 and 1 µL Poly(dI-dC) and desired concentration of Fur protein, the final concentration of Fur protein were 0, 0.1, 1, and 10 µg in four lanes. 16S rDNA was used as a negative control. DNA fragments to be identified were amplified by biotin labeled primer (Sangon, Shanghai, China). The length of DNA fragments ranged from 350 to 420 bp.

#### Assessment of Virulence *In vivo*

One-day-old Cherry Valley ducklings obtained from the Wuhan Duck Farm (Wuhan, China) housed in cages under 12-h light/dark cycle, at controlled temperature (28–30◦C) and free access to food and water during the whole course of this study. Care and maintenance of all animals were in line with the standards of Institutional Animal Care. This experiment was approved by the Institutional Animal Experimental Committee of the Veterinary Faculty of Huazhong Agricultural University.

To determine the role of fur in virulence, the median lethal dose (LD50) of the deletion mutant RA-YM 1fur strain, the complemented mutant RA-YM 1fur strain and the wild-type RA-YM was measured using the Reed–Muench method (Reed and Muench, 1938). For each wild type, mutant and complemented strains, 12-day-old ducklings were evenly divided into five groups (10 ducklings/group). All five groups were injected intramuscularly with 1.0 × 10<sup>4</sup> , 1.0 × 10<sup>5</sup> , 1.0 × 10<sup>6</sup> , 1.0 × 10<sup>7</sup> , and 1.0 × 10<sup>8</sup> colony forming units (CFU) of wild type strain, respectively. Similarly, mutant and complemented strains were injected to respective groups of ducklings for the evaluation of LD50. Moribund ducklings were killed humanely and counted as dead. Dead ducklings were identified for the presence of RA. Mortality of the ducklings was recorded daily for a period of 10 days.

A comparative analysis of bacterial load in the blood of ducklings infected with mutants and wild type was made. Blood and target organs (brain, liver, heart and spleen) were collected at 24 and 48 h post-inoculation (five ducklings per group at each time-point). The target organs were homogenized with PBS to obtain supernatant. Blood and supernatant were plated on TSB agar plates for bacterial count with a 10-fold dilution method. In addition, the degree of lesions developing on the liver, spleen, heart and brain by the wild-type and the 1fur mutant strains were also recorded. For pathological investigations, all tissues were immersed in 10% formalin solution, embedded in paraffin section and stained with hematoxylin and eosin (H E). The pathological findings of the wild-type and the RA-YM 1fur mutant were compared.

#### RESULTS

#### Characterization of *R. anatipestifer* 1*fur* Mutant and RA-YM 1*fur* Complemented Mutant Strain

As homologous recombination follows a two-step procedure, the selection of the R. anatipestifer 1fur mutant was carried out in two steps (Stibitz, 1994). The selection of mutant with Spc resistance was initially carried out, followed by the expression of mpheS gene. The function of lacZ gene could directly confirm whether the plasmid had been excised. The pheS gene was engineered by substituting alternative bases at numerous positions. The sequence similarity rate between wildtype pheS and mpheS was 71%. However, no difference was observed in amino acid sequence with the exception of the A301G mutation. In this study, the mpheS gene was driven by the promoter of RA rpsL gene. The first process was obtained by growing RA-YM strains on TSA medium containing Spc resistance. The first process obtained the merodiploid strains, which harbored the suicide vector. Then, the merodiploid strains were screened on TSA medium containing 0.2% cPhe and Xgal to obtain the deletion mutant. The merodiploid strains grew on the plate containing Spc but had no growth on the agar plate with 0.2% cPhe. The wild strains could be grown on the plate contains cPhe (**Figures 1A2,A3**). In addition, the merodiploid strains appeared as blue colonies while the wild type strain was of a white color on the plate containing X-gal (**Figures 1A4,A5**). This finding demonstrated the effectiveness of the counter-selectable markers pheS in RA. A suicide vector pRE-lacZ-mpheS-spc-fur, containing mutated pheS as a counterselectable marker (**Figure 1A1**) was constructed and successfully transformed into RA-YM strain to generate RA-YM 1fur strain.

To determine whether the plasmid replicated and deleted fur gene after homologous recombination, no PCR amplification of fur gene and smaller LR fragment size (1,540 bp) from RA-YM 1fur strain as compared to larger LR fragment size (2,008 bp) from RA-YM strain (**Figure 1C**) confirmed the plasmid activity and recombination. Similarly, development of a recombinant RA-YM 1fur complemented mutant strain was confirmed by PCR amplification of fur and spc genes as shown in **Figure 1D**.

#### Transcriptional Response in Iron Enriched and Restricted Environment

A comparison of gene expression regulated by iron and/or Fur in wild type (WT) and mutant (1fur) strains grown in iron rich (+Fe) and iron restricted condition(−Fe) was exclusively established and comparison of RA-YM 1fur deletion mutant strain with wild type RA-YM was also performed. In our

from RA-YM 1fur deletion mutant strain; Lan 3: Amplification of fur gene from RA-YM; Lane 4: Amplification of fur gene from RA-YM 1fur deletion mutant strain; (D) The PCR amplification of RA-YM 1fur mutant complemented strain. Lane M: DL2000 DNA Marker; Lan 1: Amplification of fur gene wild type RA-YM; Lane 2: Amplification of fur gene RA-YM 1fur complemented mutant strain; Lane 3: Amplification of spc gene from wild type RA-YM; Lane 4: spc gene amplification from the RA-YM 1fur complementary mutant strain.

experiments, the significance of differentially expressed genes was estimated by the false discovery rate (FDR) and was considered significant if FDR < 0.001 and the |log2Ratio| >1(Ernst et al., 2005a; Ledala et al., 2010). In total, 25 genes were downregulated and 45 genes were upregulated by iron when grown in ironrestricted conditions, in both parent and mutant. Seventeen genes were directly regulated by Fur. The expression of eight genes randomly selected regulated by iron and Fur was confirmed by real-time PCR in RA-YM and the 1fur mutant (**Figure 7**). The real-time PCR result was in accordance to the transcriptional data.

#### Downregulation of Genes by Iron

Exclusively, 25 genes were downregulated when grown without iron in medium in both wild type RA-YM and RA-YM 1fur deletion mutant strains. Furthermore, ratio (WT-Fe/WT+Fe ratio and 1fur-Fe/1fur+Fe ratio) of gene expression were calculated in both iron-restricted (−Fe) and iron-rich (+Fe) conditions, which were ≥2. Of the 25 downregulated genes, five genes encoded proteins which acted as transporters; six genes encoded enzymes which participated in tricarboxylic acid cycle; six were involved in oxidation-reduction; six genes encoded hypothetical proteins and two genes actively participated in amino acid biosynthesis (**Table 2**).

#### Upregulation of Genes by Iron

Similarly, 45 genes were upregulated by iron both in the parent and the 1fur mutant strain when iron was restricted. Gene ratio (WT-Fe/WT+Fe ratio and 1fur-Fe/1fur+Fe ratio) was calculated ≥2. Among the 45 upregulated genes, nine genes participated in amino acids and cofactor biosynthesis; six were involved in cell envelope and surface structure formation; eight genes were involved in protein synthesis; transport and binding protein were encoded by six genes; three genes encoded regulators; cellular processes were regulated by three genes, and ten encoded hypothetical proteins (**Table 3**).

#### TABLE 2 | Genes downregulated by iron in response to iron restricted condition.


#### TABLE 3 | Genes upregulated by iron in response to iron restricted condition.


#### Gene Regulation by Fur under Iron Restricted Conditions

In almost all bacteria, Fur acted as a negative regulator. Genes directly regulated by Fur were observed to be upregulated when iron was restricted and when fur was mutated. The ratio (WT-Fe/WT+Fe ratio, 1fur-Fe/1fur+Fe ratio and 1fur+Fe/WT+Fe ratio) of 17 genes regulated by Fur was ≥2 (**Table 4**). Of the 17 genes directly regulated by Fur, five genes contributed in iron acquisition; two were involved in oxidation-reduction; one gene participated in activation of type IX secretion system (T9SS); the functions of six genes remained unknown, and three encoded hypothetical proteins (**Table 4**).

#### Identification of Putative Binding Sequences of Fur-Box Binding to Fur Protein

The putative Fur binding sequence and the distance from the start condon of the genes regulated by Fur is shown in **Table 5**. Promoter sequences were analyzed using software RegPredict and ClustalW for identification of putative binding sequence of Fur protein which was 19 bp long and sequence was predicted as 5′ -ATTTAGAATTATTCTAAAT-3′ (**Figure 8A**). Therefore, the Fur binding sequence might be located within 100 bp of the translation initiation codon of the regulated genes (**Table 5**). To verify the putative role of the Furbox sequence, the promoters of hmuR, sprT, RAYM\_01847, RAYM\_09824 were selected for electrophoretic mobility shift assay (EMSA). Our findings illustrated that purified Fur protein could bind to the DNA fragment containing the putative Fur-box (**Figure 8B**).

#### *In vivo* Evaluation of Virulence of *R. anatipestifer* 1*fur* Deletion Mutants

The LD<sup>50</sup> values of RA-YM, RA-YM 1fur deletion mutant and RA-YM 1fur complemented strain were recorded as 2.0 × 10<sup>6</sup> CFU, 1.6 × 10<sup>8</sup> CFU, 1.2 × 107CFU, respectively. LD<sup>50</sup> counted for wild type as compared to RA-YM 1fur deletion mutant was approximately 80 times higher, whereas no significant difference was observed as in the case of wild type in comparison to RA-YM 1fur complemented mutant that was six times higher. Due to slight difference between wild type and RA-YM 1fur complemented mutant, a further comparison was established only between wild type and RA-YM 1fur deletion mutant strains. As the fur gene was disrupted in RA-YM 1fur deletion mutant strains resulted in attenuation of virulence of RA. However, virulence to ducklings was partially restored when the mutant was complemented with the plasmid pRES-JXrep-spc.

Microbiological analysis of heart, brain, liver and spleen showed that bacterial load was higher in wild type RA-YM strain compared to RA-YM 1fur deletion mutant strain at 24 and 48 h post-infection collection, a detailed comparison is shown in **Figures 2A,B**. Correspondingly, the pathological investigations illustrated that lesions were more significant in wild type as compared to mutant strains. The epicardial tissue of ducklings infected with wild type bacteria consisted of a higher degree of fibrinous exudate and inflammatory cell infiltration as compared to infection with mutant pathogens after 24 and 48 h (**Figure 3**). The lesions in brain tissue after both 24 and 48 h, the subarachnoid space was examined where mild inflammatory cell infiltration was noted in case of RA-YM 1fur deletion mutant strain as compared to wild type RA-YM strains (**Figure 4**). Similarly, a large number of hepatocytes expressed fatty degeneration and slight fibrotic

#### TABLE 4 | Putative genes regulated by Fur under iron restricted condition.


TABLE 5 | Identification of putative Fur binding sequences (Fur boxes).


effusion when infected with wild type RA-YM bacteria, whereas, such lesions were hardly observed in ducklings infected with RA-YM 1fur mutant bacteria after 24 h inoculation. A higher degree of fibrotic effusions in liver tissue was observed after 48 h of infection with wild type pathogens, in comparison to mutant pathogens. Collectively, severe hepatic congestion was noticed in ducklings infected with wild type RA-YM strains (**Figure 5**). Likely, both splenomegaly and congestion of spleen was observed in ducklings after 24 and 48 h in case of infection with wild type pathogens, whereas only splenomegaly was noticed in case of infection with mutant pathogens (**Figure 6**). Conclusively, all groups of ducklings, inoculated with wild type RA-YM strains, were severely infected in comparison to those inoculated with RA-YM 1fur deletion mutant strains. The control group of ducklings showed no significant pathological lesions.

#### DISCUSSION

In the present study, we employed the suicide vector pRE-lacZmpheS-spc-fur to construct an unmarked mutant RA-YM 1fur successfully. Using this technique, the traditional method of mutant development by inducing antibiotic resistance can be circumvented, and influence on the expression of downstream genes can be minimized in some cases. However, the expression of the fur downstream gene RAYM\_04841 remained unchanged. The counter-selectable marker pheS in combination with lacZ have been successfully developed for unmarked gene deletion in RA. Previously, pheS had been also successfully used in Burkholderia (Barrett et al., 2008) and E. faecalis (Kristich et al., 2007), and S. mutans (Xie et al., 2011). The technique was developed to determine the role of fur in the pathogenicity of RA. Indeed, the virulence of RA-YM 1fur mutant strain was attenuated in comparison to wild type and virulence was partially restored when RA-YM 1fur mutant strain was complemented with the plasmid pRES-JXrep-spc. Moreover, bacterial load in different tissues was significantly decreased in RA-YM 1fur mutant infection as compared to wild type strain. Similarly significantly mild lesions were observed in case of RA-YM 1fur mutant infection in comparison to wild type infection. Conclusively, it was observed that Fur regulated virulence factors of RA infection. Indeed, the role of Fur with respect to virulence has been previously examined in Vibrio cholerae (Mey et al., 2005) and Staphylococcus aureus (Johnson et al., 2011). Consequently, findings of this study depicted the role of fur in virulence of RA.

FIGURE 3 | Histopathological diagram of heart. (A1) The blank control group after 24 h, (A2) The group with wild type RA-YM strain after 24 h, (A3) The group with RA-YM 1fur deletion mutant strain after 24 h. (B1) The blank control group after 48 h, (B2) The group with wild type RA-YM strain after 48 h, (B3) The group with RA-YM 1fur deletion mutant strain after 48 h.

RA-YM 1fur deletion mutant strain after 24 h. (B1) The blank control group after 48 h, (B2) The group with wild type RA-YM strain after 48 h, (B3) The group with RA-YM 1fur deletion mutant strain after 48 h.

In the current study, we also recorded the expression of the genes downregulated by iron under iron-restricted conditions. Among those genes, six downregulated genes were involved in regulation of tricarboxylic acid (TCA) cycle, which play an important role in metabolism, energy generation and synthesis of precursors (Vuoristo et al., 2016). Under iron-restricted conditions, certain key enzymes of the TCA cycle of RA-YM 1fur strain were downregulated, which included succinate dehydrogenase (SDH) subunit (RAYM\_01977, RAYM\_01982, RAYM\_01987), fumarate hydratase (RAYM\_00925) and aconitase. SDH is involved in the respiratory chain and Krebs cycle of bacteria (Yankovskaya et al., 2003). Similarly, glyoxylate bypass pathways are also repressed in Yersinia pestis iron-restricted conditions (Pieper et al., 2010). In previous reports, downregulation of SDH, fumurase and aconitase were reported in E. coli (Massé et al., 2005) and Bacillus subtilis (Gaballa et al., 2008) in iron-sparing conditions. This phenomenon may be accounted for as the bacteria utilize an alternative iron-independent pathway of the TCA cycle and repressed numerous of iron-containing proteins under

RA-YM 1fur deletion mutant strain after 48 h.

iron–restricted conditions. Moreover, cysteine synthase A (cysK, a significant enzyme of cysteine biosynthesis) and serine O-acetyltransferase (cysE, catalyzes the acetylation of L-serine to O-acetyl-L-serine) involved in amino biosynthesis were also downregulated due to iron deficiency. This inhibition regulates the conversion of available serine to siderophore, enterobactin, which thereby increases iron acquisition (Salvail et al., 2010). The cytochrome c oxidase (Cco) family related to oxidation-reduction was also inhibited in response to iron limitation, which is comprised of four subunits, CcoN, CcoO, CcoP, and CcoQ, which act as the terminal enzyme of respiratory chain (Ahn et al., 2015; Steininger et al., 2016). The Cco family, a member of heme-copper oxidase superfamily, may play a role in iron- restricted conditions. In Pseudomonas stutzeri, the Cco family has also been reported as an essential element for nitrogen-fixing (Nyquist et al., 2001; Xie et al., 2014). Our data showed that nitrogen-fixing associated genes (RAYM\_07584, RAYM\_07589) were repressed in iron-restricted conditions. In conclusion, iron-sparing responses, which means the repression of iron-dependent genes when iron is deficient, was the vital

reason of the genes downregulated by iron in iron restricted conditions.

Among genes induced by iron, the genes (RAYM\_00510, RAYM\_00515) associated with iron acquisition were upregulated, which included Ferrous iron transport protein A (FeoA) and Ferrous iron transport protein B (FeoB). The feoB gene encodes an inner membrane Fe (II) transporter in multiple bacteria, such as E. coli and V. cholerae whereas feoA gene was demonstrated crucial for FeoB uptake of Fe(II) (Marlovits et al., 2002; Kim et al., 2012). Other genes involved in the iron-acquisition system were also upregulated, including Ferritin (RAYM\_01160), ABC transporter related protein (RAYM\_06607), TonB-dependent receptor (RAYM\_04481). This is an apparent response to iron starvation in bacteria, which has also been demonstrated in Klebsiella pneumonia and Listeria monocytogenes (Ledala et al., 2010; Lin et al., 2011). Similarly, a group of genes contributed to the biosynthetic process of iron-sulfur (Fe-S) were also regulated by iron. In bacteria, Fe-S machinery is comprised of a nitrogen-fixing NifU domain protein (RAYM\_01100), SUF system protein (RAYM\_01495, RAYM\_06457, RAYM\_06467, RAYM\_06507). Similar changes of NIF and SUF systems have been confirmed in E. coli in previous studies (Outten et al., 2004). Moreover, the genes Phosphoserine aminotransferase (RAYM\_04219), D-3-phosphoglycerate-dehydrogenase (RAYM\_04224) controlling the shikimate pathway were upregulated. D-3-phosphoglycerate-dehydrogenase (pabB) converts chorismate to 4-amino-4-deoxychorismate (ADC) and phosphoserine aminotransferase (pabC) converts ADC to p-aminobenzoate (PABA) and pyruvate (Green et al., 1992). Shikimate pathway catalyzes serine to form siderophore, enterobactin (Prévost et al., 2007). The upregulation of the pabB and pabC in iron-limited conditions results in high levels of both aromatic amino acids and phenolate siderophore (Lemaître et al., 2014). Moreover, in our data, sigma factor protein (RAYM \_00365) was upregulated by iron. Iron-starvation sigmas, a extracytoplamic function (ECFs) subfamily, have been demonstrated previously to play a role in iron acquisition in P. aeruginosa (Visca et al., 2002). In conclusion, the genes induced by iron are involved in iron-acquisition, some metabolic pathways and several transcriptional regulation factors. The series of regulatory responses under low-iron conditions resulted in increasing iron acquisition.

Among the genes regulated by Fur under iron-restricted conditions, Hmu system, comprised of hmuY (RAYM\_06175) and hmuR (RAYM\_06180), was observed to be regulated by Fur. The homologous Fur-box sequence has been identified upstream of the hmuY start codon in Porphyromonas gingivalis (Simpson et al., 2000). HmuR, the TonB-dependent receptor for ferric, has previously been illustrated to be regulated by Fur in Y. pestis (Branger et al., 2010). Furthermore, genes (RAYM\_01847, RAYM\_09824, and RAYM\_09779) encoding TonB-dependent outer membrane proteins are also regulated by Fur under iron-restricted conditions. The accessory proteins of TonB system can transduce energy in Gram-negative bacteria (Postle and Kadner, 2003; Lim et al., 2012). Our data confirmed that TonB-ExbB-ExbD system was a significant component for ferric enterobactin acquisition, which was previously reported in Campylobacter (Zeng et al., 2013). In addition, RAYM\_00450 (oxidoreductase), RAYM\_03864 (3-hydroxyacyl-CoA dehydrogenase) were also regulated by Fur which was demonstrated in previous study as Fur regulated the response to oxidative stress in C. jejuni (Holmes et al., 2005). Interestingly, sprT gene (RAYM\_03924), a component of the type IX secretion system (T9SS) was also regulated by Fur. Parallel findings were recorded in Flavobacterium johnsoniae (Kharade and McBride, 2015). T9SS was also characterized as a novel protein secretion system mediated outer membrane translocation to the cell surface in P. gingivalis (de Diego et al., 2016). In conclusion, Fur regulated processes included iron acquisition, oxidation and reduction, and regulation of some components of T9SS.

Fur has been reported to act as a negative regulator. It was demonstrated that Fur protein, together with iron, can bind to a consensus sequence, resulting in transcriptional inhibition

#### REFERENCES

Ahn, Y. O., Lee, H. J., Kaluka, D., Yeh, S. R., Rousseau, D. L., Ädelroth, P., et al. (2015). The two transmembrane helices of CcoP are sufficient for assembly of the cbb3-type heme-copper oxygen reductase from Vibrio cholerae. Biochim. Biophys. Acta 1847, 1231–1239. doi: 10.1016/j.bbabio.2015.06.013

(Baichoo and Helmann, 2002) which had been described in E. coli, P. aeruginosa, Helicobacter pylori, and H. hepaticus (Escolar et al., 1998; Vasil and Ochsner, 1999; Belzer et al., 2007; Pich et al., 2012). The Fur-box consensus sequence 5′ - GATAATGATAATCATTATC-3, was slightly different among bacteria. It was identified as an adjacent hexamer unit of the sequence 5′ -GATTAT-3′ or three repeat of the NATWAT (Lavrrar and McIntosh, 2003). Typically, the Fur binding sequence was located within 150 bp of the translation initiation codon of the regulated genes (Grifantini et al., 2003). The putative Fur-box sequence of RA-YM in our study was predicted as 5′ - ATTTAGAATTATTCTAAAT- 3′ , and the sequences might be located within 100 bp of the translation initiation codon of regulated genes which could be reasoned due to the unique promoter of the strain. RA-YM belongs to the Flavobacterium, where promoter have −7 and −33 consensus elements, whereas the promotor of E. coli has −10 and −35 consensus elements (Chen et al., 2007).

In summary, our work showed that pheS acted as effective counter-selectable marker for conjugal transfer. We successfully constructed an unmarked deletion mutant of RA with the suicide vector pRE-lacZ-mpheS-spc. In addition, we elucidated the role of the fur gene in virulence of RA. Furthermore, we screened out the genes regulated by iron and Fur. The putative Fur-box sequence of RA was also predicted. Conclusively, this was a comprehensive study on the metabolism of R. anatipestifer which may help facilitate the control of this pathogen.

#### AUTHOR CONTRIBUTIONS

YG designed, performed the experimentation, data analysis and wrote the manuscript. DH, JG, XL, and JYG performed experimentations. XW, YX, HJ, ML, and ZL designed and contributed to experimental work. ZZ and DB designed, analyzed the data and revised the manuscript. All authors read and approved the final manuscript.

#### ACKNOWLEDGMENTS

This work was supported by the National Natural Science Foundation of China (31201933) and the Fundamental Research Founds for the Central Universities (52902-0900206127).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fcimb. 2017.00382/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 © 2017 Guo, Hu, Guo, Li, Guo, Wang, Xiao, Jin, Liu, Li, Bi and Zhou. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

## Genetic and Dietary Iron Overload Differentially Affect the Course of Salmonella Typhimurium Infection

Manfred Nairz <sup>1</sup> \*, Andrea Schroll <sup>1</sup> , David Haschka<sup>1</sup> , Stefanie Dichtl <sup>1</sup> , Piotr Tymoszuk <sup>1</sup> , Egon Demetz <sup>1</sup> , Patrizia Moser <sup>2</sup> , Hubertus Haas <sup>3</sup> , Ferric C. Fang4, 5, Igor Theurl <sup>1</sup> and Günter Weiss <sup>1</sup>

*<sup>1</sup> Department of Internal Medicine II, Infectious Diseases, Immunology, Rheumatology, Pneumology, Medical University of Innsbruck, Innsbruck, Austria, <sup>2</sup> Department of Pathology, Medical University of Innsbruck, Innsbruck, Austria, <sup>3</sup> Division of Molecular Microbiology, Biocenter, Medical University of Innsbruck, Innsbruck, Austria, <sup>4</sup> Department of Laboratory Medicine, University of Washington, Seattle, WA, USA, <sup>5</sup> Department of Microbiology, University of Washington, Seattle, WA, USA*

Genetic and dietary forms of iron overload have distinctive clinical and pathophysiological features. HFE-associated hereditary hemochromatosis is characterized by overwhelming intestinal iron absorption, parenchymal iron deposition, and macrophage iron depletion. In contrast, excessive dietary iron intake results in iron deposition in macrophages. However, the functional consequences of genetic and dietary iron overload for the control of microbes are incompletely understood. Using *Hfe*+/<sup>+</sup> and *Hfe*−/<sup>−</sup> mice in combination with oral iron overload in a model of *Salmonella enterica* serovar Typhimurium infection, we found animals of either genotype to induce hepcidin antimicrobial peptide expression and hypoferremia following systemic infection in an Hfe-independent manner. As predicted, *Hfe*−/<sup>−</sup> mice, a model of hereditary hemochromatosis, displayed reduced spleen iron content, which translated into improved control of *Salmonella* replication. *Salmonella* adapted to the iron-poor microenvironment in the spleens of *Hfe*−/<sup>−</sup> mice by inducing the expression of its siderophore iron-uptake machinery. Dietary iron loading resulted in higher bacterial numbers in both WT and *Hfe*−/<sup>−</sup> mice, although Hfe deficiency still resulted in better pathogen control and improved survival. This suggests that Hfe deficiency may exert protective effects in addition to the control of iron availability for intracellular bacteria. Our data show that a dynamic adaptation of iron metabolism in both immune cells and microbes shapes the host-pathogen interaction in the setting of systemic *Salmonella* infection. Moreover, Hfe-associated iron overload and dietary iron excess result in different outcomes in infection, indicating that tissue and cellular iron distribution determines the susceptibility to infection with specific pathogens.

#### Keywords: iron, macrophage, hepcidin, lipocalin, Salmonella, infection, siderophore

### INTRODUCTION

HFE encodes an atypical MHC class I molecule which plays a major role in the regulation of iron homeostasis under basal conditions (Feder et al., 1996; Ludwiczek et al., 2004). HFE mutations, especially the homozygous C282Y substitution, result in type I (AKA classical) hereditary hemochromatosis (HH) (Camaschella et al., 2002; Pietrangelo, 2004; Weiss, 2010), the

#### Edited by:

*Susu M. Zughaier, Emory University, USA*

#### Reviewed by:

*Manuela Raffatellu, University of California, Irvine, USA Susan M. Bueno, Pontifical Catholic University of Chile, Chile*

> \*Correspondence: *Manfred Nairz manfred.nairz@i-med.ac.at*

Received: *27 January 2017* Accepted: *20 March 2017* Published: *11 April 2017*

#### Citation:

*Nairz M, Schroll A, Haschka D, Dichtl S, Tymoszuk P, Demetz E, Moser P, Haas H, Fang FC, Theurl I and Weiss G (2017) Genetic and Dietary Iron Overload Differentially Affect the Course of Salmonella Typhimurium Infection. Front. Cell. Infect. Microbiol. 7:110. doi: 10.3389/fcimb.2017.00110*

most frequent form of HH mainly found in people of Northern or Western European ancestry. HH is characterized by reduced serum levels of the antimicrobial peptide Hamp (hepcidin) and increased duodenal absorption of iron via divalent metal transporter 1 (Dmt1) and ferroportin 1 (Fpn1) despite progressive iron overload in parenchymal organs including the liver, pancreas, and heart (Zoller et al., 1999, 2001; Bridle et al., 2003; Pietrangelo, 2004; Bardou-Jacquet et al., 2013). The precise role of the HFE protein, however, remains incompletely understood. HFE binds to transferrin receptor 1 (TfR1) thus lowering its affinity for iron-laden transferrin (Feder et al., 1998; Lebrón et al., 1998; Bennett et al., 2000). This interaction controls cellular iron acquisition while also modifying the expression of the key iron-regulatory hormone Hamp (Ahmad et al., 2002; Nicolas et al., 2003; Ludwiczek et al., 2005; Vujic Spasic et al., 2008). The latter mechanism involves the sensing of circulating iron levels by TfR1 and TfR2, which reciprocally complex with HFE expressed on hepatocytes (Schmidt et al., 2008; Wallace et al., 2009). Mutations in HFE (or TFR2) impair this iron-sensing mechanism, resulting in the insufficient generation of Hamp and increased iron absorption (Goswami and Andrews, 2006; D'Alessio et al., 2012). Of note, macrophages lacking HFE display an iron-poor phenotype which has been attributed to enhanced iron export (Cairo et al., 1997; Drakesmith et al., 2002; Wang et al., 2003).

Systemic iron availability, erythropoietic iron demand, hypoxia, hormones, and inflammatory signals are key factors that modulate the production of the iron homeostatic regulator Hamp (Nemeth et al., 2004a; Bozzini et al., 2008; Theurl et al., 2010; Armitage et al., 2011; Kautz et al., 2014; Nairz et al., 2014; Canali et al., 2017). Hamp controls iron homeostasis upon binding Fpn1, which triggers Fpn1 internalization, degradation (Nemeth et al., 2004b) and blockade of iron efflux from duodenal enterocytes and macrophages, which recycle iron from senescent erythrocytes. Inflammation-driven Hamp induction thus causes iron sequestration within the mononuclear phagocyte system (MPS), which limits iron availability for extracellular pathogens (Bridle et al., 2003; Ludwiczek et al., 2003; Ganz, 2005; Theurl et al., 2008a).

In infections with the intracellular bacterium Salmonella enterica serovar Typhimurium, macrophages constitute an important habitat for pathogen replication and persistence (Malik-Kale et al., 2011). Because many bacteria are highly dependent on a sufficient supply of iron for their growth and pathogenicity, macrophage iron homeostasis is an important determinant of disease outcome (Nairz et al., 2014). On one hand, macrophage iron overload is associated with the inhibition of IFN-γ-driven antimicrobial immune effector pathways such as nitric oxide synthase 2 (Nos2) expression, resulting in impaired control of intracellular microbes (Weiss et al., 1994; Mencacci et al., 1997; Oexle et al., 2003). On the other hand, severe iron depletion of the host may result in reduced generation of ROS, which also impairs host defenses. In parallel, iron withholding from pathogens constitutes an efficient host defense strategy (Soares and Weiss, 2015). However, macrophages also contribute to host defense by the production of T-cell stimulatory cytokines and antimicrobial peptides (Graziadei et al., 1997). One of the latter, lipocalin 2 (Lcn2; also known as neutrophil gelatinase-associated lipocalin, siderocalin or 24p3), is secreted by neutrophils and macrophages in response to LPS, IL-1ß, IL-17, and IL-22 (Flo et al., 2004; Shen et al., 2006). In its best characterized function, Lcn2 captures iron-laden bacterial siderophores, small molecules that are enzymatically synthesized and actively secreted by many microbes to bind ferric iron with extraordinarily high affinity (Bachman et al., 2009). Lcn2-sensitive siderophores include enterobactin, carboxymycobactins, and bacillibactin. Upon neutralization of these siderophores, Lcn2 contributes to innate resistance against a range of pathogenic bacteria including enterobacteriaceae, mycobacteria and Bacillus anthracis by limiting their access to iron (Flo et al., 2004; Berger et al., 2006).

Salmonella Typhimurium, a facultative intracellular microbe, needs to gain sufficient access to host iron resources as a prerequisite for replication and virulence (Leung and Finlay, 1991; Vazquez-Torres et al., 1999). To acquire the metal from the host and within infected macrophages, Salmonella has evolved both siderophore-dependent and -independent strategies. Salmonella synthesizes catecholate-type siderophores such as enterochelin and salmochelins to capture and internalize ferric iron via siderophore receptors (Bäumler et al., 1998; Rabsch et al., 2003; Fischbach et al., 2005). Alternatively, Salmonella can incorporate non-siderophore-bound ionic iron using the Feo transport system. In addition, the SitABCD system, whose primary function is bacterial manganese import, may contribute through low-affinity uptake of iron (Zaharik et al., 2004). All three pathways of bacterial iron uptake are linked to Salmonella virulence (Tsolis et al., 1996; Janakiraman and Slauch, 2000; Boyer et al., 2002; Crouch et al., 2008; Kim et al., 2013).

Given the central importance of iron for the growth and proliferation of intracellular pathogens such as Salmonella and the important role of Hfe in the regulation of systemic iron balance, we performed experiments to assess the influence of Hfe and/or dietary iron overload on host iron homeostasis and immunity in response to S. Typhimurium infection. This is of specific interest because Hfe results in macrophage iron depletion whereas dietary iron overload leads to iron accumulation within the MPS.

### MATERIALS AND METHODS

#### Salmonella Infection In vivo

All animal experiments described were performed in accordance with Austrian legal requirements. Design of the animal experiments was approved by the Austrian Federal Ministry of Science and Research (approvals BMWF-66.011/0074- C/GT/2007 and /0154-II/3b/2010). Mice were maintained at the central animal facilities of the Medical University of Innsbruck and given free access to water and food. Hfe−/<sup>−</sup> mice were generated as described (Bahram et al., 1999; Flo et al., 2004), crossed back on a C57BL/6 background for at least 10 generations and transferred to the SPF unit of the local Animal Facility by means of embryonic transfer. Hfe+/<sup>−</sup> mice were intercrossed and offspring were genotyped using the following primers (obtained from Microsynth): Hfe fw: 5′ -GAATTAACA

#### GGCCGTTTCTAAAG-3′ , Hfe rev: 5′ -CTTGGAGTAGTGGCT CACACT-3′ , Hfe neo: 5′ -GAGATCAGCAGCCTCTGTTCC-3′ .

For in vivo infection experiments (Supplementary Figure 1), male mice were used at 20–26 week of age and fed either an iron-enriched diet (C1038 from Altromin) supplemented with 25 g/kg carbonyl iron (Sigma) or a standard diet (180 mg Fe/kg, C1000 from Altromin) 3 week before and during infection. Mice were infected i.p. with 500 CFU S. Typhimurium strain ATCC14028 suspended in 200 µl PBS. Animals were monitored 3 times daily for 10 days for signs of illness, and moribund mice were euthanized. Forty-eight and ninety-six hours post-infection, mice were randomly selected for the determination of colony counts. Bacterial load in livers and spleens was determined by plating serial dilutions of organ homogenates on LB agar under sterile conditions. Mice selected for the determination of colony counts were not considered for the recording of survival times.

#### Blood Counts

Blood samples were drawn under anesthesia by retroorbital puncture and collected in heparinized tubes. An aliquot of heparinized blood was used for complete blood count analysis on a Vet-ABC Animal blood counter (Scil animal care company GmbH).

#### Measurement of Iron Parameters

Serum iron was measured using the QuantiChrom Iron Assay kit (BioAssay Systems) according to the manufacturer's instructions. Serum FT was measured by a specific ELISA kit (LifeSpan BioSciences) according to the manufacturer's protocol (Theurl et al., 2016). Total tissue iron content was measured as described (Sonnweber et al., 2012).

#### Histology

Histological examinations of tissues were performed on formalin-fixed tissue sections stained with hematoxylin and eosin (HE) according to a standard protocol (Nairz et al., 2011). Images with HE staining were acquired using a Nikon-Eclipse 80i microscope equipped with a 4x objective with a 0.10 numerical aperture. Image acquisition was performed using NIS-Elements BR3 software.

#### Cell Culture, Salmonella Infection In vitro and Determination of Bacterial Iron Acquisition

Thioglycolate-elicited primary peritoneal macrophages were harvested as described (Schleicher et al., 2005) from C57BL/6 mice of indicated genotypes (detailed below), matched for sex and age, and cultured in RPMI (purchased from Biochrom AG) containing 5% heat-inactivated fetal calf serum (FCS; from PAA), 100 U/mL penicillin, 0.1 mg/mL streptomycin and 10 mM HEPES (all from Sigma). After a 24 h incubation period, macrophages were extensively washed with phosphatebuffered saline (PBS purchased from Invitrogen) and incubated in complete RPMI without antibiotics. Only cell preparations of at least 90–95% purity, as determined by F4/80 surface expression in FACS analysis, were used for subsequent experiments. Macrophages were infected with S. Typhimurium ATCC14028 at a multiplicity of infection (MOI) of 10.

Measurement of bacterial iron acquisition was performed as described elsewhere (Nairz et al., 2008). Briefly, Salmonellainfected macrophages were washed three times and resuspended in serum-free HEPES-buffered RPMI. After the addition of 5µM <sup>59</sup>Fe as citrate (NTBI) or loaded onto human apotransferrin (TBI; Sigma), cells were incubated for an additional 8 h. Intracellular bacilli were harvested according to a modified protocol as described (Olakanmi et al., 2002; Nairz et al., 2008). An aliquot of the bacterial suspension was plated in serial dilutions onto agar plates to quantify released bacteria, while the remaining volume was filtered through centrifugal filter devices with a PDVF membrane of 0.22µm pore size (Millipore). Filters containing the trapped bacteria were used to measure Salmonella-associated <sup>59</sup>Fe with a γ-counter. No association of <sup>59</sup>Fe to S. Typhimurium that had been heat-inactivated at 70◦C for 20 min could be detected.

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

Preparation of total RNA and quantification of mRNA expression by Taqman <sup>R</sup> or SYBR Green <sup>R</sup> RT-PCR following reverse transcription was performed exactly as described (Crawford et al., 2016). Murine primers and probes (Microsynth), the latter carrying 5′ -FAM and 3′ -BHQ1 labels, were used as described elsewhere (Ludwiczek et al., 2005; Theurl et al., 2008b). Bacterial primers and probes have been described (Bearson et al., 2008; Crawford et al., 2016).

#### Statistical Analysis

Statistical analysis was carried out using a SPSS statistical package. Calculations for statistical differences between various groups were carried out by ANOVA and Tukey's correction for multiple tests. Otherwise, a two-tailed unpaired Student's t-test was used. For comparison of survival between subgroups, the Wilcoxon (Gehan) statistic was used. Non-parametric variables (CFU and serum FT) were log-transformed prior to testing. P < 0.05 was used to determine statistical significance, 0.05 ≤ P < 1.0 was considered a statistical trend and depicted.

### RESULTS

#### Influence of Hfe, Dietary Iron Challenge and Salmonella Infection on Iron Parameters

To better understand the influence of Hfe and dietary iron loading on iron homeostasis and the outcome of infection, we used a well-established model of systemic Salmonella infection. Wildtype (WT) C57BL/6 (Hfe+/+) and congenic Hfe−/<sup>−</sup> mice were fed either a standard rodent diet with adequate iron content (IA) or an iron-enriched (IE) diet for 3 weeks prior to and during infection. Congenic WT and Hfe−/<sup>−</sup> mice were then systemically infected with 500 colony-forming units (CFU) of Salmonella enterica serovar Typhimurium ATCC14028 (S. Typhimurium; S. Tm.) via intraperitoneal (i.p.) injection (as delineated in Supplementary Figure 1). Mock-infected controls received a single i.p. injection of PBS as a control (Ctrl.). Animals were monitored for up to 10 days. On days 2 and 4 post-infection, randomly selected animals were sacrificed and bacterial loads (days 2 and 4), erythroid (Supplementary Figures 2B,C) and iron indices as well as the expression of iron metabolic genes (day 2) were evaluated.

As predicted, under control conditions Hfe−/<sup>−</sup> mice fed an iron-adequate (IA) diet showed elevated serum iron and serum ferritin (FT) levels as compared to Hfe+/<sup>+</sup> controls. Both Hfe+/<sup>+</sup> and Hfe−/<sup>−</sup> responded to systemic infection with a reduction of serum iron concentrations (hypoferremia) (**Figure 1A**). Of interest, serum iron levels increased upon dietary iron challenge independent of the Hfe genotype, as Hfe−/<sup>−</sup> and congenic Hfe+/<sup>+</sup> mice displayed comparable serum iron levels on an IE diet. Intriguingly, Hfe+/<sup>+</sup> and Hfe−/<sup>−</sup> mice maintained on an IE diet prior to and during Salmonella infection had even higher serum iron levels as compared to uninfected animals on an IE diet and did not mount a hypoferremic response. Serum FT levels were dramatically increased during dietary iron overload, whereas the stimulatory effect of Salmonella infection on serum FT levels was minimal (**Figure 1B**).

While Hfe−/<sup>−</sup> mice tended to have lower hepatic Hamp mRNA expression as compared to congenic WT mice, the

(boxes), and minimum/maximum ranges. Statistical significant differences within each diet group are indicated. Additional letters represent statistically significant differences (*P* < 0.05) as follows: (a) *Hfe*+/<sup>+</sup> Ctrl. IA vs. *Hfe*+/<sup>+</sup> Ctrl. IE; (b) *Hfe*−/<sup>−</sup> Ctrl. IA vs. *Hfe*−/<sup>−</sup> Ctrl. IE; (c) *Hfe*+/<sup>+</sup> *S*. Tm. IA vs. *Hfe*+/<sup>+</sup> *S*. Tm. IE; *Hfe*−/<sup>−</sup> *S*. Tm. IA vs. *Hfe*−/<sup>−</sup> *S*. Tm. IE. *n* = 7–10 per group.

induced Hamp expression in response to infection or dietary iron overload remained intact in Hfe−/<sup>−</sup> mice compared to WT littermates (**Figure 1C**). As expected, Hfe−/<sup>−</sup> mice had an elevated total iron content in the liver, and the IE diet resulted in hepatic iron accumulation (**Figure 1D**). Serum IL-6 concentrations were unaffected by the Hfe genotype (Supplementary Figure 2A).

As previously shown (Cairo et al., 1997; Nairz et al., 2009), Hfe−/<sup>−</sup> mice had reduced total iron content in the spleen is comparison to Hfe+/<sup>+</sup> mice (**Figure 2A**). Whereas, dietary iron challenge resulted in an increase in spleen iron levels, Salmonella infection caused a small yet significant reduction. Splenic Hamp mRNA expression was not significantly affected by either dietary iron content or infection (**Figure 2B**). In contrast, Fpn1 mRNA expression increased in response to Salmonella infection but was not affected by dietary iron overload or Hfe genotype (**Figure 2C**). Dmt1 and TfR1 mRNA levels were negatively affected by oral iron challenge (**Figures 2D,E**). Concurrent Salmonella infection reverted Dmt1 expression to basal levels, while TfR1 expression remained suppressed. However, there was no substantial influence of Hfe genotype on expression of these iron acquisition molecules. Splenic Lcn2 receptor (LcnR) expression was significantly reduced following Salmonella infection (**Figure 2F**), and Lcn2 mRNA expression in the spleen was higher in Hfe−/<sup>−</sup> as compared to congenic WT mice (**Figure 2G**).

#### Influence of Dietary Iron Content on the Course of Salmonella Typhimurium Infection in WT and Hfe−/<sup>−</sup> Mice

We next studied the influence of dietary iron overload on disease progression in systemic Salmonella infection in Hfe−/<sup>−</sup> and congenic C57BL/6 WT animals. All WT mice died by day 8 of infection independent of their dietary iron content, but animals on an iron-enriched (IE) diet succumbed 1–2 days earlier (**Figure 3A**). Of note, 33% of Hfe−/<sup>−</sup> mice on an ironadequate (IA) or iron-enriched (IE) diet (3 of 9 mice in each group) survived the infection beyond day 10 of the observation period. Hepatic and splenic microbial loads of randomly selected animals were quantified on days 2 and 4 of infection. Dietary iron overload significantly increased the bacterial load in both organs in WT mice as well as in Hfe−/<sup>−</sup> animals 2 days postinfection (**Figures 3B,C**). By day 4 of infection, Hfe−/<sup>−</sup> mice fed an IE diet controlled microbial replication as efficiently as their WT littermates maintained on an IA diet (Supplementary Figures 3A,B). Moreover, tissue sections obtained on day 4 post-infection revealed that WT mice on an IA diet had microabscesses in the liver, which were partly confluent. WT mice on an IE diet also exhibited hepatic macroabscesses. In contrast, hardly any microabscesses were observed in the livers of Hfe−/<sup>−</sup> mice on an IA diet, and only solitary lesions were visible in Hfe−/<sup>−</sup> mice on an IE diet (**Figure 4A**). Similar observations were made in the spleens of Salmonella-infected mice on day 4 post-infection. Only Hfe−/<sup>−</sup> mice on an IA diet had a relatively normal spleen size (**Figure 4B**). These histopathologic findings were accompanied by corresponding alterations in spleen weight (Supplementary Figures 4A,B).

mRNA levels of iron metabolic genes (B–G) in the spleen as determined by quantitative RT-PCR were measured after 48 h. Hamp (B), Fpn1 (C), Dmt1 (D), TfR1 (E), LcnR (F), and Lcn2 (G) mRNA levels were determined relative to the house-keeping gene Hprt at baseline and 48 h post-infection. Data were compared by means of ANOVA with Tukey's *post hoc* test. Values are depicted as lower quartile, median and upper quartile (boxes), and minimum/maximum ranges and only statistically significant differences are indicated exactly as described for Figure 1. Additional letters represent *(Continued)*

#### FIGURE 2 | Continued

statistically significant differences (*P* < 0.05) as follows: (a) *Hfe*+/<sup>+</sup> Ctrl. IA vs. *Hfe*+/<sup>+</sup> Ctrl. IE; (b) *Hfe*−/<sup>−</sup> Ctrl. IA vs. *Hfe*−/<sup>−</sup> Ctrl. IE; (c) *Hfe*+/<sup>+</sup> *S*. Tm. IA vs. *Hfe*+/<sup>+</sup> *S*. Tm. IE; *Hfe*−/<sup>−</sup> *S*. Tm. IA vs. *Hfe*−/<sup>−</sup> *S*. Tm. IE. *n* = 7–10 per group.

#### Classical Innate Immune Functions Are Hfe-Independent

To better define the role of Hfe and dietary iron overload in innate immune function, cytokine and antimicrobial effector system expression was measured on day 2 post-infection. Splenic mRNA levels of TNF, IL-1ß, IL-6, Nos2, and the p47 subunit of the NADPH oxidase (phox) were not affected by the Hfegenotype (**Figures 5A–E**). Increased expression of NOS2 and IL-1ß in mice receiving an IE diet relative to those receiving an IA diet paralleled the increased number of bacteria isolated from the spleens of these groups.

#### Salmonella Adapts to the Iron-Restricted Myeloid Compartment of Hfe−/<sup>−</sup> Mice

The expression of bacterial iron uptake genes was measured in the spleens of Salmonella-infected mice on day 2 post-infection. We found that multiple genes involved in iron uptake were expressed at higher levels in the spleens of Hfe−/<sup>−</sup> mice on an IA diet as compared to the spleens of WT mice receiving the same diet. These genes encoded outer membrane siderophore receptors IroN, FepA, and CirA, as well as the siderophore exporter IroC (**Figures 6A–D**). In contrast, expression of FeoB (**Figure 6E**) and SitB (**Figure 6F**) were not substantially affected

by the Hfe status of the host. Notably, no differential induction of bacterial iron genes was observed when WT and Hfe−/<sup>−</sup> mice were on an IE diet, which resulted in splenic iron overload.

In keeping with the induction of siderophore-mediated iron uptake pathways in Salmonella residing in the spleens of IA diet-fed Hfe−/<sup>−</sup> mice in vivo, we observed that the uptake of host derived <sup>59</sup>Fe by intracellular Salmonella, provided as NTBI or TBI, was reduced in Hfe−/<sup>−</sup> peritoneal macrophages infected in vitro relative to congenic WT macrophages (**Figures 7A,B**).

#### DISCUSSION

Salmonella Typhimurium causes a systemic disease in mice characterized by a tropism for and replication within professional phagocytes (Richter-Dahlfors et al., 1997; Coburn et al., 2007). S. Typhimurium invades its preferred host cell type both by phagocytic uptake and active invasion (Pfeifer et al., 1999). Virulence factors such as those encoded by Salmonella Pathogenicity Island-2 are essential for both intracelullar survival and virulence (Hensel et al., 1995, 1998), suggesting that the ability to infect and replicate within macrophages provides a major benefit for the pathogen (Leung and Finlay, 1991). Nutrient availability within this host cell niche is therefore an important factor in Salmonella pathogenesis (Carver, 2014).

Iron is one of the essential nutrients that hold a central position in the interplay of host and pathogen (Weinberg, 1974; Schaible and Kaufmann, 2005; Skaar, 2009; Nairz et al., 2010; Drakesmith and Prentice, 2012; Ganz and Nemeth, 2015; Soares and Weiss, 2015). Sufficient access to this trace element is therefore a major determinant of the outcome of Salmonella infection. In general, the host response to any bacterial infection involves the restriction of serum iron levels (hypoferremia) through a combined limitation of intestinal iron absorption and macrophage iron recycling. Hamp and its receptor Fpn1 are primary mediators of the hypoferremia, and thereby influence Salmonella-host interactions (Nairz et al., 2013; Kim et al., 2014; Armitage et al., 2016). However, a range of additional genetic and environmental factors also influence bacterial iron availability during Salmonella infections.

The data presented herein suggest that both the local and systemic availability of iron within the mammalian host affect

was measured by qPCR. Expression of *iroN* (A), *fepA* (B), *cirA* (C), *iroC* (D), *feoB* (E), and *sitB* (F) was determined relative to the housekeeping gene *gyrB*. Data were compared by means of ANOVA with Tukey's *post hoc* test. Values are depicted as lower quartile, median and upper quartile (boxes), and minimum/maximum ranges and statistical significances are indicated. *n* = 14–17 per group.

infection outcome (Nairz et al., 2015a). Hfe deficiency results in reduced Hamp production and increased serum iron levels (Pietrangelo, 2004). Enhanced bacterial replication might be anticipated in an Hfe-deficient host as a result of increased iron availability, and this has been experimentally demonstrated for pathogens such as Vibrio vulnificus, Yersinia enterocolitica, and Yersinia pestis in the setting of HH (Quenee et al., 2012; Arezes et al., 2015; Miller et al., 2016). However, the opposite is true in the context of infection with an intracellular bacterium such as S. Typhimurium because Hfe-deficient macrophages are ironpoor and thus provide an inferior niche for bacterial replication (Nairz et al., 2009). We find that Hfe-deficient macrophages restrict the availability of both non-transferrin-bound (NTBI) and transferrin-bound (TBI) iron to intracellular Salmonella more efficiently than WT macrophages. This suggests that the underlying mechanism is independent of Dmt1 and TfR1 and may be attributable to differential iron turnover or efflux. While Fpn1 constitutes the primary pathway for the cellular release of ferrous iron, iron may also be exported via alternative pathways (Devireddy et al., 2005; Keel et al., 2008; Du and Galán, 2009; Nairz et al., 2015b; Lok et al., 2016).

Slc11a1 (also known as Nramp1) has long been known to influence the course of infection with S. Typhimurium and certain species of Mycobacterium and Leishmania (Vidal et al., 1993; Atkinson et al., 1997; Blackwell et al., 2003). Although these pathogens are taxonomically unrelated, they share the features of infecting macrophages, persisting in phagolysosomes and depending on iron. Slc11a1 is incorporated into the phagolysosomal membrane and shifts iron and other divalent ions out of this compartment, thus withdrawing it from phagocytosed microbes (Vidal et al., 1995; Jabado et al., 2000; Wyllie et al., 2002; Fritsche et al., 2007; Valdez et al., 2008). We used C57BL/6 mice for our studies, which carry two dysfunctional Slc11a1 alleles. Therefore, the phenotypes observed in our studies cannot be attributed to this transporter. Moreover, our findings are unlikely to be specific for infections with S. Typhimurium but are also relevant to other iron-dependent intracellular pathogens such as Chlamydia, Legionella and Listeria (Paradkar et al., 2008; Bellmann-Weiler et al., 2010, 2013; Haschka et al., 2015), as well as Mycobacterium and Leishmania. Accordingly, we note that Hfe deficiency impairs the growth of

Mycobacterium tuberculosis in human macrophages (Olakanmi et al., 2007).

The iron content of macrophages is influenced by several mechanisms including iron levels in the extracellular microenvironment, expression of iron importers, and exporters, and the rates of erythrophagocytosis and heme-iron recycling (Canonne-Hergaux et al., 1999; Mitterstiller et al., 2016; Theurl et al., 2016). Our models of genetic (i.e., Hfe-associated) and dietary iron overload had different effects on macrophage iron content. Long-term oral iron overload results in increased iron content in virtually all cell types expressing Dmt1 and/or TfR1. In contrast, Hfe deficiency spares the myeloid compartment from iron. We found that oral iron overload results in an increased bacterial load in the spleens and livers of Salmonellainfected mice and that Hfe deficiency reduces the bacterial load. Of note, oral iron overload in the setting of Hfe deficiency resulted in intermediate pathogen numbers in spleen and liver on days 2 and 4 of infection. In contrast, the survival of Hfe−/<sup>−</sup> mice over a 10 day period was not affected by dietary iron. Hfe−/<sup>−</sup> mice with dietary iron overload remain more resistant to Salmonella infection than WT mice receiving the same diet, with reduced organ loads and increased survival (**Figure 3**) despite the alleviation of bacterial iron-deprivation by dietary iron supplementation as measured by siderophore gene expression (**Figure 6**). Albeit somewhat unexpected, these findings suggest that in the later stages of Salmonella infection, Hfe plays an immunoregulatory function that is independent of its effect on bacterial iron-restriction. However, the expression of innate immune genes known to mediate host defense against S. Typhimurium was not different between Hfe−/<sup>−</sup> and Hfe+/<sup>+</sup> mice (Vazquez-Torres et al., 2000; Vázquez-Torres et al., 2001), nor were differences in T cell-mediated pathways associated with immunity against S. Typhimurium such as IL-12, IFN-γ, IL-17, or IL-22, observed (data not shown) (Berger et al., 2006; Raffatellu et al., 2008; Saiga et al., 2008; Schulz et al., 2008; Chan et al., 2009; Godinez et al., 2009; Srinivasan et al., 2012). Given that Lcn2 was expressed at higher levels in Hfe−/<sup>−</sup> mice, it is possible that one of the siderophore-independent effects of Lcn2 may play a role. Lcn2 is a chemoattractant for neutrophils, but this is unlikely to account for a survival difference beyond day 3 of infection (Schroll et al., 2012). Furthermore, Lcn2 promotes macrophage antibacterial effector mechanisms including TNF, IL-6, and Nos2, but these were not observed to be differentially expressed on day 2 or 4 of infection of Hfe−/<sup>−</sup> and Hfe+/<sup>+</sup> mice (Nairz et al., 2015b). It is conceivable that the survival of mice in the late stages of systemic Salmonella infection is directly or indirectly influenced by the intestinal microbiome, which is modulated by Lcn2 (Raffatellu et al., 2009; Deriu et al., 2013; Moschen et al., 2016). Alternatively, the comparable survival of Hfe−/<sup>−</sup> mice on an IE on IA diet may involve an Lcn2-independent mechanism beyond innate immunity. For instance, Hfe deficiency may have beneficial effects on apoptosis, ferroptosis, autophagy, or the oxidative stress response within or outside of the myeloid compartment that is independent of dietary iron. If a vital organ system were to be involved, an effect on host survival would be a plausible. An unbiased approach such as RNA-sequencing may be required to identify such a mechanism.

A central and novel finding of our study is that both the host and the microbe adapt their iron metabolism during infection. S. Typhimurium expressed genes required for siderophoremediated iron uptake in vivo in the iron-poor spleens of Hfe−/<sup>−</sup> mice. This induction was specifically abrogated by dietary iron overload. This observation raises the question whether virulence factors other than siderophore genes may have been repressed in the setting of dietary iron overload to enhance the survival of Hfe−/<sup>−</sup> mice. This possibility is supported by the known cross-regulation of bacterial iron homeostasis and virulence gene expression (Zaharik et al., 2002). Our data on the upregulation of bacterial iron uptake genes are further in line with the specific induction of iron import mechanisms reported for Neisseria gonorrhoeae residing within human monocytes (Zughaier et al., 2014). Both studies thus support the concept that both host myeloid cells and facultatively intracellular bacteria actively compete for iron as essential nutrient.

In summary, the present study highlights the central role of macrophage iron homeostasis in the outcome of infections with iron-dependent intracellular microbes and the differential effects of genetic and dietary iron overload. We also demonstrate that Hfe is not required for the induction of hypoferremia in infected animals on an iron-replete diet. Nevertheless, the Hfe mutation alters the iron content of macrophages, which renders the host more resistant to infections with the intracellular pathogen S. Typhimurium. The selective pressure imposed by intracellular pathogens may have contributed to the evolutionary conservation of the HFE C282Y mutation, accounting for its high allelic frequency in Caucasians (Datz et al., 1998; Moalem et al., 2004).

#### AUTHOR CONTRIBUTIONS

MN planned and performed experiments, analyzed the data and wrote the manuscript. AS, DH, SD, PT, ED, and PM performed experiments. HH, FF, and IT interpreted results and edited the manuscript. GW conceived the study, analyzed the data and wrote the manuscript.

#### ACKNOWLEDGMENTS

We are grateful to Sylvia Berger, Ines Brosch, Sabine Engl, and Markus Seifert for excellent technical support. This work was

#### REFERENCES


supported by the "Verein zur Förderung von Forschung und Weiterbildung in molekularer Immunologie und Infektiologie" and by grants from the Austrian Research Fund FWF (W1253- Doctoral college-HOROS to GW, HH, and SD as well as TRP-188 to GW). FF receives support from the National Institues of Health (AI118962).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fcimb. 2017.00110/full#supplementary-material


essential for virulence of Salmonella enterica serovar Typhimurium. Mol. Microbiol. 67, 971–983. doi: 10.1111/j.1365-2958.2007.06089.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 © 2017 Nairz, Schroll, Haschka, Dichtl, Tymoszuk, Demetz, Moser, Haas, Fang, Theurl and Weiss. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

## *Mycobacterium tuberculosis* Glyceraldehyde-3-Phosphate Dehydrogenase (GAPDH) Functions as a Receptor for Human Lactoferrin

Himanshu Malhotra1‡, Anil Patidar 1‡, Vishant M. Boradia2 †‡, Rajender Kumar 3 † , Rakesh D. Nimbalkar <sup>3</sup> , Ajay Kumar <sup>2</sup> , Zahid Gani <sup>2</sup> , Rajbeer Kaur <sup>2</sup> , Prabha Garg<sup>3</sup> , Manoj Raje<sup>1</sup> and Chaaya I. Raje<sup>2</sup> \*

<sup>1</sup> Cell Biology and Immunology, Council of Scientific and Industrial Research-Institute of Microbial Technology, Chandigarh, India, <sup>2</sup> Department of Biotechnology, National Institute of Pharmaceutical Education and Research, Punjab, India, <sup>3</sup> Department of Pharmacoinformatics, National Institute of Pharmaceutical Education and Research, Punjab, India

*Edited by:* Susu M. Zughaier, Emory University, United States

#### *Reviewed by:*

Ashu Sharma, University at Buffalo, United States Teresa Olczak, University of Wrocław, Poland

> *\*Correspondence:* Chaaya I. Raje chaaya@niper.ac.in

#### *† Present Address:*

Vishant M. Boradia, Department of Microbiology and Immunology, University of Maryland School of Medicine, Baltimore, MD, United States Rajender Kumar, Architecture et Fonction des Macromolécules Biologiques UMR 7257 Case 932, Marseille, France

‡ These authors have contributed equally to this work.

> *Received:* 21 March 2017 *Accepted:* 26 May 2017 *Published:* 08 June 2017

#### *Citation:*

Malhotra H, Patidar A, Boradia VM, Kumar R, Nimbalkar RD, Kumar A, Gani Z, Kaur R, Garg P, Raje M and Raje CI (2017) Mycobacterium tuberculosis Glyceraldehyde-3-Phosphate Dehydrogenase (GAPDH) Functions as a Receptor for Human Lactoferrin. Front. Cell. Infect. Microbiol. 7:245. doi: 10.3389/fcimb.2017.00245 Iron is crucial for the survival of living cells, particularly the human pathogen Mycobacterium tuberculosis (M.tb) which uses multiple strategies to acquire and store iron. M.tb synthesizes high affinity iron chelators (siderophores), these extract iron from host iron carrier proteins such as transferrin (Tf) and lactoferrin (Lf). Recent studies have revealed that M.tb may also relocate several housekeeping proteins to the cell surface for capture and internalization of host iron carrier protein transferrin. One of the identified receptors is the glycolytic enzyme Glyceraldehyde-3-phosphate dehydrogenase (GAPDH). This conserved multifunctional protein has been identified as a virulence factor in several other bacterial species. Considering the close structural and functional homology between the two major human iron carrier proteins (Tf and Lf) and the fact that Lf is abundantly present in lung fluid (unlike Tf which is present in plasma), we evaluated whether GAPDH also functions as a dual receptor for Lf. The current study demonstrates that human Lf is sequestered at the bacterial surface by GAPDH. The affinity of Lf-GAPDH (31.7 ± 1.68 nM) is higher as compared to Tf-GAPDH (160 ± 24 nM). Two GAPDH mutants were analyzed for their enzymatic activity and interaction with Lf. Lastly, the present computational studies offer the first significant insights for the 3D structure of monomers and assembled tetramer with the associated co-factor NAD+. Sequence analysis and structural modeling identified the surface exposed, evolutionarily conserved and functional residues and predicted the effect of mutagenesis on GAPDH.

Keywords: *Mycobacterium tuberculosis*, iron, transferrin, lactoferrin, glyceraldehyde-3-phosphate dehydrogenase (GAPDH)

### INTRODUCTION

Iron acquisition is vital for the survival of Mycobacterium tuberculosis, the causative agent of tuberculosis. In order to survive and replicate within the iron restricted intracellular environment of the host cell, these pathogens utilize multiple strategies to pilfer iron from host resources. In the mammalian host iron remains tightly sequestered to either storage proteins or to transport proteins such as Tf and Lf. One strategy used by M.tb to acquire iron is to synthesize high affinity chelators known as siderophores (De Voss et al., 1999; Ratledge, 2004; Banerjee et al., 2011). Recently an alternate siderophore–independent pathway has been identified wherein human holo-Tf is directly captured at the bacterial surface and subsequently internalized. The M.tb Tf receptors identified included several conserved proteins namely Glyceraldehyde-3-phosphate dehydrogenase (GAPDH, Rv1436); Lactate dehydrogenase (Rv1872c); Iron regulated Elongation factor tu (Rv0685); Acyl desaturase (Rv0824c); 50S ribosomal protein L2rplB (Rv0704); 50S ribosomal protein L1rplA (Rv0641) (Boradia et al., 2014). The essential house-keeping glycolytic enzyme GAPDH operates as a multifunctional protein in both eukaryotes (Sirover, 1999) and prokaryotes including M.tb (Bermudez et al., 1996; Carroll et al., 2010). Earlier GAPDH has been identified as a dual receptor for both Tf and Lf in mammalian cells (Raje et al., 2007; Rawat et al., 2012). GAPDH therefore plays analogous roles in Tf acquisition of iron by M.tb as well as its human host (Boradia et al., 2014).

Tf is predominantly present in the blood and is involved in iron transport to cells, while Lf is abundantly present in human milk, mucosal secretions, and neutrophil secretory granules. The two iron carrier proteins Lf and Tf share significant sequence identity (60%) and have a conserved bi-lobed structure that each bind single atom of Fe3+. Lf plays a vital role in iron sequestration, transport, and as an immunomodulator. Several studies have revealed that it plays a protective role by enhancing phagocytosis, inhibiting biofilm formation, and preventing of microbe-host interactions (Testa, 2002; Siqueiros-Cendón et al., 2014).

While Tf iron acquisition in M.tb has been extensively assessed, limited information is available regarding acquisition of Lf iron in this pathogen. A few previous reports have indicated that M.tb bacilli in culture and intracellular bacilli can acquire Lf associated iron. Infact these studies have shown that iron is acquired several fold more efficiently from Lf than Tf (Olakanmi et al., 2004), however the exact mechanisms involved remain unknown.

In the current study using in vitro and cell based models, we demonstrate for the first time that M.tb acquires iron by utilizing GAPDH as receptor for Lf. Ligand binding analysis demonstrated that GAPDH had a greater affinity for Lf than Tf. Lf uptake by M.tb was evident in virulent, attenuated, and even siderophore negative M.tb strains with Lf being trafficked to the intraphagosomal bacilli. Taken together, these findings suggest that iron acquisition from Lf is independent of the siderophore pathway. Mutagenesis and in silico studies provided an insight into the structure and function of GAPDH. This study also provides the first 3D model for M.tb GAPDH and confirms that enzyme activity and Lf binding are unrelated.

#### MATERIALS AND METHODS

#### Plasmids and Strains

The M.tb H37Rv 1mbtB::hyg, siderophore knockout strain was received as a kind gift from Dr. CE Barry, Tuberculosis Research Section, Laboratory of Host Defenses, National Institute of Allergy and Infectious Disease, Rockville, Maryland (De Voss et al., 2000). M.tb H37Ra and H37Rv strains were transformed with pSC301 plasmid (Cowley and Av-Gay, 2001) to generate the M.tb H37Ra-GFP and M.tb H37Rv-GFP strains. M.tb H37Ra strains expressing mCherry or GAPDH-mCherry were prepared previously in our laboratory (Boradia et al., 2014).

The full length M.tb GAPDH gene was cloned and expressed in M.tb H37Ra to obtain recombinant wild type GAPDH (wt rGAPDH) as described previously (Boradia et al., 2016). During plasmid screening two point mutations were detected (i) an Arginine to Serine mutation at position 142 (N142S) and (ii) a Proline to Leucine mutation at position 295 (P295L). Each plasmid was individually transformed into M.tb H37Ra using standard methods (Wards and Collins, 1996), recombinant proteins were purified as described previously (Boradia et al., 2016), the mutant proteins are referred to as rGAPDH(N142S) and rGAPDH(P295L) respectively.

#### Conjugation of Proteins

Human Lf (Sigma Chemical Co.) was biotinylated using Sulfo-NHS-LC-Biotin as per the manufacturer's instructions (Pierce). Hilyte Fluor (HF)-488 was conjugated to Lf using AnaTagTM HiLyte FluorTM 488 Protein Labeling Kit according to the manufacturer's instructions.

#### Surface Binding of Holo-Lf-HiLyte Fluor488 (Lf-HF-488) by FACS Analysis

M.tb H37Ra cells were cultured to log phase, 2 × 10<sup>8</sup> cells were used per assay. Cells were washed with PBS and blocked for 1 h at 4◦C with PBS containing 2% BSA. Cells were then incubated with 10 µg Lf-HF488 alone or in the presence of 100-fold excess unlabeled human holo-Lf at 4◦C for 2 h. Finally, cells were washed extensively with PBS and the fluorescence data of 10<sup>4</sup> cells per assay was acquired using a Guava 8HT flow cytometer. Experiments were repeated thrice, statistical significance was determined by Student's t-test.

#### Cell Surface Binding of Lf-Gold by Transmission Electron Microscopy (TEM)

To assess Lf binding, M.tb H37Ra bacilli were cultured to log phase, 2 × 10<sup>8</sup> cells were used per assay. Cells were washed with PBS and blocked in PBS containing 2% casein at 4◦C for 1 h. Cells were then incubated with 10 µg of Lf- gold (20 nm) in PBS containing 0.2% casein at 4◦C for 2 h. Cells were washed and fixed in Karnovsky's fixative for 30 min, followed by two washes with 5 mM NaCl. Finally, cells were resuspended in PBS, placed on carbon coated grids and viewed in a JEOL 2100 Transmission Electron Microscope (TEM). As a negative control, samples were similarly incubated with streptavidin gold conjugate (20 nm, Sigma) instead of Lf. Gold particles were prepared as described previously (Boradia et al., 2014).

#### Lf Surface Binding and Uptake into *M.tb* H37Ra Bacilli Overexpressing GAPDH mCherry and mCherry

M.tb H37Ra mCherry and M.tb H37Ra GAPDH mCherry expressing cells were cultured to log phase, 1 × 10<sup>8</sup> cells were used per assay. Cells were washed with iron free Sauton's media and blocked for 1 h at 4◦C with 2% BSA in PBS. Cells were then incubated with Lf conjugated to Alexa fluor 647 (Lf-A647, 10 µg per 100 µl of iron free Sauton's media) at 4◦C for 2 h. Finally, cells were washed extensively with iron free Sauton's media and fluorescence data of 10<sup>4</sup> cells per assay was acquired using a BD FACS Verse instrument. Experiments were repeated twice in duplicates, statistical significance was analyzed by Student's t-test. Data is represented as an overlay histogram and as a bar graph of MFI values.

For uptake experiments, cells were processed as before, blocking was done with SFM containing 2% BSA for 30 min at 4◦C. Subsequently, bacteria were incubated with 10 µg Lf-A647 at 37◦C for 2 h. Cells were then washed twice and treated with 0.1% pronase at 4◦C for 5 min on ice to remove any residual surface bound proteins. Finally, cells were washed extensively with iron free Sauton's media and fluorescence data of 10<sup>4</sup> cells per assay was acquired using a BD FACS Verse instrument. Experiments were repeated twice in duplicates, statistical significance was analyzed by Student's t-test. Data is represented as an overlay histogram and as a bar graph of MFI values.

#### Evaluation of Lf-Iron Uptake

Log phase M.tb H37Ra mCh and M.tb H37Ra GAPDH-mCh cells, (2 × 10<sup>8</sup> cells / assay) were washed twice with Sauton's media without iron followed by incubation with the same media for 2 h at 37◦C. Cells were then washed and incubated for a further 1 h at 37◦C with media containing 50 µg of Lf-55Fe in 100 µl Sauton's media without iron. Cells were washed thrice with iron-free media, centrifuged, and resuspended in 3.0 ml of scintillation fluid. The radioactivity associated with the cells was measured using a beta counter (Perkin Elmer), counts were normalized to cell number. Statistical significance was estimated by Student's t-test (n = 6) (Boradia et al., 2014).

#### FRET Analysis of Lf-GAPDH Interaction

Log phase M.tb H37Ra mCh and M.tb H37Ra GAPDH-mCh cells, (2 × 10<sup>8</sup> cells/assay) were washed and blocked with 2% BSA followed by incubation with 20 µg Lf-FITC in 100 µl PBS, for 2 h at 4◦C. Samples were fixed in 4% buffered paraformaldehyde for 20 min and FRET analysis was carried out by the acceptor photobleaching method on a Nikon A1R confocal microscope as described previously (Raje et al., 2007). The intensities of a total of 70 regions of interest from multiple, randomly selected fields were measured before and after photobleaching, and FRET efficiency was calculated (Boradia et al., 2014). Statistical analysis was done by Student's t-test.

#### Lf Binding on *M.tb* H37Rv and *M.tb* H37Rv1mbtB::hyg Bacilli

M.tb H37Rv and M.tb H37Rv 1mbtB::hyg strains were cultured to log phase, and 1 × 10<sup>8</sup> cells were used per assay. Cells were washed with phosphate-buffered saline (PBS) and blocked for 1 h at 4◦C with 2% BSA in SFM (serum free media). Cells were then incubated with Lf-A647 (10 µg per 100 µl of SFM) alone or in the presence of 200-fold excess unlabeled human Lf at 4◦C for 2 h. Finally, cells were washed extensively with SFM and fluorescence data of 10<sup>4</sup> cells per assay was acquired using a BD FACS Accuri instrument. Experiments were repeated thrice in duplicates. Statistical significance was analyzed by Student's t-test.

#### Lf Uptake in *M.tb* H37Rv and *M.tb* H37Rv1mbtB::hyg Bacilli

Cells **(**1 × 10<sup>8</sup> bacilli/assay) were blocked with SFM containing 2% BSA for 30 min at 4◦C. Subsequently, bacteria were incubated with 10 µg Lf-A647. After incubation at 37◦C for 2 h, cells were treated with 0.1% pronase at 4◦C for 5 min to remove any residual surface bound proteins. Cells were washed extensively with SFM and fluorescence data of 10<sup>4</sup> cells per assay was acquired using a BD FACS Verse instrument, experiments were repeated thrice in triplicates. Statistical significance was analyzed by Student's t-test.

#### Trafficking of Lf to Intracellular Bacilli

The human macrophage cell line, THP-1, was utilized as an infection model. To study trafficking of Lf to bacteria resident within macrophages, THP-1 cells were plated into Matek glass bottom petri dishes (3 × 10<sup>5</sup> cells per dish). Cells were activated with 25 ng/ml PMA (Sigma) for 24 h and rested for an additional 24 h. Cells were then shifted to antibiotic-free RPMI media containing 10% fetal calf serum and infected with bacilli at a ratio of 1:20 (THP-1: bacteria) using log phase cultures of M.tb H37Rv-GFP, M.tb H37Ra mCherry, or M.tb H37Ra GAPDH-mCherry. After 6 h, cells were washed with serum-free media (SFM) to remove non-phagocytosed bacteria. A second wash with SFM was carried out after 24 h of infection and cells were shifted to SFM for 30 min at 37◦C. Cells were then washed and incubated with 10 µg of Lf-A647 in 100 µl SFM media at 4◦C for 30 min. Subsequently, for internalization of bound Lf cells were shifted to 37◦C, 5% CO<sup>2</sup> for 1 h. After extensive rinsing with SFM, cells were fixed with 2% paraformaldehyde for 15 min and imaged on a Nikon A1R confocal microscope using a 60X oil immersion objective and an aperture of 1 airy unit.

#### Live Cell Imaging of *M.tb* H37Ra-GFP Infected THP-1 Cells

THP-1 cells were plated and processed for infection with log phase cultures of M.tb H37Ra-GFP and incubation with Lf-A647 as described above. Subsequently cells were imaged live on a Nikon A1R confocal microscope equipped with an Okolab bold line water-jacket cryo topstage CO<sup>2</sup> incubator system (Okolab Italy).

#### Immuno-Gold Labeling TEM to Detect Internalized Lf in *M.tb* H37Rv

Cells **(**2 × 10<sup>9</sup> bacilli per sample) were blocked with SFM containing 2% BSA for 30 min at 4◦C. Subsequently, bacteria were incubated with 1 mg of Lf-conjugated with 20 nm gold particles for 2 h at 37◦C. Subsequently cells were treated with 0.1% pronase at 4◦C for 10 min to remove any residual surface bound proteins. Cells were then fixed, embedded in epoxy resin and ultrathin sections cut for visualization in TEM essentially as described previously (Boradia et al., 2014).

#### Expression and Purification of Mutants and GAPDH Enzymatic Activity

Both mutant forms of the protein were expressed and purified as described previously for wt rGAPDH (Boradia et al., 2016). Purification was confirmed by western blotting, detection was done after incubation with Mouse anti-His (Sigma) (1:3,000) for 1 h followed by incubation with anti-Mouse IgG HRP (Sigma) (1:8,000) for 1 h. The enzyme activity of wt rGAPDH, rGAPDH(N142S), and rGAPDH(P295L) purified from cytosol fraction was studied by measuring the increase in the absorbance at 340 nm due to oxidative reduction of NAD<sup>+</sup> to NADH. The reaction mixture containing 200 µl of enzyme assay buffer (50 mM HEPES, 10 mM sodium arsenate, and 5 mM EDTA, pH 8.5), 1 mM NAD<sup>+</sup> and 2 mM glyceraldehyde-3-phosphate (G-3-P) was added to wells containing 100 ng of each purified enzyme at 25◦C. Enzyme activity was measured at 340 nm for 5 min on a Tecan Infinity M200 multimode microplate reader. Negative controls were set up without the specific substrate G-3-P and their values were subtracted from the final absorbance. The experiment was repeated thrice in triplicates. The enzymatic activity of wt rGAPDH was taken as 100% and data was plotted as % residual activity ± SD. Statistical analysis was done using Student's t-test.

#### Interaction of GAPDH Mutants with Human Lf

Lf (3 µg), BSA (5 µg), wt rGAPDH (1 µg), rGAPDH (N142S) (1 µg), or rGAPDH (P295L) (1 µg) were resolved on 10% SDS-PAGE and then transferred to nitrocellulose membrane. The membrane was then blocked with 5% BSA and probed with 10 µg/ml of either wt rGAPDH, rGAPDH(N142S) or rGAPDH(P295L) for 2 h, followed by incubation with mouse anti-His (Sigma) (1:3,000) for 1 h and anti-mouse IgG HRP (Sigma) (1:8,000) for 1 h. Finally, blots were washed and developed with TMB/H2O2; the experiment was repeated three times.

#### Analysis of Lf-GAPDH Interaction by Far Western Blotting

Recombinant M.tb H37Rv GAPDH, wild type (wt rGAPDH) was prepared as described previously (Boradia et al., 2016). Lf, BSA (3 µg each) and wt rGAPDH (400 ng) were run on a 10% SDS-PAGE and transferred to nitrocellulose membrane. The membrane was probed with 10 µg/ml of wt rGAPDH for 2 h, followed by incubation with rabbit anti-GAPDH (1:1,000) for 1 h and anti-rabbit IgG HRP (1:16,000) for 1 h; blots were developed with TMB/H2O2, experiments were repeated thrice.

#### Determination of Affinity of Interaction by Solid Phase Binding Assay

The affinity of the GAPDH and Lf interaction was estimated using a plate based solid phase binding assay. Polystyrene wells were coated overnight at 4◦C using 2 µg/well of either wt rGAPDH, rGAPDH(N142S) or rGAPDH(P295L) mutant proteins in PBS (pH 7.4) and blocked using 5% cold fish skin gelatin prepared in PBS for 4 h at room temperature. GAPDH coated wells were then incubated at 25◦C for 2 h with increasing concentrations (7.5−1,000 nM) of biotinylated Lf in PBS containing 1% fish skin gelatin. Control wells were incubated with buffer alone. Extensive washing with PBST was done out to remove any unbound Lf. Bound Lf was detected using primary antibody Mouse anti-Biotin (1:5,000), secondary antibody anti-Mouse IgG HRP (1:20,000) followed by the addition of TMB-H2O<sup>2</sup> for development of color. Data was plotted as absorbance at 650 nm vs. concentration of biotinylated Lf after subtracting OD values of control wells. Experiments were performed thrice (with triplicates) and data was plotted for non-linear fit, one-site specific binding using Graph Pad Prism <sup>R</sup> .

#### Sequence Conservation Analysis of GAPDH

Initial evolutionary conservation sequence analysis to identify functionally and structurally important residues in M.tb GAPDH were performed by the ConSurf Program (Ashkenazy et al., 2010, 2016). The primary sequence of wild type, N142S and P295L mutants as well as M.tb GAPDH modeled structure were considered. For query sequences and structure the CSI-BLAST (Context-Specific Iterated-Basic Local Alignment Search Tool) homolog search algorithm were used with 0.00001 as E-value cutoff, set at a maximal 95% identity and minimal identity of 35% between sequences for homology searches. The protein database UNIREF-90 (http://www.uniprot.org/help/uniref) was utilized; all other parameters were kept at default values for calculation of conservation scores. The primary goal was to reveal the highly conserved regions and functionally and structurally important residues in M.tb GAPDH.

#### Comparative Modeling of GAPDH

GAPDH is a homo-tetramer composed of four identical 36 kDa subunits. The structure of GAPDH alongwith its associated cofactor NAD<sup>+</sup> has been reported for other homologous organisms. The sequence of M.tb H37Rv GAPDH was retrieved from NCBI (GenBank: CAB09248.1), the protein sequence is composed of 339 amino acids. The sequence was analyzed by NCBI protein BLAST search against Protein Data Bank (PDB) in order to get a template with a suitable identity. Tetrameric GAPDH from Bacillus stearothermophilus (PDB ID: 1GD1) was identified as the best template and was utilized for homology modeling of GAPDH using Modeller9v program (Webb and Sali, 2014).

#### Assembly of Monomer Units and Analysis of Mutants

GAPDH homo-tetramer assembled from four identical 36 kDa subunits was done using superimposition tool of Chimera program (Pettersen et al., 2004; Meng et al., 2006). The template (PDB ID: 1GD1) was assembled into its homotetrameric structure, each unit of modeled M.tb GAPDH was then superimposed upon this template. The assembled model was energy minimized to correct the unfavorable bond lengths, bond angles, torsion angles, and contacts. Subsequently, the final model was verified and validated using RAMPAGE Ramachandran plot analysis server (http://mordred.bioc.cam. ac.uk/~rapper/rampage.php) and native protein folding energy assessment by PRoSA program (https://prosa.services.came.sbg. ac.at/prosa.php). The position of the two mutations was analyzed in context of the GAPDH tetramer and its proximity to the NAD<sup>+</sup> binding site.

#### Analysis of Mutants by Computational Methods

Two selected mutations (N142S and P295L) were experimentally tested for the two functions attributed to GAPDH i.e., (i) enzyme activity and (ii) ability to bind Lf. Of these Asparagine 142 is a conserved residue and mutation to serine resulted in a loss of enzyme activity. However, both mutants retained their ability to bind Lf in far-western blots. Both mutants were explored to predict the effect of these alterations on protein stability and proximity to substrate binding site. The effect of mutations on protein stability were analyzed using DUET (Pires et al., 2014), mCSM and Site Directed Mutator (SDM) (Worth et al., 2011).

#### RESULTS

#### Cell Surface Binding of Lf on *M.tb* H37Ra Cells

Flow cytometry clearly demonstrated the binding of Lf to M.tb H37Ra cells which was significantly inhibited (p < 0.0001) in the presence of 100-fold excess of unlabeled Lf, indicative of a specific interaction (**Figures 1A,B**). The binding of Lf-gold particles conjugated particles to M.tb H37Ra was confirmed by TEM analysis (**Figures 1C–E**).

#### GAPDH Mediated Lf-Iron Delivery

M.tb H37Ra strains overexpressing GAPDH-mCherry (GAPDHmCh) or mCherry (mCh) alone were utilized to confirm whether Lf iron is GAPDH dependent. The overexpression of GAPDH in M.tb H37Ra resulted in enhanced binding (**Figures 2A,B**) and uptake of Lf when compared to control cells that over express mCherry alone (p < 0.0005) (**Figures 2C,D**). A correlated increase in iron uptake was also evident in the M.tb GAPDHmCh strain as compared to the mCh strain (**Figure 2E**).

#### Confirmation of GAPDH-Lf Interaction on the *M.tb* Cell Surface

To confirm that GAPDH and Lf interact on the bacterial cell surface we carried out acceptor photobleaching Förster resonance energy transfer (FRET) analysis using M.tb H37Ra cells expressing GAPDH-mCh or mCh alone. This assay is based on the principle of two interacting partners which are fluorescently tagged. The fluorophores are selected such that the emission spectra of the donor overlaps with the absorption

NIS-Elements (Nikon) software. The % intensity change of the Lf-FITC signal was significantly higher in cells expressing GAPDH-mCh as compared to mCh alone (\*\*\*p < 0.0001, Student's t-test, n = 70).

spectra of the acceptor. In the current study if Lf-FITC and GAPDH-mCherry interact, when excitation is done with the wavelength specific for Lf-FITC a strong fluorescence of mCherry is evident. When mCherry is destroyed (by photobleaching) an increase in donor signal is visible because the acceptor is no longer available to parasitize the light emitted from Lf-FITC

expressing mCh alone. (H) Quantitative analysis of FRET signal of 70 regions of interest from multiple, randomly selected fields were analyzed from both test and control samples. The FRET efficiency change ± s.d. of the donor and acceptor signals after photobleaching GAPDH-mCh and mCh was determined using

(Kenworthy, 2001). In the present study a significant increase in donor (Lf-FITC) fluorescence intensity was observed upon photobleaching of acceptor (mCh) in the M.tb H37Ra GAPDHmCh strain (**Figures 2F,H**) as compared to the strain expressing mCh alone (**Figures 2G,H**). The occurrence of FRET between the GAPDH and Lf indicates that they are at proximity of 1–10 nm (Kenworthy, 2001).

#### Cell Surface Binding and Uptake of Lf by *M.tb* H37Rv and *M.tb* H37Rv 1mbtB::hygr Strain

Both strains demonstrated cell surface binding of human Lf at 4◦C that was competitively inhibited by the presence of 200X unlabeled Lf (**Figures 3A,B**). Lf binding was comparable in wild type and siderophore knockout strains (**Figure 3C**). Internalization of labeled Lf was also evident in both strains after incubation at 37◦C (**Figure 3D**) indicating that GAPDH mediated Lf uptake is siderophore independent.

#### Internalization of Lf into *M.tb* H37Rv

Bacilli were analyzed by TEM to assess the internalization of Lfgold conjugated particles, which were observed within the cell cytosol (**Figures 3E–G**).

#### Trafficking of Lf to Intraphagosomal *M.tb*

THP-1 cells were infected with either (i) M.tb H37Ra mCh (ii) M.tb H37Ra GAPDH-mCh or (iii) M.tb H37Rv-GFP strains (**Figures 4A–C** respectively). The trafficking of Lf-A647 to the intraphagosomal bacilli was assessed by confocal microscopy. In all cases the co-localization of Lf-A647 and bacilli was observed, live cell imaging of infected cells confirmed the endocytosis and subsequent co-localization of Lf with intraphagosomal M.tb H37Ra-GFP (**Movie 1**).

#### Expression and Purification of rGAPDH (N142S) and rGAPDH (P295L)

Both proteins were expressed using M.tb H37Ra as a host strain. The proteins were purified using Ni-NTA affinity chromatography and purity confirmed by western blotting with mouse anti-His antibody (**Figures 5A,B**).

#### Enzyme Activity of Mutants and Interaction with Lf

The N142S mutant showed a significant loss of activity (p < 0.0001), only 45% enzyme activity was retained as compared to the wild type enzyme. In contrast the P295L mutant protein demonstrated enzyme activity comparable to wt rGAPDH (**Figure 5C**). The GAPDH-Lf interaction was confirmed in vitro by far western blotting using recombinant M.tb rGAPDH (rGAPDH) (**Figure 5D**). Resolved Lf was able to capture rGAPDH as evident by far western blotting, rGAPDH, and BSA were utilized as positive and negative controls respectively. The N142S, P295L mutant proteins also retained their ability to bind Lf in far western blots (**Figures 5E,F**). ELISA was carried out using 125 nM Lf to obtain quantitative data, no significant difference was observed between the three proteins (**Figure 5G**)

#### *M.tb* GAPDH and Lf Binding Affinity

Lf binding to rGAPDH and mutants was observed to steadily increase with increasing concentration of Lf with saturation at 500 nM (**Figure 5H**). The K<sup>D</sup> value of GAPDH-Lf interaction was estimated to be 31.7 ± 1.68 nM, a 5-fold higher affinity as compared to M.tb GAPDH and Tf (160 ± 24 nM). No significant difference in the affinity values were evident between wt and mutant proteins values. The N142S and P295L mutants were 38.22 ± 3.08 and 37.74 ± 2.63 nM respectively (**Figure 5H**).

#### Sequence Conservation of GAPDH

The M.tb GAPDH sequence was analyzed for its evolutionary conservation, it was observed that key active site residues as well as other surface exposed residues are highly conserved. Since it is known that GAPDH from multiple species binds to Tf (Modun and Williams, 1999), there is a high probability that the conserved regions (including active sites, co-factor binding site, and surface regions) may contribute to the interaction between Lf and Tf. Consurf server based evolutionary conservation analysis was carried out for the wt GAPDH sequence and the two point mutations N142S and P295L (http://consurf.tau.ac.il/ 2016/). Analysis of Asparagine 142 (N142) revealed that it is a highly conserved and exposed functional residue while Proline at 295 position is an exposed but variable residue (**Figure 6A**).

#### Comparative Homology Modeling of GAPDH and Mutant Analysis

The comparative modeling of M.tb GAPDH was performed, using highly identical template of GAPDH from B. stearothermophilus (PDB ID: 1GD1). The two sequences demonstrate a sequence identity of 60%. The modeled 3D structure of GAPDH was validated by the RAMPAGE Ramachandran plot analysis server (http://mordred.bioc.cam.ac.uk/ rapper/rampage.php). The quality of protein folds of GAPDH homology model was evaluated using ProSA protein structure analysis web server (https://prosa.services.came.sbg.ac.at/prosa.php). Modeled 3D structure aligned with the template (PDB ID: 1GD1) and assembled structural model was developed using UCSF Chimera software (https://www.cgl.ucsf.edu/chimera/), the RMSD score was 0.238 (**Figures 6B,C**).

Although, in silico analysis of the mutant N142S suggests that this mutation may result in a non-significant structural change, since Serine at position 142 also shows a similar characteristic to the wild type residues N142, experimentally the mutation resulted in a loss of enzyme activity. Several computational methods have been developed for predicting the stability of mutant proteins based on sequence, structure, and Gibbs free energy estimation, this combined feature based approach provides better prediction of changes due to mutations (Kulshreshtha et al., 2016). In the current study, two different methods were used to predict the stability change of mutated proteins, predictions also assessed the sequence conservation. A combined sequence conservation and protein modeling study indicates Asparagine (N142) is a highly conserved and exposed residue that is located in a flexible loop region of GAPDH. Predicted stability change energies in the mutant (11G)

indicate destabilization (**Table 1**). Structural analysis indicates that in M.tb GAPDH, the Glycine-Valine-Asparagine tripeptide sequence (at positions 140, 141, and 142 respectively) is also highly conserved in several GAPDH homologs. Additionally, it was observed that Asparagine 142 also has hydrogen bonding interactions with another highly conserved tripeptide region comprised of residues 136 to 138 (**Figures 6**, **7C**). This tripeptide is composed of the sequence Threonine136-X-Valine138 where X indicates a variable residue. In M.tb GAPDH the intervening residue is Isoleucine at position 137 (**Figure 6**). Replacement of Asparagine with serine results in a loss of hydrogen bonding with Threonine 136 (**Figure 7D**). Computational modeling of the

FIGURE 4 | Trafficking of Lf to Intraphagosomal bacilli: THP-1 cells were infected with (A) M.tb H37Ra mCh (B) M.tb H37Ra GAPDH-mCh or (C) M.tb H37Rv GFP strains. The trafficking of Lf-A647 to the phagosome and co-localization with the intracellular bacteria are evident.

N142S mutant revealed conformational changes in this region during loop refinement, suggesting that this mutation may alter the stability of the molecule or its interaction with a helix that lies in close proximity (**Figures 7A,B**). While biochemical assays confirm a loss of enzyme activity, detailed analysis by co-crystallization, molecular modeling, extensive dynamic simulation, and mutagenesis would be essential to ascertain the relevance of these residues (if any) in the GAPDH-Lf/Tf interaction. In contrast the Proline 295 residue occurs in a variable loop region adjoining a highly conserved helix, mutation to Leucine does not result in a structurally significant change as confirmed by experimental data (**Figures 7E,F**).

#### DISCUSSION

Due to its intracellular location, M. tuberculosis faces a unique challenge in acquiring iron from within the host cell. It adopts multiple strategies to acquire this essential micronutrient from host iron carrier/storage proteins such as Tf, Lf, ferritin, and hemoglobin. Among these, two well-established processes involve the synthesis of siderophores and hemophores (De Voss et al., 2000; Tullius et al., 2011). Tf and Lf are conserved iron carrier proteins that share a high degree of sequence identity (60%), they are also structurally conserved. Iron acquisition from human Tf and Lf by the siderophores mycobactin and carboxymycobactin is well-characterized. Iron is chelated by the siderophore and transported via specific iron regulated transporters (Ryndak et al., 2010; Banerjee et al., 2011). However, the recombinant strains M.tb H37Rv 1mbt:: hyg and BCG(mbt)30 that do not express siderophores retain the ability to acquire Tf iron in vivo, indicating the presence of a siderophore independent pathway (De Voss et al., 2000; Tullius et al., 2008). Similar studies with Lf have not been evaluated. Although M.tb thrive in the lung, an environment rich in Lf rather than Tf, relatively less is known about Lf mediated iron uptake by M.tb. While previous studies have demonstrated that M.tb acquires iron from Lf in broth cultures and also within host macrophage cells the exact process is yet to be defined.

In pathogens such as S.aureus, S.epidermidis, Neisseria gonorrhea, N.meningitidis, H. influenzae iron acquisition from Tf and Lf can occur by the direct capture of these proteins by

Determination of affinity of interaction (KD) between rGAPDH, rGAPDH (N142S),or rGAPDH (P295L) with Lf.

specific surface receptors (Williams and Griffiths, 1992; Gray-Owen and Schyvers, 1996; Modun and Williams, 1999; Taylor and Heinrichs, 2002). Recently our group identified multiple cell surface receptors for human Tf on the surface of M.tb. These include evolutionarily conserved metabolic enzymes such as GAPDH and LDH (Boradia et al., 2014). GAPDH is a highly

TABLE 1 | Analysis of mutants by computational methods.


∆∆G in Kcal/mol, negative values indicate destabilization and positive values indicate stability of protein structure.

mCSM: Predicts the effect of mutations in proteins using graph-based signatures.

SDM: Site Directed Mutator is a statistical potential energy function method.

DUET: Protein Stability Change upon Mutation methods.

conserved glycolytic enzyme that is present in all organisms. In human macrophages GAPDH functions as a dual receptor for both Tf and Lf (Raje et al., 2007; Rawat et al., 2012), moreover M.tb and human GAPDH are ∼45% identical. Considering all these features we evaluated whether M.tb GAPDH can also function as a dual receptor for Lf. Preliminary studies using M.tb H37Ra cells revealed the specific binding of Lf to the bacterial surface. In vitro experiments confirmed that GAPDH is the receptor involved and overexpression of GAPDH in M.tb H37Ra cells resulted in a significant increase in binding and uptake of Lf as well as incorporation of associated iron. As reported by us in earlier studies involving GAPDH-Tf uptake (Boradia et al., 2014), Lf was internalized into the bacterium. At present the detailed mechanism of this process remains undefined.

Lf binding and uptake was also evident on the surface of virulent M.tb H37Rv and on the M.tb H37Rv1mbtB::hyg strain that is incapable of synthesizing siderophores. These findings are in agreement with previous reports demonstrating the intracellular survival and replication of siderophore negative strains within macrophages and in mice (De Voss et al., 2000; Tullius et al., 2008) indicative of the presence of alternate iron acquisition pathways.

To confirm whether this mechanism operates to deliver Lf to intraphagosomal bacilli, infected THP-1 cells were used as a model system. Confocal microscopy and live cell imaging studies clearly reveal the endocytosis and eventual co-localization of labeled Lf to the bacilli. Previous studies have evaluated the uptake of iron from endogenous and exogenous sources. In the endogenous model macrophages were preloaded with iron from Tf, Lf, or chelates and subsequently infected with M.tb. Iron uptake by the macrophages was comparable, but iron acquisition by M.tb was marginally more in cells preloaded with Lf-iron. In the case of experiments utilizing the exogenous model, cells were first infected with M.tb and then provided exogenous Tf, Lf, or Fe-citrate. When iron uptake by resident intracellular bacilli was evaluated, it was found that M.tb preferentially acquired Lf-iron as compared to Tf or Fe-citrate. Infected macrophages acquired 30-fold more iron from Lf and 3-fold more iron from citrate as compared to Tf. This suggests that if Lf and Tf are present in the external milieu, Lf is vastly more efficient at iron delivery to intra-phagosomal M.tb (Olakanmi et al., 2002, 2004). Our current study revealed that M.tb GAPDH possesses a 5-fold greater affinity for Lf as compared to Tf. Moreover, the Lf based iron uptake in the GAPDH mCh strain was evident within 1 h as opposed to 3 h in the case of Tf-iron (Boradia et al., 2014).

The crystal structure of M.tb GAPDH and structural details of the Lf/Tf-GAPDH interaction are at present unknown. In an effort to identify the key residues that may be involved in this interaction, analysis of two GAPDH mutants was carried out. Mutations of Asparagine at position 142 to Serine (N142S) and a Proline to Leucine alteration at position 295 (P295L) were assessed for their enzyme activity and ability to bind Lf. The N142S mutation demonstrated a drastic loss of enzyme activity as compared to wt GAPDH. The affinity of wild type protein and mutants for Lf were comparable (**Figure 5H**). This finding suggests that the two functions of GAPDH i.e., enzyme activity and Lf binding are independent of each other. Finally, the position of these residues within the GAPDH molecule was evaluated by in silico approaches. The sequence was evaluated for conserved, variable, exposed, or internal residues; a homology model was constructed for both the monomer and the tetrameric enzyme. It was found that N142 is a highly conserved, exposed residue that is present in a flexible loop region of GAPDH. Moreover, in M.tb GAPDH, the Glycine-Valine-Asparagine tripeptide sequence (at positions 140, 141, and 142 respectively) is highly conserved in several GAPDH homologs. Computational analysis suggested that this mutation may alter the stability of the molecule or its interaction with a helix that lies in close proximity (**Figures 7A,B**). Analysis also revealed that the interaction of N142 with conserved residues T136 and V138, hydrogen bonding with T136 was disrupted in the N142S mutant (**Figures 7C,D**), which may contribute to the observed loss of enzyme activity. However, detailed analysis by co-crystallization, molecular modeling, dynamic simulation, and mutagenesis would be necessary to ascertain the exact residues involved in the GAPDH-Lf/Tf interaction.

It is known that internalization of Tf into mammalian cells including macrophages can occur via the three identified receptors TfR1, TfR2 (absent on macrophages), and GAPDH (Jandl et al., 1959; Kawabata et al., 1999; Raje et al., 2007). Cell surface and soluble GAPDH are well-characterized routes for Lf uptake in macrophages (Rawat et al., 2012; Chauhan et al., 2015). Other known Lf receptors are Lipoprotein receptor-related protein (LRP) (Grey et al., 2004) and CD14 (Baveye et al., 2000). In the context of M.tb infection, further studies are essential to explore which of these pathways deliver iron to the phagosome and where the iron carrier molecules are transferred to the bacterial receptors. An alternate possibility is that unidentified, secreted bacterial proteins may intersect the endocytic pathway to capture and recruit Lf to the phagosome.

In conclusion, the current study reveals for the first time that M.tb can efficiently acquire iron from Lf utilizing GAPDH. These findings are relevant since lung fluids are rich in Lf and it is also an essential component of the innate immune response. The use of GAPDH as a single receptor that can acquire iron from both Tf and Lf indicates the ability of the bacilli to adapt to the external milieu. This study also brings into consideration the complex inter-relationship between the host and pathogen wherein GAPDH from both sources play parallel roles in iron acquisition. In siderophore mediated iron uptake, the host synthesizes siderocalins as a protective mechanism to inhibit iron uptake (Goetz et al., 2002; Flo et al., 2004; Berger

showing the replacement of Proline with Leucine.

the host cell. Unraveling the contribution of host and bacterial components in this alternate pathway could offer an insight into iron acquisition by M.tb and perhaps other pathogens.

#### AVAILABILITY OF DATA AND SUPPORTING MATERIAL

The data supporting the research findings are available in the manuscript.

#### AUTHOR CONTRIBUTIONS

CR and MR initiated the project. HM, AP, VB, MR, and CR designed the experiments, analyzed the data, and prepared the manuscript. RKu, RN, and PG designed and analyzed the in silico component of this study. HM, AP, VB, AK, and ZG were involved with reagent preparation, data acquisition, and analysis. RKa, contributed to reagent preparation and standardization.

#### FUNDING

The financial assistance provided by Department of Science and Technology (DST) EMR/2016/001937, National Women Bioscientist Award (2014), Department of Biotechnology (DBT) SAN No.102/IFD/SAN/2599/2016-17, and Council of Scientific Research (CSIR), Govt. of India is sincerely acknowledged.

#### REFERENCES


#### ACKNOWLEDGMENTS

We would like to thank Dr. CE. Barry, Tuberculosis Research Section, Laboratory of Host Defenses, National Institute of Allergy and Infectious Disease, Rockville, Maryland for generously providing us the M.tb H37Rv 1mbt::hyg siderophore knockout strain. Mr. Ranvir Singh is acknowledged for the skillful technical assistance provided for experiments. VB was a recipient of Research Associateship, from Council of Scientific and Industrial Research (CSIR), Government of India. RN, AK, and RKa were recipients of research fellowships provided by NIPER, SAS Nagar. RKu, HM, and AP were recipients of research fellowships from CSIR, University Grants Commission (UGC) and Department of Biotechnology (DBT) respectively. ZG was a recipient of research fellowship from Department of Biotechnology, Government of India.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fcimb. 2017.00245/full#supplementary-material

Movie 1 | Trafficking and co-localization of Lf-A647 with intraphagosomal M.tb H37Ra-GFP.


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

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

## Role of Bacterioferritin & Ferritin in *M. tuberculosis* Pathogenesis and Drug Resistance: A Future Perspective by Interactomic Approach

#### Divakar Sharma\* and Deepa Bisht

*Department of Biochemistry, National JALMA Institute for Leprosy and Other Mycobacterial Diseases, Agra, India*

Tuberculosis is caused by *Mycobacterium tuberculosis*, one of the most successful and deadliest human pathogen. Aminoglycosides resistance leads to emergence of extremely drug resistant strains of *M. tuberculosis*. Iron is crucial for the biological functions of the cells. Iron assimilation, storage and their utilization is not only involved in pathogenesis but also in emergence of drug resistance strains. We previously reported that iron storing proteins (bacterioferritin and ferritin) were found to be overexpressed in aminoglycosides resistant isolates. In this study we performed the STRING analysis of bacterioferritin & ferritin proteins and predicted their interactive partners [ferrochelatase (hemH), Rv1877 (hypothetical protein/probable conserved integral membrane protein), uroporphyrinogen decarboxylase (hemE) trigger factor (tig), transcriptional regulatory protein (MT3948), hypothetical protein (MT1928), glnA3 (glutamine synthetase), molecular chaperone GroEL (groEL1 & hsp65), and hypothetical protein (MT3947)]. We suggested that interactive partners of bacterioferritin and ferritin are directly or indirectly involved in *M. tuberculosis* growth, homeostasis, iron assimilation, virulence, resistance, and stresses.

#### *Edited by:*

*Susu M. Zughaier, Emory University, United States*

#### *Reviewed by:*

*Jiaoyu Deng, Wuhan Institute of Virology (CAS), China Anna Upton, TB Alliance, United States*

> *\*Correspondence: Divakar Sharma divakarsharma88@gmail.com*

*Received: 31 March 2017 Accepted: 24 May 2017 Published: 08 June 2017*

#### *Citation:*

*Sharma D and Bisht D (2017) Role of Bacterioferritin & Ferritin in M. tuberculosis Pathogenesis and Drug Resistance: A Future Perspective by Interactomic Approach. Front. Cell. Infect. Microbiol. 7:240. doi: 10.3389/fcimb.2017.00240* Keywords: bacterioferritin, ferritin, protein-protein interaction, aminoglycosides resistance, *M. tuberculosis*

#### INTRODUCTION

Mycobacterium tuberculosis is the causing factor of tuberculosis (TB) etiology and remains one of the top 10 causes of death worldwide in 2015. Recently WHO reported 10.4 million new TB cases and 1.8 million deaths worldwide (WHO Report, 2016). Due to disappointment of BCG vaccine in adults, chemotherapy through potent anti-TB drugs is the last option to reduce prevalence and mortality but unfortunately, emergence of drug resistant tuberculosis such as multi drug resistant tuberculosis (MDR-TB), extensively drug resistant tuberculosis (XDR-TB) and totally drug resistant tuberculosis (TDR-TB) have made chemotherapy complicated. Aminoglycosides are the second line drugs of choice used especially for the treatment of MDR-TB along with the fluoroquinolones. Aminoglycosides and fluoroquinolones are the only option for the treatment of MDR-TB (with some side effects as well as less efficacy as compared to first line drugs). When the usual treatments are not possible than recently approved drugs (bedaquiline and delamanid) have been used for the treatment of MDR TB, XDR-TB and TDR-TB (with more side effects as compared to first and second line drugs). In M. tuberculosis they typically inhibit protein synthesis by interacting with protein translational machinery. Two dimensional gel electrophoresis coupled with mass spectrometry is the best accepted approach for expression proteomics (Lata et al., 2015a,b; Sharma et al., 2015b, 2016b; Sharma and Bisht, 2016). Since last decade a panel of proteomics and bioinformatics studies related to aminoglycosides resistance have been accumulated (Sharma et al., 2010, 2014, 2015b,a, 2016a,b; Kumar et al., 2013; Sharma and Bisht, 2017a,b). Here we emphasized on the M. tuberculosis iron storage proteins (bacterioferritin and ferritin) and their interactive protein partners which might be involved in pathogenesis, virulence and drug resistance.

Iron is an essential entity for metabolism of the biological cells and hence crucial for the chemistry of life. Iron assimilation, storage and their utilization play a crucial role not only in pathogenesis/pathobiology (growth, survival, virulence and latency) but also in emergence of aminoglycosides drug resistance strains of M. tuberculosis (Reddy et al., 2012; Kumar et al., 2013; Sharma et al., 2015b). Recently, Khare et al. (2017) reported that iron storage proteins are involved in maintaining iron homeostasis in M. tuberculosis (Khare et al., 2017). Bacterioferritin (Rv1876) and ferritin (Rv3841) are unique to maintain iron storage as well as homeostasis in M. tuberculosis. In mycobacteria, various genes products and their interactive partners required for high affinity iron acquisition have been identified such as siderophore production, uptake of ferricsiderophores, production of iron storage proteins and uptake of heme. Production, storage and function of iron uptake mechanisms are controlled by a regulatory protein IdeR (Gold et al., 2001). Heme is the preferable iron source for M. tuberculosis and its acquisition is done by a biosynthetic enzyme ferrochelatase (Rv1485). Rv1485 catalyzes the last step of heme biosynthesis in which iron is interleaved to protoporfhyrin IX to form protoheame (Dailey and Dailey, 2002). It is essential because it supplies the heme which is a preferred iron source for M. tuberculosis (Parish et al., 2005) and serves as a cofactor for various metabolic enzymes such as catalase-peroxidase and DosS/DosT two component system [active sites contains heme] (Svistunenko, 2005).

In pathogenic mycobacteria, usually high affinity systems are essential for maintenance of an infection, virulence and resistance. Greater definition of the functions of both the identified genes and their products, ferritin (bfrB) and bacterioferritin (bfrA) will refine our understanding of mycobacterial iron acquisition and the interplay between components of the iron systems have increased intensities under iron-rich and decreased intensities under iron deprived conditions (Pandey and Rodriguez, 2012). Under iron-rich conditions, M. tuberculosis represses iron acquisition and induces iron storage proteins suggesting the significant role of iron storage proteins in iron homeostasis. M. tuberculosis usually synthesizes two iron storage proteins: ferritin (bfrB) and bacterioferritin (bfrA). BfrB is mandatory to overcome iron limitation and defense against oxidative stress, whereas bfrA is superfluous for victorious adaptation to those stresses. Studies of M. tuberculosis lacking bfrB gene (which encode the ferritin an iron storage protein) reported increased intracellular concentration of iron. They depicted that increasing iron concentration (absence of bacterioferritin and ferritin) decreased resistance to various anti-TB drugs including aminoglycosides as well as fluoroquinolones (Pandey and Rodriguez, 2012) and suggested that ferritin and bacterioferritin are not only mandatory to maintain iron homeostasis but also make M. tuberculosis resistant to aminoglycosides. Kurthkoti et al suggested that iron-dependent regulator IdeR induces ferritin (bfrB) expression by alleviating Lsr2 repression in M. tuberculosis (Kurthkoti et al., 2015). WhiB7 (Rv3197A) is Fe-S cluster-bound protein (transcriptional regulatory protein), which not only associates with aminoglycosides resistance but also predicted to be in the IdeR|Rv2711 regulon (Gold et al., 2001; Morris et al., 2005). Recently Kuberl et al. (2016) suggested that pupylation machinery maintains the iron homeostasis by targeting iron storage proteins (Kuberl et al., 2016).

In our previous studies of expression proteomics we have discovered, that bacterioferritin (Rv1876) and ferritin (Rv3841) were overexpressed in aminoglycosides (amikacin and kanamycin) resistant M. tuberculosis clinical isolates (Kumar et al., 2013; Sharma et al., 2015b). Molecular docking revealed that aminoglycosides drugs (AK and KM) bind to conserved bacterioferritin domain of Rv1876 and ferritin domain of Rv3841 and suggested that overexpression of these proteins might be to neutralize/modulate the drug effect and could be involved in aminoglycosides resistance mechanisms of M. tuberculosis (Kumar et al., 2013; Sharma et al., 2015b). Although the enzymes and their connected pathways involved in iron metabolism in M. tuberculosis are well recognized, still our information related to iron transportation, trafficking, iron dependent post-transcriptional & translational regulations and proteinprotein interactions in mycobacteria is inadequate. Recently Sharma et al. (2016a) reported that inducible overexpression of recombinant ferritin in E. coli (BL21) leads to shift in MIC of AK & KM and suggested their probable roles in conferring aminoglycosides resistance (Sharma et al., 2016a). Consequently, proteins involved in iron storage, assimilation, regulation, uptake and their utilization can be a promising anti-mycobacterial targets against the drug resistant tuberculosis.

#### BACTERIOFERRITIN AND FERRITIN PROTEIN-PROTEIN INTERACTION: UNLOCK THE SECRETS OF IRON RELATED PATHWAYS IN AMINOGLYCOSIDES DRUG RESISTANCE

STRING-10 is an online server which is used to predict the interacting partners of the iron storage proteins [bacterioferritin (Rv1876) and ferritin (Rv3841)] (Sharma et al., 2016a,b; Sharma and Bisht, 2017a). "STRING uses a combination of prediction approaches and an integration of other information (neighborhood, transferred neighborhood, gene fusion, co-occurrence, co-expression, experiments, databases, text mining). Network was made at medium confidence level (0.400) allowing all active prediction methods, which corresponds to approximately 50% possibility of

association" (Sharma et al., 2016a,b). In the network display, each node represents a protein, and each edge represents an interaction.

STRING analysis predicted (**Figure 1**) that ferrochelatase (hemH), Rv1877 (hypothetical protein/probable conserved integral membrane protein), uroporphyrinogen decarboxylase (hemE) trigger factor (tig), transcriptional regulatory protein (MT3948), hypothetical protein (MT1928), glnA3 (glutamine synthetase), molecular chaperone GroEL (groEL1 & hsp65), and hypothetical protein (MT3947) were functional partners of the bacterioferritin and ferritin at medium confidence level (0.400) allowing all active prediction methods. Interactome analysis suggested that molecular chaperone GroEL (groEL1 & hsp65) and trigger factor (tig) were co-expressed as well as experimentally reported. In silico analysis predicted trigger factor (tig) protein as a novel target for interaction with ferritin protein. Trigger factor (tig) is not only involved in eliciting the expression of proteins & their export but also helps in maintaining the open conformation of newly synthesized protein (chaperone activity).

#### PREDICTION OF PUPYLATION SITES

Using the default threshold (medium) with cutoff 2.452; GPS-PUP predicted pupylation sites in bacterioferritin and ferritin proteins which were tabulated in **Table 1**.

TABLE 1 | Predicted / identified pupylation sites in bacterioferritin and ferritin proteins.


*K-Lysine undergoes pupylation.*

#### DISCUSSION

Emergence of extensively-drug resistant tuberculosis (XDR-TB) is the consequence of interrupted treatment of multi-drug resistant tuberculosis (MDR-TB) with second line antitubercular drugs (aminoglycosides and fluoroquinolones). Proteins engage in iron storage, assimilation, regulation, uptake and their utilization could not only be involved in M. tuberculosis pathobiology, growth, virulence and latency but also in aminoglycosides drug resistance and might be potential anti-mycobacterial drug target against the drug resistant tuberculosis (Reddy et al., 2012; Kumar et al., 2013; Sharma et al., 2015b, 2016a; Khare et al., 2017). Pandey and Rodriguez also suggested that ferritin is not only mandatory to maintain iron homeostasis in M. tuberculosis but also ferritin deficient bacilli are more susceptible to killing by antibiotics (Pandey and Rodriguez, 2012). Our previous studies reported that bacterioferritin (Rv1876) and ferritin (Rv3841) were overexpressed in aminoglycosides (amikacin and kanamycin) resistant M. tuberculosis clinical isolates and suggested their involvement in resistance (Kumar et al., 2013; Sharma et al., 2015b). Recently Sharma et al. (2016a) reported that inducible over expression of recombinant ferritin in E. coli (BL21) increased the MIC shift of AK & KM and make the bacteria more resistant against the aminoglycosides drugs (Sharma et al., 2016a). These findings suggested its probable roles in conferring resistance.

Interactome analysis of bacterioferritin (Rv1876) and ferritin (Rv3841) by STRING-10 also suggested that bacterioferritin and ferritin protein interacted with their partners such as ferrochelatase, hypothetical protein (Rv1877), uroporphyrinogen decarboxylase, trigger factor, transcriptional regulatory protein (MT3948), hypothetical protein (MT1928), glutamine synthetase, Molecular chaperones (groEL1 & Hsp65), and hypothetical protein MT3947 which were involved in intermediary metabolism and respiration, cell wall and cell processes, virulence, detoxification, adaptation, conserved hypotheticals and regulatory proteins. Hypothetical protein (Rv1877), a Probable conserved integral membrane protein, ferrochelatase and trigger factor might be potential novel targets predicted by bacterioferritin and ferritin proteinprotein interaction. Rv1877 having 14-transmembrane helixes (TMH), possibly involved in transport of drug across the membrane and could be a potential drug target against the aminoglycosides drug resistance. Rv1877 is the homologous of lfrA and deletion of this gene increased the susceptibility against various antibiotics including the aminoglycosides in mycobacteria (Li et al., 2004). Recently, Mehra et al. (2016) reported that Rv1877 is the part of ABC transporter and predicted its involvement in drug resistance (Mehra et al., 2016). Ferrochelatase involved in heme biosynthetic pathways is a preferred iron source for M. tuberculosis. Trigger factor which not only having chaperone activity but also involved in eliciting the protein expression and their export which could maintain the 3D conformation of newly synthesized proteins and prevents misfolding as well as promotes the refolding of unfolded polypeptides generated under stressed conditions. Bhuwan et al. (2016) reported and validated the STRING in the interaction of RipA with Chaperone MoxR1 (Bhuwan et al., 2016). Sharma et al. (2015b) reported that trigger factor, bacterioferritin and ferritin were overexpressed in aminoglycosides resistant M. tuberculosis clinical isolates

#### REFERENCES


(Sharma et al., 2015b) and suggested that trigger factor might trigger expression of bacterioferritin and ferritin and their interactive partners which could be involved in aminoglycosides drug resistance. We suggested that cumulative effect of iron storage proteins (bacterioferritin and ferritin) and their interactive partners {hypothetical protein (Rv1877), ferrochelatase, trigger factor and others} might be involved in various stress, and aminoglycosides drug resistance. Predicted pupylation sites in bacterioferritin and ferritin also suggested its involvement not only in iron homeostasis but also in aminoglycosides resistance (Kuberl et al., 2016; Sharma et al., 2016a). Further detailed and in-depth investigations of these interactome and pupylome could explore the aminoglycosides resistance and might be used as potential drug targets against this issue.

#### AUTHOR CONTRIBUTIONS

DS design the concept and wrote the manuscript. DS and DB finalized the manuscript.

#### ACKNOWLEDGMENTS

The authors are grateful to Director, NJIL, & OMD for the support. DS is ICMR-PDFs (ICMR, New Delhi).


sensitive clinical isolates of Mycobacterium tuberculosis. Proteome Sci. 8:59. doi: 10.1186/1477-5956-8-59


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

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

digital media

of impactful research

article's readership