# FUNDAMENTALS OF, AND APPLICATIONS BASED ON, QUORUM SENSING AND QUORUM SENSING INTERFERENCE

EDITED BY : Cristina García-Aljaro, Ana Otero and Tom Defoirdt PUBLISHED IN : Frontiers in Microbiology

#### Frontiers eBook Copyright Statement

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

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

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

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

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

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

ISSN 1664-8714 ISBN 978-2-88963-381-4 DOI 10.3389/978-2-88963-381-4

### 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

## FUNDAMENTALS OF, AND APPLICATIONS BASED ON, QUORUM SENSING AND QUORUM SENSING INTERFERENCE

Topic Editors:

Cristina García-Aljaro, University of Barcelona, Spain Ana Otero, University of Santiago de Compostela, Spain Tom Defoirdt, Ghent University, Belgium

### Background

Bacteria use quorum sensing (QS) circuits to coordinate various activities (among which biofilm formation and the expression of virulence factors) based on the presence of signaling molecules. Different families of signal molecules have been identified in Gram positive and Gram negative bacteria (e.g. autoinducer peptides and acyl homoserine lactones). Similarly, different quorum sensing antagonists interfering with these system have been found in nature, promoting a new and promising field of research, quorum sensing interference. One of the most intensively studied applications of quorum sensing interference is its use as an alternative or synergycally with antibiotics to fight (antibiotic-resistant) bacterial pathogens. Many studies have been published claiming quorum sensing inhibitory activity of natural and synthetic compounds. However, after decades of research, several questions regarding the suitability of this approach to fight bacterial pathogens remain unanswered, including the risk that pathogens will develop resistance against quorum quenching. Meanwhile, the interest in quorum sensing has increased considerably, and this has broadened the fields where it can find biotechnological, environmental and industrial applications, such as anti biofouling, steering fermentations, bioremediation and wastewater treatment.

### Goal and scope

The goal of this Research Topic is to broaden the knowledge of the phenotypes regulated by quorum sensing and the advances in quorum sensing interference. Deciphering microorganism language and the different phenotypes regulated by microbial signalling systems is a frontier for the development of new tools for the management of microorganisms to fulfil human needs with a broad application in different areas such as medicine, environmental sciences and industry.

Citation: García-Aljaro, C., Otero, A., Defoirdt, T., eds. (2020). Fundamentals of, and Applications Based on, Quorum Sensing and Quorum Sensing Interference. Lausanne: Frontiers Media SA. doi: 10.3389/978-2-88963-381-4

# Table of Contents

*05 Reducing Quorum Sensing-Mediated Virulence Factor Expression and Biofilm Formation in* Hafnia alvei *by Using the Potential Quorum Sensing Inhibitor L-Carvone*

Tingting Li, Yongchao Mei, Binbin He, Xiaojia Sun and Jianrong Li

*16 Salinity-Mediated Increment in Sulfate Reduction, Biofilm Formation, and Quorum Sensing: A Potential Connection Between Quorum Sensing and Sulfate Reduction?*

Krishnakumar Sivakumar, Giantommaso Scarascia, Noor Zaouri, Tiannyu Wang, Anna H. Kaksonen and Pei-Ying Hong

*29 Novel* N*-Acyl Homoserine Lactone-Degrading Bacteria Isolated From Penicillin-Contaminated Environments and Their Quorum-Quenching Activities*

Hiroyuki Kusada, Yu Zhang, Hideyuki Tamaki, Nobutada Kimura and Yoichi Kamagata


Tian-Nyu Wang, Qing-Tian Guan, Arnab Pain, Anna H. Kaksonen and Pei-Ying Hong


Jun Ling, Runjie Zhu, Pedro Laborda, Tianping Jiang, Yifan Jia, Yangyang Zhao and Fengquan Liu


Shaomin Yan and Guang Wu

*141 Bacterial–Fungal Interactions in the Kelp Endomicrobiota Drive Autoinducer-2 Quorum Sensing*

Anne Tourneroche, Raphaël Lami, Cédric Hubas, Elodie Blanchet, Marine Vallet, Karine Escoubeyrou, Alain Paris and Soizic Prado


Sonia Mion, Benjamin Rémy, Laure Plener, Fabienne Brégeon, Eric Chabrière and David Daudé


Antonio Pedro Ricomini Filho, Rabia Khan, Heidi Aarø Åmdal and Fernanda C. Petersen

*205 Nitric Oxide Enters Quorum Sensing via the H-NOX Signaling Pathway in*  Vibrio parahaemolyticus

Takahiro Ueno, Jonathan T. Fischer and Elizabeth M. Boon


Pai Peng, Mara Baldry, Bengt H. Gless, Martin S. Bojer, Carmen Espinosa-Gongora, Sharmin J. Baig, Paal S. Andersen, Christian A. Olsen and Hanne Ingmer

	- Hema Bhagavathi Sarveswari and Adline Princy Solomon

Luis Esaú López-Jácome, Georgina Garza-Ramos, Melissa Hernández-Durán, Rafael Franco-Cendejas, Daniel Loarca, Daniel Romero-Martínez, Phuong Thi Dong Nguyen, Toshinari Maeda, Bertha González-Pedrajo, Miguel Díaz-Guerrero, Jorge Luis Sánchez-Reyes, Dánae Díaz-Ramírez and Rodolfo García-Contreras

# Reducing Quorum Sensing-Mediated Virulence Factor Expression and Biofilm Formation in Hafnia alvei by Using the Potential Quorum Sensing Inhibitor L-Carvone

#### Tingting Li<sup>1</sup> , Yongchao Mei2,3, Binbin He2,3, Xiaojia Sun2,3 and Jianrong Li2,3 \*

<sup>1</sup> Key Laboratory of Biotechnology and Bioresources Utilization (Dalian Minzu University), Ministry of Education, Dalian, China, <sup>2</sup> College of Food Science and Technology, Bohai University, Jinzhou, China, <sup>3</sup> National & Local Joint Engineering Research Center of Storage, Processing and Safety Control Technology for Fresh Agricultural and Aquatic Products, Jinzhou, China

Edited by:

Tom Defoirdt, Ghent University, Belgium

#### Reviewed by:

Raphaël Lami, Sorbonne Université, France Vipin Chandra Kalia, Konkuk University, South Korea

> \*Correspondence: Jianrong Li lijianrong@bhu.edu.cn

#### Specialty section:

This article was submitted to Infectious Diseases, a section of the journal Frontiers in Microbiology

Received: 07 October 2018 Accepted: 21 December 2018 Published: 09 January 2019

#### Citation:

Li T, Mei Y, He B, Sun X and Li J (2019) Reducing Quorum Sensing-Mediated Virulence Factor Expression and Biofilm Formation in Hafnia alvei by Using the Potential Quorum Sensing Inhibitor L-Carvone. Front. Microbiol. 9:3324. doi: 10.3389/fmicb.2018.03324 Quorum sensing (QS), one of the most remarkable microbiological discoveries, is considered a global gene regulatory mechanism for various traits in bacteria, including virulence and spoilage. Hafnia alvei, an opportunistic pathogen and a dominant psychrophile, uses the lux-type QS system to regulate the production of virulence factors and biofilms, which are harmful to the food industry. Based on the QS interference approach, this study aimed to reveal the efficacy of L-carvone at sublethal concentrations on QS-regulated virulence factors and biofilm formation in H. alvei. QS inhibitory activity was demonstrated by the reduction in swinging motility (61.49%), swarming motility (74.94%), biofilm formation (52.41%) and acyl-homoserine lactone (AHL) production (0.5 µL/mL). Additionally, in silico analysis and RT-qPCR studies for AHL synthase HalI and QS transcriptional regulator HalR revealed a plausible molecular mechanism for QS inhibition by L-carvone. These findings suggest that L-carvone (a main component of spearmint essential oils) could be used as a novel quorum sensing inhibitor to control H. alvei in the food industry.

Keywords: Hafnia alvei, L-carvone, quorum sensing inhibitor, in silico analysis, RT-qPCR

### INTRODUCTION

Hafnia alvei is a Gram-negative, facultatively anaerobic, rod-shaped, motile bacterium of the family Enterobacteriaceae; it is an opportunistic pathogen and a dominant psychrophile found in putrid food (Vivas et al., 2008). It has been widely isolated from different food products, such as raw meat, dairy and aquatic products, and specially from various packed food products stored at low temperatures (Kennedy et al., 2010; Chen et al., 2011; Tan et al., 2014). Based on these characteristics, H. alvei is often considered a specific spoilage organism (SSO) that causes severe nutrition and safety problems in these food matrices by producing extracellular enzymes and siderophores, and forming biofilms. Recent studies have described the key roles of virulence factors

and biofilm production in H. alvei, which is regulated by quorum sensing (QS) systems (Viana et al., 2009; Hou et al., 2017b).

Quorum sensing is a process that allows single-cell organisms (like bacteria) cooperate, communicate, and act collectively. By this process, they can produce, release, detect, and establish connections with small chemical molecules called autoinducers, which in Gram-negative bacteria are acyl-homoserine lactones (AHLs). Thus, AHL-mediated QS systems are usually composed of the LuxI-type autoinducer synthetase, and cytoplasmic LuxR-type proteins, which are receptors activated by AHLs (Ng and Bassler, 2009). Among these two proteins, LuxR-type proteins have more complex functions. With the increase of bacterial density (achieving a certain threshold), the ligand-binding domain of the LuxR-type proteins will bind with AHLs, which will cause changes in the protein conformation and stimulate the formation of AI-LuxR compound proteins, and lead to the binding of the DNA-binding domain of these compound proteins to target genes, thereby regulating the expression of bacterial multiplex phenotypic features such as virulence factors and biofilms (Almeida et al., 2016). Therefore, interfering with AHL-mediated QS systems by using certain compounds, generally called QS inhibitors (QSIs), may be a better strategy to prevent bacterial food spoilage. Compared to antibiotics and antiseptics, QSIs aim to make the bacteria 'surrender' instead of killing them, which would weaken them from having resistance (Defoirdt, 2017).

Nowadays, many synthetic and natural products have been called QSIs; however, only a few of them have a therapeutic value, due to the instability or high toxicity of most other compounds (Defoirdt et al., 2013). Due to the property of low toxicity, some natural compounds from spice plants have been widely used as antimicrobial agents in the food industry, such as curcumin, vanillin, menthol, and cinnamaldehyde (Fitzgerald et al., 2003; Husain et al., 2015; Ding et al., 2017). Therefore, extracting natural compounds from spice plants to obtain effective QSIs has become a promising research hotspot. L-Carvone (or (R4)-(-)-carvone), a monoterpene, is the main component of spearmint essential oils from traditional spice plants and medical herbs. It is widely applied in the food field; it is used to enhance the fragrance and flavor in cooking, and in the beverage and the chewing gum industries (de Carvalho and da Fonseca, 2006). In many studies, L-carvone has been reported as an antimicrobial agent for foodborne pathogenic microorganisms (Friedman et al., 2002; Porfírio et al., 2017); however, there is still limited information about the relationship between spoilage bacteria and QSIs. Therefore, our study involves the characterization of the L-carvone-mediated inhibition of the QS activity of the biosensor strain Chromobacterium violaceum CV026, and subsequently, the determination of the effect of L-carvone on virulence factor and biofilm production in the spoilage bacterium H. alvei. Additionally, we further investigated the underlying mechanism of L-carvone as a potential QSI in H. alvei, by using the in silico analysis and RT-qPCR techniques. In this regard, the study has provided new information about the application of L-carvone as potential QSI and reference values for the effective control of spoilage bacteria.

### MATERIALS AND METHODS

### Reagents, Bacterial Strains, and Growth Conditions

<sup>L</sup>-Carvone (≥99% purity) and AHL standards including C4-HSL, C6-HSL, C8-HSL, C10-HSL, C12-HSL, and C14-HSL were obtained from Sigma-Aldrich (United States). The molecular biology reagents were purchased from Thermo Fisher Scientific (Shanghai, China). Other chemical reagents used in this study were of analytical grade, except for methanol (Chromatographic grade). The bacterial strains used in this study were C. violaceum CV026 and H. alvei Ha-01, as an AHL-reporter organism and a test strain, respectively. C. violaceum CV026 was provided by Dr. Yang (Xinjiang Shihezi University, Xinjiang, China) and H. alvei (ATCC 13337) Ha-01 was originally isolated and identified from putrid turbot by our group. C. violaceum CV026 was a mini-Tn5 mutant derived from C. violaceum ATCC 31532; it was kanamycin-resistant. It could respond only when exogenous AHLs were present, after which it produced the characteristic violet pigment, violacein. Both the strains were overnight cultured in Luria-Bertani (LB) broth (Qingdao Hopebio Co., Ltd., China), at 28◦C and 160 rpm; however, the LB broth culture medium for CV026 required 20 µg/mL kanamycin.

### Antibacterial Assay

### Determination of the Minimum Inhibitory Concentration (MIC) of L-Carvone

The MIC of L-carvone against the selected bacteria was determined using the Oxford cup assay method, as described by Diao et al. (2014). Overnight-cultured (OD<sup>600</sup> = 0.5, 250 µL) C. violaceum CV026 or H. alvei was inoculated in LB nutrient agar (25 mL) and poured into a plate that accommodated two autoclaved Oxford cups, which were removed when the agar solidified. Two hundred microliters of L-carvone (diluted to 2.0, 1.0, 0.5, 0.25, 0.125, and 0.0625 µL/mL using sterile water) were added to the wells, while sterile water served as the control. The plates were incubated at 28◦C for 36 h and the bacterial growth states were observed. The minimum concentration at which there was no visible growth was defined as the MIC. Then, sub-MICs were selected for the further experiments using the above strains.

### Determination of QSI Activity Violacein Inhibitory Activity

The violacein inhibitory activity was determined by adopting the method described by Ia et al. (2012), with slight modifications. Overnight-cultured C. violaceum CV026 (250 µL) was inoculated in LB nutrient agar (25 mL) containing 10 µL of exogenous AHLs (C6-HSL, 2 mg/mL). Afterwards, 200 µL of L-carvone at the sub-MICs was added to each well (diameter, 6 mm) on the plates, while 200 µL of sterilized water was used as the negative control. The plates were incubated at 28◦C for 24 h, and the bacterial growth status was observed. Once no violet pigment was produced around the well, the violacein inhibitory activity was determined.

### Quantitative Analysis of Violacein Production

fmicb-09-03324 January 7, 2019 Time: 17:36 # 3

Violacein produced by C. violaceum CV026 exposed to different concentrations (0.5, 0.25, 0.125, and 0.0625 µL/mL) of L-carvone was quantified as previously described by Choo et al. (2006). Different concentrations of L-carvone (described above) were mixed in 10 mL of LB broth containing 20 µg/mL C6-HSL, along with C. violaceum CV026 overnight cultures, and incubated at 28◦C for 48 h with shaking (160 rpm). At the same time, a similar experiment without C6-HSL was performed, and the OD<sup>595</sup> was measured to determine the effect of the above concentrations of L-carvone on the growth of the CV026.

The violacein pigment was extracted according to the method described by Kumar et al. (2015) with modifications. The cultures in each treatment group were vortexed, and 300 µL of these mixed cultures were taken in 1.5-mL tubes (Eppendorf). They were lysed (for 15 s) using 10% sodium dodecyl sulfate (SDS, 150 µL) at room temperature, and then, extracted (for 5 s) using butyl alcohol (600 µL). Finally, this solution was centrifuged (9,000 g for 5 min); violacein was contained in the organic layer. Then, the OD<sup>595</sup> of each supernatant was measured in a 96-well microtiter plate.

### Assay for Biofilm Formation

The 1.5-mL Eppendorf tubes (polypropylene material) were autoclaved, and the H. alvei overnight cultures (100 µL) were inoculated in 1 mL of LB broth containing various concentrations (0.5, 0.25, 0.125, and 0.0625 µL/mL) of L-carvone. Sterile water or 20 µg/mL C6-HSL was used as the negative control or positive control (absence of L-carvone), respectively. The tubes were statically incubated at 28◦C for 48 h. Then, the determination of biofilm was performed as described previously (Rode et al., 2007), with minor modifications. The cultures were discarded, and each tube was rinsed thrice with sterile water. The tubes were then naturally dried for 40 min and stained with 1 mL of 0.1% crystal violet (w/v) for 15 min at room temperature. After washing with sterile water, the biofilms were extracted using 33% acetic acid. The biofilm solutions were then transferred to a clean 96-well plate, and the OD<sup>595</sup> values were measured using microplate photometers (Bio-Rad, United States).

### Visualization of Biofilms by CLSM and SEM

To pre-form the biofilms, pieces of zinc (6 mm × 6 mm × 0.2 mm) were polished and immersed in LB broth containing sub-MICs of L-carvone or 20 µg/mL of C6-HSL in 90-mm sterile plates (Thermo, United States). Overnight cultures of H. alvei (OD<sup>600</sup> = 0.5, 100 µL) were inoculated in these plates and then statically incubated. After cultivation (at 28◦C for 48 h), a piece of zinc (with an adhered biofilm) was transferred to a clean sterile plate and washed thrice with sterile phosphate buffer saline (PBS, pH 7.4) to remove the planktonic cells. For visualization by confocal laser scanning microscopy (CLSM), this zinc piece was stained with 0.01% (w/v) acridine orange (AO, dissolved in PBS) for 15 min in the dark. Then, the excessive AO was removed by washing with PBS, followed by fixing with antifade mounting medium Fluoromount-GTM (Yeasen, China) for 15 min under the same conditions. Finally, the samples were observed by CLSM (Leica SP5, German) (emission: 525 nm, excitation: 488 nm). For visualization analysis by scanning electron microscopy (SEM), the zinc piece was soaked in 2.5% glutaraldehyde (v/v) at 4◦C for 5 h, dehydrated in graded ethanol (15 min for each grade). Subsequently, the SEM sample was obtained after drying with sterile air.

### Swimming and Swarming Motility Assay

Motility experiments were performed on swimming (1% [w/v] tryptone, 0.5% [w/v] NaCl, and 0.3% [w/v] agar) or swarming (1% [w/v] peptone, 0.5% [w/v] NaCl, 0.5% [w/v] D-fructose, and 0.6% [w/v] agar) agar plates, as previously described (de la Fuente-Núñez et al., 2012), but with some modifications. These agar plates were supplemented with different concentrations (0.5, 0.25, 0.125, and 0.0625 µL/mL) of the L-carvone, before the agar solidified. Then, 5 µL of H. alvei overnight cultures (OD<sup>600</sup> = 0.5) was inoculated at the center of the solidified plates, and the plate was incubated statically at 28◦C for 48 h. The motility of H. alvei was evaluated by measuring the diameter of the swimming and swarming colonies. Plates supplemented with sterile water or 20 µg/mL C6-HSL were used as a negative or positive control, respectively. At least three independent experiments for motility assays were performed.

## AHL Analysis by GC-MS

### AHLs Extraction

Dilutions (1/100) of H. alvei overnight cultures were incubated in LB broth (200 mL) in the presence of L-carvone (0.5, 0.25, 0.125, and 0.0625 µL/mL) for 24 h at 28◦C in an Erlenmeyer flask. Bacterial cells were removed by centrifugation (9,000 g for 15 min). The supernatants were extracted using ethyl acetate supplemented with 0.1% acetic acid thrice, and then, the organic phases were evaporated using a rotary evaporator. The residues were re-dissolved in methanol (1 mL) and filtered through a 0.22-µm membrane (FilterBio, China) for GC-MS detection. For comparison, LB broth in the absence of L-carvone was used. C14-HSL, as an internal standard, was added to each of the AHL samples at a concentration of 5 µg/mL.

### GC-MS Detection

The AHL samples of H. alvei were further analyzed by GC-MS (7890N/5975, Agilent, United States), according to the method described by Zhu et al. (2016). A HP-5 MS capillary column (30 m length × 0.25 mm internal diameter × 0.25 µm film thickness) was used for the chromatographic separation of the AHLs. The injection volume was 1 µL, using a slit ratio of 50:1. The injector temperature was maintained at 200◦C and the oven temperature was automated from 150 to 220◦C at a rate of 10◦C/min, followed by a 5◦C/min increase to 250◦C, and from 250 to 252.5◦C at 0.5◦C/min, with helium as the carrier gas, at a flow rate of 1 mL/min. The mass spectrometry conditions were as follows: the electron ionization source was set to 70 eV, the MS Quad temperature was 150◦C, the emission current was 500 µA, the MS Source temperature was 230◦C. Data were acquired using the full-scan mode (m/z 35–800) and selected ion monitoring (SIM) mode (m/z 143).

### RT-qPCR Analysis

fmicb-09-03324 January 7, 2019 Time: 17:36 # 4

Hafnia alvei was cultured in LB broth with sub-MICs of <sup>L</sup>-carvone at 28◦C for growth until the logarithmic phase. H. alvei cultured without L-carvone was used as a negative control, while H. alvei cultured with C6-HSL was used as a positive control. Total RNA was isolated from H. alvei using TRIzol Reagent (Thermo Scientific, United States), according to the manufacturer's guidelines. The quality of the isolated RNA was checked using standard agarose gel electrophoresis. The single-stranded cDNA was prepared in accordance with the protocol of RevertAid First Strand cDNA Synthesis Kit & DNase I (Thermo Scientific, United States), as stated by the manufacturer. The RT-qPCR experiment was performed by using BIO-RAD CFX ConnectTM Real-Time PCR test system (BIO-RAD, United States) and Power SYBR <sup>R</sup> Green PCR Master Mix (Applied Biosystems, United States). The sequences of primers are listed in **Table 1**. The housekeeping gene 16S rRNA was used as an internal reference. The conditions for RT-qPCR were as follows: initial denaturation at 95◦C for 3 min, and 95◦C for 10 s, 55◦C for 20 s for annealing, 72◦C for 20 s for extension, and 75◦C for 5 s for collecting the fluorescence signal; 40 cycles were run. The melt curve was established in the range of 65–95◦C. The relative expression of the objective genes was calculated by using the 2−11CT method, as previously described by Livak and Schmittgen (2001).

### In silico Analysis

The lux-type protein sequences of H. alvei were HalI (acyl-homoserine-lactone synthase, AAP30849.1) and HalR (transcriptional regulators, AAP30848.1) and downloaded from the NCBI<sup>1</sup> . The models of these proteins were built and assessed the online tools SWISS-MODEL<sup>2</sup> (Bertoni et al., 2017; Bienert et al., 2017; Waterhouse et al., 2018) for docking studies.

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

<sup>2</sup>https://www.swissmodel.expasy.org/

TABLE 1 | List of target genes and their respective primers used for RT-qPCR analysis.


The lux-type genes in H. alvei were halI and halR, respectively.

The water molecules associated with the protein model were removed and the missing hydrogen atoms were supplemented using Clean Protein module of Discovery Studio (DS). The 3D structures of the ligands including L-carvone, halogenated furanone C30 (a known QSI), and C6-HSL were downloaded from the ZINC 12 database<sup>3</sup> and optimized in DS to obtain their possible lowest-energy conformations. The binding spheres that covered the active site residues were also obtained with DS, using the Define and Edit Binding Site module. Finally, docking of the ligands was subsequently performed using the Libdock algorithm.

### Statistical Analysis

Each experiment was performed in triplicate, and the data were presented as the mean values ± SD. The data were analyzed by one-way analysis of variance (ANOVA) along with Tukey test correction using the software SPSS Statistics 20.0. Graphs were constructed using Origin Pro 9.0. Differences with a p-value < 0.05 were considered significant.

## RESULTS

### Minimum Inhibitory Concentration (MIC) of L-Carvone

The MIC of L-carvone, with concentrations ranging from 0.0625 to 2 µL/mL, was estimated by the Oxford cup method. It was observed that the MIC of L-carvone for C. violaceum CV026 was 1.0 µL/mL; L-carvone did not influence the growth of H. alvei at the same concentration. Therefore, the sub-MICs (0.5, 0.25, 0.125, and 0.0625 µL/mL) were selected for further experiments in this study.

### Effect of L-Carvone on Violacein Production in C. violaceum CV026

To determine whether L-carvone at the sub-MICs inhibited violacein production in CV026, two assays were performed. **Figure 1A** shows that a clear inhibitory zone was observed around the well on the purple pigment plate due to L-carvone; however, the control, sterile water, did not inhibit pigment production. Furthermore, the quantitative results of violacein production were obtained. L-Carvone showed a dose-dependent QSI activity and did not significantly inhibit bacterial growth at the sub-MICs (**Figure 1B**). The minimum violacein production rate (OD treatment group/OD control group) was only 48.25% at a 0.5 µL/mL concentration of L-carvone.

### Effect of L-Carvone on Biofilm Formation in H. alvei

The results of biofilm formation after treatment with different concentrations of L-carvone are presented in **Table 2**. A minimum biofilm inhibition of 13.43% was observed when H. alvei was cultured with L-carvone at 0.0625 µL/mL;

<sup>3</sup>http://zinc.docking.org/

TABLE 2 | Inhibitory activity of L-carvone on biofilm formation by H. alvei (mean ± standard deviation).


<sup>a</sup>Expressed as OD<sup>595</sup> after incubation with crystal violet. <sup>b</sup>The inhibitory rate = (OD control group − OD treated group)/OD control group. <sup>c</sup>−gSignificantly different means (P < 0.05).

a maximum biofilm inhibition of 52.41% was observed at am L-carvone concentration of 0.5 µL/mL. In contrast, biofilm formation in the C6-HSL-treated group was visibly higher than that in the control group, which proves that the biofilm formation of H. alvei is positively regulated by the AHL-based QS system.

In this study, the biofilm states of the H. alvei strain in the presence of various concentrations of L-carvone were also observed by CLSM and SEM. The CLSM images showed thick and dense biofilms after C6-HSL treatment, compared with the control group (**Figure 2A**), whereas L-carvone treatment significantly removed the microbes attached to the zinc surface (**Figure 2A**). The SEM images displayed similar results and showed a major disruption to the biofilm architecture as well as the reduction of the biofilm matrix (**Figure 2B**).

### Effect of L-Carvone on Swimming and Swarming Motility of H. alvei

The migration distance of H. alvei grown on swimming and swarming agar plates at 28◦C for 48 h is shown in **Figure 3**. The treatment of H. alvei with sub-MICs of L-carvone reduced the swimming motility significantly; the level of swimming motility inhibition due to L-carvone (0.0625–0.5 µL/mL) was 12.43–61.49%, as depicted in **Supplementary Table S1**. Similarly, swarming migration of H. alvei was also impaired considerably (23.29–74.94%) after treatment with L-carvone (**Supplementary Table S1**). However, the treatment of H. alvei with C6-HSL promoted its motility.

### Effect of L-Carvone on AHL Production in H. alvei

To investigate the effect of L-carvone on AHL production of test strain, the AHLs in the ethyl acetate crude extract of H. alvei were analyzed using GC-MS. After the AHL standards were separated individually, and their retention times were identified (**Supplementary Figure S1A**), we calculated the relative quantity of AHLs in the crude extracts based on the ratio of the peak area of the samples to that of the internal standard (C14-HSL). The AHL types observed in the H. alvei crude extracts were C6-HSL and C8-HSL, at concentrations of 2.16 ± 0.06 and 2.27 ± 0.12 µg/mL, respectively. Treatment with L-carvone significantly reduced the AHL production (**Supplementary Figure S1C**); when treated with 0.5 µL/mL L-carvone, the minimal concentrations of C6-HSL and C8-HSL decreased to 0.16 ± 0.09 and 0.97 ± 0.04 µg/mL, respectively (**Supplementary Figure S1B**).

### RT-qPCR

The RT-qPCR experiments were performed to understand the effect of L-carvone on the expression level of QS-regulated genes in H. alvei. The selected genes were lux-type genes, named halI and halR, respectively. Because in this QS system, the halI gene regulated AHL biosynthesis by encoding HalI (the AHL synthase), the halR gene responded to the corresponding AHL by encoding HalR (the transcriptional regulator), and further regulated the transcription of the downstream genes. The results obtained in this study show that L-carvone could selectively affect the QS system by significantly downregulating the relative expression levels of halI and halR (**Figure 4**). C6-HSL, which was used as the positive control, could significantly upregulate the expression of the selected genes. Melt and amplification curves of the genes were established in **Supplementary Figure S3**.

### Homology Modeling and Model Assessment

At present, the three-dimensional (3D) structures of the HalI and HalR proteins have not yet been analyzed; therefore, homology modeling, based on the online tool SWISS-MODEL, was utilized to solve this problem. Modeling templates were matched using the amino acid sequences of HalI (acyl-homoserine-lactone synthase, AAP30849.1) and HalR (transcriptional regulators, AAP30848.1); the top 50 templates of each protein were obtained. The three best models of HalI and HalI proteins are listed in **Supplementary Table S2**, based on sequence similarities and the GMQE scores.

Model qualities were assessed by using QMEAN, which is a composite estimator and provides both global and local absolute quality estimates for models (Benkert et al., 2011). QMEAN Z-Scores of around zero are an indication of a high quality for a model; however, scores of −4.0 or lower indicated a low quality. Therefore, the results in **Supplementary Figures S2A–D** show that the best models for HalI and HalR proteins were the 1k4j.1.A (score of −1.85) and 5l07.1.B (score of −1.37) models, respectively; they were able to efficiently predict the 3D structures of Lux-type proteins in H. alvei.

### In silico Analysis

In silico analysis studies of L-carvone provide an insight into the binding affinity of this potential QSI, with the model of HalI and HalR protein. For the QS transcriptional receptor HalR protein, the halogenated furanone C30 and C6-HSL were docked as the control ligands. As shown in **Supplementary Table S3**, L-carvone docked with the active site of the HalI

protein of H. alvei, with a LibDock score of 71.0676. Moreover, L-carvone (LibDock score of 66.7963) showed a better affinity toward HalR than the standard QSI, halogenated furanone C30 (LibDock score of 52.7221). However, both the ligands were observed to have a lower affinity toward HalR than the natural ligand C6-HSL (LibDock score of 84.7765). **Figure 4** depicts the possible mechanism of the action of L-carvone in attenuating QS-regulated virulence factor and biofilm production in H. alvei.

### DISCUSSION

There is increasing evidence that plant essential oils can act as potential QSIs, to reduce QS-mediated production of virulence factors and biofilms in microorganisms, and provide a new insight into controlling microbial communities (Zhang et al., 2018). Our data support this notion, revealing a potential QSI, L-carvone (the main component of spearmint essential oil), which interferes with violacein expression in C. violaceum CV026 and enters H. alvei, reducing its motility, biofilm formation, and expression of QS-related genes.

In this study, originally, sub-MICs of L-carvone were tested for their QSI activity using the CV026 strain (**Figure 1**). The biosensor strain CV026 can only respond to exogenous short-chain AHLs through the cytoplasmic transcription factor CviR (a LuxR homolog), which activates the expression of violacein in combination with the AHLs (McClean et al., 1997). Many studies have revealed that the reduction of violacein production without the growth of CV026 being affected is considered a direct evidence for the interference of the QS system (Venkadesaperumal et al., 2016; Liu et al., 2017). Based on the above evidences, this work further explored the QS interference activity of L-carvone on H. alvei, since the QS-mediated production of virulence factors and biofilms plays a key role in the growth of this spoilage organism (Hou et al., 2017b).

The bacterial cells in biofilms are more resistant to antiseptics and food processing conditions; this is likely to cause serious food safety issues (Bai and Rai, 2011). Consequently, studies on preventing biofilm formation are garnering special interest. Previous studies have indicated that the effects of L-carvone on biofilms of Gram-positive and Gram-negative bacteria are possibly different. Soumya et al. (2011) reported that the sub-MICs of L-carvone could reduce biofilm formation in Pseudomonas aeruginosa as a natural QS-inhibitory compound. However, in case of Gram-positive bacteria, the study by Leonard et al. (2010) indicated that carvone could increase biofilm production in Listeria monocytogenes, rather than inhibiting its production. Interestingly, Oliveroverbel et al. (2014) also showed that carvone could inhibit violacein and pyocyanin production in

C. violaceum and P. aeruginosa, respectively, by interfering with their QS systems, and found that this inhibition was produced by its levorotary analog. Herein, for the first time, we have reported that at sub-MICs, L-carvone, a potential natural QSI, could significantly reduce biofilm formation by H. alvei at 28◦C on polypropylene and zinc surfaces (**Table 2**

and **Figure 2**). This result was similar to those of the reports of Soumya and Oliveroverbel. A maximum inhibition of 52.41% was observed using a microplate photometer. Furthermore, in situ analysis of the biofilm matrix performed using SEM and LCSM was able to provide further information on the structure of the formed biofilms following different treatments (Azeredo et al., 2017). As reported by Gross (2017), biofilms were seen as 'Microbial cities' that included both infrastructure (generally embedded in polysaccharide matrixes) and social communication. CLSM and SEM images in our study clearly displayed a major disruption to this infrastructure and the reduction to the biofilm matrix (**Figure 2**). These results were similar to those of the study by Zhou et al. (2018), who found the Hordenine (a sprouting barley extract) could act as a novel QSI and inhibit biofilm formation in P. aeruginosa.

Quorum sensing-regulated flagellar-dependent motility (like swimming and swarming) is closely associated with biofilm formation. In addition, this motility (QS-regulated flagellar-dependent motility) is considered as a virulence factor because of its fundamental role in adhesion, colonization, and virulence expression of pathogens (Atkinson et al., 2006; Bluskadosh et al., 2013). Therefore, a decrease in motility would likely control the biofilm formation of H. alvei and weaken its infection ability. In the present study, treatments with L-carvone dose dependently inhibited the migration capacity of H. alvei. An L-carvone concentration of 0.5 µL/mL showed that the maximum inhibition levels of swimming and swarming motility were 61.49 and 74.94%, respectively. These results are consistent with those from an earlier study by Hou et al. (2017a), who demonstrated a significant inhibition of motility in H. alvei by the food additive dihydrocoumarin.

Due to the essential role of AHLs on the QS system, the effects of L-carvone treatment were characterized using GC-MS, and major changes in the AHL production by the H. alvei strain were observed. GC-MS, with the electron ionization mode, is a powerful tool for the rapid, easy, and selective determination of the AHL levels (Cataldi et al., 2004). The results indicated that L-carvone was able to significantly inhibit the production of both the primary AHLs (C6-HSL and C8-HSL) in H. alvei, especially reducing the C6-HSL production from 2.16 to 0.16 µg/mL. Similarly, Luciardi et al. (2016) found that volatiles from food and medicinal plants could interfere with QS-mediated virulence expression in P. aeruginosa by reducing the biosynthesis of AHLs.

Quorum sensing in Gram-negative bacteria is predominantly controlled by LuxI/R-type proteins, which regulate the production of AHLs, expression of virulence factors, and formation of biofilms (Fuqua et al., 1994). To investigate the inhibitory mechanism of L-carvone on the QS system of H. alvei, relevant protein-molecular interactions were firstly evaluated by in silico analysis. According to the in silico results, we noticed high LibDock scores of the docking of L-carvone with the HalI (LuxI-type protein) and HalR (the LuxR-type protein) of H. alvei (**Supplementary Table S3**). In HalI, Lcarvone was well embedded into a cavity in the vicinity of the active site, the key residues of which included ARG16, SER17, VAL15, ARG31, TRP34, ARG24, LYS26, and LEU23. Simultaneously, L-carvone formed three hydrogen bonds with ARG16, SER17, and VAL15 and showed a hydrophobic behavior with the other residues, as shown in **Figures 5A,a**. In HalR, L-carvone formed one hydrogen bond with SER101 and interacted with other residues (TRP82, VAL69, ALA32, TYR50, TRP54, TYR58, and SER101) via the hydrophobic effect (**Figures 5B,b**). C6-HSL, as a positive control, formed two hydrogen bonds with SER101 and ASP67 (**Figures 5C,c**). However, the standard QSI, halogenated furanone C30, as a negative control, did not form any hydrogen bonds with HalR (**Figures 5D,d**).

The hydrogen-bonding interactions are considered to play a major role in the process where ligands dock with the LuxR-type receptor (Gerdt et al., 2015). In our study, L-carvone showed a better in silico affinity toward HalR than the halogenated furanone C30, because of a higher LibDock score and additional hydrogen bonds. L-carvone and C6-HSL can form hydrogen bonds with the HalR protein at a common site, SER101, indicating a possible competitive action between them. Combined with the GC-MS results, these data confirm that the inhibitory mechanism of L-carvone on the QS system of H. alvei might involve the interaction of L-carvone with the HalI protein and subsequent interference of AHL biosynthesis in H. alvei. In addition, we also characterized the effects of L-carvone treatment using transcriptomics, and observed that the halI and halR genes were significantly downregulated in H. alvei, similar to the results reported in a previous research study (Zhou et al., 2018). The RT-qPCR results were consistent with those of the in silico analysis, which enhanced the credibility of the QS inhibitory mechanism of L-carvone.

### CONCLUSION

In summary, the present study demonstrates that L-carvone had a significant inhibitory activity on the QS system by reducing the AHL-mediated production of virulence factors and biofilm formation in H. alvei. More specifically, Lcarvone combined with the AHL synthase HalI via hydrogen bonds, which led to the disruption of AHL biosynthesis. Understanding the roles and functions of QS in food ecosystems can help in preventing the colonization of food surfaces, toxin formation, and proliferation of food-related bacteria. Therefore, L-carvone, with a QS inhibitory activity, is a promising agent for controlling foodborne pathogens and improving food safety.

### AUTHOR CONTRIBUTIONS

TL and YM contributed to the conception of the study. YM performed the data analyses and wrote the manuscript. BH and XS contributed significantly to analysis and manuscript

preparation. JL helped to perform the analysis with constructive discussions. All authors contributed to manuscript revision, read and approved the submitted version.

### FUNDING

This study was supported by a grant from the National Natural Science Foundation of China (No. 31471639) and the National

### REFERENCES


Key R&D Program of China (Nos. 2018YFD0400601 and 2017YFD0400106).

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2018.03324/full#supplementary-material


cell-to-cell communication as potential antipathogenic drugs. Evid. Based. Compl. Alt. 121, 1–8. doi: 10.1155/2016/7890260


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

Copyright © 2019 Li, Mei, He, Sun and Li. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) 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.

fmicb-09-03324 January 7, 2019 Time: 17:36 # 11

# Salinity-Mediated Increment in Sulfate Reduction, Biofilm Formation, and Quorum Sensing: A Potential Connection Between Quorum Sensing and Sulfate Reduction?

Krishnakumar Sivakumar<sup>1</sup> , Giantommaso Scarascia<sup>1</sup> , Noor Zaouri<sup>1</sup> , Tiannyu Wang<sup>1</sup> , Anna H. Kaksonen<sup>2</sup> and Pei-Ying Hong<sup>1</sup> \*

<sup>1</sup> Water Desalination and Reuse Center, Biological and Environmental Sciences and Engineering Division, King Abdullah University of Science and Technology, Thuwal, Saudi Arabia, <sup>2</sup> Land and Water, Commonwealth Scientific and Industrial Research Organization, Floreat, WA, Australia

#### Edited by:

Ana Maria Otero, University of Santiago de Compostela, Spain

#### Reviewed by:

Aindrila Mukhopadhyay, Lawrence Berkeley National Laboratory (DOE), United States Xiao-Hua Zhang, Ocean University of China, China

> \*Correspondence: Pei-Ying Hong peiying.hong@kaust.edu.sa

#### Specialty section:

This article was submitted to Microbial Physiology and Metabolism, a section of the journal Frontiers in Microbiology

> Received: 19 October 2018 Accepted: 23 January 2019 Published: 06 February 2019

#### Citation:

Sivakumar K, Scarascia G, Zaouri N, Wang T, Kaksonen AH and Hong P-Y (2019) Salinity-Mediated Increment in Sulfate Reduction, Biofilm Formation, and Quorum Sensing: A Potential Connection Between Quorum Sensing and Sulfate Reduction? Front. Microbiol. 10:188. doi: 10.3389/fmicb.2019.00188 Biocorrosion in marine environment is often associated with biofilms of sulfate reducing bacteria (SRB). However, not much information is available on the mechanism underlying exacerbated rates of SRB-mediated biocorrosion under saline conditions. Using Desulfovibrio (D.) vulgaris and Desulfobacterium (Db.) corrodens as model SRBs, the enhancement effects of salinity on sulfate reduction, N-acyl homoserine lactone (AHL) production and biofilm formation by SRBs were demonstrated. Under saline conditions, D. vulgaris and Db. corrodens exhibited significantly higher specific sulfate reduction and specific AHL production rates as well as elevated rates of biofilm formation compared to freshwater medium. Salinity-induced enhancement traits were also confirmed at transcript level through reverse transcription quantitative polymerase chain reaction (RTqPCR) approach, which showed salinity-influenced increase in the expression of genes associated with carbon metabolism, sulfate reduction, biofilm formation and histidine kinase signal transduction. In addition, by deploying quorum sensing (QS) inhibitors, a potential connection between sulfate reduction and AHL production under saline conditions was demonstrated, which is most significant during early stages of sulfate metabolism. The findings collectively revealed the interconnection between QS, sulfate reduction and biofilm formation among SRBs, and implied the potential of deploying quorum quenching approaches to control SRB-based biocorrosion in saline conditions.

Keywords: salinity, biological sulfate reduction, biocorrosion, Desulfovibrio vulgaris, Desulfobacterium corrodens, quorum sensing inhibitors

## INTRODUCTION

Limited availability of freshwater has led to the use of seawater in several industrial applications. High chloride and sulfate content in seawater coupled with biochemical reactions mediated by microorganisms accelerates the rate of biocorrosion in marine environments. Among these microorganisms, sulfate reducing bacteria (SRB) play a crucial role in biocorrosion and biofouling

through biofilm formation, hydrogen sulfide production and extracellular electron transfer (Beech et al., 2005; Kuang et al., 2007; Zhang et al., 2011; Kato, 2016; Scarascia et al., 2016).

Biocorrosion in marine environment has often been associated with SRB biofilms (Beech and Sunner, 2004). In Desulfovibrio vulgaris (an SRB) biofilm-associated cells, upregulation of hydrogenases and cytochrome c533, both of which act as electron conduits, suggest the role of SRB biofilms in microbialinduced corrosion (Pereira et al., 2011; Clark et al., 2012; Scarascia et al., 2016). Recent genomic studies have shown that D. vulgaris biofilm-associated cells often exhibit high levels of gene expression heterogeneity related to exopolysaccharide synthesis, histidine kinases involved in biofilm formation as well as hydrogenases and cytochromes (Zhang et al., 2007; Caffrey et al., 2008; Krumholz et al., 2015; Qi et al., 2016). Earlier studies have also reported on induction of putative formate dehydrogenases and Ech hydrogenases under saline conditions (Mukhopadhyay et al., 2006; Clark et al., 2012). Another study found that high levels of salinity (35 g/L NaCl) did not compromise the metabolic activity of carbon steel-associated SRB biofilms, which in turn exacerbated the rate of biocorrosion (De França et al., 2000). Taken together, it is hypothesized that salinity accelerates biocorrosion by inducing SRB-mediated biofilm formation and sulfate reduction at the gene expression level.

Earlier studies have already established the correlation between biofilm formation and quorum sensing (QS) (Davies et al., 1998; Hammer and Bassler, 2003; Parsek and Greenberg, 2005). It is therefore hypothesized that the increase in SRB biofilm formation and sulfate reduction in the saline environment would be associated with QS mechanisms. Previous studies have reported on the production of QS signal molecules such as N-acyl homoserine lactones (AHLs) [N-hexanoylhomoserine lactone (C6-HSL) to N-dodecanoyl-homoserine lactone (C12-HSL)] by SRB species (Decho et al., 2009, 2010). Compared to other bacterial species such as Vibrio sp. and Pseudomonas sp., relatively little information is available on QS in SRB.

Extensive genomic mining of Desulfovibrio species mainly revealed the presence of proteins homologous to putative QS receptor proteins such as LuxR. However, since synthases were not discovered from genomic mining of SRBs, SRBbased LuxR proteins may be simply orphan receptors and hence, may or may not be involved in QS (Scarascia et al., 2016). Comprehensive genomic analysis of Desulfovibrio species has also revealed the presence of several two-component signal transduction systems, whose exact function in SRB biofilm formation is relatively unknown (Kawaguchi et al., 2008; Decho et al., 2010; Scarascia et al., 2016). It has been speculated that sensor histidine kinases, which dominate these signal transduction systems might be linked with cell–cell communication within SRB biofilms (Zhang et al., 2007; Rajeev et al., 2011). Hence, the exact mechanism of QS in SRBs as well as its linkage to sulfate reduction is largely unknown and it would be interesting to investigate the connection between QS, sulfate reduction and biofilm formation by SRBs under saline conditions.

To explore the connection between QS, sulfate reduction and biofilm formation by SRBs under saline conditions, Desulfovibrio (D.) vulgaris Hildenborough and Desulfobacterium (Db.)corrodens were used as model SRBs in this study. D. vulgaris is a well-studied SRB with its entire genome sequenced and annotated, whereas Db. corrodens is a highly corrosive SRB well suited to iron-rich environments, whose genome has been annotated but with no evidence on the presence of QS-based gene homologs (Bryant et al., 1977; Dinh et al., 2004; Heidelberg et al., 2004; Clark et al., 2007; Gittel et al., 2010). Both species were propagated in either saline or freshwater media in the presence of lactate and Na2SO<sup>4</sup> as electron donor and acceptor, respectively. Enhanced rates of sulfate reduction, AHL production and biofilm formation by D. vulgaris and Db. corrodens were observed under saline conditions. To further understand the influence of salinity on SRB at the gene expression level, we quantified transcript levels of genes related to sulfate reduction, carbon utilization, biofilm formation-based hydrogenases and cytochromes as well as histidine kinases involved in cell–cell communication. The results demonstrated that transcript levels of all selected genes were significantly upregulated under saline conditions. Hence, salinity has a pronounced effect on sulfate reduction, biofilm formation and AHL production at genetic level by both planktonic cells and biofilms of SRB. Further, by deploying QS inhibitors, it was demonstrated that the correlation between QS and sulfate reduction displayed by SRBs is most significant during early stages of sulfate metabolism. The findings suggest that QSI could be deployed as potential biocides to inhibit SRB biofilm-mediated biocorrosion during the early phases of biofilm formation but not on mature SRB biofilm.

### MATERIALS AND METHODS

### Bacterial Strains, Media, and Culture Conditions

Desulfovibrio vulgaris Hildenborough (Heidelberg et al., 2004) and Desulfobacterium corrodens (DSM 15630) were propagated in either saline or freshwater media recommended by Leibniz Institute DSMZ, German Collection of Microorganisms and Cell Cultures. D. vulgaris strain used in this study harbors its 200 kbp native plasmid pDV1, whose presence has been reported to be crucial in its biofilm formation and maintenance (Clark et al., 2007). Saline medium (modified DSMZ medium 141) had the following composition (concentration in g/L) (salinity = 25.9g/L): KCl, 0.34; MgCl2.6H2O, 4.00; NH4Cl, 0.25; CaCl2.2H2O, 0.14; K2HPO4, 0.14; NaCl, 20; yeast extract, 1; tryptone, 1. Dissolved ingredients were initially autoclaved and then supplemented with 5 g/L NaHCO<sup>3</sup> and 10 mL/L of DSMZ-141 vitamin solution (10×) and DSMZ-141 trace element solution (10×) from their respective filter-sterilized stock solutions (Bajracharya et al., 2015, 2017). Freshwater medium (modified DSMZ 641) had the following composition (concentration in g/L) (salinity = 4.17 g/L): MgCl2, 2; K2HPO4, 0.50; NH4Cl, 1; CaCl2, 0.75; yeast extract, 1; tryptone, 1. The freshwater medium was autoclaved and further supplemented with 5 g/L NaHCO3, 10 mL/L of DSMZ-141 vitamin solution

(10×) and 1 mL/L of SL-10 trace element solution (10×) (from DSMZ medium 503) (Kádár et al., 2003; Ünal et al., 2012). Sodium lactate and Na2SO<sup>4</sup> at final respective concentrations of 20 mM (2.24 g/L) and 10 mM (1.42 g/L) were added to both media to serve as electron donor and acceptor, respectively (Bryant et al., 1977; McInerney and Bryant, 1981; Krumholz et al., 2015). The pH of the saline and freshwater media was adjusted to 7.30, and both media were filtered through 0.25 µm syringe filter prior transferring to sterile autoclaved anaerobic tubes. All the tubes were sealed with butyl rubber stoppers and then maintained under anaerobic environment by purging the media with N<sup>2</sup> for 10 min. Further, 0.50 g/L Na2S.9H2O was added to saline and freshwater media inside anaerobic chamber (Coy Laboratory Products Inc., Grass Lake, MI, United States). All cultures were incubated at 30◦C on a rotary shaker.

### Sulfate Analysis

Sulfate concentration in extracted samples (diluted 100×) were quantified using a Dionex ICS-1600 Ion Chromatography system (Dionex Corp., Sunnyvale, CA, United States) equipped with a high-pressure pump, a sample auto-injector, a guard and separator column, a chemical suppressor, a conductivity cell and a data collection system with KOH as the eluent. Data collection and processing were regulated by software Chromeleon 7.0 (Dionex Corp., Sunnyvale, CA, United States) (Altland and Locke, 2012; Uzhel et al., 2016).

### Quantification of Cell Density

Cell density of D. vulgaris and Db. corrodens propagated in saline and freshwater media were measured using Accuri C6 Flow cytometry system (BD Bioscience, Franklin Lakes, NJ, United States) using protocols described earlier (Cheng et al., 2016). Cell pellets harvested through centrifugation (12,000 × g, 15 min) was washed (two times) with 0.9% NaCl. Prior to flow cytometry, diluted cell suspensions (105– 10<sup>6</sup> cells/mL) were stained with SYBR green (Invitrogen AG, Bazel, Switzerland), diluted 10<sup>4</sup> -times from stock concentration (10<sup>4</sup> -fold concentrated in DMSO) (Marie et al., 1997; Noble and Fuhrman, 1998). After staining, cells were incubated at room temperature under dark conditions for 15 min. About 50 µL of stained cells were injected at a flow rate of 35 µL/min to Accuri C6 Flow cytometry system and then excited at 488 nm to enumerate the cell density. In order to evaluate differences in morphological changes between saline and freshwater media, black spots within the flow cytometry gating region were observed. No significant change was observed, which suggested no change in morphology between saline and freshwater media.

### Extraction and Quantification of Total N-Acyl Homoserine Lactones

For extraction of total AHLs, cell-free extracts of both SRB grown in saline and freshwater media were collected by centrifugation (12,000 × g, 15 min). Cell-free extracts were re-concentrated initially through lyophilization and then by resuspending the lyophilized fraction to 1/10th of the initial extracted volume in autoclaved H2O (pH 6.7). Total AHLs were quantitatively determined using a bioassay with beta-glo (Promega, United States) as luminescence substrate (Kawaguchi et al., 2008; Decho et al., 2009). Bioluminescence assay was conducted using a flat white 96-well plate (Greiner Bio-One, Sigma-Aldrich, MI, United States). Briefly, 20 µL of samples were mixed with 80 µL of Agrobacterium (A.) tumefaciens NT1 biosensor prepared in AT medium (Kawaguchi et al., 2008). After incubation for 90 min at 30◦C, 100 µL of Beta-Glo reagent were added into each well of the 96-well plate. After incubation for 30 min at room temperature, bioluminescence intensity of each sample was recorded using a microplate reader (TECAN M200, M200, Männedorf, Switzerland) (Kawaguchi et al., 2008; Decho et al., 2009). Biosensor A. tumefaciens NT1 harbors ß-galactosidase enzyme, whose expression is regulated by the presence of AHLs. Beta-galactosidase cleaves beta-glo substrate to form luciferin in the presence of AHLs, which generates luminescence (Kawaguchi et al., 2008). Each bioassay was conducted in triplicate to assess reproducibility. Different AHLs such as N-butanoyl-homoserine lactone (C4-HSL), N-hexanoylhomoserine lactone (C6-HSL), N-octanoyl-homoserine lactone (C8-HSL), N-decanoyl-homoserine lactone (C10-HSL), N-dodecanoyl-homoserine lactone (C12-HSL), N-tetradecanoyl homoserine lactone (C14-HSL), N-hexadecanoyl homoserine lactone (C16-HSL) and N-octadecanoyl homoserine lactone (C18-HSL) were used to prepare standard curves to optimize the bioluminescence assay (**Supplementary Figure S1A**). Dominant AHLs produced by SRBs were analyzed using liquid chromatography (LC) – mass spectrometry (MS)/MS (Agilent Technologies, Santa Clara, CA, United States) using protocols described elsewhere (Ortori et al., 2011) (**Supplementary Figure S1B**). For LC-MS/MS analysis, a part of cell-free extract was mixed with equal volumes of dichloromethane. AHL extraction procedure was repeated three times, and the organic solvent was evaporated to complete dryness using anhydrous Na2SO4. Dried samples were re-dissolved in methanol prior to analysis (McClean et al., 1997).

### Effect of Salinity on Sulfate Reduction and AHL Production

To elucidate effects of salinity on planktonic cells, D. vulgaris and Db. corrodens were propagated in saline and freshwater media using lactate and Na2SO<sup>4</sup> as electron donor and electron acceptor respectively. Test conditions and media composition were as described earlier. A working volume of 22 mL was maintained in each anaerobic tube. Three biological replicates were used for each test conditions. About 1 mL of culture was extracted from anaerobic tubes every 24 h, and then centrifuged at 12,000 × g for 15 min. Harvested cell pellets were used to enumerate cell density with flow cytometry (Cheng et al., 2016). Cell-free supernatant was used to quantify sulfate concentration (100× diluted) and total AHLs produced, corresponding to each time interval. The effects of salinity were quantified in terms of specific sulfate reduction rate and specific AHL production rate (Fründ and Cohen, 1992; Detmers et al., 2001; Flodgaard et al., 2003; Bruhn et al., 2004). Specific sulfate reduction rate was defined in terms of total amount (µmoles) of sulfate reduced per cell per unit time

(µmoles of sulfate/cell/h), whereas specific AHL production rate was expressed in terms of the total amount (nmoles) of AHLs synthesized per cells per unit time (nmoles of AHLs/cell/h).

### Effect of Salinity on Biofilm Formation

To elucidate effects of salinity on SRB biofilm formation, a static biofilm assay was conducted on D. vulgaris and Db. corrodens biofilms cultivated on a polystyrene flat bottom 96-well plate (Costar, Corning Inc., Corning, NY, United States). A total of eight biological replicates, with three technical replicates per each biological replicate, were used for this study. The biofilms were propagated using both saline and freshwater media. Biofilms were incubated at 30◦C for 144 h within anaerobic chamber (Coy Laboratory Products Inc., Grass Lake, MI, United States). After 8 days, planktonic cells were removed and cells attached to the bottom of wells were washed with sterile 0.9% NaCl. Attached cells were then stained with 100 µL of 1% crystal violet (CV) reagent. After staining, cells were incubated at room temperature for 15 min. Excess CV was removed from each well, which was then air dried. Attached cells were then resuspended in 100 µL of 96% ethanol. Biofilm biomass was quantified in terms of OD<sup>590</sup> using a microplate reader (SpectraMax 340PC384, Molecular Devices, CA, United States).

### Reverse Transcription Quantitative Polymerase Chain Reaction (RT-qPCR)

Reverse transcription quantitative polymerase chain reaction (RT-qPCR)-based approach was selected to quantify the expression of target genes associated with sulfate utilization, carbon and energy metabolism as well as biofilm formation in D. vulgaris. The complete list of selected genes with their annotated functions and primers are listed in **Supplementary Table S1**. A detailed explanation for the selection of these target genes are also provided for as **Supplementary Information 1**. D. vulgaris biofilms were propagated in anaerobic serum bottles (working volume of 120 mL) using both saline and freshwater media, with three biological replicates each. Submerged fed-batch biofilm reactors were used to propagate biofilms on cellulose acetate (CA) coupons (5 mm × 5 mm) fastened together on a sterile 4" 22G needle (Air-Tite, Virginia Beach, VA, United States) and each reactor had three of such networks (five coupons per network) of cellulose acetate coupons. Tests were conducted in three phases, with each phase lasting for 7 days. At the end of each phase, half of the spent medium was replaced with fresh medium. At the end of the final phase, planktonic cells and CA membrane coupon-bound biofilms in the reactor were harvested separately. Briefly, coupons from each reactor were placed in 10 mL 0.9% NaCl and then individually subjected to ultrasonication at 25% amplitude with 2 s pulsating intervals for 3 min using a water-bath sonicator (Q500, Qsonica, Newton, CT, United States) (Cheng et al., 2016). Dispersed biofilm cells as well as freely suspended planktonic cells in the reactor were harvested through centrifugation (8,000 × g, 10 min) and used for extracting RNA after treatment with RNA protect (Qiagen, Hilden, Germany). The cell-RNA protect mixture was incubated for 5 min at room temperature and then centrifuged at 8,000 × g for 15 min. RNA-protected treated cell pellets were then stored at −80◦C until RNA extraction. RNA extraction was performed using RNeasy Mini kits (Qiagen, Hilden, Germany) according to the manufacturer's protocol and RNA concentration was quantified using the Qubit 2.0 fluorometer (Thermo Fisher Scientific, San Jose, CA, United States) (Jumat et al., 2018). 1 µg of RNA extracts from biofilms were used as template for the synthesis of complementary DNA (cDNA) for RT-qPCR based on previously described protocols (Jumat et al., 2018). Target genes were amplified from the D. vulgaris genome using polymerase chain reaction (PCR). PCR products were cloned to pCR2.1 cloning vectors (Thermo Fisher Scientific, San Jose, CA, United States) and then transformed to E. coli TOP10 cells (Thermo Fisher Scientific, San Jose, CA, United States). The plasmids encoding each respective target gene were extracted using Plasmid Miniprep protocol (Promega, Madison, WI, United States). Based on the empirical relationship between plasmid DNA concentration, insert and vector size, the plasmid copy number was calculated. Plasmid DNA were then subjected to successive 10-fold serial dilutions to prepare the standard curve between the threshold cycle (CT) and plasmid DNA copy number. The amplification efficiency and regression coefficient (R 2 ) corresponding to standard curves for each target gene are provided in **Supplementary Table S1**. The volumes of reagents used for RT-qPCR were as follows: Fast SYBR Green Master Mix (Thermo Fisher Scientific, San Jose, CA, United States), 10 µL; forward and reverse primers, 0.4 µL each; cDNA template, 1 µL and PCR grade H2O, 8.2 µL. RT-qPCR was conducted using Applied Biosystems <sup>R</sup> QuantStudio 3 Real-Time PCR system (Thermo Fisher Scientific, San Jose, CA, United States). RT-qPCR cycle also included a melting curve analysis through an increase in temperature from 60 to 95◦C for 5 s at 0.5◦C interval. The copy numbers of each target gene estimated from the RT-qPCR standard curve were normalized against the single copy housekeeping gene Recombinase A recA (DVU1090). recA has displayed lower levels of gene expression heterogeneity in D. vulgaris biofilm growth mode, compared to other internal reference gene such as 16S rRNA (DV16SA) and glyceraldehyde 3-phosphate dehydrogenase (DVU0565), in accordance with previous studies (Zhang et al., 2007; Clark et al., 2012; Qi et al., 2014, 2016). Threshold cycle data values of recA extracted from D. vulgaris planktonic cells and biofilms are shown in **Supplementary Table S2**, and further demonstrated a low level of gene expression heterogeneity in this study.

### Quantification of Extracellular Polysaccharides and Proteins

Extracellular polysaccharides and proteins (EPS) from cellulose acetate membrane coupons-attached biofilms were detached through ultrasonication in 10 mL 0.9% NaCl, as mentioned in the previous section. After ultrasonication, the suspension harboring detached cells and EPS from CA membrane coupons was centrifuged (8,000 × g, 10 min). 0.25 µm syringe-filtered cell-free supernatant was used for quantification of EPS using liquid chromatography with organic carbon detector (LC-OCD) model-8 (DOC-Labor, Germany) equipped with a Toyopearl

size exclusion chromatography column TSK HW50S (Tosoh, Japan) (Dimension: 250 mm × 20 mm, particle size: 20–40 µm). Polysaccharides and proteins from total EPS were fractionated and resolved using organic carbon detector (OCD) and organic nitrogen detector (OND). ChromCALC uni software was used to determine the concentration of each fraction in organic matter below the curve based on the integration of the defined area (Huber et al., 2011; Stewart et al., 2013).

### Evaluation of the Impacts of Quorum Sensing Inhibitors on SRB Planktonic Cells and Biofilms

To gain a better understanding of the role of QS in events leading to biocorrosion, three quorum sensing inhibitors (QSIs), (5Z)-4-bromo-5-(bromoethylene)-3-butyl-Z(5H) furanone (bromofuranone), 3-oxo-D12-N-(2-oxocyclohexyl) dodecanamide (3-oxo-N) and γ-aminobutyric acid (GABA) (Sigma-Aldrich, MI, United States) were added to saline medium harboring D. vulgaris and Db. corrodens planktonic cells at different concentrations. Sub-inhibitory concentrations of QSI were selected from previous studies based on the effect of QSI on growth kinetics and biofilm formation (Hentzer et al., 2003; Smith et al., 2003; Ren and Wood, 2004; Chevrot et al., 2006). QSIs were deployed at sub-inhibitory concentrations (bromofuranone, 40 µM; 3-oxo-N, 20 µM; and GABA, 1 mM) and also at concentrations higher than sub-inhibitory concentrations, as follows: Bromofuranone (µM) – 80, 120, 160; 3-oxo-N (µM) – 40, 80, 120, 160; GABA (mM) – 1, 2, 5, 10, 20, 50. Planktonic SRB cells in saline medium without the addition of any QSI were used as a control in this study. Three biological replicates were maintained for each test condition, in a working volume of 22 mL. The effects of QSI on sulfate reduction, AHL production and specific growth rate were quantified. All concentrations of QSI described above were also used for biofilms to determine their effects on SRB biofilm formation.

### Statistical Analysis

All statistical assays were performed using Data Analysis tool on Microsoft Excel 2017. The degree of correlation in kinetics involving sulfate reduction, AHL production and cell density for all test conditions was measured in terms of Spearman's rank correlation coefficient. The statistical significance tests were performed using two-tailed t-test on Microsoft Excel 2017.

### RESULTS

### Salinity Enhances Biofilm Formation by D. vulgaris and Db. corrodens but Does Not Promote Growth

D. vulgaris exhibited similar specific growth rates under saline and freshwater conditions, respectively (saline, 0.17 ± 0.02/d; freshwater, 0.14 ± 0.02/d; p = 0.15) (**Supplementary Figure S2A**). Similar trend was also observed for Db. corrodens (saline, 0.17 ± 0.03/d; freshwater, 0.16 ± 0.02/d; p = 0.32) (**Supplementary Figure S2B**). However, salinity significantly improved biofilm formation by D. vulgaris and Db. corrodens. Under saline conditions, D. vulgaris produced 1.5-times higher biofilm biomass than freshwater conditions (saline, OD<sup>590</sup> = 1.80 ± 0.48; freshwater, OD<sup>590</sup> = 1.17 ± 0.38; p = 7.65 × 10−<sup>6</sup> ). Similarly, salinity also enhanced biofilm biomass of Db. corrodens by1.6-times (saline, OD<sup>590</sup> = 1.64 ± 0.43; freshwater, OD<sup>590</sup> = 1.03 ± 0.18; p = 2.77 × 10−<sup>6</sup> ). In addition, higher polysaccharide to protein ratio for both D. vulgaris (2.05-folds) (saline, 0.76 ± 0.11 µg/µg; freshwater, 0.37 ± 0.05 µg/µg; p = 0.02) and Db. corrodens (2.0-folds) (saline, 0.56 ± 0.01 µg/µg; freshwater, 0.28 ± 0.06 µg/µg, p = 0.03) biofilms was observed under saline conditions.

### Salinity Enhances Sulfate Reduction by D. vulgaris and Db. corrodens

Under saline conditions, D. vulgaris displayed significantly higher specific sulfate reduction rate (ca. 1.4- to 2.5-times, p < 0.05) during early (24 h), middle (48–72 h) and late exponential phases (96–120 h) as well as stationary phase (144–168 h) (**Figure 1A**) compared to that observed under freshwater conditions. Similarly, high specific sulfate reduction rates (ca. 1.3- to 2.3-times, p < 0.05) were observed for Db. corrodens under saline conditions during exponential phases (**Figure 1B**). However, Db. corrodens exhibited similar specific sulfate reduction rates during stationary phase (144–168 h, p > 0.05) under saline and freshwater conditions (**Figure 1B**).

### Salinity Increases AHL Production by D. vulgaris and Db. corrodens

In saline medium, total AHLs for D. vulgaris ranged from 20 to 27 nM (**Supplementary Figure S2C**) and 6 to 10 nM for Db. corrodens during exponential and stationary phases. This amount is higher, when compared to freshwater conditions (D. vulgaris, 12–14 nM; Db. corrodens, 4–6 nM) (**Supplementary Figures S2C**,**D**). Under saline conditions, D. vulgaris exhibited ca. three to four times higher specific AHL production rate for all growth phases as compared to freshwater conditions (p < 0.05, **Figure 1C**). In the case of Db. corrodens, an increase in specific AHL production rate by ca. 1.5- to 2-times was observed between early to mid-exponential phases in saline medium (p < 0.05) but the significant difference was no longer apparent in the latter growth phases (**Figure 1D**).

### High Correlation Between Sulfate Reduction and AHL Production Under Saline Conditions

A higher correlation between specific sulfate reduction rate and specific AHL production rate was observed for D. vulgaris (R <sup>2</sup> = 0.87; p = 0.01) under saline conditions compared to freshwater conditions (R <sup>2</sup> = 0.75; p = 0.01). Similarly, Db. corrodens exhibited higher correlation between specific sulfate reduction rate and specific AHL production rate in saline medium (R <sup>2</sup> = 0.93; p = 0.03) compared to freshwater medium (R <sup>2</sup> = 0.73; p = 0.01). In addition, both D. vulgaris and Db. corrodens displayed a higher correlation between

sulfate reduction and AHL production during early and midexponential phase (R <sup>2</sup> ≥ 0.79; p < 0.05) compared to late exponential and stationary phases (R <sup>2</sup> ≤ 0.63; p < 0.05) in saline medium.

### RT-qPCR Analysis Reveals an Increase in the Expression Levels of Targeted Genes Under Saline Conditions

RT-qPCR was conducted to quantify the abundance of specific genes related to sulfate reduction, carbon metabolism, hydrogenases and cytochromes, exopolysaccharide synthesis and signal response regulator in D. vulgaris biofilms propagated under saline and freshwater conditions. Key functions of all the respective genes are listed in **Supplementary Table S1**. Compared to freshwater conditions, expression levels of genes related to lactate metabolism like lactate dehydrogenase ldh (2.04-folds; p = 0.03) and pyruvate: ferredoxin oxidoreductase DVU3025 (2.19-folds; p = 0.04) were significantly upregulated under saline conditions (**Figure 2A**). High relative expression of genes involved in pyruvate and formate cycling such as pyruvate formate lyase DVU2272 (4.96-folds; p = 0.02) and formate dehydrogenases DVU0588 (7.71-folds; p = 0.02) was also detected under saline conditions (**Figure 2A**). The expression of all dissimilatory sulfite reductase subunits such as dissimilatory sulfite reductase alpha subunit dsrA (25-folds; p = 0.02), dsrB (1.92-folds; p = 0.05) and dsrC (2.49-folds; p = 0.05) was upregulated in saline conditions in D. vulgaris biofilms (**Figure 2B**). Although, no significant induction was detected for adenosine 5<sup>0</sup> -phosphosulfate reductase aprA and aprB (p > 0.05) (**Figure 2B**), sulfate adenyltransferase Sat, a key player in sulfate reduction was significantly upregulated (3.50-folds; p = 0.03) under saline conditions (**Figure 2B**). High abundance of periplasmic Fe hydrogenase alpha subunit hydA (2.73-folds; p = 7.40 × 10−<sup>5</sup> ), NiFe hydrogenase alpha subunit hynA-1 (5.61-folds; p = 2.20 × 10−<sup>3</sup> ) and NiFeSe hydrogenase hysA-1 (4.12-folds; p = 0.02) as well as Ech hydrogenases echE (14.88-folds; p = 0.03), echF (7.45-folds; p = 0.03) and cytochrome c553 DVU1817 (5.22-folds; p = 6.30 × 10−<sup>3</sup> ) was also observed under saline conditions (**Figure 2C**). In addition, c3-type cytochromes harboring heme groups such as DVU3171 (5.14 folds; p = 7.07 × 10−<sup>3</sup> ), DVU2524 (4.92-folds; p = 8.03 × 10−<sup>3</sup> ) and DVU2809 (8.57-folds; p = 0.04) were also found to be significantly upregulated in the presence of salinity (**Figure 2C**). Lastly, salinity enhanced the expression of DVU0281 (3.58 folds; p = 0.04), which encodes for exopolysaccharide synthesis and DVU3062 (2.47-folds; p = 9.60 × 10−<sup>3</sup> ), a histidine kinase involved in intracellular communication (**Figure 2D**). Apparently, expression of target genes was also found to be upregulated in case of D. vulgaris planktonic cells extracted from the biofilm reactor (**Supplementary Figure S3**).

### Quorum Sensing Inhibitors Decrease Specific Growth Rates and Biofilm Formation of SRB in Saline Media

To further comprehend and establish the linkage between QS and sulfate reduction in SRB, D. vulgaris and Db. corrodens were propagated in saline media in the presence and absence of QSIs. **Figure 3** shows specific growth rate and biofilm biomass of D. vulgaris and Db. corrodens in saline medium in the presence and absence of QSI. Specific growth rate of D. vulgaris decreased

FIGURE 2 | RT-qPCR analysis of relative expression of selected genes related to carbon metabolism, sulfate reduction, electron transfer and biofilm formation in D. vulgaris biofilms under saline and freshwater conditions. (A) Relative expression of carbon metabolism enzymes lactate dehydrogenase ldh, pyruvate formate lyase DVU2272 and pyruvate dehydrogenase DVU3025 in the primary left y-axis, and formate dehydrogenase DUV0588 in the secondary right y-axis. (B) Relative expression of dissimilatory sulfite reductase dsrB, dsrC, adenosine 5<sup>0</sup> -phosphosulfate reductase aprA, aprB and pyrophosphatase ppaC in the primary left y-axis, and dissimilatory sulfite reductase alpha subunit dsrA and sulfate adenylytransferase Sat in the secondary right y-axis. (C) Relative expression of Fe hydrogenase hydA, NiFeSe hydrogenase hysA-1, Ech hydrogenases echE, formate dehydrogenase DVU1817and c3-type cytochromes DVU3171 and DVU2524 in the primary left y-axis, as well as NiFe hydrogenase hynA-1, Ech hydrogenase echF, and c3-type cytochrome DVU2809 in the secondary right y-axis. (D) Relative expression of exopolysaccharide synthesis protein DVU0281 in the primary left y-axis and sensor histidine kinase response regulator DVU3062 in the secondary right y-axis. Relative expression refers to the transcript level of a specific gene normalized with that of reference gene recA. Results are presented as mean ± standard deviation (n = 3). Significant difference: <sup>∗</sup>p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001.

significantly (ca. 1.41 to 2.65-times; p < 0.05) in the presence of bromofuranone ≥ 80 µM compared to the control (**Figure 3A**). The addition of 3-oxo-N ≥ 40 µM (ca. 1.72- to 2.71-times; p < 0.05) (**Figure 3B**) and GABA ≥ 5 mM (ca. 1.43- to 2.08 times; p < 0.05) (**Figure 3C**) resulted in similar decrease of specific growth rates of D. vulgaris. Likewise, at similar inhibitory concentrations, bromofuranone (ca. 2.35-times; p < 0.05), 3 oxo-N (ca. 1.33 to 2.32-times; p < 0.05) and GABA (ca. 1.84 to 2.70-times; p < 0.05) significantly decreased the specific growth rate of Db. corrodens (**Figure 3C**). Biofilm formation by D. vulgaris and Db. corrodens was compromised (p < 0.05) even at bromofuranone ≤ 40 µM, 3-oxo-N ≤ 20 µM and GABA ≤ 2 mM, as illustrated by the sharp decrease in biofilm biomass compared to control (**Figure 3**).

### Quorum Sensing Inhibitors Inhibit Sulfate Reduction by SRBs in Saline Media

The effect of QSI with increasing concentrations on specific sulfate reduction rate during early, middle, late exponential and stationary phases is shown in **Figure 4**. Addition of bromofuranone ≥ 80 µM significantly decreased the specific sulfate reduction rate of D. vulgaris to ca. 0.52- to 0.71-times that of control (p < 0.05) (**Figure 4A** and **Supplementary Table S3**), while specific reduction rate of Db. corrodens decreased to ca. 0.72- to 0.84-times that of control (p < 0.05) during exponential phase (**Figure 4B** and **Supplementary Table S3**). In the presence of 3-oxo-N ≥ 40 µM, the specific sulfate reduction rate of D. vulgaris dropped to ca. 0.62 to 0.78-times of control (p < 0.05) (**Figure 4C** and **Supplementary Table S3**). The same is observed for Db. corrodens in the presence of 3-oxo-N during exponential phase (**Figure 4D** and **Supplementary Table S3**). Similarly, the specific sulfate reduction rate of D. vulgaris was ca. 0.58 to 0.83 times of control (p < 0.05) (**Figure 4E** and **Supplementary Table S3**) and that of Db. corrodens was ca. 0.60 to 0.84-times of control (p < 0.05) (**Figure 4F** and **Supplementary Table S3**) when exposed to GABA ≥ 5 mM. During stationary phase, decrease in specific sulfate reduction rate displayed by QSItreated D. vulgaris and Db. corrodens was marginal (ca. 0.75- to 0.95-times of control; p > 0.05) compared to exponential phase (**Figure 4** and **Supplementary Table S3**).

### Quorum Sensing Inhibitors Inhibit AHL Production by SRBs in Saline Media

Addition of bromofuranone (≥80 µM), 3-oxo-N (≥40 µM), and GABA (≥5 mM) to D. vulgaris reduced the specific AHL production rate to ca. 0.2- to 0.40-times of control during middle and late-exponential phases (p < 0.05) and to ca. <0.25-times of control during stationary phase (p < 0.05) (**Figures 5A,C,E** and **Supplementary Table S4**). Similarly, bromofuranone (≥80 µM), 3-oxo-N (≥40 µM) and GABA (>5 mM) considerably decreased

FIGURE 3 | Quorum sensing inhibitors (QSIs) and the effect on specific growth rate and biofilm formation of D. vulgaris and Db. corrodens in saline media. (A) Effect of bromofuranone on specific growth rate and biofilm formation of D. vulgaris (upper panel) and Db. corrodens (lower panel). (B) Effect of 3-oxo-N on specific growth rate and biofilm formation of D. vulgaris (upper panel) and Db. corrodens (lower panel). (C) Effect of γ-aminobutyric acid (GABA) on specific growth rate and biofilm formation of D. vulgaris (upper panel) and Db. corrodens (lower panel). Bar chart illustrates specific growth rate plot and dotted line scatter plot illustrates biofilm biomass plot. Results are presented as mean ± standard deviation (n = 3). Significant difference in specific growth rate: <sup>∗</sup>p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001.

FIGURE 4 | Quorum sensing inhibitors (QSIs) and the effect on sulfate reduction by D. vulgaris and Db. corrodens in saline media. Specific sulfate reduction rate of QSI-treated D. vulgaris and Db. corrodens in saline medium was normalized with that of control (no QSI added) and plotted along y-axis. (A) Effect of bromofuranone on specific sulfate reduction rate exhibited by D. vulgaris planktonic cells in saline medium. (B) Effect of bromofuranone on specific sulfate reduction rate exhibited by Db. corrodens planktonic cells in saline medium. (C) Effect of 3-oxo-N on specific sulfate reduction rate exhibited by D. vulgaris planktonic cells in saline medium. (D) Effect of 3-oxo-N on specific sulfate reduction rate exhibited by Db. corrodens planktonic cells in saline medium. (E) Effect of γ-aminobutyric acid (GABA) on specific sulfate reduction rate exhibited by D. vulgaris planktonic cells in saline medium. (F) Effect of GABA on specific sulfate reduction rate exhibited by Db. corrodens planktonic cells in saline medium. Results are presented as mean ± standard deviation (n = 3). Early exp. corresponds to early exponential phase; Middle exp. corresponds to middle exponential phase; Late exp. corresponds to late exponential phase.

the specific AHL production rate of Db. corrodens during middle and late exponential phases (ca. <0.50-times of control; p < 0.05) and stationary phase (ca. <0.40-times of control; p < 0.05) (**Figures 5B,D,F** and **Supplementary Table S4**).

Increasing concentrations of QSIs (bromofuranone ≥ 80 µM; 3-oxo-N ≥ 40 µM and GABA ≥ 5 mM) decreased the overall correlation (R 2 ) between specific sulfate reduction rate and specific AHL production rate from 0.87 to 0.57–0.78 for D. vulgaris (p < 0.05) and to 0.54–0.73 for Db. corrodens (p < 0.05). Likewise, QSIs also decreased the correlation between specific sulfate reduction rate and specific AHL production rate during early and mid-exponential phases from 0.79 to a range of 0.27–0.56. This decrease in correlation was more apparent in the exponential phases compared to that observed during stationary phase.

### DISCUSSION

Salinity is a key factor regulating the corrosion potential of a particular matrix. Increasing levels of salinity shifts corrosion potential in negative direction and hence, is often accompanied with increase in corrosion rates (Mansfeld et al., 2002). At the same time, saline environment favors the proliferation of SRBs such as D. vulgaris and Db. corrodens because of their ability to tolerate high salt stress (Lovley and Phillips, 1994; Blessing et al., 2001; Mukhopadhyay et al., 2006). An earlier study reported an increase in SRB cell numbers when salinity was increased from 13 g/L to 35 g/L, and a decline in SRB numbers as salinity increased further from 35 g/L to 80 g/L (hypersaline range). Coincidentally, biocorrosion rate was also highest when salinity was 35 g/L and when SRB were most abundant (De França et al., 2000). However, the earlier study only reported the overall sulfate reduction rates and did not normalize against cell numbers to obtain the specific sulfate reduction rates that would be more indicative of the sulfate reduction activity per cell.

In this study, it was first observed that the biofilm formation by D. vulgaris and Db. corrodens was higher in saline media than in freshwater media even though the specific growth rates of both SRB in both media were similar. It was then observed that the specific sulfate reduction rates were also higher in the saline media than in the freshwater media (**Figures 1A,B**), and that the higher specific sulfate reduction rate was accounted for by a higher expression of sulfate reduction genes in the saline media than in the freshwater media (**Figure 2**).

Although the mechanisms triggering the increased expression of sulfate reduction genes under saline conditions are not known, we infer that certain genes with possible dual roles in salinity tolerance and sulfate reduction were triggered by salinity. For example, oxidoreductases are often reported to be regulated by increasing saline content in media since oxidoreductases either serve as sensors or contribute to bacterial tolerance under saline environments (Bhatt and Weingart, 2008; Pumirat et al., 2010). Based on previous studies, the expression levels of NADHdependent oxidoreductases such as lactate dehydrogenase, formate dehydrogenase, and succinate dehydrogenase were upregulated under saline conditions (Fu et al., 1989; Weerakoon et al., 2009; Pumirat et al., 2014). This corroborates with the finding of this study related to the upregulation of lactate dehydrogenase ldh, formate dehydrogenase DVU0588 and DVU1817 and pyruvate dehydrogenase DVU3025 by D. vulgaris biofilm cells under saline conditions. The increased expression of ldh and DVU3025 under saline conditions can subsequently lead to improved electron flow and overall metabolic activity (Heidelberg et al., 2004; Keller and Wall, 2011). Furthermore, sulfate reductive enzymes have been reported to be highly dependent on carbon metabolism genes (Pereira et al., 2008; Keller and Wall, 2011). This might explain the increase in specific sulfate reduction rates of D. vulgaris in saline media compared to freshwater media.

In addition, the upregulation of Ech hydrogenases as well as c3-type cytochromes likely suggest an improved electron flow within D. vulgaris under saline environment. The induction of formate dehydrogenases and Ech hydrogenases under saline conditions is consistent with that reported by earlier studies (Mukhopadhyay et al., 2006; Clark et al., 2012). It is therefore inferred that salinity elevates the expression of carbon metabolism enzymes and electron transfer machinery within SRB, which in turn leads to enhanced specific sulfate reduction rates as observed in this study and indirectly accelerating rates of SRB-mediated biocorrosion in seawater environments.

Coincidentally, the increase in both biofilm formation and specific sulfate reduction rates in both SRB species were observed along with an increase in the specific AHL production rates (**Figures 1C,D**), suggesting a potential connection between AHL and sulfate reduction. This is especially so during the early and mid-exponential phases, likely when carbon metabolism of SRBs is most active. Previous studies have demonstrated the production of AHLs by SRB (Decho et al., 2009) but their exact role within SRB was not elucidated. Signal molecules extracted from SRB within microbial mats have been implicated to be the driving force for metabolic activities and interspecies interactions within microbial mats (Decho et al., 2009, 2010) but no prior studies have demonstrated the inter-connection between AHL production, biofilm formation and sulfate reduction.

This study has demonstrated a potential link between AHL production, biofilm formation and sulfate reduction among SRBs under saline conditions. To an extent, enhanced expression of biofilm related genes and sulfate reductive enzymes under saline conditions allude toward interconnection between QS and sulfate reduction at transcriptomic level. However, the choice of biofilm related genes considered for this study might be limited to underpin the nature of this interconnection at molecular level. Based on our findings, it could be speculated that addition of AHLs extracted from SRBs might improve the overall specific sulfate reduction rate by D. vulgaris and Db. corrodens. Further studies that monitor the effects of exogenously added AHLs on sulfate reduction, possibly using transcriptomics approaches, could provide a more comprehensive means to establish the nature of interconnection between QS and sulfate reduction among SRBs.

Nevertheless, this study attempts to further verify the connection between AHL and sulfate reduction by applying QSI at varying concentrations. A decrease in specific sulfate reduction

rates was observed during the exponential phase of SRB (**Figure 4** and **Supplementary Table S3**). This reduction in specific sulfate reduction rate was also accompanied by a considerable decline in both specific AHL production rate (**Figure 5** and **Supplementary Table S4**) and biofilm formation (**Figure 3**). The effect imposed by QSI was however not apparent during the late exponential or stationary phase. The findings collectively suggest that QS pathway could contribute in enhancing specific sulfate reduction rates of metabolically active SRB that propagate in saline environment. However, the exact pathways modulated by the QS mechanisms remain unknown.

Previous studies have reported the use of natural or synthetic compounds in saline conditions to quench QS in bacteria such as Vibrio (V.) harveyi, V. vulnificus, Halomonas pacific and complex microbial community attached to reverse osmosis membrane (Shen et al., 2006; Liaqat et al., 2014; Mai et al., 2015; Santhakumari et al., 2016). In those instances, the use of QSI demonstrated strong inhibition on biofilm formation. However, those studies did not evaluate if QSI approaches would be suitable to inhibit SRBs, and the associated sulfate reduction and biocorrosion rates. This is likely due to the lack of understanding on whether QS is indeed present among SRBs (Scarascia et al., 2016) and if present, whether there is a correlation with specific sulfate reduction rates. This study demonstrated that SRB biofilms are highly susceptible to QSI application, and a consequential decrease in specific sulfate reduction rates can indeed be achieved. Hence, QSI could be deployed as potential biocides to inhibit SRB biofilm-mediated biocorrosion during the early phases of biofilm formation. The efficacy of QSI however may be low on mature SRBs and biofilm.

### CONCLUSION

In summary, by using D. vulgaris and Db. corrodens as model SRBs, we showed that saline conditions significantly increase the rates of specific sulfate reduction, AHL production and biofilm formation by D. vulgaris and Db. corrodens. By deploying QSIs, a potential connection between sulfate reduction and AHL production under saline conditions was demonstrated, which is most significant during early stages of sulfate metabolism. Insights from this study revealed the interconnection between QS, sulfate reduction and biofilm formation among SRBs. Furthermore, this study showed quorum quenching molecules could be deployed as an environmentally benign approach to control SRB at the early stages of growth and biofilm formation.

### AUTHOR CONTRIBUTIONS

KS designed and performed the experiments, data analysis and wrote the manuscript. GS contributed to extraction and analysis of total AHLs and sulfate reduction. TW developed the protocol for quantification and analysis of total AHLs using

bioluminescence assay and LC-MS. NZ conducted quantification of extracellular polysaccharides and proteins. AK provided advice for cultivating sulfate reducing bacteria and comments on the manuscript. P-YH conceived and designed the experiments, analysis and interpretation of data, wrote the manuscript, supervised the research, and provided reagents and materials.

### FUNDING

The research reported in this publication was supported by CRG funding URF/1/2982-01-01 from King Abdullah University of Science and Technology (KAUST) awarded to P-YH.

### REFERENCES


### ACKNOWLEDGMENTS

AK thanks KAUST and CSIRO Land and Water for financial support. We would also like to thank Xiang Zhao in Bioscience Core Lab, King Abdullah University of Science and Technology for technical assistance in conducting RT-qPCR.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2019.00188/full#supplementary-material

protein filaments. Environ. Microbiol. 9, 2844–2854. doi: 10.1111/j.1462-2920. 2007.01398.x



molecules using LC-MS/MS. Anal. Bioanal. Chem. 399, 839–850. doi: 10.1007/ s00216-010-4341-0



**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 © 2019 Sivakumar, Scarascia, Zaouri, Wang, Kaksonen and Hong. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Novel N-Acyl Homoserine Lactone-Degrading Bacteria Isolated From Penicillin-Contaminated Environments and Their Quorum-Quenching Activities

Hiroyuki Kusada<sup>1</sup>† , Yu Zhang1,2† , Hideyuki Tamaki1,3 \*, Nobutada Kimura<sup>1</sup> and Yoichi Kamagata<sup>1</sup> \*

<sup>1</sup> Bioproduction Research Institute, National Institute of Advanced Industrial Science and Technology, Tsukuba, Japan, <sup>2</sup> State Key Laboratory of Environmental Aquatic Chemistry, Research Center for Eco-Environmental Sciences, University of Chinese Academy of Sciences, Chinese Academy of Sciences, Beijing, China, <sup>3</sup> JST ERATO Nomura Microbial Community Control Project, University of Tsukuba, Tsukuba, Japan

#### Edited by:

Tom Defoirdt, Ghent University, Belgium

#### Reviewed by:

Alan W. Decho, University of South Carolina, United States Mikael Elias, University of Minnesota Twin Cities, United States

#### \*Correspondence:

Hideyuki Tamaki tamaki-hideyuki@aist.go.jp Yoichi Kamagata y.kamagata@aist.go.jp †These authors have contributed equally to this work

#### Specialty section:

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

Received: 23 October 2018 Accepted: 20 February 2019 Published: 14 March 2019

#### Citation:

Kusada H, Zhang Y, Tamaki H, Kimura N and Kamagata Y (2019) Novel N-Acyl Homoserine Lactone-Degrading Bacteria Isolated From Penicillin-Contaminated Environments and Their Quorum-Quenching Activities. Front. Microbiol. 10:455. doi: 10.3389/fmicb.2019.00455 N-Acyl homoserine lactones (AHLs) are signaling molecules used in the quorum sensing (QS) of Gram-negative bacteria. Some bacteria interfere with the QS system using AHL-inactivating enzymes, commonly known as quorum-quenching (QQ) enzymes. We have recently isolated a new QQ bacterium showing high resistance to multiple β-lactam antibiotics, and its QQ enzyme (MacQ) confers β-lactam antibiotic resistance and exhibits QQ activities. This observation suggests the possibility of isolating novel QQ bacteria from β-lactam antibiotic-resistant bacteria. In this direction, we attempted to isolate penicillin G (PENG)-resistant bacteria from penicillin-contaminated river sediments and activated sludge treating penicillin-containing wastewater and characterize their QQ activities. Of 19 PENG-resistant isolates, six isolates showed high QQ activity toward a broad range of AHLs, including AHLs with 3-oxo substituents. Five of the six AHL-degraders showed AHL-acylase activity and hydrolyzed the amide bond of AHLs, whereas the remaining one strain did not show AHL-acylase activity, suggesting that this isolate may likely possess alternative degradation mechanism such as AHL-lactonase activity hydrolyzing the lactone ring of AHLs. The 16S rRNA gene sequence analysis results categorized these six AHL-degrading isolates into at least five genera, namely, Sphingomonas (Alphaproteobacteria), Diaphorobacter (Betaproteobacteria), Acidovorax (Betaproteobacteria), Stenotrophomonas (Gammaproteobacteria), and Mycobacterium (Actinobacteria); of these, Mycobacterium sp. M1 has never been known as QQ bacteria. Moreover, multiple β-lactam antibiotics showed high minimum inhibitory concentrations (MICs) when tested against all of isolates. These results strongly demonstrate that a wide variety of β-lactam antibiotic-resistant bacteria possess QQ activities. Although the genetic and enzymatic elements are yet unclear, this study may infer the functional and evolutionary correlation between β-lactam antibiotic resistance and QQ activities.

Keywords: β-lactam antibiotic resistance, quorum sensing, quorum quenching, AHL-acylase, AHL-lactonase

## INTRODUCTION

fmicb-10-00455 March 12, 2019 Time: 19:13 # 2

Bacteria communicate with one another using chemical signaling molecules. The sensing of auto-inducers allows bacteria to distinguish between low and high cell population densities as well as to adjust the gene expression in response to changes in cell number. This process, termed as quorum sensing (QS), allows bacterial cells to coordinately control the gene expression in the community. N-Acyl homoserine lactone (AHL)-dependent QS has been known to regulate many bacterial behaviors such as virulence (Givskov et al., 1998; Lindum et al., 1998; Burr et al., 2006; Xu et al., 2006) and biofilm formation (Hammer and Bassler, 2003; Labbate et al., 2004).

Over the past decade, several studies have been directed to understand the phenomenon of quorum quenching (QQ), a signal interference process that attenuates QS systems. QQ is a typical characteristic of a variety of organisms that degrade AHLs by enzymatic reactions (Dong and Zhang, 2005). Microbial communities harbor counter constituents and exhibit mechanisms as one of their survival strategies that hinder or compete with QS bacteria. Phylogenetically diverse AHL-inactivating bacteria that belong to the phyla Proteobacteria (genera Acidovorax, Acinetobacter, Achromobacter, Alcaligenes, Alteromonas, Agrobacterium, Bosea, Brevundimonas, Comamonas, Delftia, Diaphorobacter, Klebsiella, Mesorhizobium, Ochrobactrum, Pseudomonas, Ralstonia, Roseomonas, Shewanella, Sphingomonas, Stenotrophomonas, and Variovorax), Bacteroidetes (Chryseobacterium, Flaviramulus, and Tenacibaculum), and Cyanobacteria (Nostoc) have been isolated and characterized as QQ bacteria (Leadbetter and Greenberg, 2000; Flagan et al., 2003; Hu et al., 2003; Huang et al., 2003; Lin et al., 2003; Park et al., 2003; Uroz et al., 2003; Sio et al., 2006; Yoon et al., 2006; Romero et al., 2008, 2010; Chan et al., 2011; Christiaen et al., 2011; Mahmoudi et al., 2011; Chen et al., 2012; Wang et al., 2012; Zhang et al., 2013; Torres et al., 2016; Kusada et al., 2017). Furthermore, Gram-positive bacteria within the phyla, Actinobacteria (Arthrobacter, Microbacterium, Nocardioides, Rhodococcus, Staphylococcus, and Streptomyces), Deinococcus-Thermus (Deinococcus), and Firmicutes (Bacillus and Solibacillus) have been found to exhibit QQ activities, indicating that phylogenetically diverse bacteria may quench the AHL-based QS (Dong et al., 2000; Lee et al., 2002; Park et al., 2003; Uroz et al., 2003; d'Angelo-Picard et al., 2005; Wang et al., 2010; Morohoshi et al., 2012; Koch et al., 2014; Chan et al., 2015).

To date, two different types of QQ enzymes have been identified, namely, AHL-lactonase and AHL-acylase. In the earliest study, AHL-lactonase gene (aiiA) from Bacillus species strain 240B1 was cloned and shown to encode the lactonase enzyme that hydrolyzes the ester bond of the lactone ring to produce acyl homoserine (**Figure 1A**) (Dong et al., 2000). On the other hand, Variovorax paradoxus strain was found to degrade AHLs by an acylase, wherein the amide bond between the homoserine lactone (HSL) ring and the acyl chain was cleaved to release HSL and fatty acid (**Figure 1B**) (Leadbetter and Greenberg, 2000). Database retrieval by Kalia et al. (2011) for homologs of the characterized AHL-lactonases and -acylases in complete bacterial genomes have shown that the relatives of these enzymes are widespread in a diverse array of organisms, suggestive of the ubiquity of QQ systems in natural microbial communities (Kalia et al., 2011).

We have recently isolated a novel AHL-degrading bacterium, Acidovorax sp. MR-S7, that exhibits high resistance to multiple β-lactam antibiotics (Kusada et al., 2017). Our study revealed a novel AHL-acylase (MacQ) from MR-S7 that confers β-lactam antibiotic resistance and exhibits QQ activity (Kusada et al., 2017). Therefore, we hypothesize that functionally novel and hitherto-unidentified QQ bacteria may be present among multiple β-lactam antibiotic resistant bacteria. However, very little is known about organisms that exhibit both QQ activities and β-lactam antibiotic resistance. Such microorganisms are very rare, and only two strains, Pseudomonas aeruginosa PAO1 and Acidovorax sp. MR-S7, have been reported so far. Here, we report the characterization of the QQ activity and antibiotic resistance of multiple newly isolated β-lactam antibiotic-resistant bacteria and discuss their phylogenetic relevance with other known QQ bacteria.

### MATERIALS AND METHODS

### Sample Description

The environmental samples were obtained from the wastewater treatment plants in the penicillin G (PENG) production facility of the North China Pharmaceutical Group Corporation (NCPGC) and the receiving river, Wangyang River in Hebei Province, China. The wastewater from the PENG production factory is discharged into the Wangyang River after biological treatment (activated sludge treatment system), including anaerobic treatment, hydrolyzation and acidification, primary aerobic treatment, and secondary aerobic treatment. The average hydraulic residence time for each unit is about 30 h. The annual output of the excess sludge from the wastewater treatment plant (WWTP), which has been in operation since the 1990s, is 1,200 tons (dry weight). The wastewater in this plant contains 80% of wastewater from PENG production and 20% wastewater from the production of other antibiotics (cefalexin [CEFL], cefadroxil [CEFD], ampicillin [AMP], and amoxicillin [AMO]). The detected concentrations of PENG for activated sludge, river water, and sediment samples were 0.076 mg/kg, 0.000031 mg/L, and no detection, respectively. These low concentrations of PENG after the treatment clearly indicate the wastewater from PENG production could be properly treated by the activated sludge treatment system, suggesting that PENG-degrading bacteria would be present in the activated sludge sample.

The activated sludge and river sediment samples were obtained from the sludge concentration tanks that collect sludge samples from all of the biological treatment reactors and Wangyang River near the WWTP discharge outlet, respectively. The samples were collected in brown glass bottles that had been successively washed with tap water, ultra-pure water, and hexane and stored at 4◦C in the dark.

### Isolation of Antibiotic-Resistant Bacteria Using Gellan Gum Medium

In this study, gellan gum-based media were used for the cultivation and isolation of antibiotic-resistant and AHL-degrading bacteria from the environmental samples, as gellan was found to be more effective than agar for the culturing of a diverse array of uncultured microorganisms (Tamaki et al., 2005, 2009). Activated sludge and sediment samples (500 µL) were suspended in sterile water and subjected to 10-fold serial dilutions. A series of medium plates (R2A-gellan gum) supplemented with 15, 30, 50, and 100 µg/mL of PENG were inoculated with 100-µL aliquots from different dilutions and incubated at 20◦C for 4 weeks in the dark under aerobic conditions. Individual PENG-resistant colonies were purified thrice using fresh medium plates supplemented with 100 µg/mL of PENG and stored in 20% glycerol at −80◦C.

The composition of R2A was as follows (per liter): 0.5 g each of yeast extract, peptone, acid hydrolysate of casein, glucose, and soluble starch; 0.3 g each of dipotassium phosphate and sodium pyruvate; and 0.05 g of magnesium sulfate. The pH values of these media were adjusted to 7.0 with 10 mM potassium phosphate buffer. The media were solidified with gellan gum (Wako, Tokyo, Japan) at a final concentration of 1.0%.

### Identification and Phylogenic Analysis of Antibiotic-Resistant Isolates and AHL-Degrading Bacterial Strains

Phylogenetic identification of the antibiotic-resistant isolates was performed using the 16S rRNA gene sequencing analysis. DNA templates for polymerase chain reaction (PCR) amplification from isolates were extracted by FastDNA <sup>R</sup> Spin Kit (MP Biomedicals, Illkirch, France). The 16S rRNA genes of the isolates were PCR-amplified from the colonies using the primers 27F (5<sup>0</sup> -AGATTTGATCCTGGCTCAG-3<sup>0</sup> ) and 1492R (50 -GGTTACCTTGTTACGACTT-3<sup>0</sup> ). The PCR conditions included denaturation at 95◦C for 9 min, followed by 40 cycles at 95◦C for 1 min, 50◦C for 1 min, and 72◦C for 2 min. The final extension was performed at 72◦C for 10 min. PCR products were purified with a MicroSpin S-400 HR column and used as templates for sequencing. Sequencing was performed with the primer 907R (5<sup>0</sup> -CCGTCAATTCMTTTGAGTTT-3<sup>0</sup> ), a DTCS-Quick Start kit (Beckman Coulter, Fullerton, CA, United States), and a CEQ-2000 automated sequence analyzer (Beckman). The sequences of the resistant bacterial 16S rRNA gene clones with a range of about 500–600 bases were determined. All the 16S rRNA gene sequences of the antibiotic-resistant isolates were compared with those in the GenBank database<sup>1</sup> using the BLAST program (Altschul et al., 1990). For AHL-degrading bacteria, almost full 16S rRNA gene sequences (approximately 1,500 bp) were determined using primers 27F, 530F (5<sup>0</sup> -GTGCCAGCMGCCGCGG-3<sup>0</sup> ), 907R, 1100F (5<sup>0</sup> -AAGTCCCGCAACGAGCGCA-3<sup>0</sup> ), and 1492R. Multiple alignments of the 16S rRNA gene sequences of PENG-resistant isolates capable of degrading AHLs were performed with previously known QQ bacteria. The phylogenetic tree was constructed by neighbor-joining method using MEGA software (Tamura et al., 2011). Bootstrap values were estimated using neighbor-joining and maximum-likelihood methods (each 1,000 replications).

### Bioassay of AHL-Degrading Activity

The tested AHL compounds included non-substituted C6-HSL, C8-HSL, C10-HSL, C12-HSL, and C14-HSL as well as substituted 3-oxo-C6-HSL, 3-oxo-C8-HSL, 3-oxo-C10-HSL, 3-oxo-C12-HSL, and 3-oxo-C14-HSL. AHL degradation assay was performed using AHL-detectable reporter (biosensor) strains, Escherichia coli JB525-MT102 (pJBA132) and P. putida F117 (pKR-C12) (Andersen et al., 2001; Steidle et al., 2001). Briefly, exogenous permeable AHL molecules bind to a LuxR-type response

<sup>1</sup>www.ncbi.nlm.nih.gov/BLAST

regulator protein and constitute AHL-LuxR complex within biosensor strains. The AHL-LuxR complex binds to the promoter region of a GFP reporter gene. Therefore, the degradation of AHLs by the sample could be characterized with a decrease or extinction in GFP fluorescence.

For the whole-cell assay to determine the AHL-degrading ability, 3-day cultures of the isolates were washed and re-suspended in 100 mM potassium phosphate buffer (pH 6.5). A total volume of 50 µL of the cell re-suspension and an equal volume of AHL (final concentration, 20 µM) were mixed and the mixture was incubated at 30◦C in the dark with gentle agitation. The samples were treated with ultraviolet irradiation for 1 h to stop the reaction, and the reaction mixtures were diluted to an appropriate concentration and loaded into the wells of a 96-well microtiter plate. The biosensor strains were added into each well and the response of the biosensors after 4 h of incubation was analyzed with a SPECTRAmax <sup>R</sup> GEMINI XS Microplate Spectrofluorometer (Molecular Devices, Sunnyvale, CA, United States). Experiments were performed in triplicates.

### Identification of AHL Degradation Metabolites

The activity of AHL-acylase was demonstrated with high-performance liquid chromatography (HPLC) analysis of the reaction mixtures containing chemically derivatized HSL rings using DANSYL chloride (5-dimethylamino-1-naphthalene sulfonyl chloride), as previously described (Lin et al., 2003). In brief, full-grown cultures of each isolate were mixed with 3 mM C10-HSL and incubated at 30◦C for 3 h. The digestion mixtures were extracted thrice with equal volumes of ethyl acetate, and the extracted organic phases were evaporated to dryness. The samples were re-dissolved in 200 µL of methanol, and the resulting 100 µL solutions were reacted with an equal volume of DANSYL chloride (Tokyo Chemical Industry Co., Ltd., Tokyo, Japan, 2.5 mg/mL in acetone) at 40◦C for 4 h. After evaporation to dryness, 50 µL of 0.2 M HCl was added to the sample to hydrolyze any excess DANSYL chloride. For HPLC analysis, the samples were introduced onto a Develosil ODS-UG-3 column (4.6 × 150 mm, Nomura Chemicals, Aichi, Japan). Fractions were isocratically eluted with 50:50 methanol–water (v/v) at a flow rate of 0.5 mL/min (Shimadzu SPD-6AV UV-VIS spectrophotometric detector, Shimadzu C-R6A chromatopac and Shimadzu SCL-6B system controller). A control experiment was performed with phosphate-buffered saline (PBS) instead of the strain solution. Quorum-quenching bacterium Acidovorax sp. strain MR-S7 known to possess AHL-acylase activity (Kusada et al., 2017) and 2 mM HSL standard (Sigma) were used as positive controls.

### Bioassay for AHL Production Activities

We performed the AHL production assay using GFP-based biosensor strain. In brief, the overnight culture fluids of isolates were extracted with equal volumes of ethyl acetate. The resulting liquid extractions were dispensed into the wells of a 96-well microtiter plate (Becton Dickinson, Franklin Lakes, NJ, United States) and treated with 50 µL of a fivefold-diluted overnight culture of the biosensor strain. The plate was statically incubated at 30◦C for 4 h to induce detectable GFP expression from the reporter cell. E. coli strain (non-AHL producer) and C10-HSL solution were used as the negative and positive control, respectively. Experiments were performed in triplicate for each strain.

### Antibiotic Susceptibility Assay

Minimum inhibitory concentrations (MICs) of AHL-degrading bacteria were determined using microtiter plate dilution assays in R2A broth with about 1 × 10<sup>5</sup> cells/well, as previously described (Yajko et al., 1987; Riesenfeld et al., 2004). MICs were determined after 1, 2, and 3 days of incubation at 30◦C in the dark. MICs were read using a Multiskan <sup>R</sup> Spectrum microplate spectrophotometer (Thermo Labsystems, Vantaa, Finland) and was defined as the lowest concentration of an antimicrobial agent at which the organism showed no visible growth (Yajko et al., 1987). E. coli strain EPI300TM (Epicentre, Madison, WI, United States) was used as a negative control. The criteria used for the interpretation of antimicrobial susceptibility were based upon the achievable levels of antimicrobial agents. The tested antibiotics were PENG, AMP, AMO, carbenicillin (CAR), piperacillin (PIP), CEFL, and CEFD. All these antibiotics are in common use, and the environmental samples used in the present study were polluted with wastewater from PENG, AMP, AMO, CEFL, and CEFD production. The antibiotics were tested at concentrations of 8, 16, 32, 64, 125, 250, and 500 µg/mL.

### Chemicals

N-Hexanoyl-L-homoserine lactone (C6-HSL), N-octanoyl-L-homoserine lactone (C8-HSL), N-decanoyl-L-homoserine lactone (C10-HSL), N-dodecanoyl-L-homoserine lactone (C12-HSL), N-(3-oxo-hexanoyl)-L-homoserine lactone (3-oxo-C6-HSL), N-(3-oxo-octanoyl)-L-homoserine lactone (3-oxo-C8-HSL), and HSL standards were purchased from Sigma. N-(3-oxo-decanoyl)-L-homoserine lactone (3-oxo-C10-HSL), N-(3-oxo-dodecanoyl)-L-homoserine lactone (3-oxo-C12-HSL), and N-(3-oxo-tetradecanoyl)-L-homoserine lactone (3-oxo-C14-HSL) were obtained from the Nottingham University in England.

### Nucleotide Sequence Accession Numbers

The nucleotide sequences reported in this study were deposited in the GenBank database with accession numbers from AB646301 to AB646320.

### RESULTS AND DISCUSSION

### Isolation of Antibiotic-Resistant Bacteria From Penicillin-Contaminated Environmental Samples

To isolate phylogenetically diverse PENG-resistant bacteria from the environmental samples polluted with wastewater from PENG production, we used gellan gum-solidified media supplemented

TABLE 1 | Phylogenetic affiliations of microbes grown on PENG amended medium on the basis of 16S rRNA gene sequences by using the BLAST program in the GenBank database.


<sup>a</sup> S: Activated sludge; M: River sediment. As for AHL-degrading bacteria, the almost full 16S rRNA gene sequences (≈1,500 bp) were determined.

with antibiotics (8, 16, 32, 64, 125, 250, and 500 µg/mL), as gellan was shown to be effective for cultivating phylogenetically novel and diverse bacteria (Tamaki et al., 2005, 2009). Ten and nine PENG-resistant individual colonies with characteristic morphologies and colors were isolated from the activated sludge and river sediment samples, respectively. The results of the 16S rRNA gene sequencing showed that these PENG-resistant isolates belong to at least 12 different genera across Alphaproteobacteria, Betaproteobacteria, Gammaproteobacteria, Actinobacteria, and Bacteroidetes. Of these, three isolates (S5, M9, and S15) showed low 16S rRNA gene sequence similarities (<97%) to any known bacterial species (**Table 1**).

### Exploration of Novel AHL-Degrading Bacteria Among the PENG-Resistant Isolates

The PENG-resistant isolates were tested for QQ activity using GFP-based AHL biosensors. Based on the preliminary screening using biosensors toward 3-oxo-C6-HSL and 3-oxo-C12-HSL, six of the PENG-resistant strains exhibited AHL-degrading activity, as observed with biosensors toward 3-oxo-C12-HSL after a 15 h incubation period (**Supplementary Table 1**). Thereafter, we selected the six resistant strains harboring high AHL-degrading ability (more than 50% degradation of the initial AHL) from the preliminary screening to further investigate QQ behaviors. Six isolates showed high capabilities of inactivating a broad range of AHLs, including AHLs with 3-oxo substituents. In particular, these strains exhibited higher QQ activities toward AHLs with long acyl chains than those with short acyl chains (**Table 2**).

Phylogenetic analysis based on almost full-length 16S rRNA gene sequences showed that the six QQ isolates were associated with three Gram-negative taxa, Alphaproteobacteria (Sphingomonas sp. S1), Betaproteobacteria (Acidovorax sp.

#### TABLE 2 | AHL-degrading behaviors of PENG resistant isolates<sup>a</sup> .


<sup>a</sup>Biosensors for AHLs: E. coli JB525-MT102 (pJBA132) was used as s sensor strain for detecting C6-HSL, C8-HSL, C10-HSL, C12-HSL, 3-oxo-C6-HSL (OC6), 3-oxo-C8-HSL (OC8), 3-oxo-C10-HSL (OC10), 3-oxo-C12-HSL (OC12), and P. putida F117 (pKR-C12) was used for detecting 3-oxo-C14-HSL (OC14). The amount of remaining AHL was measured using bioassay after cultivation in a medium containing 20 µM AHL and biosensors for 4 h. The number of plus symbol relates to the amount of AHL degraded: +, 20–50% degradation of initial AHL; ++, 50–80%; +++, 80–100%.

with boldface.

M2, Acidovorax sp. M6 and Diaphorobacter sp. S2), and Gammaproteobacteria (Stenotrophomonas sp. S17) and one Gram-positive taxon, Actinobacteria (Mycobacterium sp. M1) (**Figure 2**). Although phylogenetically diverse QQ bacteria have been isolated so far, a Gram-positive bacterium within the genus Mycobacterium has never been shown to exhibit QQ activity.

To determine whether the six AHL degraders produce their own AHL isomers, a GFP-based biosensor strain was used (see method). The ethyl acetate extracts of six isolates as well as the negative control (non-AHL producer E. coli) showed low values of GFP fluorescence (**Supplementary Figure 1**). This result indicates that these six isolates lack the ability to produce biosensor-detectable AHL-like compounds. Although these six isolates were unable to produce AHLs, they might have the ability to sense exogenous AHLs produced by other bacteria, and the resulting QS system would enhance biofilm formation and antibiotic resistance. Indeed, our previous study demonstrated that a wide variety of exogenous AHLs induced biofilm formation of non-AHL producer, Acidovorax sp. strain MR-S7, and the MICs of AHL-supplemented (biofilm-forming) strain MR-S7 showed 5–10 folds higher resistance to various antibiotics (Kusada et al., 2014). Most human infections are localized within biofilms, where QS signaling is much more efficient due to localization of AHLs. The genetic information of these isolates (e.g., luxR gene encoding the AHL-responsive transcriptional regulator) and biochemical experiments provided in the future study would further verify the effect of QS on biofilm formation and antibiotic resistance.

### Identification of AHL Degradation Products

To demonstrate the AHL degradation mechanism of these six isolates, HPLC analysis was performed to detect the presence of HSL ring generated by AHL-acylase activities. To determine whether HSL was released as an AHL degradation product, samples of the reaction mixture were treated with DANSYL chloride and analyzed by HPLC. The LC retention time of DANSYL chloride was 4.19 min, whereas that of the dansylated digestion mixtures from isolates (S1, S2, S17, M2, and M6) was around 6.0 min, identical to the retention time of the standard control of dansylated HSL and the dansylated digestion mixture from Acidovorax sp. MR-S7

known to degrade AHLs by AHL-acylase activity (**Figure 3** and **Supplementary Figure 2**). These results indicate that the five isolates were capable of degrading C10-HSL with their AHL-acylase activities. Note that the remaining one isolate, strain M1, was able to degrade a wide range of AHLs but did not show AHL-acylase activity in the HPLC assay. This likely suggests that the strain M1 may possess the distinct degradation mechanism such as AHL-lactonase activity, though further investigation would be needed to reveal their QQ activities using purified enzymes.

β-Lactam antibiotics and AHLs are structurally similar as both compounds have ring structures and acyl side chains. In recent years, AHL-acylase enzymes were found to be also phylogenetically and structurally similar to β-lactam antibiotic resistance enzymes, β-lactam acylases. Besides, our group and other two research groups reported that AHL-acylases (MacQ, AhlM, and KcPGA) could function as β-lactam acylase, and indeed degrade PENG via hydrolysis of the amide bond (Park et al., 2005; Mukherji et al., 2014; Kusada et al., 2017). Furthermore, we recently solved the X-ray crystal structure of MacQ, and found that the degradation products of C10-HSL and PENG by MacQ were commonly accommodated in the same hydrophobic active-site pocket, indicating that both compounds were hydrolyzed by MacQ in the same degradation mechanism (Yasutake et al., 2017). Perhaps, some AHL-acylases (β-lactam acylases) may have broad substrate specificity that appears to be structurally indistinguishable from both AHL-signals and β-lactam antibiotics. Further genetic and biochemical analyses of other AHL-acylases (β-lactam acylases) will be required to more fully assess the mechanism and relationships between QQ and antibiotic resistance.

### β-Lactam Antibiotic Resistance Assay

We measured the MICs of seven different β-lactam antibiotics, including PENG, AMP, AMO, CAR, PIP, CEFL, and CEFD, toward the six AHL-degrading isolates (**Table 3**). In comparison with the control strain E. coli EPI300TM, all of the AHL-degrading isolates displayed a high resistance profile to almost all β-lactam antibiotics examined. In addition, the comparison of the antibiotic resistance activities of the six isolates with those of the previously identified multiple β-lactam antibiotic resistant pathogens revealed that our isolates displayed comparable or even greater values of MICs to all of the seven β-lactam antibiotics tested (John and McNeill, 1980; Colom et al., 1995; Stapleton et al., 1995; Lauretti et al., 1999; Afzal-Shah et al., 2001; Dubois et al., 2002). In particular, strains S17 and M2 displayed high



<sup>a</sup> Penicillin G (PENG); ampicillin (AMP); amoxicillin (AMO); carbenicillin (CAR); piperacillin (PIP); cefalexin (CEFL); cefadroxil (CEFD). The MIC (minimum inhibitory concentration) was defined as the lowest concentration of antibiotics at which the organism showed no visible growth.

abilities to resist broad types of β-lactam antibiotics. The MICs for these strains were over 500 µg/mL of PENG, AMP, AMO, CAR, PIP, CEFL, and CEFD. These values were at least 31.3 to 62.5-fold higher than the values observed for the control strain E. coli EPI300TM (<8–16 µg/mL) (**Table 3**). Furthermore, the MICs of AMP, PIP, and CAR were over 30-fold higher for the six AHL-degrading isolates than for the control strain E. coli EPI300TM.

In total, six AHL-degrading isolates exhibited high antibiotic resistance and broad substrate specificities toward multiple β-lactams as well as AHL isomers. The knowledge relevant to such organisms possessing both multiple β-lactam antibiotic resistance and AHL degradation activities has been limited. Two multiple β-lactam antibiotic resistant bacteria, P. aeruginosa PAO1 and Acidovorax sp. MR-S7, have been reported to be able to degrade AHLs. In comparison to P. aeruginosa PAO1, our six isolates showed much higher resistance to β-lactams tested (MICs of AMP, CAR, and PIP toward strain PAO1 were 256, 64, and 4 µg/mL, respectively). Given that these six isolates degrading multiple β-lactams and AHLs were found across two phyla and four classes, our findings implicate a possibility that this versatile phenotypic bi-functionality may be more broadly present in bacteria dwelling in natural ecosystems and provide new insight into the diversity of organisms that simultaneously exhibit both multiple β-lactam antibiotic resistance and QQ activities. To further clarify the functional and evolutionary correlation between β-lactam antibiotic resistance and QQ activities, studies are warranted to identify the genetic elements responsible for conferring multiple β-lactam antibiotic resistance and QQ activities.

### CONCLUSION

In this study, we successfully isolated six novel QQ bacteria from β-lactam antibiotic-resistant isolates obtained from

### REFERENCES


PENG-polluted environmental samples and characterized their AHL- and β-lactam antibiotic-degrading properties. This study expands the diversity of organisms that possess two different and important physiological functions, QQ activity and β-lactam antibiotic resistance. In addition, our study provides a novel screening strategy for the identification of AHL-degrading and β-lactam antibiotic-resistant bacteria previously unidentified. Taken together with previous studies, our findings provide additional evidence that AHL-degrading bacteria may be the potential multiple β-lactam antibiotic resistant candidates that have long been overlooked.

### AUTHOR CONTRIBUTIONS

YZ, HT, NK, and YK conceived the study and designed experiments. HK, YZ, NK, and HT performed experiments and analyzed the data. HK, HT, NK, and YK wrote the manuscript. All authors contributed to the discussion of the results obtained in this study, and reviewed and edited the manuscript.

### FUNDING

This work was supported by the Ministry of Education, Culture, Sports, Science and Technology with a JSPS fellowship. This work was also partly supported by JST ERATO Grant Number JPMJER1502, Japan.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2019.00455/full#supplementary-material




bacterium. Int. J. Syst. Evol. Microbiol. 56(Pt 8), 1783–1787. doi: 10.1099/ijs.0. 64120-0

Zhang, Y., Tang, K., Shi, X., and Zhang, X. H. (2013). Flaviramulus ichthyoenteri sp. nov., an N-acylhomoserine lactone-degrading bacterium isolated from the intestine of a flounder (Paralichthys olivaceus), and emended descriptions of the genus Flaviramulus and Flaviramulus basaltis. Int. J. Syst. Evol. Microbiol. 63(Pt 12), 4477–4483. doi: 10.1099/ijs.0.053744-0

**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 © 2019 Kusada, Zhang, Tamaki, Kimura and Kamagata. This is an openaccess article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Signal Disruption Leads to Changes in Bacterial Community Population

Michael Schwab1,2, Celine Bergonzi1,2, Jonathan Sakkos<sup>3</sup> , Christopher Staley2,4 , Qian Zhang2,5, Michael J. Sadowsky2,5,6, Alptekin Aksan2,3 and Mikael Elias1,2 \*

<sup>1</sup> Department of Biochemistry, Molecular Biology and Biophysics, University of Minnesota, Twin Cities, St. Paul, MN, United States, <sup>2</sup> Biotechnology Institute, University of Minnesota, Twin Cities, St. Paul, MN, United States, <sup>3</sup> Department of Mechanical Engineering, University of Minnesota, Twin Cities, St. Paul, MN, United States, <sup>4</sup> Department of Surgery, University of Minnesota, Twin Cities, St. Paul, MN, United States, <sup>5</sup> Department of Soil, Water, and Climate, University of Minnesota, Twin Cities, St. Paul, MN, United States, <sup>6</sup> Department of Plant and Microbial Biology, University of Minnesota, Twin Cities, St. Paul, MN, United States

The disruption of bacterial signaling (quorum quenching) has been proven to be an innovative approach to influence the behavior of bacteria. In particular, lactonase enzymes that are capable of hydrolyzing the N-acyl homoserine lactone (AHL) molecules used by numerous bacteria, were reported to inhibit biofilm formation, including those of freshwater microbial communities. However, insights and tools are currently lacking to characterize, understand and explain the effects of signal disruption on complex microbial communities. Here, we produced silica capsules containing an engineered lactonase that exhibits quorum quenching activity. Capsules were used to design a filtration cartridge to selectively degrade AHLs from a recirculating bioreactor. The growth of a complex microbial community in the bioreactor, in the presence or absence of lactonase, was monitored over a 3-week period. Dynamic population analysis revealed that signal disruption using a quorum quenching lactonase can effectively reduce biofilm formation in the recirculating bioreactor system and that biofilm inhibition is concomitant to drastic changes in the composition, diversity and abundance of soil bacterial communities within these biofilms. Effects of the quorum quenching lactonase on the suspension community also affected the microbial composition, suggesting that effects of signal disruption are not limited to biofilm populations. This unexpected finding is evidence for the importance of signaling in the competition between bacteria within communities. This study provides foundational tools and data for the investigation of the importance of AHL-based signaling in the context of complex microbial communities.

Keywords: quorum sensing, lactonase, biofilm, microbial community, silica encapsulation

### INTRODUCTION

Bacterial quorum sensing (QS) is among the most prominent and studied communication systems used by bacteria (Bassler, 1999). Numerous bacteria produce and utilize chemical signal molecules to coordinate their behavior in a cell density-dependent manner (Miller and Bassler, 2001; LaSarre and Federle, 2013). Bacterial QS has been shown to also regulate virulence and biofilm formation (LaSarre and Federle, 2013). Biofilms are comprised of a hydrated matrix of polysaccharides, proteins and nucleic acids that ultimately allow bacteria to attach to surfaces and live in complex community structures (Costerton et al., 1999). These structured communities enable a multicellular-like existence that is distinct from the planktonic state (Stewart and Costerton, 2001).

#### Edited by:

Tom Defoirdt, Ghent University, Belgium

#### Reviewed by:

Chien-Yi Chang, University of Bradford, United Kingdom Qian Yang, Yellow Sea Fisheries Research Institute (CAFS), China

> \*Correspondence: Mikael Elias mhelias@umn.edu

#### Specialty section:

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

Received: 20 November 2018 Accepted: 11 March 2019 Published: 29 March 2019

#### Citation:

Schwab M, Bergonzi C, Sakkos J, Staley C, Zhang Q, Sadowsky MJ, Aksan A and Elias M (2019) Signal Disruption Leads to Changes in Bacterial Community Population. Front. Microbiol. 10:611. doi: 10.3389/fmicb.2019.00611

Quorum quenching (QQ) relates to all processes that can interfere with QS (McClean et al., 1997). QQ is a strategy that is not aimed at killing bacteria or at limiting growth, but rather at controlling or changing the expression of different functions (Uroz et al., 2009). Consequently, QQ enzymes are naturally capable of interfering with this QS via the enzymatic degradation of autoinducer molecules (Zhang, 2003; LaSarre and Federle, 2013). This has been studied in the case of autoinducer-1, N-acyl homoserine lactones (AHLs) (Dong et al., 2000, 2001; Oh et al., 2012; Kim et al., 2013; Bzdrenga et al., 2016). Indeed, the disruption of bacterial signaling using QQ enzymes was previously shown to inhibit the production of virulence factors and biofilm production by numerous pathogens, both in vitro (Dong et al., 2000; Chow et al., 2014; Hraiech et al., 2014; Guendouze et al., 2017; Zhang et al., 2017) and in vivo (Dong et al., 2000; Hraiech et al., 2014). These properties have been instrumental in making QQ enzymes prime candidates for bacterial control in numerous fields of application. However, to achieve this goal, effort is required to overcome practical issues with use of QQ enzyme technology, such as low activity levels, activity at low or high temperatures, environmental stability, and production costs (Lee et al., 2016; Rémy et al., 2016).

A promising candidate to overcome the intrinsic limitations in current enzymes is the lactonase SsoPox, isolated from the hyperthermophilic crenarcheon Sulfolobus solfataricus (Merone et al., 2005; Elias et al., 2007, 2008). This enzyme belongs to the Phosphotriesterase-Like Lactonase family (Afriat et al., 2006; Elias and Tawfik, 2012), and naturally hydrolyzes a broad range of AHLs, from C6 AHL to 3-oxo C12 AHL (Hiblot et al., 2013). The SsoPox was shown to disrupt bacterial QS in vitro and in vivo (Hraiech et al., 2014; Guendouze et al., 2017). Additionally, this lactonase was reported to be catalytically active over a wide range of temperatures, from −19◦C to 70◦C (Merone et al., 2005; Rémy et al., 2016). Interestingly, this lactonase exhibits exceptional thermal stability (T<sup>m</sup> = 106◦C), resistance to denaturing agents, organic solvents, detergents, radiation, and proteases (Hiblot et al., 2012a; Rémy et al., 2016). A crystal structure of SsoPox revealed the critical importance of residue W263, interacting with the bound lactone ring of the AHL molecule (Elias et al., 2007, 2008; Del Vecchio et al., 2009). Mutation of W263I allowed for generation of variants with even greater lactonase catalytic activity (Hiblot et al., 2013; Jacquet et al., 2017).

While the substrate specificity of several lactonases has been determined (Hiblot et al., 2012b, 2013, 2015; Bzdrenga et al., 2014; Mascarenhas et al., 2015; Tang et al., 2015; Bergonzi et al., 2016, 2017), the types of bacteria that can be controlled by these enzymes is unclear. Indeed, AHL-based QS and effects of QS interference were mostly described in Gram-negative bacteria (Chow et al., 2014; Hraiech et al., 2014; Ivanova et al., 2015a,b; Guendouze et al., 2017). Other studies report activity of lactonase of bacterial strains that are not known as using AHLs (Ivanova et al., 2015a; Mayer et al., 2015). The presence of bacteria expressing lactonases was shown to reduce biofouling in a membrane bioreactor (MBR) (Oh et al., 2012; Kim et al., 2013, 2015), as well as affect the community structure of microbes attached to the membrane (Jo et al., 2016). Despite this, tools and insights are missing to adequately decipher the mechanisms underlying these observations.

In order to determine the effects of AHL degradation in the context of a complex microbial community, we used a silica gel, bead-based, bioencapsulation technique. Silica is a cytocompatible material in which bacteria or their enzymes can be physically confined, retained within the matrix, and protected from the environment (Reátegui et al., 2012; Mutlu et al., 2013, 2015; Aukema et al., 2014; Sakkos et al., 2016, 2017). Here, we used encapsulated Escherichia coli cells overexpressing the lactonase SsoPox W263I to produce enzymatically active beads. Because SsoPox is a hydrolase, and do not need any other recycling co-factor than water molecules, it does not need the cells to remain alive to maintain its catalytic activity. In fact, in this strategy, cells are used as "bags of enzymes." Encapsulation of bacteria overexpressing stable, engineered lactonases combines the intrinsic properties of the SsoPox enzyme, the lower production costs associated with the use of cells instead of purified enzyme, and a robust, permeable silica structure facilitating the integration of this enzyme in water treatment systems.

Catalytically active capsules were used as an enzymatic filtration matrix to degrade AHL signaling molecules produced by a complex soil microbial community cultured in a recirculating system. We determined that the presence of the lactonase in the filtration beads leads to a very significant reduction of biofilm formation over the course of the experiment (21 days) and that this reduction is associated with a change of the microbial population forming the biofilm. This experimental system opens up a new way to study the importance of bacterial signaling in complex microbial communities, the effects of signal disruption using lactonases, and highlights the potential of these enzymes to serve in water treatment processes, including a recirculating system.

### MATERIALS AND METHODS

### Preparation of Silica Lactonase-Containing Beads

The QQ lactonase SsoPox W263I and a negative control protein, an inactive mutant SsoPox 5A8 [carrying the mutations V27G/P67Q/L72C/Y97S/Y99A/T177D/R223L/L226Q/L228M/W 263H, obtained previously (Hiblot et al., 2013; Bergonzi et al., 2018)] were overexpressed in E. coli BL21-pGro7 as previously described using the autoinducing media ZYP (Hiblot et al., 2012a, 2013; Jacquet et al., 2017). The use of the 5A8 mutant allows for the control of the used plasmids and protein expression, allowing us to link observations to the enzymatic activity of SsoPox W263I. Cultures were grown to OD600 nm = 0.8 at 37◦C, while shaking at 200 RPM, and after overnight induction of the lactonase (18◦C, 0.2% <sup>L</sup>-arabinose, 200 RPM shaking), cells overexpressing proteins were centrifuged at 4,400 × g for 20 min at 4◦C. Cells were re-suspended in 100 mM potassium phosphate buffer, pH 7, at a concentration of 0.4 g/mL wet weight to provide 0.2 g/mL for the 1× lactonase beads. Gel

beads (1 mm diameter) containing the lactonase/control bacteria cultures were made using a dripping method while gelation occurred, using a method similar to a previously used protocol (Mutlu et al., 2013). Polyethylene glycol (PEG, 400 mg) with an average molecular weight of 10,000 Da, was mixed with 4 mL acetic acid (0.01 M) until the PEG dissolved. A 2.5 mL aliquot of tetramethyl orthosilicate (TMOS) was added and allowed to stir for 30 min until the solution became clear. One milliliter of cell suspension (0.2 g/mL) was mixed with the PEG/TMOS/acetic acid solution and gelation occurred within a few minutes. The bacteria-encapsulated beads (8 mL) were added directly to empty 10 mL chromatography columns to create filtration cartridges. Encapsulated bacteria can remain variable for several weeks. A report on encapsulation using similar gels show a reduction of ∼93% of viable E. coli cells after 3 weeks (Reátegui et al., 2012). Cell leakage from the gel is possible, but was previously found to be non-significant in similar gels (Reátegui et al., 2012). A filter (GE Healthcare) at the outlet of the column ensured the amount of beads present in the column would be constant throughout the duration of the experiment. Two different types of bacteria-encapsulated silica beads were produced, (1) beads where E. coli cells overexpressing the lactonase SsoPox W263I were entrapped (lactonase beads) and (2) beads where E. coli cells overexpressing the control protein (inactive mutant 5A8) were entrapped (control beads). These beads were used to produce three distinct filtration cartridges: (a) the 2× lactonase cartridge, containing only 8 mL of lactonase beads, (b) a control cartridge containing only 8 mL of control beads, and (c) a 1× lactonase cartridge containing a 1:1 ratio of lactonase beads and control beads (4 mL of each).

### Kinetic Assays of Lactonase in Silica Gel

Lactonase-containing silica gel solution, and the control protein, were poured into individual wells of 96 well microplates and allowed to solidify. Enzyme activity was quantified over a ∼7 months period (28 weeks) in buffer. Each well contained 75 µL of gel and was stored at 4◦C in the presence of the pte buffer (50 mM HEPES, pH 8, 150 mM NaCl, 0.2 mM CoCl2) or the lactonase buffer (2.5 mM Bicine pH 8.3, 150 mM NaCl, 0.2 mM CoCl2, 0.2 mM cresol purple, and 0.5% DMSO). The gel plus the buffer volume was 200 µL providing a 6.2 mm path length. Enzyme kinetics were measured by using a microplate reader (Synergy HT, BioTek, United States; Gen5.1 software), facilitated by a high level of gel transparency. The kinetics of lactonase activity were determined as previously described (Hiblot et al., 2012b, 2015; Bergonzi et al., 2016, 2017). Lactonase activity was expressed in enzymatic units defined as µM of substrate hydrolyzed per min per mg of cells (wet weight). All kinetic measurements were performed as triplicates. The activities of lactonase and phosphotriesterase were corrected by subtracting activities control gels (containing E. coli cells overexpressing mutant 5A8). The chromogenic substrate paraoxon was used as a proxy for the enzyme activity in order to evaluate the durability of gels over time. Assays were done as previously described (Hiblot et al., 2012a; Gotthard et al., 2013; Jacquet et al., 2017) and were performed using 10 µL of 20 mM paraoxon (1 mM final concentration) and a 200 µL final reaction volume. The paraoxon

degradation product (paranitrophenolate) was directly measured at 405 nm (ε = 17,000 M−<sup>1</sup> cm−<sup>1</sup> at pH 8.0). Activity over time was normalized to the measured activity at day 0.

### Flow-Through Recirculating Bioreactor System

The flow-through system used in this study consisted of three 3 L tanks set up in parallel. The parallel circuit was achieved through the use of a multi-channel peristaltic pump (Masterflex L/S, Cole-Parmer, United States) (**Figure 1** and **Supplementary Figure S1**). A peristaltic pump provided an even flow rate of 18 mL/min to each tank. The flow-through filtration cassette consists of a 10 mL chromatography column filled with QQ gel beads or control beads. Each tank contained three liters of 15× diluted LB medium in water. A pre-separated 96-stripwell plate was submerged to the bottom so that individual wells could be harvested for biofilm quantification or DNA extraction. At the bottom of each bioreactor were 22 mm square microscope cover slips to be later used for biofilm imaging. For the inoculum, ∼5 g of soil (disturbed Waukegan Silt loam; sampled in march) was suspended in 40 mL of water. Soil community was chosen as a model system to this study. The suspension was centrifuged for 5 min at 500 × g and 200 µL of the cloudy supernatant was added to 30 mL of LB medium and allowed to grow for 16 h at 37◦C, with shaking at 200 RPM. A 10 mL aliquot of this culture was inoculated into each tank and the system was allowed to run for 21 days at room temperature. During this time, measurements were taken every day to monitor OD600 nm and pH of the water, and amount of biofilm present on multiple coverslip surfaces.

A second, independent experiment was performed similarly using soil samples from the same location (sampled in October)

using duplicate bioreactors. The system was allowed to run for 10 days, and samples were taken at days 3 and 10.

Attempts to measure the AHLs concentration in the culture media using the sensor Chromobacterium violaceum CV026 have failed, due to the apparent toxicity of the supernatants to the biosensor.

### Effects of the Lactonase on a Complex Planktonic Community

The soil inoculum was used to inoculate 5 mL triplicate cultures (15× diluted LB medium) in 50 mL conical tubes. Tubes were incubated at 25◦C and treated by adding to the culture media either inactive mutant SsoPox 5A8 enzyme or the improved mutant SsoPox W263I enzymes to a final concentration of 0.5 mg/mL. Samples were collected for DNA extraction after 3 and 7 days. In this experimental conditions, no biofilm could be visualized or quantified using crystal violet.

### Biofilm Quantification

The submerged 96-stripwell plates were pre-broken so that individual wells could be extracted every day for measurements. Individual wells were extracted, in duplicate, for biofilm formation by using crystal violet biofilm quantification at 550 nm as previously described (Hraiech et al., 2014). To assess planktonic growth in the tanks, optical density of 200 µL samples were measured by using a 96 well plate spectrophotometer at 600 nm.

### pH Measurements

pH values were measured with a meter (Orion Star A214, Thermo Fisher Scientific, United States). The pH of each bioreactor was monitored throughout the experiment (**Supplementary Figure S2**).

### Sample Preparation for Imaging

Microscope cover glass samples were harvested for biofilm visualization analysis by using a Zeiss confocal microscope (West Germany, Cell Observer SD). The cover slips were fixed with 2% paraformaldehyde in 1× PBS for 1 h at room temperature, rinsed twice with 1× PBS, and fixed in a solution of 50% 1× PBS containing 50% EtOH. Samples were stored at −20◦C for later processing. For imaging analyses, the stored samples were washed twice with 1× PBS and stained with 1× SYBR Gold nucleic acid stain (Thermo Fisher, United States) for 10 min, washed with 100% EtOH, and mounted onto microscope slides for fluorescence analysis. A 1:4 mixture of Citifluor:Vectashield was used as the mounting medium.

### Microbial Community Analysis

Submerged wells from the stripwell plates were drained of excess cells/water and biofilm was scraped from the polypropylene well and into Powerbead <sup>R</sup> tubes for DNA extraction (DNeasy PowerSoil <sup>R</sup> DNA Extraction Kit, QIAGEN, Hilden, Germany). Purified DNA samples were submitted to the University of Minnesota's Genomics Center for 16S rRNA sequencing on the MiSeq platform at the University of Minnesota Genomics Center (Minneapolis, MN, United States). Each sample underwent amplification, dual-indexing, normalization, pooling, size selection, and a final QC prior to sequencing. The V4 region of the 16S rRNA gene was amplified by using primer set 515f (5<sup>0</sup> – GTG CCA GCM GCC GCG GTA A – 3<sup>0</sup> ) and 806R (5<sup>0</sup> – GGA CTA CHV GGG TWT CTA AT – 3<sup>0</sup> ). Negative (sterile water) controls were included throughout amplification and sequencing. Sequencing data were deposited to the Sequence Read Archive (SRA) under the accession number SRP156219.

### Sequencing Data Analysis

All samples were processed using Mothur v1.35.1 (Schloss et al., 2009). Forward and reverse reads were paired-end joined using fastq-join software (Aronesty, 2013), and subjected to quality trimming using parameters previously described (Staley et al., 2015). High quality sequences were aligned against the SILVA database v.119.v4 (Pruesse et al., 2007), subjected to a 2% pre-cluster to remove sequence errors (Huse et al., 2010), and chimeras were removed using UCHIME v. 4.2.40 (Edgar et al., 2011). Operational taxonomic units (OTUs) were classified at 97% similarity using the furthest-neighbor algorithm and classified against the Ribosomal Database Project v.11.5 (Cole et al., 2008). For each sample, 1,250 sequences were used for final analyses. Alpha diversity indices were analyzed through the Mothur v1.35.1 program. Genus-level identification was achieved for the composition of the bacterial community. Analysis of similarity (ANOSIM) and analysis of molecular variances (AMOVA) were used to evaluate the beta diversity (community composition) among samples using Bray–Curtis dissimilarity matrices (BC) (Bray and Curtis, 1957; Excoffier et al., 1992; Clarke, 1993). Ordination of Bray–Curtis matrices was performed using principal coordinate analysis (PCoA) to further analyze diversity of sample days throughout the tank (Anderson and Willis, 2003). Pearson correlation was performed using Mothur v1.35.1 program to evaluate the correlation between the genera abundance under temporal change in different treatments. To visualize the distribution of taxonomies and diversities in microbial communities among the samples, "ggplot2" package in R v3.3.1 was used with rarefied relative abundance and OTUs at genus level (Aislabie et al., 2013). All of statistical analyses were done using α = 0.05. Graphpad prism software was used to calculate student unpaired, two-tailed t-test values.

## RESULTS AND DISCUSSION

### Engineered Lactonase-Expressing Cells Entrapped in silica Capsules

Silica encapsulation is a method of choice for entrapping enzymes or cells due to their compatibility with biological molecules, mechanical properties, durability, stability, cost, and easy synthesis. Silica gels have been previously used to encapsulate bioreactive bacteria for bioremediation (Reátegui et al., 2012; Aukema et al., 2014; Sakkos et al., 2016, 2017). While most encapsulated bacteria may remain viable through the process of making the gels (Benson et al., 2018), it is likely to be unnecessary in this study, since the lactonase SsoPox

is a metalloenzyme that only requires a water molecule as the nucleophile for the hydrolytic reaction (Elias et al., 2008). Therefore, cells can be viewed as "bags of enzymes" that disrupt the AHL signaling molecules produced by bacteria.

The engineered silica gels showed catalytic activity against lactones including lactones with short aliphatic chains (i.e., C6- AHL and γ-heptanoic lactone) and lactones with long acyl chains (i.e., C8-AHL and γ-undecanoic lactone). This observation is consistent with the enzyme activity in solution (**Figure 2A**), that is reported to degraded AHLs ranging from C6 to 3-oxo C12 AHL, and oxonolactone of numerous chain lengths with similar catalytic efficiencies (Hiblot et al., 2013). The measured activity demonstrates that the lactonase overproduced in E. coli cells is active inside the beads, and that different types of lactones can access its active site. The lactonase assay used in this study was pH-based and was previously described by us (Hiblot et al., 2012b; Bergonzi et al., 2016, 2017). While this assay allows for the monitoring of the lactone ring opening (generating a proton), it requires significant optimization of the activity buffer for each measurement due to the buffering capacity of the gel.

In order to better and more conveniently evaluate the durability of the silica gels over time, we used the chromogenic substrate paraoxon as a proxy for SsoPox activity, as previously reported (Elias et al., 2008; Hiblot et al., 2012a, 2013).

Results in **Figure 2B** show that the lactonase-containing gel remains active for at least 39 weeks (∼9 months) in solution. This is 5 months longer than previous studies on atrazine degradation that were performed with a different enzyme but in similar conditions (Reátegui et al., 2012). The observed durability is consistent with the extreme stability of SsoPox W263I, that remains stable for >300 days (∼10 months) at 25◦C as a purified protein sample (Rémy et al., 2016). Interestingly, the activity of the enzymatic gel at T<sup>0</sup> increases over the course of the first 5 weeks of the experiment (approximately threefold). This may be caused by a change in the structure or porosity of the silica gel that could lead to an increased diffusion of the substrate into the enzyme active sites and may suggest that our current gel formulation could be optimized in future studies. Our success in obtaining silica gels containing engineered, overexpressed lactonases opens up a lot of new possibilities to study signal disruption in microbial communities.

Control of expression level, and the ability to engineer the lactonase, will be useful to optimize QQ in complex contexts. Additionally, because this technology in practice does not require purified enzyme, it may allow for the production of highly potent, specific beads to inhibit biofilms and biofouling in water filtration systems, at a low cost.

### Silica Beads Containing Lactonase Enzyme Inhibit Biofilm Formation of Complex Microbial Communities in a Water Recirculating System

In this study, a water recirculating system was used to examine the influence of lactonase on soil bacterial community structure in planktonic cells. The water within the tank was pumped through a filtration cartridge (**Figure 1** and **Supplementary Figure S1**) containing different porous, lactonase-containing silica capsules (see section "Materials and Methods"). This experimental design was based on the hypothesis that water soluble AHLs produced by the microbial community growing in the tank would be filtered through the cartridge, and degraded by the lactonase enzyme.

The effects of the lactonase enzyme in the filtration cartridge on the microbial community was monitored at different levels by examining the pH of the tank medium and the optical density at 600 nm. The pH of the tank media increased from a starting value of ∼6.2 to a final value of ∼8.0 in all three experimental setups (**Supplementary Figure S2A**). Similarly, the OD600 nm, used as a proxy for cell density, slightly increased over the course

(B) Activity of the enzymatic silica gels over time, using the chromogenic substrate paraoxon as a proxy for enzyme activity.

FIGURE 3 | Presence of lactonase reduces biofilm formation of the microbial community in the bioreactors. (A) Bioreactor parameters over the time course of the experiment (21 days). Measurements (duplicates) were performed on the three distinct bioreactors equipped with different filtration cartridges: the 2× lactonase cartridge, containing only lactonase beads (blue line), the control cartridge containing only control beads (dark line), and the 1× lactonase cartridge containing a 1:1 ratio of lactonase beads and control beads (green line). Biofilm quantification in submerged wells as quantified by crystal violet binding measured at 550 nm. (B) Glass slips submerged in the bioreactors were stained using SYBR gold DNA stain and visualized using a 60× magnification.

of the experiment in a similar fashion in the three bioreactors (**Supplementary Figure S2B**).

Biofilm formed and was quantified over the time course of the experiment in the bioreactors (**Figure 3A** and **Supplementary Figure S3**). Submerged plastic wells were sampled and assayed using crystal violet dye (**Supplementary Figure S4**). Measurements indicated that biofilms were slowly forming during the first 11 days of the experiments, and then accelerated in all bioreactors. Interestingly, there were no significant differences in biofilm quantification during the first 11 days of the experiment between filtration systems using lactonase or control beads. However, after 23 days in the presence of the lactonase beads, biofilm was reduced to ∼50% of that formed in the control. Biofilm reduction might even be greater since OD600 nm measurements reached saturation in the control experiment. The degree of biofilm reduction is consistent with the observed reduction in biofilm dry weight in tubing [49–44% when comparing control and 2× lactonases after 21 days for precolumn and post column tubing, respectively (**Supplementary Figure S5**)]. Remarkably, inhibition of biofilm was observed as a function of lactonase concentration in the cartridge: inhibition was larger in the 2× concentration compared to the 1× concentration (∼50% and ∼30%, respectively) (**Figure 3A**). The reasons accounting for this enzyme dose dependence are unclear at this stage, and will be the subject of subsequent investigations in greater details.

Biofilm formation in the bioreactors was imaged in both the early and late stages of the experiment [days 4 and 20, respectively (**Figure 3B**)]. DNA staining of the submerged microscope slips revealed that the presence of lactonase in the filtration cartridge led to a reduction in the adhesion of cells to the coverslip surfaces. While we were unable to visualize the matrix in this experiment, it was apparent that the biofilms in the control tanks had more structure and maturity than did those formed in the lactonase treated tank. Interestingly, the reduction of cell attachment was also observed in the early stage of biofilm formation (by day 4) and may indicate the importance of signaling in both the biofilm attachment and maturation steps (Parsek and Greenberg, 2005). The presence of bacteria in our tested community, such as Pseudomonas and Aeromonas that are known to form biofilms, to utilize AHL-based QS and to be inhibited by lactonases (Cao et al., 2012; Guendouze et al., 2017), may partly explain the observed biofilm inhibition.

The ability of lactonase to inhibit complex biofilms was first evidenced using encapsulated microbes naturally expressing lactonases in MBR systems (Kim et al., 2013). Here, our study provides a comprehensive analysis of the dynamics of the effect of lactonase on a soil community, and with two different doses of the enzymatic quencher. Additionally, the demonstration in this study of the ability of entrapped, overexpressed and improved lactonases to inhibit biofilm formation in a recirculating system opens new perspectives for enzyme technology use in water treatment. It also raises questions about the specific mechanism of action of entrapped lactonases on the microbial community signaling. Because lactonase enzymes degrade the secreted signaling molecules (AHLs), no physical contact between the enzyme molecules and bacteria might be needed for its action. This hypothesis is consistent with our bioreactor design and our observed inhibition of biofilm formation. Questions concerning the diffusion ability of AHLs in various media will need to be investigated, as it may modulate the "action range" of the various AHLs, and consequently, of lactonases.

### The Presence of a Lactonase Induces Changes in the Biofilm Microbial Composition

DNA from biofilm samples obtained from the three different bioreactors were isolated and submitted for amplicon sequencing of 16S rRNA (**Figure 4**). Samples were collected over the time course of the experiment to evaluate the population dynamics. Given the samples had relatively low diversity, we analyzed 1,250 sequences from each sample. This low diversity may be due

to the pre-culture step of soil samples that reduced microbial diversity, as described in other examples (Tabacchioni et al., 2000; Gorski, 2012). Taxonomic community composition, classified to genus, (**Figure 4**) and principle coordinate analysis (**Figure 5** and **Supplementary Figure S6**) indicated that microbial communities in all three replicate systems were very similar in the early stages of the experiment (ANOSIM, p > 0.05). For example, at day 4, Aeromonas represents 59.76, 63.20 and 69.52% of 2×, 1×, and control treatments, respectively. This was expected because all three bioreactors were inoculated with the same starting culture. However, notable differences in biofilm microbial communities were seen from day 7. By day 7, 42.2% of the microbial communities in biofilms from bioreactors treated with the highest concentration of lactonase (2×) were comprised by Aeromonas. In contrast, this bacterium comprised 79.7% and 81.7% of the community in the lactonase (1×) and control treatments, respectively (**Figure 4**). Principle coordinate analysis also highlighted that community composition starts to separate from control by day 7 (ANOSIM, R = 0.625, p = 0.036) (**Figure 5**).

Other notable differences include the relative populations of Stenotrophomonas (Gram-negative), Pseudomonas (Gramnegative), and Clostridium XIVa (Gram-positive) (**Figure 4**). For instance, the 2× lactonase bioreactor showed that Stenotrophomonas appeared much earlier than that seen in the 1× lactonase and control bioreactors (on day 14). In the 1× lactonase bioreactor, however, we observed an increase in the Pseudomonas population by day 11 and a gradual increase in its abundance within the community throughout the rest of the experiment. Lastly, in the later part of the experiment (days 9–18), the control bioreactor hosted a larger Clostridium population than did the two other bioreactors.

These experiments were repeated as an independent experiment performed in duplicates with a community from the same soil location but sampled at a different time of the year. We used the 1× lactonase concentration, and sampled two time points (day 3 and day 10) (**Supplementary Figure S7**). Interestingly, despite the different starting community composition, we can observe similar features. Notable differences are seen by day 10, including an increase of the Stenotrophomonas population in the presence of the enzyme (8% versus 1.5% in control). The Pseudomonas, specifically observed in the presence of lactonase as described above, are also favored in this independent run (15.7% versus 1.4%, p < 0.05). Lastly, the Clostridium group also varies (4.5% versus 14.2%; p < 0.05) and is less abundant in presence of lactonase than in the control bioreactor. These three groups are similarly affected by signal disruption using the lactonase in both independent experiments.

These data reveal that the presence of the lactonase enzyme of the filtration cassette lead to changes in the composition of the communities that occurred rapidly and were persistent throughout the experiment. Data show that these alterations of microbial communities occur in a lactonase dose-dependent fashion, and can be observed in biofilm communities.

### The Presence of a Lactonase Induces Changes in the Microbial Composition in Suspension Culture

Experiments performed with the same starting community in smaller volumes (5 mL) of suspension cultures and in presence of the inactive 5A8 mutant or the lactonase SsoPox W263I reveal similar alteration of the composition of the microbial community. Whereas in this experimental setup, no biofilm formation would be detected within the time frame (7 days), the analysis of sequencing data of culture supernatant samples to the genus-level (**Supplementary Figure S8**), PCoA (**Figure 5C**) and statistics (**Supplementary Tables S1**, **S2**) reveal that these suspension microbial communities are significantly different between treatments (ANOSIM, R = 0.94, p < 0.001). In fact, the compositions of the microbial population in presence or in

absence of quorum quencher are so different that comparisons between communities are difficult. Nevertheless, we observe again that Pseudomonas are more abundant in the presence of the lactonase than in the control experiments (24.9% versus 14%; p < 0.05). These experiments reveal that the effects of signal disruption using the lactonase SsoPox W263I robustly alters the composition of the treated microbial community and that changes are not limited to biofilms, but also affect suspension communities.

### Presence of a Lactonase Modulates Diversity Within Genera but Not the Community Diversity

Analysis of the relative abundance and diversity of genera distinctly highlighted population changes as a function of lactonase concentration and time of incubation (**Figure 6**). Overall, this analysis showed that while the presence of the lactonase induced changes in the relative abundance and diversity of some bacterial genera, it did not significantly alter overall community diversity. This is further evidenced by Shannon indices and observed species counts that are both slowly increasing over the time course of the experiments in all three bioreactors in a similar way (**Supplementary Figure S9**). Additionally, this analysis (**Figure 6**) highlights that Stenotrophomonas, Pseudomonas and Clostridium XIVa are specifically enriched over time in the 2× lactonase (Pearson correlation, r = 0.95, P = 0.001), the 1× lactonase (Pearson correlation, r = 0.90, P = 0.002), and the control bioreactors (Pearson correlation, r = 0.77, P = 0.027), respectively. This abundance increase is concomitant with an increase in their diversity.

Bacterial community compositions revealed that a few genera in low abundance were specific to treatments. For example, Propionispora were only detected in flow systems using lactonase in the filtration cartridge, whereas Acetivibrio were only detected

in the control bioreactor. Other microbial community biases were also noted: Achromobacter was more abundant in tanks using lactonases, as compared to controls. In contrast, the abundance and diversity of Sporomusa decreased as the concentration of lactonase increased.

Changes in microbial communities in the presence of lactonase were recently observed in the context of membrane biofouling (Jo et al., 2016), as well as fish gut microbiomes (Zhou et al., 2016). In this study, we show the dynamics of community composition changes in a water recirculating system and observe that these changes are concomitant with biofilm inhibition. Furthermore, these changes depend on the enzyme concentration and are related to the abundance and diversity of genera, but not the overall diversity of the community. The mechanism(s) underpinning the ability of QQ lactonases to affect complex microbial communities are unknown. Complete QS circuits (a synthase and a receptor) were previously reported to be found only in Proteobacteria (Case et al., 2008). Within the bacterial genera detected in this study, some are known to: (1) produce AHLs and utilize them for sensing [i.e., Pseudomonas, Aeromonas, Yersinia, (Venturi, 2005; Medina-Martínez et al., 2007; Ortori et al., 2007; Khajanchi et al., 2011)], (2) be capable of producing AHLs [i.e., Enterobacter (Yin et al., 2012; Ochiai et al., 2013)], and (3) be capable of sensing AHLs [i.e., Stenotrophomonas, Escherichia, Shigella (Soares and Ahmer, 2011; Martínez et al., 2015; Taghadosi et al., 2015; Lu et al., 2017)] (**Supplementary Table S3**). Others, however, are known to not produce, use, or sense AHLs (i.e., Clostridium). Additionally, the relationship between the AHL signal disruption by a lactonase and some genera is not straight forward as indicated by the increase of Pseudomonas in the community present in the 1× lactonase system. Furthermore, it is intriguing to note that Clostridium XIVa, despite being a Gram-positive bacterium that does not produce and/or sense AHLs, is reduced in the presence of the lactonase. This observation might echo previous studies describing the ability of lactonase to inhibit the biofilm of Staphylococcus aureus and Escherichia coli (Ivanova et al., 2015a; Mayer et al., 2015). Mechanisms explaining these observations are lacking. Yet, this study is a first comprehensive, time-resolved, statistically significant description on the effects of a lactonase on microbial community structures, and will inform the understanding of these complex interactions.

### CONCLUSION

Our study demonstrates that lactonase-containing beads reduce biofilm formation by a complex soil microbial community, in a dose-dependent manner. Biofilm inhibition was observed, despite the presence of abundant microbes that are not known for using or sensing AHLs, such as Clostridium. Sequencing analysis revealed that biofilm inhibition was concomitant to a change in the microbial community composition on the surface. Dynamic population analysis shows that the bias introduced by AHL signal disruption occurs rapidly and is persistent over the time course of the experiment. We show in three independent experiments that signal disruption using a lactonase robustly and significantly changes the composition of microbial communities. We show that changes related to the relative proportion of some genera, but may also reflect in the observed specific presence or absence of genera in the biofilm. Therefore, signal disruption using a lactonase may have global effects on microbial populations, and not only inhibit bacteria utilizing AHLs for signaling.

Additionally, we find that signal disruption also lead to changes in the composition of suspension communities. This suggests that the importance of AHLs signaling extends beyond biofilm formation. In fact, this unexpected finding likely points to the importance of signaling in the competition between bacteria within communities (Diggle et al., 2007; Sandoz et al., 2007; Defoirdt et al., 2010).

Finally, the system used in this study provides a unique platform to study the importance of bacterial signaling, and the effects of signal disruption on complex microbial communities of multiple origins. We strongly feel that these findings and tools will pave the way for future investigations exploring the potential use of QQ enzymes in the water treatment arena, as well as the importance of signaling in complex microbial communities.

### DATA AVAILABILITY

The datasets generated for this study can be found in Sequence Read Archive, SRP156219.

### AUTHOR CONTRIBUTIONS

ME conceived and designed the work. MS, CB, and JS performed the experiments. ME, MS, CB, CS, QZ, MJS, and AA analyzed the data. QZ, MJS, and CS performed the statistical analysis. ME, CB, and MS wrote the first draft. ME, MS, JS, QZ, and CS wrote sections of the manuscript. ME, MJS, and AA critically revised the manuscript. All authors read and approved the final manuscript.

## FUNDING

This work was supported by the MnDrive Initiative, the MnDrive demonstration grant, the BTI Biocatalysis Initiative, and the BARD grant IS-4960-16 FR to ME.

### ACKNOWLEDGMENTS

We thank the University of Minnesota Genomics Center for the deep sequencing. We also thank Dr. David Daude and Dr. Eric Chabriere for fruitful discussions.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb.2019. 00611/full#supplementary-material

### REFERENCES

fmicb-10-00611 March 28, 2019 Time: 10:50 # 11


urinary catheters. ACS Appl. Mater. Interfaces 7, 27066–27077. doi: 10.1021/ acsami.5b09489



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

Copyright © 2019 Schwab, Bergonzi, Sakkos, Staley, Zhang, Sadowsky, Aksan and Elias. 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.

# Discovering, Characterizing, and Applying Acyl Homoserine Lactone-Quenching Enzymes to Mitigate Microbe-Associated Problems Under Saline Conditions

#### Tian-Nyu Wang<sup>1</sup> , Qing-Tian Guan<sup>2</sup> , Arnab Pain<sup>2</sup> , Anna H. Kaksonen<sup>3</sup> and Pei-Ying Hong<sup>1</sup> \*

<sup>1</sup> Water Desalination and Reuse Center, Biological and Environmental Science and Engineering Division, King Abdullah University of Science and Technology, Thuwal, Saudi Arabia, <sup>2</sup> Pathogen Genomics Laboratory, Division of Biological and Environmental Science and Engineering, King Abdullah University of Science and Technology, Thuwal, Saudi Arabia, <sup>3</sup> CSIRO Land and Water, Floreat, WA, Australia

#### Edited by:

Ana Maria Otero, University of Santiago de Compostela, Spain

### Reviewed by:

Xiuzhu Dong, Institute of Microbiology (CAS), China Manuel Romero, University of Nottingham, United Kingdom

\*Correspondence:

Pei-Ying Hong peiying.hong@kaust.edu.sa

#### Specialty section:

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

Received: 19 December 2018 Accepted: 01 April 2019 Published: 17 April 2019

#### Citation:

Wang T-N, Guan Q-T, Pain A, Kaksonen AH and Hong P-Y (2019) Discovering, Characterizing, and Applying Acyl Homoserine Lactone-Quenching Enzymes to Mitigate Microbe-Associated Problems Under Saline Conditions. Front. Microbiol. 10:823. doi: 10.3389/fmicb.2019.00823 Quorum quenching (QQ) is proposed as a new strategy for mitigating microbeassociated problems (e.g., fouling, biocorrosion). However, most QQ agents reported to date have not been evaluated for their quenching efficacies under conditions representative of seawater desalination plants, cooling towers or marine aquaculture. In this study, bacterial strains were isolated from Saudi Arabian coastal environments and screened for acyl homoserine lactone (AHL)-quenching activities. Five AHL quenching bacterial isolates from the genera Pseudoalteromonas, Pontibacillus, and Altererythrobacter exhibited high AHL-quenching activity at a salinity level of 58 g/L and a pH of 7.8 at 50◦C. This result demonstrates the potential use of these QQ bacteria in mitigating microbe-associated problems under saline and alkaline conditions at high (>37◦C) temperatures. Further characterizations of the QQ efficacies revealed two bacterial isolates, namely, Pseudoalteromonas sp. L11 and Altererythrobacter sp. S1-5, which could possess enzymatic QQ activity. The genome sequences of L11 and S1-5 with a homologous screening against reported AHL quenching genes suggest the existence of four possible QQ coding genes in each strain. Specifically, two novel AHL enzymes, AiiAS1−<sup>5</sup> and EstS1−<sup>5</sup> from Altererythrobacter sp. S1-5, both contain signal peptides and exhibit QQ activity over a broad range of pH, salinity, and temperature values. In particular, AiiAS1−<sup>5</sup> demonstrated activity against a wide spectrum of AHL molecules. When tested against three bacterial species, namely, Aeromonas hydrophila, Pseudomonas aeruginosa, and Vibrio alginolyticus, AiiAS1−<sup>5</sup> was able to inhibit the motility of all three species under saline conditions. The biofilm formation associated with P. aeruginosa was also significantly inhibited by AiiAS1−5. AiiAS1−<sup>5</sup> also reduced the quorum sensing-mediated virulence traits of A. hydrophila, P. aeruginosa, and V. alginolyticus during the mid and late exponential phases of cell growth. The enzyme did not impose any detrimental effects on cell growth, suggesting a lower potential for the target bacterium to develop resistance over long-term exposure. Overall, this study suggested that some QQ enzymes obtained from the bacteria that inhabit saline environments under high temperatures have potential applications in the mitigation of microbe-associated problems.

Keywords: quorum quenching enzyme, biofouling, salinity, virulence inhibition, AHL, bacterial motility, Altererythrobacter

### INTRODUCTION

fmicb-10-00823 April 15, 2019 Time: 20:6 # 2

Seawater is increasingly being used as an alternative water resource for various purposes to alleviate water demands in water-stressed countries (Elimelech and Phillip, 2011). However, marine microorganisms can attach onto surfaces, propagate and establish biofilm matrices, and frequently, this attachment can result in detrimental consequences. For example, biofilm formation can foul membranes in a seawater desalination plant, in turn reducing the flux (Al-Ahmad et al., 2000). Biofilm formation can also accelerate the biocorrosion of metal pipelines (Enning et al., 2012). Marine pathogens form biofilms on fish and shrimp, which can lead to the mortality and morbidity of these livestock and cause economic losses in marine aquaculture (Mizan et al., 2015). In all instances, biofilm formation increases the capital and operational costs associated with seawater usage.

Conventional antifouling strategies include the use of toxic biocides and coating materials such as tributyltin, copper, chlorine and ozone (Amara et al., 2018). However, there are health, safety and environmental concerns associated with the incessant use of these chemicals. Recently, a quorum quenching (QQ) strategy was proposed as an eco-friendly way to inhibit biofouling by blocking the cell-to-cell communication ability of bacteria (which is also known as quorum sensing, or QS). QS is a cell density-dependent regulatory mechanism used by bacteria to coordinate group behavior in response to QS signals secreted by the cell population. The concentration of QS signals increases as the cell population grows, which, upon reaching a certain threshold value, will trigger the expression of certain genes, including one related to pathogenicity, biofilm formation, spore germination and other functions (Defoirdt, 2018). QQ bacteria and QQ enzymes have been demonstrated to be effective in the membrane fouling mitigation of lab-scale membrane bioreactors (MBRs) used for wastewater treatment (Lee et al., 2018; Oh and Lee, 2018), and in a pilot-scale MBR (Lee et al., 2016).

Several studies have also presented the QQ strategy for mitigating biofouling in marine environments or in seawater desalination plants (Dobretsov et al., 2011; Katebian et al., 2016, 2018). The QQs discovered and applied to date have primarily been restricted to QS inhibitors (e.g., vanillin, cinnamaldehyde, and kojic acid). Nevertheless, several QQ bacteria were tested for biocontrol in marine environments. Tinh et al. (2008) enriched a complex bacterial consortium that exhibited AHL degradation activity, and they introduced this enrichment culture to colonize larval fish guts and demonstrated an improved survival rate from AHL-induced virulence traits by opportunistic bacteria. Torres et al. (2016) further screened for QQ enzymatic activity among 450 bacterial strains that were isolated from a mollusk hatchery, and they identified Alteromonas stellipolaris PQQ-42 as a potential AHL-degrading bacterium that increased the survival rate of corals against Vibrio. QQ bacteria and enzymes were subsequently shown to be widely distributed in marine sources (Romero et al., 2011, 2012). Marine isolates belonging to the Erythrobacter, Labrenzia, and Bacterioplanes genera were capable of degrading AHL molecules (Rehman and Leiknes, 2018). However, none of these studies took advantage of the special traits of marine QQ enzyme-secreting bacteria to mitigate marine biofouling. Moreover, these studies did not evaluate if the QQ enzymes were able to mitigate biofouling under the harsh environmental conditions representative of industrial seawater applications. These environmental conditions include high salinity (of up to 58 g/L), high temperatures (of up to 45◦C) and an alkaline pH ranging from 7.2 to 8.0 (Scarascia et al., 2016).

To address this knowledge gap, during this study, bacteria isolated from a Saudi coastal habitat and the Red Sea were first screened for their AHL-quenching ability. Five AHLquenching bacterial isolates from the genera Pseudoalteromonas, Pontibacillus, and Altererythrobacter were determined to exhibit high AHL quenching at a salinity value of 58 g/L and a pH of 7.8 at 50◦C. The genomes of two bacterial isolates were sequenced, and the potential QQ genes from the genomes were screened based on their homologies with reported QQ genes. The possible AHL-quenching genes were verified by expressing the potential QQ genes in recombinant E. coli to obtain enzymatically active recombinant proteins for testing. Subsequently, two enzymes (N-acyl homoserine lactonase, AiiA and esterase, Est) from Altererythrobacter sp. S1-5 were biochemically characterized at different pH, salinity and temperature. The AiiA from Altererythrobacter sp. S1-5 was further demonstrated to inhibit marine biofilm formation and the virulence of the opportunistic bacteria Aeromonas hydrophila, Pseudomonas aeruginosa, and Vibrio alginolyticus under saline conditions. The findings from this study demonstrate the potential feasibility of using QQ bacteria and enzymes to mitigate biofouling under saline conditions. Specifically, it is the first study to demonstrate the presence of AHL-quenching activity in Altererythrobacter.

### MATERIALS AND METHODS

### Sample Collection, Strain Isolation, and Identification

High-salinity artificial seawater (**HSAS**: 51.5 g/L NaCl, 0.74 g/L KCl, 0.99 g/L CaCl2, 2.85 g/L MgCl2, and 1.92 g/L MgSO4;

salinity: 58 g/L), high-salinity marine broth medium (**HSMB**: **HSAS** containing 1 g/L yeast extract, 5 g/L peptone), and highsalinity marine agar (**HSMA**: **HSMB** with 15 g/L agar) were used for strain isolation and screening. Marine aquaculture sludge from the Jeddah Fisheries Research Center (JFRC) as well as beach sand and seawater from the Red Sea were collected for the isolation of AHL-quenching bacteria. A 50 mL aliquot of seawater was transferred into a sterile flask. For the sludge and sand, 15 g of each sample was individually poured into separate sterile flasks that contained 50 mL of sterile **HSAS** and 2 g of glass beads (Sigma, St. Louis, MO, United States; diameter: 5 mm). All the inoculated flasks were cultivated at 40◦C and 180 rpm for 24 h. Following that incubation, the cultures were serially diluted with **HSAS** and plated onto **HSMA** plates. The agar plates were further incubated at 40◦C for 24 h. Colonies showing different morphologies were picked and purified by plate streaking.

The bacterial isolates were identified through partial length 16S rRNA gene sequencing. A single colony of marine isolates was scraped with sterile toothpicks, suspended in 20 µL of sterile water and heated at 95◦C for 5 min to achieve cell lysis. One microliter of supernatant was used as the DNA template for polymerase chain reactions (PCRs). A partial 16S rRNA gene was amplified using the universal primers 11F (5<sup>0</sup> -GTTYGATYCTGG CTCAG-3<sup>0</sup> ) and 1492R (5<sup>0</sup> -GGYTACCTTGTTACGACTT-3<sup>0</sup> ). The PCR was performed at 95◦C for 5 min, followed by 35 cycles of 30 s at 95◦C, 30 s at 52◦C and 2 min at 72◦C, and a final elongation for 5 min at 72◦C. The PCR products were purified using a Wizard SV Gel and PCR Clean-Up System (Promega, Madison, WI, United States) and submitted to the KAUST Bioscience core lab for Sanger-based sequencing using the primers 11F, 1492R, 338F (5<sup>0</sup> -ACTCCTACGGGAGGCAGCAG-3 0 ), and 592R (5<sup>0</sup> -GWATTACCGCGGCKGCTG-3<sup>0</sup> ). The sequences were paired and searched against the NCBI GenBank database using the BLASTn search algorithm.

### Screening for AHL Quenching Strains From Marine Isolates Under High Salinity

Single colonies of tested strains were inoculated into individual wells of sterile 96-well microtiter plates, and each well contained 200 µL of autoclaved **HSMB** medium. The 96-well plates were incubated at 50◦C and 120 rpm for 24 h. After the incubation, the plates were centrifuged at 2,300 × g for 10 min. The supernatant in each well was discarded and the cell pellet was washed once with HSAS. The centrifugation process was repeated and the resulting cell pellets were individually resuspended in 200 µL of **HSAS** containing an AHL mixture (0.29 mM C4-HSL, 0.25 mM C6-HSL, 0.22 mM C8-HSL and 0.17 mM 3-oxo-C12-HSL). After incubating at 50◦C for 24 h, the culture was centrifuged again at 2300 × g for 10 min. Ten microliters of supernatant was collected for residual AHL quantification using an Agrobacterium tumefaciens bioassay (Tang et al., 2013). In brief, A. tumefaciens NT1 (traR, tra::lacZ749) (Piper et al., 1993) was grown in **AT** minimal medium (10.7 g/L KH2PO4, 2 g/L (NH4)2SO4, 78 mg/L MgSO4, 13 mg/L CaCl2.H2O, 5 mg/L FeSO4.7H2O, 1.4 mg/L MnSO4.H2O, pH 6.7, and 0.5% filtered glucose, pH = 6.7) at 28◦C and 150 rpm overnight. The culture was diluted 1:500 into fresh **AT** minimal medium containing 250 µg/mL 5-bromo-4 chloro-3-indolyl-β-D-galactopyranoside (X-gal, Sigma, St. Louis, MO, United States) to make the A. tumefaciens bioassay solution. Ten microliters of each sample used to analyze the residual AHL concentrations were pipetted into individual wells of a sterile 96 well-plate, with each well containing 190 µL of A. tumefaciens bioassay solution. After an incubation at 28◦C for 12 h, the absorbance of each well at 492 nm and 630 nm was recorded using a SpectraMax 340 PC 384 microplate reader (Molecular Devices LLC, San Jose, CA, United States). The absorbances at 492 nm and 630 nm are a combination of absorption and light scattering by indigo (which is an X-gal degradation product) and biosensor cells (Tang et al., 2013). The residual AHL activity in each sample was expressed as normalized β-galactosidase activity as described in Tang et al. (2013). The quantification was performed in triplicate. The same concentration of AHL mixture in abiotic HSAS solution was incubated under the same conditions described above and used as a negative control when determining the residual AHL activity. The relative AHL quenching efficiency (QE) for each strain was determined using Eq. (1) as follows:

$$QE = \frac{AHLn - AHLs}{AHLn} \times 100\% \tag{1}$$

where AHLn denotes the residual AHL activity of the negative control after the reaction and AHLs denotes the residual AHL activity of bacterial samples after the reaction with potential QQ enzymes. It was previously shown that AHLs can be rather unstable in alkaline and high temperature environments (Yates et al., 2002). Therefore, to eliminate the instability of AHLs caused by abiotic factors after long-term reactions and to denote the AHL activity changes arising from the potential QQ enzymes more accurately, we use AHLn instead of the initial AHL activity spiked into the prior reaction mixture. Strains with a relative AHL QE ≥ 90% were selected for further QQ enzyme screening.

### Localization of AHL Quenching Enzyme in AHL Quenching Isolates

Artificial seawater (**AS**: 29.5 g/L NaCl, 0.74 g/L KCl, 0.99 g/L CaCl2, 2.85 g/L MgCl2, and 1.92 g/L MgSO4; salinity: 36 g/L), marine broth medium (**MB**: **AS** with 1 g/L yeast extract, 5 g/L peptone), and marine agar (**MA**: **MB** with 15 g/L agar) were used for the AHL quenching enzyme experiment. Selected strains were grown in **MB** medium at 37◦C overnight. To obtain the potential AHL quenching enzymes present in the intracellular fraction of each bacterial isolate, 5 mL of pure culture was centrifuged at 20,000 × g for 10 min, and the cell pellet was suspended in phosphate-buffered saline (PBS) (Fisher Scientific, Hampton, NH, United States). The cells were lysed using a Q500 sonicator (Qsonica, Newtown, CT, United States) at a 45% amplitude for 5 min, with repetitive 15 s pulsating sonication at 45 s intervals. The fraction was centrifuged at 10,000 × g for 10 min and filtered through a 0.2 µm cellulose acetate membrane to obtain the lysed cellular extract as filtrate. To obtain the potential AHL-quenching enzymes secreted extracellularly by the bacterial isolates, a cell pellet made from 5 mL of overnight

culture was resuspended in 5 mL of **AS**. The cell suspension was incubated at 37◦C for 12 h and centrifuged at 20,000 × g for 10 min to obtain the supernatant fraction. The cellular extract and supernatant fraction of each strain were fractionated separately using centrifugal filters (MWCO: 10 kDa, Merck, Darmstadt, Germany) prior to the determination of the potential AHL-quenching activity. To demonstrate whether the quenching activity arose from enzymatic activity, the cellular extract and supernatant fraction were also heat-inactivated at 100◦C for 30 min and tested for AHL quenching activity. Each sample was mixed with an AHL mixture (0.15 mM C4-HSL, 0.13 mM C6-HSL, 0.11 mM C8-HSL, and 0.08 mM 3-oxo-C12-HSL) and incubated at 37◦C for 18 h for the residual AHL determination. PBS buffer and **AS** were used to replace the cell extract and supernatant to create the negative control.

### Genome Sequencing of AHL-Quenching Strains

Two strains (L11 and S1-5) that exhibit potential enzymatic AHLquenching activity were sequenced for protein-coding genes (CDS) that encode the AHL quenching enzyme. The genomic DNA from L11 and S1-5 was extracted with a QIAGEN Genomic Tips kit (Qiagen, Hilden, Germany). A 20 kb DNA library was prepared for each strain by genomic DNA fragmentation with a Pippin HT size-selection system (Sage Science, Beverly, MA, United States). Single molecular real-time (SMRT) sequencing was performed with a PacBio RS II platform (Pacific Biosciences, Menlo Park, CA, United States). The raw reads were assembled into contigs using a Canu assembler (Koren et al., 2017). The assembled contigs were examined for integrity using dotplots with a GEnome PAir-Rapid Dotter tool (Krumsiek et al., 2007). The assembled contigs were annotated with a RASTtk server (Brettin et al., 2015). The potential AHL quenching ORFs in L11 and S1-5 were screened with NCBI tBLASTn using published AHL quenching genes as reference genes. A structure-based multiple sequence alignment of potential ORFs and published QQ sequences was performed using ESPript (Gouet et al., 1999). Highly homologous ORFs (identify ≥ 25%, coverage ≥ 40%) that share a conserved domain with the reported QQ protein were selected for gene expression evaluation and to test for quenching activity.

### Expression of AHL Quenching Genes in Escherichia coli

Eight potential QQ gene sequences from strains L11 and S1- 5 were amplified with Q5-hot start polymerase (New England BioLabs, Beverly, MA, United States) using the primers listed in **Supplementary Table 1**. The PCR amplicons of the QQ genes were either ligated into a pTrcHis A vector (Thermo Fisher, Waltham, MA, United States) by T4 DNA ligase (Thermo Fisher Waltham, MA, United States) or into pET-20b (+) vectors (Merck, Darmstadt, Germany) by Gibson Assembly Master Mix (New England BioLabs, Beverly, MA, United States). Recombinant pTricHis A and pET-20b(+) vectors with correct insertions were transformed into E. coli TOP10 (Thermo Fisher, Waltham, MA, United States) and E. coli BL21 (DE3) (Merck, Darmstadt, Germany), respectively, for protein expression. The blank vectors pTricHisA and pET-20b(+) were also transformed into E. coli TOP10 and E. coli BL21 (DE3) and denoted as negative recombinant controls (NC-1 and NC-2). The AiiA gene from Bacillus mycoides ATCC 6462 was inserted into pET-20b(+) and expressed in E. coli BL21 (DE3) as a positive control sample for QQ activity evaluation.

Overnight cultures of recombinant strains were diluted 1:100 into fresh LB medium supplemented with 100 µg/mL ampicillin (Sigma, St. Louis, MO, United States). The culture was incubated at 30◦C and 160 rpm until the cells reached exponential growth (OD<sup>600</sup> = 0.8). Subsequently, induction was initiated by adding 0.1 mM isopropyl β-D-1-thiogalactopyranoside (IPTG, Thermo Fisher, Waltham, MA, United States) to the culture prior to further incubation at 30◦C 120 rpm for 16 h. After the induction, the cell pellet was collected and suspended in PBS buffer. The cell suspension was sonicated, and the supernatant was passed through a 0.2 µm filter to remove any cell debris. The crude enzyme was subjected to SDS-PAGE (sodium dodecyl sulfate polyacrylamide gel electrophoresis) analysis. Each target protein band was excised from the SDS-PAGE gel and fragmented into short peptides by in-gel digestion treatment with trypsin (Promega, Madison, WI, United States). The digested peptides were subjected to NanoLC MS/MS analysis using a nanopump UltiMate 3000 ultra-high-performance liquid chromatography (UHPLC) binary HPLC system coupled to a Q-Exactive HF mass spectrometer (Thermo Fisher, Waltham, MA, United States). The target proteins were identified by searching against the Swiss-Prot protein sequence database using the Mascot v 2.6 search engine from Matrix Science.

To evaluate the quenching performance of recombinant QQ enzymes, a reaction system containing 10 µM 3-oxo-C12-HSL, 50 µL of crude enzyme, and PBS buffer was incubated at 37◦C for 3 h. Ten microliters of reaction mixture was tested for residual AHL activity using the A. tumefaciens bioassay. The crude extract of the control strain (NC-1 and NC-2) was used as a negative control. The substrate specificity of the crude enzyme was tested with C4-HSL, C6-HSL, C8-HSL, C10-HSL, C12-HSL, 3-oxo-C6- HSL, 3-oxo-C8-HSL, and 3-oxo-C12-HSL (Sigma, St. Louis, MO, United States). The crude enzyme was mixed with different AHL molecules in PBS buffer, and after the reaction at 37◦C, the residual AHL was determined by bioassay.

## Purification of Recombinant QQ Enzyme

The recombinant strain that was verified to possess AHLquenching activity was induced in 100 mL of LB medium. The crude enzyme was prepared as described above and the recombinant enzyme was purified by affinity chromatography using a 5 mL HisTrap HP column (GE Healthcare, Piscataway, NJ, United States). The column was equilibrated with binding buffer (20 mM sodium phosphate, pH 7.4, 0.5 M NaCl, and 20 mM imidazole). The crude enzyme was diluted with binding buffer and loaded onto the column. The target protein was eluted with 200 mM imidazole; the eluted fractions were pooled and dialyzed in a dialysis tubing (Fisher Scientific, Hampton, NH, United States) with a molecular weight cutoff (MWCO) of 3.5 kDa. The PBS buffer was used as the dialysis medium, and

the entire procedure was performed three times to remove the extra imidazole and salt. The solution was then concentrated using a centrifugal filter with a MWCO of 3.5 kDa (Merck, Darmstadt, Germany).

### Biochemical Characterization of Recombinant QQ Enzymes

The optimal pH of the QQ enzyme was tested by incubating a purified enzyme with 5 µM oxo-C12-HSL in 0.1 M citratephosphate buffer (pH 3.0–8.0) and 0.1 M glycine-NaOH buffer (pH 9.0–10.0) at 37◦C for 1 h. The effect of the salinity on the enzyme activity was evaluated by incubating the enzyme with 5 µM oxo-C12-HSL in PBS buffer (pH 7.4) under different concentrations of sodium chloride (0 M, 0.1 M, 0.5 M, 1 M, 1.5 M, and 2 M) for 1 h. The optimal temperature for the recombinant enzyme activity was determined by mixing the enzyme with 5 µM oxo-C12-HSL under the optimal pH for each enzyme and under different temperatures ranging from 20 to 100◦C for 10 min. The same amount of PBS buffer was used instead of the purified enzyme as a negative control for each test. After the reaction, the residual AHL was quantified using a liquid chromatography-tandem mass spectrometry (LC-MS/MS) system. All the reactions were performed in triplicate. For the biochemical characterization of the NAD(P)-dependent enzymes SDRS1−<sup>5</sup> and SDRL11, 1 mM NADPH or NADH (Sigma, St. Louis, MO, United States) was added to each reaction to facilitate the enzymatic activity.

### AHL Quantification by LC-MS/MS

The AHL molecule oxo-C12-HSL was determined with an Agilent 1260 Infinity quaternary liquid chromatograph (Agilent, Santa Clara, CA, United States) equipped with an Agilent Pursuit C18 column (3 µm particle size, 2.0 × 150 mm) and SCIEX Q-TRAP 5500 mass spectrometer (AB SCIEX, Foster City, CA, United States). The separation was performed at room temperature at a mobile phase flow rate of 250 µL/min using the following gradient elution profile: t = 0 min, 95% solution A (LC-MS grade water with 0.1% formic acid), 5% solution B (LC-MS grade methanol); t = 2 min, 95% solution A, 5% solution B; t = 4 min, 50% solution A, 50% solution B; t = 6 min, 5% solution A, 95% solution B; t = 10 min, 5% solution A, 95% solution B; t = 12 min, 95% solution A, 5% solution B; t = 18 min, 95% solution A, 5% solution B. The detection was performed in positive ion mode using the parameters listed in **Supplementary Table 2**.

### Enzymatic Effect on the Motility and Biofilm Formation of Opportunistic Marine Pathogens

The effect of AiiAS1−<sup>5</sup> on the motility and biofilm formation of Aeromonas hydrophila, Pseudomonas aeruginosa, and Vibrio alginolyticus (**Supplementary Table 3**), was investigated. Five microliters of overnight culture for each marine strain was spotted in the center of the plate containing marine brothsoft agar (0.3% and 0.5% agar) prior to the addition of 5 µL of purified AiiAS1−<sup>5</sup> (0.2 µg). The plate was cultured at 40◦C overnight. The slightly higher mesophilic temperature of 40◦C was chosen to mimic the temperature experienced by bacteria during industrial processes (e.g., in cooling towers, seawater desalination plants and tropical marine aquaculture). The experiment was repeated six times. The migration of the culture was evaluated by calculating the diameters of the haloes around the spotted area. In addition, the effect of the AiiAS1−<sup>5</sup> on the biofilm formation was also evaluated. Single colonies of A. hydrophila, P. aeruginosa, and V. alginolyticus were cultured in **AS** with 2% peptone and incubated at 40◦C overnight. Thereafter, 2 × 10<sup>9</sup> colony forming units (CFU)/mL of cells were mixed with purified AiiAS1−<sup>5</sup> (0, 5, 10, and 20 µg/mL). The cultures were transferred into 96-well microtiter plates and further incubated at 40◦C for 24 h. Thereafter, the supernatants were carefully removed, and the biofilms were stained with 0.2% crystal violet at room temperature for 30 min. Extra dye was removed with **AS** and the stained cells were solubilized with 200 µL of 30% acetic acid (O 0 Toole, 2011). The absorbance of the solution was measured at 595 nm. PBS buffer was used instead of purified AiiAS1−<sup>5</sup> as a negative control.

### Transcription of Virulence Genes in Marine Strains in the Presence of Purified AiiAS1−<sup>5</sup>

A. hydrophila, P. aeruginosa, and V. alginolyticus were grown in marine media at 37◦C and 180 rpm overnight. The overnight culture was diluted 1:100 in **MB** medium as mentioned before (salinity: 36 g/L), and 50 µg/mL of purified AiiAS1−<sup>5</sup> was added to the pathogenic strain culture at the inoculation time. The bacterial cultures were grown at 37◦C and 180 rpm. The growth curve of the marine strain with AiiAS1−<sup>5</sup> or PBS buffer was monitored based on the optical density at 600 nm. The bacterial cultures were sampled at the mid-exponential, late exponential and stationary phases of growth. The samples were used for RNA extraction (RNeasy Mini Kit, Qiagen). cDNA was synthesized with a SuperScript III First-Strand Synthesis Supermix (Thermo Fisher, Waltham, MA, United States). The transcriptional level of each gene (**Supplementary Table 4**) was calculated using the relative standard curve method. To obtain the standard curve, the PCR amplification product of each gene was first cloned into a PCR Blunt II-TOPO vector (Thermo Fisher, Waltham, MA, United States). Plasmid DNA containing the target gene was serially diluted based on the copy number and used as the standard. Real-time PCR was performed on a 7900HT Fast Real-Time PCR system (Thermo Fisher, Waltham, MA, United States). The reaction system contained 5 µL of Fast SYBR Green master mix (Thermo Fisher, Waltham, MA, United States), 0.2 µL each of the forward and reverse primers (10 µM), and 2 µL of cDNA template (2 ng/µL). The quantification of the target gene was normalized to the reference gene (rpoB) in each sample.

### Statistical Analysis

The statistical analysis was performed with Minitab 17. A paired t-test was used to determine the statistical significance of the difference. The difference was defined as statistically significant when P < 0.05.

### RESULTS

fmicb-10-00823 April 15, 2019 Time: 20:6 # 6

### Bacterial Screening From Marine Sources and the AHL Quenching Test

A total of 51 bacterial isolates that can grow at a high salinity (58 g/L) and a high temperature of 50◦C were obtained for QQ screening (**Supplementary Table 5**). The AHL mixture (0.29 mM C4-HSL, 0.25 mM C6-HSL, 0.22 mM C8-HSL and 0.17 mM 3-oxo-C12-HSL) was used to screen the QQ bacteria under high salinity (58 g/L) and a high temperature (50◦C). These AHL quenchers belong to the following three phyla: Firmicutes (Staphylococcus, Bacillus, Halobacillus, Virgibacillus, Pontibacillus, Aquibacillus, and Thalassobacillus), Bacteroidetes (Tamlana, Tenacibaculum, Vibrio, and Mesoflavibacter), and Proteobacteria (Delftia, Bacterioplanes, Altererythrobacter, Devosia, Halomonas, and Pseudoalteromonas). Among the QQ isolates, five showed ≥90% relative AHL QE toward AHL mixtures (**Supplementary Table 1** and **Supplementary Figure 1**). These five QS-quenching bacteria belong to the Pseudomonadaceae (Proteobacteria), Bacillaceae (Firmicutes), and Erythrobacteraceae (Proteobacteria), with three of them belonging to the genus Altererythrobacter in the Erythrobacteraceae (**Table 1**).

### Size Fractionation of the Bacterial Cell Extract and Supernatant Contents to Identify the AHL-Quenching Enzymes

Both the cell extract and supernatant fraction obtained from the five bacterial isolates showed the ability to inhibit AHL (**Figure 1**), with the cell extract having a higher AHL inhibition performance than the supernatant (**Figures 1A,C**). After heat inactivation, the cell extracts of L11 and S1-5 showed decreased AHL inhibition activity, while no decrease in AHL inhibition activity was detected for the other isolates (**Figure 1A**). This result corresponds with the size fraction experiment for the L11 and S1-5 cell extract (**Figure 1B**), in which a higher relative AHL QE was observed in the >10 kDa portion. This size fraction was generally found to contain potential QQ enzymes (Czajkowski and Jafra, 2009). For the other isolates, the relative AHL quenching activity was higher in the <10 kDa portion, suggesting the presence of quorum signal inhibitors (QSIs). The supernatant of the isolates maintained the same AHL quenching activity before and after heat inactivation (**Figure 1C**), and the <10 kDa fractions in the supernatant contributed more to the quenching effect than the larger size fractions (**Figure 1D**). Both the cell extract and the supernatant of the five strains showed similar QQ performance at 50◦C (**Supplementary Figures 2A,C**) except for the presence of QQ activity found in the >3 kDa fraction of the S1-5 supernatant (**Supplementary Figures 2B,D**).

### Genomic Sequencing of the Two Bacterial Isolates S1-5 and L11 That Potentially Possess Enzymatic QQ Activity

One closed contig was assembled from S1-5 raw reads, and it had a sequence length of 3.35 Mbp and a GC content of 66.3%. Two contigs were assembled from the L11 raw reads; both contigs were predicted to be closed, with no gap detected (**Supplementary Figure 3**). Contig 1 has a sequence length of 2.89 Mbp (GC content of 46%) and Contig 2 has a full length of 0.64 Mbp (GC content of 45.2%). The sequencing files were deposited in the European Nucleotide Archive (ENA) under study accession number PRJEB30480. Eight potential QQ enzyme genes from S1-5 and L11 (**Table 2**) were classified as N-acyl homoserine lactonase (AiiA), oxidoreductase (SDR), esterase (Est), hydrolase (Hyd), a penicillin acylase family protein (PvdQ) and a hypothetical protein (HP). These ORFs all shared a conserved motif with the reference AHL-quenching enzymes (**Supplementary Figure 4**).

### Verification of Identified QQ Enzymes for AHL Quenching Activity and Specificity

The identified ORFs were expressed in recombinant Escherichia coli. The crude enzyme and cell lysate from recombinant E. coli, along with the E. coli host that contained blank expression vectors (NC-1 and NC-2), were individually mixed with 5 µM 3-oxo-C12-HSL and tested for residual AHL concentration. Four ORFs, namely, AiiAS1−5, SDRS1−5, Ests1−<sup>5</sup> and SDRL11, were verified to be AHL quenchers. The enzymes showed 26.5%, 40.1%, 13.8%, and 38.0% relative AHL QE compared with the negative controls (**Figure 2**). The positive control made up of AiiA from B. mycoides ATCC 6462 showed a 37.9% relative AHL QE compared with the negative control. The level of activity exhibited by the positive control was comparable to that observed for both SDRs from L11 and S1-5. The AHL specificity test further showed that AiiAS1−<sup>5</sup> showed catalytic activity toward all the tested AHLs, and EstS1−<sup>5</sup> can quench all the AHLs except C4-HSL and C6-HSL. Both SDRS1−<sup>5</sup> and SDRL11 can quench all

TABLE 1 | Marine bacteria that showed 90–100% relative AHL quenching efficiency under saline condition and at high temperature.


the 3-oxo-AHLs. Moreover, SDRL11 showed a quenching effect on C8-HSL (**Supplementary Table 6**).

### Biochemical Characterization of Recombinant QQ Enzymes

AiiAS1−<sup>5</sup> and EstS1−5, SDRS1−<sup>5</sup> and SDRL11 were further induced and purified for biochemical characterization at varying pH values, salt concentrations and temperatures. However, during the course of this 1-month experiment, the purified SDRS1−<sup>5</sup> and SDRL11 did not exhibit good stability while they were in storage compared to AiiAS1−<sup>5</sup> and EstS1−5. The purified SDRS1−<sup>5</sup> and SDRL11 consistently did not show any AHL-quenching activity after storage despite the addition of NADH or NADPH (data not shown). Hence, the biochemical characterization of only AiiAS1−<sup>5</sup> and EstS1−<sup>5</sup> was performed using 3-oxo-C12- HSL as the substrate. AiiAS1−<sup>5</sup> exhibited >25% relative activity against 3-oxo-C12-HSL between pH 7 and 9, with the highest activity observed at pH 8.0 (**Figure 3A**). AiiAS1−<sup>5</sup> showed the maximum relative activity in PBS buffer, but it was still able to retain more than 63% of its activity at 0.1–2 M NaCl concentrations (**Figure 3B**). AiiAS1−<sup>5</sup> showed the highest activity at 0.5 M KCl, and it maintained >70% activity at 0–2 M KCl (**Supplementary Figure 5A**). The relative activity of AiiAS1−<sup>5</sup> increased from 10 to 50◦C (the optimal temperature range) before decreasing rapidly to 7.8%, when the temperature was further increased to 70◦C. The enzyme was fully inactivated at 80◦C (**Figure 3C**). EstS1−<sup>5</sup> exhibited the highest relative activity at pH 9.0, and it retained 26.9% of its activity at pH 10.0 (**Figure 3D**). EstS1−<sup>5</sup> showed optimal activity at 0.1 M NaCl, and it retained more than 83% of its activity at NaCl concentrations between 0 and 0.5 M (**Figure 3E**). The relative activity of EstS1−<sup>5</sup> decreased with the increasing KCl concentration (**Supplementary Figure 5B**). The enzyme showed the highest activity at both 30 and 40◦C, and it was still able to maintain >30% of its relative activity at high temperatures of 50 and 60◦C (**Figure 3F**).

### Effect of Purified AiiAS1−<sup>5</sup> on Marine Bacteria

Purified AiiAS1−<sup>5</sup> was selected for further testing due to its stability, broad substrate specificity and high enzymatic activity in comparison to EstS1−5. The swarming and swimming of A. hydrophila, P. aeruginosa, and V. alginolyticus all decreased in the presence of AiiAS1−<sup>5</sup> compared with the negative control (**Figure 4A** and **Supplementary Figure 6**). Purified AiiAS1−<sup>5</sup> reduced swarming more significantly than swimming. There were observed reductions of 48.9%, 35.4%, and 70.2% in swarming in A. hydrophila, P. aeruginosa, and V. alginolyticus, respectively, compared to their respective controls, which were not exposed to the enzyme. Purified AiiAS1−<sup>5</sup> reduced swimming


TABLE 2|Screening of potential ORFs from strains L11 and S1-5 using the reported AHL quenching enzymes. consistently by only 20% for all three strains compared to the controls (**Figure 4A**). An inhibitory effect on biofilm formation was observed in P. aeruginosa, in which AiiAS1−<sup>5</sup> significantly inhibited biofilm formation by 46.9%, 79.8%, and 77.0% at 5 µg/mL, 10 µg/mL, and 20 µg/mL of enzyme (p < 0.001 compared to the control). The P. aeruginosa biofilm inhibitory effects at 10 µg/mL and 20 µg/mL AiiAS1−<sup>5</sup> were similar (p > 0.05). By contrast, a 32.2% biofilm inhibition in V. alginolyticus was only observed when 20 µg/mL of AiiAS1−<sup>5</sup> was applied (p = 0.04). AiiAS1−<sup>5</sup> has no effect on A. hydrophila at all the tested levels of the purified enzyme (p-value > 0.05 compared to the control, **Figure 4B**).

### Transcription of the Virulence Gene in Marine Strains in the Presence of AiiAS1−<sup>5</sup>

The growth curve of three marine bacteria (A. hydrophila, P. aeruginosa, and V. alginolyticus) showed no difference in the presence or absence of purified AiiAS1−<sup>5</sup> (**Figure 5A**). At midexponential phase, the transcription levels of the QS-mediated global regulators AhyR and LasR for A. hydrophila (**Figure 5B**) and P. aeruginosa (**Figure 5C**) were not significantly inhibited by AiiAS1−<sup>5</sup> (p > 0.05). However, AiiAS1−<sup>5</sup> significantly inhibited AhyR and LasR expression by 56.1% (p < 0.0001) and 69.0% (p = 0.0003) compared to the sample treated with PBS (control) during the late exponential phase. However, the inhibitory effects on both regulators were no longer observed during the stationary phase. The downstream virulence traits AprA, LasB, and ToxA were regulated by the global regulator LsrR, and therefore, they exhibited a similar trend as that observed for LasR in P. aeruginosa (**Figure 5C**). AprA is an exception to this trend, in which an inhibition of 22.5% was still observed at the stationary phase (**Figure 5C**, p = 0.01). By contrast, for V. alginolyticus, an inhibitory effect was only observed on LuxR transcription at the mid-exponential phase and not at the late exponential and stationary phases. This result likewise detrimentally affected the expression of the Pep virulence gene in V. alginolyticus only at the mid-exponential phase (p = 0.02) (**Figure 5D**).

## DISCUSSION

Microbe-associated problems in the marine environment (e.g., seawater desalination plants, seawater cooling towers, and aquaculture) are conventionally mitigated by means of biocides or antibiotics that cause health, safety and environmental concerns. QS inhibitors are increasingly being explored as alternative agents to mitigate these problems (Dobretsov et al., 2011). However, research has shown that bacteria can develop resistance to QS inhibitors (Maeda et al., 2012; García-Contreras et al., 2013), rendering the treatment ineffective over the long term. In comparison, QQ enzymes that degrade extracellular QS signals are viewed as imposing less of a burden on cellular metabolism (Fetzner, 2015), and hence, they minimize the development of resistance toward these greener inhibitory agents. In most instances, QQ enzymes or bacteria with enzymatic activity from marine sources (Huang et al., 2012; Mayer et al.,

fmicb-10-00823 April 15, 2019 Time: 20:6 # 8

Potential QQ ORFs were screened by tBLASTn using known QQ enzymes published as references.

FIGURE 2 | AHL quenching activity comparison of eight potential protein-coding genes expressed in recombinant Escherichia coli cells. The relative AHL quenching efficiency of each sample was calculated based on Eq. (1), as stated in the "Materials and Methods" section. The E. coli TOP10 carrying vector pTricHisA and E. coli BL21 (DE3) carrying pET-20b(+) were used as negative controls. AiiAS1−<sup>5</sup> (acyl homoserine lactonase), SDRS1−<sup>5</sup> (oxidoreductase), EstS1−<sup>5</sup> (esterase), and HydS1−<sup>5</sup> (hydrolase) are possible AHL quenching ORFs from Altererythrobacter sp. S1-5; HydL11 (hydrolase), PvdQL11 (penicillin acylase family protein), SDRL11 (oxidoreductase), and HPL11 (hypothetical protein) are possible AHL quenching ORFs from Pseudoalteromonas sp. L11. PC denotes a positive control, which is obtained by inserting an AiiA gene from Bacillus mycoides ATCC 6462 into pET-20b(+) and expressing it in E. coli BL21 (DE3). Two independent biological replicates with three technical replicates in each biological replicate were performed for this experiment.

FIGURE 3 | Biochemical properties of purified AiiAS1−<sup>5</sup> and EstS1−5. The relative enzyme activity of AiiAS1−<sup>5</sup> in/at different (A) pH buffers (0.1 M citrate-phosphate buffer, pH 3.0–8.0; 0.1 M glycine-NaOH buffer, pH 9.0–10.0), (B) NaCl concentrations (0, 0.1, 0.5, 1, and 2 M), and (C) temperatures of 10–80◦C. The relative enzyme activity of EstS1−<sup>5</sup> in/at different (D) pH buffers, (E) NaCl concentrations, and (F) temperatures. The relative enzyme activity of the purified enzymes was normalized using the activity of the purified enzymes at the optimal pH, salinity, and temperature.

2015; Tang et al., 2015; Liu et al., 2017; Rehman and Leiknes, 2018) were demonstrated for their efficiencies to quench QS in only minimal or nutrient medium, which deviate significantly from conditions in marine industrial systems such as seawater desalination plants (salinity: 46,400 ppm, temperature: 22– 33◦C, and pH 8.1–8.3) (Khawaji et al., 2007) and cooling

seawater (AS) with 2% peptone extracted from casein in 96-well polystyrene plates at 40◦C for 24 h. <sup>∗</sup>P < 0.05; ∗∗P < 0.01; ∗∗∗P < 0.001.

found in Supplementary Table 4. The results are presented as the means ± standard deviation (n = 3). <sup>∗</sup>P < 0.05; ∗∗P < 0.01; ∗∗∗P < 0.001.

towers (e.g., salinity: >35,000 ppm, and temperature 32–48◦C) (Al-Bloushi et al., 2018).

In this study, five bacterial isolates belonging to Pseudoalteromonas, Pontibacillus, and Altererythrobacter demonstrated high QQ activity at a salinity of 58 g/L, 50◦C, and a pH of 7.8. To the best of our knowledge, this is the first report on the existence of QQ activity in Pontibacillus and Altererythrobacter. The enzymatic QQ activity was primarily discovered in the intracellular fraction of Pseudoalteromonas sp. L11 and to a certain extent, it was also found in Altererythrobacter sp. S1-5 (**Figure 1**). So far, many bacteria with either intracellular or extracellular QQ enzymatic activity (Kim et al., 2014; Fetzner, 2015) have been reported, but the extracellular activity is more feasible. For example, Cheong et al. (2013) entrapped the bacteria in a vessel that allows only the extracellular enzymes to pass through and react with AHL in a lab-scale membrane bioreactor used for wastewater treatment. This exclusion eliminated the need for an additional step of lysing the bacterial hosts to retrieve

the intracellular enzymes. However, no similar demonstration has been performed in saline environments.

To look for possible QQ enzymes for marine application purposes, the genes identified in L11 and S1-5 that shared homologies with the existing known QQ genes were first expressed in recombinant E. coli, purified and characterized further. Using this approach, we found four protein coding genes that shared high homology with the existing QQ genes in Altererythrobacter sp. S1-5, but only three demonstrated QQ enzymatic activity when expressed in recombinant E. coli. Similarly, we found four homologous QQ ORFs in L11. Both PvdQL11 and HPL11 shared 24% and 48% of their identity with the reported pfmA and QQ-16d enzymes from the same genera (Weiland-Bräuer et al., 2015; Liu et al., 2017). However, a further examination of each of the individual purified enzymes from L11 showed that only SDRL<sup>11</sup> was positive for AHL quenching activity. This may be because the enzyme from the halophilic microorganism was folded incorrectly or maintained poor stability during recombinant expression in its mesophilic host under low-salt conditions (Madern et al., 2000). Among the three ORFs in Altererythrobacter S1-5 that showed QQ activity, an N-terminal peptide was predicted for both AiiAS1−<sup>5</sup> and EstS1−5, suggesting the possibility that the extracellular enzymes would be feasible for use in practical applications. Most QQ enzymes reported thus far do not have a signal peptide, except for AiiA in the marine organisms Muricauda olearia (Tang et al., 2015) and Erythrobacter flavus (Rehman and Leiknes, 2018) and dlhR in Rhizobium sp. (Krysciak et al., 2011).

After purification, only the AiiAS1−<sup>5</sup> and EstS1−<sup>5</sup> belonging to the AHL lactonase showed a capacity to degrade AHL. To date, the reported AHL lactonases primarily belong to the metallo-β-lactamase superfamily, the phosphotriesterase family and the α/β hydrolase family. The phosphotriesterase family from the archaeon Sulfolobus solfataricus and the crenarchaeon Vulcanisaeta moutnovskia (Hiblot et al., 2015) are the most thermophilic QQ enzyme group discovered so far, and they display activity at temperatures of 85–95◦C (Merone et al., 2005; Hiblot et al., 2012). The Est816 from the α/β hydrolase family screened from the Turban Basin metagenome can degrade AHL at an optimal temperature of 60◦C. Compared to the other two families of AHL lactonases, the metalloβ-lactamase superfamily has a mild optimal catalytic temperature range from 30 to 50◦C (Seo et al., 2011; Cao et al., 2012; Sakr et al., 2013; Pedroza et al., 2014; Vinoj et al., 2014; Tang et al., 2015). The AiiT from the thermophilic bacteria Thermaerobacter marianensis is an exception, in that it exhibited optimal QQ activity at 60–80◦C (Morohoshi et al., 2015). Similarly, the purified AiiAS1−<sup>5</sup> and EstS1−<sup>5</sup> obtained during this study can degrade AHL at temperatures of 50◦C and 30– 40◦C, respectively.

To assess if these enzymes can work under saline conditions, the activity of the two enzymes was tested in different concentrations of NaCl and KCl. The AiiAS1−<sup>5</sup> maintained >63% and >70% relative activity in the presence of 0–2 M NaCl and KCl, respectively (**Figure 3** and **Supplementary Figure 5**), which implied high enzyme robustness under saline conditions. Halophilic proteins usually share common physical features, namely, a high percentage of negatively charged amino acids along the surface area, a high number of salt bridges, low hydrophobicity, and a flexible protein structure (Madern et al., 2000; Takahashi et al., 2018). The 3D structural modeling of AiiAS1−<sup>5</sup> (without a signal sequence) suggested that there was a high percentage of negatively charged amino acids along the surface area (**Supplementary Figure 7**). Twentyfour salt bridges were predicted from AiiAS1−<sup>5</sup> (**Supplementary Table 7**), which is higher than that predicted from the AiiA described in an earlier study (Easwaran et al., 2015). This physical property of AiiAS1−<sup>5</sup> can perhaps account for its ability to exhibit activity under saline conditions, and it is consistent with that of another AiiA (TSAWB) from Bacillus sp. that can tolerate up to a 5% salt solution (Easwaran et al., 2015).

AiiAS1−<sup>5</sup> outperformed the other enzymes discovered in this study, and it was further chosen to demonstrate its QQ efficacy against A. hydrophila, P. aeruginosa, and V. alginolyticus. Both P. aeruginosa and A. hydrophila were previously reported to be commonly associated with fouled membranes in a seawater desalination plant (Hong et al., 2016; Yap et al., 2017; Nagaraj et al., 2018), while V. alginolyticus (Bermont-Bouis et al., 2007) and P. aeruginosa (Hamzah et al., 2014) were reportedly possibly linked to pipeline biocorrosion by seawater. Furthermore, these three bacterial species constitute the dominant pathogenic bacterial group reported in marine aquaculture (Sindermann, 1984). It was observed that AiiAS1−<sup>5</sup> did not impose detrimental effects on cell growth, which reiterates its lower possibility of developing resistance against QQ enzymes over long-term usage. Instead, purified AiiAS1−<sup>5</sup> showed an inhibitory effect against bacterial motility, biofilm formation and virulence transcription in three marine bacteria under saline conditions (salinity: 36 g/L). The biofilm inhibitory effect of AiiAS1−<sup>5</sup> in P. aeruginosa is more significant than that of the other two bacteria. This distinction was likely caused by the substrate affinity of AiiAS1−<sup>5</sup> toward the different AHLs secreted by the tested strains (**Supplementary Table 3**). A 10 µg/mL concentration of AiiAS1−<sup>5</sup> inhibited the biofilm formation of P. aeruginosa to ca. 80% under saline conditions. Similarly, purified AHL-lactonase (100 µg/mL) from Enterobacter aerogenes VT66 inhibited >70% of the P. aeruginosa PAO1 biofilm (Rajesh and Rai, 2015). Labrenzia sp. VG12 cells with AHL lactonase activity also reduced the P. aeruginosa biofilm by 25% (Rehman and Leiknes, 2018), although the exact enzymatic concentration used here was not made known. These earlier studies, along with that reported here, demonstrated the inhibition effect of AHL lactonases on QS-regulated biofilm formation in P. aeruginosa.

In most instances, the transcribed levels of QS-coordinated receptor genes and virulence genes were significantly decreased by AiiAS1−<sup>5</sup> at the late exponential phases of A. hydrophila and P. aeruginosa (**Figure 5**). However, the inhibitory effect was no longer observed at the stationary phase. This observation is consistent with that reported in an earlier study (Park et al., 2005; Sivakumar et al., 2019), and it was probably due

to the accumulation of enzymatic inhibitors in the culture over time, or to a decrease in the metabolic activities of the cultures at the stationary phase. QQ enzymes should therefore be deployed as an environmentally benign approach for controlling microbe-associated problems at the early stages of growth and biofilm formation.

### CONCLUSION

In this study, bacterial species that proliferate under saline conditions were isolated and screened to select the ones that are positive for QQ under high salinity and high temperatures. Further genomic and biochemical characterization revealed a particularly promising AiiA from the Altererythrobacter sp. S1-5 that demonstrated a broad AHL substrate specificity range and enzymatic activity at an optimal pH of 8 and 50◦C. AiiAS1−<sup>5</sup> was also able to maintain good relative QQ activity with increasing salt concentrations up to 2 M. Its QQ efficacy against P. aeruginosa, A. hydrophila, and V. alginolyticus was further demonstrated, in that the motility traits and virulence gene cascades were detrimentally impacted, especially at the mid to late exponential phases of bacterial growth.

These findings collectively suggest that this QQ enzyme would be feasible for use in mitigating membrane biofouling under saline conditions and/or QS-associated pathogenic infections in marine aquaculture.

### REFERENCES


### AUTHOR CONTRIBUTIONS

T-NW designed and performed the experiments, performed the data analysis, and wrote the manuscript with P-YH, who also supervised the research and provided reagents and materials. Q-TG and AP assisted with the genomic DNA sequence assembly. AK helped to write the manuscript.

### FUNDING

This work was funded by the King Abdullah University of Science and Technology (KAUST) Competitive Research Grant 2017 (URF/1/2982-01-01) awarded to P-YH.

### ACKNOWLEDGMENTS

AK thanked for the support by CSIRO Land and Water.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2019.00823/full#supplementary-material




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

Copyright © 2019 Wang, Guan, Pain, Kaksonen and Hong. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Effect of Quercetin Rich Onion Extracts on Bacterial Quorum Sensing

B. X. V. Quecan<sup>1</sup> , J. T. C. Santos<sup>1</sup> , M. L. C. Rivera<sup>1</sup> , N. M. A. Hassimotto<sup>1</sup> , F. A. Almeida<sup>2</sup> and U. M. Pinto<sup>1</sup> \* †

<sup>1</sup> Food Research Center (FoRC), Faculty of Pharmaceutical Sciences, University of São Paulo, São Paulo, Brazil, <sup>2</sup> Department of Nutrition, Federal University of Juiz de Fora, Governador Valadares, Brazil

#### Edited by:

Ana Maria Otero, University of Santiago de Compostela, Spain

#### Reviewed by:

Jin Zhou, Tsinghua University, China Tania Pozzo, University of California, Davis, United States

#### \*Correspondence:

U. M. Pinto uelintonpinto@usp.br; uelintonpinto@gmail.com

#### †Present address:

U. M. Pinto, Harvard Medical School, Massachusetts General Hospital, Boston, MA, United States

#### Specialty section:

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

Received: 04 February 2019 Accepted: 04 April 2019 Published: 24 April 2019

#### Citation:

Quecan BXV, Santos JTC, Rivera MLC, Hassimotto NMA, Almeida FA and Pinto UM (2019) Effect of Quercetin Rich Onion Extracts on Bacterial Quorum Sensing. Front. Microbiol. 10:867. doi: 10.3389/fmicb.2019.00867 Quorum sensing (QS) regulates bacterial gene expression and studies suggest quercetin, a flavonol found in onion, as a QS inhibitor. There are no studies showing the anti-QS activity of plants containing quercetin in its native glycosylated forms. This study aimed to evaluate the antimicrobial and anti-QS potential of organic extracts of onion varieties and its representative phenolic compounds quercetin aglycone and quercetin 3-β-D-glucoside in the QS model bacteria Chromobacterium violaceum ATCC 12472, Pseudomonas aeruginosa PAO1, and Serratia marcescens MG1. Three phenolic extracts were obtained: red onion extract in methanol acidified with 2.5% acetic acid (RO-1), white onion extract in methanol (WO-1) and white onion extract in methanol ammonium (WO-2). Quercetin 4-O-glucoside and quercetin 3,4-O-diglucoside were identified as the predominant compounds in both onion varieties using HPLC-DAD and LC-ESI-MS/MS. However, quercetin aglycone, cyanidin 3-O-glucoside and quercetin glycoside were identified only in RO-1. The three extracts showed minimum inhibitory concentration (MIC) values equal to or above 125 µg/ml of dried extract. Violacein production was significantly reduced by RO-1 and quercetin aglycone, but not by quercetin 3-β-D-glucoside. Motility in P. aeruginosa PAO1 was inhibited by RO-1, while WO-2 inhibited S. marcescens MG1 motility only in high concentration. Quercetin aglycone and quercetin 3-β-D-glucoside were effective at inhibiting motility in P. aeruginosa PAO1 and S. marcescens MG1. Surprisingly, biofilm formation was not affected by any extracts or the quercetins tested at sub-MIC concentrations. In silico studies suggested a better interaction and placement of quercetin aglycone in the structures of the CviR protein of C. violaceum ATCC 12472 than the glycosylated compound which corroborates the better inhibitory effect of the former over violacein production. On the other hand, the two quercetins were well placed in the AHLs binding pockets of the LasR protein of P. aeruginosa PAO1. Overall onion extracts and quercetin presented antimicrobial activity, and interference on QS regulated production of violacein and swarming motility.

Keywords: quorum sensing, antimicrobial activity, onion, quorum quenching, phenolic compounds, glycosylation

### INTRODUCTION

fmicb-10-00867 April 24, 2019 Time: 14:18 # 2

Quorum sensing (QS) is a bacterial communication that uses signaling molecules known as autoinducers that accumulate in the medium according to population density (Fuqua et al., 1994; Whitehead et al., 2001; Lazdunski et al., 2004; Waters and Bassler, 2005). Signaling in Grampositive microorganisms is mediated by low molecular weight peptides known as autoinducer peptides (AIPs) (Miller and Bassler, 2001). Other molecules such as autoinducer-2 (AI-2) are associated with most bacterial species allowing intra and interspecific communication (Miller and Bassler, 2001; Fuqua and Greenberg, 2002; Bai and Rai, 2011). Molecules such as quinolones, diketopiperazines and indole hydroxyketones can also function as communication cues (Waters and Bassler, 2005; Platt and Fuqua, 2010; Papenfort and Bassler, 2016).

In Gram-negative bacteria, signaling is usually mediated by acyl homoserine lactone (AHL) molecules, known as autoinducer-1 (AI-1) (Skandamis and Nychas, 2012). These molecules are composed of a fatty acid chain attached to a lactone ring by an amide bond. The variation that exists between the molecules of AHL occurs both by the size and the composition of the fatty acids that have a variation from 4 to 18 carbons and have some substitutions in the chain (Whitehead et al., 2001; Lazdunski et al., 2004; La Sarre and Federle, 2013). This mechanism was described in the 1970s in two species of bioluminescent marine bacteria: Allivibrio fischeri and Vibrio harveyi (Nealson and Hastings, 1979). In addition to these bacteria, there are other model microorganisms such as Chromobacterium violaceum, Pseudomonas aeruginosa, Agrobacterium tumefaciens, Erwinia carotovora, and Serratia liquefaciens in which QS has been well elucidated (Miller and Bassler, 2001; Waters and Bassler, 2005). There is great interest in these microorganisms as models to study QS, since many of the phenotypes are easily measured and are specifically regulated by QS.

In several bacteria QS regulates a range of phenotypes, coordinating a group behavior that controls the expression of virulence factors, extracellular enzymes, biofilm formation, secondary metabolites, motility, among others (Whitehead et al., 2001; Waters and Bassler, 2005; Skandamis and Nychas, 2012). Many of these phenotypes can impact food spoilage, making the product undesirable or unacceptable for consumption. As an example, the expression of some microbial extracellular enzymes like proteases, pectinases and lipases is regulated by QS (Ammor et al., 2008; Martins et al., 2018). Therefore, researchers have tried to find strategies to disrupt this communication using inhibitory compounds and consequently improve food quality and safety (Bai and Rai, 2011; Skandamis and Nychas, 2012).

Many studies have shown the potential of plant organic extracts rich in phenolic compounds to interfere with QS in different bacteria. These compounds constitute a diverse group of chemical substances, with different chemical activities, important for plant reproduction, growth, and protection against pathogens attack (Martínez et al., 2002). They can be classified depending on the ring number and the type of elements that bind them into phenolic acids, stilbenes, lignans, and flavonoids (Rodrigues et al., 2016).

The last group is an important class of natural products with polyphenol structure, widely found in fruits and vegetables (Panche et al., 2016). Its basic structural feature is the 2 phenyl-benzo-α-pyran compound which consists in two benzene rings (A and B) attached through a heterocyclic pyran ring (C) (Cushnie and Lamb, 2005). There is great interest in flavonoids because of their anti-inflammatory, antimicrobial, antioxidant and antitumor properties, among others (Cushnie and Lamb, 2005; Silveira, 2012; Rodrigues et al., 2016). In addition, flavonoids have also gained importance as potential inhibitors of the QS system. Different flavonoids such as taxifolin, kaempferol, naringenin, apigenin, baicalein, and others have demonstrated their ability to interfere in the QS system of microorganisms such as P. aeruginosa PAO1 and C. violaceum CV026 (Vandeputte et al., 2011), changing the transcription of QS-controlled target promoters and inhibiting the production of virulence factors (Paczkowski et al., 2017).

One of the most representative flavonoids found in high concentrations in foods, especially onion (284–486 mg/kg) is quercetin (Behling et al., 2008). Different studies showed the inhibitory potential of this compound against some microorganisms with phenotypes regulated by QS. A research performed by Gopu et al. (2015) evaluated the ability of quercetin against the QS biosensor strain C. violaceum CV026 and tested the anti-biofilm property of the compound against food-borne pathogens such as Bacillus spp., Pseudomonas spp. Salmonella spp., Campylobacter jejuni, and Yersinia enterocolitica. The results showed that quercetin inhibited violacein production in all the concentrations tested and additionally had a significant reduction of other phenotypes such as biofilm formation, exopolysaccharides, alginate production and motility in the compound's presence (Gopu et al., 2015). Another study showed the effect of quercetin on biofilm formation and virulence factors' production by P. aeruginosa PAO1 (Ouyang et al., 2016). The authors observed that quercetin had a significant inhibition on biofilm formation, pyocyanin, protease and elastase production. It was also observed that the expression of lasI, lasR, rhII, and rhIR genes was significantly reduced in response to quercetin (Ouyang et al., 2016).

Different types of quercetins such as quercetin aglycone, quercetin 4-glucoside, quercetin 3,4-O-diglucoside, quercetin 7,4-diglucoside, quercetin 3-glucosideglucoside and quercetin 5-glucoside are found in onion (Allium cepa Lineu). The anthocyanin cyanidin has also been identified in purple onion cultivars that give reddish or purple coloration to the bulbs (Lombard et al., 2005). The amount of quercetin in onions varies according to the color and type of bulb, being distributed mainly in the skins and outer rings (Arabbi et al., 2004; Lombard et al., 2005; Corzo Martínez et al., 2007; Kwak et al., 2017).

Studies have suggested that quercetin, a flavonol present in high concentrations in onion (Allium cepa), presents anti-QS properties against some Gram-negative microorganisms.

However, there are no studies showing the anti-QS activity of plants containing quercetin in its native glycosylated forms. Thus, the objective of this work was to assess the potential presented by onion extracts to interfere with bacterial cell-tocell communication.

### MATERIALS AND METHODS

### Bacterial Strains and Culture Conditions

The microorganisms used in this work were Chromobacterium violaceum ATCC 12472 (30◦C/24 h), Pseudomonas aeruginosa PAO1 (37◦C/24 h), and Serratia marcescens MG1 (30◦C/24 h). All cultures were grown in Luria Bertani (LB) agar or broth containing peptone 1%, yeast extract 0.5%, sodium chloride 0.5% with 1.2% agar, as needed.

### Preparation, Extraction, and Characterization of Phenolic Compounds of Onion Varieties Preparation of Extracts

The extracts were prepared in the Laboratory of Chemistry, Biochemistry and Molecular Biology of Food in the Faculty of Pharmaceutical Sciences of the University of São Paulo. Samples of 5 kg of white and red onion (Allium cepa) were purchased from Companhia de Entrepostos e Armazéns Gerais de São Paulo (CEAGESP) warehouse. The samples were selected, cut and frozen with liquid nitrogen and stored at −80◦C until use. For the analysis, 20 g of each onion variety were homogenized for 1 min using Ultra-Turrax (Polytron-Kinematica GmbH, Kriens-Luzern, Switzerland) in 100 ml of 70% methanol for white onion and 70% methanol acidified with 5% acetic acid for red onion due to its content of anthocyanins. Then, the samples were vacuum filtered, and the residue was recovered, repeating the process twice using 50 ml of the respective solvent. The obtained extracts were pooled and concentrated in a rotary evaporator (Rotavapor 120, Büchi, Flawil, Switzerland) at a temperature of 40◦C until complete methanol removal, in order to use it for the solid phase separation step.

### Solid Phase Extraction

Methanol free samples were loaded in a column with 1 g of polyamide (CC 6, Macherey-Nagel, Germany), prepared in a syringe of 6 ml and preconditioned passing 20 ml of methanol and 60 ml of distilled water. After application of the white onion extract, the column was washed with 20 ml of water and the elution of the flavonoids was performed with 50 ml of methanol and 50 ml of methanol: ammonium (95.5: 0.5 v/v), named WO-1 and WO-2 extracts, respectively. For red onion, the elution of the flavonoids was performed with 50 ml methanol acidified with 2.5% acetic acid, naming the extract as RO-1. The obtained eluates were completely dried in a rotary evaporator at 40◦C and suspended in 1 ml of methanol. These extracts were used for the identification and quantification of total phenolic compounds using high-performance liquid chromatography with diode array detector (HPLC-DAD) and liquid chromatography-electrospray ionization-tandem mass spectrometry (LC-ESI-MS/MS).

### High-Performance Liquid Chromatography With Diode Array Detector (HPLC-DAD)

Quantification and partial identification of flavonoids were conducted using HPLC-DAD. The chromatograph (Infinity 1120 model, Agilent, Germany) used was equipped with automatic sample injector, quaternary pump and DAD, controlled by Agilent's own software. The column used was Prodigy 5 (ODS3 250 × 4.60 mm, Phenomenex Ltd., United Kingdom) with a flow rate of 1 ml/min, 25◦C. The elution was performed with a solvent gradient with the following elements: A: water with 0.5% formic acid; B: Acetonitrile with 0.5% formic acid. The concentration gradient of the solvents was made with 8% of B at the beginning, 10% in 5 min, 17% in 10 min, 25% in 15 min, 50% in 25 min, 90% in 30 min, 50% in 32 min, and 8% in 35 min (running time, 35 min). The run was monitored with the following wavelengths: 270, 370, and 525 nm and peak identification was performed comparing the retention time and similarity with the absorption spectra of commercial patterns and the spectra contained in the equipment library, previously inserted in the method. The identification was also performed according to the sequence of elution according to Pérez Gregorio et al. (2011). For the quantification the following flavonoid standards were used: quercetin 3-O-glucoside, isorhamnetin and cyanidin 3-O-glucoside (Extrasynthese, Genay, France). All quercetin derivates were quantified and values expressed as quercetin 3-Oglucoside. All isorhamnetin derivates were quantified and values expressed as isorhamnetin equivalent. Cyanidin 3-O-glucoside was quantified, and value expressed as cyanidin 3-O-glucoside.

### Liquid Chromatography-Electrospray Ionization-Tandem Mass Spectrometry (LC-ESI-MS/MS)

The identification of flavonoids and other phenolic compounds was conducted in the liquid chromatography (LC) (Prominence model, Shimadzu, Japan) linked to a mass spectrometer ion trap (Esquire HCT model, Bruker Daltonics, Germany) and electrospray ionization interface (ESI). The separation conditions were the same as those used for the HPLC-DAD, described in section High-Performance Liquid Chromatography With Diode Array Detector (HPLC-DAD). After passage through the DAD, the flow was changed to 0.2 ml/min to the passage in the mass spectrometer. The ESI was maintained in positive mode. The mass detector was programmed to perform full scan between m/z 100–1000. The ionization energy for the positive mode was 3500 V. The identity of the compounds was evaluated by comparing the mass spectrum obtained with the commercial standards and, or literature data (Lee and Mitchell, 2011). To confirm the identity, the HPLC retention time of commercial flavonoid standards (quercetin 3- O-glucoside, quercetin aglycone, and cyanidin 3-O-glucoside) was used for comparison.

## Antimicrobial Activity of the Extracts and Isolated Compounds of Onions

### Minimal Inhibitory Concentration of the Extracts

The minimal inhibitory concentration (MIC) of each extract was determined using the broth microdilution method, according to

the methodology of Wiegand et al. (2008), with modifications. The extracts suspended in LB broth were tested in a 96-well plate. Cultures of C. violaceum ATCC 12472, P. aeruginosa PAO1, and S. marcescens MG1 were grown overnight on plates with LB agar, suspended in saline solution 0.85% and adjusted using a solution of McFarland 0.5 to reach a concentration of approximately 1 × 10<sup>8</sup> CFU/ml. Subsequently, each culture was diluted in LB broth in a proportion of 1:100 and 50 µl of this dilution were placed in each well, to attain the final concentration ranging from 31.2 to 125 µg/ml of extract. The controls were bacterial culture in LB broth without extracts, the broth with each of the extracts in each of the concentrations tested without bacteria, and a sterility control. The QS inhibition tests were prepared with sub-MIC concentrations to ensure that the extracts did not interfere with bacterial growth. Bacterial growth was evaluated following the same procedure for the MIC. Optical density at 595 nm (OD 595 nm) was determined each 3 h during a total time of 24 h using the spectrophotometer (Multiskan FC, Thermo Fisher Scientific, Finland). Quercetin aglycone and quercetin 3-β-D-glucoside were also evaluated as onion's representative isolated compounds.

### Quorum Sensing Modulation Assays by Extracts and Isolated Compounds of Onions

### Violacein Production in C. violaceum ATCC 12472

The test was performed according to Tan et al. (2012, 2013), with modifications. C. violaceum ATCC 12472 was grown overnight following the same parameters as in the section Minimal Inhibitory Concentration of the Extracts. In concentrations ranging from 7.8 to 31.2 µg/ml. A 96-well plate was incubated at 30◦C, 120 rpm for 24 h and then the plates were completely dried at 60◦C. Subsequently, 100 µl of dimethyl sulfoxide (DMSO) were added to each well, keeping the plate with agitation at 120 rpm for 12 h, approximately. The OD 595 nm was measured using the spectrophotometer (Multiskan FC, Thermo Fisher Scientific, Finland). The controls used in the test were the same as in the section Minimal Inhibitory Concentration of the Extracts. Quercetin aglycone and quercetin 3-β-D-glucoside were also evaluated for their anti-QS activity for being representative isolated compounds found in the extracts.

### Swarming Motility by P. aeruginosa PAO1 and S. marcescens MG1

Swarming motility was tested using semi-solid LB medium prepared with 0.5% agar, as described by Oliveira et al. (2016). Aliquots of the extracts giving final concentrations of 31.2, 62.5, and 125 µg/ml were placed in sterile Petri dishes of 49 × 9 mm and then 10 ml of the molten agar were added. For the swarming test 2 µl of the overnight grown bacteria were point inoculated at the center of the agar. Once the inoculum was dried, about 20 min after inoculation, the plates were closed and incubated at 37◦C for 24 h for P. aeruginosa PAO1 and at 30◦C for 24 h for S. marcescens MG1. Inhibition of swarming motility was considered when a visual reduction of the swarm was observed in presence of the extracts. Quercetin aglycone and quercetin 3-β-D-glucoside were also evaluated for their anti-QS activity. Synthetic furanone C-30 (≥97.0% of purity; Sigma-Aldrich, Brazil) (Z-)-4-Bromo-5- (bromomethylene)-2(5H)-furanone, was used as positive control for motility inhibition at 100 µM (Oliveira et al., 2016).

### Biofilm Formation in P. aeruginosa PAO1 and S. marcescens MG1

The effect of onion extracts on biofilm formation was assessed in a 96-well plate as it was described by Borges et al. (2012), with modifications. An aliquot of 20 µl of the overnight cultures adjusted according to McFarland solution 0.5 were inoculated into LB broth with 31.2, 62.5, and 125 µg/ml of extract, completing a final volume of 200 µl. The cultures were incubated at 37◦C for 24 h when using P. aeruginosa PAO1 and at 30◦C for 24 h when evaluating S. marcescens MG1. Thereafter, nonadherent bacteria were removed by washing with 200 µl of saline solution 0.85% and adherent bacteria were fixed with 200 µl of methanol 99% for 15 min, following removal of the solvent. Then 200 µl of crystal violet solution 0.3% (w/v) were added to the well for 5 min. The wells were washed with sterile water to remove excess stain and the crystal violet bound to the biofilm was extracted with glacial acetic acid 33% (v/v). The OD 595 nm of the crystal violet solution was measured using the spectrophotometer (Multiskan FC, Thermo Fisher Scientific, Finland). Quercetin aglycone and quercetin 3-β-D-glucoside were also evaluated.

To confirm biofilm production in the case of the microorganism P. aeruginosa PAO1, a test was performed according to Minei et al. (2008) with modifications. The 96-well plates with cultures grown overnight in LB broth and the different compounds to be tested were incubated, as previously mentioned. Planktonic cells were removed with 200 µl of sterile saline solution 0.85%. Shortly, the adhered cells and biofilm were removed manually scrubbing the walls of each well with a sterile swab until the biofilm was completely removed, and then the swab was transferred to a tube containing 10 ml of saline solution 0.85% and vortexed for 1 min. Serial dilutions were made, and an inoculum of 20 µl was plated using the drop plate method in LB agar following incubation at 37◦C monitoring the plates until the appearance of the micro-colonies. After that, cells were counted, and the results were expressed as Log<sup>10</sup> CFU/biofilm formed into the well.

### Molecular Docking of Quercetin Molecules With CviR and LasR Proteins

Docking studies were performed according to Almeida et al. (2016) and Almeida et al. (2018). In brief, the crystallized structures of CviR protein of C. violaceum ATCC 12472 (PDB: 3QP6 and 3QP8; Chen et al., 2011) and LasR protein of P. aeruginosa PAO1 (PDB: 2UV0, 6D6A, 6D6L, 6D6O, and 6D6P; Bottomley et al., 2007; O'Reilly et al., 2018) with different ligands were obtained in the RCSB Protein Data

Bank database (PDB)<sup>1</sup> . Then, the molecular docking was performed between these proteins and N-(3-hydroxydecanoyl)- DL-homoserine lactone (3-OH-C10-HSL; Pubchem CID: 71353010), N-(3-oxododecanoyl)-L-homoserine lactone (3 oxo-C12-HSL; Pubchem CID: 3246941), quercetin (quercetin aglycone; Pubchem CID: 5280343), quercetin 3,4-O-diglucoside (quercetin 3-β-D-glucoside; Pubchem CID: 5280804) and 4-bromo-5-(bromomethylene)-2(5H)-furanone (Furanone C-30; Pubchem CID: 10131246) using the "Dock Ligands" tool of the CLC Drug Discovery Workbench 4.0 software<sup>2</sup> , with 1000 interactions for each compound and the conformation of the compounds was changed during the docking via rotation around flexible bonds. The generated score mimics the potential energy change when the protein and the compound come together based on hydrogen bonds, metal ions and steric interactions, where lower scores (more negative) correspond to higher binding affinities. The five best scores of the docking of each compound were selected, allowing the inspection of the binding sites of CviR and LasR proteins with each compound (Almeida et al., 2016, 2018).

### Statistical Analysis

All experiments were performed at least three times. The data represent the means of the repetitions and their differences with respect to the controls. All data were subjected to analysis of variance (ANOVA) followed by Tukey's test using the Statistical Analysis System and Genetics Software (Ferreira, 2011). A p < 0.05 was considered to be statistically significant.

### RESULTS

### Characterization of the Phenolic Compounds Present in Onion Samples

Chromatograms obtained by HPLC-DAD of flavonoids from red and white onion extracts are shown in **Figure 1** and their respective identification detailed in **Table 1**. The LC-ESI-MS/MS spectra of chromatographic peaks obtained from red and white onions are also shown in **Supplementary Figure S1**. The major flavonol identified in the two onion varieties was quercetin 4- O-glucoside. The second major flavonol found was identified as quercetin 3,4-O-diglucoside. In addition, an anthocyanin in RO-1 extract was found. Cyanidin 3-O-glucoside, responsible for purple pigmentation of red onion, was presented with molecular ion [M]+ at m/z 449 and characteristic MS2 fragment at m/z 287 ([M]<sup>+</sup> – 162). Concentrations of flavonoids from red and white onion extracts are shown in **Table 2**. Quercetin 4-O-glucoside and quercetin 3,4-O-diglucoside corresponded to 52 and 37% of total flavonoids in the WO-1 extract. The WO-2 extract was primarily composed of quercetin 4-O-glucoside. On the other hand, in RO-1 extract, the main flavonoids were quercetin 4-Oglucoside and quercetin aglycone, representing 57 and 20% of total flavonoids.

### Determination of Minimum Inhibitory Concentration (MIC) and Microbial Growth Curves in the Presence of Extracts and Isolated Compounds of Onions

The MIC results of red and white onion extract of C. violaceum ATCC 12472, P. aeruginosa PAO1, and S. marcescens MG1 are presented in **Table 3**. For QS inhibition experiments we used sub-MIC concentrations that did not affect microbial growth, according to growth curves. For C. violaceum ATCC 12472 both the RO-1 and the WO-2 extract had an inhibitory effect in a concentration of 125 µg/ml. In addition, bacterial multiplication was slightly affected in relation to the control for the two types of onion extracts in the concentration of 62 µg/ml, showing a partial inhibition of the growth. Thus, QS inhibition experiments were performed using concentrations below 62 µg/ml to avoid toxic effects. In the case of P. aeruginosa PAO1 bacteria grew, similarly, to the control in almost all extract concentrations tested. Only the WO-2 extract had a delayed exponential phase, compared to the control at 125 µg/ml. A similar trend was observed for S. marcescens MG1.

Growth curves were also performed in order to check the effect of quercetin aglycone and quercetin 3-β-D-glucoside on the microorganisms used in this study (**Table 3**). These compounds were chosen because the first one was found in the literature as a potential QS inhibitor besides being present in the RO-1 extract and the second as a representative compound of the glycosylated forms of quercetin found in both types of onion in the present study. The MIC for the two compounds was greater than 125 µg/ml. Additionally, for C. violaceum ATCC 12472 a partial inhibition of growth was observed at the concentration of 125 µg/ml of quercetin 3-β-D-glucoside, therefore we used lower concentrations to avoid toxic effects.

### Determination of Anti-QS Activity of Extracts and Isolated Compounds of Onions

### Effect on Violacein Production in C. violaceum ATCC 12472

**Figure 2** shows the effect of white and red onion extracts on violacein production in C. violaceum ATCC 12472. The production of violacein was statistically inhibited in the presence of 31.2 µg/ml of RO-1 extract when compared to the control (p < 0.05). On the other hand, WO-1 and WO-2 extracts did not influence the production of violacein, even though WO-1 presented an inhibitory tendency. It is noteworthy that only the RO-1 extract contains quercetin aglycone (**Table 2**).

As quercetin aglycone and quercetin 3-β-D-glucoside were molecules identified in the extracts and as the aglycone form has been reported as a potential QS inhibitor, the effect of onion extracts was compared to the effect of these two molecules **Figure 2**. For the aglycone form, the results showed that there was a significant inhibition of violacein production (p < 0.05). In contrast, quercetin 3-β-D-glucoside showed no significant inhibition of pigment production, even though a tendency

<sup>1</sup>http://www.rcsb.org/pdb/home/home.do

<sup>2</sup>https://www.qiagenbioinformatics.com/

FIGURE 1 | Chromatogram obtained by HPLC-DAD in wavelengths of 370 (A) and 525 nm (B) of red and white onion. (A) Red onion, (B) White onion, (C) Quercetin 3-O-glucoside standard. Peaks identified: peak 1 – Quercetin 3,4-O-diglucoside; peak 2 – Isorhamnetin 3,4<sup>0</sup> -diglucoside, peak 3 – Quercetin 3-O-glucoside; peak 4 – Quercetin 4<sup>0</sup> -O-glucoside; peak 5 – Isorhamnetin 4- glucoside; peak 6 – Quercetin glycoside; peak 7 – Quercetin aglycone; peak 8 –Cyanidin 3-O glucoside. Identification shown in Table 1.

TABLE 1 | Mass spectra of flavonoids in positive mode from red and white onion extracts obtained by LC-ESI-MS/MS.


RT, retention time. <sup>∗</sup> Identity confirmed with commercial standard.


–, not detected; RO-1, red onion extract in methanol acidified with 2.5% acetic acid; WO-1, white onion extract in methanol; WO-2, white onion extract in methanol ammonium. All quercetin derivates were expressed as equivalents of quercetin 3- O-glucoside. All isorhamnetin derivates were expressed as isorhamnetin. All results were expressed as mg/100 mg of dry extract. In bold, the major compounds identified.


RO-1, red onion extract in methanol acidified with 2.5% acetic acid; WO-1, white onion extract in methanol; WO-2, white onion extract in methanol ammonium.

can be observed, possibly explaining the low anti-QS activity of the extracts that had glycosylated forms of quercetin as major compounds.

### Effect on Swarming Motility of P. aeruginosa PAO1 and S. marcescens MG1

The results of the effect of onion extracts on swarming motility of P. aeruginosa PAO1 are shown in **Figure 3A**. The RO-1 extract significantly reduced motility in the tested concentrations, and the control with furanone C-30 demonstrated the best phenotype inhibition. The other extracts did not present significant inhibition in this assay.

For S. marcescens MG1, inhibition of swarming motility was clearly observed at the concentration of 125 µg/ml of the WO-2 extract **Figure 3B**. The other extracts showed no significant inhibition, despite a trend observed in higher concentrations. Additionally, our assays with quercetin aglycone and quercetin 3-β-D-glucoside revealed a significant inhibition of swarming motility in both bacteria (p < 0.05) (**Figures 3A,B**). The violacein production tests showed that quercetin aglycone had better inhibitory activity than the glycosylated quercetin in C. violaceum ATCC12472. However, the results from the swarming motility assay showed that both types of quercetin were able to inhibit bacterial motility on agar plates.

### Effect on Biofilm Formation of P. aeruginosa PAO1 and S. marcescens MG1

Biofilm production was not significantly inhibited by any of the extracts as shown in **Figure 4A** for P. aeruginosa PAO1 and **Figure 4C** for S. marcescens MG1. Tests with quercetin aglycone and quercetin 3-β-D-glucoside in P. aeruginosa PAO1 showed inhibition at some concentrations, but paradoxically no inhibition was observed in the highest concentration tested. As

this result was somewhat contradictory, we decided to use an additional technique to measure biofilm formation by counting viable cells recovered from the biofilms. However, the results of these counts did not reveal any inhibition of biofilm formation in any of the tested concentrations (**Figure 4B**), meaning that there was no difference in the counts of viable cells recovered from the biofilms at different concentrations of both types of quercetins. **Figure 4C**, shows the results of biofilm formation of S. marcescens MG1, revealing little to no inhibition by the tested molecules.

### Molecular Docking of Quercetin Molecules With CviR and LasR Protein

All the evaluated compounds were able to bind in the evaluated structures of the CviR and LasR proteins and the binding affinities and binding residues are shown in **Tables 4**, **5**. The 3-OH-C10- HSL presented the highest binding affinities for the two structures of CviR protein of C. violaceum ATCC 12472, 3QP6, and 3QP8. The quercetin aglycone presented lower binding affinities than this AHL and higher than quercetin 3-β-D-glucoside and furanone C-30 (**Table 4**). The quercetin aglycone bound to M135 and S155 residues from the two structures of the CviR protein evaluated, differently from quercetin 3-β-D-glucoside that bound at different sites (**Table 4** and **Figure 5**). In addition, the S155 residue was a common binding site for quercetin aglycone and 3-OH-C10-HSL in these structures (**Table 4** and **Figures 5A,B**). In 3QP8 structure, the flavonoid structure of quercetin 3-β-Dglucoside bound in the S89 residue and glucoside structure in residue Y88, N92, and A94 (**Table 4**). On the other hand, the different structures of the LasR protein of P. aeruginosa PAO1 showed variations of the binding affinities for the four evaluated compounds. The 3-oxo-C12-HSL showed the highest binding affinities for the structures 2UV0 and 6D6A and the quercetin 3-β-D-glucoside the highest binding affinities for the structures 6D6L, 6D6O, and 6D6P (**Table 5**). The T75, T115, and S129 residues were common binding sites for the two quercetins, 3 oxo-C12-HSL and furanone C-30 (**Table 5** and **Figure 6**). These residues were also common binding sites for flavonoid structure of quercetin 3-β-D-glucoside (**Table 5**). On the other hand, the glucoside structure of this quercetin was able to bind in specific amino acid residues, such as G38, Y47, Y64, V76, L125, and A127 (**Table 5**). However, the inspection of the binding sites of CviR and LasR protein with these compounds showed that

TABLE 4 | Results from molecular docking of structures of CviR protein of C. violaceum ATCC 12472 with selected compounds.


Amino acid residues that bind only to the glucoside structure of quercetin 3-β-D-glucoside are in bold.

quercetin 3-β-D-glucoside was unable to accommodate in the pocket of the two structures of CviR protein of C. violaceum ATCC 12472 (**Figures 5**, **6**).

microorganisms. In the present study, we evaluated the effect of onion organic extracts and representative isolated compounds in QS model bacteria.

### DISCUSSION

Different cellular functions that affect food spoilage are influenced by signaling molecules accumulated as a function of QS. Consequently, many researchers have attempted to find alternatives that can inhibit this communication using natural sources that may reduce the virulence capacity of

First, we identified the different phenolic compounds present in the organic extracts. The results showed that different types of glycosylated quercetin were found in both onion varieties. Studies have shown that flavonoids such as quercetin 4-O-glucoside and quercetin 3,4-O-diglucoside are the major compounds found in onion and compounds derived from kaempferol and isorhamnetin were identified as minor flavonoids (Slimestad et al., 2007; Lee et al., 2011; Pérez Gregorio et al., 2011). In addition, cyanidin 3-O-glucoside was the


5|ResultsfrommoleculardockingofstructuresofLasRproteinofP.aeruginosaPAO1withselectedcompounds. main anthocyanin present in red onion (Pérez Gregorio et al., 2011). Another study by Arabbi et al. (2004) has also shown that significant amounts of quercetin aglycone are found in concentrations of 48–56 mg/100 g on white onions and amounts of 38–94 mg/100 g in red onions. In addition, they reported that anthocyanin cyanidin was found in varieties of red onion contributing with 9.2% of total flavonoids (Arabbi et al., 2004).

The MIC of the extracts was equal to or greater than 125 µg/ml of dry extract for all the evaluated microorganisms. These results are related to those of Gopu et al. (2015), in which quercetin aglycone, one of the compounds found in the present study, showed a MIC value of 120 µg/ml for C. violaceum CV026. Another study by Al-Yousef et al. (2017) evaluated the ethyl acetate fraction of onion peel and its major compound quercetin 4-O-β-D-glucopyranoside as a possible QS inhibitor, finding a MIC value for C. violaceum ATCC 12472 of 500 µg/ml (Al-Yousef et al., 2017). Thus, it is possible that higher concentrations of onion extracts and quercetin are needed in order to fully inhibit the growth of the bacteria evaluated in the present study.

Quorum sensing regulates violacein production, a characteristic violet pigment produced by C. violaceum, which is induced by some types of N-acyl homoserine lactone molecules (Stauff and Bassler, 2011). Curiously, the strain used in the present work, C. violaceum ATCC 12472, is induced by N-(3-hydroxydecanoyl)-L-homoserine lactone, differing from the biosensor strain C. violaceum CV026 which is induced by N-hexanoyl-L-homoserine lactone (Morohoshi et al., 2008). We observed that neither WO-1 and WO-2 extracts nor quercetin 3-β-D-glucoside significantly inhibited violacein production, even though a tendency for an inhibitory effect can be observed **(Figure 2**). On the other hand, the RO-1 extract and quercetin aglycone significantly inhibited violacein production.

Several studies indicated violacein inhibition by different extracts. For instance, Oliveira et al. (2016) showed that the phenolic extract of Rubus rosaefolius (wild strawberry) reduced violacein production by up to 88%, especially in the concentration of 118.60 mg GAE/L, showing a higher inhibition than furanone, positive control for this experiment, which inhibited 68.6%. The same authors found that the enriched extract in phenolic compounds of Malpighia emarginata (acerola) significantly inhibited violacein production in all sub-MIC concentrations evaluated (Oliveira et al., 2017). Rodrigues et al. (2016) also showed that the phenolic extract of Eugenia brasiliensis (grumixama) presented a significant inhibition of violacein production in C. violaceum ATCC 6357. However, in these studies, no identification of which phenolic compounds specifically inhibited the phenotypes was performed. Therefore, the compounds that inhibited violacein production are likely different from those of the present study. In a work by Song et al. (2018), coral symbiotic bacteria were screened for their ability to inhibit violacein production in C. violaceum ATCC 12472, with 15% of the isolates presenting QS inhibition. Furthermore, the authors showed that rhodamine isothiocyanate which is produced by one of the isolates characterized as Vibrio alginolyticus was involved in the disruption of QS in P. aeruginosa PAO1.

fmicb-10-00867 April 24, 2019 Time: 14:18 # 10

FIGURE 5 | Molecular docking of 3QP8 structure of CviR protein of C. violaceum ATCC 12472 with 3-OH-C10-HSL, quercetin aglycone, quercetin 3-β-D-glucoside and furanone C-30. (A–D) surface representation of 3QP8 structure of CviR protein of C. violaceum ATCC 12472, (E–H) surface and backbone representations and (I–L) backbone representation with hydrogen bond between the amino acid residues and compounds evaluated. Gray surface representation, CviR protein; Red surface representation, 3-OH-C10-HSL; Yellow surface representation, quercetin aglycone; Green surface representation, quercetin 3-β-D-glucoside; Purple surface representation, furanone C-30; Gray backbone representation, CviR protein; Black arrow indicates the binding site; Yellow arrow, 3-OH-C10-HSL or quercetin aglycone or quercetin 3-β-D-glucoside or furanone C-30; Blue dashed line, hydrogen bond.

For quercetin aglycone, the results showed that there was a significant inhibition of violacein production (p < 0.05). The results were comparable to those found by Gopu et al. (2015), who reported that in the presence of quercetin aglycone violacein production in C. violaceum CV026 was inhibited by up to 83.2% in a concentration of 80 µg/ml. We found inhibition in a concentration ranging from 15.6 to 125 µg/ml, even though we used a different strain of C. violaceum ATCC 12472. In contrast, quercetin 3-β-D-glucoside showed no significant inhibition of pigment production, possibly explaining the low anti-QS activity of the extracts that had glycosylated forms of quercetin as major compounds. This result indicates that the glycosylation of the molecule, or even other types of changes in the structure, could modify the antimicrobial and anti-QS activity of a phenolic compound. This hypothesis is supported by other studies reporting that changes in the flavone structure influence the biological activity of flavonoids (Xiao et al., 2014; Paczkowski et al., 2017; Xiao, 2017).

We have also performed molecular docking of quercetin aglycone and quercetin 3-β-D-glucoside with the QS transcription activator CviR protein of C. violaceum ATCC 12472 (**Figure 5**). The quercetin aglycone accommodates in the structure of the protein in a similar fashion to the autoinducer 3-OH-C10-HSL (**Figures 5A,B**), while the glycosylated quercetin presents an overall different molecular interaction with the protein, as observed by a larger portion of the bulky glycosylated quercetin molecule being exposed to the exterior of the structure (**Figure 5C**). In addition, the two quercetins bound at different amino acid residues, as well as the glucoside structure of quercetin 3-β-D-glucoside bound to other specific residues (**Figures 5B,C**). However, only quercetin aglycone and 3- OH-C10-HSL showed common binding site, suggesting that these compounds can compete to bind to the CviR protein of C. violaceum ATCC 12472 (**Figures 5A,B**).

The effect of the extracts on swarming motility was also evaluated. The expression of some virulence factors such as biofilm formation is associated with motility (Al-Yousef et al., 2017). Therefore, interferences in this phenotype can affect a microorganism's pathogenicity. We observed for P. aeruginosa PAO1 a motility inhibition by RO-1 in all tested concentrations. On the other hand, S. marcescens MG1 swarming was inhibited by WO-2 only in the concentration of 125 µg/ml. Furanone C-30 was used as a positive control for swarming motility inhibition in the concentration of 100 µM and, as expected, presented the best phenotype inhibition, corroborating previous findings (Manefield et al., 2002; Hentzer et al., 2003). The other extracts did not present significant inhibition of the phenotype. These results are related to those obtained by Husain et al. (2015) in which the essential oil of Mentha piperita inhibited the swarming motility of P. aeruginosa PAO1. Vattem et al. (2007) also evaluated the effect of sub-lethal concentrations of phytochemicals of common fruits, herbs and spice extracts, demonstrating that they decreased P. aeruginosa PAO1 swarming motility by approximately 50% (Vattem et al., 2007). This behavior was also replicated in other bacteria, as in the study of Oliveira et al. (2016), which demonstrated that wild strawberry phenolic extract inhibited the swarming motility of a strain of S. marcescens and A. hydrophila, two bacteria found in refrigerated food products, besides inhibiting the production of prodigiosin, a red pigment found in S. marcescens, regulated by QS (Oliveira et al., 2016).

Our results also revealed that the two types of quercetin showed swarming inhibitory activity in P. aeruginosa PAO1 and S. marcescens MG1. Molecular docking of these quercetins with LasR protein of P. aeruginosa PAO1 revealed that they all could interact and accommodate in the pocket of the different structures of this protein (**Figures 6B,C**). The two quercetins and 3-oxo-C12-HSL, an autoinducer synthesized by P. aeruginosa PAO1, showed common binding sites (**Figures 6A– C**). In quercetin 3-β-D-glucoside, the flavonoid structure bound to these common amino acid residues, unlike its glucoside structure (**Table 5**). These results suggest that these quercetins could compete with autoinducer to bind to the LasR protein of P. aeruginosa PAO1. Our results confirm a previous docking study performed by Gopu et al. (2015) with quercetin aglycone and LasR protein. Biochemical studies such as those performed by Paczkowski et al. (2017) with purified LasR protein and different types of quercetin molecules should be performed in order to confirm these findings.

It is important to highlight that no study has evaluated in detail the effects of red and white onion extracts on QS regulated phenotypes. In a work by Rasmussen et al. (2005), libraries of plant extracts and isolated chemical compounds were created to evaluate which of these had anti-QS activity, using a selection system of QS inhibitors called QSIS. They evaluated extracts of spring and brown onion but there was no apparent QS inhibition in their assays. The answer to the absence of inhibition may be related to the extracts concentration, identity of the extracted compounds, extraction method and the systems used to detect the anti-QS activity. It could also point to the fact that their extract could be enriched in glycosylated phenolic compounds which we suggest have lower anti-QS activity.

Finally, we analyzed the effect of the extracts on biofilm production. Biofilms are known as microbial communities that adhere to surfaces and are protected by an adherent polymeric matrix (Hentzer et al., 2002). Studies have shown that QS communication plays an important role in the maturation process of these cellular aggregates (Hammer and Bassler, 2003;

FIGURE 6 | Molecular docking of 2UV0 structure of LasR protein of P. aeruginosa PAO1 with 3-oxo-C12-HSL, quercetin aglycone, quercetin 3-β-D-glucoside and furanone C-30. (A–D) surface representation of 2UV0 structure of LasR protein of P. aeruginosa PAO1, (E–H) surface and backbone representations and (I–L) backbone representation with hydrogen bond between the amino acid residues and compounds evaluated. Gray surface representation, LasR protein; Red surface representation, 3-oxo-C12-HSL; Yellow surface representation, quercetin aglycone; Green surface representation, quercetin 3-β-D-glucoside; Purple surface representation, furanone C-30; Gray backbone representation, LasR protein; Black arrow indicates the binding site; Yellow arrow, 3-oxo-C12-HSL or quercetin aglycone or quercetin 3-β-D-glucoside or furanone C-30; Blue dashed line, hydrogen bond.

Morohoshi et al., 2007). Our experiments indicate that the extracts did not inhibit biofilm formation by P. aeruginosa PAO1 and S. marcescens MG1 at any given concentration (**Figures 4A,C**). However, experiments with quercetin and quercetin 3-β-D-glucoside showed conflicting results. For instance, the crystal violet assay suggested inhibition at concentration of 31.2 and 62.5 µg/ml of the glycosylated quercetin, and with the motility assay we also observed inhibition by both types of quercetins in the two evaluated microorganisms. But in the case of P. aeruginosa PAO1 there was no inhibition of biofilm at the concentration of 125 µg/ml (**Figure 4A**). We attempted to confirm these results by counting viable cells recovered from biofilms under these conditions, but no inhibition was observed at any of the tested concentrations (**Figure 4B**). Overall, these results suggest that neither the extracts, nor the quercetins evaluated in this study presented potential to inhibit biofilm formation at concentrations that supposedly inhibit QS. Therefore, we encourage the use of different methods in order to confirm a possible QS inhibiting candidate.

Generally, our results disagree with those of Al-Yousef et al. (2017). These authors evaluated the ethyl acetate fraction of onion peel and its major compound quercetin 4-O-β-Dglucopyranoside and found an inhibition of up to 64% in biofilm formation by P. aeruginosa. Interestingly, they used higher concentrations of extract and quercetin (up to 400 µg/ml) than in the present study where the highest concentration tested in QS inhibition experiments was 125 µg/ml. Additionally, it is important to note that in the Al-Yousef study, the major compound identified was quercetin 4-O-β-D-glucopyranoside, which in its structure has a sugar molecule attached to C4 of B ring in the flavone group, while the majority of glycosylated compounds present in the onion extracts evaluated in the present work have a sugar moiety attached to C3, taking into account that these structural changes would interfere with the effectiveness of the compound as an inhibitor.

Ouyang et al. (2016) evaluated the effect of quercetin aglycone on biofilm formation of P. aeruginosa PAO1 in sub-MIC concentrations. Their results showed that quercetin aglycone inhibited biofilm formation in concentrations ranging from 8 to 64 µg/ml. The concentration of 16 µg/ml had the best inhibitory effect (around 50% inhibition), similarly to azithromycin at 32 µg/ml, an antibiotic used in clinical treatments (Ouyang et al., 2016). Interestingly, when comparing their data to the results of the present study, inhibition was not consistent. For instance, concentrations of 32 and 64 µg/ml of quercetin presented less inhibition than 16 µg/ml. The reasons for these inconsistent results could be related to the hormesis effect (Calabrese, 2008), even though more studies are needed to confirm such hypothesis.

Another research conducted by Paczkowski et al. (2017) showed that different flavonoids, including quercetin, could specifically inhibit QS in P. aeruginosa PAO1. Structure-activity analyses demonstrated that the presence of two hydroxyl groups on the A ring of flavone, one at position 7 and at least one at any other position, are required for a potent inhibition of LasR/RhIR receptors. In addition, the authors have shown that rings B and C can accommodate many substitutions, with exception of methyl groups in B ring, which are not tolerated, because they are too bulky. Biochemical analysis revealed that flavonoids function in a non-competitive way to prevent binding of QS receptors to DNA, altering the transcription of QS-controlled target promoters and suppressing the production of virulence factors (Paczkowski et al., 2017). However, the study evaluated non-glycosylated compounds; therefore, the question whether these modifications would work for compounds that have sugar substitutions attached to any of the flavonoid rings still remains.

Overall, quercetin aglycone had better inhibitory activity over QS in C. violaceum ATCC 12472 than quercetin 3-β-D-glucoside. On the other hand, the two quercetins inhibited motility of P. aeruginosa PAO1 and S. marcescens MG1. The results of the biological and in silico analyses confirmed that the structure of the compounds interferes with anti-QS activity indicating that the low activities of organic extracts of onion varieties may be related to the glycosylation of phenolic compounds. In addition, the response of the tested bacteria may be different as a function of the amino acid variations and structures of the receptor QS proteins. It would be interesting to test different phenolic compounds with or without modifications and in different bacteria to strengthen these findings. Xiao et al. (2014) and Xiao (2017) demonstrated that glycosylation generally reduces the bioactivity of flavonoids. This phenomenon has been observed for different properties including antioxidant, anti-inflammatory, antibacterial and antifungal activities.

### CONCLUSION

The effect of organic extracts of red and white onion and their major constituents on QS controlled phenotypes has been evaluated. Different glycosylated quercetins were found in both onion varieties. The onion organic extracts showed inhibition of violacein production and swarming motility. Violacein production was significantly inhibited by quercetin aglycone, while glycosylated and quercetin aglycone inhibited motility of P. aeruginosa PAO1 and S. marcescens MG1. In silico studies suggested a better interaction and accommodation of quercetin aglycone in the structures of the CviR protein of C. violaceum ATCC 12472 than the glycosylated compound.

On the other hand, the two quercetins were able to bind and accommodate in the pocket of LasR protein of P. aeruginosa PAO1. Surprisingly, biofilm formation was not affected by any extracts or the quercetins tested in this study. These results suggest that the extracts and isolated compounds of onions could interfere in the antimicrobial and anti-QS activity, but interference on QS was limited to violacein production and swarming motility.

### AUTHOR CONTRIBUTIONS

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

### FUNDING

This research was funded by a grant from CNPq-Brazil (457794/2014-3) and supported by the Food Research Center – FoRC (2013/07914-8).

### REFERENCES


### ACKNOWLEDGMENTS

UP acknowledges a grant from CNPq-Brazil (457794/2014- 3). We thank the São Paulo Research Foundation (FAPESP) for financial support to the Food Research Center – FoRC (2013/07914-8). BQ and MR thank CNPq-Brazil for providing scholarships. We acknowledge the CLC bio of the QIAGEN Company which licensed the CLC Drug Discovery Workbench 4.0 software.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2019.00867/full#supplementary-material

FIGURE S1 | LC(ESI)-MS/MS spectra of chromatographic peaks obtained from red and white onions. (A) Peak 1 – Quercetin 3,4-O-diglucoside; (B) peak 2 – Isorhamnetin 3,4<sup>0</sup> -diglucoside; (C) peak 3 – Quercetin 3-O-glucoside; (D) peak 4 – Quercetin 4<sup>0</sup> -O-glucoside; (E) peak 5 – Isorhamnetin4-glucoside; (F) peak 6 –Quercetin glycoside; (G) peak 7 – Quercetin aglycone, and (H) peak 8 – Cyanidin 3-O-glucoside in positive ion mode.


evaluation, and long-term storage stability. J. Agric. Food Chem. 59, 857–863. doi: 10.1021/jf1033587


of a novel genetic system, the QSI selector. J. Bacteriol. 187, 1799–1814. doi: 10.1128/JB.187.5.1799-1814.2005


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

Copyright © 2019 Quecan, Santos, Rivera, Hassimotto, Almeida and Pinto. 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.

# Relationship Between Quorum Sensing and Secretion Systems

Rocio Trastoy Pena<sup>1</sup>† , Lucia Blasco<sup>1</sup>† , Antón Ambroa<sup>1</sup> , Bertha González-Pedrajo<sup>2</sup> , Laura Fernández-García<sup>1</sup> , Maria López<sup>1</sup> , Ines Bleriot<sup>1</sup> , German Bou<sup>1</sup> , Rodolfo García-Contreras<sup>3</sup> \* † , Thomas Keith Wood<sup>4</sup> and Maria Tomás<sup>1</sup> \* †

<sup>1</sup> Deapartamento de Microbiología y Parasitología, Complejo Hospitalario Universitario A Coruña (CHUAC), Instituto de Investigación Biomédica (INIBIC), Universidad de A Coruña (UDC), A Coruña, Spain, <sup>2</sup> Departamento de Genética Molecular, Instituto de Fisiología Celular, Universidad Nacional Autónoma de México, Mexico City, Mexico, <sup>3</sup> Departamento de Microbiología y Parasitología, Facultad de Medicina, Universidad Nacional Autónoma de México, Mexico City, Mexico, <sup>4</sup> Department of Chemical Engineering, Pennsylvania State University, University Park, PA, United States

#### Edited by:

Tom Defoirdt, Ghent University, Belgium

#### Reviewed by:

Romé Voulhoux, UMR7255 Laboratoire d'Ingénierie des Systèmes Macromoléculaires (LISM), France Jin Zhou, Tsinghua University, China

#### \*Correspondence:

Rodolfo García-Contreras rgarc@bq.unam.mx Maria Tomás MA.del.Mar.Tomas.Carmona@ sergas.es †These authors have contributed equally to this work

#### Specialty section:

This article was submitted to Microbial Physiology and Metabolism, a section of the journal Frontiers in Microbiology

> Received: 20 February 2019 Accepted: 30 April 2019 Published: 07 June 2019

#### Citation:

Pena RT, Blasco L, Ambroa A, González-Pedrajo B, Fernández-García L, López M, Bleriot I, Bou G, García-Contreras R, Wood TK and Tomás M (2019) Relationship Between Quorum Sensing and Secretion Systems. Front. Microbiol. 10:1100. doi: 10.3389/fmicb.2019.01100 Quorum sensing (QS) is a communication mechanism between bacteria that allows specific processes to be controlled, such as biofilm formation, virulence factor expression, production of secondary metabolites and stress adaptation mechanisms such as bacterial competition systems including secretion systems (SS). These SS have an important role in bacterial communication. SS are ubiquitous; they are present in both Gram-negative and Gram-positive bacteria and in Mycobacterium sp. To date, 8 types of SS have been described (T1SS, T2SS, T3SS, T4SS, T5SS, T6SS, T7SS, and T9SS). They have global functions such as the transport of proteases, lipases, adhesins, heme-binding proteins, and amidases, and specific functions such as the synthesis of proteins in host cells, adaptation to the environment, the secretion of effectors to establish an infectious niche, transfer, absorption and release of DNA, translocation of effector proteins or DNA and autotransporter secretion. All of these functions can contribute to virulence and pathogenesis. In this review, we describe the known types of SS and discuss the ones that have been shown to be regulated by QS. Due to the large amount of information about this topic in some pathogens, we focus mainly on Pseudomonas aeruginosa and Vibrio spp.

#### Keywords: quorum, secretion, virulence, motility, competence

### INTRODUCTION

Microorganisms coexist in competitive environments with other species, and they must develop different survival strategies to compete for space, nutrients and ecological niches. Bacteria have developed several molecular mechanisms that enable them to survive under stress conditions in different environments. The general stress response (RpoS) (Battesti et al., 2011), tolerance to reactive oxygen species (ROS) (Zhao and Drlica, 2014; Van den Bergh et al., 2017), energy metabolism (cytochrome bd complex) (Korshunov and Imlay, 2010) and Tau metabolism (Javaux et al., 2007), drug efflux pumps (Blanco et al., 2016), SOS response (Baharoglu and Mazel, 2014), (p)ppGpp signaling under starvation conditions (Hauryliuk et al., 2015), toxin-antitoxin (TA) systems (Wood et al., 2013) and quorum sensing (QS), which we will discuss in detail in this review, are the main molecular mechanisms of tolerance and bacterial persistence (Harms et al., 2016; Trastoy et al., 2018).

Quorum sensing acts by monitoring cell density through chemical signals that allow communication between bacteria in order to regulate the expression of genes involved in virulence, competition, pathogenicity and resistance (Nealson et al., 1970; Hawver et al., 2016; Paul et al., 2018). In general, QS systems are species-dependent and contribute to processes such as cell maintenance, biofilm formation and horizontal gene transfer. QS also plays a role in other events involving the synchronization of the whole population such as antibiotic production (Abisado et al., 2018), natural competence (Shanker and Federle, 2017), sporulation (Rai et al., 2015) and the expression of secretion systems (SS). In this review, we will focus on the relationship between QS networks and SS in two important bacterial pathogens Pseudomonas aeruginosa and Vibrio spp.

### QS NETWORK

To explain the structure and functioning of the QS network, we will focus on Gram-negative bacteria, in which the signaling pathways are better described. In general terms, QS systems are composed of synthase proteins that produce QS signals, QS signals, and response regulators that bind QS signals and reprogram gene expression (Ng and Bassler, 2009). N-acyl homoserine lactones (AHLs) are the most common QS signals in Gram-negative bacteria (Geske et al., 2008). Other QS signals include autoinducer-2 (AI-2) in Vibrio harveyi (Surette et al., 1999), PQS (Pseudomonas quinolone signal) (Pesci et al., 1999), DSF (diffusible signaling factor) in Xanthomonas campestris (Barber et al., 1997), indole in Escherichia coli (Lee and Lee, 2010), and PAME (hydroxyl-palmitic acid methyl ester) in Ralstonia solanacearum (Flavier et al., 1997). The LuxI/LuxR QS system of Vibrio fischeri is the prototypical model system for Gram-negative bacteria (Engebrecht et al., 1983; Engebrecht and Silverman, 1984). Homologs of luxI (which encode synthase proteins) and luxR (which encode response regulators) are present in many bacteria (Case et al., 2008). AHL signals are produced inside the cell and most of them are transported freely to the local environment. When the concentration of AHL reaches a certain level outside of the cell, the molecule re-enters the cell (or binds surface receptors) and binds/activates the LuxR-type receptor to alter gene expression. AHL signals with small structural differences are involved in the process of gene regulation (Fuqua et al., 1994; Whiteley et al., 2017; Paul et al., 2018).

Pseudomonas aeruginosa possesses three well-known QS systems: LasI/LasR, RhlI/RhlR, and PQS (Pseudomonas quinolone signal)/PqsR (MvfR). The Las system consists of LasI, a synthase protein which produces the AHL N-(3-oxododecanoyl)-L-homoserine lactone (3O-C12-HSL), and LasR, the transcriptional regulator (Seed et al., 1995; Stintzi et al., 1998; Kariminik et al., 2017). Likewise, the RhlI/RhlR system produces the N-hexanoyl-L-homoserine lactone (C4-HSL) signal and the RhlR transcriptional regulator. Finally, the PQS system comprises 2-heptyl-3-hydroxy-4(1H)-quinolone (PQS signal) and the PqsR (MvfR) receptor (Xiao et al., 2006; Jimenez et al., 2012). In 2016, James and collaborators, analyzed the role of a new binding receptor for PQS signals, i.e., MexG, an inner membrane protein of the mexGHI-opmD operon and a component of a resistance-nodulation-cell division (RND) efflux pump (Hodgkinson et al., 2016).

Quorum quenching (QQ) enzymes have also been shown to be important in the functioning of QS systems (Zhang and Dong, 2004; Dong et al., 2007; Bzdrenga et al., 2017). Our research group has recently described a new QQ enzyme (AidA) which participates in the QS network in Acinetobacter baumannii clinical strains (Lopez et al., 2017b, 2018).

### SECRETION SYSTEMS

Bacterial pathogens secrete proteins through their cell membranes in a fundamental process that enables them to attack other microorganisms, evade the host immune system, produce tissue damage and invade the host cells. Secreted proteins can act as virulence factors that generate toxic products to the host cells and may also facilitate adhesion to these cells. Translocation of proteins across the phospholipid membranes is carried out by several types of SS (Green and Mecsas, 2016). SS play a significant role in bacterial communication. To date, 8 types of SS (T1SS, T2SS, T3SS, T4SS, T5SS, T6SS, T7SS, and T9SS) have been made defined on their structure, composition and activity (**Figure 1**). These differences can be attributed to the differences between Gram-negative and Gram-positive bacteria (Desvaux et al., 2009; Sato et al., 2010; Costa et al., 2015). The characteristics of each type of SS are described in detail below.

### T1SS

The type I secretion system is widely distributed in Gram-negative bacteria such as P. aeruginosa, Salmonella enterica, Neisseria meningitidis, and E. coli (Thomas et al., 2014).

The type I secretion system (T1SS), which has three structural elements (ABC transporter protein, a membrane fusion protein and an outer membrane factor), can transfer substrates across both bacterial membranes in Gram negative bacteria in a one-step process (Green and Mecsas, 2016). T1SS uses proteins as substrates, e.g., proteases and lipases of different sizes and with different functions; these proteins have a C-terminal uncleaved secretion signal which is recognized by the ABC transporter protein to form the translocation complex (Delepelaire, 2004; Kanonenberg et al., 2013).

There are two systems described so far that regulate the expression and secretion of substrates of T1SS, the Has system of S. marcescens and P. aeruginosa, and the hemolysins of Vibrio cholerae, N. meningitidis and in particular of uropathogenic E. coli (Thomas et al., 2014).

### T2SS

The type II secretion system (T2SS), which is conserved in most Gram negative bacteria, is responsible for secreting folded proteins from the periplasm. These proteins are first transported through the IM by the general secretory (Sec) or twin-arginine translocation (Tat) pathways, and then secreted from the periplasm into the extracellular medium by the T2SS (Nivaskumar and Francetic, 2014; Green and Mecsas, 2016).

The Sec pathway consists of three structural parts: a protein targeting component, a motor protein and a membrane integrated conducting channel called SecYEG translocase. This mechanism transports unfolded proteins with a hydrophobic sequence at the N-terminus. Moreover, the secreted protein either remains in the periplasm or is transported to the extracellular space. The proteins may contain a SecB-specific signal sequence for transport to the periplasm or the extracellular milieu; however, if it has the signal recognition particle (SRP) signal it can follow the SRP pathway and remain in the inner membrane (Green and Mecsas, 2016; Tsirigotaki et al., 2017).

By contrast, the Tat secretion pathway consists of 2–3 subunits, TatA and TatB, which form a unique multifunctional protein in Gram-positive bacteria, and TatC. This mechanism translocates folded proteins with a twin-arginine motif. In Gram-positive bacteria, most proteins are transported out of the cell, while in Gram-negative bacteria the protein can remain in the periplasm or it can be translocated to the extracellular space by the T2SS (Patel et al., 2014; Green and Mecsas, 2016).

The T2SS, a complex structure composed of 15 proteins, named general secretion pathway proteins (Gsp) in E. coli (Korotkov et al., 2012), Eps in V. cholera (Abendroth et al., 2009; Sloup et al., 2017) and Xcp in P. aeruginosa (Filloux et al., 1998; Robert et al., 2005), has a wide range of substrates with diverse functions, although all share one feature, an N-terminal signal which enables them pass to the periplasm via the Sec or Tat secretion mechanisms (Nivaskumar and Francetic, 2014; Green and Mecsas, 2016).

The main function of the T2SS is to acquire nutrients (Nivaskumar and Francetic, 2014). It is responsible for secreting numerous exoproteins, most of which are hydrolytic enzymes and other proteins such as toxins, adhesins and cytochromes that have various roles in respiration, biofilm formation and motility (Nivaskumar and Francetic, 2014). The T2SS has been described in various environmental strains and also human pathogens such as V. cholera (Overbye et al., 1993), P. aeruginosa, Aeromonas sp. and enterotoxigenic Escherichia coli (ETEC) (Nivaskumar and Francetic, 2014).

### T3SS

The type III secretion system (T3SS) or injectisome, is a double-membrane-embedded apparatus found in multiple pathogenic Gram-negative bacteria such as Salmonella spp., Yersinia spp., enteropathogenic and enterohemorrhagic E. coli, Shigella spp. and Pseudomonas spp. (Cornelis, 2006; Gaytan et al., 2016; Deng et al., 2017). This complex nanomachine promotes the transfer of virulence proteins called effectors from the bacterial cytoplasm into the eukaryotic cell in a single step (Galan and Waksman, 2018).

The T3SS is composed of approximately 25 proteins assembled in three main structures: the basal body, a set of rings spanning the two membranes of the bacterium; a hollow needle-shaped component through which the semi-unfolded effectors are transported (these first two structures are collectively called "needle complex"); and the translocon, made up of a hydrophilic protein that serves as a scaffold for forming a translocation pore, constituted by two hydrophobic proteins, which is inserted into the host cell membrane and through which effectors are directly translocated. A unique set of effectors is delivered by each pathogen, which subverts specific host-cell signaling pathways to allow bacterial colonization (Izore et al., 2011; Notti and Stebbins, 2016; Deng et al., 2017).

The export apparatus associated with the basal body is formed by five poly topic inner membrane proteins that are essential

for substrate secretion. This protein complex, together with a cytoplasmic sorting platform and the ATPase complex are responsible for substrate recruitment and classification, and for energizing the secretion process enabling chaperone-effector dissociation and protein unfolding for initial entry into the T3SS central channel that serves as the secretion pathway. These components are highly conserved between different T3SS systems and with the flagella, which is evolutionarily related to the injectisome (Abby and Rocha, 2012; Galan and Waksman, 2018; Lara-Tejero and Galan, 2019).

Several effectors of T3SS have been described such as ExoS, ExoT, ExoU, and ExoY in P. aeruginosa; Tir and EspE in E. coli and YopE, YopH, YopM, YopJ/P, YopO/YpkA, and YopT in Yersinia sp. (Cornelis and Van Gijsegem, 2000).

### T4SS

The type IV secretion system family is found in Gram-negative and Gram-positive bacteria as well as in Archaea. T4SS is the most cosmopolitan secretion system and differs from other SS as it is able to transfer DNA in addition to proteins (Cascales and Christie, 2003). More specifically, T4SS is capable of performing contact-dependent secretion of effector molecules into eukaryotic cells, conjugative transfer of mobile DNA elements and also exchange of DNA without any contact with the outside of the cell (Green and Mecsas, 2016; Grohmann et al., 2018). T4SS can be divided on the basis of its functionality into two subfamilies: conjugation systems and effector translocators. Conjugation systems are responsible for the transfer of antibiotic resistance genes and virulence determinants among bacteria. The effector translocators introduce virulence factors into the host cell (Christie, 2016). However, in Gram-negative bacteria T4SS has been divided into two different subfamilies: IVA and IVB. The E. coli conjugation apparatuses and VirB/D system of Agrobacterium tumefaciens are the models used to study the structure of type IVA of T4SS (Grohmann et al., 2018). The VirB/D apparatus consists of 12 proteins which form a complex envelope-spanning structure that facilitate the translocation function. Two of these proteins, VirB2 and VirB5, make up the pilus, while another three proteins act as ATPases, and VirB1 is a lytic transglycosylase (Costa et al., 2015; Green and Mecsas, 2016). The Legionella pneumophila Dot/Icm (Defective for organelle trafficking/Intracellular multiplication) system is the model used to study the IVB subfamily of T4SS (Nagai and Kubori, 2011; Grohmann et al., 2018).

### T5SS

The type V secretion system is unique because its substrates transport themselves across the outer membrane. The substrates use the Sec translocase to pass through the inner membrane to the periplasm space. Various different types of T5SS have been identified: autotransporters (T5aSS), two-partner passenger-translocators (T5bSS), trimeric autotransporters (T5cSS), hybrid autotransporters (T5dSS) and inverted autotransporters (T5eSS) (Henderson et al., 2004; Leo et al., 2012; Rojas-Lopez et al., 2017). In general, the T5SS transports proteins across the asymmetric outer membrane (OM) that contains lipopolysaccharides, through their own C-terminal translocation domain that inserts into the OM as a β-barrel to complete the secretion of the N-terminal passenger domain via the barrel pore. Several periplasmic chaperones also participate in transport through the OM, specifically the β-barrel assembly machinery (BAM complex) and the translocation and assembly module (TAM complex) facilitate protein secretion (Rojas-Lopez et al., 2017).

A T5SS has been described in human pathogens such as Bordetella pertussis and Haemophilus influenzae, which have two-partner SS and uropathogenic E. coli, which has chaperone-usher systems (Costa et al., 2015; Green and Mecsas, 2016).

YadA of Yersinia enterocolitica and SadA of Salmonella are T5SS type c (Leo et al., 2012). Intimin of E. coli and invasin of enteropathogenic Yersinia spp. are type Ve SS (Leo et al., 2012).

A self-transporter (T5aSS) (Wilhelm et al., 2007) and three T5bSS: LepA /LepB system (Kida et al., 2008), the CupB system (Ruer et al., 2008) and PdtA/PdtB system (Faure et al., 2014), have been reported in P. aeruginosa. In B. cenocepacia, four T5SS (Holden et al., 2009) have been found, two with pertactin domains and two with haemagglutinin autotransporters; this last type is also present in S. maltophilia (Ryan et al., 2009).

### T6SS

The type VI secretion system is widely represented in Gram-negative bacteria (Coulthurst, 2013; Gallique et al., 2017b). T6SS is an integrated secretion device within the membrane and it transfers substrates, which are toxic effectors to eukaryotic (Pukatzki et al., 2007) and prokaryotic cells (Russell et al., 2014). It plays a crucial role in the pathogenesis and competition among bacteria (Ho et al., 2014; Zoued et al., 2014; Costa et al., 2015; Gallique et al., 2017a). The origin of T6SS is related to bacteriophages (Leiman et al., 2009). T6SS is a huge apparatus and consists of 13 core components organized into a trans-membrane complex, a baseplate-like structure at the cytoplasmic face of the inner membrane, and a sheathed inner tube, which is the effector delivery module that is ejected to the target cell. The tube-sheath complex is assembled from the baseplate in the cytoplasm and the hollow tube is built from hexamers of the hemolysin co-regulated protein (Hcp). The sheath contracts and pushes the tube with the associated effectors into targeted cells, using a puncturing mechanism similar to the one used by the contractile tails of phages (Russell et al., 2011, 2014; Cianfanelli et al., 2016; Green and Mecsas, 2016; Galan and Waksman, 2018).

### T7SS

Type VII secretory system has been described in some Gram-positive bacteria such as Staphylococcus aureus and in species of Mycobacterium and Corynebacterium. This SS was reported for the first time in 2003 in Mycobacterium tuberculosis and it was called ESX-1 (Stanley et al., 2003), which is an important virulence factor in M. tuberculosis. To date, five T7SS have been identified in Mycobacterium sp. but the transport mechanisms across the mycobacterial membrane are almost unknown (Costa et al., 2015; Ates et al., 2016; Green and Mecsas, 2016).

Most of the substrates of T7SS belong to EscAB clan which includes six protein families: Esx, PE, PPE, LXG, DUF2563,

and DUF2580. ESAT-6 is a M. tuberculosis protein which belongs to Esx family and which is secreted with EsxB (CFP-10) (Ates et al., 2016).

### T9SS

The type IX secretion system (T9SS) or Por secretion system (PorSS) is the most recently discovered system (Lasica et al., 2017). Its function is to transport molecules across the outer membrane. Its substrates must include a Sec signal, which allows transfer of proteins through the inner membrane with the aid of the Sec system. The T9SS system has been described in almost all members of the phylum Bacteroidetes, but it has mainly been studied in oral pathogens such as Porphyromonas gingivalis and Tannerella forsythia. In P. gigivalis, the T9SS system consists of 16 proteins with structural and functional activity, and another two proteins involved in the regulation of the transport process (Sato et al., 2010; Lasica et al., 2017).

### REGULATION OF SECRETION SYSTEMS BY QUORUM SENSING NETWORKS (TABLE 1)

### Pseudomonas aeruginosa T1SS

Transcriptional studies in P. aeruginosa suggest that in this bacterium T1SS is positively regulated by QS, since the expression of its effector, the alkaline protease AprA, depends on QS. In addition, the genes of the AprA inhibitor aprI and the structural genes aprDEF also appear to be positively regulated by QS (Hentzer et al., 2003; Schuster et al., 2003; Wagner et al., 2003).

### T2SS

Three T2SS systems, the Xcp, Hxc and Txc systems, have been described in P. aeruginosa. The first of these, Xcp, secretes the QS regulated virulence factors elastase A and B (LasA and LasB) as well as the exotoxin A (ExoA) and it is itself positively regulated by QS (**Figure 2**). Accordingly, recently it was demonstrated by ChIPseq analysis that MvfR (the receptor of the PQS autoinducer) is able to directly bind xcpQ-xcpP-xcpR regions and this is related to their induction in the presence of MvfR (Maura et al., 2016).

The second T2SS, Hxc, is regulated by the availability of phosphate and secretes LapA a low-molecular weight alkaline phosphatase (Wagner et al., 2003; Michel et al., 2007). Two genes, xphA and xqhA, which encode the PaQa subunit of the Xcp functional hybrid system, have been described. These genes, which are located outside the xcp locus, are regulated by environmental conditions but not by QS, in contrast to what occurs with the rest of the Xcp system (Michel et al., 2007, 2011). In contrast to the first two systems, the third system Txc has just recently been described and so far only identified in a region of genome plasticity of the strain PA7; it is regulated by a two component system (TtsSR) and secretes the chitin binding protein CpbE (Cadoret et al., 2014).

### T3SS

Current evidence suggest that as in Vibrio spp., QS in P. aeruginosa negatively regulates the expression of T3SS, specifically the RhlI/RhlR system, as transcription of the T3SS genes and secretion of ExoS increase significantly in a rhlI mutant and return to basal levels on the addition of exogenous C4-HSL (Bleves et al., 2005; Kong et al., 2009; **Figure 2**). In agreement, the expression of exoS is also negatively regulated by QS, specifically by the RhlI/RhlR system, as well as by the stationary phase sigma factor RpoS (Hogardt et al., 2004).

The fact that the T3SS genes do not appear to be repressed by QS in some global transcriptomic studies with mutants may be explained by the presence of high calcium concentrations in the media, or by the lack of resolution of DNA microarrays (Hentzer et al., 2003; Schuster et al., 2003). More striking is the fact that some QS inhibitors like 6-gingerol and coumarin inhibit rather than increase the expression of T3SS (Zhang et al., 2018). Nevertheless, these studies were done in the presence of high calcium, and QS-independent inhibition of T3SS has not been ruled out. Moreover, a recent study in the PA01 strain, using a lasR rhlR double mutant, demonstrated that it remains virulent in a murine abscess model, despite that it does not produce QSdependent virulence factors and that the secretion of ExoT and ExoS is fully functional in this mutant. Hence the authors hypothesized that T3SS is the cause of the remaining virulence (Soto-Aceves et al., 2019).

The P. aeruginosa QS network and its T3SS are also related by the fact that VqsM, an AraC-family transcription factor, binds to both the promoter region of lasI and the promoter of exsA, which encodes a master regulator of the T3SS, regulating both mechanisms (Liang et al., 2014; **Figure 2**).

### T6SS

The T6SS is involved in iron transport, and a connection has been observed between T6SS and QS through the TseF protein, which is a substrate of T6SS and interacts with PQS (Lin et al., 2017; **Figure 2**).

In P. aeruginosa, three loci which encode T6SS have been found to be regulated by QS proteins (LasR and MvfR) (Lesic et al., 2009). Expression of the second loci, H2-T6SS, is regulated by the Las and Rhl QS systems in PAO1 strains (Sana et al., 2012; **Figure 2**) and by the direct binding of MvfR in PA14 (Maura et al., 2016).

## Vibrio sp.

### T2SS

The formation of biofilms has multifactorial regulation in V. cholerae as in other pathogens. The QS network controls directly biofilm production which is related to type II secretion system in V. cholerae (Teschler et al., 2015). Several proteins such as RbmA, RbmC and Bap1, which are involved in the formation of biofilms, are transported by T2SS. In addition, mutant strains with inactivated T2SS have reduced biofilm formation (Johnson et al., 2014; Teschler et al., 2015).


#### T3SS

In V. parahaemolyticus and V. harveyi (unlike in E. coli), both the HAI-1 and AI-2 QS systems inhibit the expression of T3SS genes (Henke and Bassler, 2004). QS also represses T3SS during V. harveyi infections of gnotobiotic brine shrimp (Ruwandeepika et al., 2015). Waters et al. (2010) have described the regulatory pathway by which QS controls T3SS. At low cell density when LuxR is repressed, which entails the derepression of two promoters of the exsBA operon and the exsA operon, ExsA activates the expression of genes that encode the structural proteins of the type III secretion system. However, when the cell density is high, LuxR directly represses transcription of the PB promoter, preventing the production of ExsA and consequently decreasing the expression of structural genes of T3SS (Waters et al., 2010; Ball et al., 2017). OpaR inhibits the T3SS1 in V. parahaemolyticus which is the most important factor in its cytotoxicity (Gode-Potratz and McCarter, 2011).

#### T6SS

Several researchers have demonstrated the regulation of T6SS by QS networks in Vibrio spp. We present the main findings in this field here. In V. alginolyticus, activation of T6SS and the QS network has been found to be coordinated by the serine/threonine kinase PpkA cascade (Yang et al., 2018). PpkA2 is autophosphorylation and it transfers the phosphate group to VstR. Phosphorylated VstR promotes the expression of both of the T6SS in V. alginolyticus through the inhibition of LuxO activity, which acts to impede the expression of LuxR, a promoter of the T6SS. LuxR inhibits the expression of the first T6SS (T6SS1) or promotes the expression of the second T6SS (T6SS2) (Yang et al., 2018).

At low cell population density, LuxO is phosphorylated, which activates the expression of specific small regulatory RNAs (sRNAs) in conjunction with alternative sigma factor σ 54 (Sheng et al., 2012). sRNAs inhibit the expression of LuxR with the help of RNA chaperone Hfq (Liu et al., 2011). However, at high cell population density, LuxO is dephosphorylated turning off the transcription of the sRNAs and allowing the translation of LuxR (Waters and Bassler, 2005; Milton, 2006). Sheng et al. (2012) also demonstrated that the expression of the hcp T6SS gene is growth phase-dependent and the QS regulators controls the haemolysin co-regulated protein, which is one of the main proteins of the T6SS functioning as an effector of the system and/or an effector binding protein (**Figure 2**). The phosphatase PppA also acts on the QS (modulating the transcription of LuxR) and the expression and secretion of hcp1 and hcp2 (Sheng et al., 2013). It is important to highlight that PppA permits the crosstalk between the two T6SS in V. alginolyticus (Sheng et al., 2013).

Rpo N (σ <sup>54</sup>) collaborates with QS in the regulation of T6SS genes. It is involved in the regulation of the expression of hcp and vgrG3 operons that encode T6SS secreted molecules, but does not control the genes that encode the structural and sheath components of T6SS (Ishikawa et al., 2009; Dong and Mekalanos, 2012).

There are a few more studies in V. cholerae related to this topic than other species. Two QS autoinducers, CAI-I (cholerae autoinducer) and AI-2 (autoinducer-2), co-operate to control the gene expression depending on the cell density

(Ng and Bassler, 2009). Two enzymes are necessary for the biosynthesis of these autoinducers: CqsA and LuxS, respectively (Schauder et al., 2001; Miller et al., 2002; Chen et al., 2002; Higgins et al., 2007). These signal molecules are detected by two sensor kinases, LuxQ (sensor of CAI-I) and CqsS (sensor of AI-2). Both pathways merge on LuxU, a phosphotranfer protein. At low cell density (LCD), the two sensor kinases phosphorylate LuxU due to the absence of their respective autoinducers. There are two histidine kinases which also contribute to the phosphorylation of LuxU: VpsS and CqsR (Jung et al., 2015). Then, LuxU transfers the phosphorylate group to a DNA-binding response regulator protein called LuxO. Phosphorylated LuxO activates the expression of sRNA molecules (known as qrr1- 4) when the cell density is low thanks to the interaction with the alternative sigma factor σ <sup>54</sup> (Freeman et al., 2000; Lenz et al., 2004). In conjunction with the RNA-binding protein Hfq, LuxO represses the expression of HapR (Lenz et al., 2004), a TetR-family global transcriptional regulator which acts on QstR (Tsou et al., 2009; Shao and Bassler, 2014; Watve et al., 2015; **Figure 2**). HapR is accumulated when the cell density is high (Lenz et al., 2004) because LuxO is not phosphorylated and transcription of the sRNAs is blocked. QstR is a master regulator of the T6SS belonging to the LuxR-type family of regulators (Jaskólska et al., 2018). QstR binds to the promoter region of the T6SS cluster inducing the expression of the genes. The regulation of the T6SS by cAMP-CRP pathway is not clear, but it is possible that it influences T6SS genes through regulation of QS and chitin-induced competency (Liang et al., 2007; Blokesch, 2012). It is known that CRP positively regulates T6SS (Ishikawa et al., 2009). Apart from the activation of QstR via QS, it is also regulated by chitin and arabinose (Lo Scrudato and Blokesch, 2012, 2013).

The expression of the three T6SS gene clusters in V. cholerae requires TfoX, CytR, HapR, and QstR for the highest level of expression (Watve et al., 2015). CytR and TfoX are required for the expression of the T6SS genes but their regulatory effects are only mediated by QstR (**Figure 2**).

### Other Pathogens T1SS

The swr QS system, which controls swarming motility, regulates the Lip secretion system, a T1SS responsible for the secretion of lipases, metalloproteases and S-layer proteins in Serratia liquefaciens MG1 (Riedel et al., 2001). The swr QS system consists of SwrI, which synthesizes C4-HSL, and SwrR, which regulates gene transcription after binding the diffusible signal C4-HSL.

QS-mediated regulation of lipB, which encodes the LipB exporter, was demonstrated in swrI mutants with luxAB insertions, in which the level of secreted proteins was lower (Riedel et al., 2001; **Figure 2**). Other relationships between T1SS and QS have also been observed. The rice pathogen recognition XA21 receptor recognizes a sulphated peptide (axY<sup>S</sup> 22) derived from the Ax21 protein (activator of XA21-mediated immunity) and confers resistance to Xanthomonas oryzae strains. Ax21 may have a key biological role because it is conserved in Xanthomonas spp., Xylella fastidiosa, and Stenotrophomonas maltophilia. Ax21 requires RaxABC TOSS (type I secretion system) for secretion and activity. The expression of rax genes which encode T1SS has been demonstrated to be QS-dependent due to the cell-density dependency (Han et al., 2011). These data indicate that Ax21 could have a role as a signaling molecule and a direct relationship between the QS network and T1SS is established (**Figure 2**; Lee et al., 2006).

### T2SS

In Xanthomonas species, QS is mediated by the diffusible signal factor (DSF). A proteomic analysis conducted in 2013 revealed 33 proteins that are controlled by DSF. Their putative functions are associated with QS and include cellular processes, intermediary metabolism, oxidative adaption, macromolecule metabolism, cell-structure, protein catabolism, and hypothetical functions (Qian et al., 2013). In this study, it was observed that three genes encoding T2SS-dependent proteins and one gene which encodes Ax21 (activator of XA21-mediated immunity)-like protein are regulated by QS and are essential for virulence-associated functions, including extracellular protease, cell motility, antioxidative ability, extracellular polysaccharide biosynthesis (EPS), colonization, and biofilm (Qian et al., 2013; **Figure 2**).

### T3SS

The relationship between QS and T3SS in E. coli was first demonstrated by Sperandio et al. (1999), who showed that expression of the locus of enterocyte effacement (LEE) operons that encode the T3SS is activated by QS in both enterohemorrhagic (EHEC) and enteropathogenic (EPEC) E. coli due to transcriptional control of the LEE operons by LuxS, which directly activates the LEE1 and LEE2 operons and indirectly activates (via the Ler regulator) the LEE3 and tir operons (**Figure 2**). These researchers proposed that activation of the T3SS by the AI-2 autoinducer synthesized by commensal E. coli resident in the large intestine could explain the high infectivity of E. coli O157: H7, which has an infectious dose of about 50 bacterial cells (Sperandio et al., 1999).

The major virulence factors of EHEC and EPEC are intimin (T5eSS), Tir (the receptor for intimin) and the three secreted proteins EspA, EspB and EspD. T3SS functions in the secretion of the Tir and Esp proteins. The LuxR-type response regulator SdiA negatively regulates the expression of EspD and intimin in the same bacterium, indicating multifactorial regulation of the T3SS by bacterial QS signals (Kanamaru et al., 2000).

Indole, which is produced by tryptophanase (TnA) in enteric bacteria and reaches high concentrations in the gut, is another signaling molecule that influences expression of T3SS in E. coli (Lee et al., 2007, 2008). Indole increases the production and secretion of the translocators EspA and EspB in EHEC O157:H7 (Hirakawa et al., 2009; **Figure 2**); hence, indole promotes the development of attaching and effacing (A/E) lesions in HeLa cells.

The involvement of the RNA chaperone protein Hfq, which also participates in QS, in T3SS expression was demonstrated in Yersinia pseudotuberculosis and Yersinia pestis (Schiano et al., 2014; **Figure 2**). Moreover, Schiano et al., 2014 have demonstrated the regulation of T3SS by QS through virulence regulators LcrF and YmoA in Y. pseudotuberculosis (Amy, 2018).

In Aeromonas hydrophila, an unique QS system, encoded in ahyR/ahyI loci, has been described (Vilches et al., 2009; Garde et al., 2010). Vilches et al. (2009) have used the A. hydrophila AH-3 strain to study the T3SS regulation. AH-3: ahyI and AH-3: ahyR mutants have reduced activity of the aopN-aopB promoter (promoter of T3SS components) compared to the wild-type strain (**Figure 2**). So they concluded that QS could be involved in the positive regulation of the production of the T3SS component in the AH-3 strain (Vilches et al., 2009).

### T4SS

In Brucella abortus, there is a clear relationship between the QS network and T4SS. For the virB operon, which encodes the T4SS regulated by VjbR, a LuxR-type QS is responsible for the virulence characteristics of B. abortus (Li et al., 2017). The virB operon is responsible for establishing the replicative niche of the bacterium once it enters the host cell. The T4SS in B. abortus, as in other bacteria, translocates effector proteins into the host cell to avoid the immune defense mechanisms, making it one of the two main virulence factors for Brucella. Arocena et al. (2010) described the binding site of VjbR to the virB operon (Li et al., 2017). Otherwise, the conjugation process between two members of the Roseobacter group mediated by T4SS, encoded in RepABC-type plasmids, is controlled by the QS network. This was demonstrated by construction of luxI mutant and the addition of external long chain AHLs, which restored the phenotype (Patzelt et al., 2013, 2016; **Figure 2**).

### T6SS

Quorum sensing has been reported to control expression of T6SS toxin-immunity systems in Burkholderia thailandensis. Moreover, a new role for T6SS in constraining the proliferation of QS mutants has been described in B. thailandensis (Majerczyk et al., 2016). Interestingly, it has been observed that T6SS effectors function as cell-to-cell signals in a Pseudomonas fluorescens MFE01 strain lacking the AHL QS pathway (Gallique et al., 2017b).

In A. hydrophila, Hcp and VgrG- two of the "core" proteins and also effectors of the T6SS- secretion have been suggested to be regulated by the AhyRI QS regulon (Khajanchi et al., 2009; **Figure 2**). Finally, our research group has described the association between T6SS machinery and the activation of the QS system by bile salts in A. baumannii clinical strains (Lopez et al., 2017a; **Figure 2**).

### T7SS

As with other bacteria, Mycobacterium spp regulate biofilm formation by QS (Virmani et al., 2018). The second messenger c-di-cGMP, an intracellular signaling molecule, coordinates

biofilm production and QS signaling (Sharma et al., 2014). Both M. tuberculosis (Kulka et al., 2012) and diverse species of non-tuberculous mycobacteria (M. smegmatis, M. marinum, M. fortuitum, M. chelonae, M. ulcerans, M. abscessus, M. avium, and M. bovis) produce biofilm depending on certain environmental conditions such as the availability of nutrients or the pH of the medium (Hall-Stoodley et al., 1998; Bardouniotis et al., 2003; Ojha et al., 2005; Marsollier et al., 2007; Johansen et al., 2009; Rhoades et al., 2009).

In the recent work of Lai et al. (2018) it was demonstrated that the espE, espF, espG, and espH genes, located in the T7SS ESX-1 operon, are crucial for sliding motility and biofilm formation in M. marinum. Esp proteins, which regulate substrate transport, are involved also in virulence. This paper clearly demonstrates the role of M. marinum T7SS in the production of biofilm which, as already mentioned, is related to QS (Lai et al., 2018).

The T7SS of S. aureus, a virulence factors export machinery, plays a key role in the promotion of bacterial survival and long-term persistence of subpopulations of staphylococci. The expression of T7SS is regulated by the bacterial interaction with host tissues (Lopez et al., 2017c) mediated by the secondary sigma factor (σB) (Schulthess et al., 2012). Schulthess et al. (2012) reported that the repression of esxA by σB is due to the transcription of sarA induced by σB, which leads to a strong repression of esxA. The activation of the esxA transcript, on the other hand, is stimulated by arlR, the response regulator of the ArlRS two-component system, SpoVG, a σ-dependent element, and the Agr quorum detection system (Schulthess et al., 2012). Agr QS system is composed by AIP (self-activating peptide), the inducer ligand of AgrC which is the receptor of the agr signal. In the case of the QS Agr system, the effector of global gene regulation is an important regulatory RNA, RNAIII (Novick and Geisinger, 2008).

### T9SS

Moreover, an important relationship between T9SS and biofilm formation has been observed in periodontopathogenic pathogens such as Capnocytophaga ochracea, Porphyromonas spp., Fusobacterium spp. and Prevotella spp. (Kita et al., 2016). In the study by Kita et al. (2016), the participation of T9SS in the formation of biofilm of C. ochracea is demonstrated. The formation of biofilm of C. ochracea is crucial for the development of dental plaque and the same happens with other periodontal pathogens, in which it has also been seen that genes related to T9SS are present. Therefore, the components of the T9SS could be potential targets to inhibit the formation of biofilm and thus avoid the formation of dental plaque (McBride and Zhu, 2013; Kita et al., 2016). However, in depth analysis of the relationship between T9SS and QS network in different pathogens is required.

### DISCUSSION

To date, the T1SS, T2SS, T3SS, T4SS, T6SS, T7SS, and T9SS SS have been found to have important relationships with QS networks. The involvement of the T1SS system (Lip B which is part of the Lip exporter) in the QS network (swr quorum system) of S. liquefaciens MG1 has been investigated (Riedel et al., 2001). In P. aeruginosa, two QS systems (lasR/lasI and rhLR/rhlI) are linked to T2SS system by microarrays and proteomic studies (Chapon-Herve et al., 1997; Wagner et al., 2003; Michel et al., 2007), and DSF-type systems are also linked to T2SS in Xanthomonas species through proteome analysis (Qian et al., 2013). The QS signal AI-2 has been associated with a T3SS system in E. coli (Sperandio et al., 1999) and Vibrio spp. (Henke and Bassler, 2004). Moreover, this T3SS system has been related to QS proteins in another two pathogens, P. aeruginosa and Yersinia spp. (Liang et al., 2014; Schiano et al., 2014).

Several T4SS (virB operon) are controlled by VjbR protein which is a LuxR-type quorum-sensing regulator in B. abortus (Arocena et al., 2010; Li et al., 2017). Moreover, in the Roseobacter group, the conjugation of plasmids, which encode T4SS, is QS-controlled and the QS system may detect a broad range of long-chain AHLs at the cell surface (Patzelt et al., 2013, 2016).

There is a wealth of information relating the T6SS to QS in pathogens such as Vibrio spp. For example, Hcp and VasH from the T6SS system in V. cholerae are involved in QS (Ishikawa et al., 2009; Zheng et al., 2010; Kitaoka et al., 2011; Leung et al., 2011; Yang et al., 2018). For Pseudomonas spp,. there are numerous works where the different T6SS are regulated by QS networks (Lesic et al., 2009; Gallique et al., 2017b; Lin et al., 2017). In other pathogens as Burkholderia thailandensis (Majerczyk et al., 2016), and in A. baumannii, the relationship between QS and SS has begun to be studied (Lopez et al., 2017a).

In M. marinum, the relationship between biofilm formation, which is tightly connected with QS, and T7SS, has been demonstrated (Lai et al., 2018). Also in S. aureus, the Agr QS network has been related to T7SS (Schulthess et al., 2012). An important relationship between T9SS and biofilm formation has been observed in periodontopathogenic pathogens (Kita et al., 2016). Finally, although the involvement of T5SS secretion system in virulence, motility and competence is well-known, these systems and their association with QS must be studied in greater depth in order to clarify their roles.

Taking P. aeruginosa as a reference point, the positive effect of QS in the expression of T1SS and T2SS could be related to the fact that this organism secretes exoproducts that are public goods (proteases and lipases); hence, it is better to produce and secrete these compounds when a high cell density is reached, since these products are costly and the benefits associated to their production are higher at high cell densities. Similarly, the T6SS, which is involved in killing competitors by contact, will be more efficient at high cell densities since the probability of finding target bacteria is higher. In contrast, the T3SS appears to be negatively regulated by QS, and this may be related to its role in an acute infection and its "inhibition by QS" may be a way to facilitate the transition to a chronic infection state. In addition to its well established role in infections, T3SS has a broader ecological role suggested by its role in killing biofilm associated Acanthamoeba castellanii amoeba (Matz et al., 2008). Furthermore, it was recently demonstrated that T3SS is susceptible of cheating by mutants that do not produce it, allowing their establishment in infections (Czechowska et al., 2014); hence, the selective forces that act over T3SS are complex.

Therefore, research into the relationship between QS and SS must be further developed in order to better understand human infections.

### AUTHOR CONTRIBUTIONS

RTP, LB, AA, BG-P, LF-G, ML, IB, and GB developed the redaction of the manuscript, figures and table. RG-C, TW, and MT designed the review, assigned writing tasks to co-authors, contributed to writing and proofread the final version.

### FUNDING

This study was funded by grant PI16/01163 awarded to MT within the State Plan for R+D+I 2013–2016 (National Plan for Scientific Research, Technological Development and

### REFERENCES


Innovation 2008–2011) and co-financed by the ISCIII-Deputy General Directorate for Evaluation and Promotion of Research – European Regional Development Fund "A way of making Europe" and Instituto de Salud Carlos III FEDER, Spanish Network for the Research in Infectious Diseases (REIPI, RD16/0016/0001 and RD16/0016/0006) and by the Study Group on Mechanisms of Action and Resistance to Antimicrobials, GEMARA (SEIMC, http://www.seimc.org/). MT was financially supported by the Miguel Servet Research Programme (SERGAS and ISCIII). RTP and LF-G were financially supported by, respectively, a post-speciality grant awarded by the Fundación Novoa Santos (CHUAC-SERGAS, Galicia) and predoctoral fellowship from the Xunta de Galicia (GAIN, Axencia de Innovación). RG-C research is supported by the grant PAPIIT-UNAM number IN214218. BG-P was supported by grants PAPIIT-UNAM IN209617 and CONACYT 284081.

Pseudomonas aeruginosa PAO1. J. Bacteriol. 187, 3898–3902. doi: 10.1128/JB. 187.11.3898-3902.2005


during pulmonary infection. Proc. Natl. Acad. Sci. U.S.A. 111, 7801–7806. doi: 10.1073/pnas.1400782111


nutrient conditions. J. Appl. Microbiol. 85(Suppl. 1), 60s–69s. doi: 10.1111/j. 1365-2672.1998.tb05284.x



aeruginosa quorum sensing regulator, MvfR. Sci. Rep. 6:34083. doi: 10.1038/ srep34083



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

Copyright © 2019 Pena, Blasco, Ambroa, González-Pedrajo, Fernández-García, López, Bleriot, Bou, García-Contreras, Wood and Tomás. 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.

# Elucidating the Hot Spot Residues of Quorum Sensing Peptidic Autoinducer PapR by Multiple Amino Acid Replacements

Avishag Yehuda<sup>1</sup> , Leyla Slamti<sup>2</sup> , Einav Malach<sup>1</sup> , Didier Lereclus<sup>2</sup> and Zvi Hayouka<sup>1</sup> \*

1 Institute of Biochemistry, Food Science and Nutrition, The Hebrew University of Jerusalem, Rehovot, Israel, <sup>2</sup> Micalis Institute, INRA, AgroParisTech, Université Paris-Saclay, Jouy-en-Josas, France

The quorum sensing (QS) system of Bacillus cereus, an opportunistic human pathogen, utilizes the autoinducing PapR peptide signal that mediates the activation of the pleiotropic virulence regulator PlcR. A set of synthetic 7-mer PapR-derived peptides (PapR7; ADLPFEF) have been shown to inhibit efficiently the PlcR regulon activity and the production of virulence factors, reflected by a loss in hemolytic activity without affecting bacterial growth. Interestingly, these first potent synthetic inhibitors involved D-amino acid or alanine replacements of three amino acids; proline, glutamic acid, and phenylalanine of the heptapeptide PapR. To better understand the role of these three positions in PlcR activity, we report herein the second generation design, synthesis, and characterization of PapR7-derived combinations, alternate double and triple alanine and D-amino acids replacement at these positions. Our findings generate a new set of nonnative PapR7-derived peptides that inhibit the PlcR regulon activity and the production of virulence factors. Using the amino acids substitution strategy, we revealed the role of proline and glutamic acid on PlcR regulon activation. Moreover, we demonstrated that the D-Glutamic acid substitution was crucial for the design of stronger PlcR antagonists. These peptides represent potent synthetic inhibitors of B. cereus QS and constitute new and readily accessible chemical tools for the study of the PlcR system. Our method might be applied to other quorum sensing systems to design new anti-virulence agents.

Keywords: quorum sensing, quorum quenching, PlcR antagonists, B. cereus group, anti-virulence peptides

### INTRODUCTION

Quorum sensing (QS) is a cell-cell communication mechanism used to coordinate bacterial group behaviors (conjugation, virulence, sporulation, or competence) by assessing cell density through the production, secretion, and detection of small signaling molecules (Dunny and Leonard, 1997; Miller and Bassler, 2001; Slamti et al., 2014). Gram-negative bacteria appear to predominantly respond to N-acyl homoserine lactones, while QS in Gram-positive species mainly relies on the secretion of auto-inducing oligopeptides to bind and activate their cognate quorum sensors. In the past decade, a rapid increase of interest in bacterial quorum sensing peptides (QSPs) has emerged. Therefore, new QSPs databases are being established to provide chemical structures overview, microbial origin

Edited by:

Tom Defoirdt, Ghent University, Belgium

#### Reviewed by:

Evelien Wynendaele, Ghent University, Belgium Johann Mignolet, Catholic University of Louvain, Belgium

> \*Correspondence: Zvi Hayouka zvi.hayouka@mail.huji.ac.il

#### Specialty section:

This article was submitted to Infectious Diseases, a section of the journal Frontiers in Microbiology

Received: 12 March 2019 Accepted: 20 May 2019 Published: 07 June 2019

#### Citation:

Yehuda A, Slamti L, Malach E, Lereclus D and Hayouka Z (2019) Elucidating the Hot Spot Residues of Quorum Sensing Peptidic Autoinducer PapR by Multiple Amino Acid Replacements. Front. Microbiol. 10:1246. doi: 10.3389/fmicb.2019.01246

and functionality responses of these QS-derived signaling peptides (Gray et al., 2013; Wynendaele et al., 2013; Rajput et al., 2016).

The QSPs binding to their cognate quorum sensors occurs either on the outside of the bacterium (by interacting with a sensor in the membrane) or in the cytoplasm of the bacterial cell. In the latter case, the quorum-sensing regulators are controlled by direct interaction with a internalized signaling peptide (Dunny and Leonard, 1997; Lazazzera et al., 1997; Gominet et al., 2001; Miller and Bassler, 2001). They have been grouped in a new family of quorum sensors termed Rap-Rgg-NprR-PrgX-PlcR (RRNPP; Declerck et al., 2007; Neiditch et al., 2017). These quorum sensors are characterized by the presence of structural tetratricopeptide repeats (TPRs) forming a peptide binding domain (Blatch and Lässle, 1999), and a helix-turnhelix (HTH) DNA-binding domain (Wintjens and Rooman, 1996) in the case of transcriptional regulators. The PrgX – cCF10 system regulates conjugation in Enterococcus faecalis (Suzuki et al., 1984; Shi et al., 2005), the Rap phosphatases-Phr peptides system control competence and sporulation in Bacillus subtilis (Lazazzera et al., 1997; Perego and Brannigan, 2001; Perego, 2013), the transcriptional regulator/peptide pairs PlcR – PapR and NprR – NprX of the Bacillus cereus group are required for virulence and necrotrophism gene expression, respectively (Slamti and Lereclus, 2002; Perchat et al., 2011; Dubois et al., 2012; Grenha et al., 2013) and the archetype transcriptional regulator of the Rgg family, namely ComR that controls competence in most mutans, suis, pyogenes, bovis and salivarius streptococci (Mashburn-Warren et al., 2010; Fontaine et al., 2015) and predation in S. salivarius (Mignolet et al., 2018). The last discovered RRNPP transcriptional regulators are the PlcRa that activate the oxidative stress response and cysteine metabolism in transition state cells in B. cereus (Huillet et al., 2012) and aimR, which coordinates viruses of SPbeta group lysis-lysogeny decisions during infection of its Bacillus host cell (Erez et al., 2017).

The RRNPP family has an important role in adaptive and virulence processes in several bacteria (Slamti et al., 2014; Neiditch et al., 2017). This clearly identifies these regulators as major targets for the search of novel strategies against bacterial infections beyond conventional treatments. Antimicrobial therapy based on quorum quenching (QQ) can interfere or block all the processes involved in quorum sensing (Amara et al., 2011; Kalia, 2013; Grandclément et al., 2015). In contrast to antibiotics or antimicrobial agents, which aim at killing bacteria or inhibiting their growth, blocking cell-tocell signaling mechanism, could attenuate bacterial pathogenicity without imposing the level of selective pressure on a bacterial population to develop resistance (Suga and Smith, 2003; Rasmussen and Givskov, 2006). A wide range of promising molecules have been already identified to inhibit QS-controlled virulence genes in Gram-negative bacteria (Hentzer et al., 2003; Galloway et al., 2012). On the other hand, except for strategies that have been investigated to inhibit the two component QS system Agr of Staphylococcus, which uses a peptide-thiolactone as the extracellular signal, the design of molecules modulating QS systems in Gram-positive bacteria has been poorly explored (Fontaine et al., 2010; Zheng et al., 2011; Tal-Gan et al., 2013a, 2014, 2016; Sully et al., 2014).

Bacillus cereus is a human opportunistic, Gram-positive sporeforming bacterial pathogen belonging to the B. cereus group (Stenfors Arnesen et al., 2008). This group comprises a number of highly phenotypically and genetically indistinguishable related species, including Bacillus thuringiensis, an insect pathogen, and Bacillus anthracis, the aetiological agent of anthrax (Helgason et al., 2000). The widespread presence of B. thuringiensis and B. cereus in soil and food, and their close relationship with B. anthracis make this group an important threat to public health (Rasko et al., 2005; Rossi et al., 2018), and a potential source of new pathogens. Indeed, B. cereus is generally regarded as a pathogen causing foodborne infections due to the production of enterotoxins such as Hbl and Nhe (Stenfors Arnesen et al., 2008), and nosocomial infections in an immuno-compromised patients (Granum and Lund, 1997; Kotiranta et al., 2000; Chu et al., 2001; Gaur et al., 2001; Bottone, 2010). B. cereus strains were also found to be responsible for severe infections resembling anthrax (Hoffmaster et al., 2004; Klee et al., 2006).

The QS system of B. cereus plays an important role in virulence (Agaisse et al., 1999; Gohar et al., 2008). B. cereus uses QS to establish infections by producing an arsenal of virulence factors, such as enterotoxins, pore-forming haemolysins, cytotoxins and various degradative enzymes (Granum and Lund, 1997; Vilas-Boas et al., 2002; Stenfors Arnesen et al., 2008; Ramarao and Sanchis, 2013). Production of most of these exported virulence factors is activated by PlcR, a 34 kDa protein that acts as a B. cereus group main virulence transcription factor (Lereclus et al., 1996; Agaisse et al., 1999; ØKstad et al., 1999; Gohar et al., 2008). Activity of PlcR depends on the binding of the signaling C-terminal heptapeptide PapR<sup>7</sup> (ADLPFEF) at the end of the exponential growth stage. PapR<sup>7</sup> is imported by the oligopeptide permease system (OppABCDF; Gominet et al., 2001), binds the tetratricopeptide repeat (TPR)-type regulatory domain of PlcR (Grenha et al., 2013) and promotes recognition of the PlcR box to transcriptional activation of the target genes (Lereclus et al., 1996; Gominet et al., 2001; Slamti and Lereclus, 2002; Bouillaut et al., 2008). This triggers a positive feedback loop that up-regulates the expression of plcR, papR and various virulence genes (Agaisse et al., 1999; Ivanova et al., 2003; Gohar et al., 2008).

The structural and molecular basis for the activation of PlcR by PapR has been the focus of several studies, which have revealed interesting insights on the PlcR – PapR interactions. The PlcR – PapR relationship has been shown to be strain specific; comparison of the amino acid sequences of PlcR and PapR from 29 different strains demonstrated the existence of four classes (I to IV) of PlcR – PapR pairs, defining four distinct pherotypes in the B. cereus group. While PapR sequences from different strains of the B. cereus group showed divergences in their three N-terminal residues, the PFEF core was more conserved (Slamti and Lereclus, 2005). In 2007, the crystal structure of the complex formed between the protein PlcR (from group I) and the C-terminal PapR<sup>5</sup> pentapeptide (LPFEF) was published (Declerck et al., 2007). According to RRNPP conserved features, each subunit of PlcR is formed of an N-terminal HTH DNA-binding domain, and a C-terminal regulatory domain

composed of five degenerated TPRs forming a peptide binding domain. Binding of PapR triggers an allosteric mechanism that leads to a drastic conformational change of the HTH domains upon the two half sites of the DNA binding site, known as PlcR-box. The LPFEF pentapeptide, PapR<sup>5</sup> was identified as the minimal peptide size required for PlcR activation (Slamti and Lereclus, 2002). However, the physiologically relevant heptapeptide PapR<sup>7</sup> displays a slightly better affinity for PlcR (Bouillaut et al., 2008; Grenha et al., 2013). In 2008, Bouillaut and co-workers established a molecular model for the complex formed between PlcR and the heptapeptide PapR<sup>7</sup> based on the crystal structure of PapR5-bound PlcR (Declerck et al., 2007; Bouillaut et al., 2008). Structural analysis and directed mutagenesis of PlcR residues suggested that: a) activation of PlcR by PapR<sup>7</sup> is triggered by the hydrophobic interactions of the leucine, and two phenylalanines with helices 5 and 7 of the TPR-containing domain of PlcR b) the central proline residue may be required for the PapR peptides to fit into the binding groove on PlcR and c) the glutamic acid of the FEF PapR<sup>7</sup> core motif may function to selectively allow PapR to bind PlcR by ionic interactions with Lys87 and 89. In a follow up study in 2013, Grenha and co-workers, have determined the crystal structure of the ternary complex DNA-PlcR-PapR7. It has been reported that both PapR<sup>7</sup> phenylalanine residues are located in hydrophobic pockets and the only specific interactions are made between the glutamate of PapR<sup>7</sup> and residues Lys89, Gln237, and Tyr275 of PlcR.

Binding of PapR to PlcR is essential to trigger QS-mediated functions in B. cereus. Thus, we recently studied the PlcR – PapR activation in B. cereus and B. thuringiensis at the molecular level. We designed, synthesized and characterized synthetic PapR 7-mer derived peptides to determine the contribution of each residue within PlcR – PapR<sup>7</sup> interactions. Our findings reveal the first set of non-native peptides that can repress the PlcR regulon and thus relevant virulence factors. Moreover, we could demonstrate that the repression is mediated by QS and regulation of PlcR expression without affecting bacterial growth (Yehuda et al., 2018). Interestingly, these first potent synthetic inhibitors involved D-amino acid or alanine replacements of either proline (P) glutamic acid (E) or phenylalanine (F) of the heptapeptide PapR (ADLPFEF). To better understand the role of these three crucial positions in PlcR activation, we report herein the second generation design, synthesis, and characterization of PapR7-derived combinations, alternate double and triple alanine and D-amino acids replacement at these positions. We propose this systematic replacement approach to elucidate other quorum quenching agents in Gram-positive bacteria.

### MATERIALS AND METHODS

### Bacterial Strains and Growth Conditions

Bacterial strains used in this study: B. thuringiensis 407 Cry<sup>−</sup> plcA0Z (Bt A'Z) and the PapR null-mutant 407 Cry<sup>−</sup> 1papR plcA'Z (Bt 1papR A'Z) strains, containing a transcriptional fusion between the promoter of plcA and the lacZ reporter gene (as described previously; Gominet et al., 2001; Slamti and Lereclus, 2002); B. cereus strain ATCC 14579 (Ivanova et al., 2003). Unless otherwise noted, cells were grown in modified LB medium (16 g/L tryptone, 8 g/L yeast extract, 5 g/L NaCl) at 37◦C and stored at −80◦C in LB containing 25% glycerol. Kanamycin (200 µg/mL) was used for the selection of B. thuringiensis.

## Solid Phase Peptide Synthesis Methodology (SPPS)

All the peptides were synthesized using standard Fmoc-based solid-phase peptide synthesis (SPPS), microwave irradiation, procedures on Rink Amide resin (substitution 0.5 mmol/g, 25 µmol) in SPE polypropylene Single-Fritted tubes. The Fmocprotecting group was then removed by treating the resin with 20% (v/v) piperidine diluted in dimethylformamide (DMF) followed by heating to 80◦C in the microwave (MARS, CEM, United States; 2-min ramp to 80◦C, 2-min hold at 80◦C) with stirring. To couple each amino-acid, Fmoc-protected amino acids (4 equiv. relative to the overall loading of the resin), were dissolved in DMF and mixed with 2-(1H-benzotriazol-1-yl)-1,1,3,3-tetramethyluronium hexafluorophosphate (HBTU; 4 equiv.) and diisopropylethylamine (DIEA; 4 equiv.). The solution was allowed to pre-activate for 5 min before being added to the resin, and heated to 70◦C in a multimode microwave (2-min ramp to 70◦C, 4-min hold at 70◦C) with stirring. After each coupling/deprotection cycle the resin was drained and washed with DMF (3 × 5 mL). Once peptide synthesis was completed, the peptide was cleaved from the resin, by mixing the resin with 3 mL cleavage cocktail of 95% trifluoroacetic acid (TFA), 2.5% triisopropylsilane (TIPS), and 2.5% deionized water for 3 h with agitation. The peptide mixture was precipitated from the TFA solution by the addition of cold ether and collected by centrifugation (Eppendorf R5810 8000 rpm for 10 min). The ether was then removed, and the peptide was dried under a stream of nitrogen, and lyophilized, before high-performance liquid chromatography (HPLC) purification.

### Peptide Purification

Crude peptides were purified and characterized with Reverse-Phase (RP)-HPLC. The crude peptides were diluted to a final concentration of 10 mg/ml in a solution of 20% acetonitrile (ACN) in water (v/v) or dimethyl sulfoxide (DMSO). A semipreparative Phenomenex Kinetex C18 (5 µm, 10 × 250 mm) was used for preparative RP-HPLC work. An analytical Phenomenex Gemini C18 column (5 µm, 4.6 mm × 250 mm, 110 Å) was used for analytical RP-HPLC work (**Supplementary Figure S1**). Standard RP-HPLC conditions were as follows: flow rates = 5 mL min−<sup>1</sup> for semi-preparative separations and 1 mL min−<sup>1</sup> for analytical separations; mobile phase A = 18 M water + 0.1% TFA; mobile phase B = ACN. Purities were determined by integration of peaks with UV detection at 220 nm using a linear gradient (first prep 5% B → 65% B over 60 min and second prep 26% B → 36% B over 20 min). The purity of the tested peptides was determined using a linear gradient (5% B → 65% B over 60 min). MALDI-TOF spectrometry (Bruker Daltonik, Germany) was used to validate the synthesized peptides molecular weight (**Supplementary Table S1**). The purified peptides were lyophilized and stored at −20◦C.

## Analysis of PlcR Regulon Expression Using β-Galactosidase Assay

### PlcR Activation Studies

fmicb-10-01246 June 6, 2019 Time: 20:11 # 4

Bt 1papR A'Z cells were grown overnight in LB medium with selective antibiotic. The cells were diluted 10−<sup>3</sup> in modified LB to a final volume of 1 liter and incubated at 37◦C with shaking (200 rpm) until onset of the stationary phase of bacterial growth (OD<sup>600</sup> 3 ± 0.5). Various concentrations of synthetic peptides were added to 2 ml aliquots of culture, which were incubated for 1 h before centrifugation (Eppendorf centrifuge R5810, 4000 rpm for 5 min) and quantification of β-galactosidase.

### Competition Studies of PapR7-Derived Peptides

Bt A'Z cells were grown overnight in LB medium. The cells were diluted 10−<sup>3</sup> in modified LB to a final volume of 1 L and incubated at 37◦C with shaking (200 rpm) until the end of the lag or late-exponential of bacterial growth (OD<sup>600</sup> 0.1 ± 0.03; 1.8 ± 0.1, respectively). Different concentrations of synthetic peptides were added to 2 ml aliquots of culture and incubated for various times (1–24 h) before centrifugation (Eppendorf centrifuge R5810, 4000 rpm for 5 min) and quantification of β-galactosidase activity.

### β-Galactosidase Assay

β-galactosidase activity was measured as described previously (Yang et al., 2017), with minor modifications. Briefly, 200 µL aliquots from 2 ml treated cultures were added in triplicate to a clear 96-well microtiter plate, and then OD<sup>600</sup> was measured and β-galactosidase activity was assayed. The final results were reported as percentage of activation, which is the ratio between the Miller units obtained after addition of the PapR<sup>7</sup> analogs. In Bt 1papR A'Z strain, the plcA promoter activity was very low and considered as a baseline. In Bt A'Z strain, the untreated bacteria were considered as 100% of activation and the results were normalized accordingly. Each assay was repeated at least three times.

### Hemolytic Assay Toward Human Red Blood Cells

Bt A'Z or B. cereus ATCC 14579 cells were grown overnight in LB medium. The cells were diluted 10−<sup>3</sup> in modified LB to a final volume of 1 liter and incubated at 37◦C with shaking (200 rpm) until the end of the lag phase of bacterial growth (OD<sup>600</sup> 0.1 ± 0.03). Different concentrations of synthetic peptides were added to 2 ml aliquots of culture and incubated for 2.5 h before centrifugation (Eppendorf centrifuge R5810, 4000 rpm for 5 min), separation and filtration (0.2 µm filter) of the supernatants of the treated cultures. Analyses of hemolytic activity were conducted as previously described using human red blood cells (Tal-Gan et al., 2013b; Lobel et al., 2015). Bacterial supernatants were serially diluted in Tris-buffered saline (pH 7.2, 10 mM Tris–HCl, 155 mM NaCl) with 1% human red blood cells (hRBC) suspension and were incubated for 30 min at 37◦C. Hemolytic activities were measured by monitoring the absorbance at 420 nm.

### Statistical Analysis

Unless otherwise noted, the results are presented as the mean ± SEM. One-way analysis ANOVA of variance, followed by Tukey post hoc analysis was used for statistical analysis. The results were considered to be statistically significant if p < 0.01.

## RESULTS

We have previously reported the first five synthetic peptidic inhibitors of B. cereus PlcR-PapR QS system; three independent alanine amino acid replacements (PapR<sup>7</sup> – P4A, E6A, and F7A) and two D-amino acid substitutions (PapR<sup>7</sup> – dE<sup>6</sup> and dF7) showed great reduction of PlcR regulon expression and virulence factor secretion (Yehuda et al., 2018).

We initiated the current study by evaluating the three crucial positions of the heptapeptide PapR -proline (Pro4), glutamic acid (Glu6) and phenylalanine (Phe7) through systematic single or/and multiple amino acid substitution strategy. We designed, synthesized and purified a second generation set of PapR7 derived peptide combinations to further explore the structure– activity relationship delineated previously for the first-generation of peptidic analogs. This set included twelve peptides with double and triple alanine and D-amino acid replacements, at the crucial Pro4, Glu6 and Phe7 residues (**Supplementary Table S1**, **Supplementary Figure S1** and **Figure 1**).

We scanned each of PapR7-derived peptide combinations for its ability to modulate the expression of the PlcR regulon using B. thuringiensis 407 Cry<sup>−</sup> (Bt 407−) as a model bacterium for the B. cereus group. This strain cured of its plasmid is acrystalliferous and shows high phylogenic similarity with the B. cereus reference strain ATCC 14579 (Lereclus et al., 1989; Priest et al., 1994; Slamti and Lereclus, 2005). Two lacZ-based reporter strains were used in the current study, B. thuringiensis 407<sup>−</sup> plcA0Z (Bt A'Z) and PapRnull mutant B. thuringiensis 407<sup>−</sup> 1papR plcA0Z (Bt 1papR A'Z). Both reporter strains contain a transcriptional fusion between the plcA promoter region and the lacZ gene. plcA is a member of the PlcR regulon and its expression directly reflects the activity of PlcR. The activity of each PapR7-derived peptide combination was evaluated and compared to the previously described five B. cereus inhibitory synthetics PapR7-derived peptides (PapR7- P4A; E6A; F7A; dE<sup>6</sup> and dF7; Yehuda et al., 2018).

We first conducted an initial screening of all the analogs at high peptide concentration (5 µM) in order to evaluate their ability to activate the PlcR regulon to a level comparable to the synthetic PapR<sup>7</sup> signal peptide (**Figure 2**). Both first and second generations of PapR<sup>7</sup> analogous were classified by their number of amino acid replacements (alanine or D-amino); single, double and triple. Similar to our first-generation single replacement inhibitors, none of the new PapR7-derived peptide combinations were capable of activating the PlcR regulon. These findings revealed that any alanine and D-amino acid replacements, at the

(B) Sequences of second generation of PapR7-derived peptide combinations, involved double and triple alanine and D-amino acids replacement.

positions Pro4, Glu6 and Phe7 of PapR<sup>7</sup> derivatives, are critical for PlcR regulon activation. These derivatives can therefore be classified as potential candidates for the development of potent second-generation PlcR inhibitors.

3 ± 0.5, mean ± SEM, n = 9). <sup>∗</sup>p < 0.01 indicates a statistically significant difference between addition of synthetic PapR<sup>7</sup> peptide and PapR7-

We next scrutinized their ability to compete with the endogenous PapR signal peptide (in Bt A'Z reporter strain) for reducing the activation of the PlcR regulon in late-exponential phase of bacterial growth (OD<sup>600</sup> of 1.8 ± 0.1). As shown in **Figure 3A**; PapR<sup>7</sup> derivatives activities were classified by D-amino acid enantiomer or Ala replacements at either of the three PapR<sup>7</sup> crucial positions (Pro4, Glu6 and Phe7). The previously reported inhibitors, PapR7-P4A; E6A; F7A; dE<sup>6</sup> and dF7, were also included for comparison. Peptides PapR<sup>7</sup> – P4A:E6A, P4A:E6A:F7A and P4A:E6A:dF<sup>7</sup> did not show any reduction in plcA'Z activity. PapR<sup>7</sup> – E6A:F7A, E6A:dF7, P4A:F7A, P4A:dE6, P4A:dF7, PapR<sup>7</sup> – P4A:dE6:dF<sup>7</sup> and P4A:dE6:F7A were able to reduce plcA'Z activity by approximately 40%. We identified two candidate peptides that inhibited plcA'Z activation when added at late-exponential phase. Indeed, PapR<sup>7</sup> – dE6:dF<sup>7</sup> and dE6:F7A reduced plcA'Z activity by 71 and 65 %, respectively, similarly to the inhibitory activity of their parent reporter inhibitors PapR<sup>7</sup> – dE6, dF<sup>7</sup> and F7A (65, 63, and 70%, respectively). To confirm PapR<sup>7</sup> – dE6:dF<sup>7</sup> and dE6:F7A potent inhibition, we repeated the experiment with several concentrations of these inhibitors in order to determine their IC<sup>50</sup> values (**Figures 3B,C**). The results show that the new inhibitory peptides have IC<sup>50</sup> values in the low micromolar range, almost comparable to their parent reporter inhibitors IC<sup>50</sup> values (**Table 1**). Overall, we identified new potent inhibitors, PapR<sup>7</sup> – dE6:dF<sup>7</sup> and dE6:F7A, which were able to compete with endogenous PapR and inhibit PlcR regulon activity very effectively. Interestingly, these two potent inhibitors contain D-Glutamic acid replacement at position 6 of the PapR heptapeptide.

We have previously reported that inhibition through PapR<sup>7</sup> inhibitory peptidic derivatives is cell density dependent (Slamti and Lereclus, 2002; Yehuda et al., 2018). To examine the effect of bacterial cell density on the inhibition of the PlcR regulon expression, each derivative was added to Bt A'Z cells at OD<sup>600</sup> 0.1 ± 0.03, which corresponds to the early stage of exponential phase. PlcR-dependent gene expression was then quantified after 2.5 h (**Figure 4**) and after 24 h in order to assess their activity and stability over time (**Supplementary Figure S2**). As has been reported recently, all of our known parent PapR<sup>7</sup> inhibitors were able to completely block plcA'Z activation for up to 24 h under these conditions (Yehuda et al., 2018). We observed similar results with PapR<sup>7</sup> – dE6:dF<sup>7</sup> and dE6:F7A; these peptides blocked plcA'Z activation compared to the parent peptidic inhibitors for 2.5 h and up to 24 h (**Figure 4** and **Supplementary Figure S2**). Combining them with additional alanine replacement at Pro4, PapR<sup>7</sup> – P4A:dE6:dF<sup>7</sup> and P4A:dE6: F7A showed a

derived peptides.

FIGURE 3 | Competition studies with new PapR7-derived peptide combinations. (A) β-galactosidase activity of Bt A'Z induced by the addition of 2.5 µM PapR7-derived peptides, (B) PapR7- dE6:dF7, and (C) PapR7-dE6:F7A derivatives in several concentrations normalized to untreated bacterial cells at late-exponential of bacterial growth (OD<sup>600</sup> of 1.8 ± 0.1, mean ± SEM, n = 9). <sup>∗</sup>p < 0.01 indicates a statistically significant difference between untreated Bt A'Z and addition of PapR7-derived peptides. Different letters indicate statistically significant differences between PapR7-derived peptide treatments (p < 0.01).

reduction of ∼75% in plcA'Z activation. The addition of PapR<sup>7</sup> derivatives as PapR7- E6A:F7A, E6A:dF7, P4A:F7A, P4A:dE<sup>6</sup> and P4A:dF<sup>7</sup> in the early stages of the bacterial growth led to drastic reduction in plcA'Z activity, revealing a series of new peptidic inhibitors. In contrast, the non-inhibitory peptidic combinations (PapR7- P4A:E6A, P4A:E6A:F7A and P4A:E6A:dF7) did not reduce the PlcR regulon expression even in low bacterial density. Importantly, the bacterial growth was not affected by the addition of all the examined peptides (data not shown).

Throughout these competition studies, we identified a group of three non-inhibitory peptidic combinations (PapR7- P4A:E6A, P4A:E6A:F7A and P4A:E6A:dF7), seven variants of PapR<sup>7</sup> containing multiple replacements combinations with median to high inhibitory activities (PapR7- P4A:dE6:F7A, P4A:dE6:dF7, E6A:F7A, E6A:dF7, P4A:F7A, P4A:dE<sup>6</sup> and P4A:dF7) and two very potent inhibitors that abolish activation of plcA'lacZ (PapR<sup>7</sup> – dE6:dF<sup>7</sup> and dE6:F7A) and compete with endogenous PapR. These two peptides are proposed as quorum quenchers that

TABLE 1 | Comparison between IC<sup>50</sup> values of new PapR7-derived peptide combinations and their parent inhibitors, as determined by the lacZ-based reporter assay.


∗ IC<sup>50</sup> values were calculated by GraphPad Prism 8, using the non-linear inhibitor vs. normalized response method. Different letters indicate statistically significant differences between IC<sup>50</sup> values of PapR7-derived peptide combinations (p < 0.01).

FIGURE 4 | Exploring the effect of bacterial cell density on the PapR7-derived peptide combinations activity. β-galactosidase activity of Bt A'Z induced by the addition of 10 µM PapR7-derived peptides normalized to untreated bacterial cells at end lag phase of bacterial growth (OD<sup>600</sup> of 0.1 ± 0.03; mean ± SEM, n = 9). <sup>∗</sup>p < 0.01 indicates a statistically significant difference between untreated Bt A'Z and addition of PapR7-derived peptides. Different letters indicate statistically significant differences between PapR7-derived peptide treatments (p < 0.01).

do not affect the bacterial growth but inhibit the expression of the PlcR regulon.

After observing this inhibitory activity, we expanded our study to explore the effect of these second generation PapR<sup>7</sup> analogs on the production of a representative virulence factor under the control of PlcR in wild-type bacteria. Previous studies have shown that the activity of hemolysins in B. cereus is regulated by the PlcR – PapR QS system (Salamitou et al., 2000; Slamti and Lereclus, 2002; Slamti et al., 2004). Therefore, we studied

the effect of the new PapR<sup>7</sup> derivatives on the production of hemolysins in B. cereus strain ATCC 14579, a representative member of the B. cereus sensu stricto species.

We performed hemolytic activity assays toward human red blood cells in the presence of the synthetic derivatives (**Figure 5A**). From these results, we identified PapR<sup>7</sup> inhibitory peptidic analogs that were able to reduce the hemolytic activity of wild type B. cereus ATCC 14579. Interestingly, these analogs reduced the expression of hemolysin (**Figure 5A**) even more efficiently than the inhibition that was observed for Bt 407<sup>−</sup> PlcR-dependent gene expression (as shown in **Figure 4**). We quantified the hemolytic activity of the strong inhibitors group by determining their IC<sup>50</sup> values. In addition to the two strong inhibitors derivatives (PapR<sup>7</sup> – dE6:dF7, dE6:F7A), we observed great activity also for PapR<sup>7</sup> – P4A:dE6:dF<sup>7</sup> and P4A:dE6:F7A analogs. Regard to PapR<sup>7</sup> – dE6:dF<sup>7</sup> and dE6:F7A derivatives, the relative IC<sup>50</sup> value trends were highly similar to those in the lacZ-reporter assays results (**Figure 5B** and **Tables 1**, **2**). In comparison to their parent single inhibitors PapR<sup>7</sup> – dF<sup>7</sup> and F7A, the relative IC<sup>50</sup> values were in the same range (**Figures 5B,C** and **Table 2**); addition of D- Glutamic acid in the parent peptide PapR<sup>7</sup> – dF<sup>7</sup> (PapR<sup>7</sup> – dE6:dF7) slightly improved its IC<sup>50</sup> (IC<sup>50</sup> value was reduced to 1.382 compared to 3.041), while alanine substituted at position Phe7 (PapR<sup>7</sup> – dE6:F7A) did not show any effect. PapR7– P4A:dE6:dF7 and P4A:dE6:F7A exhibited lower IC<sup>50</sup> values compared to their parent single inhibitors PapR7– P4A, dF<sup>7</sup> and F7A but a higher IC<sup>50</sup> value compared to the parent peptide PapR7– dE<sup>6</sup> (**Table 2**). We observed another key feature in the new PapR<sup>7</sup> combinations; sharing specific substitutions at PapR<sup>7</sup> sequence influence their ability to inhibit PlcR activity; for example, all non-inhibitory peptidic combinations included alanine substitutions at positions Pro4 and Glu6 of PapR<sup>7</sup> peptide, while all strong inhibitors group members share replacement of Glu6 by its D-isomer. All these new PapR<sup>7</sup> combinations activity profiles may shed light on PlcR and PapR interaction. To better understand their role in PlcR activity, we divided the delineated above (**Figure 5**) PapR7-derived peptidic combinations hemolytic activities on human red blood cells to three different sets (**Figures 6A–C**).

TABLE 2 | Comparison between IC<sup>50</sup> values of new PapR7-derived peptide combinations and their parent inhibitors, as determined by the hemolytic assay.


∗ IC<sup>50</sup> values were calculated by GraphPad Prism 8, using the nonlinear inhibitor vs. normalized response method. Different letters indicate statistically significant differences between IC<sup>50</sup> values of PapR7-derived peptide combinations (p < 0.01).

### SUMMARY

### Role of Individual and Combined Pro4 or Glu6 Residue Replacements

Replacing both Pro4 and Glu6 in PapR<sup>7</sup> with alanine yielded nonactive peptidic combinations regardless of other modifications in Phe7 position (PapR7- P4A:E6A, P4A:E6A:F7A, and P4A:E6A:dF7; **Figure 6A**). PapR7- **P4A:E6A**:F7A and **P4A:E6A** are non-inhibitory peptides, while PapR7- **E6A**:F7A and **P4A**:F7A defined as medium inhibitors, reduce the hemolytic activity by ∼60%. The same trend was observed with other two non-inhibitory peptidic combinations; PapR7- **P4A:E6A**:dF<sup>7</sup> and PapR7- **P4A:E6A** compared to their disassembled peptidic combinations PapR7- **E6A**:dF7, **P4A**:dF<sup>7</sup> and PapR7- **P4A**, **E6A,** respectively. These findings verified the dependence of proline and glutamic acid residues in PapR<sup>7</sup> inhibitory activity. By modifying these two residues together to alanine the PapR7 derived inhibitors lost their antagonist features and their ability to prevent native PapR-PlcR interaction. This is in agreement with previous study (Bouillaut et al., 2008; Grenha et al., 2013), which emphasizes that both proline and glutamic acid have an important role in PlcR activation.

### Effect of Phe7 Replacement

Crystal structure of PapR7- PlcR complex showed that both PapR<sup>7</sup> phenylalanine residues (Positions 5 and 7) are located in hydrophobic pockets (Grenha et al., 2013), and involved in hydrophobic interactions with PlcR. We examined the effect of replacing Phe7 with alanine or D-amino acid on the inhibitory activities of the designed peptides. Introducing D-Phenylalanine at position 7, regardless to the modifications at position 4 and 6, enhances the inhibitory activity of PlcR (**Figure 6B**). Replacing F7 in alanine (PapR<sup>7</sup> – E6A:**F7A** or P4A:**F7A)** reduced hemolysis of red blood cells by 60%. However, these replacements combined with D-Phenylalanine yielded analogs (PapR<sup>7</sup> – E6A:**dF**<sup>7</sup> and P4A:**dF**7) with stronger antagonistic activities (approximately 83% inhibition). These results indicate that the inclusion of D-Phenylalanine may contribute to hydrophobic interactions with PlcR by preserving the aromatic ring side chain interaction.

### Importance of Glu6 and Its Stereoisomer Substitution

We observed that all the strong inhibitor group members contain the replacement of Glu6 to its D-enantiomer (**Figure 5A**). PapR<sup>7</sup> – **dE**6:F7A and **dE**6:dF<sup>7</sup> displayed similar inhibitory effect, regardless of other modifications (D-amino or alanine replacements) in position Phe7 (**Figure 6C**). In contrast, replacing Glu6 with alanine yielded two weaker PlcR antagonists, when either Phe7 combined substitutions with alanine or D-amino (PapR<sup>7</sup> – **E6A**:F7A and **E6A**:dF7). PapR<sup>7</sup> – P4A:**dE**6:F7A and P4A:**dE**6:dF<sup>7</sup> fully prevented hemolysis of red blood cells, however, replacing only Glu6 with alanine yielded two non-inhibitory peptidic combinations PapR7- P4A:**E6A**:F7A and P4A:**E6A**:dF<sup>7</sup> as was supported by IC<sup>50</sup> values (**Figure 6D**).

FIGURE 5 | New PapR7-derived peptide combinations inhibit Bc virulence factor. (A) Hemolytic activity on human red-blood cells of supernatant B. cereus ATCC 14579 treated cultures in 10 µM of PapR7-derived peptides normalized to untreated bacterial cells at end lag phase of bacterial growth (OD<sup>600</sup> of 0.1 ± 0.03; mean ± SEM, n = 9). Hemolysis inhibition dose response curves of B. cereus ATCC 14579 treated supernatant cultures in different concentrations of (B) PapR<sup>7</sup> – dE6:dF7, dE6:F7A, P4A:dE6:dF7, and P4A:dE6:F7A and (C) known inhibitors normalized to untreated bacterial culture (mean ± SEM, n = 9). <sup>∗</sup>p < 0.01 indicates a statistically significant difference between untreated B. cereus ATCC 14579 and addition of PapR7-derived peptides. Different letters indicate statistically significant differences between PapR7-derived peptide treatments (p < 0.01).

Interestingly, in an earlier study (Bouillaut et al., 2008; Grenha et al., 2013) the authors characterized the function and specific interactions of PapR glutamic acid with conserved residues in PlcR. These findings support our results about the important role of Glu6 in the activity of PlcR regulon. Indeed, replacement of L-glutamic acid of PapR7- P4A:F7A and P4A:dF<sup>7</sup>

(corresponding ADLAF**E**A and ADLAF**E**dF), with D-glutamic acid yielded two potent PlcR antagonists; PapR7- P4A:**dE**6:F7A and P4A:**dE**6:dF7. Overall, these three sets of new PapR7-derived peptide combinations support previous published studies and reveal the important role of three crucial positions at designing potent PlcR antagoinsts; Pro4, Phe7 and especially Glu6 that may function to selectively allow PapR, but not other similar autoinducers, to bind PlcR.

### CONCLUSION

The PapR-PlcR QS system is extensively involved in the pathogenesis of B. cereus, highlighting this system as an attractive target for an alternative treatment to prevent infection. We have previously reported the first five potent synthetic peptidic inhibitors of B. cereus PlcR-PapR QS system (Yehuda et al., 2018); three independent alanine amino acid replacements (PapR<sup>7</sup> - P4A, E6A, and F7A) and two D-amino acid substitutions (PapR<sup>7</sup> – dE<sup>6</sup> and dF7). We concluded that the critical residues for PapR<sup>7</sup> –PlcR interaction and PlcR activation were proline, glutamic acid and phenylalanine. To further understand their role in PlcR activity, a new set of PapR<sup>7</sup> analogs with double and triple alanine and Damino acid replacements at these positions were designed and synthesized. Multiple amino acid substitutions revealed that any replacement at these positions Pro4, Glu6 and Phe7 of PapR<sup>7</sup> derivatives, is critical for PlcR regulon activation in the 1papR mutant strain. A comprehensive competition study of all PapR7-derived peptides combinations in lateexponential phase identified four promising QS peptidic inhibitors candidates; PapR<sup>7</sup> – dE6:dF<sup>7</sup> and dE6:F7A and two other analogs with additional alanine substituted PapR<sup>7</sup> – P4A:dE6:dF<sup>7</sup> and P4A:dE6:F7A, all contained D-Glutamic acid at position 6 of the C-terminus of heptapeptide PapR. The two potent inhibitors PapR<sup>7</sup> – dE6:dF<sup>7</sup> and dE6:F7A generated similar inhibitory activity as their parent single replacement reported inhibitors PapR<sup>7</sup> – dE6, dF<sup>7</sup> and F7A with comparable IC<sup>50</sup> values ∼= 1–2.6 <sup>µ</sup>M. Our results verified previous reports that inhibition through PapR<sup>7</sup> derivatives is cell density dependent (Slamti and Lereclus, 2002; Yehuda et al., 2018). We showed that all of our new four promising QS peptidic inhibitors candidates blocked the PlcR regulon activity even after 24-h period, when they were added at an early stage of bacterial growth (PapR<sup>7</sup> – dE6:dF7, dE6:F7A, PapR<sup>7</sup> – P4A:dE6:dF7, and P4A:dE6:F7A). Moreover, by exposing the bacterial cells to these analogs at earlier stage (OD<sup>600</sup> 0.1 ± 0.03) we discovered a new series of inhibitors (PapR7- E6A:F7A, E6A:dF7, P4A:F7A, P4A:dE6, and P4A:dF7). Similar to the parent peptidic inhibitors, we hypothesized that the positive autoregulatory loop was blocked and quorum quenching was achieved throughout growth by the inhibitory multiple combinations PapR<sup>7</sup> derivatives.

We next used a human red-blood cells hemolytic assay as a direct method to assess a QS-related phenotype linked to virulence in wild-type B. cereus. The inhibitory PapR<sup>7</sup> peptidic analogs identified using the lacZ-reporter assays were even more efficient in reducing the hemolytic activity of wild type B. cereus ATCC 14579.

Our findings both corroborate and extend previous observations regarding the role of the PapR<sup>7</sup> in PlcR receptor recognition; first, we showed the important role of proline or glutamic acid residues in PapR- PlcR interactions and as key in designing strong inhibitors. Second, we demonstrated that inclusion of D-Phenylalanine at Phe7 contribute to PapR<sup>7</sup> derivatives inhibitory activities probably due to its hydrophobic features. Moreover, by interfering this Glu6 specific interactions with PlcR, we found the potential of D-Glutamic substitution at designing potent PlcR antagonist. These findings are consistent with previous study (Slamti and Lereclus, 2005) which investigated specificity and polymorphism of PlcR – PapR in the B. cereus group. Interestingly, while all the PapR sequences from different strains of the B. cereus group showed divergences in their three N-terminal residues, the E6 position was conserved. In the current study we highlighted the precise and unpredictable engineering of natural pheromone in our effort to develop new Quorum Quenching agents, reflecting the trade-off between good peptide binding and lower activation. These new non-native peptides inhibitors may be applied as chemical tools to further study the role of PlcR and other QS in all B. cereus group members. Further, our method of single and multiple amino acid replacements might be applied to other QS system to design new anti-virulence agents.

### DATA AVAILABILITY

No datasets were generated or analyzed for this study.

### AUTHOR CONTRIBUTIONS

AY, LS, and EM performed the research. DL and ZH analyzed the data and wrote the manuscript.

### FUNDING

This research project was supported by UHJ-France and the Scopus Foundation.

### ACKNOWLEDGMENTS

We would like to thank Dr. John Karas for reading and improving this manuscript.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb.2019. 01246/full#supplementary-material

### REFERENCES

fmicb-10-01246 June 6, 2019 Time: 20:11 # 10



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

Copyright © 2019 Yehuda, Slamti, Malach, Lereclus and Hayouka. This is an openaccess article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# LbDSF, the Lysobacter brunescens Quorum-Sensing System Diffusible Signaling Factor, Regulates Anti-Xanthomonas XSAC Biosynthesis, Colony Morphology, and Surface Motility

Jun Ling<sup>1</sup>† , Runjie Zhu<sup>1</sup>† , Pedro Laborda<sup>2</sup> , Tianping Jiang<sup>1</sup> , Yifan Jia<sup>1</sup> , Yangyang Zhao<sup>1</sup> and Fengquan Liu1,3 \*

1 Institute of Plant Protection, Jiangsu Academy of Agricultural Sciences, Jiangsu Key Laboratory for Food Quality and Safety-State Key Laboratory Cultivation Base of Ministry of Science and Technology, Nanjing, China, <sup>2</sup> School of Life Sciences, Nantong University, Nantong, China, <sup>3</sup> Institute of Life Sciences, Jiangsu University, Zhenjiang, China

#### Edited by:

Cristina García-Aljaro, University of Barcelona, Spain

#### Reviewed by:

Meriyem Aktas, Ruhr-Universität Bochum, Germany Shaohua Chen, South China Agricultural University, China

\*Correspondence:

Fengquan Liu fqliu20011@sina.com †These authors have contributed

#### Specialty section:

equally to this work

This article was submitted to Microbial Physiology and Metabolism, a section of the journal Frontiers in Microbiology

> Received: 14 February 2019 Accepted: 17 May 2019 Published: 18 June 2019

#### Citation:

Ling J, Zhu R, Laborda P, Jiang T, Jia Y, Zhao Y and Liu F (2019) LbDSF, the Lysobacter brunescens Quorum-Sensing System Diffusible Signaling Factor, Regulates Anti-Xanthomonas XSAC Biosynthesis, Colony Morphology, and Surface Motility. Front. Microbiol. 10:1230. doi: 10.3389/fmicb.2019.01230 Lysobacter species are emerging as novel sources of antibiotics, but the regulation of these antibiotics has not been thoroughly elucidated to date. In this work, we identified a small diffusible signaling factor (DSF) molecule (LbDSF) that regulates the biosynthesis of a novel Xanthomonas-specific antibiotic compound (XSAC) in Lysobacter brunescens OH23. LbDSF was isolated from the culture broth of L. brunescens OH23, and the chemical structure of the molecule was determined by NMR and MS. The LbDSF compound induced GUS expression in a reporter strain of Xanthomonas campestris pv. campestris FE58, which contained the gus gene under the control of a DSF-inducible engXCA promoter. LbDSF production was found to be linked to the enoyl-CoA hydratase RpfF and dependent on the two-component regulatory system RpfC (hybrid sensor histidine kinase)/RpfG (response regulator), and LbDSF production was increased 6.72 times in the 1rpfC compared to wild-type OH23. LbDSF-regulated XSAC production was dramatically decreased in 1rpfF, 1rpfC, and 1rpfG. Additionally, a significant reduction in surface motility and a number of changes in colony morphology was observed in the 1rpfF, 1rpfC, and 1rpfG compared to the wild-type OH23. The exogenous LbDSF significantly increased XSAC production in wildtype OH23 and recovered the XSAC biosynthetic ability in 1rpfF. Taken together, these results showed that LbDSF is a fatty-acid-derived DSF that positively regulates XSAC biosynthesis, cell morphology, and surface motility. Moreover, the RpfC/RpfG quorumsensing signal transduction pathway mediates XSAC biosynthesis. These findings may facilitate antibiotic production through genetic engineering in Lysobacter spp.

Keywords: diffusible signaling factor, anti-Xanthomonas compound, colony morphology, surface motility, Lysobacter brunescens

## INTRODUCTION

fmicb-10-01230 June 15, 2019 Time: 17:44 # 2

The gliding Gram-negative Lysobacter bacteria are ubiquitous freshwater and soil microorganisms; because they are fastgrowing, simple to maintain, and genetically amenable for bioengineering, they are a promising source of novel bioactive natural antibiotics (Christensen and Cook, 1978; Xie et al., 2012). Some Lysobacter species have been demonstrated to be capable of producing antibiotics, including cyclodepsipeptides, cyclic lipodepsipeptides, cephem-type β-lactams, polycyclic tetramate macrolactams, phenazines, and lactivicins, which can inhibit the growth of plant pathogens.

Lysobactin is a cyclodepsipeptide that was first reported by O'Sullivan et al. (1988). The structure of lysobactin contains five natural amino acid residues and six non-proteinogenic amino acids, forming a macrocycle. Conversely, two families of cyclic lipodepsipeptides have been isolated from Lysobacter species (Kato et al., 1997; Hashizume et al., 2001; Zhang et al., 2011), both of which contain a β-hydroxyl fatty acid linked to the peptide moiety. The WAP-8294A family consists of 12 amino acid macrocycles, whereas the tripopeptin family is formed by eight amino acid cycles and a branched acyl group varying in length from 11 to 16 carbons (Xie et al., 2012). Cephem-type β-lactam antibiotics include cephalosporins, cephabacins, and cephamycins, but only cephabacins have been isolated from Lysobacter species (Harada et al., 1984; Ono et al., 1984). Cephabacins contain a cephem nucleus, which is linked to a non-ribosomal peptide by an ester linkage. Three different polycyclic tetramate macrolactams, xanthobacin, dihydromaltophilin (HSAF), and maltophilin, have been isolated from Lysobacter gummosus OH17 and Lysobacter enzymogenes OH11. All three of these polycyclic tetramate macrolactams contain γ-butyrolactam, which is involved in the macrocyclic structure (Meyers et al., 1985; Lou et al., 2011; Qian et al., 2013; Wang et al., 2013). Lysobacter cephabacins were shown to employ a novel mode of antibacterial action. These cephabacins specifically interfere with the biosynthesis of sphingolipids by targeting ceramide synthase, which causes thickening of the cell wall due to the accumulation of sphingolipid promoters that increase the degradation of chitin and block the elongation of the hyphal tips (Li et al., 2009). Several phenazines, including myxin and iodinin, have been found in Lysobacter antibioticus OH13 (Zhao et al., 2016). These phenazines differ not only in the presence of an N-oxide bond but also in the substituent at position 1 of the heterocycle. Finally, it is worth mentioning that lactivicin has a unique structure composed of a cycloserine ring linked to a γ-lactone ring (Harada et al., 1986). HSAF was shown to have a wide range of biological activities, including antibiotic, antifungal, and anticancer activities. Our research group also described a new cyclic lipodepsipeptide, WAP-8294A2, from L. enzymogenes, which exhibits strong inhibitory activity against methicillin-resistant Staphylococcus aureus (Zhang et al., 2011). WAP-8294A2 is currently in phase I/II clinical studies. WAP-8294A2 is produced by a large non-ribosomal peptide synthetase complex of 12 modules, which forms the linear assembly of amino acids. The biosynthetic pathway leading to the formation of 6 bioactive phenazines in L. antibioticus, starting from chorismic acid, was also reported by our research group (Zhao et al., 2016). After isolation of the compounds using reversephase HPLC, it was demonstrated that these phenazines exhibited strong activity against several bacteria, including Escherichia coli, Bacillus subtilis, and Xanthomonas spp.

In some prokaryotic systems, bacteria produce exogenous chemical signaling molecules to monitor population density and regulate a wide range of biological functions, such as secondary metabolite biosynthesis, biofilm formation, colony morphology, surface motility, and virulence (Abisado et al., 2018). LuxI/LuxRtype quorum-sensing systems are very common in Gramnegative bacteria, where they regulate the expression of genes through small chemical auto-inducers (Van Houdt et al., 2007). LuxI proteins are responsible for the synthesis of N-acyl homoserine lactone (AHL) quorum-sensing signals, and LuxR proteins are considered to be the main regulatory component of AHL quorum-sensing systems. These two proteins share two conserved regions, an AHL-binding domain and a DNA-binding domain (Maddocks and Oyston, 2008). Another group of quorum-sensing signals, the diffusible signaling factor (DSF) family, has recently been reported in a range of plant and human bacterial pathogens, including Xanthomonas campestris pv. campestris (Xcc), Xylella fastidiosa, Stenotrophomonas maltophilia, and L. enzymogenes (Barber et al., 1997; Deng et al., 2011; Robert and Dow, 2011; Han et al., 2015). In the DSF system, the putative enoyl-CoA hydratase RpfF is responsible for the synthesis of the DSF, and the RpfC/RpfG twocomponent system is involved in sensing and transducing DSF signals through a conserved phosphorelay mechanism (Holly et al., 2000). Different types of DSF signal molecules have been identified by electrospray ionization mass spectrometry (ESI-MS), gas chromatography, and nuclear magnetic resonance (NMR) analysis, such as cis-11-methyl-2-dodecenoic acid (DSF), cis-dodecenoic acid (BDSF), cis-11-methyldodeca-2,5-dienoic acid (CDSF), and 13-methyltetradecanoic acid (Wang et al., 2004; He et al., 2010; Han et al., 2015). Functional analysis of rpfF and rpfC mutants in different bacterial species suggests that the general role of the DSF-signaling system in virulence modulation is conserved, but the regulatory mechanisms and DSF-dependent traits may differ among taxa (Holly et al., 2000).

In this study, we isolated a novel Lysobacter brunescens strain, OH23, and active compounds from the culture supernatant of OH23 were shown to have strong specific activity against Xanthomonas species, whereas other bacteria and fungi, including Pseudomonas syringae pv. glycinea, P. syringae pv. lachrymans, Acidovorax citrulli, E. coli, Erwinia amylovora, Botryosphaeria dothidea, Phytophthora capsica, Valsa ambiens var. Pyri, and Colletotrichum gloeosporioide, remained unaltered (**Supplementary Figure S1**). We found that the genome of L. brunescens OH23 contained gene homologs to the regulation of pathogenicity factor (rpf) gene clusters of quorum-sensing genes from L. enzymogenes OH11. DSF signaling has been shown to be involved in the synthesis of metabolites with high pharmaceutical interest in L. enzymogenes OH11. This finding prompted us to investigate whether a DSF-dependent quorum-sensing signaling pathway was also responsible for the production of Xanthomonas-specific antibiotic compounds

(XSAC). To this end, we developed single in-frame mutants of the homolog DSF genes rpfF, rpfC, and rpfG and tested their ability to both produce DSF signals and synthesize XSAC. Our results revealed that L. brunescens uses an autoregulatory mechanism similar to L. enzymogenes to control DSF biosynthesis, suggesting the production of a DSF-like molecule in L. brunescens (Holly et al., 2000). We characterized the L. brunescens DSF signal as 13-methyltetradecanoic acid and demonstrated that this molecule regulates XSAC biosynthesis, surface motility, and cell morphology through the RpfC/RpfG signaling pathway.

### MATERIALS AND METHODS

### Bacterial Strains, Vectors, and Culture Conditions

The bacterial strains and plasmids used in this study are listed in **Table 1**. E. coli DH5α λpir, K-12, and S17-1 λpir were grown in LB (10 g tryptone, 5 g yeast extract, and 10 g sodium chloride, pH 7.0–7.2, in 1 L of distilled water) at 37◦C (Sambrook, 2001). L. brunescens OH23 (stored at China General Microbiological Culture Collection Center, Beijing, CGMCC No. 13677) and the genetically engineered 1rpfF, 1rpfC, and 1rpfG mutants were grown in nutrient broth (NB) medium (5 g peptone, 1 g yeast extract, 3 g beef extract, and 10 g sucrose, pH 7.0–7.2, in 1 L of distilled water) or in nutrient broth–yeast extract–glucose

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


(NYG) medium (5 g peptone, 3 g yeast extract, and 20 g glycerol, pH 7.0–7.2, in 1 L of distilled water) at 28◦C (Atlas, 1997). X. campestris pv.campestris FE58, Xanthomonas oryzae pv. oryzae PXO99A, X. oryzae pv. oryzae RS105, X. oryzae pv. oryzae KACC10331, X. campestris pv. campestris 8004, Xanthomonas axonopodis pv. glycines 12-2, P. syringae pv. glycinea PG4180, A. citrulli DSM17060, P. syringae pv. lachrymans 814/98, and E. amylovora ATCC15580 were also grown in nutrient broth (NB) medium at 28◦C. All solid media contained 1.5% agar, and antibiotics were added at the following concentrations: 20 µg/ml rifampicin and 8 µg/ml gentamicin for L. brunescens, 50 µg/ml rifampicin, and 10 µg/ml tetracycline for X. campestris pv. campestris FE58, and 20 µg/ml gentamicin for E. coli. Sucrose was added at a final concentration of 4% for the counter-selection of in-frame deletion strains.

### Generation of In-Frame Deletion Mutants

In-frame deletion plasmids were constructed by amplifying the flanking regions of specific genes by polymerase chain reaction (PCR) using Tks Gflex DNA Polymerase (TaKaRa Bio Inc., Kusatsu, Japan) and OH23 genomic DNA as the template according to the manufacturer's instructions. Briefly, PCR amplification was performed using 30 PCR cycles consisting of denaturation at 98◦C for 10 s, annealing at 55◦C for 15 s, and elongation at 68◦C for 1 min on a Bio-Red S1000 thermal cycler. The suicide vector pJQ200SK was digested with XbaI and BamHI (Thermo Fisher Scientific), and the target PCR fragments were ligated into the suicide vector using In-Fusion HD Cloning Plus (TaKaRa Bio Inc.). The recombinant vectors were transformed into E. coli DH5αλpir and confirmed using the universal primers M13F/M13R. The resulting plasmids were introduced into L. brunescens by conjugation. The deletion mutants 1rpfF, 1rpfC, and 1rpfG were selected for double homologous recombination events because the suicide vector contained a sacB counter-selectable marker (Quandt and Hynes, 1993). Finally, all mutants were confirmed by PCR using specific primers (**Table 2**).

### Growth Measurements

Lysobacter brunescens OH23 and the rpf mutants were cultured in NB medium at 28◦C with shaking at 180 rpm until the OD<sup>600</sup> was approximately 1.0 [which corresponds to approximately 10<sup>9</sup> CFU/ml (Colony Forming Units/ml)]. Then, 1 ml of culture for each strain was transferred into 50 ml of new liquid NB medium. The cultures were then incubated at 28◦C with shaking at 180 rpm. To measure the growth, the OD<sup>600</sup> value was determined every 12 h for each culture using a BioPhotometer Plus (Eppendorf, Germany) until each culture reached stationary stage. Three replicates were performed for each treatment, and the experiment was repeated three times.

### Bioassay for DSF Activities in a DSF Reporter Strain

The fermentation and isolation of LbDSF was assessed as previously described (Han et al., 2015). After extracting the DSF-dependent quorum-sensing signals in L. brunescens OH23,


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

fmicb-10-01230 June 15, 2019 Time: 17:44 # 4

we performed a DSF bioassay, as described previously with modifications (Wang et al., 2004). Briefly, L. brunescens OH23 and the 1rpfF, 1rpfC, and 1rpfG mutants were grown in NB liquid medium (25 L liquid medium in a CRJ-50D fermenter, twice, for a total of 50 L) at 28◦C at 180 rpm until the culture reached 2 × 10<sup>9</sup> CFU/ml (OD<sup>600</sup> = approximately 2.0). The culture broth was centrifuged at 12,000 rpm for 10 min, and the supernatant was extracted with the same volume of ethyl acetate. The extracted organic phase was evaporated at 47◦C, and the dry crude extract was partitioned using methanol and petroleum ether (100 ml each, three times). The petroleum ether phase was evaporated at 47◦C, and the resulting oil (2.1 g) was dissolved in 10 ml of dimethyl sulfoxide (DMSO). The DSFreporter strain, Xcc FE58, was grown in liquid NYG medium (5.0 g peptone, 3.0 g yeast extract, 20.0 g glycerol, and 1.0 L water, pH 7.2) for 2 days until the culture reached 3 × 10<sup>9</sup> CFU/ml (OD<sup>600</sup> = approximately 3.0).

The bioassay plates were prepared using the following steps. First, 0.8 g agarose powder was added to 100 ml 0.5 × NYG liquid medium. The medium was heated until the agarose was resolved, at which point 60 µl of X-gluc (60 mg/ml) and 2 ml of the reporter strain Xcc FE58 culture (3 × 10<sup>9</sup> CFU/ml) were added into the NYG agarose medium (the medium was cooled to 42◦C before use). This medium was used to prepare the 90-mm bioassay plates. Then, 5 µl of the DSF crude extract (10 mg crude extract resuspended in 200 µl of DMSO) from L. brunescens OH23, 1rpfF, 1rpfC, or 1rpfG mutants was added to each well. The bioassay plates were incubated at 28◦C for 24 h. The quantification of LbDSF was assessed as previously described (Wang et al., 2004). In brief, DSF activity was confirmed by the presence of a blue halo around the well and measured using the formula DSF (unit/ml) = 0.134 × e (1.9919W) , where W is the diameter of the blue halo zone and the width of the blue halo zone was increased while adding more exogenous LbDSF (**Figure 2B**).

One unit of DSF is equivalent to a 1-cm-diameter blue halo zone formed by the addition of 0.18 ± 0.07 µg exogenous LbDSF. We also added DMSO to the bioassay plate as a negative control. Three replicates were performed for each treatment, and the experiment was repeated three times.

### Isolation and Identification of LbDSF

The crude DSF extract was purified using HPLC. The HPLC detection conditions were as follows: reverse-phase HPLC (Shimadzu MS-2020, Tokyo, Japan) at 210 nm using a C-18 column (250 × 4.6 mm, Phenomenex). The mobile phase was 80% methanol in H2O from 0 to 13 min, 100% methanol in H2O from 13 to 20 min, and 80% methanol in H2O from 20 to 30 min (H2O containing 0.04% trifluoroacetic acid). The fraction corresponding to LbDSF appeared at 11.5 min.

Fractions containing the desired compound were evaporated to dryness under reduced pressure. The purified compound (4.3 mg) was dissolved in chloroform and characterized by mass spectrometry and NMR. High-resolution ESI-MS of the purified compound was performed in an AB (QTRAP 6500) instrument. The samples were directly infused into a mass spectrometer and analyzed in negative ion mode using a Turbo Ion Spray source. The NMR spectra ( <sup>1</sup>H NMR and <sup>13</sup>C NMR) were recorded in CD3OD using an AVANCE 600 (Bruker Company, Germany) instrument with a standard pulse program.

<sup>1</sup>**H NMR (600 MHz, CDCl**3**):** δ 2.35 (t, 2H, J = 7.8 Hz), 1.63 (qt, 2H, J = 7.2 Hz), 1.51 (sept, 1 H, J = 6.6 Hz), 1.37–1.24 (m, 16 H), 1.15 (m, 2 H), and 0.86 (d, 6 H, J = 6.6 Hz).

<sup>13</sup>**C NMR (150 MHz, CDCl**3**):** 179.97, 39.05, 34.01, 29.93, 29.69, 29.63, 29.59, 29.42, 29.23, 29.05, 27.97, 27.41, 24.67, and 22.65.

**MS (ESI):** calculated for C15H29O<sup>2</sup> [M-H]<sup>−</sup> 241.2173, found 241.2.

### Effect of the DSF-Dependent Quorum-Sensing System on XSAC Production

The ability of the wild-type and DSF mutants to produce XSAC was measured by an anti-Xanthomonas activity assay (diameter of inhibition zone) and analyzed using the agar diffusion method as described below. Pathogenic strains of Xanthomonas, including X. oryzae pv. oryzae PXO99A, X. oryzae pv. oryzae RS105, X. oryzae pv. oryzae KACC10331, X. campestris pv. campestris 8004, and X. axonopodis pv. glycines 12-2, were individually incubated in NB liquid medium until the culture OD<sup>600</sup> was approximately 1.0. L. brunescens OH23, and its derived mutants (1rpfF, 1rpfC, and 1rpfG mutants) were incubated in NB liquid medium or NB liquid medium supplemented with 2 µM LbDSF until approximately 1.5–2.0 × 10<sup>9</sup> CFU/ml (OD<sup>600</sup> = approximately 1.5–2.0). The cultures of L. brunescens OH23, 1rpfF, 1rpfC, and 1rpfG were centrifuged, and the supernatants were incubated at 85◦C for 30 min. For each pathogenic strain of Xanthomonas (**Table 1** and **Supplementary Table S1**), 100 ml of liquefied NB solid medium was incubated at 45◦C for 30 min, mixed with 10<sup>8</sup> cells, and then poured into plates. Then, 30 µl of supernatant from the cultures was spotted onto the selective plates. All plates were cultured at 28◦C, and the zones of inhibition on the plates were photographed and compared after 2 days. XSAC production was confirmed by the diameter of the inhibition zone, and all of the data regarding the inhibition zone were subtracted from the diameter of the oxford cup. Three replicates were performed for each treatment, and the experiment was repeated three times.

### Pathogenicity Assays in vivo

Oryza sativa ssp. indica rice cultivars IR24 and X. oryzae pv. oryzae PXO99<sup>A</sup> were used in the pathogenicity assay. O. sativa ssp. indica rice cultivars IR24 were grown under a 12-h light/dark cycle at 25◦C with approximately 70% relative humidity for 2 months. The preparation of the supernatants from cultures of L. brunescens OH23, 1rpfF, 1rpfC, and 1rpfG is described above. The leaves of IR24 plants were detached and dipped in the liquid culture of X. oryzae pv. oryzae PXO99<sup>A</sup> at a concentration of approximately 0.5 × 10<sup>9</sup> CFU/ml (OD<sup>600</sup> = approximately 0.5) for 1 h. To determine the XSAC production and activity of wild-type OH23, 1rpfF, 1rpfC, and 1rpfG, and chemically complemented strains, X. oryzae pv. oryzae PXO99A-infected rice leaves were treated with the respective supernatants every 24 h. For the negative control, X. oryzae pv. oryzae PXO99Ainfected rice leaves were treated with NB medium. IR24 plants were grown in a glasshouse with the same conditions as above, and lesion lengths were measured and photographed at 7 dpi. Three replicates were performed for each treatment, and the experiment was repeated three times.

### Detection and Comparison of Colony Morphology in L. brunescens

Single colonies of wild-type OH23, 1rpfF, 1rpfC, and 1rpfG strains were inoculated on NYG plates or NYG plates containing 5 µM LbDSF for 3 days at 28◦C. Then, the colony morphology of each strain was photographed and compared. Three replicates were used for each treatment, and the experiment was repeated three times.

### Observation of Surface Motility

The surface motility assay of L. brunescens wild-type OH23, 1rpfF, 1rpfC, and 1rpfG strains was performed as previously described (Song et al., 2017). Briefly, NB semi-solid medium containing 0.3% agar was used for surface motility assays, and 2.5 µl of L. brunescens wild-type OH23 or the derived mutants (10<sup>9</sup> CFU/ml, OD<sup>600</sup> was approximately 1.0 for all strains) was spotted onto the surface of NB semi-solid medium plates or NB semi-solid medium plates containing 5 µM LbDSF. The plates were incubated at 28◦C for 4 days, and then the surface motility of each strain was photographed, measured, and compared. Three replicates were performed for each treatment, and the experiment was repeated three times.

### RNA Extraction, Reverse Transcription PCR, and Real-Time-PCR

Lysobacter brunescens OH23, 1rpfF, 1rpfC, and 1rpfG mutants were each grown in 5 ml of NA medium to an OD<sup>600</sup> of approximately 1.0. Three milliliters of cells was transferred into a sterile centrifuge tube and centrifuged for 3 min at 12,000 rpm. RNA was extracted from the strains using TRIzol solution (TaKaRa Biocompany) following the manufacturer's instructions. For DNA removal and reverse transcription PCR, the PrimerScript RT Reagent Kit with the gDNA Eraser Kit (TaKaRa Biocompany) was used in this study. For the realtime PCR assay, a QuantStudio 6 Flex Real-Time PCR System (Thermo Fisher Scientific) was used to detect gene expression. The gene expression was calculated by the 2−11CT method. The primers for real-time PCR are listed in **Supplementary Table S2**. Three replicates were performed for each treatment, and the experiment was repeated three times.

### Data Analysis

Statistical analyses were calculated using SPSS (Statistical Package, Version 21.0). The variables were subjected to Student's t-test and tested for significance at P < 0.05 (<sup>∗</sup> ), P < 0.01 (∗∗), P < 0.001 (∗∗∗), and P < 0.0001 (∗∗∗∗).

### RESULTS AND DISCUSSION

### Identification of the Small Signaling Molecule LbDSF in L. brunescens

To identify the structure of the DSF signal in L. brunescens OH23, we harvested the L. brunescens culture from a 50-L fermenter system (CRJ-50D), collecting 25 L each time for a total of 50 L, and we extracted the crude DSF from the supernatant using the same method as previously described (Han et al., 2015). The methods of separation and identification of the DSF from OH23 were described previously (Wang et al., 2004). Briefly, we collected samples every 2 min from HPLC, concentrated every fraction by evaporation, and added 1 µg of every sample

into the DSF bioassay plates containing Xcc FE58 and X-gluc. The formation of a blue halo in the plates indicated the induction ability of each fraction. The results showed that the samples from 10.01 to 12.00 min induced gusA expression (**Figures 1A,B**). The DSF activity from 10.01 to 12.00 min was 0.53 ± 0.20 units. Furthermore, PengXCA-gusA was minimally induced by the addition of the samples from 0.01–10.00 or 12.01–16.00 min, and blue halo zones were not observed. To further confirm these results, we purified 1 µg of compound from these four fractions and detected the induction abilities of each fraction using the DSF bioassay system. The results of this assay revealed that PengXCA-gusA activation was induced only by adding the purified compound from fraction c, which exhibited a DSF activity of 1.98 ± 0.69 units.

We also examined whether the compound from these fractions enhanced the anti-Xanthomonas activity in L. brunescens OH23. We collected, dried, and dissolved every fraction in DMSO. The supernatant from strain OH23 formed a zone of inhibition in the growth of X. oryzae pv. oryzae RS105 that was 2.15 ± 0.09 cm in diameter. The zone of inhibition increased to 2.77 ± 0.12 cm after the addition of 2.0 µM purified fraction c (**Supplementary Figure S3**). Moreover, the zone of inhibition did not significantly increase when the purified compounds from fraction a, fraction b, fraction d, or the DMSO control were added (**Supplementary Figure S3**). These results demonstrated that only the compounds from fraction c, collected at 11.5 min, enhanced the inhibition activities in strain OH23 (**Supplementary Figure S3**).

To identify the structure of the unidentified DSF compounds, we first collected approximately 2.6 mg of the compounds exhibiting the highest DSF activity (fraction c, 11.5 min) from 50 L of crude supernatant and resuspended it in chloroform for NMR analysis. After purification of the compounds from fraction c, mass spectrometry analysis of the purified compound revealed a main m/z fraction at 241.2 Da, which was consistent with the expected molecular weight of LeDSF3, [M-H]<sup>−</sup> = 241.2173 Da (**Figure 1C**). In contrast with most DSFs, the structure of LeDSF3 consisted of a saturated aliphatic chain, which was formed by a tetradecanoic acid structure with a methyl group linked to carbon 13. To confirm the presence of LeDSF3, <sup>1</sup>H and <sup>13</sup>C NMR spectra were collected, and all signals were assigned (**Supplementary Figures S2A,B**). In agreement with the expected structure, no signal was detected between 7 and 5 ppm in the <sup>1</sup>H NMR spectrum, indicating that the purified compound was not an unsaturated aliphatic chain. A doublet with a 6.6-Hz coupling constant appeared at 0.86 ppm in the <sup>1</sup>H NMR spectrum, corresponding to the methyl groups. The protons in the alpha and beta position to the carboxylic acid were observed at 2.35 and 1.63 ppm as a triple and a quintuplet, respectively, whereas the proton of carbon **13** appeared at 1.51 ppm as a septuplet. <sup>13</sup>C NMR revealed a signal at 179.97 ppm, corresponding to the carboxylic acid. Carbon **2** was detected at 34.01 ppm, whereas the methyl groups appeared at 22.65 ppm in the <sup>13</sup>C NMR spectrum. In agreement with the <sup>1</sup>H NMR spectrum, no signal was detected between 180 and 120 ppm in the <sup>13</sup>C NMR spectrum, discarding the possibility of an unsaturated chain. Therefore, the primary compound from fraction c, named LbDSF, was determined to be 13-methyltetradecanoic acid (**Figure 1C**), which was originally reported in L. enzymogenes (Han et al., 2015).

### Identification of the rpf Gene Cluster in L. brunescens

As mentioned above, DSF family signals have been reported in different Xanthomonas bacteria. As a result, we compared and sequenced the genome of OH23 to identify genes potentially related to the LbDSF biosynthesis pathway. By blasting all rpf operon protein sequences<sup>1</sup> , we found that rpf proteins were highly conserved in Lysobacter spp. NCBI BLASTp analysis revealed that the Rpf proteins from L. brunescens had the highest homology to those of L. enzymogenes (taxid: 69), and the levels of protein identity were 72% (rpfF), 61% (rpfC), and 89% (rpfG). According to our BLASTp results, RpfF is an enoyl-CoA hydratase involved in the biosynthesis of DSF (**Figure 2A**). Accordingly, the DSF activity by wild-type OH23 was 0.88 ± 0.28 units, and the mutation of the rpfF gene abolished DSF production in L. brunescens (**Figure 2B**). RpfC is the hybrid histidine sensor kinase of the Rpf two-component regulatory system, which negatively regulates DSF production. Therefore, DSF accumulated in the 1rpfC mutant (**Figures 2A,B**), and DSF production of 1rpfC was 5.91 ± 2.29 units. RpfG is the response regulator of the Rpf two-component regulatory system and is responsible for signal transduction. Compared to wild-type OH23, DSF production was not altered in the 1rpfG mutant (**Figures 2A,B**), and the DSF activity of the 1rpfG was 0.63 ± 0.14 units.

Mutation of rpfF abolished the ability to produce DSFs in X. campestris pv campestris, which suggested that rpfF were essential for DSF production (Deng et al., 2011). In X. oryzae pv oryzae, the DSF production increased in the mutant of rpfC, and the DSF sensor kinase RpfC negatively regulated the biosynthesis of DSF (He et al., 2010; Deng et al., 2011). Taken together, these results indicate that the model for DSF production and signal transduction in L. brunescens is highly similar to that of L. enzymogenes and X. oryzae pv oryzae (He et al., 2010; Qian et al., 2013). In addition, the growth of wild-type OH23 and the rpf mutants was measured. The results demonstrated that the rpf mutants had growth similar to that of the wild-type strain under the tested conditions (**Figure 2C**).

### DSF-Dependent Quorum-Sensing System Positively Regulates Antibiotic Biosynthesis Through the Small Signaling Molecule LbDSF and the RpfC/RpfG Two-Component System

In previous findings, the genus Xanthomonas and several Gram-negative bacteria used the DSF-dependent quorumsensing system to mediate a diverse range of physiological activities related to virulence, motility, biofilm, and extracellular enzyme, and the DSF-dependent quorum-sensing system was also required for the biosynthesis of HSAF in L. enzymogenes (Qian et al., 2013; Guo et al., 2019). To investigate the

<sup>1</sup>https://blast.ncbi.nlm.nih.gov/Blast.cgi

function of the DSF-dependent quorum-sensing system in the biosynthesis of XSAC, we tested the anti-Xanthomonas abilities of the OH23 wild-type, 1rpfF,1rpfG, 1rpfC, and chemically complemented strains (mutants supplemented with LbDSF). As shown in **Figures 3A,B** and **Supplementary Figures S4–S7**, wild-type OH23 inhibited the growth of X. oryzae pv. oryzae

OH23 and its corresponding rpf mutants.

FIGURE 3 | Anti-Xanthomonas activity of L. brunescens OH23 and its corresponding rpf mutants with or without LbDSF addition. (A) Supernatants from L. brunescens OH23 or rpf mutant cultures with or without exogenous LbDSF supplementation were tested for their antimicrobial activity against X. oryzae pv. oryzae PXO99<sup>A</sup> . (B) The analysis of the images of X. oryzae pv. oryzae PXO99<sup>A</sup> growth inhibition zones shown in A. (C) Representative images of IR24 leaves infected with X. oryzae pv. oryzae PXO99<sup>A</sup> and treated with test supernatants. CK is the control, which was treated with water every 24 h. Images were taken 7 days post-inoculation (dpi). (D) Analysis of the lesion lengths on IR24 rice leaves caused by X. oryzae pv. oryzae PXO99<sup>A</sup> infection with or without treatment with test supernatants, as shown in C. Different numbers of star (<sup>∗</sup> ) above the bars indicate a significant difference between the wild-type strain OH23 and mutant strains (ns: not sigificant, ∗∗∗P < 0.001; ∗∗∗∗P < 0.001, t-test).

PXO99<sup>A</sup> (Xoo PXO99A), X. oryzae pv. oryzae RS105 (Xoo RS105), X. oryzae pv. oryzae KACC 10331 (Xoo KACC 10331), X. campestris pv. campestris 8004 (Xcc 8004), and X. axonopodis pv. glycines 12-2 (Xag 12-2). The diameters of the inhibition zones were 1.22 ± 0.10 cm, 1.75 ± 0.15 cm, 1.95 ± 0.24 cm, 0.29 ± 0.09 cm, and 1.50 ± 0.14 cm, respectively. The 1rpfF,1rpfG, and 1rpfC mutants lost the ability to inhibit all tested Xanthomonas strains, demonstrating that these mutants were impaired in the biosynthesis of XSAC (**Figure 3A** and **Supplementary Figures S4–S7**). To determine the genes relevant to the XSAC biosynthesis, we analyzed the putative operon that related to the biosynthesis of XSAC (data not shown), and 8 genes of 13 genes were indispensable for the biosynthesis of XSAC (**Supplementary Figures S8A,B**). Furthermore, the q-PCR results shown that the expression of all 13 genes, the genes for XSAC biosynthesis, were dramatically reduced in the 1rpfF mutant (**Supplementary Figure S8C**).

In L. enzymogenes, HSAF biosynthesis gene pks-nrps expression was reduced ∼10 times, and HSAF production was also dramatically reduced in the 1rpfFOH11 mutant (Qian et al., 2013). Taken together, the DSF quorum-sensing XSAC biosynthesis regulation model in L. brunescens OH23 was similar to the DSF quorum-sensing HSAF biosynthesis regulation model in L. enzymogenes OH11, and the DSF quorum-sensing system positively regulated the biosynthesis of XSAC (Qian et al., 2013; Han et al., 2015).

To further address whether the small signaling molecule LbDSF restored XSAC biosynthesis in the rpf mutants, we added LbDSF (2 µM) to the mutant culture in liquid NB medium, incubated the culture at 28◦C at 180 rpm, and then tested the anti-Xanthomonas activity of the supernatant. When LbDSF was added into the cultures, the growth inhibitory ability of wild-type OH23 increased 35.01, 19.54, 8.97, 16.86, and 27.15% against Xoo PXO99A, Xoo RS105, Xoo KACC 10331, Xcc 8004, and Xag 12- 2, respectively (**Figure 3A** and **Supplementary Figures S4–S7**). LbDSF addition restored the anti-Xanthomonas activity in the 1rpfF, and the diameter of the zones of inhibition were 0.84 ± 0.22 cm, 1.71 ± 0.16 cm, 0.36 ± 0.10 cm, 0.25 ± 0.11 cm, and 1.18 ± 0.16 cm against Xoo PXO99A, Xoo RS105, Xoo KACC 10331, Xcc 8004, and Xag 12-2, respectively (**Figure 3A** and **Supplementary Figures S4–S7**). Additionally, exogenous LbDSF had no effect on the anti-Xanthomonas activity in 1rpfC and 1rpfG (**Figure 3A**). These results were consistent with the fact that the small signaling molecule LbDSF is a transduction signal in the L. brunescens rpf system, where RpfF is involved in DSF biosynthesis, and RpfC/RpfG is the two-component system involved in signal transduction.

We then investigated whether all of the supernatants mentioned above had inhibitory activities in vivo. Xoo PXO99Ainfected IR24 rice leaves were treated with the supernatants every 24 h, and lesion lengths were measured at 7 dpi. As shown in **Figures 3C,D**, the negative control was only treated with NB medium, and the lesion length was 2.92 ± 0.41 cm. However, the wild-type OH23 treatment group showed significantly decreased lesion lengths (0.70 ± 0.21 cm). The addition of exogenous LbDSF or DMSO to the culture of wild-type OH23 was shown to significantly influence the lesion length relative to the negative control. For the 1rpfF treatment group, exogenous LbDSF restored the anti-Xanthomonas activity, and the lesion length was 0.83 ± 0.23 cm. However, both the 1rpfG and 1rpfC treatment

groups exhibited similar lesion lengths as the negative control. The results from the in vivo plant assays were consistent with the results from the anti-Xanthomonas ability on plates.

In L. enzymogenes, LeDSF3 was 13-methyltetradecanoic acid, and acted as an extracellular signal to positively regulate the biosynthesis of HSAF; the two-component regulatory system RpfC/RpfG sensed and transduced LeDSF3; and the global regulator Clp was downstream of the LeDSF quorumsensing system and also played a positive role in regulating the biosynthesis of HSAF and WAP-8294A2 (Han et al., 2015; Xu et al., 2016). To further investigate the function of clp in L. brunescens OH23, the ClpOH11 amino acid sequence was compared with the draft genome sequence of L. brunescens OH23, and Peg.2300 (Clp) shared 83% similarity to that of ClpOH11 at the amino acid level (**Supplementary Figure S9A**). Next, the anti-Xanthomonas abilities of 1clp and its complementary strain were tested, and clp had no significant effect on biosynthesis of XSAC (**Supplementary Figure S9B**), which revealed that Clp was not involved in the regulation of XSAC biosynthesis. Taken together, these findings indicate that the structure of LbDSF and the regulation of RpfC/RpfG L. brunescens were similar to the LeDSF quorum-sensing HSAF biosynthesis regulation model in L. enzymogenes and a potential novel transcription regulator in the DSF-dependent quorum-sensing system that regulated XSAC biosynthesis in L. brunescens.

### DSF-Dependent Quorum-Sensing System Affects Colony Morphology in L. brunescens

The DSF quorum-sensing system was shown to influence colony morphology in L. enzymogenes (Qian et al., 2013). To investigate the function of the DSF-dependent quorum-sensing system in modulating colony morphology in L. brunescens, we tested the wild-type OH23 strain and the 1rpfF, 1rpfC, and 1rpfG mutants. As shown in **Figure 4**, the wild-type OH23 displayed round colonies with lobular and spiculated boundaries on NYG plates with an average diameter size of 1.87 ± 0.05 cm. However, the 1rpfF, 1rpfC, and 1rpfG mutants showed smooth colony appearance with an average diameter size of 0.64 ± 0.08 cm under the same growth conditions.

Exogenous LbDSF (2 µM) restored the colony morphology in the 1rpfF when compared to the wild-type OH23 with an average diameter size of 1.32 ± 0.08 cm, whereas the addition of exogenous LbDSF did not influence the colony morphology of the 1rpfC and 1rpfG (**Figure 4**). These results indicate that the DSFdependent quorum-sensing system is involved in the regulation of colony morphology in L. brunescens.

### DSF Signaling Controls Surface Motility Through Type IV Pili (T4P) in L. brunescens

The DSF-dependent quorum-sensing system was also shown to affect cell motility in L. enzymogenes (Qian et al., 2013). To investigate the function of the DSF-dependent quorumsensing system in regulating motility in L. brunescens, we tested

the motility of the wild-type OH23, 1rpfF, 1rpfC, and 1rpfG strains. Since OH23, unlike L. enzymogenes OH11, does not exhibit twitching motility (data not shown), we analyzed the surface motility of wild-type OH23 and its derivative mutants on NB semi-solid (0.3% agar) motility medium plates for 4 days incubated at 28◦C.

NYG plates contained 5 µM LbDSF. Each treatment was completed in

triplicate, and the experiment was repeated three times.

As shown in **Figure 5A**, the wild-type OH23 strain was motile in motility medium plates with a typical circular dissemination pattern from the point of inoculation that was 2.78 ± 0.33 cm in size. However, the surface motility of 1rpfF, 1rpfC, or 1rpfG was substantially reduced. These mutants only reached an average surface motility diameter of approximately 0.82 ± 0.11 cm in 4 days, indicating that the mutants had a 73.72% reduction in the diameter of their surface motility zone compared to the wildtype OH23. Exogenous LbDSF (2 µM) restored surface motility in the 1rpfF strain compared to the wild-type OH23 strain with an average surface motility diameter of 2.42 ± 0.10 cm. However, exogenous LbDSF did not exhibit any effect on the surface motility in the 1rpfC and 1rpfG mutants. These results indicate that the DSF-dependent quorum-sensing system is involved in the regulation of surface motility in L. brunescens.

Type IV Pili has been shown to be important for surface motility in diverse bacteria. Since the pilA gene encodes the major pilin subunit of T4P (Mattick, 2002; Burdman et al., 2011; Wang et al., 2014), we measured pilA<sup>1</sup> expression in the wild-type OH23, 1rpfF, 1rpfC, and 1rpfG. As shown

in **Figure 5B**, pilA<sup>1</sup> expression was dramatically decreased in 1rpfF, 1rpfC, and 1rpfG. The addition of exogenous LbDSF (2 µM) partially restored pilA<sup>1</sup> expression in the 1rpfF compared to wild-type OH23, while exogenous LbDSF did not influence pilA<sup>1</sup> expression in the 1rpfC and 1rpfG. These results indicate that the DSF-dependent quorum-sensing system is involved in the regulation of pilA<sup>1</sup> expression in L. brunescens.

biosynthesis, surface motility, and cell morphology.

The DSF-dependent quorum-sensing system in the Xanthomonas genus differs substantially between species and has been shown to positively regulate virulence, biofilm formation, EPS biosynthesis, and adaption. Moreover, specific DSF molecules were related to specific antibiotic HSAF biosynthesis (He et al., 2010; Han et al., 2015). The Rpf system was shown to be activated by adding specific DSF molecules, and EPS production, extracellular xylanase activity, or antibiotic HSAF biosynthesis were restored (He et al., 2010; Han et al., 2015).

## CONCLUSION

In this study, we report the role of a quorum-sensing system in the production of a novel XSAC in the ubiquitous environmental bacterium L. brunescens (**Figure 6**). Our data revealed that L. brunescens OH23 uses a DSF-dependent quorum-sensing molecule to regulate XSAC production. We characterized this DSF compound as 13-methyltetradecanoic acid. This extracellular signal is produced by RpfF and transduced by the RpfC/RpfG two-component regulatory system. This molecule also regulates surface motility and colony morphology. Our findings will be useful in applied genetics and molecular biotechnology, thereby facilitating the improvement of antibiotic production in Lysobacter spp., which can potentially be used in the agricultural industry. Furthermore, the tight control that our identified DSF molecule exerts on XSAC expression suggests that Lysobacter spp. may produce other quorum-sensing signals that can induce the production of additional novel bioactive compounds.

### DATA AVAILABILITY

The datasets generated for this study can be found in Genbank, MK532476, MK532477, and MK532478.

## AUTHOR CONTRIBUTIONS

JL, RZ, PL, TJ, YJ, and YZ conducted the experiments. FL designed and conducted the experiments. JL, PL, and FL contributed to the writing of the manuscript. FL revised the manuscript.

### FUNDING

This study was supported by the Jiangsu Agricultural Science and Technology Innovation Funds [CX (16) 1049], the Jiangsu Provincial Key Technology Support Programme (BE2015354), the "948" Project of the Ministry of Agriculture (2014- Z24), the Earmarked Fund for China Agriculture Research System (CARS-28-16), and the China Postdoctoral Science Foundation (2017M610310).

### REFERENCES

fmicb-10-01230 June 15, 2019 Time: 17:44 # 13

Abisado, R. G., Benomar, S., Klaus, J. R., Dandekar, A. A., and Chandler, J. R. (2018). Bacterial quorum sensing and microbial community interactions. mBio 9:e02331-17. doi: 10.1128/mBio.02331-17

Atlas, R. M. (1997). Handbook of microbiological media. Q. Rev. Biol. 2, 364–365.


### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2019.01230/full#supplementary-material


enzymogenes OH11. Biol. Control 120, 52–58. doi: 10.1016/j.biocontrol.2016. 08.006


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

Copyright © 2019 Ling, Zhu, Laborda, Jiang, Jia, Zhao and Liu. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(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.

# AHLs Regulate Biofilm Formation and Swimming Motility of Hafnia alvei H4

Yao lei Zhu1,2, Hong man Hou1,2 \*, Gong liang Zhang1,2, Yi fang Wang1,2 and Hong shun Hao<sup>2</sup>

<sup>1</sup> School of Food Science and Technology, Dalian Polytechnic University, Dalian, China, <sup>2</sup> Liaoning Key Lab for Aquatic Processing Quality and Safety, Dalian, China

The aim of this study was to evaluate the role of N-acyl homoserine lactones (AHLs) in the regulation of swimming motility of Hafnia alvei H4 and its biofilm formation on 96-well plate, glass and stainless-steel surfaces. The luxI gene, which codes for an enzyme involved in AHL synthesis, was deleted to generate a luxI mutant (1luxI). The mutant produced no AHL, and the relative expression of the luxR gene was significantly (P < 0.05) decreased. In addition, qRT-PCR analysis showed that the relative expression of the luxR gene in 1luxI was stimulated by the presence of exogenous AHLs (C4-HSL, C6-HSL, and 3-o-C8-HSL) added at concentrations ranging from of 50–250 µg/ml. Among the three AHLs, C6-HSL had the strongest effect. The ability of 1luxI to form biofilm on 96-well plate, glass and stainless-steel surfaces was significantly reduced (P < 0.05) compared with the wild type (WT), but was increased when provided with 150 µg/ml C4-HSL, whereas C6-HSL and 3-o-C8-HSL had no effect. Scanning electron microscopy analysis of the biofilm revealed less bacteria adhering to the surface of stainless-steel and fewer filaments were found binding to the cells compared with the WT. Furthermore, 1luxI also exhibited significant (P < 0.05) decrease in the expression of biofilm- and swimming motility-related genes, flgA, motA and cheA, consistent with the results observed for biofilm formation and swimming motility. Taken together, the results suggested that in H. alvei H4, C4-HSL may act as an important molecular signal through regulating the ability of the cells to form biofilm, as well as through regulating the swimming motility of the cell, and this could provide a new way to control these phenotypes of H. alvei in food processing.

Keywords: Hafnia alvei, AHLs, quorum sensing, biofilm, swimming motility

## INTRODUCTION

Quorum sensing (QS) is a cell-to-cell communication system used by bacteria, and it is widely by both Gram-negative and Gram-positive bacteria (Ammor et al., 2008). Bacteria secret several kinds of chemical compounds that can act as signaling molecules [autoinducers (AIs)]. N-acyl homoserine lactone (AHL), also known as AI-1, is secreted by Gram-negative bacteria, and the communication mechanism of this compound involves AHL synthase (LuxI) and the transcription factor LuxR, which is responsible for controlling gene expression in the presence of AHLs. The

#### Edited by:

Tom Defoirdt, Ghent University, Belgium

#### Reviewed by:

Maria Cristina D. Vanetti, Universidade Federal de Viçosa, Brazil Julia Van Kessel, Indiana University Bloomington, United States

> \*Correspondence: Hong man Hou houhongman@dlpu.edu.cn

#### Specialty section:

This article was submitted to Microbial Physiology and Metabolism, a section of the journal Frontiers in Microbiology

> Received: 29 November 2018 Accepted: 28 May 2019 Published: 19 June 2019

#### Citation:

Zhu Yl, Hou Hm, Zhang Gl, Wang Yf and Hao Hs (2019) AHLs Regulate Biofilm Formation and Swimming Motility of Hafnia alvei H4. Front. Microbiol. 10:1330. doi: 10.3389/fmicb.2019.01330

**121**

LuxI/LuxR system has become the model system of AHLsmediated quorum sensing, and the quorum-sensing system of Gram-negative is based on this system. LuxI synthesizes AHLs and it is encoded by the luxI gene. LuxR is encoded by the luxR gene, and it acts by binding to AHLs, thereby stimulating the expression of these genes in the presence of AHLs. The LuxI/LuxR complex is responsible for the up- or down-regulation of multiple target genes, such as those that code for pectinase, cellulase, and protease (Swift et al., 2001). Autoinducer-2 (AI-2) is synthesized from 4,5-dihydroxy-2,3-pentanedione (DPD) by LuxS, and it is used by Gram-negative and Gram-positive bacteria in interspecies communication. Peptides and derived peptides, generally serve as signaling molecules in Gram-positive bacteria (Bai and Rai, 2011).

Biofilm is a bacterial self-protection growth pattern and it is formed by the aggregation of bacterial cells within an extracellular matrix, which is mainly made of exopolimers (EPS) (Wang J. et al., 2016), and the adherence of bacterial cells to a solid surface depends on the EPS that the cells secret (Jung et al., 2013). In general, some pathogens and spoilage bacteria can adhere to the solid surfaces that can come into contact with food, such as the surfaces of food processing machines and packaging materials. These bacteria may then form biofilms, and the biofilms will allow the cells to become more resistant to cleaning treatments, and enable them to contaminate the food during subsequent processing (Gounadaki et al., 2008; Bai and Rai, 2011). This will effectively facilitate the transmission of the bacteria to the consumers via the contaminated food, eventually causing infections. Biofilms have been recognized as a frequent source of bacterial infections (Costerton et al., 1999). According to a report by Janssens et al. (2008), nearly 80% of persistent bacterial infections in the US were found to be related to biofilms. The formation of biofilm is a multi-step process, which consists of initial attachment, irreversible attachment, early development of biofilm architecture (microcolony formation), maturation and dispersion (Srey et al., 2013). Quorum sensing appears to be involved in all the steps of the process. Promotion and inhibition of biofilm formation by exogenous AHLs have been reported for Shewanella baltica (Zhao et al., 2016), Serratia A2 and Aeromonas B1 (Zhang et al., 2016), Vibrio parahaemolyticus (Bai and Rai, 2016), and Pseudomonas sp. HF-1 (Wang et al., 2012), suggesting that QS has a regulatory role in biofilm formation.

H. alvei is a Gram-negative, short-rod-shaped, flagellated bacterium that belongs to the family Enterobacteriaceae, which is considered as an opportunistic pathogen of humans and animals (Tan et al., 2014). However, despite being classified as a member of the Enterobacteriaceae family, H. alvei is still far from being virulent and pathogenic (Vivas et al., 2008). H. alvei is a common bacterial food contaminant (Liu et al., 2006), and it has been frequently isolated from spoiled food products, especially in chill-stored proteinaceous raw food, like refrigerated spherical fish paste (Tan et al., 2014), vacuumpacked beef (Bruhn et al., 2004) and raw milk (Viana et al., 2009). The strong tendency of H. alvei to adhere to solid surface and to form biofilm has been reported by Viana et al. (2009) and Hou et al. (2017), and it is considered to be a potentially important factor that causes food contamination and food spoilage. Therefore, it is necessary to look for effective ways to control biofilm formation. To our knowledge, fewer studies have studied the regulatory mechanism of quorum sensing of H. alvei with respect to biofilm formation and the motility of the cells in an artificial medium. Understanding more about the mechanism by which quorum sensing can impact biofilm formation will open up a new way to tackle the problem of food contamination by bacteria, and help safeguard better food quality and prevent food-poising.

In our previous study, we isolated a strain of H. alvei (H. alvei H4) from spoiled instant sea cucumber, and identified three kinds of AHLs secreted by this bacterium. These AHLs are C4-HSL, C6- HSL, and 3-o-C8-HSL. In addition, we also detected a significant influence of AHLs on the biofilm formation of H. alvei H4 (Hou et al., 2017). In this study, a luxI mutant of H. alvei H4 was constructed to conduct further research on the regulatory roles of C4-HSL, C6-HSL, and 3-o-C8-HSL in biofilm formation and swimming motility of H. alvei H4.

### MATERIALS AND METHODS

### Bacterial Strains and Culture Conditions

The bacterial strains used in this study are presented in **Table 1**. Chromobacterium violaceum CV026, and H. alvei H4 were routinely cultured at 30◦C, while Escherichia coli was grown at 37◦C. All strains were grown in LB medium (Luria Bertani, 10 g/l tryptone, 5 g/l yeast extract, 10 g/l NaCl) supplemented with antibiotics where appropriate (50 µg/ml ampicillin and 34 µg/ml chloramphenicol in the case of E. coli culture or 20 µg/ml chloramphenicol for the H. alvei H4 mutant).

### Construction of 1luxI Strain

To construct a luxI-deficient strain of H. alvei H4, a chloramphenicol resistance marker (Cm<sup>R</sup> ) was inserted into the genomic DNA of H. alvei H4 at the luxI locus. Briefly, a 608-bp upstream homologous recombination arm and a 654-bp downstream homologous recombination arm of the luxI gene were amplified from the gDNA of H. alvei H4 and then cloned into the plasmid pUC19 to yield the construct pUC19-1luxI. The two DNA fragments were linked by a BamHI restriction

TABLE 1 | Strains and plasmids used in this study.


site. The chloramphenicol resistance maker (Cm<sup>R</sup> ) was amplified from the plasmid pKD3, and then cloned into pUC19-1luxI at a site between the upstream and downstream homologous recombination arm fragments, yielding the construct pUC19- 1luxI::Cm. Target fragment in pUC19-1luxI::Cm was subcloned and ligated into the suicide plasmid pCVD442 to yield the construct pCVD442-1luxI::Cm, which was then introduced into E. coliβ2155 by electroplating. Conjugation between E. coliβ2155 harboring pCVD442-1luxI::Cm and H. alvei H4 was then performed. The mutant colonies obtained were verified by PCR and DNA sequencing. Primers used in this section see **Table 2**.

### AHLs Production

Chromobacterium violaceum CV026 was used as the biosensor strain to detect the production of AHLs by H. alvei H4 WT and 1luxI, since in the presence of AHLs, C. violaceum CV026 would produce a purple pigment that could be easily detected. This assay was performed as described by Viana et al. (2009). Briefly, AHLs were extracted from the corresponding bacterial culture. Aliquot (100 ml) of an overnight culture was centrifuged at 8000 × g for 15 min at 4◦C, and the supernatant was added to an equal volume of ethyl acetate containing 0.1% acetic acid (v/v) followed by thorough mixing. The mixture was then incubated at 25◦C for 2 h with shaking at 180 rpm. After that, the ethyl acetate layer was removed and freeze-dried under vacuum, and the residue was dissolved in 1 ml ultra-pure water. To prepare the plates for the assay, 80 ml LB agar medium was cooled to about 60◦C, mixed with 20 ml overnight culture of CV026, and then poured into sterile plates (20 ml per plate). Holes were punched into the solidified medium at the center of each plate using a sterile 1 ml pipette tip, and 60 µl of the AHL extract was dispensed into the hole. The plates was incubated at 30◦C until the purple zone appeared around the point of AHL application.

### Biofilm Formation on 96-Well-Plate

Biofilm formation was measured using the microplate assay as described by Hou et al. (2017), but with some modifications. Briefly, culture of H. alvei H4 WT and 1luxI were grown to an OD600 nm value of 1.0, and then diluted 100-fold in LB medium. The diluted culture was dispensed into a 96-well plate polypropylene microtiter plate (Corning, NY, United States) using 200 µl per well. After incubation at 30◦C for 24 h, the OD600 nm of the culture was measured. After that, the

TABLE 2 | Primers used for the construction of 1luxI.

culture medium was removed and washed off unattached cells by washing each well with 250 µl of 10 mM PBS (pH 7.2). A total of three washes were performed. This was followed by the addition of 250 µl anhydrous methanol and a 15-min incubation to fix the cells. Subsequently, 250 µl of 0.1% (v/v) crystal violet solution was added to each well and the plate was incubated at room temperature for 15 min and rinsed three times with deionized water (250 µl per rinse). The crystal violet was dissolved by the addition of 200 µl 33% glacial acetic acid followed by shaking at 300 rpm for 15 min. Biofilm formation was finally analyzed by measuring the absorbance of the plate at OD590 nm using a Spectra M2 spectrophotometer (Molecular Devices, United States). Biofilm formed by 1luxI in the presence of 150 µg/ml C4-HSL, 200 µg/ml C6-HSL, or 100 µg/ml 3-o-C8- HSL was detected essentially as described above.

### Biofilm Formation on Glass Surface

To examine the effect of quorum sensing on biofilms formed by H. alvei H4 on the glass surface, biofilm formation was investigated as described by Cai et al. (2018) with some modifications. Briefly, an overnight bacterial culture was diluted 100-fold in fresh LB medium, and 2 ml of the diluted culture was added to a glass tube (1 cm × 10 cm). As a control, 2 ml of LB was added to a separate glass tube. Both tubes were incubated at 30◦C for 24 h without shaking. After incubation, the culture medium was gently removed, and each tube was washed three times with 2.5 ml PBS, and then with 2.5 ml anhydrous methanol followed by drying at 60◦C for 15 min. The tube was then stained with 2.5 ml 0.1% CV (v/v) for 15 min at room temperature followed by three washes with 2.5 ml ultra-pure water and drying at 60◦C. The remaining CV on the inner surface of the tube was dissolved in 1.0 ml of 33% glacial acetic acid (v/v) with vortexing, and 200 µl of the sample was transferred to a new 96-well plate and the optical density of the sample at 590 nm was measured with a microplate reader (Spectra M2; Molecular Devices, Sunnyvale, CA, United States).

### Biofilm Formation on Stainless Steel

Stainless steels (type: 304) were cut into strips (1 cm × 3 cm × 0.2 mm) and processed as described by Tapia-Rodriguez et al. (2017). An overnight culture of H. alvei H4 WT and 1luxI were diluted 100-fold with LB medium to a cell density of about 10<sup>6</sup> CFU/ml. Aliquot (10 ml) of the diluted culture was placed in a test tube (1.5 cm × 10 cm) containing


a stainless-steel strip and incubated at 30◦C for 24 h without shaking. After incubation, the strip was transferred to a small test tube (1.5 cm × 5 cm) containing 5 ml PBS to wash off unattached cells. This washing step was repeated three times and the strip was then transferred to a 10 ml-tube and sonicated for 1.5 min in an ultrasonication bath (power, 300 W, 37 kHz 37; Elma; Elmasonic P; Germany) to disperse the biofilm as previously described (Jahid et al., 2015). Finally, the bacterial suspension was vortexed and serially diluted with 0.85% NaCl solution, and then spread onto an LB agar plate. The plate was incubated at 30◦C for 24 h and the colonies appearing on the plate were then counted.

### Biofilms Detected by SEM

H. alvei H4 WT and 1luxI were separately incubated in LB medium for 16 h at 30◦C with shaking at 150 rpm. The culture was then diluted to a final density of about 1.0 × 10<sup>7</sup> CFU/ml. First, 1 ml of the diluted culture was dispensed onto a sterile stainless steel (1 cm × 1 cm × 0.2 mm) placed inside a 24-well microtiter plate and cultured at 30◦C for 24 h. The biofilm deposited on the stainless steel was then was prepared for scanning electron microscopy (SEM) analysis according to the method described by Nithya et al. (2010). Briefly, the stainless steel was gently washed with sterile 0.1 M PBS, and the biofilm was fixed with 2.5% glutaraldehyde solution for 2 h followed by washing with 0.1 M PBS. The stainless steel was subjected to dehydration under a graded series of tert-butanol. This was followed by critical point drying, gold sputtering and SEM observation (Quanta 450, Waltham, MA, United States) performed at 3.0 kV under 5000 and 50000× magnifications as described previously (Hou et al., 2018).

### Swimming Motility Assay

To measure the swimming motility of H. alvei H4 and 1luxI, motility agar (2.6 g/l agar, 10 g/l tryptone, 5 g/l NaCl) (Zhu and Winans, 1998) was used. H. alvei H4 or 1luxI was incubated in LB medium at 30◦C until the OD600 nm of the cultures reached 1.0. Aliquot (3 µL) of the culture was spotted at the center of the motility agar plate followed by incubation at 30◦C. The extent of motility was assessed by measuring the diameter of the zone spread from the point of inoculation. To detect the effect of AHLs on the swimming motility of 1luxI, C4-HSL, C6-HSL, and 3-o-C8-HSL were added to separate motility agar plates. The final concentrations of C4-HSL, C6-HSL, and 3-o-C8-HSL in the plates were 150, 200, and 100 µg/ml, respectively. Swimming motility was determined as described above.

### Real-Time PCR Assay

H. alvei H4 WT and 1luxI was cultured in LB medium the absence or presence of AHLs (C4-HSL, C6-HSL, and 3-o-C8-HSL) at 30◦C with shaking at 150 rpm for 16 h. When the OD600 nm of the cultures reached 1.65, the cells were harvested and subjected to total RNA extraction using a RNAprep Pure Bacteria Kit (DP430, TIANGEN, Beijing). About 500 ng of total RNA was reversely transcribed into cDNA using a PrimeScriptTM Reagent kit (RR047, Takara, Japan). The reaction mixture consisted of 2 µL template cDNA, 12.5 µL SYBR Premix Ex TaqTM (RR420, Takara, Japan), 0.5 µL each of the forward and reverse primers (10 mM) (**Table 3**), and 9.5 µL RNA free water. Amplification was performed with a Step-One Thermal Cycler (Applied Biosystems, United States) and consisted of 40 cycles of denaturation at 95◦C for 15 s, annealing at 95◦C for 30 s, and extension at 60◦C for 45 s. The 16S rRNA gene was used as housekeeping control. The result was analyzed by 2−11CT method (Schmittgen and Livak, 2008).

### Statistical Analysis

Three replicate trials were carried out for each sample, and all the experiments were repeated three times. Data were analyzed by one-way analysis of variance (ANOVA) using SPSS18.0 software, and expressed as means ± standard deviations (SDs). Statistical significance was considered at the P < 0.05 level. All graphs were drawn with OriginPro 8.6 software.

## RESULTS

### AHLs Production by luxI Mutant

Pigment production assay showed a violet zone in the center of the plate when CV026 was incubated with the ethyl acetate extract prepared from the culture supernatant of H. alvei H4 but not from the culture supernatant of 1luxI (**Figure 1**), indicating the lack of AHLs production by 1luxI. The assay demonstrated the dependence of AHLs production on a functional luxI gene.

TABLE 3 | Primers used for RT-PCR.


CV026 was cultured in the absence of ethyl acetate extract prepared from LB medium (A), H. alvei H4 culture supernatant (B), and 1luxI culture supernatant (C).

### Response of luxR to AHLs

fmicb-10-01330 June 17, 2019 Time: 17:32 # 5

The effect of luxI mutation was further investigated by measuring the level of luxR mRNA in 1luxI and compared it with that of WT strain. 1luxI exhibited about fivefold reduction in the level of luxR mRNA relative the WT (**Figure 2A**), which clearly suggested that the expression of the luxR gene in the mutant was significantly (P < 0.05) inhibited. However, in the presence of exogenous AHLs, the level of luxR mRNA gradually increased with increasing concentrations of AHLs, suggesting that the expression of luxR could be stimulated by the presence of AHLs (**Figure 2B**). Maximum increase in luxR mRNA level stimulated by AHL ranged from 5.5-fold in the case of C4-HSL to 6.5-fold in the case of C6-HSL, while 3 o-C8-HSL yielded somewhat lower increase (4.5-fold). Thus, C6-HSL appeared to exert the strongest stimulatory effect on the expression of luxR.

### Biofilm Formation

Growth of H. alvei H4 was not affected by the deletion of the luxI gene, as shown by the similar growth curves between WT and 1luxI in the absence of AHLs (**Figure 3A**). Furthermore, the addition of AHLs to the culture of 1luxI also had not obvious effect on its growth. This suggested that growth of the bacterial cells was not dependent on the product of the luxI gene. However, deletion of the luxI gene had a significant effect on biofilm formation by H. alvei H4, as shown by the significantly higher level of biofilm formed by the wild type (WT) compared with 1luxI (**Figure 3B**). Addition of AHLs to the culture of 1luxI appeared to cause increase in biofilm formation. The extent of biofilm formation of 1luxI in the presence of C4-HSL was significantly (P < 0.05) enhanced and almost similar to that of the WT strain. However, C6-HSL seemed to

have no effect on the biofilm formation of 1luxI, causing no significant increase. Obvious promotion (P < 0.05) of biofilm formation on 1luxI were also achieved at the presence of 3-o-C8-HSL, but still far less than that of C4-HSL on 1luxI and WT strain. The result suggested that C4-HSL could be the AHL with major influence on biofilm formation.

### Biofilms on Glass

An intense zone of purple stain was found on the wall of the glass tube used to culture WT strain, indicating the presence of biofilm (**Figure 4A**). On the other hand, only very slight purple staining was found on the wall of the glass tube that was used to culture 1luxI, suggesting a lack of biofilm being formed by 1luxI. Addition of C4-HSL to the 1luxI culture resulted in obvious enhancement of the purple zone, but this did not occur when either C6-HSL or 3-o-C8-HSL was added to the culture. The result clearly suggested that the capacity of 1luxI to form biofilm was greatly reduced compared with its WT strain counterpart, and that only C4-HSL was able to restore its biofilm formation capacity.

### Biofilms on Stainless Steel

The mutant 1luxI was also compared with its WT counterpart in term of the ability to form biofilm on the surface of stainless steel. Significantly (P < 0.05) less colonies of 1luxI were found on the surface of the stainless steel compared with the WT strain, but when 1luxI was cultured in the presence of C4-HSL, the number of colonies found on the surface of the stainless steel was significantly (P < 0.05) increased, and being comparable with that of WT strain (**Figure 5**). However, as in the case of biofilm formed on glass surface, C6-HSL and 3-o-C8-HSL had no obvious effect on the ability of 1luxI to form biofilm on stainless steel surface.

### Biofilm by SEM

Biofilm formation of H. alvei H4 WT and 1luxI on stainless-steel surface was detected by SEM. Obvious reduction in the number of 1luxI cells on stainless-steel

FIGURE 4 | Biofilm formation by H. alvei strains. The image shows the biofilms formed on glass surface by WT and 1luxI with and without AHLs. The plot shows the extent of biofilm formed on the glass surface as quantitated by OD<sup>590</sup> nm. NC, negative control. Data are the means ± SEMs (n = 3). Different letters above the columns indicate differences at the P < 0.05 level.

surface was observed, but the number of 1luxI cells was restored when supplied with 150 µg/ml C4-HSL (**Figure 6**). In addition, WT cells appeared to aggregate and adhered to the surface, and produced evident EPS filaments. In contrast, 1luxI tended to adhere to the surface as individual cells. When 1luxI was exposed to 150 µg/ml C4-HSL, the cells also formed aggregates and adhered to the surface of stainless steel, these cells also produced a lot of EPS.

### Swimming Motility

As shown in **Figures 2B–D**, the concentrations of AHLs, which induce the relative expression of luxR gene to the maximum respectively, were chosen for swimming motility assay, and swimming motility was determined by the expansion of the bacterial colony from the point of application to a greater diameter on the surface of an agar plate. Compared with WT, 1luxI suffered significant (P < 0.05) loss in swimming motility in the absence of exogenous AHLs (**Figure 7**). However, upon addition of C4-HSL, the swimming motility of 1luxI increased and became significantly higher than the level of WT. Addition of C6-HSL also had some enhancing effect on the swimming motility of 1luxI, but it remained obviously reduced compared with WT. In contrast, the addition of 3-o-C8-HSL appeared to have no real effect on the swimming motility of 1luxI.

### Expression of flgA, motA, and cheA

To further study the molecular mechanism of quorum sensing on biofilm formation, three biofilm formation and swimming motility related genes, flgA, motA and cheA, were chosen for analysis using qRT-PCR. The mRNA levels of all three genes in 1luxI in the absence of exogenous AHLs were significantly (P < 0.05) reduced compared with those of the WT, with motA being the most severely affected gene (**Figure 8**). Addition of C4-HSL (150 µg/ml) to the 1luxI culture restored the mRNA levels of the three genes to the levels comparable with those of the WT strain, except for the cheA gene, which showed lesser, but still significant increase. Similarly, the addition of C6-HSL (200 µg/ml) had some enhancing effect on the mRNA levels of motA and cheA, whereas the addition of 3-o-C8-HSL (100 µg/ml)

appeared to increase the mRNA levels of all three genes, but the increases still fell short of those achieved by C4-HSL. The result suggested that AHL could stimulate the expression of the motA and cheA genes, and to a lesser extent, the expression of the flgA gene.

### DISCUSSION

Mutant strain of H. alvei H4 with defective luxI gene was successfully constructed in this study, and the effect of this mutation on the quorum-sensing system was characterized with respect to AHL production, biofilm formation, and swimming motility. The luxI gene was generally recognized as AHL synthase gene, which controls the synthesis of AHLs (Waters and Bassler, 2005; Morohoshi et al., 2007). Indeed, the luxI mutant produced no detectable AHLs as assayed with the biosensor strain CV026 (**Figure 1**), consistent with the results reported for Aeromonas hydrophila and Acinetobacter nosocomialis (Khajanchi et al., 2009; Oh and Choi, 2015). The complementing strain of luxI was also constructed, and no differences in AHLs production, biofilm formation and swimming motility were found between comp-1luxI and WT strain (**Supplementary Figure 1**).

It is known to us that when local concentration of AHLs is high enough, AHLs would diffuse back into the cell, and induce the expression of luxR gene (Waters and Bassler, 2005; Bai and Rai, 2011). Expression of the luxR gene in the AHL-deficient strain of H. alvei H4 was found to decrease significantly, suggesting that expression of the luxR gene might require induction by AHLs. This hypothesis was confirmed by significant increase in luxR expression in the mutant when it was cultured in the presence of exogenous AHL (C4-HSL, C6-HSL, or 3-o-C8-HSL). The stimulating effect on luxR gene increased with the increase of AHLs concentration, causing the increase in luxR gene expression consequently. Similar result has also been obtained for Aliivibrio fischeri, whereby expression of the luxR gene was observed in the presence of 3-o-C6-HSL, with higher concentrations of the compound leading to more expression of the gene (Ramalho et al., 2016). In addition, there is also a threshold concentration of AHLs, and QS regulon is depressed when the concentration of AHLs exceeds this threshold, as a result of causing the reduction of luxR gene expression (Swift et al., 2001). Similar changes of sinR gene expression affected by AHLs were also obtained by Gao et al. (2012) in Sinorhizobium meliloti. Besides, response of CviR to autoinducer were found depressed when exposed to high concentration AHLs (Swem et al., 2009) in C. violaceum, consistent to the results achieved in the present study (**Figures 2B–D**).

Biofilm formation by the luxI mutant was inhibited, and since the mutant was deficient in AHL, this indicated that biofilm formation might be modulated by AHLs, and is associated with AHL-mediated QS. As reported by Ammor et al. (2008), bacterial phenotype such as resistance to antimicrobial compounds, biofilm formation, bioluminescence; pigment production, virulence gene expression, swimming motility and production of degradative extracellular proteases are regulated by the QS system. Furthermore, the addition of C4- HSL to the 1luxI culture restored the ability of the mutant cells to form biofilm, while the addition of C6-HSL and 3-o-C8-HSL appeared to have no significant effect, suggesting that the C4- HSL might play an important role in the regulation of biofilm formation in H. alvei H4. Similar result has been reported for A. hydrophila, in which biofilm formation and protease activity, which are inhibited in an AHL-deficient strain, can be stimulated by the addition of C4-HSL to the culture (Khajanchi et al., 2009). Niu et al. (2008) investigated the regulatory role of C12- HSL in the biofilm formation of Acinetobacter baumannii M2 and showed that biofilm formation is obviously inhibited in an

Zhu et al. Quorum Sensing of Hafnia alvei

abaI mutant, but can be restored by the addition of C12-HSL to the culture. Similarly, in Acinetobacter sp. strain DR1, C12- HSL was also found to play the same role as in A. baumannii M2 (Kang and Park, 2010). In P. aeruginosa, the LasR protein (a LuxR homologous protein) is soluble and stable only when it is produced in the presence of its cognate AHL ligand 3 o-C12-HSL, as the protein is becoming less soluble if it is produced in the presence of a different AHL, such as 3-o-C8- HSL or 3-o-C6-HSL, while soluble form is not produced in the presence of C4-HSL (Bottomley et al., 2007). Hence, C4- HSL might be the cognate ligand of LuxR in H. alvei H4, and it may possibly bind to the LuxR protein and play a regulatory role. In P. syringae and S. meliloti, the production of extracellular polysaccharide has been shown to be regulated by QS (Marketon et al., 2003; Quiñones et al., 2005). Therefore, the reduction in the numbers 1luxI cells adhering to polystyrene, glass and stainless-steel surfaces that we observed (**Figures 3–5**) might be caused by the reduction in biofilm formation and extracellular polysaccharide secretion. According to the results achieved in the present research and the studies reported by Nasser et al. (1998) and Andersson et al. (2000), different AHLs in bacteria may regulate different phenotypes, for H. alvei H4, C4-HSL seems to have a more important effect on regulating biofilm formation and swimming motility than C6- HSL and 3-o-C8-HSL, however, in other phenotypes regulated by quorum sensing system, C6-HSL and 3-o-C8-HSL might act as regulators.

Another important factor in the biofilm formation of bacteria is the extent of swimming motility, which contributes to the early development of the biofilm architecture (microcolony formation) (Srey et al., 2013). Bacterial motility depends mainly on the movement of flagella and Brownian motion (Li et al., 2008). Furthermore, some genes such as the flgA, motA, and cheA genes have been shown to be involved in biofilm formation and swimming motility (Lee et al., 2015). The data on swimming motility assay and qRT-PCR analysis of the expression of these three motility-related and biofilm-related genes, flgA, motA, and cheA (**Figure 8**), appeared to suggest that swimming motility might be regulated by regulating the expression of these genes. Since 1luxI exhibited significantly less swimming motility and lower expression of the flgA, motA, and cheA genes than WT strain in the absence of exogenous AHLs, these two factors might also be the cause of poor biofilm formation displayed by the mutant. Similar results have also been reported by Gurich and González (2009), whereby deletion of the sinI gene in

### REFERENCES


S. meliloti can result in significant reduction in the expression of the motility genes flaA and flaB, but such reduction can be reversed by the presence of AHLs. In contrast to our data, Wang T. et al. (2016) demonstrated that in Acidovorax citrulli, biofilm formation, motility and adherence of bacterial cells to a solid surface can be significantly promoted when the accR/I gene is defective. Furthermore, in A. nosocomialis, both biofilm formation and motility were found to be modulated by the luxR homologous gene anoR (Oh and Choi, 2015), and not the anoI gene. Therefore, the regulation of QS in biofilm formation their swimming motility might be different for different bacteria.

### CONCLUSION

In H. alvei, AHL-mediated QS system plays a key part in swimming motility and biofilm formation on different solid surfaces. We have shown here that among the different AHLs tested, C4-HSL appeared to exert a more significant impact on the modulation of these properties of the cells. Further study focusing on the regulation mechanism of the luxR/I gene, and the screening of effective QS inhibitors for H. alvei, and the application of these inhibitors to food production may bring about economic benefits as well as preventing the spread of food-related incident of infection among the public.

### AUTHOR CONTRIBUTIONS

HMH and GZ designed this study. YZ conducted the experiments. YZ, HMH, and YW performed the data analyses. HSH, GZ, and YZ drafted and revised the manuscript. All authors read and approved the final version of this manuscript.

### FUNDING

This work was supported financially by "The National Natural Science Foundation of China (Grant No. 31871895)."

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2019.01330/full#supplementary-material



of exogenous AHLs on bacterial phenotype. J. Gen. Appl. Microbiol. 62, 60–67. doi: 10.2323/jgam.62.60


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

Copyright © 2019 Zhu, Hou, Zhang, Wang and Hao. 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.

# Can Biofilm Be Reversed Through Quorum Sensing in Pseudomonas aeruginosa?

### Shaomin Yan and Guang Wu\*

State Key Laboratory of Non-Food Biomass and Enzyme Technology, National Engineering Research Center for Non-Food Biorefinery, Guangxi Key Laboratory of Biorefinery, Guangxi Biomass Engineering Technology Research Center, Guangxi Academy of Sciences, Nanning, China

Pseudomonas aeruginosa is a Gram-negative bacterium causing diseases in plants,

#### Edited by:

Cristina García-Aljaro, University of Barcelona, Spain

#### Reviewed by:

Rodolfo García-Contreras, National Autonomous University of Mexico, Mexico Akanksha Singh, Central Institute of Medicinal and Aromatic Plants (CIMAP), India

\*Correspondence:

Guang Wu hongguanglishibahao@yahoo.com; hongguanglishibahao@gxas.cn

#### Specialty section:

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

Received: 02 April 2019 Accepted: 25 June 2019 Published: 23 July 2019

#### Citation:

Yan S and Wu G (2019) Can Biofilm Be Reversed Through Quorum Sensing in Pseudomonas aeruginosa? Front. Microbiol. 10:1582. doi: 10.3389/fmicb.2019.01582 animals, and humans, and its drug resistance is a major concern in medical care. Biofilms play an important role in P. aeruginosa drug resistance. Three factors are most important to induce biofilm: quorum sensing (QS), bis-(3<sup>0</sup> -50 )-cyclic diguanosine monophosphate (c-di-GMP), and small RNAs (sRNAs). P. aeruginosa has its own specific QS system (PQS) besides two common QS systems, LasI–LasR and RhlI–RhlR, in bacteria. PQS is interesting not only because there is a negative regulation from RhlR to pqsR but also because the null mutation in PQS leads to a reduced biofilm formation. Furthermore, P. aeruginosa dispersed cells have physiological features that are distinct between the planktonic cells and biofilm cells. In response to a low concentration of c-di-GMP, P. aeruginosa cells can disperse from the biofilms to become planktonic cells. These raise an interesting hypothesis of whether biofilm can be reversed through the QS mechanism in P. aeruginosa. Although a single factor is certainly not sufficient to prevent the biofilm formation, it necessarily explores such possibility. In this hypothesis, the literature is analyzed to determine the negative regulation pathways, and then the transcriptomic data are analyzed to determine whether this hypothesis is workable or not. Unexpectedly, the transcriptomic data reveal a negative regulation between lasI and psqR. Also, the individual cases from transcriptomic data demonstrate the negative regulations of PQS with laslI, laslR, rhlI, and rhlR under different experiments. Based on our analyses, possible strategies to reverse biofilm formation are proposed and their clinic implications are addressed.

Keywords: biofilm, P. aeruginosa, quorum sensing, transcriptome, positive feedback, negative feedback

## INTRODUCTION

Pseudomonas aeruginosa is a Gram-negative bacterium living in soil and water. Being an opportunistic pathogen, P. aeruginosa can cause the bacterial soft rot in plants (Rahme et al., 2000; Walker et al., 2004), and diseases in animals (Ferris et al., 2017; Vingopoulou et al., 2018) and humans, including eye (Willcox, 2007), burn wound (Church et al., 2006), acute and chronic pulmonary infections, where cystic fibrosis is associated with substantial morbidity and mortality (Elborn, 2016; Klockgether and Tümmler, 2017).

Therefore, P. aeruginosa is a major concern in medical care because of its drug resistance against the traditional antibiotic therapy (Buhl et al., 2015; Oliver et al., 2015), that is particularly problematic for immunocompromised patients and the elderly in nosocomial environments (Xia et al., 2016). P. aeruginosa brings about its drug resistance through hydrolyzation of antibiotics with carbapenemases or extended-spectrum β-lactamases or ApmR (Vatcheva-Dobrevska et al., 2013; Fisher and Mobashery, 2014; Hakemi Vala et al., 2014), the low permeability of outer membrane (Eren et al., 2013; Zgurskaya et al., 2015), the multidrug efflux (Poole, 2004; Aghazadeh et al., 2014), etc. Also, the biofilm is an important player in P. aeruginosa drug resistance (Mah et al., 2003) because the dense extracellular matrix of biofilms reduces the efficacy of detergents and antibiotics (Mah et al., 2003). Such resistance could be increased a thousand times in some cases (Stewart and Costerton, 2001).

The dispersal of cells from the biofilm colony is a crucial and unique stage for biofilms to spread and colonize new surfaces (Monroe, 2007) and for the transition of dispersed cells from the biofilm to the planktonic growth phase. Could it be possible to stop the biofilm from happening, or reserve the biofilm back to the planktonic phenotype, or eradicate the biofilm in bacteria?

Theoretically, this hypothesis could be possible for P. aeruginosa, because its dispersed cells have physiological features that are distinct between the planktonic and the biofilm cells (Chua et al., 2014, 2015). In response to a low concentration of c-di-GMP, P. aeruginosa cells can disperse from the biofilm to become the planktonic cells. The drug resistance is not stronger in the biofilm cells than in the stationary-phase planktonic cells, but is stronger than in the logarithmic-phase planktonic cells (Spoering and Lewis, 2001). Additionally, P. aeruginosa produces cis-2-decenoic acid, which is a fatty acid messenger and induces dispersion and inhibits the growth of biofilm colonies (Davies and Marques, 2009). Furthermore, nitric oxide triggers the dispersal of biofilms in P. aeruginosa (Barraud et al., 2006), leading to the treatment of chronic infections in cystic fibrosis (Howlin et al., 2017).

The formation of biofilm is induced and regulated by numerous genes and environmental factors (Fazli et al., 2014), of which three are most important. The first one is the quorum sensing (QS), because QS controls about 10% genes in P. aeruginosa (Wagner et al., 2003), including many genes that are actively involved in the biofilm development and dispersal, although they are unlikely to be involved in the attachment and the initial of biofilm growth (Davies et al., 1998). The second one is the bis-(3<sup>0</sup> -50 )-cyclic diguanosine monophosphate (c-di-GMP), because its signaling network is the most complex secondary signaling system in bacteria (Hengge, 2009) and has the responsibility to decide whether bacteria adopt either planktonic or biofilm phenotype (Jenal and Malone, 2006). The third one is the small RNAs (sRNAs) although their role in biofilm is yet to be clear (Wolska et al., 2016).

Indeed, QS has a close relationship with biofilm (Wolska et al., 2016). It controls the synthesis of rhamnolipids that maintain the channels (Stoodley et al., 1994) for distributing nutrient and oxygen and removing waste products in mushroom-shaped structures (Davey and O'Toole, 2000). The channels can help in the release of a large amount of eDNA due to the autolysis of subpopulation of bacteria (Allesen-Holm et al., 2006) at the late stage of biofilm development. Various components of the biofilm matrix, such as extracellular DNA (eDNA), exopolysaccharides (EPS) and glucan, are closely related to biofilm matrix dynamics and bacterial virulence (Rainey et al., 2019). Also, there are other virulence factors, which play an important role in the QS regulation and biofilm formation. For example, pyocyanin promotes eDNA release and facilitates the biofilm formation (Klare et al., 2016).

It is worth reviewing literature to explore whether the biofilm is theoretically reversible through QS in P. aeruginosa, not only because P. aeruginosa is a causal organism of important health ailments but also because P. aeruginosa is a commonly used biofilm model organism (Rasamiravaka et al., 2015). More importantly, the synthesis of rhamnolipid in P. aeruginosa occurs at its late-exponential and stationary phases (Guerra-Santos et al., 1986). Rhamnolipid helps bacteria to utilize long-chain fatty acids as sources of carbon (Ochsner et al., 1994a) so it plays an important role in the biofilm formation (Stoodley et al., 1994; Davey and O'Toole, 2000; Allesen-Holm et al., 2006).

Reversing of biofilms could be plausible because QS is a target in many different circumstances such as attenuate virulence (Chan et al., 2015), bacterial metabolism (Goo et al., 2015), bacterial response to antibiotics (Rasamiravaka and El Jaziri, 2016), and therapy (LaSarre and Federle, 2013). Besides, the mechanism to form biofilms in P. aeruginosa is definitely different from other bacteria such as P. putida, P. fluorescens, Staphylococcus aureus, and Vibrio cholera (Wolska et al., 2016).

Needless to say, the reversing of biofilms is related to multiple factors, so a single factor such as QS could have very limited effects. However, we should theoretically explore those possibilities one by one at initial stage in view of the importance of biofilms in clinical meanings.

### POSITIVE AND NEGATIVE REGULATIONS IN QS

If we wish to reverse the biofilm through the QS, we need to find out whether the QS is reversible or not. So far overwhelmed evidence suggests that the QS is a positive feedback system, which implies that it is impossible to stop the QS once the QS is initiated. However, we have yet to know whether the ending point of QS is the biofilm formation? If this is the case, the stop of QS will either reverse the biofilm or stop the biofilm formation. To answer this issue, it is necessary to find out the negative regulation (feedback) in QS.

The QS is a cell-to-cell communication by means of production, detection, and response of chemical compounds, autoinducers, and thus the QS changes an individual or a population behavior upon the concentration of autoinducers, which are subject to the cell density (Fuqua et al., 1994).

Pseudomonas aeruginosa has three QS systems. (i) LasI– LasR that is related to the synthesis and the use of N-(3-oxododecanoyl)-L-homoserine lactone (3OC12-HL) (Passador et al., 1993; Pearson et al., 1994), whose concentration is ranged from

1 to 5 µM (Pearson et al., 1994, 1995) (brown color items in **Figure 1**). (ii) RhlI-RhlR that is related to the synthesis and the use of N-(butyryl)-L-homoserine lactone (BHL) (Pearson et al., 1995), whose concentration is about 10 µM (Pearson et al., 1995) (yellow color items in **Figure 1**). (iii) Pseudomonas quinolone signal (PQS)-based QS, PqsABCDH-PqsR that is related to the synthesis and the use of 2-heptyl-3-hydroxy-4-quinolone (HHQ) (Mashburn-Warren et al., 2008; Kulkarni and Jagannadham, 2014), whose concentration is about 6 µM (Pesci et al., 1999) (green color items in **Figure 1**). The first two QS systems essentially are N-acylated homoserine lactone (AHL)-based QS systems (Pesci et al., 1997) and exist in many bacteria.

The sophisticated QS systems in P. aeruginosa are described as follows. (i) LasI produces 3OC12-HL, which acts on LasR (Gambello and Iglewski, 1991; Pearson et al., 1994) (the upward brown arrow from lasR to LasR on the left side of **Figure 1**). (ii) LasR acts on aprA (Gambello et al., 1993), lasA (Toder et al., 1991) and toxA (Gambello and Iglewski, 1991; Gambello et al., 1993; Passador et al., 1993) (the downward brown arrow on the far left side in **Figure 1**). (iii) Both LasI and LasR act on lasB (Pearson et al., 1994, 1995) through 3OC12-HL, whose half-maximal expression needs 1.0 nM (Seed et al., 1995) (brown symbols on the left side of **Figure 1**). (iv) RhlI produces BHL, which acts on RhlR (Pearson et al., 1995, 1997) (the bright yellow arrow on the right side of **Figure 1**). (v) RhlR acts on pyocyanin synthesis (Meighen, 1991; Ochsner et al., 1994b; Brint and Ohman, 1995) (the long yellow arrow on the middle of **Figure 1**), lasA (Brint and Ohman, 1995) (the yellow arrow on the middle left of **Figure 1**), and rpoS (Latifi et al., 1996) (the yellow arrow on the upper right corner of **Figure 1**). (vi) Both RhlI and RhlR act on lasB through BHL (Brint and Ohman, 1995) (the yellow arrow on the upper right part of **Figure 1**), and rhlABR (Ochsner and Reiser, 1995) (the small yellow arrow on the middle of **Figure 1**), where rhlAB encodes rhamnosyltransferase (Ochsner et al., 1994a) (two yellow arrows on the upper middle part of **Figure 1**) together with rhlR positively regulate rhamnolipid synthesis (Ochsner et al., 1994b) (the yellow arrow on the middle upper part of **Figure 1**). (vii) LasR and RhlR positively regulate the synthesis of hydrogen cyanide (Pessi and Haas, 2000) (the downward yellow arrow on the lower middle part of **Figure 1**).

Still, **Figure 1** displays the effects of PQS-based QS on their targets. (i) PqsABCDH produces HHQ requiring phnA and phnB through anthranilate (Gallagher et al., 2002) (the green curly lines on the middle right part of **Figure 1**), then HHQ acts on PqsR (Cao et al., 2001), regulating the production of elastase, PA-IL lectin, pyocyanin and rhamnolipid (Pesci et al., 1999; McKnight et al., 2000; Gallagher et al., 2002; Lee and Zhang, 2015) (the green lines from the lower right corner in **Figure 1**). (ii) PqsE positively acts on biosynthesis of various virulent factors, which is independent of HHQ or any compounds produced related to the function of pqsABCDE operon although the expression of pqsE and PqsE are controlled by HHQ and PqsR (Farrow, et al., 2008) (dashed green line on the lower right part of **Figure 1**). (iii) PqsR– HHQ is involved in iron homeostasis (Bredenbruch et al., 2006; Oglesby et al., 2008) (the lowest green line in **Figure 1**).

A positive feedback can be found in each of three QS systems. (i) The first positive feedback goes from LasR–3OC12-HL to LasI through lasI, whose half-maximal expression needs 0.1 nM 3OC12-HL (Seed et al., 1995) (light gray ellipse on the left part of **Figure 1**). (ii) The second positive feedback goes from RhlR–HL to RhlI through rhlI (Ochsner and Reiser, 1995) (light gray ellipse on the upper right part of **Figure 1**). (iii) The third positive feedback goes from PqsR–HHQ to pqsABCDE and phnAB operons (Cao et al., 2001; Gallagher et al., 2002; Wade et al., 2005) (light gray ellipse on the middle right part of **Figure 1**).

The relationship among three QS systems in P. aeruginosa is positive in the following regulations. (i) LasR positively regulates HHQ through the complex LasR–3OC12-HL on pqsH (Pesci et al., 1999; Schertzer et al., 2009) (brown arrow on the right middle part of **Figure 1**). (ii) RhlR positively regulates HHQ through PqsE (Pesci et al., 1999) (two arrow-blue lines on the middle right part of **Figure 1**). (iii) LasR positively regulates rhlR through the complex LasR–3OC12-HL (Latifi et al., 1996; Pesci et al., 1997) and rhlI (Latifi et al., 1996) (the brown horizontal line with two arrows in **Figure 1**). (iv) HHQ strongly acts on rhlI with BHL (two arrow-blue lines in middle right part of **Figure 1**) but weakly acts on lasR and rhlR (McKnight et al., 2000). (v) RhlR positively regulates PqsE, whose overexpression leads to a high rhamnolipid production (Farrow, et al., 2008) (the yellow arrow on the right middle part of **Figure 1**). (vi) PqsE changes the function of RhlR rather than that of BHL (Farrow, et al., 2008) (two arrow-blue lines on the middle right part of **Figure 1**). (vii) LasR/3OC12-HL controls pqsR (Camilli and Bassler, 2006) (the end arrow of brown horizontal line in **Figure 1**).

In fact, there is a negative regulation among QS systems, namely, RhlR negatively regulates pqsR in P. aeruginosa (Pesci et al., 1997; Wade et al., 2005), or RhlR and BHL together negatively affect the production of HHQ and other quinolones through pqsR to pqsABCDE operon transcription (McGrath et al., 2004; Jensen et al., 2006; Xiao et al., 2006b) (the yellow arrow from RhlR–BHL to the yellow horizontal line to the right end with downward dash end in **Figure 1**). On the other hand, a negatively regulatory pathway is not so sure (the yellow end line highlighted with a red star on the middle of **Figure 1**).

### CAN THIS NEGATIVE REGULATION WORK?

As PQS-based QS is so particularly relevant to Pseudomonas, its significance should not be ignored. This is because the null mutation in PQS leads to a reduced biofilm formation and decreased the productions of pyocyanin, elastase, PA-IL lectin and rhamnolipids (Rahme et al., 1997, 2000; Cao et al., 2001; Diggle et al., 2003). Indeed, PQS directly or indirectly controls 92 or 143 genes as shown in two transcriptomic analyses (Deziel et al., 2005; Bredenbruch et al., 2006). By contrast, the other two QS systems together influence the expression in 200-plus genes (Whiteley et al., 1999).

For PQS, it does not reach its maximal production until the late stationary phase of growth (McKnight et al., 2000). This implies that HHQ is not involved in sensing the cell density, so the observation that the QS response is not reversed for

small decreases in population density in P. aeruginosa (Williams and Camara, 2009) is not the failure of PQS. An important time interval appears between QS systems, i.e., BHL is produced during the log phase of growth but HHQ is produced during late time in the stationary phase of growth (McKnight et al., 2000), so the positive regulation of HHQ on rhlI is more likely to be related to the second round of RhlI cycle. If HHQ would not function at this time interval, perhaps the QS would stop.

Another promising point is that the phenazine production requires HHQ in P. aeruginosa (McKnight et al., 2000; Mavrodi et al., 2001). In fact, phenazines may have a significant ecological impact on the biofilm formation in P. aeruginosa as well as other bacteria persisting in biofilms mixed with P. aeruginosa. Through affecting H2O<sup>2</sup> generation, phenazines bring about the lysis of competing bacterial cells in mixed biofilms and the subsequent eDNA release (Das and Manefield, 2013).

Perhaps, one of the best ways to explore the possibility of whether the QS is reversible through PQS in P. aeruginosa is to analyze the transcriptomic data in order to find some common patterns. Accordingly, we analyzed the transcriptomic data on Affymetrix P. aeruginosa array with 5549 P. aeruginosa genes, platform GPL84, from Gene Expression Omnibus (GEO) (Edgar et al., 2002; Barrett et al., 2013), including all the data in 104 publications (**Supplementary Information**) with 274 datasets. Each dataset represents the response to a specifically experimental condition. With these all available transcriptomic data, we wish to determine if PQS could be depressed under different experimental conditions.

**Table 1** shows correlation coefficients between any two genes of three QS systems. The rationale is that there are up-regulations and down-regulations in transcriptomic data. The correlation between two genes, which are both up-regulated or both downregulated, would suggest a positive regulation with a positive correlation coefficient. By contrast, the correlation between two genes, which are regulated oppositely, would suggest a negative regulation with a negative correlation coefficient.

Based upon the correlations within a single QS system in **Table 1**, the correlations between lasI and lasR, and between rhlI and rhlR confirm their auto-induction relationships (Gambello and Iglewski, 1991; Pearson et al., 1994) within each QS system. No negative correlation is found between the QS genes in the same QS system. As pqsR is named mvfR in gene bank, the auto-induction relationship with the rest of PQS genes are not very evident as the paired correlations between pdsA, pqsB, pqsC, pqsD, pqsE, but all paired correlations suggest a positive regulation within PQS system (Gallagher et al., 2002). Based upon the correlations between two QS systems in **Table 1**, the results conform there is a positive regulation between lasI–lasR and rhlI–rhlR, and between rhlI-rhlR and pqsABCDE. However, an undocumented negative regulation is revealed between lasI


and pqsR/mvfR using these transcriptomic data. Could it be a potential pathway to reverse the biofilm formation?

Furthermore, the responses of QS systems are analyzed under different transcriptomic experiments, and classified as down-regulation, down-regulation/no response, no response, no response/up-regulation, up-regulation and mixed responses. **Figure 2** shows such analysis according to 94 transcriptomic experiments. No response on three QS systems was found in 30 transcriptomic experiments (the intersection of three circles in **Figure 2**). Both LasI-LasR and RhlI-RhlR have the same response in 29 transcriptomic experiments (the intersection of two upper circles in **Figure 2**), suggesting a good cooperation between them. By the contrast, only five and six transcriptomic experiments show the same response for PQS with LasI-LasR and RhlI-RhlR (the intersection of two upper and lower circles in **Figure 2**), respectively. The same response in both LasI-LasR and PQS systems includes no response in GSE24784, GSE26142, GSE35248, and GSE39044, and no response/up-regulation in GSE22684, indicating few positive impact of LasI-LasR on PQS. The same response in both RhlI-RhlR and PQS systems includes: down-regulation in GSE9255; down-regulation/no response in GSE5887; no response in GSE17179 and GSE61925; and no response/up-regulation in GSE65882 and GSE7402. Thus, the results from Venn diagram indicate that RhlI-RhlR has weak impacts on PQS. **Figure 2** demonstrates the responses of 30, 29, and 53 transcriptomic experiments solely in LasI-LasR, RhlI-RhlR, and PQS, respectively, of which their response ranges from down-regulation to mixed response.

Finally, the negative regulation between different QS systems is found in four transcriptomic experiments (GSE4152, GSE8408, GSE6122, and GSE17296). In the study on Australian clonal strain (AES-1) in patients with cystic fibrosis in GSE6122 (Manos et al., 2009), lasI, rhlI, and rhlR were down-regulated while pqsA, pqsB, pqsC, pqsD, and pqsE were up-regulated. This highlights the PQS remarkable effect on the biofilm formation and enhanced infectivity. Another three

transcriptomic experiments show that pqsA, pqsB, pqsC, pqsD, and pqsE were down-regulated whereas rhlI and rhlR were upregulated (Teitzel et al., 2006; Tralau et al., 2007; Kawakami et al., 2010), of which lasI was up-regulated in copper-stressed (Teitzel et al., 2006), and both lasI and lasR were up-regulated in sulfate limitation (Tralau et al., 2007). Therefore, RhlI-RhlR does have a negative regulation on PQS (McGrath et al., 2004; Wade et al., 2005; Xiao et al., 2006a).

### CONCLUSION

In this hypothesis, we apply the transcriptomic data to verify the hypothesis of whether the biofilm can be reversed in P. aeruginosa through QS because there are negative regulations between PQS and RhlI-RhlR. Interestingly, the transcriptomic data from 104 publications reveal a negative regulation between lasI and psqR, rendering a support to the hypothesis. Individual cases from transcriptomic data under different experiments demonstrate the negative regulations of PQS with laslI, laslR, rhlI, and rhlR.

In general, the relationships among different QS systems reveal positive regulations, which act together to promote the biofilm formation. However, the present analyses from literature and transcriptomic data provide the evidence that both LaslI-LaslR and RhlI-RhlR systems have negatively regulatory effects on PQS system. This is very important because these negative regulations lay the foundation for the biofilm reversion through QS. Although the exact pathways are still not fully discovered, the N-acylated homoserine lactone (AHL)-based QS systems can influence PQS-based QS system by inhibiting the expression of pqsABCDE operon and pqsR, resulting in the reduction of HHQ and PqsR synthesis. Consequently, the low concentration of PQS related products cannot maintain the biofilm, leading to its reversion. On the other hand, the down-regulated PQSbased QS system cannot perform well their function of positive regulations on LaslI-LaslR and RhlI-RhlR systems, which will further affect the biofilm formation, especially in the second round of RhlI cycle. Surely, there are other factors that play roles in the formation of drug-resistant multicellular biofilms, such as c-di-GMP. As mentioned in section "Introduction," this signal can govern bacterial cells to adopt either planktonic phenotype or biofilm formation (Hengge, 2009). Recent study demonstrates that high levels of cAMP lead to the decrease of c-di-GMP

### REFERENCES


content, which inhibits the biofilm formation in P. aeruginosa (Almblad et al., 2019).

In clinic therapeutics for infectious diseases, antibiotic resistance has been spreading widely and rapidly, which becomes a major challenge for modern medicine. The strategy of interfering the biofilm formation is effective through bacterial cell-to-cell communication, especially with QS system (Soheili et al., 2019). More recently, QS inhibitors are drawing great attention in blocking the pathogenicity from P. aeruginosa (Calvert et al., 2018; Schütz and Empting, 2018). The transcriptional regulator PqsR becomes an attractive object and is considered to be one of the most appropriate targets. Currently, QS regulation mechanism in P. aeruginosa is mainly related to positive and negative regulation between QS systems. Clearly, exploration of regulation beyond QS should get attention in future.

### DATA AVAILABILITY

All datasets generated or analyzed for this study are included in the manuscript and the **Supplementary Files**.

### AUTHOR CONTRIBUTIONS

GW designed the work. Both authors prepared and approved the manuscript.

### FUNDING

This work was partly supported by the National Natural Science Foundation of China (Nos. 31460296 and 31560315) and the Key Project of Guangxi Scientific Research and Technology Development Plan (AB17190534).

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2019.01582/full#supplementary-material





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

Copyright © 2019 Yan and Wu. 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.

# Bacterial–Fungal Interactions in the Kelp Endomicrobiota Drive Autoinducer-2 Quorum Sensing

Anne Tourneroche<sup>1</sup> , Raphaël Lami<sup>2</sup> \*, Cédric Hubas<sup>3</sup> , Elodie Blanchet<sup>2</sup> , Marine Vallet<sup>1</sup> , Karine Escoubeyrou<sup>4</sup> , Alain Paris<sup>1</sup> and Soizic Prado<sup>1</sup> \*

<sup>1</sup> Unité Molécules de Communication et Adaptation des Microorganismes (MCAM), Muséum National d'Histoire Naturelle (MNHN), Centre National de la Recherche Scientifique (CNRS), CP 54, Paris, France, <sup>2</sup> CNRS, Laboratoire de Biodiversité et Biotechnologies Microbiennes (LBBM), USR3579, Observatoire Océanologique de Banyuls, Sorbonne Université, Banyuls-sur-Mer, France, <sup>3</sup> Muséum National d'Histoire Naturelle, UMR BOREA 7208 MNHN-Sorbonne Université-CNRS-UCN-UA-IRD, Station Marine de Concarneau, Paris, France, <sup>4</sup> CNRS, Observatoire Océanologique de Banyuls, Sorbonne Université, FR3724, Banyuls-sur-Mer, France

#### Edited by:

Tom Defoirdt, Ghent University, Belgium

### Reviewed by:

Qian Yang, Yellow Sea Fisheries Research Institute (CAFS), China Natrah Fatin Mohd Ikhsan, Putra Malaysia University, Malaysia

#### \*Correspondence:

Raphaël Lami lami@obs-banyuls.fr Soizic Prado soizic.prado@mnhn.fr

#### Specialty section:

This article was submitted to Microbial Physiology and Metabolism, a section of the journal Frontiers in Microbiology

> Received: 16 April 2019 Accepted: 09 July 2019 Published: 31 July 2019

#### Citation:

Tourneroche A, Lami R, Hubas C, Blanchet E, Vallet M, Escoubeyrou K, Paris A and Prado S (2019) Bacterial–Fungal Interactions in the Kelp Endomicrobiota Drive Autoinducer-2 Quorum Sensing. Front. Microbiol. 10:1693. doi: 10.3389/fmicb.2019.01693 Brown macroalgae are an essential component of temperate coastal ecosystems and a growing economic sector. They harbor diverse microbial communities that regulate algal development and health. This algal holobiont is dynamic and achieves equilibrium via a complex network of microbial and host interactions. We now report that bacterial and fungal endophytes associated with four brown algae (Ascophyllum nodosum, Pelvetia canaliculata, Laminaria digitata, and Saccharina latissima) produce metabolites that interfere with bacterial autoinducer-2 quorum sensing, a signaling system implicated in virulence and host colonization. Additionally, we performed co-culture experiments combined to a metabolomic approach and demonstrated that microbial interactions influence production of metabolites, including metabolites involved in quorum sensing. Collectively, the data highlight autoinducer-2 quorum sensing as a key metabolite in the complex network of interactions within the algal holobiont.

Keywords: quorum sensing (QS), AI-2, bacterial–fungal interaction, kelp microbiota, algal holobiont

### INTRODUCTION

Large marine brown algae, i.e., kelp, are an essential component of temperate coastal ecosystems. Indeed, these organisms are important primary producers that generate a specific habitat and thereby shape coastal marine life (Egan et al., 2013). In addition, kelp farming has been a growing economic sector over the last decades (FAO, 2018).

Like most eukaryotes, macroalgae are colonized by various microorganisms (the microbiota) that interact with them throughout the life cycle, and that modify their physiology (Wahl et al., 2012; Egan et al., 2013; Singh and Reddy, 2014). For example, commensal bacteria have profound effects on seaweed development, nutrition, and defense (Wahl et al., 2012; Singh and Reddy, 2014; Tapia et al., 2016). Algal tissues are also asymptomatically colonized by filamentous fungi (Debbab et al., 2012), although these fungi and their role are yet to be fully characterized (Fries, 1979; Zuccaro et al., 2003, 2008; Loque et al., 2010; Jones et al., 2012). Previously, we isolated and characterized the molecular diversity of cultivable fungi in different parts of the brown algae Ascophyllum nodosum, Pelvetia canaliculata, Laminaria digitata, and Saccharina latissima. We also found that metabolites produced by endophytic fungi are key mediators of interactions among macroalgae, their fungal microbiota, and protistan pathogens (Vallet et al., 2018).

Endophytic bacteria are likely to interact with fungi in the algal host to maintain the host-microbiota equilibrium and thus contribute to host health (Deveau et al., 2018; Hassani et al., 2018). Indeed, structural changes in the microbiota, i.e., dysbiosis, have been linked to disease in marine organisms and seaweeds (Fernandes et al., 2012; Zozaya-Valdes et al., 2015; Egan and Gardiner, 2016). However, the mechanisms underlying bacterial– fungal homeostasis remain unclear, although they appear crucial to macroalgal physiology not only in nature but also in farms, highlighting their importance in light of intensifying algal culture (Gachon et al., 2010).

Bacterial–fungal interactions consist of multiple and concomitant mechanisms ranging from nutrient competition to antibiosis, many of which depend on chemical signaling. Quorum sensing, which allows bacteria to coordinate gene expression based on the density of specific signaling molecules, is of particular interest, since it is essential for virulence, colonization, biofilm formation, and toxin production (Atkinson and Williams, 2009). Indeed, various types of quorum signals, also known as auto-inducers (AI), are already known. Some have been found in only one genus whereas others, such as type 1 (AI-1) or type 2 (AI-2), are present in various bacterial genera. Indeed, AI-2 molecules appear widespread among prokaryotes, and is produced by over 50% of sequenced bacterial species, including both Gram-positive and Gram-negative species (Hammer and Bassler, 2003; Federle, 2009). AI-2 molecules are derived from a common precursor, (S)-4,5 dihydroxypentane-2,3-dione (DPD), which is synthesized by the enzyme LuxS. By spontaneous cyclization, this precursor is transformed to 4-hydroxy-5-methyl-3(2H)furanone, (2R,4S)- 2-methyl-2,3,3,4-tetrahydroxytetrahydrofuran, and, especially in marine environments, to the furanosyl borate diester (Hardie and Heurlier, 2008).

Various marine bacteria were found to produce AI-2 (Bodor et al., 2008; Doberva et al., 2015; Pérez-Rodríguez et al., 2015), although it is considered in some species to be a metabolic by product and not a signaling molecule (Rezzonico and Duffy, 2008). Nevertheless, many studies showed that AI-2 regulates niche-specific behaviors in commensal and pathogenic bacteria, including biofilm formation or dispersion, cell division, virulence, bioluminescence, and motility (Hammer and Bassler, 2003; Hardie and Heurlier, 2008; Federle, 2009). Accordingly, secreted AI-2 is now recognized as a key signaling molecule that affects bacterial behavior at species and community level (Whiteley et al., 2017).

Strikingly, quorum sensing molecules are not exclusively produced by bacteria. Indeed, these compounds were also shown to regulate fungal morphogenesis, germination, apoptosis, biofilm development, or pathogenicity (Wongsuk et al., 2016). Importantly, some recent studies showed that both bacterial and fungal quorum sensing compounds mediate cross-kingdom signaling. For instance, eukaryotes may interfere with bacterial quorum sensing (Martín-Rodríguez et al., 2014; Ismail et al., 2016), while bacteria may react to fungal quorum signals (Cugini et al., 2007; Fourie et al., 2016). Cross-kingdom signaling is also modulated by quorum sensing inhibitors, i.e., quorum quenchers (Grandclément et al., 2016; Rolland et al., 2016). To date, halogenated furanones synthesized by the marine red algae Delisea pulchra are the best-studied naturally occurring quorum sensing inhibitors in eukaryotes. These compounds regulate bacterial colonization of algal surfaces by interfering with AI-1 and AI-2 (Defoirdt et al., 2007; Harder et al., 2012).

Despite these breakthroughs, the role of AI-2 quorum sensing in marine environments in general, and in holobionts in particular, remains poorly characterized (Doberva et al., 2015; Hmelo, 2017). One important shortcoming is the lack of quantitative measurements, since AI-2 itself is difficult to quantify (Wang et al., 2018). In this study, we hypothesized that AI-2 quorum sensing is involved in interspecies chemical signaling among endophytic fungi and bacteria in seaweeds. Accordingly, we first isolated and molecularly characterized the cultivable bacteria associated with the brown algae A. nodosum, P. canaliculata, L. digitata, and S. latissima. These bacteria were then investigated for their ability to produce or inhibit production of AI-2 along with fungi previously isolated from the same samples (Vallet et al., 2018). Co-cultures experiments, between several fungal and bacterial endophytic strains isolated from S. latissima microbiota combined with metabolomics approach pointed out that inter-species interactions involve metabolites production that modulates AI-2 production. Altogether these results suggest that dynamic interactions driven by microbial metabolites may occur within the microbiota and impact AI-2 QS signaling.

### MATERIALS AND METHODS

### Sampling and Endophyte Isolation

Fungi and bacteria were previously isolated from the brown algae L. digitata, S. latissima, A. nodosum, and P. canaliculata. L. digitata, and A. nodosum were collected in triplicate in Roscoff, France, in January 2013. Samples of all four species were also collected in triplicate in Oban, Scotland, in July 2013. Algae were surface-sterilized with 70% ethanol and 0.1% sodium hypochlorite, and cut into small pieces. Around 4,600 of these pieces were then aseptically transferred to different solid media, using at least 10 replicates from each algal part on each type of medium (**Supplementary Table S1**). Resulting cultures were then grown and preserved following previously described protocols (Vallet et al., 2018).

### Taxonomic Identification of Endophytic Bacteria

Genomic DNA was extracted with Wizard <sup>R</sup> Genomic DNA Purification Kit (Promega, Charbonnières-les-Bains, France) from single colonies of 209 bacterial isolates that were grown in marine broth. 16S rRNA genes were then amplified using 2× KAPA2G Ready Mix (Clinisciences, Nanterre, France), 1 µL bacterial DNA, and the universal primers 27F mod (50 -AGRGTTTGATCMTGGCTCAG-3<sup>0</sup> ) and 1492R mod (5<sup>0</sup> - TACGGYTACCTTGTTAYGACTT-3<sup>0</sup> ). Targets were amplified over one cycle of denaturation at 94◦C for 5 min, 35 cycles at 94◦C for 15 s, 50◦C for 15 s, and 72◦C for 20 s, and final extension at 72◦C for 10 min. PCR products were sequenced

by Sanger sequencing on the Bio2Mar platform (Observatoire Océanologique, Banyuls-sur-Mer, France), using primer 907R (Eurofins MWG Operon, Ebersberg, Germany).

The quality of each sequence was checked manually and the closest match in NCBI databases was determined by BLAST (Altschul et al., 1990). Further, sequences were aligned in Muscle, as implemented in MEGA 7.0 (Edgar, 2004; Kumar et al., 2016). Alignments were reviewed manually to verify mismatches, and a phylogenetic tree was constructed by maximum likelihood using the K2, G+I model. The reliability of each node in the tree was assessed by bootstrapping over 500 replicates.

### Screening for the Production of Quorum Sensing Mediators

The QS bioluminescent reporter strain Vibrio campbellii MM32 (**luxN**::Cm, **luxS**::Tn5Kan) was used to detect AI-2 in bacterial and fungal extracts, as previously described (Miller et al., 2004). The receptor luxN is mutated in this strain to abolish sensing of acyl homoserine lactones, while the synthase gene luxS is mutated to abolish AI-2 production but not sensing. It was previously constructed by introducing luxS::Tn5Kan onto the chromosome of strain JAF305 (luxN::Cm) (Bassler et al., 1993; Freeman and Bassler, 1999).

To obtain bacterial supernatants, 1 mL was collected from each of 209 bacterial cultures grown for 24 h at 22◦C in marine broth. Samples were then centrifuged at 17,000 × g for 10 min, and resulting supernatants were filtered at 0.22 µm. To obtain fungal extracts, 43 fungal isolates were grown for 3 weeks at 19◦C in MEA/ASW medium, and extracted three times with ethyl acetate. A detailed recipe of the medium is provided in **Supplementary Table S1**. Extracts were tested at a final concentration of 250 µg/mL, with final concentration of DMSO 2.5%.

Bacterial supernatants and fungal extracts were tested for the production of molecules interfering with AI-2 quorum sensing. Briefly, 20 µL of test samples and corresponding controls were mixed with 180 µL of V. campbellii MM32 diluted 1:5,000 and incubated at 30◦C and 100 rpm. Luminescence and cell density (OD620) were measured after 24 h. Data were collected in triplicate, and luminescence change was calculated as (lumiSN/<sup>E</sup> – lumiControl)/lumiControl, where lumiSN/<sup>E</sup> is bioluminescence (normalized to cell density) from the reporter strain in the presence of supernatant or extract, and lumiControl is bioluminescence (normalized to cell density) in the presence of either marine broth (when supernatants are tested) or DMSO (when extracts are tested).

### Quantification of AI-2 Precursor by LC–MS/MS

Due to low ionization potential and instability, AI-2 and DPD are not directly detectable by mass spectrometry (MS). However, quinoxaline derivatives of DPD, obtained by reaction with 4,5-dimethyl-1,2-phenylenediamine, are detectable by LC– MS/MS. DPD was thus quantified in bacterial supernatants after performing a derivatization reaction as described (Xu et al., 2017). Briefly, triplicate DPD standard solutions with concentration 2.6 nM–26 µM were obtained by diluting a stock solution of DPD (16.64 nM) in marine broth. To obtain quinoxaline derivatives, 250 µL of standard solution or supernatant was reacted with 250 µL of 0.1 mg/mL 4,5-dimethyl-1,2-phenylenediamine (Sigma, St. Louis, MO, United States) in 0.1M HCl. Samples were thoroughly mixed for 1 min, and incubated for 5 h at 25◦C with agitation. Samples were then desalted with two volumes of water using Sep-Pak C18 SPE cartridges (Waters, Beverly, MA, United States), and eluted with two volumes of acetonitrile. Subsequently, samples were analyzed by LC–MS/MS on a Dionex Ultimate 3000 HPLC system coupled to a Q ExactiveTM Focus mass spectrometer (Thermo Fisher Scientific, Waltham, MA, United States) and fitted with an electrospray ionization source and a Hypersil GOLD C18 column (2.1 mm × 150 mm, 1.9 µm particle size; Thermo Scientific, Waltham, MA, United States) operating at 20◦C. In this system, eluates are introduced directly into the mass spectrometer. LC– MS parameters are detailed in **Supplementary Table S1**. Data were collected using Xcalibur, in parallel reaction monitoring mode targeting the precursor ion at m/z 233.1285. The product ion at m/z 186.1140 was used for quantification.

### AI-2 Antagonist Activity in Fungal Extracts

To test AI-2 antagonist activity in fungal extracts, 20 µL samples were reacted as described with 180 µL of V. campbellii MM32 diluted 1:5,000 in marine broth and supplemented with 2 µM DPD (purchased from Rita Ventura's research group at ITQB, Oeiras, Portugal). To confirm that loss of luminescence, if any, was not due to cytotoxicity, 100 µL of culture was reacted with 30 µL of 0.01% resazurin (Graça et al., 2013) immediately after measurement of luminescence, and fluorescence (λex: 530 nm, λem: 590 nm) was measured after incubating for 4 h at 30◦C with agitation. As control, a 96-well plate containing 20 µL of fungal extract in marine broth was assayed in the same manner to assess background luminescence, fluorescence, and absorbance.

### Co-culture Experiment: Culture Conditions and Impact on Quorum Sensing

Fungal and bacterial isolates were co-cultured in triplicate in marine broth supplemented with 10% malt extract, 4% glucose, and 1.5% agar, adjusted to pH 7, and plated. Plates were inoculated with 2 mL of a mixture of 2 × 10<sup>4</sup> fungal spores and bacteria diluted to OD 0.1. Corresponding monocultures were prepared in triplicate in the same manner. Cultures were then incubated for 21 days at 19◦C and on a 12-h light/dark cycle. Petri dishes containing only culture medium (n = 3) were used as blank. Cultures and corresponding controls were extracted with ethyl acetate for 30 min, in a sonicator at room temperature. Samples were then filtered through a filter paper, and dried under vacuum using a centrifugal evaporator. Extracts were tested for their impact on quorum sensing at a final concentration of 250 µg/mL, with final concentration of DMSO 2.5%. Briefly, 20 µL of extract were mixed with 180 µL of V. campbellii MM32 diluted 1:5,000 and incubated at 30◦C and

100 rpm. Luminescence and cell density (OD620) were measured after 24 h. Data were collected in triplicate, and luminescence change was calculated as (lumi<sup>E</sup> – lumiControl)/lumiControl, where lumi<sup>E</sup> is bioluminescence (normalized to cell density) from the reporter strain in the presence of extract, and lumiControl is bioluminescence (normalized to cell density) in the presence of DMSO. To assess the viability of the biosensor, 100 µL of culture was reacted with 30 µL of 0.01% resazurin (Graça et al., 2013) immediately after measurement of luminescence, and fluorescence (λex: 530 nm, λem: 590 nm) was measured after incubating for 4 h at 30◦C with agitation.

### Co-culture Experiment: LC–MS-Based Metabolomic Analysis

Dried extracts were solubilized in methanol at 0.5 mg/mL, and analyzed by HPLC–MS in one batch and in a random sequence. Samples were loaded onto a Dionex Ultimate 3000 HPLC system fitted with a C18 AcclaimTM RSLC PolarAdvantage II column (2.1 mm × 100 mm, 2.2 µm pore size; Thermo Scientific, Waltham, MA, United States) operating at 40◦C, and coupled to a Maxis IITM QTOF mass spectrometer (Bruker, Bellerica, MA, United States) with an electrospray ionization source. Data were acquired with Data Analysis software. LC–MS parameters are listed in **Supplementary Table S1**. Raw LC–MS data were calibrated and converted to netCDF format using Data Analysis software (Bruker), and processed using the R package XCMS (Smith et al., 2006). Based on analytical conditions and raw data characteristics, final peak picking parameters were method = 'centWave,' ppm = 10, and peak width = c(5,20), while final grouping parameters were bw = 5, mzwid = 0.015, and retention time correction method = 'obiwarp.' Other parameters were set to default values. To limit noise from compounds already present in culture media, the dataset was filtered with an in-house script to retain only those features with intensity in at least one sample more than fivefold its average intensity in blank samples.

### Statistical Procedures

All analyses and graphs were performed using the R statistical framework (R Core Team, 2019). Van der Waerden tests followed by a post hoc test using the Fischer's Least Significant Difference (LSD) criterion was performed to test the contrasting effect of mono and co-culture on QS activity. Multivariate analyses were done using the R library mixOmics (Rohart et al., 2017).

### RESULTS

### Diversity of Cultivable Endophytic Bacteria From Brown Algae

A total of 209 bacterial isolates was obtained, and classified based according to 16S rRNA genes into 4 phyla, 12 orders, 19 families, 27 genera, and 88 taxonomically unique units (**Figures 1**, **2** and **Supplementary Table S2**). The most abundant phyla in L. digitata and S. latissima were Firmicutes (comprising 47 and 35% of all isolates, respectively) and Proteobacteria (39 and 53%). In P. canaliculata, Proteobacteria, and Actinobacteria were

predominant (44 and 38%), whereas Firmicutes accounted for 77% of bacteria isolated from A. nodosum. Gammaproteobacteria was between 50 and 100% of all Proteobacteria depending on algal species, although Alphaproteobacteria was occasionally present. On the other hand, Bacillus and Pseudoalteromonas were the most abundant genera in S. latissima (33 and 30%) and L. digitata (47 and 22%). Isolates from P. canaliculata were mostly Pseudoalteromonas, Rhodococcus, and Bacillus (19, 16, and 13%), whereas Bacillus was dominant in A. nodosum (63%). Rhodococcus, Bacillus, Cobetia, and Pseudoalteromonas were isolated from all four algal species.

### Production of AI-2 Compounds by Bacterial Endophytes

Most (86%) bacterial endophytes isolated from brown algae elicited an increase in luminescence from V. campbellii MM32, a quorum sensing reporter strain (**Table 1** and **Supplementary Data S1**). Strikingly, 10% of isolates boosted luminescence by over 50%, including a Kocuria isolate, six Bacillus isolates, and 13 Proteobacteria. Indeed, a Marinomonas isolate and five Cobetia isolates increased luminescence by at least 100%. In contrast, all but one Pseudoalteromonas isolate increased luminescence by less than 50%.

To confirm these results, DPD production by Marinomonas 424 and Cobetia 352b, two of the strongest inducers of luminescence, was quantified by LC–MS/MS. For comparison, DPD production was also quantified in Pseudoalteromonas 352a, which was co-isolated with Cobetia 352b and is a weak inducer of luminescence. Interestingly, culture supernatants from Marinomonas 424 and Cobetia 352b contained similar levels of DPD (692 and 585 nM, respectively), whereas a lower amount was found in the supernatant from Pseudoalteromonas 352a (66 nM). Thus, our results based on LC–MS/MS confirmed the quorum sensing patterns observed using the biosensor-based approach (**Figure 3**).

### Impact of Metabolites of Endophytic Fungi on Quorum Sensing

As presented in **Figure 4**, extracts from 13 fungi boosted the luminescence from the biosensor. Of these, eight (Verticillium biguttatum AN130T, Chaetomium globosum LD13H, Microsphaeropsis olivacea LD50H, Botryotinia fuckeliana LD535H, Leptosphaeria marina SL457T, and Diaporthe eres SL473T) increased the luminescence by over 50%. Conversely, extracts from 30 fungi strongly diminished the luminescence, as shown in **Figure 5**. For 14 of these extracts, the loss of luminescence was due to toxicity against the biosensor V. campbellii MM32. In contrast, the other 16 extracts evidently inhibited AI-2, blocking the effects of 2 µM DPD but with very limited impact on the biosensor V. campbellii MM32 metabolism (**Figure 5**).

### Impact of Bacterial–Fungal Interactions on Quorum Sensing

Metabolites produced in co-cultures of Cladosporium SL405T with Pseudoalteromonas 352a or Cobetia 352b reduced luminescence from V. campbellii MM32 by 44 and 20%, respectively. In contrast, metabolites from monocultures of Cladosporium SL405T decreased luminescence only slightly (12%), whereas metabolites from monocultures of either bacterium elicited a stronger decrease in luminescence (49%) (**Supplementary Data S3**). These results highlight the contrasting effects of co-cultures and monocultures on quorum sensing supported by a non-parametric test for independent



Numbers are bacterial isolates per genus.

samples (Van Der Waerden test), followed by a post hoc test using the Fischer's Least Significant Difference (LSD) criterion (**Figure 6**). Furthermore, the partial least squares discriminant analysis of samples covering the different co-culture conditions characterized through 4,221 metabolomic features collected by LC–MS showed that each monoculture or co-culture is characterized by a specific set of features corresponding to a unique set of metabolites (**Supplementary Data S2**). Sparse partial least squares discriminant analysis also identified four latent variables of 180 features (720 features in total) that discriminate a culture from all others (**Supplementary Data S2**). In a second round of partial least squares discriminant analysis based only on these 720 features (**Figure 7**), the co-cultures were differentiated in the third dimension both from each other and from every monoculture. Analysis of variance of scores get on the 3rd dimension confirmed that selected features significantly separate co-cultures from each other and from monocultures (F2,<sup>12</sup> = 167.2, p < 1.7 × 10−<sup>9</sup> ). This result implies that bacterial–fungal interactions impact very significantly metabolite production in specific ways.

Features that distinguish a culture and that are altered more than 10-fold over other cultures are listed in **Supplementary Table S3**. When compared to the corresponding monocultures, 28 and 5 of such features were identified in co-cultures of Cladosporium SL405T with Cobetia 352b and Pseudoalteromonas 352a, respectively. Conversely, 120 and 87 of such features were identified in Cobetia 352b and Pseudoalteromonas 352a monocultures when compared to the corresponding co-cultures with Cladosporium SL405T. Moreover, 93 features diminished by at least a factor of 10 in cocultures when compared to Cladosporium SL405T monocultures. Unfortunately, top-ranked metabolites were not identified by annotation against ISDB (Allard et al., 2016), GNPS, and MassBank. Similarly, identification against SIRIUS 4.0. (Böcker and Rasche, 2008) and Pubchem was inconclusive.

### Search of a Multivariate Link Between Luminescence Measurements and MS Metabolomic Variables

A link between the global response given by the luminescence representing an integrated measurement of the QS and the metabolome in the different mono or co-culture conditions was obtained thanks a PLS-based regression (**Figure 8**).

Complementary filters such as (i) VIP above 1.20 with a VIP standard error coming from repeated cross-validation calculus lower than the VIP value for the variable, and (ii) absolute value of the correlation between primary metabolomic variables and the predicted luminescence response above 0.75, were used to select from the initial set of 4221 variables a subset of 521 variables. The comprehensive heatmap obtained (**Figure 9**) showcased nine and one variables displaying a significantly higher mean

value for co-culture between Cla and Co and Cla and Ps, respectively (**Figure 10**). All these 10 significant variables are supposed to be induced in these co-culture conditions when compared to mono-culture conditions as revealed by the multiple comparison of means based on the Student-Newman-Keuls test. Indeed, these variables are of prime importance candidates that would be putatively involved in some metabolic pathways explaining the QS event.

### DISCUSSION

The data highlight the large diversity of cultivable bacterial endophytes associated with healthy L. digitata, S. latissima, A. nodosum, and P. canaliculata (**Figures 1**, **2**). The four phyla (Proteobacteria, Firmicutes, Actinobacteria, and Bacteroidetes) and 27 genera that were identified are consistent with previous surveys of cultivable bacteria associated with macroalgae (Wiese et al., 2009; Hollants et al., 2013; KleinJan et al., 2017). Notably, the cultivable fraction of bacterial communities appears to vary depending on algal species, sampling site, and algal tissue. Nevertheless, bacteria classified as Cobetia and Pseudoalteromonas were isolated from every algal species, tissue, and sampling site. Moreover, these genera are the most frequently isolated from S. latissima, apart from Bacillus. Of note, Vibrio and Flavobacterium were not isolated from our samples, even though these are frequently isolated from brown algae (Hollants et al., 2013; Albakosh et al., 2016).

Among the isolated bacterial endophytes, 86% were found to produce AI-2, which triggers quorum sensing in the reporter strain V. campbellii MM32 (**Table 1**). Cobetia, Marinomonas, and Erwinia were the strongest inducers (>100% induction) of quorum sensing, whereas 93 and 73% of isolated Bacillus and Pseudoalteromonas induced quorum sensing only weakly (<50% induction). These results were confirmed by quantifying DPD, an AI-2 precursor, in bacterial supernatants using tandem mass spectrometry. As shown in **Figure 3**, Cobetia 352b and Marinomonas 424 produced around 700 nM of DPD, suggesting that these strains engage neighboring bacteria with AI-2 receptors. Pseudoalteromonas 352a also produced DPD but to a lesser extent (66 nM). Interestingly, fungal endophytes isolated along with these bacteria positively or negatively modulated AI-2 quorum sensing (**Figures 4**, **5**). These results provide more evidence that AI-2 quorum sensing is involved in interkingdom signaling.

While inhibitory activity against quorum sensing was previously detected in marine fungi (Martín-Rodríguez et al., 2014), this is the first demonstration, to the best of our

FIGURE 7 | Visualization of samples using the first four latent variables from partial least squares discriminant analysis of 720 selected features. Ps, Pseudoalteromonas monocultures (violet); Co, Cobetia monocultures (green); Cla, Cladosporium monocultures (blue); Cla-Ps, Cladosporium-Pseudoalteromonas co-cultures (gray); Cla-Co, Cladosporium-Cobetia co-cultures (orange). Ellipses represent 95% confidence intervals.

knowledge, that fungal metabolites may also enhance quorum sensing. However, such result is not surprising, as this effect was previously observed in metabolites from some other types of eukaryotes such as Chlamydomonas reinhardtii and Chlorella spp. microalgae (Rolland et al., 2016). Collectively, these findings suggest a key role for AI-2 signaling among endophytes of brown algae. On the other hand, we note that 14 fungal endophytes are strongly antimicrobial against the Vibrio biosensor (**Figure 5**),

and thus may be similarly active against V. harveyi, a prominent pathogen in aquaculture (Zhang and Li, 2016).

As the fungus Cladosporium SL405T and the bacteria Pseudoalteromonas 352a and Cobetia 352b were isolated from the same holobiont (S. latissima), they were characterized in monoculture and in co-culture, with a view to assess the impact of fungal–bacterial interactions on metabolite production and quorum sensing. The data indicate that these isolates produce quorum sensing or quorum sensing-modulating compounds. Of note, these genera are frequently isolated from macroalgae (Hollants et al., 2013; Hulikere et al., 2016; Li et al., 2017). Pseudoalteromonas is of particular interest, as it was shown to produce antimicrobials or bioactive molecules against algal spores, invertebrate larvae, fungi, and other bacteria. Such molecules may help the host against surface colonization by these organisms (Holmström et al., 2002; Richards et al., 2017). Also, Pseudoalteromonas was implicated in Hole-Rotten disease (Wang et al., 2007).

As shown in **Figures 6**, **7**, co-cultures of Cladosporium SL405T with two different bacteria produce different metabolites. These metabolites are also different from those produced by corresponding monocultures. For example, 28 and 5 metabolic features were at least 10-fold more abundant in co-cultures of Cladosporium SL405T with Cobetia 352b and Pseudoalteromonas 352a than in corresponding monocultures. This result implies that microbial interactions induce production of specific metabolites (**Supplementary Table S3**). Conversely, 120 and 87 features were at least 10-fold more abundant in Cobetia 352b monocultures and Pseudoalteromonas 352a monocultures than in co-cultures, suggesting either that production of these metabolites is inhibited by microbial interaction, or that these metabolites are degraded in co-cultures. Similarly, 105 and 108 features (of which 93 are common) were at least 10-fold more abundant in Cladosporium SL405T monocultures than in cocultures with Cobetia 352b and Pseudoalteromonas 352a.

Taken together, these results demonstrate that metabolomes in co-cultures fundamentally differ from metabolomes in monocultures due to microbial interactions. Unfortunately, chemical characterization of culture-specific metabolites was not possible since none matched known natural products. Identification of the source of metabolites in co-cultures also remains a major challenge, since the structure of such metabolites and other related biochemical information would be required (Xu et al., 2018). Nevertheless, we found that different cultures have variable impact on quorum sensing (**Figure 6**), such that metabolites obtained from co-cultures of the same fungus with two different bacteria clearly display contrasting effects on quorum sensing (**Figure 6**). For instance, metabolites from

a co-culture of Cladosporium SL405T and Pseudoalteromonas 352a diminish luminescence from the biosensor by 40%, while metabolites from a co-culture of Cladosporium SL405T and Cobetia 352b led to a loss of only 20%. The link between the impact on quorum sensing and the metabolites present in different culture conditions was strengthened by the PLSregression based analysis highlighting 10 variables, highly correlated with the predicted luminescence response (absolute value of the correlation above 0.75), and specially induced in the co-culture conditions when compared to mono-culture conditions (**Figure 10**).

Collectively, our data provide the first evidence of quorum sensing and quorum quenching in bacterial and fungal endophytes of brown algae. These results highlight the importance of chemical communication among microbial components of a holobiont. Indeed, our laboratory model clearly demonstrates the impact of interspecies interactions on the production of metabolites, including metabolites involved in quorum quenching or in antagonizing other microorganisms. Our model also demonstrates the various phenotypes that may be observed in a given fungal or bacterial strain depending on environmental conditions. Hence, these results provide a glimpse of the complexity of molecular dialogs in the holobiont, and how this may impact host fitness.

Accordingly, the data also highlight the need to fully understand the functional role of all microbial members in the seaweed holobiont and their impact on algal fitness either in nature or in farms. Indeed, quorum sensing is already known to significantly affect the expression of virulence genes in aquaculture pathogens (Zhao et al., 2015). Quorum sensing was also demonstrated to control microbial colonization in the red algae Delisea pulchra, notably by release of halogenated furanone, which inhibits pathogenic epiphytic bacteria such as Nautella sp. (Harder et al., 2012). Hence, quorum sensing represents a very promising target in future studies of approaches to limit pathogenic effects in algae.

### DATA AVAILABILITY

fmicb-10-01693 July 30, 2019 Time: 15:54 # 12

The datasets generated for this study can be found in GenBank, SUB4670131.

### AUTHOR CONTRIBUTIONS

AT carried out the experiments. AT, AP, RL, CH, KE, and SP analyzed the data. AT, RL, and SP wrote the manuscript. All authors conceived this study.

### FUNDING

This work was supported by ATM "Microorganisms" grant from the Natural History Museum of Paris (SP), EMBRC project

### REFERENCES


No. 262280 and PEPS ExoMod CNRS grant (SP), EC2CO Roseocom CNRS grant (RL), and the Convergence Sorbonne Université program (RL and SP). This work was funded through the IDEX SUPER (AT).

### ACKNOWLEDGMENTS

We acknowledge the Scottish Association for Marine Science (SAMS) and the Biological Station of Roscoff for hosting the field missions. We would like to greatly thank Philippe Potin (Marine Biology Station of Roscoff, France) for taxonomic identification of the brown macroalgae species during the field mission. We are grateful to the Bio2Mar platform (http://bio2mar.obs-banyuls.fr) for providing technical help and support.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2019.01693/full#supplementary-material

sensing-regulated gene expression in Vibrio Harveyi by decreasing the DNA-binding activity of the transcriptional regulator protein LuxR. Environ. Microbiol. 9, 2486–2495. doi: 10.1111/j.1462-2920.2007. 01367.x



signal in epsilon proteobacteria. ISME J. 9, 1222–1234. doi: 10.1038/ismej. 2014.214



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

Copyright © 2019 Tourneroche, Lami, Hubas, Blanchet, Vallet, Escoubeyrou, Paris and Prado. 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.

# *In silico* Analysis Reveals Distribution of Quorum Sensing Genes and Consistent Presence of LuxR Solos in the *Pandoraea* Species

*Kah-Ooi Chua1 , Wah-Seng See-Too1 , Robson Ee1 , Yan-Lue Lim1 , Wai-Fong Yin1 and Kok-Gan Chan1,2 \**

*1 Division of Genetics and Molecular Biology, Institute of Biological Sciences, Faculty of Science, University of Malaya, Kuala Lumpur, Malaysia, 2 International Genome Centre, Jiangsu University, Zhenjiang, China*

#### *Edited by:*

*Ana Maria Otero, University of Santiago de Compostela, Spain*

#### *Reviewed by:*

*Navneet Rai, University of California, Davis, United States Ralf Heermann, Johannes Gutenberg University Mainz, Germany*

> *\*Correspondence: Kok-Gan Chan kokgan@um.edu.my*

#### *Specialty section:*

*This article was submitted to Infectious Diseases, a section of the journal Frontiers in Microbiology*

*Received: 26 February 2019 Accepted: 16 July 2019 Published: 06 August 2019*

#### *Citation:*

*Chua K-O, See-Too W-S, Ee R, Lim Y-L, Yin W-F and Chan K-G (2019) In silico Analysis Reveals Distribution of Quorum Sensing Genes and Consistent Presence of LuxR Solos in the Pandoraea Species. Front. Microbiol. 10:1758. doi: 10.3389/fmicb.2019.01758*

The most common quorum sensing (QS) system in Gram-negative bacteria consists of signaling molecules called *N-*acyl-homoserine lactones (AHLs), which are synthesized by an enzyme AHL synthase (LuxI) and detected by a transcriptional regulator (LuxR) that are usually located in close proximity. However, many recent studies have also evidenced the presence of LuxR solos that are LuxR-related proteins in Proteobacteria that are devoid of a cognate LuxI AHL synthase. *Pandoraea* species are opportunistic pathogens frequently isolated from sputum specimens of cystic fibrosis (CF) patients. We have previously shown that *P. pnomenusa* strains possess QS activity. In this study, we examined the presence of QS activity in all type strains of *Pandoraea* species and acquired their complete genome sequences for holistic bioinformatics analyses of QS-related genes. Only four out of nine type strains (*P. pnomenusa*, *P. sputorum*, *P. oxalativorans,* and *P. vervacti*) showed QS activity, and C8-HSL was the only AHL detected. A total of 10 canonical *luxI*s with adjacent *luxR*s were predicted by bioinformatics from the complete genomes of aforementioned species and publicly available *Pandoraea* genomes. No orphan *luxI* was identified in any of the genomes. However, genes for two LuxR solos (LuxR2 and LuxR3 solos) were identified in all *Pandoraea* genomes (except two draft genomes with one LuxR solo gene), and *P. thiooxydans* was the only species that harbored no QS-related activity and genes. Except the canonical LuxR genes, LuxIs and LuxR solos of *Pandoraea* species were distantly related to the other well-characterized QS genes based on phylogenetic clustering. LuxR2 and LuxR3 solos might represent two novel evolutionary branches of LuxR system as they were found exclusively only in the genus. As a few *luxR* solos were located in close proximity with prophage sequence regions in the genomes, we thus postulated that these *luxR* solos could be transmitted into genus *Pandoraea* by transduction process mediated by bacteriophage. The bioinformatics approach developed in this study forms the basis for further characterization of closely related species. Overall, our findings improve the current understanding of QS in *Pandoraea* species, which is a potential pharmacological target in battling *Pandoraea* infections in CF patients.

Keywords: cystic fibrosis, type strains, single molecule real-time sequencing, quorum sensing, LuxR solos

## INTRODUCTION

Cystic fibrosis (CF) results from mutations in the cystic fibrosis transmembrane conductance regulator (CFTR) gene that functions in modulating chloride ion transport across epithelial cells (Trapnell et al., 1991; Pier et al., 1996). As a consequence to this gene abnormality, majority of CF patients suffer from secretion of thick and viscous mucus in their respiratory tracts. These copious respiratory secretions become the breeding ground for microorganisms, which lead to both chronic and transient pulmonary infections, inflammation, obstruction of airways, and ultimately life-threatening pulmonary dysfunction (Gibson et al., 2003). *Staphylococcus aureus*, *Pseudomonas aeruginosa*, *Burkholderia cepacia,* and a spectrum of other Gram-negative bacteria are frequently associated with bacterial lung infections in CF patients (Gibson et al., 2003; LiPuma, 2010). However, recent reports revealed unprecedented infections by a number of bacteria and *Pandoraea* species are among the novel bacteria associated with pulmonary infections in CF patients (Atkinson et al., 2006; Davies and Rubin, 2007).

The genus *Pandoraea* was proposed to accommodate a group of isolates cultured from sputum specimens of CF patients that were initially misidentified as *B. cepacia* and genus *Ralstonia*. In the process of taxonomical characterization, some members of genus *Burkholderia* are reclassified into *Pandoraea* based on genotypic characteristics as well (Coenye et al., 2000, 2001). Members of genus *Pandoraea* are commonly recovered from sputum specimens of patients with cystic fibrosis, but some species were isolated from various environmental sources too. These bacteria have been considered as emerging multi-drug resistant pathogens in the context of cystic fibrosis (Davies and Rubin, 2007), but our understanding about the epidemics of *Pandoraea* species remains scarce.

Bacterial cells are able to interact with one another *via* production and release of diffusible signaling molecules into their living environment. Detection of such molecules enables bacteria to coordinate gene expression in response to both high and low cell population densities. The process is termed as quorum sensing (QS) or bacterial cell-to-cell communication (Williams, 2007). The canonical LuxI/R QS system is one of the most studied QS systems in bacteria. In this system, the responsible signaling molecules are *N*-acyl homoserine lactones (AHLs) that are produced by an AHL synthase, LuxI activates a cognate transcriptional regulator, LuxR if the concentration of AHLs achieves a threshold. Upon activation, LuxR binds to the promoters or regulators of targeted genes in response to the cell density and causes coordinated gene expression in the bacterial population (Fuqua and Greenberg, 2002; Williams, 2007). *Via* this system, bacteria regulate a variety of activities including biofilm formation, production of extracellular enzymes, regulation of virulence genes, and so on.

With the advancement in DNA sequencing technologies, novel subgroups of *luxI* and *luxR* homologs have been identified in numerous bacterial species. While it led to reports that most typical *luxI*/*R* QS systems have both genes involved located almost adjacent to each other, additional *luxR* homologs that do not pair with a cognate *luxI* are frequently found. These unpaired *luxR* homologs that are termed as *luxR* solos possess modular homologies to the canonical LuxR with an AHL-binding domain at their N-terminus and a DNA-binding helix-turn-helix (HTH) domain at the C-terminus (Subramoni and Venturi, 2009). Bioinformatics prediction of QS genes in proteobacterial genomes had revealed the presence of numerous additional orphan *luxR* homologs with no *luxI* homologs in close proximity (Fuqua, 2006). In addition to their widespread distributions in proteobacteria, some of these LuxR solos are phylogenetically related and several surveys provided evidence on clustering of LuxR solos into different functionally relevant groups (Brameyer et al., 2014; Gan et al., 2015; Subramoni et al., 2015). It is believed that the presence of additional LuxR solos increases the range of gene regulatory activities by responding to self-produced AHLs or eavesdropping on exogenous AHLs and even other signaling molecules produced by other species (Hudaiberdiev et al., 2015; Subramoni et al., 2015). Interestingly, some LuxR solos harbored by non-QS bacteria are responding to non-AHL signaling molecules such as OryR of *Xanthomonas oryzae* pv. oryzae and XccR of *Xanthomonas campestris* pv. campestris that are capable of interacting with plant signaling molecules and play essential roles in their pathogenicity (Zhang et al., 2007; Ferluga and Venturi, 2009).

Members of genus *Pandoraea* have been reported with QS activity and are able to communicate *via* the production of AHLs (Ee et al., 2014). In this study, we investigated (1) if AHL-mediated QS is a common activity employed by all type strains of *Pandoraea* species and (2) the distribution of QS genes in *Pandoraea* genus. Our work was initiated by obtaining type strains of all species of the genus from culture collection to characterize their QS activity before we sequenced their complete genomes to provide molecular data on distributions and phylogenetic relationships of QS genes in the study species. A systematic bioinformatics prediction workflow was developed for the identification of LuxI and LuxR of *Pandoraea* species. Our findings indicated that AHL synthases of genus *Pandoraea* represent a novel evolutionary branch of QS system. We also identified the presence of two conserved LuxR solos in most members of *Pandoraea* genus, which prompted us to further discuss the acquisition mechanisms and possible roles of these LuxR solos in this study.

### MATERIALS AND METHODS

### Bacterial Strains and Culture Conditions

All type strains of species in *Pandoraea* genus were acquired from Leibniz Institute-Deutsche Sammlung von Mikroorganismen und Zellkulturen (DSMZ) culture collection. All bacteria strains were maintained in media and condition as listed in **Supplementary Table S1**.

### Detection of Quorum Sensing Activity Using CVO26 Bioassay

Preliminary detection of QS activity was conducted by CVO26 bioassay in which the bacterial samples were streaked perpendicularly to CVO26 biosensor on Luria-Bertani (LB) agar and incubated in 28°C for 24 h. The CVO26 biosensor is useful for the detection of short chain AHLs in the range of C4-HSL to C8-HSL (McClean et al., 1997). Positive result was observed with purple pigmentation of viocalein forming on the CVO26 biosensor. Positive and negative controls were set up with *E. carotovora* GS101 and *E. carotovora* PNP22, respectively.

### Extraction of Acyl-Homoserine Lactone Signaling Molecules

All the strains were cultured in LB broth buffered with 50 mM 3-(N-morpholino) propanesulfonic acid (MOPS; pH 5.5) in their respective optimum culturing temperature for 24 h with 220 rpm agitation prior to AHL extraction. Spent culture supernatants were mixed thoroughly with an equal volume of 0.1% v/v glacial acetic acid-acidified ethyl acetate solvent until biphasic layers were formed, and the upper immiscible solvent layer was transferred out. Similar extraction was performed twice, and the organic solvent containing AHL extract was desiccated completely for mass spectrometry analysis.

### Multiple Reaction Monitoring Mass Spectrometry Analysis

Desiccated AHL extracts were suspended with acetonitrile solvent prior to sample loading into an Agilent 1,290 Infinity LC system (Agilent Technologies Inc., Santa Clara, CA, USA). The liquid chromatography (LC) system was comprised of an Agilent ZORBAX Rapid Resolution High Definition SB-C18 Threaded Column (2.1 mm × 50 mm, 1.8 μm particle size) operated at 500 μl/min flow rate, 37°C with solvent A (0.1% formic acid buffered water) and solvent B (0.1% formic acid buffered acetonitrile) as the mobile phases. Three-step elution was performed with 7 min of linear gradient profile of 20–70% of solvent B, followed by 5 min of isocratic profile of 80% of solvent B, and 3 min of gradient profile of 80–20% of solvent B.

The LC-separated compounds were detected by electrospray ionization trap mass spectrometry (ESI-MS) using Agilent 6,490 Triple Quadrupole LC/MS system under positive-ion mode. The electrospray used nitrogen as nebulizing gas (pressure set to 20 p.s.i) and drying gas (flow set to 11 ml/h). The desolvation temperature was 200°C, and probe capillary voltage was set at 3 kV. AHL profiles were characterized using multiple reaction monitoring (MRM; Gould et al., 2006) by comparison of retention times and *m/z* transitions with those of the synthetic AHLs. A total of 10 synthetic AHLs varying in substitution oxo-group at C3 position (e.g., C6-HSL and 3-oxo-C6-HSL) and carbon length (ranging from C4-HSL to C12-HSL) were loaded in the MS analysis for reference as listed in **Supplementary Table S2**. The ions monitored in Q1 include the AHL precursor ion [M+H]+ , whereas both the lactone moiety at *m*/*z* 102 and the acyl moiety [M+H−101]+ were monitored in Q3. Blanks (acetonitrile) were analyzed as control (data not shown). Data analysis was performed using Agilent MassHunter software.

### Genome Sequencing

Genomic DNA (gDNA) was extracted using MasterPure DNA Purification Kit (EpiCenter, CA, USA) according to the manufacturer's instruction, and the quality of gDNA was assessed using gel electrophoresis, NanoDrop 2000 UV-Vis spectrophotometer (Thermo Scientific, MA, USA) and Qubit 2.0 fluorometer (Life Technologies, MA, USA), respectively. Whole genome sequencing was performed using PacBio (Pacific Biosciences, CA, USA) Single-Molecule Real Time (SMRT) sequencing technology.

### Genome Assembly, Circularization, and Annotation

Raw data generated were assembled using hierarchical genome assembly process (HGAP) assembler. Circularity of genomes was assessed using Contiguity (Sullivan et al., 2015; Lim et al., 2015a), and the precise location of the overlapping region was determined using Gepard (Krumsiek et al., 2007; Lim et al., 2015b) prior to genome circularization using Minimus2 pipeline in the AMOS software package to generate the blunt-ended circular genomes with complete closure. The presence of plasmid was distinguished from the chromosomal genome, and functional annotation was performed using NCBI Prokaryotic Genome Annotation Pipeline (PGAP), Rapid Prokaryotic Genome Annotation (Prokka; Seemann, 2014), Rapid Annotation Search Tool (RAST; Aziz et al., 2008), KEGG database (Kanehisa et al., 2016), and IMG ER (Markowitz et al., 2009).

### Systematic Bioinformatics Prediction of LuxI and LuxR

A systematic bioinformatics prediction and identification of LuxI and LuxR was employed in this study as presented in **Figure 1**. In short, translated proteomes were local blast against LuxI and LuxR databases downloaded from Uniprot and NCBI non-redundant protein database. Conserved domains of the putative LuxI (cl7182, N-acyltransferase superfamily) and LuxR (cd06170, C-terminal DNA-binding domain of LuxR-like proteins; PF03472, autoinducer-binding domain) were identified using conserved domain database (CDD) search tool in NCBI. All short-listed candidates were then scanned for signature protein family (Pfam) domain present in all LuxI (PF00765, autoinducer synthase), and LuxR (PF03472, autoinducer-binding domain). Lastly, a comprehensive InterProScan was conducted to provide high confidence authenticity of the identified LuxI and LuxR in which all LuxIs must contain three signature LuxI domains, IPR016181 (acyl-CoA N-acyltransferase), IPR001690 (autoinducer synthase), and IPR018311 (autoinducer synthesis, conserved site), while LuxR must contain four signature LuxR domains, IPR005143 (transcription factor LuxR-like, autoinducer-binding domain), IPR011991 (winged helix-turn-helix DNA-binding domain), IPR016032 (signal transduction response regulator, C-terminal effector), and IPR000792 (transcription regulator LuxR, C-terminal).

Multiple alignment of LuxI was performed with numbering relative to curated LuxI sequence of *Vibrio fischeri* (Entry: P12747) retrieved from Swiss-Prot database to determine the 10 conserved residues (R25, F29, W35, E44, D46, D49, R70, F84, E101, and E104) present in all LuxI (Fuqua and Greenberg, 2002). Multiple alignment of LuxR was performed with numbering relative to curated TraR sequence of *Agrobacterium tumefaciens* (Entry: P33905) retrieved from Swiss-Prot database to determine

the nine signature conserved residues (W57, Y61, D70, P71 W85, and G113, which are key amino acids in autoinducerbinding domain, and E178, L182, and G188, which are three key amino acids in DNA-binding domain) found in all LuxR (Subramoni et al., 2015).

### Average Nucleotide Identity and Phylogenomics Analysis

Average nucleotide identity (ANI) analysis by Goris et al. (2007) was performed using ANI calculator1 . Genome alignments were performed with a minimum length of 700 bp and a minimum identity of 70%. Genome fragments options were set with a window size of 1,000 bp and a step size of 200 bp. Phylogenomics tree was constructed using neighbor joining (NJ) method in MEGA 6.06 in which pairwise distances were acquired from ANI analysis (Saitou and Nei, 1987; Tamura et al., 2013).

### Maximum Likelihood Phylogenetic and Pairwise Identity Matrix Analyses

Evolutionary analyses of QS genes were performed using MEGA 6.06 in which all *in silico* functionally validated amino acid sequences were aligned using MUSCLE algorithm (Edgar, 2004; Tamura et al., 2013). Phylogenetic analyses were performed using maximum likelihood (ML) method with Jones-Taylor-Thorton (JTT) model and 1,000 replications of bootstrap analysis (Jones et al., 1992). Initial tree was constructed automatically using NJ and BioNJ algorithms with nearest-neighbor-interchange (NNI) method. All gaps and missing data were not included

1

in the analyses. Pairwise identity matrix analyses were performed using Sequence Demarcation Tool (SDT) version 1.2 (Muhire et al., 2014). Sequences alignment was performed using MUSCLE algorithm, and data were presented in three color modes.

## RESULTS AND DISCUSSION

### Detection and Characterization of Quorum Sensing Activity in *Pandoraea* Genus

Following our previous discovery on QS activity in *P. pnomenusa* RB-38 and RB-44 (Han-Jen et al., 2013; Ee et al., 2014), we hypothesized that QS could be a common activity employed by all members in *Pandoraea* genus. To prove our hypothesis, all nine type strains of *Pandoraea* were first screened for their QS activity using CVO26 biosensor (McClean et al., 1997), and other possible AHLs were also extracted from spent culture prior to characterization of AHL signaling molecules using MRM MS analysis. Of nine strains, only four (*P. pnomenusa, P. sputorum, P. oxalativorans,* and *P. vervacti*) activated the CVO26 biosensor, and C8-HSL was the only AHL detected in MRM MS analysis (**Supplementary Figure S1**). Interestingly, not all clinically isolated strains were detected positive for AHL production as *P. apista* DSM 16535T and *P. pulmonicola* DSM 16583T have no QS activity detected.

### Complete Genome Sequencing of Nine Type Strains of *Pandoraea* Species

As QS is not a common activity employed by all *Pandoraea* species, we questioned (1) if other non-AHL-producing

http://enve-omics.ce.gatech.edu/ani/

*Pandoraea* species (*P. apista, P. pulmonicola, P. norimbergensis, P. faecigallinarum,* and *P. thiooxydans*) are actually possessing mutated *luxI* and/or *luxR*, incapable of producing or detecting AHL (Sandoz et al., 2007) and (2) if they harbor *luxR* solo in their genomes? To prove these hypotheses, we sequenced the complete genomes of all nine type strains of *Pandoraea* using SMRT sequencing technology to facilitate the identification of the QS genes in the genomes. HGAP assembler was employed to assemble all genomes to complete closure, and circularization was performed to provide a high confidence genomic size of each *Pandoraea* strains. Plasmids were distinguished from the chromosomal DNA, and designation code was provided for each strain (**Table 1**). Besides, all *Pandoraea* genomes available in GenBank were also retrieved for investigation (**Supplementary Table S3**).

Genome sizes of species in *Pandoraea* genus range between 4.5 and 6.2 Mb with genomes of *P. thiooxydans* DSM 25325 T and *P. norimbergensis* DSM 11628T representing the smallest and largest, respectively (**Table 1**). The G + C content of these genomes varies from 62.63 to 64.9%. The presence of plasmids was identified in four of nine type strains. Notably, *P. apista* DSM 16535T is the only one harboring plasmid of five clinically isolated strains. The other three type strains harboring plasmid are *P. faecigallinarum* DSM 23572T , *P. oxalativorans* DSM 23570T , and *P. vervacti* DSM 23571T that were isolated from oxalateenriched cultures from different environments (**Table 1**; Sahin et al., 2011).

ANI analysis was subsequently performed to investigate the genetic and evolutionary distances of all *Pandoraea* species (**Supplementary Table S4**). Generally, 95% of ANI value is the accepted cut-off threshold for species-species delineation (Richter and Rosselló-Móra, 2009). The *Pandoraea* genomes in this study formed several clusters on a phylogenomic tree constructed using neighbor-joining algorithm (**Figure 2**). In Cluster 1 that included several *P. pnomenusa* and *P. pulmonicola* DSM 16583T , we observed that the clinically isolated strains *P. pnomenusa* DSM 16536T were distinguished from the other *P. pnomenusa* that were obtained from the environments by slightly further distances.

By contrast, *P. sputorum* DSM 21091T was the only clinically isolated species that was clustered together with *P. oxalativorans* DSM 23570T , *P. faecigallinarum* DSM 23572T , and *P. vervacti* DSM 23571T that were isolated from various environments despite being obtained from different origins. The strains *P. norimbergensis* DSM 11628T and *P. thiooxydans* were found to be most distantly related to all other *Pandoraea* species (<85 and <79% ANI values, respectively; **Supplementary Table S4**) forming outgroups in the phylogenomic analysis of *Pandoraea* genus (**Figure 2**).

Results from the ANI analysis also suggested reclassification of several *Pandoraea* strains with uncertain taxonomic status. We deduce from the analysis that *Pandoraea* sp. E26 could be reclassified as *P. pnomenusa* E26 as it shares 99% ANI value with all *P. pnomenusa*. On the other hand, *P. pnomenusa* strains 6,399 and 7,641 that demonstrated ANI value <90% against all other *P. pnomenusa* suggested that reclassification might be necessary. This is supported by the observations that both the strains exhibited the highest ANI values with *P. apista* species (>88% ANI values; **Supplementary Table S4**) and were placed out of Cluster 1 (consisted of all other *P. pnomenusa* and *P. pulmonicola* DSM 16583T) in phylogenomic tree (**Figure 2**). Unfortunately, the taxonomy of *Pandoraea* strains SD6-2 and B-6 remained questionable as both of them were having low ANI values (<88 and <86%, respectively; **Supplementary Table S4**) with all other *Pandoraea* in this study.

### Identification of *luxI* and *luxR* Homologs in Genomes of Genus *Pandoraea*

To provide high confidence in authenticity of all LuxI and LuxR identified in this study, we created a stringent and effective systematic bioinformatics prediction of LuxI and LuxR as presented in **Figure 1**. A total of 10 *luxI*s were identified in genomes of *P. pnomenusa, P. sputorum, P. oxalativorans,* and *P. vervacti* (**Supplementary Table S5**). A typical authentic LuxI contains three signature InterPro domains (IPR016181, IPR001690, and IPR018311) and 10 signature conserved residues. Although all 10 LuxIs of *Pandoraea* species identified were found to contain only domains IPR016181 and IPR001690 (**Figure 3**), our previous gene cloning data of *PpnI* RB38 confirmed that LuxI of *Pandoraea* species could function properly despite the absence of domain IPR018311 (Lim et al., 2015a). Multiple alignment analysis of LuxI also revealed a consistent profile of signature conserved residues in all LuxIs of *Pandoraea* species, thus concordantly supported the evidence that these are authentic functional LuxI for the production of C8-HSL (**Supplementary Table S6**). No orphan *luxI* was identified in any of the *Pandoraea* genomes in this study.

Intriguingly, besides the expected canonical *luxR*, two additional *luxR* solos (named *luxR*2 solo and *luxR*3 solo) were identified in most *Pandoraea* genomes (**Supplementary Table S5**). *P. thiooxydans* DSM 25325T is the only exception and does not harbor any canonical *luxI*/*R1* and *luxR* solo in its genome (**Supplementary Table S5**). *P. apista* TF81 and *Pandoraea* sp. E26 were also found to harbor only *luxR2* solo. However, we hypothesized that *luxR*3 solo could be missing in the gap of their draft genomes. All canonical LuxR and LuxR solos identified in this study contained all the four signature InterPro domains (IPR005143, IPR011991, IPR016032, and IPR000792), and all nine signature conserved residues (six key amino acids in autoinducer-binding domain and three key amino acids in DNA-binding domain) found in typical LuxR (**Supplementary Table S7**; Subramoni et al., 2015).

To date, there have been reports on LuxR solos responding to non-AHL signals. For examples, the PluR of *Photorhabdus luminescens* senses α-pyrone (Brachmann et al., 2013), while PauR of *P. asymbiotica* detects dialkylresorcinols and cyclohexanediones (Brameyer et al., 2015) signaling molecules instead of AHLs. These non-AHL-binding LuxRs, however, harbor substitutions in the conserved amino acid motif of autoinducer-binding domain compared to that in AHL sensors (Brameyer and Heermann, 2015). The autoinducer-binding domain of all LuxR solos in *Pandoraea* species contained


TABLE 1| Designation code, sequencing information and general features of nine circularized genomes of *Pandoraea* type species.

FIGURE 2 | Phylogenomic analysis depicting the genetic and evolutionary distances of all *Pandoraea* species. In general, *Pandoraea* genus was separated into five distinct clusters: Cluster 1 (*P. pnomenusa*, *P. pulmonicola*), Cluster 2 (*P. sputorum*, *P. oxalativorans*, *P. faecigallinarum*, *and P. vervacti*), Cluster 3 (*P. apista*), Cluster 4 (*P. norimbergensis*), and Cluster 5 (*P. thiooxydans*). Bar, 0.2 substitutions per nucleotide position.

the six conserved amino acids (W57, Y61, D70, P71, W85, and G113) with respect to TraR (**Supplementary Table S7**) and thus reflected a conserved motif for AHL-binding LuxR proteins.

From our analysis, we noticed that majority of the annotation pipelines often annotated *luxR*2 and *luxR*3 solo genes as hypothetical proteins making it a challenge in their identification process. Hence, we employed an *in silico* systematic bioinformatics prediction of these genes to aid in future identification of *luxR*2 and *luxR*3 solos. For nomenclature purpose, the gene products of canonical *luxI*/*R* identified in *Pandoraea* genomes were given designation with the first alphabet of the genus followed by the first two alphabet of species, for instances, PpnI, LuxI of *P. pnomenusa,* and PspR, LuxR of *P. sputorum* (**Supplementary Table S5**). Additionally, to differentiate between canonical LuxR and LuxR solos, canonical LuxR were given designation as LuxR1 (e.g., PpnI/ R1 and PspI/R1), while LuxR solos were given designation as LuxR2 and LuxR3 solos. Gene designations, accession numbers, amino acid length, GC content, and genetic orientation of all canonical LuxI/R1 and LuxR solos of *Pandoraea* species are presented in **Supplementary Table S5**. No QS gene was found on plasmid.

### LuxI of *Pandoraea* Species Represents a Novel Evolutionary Branch of Quorum Sensing System

To determine the relatedness of LuxI in *Pandoraea* genus with the other well-characterized LuxI, phylogenetic and pairwise identity matric analyses were conducted. **Figure 4** shows the recent phylogenetic tree of LuxI from several groups that are most closely related to the LuxI in *Pandoraea* genus, together with pairwise identity matrix analysis. The analyses revealed that LuxI of *Pandoraea* species were highly conserved in *Pandoraea* genus forming a distinct cluster separated from the LuxI of *Burkholderia* species, SolI of *Ralstonia solanacearum*, and RhlI and LasI of *Pseudomonas aeruginosa*, representing a novel evolutionary branch of QS system (**Figure 4**). *Ralstonia* and *Burkholderia* are closely related genera to *Pandoraea*, and they shared highly similar phenotypic profiles that often resulted in the misidentification of *Pandoraea* species (Coenye et al., 2001; Henry et al., 2001). As *Pandoraea* were also predominantly recovered from CF patients, QS genes of *P. aeruginosa* (model organism for QS and CF patients) were also included in the analysis (Barr et al., 2015).

While analysis on the amino acid pairwise identity revealed that similarity of PpnI of different *P. pnomenusa* strains can be as high as 90%, pairwise identity shared between the LuxI of different *Pandoraea* species varied from about 50–90%, even though they were all placed in the same cluster in phylogenetic tree of LuxI (**Figure 4**). When compared to the LuxI of other genera, the LuxI of *Pandoraea* species are only 41–55% in pairwise identity with the well-characterized CepI of *Burkholderia cenocepacia* that catalyzes primarily the synthesis of C8-HSL and a minority of C6-HSL (Lewenza et al., 1999; Lewenza and Sokol, 2001); 48–55% in pairwise identity with SolI, which catalyzes the synthesis of C6-HSL and C10-HSL (Flavier et al., 1997); 26–41% in pairwise identity with CciI, which catalyzes primarily the synthesis of C6-HSL and a minority of C8-HSL (Malott et al., 2005); and lastly, about 26–33% pairwise identity with RhlI and LasI, which catalyzes primarily the synthesis of C4-HSL and 3-oxo-C12-HSL (Latifi et al., 1996). Besides, we also identified the 20 bp lux box located in the upstream region of the *luxI* of *Pandoraea* species

percentages of 1,000 replications) greater than 50%. Bar, 0.2 substitutions per amino acid position.

(**Supplementary Figure S2**). The lux box is a 20 bp palindromic sequence located upstream in the promoter region of *luxI,* which is required for the binding of AHL-activated LuxR (Lewenza et al., 1999). All *luxI* of *Pandoraea* species shared consensus in 14 of 20 lux box sequence (**Supplementary Figure S2**).

### Canonical LuxR1 and LuxR Solos in *Pandoraea* Genus

Phylogenetic and pairwise identity matric analyses performed on all identified canonical LuxR1 of genus *Pandoraea* demonstrated close clustering with CepR from genus *Burkholderia* and SolR from genus *Ralstonia* with 96% bootstrap value (**Figure 5**). Similar to CepI of genus *Burkholderia*, the LuxI of *Pandoraea* produce C8-HSL, but SolI from *Ralstonia* produces two short-chain AHL signals (C6-HSL and C10-HSL). As all identified *luxR*1 of *Pandoraea* are located adjacent to *luxI*, it is believed that the primary function of LuxR1 is for the detection of C8-HSL produced by its canonical LuxI. A comparison on the amino acid sequences of multiple LuxR groups revealed the highest pairwise identity among canonical LuxR1 of *Pandoraea* (71–100%) but lower identity to all the LuxR of other groups (<41%), including the LuxR2 and LuxR3 solos in *Pandoraea*.

Intriguingly, LuxR2 and LuxR3 solos formed two separated clusters on phylogenetic tree with the canonical LuxR identified in *Pandoraea* genus. Both the LuxR solos were distinctive from each other and had CciR of *Burkholderia cenocepacia* and LasR of *P. aeruginosa* as outgroups of the clusters, respectively (**Figure 5**). The LuxR2 solos are highly conserved in the genus *Pandoraea* showing >85% in amino acid pairwise identity among different species, as compared to the canonical LuxR1 (>71% in pairwise identity) and LuxR3 solos (>56% in pairwise identity) (**Figure 5**). From the phylogenetic analysis, it might imply that these LuxR solos in *Pandoraea* represent two novel evolutionary branches of LuxR in QS system. This is supported by a comprehensive search in various databases, which did not return significant matches with any other species and thus indicated that LuxR2 and LuxR3 solos were found exclusively only in *Pandoraea* species.

The widespread distribution of LuxR solos in almost every *Pandoraea* species (except *P. thiooxydans* DSM 25325T)

indicated that they could be playing potential roles in survival and persistence of these species. For *Pandoraea* species that possess QS activity (*P. pnomenusa, P. sputorum, P. oxalativorans,* and *P. vervacti*), additional LuxR solos could function in detecting endogenous AHL signals produced by the AHL synthase to increase the regulatory targets of the complete canonical LuxI/R QS system. Notably, QS positive *P. pnomenusa* and *P. sputorum,* which are clinically isolated, might possess LuxR solos for their survival and persistence in respiratory tracts of CF patients as well as regulation of virulence factors. Similar phenomenon was observed in QscR solo of *Pseudomonas aeruginosa,* which is a LuxR solo that responds to endogenous 3-oxo-C12-HSL produced by LasI to control the timing of AHL production in the species for regulating expression of virulence factors. A study on *qscR* mutant demonstrated that it is hypervirulent in killing its host indicating that QscR solo is important for efficient regulation of QS-mediated virulence factors (Chugani et al., 2001).

In addition, the LuxR solos in *Pandoraea* species could be essential for detecting exogenous AHLs produced by neighboring species, especially for *Pandoraea* species that do not own a LuxI/R AHL system. In this study, AHL production was not observed in *P. apista* DSM 16535T and *P. pulmonicola* DSM 21091T that were isolated from sputa of CF patients. The presence of LuxR solos in these strains could be responsible for eavesdropping by detecting exogenous AHL molecules produced by *P. aeruginosa* that is chronically colonizing the respiratory tracts of CF patients. It is also noteworthy that many Gram-negative bacteria with QS activity such as *Burkholderia* and *Ralstonia* are common pathogens causing lung infections in CF patients. In fact, there are bacteria that possess LuxR solos even though they do not harbor any type of AHL synthase such as *Escherichia coli* and *Salmonella enterica* serovar Typhimurium. These bacteria carry a LuxR homolog, and SdiA was reported able to detect and respond to AHL signaling molecules produced by other bacterial species to activate their gene expression (Ahmer, 2004).


FIGURE 6 | Comparative gene mapping of all QS genes in type strains of *Pandoraea* species. All QS genes were highly conserved at syntenic genomic location. (A) Canonical *luxI* and *luxR*1 were found be convergently inverted and located upstream of alcohol dehydrogenase and ABC transporter ATP-binding protein. (B) *luxR*2 solos located between LysR transcriptional regulators and RND transporter. (C) *luxR*3 solo located downstream of cytochrome c oxidase subunits I and II and a membrane protein.

### Comparative Gene Mapping of All Quorum Sensing Genes and Putative Acquisition Mechanism of LuxR Solos in Type Strains of *Pandoraea*

Since this is the first documentation of *luxR*2 and *luxR*3 solos in *Pandoraea* genus, we are determined to investigate the acquisition mechanism of these genes in *Pandoraea* genus. Hence, we performed comparative gene mapping to study the degree of conservation of all QS genes. All QS genes of *Pandoraea* were found to be highly conserved at syntenic genomic locations (**Figure 6**): all canonical *luxI*/*R*1 were found to be convergently inverted (only *luxI*/*R*1 of *P. sputorum* and *P. norimbergensis* overlapped with each other) and located upstream of an alcohol dehydrogenase and an ABC transporter ATP-binding protein (**Figure 6A**); *luxR*2 solos were consistently located between a LysR transcriptional regulators and a RND transporter (**Figure 6B**); and *luxR*3 solos were always found located downstream of cytochrome c oxidase subunits I and II and a membrane protein (**Figure 6C**). As there are hypothetical proteins located in the immediate upstream of *luxR*2 and *luxR*3, we questioned if these hypothetical proteins could be the canonical *luxI* that have mutated and lost its function or domain. However, after a comprehensive domain prediction was performed on these hypothetical proteins, there was no residue of *luxI* in these hypothetical proteins.

Subsequently, we also performed an extensive search for the presence of any QS genes in genomic island, prophages, and mobile genetic element regions to determine the possibility of horizontal gene transfer event. No QS gene was found on any genomic island, and no residue of transposase was found in close proximity of all QS genes. Although no QS gene was found within any intact prophage region, there are, however, few *luxR* solos that were found in close proximity with incomplete and intact prophage sequences, such as 53,482 bp between PpuR2 (63.0% GC content; 544,114–544,881 bp) with an incomplete prophage region 1 (63.7% GC content; 597,363– 605,870 bp); 62,330 bp between PoxR2 (60.0% GC content; 478,160–478,927 bp) with an incomplete prophage region 1 (62.91% GC content; 533,510–541,257 bp); and 6,231 bp between PoxR3 (65.2% GC content; 2,404,930–2,405,844 bp) with an intact prophage region 8 (63.2% GC content; 2,391,675– 2,398,699 bp). These observations suggested that *luxR*2 and *luxR*3 solos could be transmitted into *Pandoraea* genus by transduction event mediated by prophage. However, parts of these prophage sequences might be lost during evolution. Various QS-related genes had been reported in the genomes of bacteriophages including homologs of accessory gene regulator (*agr*) in the genome of *Clostridium difficile* phage phiCDHM1 (Hargreaves et al., 2014) and regulatory protein LuxR in the *Azospirillum brasilense* Cd bacteriophage's genome (Boyer et al., 2008).

QS activity in *Pandoraea* species has been related to the regulation of virulence factors, biofilm formation, extracellular enzymes production, antibiotic resistance, and various other lethal traits. Although not all *Pandoraea* species exhibit QS activity, findings in this study revealed that almost every *Pandoraea* species (except *P. thiooxydans* DSM 25325T) possess LuxR solos genes in their genomes. The repertoire of LuxR solos in the genus increases the range of gene regulatory activities and is anticipated to play roles in QS-dependent regulation of phenotypic functions, which should be investigated further. The data presented are also useful in future application including quorum quenching (QQ) study that attempts to disrupt the bacterial cell-to-cell communication of *Pandoraea* species through QS (See-Too et al., 2018). QQ has been suggested as alternative antibacterial strategy to antibiotics, which might lead to emergence of multi-drug resistant bacteria (Tang and Zhang, 2014). Last but not least, we hope that findings from this study contribute to further research to elucidate the downstream roles of QS genes in *Pandoraea* species, including their LuxR solos.

## CONCLUSIONS

Multiple species of the genus *Pandoraea* were frequently isolated from sputum samples of CF patients from all over the world, and *Pandoraea* species are identified as emerging pulmonary pathogen associated with CF. While some species were obtained from the environments, clinically isolated species such as *P. pnomenusa* has also been recovered from soils in the environment. This suggests the ubiquitous nature of this group of bacteria, and they are thus identified as opportunistic pathogens. The recent report on the QS activity in *P. pnomenusa* rapidly caught the attention of the scientific community as QS systems have been linked to the regulation of virulence factors, antibiotic resistance, and various traits that are dangerous to patients. Although this study revealed that only four type strains of nine species of genus *Pandoraea* possess AHL-based QS activity, we also reported the presence of two highly conserved *luxR* solos in most of their genomes. Our analyses had revealed that these LuxR solos belonged to different clusters of novel evolutionary branches in QS systems. We hypothesize that these LuxR solos in *Pandoraea* could potentially be responsive to AHLs or different signals produced by neighboring species and coordinate regulation of gene expression, thus playing important roles in the infection process and persistence of these pathogens in cystic fibrosis patients. In the process, we developed an *in silico* systematic bioinformatics prediction workflow, which is useful for LuxI and LuxR genes identification of other species. To summarize, this study lays the foundation for future study on QS systems of *Pandoraea* as a potential antimicrobial target in the treatment of *Pandoraea* infections.

### DATA AVAILABILITY

Publicly available datasets were analyzed in this study. This data can be found here: https://www.ncbi.nlm.nih.gov/genome/.

### AUTHOR CONTRIBUTIONS

KG-C and WF-Y conceived and designed the experiment. RE and YL-L conducted the experiments. KO-C, WS-ST, RE, and YL-L conducted the data analyses. KO-C, WS-ST, and RE wrote the manuscript. All authors read and approved the manuscript.

### FUNDING

This work was supported by University of Malaya Research Grants (PG263-2016A and FP022-2018A), University of Malaya High Impact Research Grants (UM-MOHE HIR Grant UM.C/625/1/HIR/MOHE/ CHAN/14/1, Grant No. H-50001-A000027; UM-MOHE HIR Grant UM.C/625/1/HIR/ MOHE/CHAN/01, Grant No. A-000001-50001) awarded to

### REFERENCES


KG-C, Postgraduate Research (PPP) Grant (Grant No. PG084- 2015B) awarded to RE and (PG089-2015B) to KO-C.

### ACKNOWLEDGMENTS

KO-C thanks MyBrain15 Postgraduate Scholarship Programme for the scholarship (MyPhD, KPT(B)900909146137) awarded. WS-ST thanks the Bright Sparks Program of the University of Malaya for scholarship awarded.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb.2019.01758/ full#supplementary-material


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

*Copyright © 2019 Chua, See-Too, Ee, Lim, Yin and Chan. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Quorum Quenching Lactonase Strengthens Bacteriophage and Antibiotic Arsenal Against Pseudomonas aeruginosa Clinical Isolates

Sonia Mion<sup>1</sup> , Benjamin Rémy1,2, Laure Plener<sup>2</sup> , Fabienne Brégeon1,3, Eric Chabrière<sup>1</sup> \* and David Daudé<sup>2</sup> \*

<sup>1</sup> Aix-Marseille University, IRD, APHM, MEPHI, IHU-Méditerranée Infection, Marseille, France, <sup>2</sup> Gene&GreenTK, Marseille, France, <sup>3</sup> Service des Explorations Fonctionnelles Respiratoires Centre Hospitalo Universitaire Nord, Pôle Cardio-Vasculaire et Thoracique, Assistance Publique des Hôpitaux de Marseille, Marseille, France

### Edited by:

Ana Maria Otero, University of Santiago de Compostela, Spain

#### Reviewed by:

Rodolfo García-Contreras, National Autonomous University of Mexico, Mexico Jin Zhou, Graduate School at Shenzhen, Tsinghua University, China

\*Correspondence:

Eric Chabrière eric.chabriere@univ-amu.fr David Daudé david.daude@gene-greentk.com

#### Specialty section:

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

Received: 23 May 2019 Accepted: 20 August 2019 Published: 03 September 2019

#### Citation:

Mion S, Rémy B, Plener L, Brégeon F, Chabrière E and Daudé D (2019) Quorum Quenching Lactonase Strengthens Bacteriophage and Antibiotic Arsenal Against Pseudomonas aeruginosa Clinical Isolates. Front. Microbiol. 10:2049. doi: 10.3389/fmicb.2019.02049 Many bacteria use quorum sensing (QS), a bacterial communication system based on the diffusion and perception of small signaling molecules, to synchronize their behavior in a cell-density dependent manner. QS regulates the expression of many genes associated with virulence factor production and biofilm formation. This latter is known to be involved in antibiotic and phage resistance mechanisms. Therefore, disrupting QS, a strategy known as quorum quenching (QQ), appears to be an interesting way to reduce bacterial virulence and increase antibiotic and phage treatment efficiency. In this study, the ability of the QQ enzyme SsoPox-W263I, a lactonase able to degrade acyl-homoserine lactones, was investigated for quenching both virulence and biofilm formation in clinical isolates of Pseudomonas aeruginosa from diabetic foot ulcers, as well as in the PA14 model strain. These strains were further evolved to resist to bacteriophage cocktails. Overall, 10 antibiotics or bacteriophage resistant strains were evaluated and SsoPox-W263I was shown to decrease pyocyanin, protease and elastase production in all strains. Furthermore, a reduction of more than 70% of biofilm formation was achieved in six out of ten strains. This anti-virulence potential was confirmed in vivo using an amoeba infection model, showing enhanced susceptibility toward amoeba of nine out of ten P. aeruginosa isolates upon QQ. This amoeba model was further used to demonstrate the ability of SsoPox-W263I to enhance the susceptibility of sensitive and phage resistant bacteria to bacteriophage and antibiotic.

Keywords: quorum quenching, quorum sensing, lactonase, AHL, Pseudomonas aeruginosa, multidrug resistant bacteria, antibiotics, bacteriophages

## INTRODUCTION

Pseudomonas aeruginosa is an opportunistic human pathogen involved in numerous diseases from otitis to keratitis, wound and burn infections, pneumonia and urinary tract infections (Driscoll et al., 2007). P. aeruginosa isolates are among the most frequently found antibiotic-resistant pathogens involved in diabetic foot infections, their presence is usually associated with morbidity

(Ertugrul et al., 2012). Surveillance of P. aeruginosa infections has revealed trends of increasing resistance to antibiotic treatments (Breidenstein et al., 2011) and the priority to support research and development of effective drugs against antibiotic-resistant P. aeruginosa was recently defined as critical by the World Health Organization (WHO) (Tacconelli et al., 2018).

This last decade, bacteriophage (phage) therapy has regained interest as a new weapon to treat antibiotic-resistant infections (Rolain et al., 2015; Domingo-Calap et al., 2016) and is under consideration to treat P. aeruginosa infections. Phages are the most abundant predator of bacteria in nature (Suttle, 2005) and have been domesticated, especially in Eastern Europe, to treat enteric infections, such as cystic fibrosis (Rolain et al., 2015), dysentery (Salmond and Fineran, 2015), diabetic foot infection, chronic osteomyelitis and other surgical and wound infections (Kutter et al., 2010). Phages offer the advantage of specifically targeting their host bacteria by being harmless to the commensal flora (Loc-Carrillo and Abedon, 2011) and human cells (Domingo-Calap et al., 2016). In western countries, clinical trials have been conducted to assay the therapeutic potential of phages (Rhoads et al., 2009; Wright et al., 2009; Kakasis and Panitsa, 2018), for example to treat P. aeruginosa infected burns (Jault et al., 2018). However, as for antibiotics, phages suffer resistance phenomena that may hinder the development of bacteriophage-based therapy (Jault et al., 2018). Finding new therapeutic strategies to limit bacterial resistance is thus of great interest.

In this way, another alternative to treat P. aeruginosa infections aims to highjack its communication system referred to as quorum sensing (QS). In P. aeruginosa, QS mainly relies on the secretion and perception of N-acyl homoserine lactones (AHL) to orchestrate its behavior, including virulence and biofilm formation as well as the CRISPR-Cas defense system, in a celldensity dependent manner (Bassler and Losick, 2006; Høyland-Kroghsbo et al., 2017). Interconnections between bacterial QS and susceptibility to bacteriophage infections were also identified (Qin et al., 2017; Saucedo-Mora et al., 2017). Disrupting QS, a strategy referred to as quorum quenching (QQ), is highly attractive to counteract bacterial virulence by using QS inhibitors (QSI) or QQ enzymes (QQE). Among QQE, special attention has been paid to the robust lactonase SsoPox-W263I (Hiblot et al., 2013) that was proved to efficiently inhibit virulence in vitro in model and clinical strains of P. aeruginosa (Guendouze et al., 2017) and to drastically decrease mortality in a rat pulmonary infection model (Hraiech et al., 2014). Furthermore, the impact of SsoPox-W263I on CRISPR-Cas gene expression of P. aeruginosa was recently demonstrated in both model and clinical strains suggesting that enzymatic QQ may modulate bacterial susceptibility to bacteriophages (Mion et al., 2019).

Here, we consider the use of SsoPox-W263I in clinical isolates of P. aeruginosa from diabetic foot ulcers together with the model strain PA14 and further evolved these strains to resist bacteriophages. Knowing that the evolutionary selection of phage resistance in bacteria can induce phenotypic shifts (Labrie et al., 2010), we focused our interest on determining the potential of QQ to control antibiotic and phage-resistant bacteria. Our results show that SsoPox-W263I is efficient to decrease virulence or biofilm in these multi-resistant strains, both in vitro and in vivo using an amoeba model. In the last part, we investigate the QQE treatment to enhance the therapeutic effect of antibiotic and phage when used as a co-treatment in the amoeba infection model. This study shows that QQ is an interesting strategy to treat bacterial infections that can be used to strengthen the effect of antimicrobial treatments.

## MATERIALS AND METHODS

### Bacterial Strains and Growth Conditions

Experiments were conducted using model strain PA14 and three clinical isolates of P. aeruginosa, isolated from diabetic patients of the Nimes University Hospital presenting diabetic foot infections. All the patients received an oral information, were anonymized and gave a non-opposition statement to bacterial storage. This study was approved by the local ethics committee (South Mediterranean III) and was carried out in accordance with the Declaration of Helsinki as revised in 2008. Clinical isolates of P. aeruginosa and model strain PA14 (UCBPP-PA14) were inoculated from a single colony and pre-cultivated during 6 h at 37◦C in Luria Bertani (LB) medium (10 g l−<sup>1</sup> NaCl, 10 g l−<sup>1</sup> Tryptone, 5 g l−<sup>1</sup> yeast extract) with agitation at 650 rpm. Then, precultures were diluted by a 1,000 factor in MOPS minimal medium complemented with nitrogen (15 mM NH4Cl), iron (5 µM Fe2SO4), phosphate (4 mM K2HPO4) and glutamate (25 mM) as carbon source (MOPS glutamate) (Welsh and Blackwell, 2016) and cultures were incubated at 37◦C under agitation at 650 rpm. Enzymes were added, when indicated, at 0.5 mg ml−<sup>1</sup> . For virulence factor and biofilm formation analysis cultures were incubated during 20 h.

### Phage Production and Isolation

A bacteriophage cocktail (Intesti-bacteriophage, Microgen, Russia) was used for this study. Isolated phage 8Intesti-PA14 corresponds to the isolation and concentration of plaque-forming unit (PFU) formed by the phage cocktail on P. aeruginosa PA14 according to the following protocol.

The double agar overlay plaque assay was used to determine the phage titer and isolate phages from phage cocktail (Kropinski et al., 2009). 500 µl of an overnight culture of bacteria was added to 4.5 ml of molten soft LB-agar (0.75%) and overlaid onto a hard LB-agar plate. Once dry, 10 µl drops of phage cocktail were spotted on the soft LB-agar layer. Plates were incubated overnight at 37◦C. Lytic plaques were collected and suspended in 500 µl of MgSO<sup>4</sup> (10 mM). After chloroform treatment and 10 min of centrifugation at 4,500 g, the supernatant was filtered at 0.22 µm. The resulting phage suspension was again spotted on a double-layer plate, and the experiment was repeated until a 10<sup>8</sup> PFU ml−<sup>1</sup> suspension was obtained. To estimate the phage titer, serial dilutions (from 10<sup>0</sup> to 10−10) of phage suspension were performed in MgSO<sup>4</sup> (10 mM), then 10 µl of each dilution was spotted on the double-layer plate. The plate was incubated overnight at 37◦C. The titer (PFU ml−<sup>1</sup> ) was determined by the calculation of lytic PFU for 1 ml of phage suspension.

### Isolation of Phage Resistant P. aeruginosa Bacteria

fmicb-10-02049 August 31, 2019 Time: 16:11 # 3

Precultures were performed by inoculating one colony of the initial strains in LB medium during 6–8 h, then the precultures were diluted by a 10,000 factor in MOPS glutamate, 3 ml were transferred in each well of a 12-well plate. 10, 50, or 100 µl of bacteriophage containing cocktail (Intestibacteriophage, Ìicrogen, Russia) were added to each well. The plate was incubated at 37◦C overnight under agitation at 650 rpm. The OD 600 nm was then measured using a plate reader (Synergy HT, BioTek). Cultures showing complete lysis were diluted in a solution of phosphate saline buffer (PBS) and 20% glycerol. The dilutions were then plated on LB agar. Plates were incubated at 37◦C overnight.

After 16 h of incubation, the isolated colonies were picked, cultivated in LB medium and stocked in 20% glycerol at −150◦C. The tolerance of the isolated bacteria to the phage cocktail was verified by exposing the strains to the same conditions as those used for their isolation and by verifying that the final OD 600 nm was indeed higher for the resistant strain than for the parental strain.

### Protein Production and Purification

Two SsoPox variants namely W263I and 5A8 were used as QQ and inactive enzymes, respectively. Productions were realized as previously described (Hiblot et al., 2012, 2013; Hraiech et al., 2014). Briefly, Escherichia coli BL21 (DE3)-pGro7/GroEL cells (TaKaRa), carrying plasmid pET22b-SsoPox-W263I or pET22b-SsoPox-5A8, were cultivated in ZYP-5052 medium complemented with 100 µg ml−<sup>1</sup> ampicillin and 34 µg ml−<sup>1</sup> chloramphenicol at 37◦C until OD 600 nm reached 0.8–1. The expression of chaperone proteins was induced by adding Larabinose at a final concentration of 0.2% (w/v). At the same time, the temperature was reduced to 23◦C and 0.2 mM of CoCl<sup>2</sup> was added. After 20 h of incubation, the cells were harvested by centrifugation (4,400 g, 4◦C, 20 min), the pellet was resuspended in lysis buffer [50 mM HEPES pH 8, 150 mM NaCl, 0.25 mg ml−<sup>1</sup> lysozyme, 0.1 mM Phenylmethylsufonyl fluoride (PMSF) and 10 mg ml−<sup>1</sup> DNaseI] and stored at −80◦C during 16 h. Frozen cells were thawed at 37◦C during 15 min and lysed by three steps of 30 s sonication (QSonica sonicator Q700, amplitude at 45). Cell debris were removed by centrifugation (21,000 g, 4◦C, 15 min). Crude extract was incubated during 30 min at 80◦C and then centrifuged to precipitate E. coli proteins (21,000 g, 4◦C, 30 min). The enzyme was then concentrated by overnight incubation at 4◦C in 75% ammonium sulfate. After resuspension in activity buffer (50 mM HEPES pH 8, 150 mM NaCl, 20.2 mM CoCl2) ammonium sulfate was eliminated by desalting (HiPrep 26/10 desalting, GE Healthcare, ÄKTA Avant). The protein sample obtained was concentrated to 2 ml and then loaded on a size-exclusion chromatography column and purified to homogeneity (HiLoad 16/600 SuperdexTM 75pg, GE Healthcare, ÄKTA Avant). Protein purity was checked by migration on 10% SDS-PAGE and protein concentration was measured using a spectrophotometer NanoDrop 2000 (Thermo Scientific).

### Antibiograms

Antibiotic sensitivity of the strains was determined on a Mueller Hinton agar (BioMerieux). The disk diffusion method was realized using the following antibiotics: amikacin (30 µg), cefepime (30 µg), ceftazidime (30 µg), ciprofloxacin (5 µg), doxycycline (30 µg), fosfomycin (50 µg), imipenem (10 µg), nitrofurantoin (300 µg), piperacillin/tazobactam (85 µg), ticarcillin (75 µg), ticarcillin/clavulanate (85 µg), trimethoprim/sulfamethoxazole (25 µg), tobramycin (10 µg), rifampicin (30 µg). The results were interpreted according to the EUCAST guidelines (European Society of Clinical Microbiology and Infectious Diseases, 2018<sup>1</sup> ) using the Scan <sup>R</sup> 1200 (Interscience) (Diop et al., 2016).

### Analysis of Virulence Factor Production

Virulence factor productions for the different strains were determined in vitro after a 20-h culture in presence of 0.5 mg ml−<sup>1</sup> SsoPox-W263I or inactive mutant SsoPox-5A8 as control.

### Pyocyanin Production

Cell-free culture supernatants were prepared by centrifugation for 5 min at 12,000 g. Pyocyanin was extracted by mixing 500 µl of cell-free supernatant with 250 µl of chloroform (Price-Whelan et al., 2007). After vortexing and 5 min of centrifugation at 12,000 g, 200 µl of the bottom chloroform phase were transferred into a quartz 96-well plate. The absorbance was measured at 690 nm using a plate reader (Synergy HT, BioTek).

### Proteolytic Activity

The protease activity was measured using azocasein (Sigma) (Chessa et al., 2000). 675 µl of PBS solution pH 7 were mixed with 50 µl of azocasein (30 mg ml−<sup>1</sup> in water) and 25 µl of cellfree supernatant. After 2 h of incubation at 37◦C, 125 µl of 20% (w/v) trichloroacetic acid were added to stop the reaction. The solution was then centrifugated for 10 min at 10,000 g and the absorbance of 200 µl of the supernatant was measured at 366 nm using a plate reader (Synergy HT, BioTek).

### Elastolytic Activity

Elastase B activity was measured using elastin-Congo red conjugate (Sigma) degradation assay (Smith et al., 2003). 50 µl of cell-free supernatant were added to 150 µl of elastin-Congo red solution (5 mg ml−<sup>1</sup> in 10 mM Tris–HCl and 1 mM CaCl<sup>2</sup> buffer at pH 7.2) into a 96-well plate. The plate was then sealed with an aluminum membrane and incubated during 24 h at 37◦C under agitation (300 rpm). After sedimentation of undigested elastin-Congo red conjugate 100 µl of the upper phase was transferred to an empty well and the absorbance was measured at 490 nm using a plate reader (Synergy HT, BioTek).

### Biofilm Formation

The biofilm formed in each well was quantified using crystal violet (Sigma) staining as previously described (Stepanovic et al., ´ 2000). Briefly, planktonic cells (non-attached) were first removed

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

by washing the wells with 4 ml of PBS. The plates were then dried at 37◦C and the biofilm was stained by adding 4 ml of crystal violet 0.05%, incubated for 3 min under agitation at 150 rpm. Then, the crystal violet was removed and each well was rinsed with 4 ml of PBS. Crystal violet was then resolubilized by adding 3 ml of ethanol 96%. 200 µl of the solution was transferred to a 96-well plate and the final concentration of crystal violet was measured at OD 595 nm using a plate reader (Synergy HT, BioTek).

### Virulence Assay Toward Amoeba

In vivo virulence assay was adapted from a previously described procedure using P. aeruginosa and Acanthamoeba polyphaga Linc AP1 (Fenner et al., 2006). Briefly, 3 ml of bacterial culture were pelleted down and resuspended in Page's amoeba saline (PAS) buffer (2 mM NaCl, 16 µM MgSO4, 27 µM CaCl2, 0.53 mM Na2HPO4, 1 mM KH2PO4, pH 6.9) after a culture in 6-well plates (NuncTM, Thermo Scientific). A. polyphaga Linc AP1 was cultivated during 2–3 days into peptone yeast extract glucose (20 g l−<sup>1</sup> proteose peptone, 2 g l−<sup>1</sup> yeast extract, 0.1 M glucose, 4 mM MgSO4, 0.53 mM CaCl2, 3.4 mM sodium citrate, 50 µM (NH4)2Fe(SO4)2, 2.5 mM KH2PO4, 1.3 mM Na2HPO4, pH 6.8) medium at 28◦C (Fenner et al., 2006). Amoeba cells were recovered after centrifugation at 750 g and resuspended into PAS buffer to 10<sup>5</sup> cells µl −1 . Then, 1 ml of bacterial suspension was spread on a PAS agar plate and was left to dry at room temperature. At the center of each plate, 5 µl of A. polyphaga were spotted and dried at room temperature. Then, plates were incubated at 30◦C over 7 days and amoeba propagation was followed by directly measuring the central spot with a ruler.

To test the combinatory effect of enzymatic and antibiotic treatments, the MOPS bacterial culture was treated with 10 µg ml−<sup>1</sup> of enzyme (either SsoPox-W263I or inactive variant SsoPox-5A8 as control) and 25 µg ml−<sup>1</sup> ciprofloxacin was added to PAS buffer during the resuspension step.

To test the combinatory effect of enzymatic and phage treatments, the MOPS bacterial culture was treated with 10 µg ml−<sup>1</sup> of enzyme (either SsoPox-W263I or inactive variant SsoPox-5A8 as control) and 10<sup>7</sup> PFU ml−<sup>1</sup> of 8Intesti-PA14 phage was added to PAS buffer during the resuspension step.

### RESULTS

### Evaluating Sensitivity of Clinical Isolates to Antibiotics and Bacteriophage Cocktail

PA14 is a model strain originally isolated from a burn wound (Soyza et al., 2013) and B10, C5, and C11 were isolated from diabetic foot ulcerations (Guendouze et al., 2017). Antibiotic susceptibility of the strains was evaluated using the disk diffusion method with 14 antibiotics and analyzed according the EUCAST recommendations (**Figure 1A**). The strains were non-susceptible (resistant or intermediate) to at least two different antibiotics tested belonging to rifamycin, sulfonamide or nitrofuran classes. All strains were found sensitive to the tested agents in β-lactam, aminoglycosides and fosfomycin antimicrobial classes which are commonly used to fight pseudomonal infections. In addition to antibiotic sensitivity, the impact of a commercial bacteriophage cocktail on the strains was evaluated. Interestingly, all the strains were sensitive to the cocktail resulting in drastic decreases in cell density (**Supplementary Figure 1**). These results confirm that bacteriophage-based therapy may constitute an alternative to antibiotherapies in case of resistant infections.

### Isolation and Characterization of Phage-Resistant Variants

Although bacteriophages were virulent to all four strains, resistance phenomena were rapidly observed after exposure of PA14, B10, C5, and C11 to three different concentrations of phage cocktail for 16 h. For PA14, B10, and C11 one mutant was isolated from each strain: PA14R1, B10R1, and C11R1, respectively. Three different mutants, presenting different phenotypes, were isolated from C5: C5R1, C5R2, and C5R3. The newly isolated mutants were cultured in the presence of the same amount of phage cocktail as that used for their isolation to confirm their resistance (i.e., 100 µl for PA14 and PA14R1, 50 µl for B10 and B10R1 and 10 µl for C5, C5R1, C5R2, C5R3, C11, and C11R1). Cell density was compared to the initial strains in the presence of phages after a 16-h culture by measuring the OD 600 nm (**Figure 2**). In the presence of phages, growth was 4 to 7 times higher for all mutants than for parental strains (**Figure 2**). The resistance of the isolated mutants against the phage cocktail was thereby clearly highlighted (**Figure 2**). Antibiotic sensitivity patterns of the phage-resistant strains were further evaluated (**Figure 1B**). As for parental isolates, phage resistant strains were non-susceptible (resistant or intermediate) to at least two different antibiotics tested belonging to rifamycin, sulfonamide or nitrofuran classes. Phage-resistant strains were found to be sensitive to the tested agents in β-lactam, aminoglycosides, cephalosporins, carbapenems and fosfomycin antimicrobial classes. As noticed for B10, B10R1 showed intermediate resistance to rifampicin while all other strains were resistant to this antibiotic. Interestingly, the acquisition of bacteriophage resistance was detrimental to antibiotic resistance in the resistant clones isolated from C5. C5R1, C5R2, and C5R3 lost their resistance against doxycycline and ciprofloxacin, C5R1 being also sensitive to nitrofurantoin conversely to C5, C5R2, or C5R3. Similarly, C11R1 exhibited a lower tolerance to trimethoprim/sulfamethoxazole and doxycycline than C11.

### Quenching Virulence Factors of Antibiotic or Phage Resistant Clones in vitro

The QQ effect of SsoPox-W263I on P. aeruginosa isolates PA14, C5, C11, B10 and their bacteriophage resistant counterparts was investigated by measuring the production of three typical virulence factors in vitro: pyocyanin, protease, elastase as well as biofilm production. Under the tested conditions, the addition of SsoPox-W263I significantly reduced the three virulence factors for all strains compared to controls (**Figures 3A–C**). All tested strains produced detectable levels of pyocyanin and elastolytic


FIGURE 1 | Interpretative zone diameters (mm) of 14 antibiotics used on clinical isolates of Pseudomonas aeruginosa (A) and their associated phage-resistant mutants (B). Susceptibility is expressed as Resistant (red), Intermediate (orange), Sensitive (green).

activity. Pyocyanin levels were reduced by more than 75% in all strains upon enzymatic treatment, while elastase levels were reduced by at least 60% in nine out of ten strains with the addition of SsoPox-W263I (C5 elastase level being reduced by only 25%). Proteolytic activity was detectable for all tested strains but one: the parental isolate C5 which did not produce sufficient levels of proteases with or without SsoPox-W263I treatment to be detected. Conversely, the three phage-resistant derivatives obtained from C5 had each detectable protease activity in the absence of enzyme and this activity was completely extinguished by the addition of SsoPox-W263I (**Figure 3A**). Biofilm production was significantly reduced by more than 70% in six strains when treated with SsoPox-W263I (**Figure 3D**). Interestingly, the production of these four factors differed

between the parental strains and their phage-resistant derivatives (**Figure 3**). It appears that the selection for phage-resistant clones did not result in the selection for mutants with only increased or only decreased biofilm formation. Similarly phageresistant mutants did not all increase nor decrease virulence factor secretion as compared to their parental strains. Thereby no common trade-off due to the selection of phage resistant bacteria can be drawn at the level of these four phenotypic traits. Altogether, these results show the efficiency of SsoPox-W263I in reducing the amount of three virulence factors characteristic of P. aeruginosa and modulating the production of biofilm in vitro in both antibiotic and phage-resistant isolates.

### Evaluation of Quorum Quenching in Amoeba, Virulence Model

To further confirm the potential of SsoPox-W263I to decrease the virulence of bacterial isolates, the in vivo protecting effect of the enzyme was assayed using the amoeba A. polyphaga. The virulence of treated and control bacteria toward amoeba was assayed by measuring the propagation of A. polyphaga on a plate flooded by a pretreated bacterial lawn (Fenner et al., 2006).

Overall, SsoPox-W263I treatment decreased the virulence toward A. polyphaga of 9 out of 10 strains. No effect of QQ was observed for C11, for which virulence toward A. polyphaga remained unchanged with or without SsoPox-W263I treatment, but its phage resisting mutant C11R1 recovered a high sensitivity to the amoeba upon treatment (**Figure 4**). C5, C5R1, and PA14R1 were initially not virulent enough to prevent the propagation of amoeba in the control condition; however, treatment by SsoPox-W263I significantly enhanced its expansion (**Figure 4**). Consistently with in vitro observations on virulence factor production, the results obtained in vivo confirmed that parental and resistant strains behave differently, especially for PA14 and C5 (**Figure 4**). SsoPox-W263I treatment showed a benefic effect in all but one case and no negative effects were observed, highlighting the efficiency of the QQ treatment in reducing the virulence of antibiotic and phage resistant isolates in an in vivo model.

### Combined Effect of Enzymatic and Antimicrobial Treatment

Considering that PA14 displayed antibiogram comparable to most clinical isolates tested in this study, PA14 and its phage resistant mutant PA14R1 were further used as representative candidates to evaluate the combined effect of SsoPox-W263I and antibiotic treatment using the amoeba infection model (**Figure 5**), then the combined effect of SsoPox-W263I and phage treatment was also assayed (**Figure 6**). As observed in our first experiment, with an enzyme concentration of 500 µg ml−<sup>1</sup> , QQ increases the sensitivity toward amoeba. Hence, to better assay the combined effect of enzymatic and antimicrobial treatments, lower doses of SsoPox-W263I were considered. Thus, following a dose response experiment on PA14, 10 µg ml−<sup>1</sup> of SsoPox-W263I was used with antimicrobial as it had a lower impact on the virulence (**Supplementary Figure 2**).

Synergistic effect was observed with a treatment of SsoPox-W263I and 25 mg ml−<sup>1</sup> ciprofloxacin on PA14, PA14R1 (**Figure 5**). At these concentrations, each treatment alone had no effect on the virulence of PA14 against the amoeba and the amoeba could not grow. However, with the combined treatment (ciprofloxacin + SsoPox-W263I), the amoeba was able to completely colonize the Petri dish in 6 and 7 days, respectively, for PA14. PA14R1 virulence was slightly impacted by the antibiotic treatment alone or the QQE treatment alone, but the highest effect was observed with the combined treatment: the Petri dish limit (8.5 cm) was reached 1 day earlier than with the QQE treatment alone, and the growth of amoeba started after 1 day with the combined treatment against 4 days with ciprofloxacin alone.

As the composition of the commercial cocktail was toxic for amoeba, PA14 lytic phages were purified from the cocktail and concentrated to 10<sup>8</sup> PFU ml−<sup>1</sup> . To test the combined effect of SsoPox-W263I and phages, 8Intesti-PA14 was used alone or in combination with the enzyme against PA14 and PA14R1. The isolated phages did not impact the growth of the amoeba. For both PA14 and PA14R1, amoeba growth was faster with SsoPox-W263I alone, yet the combined treatment led to a further increase of amoeba growth (**Figure 6**).

## DISCUSSION

In this study, four strains of P. aeruginosa were used, including three clinical isolates from diabetic foot infections. The antibiotic resistance profiles of the strains revealed that all the clinical isolates presented a significant tolerance to rifampicin, trimethoprim/sulfamethoxazole and nitrofurantoin, confirming previous observations regarding the increasing rate of multi-resistance in diabetic foot infections in the recent years (Lipsky, 2016).

To address antibioresistance issues, bacteriophages and QQ have emerged as promising therapeutic approaches. As with antibiotics, the use of bacteriophages suffers from rapid resistance phenomena such as the formation of a biofilm, the modification of phage receptor expression that can reduce phage entry (Chapman-McQuiston and Wu, 2008), or the cell adaptive inducible CRISPR-Cas (clustered regularly interspaced short palindromic repeat and CRISPR associated proteins) system that recognizes and degrades phage DNA (Barrangou et al., 2007). Interestingly, it has been extensively demonstrated that QQ reduces biofilm formation in P. aeruginosa, thereby increasing antimicrobial treatment efficacy, and a recent study has underlined that QS disruption can also decrease phage resistance by inhibiting QS stimulation of the CRISPR-Cas system (Høyland-Kroghsbo et al., 2017). The efficacy of SsoPox-W263I to modulate CRISPR-Cas system regulation of proteobacteria, including P. aeruginosa, was recently demonstrated (Mion et al., 2019). In addition, recent studies have shown that even without involvement of CRISPR-Cas defense, the phage infection outcome could be different in QS-deficient mutant of P. aeruginosa than in wild type strains (Qin et al., 2017; Saucedo-Mora et al., 2017). As antibiotic or phage resistance can induce

FIGURE 5 | Combinatory effect of SsoPox-W263I and ciprofloxacin on the virulence of clinical isolates and phage resistant mutants against A. polyphaga Linc AP1. For each strain, curves represent the mean diameter of amoeba at different days after incubation in presence of bacteria without treatment (red) or treated with 25 mg ml−<sup>1</sup> ciprofloxacin (orange), 10 µg ml−<sup>1</sup> SsoPox-W263I (blue), or 10 µg ml−<sup>1</sup> SsoPox-W263I and 25 mg ml−<sup>1</sup> ciprofloxacin (green). Error bars represent the standard deviations of three experiments.

life-threatening complications, we evaluated the potential of QQ to act as an alternative therapeutic approach.

First, we selected six mutants derived from the four initial strains as phage resistant toward a commercial phage cocktail. The virulence profiles of each strain were then evaluated and the efficacy of three therapeutic approaches (antibiotics with ciprofloxacin, bacteriophages and QQ with SsoPox-W263I) on the different strains was further assessed alone or in combinations.

In several cases bacteriophage resistance did not affect antibiotic resistance, although the selection pressure induced by the phage cocktail resulted in a loss of resistance to ciprofloxacin and doxycycline in the mutant strains isolated from C5. This evolutionary trade-off is coherent with a previous study, where the phage OMKO1 which targets OprM, the porin of a multi-drug efflux system of P. aeruginosa as receptor-binding site, selected resistant bacteria harboring a change in the efflux mechanism with increased sensitivity to different antibiotic classes, such as tetracycline and fluoroquinolone (Chan et al., 2016).

To determine the potential of QQ treatment on the different strains, four QS-regulated traits, biofilm formation and the production of three virulence factors (pyocyanin, protease and elastase) were measured in vitro. As reported in other studies, some phage resistant bacteria such as C5R1, C5R2, and C5R3 exhibited higher level of virulence factor production than the initial strain (Hosseinidoust et al., 2013). The ability of SsoPox-W263I as QQ agent to decrease virulence in the different strains was also assayed. The QQ treatment significantly reduced the production of the virulence factors for the initial strains as well as their phage resistant mutants, showing that bacteria resistant to phage treatment conserved a functional QS system which can be inhibited by enzyme-mediated QQ. Strains harboring a higher virulence profile were also efficiently quenched using SsoPox-W263I. In order to correlate in vitro production of virulence factors, biofilm formation and in vivo virulence, an amoeba-based assay was developed. Amoeba are eukaryotic organisms able to feed on bacteria. Their phagocytosis and digestion mechanisms are similar to those of macrophage bacterial elimination (Greub and Raoult, 2004), thus amoeba are frequently used to test in vivo virulence of bacteria (Rémy et al., 2018). The model of A. polyphaga feeding on P. aeruginosa, where the growth of the amoeba directly indicates the pathogenicity level of the bacteria, was chosen to assess the efficiency of QQ treatment with SsoPox-W263I. The results showed that the inhibition of virulence by the QQ treatment was efficient for 9 out of 10 strains, allowing the

growth of the amoeba, independently from the resistance profiles to antibiotics or bacteriophages. Interestingly, P. aeruginosa was previously proved to use type III secretion system (T3SS) to kill biofilm-associated amoebae potentially suggesting that differences observed upon enzymatic treatment could be related in the modification of T3SS regulation (Matz et al., 2008). Although variable enzyme effects on virulence factor production or biofilm formation were observed in vitro depending on the strains, our results indicate that QQ is efficient in vivo on most of the antibiotic or phage resistant clinical isolates. This highlights the potential of enzymatic QQ treatment as an interesting alternative in case of therapeutic dead ends.

Furthermore, we evaluated the potential of SsoPox-W263I to act as a complement of antibiotics or bacteriophages to counteract P. aeruginosa virulence toward A. polyphaga. Using lower concentrations of SsoPox-W263I, the combination of QQ and ciprofloxacin enhanced the growth of the amoeba for ciprofloxacin-sensitive strains PA14 and PA14R1 as compared to the antibiotic or the enzymatic treatment alone. Consistently with previous reports using the lactonase from Bacillus sp. ZA12 with ciprofloxacin in a murine burn infection model (Gupta et al., 2015), our results show that enzymatic QQ works in synergy with fluoroquinolones in sensitive strains decreasing efficiently the amount of antibiotics required to fight bacterial infections.

In addition to antibiotics, the synergy of SsoPox-W263I with bacteriophages was underlined. Synergistic effects were observed when P. aeruginosa PA14 was treated by the combined actions of isolated phage 8Intesti-PA14 and SsoPox-W263I. Interestingly, the synergy was also observed on PA14R1 strain, which was resistant to the phage cocktail. In concordance with these observations, recent reports showed close relationships between QS and phage tolerance mechanisms in P. aeruginosa (Mion et al., 2018).

Altogether, the results obtained in vitro and in vivo show that SsoPox-W263I is efficient to decrease bacterial virulence in model and clinical isolates of P. aeruginosa and constitute a proof of concept suggesting that enzymatic QQ can strengthen the therapeutic arsenal available against P. aeruginosa infections by enhancing the efficiency of available treatments including bacteriophages or antibiotics. In addition, it has been shown that this enzyme, issued from an extremophilic organism, resists harsh industrial conditions (Rémy et al., 2016) confirming its tremendous potential for biopharmaceutical applications and

### REFERENCES


should now be evaluated on mammalian models in order to reach clinical trials.

### DATA AVAILABILITY

The datasets generated for this study are available on request to the corresponding author.

### AUTHOR CONTRIBUTIONS

SM, BR, LP, FB, DD, and EC designed the study and wrote the manuscript. SM, BR, and LP performed the experiments. SM, BR, LP, and DD analyzed the data.

### FUNDING

SM is a Ph.D. student granted by the Direction Générale de l'Armement (DGA). BR received a Ph.D grant from the "Emplois Jeunes Doctorants" program of Région Provence-Alpes-Côte d'Azur (PACA, France). This work was supported by Investissements d'avenir program (Méditerranée Infection 10-IAHU-03) of the French Agence Nationale de la Recherche (ANR). This work also received support from RAPID (LACTO-TEX) program from the Direction Générale de l'Armement (DGA).

### ACKNOWLEDGMENTS

We thank Dr. M. Ansaldi and Dr. P. Masson for providing phages; Dr. G. Dubourg for the antibiogram experiment; Prof. B. La Scola for providing amoeba and Prof. J-P Lavigne for providing strains. We also thank Dr. Mikael Elias for fruitful discussions.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2019.02049/full#supplementary-material



**Conflict of Interest Statement:** EC has a patent WO2014167140 A1 licensed to Gene&GreenTK. BR, LP, DD, and EC report personal fees from Gene&GreenTK during the conduct of the study.

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

Copyright © 2019 Mion, Rémy, Plener, Brégeon, Chabrière and Daudé. 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.

#### Edited by:

Ana Maria Otero, University of Santiago de Compostela, Spain

#### Reviewed by:

Jürgen Tomasch, Helmholtz Center for Infection Research, Helmholtz Association of German Research Centers (HZ), Germany Raphaël Lami, Sorbonne Universités, France

> \*Correspondence: Fergal O'Gara f.ogara@ucc.ie

#### †Present address:

José A. Gutiérrez-Barranquero, Facultad de Ciencias, Departamento de Microbiología, Instituto de Hortofruticultura Subtropical y Mediterránea La Mayora (IHSM-UMA-CSIC), Universidad de Málaga, Málaga, Spain Ronan R. McCarthy, Department of Life Sciences, Brunel University London, London, United Kingdom

#### Specialty section:

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

Received: 20 May 2019 Accepted: 29 August 2019 Published: 11 September 2019

#### Citation:

Reen FJ, Gutiérrez-Barranquero JA, McCarthy RR, Woods DF, Scarciglia S, Adams C, Fog Nielsen K, Gram L and O'Gara F (2019) Quorum Sensing Signaling Alters Virulence Potential and Population Dynamics in Complex Microbiome-Host Interactomes. Front. Microbiol. 10:2131. doi: 10.3389/fmicb.2019.02131

# Quorum Sensing Signaling Alters Virulence Potential and Population Dynamics in Complex Microbiome-Host Interactomes

F. Jerry Reen1,2, José A. Gutiérrez-Barranquero<sup>1</sup>† , Ronan R. McCarthy<sup>1</sup>† , David F. Woods<sup>1</sup> , Sara Scarciglia<sup>1</sup> , Claire Adams<sup>1</sup> , Kristian Fog Nielsen<sup>3</sup> , Lone Gram<sup>3</sup> and Fergal O'Gara1,4,5 \*

<sup>1</sup> BIOMERIT Research Centre, School of Microbiology, University College Cork, Cork, Ireland, <sup>2</sup> School of Microbiology, University College Cork, Cork, Ireland, <sup>3</sup> Department of Biotechnology and Biomedicine, Technical University of Denmark, Lyngby, Denmark, <sup>4</sup> Telethon Kids Institute, Perth Children's Hospital, Perth, WA, Australia, <sup>5</sup> School of Pharmacy and Biomedical Sciences, Curtin Health Innovation Research Institute, Curtin University, Perth, WA, Australia

Despite the discovery of the first N-acyl homoserine lactone (AHL) based quorum sensing (QS) in the marine environment, relatively little is known about the abundance, nature and diversity of AHL QS systems in this diverse ecosystem. Establishing the prevalence and diversity of AHL QS systems and how they may influence population dynamics within the marine ecosystem, may give a greater insight into the evolution of AHLs as signaling molecules in this important and largely unexplored niche. Microbiome profiling of Stelletta normani and BD1268 sponge samples identified several potential QS active genera. Subsequent biosensor-based screening of a library of 650 marine sponge bacterial isolates identified 10 isolates that could activate at least one of three AHL biosensor strains. Each was further validated and profiled by Ultra-High Performance Liquid Chromatography Mass Spectrometry, with AHLs being detected in 8 out of 10 isolate extracts. Co-culture of QS active isolates with S. normani marine sponge samples led to the isolation of genera such as Pseudomonas and Paenibacillus, both of which were low abundance in the S. normani microbiome. Surprisingly however, addition of AHLs to isolates harvested following co-culture did not measurably affect either growth or biofilm of these strains. Addition of supernatants from QS active strains did however impact significantly on biofilm formation of the marine Bacillus sp. CH8a sporeforming strain suggesting a role for QS systems in moderating the microbemicrobe interaction in marine sponges. Genome sequencing and phylogenetic analysis of a QS positive Psychrobacter isolate identified several QS associated systems, although no classical QS synthase gene was identified. The stark contrast between the biodiverse sponge microbiome and the relatively limited diversity that was observed on standard culture media, even in the presence of QS active compounds, serves to underscore the extent of diversity that remains to be brought into culture.

Keywords: quorum sensing (QS), microbiome, marine sponge-associated bacteria, cell–cell communication, acyl homoserine lactone (AHL)

#### Reen et al. Quorum Sensing in Complex Interactomes

### INTRODUCTION

fmicb-10-02131 September 9, 2019 Time: 15:15 # 2

The marine ecosystem is considered to be an underexplored resource for the study of bacterial interactions within eukaryotic hosts. Despite a number of well-studied examples of bacterial interactions within marine hosts such as the density dependent production of luminescence by Aliivibrio fischeri within the light organ of Euprymna scolopes, relatively little is known about the interactions that occur within marine microbial communities (Hmelo, 2017). This is particularly true in the case of the ancient invertebrate, the marine sponge. Marine sponges are sessile filter feeders that consume bacteria and other marine matter (Taylor et al., 2007). Bacteria can inhabit the mesophyll matrix of these invertebrates with almost 60% of the biomass of a marine sponge being comprised of bacterial endosymbionts (Wang, 2006). This symbiotic relationship is mutually beneficial whereby bacteria are provided with a sheltered nutrient rich environment and the marine sponges acquire limiting nutrients from the microflora (Mohamed et al., 2008; Blunt et al., 2009; Mayer et al., 2010, 2011). Within the dense polymicrobial environment of a marine sponge, bacteria can engage in a form of chemical communication termed quorum sensing (QS) (Taylor et al., 2004a; Diggle et al., 2007; Hmelo, 2017). Several classes of QS signaling system are known, with autoinducer peptides favored by gram positive bacteria while N-acyl homoserine lactones (AHLs) predominate within gram negative bacteria (Whiteley et al., 2017). AHLs are capable of activating an autoinducing transcriptional regulator which controls the transcription of target genes involved in a wide variety of cellular processes including the production of virulence determinants (Diggle et al., 2007). There are relatively few studies on the prevalence of functional AHL based QS systems within microorganisms inhabiting marine sponges (Taylor et al., 2004b; Mohamed et al., 2008; Cuadrado-Silva et al., 2013; Britstein et al., 2018). A number of studies have focused on the identification of homologs of genes associated with QS pathways (Zan et al., 2011). However, sequence based approaches provide limited information on the functionality of these homologous systems, which for the most part remains to be determined. This homology-based approach is also limited by the lack of nucleotide sequence homology among AHL synthases and AHL responsive transcriptional regulators (Steindler and Venturi, 2007). More recently, screening of marine sponges for AHL signals has revealed a rich diversity likely encoded by the microbial communities residing in those sponges (Britstein et al., 2018). Given the difficulties faced in bringing marine sponge biodiversity into culture, it is intriguing to speculate that these signals may play a role in moderating the dynamics of the microbial communities within which they operate.

To gain more insight into the relevance of AHL based QS systems within the microbiota inhabiting marine sponges, bacterial sponge isolates were screened for the production of AHLs using classical AHL reporter strains. A total of 10 QS producing isolates were identified and characterized for AHL production. Co-culture of QS positive isolates with marine sponge samples resulted in increased culturable plate diversity from these communities, although no new genera were identified. While addition of AHLs alone did not influence growth or biofilm in the marine sponge isolates, supernatants from several QS positive isolates suppressed biofilm formation in the marine sponge Bacillus sp. CH8a sporeforming strain. This suggests that the anti-biofilm activity of the QS active supernatants may be mediated downstream of intact QS signaling systems in the producing isolates. Genome sequencing of a QS positive Psychrobacter sp. isolate identified in this study revealed the presence of LuxR DNA binding domains. However, there was no evidence of a LuxR autoinducer domain or an AHL synthase domain in this or any other sequenced Psychrobacter genome. Further establishing the prevalence, structure and diversity of AHL based QS systems will give a better understanding of the role of AHL signaling in the marine ecosystem, potentially unlocking some of the natural biodiversity encoded therein.

### MATERIALS AND METHODS

### Sponge Collection

Bacteria had previously been isolated from sponge genera including Hexactinellida, Stelletta, Lissodendoryx, Poecillastra, Inflatella. These sponges were collected using a remote operated vehicle on board the Celtic Explorer research vessel, 300 nautical miles off the west coast of Ireland as part of the marine biodiscovery cruise, May 2010. Sponge samples from the Amphilectus genus were collected in Gurraig Sound Kilkieran Bay, Galway (Kennedy et al., 2008). Whole sponge samples were rinsed with sterile artificial sea water (ASW, 3.33% (w/v) artificial sea salts, Instant Ocean) and immediately frozen at −80◦C on board the ship until further processing. A sample of sponge tissue (1 g) was homogenized by grinding with a sterile porcelain pestle and mortar in 9 ml of sterile ASW. The sponge homogenate was subsequently serially diluted in ASW to 10−<sup>5</sup> and 100 µl aliquots of the different dilutions were plated onto Marine agar (MA) (Difco, United Kingdom) and SYP-SW Agar (1% (w/v) starch, 0.4% (w/v) yeast extract, 0.2% (w/v) peptone and 3.33% (w/v) artificial sea salts, 1.5% (w/v) agar. Distinct morphologies were collected to form a library of approximately 650 isolates for subsequent screening and characterization.

### Sponge Microbiome Profiling

As above, two independent samples of each sponge tissue (1 g) were homogenized by grinding with a sterile porcelain pestle and mortar in 10 ml of sterile PBS. DNA was extracted from 100 µl of these samples using the MoBio DNA extraction kit (MoBio) as per manufacturer's instructions. The extracted gDNA was used as a template to amplify the v3-v5 region of the 16S rRNA gene, these amplicons were sequenced to 2 × 300 bp on a Next Gen Illumina MiSeq (V3) platform. Prior to the microbiome analysis, raw reads were demultiplexed based on inline-barcode sequences. The reads were processed using Minimum Entropy Decomposition (Eren et al., 2015). To assign taxonomic information to each Operational Taxonomic Unit (OTU), BLAST alignments of representative sequences

to the NCBI database were performed. Further processing of OTUs was performed using the QIIME software package (version 1.8.0<sup>1</sup> ). Microbiome data has been uploaded on the NCBI Sequence Read Archive (SRA) database (BioProject No. PRJNA555824).

### Quorum Sensing Biosensor Assay

The following reporter strains were used to identify AHL production. Short chain AHLs were detected using the Serratia marcescens SP19 (Poulter et al., 2010). Upon production of short chain AHLs S. marcescens SP19 produces a red pigment, prodigiosin. S. marcescens SP19 is an AHL deficient mutant that only produces prodigiosin in response exogenous AHLs. Marine strains were cultured on Marine Broth (Difco, United Kingdom) supplemented with agar (1.5% w/v) for 72 h at 23◦C. They were then overlaid with soft LB agar (0.1% agar) inoculated with S. marcescens SP19 at an OD600 nm of 0.5. Overlaid plates were incubated at 30◦C overnight. QS was identified by prodigiosin production. As a positive control, 20 µM C4-HSL (Sigma, United Kingdom) was used. C4-C8 chain AHLs were detected using Chromobacterium violaceum CV026. C. violaceum CV026 produces a purple compound called violacein in a QS dependent manner. C. violaceum CV026 contains a transposon in its indigenous AHL synthase gene thus it only produces violacein in response to exogenous AHLs (McClean et al., 1997). Marine strains were cultured on Marine Agar (Difco, United Kingdom) for 72 h at 30◦C. They were then overlaid with soft LB agar (0.1% agar) inoculated with C. violaceum CV026 at an OD600nm of 0.5. Overlaid plates were incubated at 30◦C overnight. QS was identified by violacein production. 20 µM C8-HSL (Sigma, United Kingdom) was used as a positive control. The broad range Agrobacterium tumefaciens NTL4 biosensor was used for the detection of longer chain AHLs. It contains a plasmid pZLR4 carrying a traG:lacZ reporter fusion (Farrand et al., 1996; Yin et al., 2012). In response to exogenous AHLs the lacZ gene is transcribed resulting in the degradation of 5-bromo-4-chloro-3-indolyl-β-D-galactopyranoside in the media. The previously described plating protocol was observed except soft LB was supplemented with 50 µg/ml 5-bromo-4-chloro-3-indolyl-β-Dgalactopyranoside. QS was identified by the breakdown of 5-bromo-4-chloro-3-indolyl-β-D-galactopyranoside resulting in a blue color. A concentration of 20 µM C10-HSL (Sigma, United Kingdom) was used as a positive control.

### Strain Identification

Genomic DNA from bacterial isolates was extracted using the MoBio UltraClean DNA extraction kit (MoBio, United States) following manufacturers' guidelines. Strains were identified by PCR amplification of 16S rRNA genes which was carried out as described (Chan et al., 2010) using the universal primer pairs 27F (5<sup>0</sup> -AGAGTTTGATCMTGGCTCAG-3<sup>0</sup> ) and 1525R (5<sup>0</sup> -AAGGAGGTGWTCCARCC-3<sup>0</sup> ). PCR products were sequenced by MWG Eurofins, United Kingdom. Sequence identification was performed using BlastN and all sequences

<sup>1</sup>http://qiime.org/

were submitted to the NCBI database (Accession numbers: MN209943-MN209952).

### Extraction and TLC Analysis of Culture Supernatants

Extracts for thin layer chromatography (TLC) and ultrahigh performance liquid chromatography-high resolution mass spectrometry (UHPLC-HRMS) were prepared from 200 ml cultures in Marine Broth (Difco, United Kingdom) that had been incubated for 72 h at 30◦C and 180 rpm. Bacterial cells were pelleted by centrifugation at 4,000 rpm for 7 min and the supernatant was filter sterilized using a Nalgene <sup>R</sup> vacuum filtration system (0.2 µm, Sigma-Aldrich <sup>R</sup> , Germany). The supernatant was incubated with 1:1 volume of acidified ethyl acetate (1% formic acid) for 10 min at RT shaking at 180 rpm. The ethyl acetate phase was taken and dried using rotary evaporation. Residues were resuspended in 1 ml ethyl acetate and stored at −20◦C. Ethyl acetate extracts were tested for QS activity by spotting 20 µL on C18 reverse phase TLC plates (20 cm × 20 cm TLC aluminum plates, Millipore, United Kingdom) with a methanol:water 7:3 (v/v) mobile phase. Once dried, TLC plates were overlaid with 25 ml of soft LB agar (0.1% agar) inoculated with the biosensor at an OD600nm of 0.5. Following confirmation of extract activity, extracts were dried using nitrogen evaporation and analyzed by UHPLC-HRMS for identification.

### UHPLC-HRMS Profiling of QS-Active Supernatant Extracts

Samples were resuspended in approx. 150 µl 50:50 (vol/vol) acetonitrile (ACN)-water 50:50 (vol/vol) and 2 µl subsamples analyzed by UHPLC-HRMS. This was done on an Agilent Infinity 1290 UHPLC system (Agilent Technologies, Santa Clara, CA, United States) coupled to an Agilent 6550 QTOF MS operated in positive electrospray (ESI) mode, scanning m/z 50-1700. A lock mass solution of 10 µM Hexakis(2,2,3,3 tetrafluoropropoxy)phosphazene (Apollo Scientific Ltd., Cheshire, United Kingdom) dissolved in 95% acetonitrile and infused in the secondary ESI sprayer using an extra LC pump at a flow of 20 µl/min, and the [M + H]<sup>+</sup> at m/z 922.0098 used as lock mass, resulting in a mass accuracy better than 3 ppm (deviation relative to theoretical m/z value).

An Agilent Poroshell 120 phenyl-hexyl column (2.1 × 150 mm, 2.7 µm), held at 60◦C was used for separation. A linear gradient at 0.35 ml/min, consisting of water and ACN both buffered with 20 mM formic acid was started at 10% ACN and increased to 100% after 15 min, where it was held for 2 min. It was subsequently returned to 10% ACN in 0.1 min and maintained for 3 min (Kildgaard et al., 2014).

A reference standard mixture containing the following HSLs were included in the analytical sequence: C4, C6, C8, C10, C12, Oxo-C6, Oxo-C8, Oxo-C10, Oxo-C12, OH-C6, OH-C8, OH-C10, and OH-C12. Using the Agilent MassHunterQuant software, extracted ion chromatograms the [M + H]<sup>+</sup> and [M + Na]<sup>+</sup> ions ± 10 ppm were used for identification, along with the isotopic pattern (Kildgaard et al., 2014), and finally a retention time match ± 0.01 min.

### Growth Analysis of Marine Isolates

fmicb-10-02131 September 9, 2019 Time: 15:15 # 4

Isolates with distinct morphologies that were cultured on QStreated plates and identified by 16S rRNA sequencing were grown on marine agar for 72 h at 23◦C. Cells were transferred into fresh marine broth, OD600 nm 0.05, in the presence and absence of 50 nM–10 µM of 3-oxo-C12 HSL, with DMSO as carrier control. Growth was measured spectrophotometrically at OD600 nm in honeycomb plates incubated at 23◦C on a Bioscreen-C automated growth curves analysis system (Growth Curves USA).

### Biofilm Assays

Bacillus sp. CH8a was grown in marine broth (MB) at 23◦C overnight with shaking (Phelan et al., 2012). In order to monitor the impact on biofilm formation, 500 µl of cell-free supernatant (CFS) from the QS strain 3-day cultures was added to 500 µl of test cultures at OD600 nm 0.1, grown in MB. As controls, 500 µl of fresh media was added to 500 µl of test cultures (media control) and 500 µl of the CFS of test cultures were added to 500 µl of corresponding test culture. In all cases, CFS was obtained by centrifugation at 8,000 rpm for 5 min followed by filtration through a 0.2 µm sterile filter. Bacillus sp. CH8a biofilms were incubated for 2 days at 23◦C. Unattached cells were aspirated out of all wells which were washed once with 1 ml of sterile water. Attached cells were quantified using 0.1% (w/v) crystal violet. For marine isolate biofilm analysis, 3-oxo-C12-HSL was added at 10 µM to media and processed as described above.

### Co-culture Sponge Enrichment Assays

A simple co-culture system was designed to expose marine sponge homogenate to signals and metabolites from actively growing QS active isolates. Costar Spin-X 0.22 µm cellulose acetate centrifuge tube filters (Corning) consisting of an upper chamber with a filter base and a lower receptacle chamber was used for this purpose. QS active isolates were grown in nutrient media in the top chamber, while marine sponge homogenate was incubated in the lower chamber. While the marine sponge homogenate was physically separate from the QS active isolate at all times, transfer of small molecular signals and metabolites between both chambers was possible. In order to establish the validity of the co-culture system for QS transfer between the upper and lower chambers, as well as the integrity of the filter system to prevent bacterial leakage, several test studies were performed. The upper chamber was removed, and 1 ml of LB broth was added to the lower chamber. The upper chamber was replaced and a known QS producing organism P. aeruginosa PA14 (500 µl) was added. The tubes were incubated at 37◦C for 24 h after which time the top chamber was removed. Aliquots of the lower chamber were tested for (a) activation of AHL biosensor strains C. violaceum and S. marcescens and (b) contaminating bacterial growth.

Quorum sensing active isolates were grown in marine broth at 23◦C for 72 h. Stelletta normani sponge sample (∼1 g) was homogenized in 10 ml of PBS. The upper chamber of the Costar Spin-X 0.22 µm cellulose acetate centrifuge tube filter (Corning) was removed, and 500 µl of sponge homogenate with 500 µl marine broth was added to the lower chamber of the filter tube. The filter was repositioned and 300 µl of QS active isolate in marine broth at OD600 nm 0.2 was added. Several controls were included: (i) 300 µl of marine broth was added to the top chamber to provide a baseline profile of bacteria that could be cultured under the standard media conditions used, (ii) each QS active isolate was prepared as above with no sponge sample in the lower chamber to control for leakage through the membrane, and (iii) media controls were included to control for inadvertent contamination (**Figure 1**). The tubes were incubated at 23◦C with gentle shaking at 50 rpm for 72 h at which time the upper chambers were carefully removed. The contents of the lower chamber were mixed gently and serially diluted. Dilutions were plated on marine agar and incubated at 23◦C for at least 72 h. Colony numbers and morphologies were profiled and distinct isolates were identified by 16S rDNA sequencing.

### Draft Genome Sequencing of QS Active Psychrobacter sp.

Total DNA of Psychrobacter sp. 230 strain was obtained using the UltraClean microbial DNA isolation kit (Mo Bio Laboratories, Inc., Carlsbad, CA, United States) and was used for DNA library preparation using a TruSeq exome library prep kit. The draft genome sequencing project of Psychrobacter sp. 230 strain was performed by the Beijing Genomics Institute (BGI, China) using the Illumina HiSeq 4000 sequencing platform involving pairedend reads with a read length of 150 bp. The superfast FASTA/Q file manipulation tool, readfq.v5 (BGI, unpublished software), was used for quality trimming. This software removes the pairedend reads with a certain proportion of low-quality bases (default, 40%; parameter setting, 6 bp), reads with a certain proportion of Ns (ambiguous bases; default, 10%; parameter setting, 10 bp), reads with adapter contamination (default, 15 bp overlapped between adapter and reads), and duplicate sequences. Thus, the high-quality-filtered reads were all 150 bp long. From a total of 5.417.936 raw paired-end reads of 150 bp length, 3.329.182 highquality reads were generated after processing with readfq.v5. Assembly was performed using SOAPdenovo 2.04 with default parameters. The sequencing depth provided 47.5 coverage of the genome. The draft genome assembly comprised 69 contigs with an N50 value of 183,949 grouped into 69 scaffolds with a total size of 3,290,930 bp and an overall GC content of 42.8%. The whole-genome shotgun project was deposited at NCBI under the accession number: NZ\_SNVH00000000.1. The raw reads obtained after processing with readfq.v5 have been submitted to NCBI SRA under the accession number SRP216019.

### Phylogenetic Analysis

To infer the phylogenetic history, nucleotide sequences were retrieved and downloaded from NCBI: GenBank<sup>2</sup> . Sequences were aligned using Clustal Omega and were cut to consistent lengths (Sievers et al., 2011). The evolutionary analysis was conducted in MEGA X using the Neighbor-Joining method (Saitou and Nei, 1987; Kumar et al., 2018). The clustering was tested using Bootstrapping with 1,000 replicates (Felsenstein, 1985). The Tajima–Nei method was used to calculate the

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

evolutionary distances (Tajima and Nei, 1984). There was a total of 1,459 positions in the final dataset.

## RESULTS

### QS Signaling Potential Within Marine Sponge Microbiomes

A range of QS active marine sponge microbial communities have been reported in recent years (Mohamed et al., 2008; Zan et al., 2012; Abbamondi et al., 2014; Britstein et al., 2018). The presence of QS systems in the sponge microbiota suggests a dynamic and ordered community that can respond to external cues and challenges. However, the extent to which QS producing bacteria colonize the marine sponge, and the role of QS within those microbial communities remains to be determined. Therefore, two distinct marine sponges were selected for microbial community profiling to establish the extent to which QS potential existed therein. Given the heterogeneity that exists in many clinical and ecological niches, with localized population profiles existing within relatively short distances of each other within a singular niche, two separate samples of each marine sponge species were selected. Homogenization and subsequent sequencing of two independently harvested triplicate samples revealed some interesting features of the respective microbiomes. Approximately 50% of the S. normani samples was identified only to the level of kingdom classification, suggesting a unique and unexplored bacterial diversity (**Supplementary Table S1**). The dominant phyla were Proteobacteria and Actinobacteria. At the genus level, Iamia [14.9% (±1.9)], Pseudoalteromonas [6.1% (±0.6)], Moraxellaceae [4.7% (±0.1)], and Nitrospira [2.5% (±0.6)] were the most abundant identifiable genera (**Figure 2A**). In contrast, the largest group identified for sponge BD1268 was the γ-Proteobacteria, representing between 85 and 94% of the OTUs in these samples. Sponge BD1268 was colonized by genera such as Candidatus Pelagibacter [12% (±4.5)], Thioprofundum [4.2% (±1.0)], and Thiohalophilus [2.4% (±0.4)]. Although the individual profiles remained relatively consistent from the perspective of the dominant phyla and families, there were differences in relative abundances between the microbiome profiles from the distinct sponge samples suggesting that heterogeneity of the population may exist within the sponge (**Figure 2A**). Diversity indexes (Shannon and Simpson) and OTU abundance was higher in the S. normani samples when compared with the respective samples from the BD1268 sponge (**Figure 2B**). As expected, independent samples clustered together based on sponge source (**Supplementary Figure S1**). Genera known to encode QS signaling systems (e.g., Pseudomonas, Halomonas, Psychrobacter) were present in the microbiomes of both sponges, although it was interesting to note that they were low in abundance when compared to the principal colonizers of the sponges. Therefore, notwithstanding the fact that genera previously shown to encode QS systems were present, the degree to which QS signaling pervades in these sponges remained to be determined.

### Identification of AHL Activities in Bacteria Isolated From Marine Sponges

The collection of morphologically diverse bacterial isolates from the S. normani sponge described above had previously been

reported (Gutierrez-Barranquero et al., 2017). Together with this, cultivation of bacteria from sponge tissue derived from other sponge families such as Hexactinellida, Lissodendoryx, Poecillastra, Inflatella, and Amphilectus had yielded a collection of bacteria from the Proteobacteria, Actinobacteria, Bacteroidetes and Firmicutes groups (Gutierrez-Barranquero et al., 2017). Within these groups a wide range of different families were represented. The most abundant families included Pseudoalteromonadaceae, Flavobacteriaceae, and Moraxellaceae. To establish the prevalence of AHL based QS within the sponge microbiome, over 650 sponge isolated bacteria were screened for AHL production. In total 10 AHL producing candidates were identified, representing approximately 0.015% of the culture collection (**Supplementary Figure S2** and **Table 1**). In order to identify the AHL positive isolates to a species level, the full length 16S rRNA (1,400 bp) sequence was determined. Among the isolates identified with AHL activity were a number of known AHL producers these included members of the Halomonas, Psychrobacter, Vibrio, Pseudoaltermonas, and Pseudomonas genera (Bruhn et al., 2005; Huang et al., 2009; Tahrioui et al., 2011; Ma et al., 2016). Despite the fact that the culture collection was populated by bacterial isolates harvested from six sponges, only four of these (Stelletta, Hexactinellida, Inflatella, and Lissodendoryx) yielded QS active isolates. The detection of QS active genera such as Pseudomonas and Halomonas from within the S. normani strains, despite the low abundance evident in the microbiome profile (**Supplementary Table S1**), is an important finding and suggests potentially a temporal role for QS active strains within communities. All QS active isolates were further validated by TLC analysis and soft agar biosensor overlay prior to selection for UHPLC-HRMS analysis and structural characterization (**Supplementary Figure S2**).

To identify the AHLs being produced by each of the QS positive candidates UHPLC-HRMS was performed


#Trace detected.

(**Supplementary Figure S2**). Ethyl acetate extracts were prepared for all isolates that had tested positive for AHL. All extracts were tested to ensure activation of respective AHL biosensors. Based on UHPLC-HRMS analysis, AHLs were identified in the majority of isolates (**Table 1**). The most abundant AHLs being produced by these isolates were 3-oxo-C10-HSL and 3-oxo-C12-HSL. Both AHLs, in addition to 3-OH-C10 HSL, were identified in extracts from Psychrobacter sp. and Pseudoalteromonas sp., the latter also producing trace amounts of 3-OH-C6 HSL. Only one isolate produced detectable amounts of C4-HSL, with the same Pseudomonas sp. isolate also producing trace amounts of C6-HSL. In some cases, no AHL traces were identified, notwithstanding the ability of the isolates to activate the respective biosensors.

### Co-culture With QS Positive Isolates Influences Culturable Outputs From Marine Sponge

The marine sponges profiled in this study appear to sustain a significant network of QS systems which may be important in modulating the population in response to external cues. Previous studies have shown that the ratio of QS and QQ strains can change dramatically in response to environmental conditions (Tan et al., 2015). Therefore, we investigated whether the secretome of the QS active isolates could impact on the profile of culturable bacteria that could be obtained from marine sponges under the standard media conditions used. Filter based microtubes were used to establish a co-culture system whereby the QS active isolate was added to a well with a porous membrane through which secreted QS signals can be transferred to the chamber below containing S. normani sponge homogenate (**Figure 3A**). In three independent studies, distinct colony morphologies were observed on plates with QS treated samples when compared to untreated controls (**Figure 3B**). The fidelity of all controls was maintained throughout the experiments with no leakage or inadvertent contamination observed.

Pseudoalteromonas and Vibrio were both cultured from the S. normani sponge homogenate in the absence of any treatments. Upon co-culture with QS active isolates, Pseudomonas, Paenibacillus and Psychrobacter were isolated in independent experiments. This was in addition to Pseudoalteromonas and Vibrio as seen on the control plates. Co-culture with QS-active isolates 12 and 411 resulted in isolation of Pseudomonas while 214 and 211 led to isolation of Paenibacillus. Psychrobacter was isolated upon co-culture with all 211, 214, and 411. No representative from these genera was observed on the untreated control plates (**Figure 3**). All three genera were represented at a very low relative abundance within the microbiome from the untreated S. normani sponge samples (Pseudomonas [0.003% (±0.003), Psychrobacter [0.6% (±0.1)], and Paenibacillus only evident at Order level of Bacillales [0.02% (±0.05)]) suggesting a change in community structure (**Supplementary Table S1**). Together, these data suggest that cell-cell communication and QS can impact the dynamics of population growth within microbiomes and influence the culture-readiness of genera in response to external cues. It was notable that the impact of QS active supernatants was not restricted to gram negative organisms. However, the vast majority of microbiome constituents were not represented in the culturable diversity on the plates. Representatives of e.g., Iamia, Nitrospira, Caldilinea, or Gaiella were not observed on either control or test plates. Therefore, modification of the media to cater for additional supplemental requirements, or indeed to reduce nutritional richness, may be required to capture these unculturable components.

### QS Phenotype Profiling

In order to understand how QS signaling might influence species dynamics within the marine sponge, we investigated whether marine isolates obtained following co-culture with QS positive marine sponge isolates responded to exogenous AHLs. As the most frequently identified AHL in this study, 3 oxo-C12 HSL was selected to assess its influence on two key QS associated phenotypes i.e., biofilm formation and growth. Somewhat surprisingly, neither was affected upon addition of 3 oxo-C12 HSL when compared to untreated samples (**Figure 4**). Addition of either 10 or 50 µM 3-oxo-C12 HSL did not impact on growth of test strains, either in exponential or stationary phase, as determined over 72 h (**Figure 4** and **Supplementary Figure S3**). Similarly, although the test isolates formed biofilms in multi-well plates, addition of either 10 or

FIGURE 3 | Co-culture sponge enrichment analysis. (A) Proof of concept using QS producing strains. Transfer of the QS active compounds confirmed by validation on Biosensor seeded plates. (B) Outcome of QS-mediated sponge enrichment assays measuring recoverable morphologies which were subsequently 16S rRNA-typed. Each datapoint refers to the number of distinct species recovered after the designated treatment. The species listed represent recoverable isolates identified in the study. Data presented represents at least three independent assays encompassing distinct S. normani sponge preparations. Statistical analysis was performed by Student's t-test. ∗∗∗p ≤ 0.001.

50 µM 3-oxo-C12 HSL did not significantly alter total attached biomass (**Figure 4**).

The ability of QS to control a range of virulence related phenotypes is well established. In addition to controlling biofilm formation and toxin secretion in important pathogens, QS is also known to elicit an antagonistic response toward co-colonizing organisms in producing strains. To test if this applied to the marine sponge QS active isolates, biofilm formation in the marine sponge sporefomer Bacillus sp. CH8a was investigated in the presence and absence of extracts from QS positive isolates.

Addition of extracts from several of the isolates led to a significant reduction in biofilm formation by CH8a when compared with the untreated control (**Figure 5**). This suggests that QS active strains would likely produce compound(s) that would moderate the behavior of co-existing microbes within the sponge microbiome.

### Genome Sequencing and Identification of LuxR Domain Proteins Within the Psychrobacter Genus

Suppression of biofilm and activation of QS biosensors suggest that QS signaling is a highly networked and evolved system in the marine sponge niche. However, very little is known about the factors involved in newly emerging QS positive species. Psychrobacter sp. have only recently been shown to be QS positive, although the molecular mechanism underpinning this activity remains to be elucidated. To assess if any proteins encoded within Psychrobacter genomes could function as LuxR homologs the available Psychrobacter genomes were investigated via the Pfam Domain Search Tool to identify proteins with specific domains associated with LuxR type transcriptional regulators, i.e., the autoinducer binding domain (PFAM03472) or the LuxR DNA binding domain (PFAM00196). A total of 70 proteins were identified in 38 genomes that possessed a LuxR type DNA binding domain (**Supplementary Table S1**). However, neither an autoinducer binding domain nor a classical autoinducer synthase domain (PFAM00765) were identified in any of the available genome sequences. This raised two possibilities; (i) a potential novel AHL synthase with a low sequence homology to known AHL synthases may be functionally active within the Psychrobacter genus, or (ii) marine Psychrobacter sp. encode novel regions of DNA carrying the capacity for AHL production. Several reports exist in the literature of QS active strains where the corresponding genetic systems encoding that activity remain to be identified. The prevalence of horizontal gene transfer, and the phenotypic and

genotypic heterogeneity that exists within communities is such that interrogating model genomes can be limited when searching for a particular functionality that may be strain specific. We considered it important to investigate the QS signaling potential that is encoded in the genome of the Psychrobacter species, particularly as we had a QS active strain with which to interrogate.

Whole genome sequencing of the Psychrobacter sp. 230 isolate from this study revealed a genome encoding 2908 genes (**Supplementary Figure S4**). The draft genome assembly comprised 69 contigs with an N50 value of 183,949 grouped into 69 scaffolds with a total size of 3,290,930 bp and an overall GC content of 42.8% (**Table 2**). The Psychrobacter sp. 230 genome was searched for Lux domains and three putative LuxR domain proteins were identified. BLASTX sequence searches and SMART domain analysis suggested that these three proteins were transcriptional response regulator proteins, with no evidence of autoinducer domains. A LysE family homoserine(lactone) translocation protein was also encoded in the genome, as was a dienelactone hydrolase family protein. However, the absence of a LuxI-like synthase gene in the Psychrobacter genome indicates that the molecular mechanism through which AHL based QS is performed in this isolate remains to be ascertained. Cluster based analysis revealed that this Pscyhrobacter sp. grouped with three other Psychrobacter sp. isolates and distinct from other members of the genus suggesting it may be a genetic outlier within Psychrobacter (**Figure 6**). Previously, Ma et al. (2016) identified a Psychrobacter sp. isolate from mangrove with QS activity that clustered with Psychrobacter sp. isolates from deep sea sediments of the east Pacific Ocean. Therefore, a functional approach may be warranted to uncover the molecular basis of QS signaling in this species.

### DISCUSSION

In this study, 650 marine sponge bacterial isolates were screened for the ability to produce AHLs. A total of 10 isolates were identified that were capable of activating AHL biosensor reporter strains. Mass spectrometry revealed that several of the isolates produced the same or similar AHLs (OC10–OC12 HSL). The capacity for AHL based signaling in the marine ecosystem has previously been reported. AHL signaling in marine snow was first described by Gram et al. (2002), with species of Roseobacter shown to be QS active. More recently, Pantoea ananatis has been reported to produce a spectrum of AHL signals in marine snow, governing extracellular enzyme production in

TABLE 2 | Genome data for Psychrobacter sp. 230 marine sponge isolate.


producing strains (Jatt et al., 2015). Since the first description of AHL based QS in A. fischeri species (Nealson and Hastings, 1979), where the LuxIR paradigm system was first identified, AHL signals have been found in a broad diversity of marine isolates (Hmelo, 2017). Rasch et al. (2007) described AHL production in Aeromonas salmonicida isolates, while a number of studies profiled members of the Vibrionaceae for AHL production (Yang et al., 2011; Purohit et al., 2013). The diversity of AHL signals that are encoded in the marine ecosystem has been highlighted by a recent study reporting AHLs with long (up to 19 carbons) and poly-hydroxylated acyl side chains (Doberva et al., 2017). At the same time, studies reporting

QS inhibition or quenching in the marine environment have also received considerable attention in recent years (Romero et al., 2012; Gutierrez-Barranquero et al., 2017; Ma et al., 2018). Primarily produced by microbial species, host derived quorum quenching (QQ) has also been described (Weiland-Brauer et al., 2019). Elucidating and profiling the extent of QS signaling within these environments is a key step in understanding the functional role played by QS in the hostmicrobe interaction.

Interspecies communication within the microbial communities of the marine sponge may offer a competitive advantage through cross-genus coordinated behavior. If several species all produce the same or similar AHLs, then they potentially could adopt community like behaviors more rapidly than species that are not part of this interspecies signaling network. This could arise from the activity threshold being reached more rapidly if several different species produce the same signaling molecule. Of course, conservation within receptor systems would also be an integral factor in moderating these responses. The prevalence of orphan LuxR receptor systems in sequenced microbial genomes highlights the complexity of signaling interactions that remain to be identified and understood (Patankar and Gonzalez, 2009). Adopting community-like behaviors through QS systems may offer a distinct competitive advantage as bacteria can attach to a form a biofilm like structure within the environment of the sponge. Community-based small molecular interactions may also be important with respect to intracellular sponge symbionts, such as the recently reported Candidatus Endohaliclona renieramycinifaciens intracellular interaction with Haliclona (Tianero et al., 2019).

The prevalence and diversity of AHLs being produced by the sponge bacterial isolates identified in this study suggests that mechanisms to inhibit these systems may also exist within the sponge microenvironment. The identification of novel compounds that are capable of inhibiting AHL based QS systems is one of the key areas of focus in the development of next generation antimicrobials. Previously, we have reported on the profiling of a subset of this collection of marine sponge isolates for quorum sensing inhibitory (QSI) or QQ activity. A total of 18/440 culturable isolates were found to encode QSI, being able to supress AHL signaling is an isolate dependent manner (Gutierrez-Barranquero et al., 2017). It was interesting to note in that study that several species possessed dual QS and QSI activities. In this current study the finding that Psychrobacter sp. isolates from the same sponge collection were also capable of QS activity suggests that community level moderation of group behavior is a highly evolved trait in the marine ecosystem. Tan et al. (2015) previously showed how the dynamics of QS and QSI/QQ producing organisms can fluctuate in response to changes in environmental conditions. It is noticeable in this regard that two species cultured from QS treated sponge homogenate, Pseudomonas and Paenibacillus, are themselves known to possess AHL signaling systems (Ma et al., 2016). Understanding the interplay between QS and QSI/QQ in the marine sponge ecosystem and the role of QQ in moderating community behavior will underpin advances in marine ecology and beyond.

The dynamics of AHL production in marine microbial communities is seen as a mechanism to enhance culturability of rare genera, many of which encode valuable biosynthetic gene clusters for natural products such as antibiotics and anti-cancer drugs (Reen et al., 2015). While co-culture with QS positive isolates did alter the profile of culturable bacteria isolated from marine sponge homogenates, they failed to introduce new genera into culture. This of course could be due to limitations in the culture conditions, including the general nature of the media used which is more conducive to the culture of fast-growing bacteria. Dilution based methods and modification of the growth conditions with regard to media, temperature, and time may provide the optimum conditions for culture of QS dependent organisms (Rygaard et al., 2017).

The absence of a LuxIR system in the QS positive Pychrobacter sp. 230 isolate would suggest that a hidden diversity to the molecular mechanisms underpinning QS signaling remains to be elucidated. This is consistent with previous reports of AinS and LuxM family autoinducer synthase enoding genes, quite distinct from their LuxI counterparts (Venturi and Subramoni, 2009). Recently, a new LuxIR based system termed TswIR has been identified in an uncultured symbiont from the Red Sea Sponge Theonella swinhoei (Britstein et al., 2016). The synthase protein TswI (COG3916) was annotated as both an autoinducer synthase and a GNAT acetyltransferase activity and while GNAT acetyltransferase proteins were identified in the Psychrobacter genomes, no members of the COG3916 family were found. Furthermore, the recent finding that LuxIR homologs can synthesize and respond to non-acyl HSL signals, serves to underscore the hidden complexity in these systems (Ahlgren et al., 2011). Two orphan Photorhabdus LuxR proteins, PluR and PauR, sense alpha-pyrones and dialkylresorcinols, respectively (Brameyer and Heermann, 2015). It is possible that other examples of non-AHL LuxR interactions may be uncovered in the future, something that would add greatly to the complexity of the signaling interactions as currently understood. The absence of homologs of these proteins in the Psychrobacter sp. 230 genome may necessitate a functional approach in order to elucidate the molecular mechanism through which AHL signaling is established in this and other marine genera.

### DATA AVAILABILITY

The datasets generated for this study can be found in the NCBI Database accession nos: MN209943-MN209952, NZ\_SNVH00000000.1, SRP216019 and PRJNA555824.

### AUTHOR CONTRIBUTIONS

FR and FO'G conceived the study. FR, JG-B, CA, DW, RM, SS, and KN performed the experimental analysis. FR wrote the

manuscript with inputs from all the authors. FR and FO'G finalized the manuscript for submission.

### FUNDING

FR and FO'G acknowledge support from Enterprise Ireland (CF-2017-0757-P) and the Health Research Board/Irish Thoracic Society (MRCG-2018-6). This research was also supported in part by grants awarded to FO'G by the European Commission (EU2020-634486-2015), Science Foundation Ireland (SSPC-2, 15/TIDA/2977), the Irish Research Council for Science, Engineering and Technology (GOIPG/2014/647), the Cystic Fibrosis Foundation, United States (OG1710), and the Health Research Board/Irish Thoracic Society (MRCG-2014-6). KN is grateful to Agilent technologies for the Thought Leader Donation of the UHPLC-QTOF system.

### ACKNOWLEDGMENTS

The authors thank Iwona Kozak and Niall Dunphy for excellent technical assistance and Jamie Deery for generation of the PCA plots.

### REFERENCES


### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2019.02131/full#supplementary-material

FIGURE S1 | PCA bi-plot cluster analysis of microbiome samples performed using R (v. 3.5.2). Visualization was performed using the ggplot and ggfortify packages. Samples from both sponges form separate clusters reflecting their distinct microbial community profiles.

FIGURE S2 | (A) Screen of Marine Sponge Isolates using QS-Biosensor Strains. (B) QS positive isolates were grown in culture flasks to confirm biosensor activation. Species level identification was achieved by 16S rRNA sequencing and subsequent BLAST analysis. (C) Extracts were validated by TLC overlay and subsequently sent for UHPLC-HRMS analysis and classification.

FIGURE S3 | (A–F) Growth profiling of isolates harvested following co-culture with QS active strains in the presence of 10 µM 3-oxo-C12-HSL or DMSO carrier control. Data presented is the average (±SEM) of two independent biological replicates with five technical replicates in each experiment performed on the BioScreen-C.

FIGURE S4 | Genome representation of the newly sequenced Psychrobacter sp. 230 isolate identified as a 3-oxo-C12-HSL producer in this study. COG functional categories are presented on the outer ring, while forward and reverse strand gene annotations are presented in the inner rings in red and blue, respectively.



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

Copyright © 2019 Reen, Gutiérrez-Barranquero, McCarthy, Woods, Scarciglia, Adams, Fog Nielsen, Gram and O'Gara. 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.

# Conserved Pheromone Production, Response and Degradation by Streptococcus mutans

Antonio Pedro Ricomini Filho<sup>1</sup>† , Rabia Khan<sup>2</sup>† , Heidi Aarø Åmdal<sup>2</sup> and Fernanda C. Petersen<sup>2</sup> \*

<sup>1</sup> Department of Physiological Science, Piracicaba Dental School, University of Campinas, Piracicaba, Brazil, <sup>2</sup> Department of Oral Biology, Faculty of Dentistry, University of Oslo, Oslo, Norway

#### Edited by:

Cristina García-Aljaro, University of Barcelona, Spain

#### Reviewed by:

Indranil Biswas, The University of Kansas, United States Jacqueline Abranches, University of Florida, United States Stephen J. Hagen, University of Florida, United States

\*Correspondence:

Fernanda C. Petersen f.c.petersen@odont.uio.no

†These authors have contributed equally to this work

#### Specialty section:

This article was submitted to Microbial Physiology and Metabolism, a section of the journal Frontiers in Microbiology

> Received: 14 May 2019 Accepted: 30 August 2019 Published: 13 September 2019

#### Citation:

Ricomini Filho AP, Khan R, Åmdal HA and Petersen FC (2019) Conserved Pheromone Production, Response and Degradation by Streptococcus mutans. Front. Microbiol. 10:2140. doi: 10.3389/fmicb.2019.02140 Streptococcus mutans, a bacterium with high cariogenic potential, coordinates competence for natural transformation and bacteriocin production via the XIP and CSP pheromones. CSP is effective in inducing bacteriocin responses but not competence in chemically defined media (CDM). This is in contrast to XIP, which is a strong inducer of competence in CDM but can also stimulate bacteriocin genes as a late response. Interconnections between the pathways activated by the two pheromones have been characterized in certain detail in S. mutans UA159, but it is mostly unknown whether such findings are representative for the species. In this study, we used bioassays based on luciferase reporters for the bacteriocin gene cipB and the alternative sigma factor sigX to investigate various S. mutans isolates for production and response to CSP and XIP pheromones in CDM. Similar to S. mutans UA159, endogenous CSP was undetectable in the culture supernatants of all tested strains. During optimization of the bioassay using the cipB reporter, we discovered that the activity of exogenous CSP used as a standard was reduced over time during S. mutans growth. Using a FRET-CSP reporter peptide, we found that S. mutans UA159 was able to degrade CSP, and that such activity was not significantly different in isogenic mutants with deletion of the protease gene htrA or the competence genes sigX, oppD, and comR. CSP cleavage was also detected in all the wild type strains, indicating that this is a conserved feature in S. mutans. For the XIP pheromone, endogenous production was observed in the supernatants of all 34 tested strains at peak concentrations in culture supernatants that varied between 200 and 26000 nM. Transformation in the presence of exogenous XIP was detected in all but one of the isolates. The efficiency of transformation varied, however, among the different strains, and for those with the highest transformation rates, endogenous XIP peak concentrations in the supernatants were above 2000 nM XIP. We conclude that XIP production and inducing effect on transformation, as well as the ability to degrade CSP, are conserved functions among different S. mutans isolates. Understanding the functionality and conservation of pheromone systems in S. mutans may lead to novel strategies to prevent or treat unbalances in oral microbiomes that may favor diseases.

Keywords: pheromone, streptococcus, competence, natural transformation, quorum-sensing, CSP, XIP, ComS

### INTRODUCTION

fmicb-10-02140 September 13, 2019 Time: 16:10 # 2

Natural genetic transformation is widely distributed in bacteria. In streptococci it occurs during a genetically programmed differentiated state called competence. During this state the bacteria become capable of taking up DNA from the environment and incorporate it into their genomes. The capacity for natural transformation has been reported for more than 80 bacterial species (Johnsborg et al., 2007; Johnston et al., 2014). Among streptococci, competence for natural transformation includes most species of the mitis, salivarius, bovis, anginosus, and mutans groups (Johnsborg et al., 2007; Fontaine et al., 2010; Desai et al., 2012; Morrison et al., 2013). The core of the machinery necessary for streptococcal transformation relies on the transcription of the conserved alternative sigma factor SigX, also known as ComX. SigX orchestrates a core response in streptococcal species characterized by the induction of 27 to 30 genes (Khan et al., 2016). The functions of the core genes are predominantly related to transformation, most of them coding for competence effector proteins for DNA binding, uptake and recombination (Li et al., 2001; Mashburn-Warren et al., 2010; Khan et al., 2016).

Streptococcus mutans is a member of the mutans group and part of the human oral microbiota. Environmental factors disturbing the ecological balance in the oral cavity, such as increased sugar intake, may favor S. mutans growth. The S. mutans acidogenic and aciduric properties may then contribute to tooth demineralization and dental caries (Marsh et al., 2011). Natural transformation was first reported in S. mutans in 1981 (Perry and Kuramitsu, 1981). This and later studies showed that transformation in S. mutans, at least in the laboratory setting, is restricted to a limited range of strains and depends on environmental conditions not yet fully understood (Perry and Kuramitsu, 1981). Studies investigating the S. mutans pan-genome have recently revealed the ubiquity of sigX and competence effector genes in S. mutans (Cornejo et al., 2013; Palmer et al., 2013; Song et al., 2013), indicating that competence may be a conserved feature in S. mutans. Moreover, the extensive horizontal gene transfer observed in the genomes of S. mutans clinical isolates (Cornejo et al., 2013) indicates that transformation occurs in their natural habitat and may be a widespread feature in the species.

In S. mutans the competent state is triggered by two linear peptides, the CSP (competence stimulating peptide) (Li et al., 2001) and the XIP (sigX-inducing peptide) (**Figure 1**; Mashburn-Warren et al., 2010). CSP was the first described pheromone and, until recently, the only one known to activate the competence system for genetic transformation in S. mutans. The comC gene, which encodes CSP, is downstream of the comDE operon, encoding the ComD histidine kinase and the ComE response regulator. At least in the synthetic form, CSP is thought to bind to the ComD histidine kinase of the ComDE twocomponent system, leading to phosphorylation of the ComE response regulator (Peterson et al., 2004). Phosphorylated ComE directly up-regulates the transcription of clusters of bacteriocinrelated genes (Petersen et al., 2006; Mashburn-Warren et al., 2010; Federle and Morrison, 2012; Khan et al., 2016). These include among others the SMU\_1914 gene (also known as nlmC and cipB), encoding mutacin V, and the putative immunity protein SMU\_1913, found in a comE downstream region. Mutacin V has been proposed to link the early CSP bacteriocininducing response to the competence response (Ween et al., 1999; van der Ploeg, 2005; Kreth et al., 2007; Dufour et al., 2011; Hung et al., 2011) by mechanisms that remain elusive. Essential for competence development is the activation of the XIP pheromone encoding gene, comS, followed by the upregulation of the alternative sigma factor SigX, the master core regulator of competence. In rich media, ComS seems to be processed to XIP inside the cells. Intracellular XIP and/or ComS then binds to ComR to activate the competence pathway (Underhill et al., 2018). The predicted CSP is in general conserved among S. mutans strains (Li et al., 2001; Petersen et al., 2006), and there are indications that S. mutans may form a single CSP pherotype (Petersen et al., 2006; Hossain and Biswas, 2012). The C-terminal of the CSP precursor (ComC) is exported outside of the cells, where it is further processed into the active 18 amino acid peptide (18-CSP) (Petersen et al., 2006; Hossain and Biswas, 2012).

The XIP pheromone system is found in several streptococci (Gardan et al., 2009; Fontaine et al., 2010; Mashburn-Warren et al., 2010; Fleuchot et al., 2011; Morrison et al., 2013). In S. mutans UA159 the XIP pheromone encoded by comS has been identified in culture supernatants grown in chemically defined medium (CDM) lacking peptides (Mashburn-Warren et al., 2010; Khan et al., 2012). XIP is produced as a propeptide (ComS) that is possibly exported and processed into the active XIP (N-GLDWWSL) (Mashburn-Warren et al., 2010; Desai et al., 2012; Khan et al., 2012). The presence of an exporter has, however, never been demonstrated, and recent studies indicate that release of XIP or ComS by autolysis is sufficient to promote intercellular communication in CDM (Kaspar et al., 2017). Under such conditions, response to the XIP depends on the Opp oligopeptide permease system, indicating that the peptide is internalized (Mashburn-Warren et al., 2010). Once inside the cells XIP is thought to bind to the Rgg-like regulator ComR to activate transcription of sigX and comS. In contrast with XIP, CSP shows no activity or low potency in triggering competence in CDM (Desai et al., 2012; Wenderska et al., 2012; Reck et al., 2015) but can still induce the expression of bacteriocin-related genes of the ComE regulon (Reck et al., 2015). When it comes to transformation, the use of synthetic CSP has led to higher levels of S. mutans transformation (Gaustad and Morrison, 1998; Petersen and Scheie, 2010) but has not extended the range of strains transformed in the absence of the synthetic pheromone. As for the synthetic XIP, evidence for competence induction and endogenous pheromone production has so far been restricted to the reference strain UA159, with few exceptions (Palmer et al., 2013).

In this study we investigated the functional conservation of the S. mutans CSP and XIP pheromone signaling systems in CDM. Extracellular CSP activity was not detected in any of the tested strains. Exposure to synthetic CSP revealed that S. mutans can indeed degrade CSP, a behavior that was conserved in the examined strains, and that at least in strain UA159 it was not abolished by deletion of sigX, comR, oppD, or the htrA protease gene. For the XIP pheromone, endogenous production

was found in all tested strains, and transformation was found in all but one of the isolates grown in CDM. The results thus indicate the presence of a single S. mutans XIP pherogroup and suggest that S. mutans has a conserved ability to suppress CSP activity in CDM.

### MATERIALS AND METHODS

### Bacterial Strains and Media

The Streptococcus mutans strains used in this study and their relevant characteristics are listed in **Table 1**. The strains used in the study were selected from the collection of strains in our laboratory, representing strains known to be transformed (UA159, V403, OMZ175, LML-2, LML-4, GS5, NG8, BM71, LT11) and others in which genetic competence has not yet been characterized. Todd-Hewitt broth (THB; Becton Dickinson) was used to grow all the strains used for DNA extraction and DNA sequencing. Chemically Defined Medium (CDM) (Mashburn-Warren et al., 2010) was used to perform the genetic transformation assays and bioassay to measure the activity of exogenously added CSP in the supernatants of S. mutans and also to verify and quantify the concentration of extracellular native XIP in culture supernatants (Desai et al., 2012; Khan et al., 2012). The carbohydrate source was glucose at 1% final concentration, unless specified. The antibiotics erythromycin, kanamycin and spectinomycin were used at final concentrations of 10, 500, and 500 µg mL−<sup>1</sup> , respectively. THB agar and THB agar supplemented with erythromycin were used to enumerate the total and transformed number of S. mutans.

### Construction of Mutants

The deletion mutants were constructed using the PCR-ligation mutagenesis strategy (Lau et al., 2002). Sequence information was obtained from the S. mutans UA159 genome. Mutants with deletion of sigX and htrA were constructed. AscI or FseI sites were incorporated into the 5<sup>0</sup> -ends of the oligonucleotide primers and both ends of the resistance cassette. The kanamycin resistance cassette was amplified using primer pair FP001-FP068. The comX flanking regions were amplified with the primers pairs

#### TABLE 1 | Strains and plasmid used in this study.

fmicb-10-02140 September 13, 2019 Time: 16:10 # 4


Erm, erythromycin; Spc, spectinomycin; Kan, kanamycin. <sup>a</sup>Genome sequenced strains. <sup>T</sup> , type strain.

FP462 (CTTGGTAGCAGGAGAGCAC), FP463 (AAAG CACAGCCTGCTTCAAT) and FP714 (TGCCGAACA CAGCAGTTAAG), FP715 (CATTCCCTCTTGTTGCCAAT). The htrA flanking regions were amplified with the primers pairs FP807 (TCCCTCCAATAACGAAGGTCA), FP808 (GGTAAGT GTTGA TATGACCCCT) and FP809 (GAAGGTAGCGTCTA TCAGCGA), FP810 (GCAGTCGAGGTTGATAGGGA). The resultant amplicons were digested with AscI or FseI whereas the kanamycin resistance cassette was digested with both enzymes. The upstream and downstream amplicons of target genes were ligated to the kanamycin cassette using T4 DNA ligase. The two ligated products were mixed and PCR amplified with distal primers. The resultant amplicons were used to transform S. mutans. Gene deletion was confirmed by PCR amplification and gel electrophoresis.

### DNA Sequence Analysis

The chromosomal DNA of S. mutans strains were isolated as previously described (Petersen and Scheie, 2000). The primers FP678 (5<sup>0</sup> -ATGCGGAAGCTAAAAAGAGC-3<sup>0</sup> ) and FP679 (5<sup>0</sup> - TCCAGTCTTCCTATCTGAGCAA-3<sup>0</sup> ) were used to amplify a region of 431 bp, which contains the tRNA transcriptional terminator of comR (9 bp stem, a 4 nt loop and a T-rich region at the 3<sup>0</sup> side of the stem-loop), the comS promoter and the comS sequence. First, a PCR was performed (20 µL reaction volume containing 2.5 mM MgCl2, 2 pmol of each primer, 0.2 mM of each dNTP, 2 µL of 10× buffer, 50– 100 ng of genomic DNA and 2 units of TaqDNA polymerase using a thermocycling profile of initial denaturing period of 3 min at 94◦C; followed by 30 cycles of 30 s at 94◦C, 30 s at 55◦C and 2.5 min at 72◦C; with a final extension period of 3 min at 72◦C and the products were visualized on a 1% agarose gel to check the amplified fragments by the primers (Petersen and Scheie, 2000). The sequencing was performed bi-directionally in a single experiment for all the samples using BigDye <sup>R</sup> Terminator v1.1 Cycle Sequencing Kit (Applied Biosystems, Foster City, CA, United States) on an ABI 3730 DNA analyzer (Applied Biosystems, Foster City, CA, United States). All the sequence analysis was done using the Sequencher 5.0 software (Gene Codes, Ann Arbor, MI, United States). The electropherograms were checked and comparative analysis of the bidirectional sequences was performed using the S. mutans UA159 sequence as a reference.

### Bioassay of CSP and XIP Activity in Culture Supernatants

The CSP bioassay was performed to measure the production of native CSP and activity of exogenously added CSP in the supernatants of S. mutans during growth, and the XIP bioassay was to verify and quantify the concentration of extracellular native XIP in culture supernatants. A PcipB-luc reporter (SM059) was used in the CSP bioassay as previously described (Khan et al., 2012), and a PsigX-luc reporter in a comS deletion background (SM091) was used to perform the XIP bioassay as described by Desai et al. (2012), with slight modifications (Khan et al., 2012).

To estimate CSP production and response, the supernatants were collected by 10× dilution of precultures at OD<sup>600</sup> of 0.6 in CDM containing 10% bovine serum albumin (BSA). When added, 250 nM CSP concentration was used. Supernatants were collected for 6 h during growth. For XIP production and response, overnight cultures of S. mutans strains were diluted in fresh CDM to an optical density at 600 nm (OD600) of 0.05 and incubated in air at 37◦C. At each hour, from the second until the 10 h of bacterial growth, the OD<sup>600</sup> was measured and 1 mL of the culture was centrifuged (10,000 g for 10 min at 4◦C) to collect the supernatant. The indicator strains SM059 and SM091 were grown in CDM to an OD<sup>600</sup> of 0.05 and stored in microcentrifuge tubes with 10% of glycerol at −80◦C. The entire assay was performed with stock cultures from the same batch. The bioassay was performed adding 10 µL of each culture supernatant to 50 µL of the indicator strain, 40 µL fresh CDM and 20 µL of 1.0 mM D-luciferin (Synchem, Felsberg-Altenburg, Germany) in a flat-bottom 96-well plate (Nunc, Rochester, NY, United States). Luminescence was measured by reading the plates in a multidetection microplate reader (Synergy HT; BioTek Instruments, Winooski, VT, United States). To quantify the XIP concentrations in the culture supernatants, standard curves were performed at the same time, replacing the culture supernatant with 10 µL of CDM-standard solutions with different XIP concentrations. CDM without synthetic peptides was used to obtain background values that were subtracted from the sample values. For strains exhibiting no XIP activity in the supernatants, the experiment was repeated at least twice. The software SigmaPlot (version 12.0, Systat Software, Inc.) was used to calculate the XIP concentrations using the relative light unit (RLU) values.

### Genetic Transformation

Overnight cultures of all S. mutans strains were diluted 1:10 in fresh CDM and incubated in 5% CO<sup>2</sup> at 37◦C. The OD<sup>600</sup> was followed until the cultures reached an absorbance value of 0.6, then 10% of glycerol was added and they were stored at −80◦C. To perform the assay the stored cultures were diluted in fresh CDM 1:10 to an OD<sup>600</sup> of 0.05 and incubated in air for 2 h at 37◦C. The OD was measured and aliquots of 150 µL of the bacterial suspensions were transferred to 1.5 mL microcentrifuge tubes. The plasmid pVA838 (**Table 1**) was added to all tubes at a final concentration of 1 µg mL−<sup>1</sup> and XIP was added in the experimental group at a final concentration of 1,000 nM. The samples were then incubated in air for an additional 4 h at 37◦C, at which time the OD was measured again and the bacterial suspensions were serially diluted (up to 1 to 10<sup>6</sup> ) in PBS. The suspensions were plated in duplicate on THB agar and THB agar supplemented with erythromycin 10 µg mL−<sup>1</sup> . The plates were incubated in 5% CO<sup>2</sup> for 48 h at 37◦C before counting the colony-forming units (CFU). The number of genetically transformed cells was divided per total CFU to obtain the values for transformation frequency. Three independent experiments were performed to evaluate transformation in CDM.

In order to evaluate genetic transformation in the nonresponsive strains, an independent experiment extending the incubation time to 10 h before plating was performed. In addition, at 6 h an extra load of pVA838 was added, which increased the final concentration to 2 µg mL−<sup>1</sup> . For the remaining non-responsive strains, transformation was attempted using the 6.3 kb aRJ02 chromosomal PCR amplicon with a kanamycin marker, which is expected to have a higher sensitivity for detection of transformability (**Table 1**; Morrison et al., 2015; Salvadori et al., 2017).

### Synthetic Pheromones

The synthetic XIP (ComS11−17; GLDWWSL) (Mashburn-Warren et al., 2010; Khan et al., 2012) and the 18-CSP (ComC26−43; SGSLSTFFRLFNRSFTQA) (Petersen et al., 2006) were synthesized by GenScript (GenScript Biotech Corporation, Piscataway, NJ, United States), both with an estimated purity of 98% (Khan et al., 2012). The stock solutions of both peptides were stored in small aliquots at −20◦C. The final concentration used in the genetic transformation assays for XIP was 1,000 nM. For FRET-assays to investigate protease activity, the 18-CSP was synthesized using methoxycoumarin-acetic-acidyl (MCA) on the N-terminal and Lys-Dinitrophenyl on the C-terminal (Dnp) (MCA-SGSLSTFFRLFNRSFTQA-Dnp; GenScript). On cleavage of the peptide by potential proteases, the Dnp quencher separates from the MCA fluorophore. The fluorophore is then activated giving an increase in fluorescence.

### FRET-18CSP Peptide Cleavage Reporter Assays

### Growth in the Presence of FRET-18CSP

Pre-cultures of S. mutans were diluted in 2 mL CDM supplemented with 2% BSA to an OD<sup>600</sup> of 0.04 to 0.05, and incubated at 37◦C in air atmosphere to an OD<sup>600</sup> of 0.06. The cultures were then distributed into the wells of a 96-well plate (115 µL in each well), and 5 µL of FRET-18CSP peptide was added (4 µM final concentration). The plate was incubated at 37◦C in air, and fluorescence and optical density at 600 nm were measured at different time points during growth in a multidetection microplate reader (Synergy, Cytation 3; excitation 325 nm, emission 392 nm).

### Measurement of FRET-18CSP Cleavage in Culture Supernatants

Supernatants from S. mutans UA159 PcipB-luc reporter (SM059) and 1htrA (SM165) were collected by centrifugation (6000 g for 10 min at 4◦C) at different time points during growth, corresponding to early, mid- and late exponential phases. The supernatants were then distributed into the wells of a 96 well plate (115 µL in each well), and 5 µL of FRET-18CSP peptide was added (4 µM final concentration). The plate was incubated at 37◦C in air atmosphere for 1 h, and fluorescence was measured in a multi-detection microplate reader (Synergy, Cytation 3; excitation 325 nm, emission 392 nm). Supernatants without FRET-18CSP were used for detection of background fluorescence, which were then subtracted from the values in the tested samples exposed to FRET-18CSP.

## RESULTS

### Bioassay for CSP Detection in CDM

In this study we used a PcipBluc reporter for detection of extracellular CSP activity, in CDM as previously described (Khan et al., 2012), except that high sensitivity was only achieved in the presence of BSA (**Figure 2A**). BSA was chosen because it is known to enhance competence of S. mutans in rich media (Ahn et al., 2006), and in S. pneumoniae, BSA prevents CSP proteolysis (Cassone et al., 2012).

The potency of CSP in inducing the activity of the cipB promoter was increased by approximately 33-fold in CDM supplemented with BSA compared with CDM alone (**Figure 2A**). BSA concentrations from 0.5 to 10% gave similar results (**Figure 2A**).

### Failure of CSP to Induce sigX in CDM Was Independent of Carbohydrate Source

Since the induction of cipB was significantly enhanced in the presence of BSA, we decided to further examine the intriguing fact that CSP fails to stimulate sigX in CDM (Desai et al., 2012; Wenderska et al., 2012; Reck et al., 2015). Our hypothesis was that increased stimulation of the early response by CSP, as observed with BSA supplementation, would enable the activation of sigX, which is a late response in complex media. We used a similar bioassay as described above, but using the promoter of sigX linked to the luciferase gene instead. The results showed that CSP concentrations up to 1000 nM failed to induce the PsigXluc reporter in CDM supplemented with BSA and glucose (**Figure 2B**).

Since the carbohydrate source, at least in complex media, has a large influence on the induction of sigX expression by CSP (Moye et al., 2016), we also investigated sigX expression in CDM supplemented with fructose and galactose (**Figures 2C,D**). CSP still failed in demonstrating sigX induction. Thus, although CSP induces strong expression of cipB in CDM supplemented with BSA and different sources of carbohydrate, it fails to induce the late response characterized by the induction of sigX expression at concentrations as high as 1000 nM.

### Endogenous CSP Activity Was Not Detected in the Supernatants of S. mutans

Lack of endogenous CSP activity in the supernatants of S. mutans UA159 grown in CDM has been previously reported by our group (Khan et al., 2012). It is, however, unknown whether this is a conserved feature in the species. We investigated 6 other strains of S. mutans, and the comC deletion mutant SM004, for the CSP activity in their supernatants by growing them in CDM in the presence of BSA (**Figure 2E**). No cipB-inducing activity was detected in culture supernatants collected at early-, mid- or late- exponential phases of growth, thus indicating that the lack of extracellular CSP activity under such growth conditions may represent a conserved feature in S. mutans.

## Suppression of CSP Activity During Growth in CDM

Supernatants of S. mutans cultures exposed to 250 nM synthetic CSP were collected at different time points during growth in CDM. CSP activity in the supernatants of S. mutans UA159 was measured by using the cipB luciferase reporter described above. Within the first 3 h, CSP had almost completely disappeared from the culture supernatants (**Figure 3A**).

Due to the recognizable role of CSP in the induction of competence, it was crucial to determine if the loss of CSP activity was dependent on the development of competence. For this, we investigated the CSP activity in the culture supernatants of UA159 deletion mutants for sigX, oppD, and comR (**Table 1**). In all of them, a dramatic reduction in activity of exogenously added CSP was observed (**Figure 3B**).

To determine whether such an inhibitory effect on CSP activity could be due to proteolytic cleavage, we supplemented the CDM medium with a FRET-18CSP reporter peptide (**Figure 3C**). We found that all three isogenic mutants (sigX, oppD, and comR) grown under such conditions promoted increase in fluorescence activity, as measured after 240 min incubation at 37◦C (**Figure 3C**), thus suggesting that the ability of S. mutans to degrade CSP does not require expression of key elements of the competence regulon. We also examined the effect of inactivating the htrA gene coding for the S. mutans HtrA protease, which in S. pneumoniae inactivates the CSP and cleaves misfolded proteins. No reduction in CSP degradation was observed for the htrA mutant (**Figure 3C**). CSP activity was indeed slightly higher for the culture supernatants of the htrA mutant at mid- and late exponential phases of growth (**Figure 3D**).

Finally, we tested whether suppression of CSP activity is extended to other S. mutans strains. In all the strains tested, reduction in CSP activity was observed, though at different levels (**Figure 3E**). When grown in the presence of the FRET-18CSP reporter peptide, an increase in fluorescence was observed for all strains (**Figure 3F**). We conclude that S. mutans suppresses CSP activity during growth in CDM, and that the mechanisms involved are active in the absence of htrA, sigX, comR or oppD, indicating the involvement of competence-regulated independent factors. The potential proteolytic activity leading to CSP inactivation seems to be conserved in S. mutans.

### XIP Activity Is Detected in the Culture Supernatants of Practically all S. mutans Strains Grown in CDM

S. mutans UA159 produces and responds to exogenous XIP in CDM (Mashburn-Warren et al., 2010; Desai et al., 2012). We investigated whether XIP production is conserved in S. mutans. We used the PsigXluc 1comS indicator (SM091) to estimate the XIP concentration in the supernatants of the different strains during growth. The minimum detection level was set at 10 nM, as determined by standard curve measurements using synthetic XIP. Our results showed the presence of extracellular XIP activity in all strains tested. The highest XIP concentrations in the supernatants of growing cultures ranged from approximately 200 to 26,000 nM (**Figure 4A**), with most showing XIP accumulation

was used as a positive control.

(A) S. mutans UA159 growth curve was measured at OD<sup>600</sup> (black line) and indicates the time points when supernatants were collected. CSP activity in CDM alone (positive control) is represented by the red dashed line and in the supernatants of S. mutans UA159 by the red solid line. (B) Loss of CSP activity in the supernatants of UA159, 1sigX, 1oppD and 1comR. (C,F) CSP cleavage measured at 240 min growth in the presence of the FRET-18CSP peptide, and recorded as relative fluorescence units (RFU) for panel (C) UA159, 1htrA, 1sigX, 1oppD, and 1comR, and (F) UA159, GW2, NG, LML-4, OMZ175, 357, and UA130. In panel (C), right upper corner, a standard curve showing degradation of FRET-18CSP by trypsin is included as a reference. (D) FRET-18CSP proteolytic activity in supernatants of UA159 and the 1htrA mutant collected at early (EP), mid- (MP) and late (LP) exponential phase of growth. RFU background values of the corresponding strains without FRET-CSP were subtracted. Error bars show standard error of mean from two to three independent experiments, with three parallels each.

strains shown in panel (A), grouped according to transformation frequency.

starting at the mid-exponential growth phase and reaching maximal values at early stationary phase (data not shown). We conclude that the growth conditions that support extracellular XIP pheromone accumulation by UA159 may also sustain XIP production by a majority of the strains.

### Thirty-Three Out of Thirty-Four Strains Were Naturally Transformable

Natural transformation without the addition of synthetic pheromone was detected in 23 of the 34 strains (data not shown). Because optimal levels of XIP may not be present at the time of competence under in vitro conditions, it was of interest to learn whether addition of synthetic XIP would result in higher transformation levels. Our results showed that synthetic XIP extended the range of transformed strains. Three of the strains with detectable XIP levels in the supernatants, including OMZ175, UAB90, and KM1 were transformed only upon addition of synthetic XIP. In the strains that were not transformed with the initial protocol, we increased the time during which they grew in the presence of synthetic XIP and donor DNA to 10 h, but even then no transformants were obtained (Murchison et al., 1986). We next used a more sensitive protocol based on the use of PCR large fragments as donor DNA (Morrison et al., 2015; Salvadori et al., 2017). With addition of the synthetic pheromone, 33 out of 34 strains were naturally transformable. Only strain At10 was not transformed, thus indicating that a majority of S. mutans are amenable to transformation. Overall, strains that exhibited lower transformation frequencies showed lower levels of XIP activity in their culture supernatants (**Figure 4B**).

### The comS Gene Was Conserved in the 34 S. mutans Strains

The S. mutans comS gene encoding the XIP precursor is essential for competence development. Given the variability in transformation efficiency observed above, and that some strains were not amenable to transformation, we investigated whether the comS gene and its putative promoter sequences were conserved in the S. mutans strains included in the study. The results showed that comS was identical in all 34 strains (**Figure 5**). Highly conserved sequences were also found in the comS putative promoter. In this region the only difference in relation to UA159 was the presence or absence of an additional adenine between the −10 element and the comS translation initiation site. The additional adenine was present in 22 out of the 34 strains evaluated. In all the strains, comS was located 57– 58 nt downstream of the tRNA transcriptional terminator of comR (SMU\_61) and the putative −10 element was 32–33 nt upstream of the comS translation initiation site. While this study was being performed, new genomic sequences for 78 more strains of S. mutans were made available on public databases (Maruyama et al., 2009). In order to compare our results with the new genomic sequences we ran BLAST using the sequence shown in **Figure 5** against the CoGe database<sup>1</sup> (Lyons et al., 2011). The comS sequences in these strains were identical to the S. mutans UA159. In the comS promoter the sequences varied in only a single base pair at the same position as that observed in 22 of the strains sequenced in our study. Taken together the results indicate that the comS gene and promoter regions are highly conserved in the S. mutans strains analyzed, suggesting that variations in transformation levels and XIP production are not correlated with differences in this region.

### DISCUSSION

In complex media, the CSP pheromone triggers the expression of bacteriocin-related genes, followed by a late response characterized by increased expression of the XIP-encoding gene comS and competence development (Kreth et al., 2005; Lemme et al., 2011; Khan et al., 2016). The mechanisms leading to activation of comS expression and competence remain unknown. It is also still unknown why CSP fails in stimulating competence in peptide-free defined medium. It has been suggested that S. mutans may perhaps produce a protease that, similar to the HtrA protease in S. pneumoniae, could inactivate the CSP (Desai et al., 2012). This possible explanation did not seem to match with the later finding that CSP is actually active in CDM, in that it can stimulate cipB and all other genes regulated by ComED (Reck et al., 2015). Our results show that S. mutans can indeed inactivate CSP by a mechanism that most probably involves proteolysis, given the results obtained with the FRET-18CSP reporter peptide. We confirmed previous results that CSP induces cipB (Reck et al., 2015) and found that a CSP concentration as low as 1 nM in the presence of BSA was sufficient to activate the early bacteriocin response. This is opposite to the CSP concentration to activate sigX, which

<sup>1</sup>http://genomevolution.org/CoGe/CoGeBlast.pl

requires values that exceed the maximum concentration used in this study (1000 nM) (data not shown). A higher threshold for sigX activation by CSP has also been observed in complex media (Son et al., 2012). A question that remains is why the threshold of CSP concentration to activate sigX, and therefore competence, is higher than for cipB activation, when CipB is thought to be the main link of CSP to sigX activation. One possibility is that at high concentrations CSP may activate or repress other two-component systems than ComED, which would then trigger a different link to competence. It is a well known phenomenon that although pheromones are usually highly specific to their cognate receptors, at high concentrations they may bind to other non-specific receptors (Hawver et al., 2016). However, several other mechanisms are possible, since the competence system is part of a signaling network of high complexity, affected by a variety of other systems not directly regulated by ComED, as recently reviewed (Kaspar and Walker, 2019). Irrespective of the mechanism, the higher threshold for CSP to induce competence indicates that CSP degradation may affect the competence response to this pheromone.

Inactivation of the S. mutans CSP by proteases produced by other streptococcal species has been known for more than 10 years (Wang and Kuramitsu, 2005). However, this is the first indication that S. mutans can also inhibit the activity of its own CSP, most probably via proteolytic activity. The genomes of S. mutans have more than 65 known or putative proteases, according to the MEROPS database<sup>2</sup> , but only HtrA has shown a role in autologous CSP cleavage. However, unlike for S. pneumoniae (Cassone et al., 2012), the HtrA protease in S. mutans was not required for CSP inactivation.

In S. pneumoniae, luciferase and gfp reporter strains for CSP activity have been successfully used to identify pneumococcal CSP pheromones in culture supernatants (Moreno-Gamez et al., 2017). Also in the seminal study that identified the S. pneumoniae competence factor, the CSP was isolated and purified from the supernatant (Havarstein et al., 1995). In S. mutans UA159 grown in complex medium, endogenous CSP has been identified by mass spectrometry in the supernatant fraction (Hossain and Biswas, 2012). Based on these findings, we hypothesized that if S. mutans produces CSP in CDM, we would probably be able to detect it in its culture supernatant. However, despite the high sensitivity of the protocol used in the present study, we could not detect any CSP-inducing activity in the supernatants of the different strains tested. Thus, if CSP is produced, it is either rapidly degraded outside of the cells or it remains associated with the cells.

In contrast to CSP, endogenous XIP activity was present in the supernatants of most strains. Moreover, activation of the XIP system in CDM enabled transformation of all but one of the 34 strains tested. The comS gene encoding the pre-processed form of XIP was indeed found in all the strains, which is in line with the results of a recent study reporting comS as part of the S. mutans core genome (Cornejo et al., 2013). In general, higher levels of XIP detection in the supernatants correlated with increased transformability. Of note, some strains that produced high concentrations of endogenous XIP needed to be stimulated by synthetic XIP to transform. This was, however, not surprising, given the fact that maximum concentrations were in most cases observed only close to the stationary phase, which is a period during which competence under laboratory conditions is already shut off. For some of the strains that produced low levels of XIP, transformation was only detected by using a large PCR amplicon as DNA donor, which is known to result in increased recovery of transformants (Morrison et al., 2015; Junges et al., 2017).

Taken together, our results indicate that the XIP system is conserved in S. mutans, and that all S. mutans strains may belong to a single pherotype. Moreover, the conserved ability of S. mutans to cleave CSP under conditions that favor accumulation of XIP may provide an adaptive advantage to S. mutans, by allowing them to fine-tune competence and bacteriocin responses in response to environmental changes. While the knowledge on pheromone signaling for any species is mostly based on models derived from limited selected strains, understanding how the system works in different strains of a species is of high relevance for the development of strategies aiming at interfering with their communication systems.

### DATA AVAILABILITY

All datasets generated for this study are included in the manuscript and or supplementary files.

### AUTHOR CONTRIBUTIONS

All authors conceptualized the manuscript, drafted and critically revised the manuscript and approved the final version of the manuscript for publication.

### FUNDING

This work was supported in part by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - Brasil (CAPES) - Finance Code 001 (AR postdoctoral grant), and by the program International Partnership for Outstanding Education, Research, and Innovation (INTPART), RCN grant number 274867.

### ACKNOWLEDGMENTS

We thank Andreas Podbielski for the kind gift of pFW5-luc and Roy Russel for providing the S. mutans strains LML-2, LML-4, LML-5, and At10. We are grateful to Anne Karin Kristoffersen for helpful assistance in DNA sequence analysis.

<sup>2</sup> https://www.ebi.ac.uk/merops/index.shtml

## REFERENCES

fmicb-10-02140 September 13, 2019 Time: 16:10 # 12



pheromones. Methods Mol. Biol. 1537, 219–232. doi: 10.1007/978-1-4939- 6685-1\_13


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

Copyright © 2019 Ricomini Filho, Khan, Åmdal and Petersen. 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.

# Nitric Oxide Enters Quorum Sensing via the H-NOX Signaling Pathway in Vibrio parahaemolyticus

#### Takahiro Ueno<sup>1</sup> , Jonathan T. Fischer<sup>1</sup> and Elizabeth M. Boon1,2 \*

<sup>1</sup> Department of Chemistry, Stony Brook University, Stony Brook, NY, United States, <sup>2</sup> Institute of Chemical Biology & Drug Discovery, Stony Brook University, Stony Brook, NY, United States

Nitric oxide (NO) plays a major role in the regulation of mammalian biological functions. In recent years, NO has also been implicated in bacterial life cycles, including in the regulation of biofilm formation, and the metabolism of the bacterial second messenger signaling molecule cyclic-di-GMP. In a previous study, we reported the discovery of an NO-responsive quorum sensing (QS) circuit in Vibrio harveyi. Here, we characterize the homologous QS pathway in Vibrio parahaemolyticus. Spectroscopic analysis shows V. parahaemolyticus H-NOX is an NO sensory protein that binds NO in 5/6 coordinated mixed manner. Further, we demonstrate that through ligation to H-NOX, NO inhibits the autophosphorylation activity of an H-NOX-associated histidine kinase (HqsK; H-NOX-associated quorum sensing kinase) that transfers phosphate to the Hpt (histidine-containing phosphotransfer protein) protein LuxU. Indeed, among the three Hpt proteins encoded by V. parahaemolyticus, HqsK transfers phosphate only to the QS-associated phosphotransfer protein LuxU. Finally, we show that NO promotes transcription of the master quorum sensing regulatory gene opaR at low cell density.

Keywords: nitric oxide, quorum sensing, H-NOX, histidine kinase, Vibrio

### INTRODUCTION

Quorum sensing (QS) is a cell-to-cell communication system utilized by bacteria to assess their population density and to coordinate population-wide changes in gene expression. Bacteria produce, secrete, and detect small signaling molecules called autoinducers (AI). Many types of autoinducers have been identified, some that are unique to a particular species, and some that are shared by multiple bacterial species (Eberhard et al., 1981; Engebrecht and Silverman, 1984; Henke and Bassler, 2004; Miller et al., 2004). Detection of AIs by a receptor protein in a QS pathway ultimately leads to changes in gene expression (Miller and Bassler, 2001). A classic example of a QS system is the well-studied LuxI-LuxR system found in several Vibrio species as well as other gramnegative bacteria (Miller and Bassler, 2001). In this system, LuxI synthesizes a homoserine lactone AI. This AI binds to the transcriptional regulator LuxR, which regulates the transcription of the luxICDABE operon. In addition to the LuxI-LuxR QS system, alternative QS circuits have also been characterized. Many QS systems in Vibrio species consist of an AI synthase, a corresponding AIsensing histidine kinase (HK), and a cytoplasmic response regulator (RR). The QS-associated HKs in many bacteria are hybrid histidine kinases containing both kinase and receiver domains. These kinases typically have both kinase and phosphatase activities, allowing them to both phosphorylate

#### Edited by:

Cristina García-Aljaro, University of Barcelona, Spain

#### Reviewed by:

Karen L. Visick, Loyola University Chicago, United States Brett Mellbye, Oregon State University, United States

\*Correspondence: Elizabeth M. Boon elizabeth.boon@stonybrook.edu

### Specialty section:

This article was submitted to Microbial Physiology and Metabolism, a section of the journal Frontiers in Microbiology

> Received: 13 June 2019 Accepted: 27 August 2019 Published: 18 September 2019

#### Citation:

Ueno T, Fischer JT and Boon EM (2019) Nitric Oxide Enters Quorum Sensing via the H-NOX Signaling Pathway in Vibrio parahaemolyticus. Front. Microbiol. 10:2108. doi: 10.3389/fmicb.2019.02108

and remove phosphate from the cognate response regulator or, as an intermediary, the histidine-containing phosphotransfer protein (Hpt) (Aiba et al., 1989; Makino et al., 1989; Tanaka et al., 1991). Frequently, the response regulators are transcription factors whose activity is modulated by phosphorylation (Hoch, 2000; Galperin, 2006; Gao et al., 2007).

Quorum sensing controls gene expression patterns as a function of population density. At low cell density, AI concentrations are also low and therefore the HKs are not complexed with the AI. Under these conditions, kinase activity predominates, and phosphate is transferred to the RR. The phosphorylated RR then facilitates changes in gene expression patterns, resulting in an adaptive cellular response. At high cell density, elevated AI concentrations drives HK binding to the AI, which switches the function of the kinase to act predominantly as a phosphatase. This ultimately leads to the removal of phosphate from the RR and causes corresponding changes in gene expression (Waters and Bassler, 2005; Bassler and Losick, 2006; Boyer and Wisniewski-Dyé, 2009).

Vibrio parahaemolyticus is a marine bacterium widely distributed in sea water around the world (McCarter, 1999; FAO/WHO, 2011), and is the leading cause of seafoodborne illnesses (Yeung and Boor, 2004; McLaughlin et al., 2005). V. parahaemolyticus produces a number of virulence factors, including the type III secretion system 1 (T3SS1), that are regulated by QS (Gode-Potratz and McCarter, 2011). V. parahaemolyticus shares a QS architecture with Vibrio harveyi, which is composed of three AIs (AI-2, HAI-1, and CAI-1), their synthases, and cognate membrane-bound sensory histidine kinases (LuxM/N, LuxS/PQ, and cqsA/cqsS, respectively) (Henke and Bassler, 2004; Ng and Bassler, 2009). All three AI-sensing kinases engage in phosphotransfer with LuxU, a histidinecontaining phosphotransfer protein (Hpt). LuxU transfers phosphate to the response regulator LuxO, a transcription factor that regulates the transcription of quorum regulatory RNAs (qrrs) (Ng and Bassler, 2009). Along with RNA-binding protein Hfq, these qrrs regulate the translation of two master QS regulatory proteins, AphA, and OpaR. AphA and OpaR are both transcription factors that regulate the expression of various genes, including many involved in motility, surface sensing, biofilm formation, and virulence (Gode-Potratz and McCarter, 2011; Zhang et al., 2012; van Kessel et al., 2013; Kalburge et al., 2017).

A recent study conducted in our laboratory identified a fourth arm in the V. harveyi QS circuit. In this pathway, the gaseous signaling molecule nitric oxide (NO) is integrated into QS via ligation to the hemoprotein H-NOX and its partner, an H-NOXassociated hybrid HK called HqsK (H-NOX-associated quorum sensing kinase) (Henares et al., 2012). Membrane-permeable NO is detected by H-NOX in the cytoplasm. The ligation state of H-NOX regulates the activity of HqsK. When NO is not present, HqsK functions as a kinase, transferring phosphate to LuxO via LuxU. Upon NO binding to H-NOX, HqsK switches its activity from a kinase to a phosphatase, causing a reverse in phosphate flow and contributing to dephosphorylation of LuxO (Henares et al., 2012). In this work, we further characterize the NO-responsive QS circuit in the pathogenic marine bacterium V. parahaemolyticus.

### MATERIALS AND METHODS

Unless otherwise noted, all the reagents were purchased in their highest available qualities and used as received.

### Bacterial Strains and Culture Method

Escherichia coli strain DH5α was used for cloning and E. coli BL21(DE3)pLysS was used for protein expression. DH5α was cultured in LB medium supplemented with 100 µg/mL ampicillin. BL21 (DE3) pLysS was cultured in 2XYT media (16 g/L Tryptone, 10 g/L yeast extract, and 5 g/L NaCl) supplemented with 100 µg/mL ampicillin and 34 µg/mL chloramphenicol. Both cultures were grown at 37◦C with 250 rpm agitation. At A600nm between 0.6 and 0.9, protein expression was induced with 100 µM isopropyl-β-Dthiogalactopyranoside (IPTG) for 15 h at 16◦C, then cells were harvested. V. parahaemolyticus strain EB101 (ATCC17802) was purchased from American Type Culture Collection and was grown by following the supplier's culture method. In V. parahaemolyticus growth curve and qPCR experiment, an overnight culture of V. parahaemolyticus was diluted 1:500 into fresh media (BD234000 Nutrient broth with 3% NaCl), supplemented with 100 µg/mL ampicillin with various DETA NONO concentrations (Cayman Chemical). Cultures were grown at 37◦C with agitation at 250 rpm. Bacterial growth was monitored by measuring A600nm and harvested at designated ODs.

### Molecular Cloning and Mutagenesis

The genome of V. parahaemolyticus strain EB101 has not been sequenced. Instead, we referred to the genome sequence of V. parahaemolyticus strain RIMD 2210633 for gene annotations and primer designing. V. parahaemolyticus genomic DNA was extracted using Zymo Research Quick-gDNA MiniPrep (D3006), by following the manufacturer's instructions. Extracted V. parahaemolyticus gDNA was used as a template to amplify Vp H-NOX (VP1877, gene ID: 1189384), HK (VP1876 gene ID: 1189383), LuxU (VP2098, gene ID: 1189609), and Hpt proteins (VP1472, gene ID: 1188978 and VP2127, gene ID: 1189639) by PCR. Primers for VP1877, VP1876 kinase domain only, VP1876 internal kinase domain only and VP2098 contained NdeI and XhoI up-stream/down-stream restriction sites, respectively. VP1472 and VP2127 primers contained NdeI and NotI restriction sites, respectively. For all PCR reactions, Phusion High-Fidelity DNA Polymerase (New England Biolabs, M0530S) was used. Amplified products were double digested then ligated into pET-20b(+) vector (Novagen) or pET-23aHis-TEV and sequenced (Stony Brook DNA sequencing facility). Site-directed mutagenesis to generate VP1876 HK H214A and D499A was carried out following the QuikChange Site-directed Mutagenesis kit protocol (Stratagene). The primer sequence used for cloning and site-directed mutagenesis will be provided upon request.

### Protein Expression and Purification

All proteins contained a 6× His tag on either the N or C-terminus and were purified by immobilized metal ion affinity chromatography using Ni-NTA agarose. Protein concentrations were determined by Bradford assay with bovine serum albumin as a standard (Bradford, 1976).

### H-NOX Complex Preparation and Electronic Microscopy

fmicb-10-02108 September 13, 2019 Time: 16:56 # 3

In an anaerobic glove bag, purified Vp H-NOX protein was incubated with 10 mM potassium ferricyanide for 5 min to make Fe (III) H-NOX. Potassium ferricyanide was then removed using PD10 desalting column (GE Healthcare). Prepared Vp Fe (III) H-NOX was incubated with 20 mM sodium dithionite for 30 min then desalted to prepare Fe (II) H-NOX. Fe (II) H-NOX was further incubated with 3 mM DPTA NONOate (Cayman Chemicals) for 30 min then desalted to prepare Fe (II) NO·H-NOX. To make CO bound H-NOX, Fe (II) H-NOX was bubbled with CO for 10 min in a closed Reacti-Vial (Thermo Fisher Scientific). Electronic spectra of all samples were measured by a Cary 100 UV-Vis spectrophotometer (Agilent) equipped with Cary temperature controller set at 20◦C. For temperature dependent 5, 6-coordinate NO·H-NOX distribution analysis, the sample's temperature was varied from 4 to 40◦C utilizing the temperature controller.

### NO Dissociation Rate Constant

This procedure has been described previously (Boon et al., 2006). Briefly, in an anaerobic glove bag, NO bound Vp Fe (II) H-NOX was diluted in 40 mM Tris-Cl, 150 mM KCl, 4 mM DTT, and 10% glycerol buffer at pH 8.0, then was rapidly mixed with an equal amount of the same buffer at 3, 30 or 300 mM of sodium dithionite saturated with CO. The absorption spectra were obtained by Cary 100 UV-Vis spectrophotometer (Agilent) equipped with a Cary temperature controller set at 20◦C. NO dissociation was monitored by tracking increasing Fe (II) CO peak at 424 nm and decreasing Fe (II) NO peak at 399 nm. The resulting data was fitted to a two-phase exponential association equation, y = y<sup>0</sup> + A<sup>1</sup> ∗ (1–e−x/t1) + A<sup>2</sup> ∗ (1–e−x/t2) to determine the NO dissociation rate constant. The experiments were repeated a minimum of three times for each sodium dithionite concentration. The resulting average koff (NO) rate constants were reported with standard error of the mean (SEM). The dissociation rate constants were independent of sodium dithionite concentrations.

### HK Autophosphorylation Assay

[γ-<sup>32</sup>P]-ATP (6000 Ci/mmol, 10 mCi/mL) was purchased from PerkinElmer Health Sciences Incorporated. All reactions were performed at room temperature. Reaction mixtures contained final concentrations of 40 mM Tris-Cl, 150 mM KCl, 4 mM DTT, 10% glycerol, 4 mM MgCl2, 5 µM histidine kinase at pH 8.0. Histidine kinase in the reaction buffer was allowed to equilibrate at room temperature for a few minutes then reactions were initiated by the addition of ATP (2 mM) with trace amount of [γ-<sup>32</sup>P]-ATP (10 µCi). For SDS-PAGE analysis, reactions were stopped by adding 5× SDS loading dye (0.25% bromophenol blue, 0.5 M dithiothreitol, 50% Glycerol, 10% sodium dodecyl sulfate, 0.25 M, pH 6.8 Tris-Cl) at indicated time points. Quenched reactions were separated by SDS-PAGE. After gel drying, the sample radioactivity was detected by a Typhoon scanner (Typhoon 9400, Amersham Biosciences) then quantified with image processing software ImageJ. For the dot blot assay, reactions were quenched with 25 mM H3PO4, and 30 µL aliquots were pipetted onto a nitrocellulose membrane in a dot blot apparatus as described previously (Fischer et al., 2017). The membrane was washed with 25 mM H3PO<sup>4</sup> and dried before exposure to a storage phosphor screen. Radioactivity in each spot was detected by a Typhoon scanner and quantified with ImageJ.

### HK Phosphatase Activity Assay

Purified Vp HK was mixed with the final concentration of 250 µM 3-O-methyl fluorescein phosphate (OMFP) in the reaction buffer (40 mM Tris–HCl, 150 mM KCl, and 10% glycerol at pH 8.0) at room temperature. HK's phosphatase activity was quantified by measuring the production of OMFP hydrolysis product, O-methylfluorescein (OMF) at 450 nm. All data acquisition was done by VICTOR X5 Multilabel Plate Reader (PerkinElmer).

### HK/LuxU Phosphotransfer Assay

The components of the reaction mixture were the same as those used in the HK autophosphorylation assay. Reactions were performed at room temperature. The final concentration of 9.4 µM HK in the reaction buffer was incubated with ATP/[γ-<sup>32</sup>P]-ATP mix solution for 40 min to autophosphorylate HK. The final concentration of 40.7 µM LuxU was then added to the reaction mixture to initiate phosphotransfer. Reactions were quenched by the addition of 5× SDS loading dye at various


<sup>a</sup>H-NOX domain from soluble guanylate cyclase from bovine lung. <sup>b</sup>H-NOX from Thermoanaerobacter tengcongensis. <sup>c</sup>H-NOX from Vibrio harveyi. <sup>d</sup>H-NOX from Nostoc punctiforme.

time points followed by gel electrophoresis, gel drying and image scanning. Histidine kinase only with ATP mix solution, LuxU only with ATP mix solutions were run along with HK + LuxU as controls.

### HK Autophosphorylation Activity Inhibition by H-NOX

fmicb-10-02108 September 13, 2019 Time: 16:56 # 4

All reactions were carried out at room temperature. The components of the reaction mixture were the same as those used in the HK autophosphorylation assay. In an anaerobic glove bag, different oxidation/ligation states of H-NOX [final concentration 69.4 µM or varying (NO·H-NOX) for the titration] were incubated with Vp HK (D499A, final concentration 5.3 µM) for 30 min. Reactions were initiated by the addition of ATP/[γ-<sup>32</sup>P]-ATP mix solution. Thirty minutes after the initiation of the reaction, reactions were quenched by adding 5× SDS loading dye, followed by SDS-PAGE, gel drying and image scanning.

### Phosphotransfer Profiling

The procedure followed what's been described by Laub (Laub et al., 2007). All reactions were performed at room temperature. The components of the reaction mixture were the same as those of the HK autophosphorylation assay. Final concentrations of 3.3 µM HK kinase domain and 3.3 µM internal kinase receiver (IKR) domain were incubated with ATP/[γ-<sup>32</sup>P]-ATP mix solution for 60 min. Final concentrations of 33 µM histidine containing phosphotransfer proteins (VP1472, VP2098, or VP2127) were added to the reaction mixtures to initiate phosphotransfer. At various time points, reactions were quenched by adding 5× SDS loading dye. Proteins were separated by gel electrophoresis followed by gel drying and image scanning.

### Phosphotransfer Specificity Test

All reactions were carried out at room temperature and the components of the reaction mixtures were the same as those used in phosphotransfer profiling. Reaction mixtures with various components, KD only, IKR only, VP1472 only, VP2098 (LuxU) only, VP2127 only, KD + IKR, KD + IKR + VP2098 (LuxU),

KD + VP1472, KD + VP2098 (LuxU), and KD + VP2127, were incubated with ATP/[γ-<sup>32</sup>P]-ATP mix solution for 60 min. Reactions were quenched by the addition of 5×

V. parahaemolyticus FeII-NO H-NOX. The absorption spectrum shows the shift in the 5/6-coordinate FeII-NO H-NOX ratio by temperature, lower temperatures favoring 6-coordinate (417 nm) and higher temperatures favoring 5-coordinate (399 nm). Absorption spectra were taken at 4◦C (–), 10◦C (···), 20◦C (- - -), 30◦C (–··–), and 40◦C (– – –).

SDS loading dye, followed by gel electrophoresis, gel drying and image scanning.

### RNA Extraction, cDNA Synthesis, and qPCR

RNA was extracted from bacterial cultures using the PureLink RNA Mini Kit (Ambion). Extracted RNA was treated with dsDNase (Thermo Fisher Scientific). cDNA synthesis and qPCR were carried out using either DyNAmo SYBR Green 2-Step qRT-PCR Kit or Maxima First Strand cDNA Synthesis Kits for RT-qPCR and DyNAmo HS SYBR Green qPCR Kit (Thermo Fisher Scientific). qPCR was carried out using secY (VP0277) as a reference gene. Data was analyzed by 11CT method (BIO-RAD). Sequences for primer sets used are as follows:

aphA forward: TCAGCGAAACTTATGGCTTG aphA reverse: GTTGAAGGCGTTGCGTAGTA opaR forward: GAAATTGCGCAAGTGTCTGT opaR reverse: ACGGACAACATGGTTGAGAA secY forward: CAGTGGTTTGGTCAGAATGG secY reverse: GGGCTAAGAGCCAAAGACAC

### RESULTS AND DISCUSSION

### Vp H-NOX Is an NO-Binding Protein

To begin investigating the NO/H-NOX-mediated QS circuit in V. parahaemolyticus, we cloned, expressed, and purified Vp H-NOX (VP1877, gene ID: 1189384) and analyzed its

spectroscopic and ligand-binding properties (**Figure 1** and **Table 1**). The absorption peaks of purified H-NOX from V. parahaemolyticus in various oxidation and ligation states are similar to those of the eukaryotic H-NOX homolog sGC, as well as V. harveyi, and other previously characterized H-NOX proteins (Stone and Marletta, 1994; Karow et al., 2004; Boon et al., 2006; Henares et al., 2012). The reduced-unligated (FeII unligated) and CO-bound (FeII-CO) Vp H-NOX complexes have Soret band maxima at 426 nm and 424 nm, respectively (**Figure 1**). Interestingly, NO-bound (FeII-NO) Vp H-NOX has two Soret peaks, at 399 and 417 nm, which indicates the protein is a mixture of 5- and 6-coordinate heme at 20◦C. In previous work on the H-NOX from Nostoc punctiforme, it was shown that the distribution of the FeII-NO coordination state is temperature dependent, with lower temperatures favoring the 6-coordinate complex and higher temperature favoring the 5 coordinate state. We tested if this was also the case for Vp H-NOX by varying the temperature from 4 to 40◦C. Indeed,

FIGURE 5 | Dependence of Vp HqsK D499A autophosphorylation on ATP concentration. (A) Dot blot assay showing radiolabeled Vp HqsK D499A bound to a nitrocellulose membrane shows HqsK autophosphorylation follows Michaelis-Menten kinetics. Each concentration of ATP was assayed in triplicate using biological replicates. (B) Quantification of Vp HqsK D499A autophosphorylation as a function of ATP concentration. The apparent K<sup>m</sup> is 35 µM. Error bars indicate the standard deviation of three biological replicates.

we observed that the distribution of the 5- and 6-coordinate complexes was temperature dependent, being predominately 6-coordinate at 4◦C with a shift toward 5-coordinate as the temperature was increased (**Figure 2**). This suggests that like Np H-NOX, a thermal equilibrium exists between 5- and 6- coordinate NO-bound Vp H-NOX, and cleavage of the Fe-His bond upon NO binding is temperature dependent. We also determined the NO dissociation rate constant for Vp H-NOX using a sodium dithionite/carbon monoxide trap (**Figure 3**; Kharitonov et al., 1997; Boon et al., 2005). We determined koff(NO) to be (4.3 ± 0.5) × 10−<sup>4</sup> s −1 at 20◦C, indicating a slow NO dissociation rate constant, similar to those of other previously characterized H-NOX proteins (**Table 1**). All of these results support that Vp H-NOX is an NObinding protein.

### Vp HqsK Is an Active Hybrid Kinase With Kinase and Phosphatase Activities

Bacterial hnoX genes often neighbor genes that code for signaling proteins, such as HKs, cyclic-di-GMP metabolizing proteins, or methyl accepting chemotaxis proteins (Iyer et al., 2003). In all H-NOX homologs characterized to date, H-NOX has been demonstrated to regulate the activity of its associated signaling protein as a function of binding to NO. In V. parahaemolyticus, hnoX is encoded in the same operon as a hybrid HK predicted to be involved in QS. We named this kinase HqsK (H-NOXassociated QS kinase). To identify whether HqsK (VP1876, gene ID: 1189383) is a functional kinase, we cloned, expressed, and purified the protein, and then conducted autophosphorylation activity assays. When HqsK is incubated with ATP containing trace [γ-<sup>32</sup>P]-ATP over time, the resulting autoradiography shows accumulating radiolabeled phosphate on the HK in a time-dependent manner, confirming kinase activity of Vp HqsK (**Figure 4**). We also tested the kinase activity as a function of ATP concentration (**Figure 5**). HqsK appears to follow Michaelis-Menten kinetics, with an apparent K<sup>m</sup> of 35 µM.

Many hybrid HKs are dual-functioning enzymes that have phosphatase activity in addition to kinase activity, in order to regulate the phosphorylation state of a partner response regulator (Parkinson and Kofoid, 1992). To test whether Vp HqsK is such an enzyme, phosphatase activity was monitored using a generic phosphatase substrate 3-O-methyl-fluorescein phosphate (OMFP) that yields O-methylfluorescein (OMF) upon dephosphorylation. Wild-type HqsK displays increasing OMF fluorescence over time, indicating Vp HqsK has phosphatase activity (**Figure 6**).

The kinase and phosphatase activities of hybrid HKs are dependent on conserved His and Asp residues, respectively (Kwon et al., 2000; Albanesi et al., 2009; Goodman et al., 2009). To demonstrate the importance of these residues for phosphatase activity, we generated H214A and D499A mutant HqsKs and repeated the phosphatase assay. Wild-type and H214A HKs have

FIGURE 7 | In vitro phosphotransfer assay of Vp HqsK to LuxU. Phosphotransfer activity was detected by SDS-PAGE and autoradiography. Stained polyacrylamide is shown for uniform protein loading (A). Vp HqsK transfers phosphate to LuxU over time in the presence of ATP (B). Time indicates the length of incubation after the addition of LuxU to HK.

FIGURE 8 | Phosphotransfer specificity test. In vitro phosphotransfer assay from HqsK to LuxU with WT, H214A, and D499A HqsK. Stained polyacrylamide gel (A) and autoradiography (B) are shown. H214A HqsK has no autophosphorylation activity due to the phosphorylation site mutation. D499A, kinase receiver domain mutant autophosphorylates, but does not phosphotransfer to receiver domain nor LuxU.

the same level of phosphatase activity as the wild-type enzyme, whereas D499A has significantly reduced phosphatase activity, indicating conserved D499 in the internal kinase receiver domain (IKR) is required for Vp HqsK phosphatase activity.

### Vp HqsK Transfers Phosphate to the Quorum Sensing Phosphotransfer Protein LuxU

To establish the phosphotransfer circuit between Vp HqsK and the QS histidine-containing phosphotransfer protein LuxU, a phosphotransfer assay was conducted. Purified HqsK was incubated with ATP containing trace [γ-<sup>32</sup>P]-ATP, followed by the addition of purified LuxU. The resulting autoradiography shows an accumulation of radiolabeled phosphate on LuxU over time, indicating Vp HqsK transfers phosphate to LuxU (**Figure 7**). We also tested HqsK/LuxU phosphotransfer using the H214A and D499A HqsK mutants to confirm the transfer is occurring, as predicted, via phosphotransfer from H214 to D499 in HqsK followed by transfer to H56 in LuxU. As expected, H214A HqsK showed neither autophosphorylation nor phosphotransfer activities, whereas D499A HqsK exhibited only autophosphorylation activity (**Figure 8**).

### LuxU Is the Cognate Phosphotransfer Protein for Vp HqsK

Vibrio parahaemolyticus has three stand-alone Hpts. As described above, we have demonstrated that Vp HqsK transfers phosphate to LuxU, but since many HKs can transfer phosphate to

FIGURE 10 | Vp HK, Hpt phosphotransfer specificity test. Various combinations of KD, IKR, and Hpt proteins were mixed to observe phosphotransfer. Stained polyacrylamide gel (A) and autoradiography (B) are shown.

multiple partners, there remained a possibility that this HqsK/LuxU phosphotransfer was not exclusive and/or that LuxU is not a cognate phosphotransfer partner for HqsK. It has been demonstrated that in vitro, an HK will have a kinetic preference for its in vivo cognate response regulator, exhibiting a much faster rate of phosphotransfer with its cognate partner (Laub et al., 2007). This preference is determined in an experiment called phosphotransfer profiling, in which either loss of phosphate from the HK or the appearance of a preferentially phosphorylated Hpt is detected using PAGE followed by autoradiography (Laub et al., 2007).

The kinase domain and internal kinase receiver domain from Vp HqsK were separately cloned and expressed to isolate the kinase and receiver domains. We also purified all three stand-alone Hpt proteins from V. parahaemolyticus: LuxU (VP2098, gene ID: 1189609), VP1472 (gene ID: 1188978), and VP2127 (gene ID: 1189639). Then each Hpt was separately added to a mixture of Vp HqsK KD and IKR domains that had been preincubated with radiolabeled ATP. The resulting autoradiography showed no apparent band intensity loss from any of the kinase domains, nor band appearance for VP1472 or VP2098. However, a band corresponding to phosphorylated LuxU appears after 15 min and increases in intensity over time, indicating the accumulation of phosphate on LuxU (**Figure 9**). To verify this accumulation of phosphate on LuxU is not due to nonspecific phosphorylation, but due to phosphotransfer through the HqsK KD/IKR domains, we conducted a phosphotransfer specificity test (**Figure 10**). The results indicate that LuxU does not phosphorylate itself, nor can it directly receive phosphate from the HqsK kinase domain, but it can only be phosphorylated via a His-Asp-His phosphotransfer through the HqsK kinase and IKR domains. Overall, these results show LuxU is the cognate phosphotransfer Hpt protein for HqsK.

### H-NOX Suppresses Vp HqsK Autophosphorylation Upon NO Binding

In the NO-responsive QS circuit in V. harveyi, the basal kinase activity of Vh HqsK was repressed by H-NOX upon NO binding (Henares et al., 2012). To test whether Vp HqsK is regulated

in a similar manner, various ligation states of Vp H-NOX were incubated with HqsK and the effects on autophosphorylation were observed. Vp HqsK and FeII-unligated, FeII-CO or FeII - NO Vp H-NOX were incubated in an anaerobic chamber, and then ATP with trace [γ-<sup>32</sup>P]-ATP was added to initiate HqsK autophosphorylation. The resulting autoradiography showed significant kinase activity inhibition with all Vp H-NOX complexes, with the most inhibition (75%) by FeII-NO H-NOX (**Figure 11**). We also titrated HqsK with FeII-NO H-NOX and observed decreasing Vp HqsK autophosphorylation in an FeII - NO H-NOX concentration-dependent manner (**Figure 12**). Our data indicate that Vp H-NOX suppresses the kinase activity of the associated HqsK upon NO binding, similar to what has been observed in V. harveyi.

### The Effect of NO on the Master Quorum Regulatory Genes opaR and aphA

In the V. parahaemolyticus QS circuit, LuxU and LuxO ultimately regulate the translation of two master quorum regulatory proteins, OpaR, and AphA (Rutherford et al., 2011). However, these master quorum regulatory proteins have also been shown to reciprocally regulate each other's gene transcription. AphA binds to the luxR promoter in V. harveyi and directly suppress transcription of luxR (Rutherford et al., 2011). In V. parahaemolyticus, OpaR binds to the aphA promoter and represses transcription of aphA (Zhang et al., 2012). These findings of transcriptional regulation of the master QS genes prompted us to investigate if NO has a role in regulating aphA and opaR transcription. Here, we analyze the effect of various concentrations of NO (delivered with the NO donor DETA NONOate) on the transcription of these master quorum regulatory proteins at low (A600nm = 0.2) cell density using qRT-PCR.

First, the growth of V. parahaemolyticus in the presence of various concentrations of DETA NONOate was monitored to determine the NO toxic threshold (**Figure 13**). The growth curve for cultures with 0, 50, 100, and 200 µM DETA NONOate [equivalent to 0, 190, 380, and 760 nM NO under these conditions (Bebok et al., 2002)] showed almost identical growth, but almost no cell growth was seen with 500 µM DETA NONOate (1900 nM NO). Accordingly, V. parahaemolyticus cultures grown with

0–200 µM DETA NONOate were used to monitor aphA and opaR transcription (**Figure 14**). The gene expression levels of opaR and aphA were quantified using secY (VP0277, gene ID: 1187744) as a reference gene.

We observed a trend of increasing transcription of opaR as a function of NO (1.22-, 1.71-, and 2.04-fold change with 50, 100, and 200 µM DETA NONOate, respectively). The foldchange of 2.04 with 200 µM NONOate was significantly different (p < 0.05) with respect to no NONOate, but the other two were not (p > 0.05). Transcript levels of aphA remained almost constant with increasing NO, although at 200 µM NONOate, there was an increase of 1.56 with respect to no NONOate.

An NO-dependent increase in both aphA and opaR transcription is unexpected since opaR suppresses aphA transcription. The mechanism of this increased aphA transcription at higher NO concentration is unknown. It is possible that although 200 µM NONOate did not hinder the bacterial growth, it is causing stress to bacteria, and increased aphA transcription is a part of the stress response. Also, it is possible that NO may be activating an unidentified signaling cascade that is turned on only at higher NO concentrations.

### DISCUSSION

Quorum sensing is a key signaling system bacteria use to orchestrate population-wide gene expression patterns appropriate for different stages of growth or environmental conditions. Bacteria employ various types of QS circuits, AIs, and AI receptors that are most suitable for their particular lifestyle or the environment in which they live. Not only do bacteria monitor the AI concentrations of their own and other bacterial species, many actively interfere with QS circuits of competing species through a process called quorum quenching (Leadbetter and Greenberg, 2000; Dong et al., 2001). Some pathogenic or symbiotic bacteria also integrate host-organismoriginated signals into QS to manage effective colonization and pathogenicity (Beck von Bodman et al., 1992; Fuqua and Winans, 1994; Hwang et al., 1994). Similarly, eukaryotes monitor bacterial AIs to detect their presence and activate an immune response (Givskov et al., 1996; Chun et al., 2004). Autoinducer-mediated QS is a dynamic communication system that spans across different bacterial species and even across kingdoms. Recent studies have revealed increasing complexity in QS signaling, indicating the growing importance of QS not only in the regulation of the bacterial life cycle, but also in our eukaryotic immune response.

Nitric oxide has been shown to promote bacterial biofilm dispersal through the regulation of the bacterial second messenger signaling molecule c-di-GMP in Shewanella woodyi (Liu et al., 2012) and Shewanella oneidensis

(Plate and Marletta, 2012). Although their mechanisms have not been identified, NO has also been shown to modulate biofilm formation in Pseudomonas aeruginosa (Barraud et al., 2006) and Nitrosomonas europaea (Schmidt et al., 2004). In V. harveyi QS circuits, the NO signal is integrated into a QS signaling cascade through H-NOX/HqsK (Henares et al., 2012). H-NOX shifts HqsK's activity from kinase to phosphatase upon NO binding. Together with other AI sensory HKs, it controls the phosphorylation state of the RR LuxO. LuxO ultimately regulates hundreds of gene expressions through two master quorum regulatory proteins LuxR (OpaR homolog) and AphA (Rutherford et al., 2011). Based on gene analysis, V. harveyi does not have an NO synthase. Thus, we hypothesize V. harveyi detects NO originated from its host organism or other NOproducing bacteria and integrates that information in its QS circuit. In this paper, we show that V. parahaemolyticus appears to also have an NO-responsive signaling pathway that feeds into the QS circuit.

The signaling cascade of V. parahaemolyticus NO-responsive QS circuit is similar to that of V. harveyi. NO is detected by a cytoplasmic H-NOX protein that switches the activity of HqsK between kinase and phosphatase. HqsK integrates the NO signal into a QS circuit to regulate the phosphorylation state of QS RR LuxO and ultimately control the transcription of two master quorum regulatory proteins OpaR and AphA. Studies on other Vibrios (V. harveyi and V. cholerae) show that low cell density promotes phosphorylation of LuxO and activation/suppression of aphA/luxR transcription. High cell density promotes dephosphorylation of LuxO and activation/suppression of luxR/aphA transcription. Our data suggest that NO functions as an AI and participates in the V. parahaemolyticus QS circuit. If OpaR and AphA's reciprocal regulatory circuit is conserved in V. parahaemolyticus, we expect increasing NO concentration would promote opaR transcription and suppress aphA transcription. Our qPCR result supported the former

### REFERENCES


part of the hypothesis. Increasing NO concentration promoted opaR transcription in a concentration-dependent manner, supporting a role for NO in the QS circuit as an AI. However, aphA transcription also increased at high NO concentration (200 µM). This discrepancy may be due to NO toxicity or may mean the presence of an unidentified NO signaling cascade in V. parahaemolyticus. We are currently investigating the mechanism behind the increased aphA transcription caused by NO and the alternative NO signaling cascade in V. parahaemolyticus.

### DATA AVAILABILITY

All datasets generated for this study are included in the manuscript and/or the supplementary files.

### AUTHOR CONTRIBUTIONS

TU and JF performed the experiments. TU, JF, and EB wrote the manuscript.

### FUNDING

This work was supported by the Stony Wold-Herbert Fund, the National Science Foundation (Grant CHE-1607532 to EB), and the National Institutes of Health (Grant GM118894-01A1 to EB).

### ACKNOWLEDGMENTS

We thank Dr. Roger Johnson and the Boon Group for their helpful discussions.



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

Copyright © 2019 Ueno, Fischer and Boon. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Insights Into Nitric Oxide Modulated Quorum Sensing Pathways

#### Ilana Heckler and Elizabeth M. Boon\* †

Department of Chemistry, Institute of Chemical Biology & Drug Discovery, Stony Brook University, Stony Brook, NY, United States

The emerging threat of drug resistant bacteria has prompted the investigation into bacterial signaling pathways responsible for pathogenesis. One such mechanism by which bacteria regulate their physiology during infection of a host is through a process known as quorum sensing (QS). Bacteria use QS to regulate community-wide gene expression in response to changes in population density. In order to sense these changes in population density, bacteria produce, secrete and detect small molecules called autoinducers. The most common signals detected by Gram-negative and Grampositive bacteria are acylated homoserine lactones and autoinducing peptides (AIPs), respectively. However, increasing evidence has supported a role for the small molecule nitric oxide (NO) in influencing QS-mediated group behaviors like bioluminescence, biofilm production, and virulence. In this review, we discuss three bacteria that have an established role for NO in influencing bacterial physiology through QS circuits. In two Vibrio species, NO has been shown to affect QS pathways upon coordination of hemoprotein sensors. Further, NO has been demonstrated to serve a protective role against staphylococcal pneumonia through S-nitrosylation of a QS regulator of virulence.

Keywords: nitric oxide, quorum sensing, gas sensing, hemoproteins, virulence, biofilm

### INTRODUCTION

Quorum sensing (QS) refers to the process by which bacteria regulate gene expression in response to small molecules called autoinducers. Autoinducers, typically acylated homoserine lactone molecules in Gram-negative bacteria or oligopeptides in Gram-positive bacteria, are synthesized inside the cell and freely diffuse across the membrane. As cell density rises, the concentration of autoinducers in the surrounding environment also increases. When a threshold level of autoinducers is reached, autoinducers bind to their respective receptors and initiate a signal transduction cascade downstream, ultimately resulting in the repression or expression of specific genes. QS has been shown to mediate several bacterial processes including bioluminescence, sporulation, competence, antibiotic production, biofilm formation, and virulence factor production (Miller and Bassler, 2001).

Bacterial receptors of autoinducers can be membrane bound hybrid histidine kinase proteins (Ng and Bassler, 2009). Hybrid histidine kinases contain a receiver domain, in addition to an input domain and kinase core. Upon binding ATP, the receptor will autophosphorylate on a conserved histidine residue in the kinase core. Phosphate is then transferred from the histidine residue to an aspartate residue on the same polypeptide. Hybrid histidine kinases rely on an intermediate histidine phosphotransferase (HPT) protein to then transfer the phosphate from the receiver domain of the kinase, to the receiver domain of a response regulator. The most

#### Edited by:

Tom Defoirdt, Ghent University, Belgium

#### Reviewed by:

Julia Van Kessel, Indiana University Bloomington, United States Gary J. Vora, United States Naval Research Laboratory, United States

#### \*Correspondence:

Elizabeth M. Boon elizabeth.boon@stonybrook.edu

### †ORCID:

Elizabeth M. Boon orcid.org/0000-0003-1891-839X

#### Specialty section:

This article was submitted to Microbial Physiology and Metabolism, a section of the journal Frontiers in Microbiology

> Received: 12 June 2019 Accepted: 05 September 2019 Published: 24 September 2019

#### Citation:

Heckler I and Boon EM (2019) Insights Into Nitric Oxide Modulated Quorum Sensing Pathways. Front. Microbiol. 10:2174. doi: 10.3389/fmicb.2019.02174

common autoinducers sensed by Gram-negative bacteria are acylated homoserine lactones, or an interfuranosyl borate diester interspecies signal named autoinducer-2 (Papenfort and Bassler, 2016). In Gram-positive bacteria, autoinducing peptides (AIPs) which are synthesized by ribosomes and secreted extracellularly by specific transporters, bind to receptor kinases at high cell density to initiate QS pathways.

Nitric oxide (NO) is a diatomic membrane-permeable gas molecule that has been implicated in a wide range of physiological processes in both eukaryotes and bacteria. Many of the interesting biological properties of NO can be attributed to its unique physical properties. Due to its size, neutral charge, and hydrophobic nature, NO is able to freely diffuse into the cytosol, where it may undergo a variety of reactions including protein S-nitrosylation and metal coordination. While at high (micromolar) concentrations, NO is cytotoxic, at low concentrations, NO has been shown to be a signaling molecule in both bacteria and eukaryotes (Murphy, 1999; Arora et al., 2015).

In mammals, NO is synthesized by nitric oxide synthases (NOS) from L-arginine, oxygen and NADPH. The first report of bacterial NOS activity came from studies of the Nocardia species (Chen and Rosazza, 1994). Although it does not appear that most bacteria synthesize NO enzymatically, bacteria can produce NO as a by-product of denitrification, through the anaerobic reduction of nitrite by nitrite reductase. NO is then subsequently reduced to nitrous oxide by NO reductase. Indeed, the expression of nitrite reductase and NO reductase in Pseudomonas aeruginosa biofilms were found to be QS-dependent, suggesting that QS is responsible for the maintenance of NO levels in this organism (Hentzer et al., 2005).

Bacteria could also respond to NO that is produced by a eukaryotic host during symbiotic or pathogenic engagement. The ability to respond to NO to mediate QS related behaviors may provide microbes with an evolutionary advantage when encountering NO during infection of a host. In addition to a role for NO as a QS signal, there may also be a connection between QS and the regulation of intracellular levels of NO.

### NO AFFECTS QUORUM SENSING PATHWAYS THROUGH LIGATION TO CYTOSOLIC HEMOPROTEIN RECEPTORS

Evidence has suggested that bacteria can respond to low, physiologically relevant levels of NO for the regulation of cellular processes such as biofilm formation and dispersion (Arora et al., 2015). In mammals, NO sensing is essential to regulating vasodilation and is carried out by the soluble guanylyl cyclase (sGC) receptor, which coordinates NO at the ferrous center of a heme cofactor. Homologs of the heme binding domain of sGC have been discovered in bacteria, suggesting a role for NO in regulating bacterial physiology (Iyer et al., 2003). The most well characterized NO sensor in prokaryotes is a family of hemoproteins named heme-nitric oxide/oxygen (H-NOX) proteins (Karow et al., 2004; Plate and Marletta, 2017). H-NOX proteins bind NO selectively, and with picomolar affinity (Tsai et al., 2012). H-NOX domains are commonly encoded in putative operons with cyclic di-guanosine monophosphate (c-di-GMP) processing enzymes and/or histidine kinase proteins. c-di-GMP is a bacterial secondary messenger molecule that plays a role in controlling the switch between motile and sessile/biofilm lifestyles. For example, in Shewanella oneidensis, NO bound H-NOX was found to inhibit the activity of an associated histidine kinase, resulting in decreased phosphorylation of a response regulator protein involved with biofilm formation through modulation of c-di-GMP levels (Plate and Marletta, 2012). Recently, NO bound H-NOX was shown to also influence bacterial biofilm formation in the marine bacterium Vibrio fischeri through inhibition of its associated histidine kinase, HahK (Thompson et al., 2019).

However, not all bacteria that have been shown to respond to NO encode for an H-NOX protein in their genome. This discrepancy led to the discovery of another bacterial heme bound NO sensing protein named NosP. NosP was first characterized in P. aeruginosa where it was found to have a role in biofilm formation (Hossain and Boon, 2017). In the human pathogen Vibrio cholerae, and in the marine pathogen Vibrio harveyi (recently reclassified as Vibrio campbellii) (Lin et al., 2010), NO has been implicated in mediating QS through pathways involving either H-NOX or NosP, and will be discussed here. In addition, our lab has found evidence for a similar NO responsive QS circuit in the gastroenteritis causing bacterium Vibrio parahaemolyticus (Ueno et al., 2019).

### Vibrio harveyi Biofilm, Bioluminescence, and Motility Are Regulated by a NO-Mediated QS Pathway

Vibrio harveyi is a bioluminescent opportunistic marine pathogen often found in tropical waters. V. harveyi bioluminescence, as well as biofilm, virulence and flagellar production, is controlled by QS. V. harveyi is thought to respond to at least three different autoinducers through the binding of three separate QS receptors (Bassler et al., 1993, 1994; Henke and Bassler, 2004). These receptors, LuxN, LuxQ, and CsqS are transmembrane hybrid histidine kinases. LuxQ also utilizes an accessory protein, LuxP, to sense its cognate autoinducer (Bassler et al., 1994; Neiditch et al., 2006). When cell density is low, the receptors LuxN, LuxQ, and CsqS autophosphorylate and subsequently transfer phosphate to a single histidine phosphotransfer protein LuxU (**Figure 1**). Phosphorylated LuxU then transfers phosphate downstream to a transcription factor LuxO which promotes the transcription of Quorum Regulatory RNAs (qrrs) (Freeman and Bassler, 1999a,b). Qrrs destabilize the mRNA encoding LuxR, the transcription regulator responsible for luciferase production and activator of biofilm related genes, consequently inhibiting bioluminescence as the light producing lux operon is not expressed (Lenz et al., 2004). However, when cell density increases, and autoinducer concentration is high, the receptors switch from kinases to phosphatases resulting in the dephosphorylation and inactivation of LuxO, and subsequent translation of LuxR leading to light production.

resulting in the transcription of four small regulatory RNA molecules (Qrrs 1-5). Qrrs inhibit the expression of LuxR, the transcription regulator responsible for luciferase production and activator of biofilm related genes. (B) QS sensing signal transduction in V. harveyi at high cell density and in the presence of NO. When cell density is high, and autoinducer concentrations rise, CqsS, LuxPQ, and LuxN act as phosphatases, dephosphorylating LuxU. Qrrs are not expressed and LuxR is transcribed, resulting in bioluminescence, and the activation of biofilm related genes. In the presence of NO, NO bound H-NOX inhibits the autophosphorylation of HqsK resulting in decreased phosphate flow downstream, in addition to the repression of flagellin proteins by an unknown mechanism.

Vibrio harveyi has also been shown to contain a NOmediated QS pathway (Henares et al., 2012). Low cell density cultures of V. harveyi exhibit increased bioluminescence in the presence of exogenous nanomolar NO, suggesting that NO can modulate QS pathways. The effect on NO on QS was attribute to a fourth hybrid histidine kinase, named Hqsk, that has both kinase and phosphatase activities and is able to transfer phosphate to LuxU. Unlike the other three previously characterized V. harveyi kinase receptors, HqsK is cytosolic, and does not contain a sensory domain in its primary sequence. Instead, HqsK activity was found to be regulated through interaction with an NO-sensitive H-NOX protein. NO-bound VhH-NOX inhibits the kinase activity of HqsK which contributes to a loss of phosphorylated LuxU and a subsequent increase in light production through the LuxU/LuxO pathway described above. It should be reiterated here that V. harveyi contains three additional AI receptors, which together overwhelm the effect of NO at high cell density (when, presumably, the other three AIs are present at high concentration); the addition of NO to high cell density cultures of V. harveyi was not found to significantly increase bioluminescence (Henares et al., 2012). This suggests that NO may influence bioluminescence only at low cell density, or perhaps be only a minor contributor to the total QS output.

In addition to bioluminescence, NO was found to influence QS regulation of biofilm and flagellar production in V. harveyi (Henares et al., 2013). Many bacteria, including V. harveyi, rely on QS to initiate a switch between a motile and sessile (or biofilm) lifestyle. At high cell density, V. harveyi enters biofilm through the negative regulation of genes involved with motility (Waters and Bassler, 2006). Analogous to a high-density state, the addition of low nanomolar (50 nM) NO to cultures of V. harveyi resulted in thicker biofilms compared to cultures grown in the absence of NO. Further, the addition of NO to a 1hnoX mutant strain lost the same biofilm enhancement phenotype as the wildtype strain, indicating that VhH-NOX is involved in the NO-mediated biofilm pathway. As described previously, H-NOX plays a role in V. harveyi QS through an interaction with HqsK. Interestingly, the addition of autoinducers to cultures of V. harveyi did not

increase biofilm to the same degree that NO addition had (Henares et al., 2013). This finding suggests that the NO/H-NOX pathway is primary in the regulation of biofilm.

The initial stage of biofilm formation, when surface attachment occurs, is correlated with a decrease in motility and is dependent on functional flagella. While functional flagella are critical for colonization and initial attachment, late stage biofilms are often made up of bacteria that have lost their flagella. In the marine bacterium V. fischeri, flagellar proteins have been previously shown to be negatively regulated by QS (Lupp and Ruby, 2005). iTRAQ proteomics analysis was performed on V. harveyi in the presence of NO. Upon addition of 50 nM NO, the relative abundance of several V. harveyi flagellin genes, was shown to be decreased, consistent with the effect of NO on biofilm observed in the same study (Henares et al., 2013). Interestingly, however, the effect of NO on biofilm and flagellin concentration was NO concentration-dependent, however; as NO concentration increased, biofilm decreased and flagellin proteins increased (50–200 nM NO was studies). Nevertheless, since the loss of flagellin is associated with biofilm formation, these experiments provide a possible mechanism by which low nanomolar NO influences QS-mediated biofilm formation through the H-NOX/HqsK pathway.

Key concept 1: NO mediates biofilm enhancement analogously to a high cell-density state in Vibrio haveryi.

### Vibrio cholerae Contains a NO Sensing Hybrid Histidine Kinase Receptor

Vibrio cholerae relies on QS for the regulation of biofilm and virulence related genes (Kovacikova and Skorupski, 2002; Hammer and Bassler, 2003). Vibrio polysaccharide genes (vps), involved in the synthesis of the exopolymeric matrix of biofilms, and the master regulator of virulence factor production, AphA, are under QS control. Four separate QS pathways are believed to work in parallel in V. cholerae to allow for the detection of multiple signals (Jung et al., 2016). Four hybrid histidine kinase receptors have been identified. Three of these receptors, CqsS, LuxP/Q, and Vc1831 (CqsR), are membrane bound and the fourth, VpsS, is cytosolic (**Figure 2**; Miller et al., 2002; Shikuma et al., 2009). When autoinducer concentrations are high, CqsS, LuxP/Q and CqsR bind their cognate autoinducers and act as phosphatases, drawing phosphate away from the common phosphotransfer protein LuxU. At high cell density, HapR, the homolog of V. harveyi's LuxR, is expressed, and represses genes responsible for biofilm and the master regulator of virulence factors, AphA (Rutherford et al., 2011). Therefore at high cell density, V. cholerae disperses from biofilm and is latent. However, when the concentration of autoinducers is small, typically at low cell density, the membrane bound receptors act as kinases and initiate phosphoryl transfer to LuxU and subsequently to the transcription factor LuxO. As in V. harveyi, phosphorylated LuxO in V. cholerae, initiates the expression of small regulatory RNA molecules qrrs (1-4) which inhibit HapR activity and also stabilize AphA (Svenningsen et al., 2008; Shao and Bassler, 2012). Repression of HapR activity derepresses expression of AphA and biofilm vps genes, resulting in increased virulence and biofilm at low cell density.

Until recently, the cognate autoinducer for the cytosolic kinase receptor VpsS was unknown, though VpsS was predicted to participate in QS based on studies demonstrating its purified receiver domain accepts phosphate from phosphorylated LuxU in vitro (Shikuma et al., 2009). Early studies of VpsS found that overexpression of vpsS in V. cholerae resulted in LuxOmediated activation of vps genes and a hyperbiofilm phenotype. LuxO activation of vps genes was found to occur by activation of the positive regulator of vps gene expression, VpsR and is independent of HapR. Recently, our lab has shown that full-length VpsS participates in phosphotransfer with LuxU (Hossain et al., 2018). Like HqsK in V. harveyi, VpsS does not contain a sensory domain, which suggests that an accessory protein is used to regulate its activity. Our lab discovered that VpsS is co-cistronic with a novel NO sensing hemoprotein called NosP. Further, we showed that purified NO-bound NosP inhibits the autophosphorylation of VpsS, which subsequently results in decreased levels of phosphorylated LuxU in vitro (Hossain et al., 2018).

These experiments suggest that V. cholerae may sense NO to regulate gene expression through a QS pathway involving LuxU and LuxO. As in V. harveyi, NO appears to act analogously to an autoinducer to mimic a high cell density state where phosphate is transferred downstream through LuxU. Interestingly, V. cholerae has been shown to disperse from biofilm in the presence of nanomolar NO (Barraud et al., 2009). A decrease in phosphate flux downstream, as a result of NO-bound NosP inhibition of VpsS autophosphorylation, would explain the biofilm dispersal phenotype of V. cholerae in the presence of NO, as vps genes responsible for biofilm formation would be repressed. The relative effect of NO-contributed phosphate flux compared to that of the three other QS pathways in V. cholerae has not been quantified. It is possible, that like with V. harveyi, NO modulation of QS occurs predominantly at low cell density and/or is a minor overall contributor to phosphate flux through LuxU/LuxO. It is also possible that NO-modulated QS represents the broader possibility that exogenous environmental signals are able to modulate QS, a departure from QS outputs dependent on only true autoinducing (and thus cell density-dependent) molecules.

### NO AND INNATE IMMUNITY

In response to a bacterial infection, the mammalian host produces high levels of NO through the upregulation of (iNOS). Nitrosative stress is an important component of the hosts innate immunity as it curbs microbial growth through the disruption of the fundamental physiological process including respiration, metabolism, and DNA replication (Murphy, 1999). However, bacteria have evolved to cope with nitrosative stress in order to circumvent host defenses during infection. In this section, we will discuss the discovery of a NO sensitive QS circuit in Staphylococcus aureus that may provide the bacterium with an evolutionary advantage when encountering high levels of NO during infection of a human host.

phosphatases, dephosphorylating LuxU. Qrrs are not expressed and HapR is transcribed, resulting in the inhibition of AphA and genes involved in vps expression. Likewise, in the presence of NO, VcNosP FeII–NO inhibits the autophosphorylation activity of VpsS and results in decreased phosphate flux through LuxU.

### Nitrosylation of a QS Regulator Inhibits Virulence of S. aureus

In addition to hemoprotein coordination, NO may exert its influence over bacterial QS pathways by protein S-nitrosylation (Kovacs and Lindermayr, 2013). NO modification of cystine residues is an important post translation modification governing protein function that has been observed in mammals, bacteria and plants. Free radical NO can react directly with thiyl radicals or may first react with an oxidant such as superoxide, oxygen or redox metals to form S-nitrosylating agents like peroxynitrite (Kovacs and Lindermayr, 2013). Not surprisingly, considering the inhibitory effect of NO on microbial survival, targets of S-nitrosylation within the genome of the commensal bacterium, S. aureus, include proteins involved in carbohydrate and lipid metabolism, tRNA and cell wall biosynthesis, DNA replication and repair, and amino acid metabolism (Urbano et al., 2018). Interestingly however, a small subset of S-nitrosylated proteins in S. aureus are responsible for antibiotic resistance and virulence. AgrA, a major transcriptional activator of virulence genes and a key component of staphylococcal QS, was found to be a target of S-nitrosylation at three separate cysteine residues C6, C123, and C199 (Urbano et al., 2018). In S. aureus, when cell density is high, AgrA is activated through phosphotransfer from the autophosphorylated AgrC receptor (**Figure 3**). When phosphosphylated, AgrA subsequently binds to several promotors responsible for the expression of virulence factors, including agrPIII, leading to positive autoregulation of the agr operon. Upon addition of a NO donor, the transcription of several AgrA targets was inhibited in a concentration dependent matter. Further, a NO insensitive mutant, in which C199 was replaced with serine residue, exhibited resistance to NO (Urbano et al., 2018). These findings suggest that NO inhibits QS-mediated virulence in S. aureus through the S-nitrosylation of AgrA cysteine residues, particularly C199. The authors also hypothesize that considering the conservation of cysteine residues across the LytTR family of regulators, NO might affect similar processes in other bacterial species through cysteine modification.

Key concept 2: S-nitrosylation of AgrA cysteine residues impedes binding to target promoters.

α-toxin is a pore forming toxin that is the major contributor of S. aureus pathogenesis (Berube and Bubeck Wardenburg, 2013). In S. aureus pneumonia, α-toxin is responsible for lung

injury and inflammation (Bartlett et al., 2008). The production of α-toxin is regulated by AgrA, as activation of agrPIII induces RNAIII, a regulatory RNA molecule responsible for stimulation of α-toxin transcription (Morfeldt et al., 1995). The Fang laboratory discovered that the QS pathway used to regulate α-toxin production in S. aureus is affected by NO (Urbano et al., 2018). Congenic iNOS knockout mice were more susceptible to S. aureus infection than iNOS-competent mice suggesting that NO serves a protective role in the host during infection. While there were no significant differences in bacterial burden in the lungs of mice with and without iNOS, significantly higher levels of α-toxin were produced in iNOS-deficient mice compared with iNOS-competent mice (Urbano et al., 2018). Further, iNOSdeficient mice infected with a S. aureus mutant, in which C55 and C199 were replaced with NO insensitive residues, exhibited no differences in α-toxin levels compared to iNOS competent mice. Taken together, the in vitro and in vivo findings provide a mechanism by which NO serves a protective role during S. aureus infection by inhibiting AgrA activation of toxin expression through S-nitrosylation of AgrA cysteines required for promotor binding. The authors speculate that a potential advantage to S. aureus, of having a NO sensitive QS circuit to represses virulence, may be to maintain a balance between a pathogenic and commensal lifestyle during colonization of the human nose. Interestingly, asymptomatic nose carriers of S. aureus were found to contain bacteria with weak expression of Agr regulated toxins (Burian et al., 2010).

Key concept 3: In a S. aureus model of infection, virulence, and not bacterial burden, is responsible for the differential disease severity between iNOS-deficient and iNOS-competent mice.

### CONCLUDING REMARKS

In this review, the influence of NO on the QS-regulated behaviors of V. harveyi, V. cholerae, and S. aureus have been discussed. In Vibrios, NO ligation to hemoproteins has been found to influence the autophosphorylation of partner histidine kinases proteins that are integrated in QS pathways. NO appears to influence QS-mediated behaviors via a different mechanism in S. aureus, namely S-nitrosylation of a QS regulator protein. It has yet to be determined whether S-nitrosylation occurs in other bacterial species (including V. harveyi and V. cholerae) as a means to regulate cell-to-cell communication. Based on the currently

available data, it is likely that additional, as of yet undiscovered, NO-sensitive QS pathways exist in other bacterial species, both Gram-negative and Gram-positive strains. Like the examples first characterized and described here, these pathways are likely to detect NO concentrations through ligation to a hemoprotein or S-nitrosylation of cysteine residues.

Major questions concerning the role of NO in QS are outstanding and should be the subject of future study. There is no current evidence that NO is truly an autoinducer, as it may not be self-synthesized. Thus, one major question for future studies is what is the source of NO detected by QS circuits? Furthermore, if exogenous (either environmental or eukaryotic) NO is informing bacterial group decision making, another major open question is why? Finally, as molecules alternative to traditional AIs, including NO and lipids, have now been shown

### REFERENCES


to modulate QS pathways, thus it may be appropriate to expand our understanding of QS from the traditionally understood role of monitoring cell density.

### AUTHOR CONTRIBUTIONS

IH and EB wrote the manuscript.

### FUNDING

This work was supported by the National Science Foundation (grant CHE-1607532 to EB) and the National Institutes of Health (grant GM118894-01A1 to EB).



out of quorum-sensing mode. Genes Dev. 22, 226–238. doi: 10.1101/gad. 1629908


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

Copyright © 2019 Heckler and Boon. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Effect of Co-inhabiting Coagulase Negative Staphylococci on S. aureus agr Quorum Sensing, Host Factor Binding, and Biofilm Formation

Pai Peng<sup>1</sup> , Mara Baldry<sup>1</sup> , Bengt H. Gless<sup>2</sup> , Martin S. Bojer<sup>1</sup> , Carmen Espinosa-Gongora<sup>1</sup> , Sharmin J. Baig<sup>3</sup> , Paal S. Andersen1,3, Christian A. Olsen<sup>2</sup> and Hanne Ingmer<sup>1</sup> \*

<sup>1</sup> Faculty of Health and Medical Sciences, Department of Veterinary and Animal Sciences, University of Copenhagen, Copenhagen, Denmark, <sup>2</sup> Faculty of Health and Medical Sciences, Department of Drug Design and Pharmacology, University of Copenhagen, Copenhagen, Denmark, <sup>3</sup> Department of Bacteria, Parasites and Fungi, Statens Serum Institut, Copenhagen, Denmark

### Edited by:

Ana Maria Otero, University of Santiago de Compostela, Spain

#### Reviewed by:

Iñigo Lasa, Universidad Pública de Navarra, Spain Joanna Nakonieczna, University of Gdansk and Medical ´ University of Gdansk, Poland ´

> \*Correspondence: Hanne Ingmer hi@sund.ku.dk

#### Specialty section:

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

Received: 28 February 2019 Accepted: 10 September 2019 Published: 27 September 2019

#### Citation:

Peng P, Baldry M, Gless BH, Bojer MS, Espinosa-Gongora C, Baig SJ, Andersen PS, Olsen CA and Ingmer H (2019) Effect of Co-inhabiting Coagulase Negative Staphylococci on S. aureus agr Quorum Sensing, Host Factor Binding, and Biofilm Formation. Front. Microbiol. 10:2212. doi: 10.3389/fmicb.2019.02212 Staphylococcus aureus is a commensal colonizer of both humans and animals, but also an opportunistic pathogen responsible for a multitude of diseases. In recent years, colonization of pigs by methicillin resistant S. aureus has become a problem with increasing numbers of humans being infected by livestock strains. In S. aureus colonization and virulence factor expression is controlled by the agr quorum sensing system, which responds to and is activated by self-generated, autoinducing peptides (AIPs). AIPs are also produced by coagulase negative staphylococci (CoNS) commonly found as commensals in both humans and animals, and interestingly, some of these inhibit S. aureus agr activity. Here, we have addressed if cross-communication occurs between S. aureus and CoNS strains isolated from pig nares, and if so, how properties such as host factor binding and biofilm formation are affected. From 25 pig nasal swabs we obtained 54 staphylococcal CoNS isolates belonging to 8 different species. Of these, none were able to induce S. aureus agr as monitored by reporter gene fusions to agr regulated genes but a number of agr-inhibiting species were identified including Staphylococcus hyicus, Staphylococcus simulans, Staphylococcus arlettae, Staphylococcus lentus, and Staphylococcus chromogenes. After establishing that the inhibitory activity was mediated via AgrC, the receptor of AIPs, we synthesized selective AIPs to explore their effect on adhesion of S. aureus to fibronectin, a host factor involved in S. aureus colonization. Here, we found that the CoNS AIPs did not affect adhesion of S. aureus except for strain 8325-4. When individual CoNS strains were co-cultured together with S. aureus we observed variable degrees of biofilm formation which did not correlate with agr interactions. Our results show that multiple CoNS species can be isolated from pig nares and that the majority of these produce AIPs that inhibit S. aureus agr. Further they show that the consequences of the interactions between CoNS and S. aureus are complex and highly strain dependent.

Keywords: Staphylococcus aureus, coagulase-negative staphylococci, colonization, agr, quorum sensing interaction, cross-talk

## INTRODUCTION

fmicb-10-02212 September 27, 2019 Time: 16:11 # 2

Staphylococcus aureus is a common colonizer and opportunistic pathogen of both animals and humans. The increasing spread of antibiotic resistance among S. aureus strains is of major concern in the treatment of staphylococcal infections, with methicillinresistant S. aureus (MRSA) in particular being a proven health risk to humans, causing skin and soft tissue infections, food poisoning, and even fatal systemic disease (Fridkin et al., 2005; Kourbatova et al., 2005; King et al., 2006). MRSA strains are commonly divided into community, hospital or livestock associated and in recent years, the transmission of livestockassociated (LA)-MRSA from animals to humans has become a public health concern particularly in Europe, North America and Asia where pig farming is extensive. Within the EU alone nearly 46% of pigs are colonized by strains of the most predominant LA-MRSA type namely the clonal complex 398 (CC398) (Khanna et al., 2008; Lewis et al., 2008; Van Duijkeren et al., 2008; Authority, 2009; Smith et al., 2009; Golding et al., 2010; Köck et al., 2013; Chuang and Huang, 2015). Studies have revealed a high prevalence of nasal MRSA carriage in pig slaughterhouse workers and pig farmers, indicating that working with MRSAcolonized pigs is the predominant risk factor (Lewis et al., 2008; Van Cleef et al., 2010).

In general, S. aureus colonization is a multifactorial process involving a number of adhesins or host binding proteins that are expressed by, and located on, the surface of the bacterium (Josse et al., 2017). Particularly fibronectin binding proteins have been reported to be important for internalization and uptake of S. aureus by keratinocytes; to be key in the adhesion of S. aureus to keratinocytes of atopic skin and also to contribute to biofilm formation by MRSA strains (Cho et al., 2001; Kintarak et al., 2004; O'Neill et al., 2008; Josse et al., 2017). In addition to colonization factors, S. aureus also expresses a multitude of toxins and other factors necessary for virulence and biofilm formation (Archer et al., 2011; Kobayashi et al., 2015). Production of both adhesins and toxins are controlled by the accessory gene regulator (agr) quorum sensing system with the former being produced at low bacterial cell densities and the latter at high cell densities (Yarwood and Schlievert, 2003). agr is composed of a two component system which senses a selfgenerated autoinducing peptide (AIP) that, by binding to the sensor histidine kinase AgrC, leads to phosphorylation of the AgrA response-regulator and expression of the main agr effector molecule, RNAIII. As cells enter stationary phase, RNAIII is responsible for the down-regulation of host binding proteins such as Protein A encoded by spa and the concomitant upregulation of toxins such as α-hemolysin encoded by hla (Queck et al., 2008; Wang et al., 2014; Le and Otto, 2015). An RNAIIIindependent agr gene regulation pathway also exists, involving AgrA-mediated expression of a family of toxins called the phenol soluble modulins (PSMs) (Periasamy et al., 2012). These PSMs are important players in biofilm formation and dispersal linking agr and biofilm formation (Boles and Horswill, 2008; Periasamy et al., 2012). Interestingly agr varies between S. aureus strains and can be divided into four groups (AgrC-I-IV) where AIPs from the corresponding group lead to self-activation whereas AIPs from other groups lead to cross-inhibition (Otto et al., 2001; Olson et al., 2014; Le and Otto, 2015). This group specificity has lead to an interest in studying the inhibitory activity of non-cognate AIPs as antivirulence sources targeting agr (Canovas et al., 2016; Tal-Gan et al., 2016).

Humans and animals are also colonized with a variety of other staphylococcal species. In contrast to S. aureus they do not produce coagulase and thus are termed the coagulase negative staphylococci (CoNS). Commonly, the CoNS are not pathogens and their presence has been suggested to influence S. aureus colonization. For example, in humans it has been proposed that Staphylococcus epidermidis may prevent colonization by S. aureus (Iwase et al., 2010), whereas in pigs, S. aureus colonization was not observed in the presence of Staphylococcus sciuri, Staphylococcus cohnii, or Staphylococcus saprophyticus (Verstappen et al., 2017). Interestingly, CoNS also encode AIPlike molecules and some of these are able to inhibit S. aureus agr (Otto et al., 1999; Canovas et al., 2016; Gless et al., 2017, 2019; Mahmmod et al., 2018). This cross-talk has been suggested to be involved in the competition between S. aureus and S. epidermidis on the skin (Otto et al., 2001) and in preventing MRSA colonization (Paharik et al., 2017).

The agr-mediated interactions between species isolated from the same host niche remains largely unexplored. In the present study we have examined which staphylococcal species cocolonize the pig nares and assessed the extent to which isolated CoNS strains are able to inhibit S. aureus agr. We have addressed if agr mediated cross-species communication affects S. aureus binding to fibronectin as well as biofilm formation, both elements that may be important for host colonization. Our results suggest extensive cross-communication between CoNS and S. aureus colonizing the same host niche. A better understanding of the role of agr cross-talk between colonizing staphylococci may provide insightful information that can be used for future exploitation in S. aureus colonization interference and anti-virulence therapy.

### MATERIALS AND METHODS

### Bacterial Strains and Growth Conditions

Strains used in this study are listed in **Table 1**. Unless otherwise stated, all bacterial strains were grown in Tryptone Soya Broth (TSB) from Oxoid, at a 1:10 volume/flask ratio, at 37◦C with shaking at 200 rpm.

### Sample Collection, Isolation, and Identification

Nasal swabs (E-Swab, Copan Diagnostics Inc., United States) were collected from the pig nasal cavity of randomly selected pigs (weighing 20–30 kg) at three organic farms in Denmark. It should be noted that no permission is required to sample the nostril of pigs according to the Danish Animal Experimentation Act § 1.2. Samples were sent within 24 h to the Department of Veterinary and Animal Sciences, University of Copenhagen, and analyzed on the day of arrival. In total, 25 samples from 25 pigs were analyzed. Swabs were suspended and diluted in saline solution and plated on SaSelectTM plates (Bio-Rad)



for staphylococcal species isolation. Species identification was carried out by matrix-assisted laser desorption ionization–time of flight mass spectrometry (MALDI-TOF MS) and tuf gene analysis by standard PCR-based methods (Hwang et al., 2011).

### β-Galactosidase Plate Assay

This assay was performed as previously described (Nielsen et al., 2010; Bojer et al., 2017). Briefly, the fused reporter strains PC203 (spa:lacZ), PC322 (hla:lacZ), and SH101F7 (rnaIII:lacZ) all in the 8325-4 strain background, were grown in TSA agar supplemented with 150 µg/mL β-galactosidase substrate 5-bromo-4-chloro-3-indolyl-b-D-galactopyranoside (X-gal) and 5 µg/mL erythromycin. Sixty microliter supernatants of the identified staphylococcal strains or TSB medium were added to premade wells in the plates. The plates were incubated at 37◦C for 10–24 h (the incubation time varies depending on the different reporter strains) until the plates appeared blue.

### β-Lactamase Assay

This method was carried out as previously described with minor modifications (Nielsen et al., 2014; Bojer et al., 2017). Briefly, the reporter strains RN10829/pagrC-I-IV (WT) and RN10829/pagrC-I-R23H (AgrC const.) were treated with a 1/10 volume of CoNS supernatant at OD600 = 0.35, followed by the addition of a 1/10 volume of AIP-I-IV containing supernatant obtained separately from S. aureus strain 8325-4, RN6607, MW2 and RN4850. For the experiment investigating whether CoNS supernatants can induce agr, external AIP-I-IV supernatants were added as activation controls only. Samples were obtained after incubating at 37◦C for 1 h, and optical density at 600 nm was recorded. Samples were stored at −80◦C before thawing to test for β-lactamase activity as described (Bojer et al., 2017). Activity was calculated as arbitrary units based on nitrocefin conversion velocities (Vmax, 1OD486 nm/time) normalized to the sample cell densities.

### Chemical Synthesis of AIPs

All AIPs were synthesized according to a previously reported protocol (Gless et al., 2017). Briefly, linear peptides were synthesized using automated 9-fluorenylmethyloxycarbonyl (Fmoc) solid-phase peptide synthesis (SPPS) on a Gly-ChemMatrix resin loaded with Fmoc-3-amino-4- (methylamino)-benzoic acid (Fmoc-MeDbz-OH). The last residue was incorporated as N-Boc protected amino acid. After SPPS, the MeDbz linker was converted to the N-acylbenzimidazolinone (Nbz) species by treating the resin with 4-nitrophenyl-chloroformate in dichloromethane followed by a solution of i-Pr2NEt in dimethylformamide. The activated Nbz-resin was then treated with a trifluoroacetic acid (TFA) solution to cleave protecting groups and after excessive washing, swelled in cyclization buffer (phosphate buffer, 0.2 M, pH 6.8 in 50% acetonitrile) and incubated at 50◦C for 2 h. The AIP containing solution was separated from the resin and the desired AIP purified by preparative reverse-phase high performance liquid chromatography (RP-HPLC). Full characterization of all synthetic AIPs has been reported previously (Gless et al., 2019). The sequences and quality of the synthetic AIPs can be found in **Supplementary Table S1**.

### Adhesion Assay

This assay was carried out as previously described (Baldry et al., 2016a). Ninety-six well plates were pre-coated with 100 µL/well of 10 µg/mL fibronectin (Fibronectin from human plasma, F2006, Sigma-Aldrich) and incubated for up to 24 h with mild shaking at 4◦C. Respective overnight cultures of S. aureus strains 8325-4, 61599 (CC398 strain), HG001, HG003 and two S. aureus pig isolates (from this study) were diluted 1:100 and grown till OD<sup>600</sup> = 0.5 in fresh TSB medium, after which the bacteria were treated with the synthesized AIPs belonging to Staphylococcus hyicus (10−<sup>4</sup> mM), Staphylococcus simulans (10−<sup>4</sup> mM), and Staphylococcus chromogenes (10−<sup>3</sup> mM) separately, and grown at 37◦C with shaking until OD<sup>600</sup> = 1.7. The concentrations resulting in 100% inhibitory effect on the agr system were chosen according to their IC<sup>50</sup> values. After removing and washing, untreated and treated S. aureus were added to fibronectin-coated wells and incubated statically at 37◦C for 1 h. To avoid the toxic effect of DMSO on bacterial growth, the final solvent concentration of DMSO was maintained at 0.2% (v/v) for all experimental and control cultures. After removing the nonadhered bacteria and washing the wells, the attached bacteria were fixed with 2.5% glutaraldehyde in PBS statically for 1 h at 37◦C. Binding activity of S. aureus was quantified by measuring the OD<sup>570</sup> absorbance of resuspension in 96% ethanol after staining with 0.1% crystal violet at room temperature for 30 min. Arbitrary binding units were calculated by dividing the crystal violet absorption OD by the bacterial cell density of 1.7.

### Static Biofilm Assay

As previously described (Nielsen et al., 2012; Goetz et al., 2017) and with minor modifications, overnight cultures were adjusted to OD600 = 0.2 in TSB and then further diluted 1:100 in 66%TSB supplemented with 0.2% glucose. A total of 200 µL of the bacterial suspensions were added to wells where either S. aureus 8325-4 WT, 8325-41agr (agr<sup>−</sup> strain), CoNS alone, or a 1:1 ratio of S. aureus + CoNS was added. After a 24–30 h incubation period, the medium was removed from each well; the plates were washed and allowed to air dry. Dried biofilms were stained with 125 µL of 0.1% crystal violet solution for 30 min, washed three times with PBS and allowed to dry. To quantify the biofilm formation, the stained biofilm was solubilized in 200 µL of 95% ethanol for 10–15 min and 100 µL were transferred to a new microtiter plate, after which the absorbance was measured at 590 nm. In this assay, three biological replicates were performed with eight technical replicates per experiment. Parallel samples were set for CFU quantification by subsequent plating on SaSelectTM plates (Bio-Rad).

### DNA Sequence Analysis

fmicb-10-02212 September 27, 2019 Time: 16:11 # 4

From purified DNA a sequencing library was generated using Nextera XT (Illumina) followed by (2 × 150 bp) paired-end sequencing on a NextSeq (Illumina) instrument. Genome sequences were de novo assembled using skesa with default settings (Souvorov et al., 2018). From assembled draft genomes the species were identified using the tuf gene, which has previously been described to discriminate between Staphylococcusspecies (McMurray et al., 2016; Strube et al., 2018).

### Statistical Analysis

Where applied, we used a 1-way ANOVA analysis (GraphPad Prism version 7.04 software; GraphPad Software Inc., La Jolla, CA, United States). Differences were considered statistically significant at P < 0.05.

### RESULTS

### Nasal Colonization of S. aureus and Other Staphylococcal Species

To investigate CoNS strains colonizing pig nares we collected nasal swabs from 25 pigs originating from Danish organic pig farms and isolated staphylococcal species on Sa SelectTM plates. In total, 384 isolates were obtained of which 75 were identified by MALDI-TOF MS. tuf gene analysis and genome sequencing were performed to further verify some of the strains (Hwang et al., 2011; Loonen et al., 2012). Of the 75 isolates 21 were identified as S. aureus; corresponding to just over half of the swabs being positive for S. aureus (52%; 13/25 pigs). The remaining 54 isolates were identified and classified into 8 CoNS species originating from 20 of the 25 pigs (**Table 2**). Staphylococcus sciuri (40%) was the most dominant amongst the CoNS isolated, followed by Staphylococcus lentus (24%), Staphylococcus xylosus (24%), S. simulans (20%), S. hyicus (16%), Staphylococcus arlettae (16%), S. chromogenes (8%), and finally Staphylococcus agnetis (4%). These results show that there is substantial variation among pigs with respect to staphylococcal colonization and that they are commonly colonized by more than one species.

## S. aureus Virulence Factor Expression Is Affected by CoNS Strains

Based on previous reports of cross-communication between S. aureus and CoNS strains we hypothesized that S. aureus interacts via agr with the surrounding microbial consortia including the resident CoNS. Therefore, the secreted products of the isolated CoNS strains were screened for their ability to modulate S. aureus agr using a previously established reporter assay. This assay is based on three reporter strains where the promoters of RNAIII, hla and spa, respectively, are fused to lacZ (Nielsen et al., 2010). Upon induction of agr such as is observed during entry into stationary growth phase, promoters of both RNAIII and hla will be induced while that of spa will be repressed. Therefore, after incorporation of the reporter strains together with the LacZ substrate into the agar plates they will become blue when containing the RNAIII and hla reporter strain fusions, but will remain colorless when containing the spa reporter strain after overnight incubation. Conversely, if an agr inhibiting compound has been added to a well in the agar plate reduced expression of hla and RNAIII but increased expression of spa will be observed. As seen in **Figure 1**, the extent to which CoNS supernatants affected the S. aureus agr (agr-I) system varied between species and in some cases even within species. Interestingly, while S. sciuri was the most prevalent species in the swabs, none of the supernatants from isolates of this species affected the S. aureus agr system. In contrast, isolates belonging to S. hyicus, S. simulans, and S. lentus species contained the most isolates with S. aureus agr modulation capabilities. These findings show that CoNS display varying ability to repress the S. aureus agr and that such repression is commonly observed.

### Effect of CoNS Strains on S. aureus agr Groups I–IV

In the agar plate assay (**Figure 1**) we had determined the inhibitory activity of CoNS strains in a S. aurues strain belonging to AgrC group I. To determine if the CoNS AIPs are able to inhibit agr in S. aureus strains carrying the AgrC groups II to IV, and to obtain a quantative measure of the inhibitory effect we employed β-lactamase reporter strains that monitor expression of the RNAIII P3 promoter in cells expressing AgrC groups I to IV. As these strains have been engineered so that they do not produce intrinsic AIPs, induction of agr requires addition of supernatants from strains producing the corresponding AIP group (Nielsen et al., 2014). In this system, the activity of the reporter strains were measured in the presence or absence of cell-free, overnight culture supernatants of our CoNS isolates. Importantly, all the CoNS supernatants that displayed an agr-inhibitory activity in the plate assay also inhibited RNAIII expression in S. aureus strains carrying agr groups I, II, and III, whereas the inhibitory potential against group IV was only marginal (**Figures 2A1– D2**). Further we confirmed the notion that the CoNS AIPs affect S. aureus agr via competitive inhibition of AgrC as we saw little to no inhibition of the P3 promoter in a reporter strain encoding a constitutively active AgrC variant of agr group I that displays kinase activity in the absence of inducing AIP (Geisinger et al., 2009) (**Figures 2E1,E2**).


TABLE 2 | Frequency and presence of eight staphylococcal species isolated from 25 pigs.

Rows indicate the presence of staphylococci in the individual pigs. Columns contain the different staphylococcal species identified.

In addition to inhibition, we were also interested in exploring whether any of the staphylococcal supernatants could induce S. aureus agr activity. To this end we tested the ability of our CoNS isolates to induce S. aureus agr using the same β-lactamase reporters of the S. aureus agr groups, but in this case the staphylococcal supernatants were used as presumptive inducers omitting induction by the cognate AIP. Our results show that none of the CoNS supernatants were capable of activating any of the four S. aureus agr groups (**Supplementary Figure S1**). These results show that CoNS strains interfere with agr induction by competing with the S. aureus AIPs for AgrC binding and that they generally have inhibitory activity toward S. aureus agr.

### Dual Species Biofilm Involving S. aureus and CoNS

As S. aureus agr is known to influence biofilm formation (Le and Otto, 2015), we asked if CoNS strains potentially producing agr repressing peptides affected biofilm formation. When grown individually, the CoNS strains were less robust at forming biofilm than S. aureus (**Figures 3**, **4** and **Supplementary Figure S2**). From these we selected 1-3 CoNS strains from each species to examine biofilm formation in the presence of S. aureus. While the biofilm biomass was quantified by crystal violet staining, bacterial composition of these dual-species biofilms was determined by inspection of CFU on SaSelectTM plates. In all, we tested biofilm formation for eight combinations where the CoNS species had no inhibitory effect on S. aureus agr (**Figure 3**), eight combinations for those CoNS species with a strong S. aureus agr inhibitory effect (**Figure 4**), and another three combinations with CoNS strains with varying agr inhibitory effects (**Supplementary Figure S2**). When examining the composition of dual-species biofilm (**Figures 3B,D**, **4B,D** and **Supplementary Figure S2B**), both species were represented. For 8 out of the 19 dual-species combinations we observed increased biofilm biomass when compared to biofilm formation by individual strains. Importantly, these grouped almost evenly into the agr cross-inhibition group (5 out of 10) and the non-inhibitory group (3 out of 9). While this data already indicated that the increased biofilm in dual species biofilms was independent of agr cross-talk, we sought to consolidate this finding further. For this we chose one strain capable of agr cross-inhibition (S. simulans No. 17) and one strain from the non-inhibitory group (S. sciuri No. 52), and analyzed their effect on biofilm formation of a S. aureus agr mutant strain. These data indicate that the absence of a functional agr in S. aureus did not influence biofilm formation when mixed with CoNs strains (**Figure 5**). Collectively our data show that the

FIGURE 1 | Effect on virulence factor expression of S. aureus by CoNS culture supernatant. TSA agar plates (with erythromycin and X-gal) containing the hla:lacZ (PC322; Ery<sup>r</sup> ) (plates A1–A4), the rnaIII:lacZ (SH101F7; Ery<sup>r</sup> ) (plates B1–B4), or the spa:lacZ (PC203; Ery<sup>r</sup> ) (plates C1–C4) reporter strains of S. aureus were used to screen the cell-free overnight culture supernatants from 55 isolates. Sixty microliter of supernatant or TSB (as a negative control) were added to the wells in the plates. The supernatants in wells are from S. hyicus (wells 1–7 and 51); S. sciuri (wells 8–16 and 52–60); S. simulans (wells 17–21, 23 and 28); S. lentus (wells 24–27 and 29–32); S. arlettae (wells 33–36); S. xylosus (wells 37–43); S. agnetis (well 44); S. chromogenes (well 45); UI shown in well 22 stands for an unidentified isolate. The plates were incubated at 37◦C for 10–24 h (until zones appeared). The assay was performed three times as biological replicates. This figure is representative of one set of screening plates.

FIGURE 4 | Dual species biofilm formation between S. aureus and CoNS strains displaying agr cross-inhibition. For dual-species biofilms, S. aureus 8325-4 (SA) was co-cultured together with one of S. hyicus (Sh, A,B), S. simulans (Ssi, C,D), S. agnetis and S. chromogenes (Sag/Sc, E,F) and biofilm biomass (A,C,E) or CFU (B,D,F) were determined as indicated by mix color bars and compared to biofilms formed by the individual species (SA indicated by gray bars and CoNS by black bars).

presence of both S. aureus and CoNS may in some instances enhance biofilm formation when compared to that formed by the individual strains, but in those cases it is unrelated to agr mediated interactions.

### Strain-Specific Enhancement of S. aureus Adherence to Fibronectin in the Presence of Synthesized CoNS AIPs

As inhibition of S. aureus agr leads to increased expression of surface adhesion proteins recognizing host factors, we were curious to see whether the addition of synthesized AIPs from CoNS would lead to increased binding of S. aureus to host factor. To address this, we studied the fibronectin binding capacity of different S. aureus strains namely the laboratory strain 8325- 4 (CC8), the livestock associated CC398 strain 61599 (Tang et al., 2017b) as well as two S. aureus strains identified from the pig nares together with either S. chromogenes (A) or S. hyicus and S. simulans (B). S. aureus A was classified as CC8 and S. aureus B as CC45. Strains belonging to both CC8 and CC45 have previously been found associated with live stock (Tang et al., 2017a). When these strains were separately treated with synthesized AIPs of S. hyicus, S. simulans, and S. chromogenes, that have been detected in a previous study (Gless et al., 2019), we observed a significant increase in S. aureus 8325-4 binding to fibronectin in the presence of the CoNS AIPs, in comparison to the vehicle (DMSO)-treated control (**Figure 6**). However, neither strain 61599 nor the pig-derived S. aureus isolates obtained in this study and tested here were affected by the presence of CoNS AIPs over the vehicle control. In consideration of the known regulatory defects of S. aureus 8325-4, we also examined the adherence of repaired strains HG001 (restored rsbU, an activator of SigB) and HG003 (restored rsbU and tcaR, an activator of protein A transcription) under the same condition (Herbert et al., 2010) (**Supplementary Figure S3**). Both of these strains showed a low adhesion to fibronectin. Thus, exposure to CoNS AIPs does not lead to increased binding to fibronectin except for strain 8325-4.

### DISCUSSION

Coagulase negative staphylococci comprise a diverse group of staphylococcal species that largely are harmless colonizers of both humans and animals. For a given host several CoNS are commonly present and the composition varies both within and between host species (Nagase et al., 2002). Likewise, pigs have been reported to be colonized by a variety of CoNS with one study describing 10 species including S. hyicus, Staphylococcus haemolyticus, Staphylococcus warneri, S. simulans, S. xylosus, and S. sciuri to be isolated at approximately equal frequency (Nagase et al., 2002). Others document higher CoNS species numbers (between 18 and 20 different CoNS) including the aforementioned, as well as S. saprophyticus and S. cohnii (Schoenfelder et al., 2017; Verstappen et al., 2017). However, both the latter studies report a marked increase in S. sciuri prevalence over the other species amounting to between 30 and 46% of the total colonizing CoNS species (Schoenfelder et al., 2017; Verstappen et al., 2017). In our investigation we also found S. sciuri to be the most prevalent CoNS being isolated from 40% of the pigs followed by S. lentus and S. xylosus. Unlike the Verstappen study though, we did not isolate S. saprophyticus or S. cohnii which were the other most prominent species identified along with S. sciuri (Verstappen et al., 2017).

Interestingly CoNS strains appear to be common producers of AIP molecules that resemble the AIPs of the S. aureus agr quorum sensing system. Analogous to the cross-inhibition of agr that occurs between S. aureus strains belonging to different AgrC subgroups, the CoNS AIPs also tend to inhibit expression of agr controlled genes. Previously we showed that out of 52 staphylococcal isolates obtained from a common

compared to biofilms formed by the individual species (SA indicated by gray bars and CoNS by black bars).

OD590. Results shown are representative of 3 independent experiments. Each bar represents the average of 8 biological replicates and the error bars represent the standard deviation. <sup>∗</sup>P < 0.05; ∗∗∗P < 0.001; ∗∗∗∗P < 0.0001.

strain collection, 37 were capable of inhibiting agr of S. aureus representing 17 different CoNS species (Canovas et al., 2016). Here, we aimed to investigate the extent to which CoNS strains isolated form the same niche environment (i.e., from individual pigs) were able to repress S. aureus agr. Our results show that out of 25 pigs we isolated 8 different CoNS species of which 24 out of 54 strains had quorum quenching properties. Interestingly out of 18 tested S. sciuri strains none were able to repress agr. Similar was observed for S. xylosus whereas for S. lentus, which was present in 24% of the pigs, agr was repressed by some strains but unaffected by others. In contrast, all isolates of both S. hyicus and S. simulans displayed strong agr repressing activity. Using a constitutively active AgrC variant we were able to show that the CoNS strains likely repress agr through production of AIP-like molecules that are secreted to the culture supernatant and compete with S. aureus AIPs for binding to S. aureus AgrC, as opposed to other agr quorum quenching mechanisms such as via AgrA binding (Baldry et al., 2016b). This notion was confirmed by synthesis of selective CoNS AIP molecules. Furthermore, we observed correlation between the inhibitory potential of individual CoNS strains against S. aureus agr group I and the inhibition exerted on groups II and III, while the inhibition pattern was not clearly reflected on group IV. Low inhibitory activity against agr group IV for entities that are highly active against other groups have been described before (Gless et al., 2019). We also examined if any of the CoNS strains were able to induce the S. aureus agr system; however, none of the strains demonstrated such activity. For P. aeruginosa, analogs of quorum sensing molecules have been reported to induce quorum sensing (Smith et al., 2003) and we have observed that synthesized AIPs of S. schleiferi and Staphylococcus hominis are capable of inducing S. aureus AgrC group IV (Gless et al., 2019). However, cross-species induction of agr appears to be a rare phenomenon and the vast majority of agr modulating compounds interfere with quorum sensing induction (Hansen et al., 2018).

Previously it has been suggested that presence or absence of CoNS strains may correlate with S. aureus colonization. For example, Verstappen et al. observed a lower frequency of S. aureus colonization in the presence of S. sciuri, S. cohnii, or S. saprophyticus. We did not identify any S. cohnii or S. saprophyticus in our sampled pigs, and out of the 10 pigs positive for S. sciuri, 4 were co-isolated with S. aureus while 6 were not. However, we did observe that all S. arlettae and S. lentus isolates were colonizing pig nares where S. aureus was not found to be present. To investigate competitive behavior between CoNS strains and S. aureus we performed a series of dual-species biofilm studies based on the notion that agr has been reported to influence both biofilm formation and dispersal (Boles and Horswill, 2008; Periasamy et al., 2012). This rational was also made interesting by the recent observations by Gonzalez et al. that S. epidermidis secreted soluble products (when added to S. aureus cultures) inhibit S. aureus biofilm formation, but when the two species are co-inoculated and grow in physical contact they are capable of forming a robust dual-species biofilm (Gonzalez et al., 2018). Our data corroborate Gonzalez's observations in that we also show robust biofilm formation in the dual-species setting with no evident out-competition of one over the other species. Such dual-species interactions can benefit both species in that they can persist in a colonizing state more robustly, as biofilms are extremely hard to eradicate by both the host and by antimicrobial therapies, and thus also providing a constant reservoir for possible S. aureus chronic infections (Archer et al., 2011). Moreover, even though we have observed the effect of interaction between S. aureus and CoNS on biofilm formation,

no correlation to agr-inhibition was seen. Further studies are needed to better understand the complexity of these interactions.

In context of antibacterial therapy, cross-talk between staphylococci and S. aureus via agr has become a topic of interest. We recently showed that agr inhibition by AIP-like molecules reduces S. aureus induced lesions in an atopic dermatitis model (Baldry et al., 2018) and this was supported by the finding that CoNS strains reduce skin barrier damage by inhibiting production of proteases and phenol-soluble modulins secreted by S. aureus (Williams et al., 2019). Another study reports a synthetic AIP from Staphylococcus caprae that dramatically reduced dermonecrotic injury caused by S. aureus and reduced cutaneous bacterial burden relative to controls (Paharik et al., 2017). However, as inhibition of agr is associated with increased expression of surface adhesion proteins favoring host adhesion and immune evasion (Novick et al., 1993), one could speculate that CoNS strains may increase the ability of S. aureus to colonize. For only one strain of S. aureus, namely 8325-4, we observed a significant increase in adhesion to human fibroncetin in the presence of CoNS AIPs. This was not seen for the livestock associated MRSA strain 61599 belonging to CC398, the two S. aureus isolates obtained from pigs in the present study, nor for the strains HG001 or HG003 that are derivatives of 8325-4 and restored by rsbU or rsbU/tcaR regulatory genes. Thus, we could not consistently demonstrate an effect of CoNS AIPs on S. aureus binding to fibronectin. In conclusion, the interactions between coagulase-negative staphylococci and S. aureus are complex and involve both agr dependent and independent factors, which future studies will be required to elucidate.

### CONCLUSION

We have conducted an investigation of the possible role of the agr cross-talk between S. aureus and CoNS strains

### REFERENCES


isolated from the same colonizing location. We show that there are substantial variations with respect to species colonization amongst the pig hosts tested, as well as in S. aureus agrmodulation capacity of isolated CoNS species. Importantly, our results document multiple interactions between S. aureus and CoNS and they suggest that S. aureus adhesion and dual-species biofilm formation can indeed be influenced by CoNS in both agr-dependent and agr-independent manners.

### DATA AVAILABILITY STATEMENT

All datasets analyzed for this study are included in the manuscript/**Supplementary Files**.

### ETHICS STATEMENT

No permission is required to sample the nostril of pigs according to the Danish Animal Experimentation Act § 1.2.

### AUTHOR CONTRIBUTIONS

PP, MB, MSB, and HI designed the study. PP, BG, SB, and CE-G conducted the experimental work. PP, MB, MSB, HI, CO, BG, PA, CE-G, and SB analyzed the data. PP, MB, MSB, BG, CO, and HI wrote the manuscript.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2019.02212/full#supplementary-material

Staphylococcus aureus," in Antimicrobial Peptides. Methods in Molecular Biology, ed. P. Hansen, (New York, NY: Humana Press), 387–394. doi: 10.1007/978-1-4939-6737-7\_28


peptides," in Proceedings of the National Academy of Sciences, PNAS.



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

Copyright © 2019 Peng, Baldry, Gless, Bojer, Espinosa-Gongora, Baig, Andersen, Olsen and Ingmer. 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.

# Profile of the Intervention Potential of the Phylum Actinobacteria Toward Quorum Sensing and Other Microbial Virulence Strategies

### Hema Bhagavathi Sarveswari and Adline Princy Solomon\*

Quorum Sensing Laboratory, Centre for Research in Infectious Diseases (CRID), School of Chemical and Biotechnology, SASTRA Deemed to be University, Thanjavur, India

The rapid dissemination of antimicrobial resistance amongst microorganisms and their deleterious effect on public health has propelled the exploration of alternative interventions that target microbial virulence rather than viability. In several microorganisms, the expression of virulence factors is controlled by quorum sensing systems. A comprehensive understanding into microbial quorum sensing systems, virulence strategies and pathogenesis has exposed potential targets whose attenuation may alleviate infectious diseases. Such virulence attenuating natural products sourced from the different phyla of bacteria from diverse ecosystems have been identified. In this review, we discuss chemical entities derived from the phylum Actinobacteria that have demonstrated the potential to inhibit microbial biofilms, enzymes, and other virulence factors both in vivo and in vitro. We also review Actinobacteria-derived compounds that can degrade quorum sensing signal molecules, and the genes encoding such molecules. As many Actinobacteria-derived compounds have been translated into pharmaceutically important agents including antibiotics, the identification of virulence attenuating compounds from this phylum exemplifies their significance as a prospective source for anti-virulent drugs.

### Edited by:

Cristina García-Aljaro, University of Barcelona, Spain

### Reviewed by:

Gloria Soberón-Chávez, National Autonomous University of Mexico, Mexico Christopher Milton Mathew Franco, Flinders University, Australia

#### \*Correspondence:

Adline Princy Solomon adlineprinzy@biotech.sastra.edu; adlineprincy@gmail.com

#### Specialty section:

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

Received: 16 April 2019 Accepted: 22 August 2019 Published: 04 October 2019

#### Citation:

Sarveswari HB and Solomon AP (2019) Profile of the Intervention Potential of the Phylum Actinobacteria Toward Quorum Sensing and Other Microbial Virulence Strategies. Front. Microbiol. 10:2073. doi: 10.3389/fmicb.2019.02073 Keywords: anti-biofilm, anti-virulence, microbial natural product, Actinobacteria, anti-pathogenic agents

### INTRODUCTION

Antimicrobials have remained the only mode of prophylaxis and therapeutics for microbial infections since its discovery. In the past century, antimicrobials have undeniably revolutionized clinical practices, laying the foundation for breakthroughs in surgeries, organs transplantations, cancer therapy, treatment of burns and trauma wounds, subsequently improving human health. However, the current antimicrobial resistance (AMR) era threatens the reversal of all breakthroughs achieved thus far (Brown and Wright, 2016; Marston et al., 2016). In the United States alone, AMR contributes to 2 million infections and 23,000 deaths per year, substantially increasing the medical expenses by up to 20 billion US dollars each year (Gelband et al., 2015; Centers for Disease Control, and Prevention, 2017). Healthcare agencies across the world have prioritized AMR, and the scientific community has proposed and developed many innovative strategies including the discovery of novel drug targets and other alternative therapeutic interventions to minimize the development of antimicrobial resistance amongst pathogens (Marston et al., 2016).

Virulence factors produced by pathogens are constructive in deteriorating host fitness during infection. A virulence factor could be a structure, or a product, or a strategy that enables the

pathogen to gain access and survive within the non-colonized region or cellular compartment of the host. Adhesins, enzymes (invasins and internalins), toxins (hemolytic, cytolytic and neurotoxins), and superantigens are some crucial virulence factors expressed by a pathogen to damage the host's physiological condition (Hill, 2012). In several pathogens, the expression of these virulence factors are regulated by a cell density-dependent signaling system called quorum sensing (QS) system (Miller and Bassler, 2001; Fetzner, 2015). QS system enables microorganisms across inter and intraspecies within a community to initiate controlled and co-ordinated behavior (Greenberg, 2003; Kaufmann et al., 2008). Although many facets of the intricate prokaryotic QS system remain undeciphered, the available knowledge on the domain's diverse QS systems provides many targets for the development of drugs that could inhibit the expression of virulence. Given the unrelatedness of virulence to the viability of a pathogen, the cultivation of resistance toward the anti-virulence agent through selective pressure is presumed to be non-existential (Clatworthy et al., 2007).

What is so paramount in the evolution of bacteria is the co-development of secondary metabolites that can disrupt the QS signal molecules and attenuate the virulence of other microorganisms. The competency to disrupt the QS signal molecules [quorum quenching (QQ)] could have been evolved in quorum sensing bacteria to remove or repurpose its own QS signal molecules, or the signal molecules of microorganisms that co-inhabit a competitive environment (Grandclément et al., 2016). Bacteria could have also evolved molecules for degrading N-acyl homoserine lactone (AHL) to utilize AHL as a sole source of carbon and nitrogen, or as armor against antibiotic-producing bacteria (Gonzalez and Keshavan, 2006).

Since the discovery of the AHL degrading enzyme AhlD (acyl homoserine lactone degradation enzyme) from Arthrobacter sp. IBN 110 (Park et al., 2003), and the demonstration of the attenuation of Erwinia carotovora pathogenesis in transgenic plants expressing autoinducer inactivating aiiA gene from Bacillus sp. (Dong et al., 2001), an array of bacterial natural components with quorum quenching properties have been reported. These include secondary metabolites produced by bacteria from various phyla including Actinobacteria, Bacteroidetes, Firmicutes, Proteobacteria, and Cyanobacteria. In this article, we review the natural compounds from the phylum Actinobacteria that have been reported to produce AHL degrading enzymes, the corresponding genes, and other Actinobacteria derived compounds that inhibits or attenuates microbial virulence both in vitro and in vivo (**Tables 1**, **2**).

### FAMILY: Micrococcaceae; GENUS: Arthrobacter

Arthrobacter was one of the first genera in the phylum Actinobacteria reported to possess a gene dedicated to the degradation of AHL. Arthrobacter sp. IBN110 demonstrated the potential to degrade AHLs of different lengths and acyl side chains including N-3-oxohexanoyl-L-homoserine lactone (OHHL), N-octanoyl-L-homoserine lactone (OHL), and N-3-oxododecanoyl-L-homoserine lactone (OdDHL) (Park et al., 2003). When OHHL producing Erwinia carotovora N98 was co-cultured with strain IBN110, the concentration of OHHL and OHHL mediated pectate lyase activity significantly reduced, indicating the potential of IBN110 to disrupt AHL. Indeed, the strain IBN110 possessed acyl homoserine lactone degradation gene (ahlD) that encoded AhlD protein with characteristic zinc-binding motif HXDH≈H≈D crucial for N-acyl homoserine lactonase (AHLase) activity (Dong et al., 2002). HPLC and mass spectrometry analysis revealed that AhlD hydrolyzed the ester bond in N-acyl homoserine lactone molecules and released the homoserine lactone ring. Multiple sequence alignment of AhlD with the other known AHLases, including AttM and AiiA revealed < 26% overall sequence similarity (Park et al., 2003).

Arthrobacter sp. PGVB1 derived arthroamide and turnagainolide A (cyclic depsipeptides) demonstrated the ability to inhibit agr signaling in a Staphylococcus aureus agr reporter strain. At 5–10 µM concentrations, the compounds suppressed the expression of the agr-dependent gene without cytotoxicity. The inhibitory concentration value (IC50) of arthroamide and turnagainolide A against Staphylococcus aureus agr reporter strain was 0.3 and 0.8 µM, respectively (Igarashi et al., 2015). The Arthrobacter B4-EPS1 exopolysaccharide abolished Pseudomonas aeruginosa biofilms at a lower concentration (about 86.1% at 50 µg/mL) than the exopolysaccharides reported from other genera (Li et al., 2015) including EPS (exopolysaccharide) from Streptococcus phocae PI80 (about 20% inhibition at 1 mg/mL) (Kanmani et al., 2011), r-EPS (released-exopolysaccharide) from Lactobacillus acidophilus A4 (about 80% inhibition at 1 mg/mL) (Kim et al., 2009), and A101 from Vibrio sp. QY101 (about 75% inhibition at 100 µg/mL) (Jiang et al., 2011). The B4-EPS1 exopolysaccharide also expressed broad-spectrum inhibitory activity against the Staphylococcus epidermidis, Enterococcus faecium, Klebsiella pneumonia, Escherichia coli, and Morganella morganii biofilms (Li et al., 2015). Dex410, a dextranase from marine Arthrobacter sp. strain (Arth410) inhibited biofilms of Streptococcus mutans with minimum biofilm inhibitory concentration (MBIC50) ranging between 1.27 and 6.35 µM/ml. Dex410 also reduced the 24 h biofilms of Streptococcus mutans with minimum biofilm reduction value (MBRC50) of 3.81– 8.89 µM/ml. This concentration was significantly lesser than the antibacterial chlorhexidine (MBRC<sup>50</sup> > 20 µM) present in the commercially available oral care products. The animal experiment showed that long term usage of Dex410 effectively prevented dental caries (Jiao et al., 2014). Arthrobacter oxydans KQ11-1 derived dextranase displayed MBIC<sup>50</sup> and MBIC<sup>90</sup> values of 2 U/m1 and 6 U/ml, respectively toward Streptococcus mutans biofilm. The MBRC<sup>50</sup> against preformed Streptococcus mutans biofilm was 5 U/ml and the dextranase decreased the thickness of the biofilm up to 36.67 µm (Wang et al., 2016).

### FAMILY: Brevibacteriaceae; GENUS: Brevibacterium

In 1959, when Grecz and his team reported the inhibitory activity of culture filtrates of Brevibacterium linens ATCC 9174 and



(Continued)


(Continued)

#### TABLE 1 | Continued


Brevibacterium linens ATCC 9175 toward the germination of Clostridium botulinum type A spores, little did they know that it was one of the earliest reports of anti-infective property ever reported from the Genus Brevibacterium (Grecz et al., 1959). In fact, it was only during the mid 2000s that the evidence of quorum sensing in Clostridium botulinum and its role in regulating the germination of botulinum spores was established (Zhao et al., 2006). Today, out of the 51 known species of Brevibacterium<sup>1</sup> only two strains from Brevibacterium casei (Brevibacterium casei MSA19 and MS104), both interestingly isolated from the marine sponge Dendrilla nigra, have been reported to produce compounds with anti-virulence property against bacterial pathogens (Kiran et al., 2010, 2016; **Table 1**).

At a concentration of 30 µg/ml, Brevibacterium casei MSA19 glycolipid affected the formation of biofilm by inhibiting the initial attachment of the bacteria mediated by pili and flagella. At a very low concentration, the Brevibacterium glycolipid significantly reduced the formation of both individual and mixed bacterial biofilms (Kiran et al., 2010; **Table 1**). Microtiter plate assay and CLSM images revealed that polyhydroxy butyrate (PHB) derived from Brevibacterium casei MSI04 suppressed the adhesion of pathogenic Vibrio species on both polystyrene

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

#### TABLE 2 | Extract from Actinobacteria displaying anti-virulence activity.


and glass surfaces at a concentration of 0.6 mg (200 µl). In fact, the PHB was most effective in inhibiting the formation of biofilm than dislodging pre-formed biofilm (Kiran et al., 2014). At 50 µg/ml concentration, PHB inhibited bioluminescence, and at 150 µg/ml reduced the formation of Vibrio campbellii PUGSK8 biofilm. Infection of Vibrio species in brine shrimp (Artemia sp.) is typically fatal, and, treatment of ≥ 50 µg/ml of PHB resulted in the elicitation of protection to shrimps up to 48 h. This research revealed that the ß-hydroxy butyric acid, an intermediate released during the PHB degradation indeed regulates the expression of the virulence factors in PUGSK8 (Kiran et al., 2016).

### FAMILY: Mycobacteriaceae; GENUS: Mycobacterium

The discovery of AHL lactonases in Mycobacterium was an outcome of exploration for the establishment of promiscuity of the divergence of bacterial phosphotriesterase (PTE), an

enzyme first discovered in Pseudomonas diminuta with efficient paraoxonase activity (Raushel and Holden, 2000; Roodveldt and Tawfik, 2005). The absence of naturally occurring specific substrate and the evolutionary elusiveness of PTE led to a BLAST search for genes homologs to Pseudomonas diminuta PTE. Three genes including two from the phylum Actinobacteria; PPH (putative parathion hydrolase) in Mycobacterium tuberculosis and AhlA (N-acyl-homoserine lactone acylase) in Rhodococcus erythropolis sharing a 34 and 28% identity and SsoPox (phosphotriesterase with natural lactonase activity) from an archeon Sulfolobus solfataricus with 31% identity were identified (Afriat et al., 2006). The PPH and AhlA have been classified as phosphotriesterase-like lactonase (PLL) from the amidohydrolase superfamily that hydrolysis substrates with either ester or amide functional groups at phosphorus and carbon centers (Seibert and Raushel, 2005). A subsequent exploration into the enzymology of PPH and AhlA revealed that the paraoxonases activity was rather a promiscuous function that could have emerged in PLLs from its progenitor lactonase activity (Afriat et al., 2006).

Expression of PPH gene in Escherichia coli in the presence of three metal ions (Zn2+, Co2<sup>+</sup> and Mn2+) prompted a 2000 fold increase in PPH's lactonase activity than the paraoxonase activity. Further research revealed that these metal ions were vital for PPH's enzymatic activity and that metal chelation inactivated PPH. The K<sup>M</sup> and kcat/K<sup>M</sup> values of PPH during the hydrolyzes of lactones ranged between e20 and 230 µM, and from 1.4 × 10<sup>4</sup> to 5 × 10<sup>5</sup> s <sup>−</sup><sup>1</sup> M−<sup>1</sup> , respectively. The kcat/K<sup>M</sup> values generally increased with six membered lactone ring and lactones with longer and more hydrophobic side chains. However, no visible lactonase activity against N-acyl thiolactone analog derived from homocysteine was observed (Afriat et al., 2006). Another orthologous of PLL, MCP (AHL lactonase from Mycobacterium avium subsp. paratuberculosis K-10), also degraded a wide range of AHLs and displayed up to 92% sequence similarity with PPH. MCP also demonstrated low paraoxonase activity indicating that the naturally occurring substrate for MCP does not contain phosphate esters. Introduction of a single point mutation in ßα loop at the carboxyl-terminal end of eighth β-strand of the MCP resulted in a mutant (N266Y) with enhanced AHL lactonase activity than the wild type MCP. The N266Y mutant (substitution of TAC for AAC at 266 codon) increased the kcat/K<sup>M</sup> values up to 4 to 32-fold for C12-HSL and C6-HSL than the wild type. Further research with the mutants including the N266 showed that a suitable amino acid substitution at the 266 residue, and its proximity to the lactone ring of AHL provide the possibility to enhance AHL lactonase activity by introducing an AHL binding geometry (Chow et al., 2009).

### FAMILY: Microbacteriaceae; GENUS: Microbacterium

Several strains of Microbacterium species isolated from potato tuber plant (Solanum tuberosum) have been reported to degrade AHLs with both short and long acyl side chains (Morohoshi et al., 2009; Wang et al., 2010, 2012). An infestation of Pectobacterium carotovorum subsp. carotovorum in potato crop results in soft rot disease, a consequence of coordinated expression of virulence factors mediated by QS signal molecule N-(3 oxohexanoyl)-L-homoserine lactone (Chatterjee et al., 1995). Two endophytic strains: Microbacterium testaceum StLB018 and Microbacterium testaceum StLB037 attenuated virulence in Pectobacterium carotovorum subsp. carotovorum NBRC 3830 without bactericidal activity (Morohoshi et al., 2009). Nucleotide sequence analysis of StLB037 revealed a complete open reading frame encoding a protein of 295 amino acids that belonged to α/ßhydrolase fold family encompassing the characteristic catalytic active site Gly-X-Ser-X-Gly (Holmquist, 2000). Named as autoinducer inactivation gene from Microbacterium testaceum (aiiM), the expression of StLB037 AiiM protein in the NBRC 3830, drastically reduced the pectinase production and also attenuated tissue maceration non-bactericidally (Morohoshi et al., 2009). HPLC analysis with fraction containing maltose binding protein-AiiM (MBP-AiiM) fusion protein and C10-HSL produced two peaks that coordinated with the standards of C10- HSL, and the opened lactone ring of C10-HSL. As this established the role of AiiM in degrading AHL, further study revealed that AiiM was not influenced by the length or the substitution of the acyl side chains. The partially purified MBP-AiiM protein exhibited relatively better activity against C12-HSL and 3-oxosubstituted AHLs than C6-HSL, C8-HSL, C10-HSL and other unsubstituted AHLs (Wang et al., 2010).

Investigation into the distribution and diversity of AiiM among the Genus Microbacterium with various strains isolated from different sources including potato plant, scarlet runner bean, rapeseed, Chinese paddy, milk, cheese, air, soil, activated sludge, imperial moth and many more, exposed that the superior level of AHL degradation exhibited by the Microbacterium strains was due to the presence of aiiM gene encoded in the chromosome of bacterium. Out of 26 Microbacterium strains included in the study, only 9 strains exhibited high degrading ability against C6-HSL, 3OC6-HSL, C10-HSL, and 3OC10-HSL. Remarkably, these strains were of potato plant origin and were positive for aiiM gene in their genetic material. The remaining 17 strains lacked the ability to degrade C6-HSL and exhibited low to relatively intermediate degrading ability against 3OC6- HSL, C10-HSL, and 3OC10-HSL. These strains were of nonpotato origin and were negative for aiiM gene in their DNA (Wang et al., 2012). Comparison of the nine aiiM positive strains with phylogenetically related Microbacterium strains (Microbacterium sp. PcRB024 and M. testaceum ATCC 15829) revealed the absence of significant AHL degrading activity or the aiiM gene in the chromosome. The aforementioned evidence led to a conclusion that the aiiM was not conserved among the Genus Microbacterium and could have spread amongst the Microbacterium strains inhabiting potato tuber ecosystem through the non-horizontal mode of transmission supposedly due to the absence of transposons flaking the aiiM (Wang et al., 2012). Although Microbacterium testaceum aiiM homologous gene with high sequence similarities have been identified in other actinobacterial strains including Rhodococcus erythropolis PR4 and Rhodococcus opacus B-4, their expression as MBP-AiiM protein lacked AHL lactonase activity. The Microbacterium

StLB037 encoded AiiM bears < 15% similarity with other known AHL lactonases including AidP, AiiA, AttM, AhlD, QsdA, QlcA, BpiB01, BpiB04 and BpiB07. The absence of conserved zincbinding domains found in AHL lactonases from metallo-ßlactamase super family and PTE family proteins affirmed the novelty and ingenuity of AiiM (Wang et al., 2010).

### FAMILY: Nocardiopsaceae; GENUS: Nocardiopsis

The culture supernatant of cold temperature adapted Nocardiopsis sp. A731, at a concentration of 20% (v/v) inhibited about 80% of V. cholerae biofilm (Augustine et al., 2012). Three novel α-pyrones; nocapyrone H **(1)**, nocapyrone I **(2)**, and nocapyrone M **(3)** (**Tables 1**, **3**), were extracted from Nocardiopsis dassonvillei subsp. dassonvillei XG-8-1 inhibited QS controlled virulence in P. aeruginosa QSIS-lasI biosensor and Chromobacterium violaceum CV026 at a concentration of 100 µg/mL (Fu et al., 2013). At 200 µg/ml concentration, the crude extract of Nocardiopsis sp. ZoA1 inhibited the formation of Staphylococcus haemolyticus 41 and Staphylococcus capitis 267 biofilms by ≥ 90% (**Table 2**). Dose-dependent biofilm inhibition assay with ZoA1extract supported the assumption that inhibition of multidrug-resistant coagulase negative staphylococci (CONS) was due to the inhibition of production of proteinaceous factors and exopolysaccharide. However, ZoA1 strain also possessed broad-spectrum antibacterial activity against Staphylococcus aureus, Bacillus subtilis, Salmonella typhi, and Vibrio cholerae (Sabu et al., 2017). The spent medium of soil Nocardiopsis sp. TRM 46200 showed ≥ 90% inhibition against the Staphylococcus epidermidis biofilms for over 24 h. The major metabolite in the culture supernatant was proteinous in nature and exhibited both antibiofilm and protease activity. The crude protein derived from TRM 46200 reduced the cell surface hydrophobicity, and also degrade DNA and the extracellular polymeric substance (EPS) of Staphylococcus epidermidis strains (ATCC 35984 and 5-121-2) (Xie et al., 2018). The culture supernatant of Nocardiopsis sp. GKU 213 inhibited biofilm formation of Staphylococcus aureus ATCC 25923 by 60% without anti-bacterial activity (Leetanasaksakul and Thamchaipenet, 2018). Zinc oxide nanosheets (ZnO NSs) produced by Nocardiopsis sp. GRG1 (KT23540) effectively inhibited the biofilms of multi-drug resistant Proteus mirabilis BDUMS1 and Escherichia coli BDUMS3 by 92 and 90%, at 20 µg/ml concentration, respectively. CLSM images and fluorescent light microscopic analysis showed that ZnO NSs disintegrated the biofilm architecture of uropathogens, by dispersing the bacterial cells leaving only fewer adherent cells and cell aggregates (Rajivgandhi et al., 2018).

### FAMILY: Nocardiceae; GENUS: Rhodococcus

While the possible presence of γ-butyrolactone dependent quorum sensing system in Rhodococcus species could be understood only by in silico genomic analysis of Rhodococcus



(Continued)



erythropolis PR4 and Rhodococcus strain RHA1, the quorum quenching mechanism of this genera is one of the wellestablished among bacteria (Wuster and Babu, 2007; Latour et al., 2013). Indeed, Rhodococcus sp. is a unique organism possessing three different mechanisms for N-acyl homoserine lactone degradation; an AHL lactonase, an oxidoreductase and an amidase (Uroz et al., 2003, 2005, 2008, 2009; Park et al., 2006), unraveling the unprecedented evolution of multiple QQ strategies within a bacterium.

In 2003, Uroz and his team demonstrated that a 'wild type' Rhodococcus erythropolis W2 can degrade C6-HSL, and attenuate the QS-regulated pathogenesis in Pectobacterium carotovorum subsp. carotovorum, a pathogen of potato tubers, without limiting or inhibiting its growth (Uroz et al., 2003). Although primarily identified on the basis of its ability to utilize 3-oxo-C6 HSL, the Rhodococcus erythropolis W2 interestingly degraded the 3-oxo derivative of acyl homoserine lactone less efficiently than the other known AHL degrading bacteria (Leadbetter and Greenberg, 2000). The broad substrate specificity, rapid AHL inactivation and interference with QS regulated pathogenesis exhibited by Rhodococcus erythropolis W2, instigated a series of studies to understand the underlying catabolic mechanism involved in AHL degradation (Uroz et al., 2003, 2005, 2009). Incubation of N-(3-oxooctanoyl)-L-homoserine lactone (3O,C8- HSL), N-(3-oxodecanoyl)-L-homoserine lactone (3O,C10-HSL), N-(3-oxododecanoyl)-L-homoserine lactone (3O,C12-HSL), N-(3-oxotetradecanoyl)-L-homoserine lactone (3O,C14-HSL) with whole cells of W2 in phosphate buffer saline resulted in the production of 3-hydroxy derivatives: 3OH,C8-HSL, 3OH,C10-HSL, 3OH,C12-HSL and 3OH,C14-HSL, respectively. This reaction was mediated by oxidoreductase activity (Uroz et al., 2005). The broad substrate specificity of oxidoreductase also catalyzed the reduction of AHL derivatives substituent with aromatic acyl side chains or without lactone ring including N-(3-oxo-6-phenylhexanoyl) homoserine lactone and 3-oxododecanamide, respectively (Uroz et al., 2005).

Interestingly, the oxidoreductase activity observed in the whole cell of Rhodococcus erythropolis W2 was absent in the culture extract. The complete elimination of unsubstituted and substituted (3-oxo or 3-hydroxy) AHLs from the incubation medium containing the culture extract of W2, suggested the presence of another mechanism to degrade AHL. This was later validated to be an acylase that catalyzed AHL degradation by releasing dansylated homoserine lactone from the incubated reaction mixture of N-(3-oxodecanoyl)-L-homoserine lactone and cell culture extract of W2 strain. The AHL acylase cleaved the amide bond of both short and long chain AHLs yielding homoserine lactone through amidolytic activity (Uroz et al., 2005).

Identification of a soil bacterium that displayed the potential to utilize AHL led to the discovery of AHL lactonase, the third mechanism for the catabolism of AHL in Rhodococcus species (Park et al., 2006). Two strains of Rhodococcus sp. LS31 and PI33 displayed different substrate specificity for N-3-oxo-hexanoyl-L-homoserine lactone (OHHL), and mass spectrometric analysis revealed that both the strains hydrolyzed the lactone ring of AHL (Park et al., 2006). Rhodococcus sp. strain LS31 degraded AHL of different lengths with different acyl side chain substitutions, contradicting the higher degrading activity exhibited by Rhodococcus erythropolis W2 against

3-oxo-substituent AHLs than unsubstituted AHLs (Uroz et al., 2003, 2005). The AHL lactonase from both the strains LS31 and PI33 destroyed AHL, while the R. erythropolis W2 attenuated the signal molecules (Park et al., 2006). Although much of the enzymology underlying Rhodococcus AHL acylase and AHL oxidoreductase has been unraveled, the genetic determinant of these enzymes still remains unknown.

QsdA, a product of the gene qsdA (quorum sensing signal degradation), was reported as the another AHL lactonase utilized by Rhodococcus erythropolis strain W2 to degrade AHL. This novel class of AHL lactonase did not show homology to any previously reported AHL degrading enzymes that were characterized from the two protein super families: Zincdependent glyoxylase and N-AHSL amidohydrolases of the β lactam acylases (Uroz et al., 2008). In fact, the QsdA belonged to the group of phosphotriesterase (PTE) like lactonase (PLL) within the amidohydrolase superfamily (Hawwa et al., 2009) that possessed the characteristic binuclear metal center inside a TIM- barrel (β/α)<sup>δ</sup> - barrel-shaped scaffold). Though initially this enzyme was described as paraoxonases due to their activity against organophosphate pesticide paraoxon (Afriat et al., 2006), later experiments showed that the enzymes also hydrolyzed lactones including the N-acyl homoserine lactones with 6 to 14 carbon in acyl side chains, irrespective of carbon 3 substitution (Uroz et al., 2008). The qsdA operon can also be utilized for the assimilation of various lactone in the milieu including the γlactone, and also for the disruption of QS signals of competitive bacteria (Latour et al., 2013). The qsdA homologue is conserved in reference strains including, Rhodococcus erythropolis DCL14 (de Carvalho and da Fonseca, 2005) and it was suggested that the detection of AHL signals or the γ- capro lactones in the environment can lead to the transcription of qsdA within qsd operon (Barbey et al., 2012, 2013). A putative transcriptional regulator homologous to TetR (QsdR) had been reported upstream of qsd operon (Latour et al., 2013), which, in the absence of AHL could bind to the promoter inhibiting the expression of qsdA. In the presence of AHL or γ-butyro lactones, the QsdR might undergo conformational change leading to the transcription of the gene qsdA (Cuthbertson and Nodwell, 2013; Barbey et al., 2018).

Attenuation of QS-regulated pathogenesis in Pectobacterium carotovorum subsp. carotovorum, a pathogen of Solanum tuberosum (potato tubers), by rhizosphere soil Rhodococcus erythropolis W2 illustrates the interaction between a QS producer, a QQ producer, and their plant host. The treatment of rhizosphere soil of potato plant with growth stimulator such as gamma- caprolactone (GCL), provoked the growth of native AHL degrading strains especially Rhodococcus erythropolis (Cirou et al., 2007, 2011). Another study with Rhodococcus sp. R138 isolated from GCL treated potato rhizosphere soil exhibited strong biocontrol activity in potato tuber assay by degrading AHL and through assimilating GCL (Cirou et al., 2011). Rhodococcus erythropolis not only increased its population in response to GCL (a natural plant molecule) but also assimilated GCL, a reaction proposed to have been catalyzed by QsdA and other rhodococcal enzymes (Cirou et al., 2012). Drastic reduction in AHL mediated virulence of Pectobacterium atrospeticum by Rhodococcus erythropolis was identified by transcriptome analysis (Kwasiborski et al., 2015). Rhodococcus sp. BH4 encapsulated within free moving alginate cell trapping beads (CEBs) quenched AHL and reduced the synthesis of extracellular matrix of biofilmforming microbial cells in membrane bioreactors. This property of quenching AHL by strain BH4, in combination with the physical friction exerted by alginate beads, has been proposed as prospective model for controlling biofouling (Kim et al., 2013).

### FAMILY: Streptomycetaceae; GENUS: Streptomyces

An AHL acylase termed as AhlM (N-acyl homoserine lactone acylase) derived from Streptomyces sp. strain M664 was the first AHL degrading enzyme characterized from the genera Streptomyces (Park et al., 2005). Discovered based on its potential to obstruct N-acyl homoserine lactone facilitated violacein production, the AHL acylase catalyzed the hydrolysis of an amide bond between homoserine lactone and acyl side chain in AHL. The active enzyme was composed of 804 amino acids that were arranged in a pattern characteristic of a penicillin acylase class of proteins belonging to Ntn hydrolase superfamily. Amino acid sequence analysis of AhlM with known AHL acylases: AiiD from Ralstonia strain XJ12B (Lin et al., 2003) and PvdQ from Pseudomonas aeruginosa (Huang et al., 2003) displayed < 35% sequence identity. Apart from the acylase activity, the AhlM also displayed deacylation activity against long acyl chain AHLs and was suggested of possessing the ability to degrade cyclic lipopeptides. At a concentration of 20 µg/ml, AhlM significantly reduced the production of elastase, total protease, and Las A protease in P. aeruginosa PAO1 (Park et al., 2005).

A metabolite phenylalanyl-ureido-citrullinyl-valinylcycloarginal termed as FA-70C1 **(4)** (**Tables 1**, **3**) isolated from Streptomyces species FA-70, strongly inhibited arggingipain (Rgp), an enzyme crucial for survival and proliferation of Porphyromonas gingivalis both in vitro and in vivo (Kadowaki et al., 1998, 2003).

Guadinomines A **(5)** and B **(6)** (**Table 3**) derived from Streptomyces K01-0509 showed dose-dependent inhibitory activity against hemolysis caused by enteropathogenic Escherichia coli (EPEC), potentially through the inhibition of type III secretion system. The inhibitory concentration (IC50) value of guadinomine B and guadinomine A was 0.007 mg/ml and 0.02 mg/ml, respectively (Iwatsuki et al., 2008). Piericidin A1 **(7)**, a major metabolite of Streptomyces sp. TOHO-Y209 and TOHO-O348, displayed an IC<sup>50</sup> value of 10 µg/ml against violacein production by C. violaceum CV026. 3<sup>0</sup> -rhamnopiericidin A1 **(8)**, and piericidin E **(9)** also expressed QSI activity but much lesser than piericidin A1 (Ooka et al., 2013). Alnumycin D **(10)**, a C-ribosylated pathway shunt product isolated from recombinant strain Streptomyces albus, effectively inhibited the biofilm and planktonic cells of Staphylococcus aureus ATCC 25923 by 12 to 22-fold higher than alnumycin A. Similarly, granaticin B, a polyketide metabolite from Streptomyces violaceoruber, could disrupt pre-formed staphylococcal biofilms. The structural similarities observed between the two compounds, including

glycosylation at the C-8 position with ribopyranosyl unit in alnumycin D and the aglycone unit through C–C bond at C-7 and C-8 positions in granaticin B, were suggested to have contributed to the biofilm inhibitory activity. In addition to this, the oxygenation pattern within the naphthoquinone ring, carbonyl oxygen atom in alnumycin D and hydroxyl group in granaticin B, were also suggested to have contributed to the anti-biofilm activity (Oja et al., 2015).

Well studied for its role in suppressing (Tzaridis et al., 2016) and treating tumors (Walsh et al., 2016; Das et al., 2017; Schmidt et al., 2017; Lamture et al., 2018), actinomycin D from Streptomyces parvulus also possessed biofilm inhibitory activity in vitro. At 0.1 µg/ml concentration, actinomycin D reduced the formation of biofilm of methicillin sensitive Staphylococcus aureus strains (ATCC 25923 and ATCC 6538) and methicillin resistant Staphylococcus aureus strain (ATCC 33591) by ≥ 70%, ≥ 80%, and ≥ 80%, respectively (Lee et al., 2016). At the same concentration, actinomycin D reduced the biomass and mean thickness of Staphylococcus aureus biofilm by 98%, and the hemolytic activity by ≥ 85%. This led to the suggestion that the inhibitory activity of actinomycin D toward Staphylococcus aureus was partly concatenated with its ability to inhibit hemolysis. Besides, Streptomyces parvulus derived actinomycin D also reduced the hydrophobicity of the staphylococcal cells, a property crucial for the bacterial adherence to the substrata (Krasowska and Sigler, 2014). The failure of the actinomycin D to disperse preformed staphylococcal biofilms highlighted the non-association of actinomycin D with protease or the staphylococcal agr QS system (Lee et al., 2016). Conversely, actinomycin D derived from Streptomyces parvulus HY026 significantly reduced the production of violacein by C. violaceum up to 90.7% at 50 µg/ml concentration. Although the potential of actinomycin D from endophytic Streptomyces parvulus (1% (v/v) concentration) to inhibit staphylococcal biofilms does seem to be more superior than the actinomycin D from Streptomyces parvulus HY026 (10% v/v concentration), the non-agr QS mediated mode of biofilm inhibition by the former strain and anti-QS activity of actinomycin D from HY026 exemplifies the outstanding functional adaptation of actinomycin D at molecular level (Miao et al., 2017; **Table 1**).

Streptomyces coelicoflavus S17 derived 1H-pyrrole-2 carboxylic acid **(11)** and docosanoic acid **(12)** (**Table 3**) significantly attenuated the virulence of P. aeruginosa PAO1 at 1 mg/ml concentration. While 1H-pyrrole-2-carboxylic acid decreased the production of elastase, protease, and pyocyanin by 96, 74, and 44%, respectively, the docosanoic acid reduced their production by 91.8, 46.1, and 64.45%, respectively. The compound 1H-pyrrole-2-carboxylic acid eliminated the expression of las genes; lasA, lasB, lasI and lasR by 88, 92, 80, and 87%, respectively. The compound also inhibited rhl/pqs cascade including pqsA, pqsR, rhlI and rhlR by 97, 78, 69, and 89%, respectively (Hassan et al., 2016). All maniwamycins from Streptomyces TOHO-M025 reduced the production of violacein by C. violaceum CV026 in a dose-dependent manner at a concentration ranging from 0.01 to 1 mg/ml. Maniwamycins D **(13)** and E **(14)** displayed higher QS inhibitory activity than C **(15)**

and F. Maniwamycin E showed IC<sup>50</sup> value of 0.12 mg/ml (Fukumoto et al., 2016).

Quercetin **(16)** from marine Streptomyces fradiae PE7 reduced the germination of Anabaena and Nostoc sp. spores by 70% at 100 µg/ml concentration (Gopikrishnan et al., 2016). The addition of culture extract from Streptomyces xanthocidicus KPP01532 (≥ 2.5 µL), reduced the violacein production by CV026 considerably. Transcriptomic analysis on the effect of purified piericidin A **(17)** and glucopiericidin A **(18)** from the KPP01532 media extract on E. carotovora subsp. atroseptica revealed that the reduction in the expression of genes encoding hydrolytic enzymes including pectate lyase (PelC), cellulase (CelV), polygalacturonase (PehA) and QS controlled virulence-associated gene (nip). Treatment of potato tubers with 50 and 100 µM of piericidin A also reduced the development of soft rot disease symptoms. Similar results were also obtained in vitro with KPP01532 glucopiericidin A (Kang et al., 2016).

Hygrocin C (an ansamycin) derived from Streptomyces sp. SCSGAA0027 displayed a biofilm inhibitory concentration (BIC80) value of 12.5 µg/ml, 25.0 µg/ml and 200 µg/ml against Bacillus amyloliquefaciens, Staphylococcus aureus and P. aeruginosa, respectively. At a dosage of 12.5 to 100 µg/ml, hygrocin C reduced pre-formed biofilms of Bacillus amyloliquefaciens by 11.73 to 54.76%. Transcriptomic analysis showed that in the presence of hygrocin C, 107 genes were upregulated, and 102 genes were downregulated. While the downregulated genes were crucial for motility including FliC and FliA (Flagellar genes), MotB (Flagellar motor protein) and two-component systems including ResE (Sensor histidine kinase ResE) and CydB (Cytochrome-bd-ubiquinol oxidase), the upregulated genes led to the mass synthesis of arginine and histidine. The unbalanced level of histidine and arginine, and the downregulation of genes essential for motility were suggested to have contributed to the repression of biofilm formation. It was also suggested that the suppression of bacteria's survival was due to the downregulation of nitric oxide dioxygenase (HmpA) (Wang et al., 2018).

Metal nanoparticles including selenium and silver nanoparticles synthesized from Streptomyces species have also been effective in attenuating virulence of microbial pathogens. Selenium nanoparticles synthesized by Streptomyces minutiscleroticus M10A62 inhibited biofilm of antibiotic-resistant strains of Acinetobacter species at a concentration of 3.2 µg/ml (Ramya et al., 2015). Silver nanoparticles from Streptomyces griseorubens AU2 suppressed the biofilm of Staphylococcus aureus ATCC 25923 and P. aeruginosa ATCC 27853 at a concentration 20 µg/ml and 10 µg/ml, respectively (Baygar and Ugur, 2017). A furonone derivative from Streptomyces sp. AT37 5-[(5E,7E,11E)-2,10-dihydroxy-9,11-dimethyl-5,7,11 tridecatrien-1-yl]-2-hydroxy-2-(1-hydro-xyethyl)-4-me-

thyl-3(2H)-furanone or antibiotic AT37-1 **(19)** exhibited minimum biofilm inhibition concentration (MBIC50) of 10– 15 µg/mL against methicillin ensitive Staphylococcus aureus (MSSA) ATCC 29523 and methicillin resistant Staphylococcus aureus (MRSA) ATCC 43300 (Driche et al., 2017). Streptorubin

B from Streptomyces sp. strain MC11024 displayed IC<sup>50</sup> value of 0.56 µM against the biofilms of m MRSA N315. Although streptorubin B inhibited the growth of MRSA N315 at 2– 4 µg/mL, the compound also exhibited anti-biofilm activity (Bauermeister et al., 2019).

At a dosage of 2.5% (v/v), the metabolites from marine Streptomyces albus A66 repressed the formation of V. harveyi biofilms by 99.3% and dispersed the mature biofilms of V. harveyi by 75.6%. The A66 metabolite was suggested to affect the development of Vibrio biofilms by attenuating the initiation and maturation stage (You et al., 2007; **Table 2**). Methanolic extract from the spent medium of Streptomyces akiyoshiensis CAA-3 inhibited staphylococcal biofilms at a concentration of 0.1 mg/ml. The extract also possessed the ability to inhibit the colonization of Staphylococcus aureus in the intestine of Caenorhabditis elegans up to 70% (**Table 2**; Bakkiyaraj and Pandian, 2010). Culture extracts of Streptomyces sp. BFI 250 at 0.01% (v/v) inhibited the biofilm formation and detachment of preformed biofilms of Staphylococcus aureus ATCC 25923 by ≥ 80% for more than 17 h. The ability to subdue both the formation and detachment of biofilms by Streptomyces sp. BFI 250 was due to the extracellular protease in the extract that was equivalent to approximately 0.1 µg of proteinase K/ml (Park et al., 2012). Extracts from Streptomyces sp. NIO 10068 spent medium reduced motility, formation of biofilm, production of pyocyanin, rhamnolipid and Las A protease, swimming and twitching by 90, 67, 45, 45, 43, 20, and 15%, respectively in P. aeruginosa ATCC 27853. Among the several active compounds including cinnamic acid, linear dipeptides N-amido-a-proline, pro-line–glycine and aromatic acids characterized from the extract of strain NIO 10068, only linear dipeptide and cinnamic acid expressed quorum sensing inhibitory (QSI) activity (Naik et al., 2013). DNA microarray analysis revealed that the spent medium of the strain BFI 230 repressed 42 genes and induced 78 genes in P. aeruginosa cells embedded within the biofilm. The 78 genes that were induced were essential for utilization of iron, biosynthesis of phenazine (phz operon), pyoverdine (pvd operon) and pyochelin (pch). At 1% (v/v) concentration, spent medium of BFI 230 repressed 90% of the P. aeruginosa biofilm. However, at this concentration other virulence factors including swarming and the production of pyoverdine and pyocyanin increased. As the transcriptomic analysis showed that the BFI 230 spent medium induced the genes for iron uptake, external addition of ferrous compounds (FeCl<sup>3</sup> and FeSO4) in the presence of the BFI 230 spent medium resulted in the restoration P. aeruginosa biofilms. The study revealed that proteins or peptides native to the Streptomyces sp. BFI 230 spent medium suppressed the formation of P. aeruginosa biofilms either indirectly interfering with the bacterium's iron utilization or through linking iron with quorum sensing system (Kim et al., 2012).

Characterization of quorum quenching activity in 63 Streptomyces soil isolates showed that 3 strains St11, St61 and St62 degraded synthetic hexanoyl homoserine lactone (HHL). The acylase was stable in the presence of heavy metals and chelating agents, and maintained a maximum catalytic activity between 20 to 50◦C up to pH 8 (Sakr et al., 2015). The extracts of Streptomyces akiyoshinensis (A3) inhibited Streptococcus pyogenes biofilms at a concentration of 10 to 50 µg/ml. The extract from Streptomyces akiyoshinensis affected the cell hydrophobicity, and the initial colonization of Streptococcus pyogenes (Nithyanand et al., 2010). About 200 µg/ml of diethyl ether extracts of Streptomycetes species A745 culture subdued the formation of V. cholerae biofilm by 60% (Augustine et al., 2012). Crude fatty acid extract from three Streptomyces isolates (Streptomyces sps isolates S8, S9, and S15) inhibited formation of Streptococcus pyogenes ATCC 19615 biofilm at a concentration of 10 µg/ml. Remarkably, the lipids found in the crude extract of these Streptomyces species influenced the secretion of extracellular proteins especially streptolysin S (Rajalakshmi et al., 2014). The extract of Streptomyces sp. SBT343 displayed BIC<sup>50</sup> value of 62.5 µg/ml toward Staphylococcus epidermidis RP62A biofilm. At 125 µg/ml, the extract subdued the formation of biofilms of MRSA, MSSA and Staphylococcus epidermidis. Physiochemical characterization of the extract revealed that the bioactive molecule(s) mediating the inhibitory activity toward staphylococcal biofilm were thermostable and non-proteinaceous in nature (Balasubramanian et al., 2017). Hexane partition of Streptomyces sp. CCB-PSK207 spent medium gradually increased the survival of P. aeruginosa PA14 infected C. elegans from 45.33 to 72.71% at the concentration ranging from 50 to 400 µg/ml. Phenotypical analysis on the expression of virulence factors of PA14 showed that the metabolites (fatty acid methyl esters) in the extract were indifferent on the formation of biofilm or on the production of protease and pyocyanin. However, restoration of the green fluorescent protein (GFP) expression in transgenic lys-7:GFP C. elegans strain SAL105 revealed that the hexane partition of CCB-PSK207 did not repress the killing of C. elegans by subduing the virulence of PA14, but rather through boosting the immunity in the nematode by inducing the expression of lysozyme 7 (lys-7) (Fatin et al., 2017). The minimum biofilm inhibitory concentration of metabolites from Streptomyces albogriseolus GIS39Ama were 312 ppm against Escherichia coli MTCC 687, 625ppm against Klebsiella pneumoniae MTCC 3384 and Vibrio cholerae MTCC 3906, and 1250 ppm against Pseudomonas aeruginosa MTCC 2453. Streptomyces albogriseolus GIS39Ama reduced the production of violacein by C. violaceum MTCC 2656 by 87.67% (Lokegaonkar and Nabar, 2017). The extract from Streptomyces sp. MC025 isolated from an unidentified red alga suppressed the formation of Staphylococcus aureus biofilm by ≥ 90% with minimal bactericidal effect on planktonic cells. Bioactivityguided fractionation of the crude extract Streptomyces sp. MC025 led to the identification of 6 bipyridines molecules, of which, collismycins C **(20)** and pyrisulfoxin A **(21)** showed inhibitory activity against MSSA ATCC 6538 at 50 µg/mL. Further studies revealed that Collismycin C was the major component initiating anti-biofilm activity by chelating Fe ions, and that the location of the OH group on bipyridines were vital for anti-biofilm activity against Staphylococcus aureus (Lee et al., 2017).

Screening of 101 marine Actinomycetes led to the discovery of Streptomyces strains that could suppress biofilms of Escherichia coli (by 61 – 80%) and Staphylococcus aureus (by 60%) (Leetanasaksakul and Thamchaipenet, 2018). Extracts from the spent medium of Streptomyces sp. TRM 41337 suppressed the

formation of Staphylococcus epidermidis (ATCC 35984 and 5- 121-2) biofilms by ≥ 90% in a dose-dependent manner for over 24 h. While the culture extracts of Streptomyces sp. TRM 41337 effectively degraded DNA of S. epidermidis, the protein metabolite from the extract reduced the cell surface hydrophobicity and degraded EPS of Staphylococcus epidermidis. Thus, it was suggested that through these properties, the crude protein was able to prevent the formation of S. epidermidis biofilm (Xie et al., 2018).

Melanin pigment (soluble and insoluble forms) purified from Streptomyces sp. ZL-24 suppressed the formation of P. aeruginosa ATCC 9027 and Staphylococcus aureus ATCC 6538 biofilms up to 67.5 and 74.6% and 79.2 and 71.7%, respectively (Wang et al., 2019). Similarly, the extract from Streptomyces griseoincarnatus

HK12 suppressed P. aeruginosa and Staphylococcus aureus biofilms by 82.657 and 78.973%, respectively. GC-MS analysis of the extract showed the presence of five active compounds including arachidic acid, erucic acid, 13Z-octadecenal **(22)**, 9Zoctadecenal and tetracosanoic acid. In silico docking of all the five active compounds with LasI of P. aeruginosa showed that the 13Z-octadecenal interacted with LasI and formed pi-alkyl bond with the conserved residues Trp33 and Phe2 in LasI. It was suggested that the downregulation of QS-regulated virulence gene was due to the small molecule mediated inhibition of LasI binding to its native ligand LasR. This study also suggested that the presence of fatty acyl molecule in HK12 spent medium could have exerted both synergistic or independent anti-biofilm activity (Kamarudheen and Rao, 2019). Four new α- pyridones (compound 16 **(23)**,17 **(24)**,19 **(25)** & 20 **(26)**) generated through chemical transformation of the compounds derived from culture extract of Streptomyces sp., inhibited the expression of P. aeruginosa QSIS-lasI biosensors at a concentration of 6.35 µg/well (Du et al., 2018).

Dimorphic fungi Candida albicans can potentiate clinically significant systemic infections due to its complex and multifactorial virulence factors including isocitrate lyase (ICL), a glyoxylate cycle enzyme (Lorenz and Fink, 2001; Ramírez and Lorenz, 2007; Mayer et al., 2013). Bahamaolide A **(27)** purified from Streptomyces sp. CNQ343 strongly inhibited the mRNA expression of ICL with an IC<sup>50</sup> value of 11.82 µM. Due to the absence of ICL in mammals, Bahamaolide A has been suggested as a promising anti-virulent agent for C. albicans (Lee et al., 2014).

Pre-exposure of C. albicans to the Streptomyces toxytricini Fz94 culture extract at a concentration of 5 g/L prevented the formation of biofilm up to 92%. At 7 g/L, the extract destroyed up to 82% of biofilms after 120 min (Sheir et al., 2017). Partially purified fractions of Streptomyces chrestomyceticus strain ADP4 strongly inhibited the secretory aspartic proteases (Saps) in C. albicans which has been shown to be vital for the formation of hyphae, phenotypic switching, adhesion, digestion of host cell membrane, and also for the evasion of host immune system by the yeast (Srivastava et al., 2017). A metabolite from Streptomyces sp. ADR1 displayed MBIC ≤ 15.625 µg/ml and < 500 µg/ml against preformed biofilm of pathogenic Staphylococcus aureus (Singh and Dubey, 2018). Khatmiamycin **(28)** and aloesaponarin II **(29)** derived from Streptomyces sp. ANK313 inhibited the motility of zoospores of Plasmopara viticola with a MIC value of 10 µg/ml and 25 10 µg/ml, respectively (Abdalla et al., 2011).

### OTHERS

Partially purified pigment from Actinomycetes C5-5Y inhibited the cell surface hydrophobicity, proteolytic and lipase activity of Streptococcus mutans and Staphylococcus aureus (**Table 2**). When treated with the pigment, cell surface hydrophobicity of these nosocomial pathogens reduced by 23 and 24% compared to the 91 and 89% hydrophobicity observed in the control cells. Furthermore, at 10 µg/ml concentration, the pigment also significantly reduced the formation of Streptococcus mutans and Staphylococcus aureus biofilms, leading to the suggestion that Actinomycetes C5- 5Y derived pigment were capable of quenching quorum sensing signals (Soundari et al., 2014). Transcriptomic

TABLE 4 | List of compounds characterized from Actinobacteria and their specific virulence inhibitory function.


analysis on the effect of cyclodepsipeptides (WS9326A and WS92326B) from Actinomycetes strain DSW812 on the VirSR system of C. perfringens, revealed that the WS9326A suppressed the expression of pfoA encoding perfringolysin O in dosedependent manner at sub-micromolar IC<sup>50</sup> concentration. As WS9326B lacked this activity, the absence of double bonds in the dehydrotyrosine of WS92326B was concluded to be crucial for the cyclodepsipeptide binding to VirS system. However, the study also showed that WS9326B effectively decreased the cytotoxicity of Staphylococcus aureus on human corneal epithelial cells significantly. WS9326A and WS9326B also repressed hemolysin production by S. aureus 8325–4 (type-I AIP), S. aureus K12 (type-II AIP) and S. aureus K9 (type-IV AIP), indicating the specificity of Actinomycetes cyclodepsipeptides toward the different auto inducing peptides (AIP). Cochinmicins II and III from Actinomycetes strains GMKU369, have also been suggested to function as an antagonist like cyclodepsipeptides due to their similarities in structure, molecular size, and hydrophobicity (Desouky et al., 2015).

### OPINION AND FUTURE PERSPECTIVE

The phylum Actinobacteria encompasses a group of organisms well known for its prodigious production of secondary metabolites with complex scaffolding and chemical entities. This actinic uniqueness has been beneficial in terms of its pharmaceutical adaptability, as clinically significant antimicrobials, anti-tumor agents, immunosuppressants, antiproliferative agents, anti-parasitic agents and herbicides than any other bacterial origin natural product. In this regard, identification of secondary metabolites from the phylum Actinobacteria with potential to attenuate virulence in other microorganisms, and the broad-spectrum specificity toward different AHLs, could be advantages for engineering the much anticipated anti-virulence drugs. Actinobacteria strains that suppressed microbial virulence have been reported majorly from marine and terrestrial environment (**Figures 1**, **2**). Over the past decade, several marine natural products (MNP) derived from various phyla of bacteria, alga, seaweeds and invertebrates exhibiting anti-virulence property including antibiofilm property have been reported. This could be the reflection

### REFERENCES


of the recent trend in exploring the metabolite profile of microbiome from uninhabited areas including arctic regions, to prevent the re-isolation of known active metabolites. While the active metabolites from the Actinobacteria have been demonstrated with virulence suppressing potential against a wide range of bacteria and yeast cells, the assays employed to evaluate the virulence inhibiting potential are very limited (**Table 4**). The Actinobacteria derived products were mainly evaluated for their potential to inhibit biofilm formation or the production of enzymes, pigments, cell hydrophobicity, and motility. Yet, many crucial virulence factors including iron uptake, immune cell evasion and suppression of host immune system should have been considered as promotion of pathogenesis by bacteria like Staphylococcus aureus is site-specific. Similarly, evaluation of the majority of actinobacterial origin anti-virulence agents has been against very limited bacterial reference strains and reporter strains particularly Staphylococcus aureus and Pseudomonas aeruginosa. Although, undeniably, these organisms are highly virulent with or without AMR, researches with a wide range of organisms especially variant cell populations such as persister cells that have been demonstrated to be the etiological agents of chronic infections would help to establish the potency of metabolites as anti-virulences. To conclude, in the evolutionary struggle for co-existence between microorganism and humans, the single-sided supremacy observed during the prodromal antibiotic era convincingly advocates requirement of multifactor approach to target pathogenesis of microorganism in the host body.

### AUTHOR CONTRIBUTIONS

Both authors contributed equally to the preparation and completion of the manuscript.

## FUNDING

The authors would like to thank the support of DST-SERB (EMR/2016/002296) and SASTRA Deemed to be University, Thanjavur, India.



from Pseudomonas aeruginosa infection through restitution of Lysozyme 7. Front. Microbiol. 8:2267. doi: 10.3389/fmicb.2017.02267



isocitrate lyase in Candida albicans. Bioorg. Med. Chem. Lett. 24, 4291–4293. doi: 10.1016/j.bmcl.2014.07.021



quenchers of quorum-sensing-regulated functions of plant-pathogenic bacteria. Microbiology 149, 1981–1989. doi: 10.1099/mic.0.26375-0


**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 © 2019 Sarveswari and Solomon. 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.

# AiiM Lactonase Strongly Reduces Quorum Sensing Controlled Virulence Factors in Clinical Strains of Pseudomonas aeruginosa Isolated From Burned Patients

#### Edited by:

Ana Maria Otero, University of Santiago de Compostela, Spain

#### Reviewed by:

Giordano Rampioni, Roma Tre University, Italy Wim J. Quax, University of Groningen, Netherlands

#### \*Correspondence:

Rodolfo García-Contreras rgarc@bq.unam.mx

#### Specialty section:

This article was submitted to Microbial Physiology and Metabolism, a section of the journal Frontiers in Microbiology

> Received: 15 May 2019 Accepted: 31 October 2019 Published: 14 November 2019

#### Citation:

López-Jácome LE, Garza-Ramos G, Hernández-Durán M, Franco-Cendejas R, Loarca D, Romero-Martínez D, Nguyen PTD, Maeda T, González-Pedrajo B, Díaz-Guerrero M, Sánchez-Reyes JL, Díaz-Ramírez D and García-Contreras R (2019) AiiM Lactonase Strongly Reduces Quorum Sensing Controlled Virulence Factors in Clinical Strains of Pseudomonas aeruginosa Isolated From Burned Patients. Front. Microbiol. 10:2657. doi: 10.3389/fmicb.2019.02657 Luis Esaú López-Jácome1,2, Georgina Garza-Ramos<sup>3</sup> , Melissa Hernández-Durán<sup>2</sup> , Rafael Franco-Cendejas<sup>2</sup> , Daniel Loarca<sup>1</sup> , Daniel Romero-Martínez<sup>3</sup> , Phuong Thi Dong Nguyen<sup>4</sup> , Toshinari Maeda<sup>4</sup> , Bertha González-Pedrajo<sup>5</sup> , Miguel Díaz-Guerrero<sup>5</sup> , Jorge Luis Sánchez-Reyes<sup>1</sup> , Dánae Díaz-Ramírez<sup>1</sup> and Rodolfo García-Contreras<sup>1</sup> \*

<sup>1</sup> Laboratorio de Bacteriología, Departamento de Microbiología y Parasitología, Facultad de Medicina, Universidad Nacional Autónoma de México, Mexico City, Mexico, <sup>2</sup> Laboratorio de Infectología, Centro Nacional de Investigación y Atención de Quemados, Instituto Nacional de Rehabilitación, Mexico City, Mexico, <sup>3</sup> Laboratorio de Fisicoquímica e Ingeniería de Proteínas, Departamento de Bioquímica, Universidad Nacional Autónoma de México, Mexico City, Mexico, <sup>4</sup> Department of Biological Functions Engineering, Gradute School of Life Sciences and System Engineering, Kyushu Institute of Technology, Kitakyushu, Japan, <sup>5</sup> Departamento de Genética Molecular, Instituto de Fisiología Celular, Universidad Nacional Autónoma de México, Mexico City, Mexico

Pseudomonas aeruginosa is an opportunistic bacterium associated with healthcare infections in intensive care units (ICUs), ventilator-associated pneumonia (VAP), surgical site infections, and burns. This bacterium causes 75% of death in burned patients, since it can develop a persistent biofilm associated with infections, express several virulence factors, and antibiotic-resistance mechanisms. Some of these virulence factors are proteases such as elastase and alkaline protease, or toxic metabolites such as pyocyanin and is one of the few microorganisms able to produce cyanide, which inhibits the cytochrome oxidase of host cells. These virulence factors are controlled by quorum sensing (QS). In this work, 30 P. aeruginosa clinical strains isolated from burned patients from a tertiary hospital in Mexico City were studied. Antibiotic susceptibility tests were done, and virulence factors (elastase, alkaline protease, HCN, and pyocyanin) were determined in presence of an N-acylhomoserine lactonase, AiiM able to hydrolyze a wide range of acyl homoserine lactones. The treatment reduced significantly the activities of elastase and alkaline protease, and the production of pyocyanin and HCN in all producer strains but not the secretion of toxins through the type III secretion system. Our work suggests that AiiM treatment may be an effective therapy to combat P. aeruginosa infection in burn patients.

Keywords: AiiM lactonase, virulence factors, Pseudomonas aeruginosa, quorum quenching, burned patients, anti-virulence therapy

### INTRODUCTION

fmicb-10-02657 November 13, 2019 Time: 14:36 # 2

Pseudomonas aeruginosa is an opportunistic bacteria associated with healthcare infections in intensive care units (ICUs), ventilator-associated pneumonia (VAP), central line-associated blood stream infections, surgical site infections (Cohen et al., 2017), burnt wounds (Fournier et al., 2016), and urinary tract infections, otitis media, and keratitis (Chatterjee et al., 2016; Olivares et al., 2016). In the United States, according to the Centers for Disease Control and Prevention, in 2013 it was estimated that every year around 51,000 health-care infections are associated to P. aeruginosa, of which 6,700 are multidrug resistant, causing 440 deaths per year (Centers for Disease Control and Prevention, 2018) 1 . Both P. aeruginosa and Acinetobacter baumannii complex are the most important, resistant and dangerous microorganisms infecting burnt patients (Tredget et al., 1992; Estahbanati et al., 2002; Turner et al., 2014; Centers for Disease Control and Prevention, 2019). Despite medicine advances, these sorts of complications are still a huge problem to solve, and as a consequence, around 75% of burned infected patients die. Burn infections related to P. aeruginosa often promote a faster deterioration allowing the spread of bacteria causing death in weeks and even in days (Mcmanus et al., 1985; Turner et al., 2014). P. aeruginosa has a wide arsenal of virulence factors that enable it to colonize and cause infections in the host, the relevance of these virulence factors has been demonstrated using P. aeruginosa strains with deficiencies in their production, leading to a reduced ability of colonizing and a lower dissemination in the host (Pavlovskis and Wretlind, 1979; Rumbaugh et al., 2009; Jimenez et al., 2012; Castillo-Juarez et al., 2015). Elastase is a metalloprotease that disrupts several proteins such as: collagen, elastin, immunoglobulins (IgA and IgG), complement components, and cytokines like interferon gamma and tumor necrosis factor alpha (Pavlovskis and Wretlind, 1979; Lyczak et al., 2000; Ben Haj Khalifa et al., 2011). Alkaline protease is also a zinc metalloprotease that inhibits phagocytosis, killing through neutrophils, opsonization, the action of the complement cascade by degrading C3b and is as well related to corneal damage (Howe and Iglewski, 1984; Ben Haj Khalifa et al., 2011; Laarman et al., 2012; Lee and Zhang, 2015). P. aeruginosa is one of the few microorganisms that can synthesize cyanide through the oxidative decarboxylation of glycine by hydrogen cyanide synthase enzyme, under microaerobic conditions (O<sup>2</sup> < 5%). HCN is a poison that inhibits respiration by inactivating cytochrome oxidase C (Huber et al., 2016). Another important virulence factor is pyocyanin, a blue phenazine, that promotes oxidative stress, which inhibits ciliary movement and delays inflammatory response due to the damage of neutrophils and apoptosis induction (Ben Haj Khalifa et al., 2011; Lee and Zhang, 2015). In burn injuries, pyocyanin plays an important role because it stimulates colonization, damage of surrounding tissue and promotes dissemination. Furthermore, P. aeruginosa strains are often multi-drug resistant, limiting treatment options in healthcare settings around the globe, owing this the World Health Organization classified P. aeruginosa as the second more threatening bacterium. Moreover, although new antibiotics are available, each time, resistance against those new drugs quickly appears (Fournier et al., 2016; Shortridge et al., 2017; Karampatakis et al., 2018; Shields et al., 2018). In many pathogenic bacteria, virulence factors are controlled by cell to cell communication known as quorum sensing (QS). P. aeruginosa has two QS systems mediated by N-acyl homoserine lactones, Las and Rhl, each one is constituted by three elements, a synthase, a signal receptor and an autoinducer signal. The Las system is formed by LasI which is the synthase, the receptor is LasR and the autoinducer is N-3-oxo-dodecanoyl-L-homoserine lactone, meanwhile the Rhl system is formed by the synthase RhlI, RhlR as receptor and the auto-inducer is N-butyryl-homoserine lactone. These two systems are hierarchically organized and each one of them controls several virulence factors. The Las system regulates the Rhl system and virulence factors such as elastase, protease, exotoxin A, alkaline protease and type II secretion system; while the Rhl system enhances the production of rhamnolipids, hydrogen cyanide, and pyocyanin (Van Delden and Iglewski, 1998; Douzi et al., 2011; Lee and Zhang, 2015). Due to the fast increase in bacterial resistance, alternative strategies such as quorum quenching (QQ) have been proposed. QQ consists of blocking or inhibiting cell to cell communication by obstructing the autoinducer synthases, the signal receptors or by degrading the autoinducers via two enzymatic strategies: disrupting the lactone ring through a lactonase or through the cleavage of the acylic tail by acylases (Defoirdt et al., 2013; Defoirdt, 2018). The aim of this work is to evaluate the activity of AiiM, a lactonase enzyme, in P. aeruginosa clinical bacterial strains isolated from burnt patients in a third level center of Mexico City, in order to find out if its utilization could eventually be proposed to treat infected burnt patients.

### MATERIALS AND METHODS

### Clinical Strains

Randomly 200 strains were selected<sup>2</sup> from the collection belonging to Infectious Diseases Laboratory at Centro Nacional de Investigación y Atención de Quemados at Instituto Nacional de Rehabilitación Luis Guillermo Ibarra, to avoid genomic redundancy pulsed field, gel electrophoresis was performed and only one strain per clonal group was selected for further experiments. Clinical strains were identified with Vitek 2 Compact <sup>R</sup> (Biomerieux, France) with Gram negative card identification, some biochemical tests included were oxidase, indole production, growth at 42◦C, arginine dihydrolase and glucose oxidation/fermentation. The origin of each clinical isolate is shown in **Supplementary Table S1**.

### Minimal Inhibitory Concentrations

Minimal inhibitory concentrations were determined according to Clinical and Laboratory Standards Institute <sup>R</sup> (CLSI) M07- A10 (CLSI, 2015) in 96-well plates. Breakpoints interpretation were made according to the M100 Performance Standards for

<sup>1</sup>https://www.cdc.gov/hai/organisms/pseudomonas.html

<sup>2</sup>www.randomization.com

Antimicrobial susceptibility testing 28th edition (CLSI, 2019). Antibiotics included were amikacin (Sigma Aldrich A1774), gentamicin (Sigma Aldrich G3632), aztreonam (Sigma Aldrich PZ0038), ceftazidime (Sigma Aldrich C3809), cefepime (Sigma Aldrich PHR1763), ciprofloxacin (Sigma Aldrich 17850), levofloxacin (Sigma Aldrich 28266), doripenem (Sigma Aldrich 32138), imipenem (Sigma Aldrich I0160), meropenem (Sigma Aldrich M2574), colistin (Sigma Aldrich C4461), and piperacillin/tazobactam (Sigma Aldrich P8396/T2820). P. aeruginosa ATCC <sup>R</sup> 27853 was used as control as according to CLSI (**Supplementary Table S1**).

## las/rhl Genes Detection

fmicb-10-02657 November 13, 2019 Time: 14:36 # 3

### DNA Extraction

Pseudomonas aeruginosa strains were cultured in 5% sheep blood agar during 18 h at 37◦C, and then one single colony was taken and lysed in an Eppendorf tube with 500 µL of TE buffer (10 mM Tris–HCl, 1 mM EDTA, pH 7.5) and were set into a heat block at 95◦C for 5 min. Tubes were centrifuged, and the supernatant was added into a new tube.

Genes related to Las and Rhl systems were amplified (**Table 1**) in a final volume of 25 µL of buffer 1X, 3 mM MgCl2, 200 µM dNTP<sup>0</sup> s, 0.2 µM primer forward and reverse, 0.026 U/µL Taq polymerase (Amplitaq Gold <sup>R</sup> DNA Polymerase, Applied Biosystems N808-0241, United States). The amplification conditions used were: 95◦C 10 min, 95◦C 30 s, 58◦C 45 s and 72◦C 50 s during 35 cycles, 72◦C 5 min and finally 4◦C (Veriti 96 Well thermal cycler, Applied Biosystems, United States). Amplification products were loaded into a 1% agarose gel stained with SYBR <sup>R</sup> green I (S7567, Life Technologies, United States) and visualized with Gel DOCTM XR + with Image LabTM software (Bio-Rad, United States). P. aeruginosa PAO1 was used as positive control in each one of the systems and 1 lasR/rhlR PAO1 as negative control of transcriptional regulators.

### AiiM Purification

AiiM construction was provided by Dr. Toshinari Maeda (Nguyen et al., 2019). Briefly 50 mL Escherichia coli M15/pQE30 AiiM was grown overnight (ON) in Luria Bertani broth with 100 µg/mL of carbenicillin (Sigma Aldrich C1389) and 50 µg/mL of kanamycin (Sigma Aldrich K1876), afterward, 10 mL of the ON cultures were taken and inoculated into 1 L of terrific broth with carbenicillin and kanamycin as above described, cultures were incubated at 37◦C 220 rpm, optical density (OD)


at 600 nm was measured until the culture reached an OD of 0.5 and immediately after, it was induced with 500 µM IPTG. The cultures were incubated at 37◦C 220 rpm for 6 h and centrifuged at 10,000 rpm for 40 min. Pellets were resuspended in 40 mL of purification buffer (50 mM NaH2PO4, 300 mM NaCl and 10% glycerol, pH adjusted to 8.0) and 500 µM PMSF (Sigma Aldrich P7626), sonicated (45% amplitude for 45 s and 2 min of rest, all this 10 times; Ultrasonic processor, Cole Parmer) and centrifuged at 10,000 rpm during 40 min. The supernatants were passed through a 0.2 nm filter and loaded onto Protino <sup>R</sup> Ni-TED resin (Macherey-Nagel, 745200.600) for purification of His-tagged proteins previously equilibrated with 3 column volumes (CV) of 20 mL of purification buffer, then filtered protein extracts were passed through the column, after that, 2 additional CV of purification buffer were passed. Protein was eluted with 150 mM imidazole (Sigma Aldrich I5513) in 2 CV of purification buffer. AiiM fractions with higher purity were selected, concentrated with polyethylene glycol 35 KDa (Sigma Aldrich 946-46) into dialysis tubing cellulose membrane (Sigma Aldrich D9777). Afterward, dialysis was done to remove imidazole using dialysis buffer (50 mM Tris, 300 mM NaCl adjusted at pH 7.5). SDS-PAGE was done to estimate the amount and purity of AiiM. Protein was quantified by its absorbance at 280 nm with NanoDrop 2000 (Thermo Fisher Scientific, United States), using an extinction coefficient (Abs 0.1%) of 1.08. Aliquots of protein were made and stored at −20◦C until they were used.

### AiiM Activity Against N-acyl Homoserine Lactones

To evaluate the HSL lytic activity of the purified AiiM, an analytical assay was developed in an Alliance HPLC system (Waters, United States) with a Symmetry (Waters, United States) C18 Column (75 mm, 3.5 mm). Both short and long acylated chains were included, 1 mM N-butyryl-DLhomoserine lactone (C4-HSL; Sigma Aldrich 09945), 1 mM N-(3 oxooctanoyl)-L-homoserine lactone (3OC8-HSL;Sigma Aldrich O1764), 1 mM N-decanoyl-DL-homoserine lactone (C10- HSL; Sigma Aldrich 17248), 1 mM N-(3-oxodecanoyl)-Lhomoserine lactone (3OC10-HSL; Sigma Aldrich O9014), and N-(3-oxododecanoyl) homoserine lactone (3OC12-HSL; Sigma Aldrich O9139). 60 mM NaOH was used as positive control since it can hydrolyze HSL molecules and the reaction was stopped with the addition of 2N HCl. Several concentrations, from 250, 100, 50, 25, 10, and 5 µg/mL of purified AiiM were tested. Time exposition was also varied; 24 h, 2 h, 1 h, 30 min, 20 min, 10 min, and 5 min. All experiments were performed in triplicate. Chromatographic conditions were: column temperature 25◦C, sample temperature 25◦C, injection volume 10 µL, flow rate 1 mL/min, detection 205 nm. Elution mixture was made with 50 mM phosphates buffer pH 2.9:acetonitrile. For C4-HSL the relation was 90:10, 3OC8-HSL 60:40, C10-HSL 60:40, 3OC10-HSL 60:40, and 3OC12-HSL 50:50.

### Bacterial Growth With and Without AiiM

Grow curves were analyzed to test the effect of AiiM on P. aeruginosa PAO1 and 1 lasR/rhlR PAO1 growth kinetics. A 50 mL flask with 5 mL of LB was inoculated with each one of the strains at an initial OD600nm of 0.05 with and without 5 µg/mL of AiiM, samples were taken each hour for 12 h.

### Virulence Factors Determination

fmicb-10-02657 November 13, 2019 Time: 14:36 # 4

Elastase was determined for clinical strains and P. aeruginosa PAO1 1lasR/rhlR and 1lasI/rhlI PAO1 mutants according to methods previously described (Ohman et al., 1980), with some modifications. ON of each clinical strain and control strains were cultured in LB at 37◦C 220 rpm and were inoculated at an initial OD600nm of 0.05 in 5 mL of LB with and without AiiM (5 µg/mL), samples were incubated for 18 h at 37◦C and 220 rpm, and centrifuged at 14,000 rpm for 2 min. 50 µL of the supernatant were taken and set into 950 µL of elastase buffer (100 mM Tris–HCl, 1 mM CaCl2, pH 7.5) with 2.5 mg of elastin-congo red (Sigma Aldrich reference E0502) as substrate. Tubes were incubated at 37◦C, 220 rpm for 2 h, centrifuged at 14,000 rpm for 5 min and the released die in the supernatant was measured at 495 nm with a spectrophotometer SmartSpec Plus (Bio-Rad, United States). All determinations were performed by triplicate.

### Alkaline Protease

Alkaline protease was determined according to methods previously described (Howe and Iglewski, 1984), with some modifications. ON of each strains were cultured into LB at 37◦C 220 rpm and were inoculated in 5 mL of LB at an initial OD600nm of 0.05 with and without AiiM (5 µg/mL). AiiM protein was added at the beginning of the cultures, samples were incubated for 18 h at 37◦C and 220 rpm, and centrifuged at 14,000 rpm for 2 min. 50 µL of supernatant were taken and added into 950 µL of protease buffer with 2.5 mg of Hide-Remazol brilliant blue R (Sigma Aldrich reference H6268) as substrate. Tubes were incubated at 37◦C, 220 rpm for 20 min centrifuged at 14,000 rpm 5 min and the supernatant was measured at 595 nm with a spectrophotometer SmartSpec Plus (Bio-Rad, United States).

### Pyocyanin

For pyocyanin production ON of each clinical and control strains were cultured in LB at 37◦C at 220 rpm and inoculated into 5 mL of LB at an initial OD600nm 0.05 with and without AiiM (5 µg/mL). AiiM protein was added at the beginning, samples were incubated 18 h at 37◦C and 220 rpm. One milliliter of supernatant was taken, centrifuged at 14,000 rpm for 5 min, and then 800 µL of supernatant were set into new 1.5 mL conic tube and 400 µL of chloroform was added, each tube was mixed in vortex for 2 min. Tubes were centrifuged for 5 min at 14,000 rpm, 300 µL of the organic phase were taken and deposited into a new tube, 800 µL of 0.2 N HCl were added and mixed for 2 min in vortex then samples were read at 520 nm (Maeda et al., 2012). P. aeruginosa PAO1,1lasR/rhlR PAO1 and 1lasI/rhlI PAO1 were used as positive and negative controls, respectively.

### HCN Determination

For HCN determination, bacteria were cultured in 3 mL of LB medium in flasks with rubber stoppers at 37◦C and 200 rpm for 18 h, after the incubation two needles were inserted in the rubber stopper, one of them was used for pumping air for 1 h and the other to collect the outflow in 5 mL of 4 M NaOH. HCN concentrations were determined as described by Gallagher and Manoil (2001), briefly, samples were mixed with a 1:1 fresh mixture of 0.1 M o-dinitrobenzene and 0.2 M p-nitrobenzaldehyde both dissolved in 2-methoxyethanol, and following 20 min of incubation at room temperature, the absorbance at 578 nm was determined and compared with a calibration curve made with KCN standards.

### AiiM Dose Response Curve and Suppression of Its Activity by Exogenous Addition of 3OC12-HSL

For these control experiments, the PAO1 reference strain and the clinical isolate P809 were used. Three independent cultures per strain were inoculated in LB medium at an initial OD600nm of 0.05 without and with AiiM at 0.5, 1, 2.5, and 5 µg/mL, and incubated 37◦C at 220 rpm, supernatants were then obtained and used for the determination of pyocyanin concentration and elastase activity (as described before). In addition another 3 cultures per strain with AiiM 0.5 µg/mL were grown to an OD600nm of ∼ 1.0, supplemented with a final concentration of 30 µM of 3OC12- HSL, and incubated until 18 h of incubation were completed, supernatants were collected and pyocyanin concentration and elastase activity determined.

### Long Chain HSL Autoinducer Detection and Its Inactivation by AiiM

To identify autoinducer production and its inactivation by AiiM, each one of the 30 clinical strains were grew up onto MacConkey agar plates, then one colony was taken and inoculated into 5 mL of LB for ON growth. After that, new cultures were inoculated at an initial OD600nm 0.05 in 5 mL of LB and grown at 37◦C, with 220 rpm shaking during 18 h, with AiiM 5 µg/mL and without AiiM enzyme. LB cultures then were centrifuged 14,000 rpm for 5 min. Supernatants were separated in new tubes. For long chain HSL detection Agrobacterium tumefaciens NT1 pZLR4 (Shaw et al., 1997) was used as a biosensor strain. Previously the biosensor strain was grown in one liter of M9 medium and incubated at 37◦C and 220 rpm for 18 h. Bacteria were concentrated by centrifugation at 12,000 rpm during 5 min to a final volume of 15 mL, then aliquots of 1 mL of concentrated bacteria were separated. M9 agar plates were prepared and before solidification 1 mL of the concentrated biosensor plus Xgal at a final concentration of 40 µg/mL (5-bromo-4-chloro-3-indolyl- ß-D-galactopyranoside, USB corporation, Cleveland, OH, United States) were added for 100 mL of M9 agar. 15 µL of each supernatant (with and without AiiM) were then added onto 6 mm filter paper sterile disk on the M9 agar. All experiments were done by triplicate. 1 mM N-decanoyl-DLhomoserine lactone (C10-HSL; Sigma Aldrich 17248), 1 mM N-(3-oxodecanoyl)-L-homoserine lactone (3OC10-HSL; Sigma Aldrich O9014) and N-(3-oxododecanoyl) homoserine lactone (3OC12-HSL; Sigma Aldrich O9139) were used as positive controls and the molecules treated with AiiM 5 µg/mL as negative controls. Plates were incubated at 28◦C and results were observed, a positive reaction associated to the production of long chain HSL was observed as a green halo and inactivation of the signals when the halo was absent.

### Type III Protein Secretion Profiles

fmicb-10-02657 November 13, 2019 Time: 14:36 # 5

For type III secretion assays P. aeruginosa strains (PAO1, PA14, and the clinical isolates H015 and P729) were grown overnight in LB medium. Bacteria were diluted 1:200 into 4 mL of a modified LB medium supplemented with 10 mM MgCl2, 0.5 mM CaCl<sup>2</sup> and 5 mM EGTA (pH 7.4) in the presence or absence of 5 µg/mL of AiiM, and grown at 37◦C to an OD600nm of 0.8 to 1.0. 1 mL of each culture was collected into a microcentrifuge tube and bacteria were pelleted by centrifugation. The resulting pellet was resuspended in 200 µl of 1× Laemmli SDS sample buffer normalized for OD600nm. The supernatant was centrifuged once again and the resulting supernatant was transferred into a clean tube. Supernatant proteins were precipitated overnight at 4◦C by adding trichloroacetic acid to a final concentration of 10%, pelleted by centrifugation and resuspended in 20 µL of 1x Laemmli SDS sample buffer containing 10% saturated Tris base normalized for OD600nm. Samples were separated by 15% SDS-PAGE, transferred onto a nitrocellulose membrane and probed for the presence of the effectors ExoS and ExoU by immunoblotting. Detection was performed using the Immobilon Western chemiluminescent HRP substrate (Millipore), and bands were visualized on X-ray films (Carestream MXB Film).

### ß-Lactams Inactivation

As ß-lactams have a lactone ring, we performed the inactivation disk method to determine whether AiiM would be able to inactivate this kind of antibiotics. Briefly, 2 mL of 0.8% isotonic saline solution (ISS) was added into sterile 12 mm × 75 mm tube, disks of 30 µg ceftazidime (Becton Dickinson, United States), 30 µg cefepime (Becton Dickinson, United States), 10 µg imipenem (Becton Dickinson, United States), and 10 µg meropenem (Becton Dickinson, United States). One set of disks containing ISS was used as negative control, another set of all antibiotics above mentioned with 5 µg/mL of AiiM, and finally, since NaOH can break the lactone ring another set of all antibiotics was used as positive control with 60 mM NaOH. Escherichia coli ATCC <sup>R</sup> 25922 was used as a pansusceptible strain. Disks were incubated for 1 and 10 min, 2 and 24 h. A suspension of 0.5 McFarland was made with E. coli ATCC <sup>R</sup> 25922 and was plated onto Müller Hinton agar (Becton Dickinson, United States), tubes were incubated at 37◦C until their use. All experiments were made by triplicate; the inhibition diameter was measured using a Vernier device.

### RESULTS

All clinical isolates were obtained from burned patients infected with P. aeruginosa. The most common burn etiology was fire (66.7%) followed by scalds (23.3%) and electrical burns (10%). Medians of hospital length of stay were 53 days (8-303), the mean total body surface area was 40% (10-85%) with a mortality rate of 26.6% (n = 8). P. aeruginosa strains were isolated from urine (n = 8), quantitative biopsies (n = 8), blood (n = 6), endotracheal aspirates (n = 3), bronchoalveolar lavage fluid (n = 2), catheter tips (n = 2), and qualitative biopsy (n = 1).

### Pseudomonas aeruginosa Antibiotic Susceptibility Patterns

Susceptibility tests were carried out for all clinical isolates with different antibiotic families including aminoglycosides, monobactams, cephalosporins, fluoroquinolones, carbapenems, lipopeptides, and ß-lactam combination agents. The strains were resistant to almost all antibiotics except colistin (**Figure 1**), resistance rates against all antibiotics families were over 50%. The highest resistance rates were for carbapenems which, until recent decades, were the most potent antibiotics against P. aeruginosa and other non-fermentative Gram negative rods; therefore colistin represents the last treatment option for these types of infections (**Supplementary Table S1**).

### Gene Amplification

In order to determine if Las/Rhl systems were present in P. aeruginosa isolated from burned patients, PCR was performed using the primers described in **Table 1**. Results showed that the genes encoding the Las system (lasI and lasR) and Rhl system (rhlI and rhlR) were found in all the clinical strains. P. aeruginosa PAO1 was used as positive control and 1lasR/rhlR PAO1 was used as negative control (data not shown).

### AiiM Purification

AiiM protein was obtained at a purity of >90% as judged by SDS-PAGE. After elution with 150 mM imidazole, fractions were collected, and those in which AiiM was present were concentrated into dialysis tubing cellulose membrane with polyethylene glycol, then protein concentration was determined and kept at −20◦C until used. The purified protein consists of a single band of ≈ 27 kDa, compared with a theoretical molecular mass of 27.2 kDa (**Supplementary Figure S1**).

### AiiM Activity Against Homoserine Lactones

In order to prove AiiM activity against acylated chains of diverse HSL, cleavage was determined by HPLC. Five HSL autoinducers with both short and long chains were tested (C4- HSL, 3OC8-HSL, C10-HSL, 3OC10-HSL, and 3OC12-HSL). First HSL alone was run to identify retention times (**Supplementary Table S2**), later the same HSL were treated with 60 mM NaOH to disrupt the lactone ring and finally HSL molecules were treated with AiiM and HPLC experiments were performed under the same conditions. Several AiiM concentrations (250 µg/mL, 100 µg/mL, 50 µg/mL, 25 µg/mL, 10 µg/mL, and 5 µg/mL) were used, in order to identify the lowest one suitable for inhibiting the expression of virulence factors. Each HSL was used at 1 mM. The lowest concentration of AiiM tested (5 µg/mL) was enough to cleave all the HSLs in 5 min (**Figure 2** and **Supplementary Figure S2**) and therefore, this concentration was used for all subsequent experiments.

### AiiM Does Not Affect Pseudomonas aeruginosa Growth

Growth curves of P. aeruginosa with and without AiiM were done to determine if its addition had any effect in the growth rates.

As expected, there was no difference in the growing dynamics between these cultures (**Supplementary Figure S3**). PAO1 and 1 lasR/rhlR strains were used as controls.

### QS-Controlled Virulence Factors Inhibition

Once it was confirmed that AiiM did not affect P. aeruginosa growth, its effect over the expression of the QS-controlled virulence factors was determined. Experiments were classified in two groups, one without AiiM addition, and the other with 5 µg/mL addition of AiiM. For elastolytic activity (**Figure 3A**), activity was found in 29 clinical samples, while for alkaline protease activity (**Figure 3B**) there were 27 producing strains; only 12 strains produced pyocyanin (**Figure 3C**), and seven strains were HCN producers (**Figure 3D**). At the same time, experiments with AiiM addition were carried out and the same virulence factors were measured. A significant decrease in the production of each virulence factor was found (elastase p = 0.000002, protease p = 0.000004, pyocyanin p = 0.001 and p = 0.008 for HCN).

In addition, AiiM dose response experiments using it at 0.5, 1, 2.5 and 5 µg/mL were performed with the reference strain PAO1 and the clinical isolate P809 measuring pyocyanin production and elastase activity, as expected the degree of inhibition of both phenotypes was dependent in the concentration of AiiM for both virulence factors (**Supplementary Figure S4**). Moreover the effect of 0.5 µg/mL was partially reversed by the addition of 30 µM of 3OC12-HSL (**Supplementary Figure S4**). In order to verify that the clinical strains had active QS systems and that the inhibitory effect in the expression of QS-dependent virulence factors exerted by AiiM was mediated by the degradation of QS signals. Identification of long chain HSL for each strain was done using the biosensor strain Agrobacterium tumefaciens NT1 pZLR4 (Shaw et al., 1997), as expected all strains were long chain HSL producers, moreover also for all strains AiiM at 5 µg/mL was enough to degrade the long chain HSL of all strains as determined with the biosensor strain (**Supplementary Figure S5**).

### AiiM Does Not Inhibit the Type III Secretion System

Despite its strong inhibitory activity against QS-controlled virulence factors, AiiM had no effect on the secretion of T3SS effectors in both PA14 and PAO1 type strains as well as in the clinical strains P729 and H015 at 5 µg/mL (**Supplementary Figure S6**) and even using 20 µg/mL (data not shown). These two clinical strains were selected as representative examples of a secretion profile similar to strains PAO1 or PA14, respectively.

### AiiM Does Not Inactivate ß-Lactam Antibiotics

Wang et al. (2010) defined AiiM as a member of the superfamily of alpha/beta hydrolases, this may represent a problem if it has the ability to inactivate the broad spectrum of ß-lactam antibiotics

as other carbapenem enzymes do, such as NDM, IMP or VIM. In order to test this, we exposed anti Pseudomonas ß-lactam antibiotics to 5 µg/mL of AiiM (**Supplementary Figures S7, S8**). Nevertheless, AiiM did not degrade any anti Pseudomonas ß-lactam antibiotics.

### DISCUSSION

Pseudomonas aeruginosa is one of the main bacteria that causes hospital acquired infections in immunocompromised patients and vulnerable ones (Azam and Khan, 2018). P. aeruginosa is one of the 12 priority multi-drug resistant bacteria according to the WHO list published in 2017 (WHO, 2017) and belongs to the ESKAPE group, together with Enterococcus faecium, Staphylococcus aureus, Klebsiella pneumoniae, A. baumannii, and Enterobacter species (Chen et al., 2018). In addition to acquired resistance mechanisms such as carbapenemases (Carmeli et al., 2016), P. aeruginosa has many intrinsic antibiotic tolerance mechanisms, for instance low permeability in its external membrane, and expression of several efflux pumps (Malhotra et al., 2018; Ferrer-Espada et al., 2019).

Burn injuries are one of the most common and devastating forms of trauma and patients with serious thermal injury require immediate specialized care in order to minimize morbidity and mortality (Church et al., 2006). P. aeruginosa is one of the most frequent bacteria associated to infection in burn patients together with A. baumannii (Li et al., 2018). In a recent Mexican study (Garza-Gonzalez et al., 2019), P. aeruginosa had around 27% of resistance to carbapenems, in a global context it was one of the main bacteria in 47 Mexican health centers in 20 states, 175/1995 strains were multi-drug-resistant, 165/1995 were possible extreme drug resistant and 87/1995 possible pandrug resistant. In our 30 isolates we had more than 60% of resistance to cephalosporins (ceftazidime and cefepime), carbapenems (dorypenem, imipenem, and meropenem), aminoglycosides (amikacin and gentamicin), fluoroquinolones (ciprofloxacin and levofloxacin), and piperacillin/tazobactam. Moreover, one strain was resistant to colistin, which is the last antibiotic resource.

AiiM showed a wide activity and was able to cut all HSL molecules tested, consistent with a previous report by Wang et al. (2010). Even though we did not analyze it in a quantitative form, we infer a strong activity of AiiM due to its ability to degrade all HSL tested within 5 min of exposition, moreover in our study 5 µg/mL were enough to break down these molecules. One of the main characteristics that a quorum quencher must fulfill is that it should not inhibit bacterial growth (Defoirdt et al., 2013; Defoirdt, 2018) and as expected, AiiM treatment did not affect P. aeruginosa growth kinetic.

In contrast, AiiM significantly reduced the four QS-dependent virulence factors tested in our study following a dose response pattern; moreover, others have demonstrated that AiiM had very good activity in a mouse model of acute pneumonia (Migiyama et al., 2013) and reduces methane production in waste sewage sludge (Nguyen et al., 2018). Although to date, the majority of studies with QQ enzymes have been performed only in type strains like PAO1, PA14 (Fetzner, 2015), recently, Guendouze et al. (2017) did the first investigation with P. aeruginosa clinical strains isolated from diabetic foot using the lactonase SsoPox with a substitution in the amino acid 263 changing a tryptophan to isoleucine, in order to increase the enzymatic activity, using 0.5 mg/mL of protein, they found some strains with certain tolerance to the SsoPox addition. In our study, AiiM reduced elastase and alkaline protease activities, pyocyanin and HCN concentrations, and we did not find any strain with tolerance against AiiM treatment, in spite that AiiM was used at a 100 times lower concentration than SsoPox. Moreover, AiiM effectivity is much higher than the effectivity of small molecule QS inhibitors such as brominated furanones and 5-fluorouracil, that cannot inhibit QS-virulence factor production of several of the clinical strains tested, and that are very toxic to some of them (García-Contreras et al., 2013, 2015; García-Contreras, 2016; García-Contreras et al., 2016; Guendouze et al., 2017).

Nevertheless, for type III secretion, no inhibition by AiiM was found, which is consistent with recent findings showing that in a 1lasR/rhlR mutant of P. aeruginosa PAO1, T3SS effector toxins are secreted at the same levels than in the wild-type strain, demonstrating that this virulence factor is not positively regulated by QS (Soto-Aceves et al., 2019), instead it may be used at low cell densities to establish infections in the host (Hauser, 2009). These highlights the importance of targeting both QS and T3SS to develop robust anti-virulence therapies (García-Contreras, 2016). Moreover, other results indicate that the inhibition of QS systems and T3SS by molecules such as coumarin (Zhang et al., 2018) must be due to independent effects over the QS systems and T3SS.

One possible limitation of the utilization of AiiM and other QQ enzymes for treating P. aeruginosa infections is the fact that lasR defective mutants are often found in infections and although in principle these mutants will produce low levels of QS-dependent virulence factors, this is not always the case due to a rewiring of the virulence factor regulation

(Morales et al., 2017). And these strains could be tolerant against the effect of QQ enzymes.

Since AiiM is a member of the alpha/beta hydrolases superfamily and several antibiotics are inactivated by metallo ßlactamases, reducing clinical options to treat infections (Hong et al., 2015), we tested if AiiM could cleave these ß-lactamase antibiotics, however, AiiM did not inactivate those tested, and hence it could be safely used in combination with them.

Although in vivo tests in burn infection models are lacking, our work suggests that AiiM treatment may be an effective addition for the treatment of P. aeruginosa infections, and since research by other groups had shown also the utility of the lactonase SsoPox against clinical isolates from diabetic foot patients in vitro (Guendouze et al., 2017) and in vivo using an amoeba model (Mion et al., 2019), and in rat pneumonia against the PAO1 strain (Hraiech et al., 2014) lactonase utilization became an strong candidate for its eventual application in the clinical practice, moreover although in vivo studies using acylases are scarce, recently it was shown the PvdQ in addition to their inhibitory properties in vitro (Sio et al., 2006) and in Caenorhabditis elegans model was also able to increase survival, reduce damage and decrease bacterial loads in a pulmonary infection mice model (Papaioannou et al., 2009; Utari et al., 2018). Hence QQ enzymes may be beneficial for the treatment of burn and lung infections as well.

### CONCLUSION

AiiM showed a strong activity against C4-HSL, 3OC8- HSL, C10-HSL, 3OC10-HSL, and 3OC12-HSL. It reduced elastase and alkaline protease activities as well as pyocyanin and HCN concentrations in all tested clinical strains of P. aeruginosa isolated from burned patients and no AiiM tolerant strain was found. However, it had no inhibitory effect against the T3SS.

### DATA AVAILABILITY STATEMENT

All datasets generated for this study are included in the article/**Supplementary Material**.

### REFERENCES


### ETHICS STATEMENT

Pseudomonas aeruginosa clinical strains used in this study were isolated as part of routine clinical hospital procedures to diagnose infection and hence ethical approval was not required, according to the National Institute of Rehabilitation ethical committee. All bacterial isolates were stored as part of laboratory and epidemiology necessities.

### AUTHOR CONTRIBUTIONS

LL-J, GG-R, MH-D, DR-M, PN, MD-G, DL, JS-R, and DD-R performed the experiments. RF-C, TM, BG-P, GG-R, LL-J, and RG-C designed the study, supervised the project and discussed the results. LL-J wrote the manuscript with input from all authors.

### FUNDING

This work was supported by Grants from the Programa de Apoyo a Proyectos de Investigación e Innovación Tecnológica (PAPIIT), the Universidad Nacional Autónoma de México number IN214218, the Consejo Nacional de Ciencia y Tecnología number SEP-CONACYT CB-A1-S-8530, and the CONACYT INFR-2015-252140 grant.

### ACKNOWLEDGMENTS

RG-C is grateful to Beatriz Meráz Rios for her assistance with some experiments. We acknowledge Dr. Norma Espinosa Sánchez and Eugenia Flores Robles for technical assistance. LL-J is a doctoral student from the Programa de Doctorado en Ciencias Biomédicas, Universidad Nacional Autónoma de México.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2019.02657/full#supplementary-material


by state-of-the-art genotyping techniques. Front. Microbiol 9:1104. doi: 10. 3389/fmicb.2018.01104



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

Copyright © 2019 López-Jácome, Garza-Ramos, Hernández-Durán, Franco-Cendejas, Loarca, Romero-Martínez, Nguyen, Maeda, González-Pedrajo, Díaz-Guerrero, Sánchez-Reyes, Díaz-Ramírez and García-Contreras. This is an openaccess article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.