INDUSTRIAL AND HOST ASSOCIATED STRESS RESPONSES IN FOOD MICROBES. IMPLICATIONS FOR FOOD TECHNOLOGY AND FOOD SAFETY

EDITED BY: Lorena Ruiz, Abram Aertsen, Christophe Nguyen-The, Michael G. Gänzle and Avelino Álvarez-Ordoñez PUBLISHED IN: Frontiers in Microbiology

#### *Frontiers Copyright Statement*

*© Copyright 2007-2017 Frontiers Media SA. All rights reserved. All content included on this site, such as text, graphics, logos, button icons, images, video/audio clips, downloads, data compilations and software, is the property of or is licensed to Frontiers Media SA ("Frontiers") or its licensees and/or subcontractors. The copyright in the text of individual articles is the property of their respective authors, subject to a license granted to Frontiers.*

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

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

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

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

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

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

ISSN 1664-8714 ISBN 978-2-88945-293-4 DOI 10.3389/978-2-88945-293-4

# 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

# **INDUSTRIAL AND HOST ASSOCIATED STRESS RESPONSES IN FOOD MICROBES. IMPLICATIONS FOR FOOD TECHNOLOGY AND FOOD SAFETY**

Topic Editors:

**Lorena Ruiz,** Instituto de Productos Lácteos de Asturias, IPLA-CSIC, Spain **Abram Aertsen,** KU Leuven, Belgium **Christophe Nguyen-The,** INRA, UMR408 Sécurité et Qualité des Produits d'Origine Végétale; Univ-Avignon, UMR408 Sécurité et Qualité des Produits d'Origine Végétale, France **Michael G. Gänzle,** University of Alberta, Canada **Avelino Álvarez-Ordoñez,** University of León, Spain

Throughout the food processing chain and after ingestion by the host, food associated bacteria have to cope with a range of stress factors such as thermal and/or non-thermal inactivation treatments, refrigeration temperatures, freeze-drying, high osmolarity, acid pH in the stomach or presence of bile salts in the intestine, that threaten bacterial survival. The accompanying plethora of microbial response and adaptation phenomena elicited by these stresses has important implications for food technology and safety. Indeed, while resistance development of pathogenic and spoilage microorganisms may impose health risks for the consumer and impart great economic losses to food industries, reduced survival of probiotic bacteria may strongly compromise their claimed health benefit attributes.

As a result, substantial research efforts have been devoted in the last decades to unravel the mechanisms underlying stress response and resistance development in food associated microorganisms in order to better predict and improve (i) the inactivation of foodborne pathogens and spoilage microorganisms on the one hand and (ii) the robustness and performance of beneficial microorganisms on the other. Moreover, the recent implementation of system-wide omics and (single-)cell biology approaches is greatly boosting our insights into the modes of action underlying microbial inactivation and survival.

This Research Topic aims to provide an avenue for dissemination of recent advances within the field of microbial stress response and adaptation, with a particular focus not only on food spoilage and pathogenic microorganisms but also on beneficial microbes in foods.

**Citation:** Ruiz, L., Aertsen, A., Nguyen-The, C., Gänzle, M. G., Álvarez-Ordoñez, A., eds. (2017). Industrial and Host Associated Stress Responses in Food Microbes. Implications for Food Technology and Food Safety. Lausanne: Frontiers Media. doi: 10.3389/978-2-88945-293-4

# Table of Contents

*06 Editorial: Industrial and Host Associated Stress Responses in Food Microbes. Implications for Food Technology and Food Safety* Lorena Ruiz, Abram Aertsen, Christophe Nguyen-The, Michael G. Gänzle and Avelino Alvarez-Ordóñez

# **CHAPTER 1. Stress responses in** *Escherichia coli*


Bram Vivijs, Abram Aertsen and Chris W. Michiels


Louise Crozier, Pete E. Hedley, Jenny Morris, Carol Wagstaff, Simon C. Andrews, Ian Toth, Robert W. Jackson and Nicola J. Holden

*70 Corrigendum: Whole-Transcriptome Analysis of Verocytotoxigenic* **Escherichia coli** *O157:H7 (Sakai) Suggests Plant-Species-Specific Metabolic Responses on Exposure to Spinach and Lettuce Extracts*

Louise Crozier, Pete E. Hedley, Jenny Morris, Carol Wagstaff, Simon C. Andrews, Ian Toth, Robert W. Jackson and Nicola J. Holden

*72 Diversity of Survival Patterns among* **Escherichia coli** *O157:H7 Genotypes Subjected to Food-Related Stress Conditions* Mohamed Elhadidy and Avelino Álvarez-Ordóñez

**CHAPTER 2. Stress responses in spore formers: Bacillus as an organism of study**

*82 High-Level Heat Resistance of Spores of* **Bacillus amyloliquefaciens** *and* **Bacillus licheniformis** *Results from the Presence of a* **spoVA** *Operon in a Tn***1546** *Transposon* Erwin M. Berendsen, Rosella A. Koning, Jos Boekhorst, Anne de Jong, Oscar P. Kuipers and Marjon H. J. Wells-Bennik

*92 Recovery of Heat Treated* **Bacillus cereus** *Spores Is Affected by Matrix Composition and Factors with Putative Functions in Damage Repair*

Alicja K. Warda, Marcel H. Tempelaars, Tjakko Abee and Masja N. Nierop Groot

*102 The Synergistic Effect of High Pressure CO2 and Nisin on Inactivation of* **Bacillus subtilis** *Spores in Aqueous Solutions*

Lei Rao, Yongtao Wang, Fang Chen and Xiaojun Liao

*111 Investigating the Inactivation Mechanism of* **Bacillus subtilis** *Spores by High Pressure CO2*

Lei Rao, Feng Zhao, Yongtao Wang, Fang Chen, Xiaosong Hu and Xiaojun Liao


Markus Kranzler, Katharina Stollewerk, Katia Rouzeau-Szynalski, Laurence Blayo, Michael Sulyok and Monika Ehling-Schulz

*145 Adaptation in* **Bacillus cereus:** *From Stress to Disease* Catherine Duport, Michel Jobin and Philippe Schmitt

# **CHAPTER 3. Stress responses in biofilm formers**

*163 The Biofilm Lifestyle Involves an Increase in Bacterial Membrane Saturated Fatty Acids*

Florence Dubois-Brissonnet, Elsa Trotier and Romain Briandet

# **CHAPTER 4. Stress responses in Listeria monocytogenes**


# **CHAPTER 5. Stress responses in the psychrotrophic microorganism Yersinia**

*212 Cold Shock Proteins: A Minireview with Special Emphasis on Csp-family of Enteropathogenic* **Yersinia**

Riikka Keto-Timonen, Nina Hietala, Eveliina Palonen, Anna Hakakorpi, Miia Lindström and Hannu Korkeala

# **CHAPTER 6. Microbial behaviour in fermented foods**

*219 Proline-Based Cyclic Dipeptides from Korean Fermented Vegetable Kimchi and from* **Leuconostoc mesenteroides** *LBP-K06 Have Activities against Multidrug-Resistant Bacteria*

Rui Liu, Andrew H. Kim, Min-Kyu Kwak and Sa-Ouk Kang

*234 Environmental Factors Affecting Microbiota Dynamics during Traditional Solid-state Fermentation of Chinese Daqu Starter*

Pan Li, Weifeng Lin, Xiong Liu, Xiaowen Wang and Lixin Luo

# **CHAPTER 7. Development onf new methodologies to study stress responses in food associated bacteria**


Wilma C. Hazeleger, Wilma F. Jacobs-Reitsma and Heidy M. W. den Besten


Lucie Léonard, Lynda Bouarab Chibane, Balkis Ouled Bouhedda, Pascal Degraeve and Nadia Oulahal

# Editorial: Industrial and Host Associated Stress Responses in Food Microbes. Implications for Food Technology and Food Safety

Lorena Ruiz 1, 2 \*, Abram Aertsen<sup>3</sup> , Christophe Nguyen-The4, 5, Michael G. Gänzle<sup>6</sup> and Avelino Alvarez-Ordóñez <sup>7</sup>

<sup>1</sup> Department of Nutrition, Food Science and Food Technology, Complutense University of Madrid, Madrid, Spain, <sup>2</sup> Department of Microbiology and Biochemistry of Dairy Products, Instituto de Productos Lácteos de Asturias-CSIC, Villaviciosa, Spain, <sup>3</sup> Laboratory of Food Microbiology, Department of Microbial and Molecular Systems, KU Leuven, Leuven, Belgium, <sup>4</sup> Institut National de la Recherche Agronomique, UMR408 Sécurité et Qualité des Produits d'Origine Végétale, Avignon, France, <sup>5</sup> Univ-Avignon, UMR408 Sécurité et Qualité des Produits d'Origine Végétale, Avignon, France, <sup>6</sup> Department of Agricultural, University of Alberta, Food and Nutritional Science, Edmonton, AB, Canada, <sup>7</sup> Department of Food Hygiene and Technology and Institute of Food Science and Technology, University of León, León, Spain

#### Edited by:

Andrea Gomez-Zavaglia, Center for Research and Development in Food Cryotechnology (CIDCA, CONICET), Argentina

#### Reviewed by:

Fausto Gardini, Università di Bologna, Italy Giulia Tabanelli, Università di Bologna, Italy Diego Garcia-Gonzalo, University of Zaragoza, Spain Séamus Fanning, University College Dublin, Ireland

> \*Correspondence: Lorena Ruiz lorena.ruiz @ipla.csic.es

#### Specialty section:

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

Received: 25 June 2017 Accepted: 28 July 2017 Published: 08 August 2017

#### Citation:

Ruiz L, Aertsen A, Nguyen-The C, Gänzle MG and Alvarez-Ordóñez A (2017) Editorial: Industrial and Host Associated Stress Responses in Food Microbes. Implications for Food Technology and Food Safety. Front. Microbiol. 8:1522. doi: 10.3389/fmicb.2017.01522 Keywords: food microbiology, spoilage bacteria, pathogens, food preservation, bacterial stress response, stress response, food safety, food quality

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

### **Industrial and Host Associated Stress Responses in Food Microbes. Implications for Food Technology and Food Safety**

Throughout the food chain, from primary production to consumption, food associated bacteria are confronted with a range of intrinsic and extrinsic factors which influence bacterial survival and activity. Intrinsic stress factors are imposed by food characteristics like high osmolarity or acidity or the dynamics of the microbial ecology of foods; and extrinsic stress factors are inflicted by food preservation methodologies, such as thermal and/or non-thermal inactivation treatments, refrigeration temperatures, or freeze-drying and host associated factors like the stomach acidity or the presence of bile salts in the intestine (Papadimitriou et al., 2016). The accompanying plethora of stress responses elicited by food associated microorganisms has important implications for food technology, quality, and safety (Alvarez-Ordoñez et al., 2015). Resistance development of pathogenic and spoilage microorganisms may impose health risks for the consumer and impart great economic losses to food industries, whereas reduced survival of probiotic or beneficial microbes may strongly compromise their functional attributes.

In the last decades substantial research efforts in the field of food microbiology have been devoted to unravel the mechanisms underlying stress responses and resistance development in food associated microorganisms in order to better predict and improve (i) the inactivation of food-borne pathogens and spoilage microorganisms and (ii) the control of the performance of starter cultures and probiotic microorganisms. Moreover, the recent implementation of system-wide omics and (single-)cell biology approaches has greatly boosted our insights into the molecular mechanisms underlying microbial inactivation and survival (Abhyankar et al., 2017). In this context, this Research Topic aims to provide an avenue to disseminate latest results on food microorganisms stress responses and adaptation, including the implementation of new methodologies for their study.

This Research Topic comprises a collection of 24 articles, including 4 reviews, 1 mini-review, 1 hypothesis and theory article, and 18 original research articles providing a broad overview of recent advances within the field of microbial stress responses and adaptation. Contributions include research on a range of pathogenic microorganisms, such as Escherichia coli, Listeria monocytogenes, Yersinia, and spore formers from the genus Bacillus, but also on microorganisms of industrial interest as starter cultures, including strains of Leuconostoc and Lactobacillus. Heat stress, acid exposure, low temperature, and high CO<sup>2</sup> pressure are among the stress factors evaluated throughout the articles presented in this Research Topic. In addition, the effect of beneficial bacteria on the inactivation of pathogenic bacteria on fermented products, the effect of fermentation associated factors in the dynamics of food microbiota, and the development of new technologies for the study of the performance and inactivation of food

this research topic as summarized below. The review by Li and Gänzle provides a comprehensive evaluation on the heat resistance of E. coli cells including information on the molecular mechanisms associated to heat tolerance variability among strains and on genetic determinants of extreme heat resistance. In addition, the increased heat tolerance following pre-exposure to other stresses such as desiccation or exposure to osmotic or acid stresses, and the effect of certain food ingredients on E. coli heat susceptibility are also discussed. Also in relation to E. coli heat stress, Huertas et al. demonstrate that the response of E. coli to thermal treatments strongly depends on the heating rates, thus providing evidence that classical estimation of heat inactivation parameters, such as D and z values, under isothermal conditions might not provide a realistic estimation of inactivation levels achieved during heat processing of food.

microorganisms have also been evaluated in articles included in

Molecular mechanisms allowing E. coli growth under mild acidic conditions were characterized by Vivijs et al. These authors identify in their elegant study new mechanisms involved in the growth of E. coli at moderately low pH, and demonstrate that these mechanisms differ from those used to withstand exposure to extremely low pH. The response of food pathogens to weak acids is very relevant in the context of food safety and preservation, since weak acids are frequently employed as food additives. In fact, in another work by Lenzi et al. the effect of weak acids on Shiga toxin production by an E. coli strain was also studied. Results presented in their work highlight the necessity to evaluate not just the bactericidal effects of technological and preservation treatments, but also other physiological characteristics of surviving bacteria, like toxin production, which might be enhanced by certain stressors.

Food components can have a strong influence on E. coli physiology and persistence as demonstrated by Crozier et al. and Crozier et al. who studied the transcriptional adaptation of this microorganism to vegetable extracts and its long term persistence on different live plants. Their results identify possible markers to be used in the development of risk-based analysis on microbial contamination of crops.

Overall, the research works included in this Research Topic reflect the existence of strong intra-species variability in E. coli tolerance to food associated stress factors. This fact was further highlighted in the work by Elhadidy and Alvarez-Ordóñez, who evaluated the response of a collection of strains to multiple stresses prevailing throughout the food chain. These authors report that E. coli serotypes more commonly associated with human disease do also exhibit higher resistance to food-associated stresses, what suggests the influence that stressors may have on the spread of this human pathogen.

Bacterial stress resistance is of particular concern in those species exhibiting specialized persistence mechanisms as is the case of sporulated bacteria and/or biofilm formers. Six original articles and one review article in this Research Topic focus on sporulated Bacillus species and one original research article focuses on characteristics of biofilm-lifestyle cells. From these, four articles study spores inactivation dynamics through heat and high pressure CO<sup>2</sup> treatments, while two articles are conducted with vegetative forms. Regarding spores, the study by Berendsen et al. identify genetic elements in several Bacillus species conferring high heat tolerance to the spores, thus providing new knowledge to design genetic-based detection methods to predict the heat tolerance of the spores produced. In another work, a strong variability in the recovery rate of heat treated spores depending on the media used was reported by Warda et al. This work also studies a range of deletion mutants with defects in spore recovery, providing new knowledge on the specific function of genes required for spore germination and emphasizing the importance of optimizing both preservation treatments and methodological procedures employed for their evaluation. Whereas, heat is the most commonly used method to inactivate spore formers, Rao et al. and Rao et al. also report in two research articles that high pressure CO<sup>2</sup> inactivates spores by damaging the spore structure, allowing them to demonstrate the synergistic effect of high pressure CO<sup>2</sup> and nisin for B. subtilis spores inactivation. These works highlight that an advanced understanding of the inactivation mechanisms can facilitate the rationale for selection of preservation treatments.

The stress response of vegetative forms of Bacillus spp. is evaluated in two original research articles. Van Beilen et al. screened for B. cereus mutants susceptible to weak organic acids commonly used as food preservatives. As a result, a gene essential for membrane homeostasis maintenance and sorbic acid tolerance was identified. In another work with vegetative forms of Bacillus, Kranzler et al. showed that heat tightly controls the production of the emetic toxin cereulide produced by B. cereus. The production of toxin and the proportion of toxin isomers differing in toxicity were shown to be strongly influenced by the temperature, allowing the authors to conclude that toxin production cannot be predicted from growth rates. This emphasizes the necessity to evaluate bacterial traits in preserved food and not just growth parameters or sole cell numbers.

One last article focused on Bacillus species is presented by Duport et al. who comprehensively review aspects of bacterial response and adaptation to gastrointestinal stress factors including acidity and fluctuating oxygen availability, and the role of these factors on enterotoxin production. They also identify knowledge gaps providing a basis to delineate future research on this topic.

Another common basis for bacterial persistence along the food chain is based on the biofilm forming capability of foodborne pathogenic and spoilage bacteria, since biofilmliving cells are more resistant to biocides and disinfection regimes. In the work by Dubois-Brisonnet et al. biofilm lifestyle in Staphylococcus, Listeria, Pseudomonas, and Salmonella species is associated with changes in membrane fatty acid composition, which are hypothesized as a biofilm-adaptive trait allowing energy saving and long-term survival. Among biofilm forming microorganisms Listeria monocytogenes represents a great concern to food safety due to the high mortality rate of listeriosis and, therefore, several articles in this Research Topic focus on this bacterium, which is a highly adaptable microorganism capable to survive under a range of stress factors. NicAogáin and O'Byrne review routes allowing L. monocytogenes inclusion in the food chain, genetic mechanisms allowing its survival under stressors encountered throughout the food chain, and describe control mechanisms, including novel inactivation approaches such as those targeting specific regulatory elements. Tremonte et al. studied in a range of foodborne pathogens, the phylogenetic conservation, and tridimensional structure prediction of a universal stress protein whose production is acid-induced in Listeria and has been related to acid tolerance. Another concerning feature regarding L. monocytogenes control is its psychrotrophic character. In this context, Saunders et al. modeled the activity of the enzyme initiating fatty acids biosynthesis at different temperatures, providing new knowledge on the basis of the homeoviscous adaptation to low temperature in L. monocytogenes and offering new targets to control L. monocytogenes survival under refrigeration temperatures.

Proteins involved in adaptation to cold shock in Yersinia, a psychrotrophic member of the Enterobacteriaceae family, capable to grow at temperatures close to 0◦C, have also been reviewed in the article by Keto-Timonen et al. In other bacteria, these proteins are involved in adaptation to a range of stress factors and, therefore, studies on their regulation and specific activity in relevant food microorganisms is warranted.

Other original research articles included in this Research Topic are related to stresses imposed by or affecting beneficial food-associated bacteria. On the one hand, Liu et al. identify bioactive cyclic peptides displaying antibacterial activity against a range of multidrug resistant bacteria in fermented Kimchi. Producing strains belong to Leuconostoc species and offer potential to prevent food contamination with undesirable microorganisms. On the other hand, Li et al. assess the dynamic evolution of microbiota composition during the two stage fermentation process of a traditional Chinese Daqu fermented product. Multivariable analysis allowed the authors to establish a correlation between environmental variables and microbial community profiles. These kind of studies will ease the selection of appropriate environmental parameters when establishing or improving industrial fermentation procedures.

Finally, four articles in this collection report the development and/or implementation of new methodologies to study microbial stress responses and monitor the efficacy of food preservation treatments. One of the issues that may compromise the efficacy of food preservation technologies is the existence of damaged cells, incapable of growth in standard laboratory media, but still alive and capable to proliferate in foods under appropriate conditions. In this context Espina et al. studied the selective media methodology used to determine the proportion of sublethally injured cells, using heat treated E. coli as a model. These authors provide a mechanistic explanation for the methodology and propose method improvements.

Similarly, Hazeleger et al. report the optimization of protocols for the detection of Campylobacter in foods. Campylobacter grow slowly during the recommended two-step enrichment process, and, therefore, can be easily outgrown by other bacteria present in the sample. In this work, authors evaluate some modifications of the recommended procedure that result in improved detection of Campylobacter specifically in samples where the presence of background flora is expected.

New applications of already existing methodologies are also described in this collection of articles. Bancalari et al. test a new potential application of impedance microbiology based methods to ease the evaluation in real time of acidifying performance in a large number of lactic acid bacteria. Finally, a review article by Léonard et al. discusses the potential, advantages and limitations of multi-parameter flow cytometry assays as a tool to study antimicrobial treatment efficacy. These authors also describe the cautions that must be taken when quantifying both inactive and damaged cells and discuss additional technologies that may be used to corroborate results obtained through flow cytometry based approaches.

This editorial summarizes the publications included in this Research Topic. We sincerely hope that this collection of articles will prompt further research on microbial stress responses and contribute to advance the knowledge on microbial physiology and ecology in foods.

# AUTHOR CONTRIBUTIONS

LR designed and wrote the Editorial with contributions from AA, CN, MG, and AA.

# ACKNOWLEDGMENTS

We would like to thank the authors and reviewers for their valuable contributions and constructive criticisms to this special issue.

# REFERENCES


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

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

# Some Like It Hot: Heat Resistance of Escherichia coli in Food

Hui Li<sup>1</sup> and Michael Gänzle1,2 \*

<sup>1</sup> Department of Agricultural, Food and Nutritional Science, University of Alberta, Edmonton, AB, Canada, <sup>2</sup> College of Bioengineering and Food Science, Hubei University of Technology, Hubei, China

Heat treatment and cooking are common interventions for reducing the numbers of vegetative cells and eliminating pathogenic microorganisms in food. Current cooking method requires the internal temperature of beef patties to reach 71◦C. However, some pathogenic Escherichia coli such as the beef isolate E. coli AW 1.7 are extremely heat resistant, questioning its inactivation by current heat interventions in beef processing. To optimize the conditions of heat treatment for effective decontaminations of pathogenic E. coli strains, sufficient estimations, and explanations are necessary on mechanisms of heat resistance of target strains. The heat resistance of E. coli depends on the variability of strains and properties of food formulations including salt and water activity. Heat induces alterations of E. coli cells including membrane, cytoplasm, ribosome and DNA, particularly on proteins including protein misfolding and aggregations. Resistant systems of E. coli act against these alterations, mainly through gene regulations of heat response including EvgA, heat shock proteins, <sup>E</sup>σ and <sup>S</sup>σ , to re-fold of misfolded proteins, and achieve antagonism to heat stress. Heat resistance can also be increased by expression of key proteins of membrane and stabilization of membrane fluidity. In addition to the contributions of the outer membrane porin NmpC and overcome of osmotic stress from compatible solutes, the new identified genomic island locus of heat resistant performs a critical role to these highly heat resistant strains. This review aims to provide an overview of current knowledge on heat resistance of E. coli, to better understand its related mechanisms and explore more effective applications of heat interventions in food industry.

#### Edited by:

Jean-Christophe Augustin, Ecole Nationale Vétérinaire d'Alfort, France

#### Reviewed by:

Louis Coroller, University of Western Brittany, France Sergio I Martinez-Monteagudo, South Dakota State University, USA

# \*Correspondence:

Michael Gänzle mgaenzle@ualberta.ca

#### Specialty section:

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

Received: 26 July 2016 Accepted: 20 October 2016 Published: 03 November 2016

#### Citation:

Li H and Gänzle M (2016) Some Like It Hot: Heat Resistance of Escherichia coli in Food. Front. Microbiol. 7:1763. doi: 10.3389/fmicb.2016.01763

Keywords: Escherichia coli, heat resistance, VTEC, food processing, protein, locus of heat resistance

# INTRODUCTION

Pasteurization and domestic cooking are common interventions for reducing the numbers of vegetative bacterial cells including pathogens in food. Heat kills vegetative bacterial cells by inactivation of cellular components, particularly membranes, proteins, and ribosomes (Tsuchido et al., 1985; Mackey et al., 1991; Mohácsi-Farkas et al., 1999; Lee and Kaletunc, 2002). Thermal food processing has an excellent record of establishing and maintaining food safety. However, consumer preferences for raw or minimally processed food, and the aim to minimize thermal degradation of nutrients are incentives to reduce the intensity of thermal processing. Moreover, fresh foods including meats and produce cannot be heated to temperature that are lethal to all pathogens, and bacterial pathogens are highly resistant to thermal processing in the dry state (Santillana Farakos et al., 2014; Syamaladevi et al., 2016). In addition, the heat resistance of pathogens is variable and

heat resistant strains may withstand thermal processes that are lethal to the majority of strains of the same species (Ng et al., 1969; Murphy et al., 1999; Dlusskaya et al., 2011).

Escherichia coli has been considered to be a relatively heat sensitive organism; however, strains of E. coli belong to the most heat resistant vegetative foodborne pathogens (**Figure 1**; Jay et al., 2005; Doyle and Beuchat, 2013). Heat resistant E. coli have D<sup>60</sup> value of more than 6 min (**Figure 1**; Liu et al., 2015; Mercer et al., 2015), and their resistance matches or exceeds Salmonella Senftenberg 755 with D<sup>60</sup> of 6.3 min (Ng et al., 1969; Baird-Parker et al., 1970) and Staphylococcus aureus with D<sup>60</sup> of 4.8−6.5 min (Jay et al., 2005; Kennedy et al., 2005; Doyle and Beuchat, 2013). Foodborne disease with E. coli has been linked to consumption of meat and meat products as well as fruits and fresh produce (Frenzen et al., 2005; Karch et al., 2005; Greig and Ravel, 2009; Yeni et al., 2015). Heat treatments for effective microbial decontamination and minimum organoleptic deterioration of foods (Woodward et al., 2002; Klaiber et al., 2005; Rajic et al., 2007) necessitate knowledge of the heat resistance of target foodborne pathogens as well as factors influencing heat resistance. This review aims to provide an overview of current knowledge on mechanisms of heat resistance of E. coli to provide novel perspectives on conventional and novel thermal processing of foods. Major mechanisms of heat resistance are active in all strains of E. coli; however, relatively few studies elucidated genetic determinants for strain-specific acquisition of heat resistance. A recently identified genomic island termed locus of heat resistance (LHR) substantially increases the heat resistance of about 2% of strains of E. coli (Mercer et al., 2015). Where appropriate, E. coli will be compared to Salmonella enterica, a closely related organisms exhibiting comparable resistance to heat.

# VARIABILITY OF RESISTANCE OF STRAINS OF E. coli TO HEAT

The D60-value of E. coli K12 is reported as 0.1 to 0.3 min (Chung et al., 2007; Jin et al., 2008; Dlusskaya et al., 2011); however, a majority of strains of E. coli exhibits D60-values exceeding that value up to 10-fold (**Figure 1**). Heat resistance is not related to the phylogenetic group, the serotype, or the virotype of E. coli (Liu et al., 2015; Mercer et al., 2015). Highly heat resistant strains of E. coli exhibit D60◦<sup>C</sup> values exceeding 10 min (Dlusskaya et al., 2011; Garcia-Hernandez et al., 2015). Genetic determinants of the variability of heat resistance between strains are only partially understood. An overview on isogenic mutant strains of E. coli and their heat resistance is shown in **Table 1**. Genes that are related to the heat shock response, including the alternative sigma factors σ <sup>H</sup> and σ E , the heat shock proteins (HSPs) IbpA/B, the alternative sigma factor σ S regulating the general stress response, the oxidative stress response regulated by SodA/B, and genes related to envelope properties including synthase of colanic acid, cyclopropane fatty acids (CFAs), NmpC and EvgA relate to heat resistance (**Table 1** and references therein). E. coli strains deficient of in σ <sup>H</sup>, σ S , SodA/B, IbpA/B, and colanic acid as well as CFAs were more sensitive to heat compared to their isogenic parental strains. Overexpression of EvgA increased heat



LB, Luria-Bertani; TY, Tryptone-yeast extract.

resistance (**Table 1**). The LHR (**Table 1**) mediates extreme heat resistance with D60-values of 10 min or higher (**Table 1**). The heat resistance of strains of E. coli also depends on the food matrix (**Table 2**). The resistance of E. coli LTH5807 to heating on mung bean, radish, or alfalfa seeds differed substantially (**Table 2**). The survival of the LHR-positive E. coli AW1.7 in beef patties cooked to 71◦C provides further evidence that the heat resistance of E. coli depends on the food matrix. Heat treatments that are considered to be lethal to E. coli thus may fail to safely eliminate contaminating E. coli (**Table 2**).

# MECHANISMS RELATED TO OUTER MEMBRANE AND MEMBRANE FLUIDITY

Cell surface structures and appendages provide the first line of defense to environmental stress. An overview of heat stress responses related to cell membranes and the periplasm is provided in **Figure 2**. Most strains of E. coli secrete extracellular polysaccharides, including colanic acid, which forms a thick mucoid matrix on the cell surface (Whitfield and Valvano, 1993; Mao et al., 2001). A colanic acid-deficient mutant of E. coli M4020, obtained by insertional disruption of the wsc genes required for colanic acid biosynthesis, was less tolerant to exposure to 55 and 60◦C than its parental strain E. coli O157:H7 W6-13 (**Table 1**), indicating that colanic acid confers heat resistance to E. coli O157:H7 (**Figure 2**) (Mao et al., 2001). Lipopolysaccharide (LPS) serves as a barrier to prevent rapid penetration of hydrophobic molecules, and is stabilized by divalent cations, particularly Mg2<sup>+</sup> and Ca2<sup>+</sup> (**Figure 2**) (Hitchener and Egan, 1977; Vaara, 1992; Hauben et al., 1998; Li et al., 2016). Expression of the outer membrane porin NmpC increased survival of E. coli GGG10 at 60◦C by 50- to 1,000-fold (**Figure 2**) (Ruan et al., 2011). The outer membrane permeabilizing polysaccharide chitosan decreased the heat resistance of E. coli in apple juice at 60◦C (Liu, 2015). The pronounced effect on heat resistance of chitosan occurred on EHEC when combined with rutin or resveratrol in beef patties, due to the greater bacterial destruction from outer membrane to cytoplasmic membrane (Nair et al., 2016).

The fluidity of the membrane influences its function (Zhang and Rock, 2008). The adjustment of membrane lipid composition and membrane fluidity by homoviscous adaptation is a major contributor to the bacterial resistance to heat stress (Sinensky, 1974; Arneborg et al., 1993; Denich et al., 2003; Yuk and Marshall, 2003; Yoon et al., 2015). Adaptive systems responding to heat stress in E. coli contribute to the stabilization of membrane-bound enzymes, and affect physical properties of the cytoplasmic membrane (Torok et al., 1997; Beney and Gervais, 2001). Remarkably, heat resistance induced by slow heating of E. coli was related to adaptation of the membrane fluidity rather than protein synthesis (Guyot et al., 2010). Heatadaptation increased the heat resistance of E. coli strains by the maintenance of the membrane in the liquid-crystalline state. The incorporation of saturated fatty acids into membrane lipids reduces membrane fluidity (Nakayama et al., 1980; Katsui et al., 1981) and consequently antagonizes the heat-induced increase in fluidity (**Figure 2**) (Quinn, 1981; De Mendoza and Cronan, 1983; Suutari and Laakso, 1994; Mejía et al., 1995; Yuk and Marshall, 2003). The heat resistant E. coli AW1.7 was characterized by a higher proportion of saturated and CFAs in the cytoplasmic membrane when compared to heat sensitive strains of E. coli

#### TABLE 2 | Examples of heat resistance of E. coli strains in food.


UDL, cell counts after treatment were under detection limit.

NC, no surviving cells after enrichment.

#Reductions depend on fat content from 15 to 35% in ground beef.

ˆThickness of beef steaks is 2.54 or 3.81 cm; initial cell counts are around 5.50 cfu/g.

∗∗Temperature is the surface temperature; cooking time refers to the time per side; initial cell counts are around 6.3−6.8 cfu/g.

(**Figure 2**) (Ruan et al., 2011). A contribution of CFAs to heat resistance of E. coli was confirmed by disruption of cfa coding for CFA synthase (Chen and Gänzle, 2016). The cfa deficient derivatives of E. coli AW1.7 and MG1655 did not produce CFAs; the unsaturated fatty acid C16:1 and C18:1 replaced CFAs in membrane lipids and the mutant strain was less resistant to heat when compared to the parent strains (**Figure 2**) (Chen and Gänzle, 2016).

#### REGULATION OF HEAT RESPONSE BY EvgA, HSPs, AND σ E

Cytoplasmic mechanisms of heat resistance relate to the effect of HSPs and compatible solutes on protein folding, and to oxidative stress (**Figure 3**). The regulation of the heat shock response of E. coli is governed by the two alternative sigma factors σ <sup>H</sup> and σ E (**Figure 3A**). The heat shock response is induced by temperatures around the growth/no-growth interface which aggravate protein misfolding but permit gene expression and protein synthesis (Lindner et al., 2008; Winkler et al., 2010; Govers et al., 2014; Lee et al., 2016). σ <sup>H</sup> and σ E are encoded by rpoH and rpoE, regulate transcription of heat-shock regulons coping with protein misfolding in the cytoplasm and the periplasm, respectively, and mediate cytoplasmic stress and envelope stress responses (Bukau, 1993). HSPs including chaperones and proteases function by holding partially unfolded proteins to prevent aggregation of heat-denatured proteins, and disaggregation of denatured proteins to allow refolding or proteolytic degradation (Parsell and Lindquist, 1993; Landini et al., 2014; Lee et al., 2016). The small HSPs IbpA and IbpB are holdases; DnaK, DnaJ, GrpE facilitate protein folding during translation, and guide aggregated proteins to the disaggregase ClpB. ClpP and other heat-shock proteases degrade aggregated proteins. The expression of HSPs is induced by σ <sup>H</sup> under sublethal heat stress and increases heat resistance of E. coli (Arsène et al., 2000). A σ <sup>H</sup> deletion in E. coli eliminated synthesis of HSPs including DnaK, GroEL, and HtpG and the resulting strain was very sensitive to exposure to 57◦C (**Table 1**). Starvation significantly enhanced the heat resistance of this strain (Jenkins et al., 1991). Small HSPs prevent protein aggregation by

solute transport proteins and the outer membrane porin NmpC contribute to heat resistance of E. coli AW1.7 (Ruan et al., 2011). Addition of antimicrobials including chitosan decreased the heat resistance due to the increased permeability of outer membrane (Liu, 2015). The master transcriptional regulator evgA is a cytoplasmic protein that increased heat resistance through activation of genes involved in periplasmic functions (Christ and Chin, 2008). The alternative sigma factors σ <sup>S</sup> and σ E also influence the properties of cell envelope (Lange and Hengge-Aronis, 1991; Bukau, 1993). LPS proteins SurA and PpiD lead to overall reduction in the level and folding of outer membrane proteins, consequently induce the periplamic heat shock response (Missiakas et al., 1996; Dartigalongue and Raina, 1998). Incorporating more saturated fatty acids such as palmitic acid and cyclopropane fatty acids (CFAs) into membrane lipids antagonizes the heat-induced increase in fluidity and achieves an ideal physical state of membrane (Katsui et al., 1981; Ruan et al., 2011; Chen and Gänzle, 2016). Disruption of cfa coding for CFA synthase of E. coli AW1.7 and MG1655 induced accumulation of the unsaturated fatty acid C16:1 and C18:1 in membrane lipids, consequently reducing the heat resistance of them (Chen and Gänzle, 2016).

heat (Jakob et al., 1993; Lee et al., 1997; Kitagawa et al., 2000; Mogk et al., 2003). Overexpression of IbpA and IbpB increased resistance not only to heat but also to superoxide (Kitagawa et al., 2000; **Table 1**). Small HSPs IbpA and IbpB prevent the aggregation of denatured endogenous proteins (Laskowska et al., 1996; Veinger et al., 1998; Kuczyñska-Wi´snik et al., 2002). The DnaK system also prevented protein aggregation induced by heat. This disaggregation is more efficient when DnaK acts in concert with ClpB (Mogk et al., 1999, 2003). However, disruption of clpA, htpG, and ibp in E. coli did not affect the viability at 50◦C (Thomas and Baneyx, 1998). The pressure resistant strains E. coli LMM1010, LMM1020, and LMM 1030 exhibit an increased basal expression of HSPs including DnaK, Lon, and ClpX; this increased expression may also account for the moderate increase of heat resistance of these strains (Hauben et al., 1997; Aertsen et al., 2004). Overall, the inducible heat shock response is a key contributor for growth of E. coli at temperature exceeding the optimum temperature of growth, but it makes only a modest contribution to the strain-specific differences of the resistance to lethal heat challenge.

Four key proteins involve in the regulation of σ E -dependent envelope stress response, including RseA, RseB, DegS, and Yael (Alba and Gross, 2004). The activity of σ E is modulated by the expression of outer membrane proteins and outer membrane proteins induce σ E activity (Mecsas et al., 1993). Moreover, deletions of LPS proteins SurA and PpiD lead to overall reduction in the level and folding of outer membrane proteins, and to the induction of the periplamic heat shock response (**Figure 2**) (Missiakas et al., 1996; Dartigalongue and Raina, 1998).

A master transcriptional regulator evgA activates genes involved in periplasmic functions, as well as in membrane and permeability functions. Its overexpression significantly increases heat resistance of E. coli (Christ and Chin, 2008; **Table 1**; **Figure 2**). The response regulator EvgA is part of a twocomponent regulatory system with sensor kinase EvgS, binding the intergenic region of evgAS and emrKY coding for efflux

pump, and regulating the expression of both operons (Kato et al., 2000). Comparison of the genome-wide transcription profile of EvgA-overexpressing and EvgA-lacking strains revealed that EvgA conferred acid resistance to E. coli (Masuda and Church, 2002). EvgA controls the expression of wide range of genes, including gadABC, hdeAB, emrKY, yhiUV, and yfdX which are related to acid resistance, osmotic adaptation, drug resistance and other functions (Nishino et al., 2003).

### REGULATION OF HEAT RESISTANCE BY σ <sup>S</sup>, AND CROSS-RESISTANCE TO ACID, OXIDATIVE, AND HIGH PRESSURE STRESS

Stationary phase cells are more resistant than exponential phase cells, mainly because of the increased expression of σ S (**Figure 3A**) (Cheville et al., 1996; Kaur et al., 1998). The σ S regulon contributes to the general stress response and increase acid, heat, and / or osmotic resistance of E. coli (Hengge-Aronis et al., 1991; Cheville et al., 1996; Robey et al., 2001; Hengge-Aronis, 2002; Allen et al., 2008; Landini et al., 2014). Adaptation to acid stress provides cross-protection to heat stress (Ryu and Beuchat, 1998; Buchanan and Edelson, 1999; Ryu and Beuchat, 1999; Mazzotta, 2001; Yuk and Marshall, 2003). For example, adaptation of enterohemorrhagic E. coli to pH 4.6 increased the heat resistance at 58◦C 2−4 fold when compared to cells grown at pH 7.0 (Buchanan and Edelson, 1999). Induction of acid resistance in E. coli O157:H7 increases levels of CFAs in the cytoplasmic membrane (Brown et al., 1997), which stabilize cells against several environmental stressors including heat (Grogan and Cronan, 1997; Chen and Gänzle, 2016). Moreover, σ S dependent gene expression increased the heat resistance of E. coli O157:H7 after adaptation to temperatures above the optimum growth temperature (Cheville et al., 1996; Yuk and Marshall, 2003; **Table 1**). Starvation of E. coli O157:H7 substantially

increased D52-values; this enhanced heat resistance was related to the expression of starvation-induced proteins UspA and GrpE (Zhang and Griffiths, 2003).

Heat induces production of O<sup>2</sup> in E. coli under aerobic conditions, possibly by disruption of the electron transport systems of the membrane, and consequently induces the manganese-containing superoxide dismutase (Privalle and Fridovich, 1987). Accumulation of reactive oxygen species after exposure to sublethal stress results in lethal damage to DNA, RNA, proteins, and lipids (Aldsworth et al., 1999; Cabiscol et al., 2000; Aertsen et al., 2005). The general stress response factor σ S also protects against oxidative stress (**Figure 3C**) (Landini et al., 2014). The σ S -regulated DNA binding protein dps binds DNA as homo-dodecamer and prevents DNA damage by oxidative stress or low pH (Choi et al., 2000; Zhao et al., 2002). The synthesis of CFAs in E. coli also increases resistance to oxidative stress (Grogan and Cronan, 1997). Proteins that are alter the resistance of E. coli to pressure-induced oxidative stress, including systems for thiol-disulfide redox homeostasis and proteins containing iron−sulfur clusters, probably also contribute against oxidative stress induced by heat (Malone et al., 2006; Charoenwong et al., 2011; Imlay, 2013; Gänzle and Liu, 2015).

Oxidative stress induced by sublethal thermal damage may also account for the phenomenon termed "viable but nonculturable state" (VBNC). VBNC cells cannot be detected by standard culture techniques but can be resuscitated under favorable conditions (Bogosian et al., 2000; Gupte et al., 2003; Morishige et al., 2013). Addition of sodium pyruvate recovered cells of E coli after heat-induced sublethal injury. This protective effect was related to the ability of pyruvate to degrade hydrogen peroxide (Czechowicz et al., 1996; Mizunoe et al., 2000). Addition of sodium pyruvate or catalase to medium agar also resuscitated VBNC Salmonella Enteritidis or Vibrio vulnificus cells, respectively, which had become sensitive to hydrogen peroxide (Bogosian et al., 2000; Morishige et al., 2013).

# EFFECTS OF SALT OR SUGAR ADDITION IN HIGH MOISTURE FOODS

The water activity of food and particularly the salt content influence the heat resistance of E. coli. E. coli responds to an increase of the osmotic pressure by accumulation or synthesis of compatible solutes, small organic solutes that balance the osmotic pressure without interfering with cytoplasmic functions (Kempf and Bremmer, 1998). High cytoplasmic concentrations of compatible solutes increase heat resistance of E. coli and other bacterial cells by stabilizing ribosomes and proteins through a mechanisms referred to as "preferential hydration" (**Figure 3B**) (Ramos et al., 1997; Lamosa et al., 2000; Pleitner et al., 2012). A reduction in water activity from 0.995 to levels between 0.98 and 0.96 in salt or sucrose solutions significantly enhanced the heat resistance of E. coli (Kaur et al., 1998). The heat resistance of several strains of E. coli was also increased by addition of 2–6% of NaCl (Garcia-Hernandez et al., 2015). Addition of 2% NaCl resulted in the accumulation of amino acids including glycine betaine and proline as major cytoplasmic solutes; accumulation of carbohydrates including glucose and trehalose occurred in response to the addition of 6% NaCl (Pleitner et al., 2012). The accumulation of solutes corresponded to an increased heat resistance of E. coli, and a higher thermal stability of ribosomes (Pleitner et al., 2012). The effect of NaCl addition on solute accumulation and heat resistance of E. coli is observed at concentrations that are typical for food systems. A critical concentration of NaCl in ground beef, about 2.7−4.7%, substantially increased heat resistance of E. coli O157:H7 at 55−62.5◦C (Juneja et al., 2015). In addition, pre-exposure to 5% NaCl at room temperature for 24 h increased the heat resistance of E. coli O157:H7 at 55◦C (Bae and Lee, 2010).

The effect of the fat content on heat resistance of E. coli is controversial. An increased fat content in food products increased the heat resistance of E. coli in some studies (Line et al., 1991; Huang et al., 1992; Ahmed et al., 1995; Smith et al., 2001; Liu et al., 2015), while other studies reported decreased resistance, no effect, or strain-specific effects (Kotrola and Conner, 1997; Vasan et al., 2014; Liu et al., 2015). The potential direct effects of fat on heat resistance of E. coli are confounded by the strong effect of fat on heat transfer in solid foods. Reduced heat transfer increases the heating times to a certain target temperature and thus profoundly affects process lethality.

# LHR AND EXTREME RESISTANCE TO HEAT

Extreme heat resistance of E. coli is conferred by the LHR (**Figure 3D**, Mercer et al., 2015). The LHR is a genomic island of about 14 kbp which encodes for 16 genes; six of these genes are unique to heat resistant strains of E. coli (Mercer et al., 2015). Acquisition of the LHR increases survival after exposure to 60◦C for 5 min by more than 7 log(cfu/mL); the LHR is thus one of the most powerful mediators of heat resistance in E. coli (**Table 1**; Mercer et al., 2015). Loss of the LHR also reduces the pressure resistance in E. coli AW1.7 (Garcia-Hernandez et al., 2015; Liu et al., 2015; Mercer et al., 2015). Remarkably, the presence of a truncated LHR in wild type strains of E. coli, or cloning of fragments of the LHR had little effect on heat resistance, indicating that the 16 genes act in concert to provide heat resistance in LHR-positive strains (Mercer et al., 2015). A genomic island with high similarity to the LHR, the Pseudomonas aeruginosa clone C-specific genomic island (PACGI-1) was characterized in Pseudomonas (Lee et al., 2015).

The 16 predicted open reading frames (ORF) within LHR encode small HSPs (Orf2 and Orf7), proteins of the YfdX family with unknown function (Orf8 and Orf9), heat shock proteases (Orf3, Orf15 and Orf16), thioredoxin (Orf12), and a sodium/hydrogen antiporter (Orf13) (Mercer et al., 2015). According to the predicted function of proteins encoded by the LHR, the genomic island may thus contribute to the turnover of misfolded or aggregated proteins, the osmotic stress response, and mitigate oxidative stress (Mercer et al., 2015). The contribution of genes encoded by the LHR to protein folding and protein turnover was confirmed in the homologous

gene cluster PACGI-1 in P. aeruginosa (Lee et al., 2015). The small HSPs sHsp20c and ClpGGI contribute to thermotolerance in P. aeruginosa through their function as holdases and disaggregating chaperones (Lee et al., 2015, 2016). Cloning of the homologous LHR proteins in E. coli, however, had no influence on the heat resistance in E. coli (Mercer et al., 2015), demonstrating that the effect of LHR-encoded genes is species specific, and that extreme heat resistance in E. coli necessitates HSPs acting in concert with other biochemical functions.

# HEAT RESISTANCE OF DESICCATED E. coli

Desiccated strains of E. coli and Salmonella are characterized by extreme resistance to physical and chemical stressors including heat (Beuchat and Scouten, 2002; Beuchat et al., 2013; Studer et al., 2013; Syamaladevi et al., 2016). Parameters for the heat inactivation of dry bacterial cells are comparable to the moist heat inactivation of bacterial endospores spores rather than pasteurization (Brandl et al., 2008; Du et al., 2010; Podolak et al., 2010). Hot air roasting of almonds even at very high temperature (130-150 ◦C) achieve less than a 4 log (cfu/g) reduction of Salmonella on almonds (Yang et al., 2010). Similarly a 2 log (cfu/g) reduction of Salmonella on dry alfalfa seeds required 10 days of treatment at 60◦C; an equivalent bactericidal effect was achieved after 5 min of treatment with wet heat at 60◦C (Jaquette et al., 1996; Neetoo and Chen, 2011).

Mechanisms of dry heat resistance are best understood for Salmonella (Podolak et al., 2010; Finn et al., 2013). The heat resistance of Salmonella at 75◦C in meat and bone meal was higher at a<sup>W</sup> 0.77 than at a<sup>W</sup> 0.88 (Riemann, 1968). Comparable to the effect of NaCl in high-moisture foods, the heat resistance of dry cells is related to the intracellular concentration of compatible solutes, including K+, glutamate and trehalose. The up-regulation of σ S , σ E , fatty acid catabolism, and formations of Fe−S clusters and filaments also contribute to the resistance to dry conditions (Finn et al., 2013). It was speculated that the extent and strength of the vibration of water molecules in dry bacteria are limited substantially because of the very low water contents. The low water content thus prevents denaturation of cytoplasmic and membrane proteins even at very high temperatures (Earnshaw et al., 1995; Archer et al., 1998). This mechanism was proposed in analogy to bacterial endospores, where the reduced core water reduces the amount of water associated with proteins, thus preventing thermal denaturation (Nicholson et al., 2000). Desiccation of bacterial

# REFERENCES


cells may also stabilize ribosomal units (Syamaladevi et al., 2016).

Several studies demonstrate that concepts and mechanisms that were identified in Salmonella are also relevant in E. coli. Desiccated VTEC survived at 70◦C for 5 h, thus exhibiting almost the same level of heat resistance as Salmonella (Hiramatsu et al., 2005). The lethality of treatments of radish seeds at 60◦C against E. coli O157:H7 increased as the a<sup>W</sup> increased from 0.25 to 0.65 and 1.0 (Kim et al., 2015). However, information on the dry heat resistance of E. coli remains limited when compared to the information on the wet heat resistance of the organisms.

# CONCLUSION

The resistance of E. coli strains to heat intervention treatments has been widely evaluated in the past decades, particularly using strains of E. coli O157: H7. Although E. coli has been considered as a relatively heat sensitive organisms, the D60- values of some strains of E. coli are increased to several minutes or even hours by the heat shock response, adaptation to salt or acid stress, acquisition of the LHR, or desiccation (**Figure 1**). About 2% of E. coli including food isolates and pathogens harbor the LHR and exhibit extreme resistance to wet heat (Mercer et al., 2015). The biochemical function of the LHR links to proteins aggregation and folding as well as thiol- and ion homeostasis, however, the mechanisms of LHR –mediated heat resistance are only partially understood. Current pathogen intervention methods or cooking recommendations may not suffice to control these highly heat resistant strains of E. coli (Dlusskaya et al., 2011; Liu et al., 2015; Mercer et al., 2015). Additional hurdles need therefore to be developed to assure the inactivation of highly heat resistant strains. Further evaluations on inactivation of heat resistant strains under improved heat interventions and mechanisms of heat resistance allow us to design more effective applications in food industry.

# AUTHOR CONTRIBUTIONS

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

# ACKNOWLEDGMENTS

HL is supported by China Scholarship Council. The Alberta Livestock and Meat Agency and Alberta Innovates Biosolutions are acknowledged for financial support.

pressure in Escherichia coli. Appl. Environ. Microbiol. 70, 2660–2666. doi: 10.1128/AEM.70.5.2660-2666.2004




in Escherichia coli. FEMS Microbiol. Lett. 184, 165–171. doi: 10.1111/j.1574- 6968.2000.tb09009.x




Escherichia coli and consequences on protein quality control and cellular ageing. EMBO J. 29, 910–923. doi: 10.1038/emboj.2009.412


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

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

# High Heating Rates Affect Greatly the Inactivation Rate of Escherichia coli

Juan-Pablo Huertas <sup>1</sup> , Arantxa Aznar <sup>1</sup> , Arturo Esnoz <sup>1</sup> , Pablo S. Fernández 1, 2 , Asunción Iguaz <sup>1</sup> , Paula M. Periago1, 2 and Alfredo Palop1, 2 \*

<sup>1</sup> Departamento de Ingeniería de Alimentos y del Equipamiento Agrícola, Escuela Técnica Superior de Ingeniería Agronómica, Universidad Politécnica de Cartagena, Cartagena, Spain, <sup>2</sup> Unidad de Microbiología y Seguridad Alimentaria, Instituto de Biotecnología Vegetal, Universidad Politécnica de Cartagena, Cartagena, Spain

Heat resistance of microorganisms can be affected by different influencing factors. Although, the effect of heating rates has been scarcely explored by the scientific community, recent researches have unraveled its important effect on the thermal resistance of different species of vegetative bacteria. Typically heating rates described in the literature ranged from 1 to 20◦C/min but the impact of much higher heating rates is unclear. The aim of this research was to explore the effect of different heating rates, such as those currently achieved in the heat exchangers used in the food industry, on the heat resistance of Escherichia coli. A pilot plant tubular heat exchanger and a thermoresistometer Mastia were used for this purpose. Results showed that fast heating rates had a deep impact on the thermal resistance of E. coli. Heating rates between 20 and 50◦C/min were achieved in the heat exchanger, which were much slower than those around 20◦C/s achieved in the thermoresistometer. In all cases, these high heating rates led to higher inactivation than expected: in the heat exchanger, for all the experiments performed, when the observed inactivation had reached about seven log cycles, the predictions estimated about 1 log cycle of inactivation; in the thermoresistometer these differences between observed and predicted values were even more than 10 times higher, from 4.07 log cycles observed to 0.34 predicted at a flow rate of 70 mL/min and a maximum heating rate of 14.7◦C/s. A quantification of the impact of the heating rates on the level of inactivation achieved was established. These results point out the important effect that the heating rate has on the thermal resistance of E. coli, with high heating rates resulting in an additional sensitization to heat and therefore an effective food safety strategy in terms of food processing.

Keywords: heat resistance, heating rate, Escherichia coli, heat exchanger, thermoresistometer

# INTRODUCTION

Microbial heat resistance studies are necessary for the safe production of heat processed foods. The knowledge provided by these studies on microbial destruction kinetics and on the mechanisms of inactivation has allowed the design and development of safe processes, eliminating the risk of foodborne pathogen and spoilage microorganisms. Also, the correct application of thermal treatments results in avoiding overprocessing of food products.

Thermal resistance of microorganisms is affected by many different factors. Some of the most influencing factors are the water activity, nutrient content, pH of the

#### Edited by:

Lorena Ruiz, Universidad Complutense de Madrid, Spain

#### Reviewed by:

Alexandra Lianou, Agricultural University of Athens, Greece Francisco Diez-Gonzalez, University of Georgia, USA

> \*Correspondence: Alfredo Palop alfredo.palop@upct.es

#### Specialty section:

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

Received: 16 May 2016 Accepted: 29 July 2016 Published: 11 August 2016

#### Citation:

Huertas J-P, Aznar A, Esnoz A, Fernández PS, Iguaz A, Periago PM and Palop A (2016) High Heating Rates Affect Greatly the Inactivation Rate of Escherichia coli. Front. Microbiol. 7:1256. doi: 10.3389/fmicb.2016.01256 heating medium, growth phase and growth temperature of the microbial culture, as well as the genus, species and even the strain within the same species. The research on these factors has been usually performed under isothermal treatment conditions. However, heat treatments applied in food industry comprise nonisothermal stages (corresponding to heating and cooling phases), which may be even more important than the isothermal stage (holding phase) in terms of inactivation of microorganisms.

One of the factors influencing heat resistance to which authors have paid less attention is the heating rate, probably because of the lack of appropriate equipment to measure this effect. This fact has led some authors to develop and use non-isothermal methods as an alternative to understand microbial inactivation kinetics under these heating conditions (Reichart, 1979; Periago et al., 1998; Fernández et al., 1999; Conesa et al., 2003; Hassani et al., 2005; Valdramidis et al., 2006; Van Derlinden et al., 2010; Esteban et al., 2013). Some of these researches (De Cordt et al., 1992; Periago et al., 1998) have shown that there are differences between the heat resistance values obtained under isothermal and non-isothermal heating conditions.

Many different heat resistance determination methods and instruments have been used (Stumbo, 1973; Brown and Ayres, 1982; Palop et al., 2012), each of them having their own advantages and drawbacks. In 2009 Conesa et al. built the thermoresistometer Mastia, where most advantages of the existing methods were incorporated. Its only limitations were that the maximum heating and cooling rates it was able to provide were about 35◦C/min, and that it worked as a batch system. These heating and cooling rates were fast enough to mimic batch heating systems, such as retorts (Lewis, 2006), but did not achieve the faster heating rates reached at continuous heating systems, such as heat exchangers. Continuous processing minimizes the exposure time of food products at high temperatures because of the high heating and cooling rates reached on these systems, reducing the adverse effects of thermal treatments on food quality and also minimizing the processing times (Tucker et al., 2002). These limitations led Huertas et al. (2015) to build a pilot plant heat exchanger, in which it was possible to mimic in-flow processes, measure the temperature and take several samples along its pipelines, enabling to build survival curves at faster heating rates. However, this heat exchanger cannot reach very high heating rates, which in some processes, such as HTST or UHT systems could be almost instantaneous. This limitation is a hindrance on the exploration of the effect of heating rate on microbial heat inactivation. Still, the thermoresistometer Mastia can be used for continuous heating processes, in which much faster heating rates could be achieved.

The objectives of this research were to explore the effect of high heating rates on the thermal inactivation of E. coli and to evaluate the thermoresistometer as a continuous heating system.

# MATERIALS AND METHODS

## Microorganisms

Escherichia coli type strain (CECT 515) was provided by the Spanish Type Culture Collection (CECT). Cells were grown overnight at 37◦C in tryptic soy broth (TSB; Scharlau Chemie) supplemented (w/v) with 0.6% yeast extract (YE; Scharlau Chemie), until the stationary phase of growth was reached.

# Heating Medium

Citrate phosphate pH 7 McIlvaine buffer was prepared as described by Dawson et al. (1974) and was stored at 0–5◦C until used.

# Determination of Heat Resistance in the Heat Exchanger

A pilot plant scale, double tube heat exchanger (Huertas et al., 2015) was used. The heating medium inoculated with the microorganism was pumped through the product pipe, as described by Huertas et al. (2015), into the system at different flow rates, 480 and 780 mL/min. The heat exchanger was programmed to raise the temperature of the product to 60 or 65◦C. The sampling points located along the heating sections enabled to measure the temperature and to take samples for microbiological analysis along with heating. Three different maximum heating rates were reached: 21◦C/min, with a flow rate of 480 mL/min and a final temperature of 60◦C and 32◦C/min, with a flow rate of 780 mL/min and a final temperature of 60◦C and 50◦C/min, with a flow rate of 780 mL/min and a final temperature of 65◦C. The experiments were repeated at least three times.

# Determination of Heat Resistance in the Thermoresistometer

The thermoresistometer Mastia (Conesa et al., 2009) was used in a continuous mode. This operating mode consisted in filling the vessel of the thermoresistometer with water and heating it at a preset temperature. Then, the process medium, already inoculated with the microorganisms, was circulated through the cooling system coil by means of a peristaltic pump (Selecta, Barcelona, Spain) at a controlled flow. In this way, the instrument works as a heat exchanger, with a constant temperature of the water used as heating fluid. The dimensions of the coil were 160 cm long (110 cm were immersed inside the thermoresistometer and 25 cm corresponded to each branch outside the instrument), 3.2 mm of inside diameter and a total volume of 12 mL. The coil was previously sterilized in situ. The input temperature of the microbial suspension, before pumping through the coil, was 20◦C in all cases. The output temperature of the microbial suspension was continuously measured after passing through the coil, by means of a thermocouple located just after the 25 cm long output branch of the coil, well outside the vessel. After this probe, a short silicon tube was placed to enable sampling for microbiological counts. When the output temperature was constant the system was in steady state. Then, suspension samples were taken and quickly cooled at room temperature. In this way, very fast heating (estimated as described in Temperature Profile and Mean Residence Time Estimation) was achieved (up to 22.5◦C/s), followed by a short holding period and an instantaneous cooling, similar to those on continuous food pasteurization treatments.

Experiments were carried out keeping a constant temperature of 65◦ or 70◦C inside the vessel of the thermoresistometer, and passing the bacterial suspension through the coil at different speeds of the peristaltic pump. At least 3 samples were taken at each flow and the numbers of survivors were determined.

# Temperature Profile and Mean Residence Time Estimation

The temperature profiles in the pilot plant heat exchanger were obtained with the Pt-100 temperature probes placed in each elbow, as described elsewhere (Huertas et al., 2015). Mean residence times were also calculated as described by Huertas et al. (2015).

The temperature profiles in the coil of the thermoresistometer were estimated according to Son and Singh (2002). The energy balance applied to a differential volume of the coil of length dl provides:

$$dq = U \cdot 2\pi \cdot r \cdot dl \cdot (T\_{tr} - T) = m \cdot c\_p \cdot dT \tag{1}$$

Solving this differential equation, the same expression proposed by Deindoerfer and Humphrey (1959) was obtained:

$$\ln\left(\frac{T - T\_{tr}}{T\_i - T\_{tr}}\right) = -\frac{U \cdot 2\pi \cdot r}{m \cdot c\_p} \cdot l \tag{2}$$

where T is the temperature of bacterial suspension inside the coil at a distance l from the inlet (◦C), Ttr is the thermoresistometer temperature (◦C), T<sup>i</sup> is the temperature of bacterial suspension at coil inlet (◦C), U is the overall heat transfer coefficient (W/m<sup>2</sup> × ◦C), r is the coil internal radius (m), l is the distance from the inlet of the coil (m), m is the mass flow rate (kg/s),c<sup>p</sup> is the specific heat capacity of the suspension (J/kg × ◦C).

The overall heat transfer coefficient (U) was estimated according to Dichfield et al. (2006). This value was used to calculate the temperature profile inside the coil.

A small decay in the temperature corresponding to the output branch of the coil, outside the instrument, was observed at flows slower than 100 mL/min, even isolating the output coil pipe. This small decay was estimated as previously stated, with Equation (2).

Mean residence times in the coil were calculated for each flow and are shown in **Tables 1**, **2**.

Enumeration of Survivors

Viable counts were based on duplicate counts, from appropriate dilutions, in tryptic soy agar (TSA; Scharlau Chemie)+ 0.6% YE. The plates were incubated for 24 h at 37◦C. Preliminary experiments showed that longer incubation times did not modify plate counts.

## Data Analysis

Experimental data in the present research were obtained under non-isothermal conditions. These experimental data were contrasted against survivor numbers predicted from D<sup>T</sup> and z values obtained under isothermal conditions in a previous study (Conesa et al., 2009). In that research, D<sup>T</sup> values were calculated from the slope of the regression line of survival curves as given by the Bigelow model (Equation 3):

$$\log N\_t = \log N\_0 - \frac{t}{D\_T} \tag{3}$$

where N<sup>t</sup> is number of microorganisms at time t and N<sup>0</sup> is the initial number of microorganisms.

To predict the number of survivors in the present research, a rate model derived from the Bigelow model (Equation 3), representing the momentary time-dependent isothermal logarithmic inactivation rate was used. This rate model considers the non-isothermal treatments as composed of successive isothermal treatments of very short (differential) duration, each one at a different temperature, and hence can be written as an ordinary differential equation as given by Equation (4):

$$\frac{d\log N}{dt} = \frac{-1}{D\_T} \tag{4}$$

with the initial condition N(0) = N0.

The calculation of the D values for each of these different temperatures was based on the dependence of D with respect to temperature, which can be described with the classic Bigelow model as given by Equation (5):

$$D\left(T\right) = \frac{D\_{T\_{ref}}}{10^{\frac{T-T\_{ref}}{\varepsilon}}}\tag{5}$$

TABLE 1 | Mean residence time, maximum heating rate, outlet temperature and number of log cycles inactivated in the thermoresistometer under a constant temperature of 65◦C at different flows vs. their corresponding predicted inactivation values.

TABLE 2 | Mean residence time, maximum heating rate, outlet temperature and number of log cycles inactivated in the thermoresistometer under a constant temperature of 70◦C at different flows vs. their corresponding predicted inactivation values.



where DTref is the D(T) value at the reference temperature (Tref), and z is the number of degrees Celsius change of temperature required to achieve a tenfold change in D-value.

Significant differences between counts of the sample replicates and the experiment repetitions were analyzed by ANOVA test (Statgraphics 5.1 plus, Manugistics Corp., Rockville, MD, USA) at the 95 % confidence level.

# RESULTS

Thermal resistance characterization of E. coli CECT 515 under isothermal heating conditions in pH 7 McIlvaine buffer was taken from a previous study. An average D<sup>60</sup> value of 0.38 min and a zvalue of 4.7◦C were obtained (Conesa et al., 2009), and were used in this research to predict the inactivation under non-isothermal heating conditions.

# Heat Resistance in the Heat Exchanger

**Figure 1** shows the heating profiles of the different experiments performed in the heat exchanger, together with the observed and predicted inactivation data. Since this heat exchanger is provided with temperature sensors and sampling points along the heating section (Huertas et al., 2015), it was possible to determine the temperature profiles and to obtain samples during the entire thermal process, which permits to follow the inactivation of the microorganisms along the heat exchanger and to observe the effect of the different heating rates on this inactivation. The maximum heating rate obtained at the beginning of the experiment was 50◦C/min, for a flow rate of 780 mL/min and a final temperature of 65◦C (**Figure 1C**) and the minimum was 21◦C/min, for a flow rate of 480 mL/min and a final temperature of 60◦C (**Figure 1A**). As it can be observed in **Figure 1**, the higher the heating rate, the earlier the inactivation of the bacterial cells starts: for the treatments with a maximum heating rate of 50◦C/min (**Figure 1C**), 32◦C/min (**Figure 1B**), and 21◦C/min (**Figure 1A**), about seven log cycles were inactivated in 60, 90, and 150 s respectively. These differences in time to inactivate were somehow expected since at higher heating rates, lethal temperatures are reached faster than at lower heating rates. Still, it is noteworthy that the inactivation obtained experimentally was much higher than the one predicted by using the isothermal data, under all the experimental conditions (**Figure 1**): in all cases, when the observed inactivation had reached about seven log cycles, the predictions were estimating about 1 log cycle of inactivation. Hence, all these non-isothermal heating profiles were more lethal than expected or, in other words, predictions were well within the fail-safe side in all cases.

# Heat Resistance in the Thermoresistometer

**Figure 2** shows the output temperature (after passing through the thermoresistometer preheated at 65◦C) for each flow and the corresponding number of survivors. **Figure 3** depicts the evolution of temperatures inside the coil for flows of 70, 77, 85, 95, and 106 mL/min, estimated by means of Equation (2), and considering the mean residence time, when the thermoresistometer was preheated at 65◦C. For flows lower than

100 mL/min the temperature decay in the output pipe of the coil, outside the instrument, was also estimated with Equation (2). At flows faster than 106 mL/min, residence times of the suspension inside the whole coil were lower than 6 s and the temperature decay in the output pipe was negligible. At flows faster than 95 mL/min, residence times were too short to reach the treatment temperature in the coil an almost no population reduction was observed, in spite of the fast heating rates (**Figure 2**). At flows between 70 and 95 mL/min, the slower the flow, the more heat inactivation was observed, since longer residence times were achieved while the output temperature was very similar in all cases (**Figure 2**).

Similar results were obtained when the thermoresistometer was preheated at 70◦C (data not shown). However, since the temperature reached inside the coil was higher, faster flows and shorter residence times were needed to achieve similar levels of inactivation.

The estimation of temperatures along the coil by means of Equation (2) enabled to predict the microbial inactivation reached by these treatments (based on heat resistance data obtained under isothermal heating by Conesa et al. (2009) together with Equations 4, 5), which are represented in **Tables 1**, **2**, together with the observed values of log cycles of inactivated bacteria. Flow rates, mean residence times, outlet temperatures and maximum heating rates reached by these treatments are also included in these tables. Heating rates between 14.7 and 22.5◦C/s (i.e., up to 1350◦C/min) were obtained at the beginning of these treatments (**Tables 1**, **2**), which are much higher than those achieved in the heat exchanger.

**Table 1** shows the results obtained for several flows when the constant temperature inside the vessel was 65◦C (those corresponding to the experiment depicted in **Figures 2**, **3**). At a flow rate of 106 mL/min scarcely 0.27 log cycles were inactivated (**Table 1**), probably because the mean residence time (6.8 s) in the coil was too short to achieve higher levels of inactivation, even when lethal temperatures (close to 61◦C; **Figure 2**) were reached. Actually, only 0.13 log cycles of inactivation were predicted for this thermal treatment (**Table 1**), but significant differences were found at this flow rate between observed and predicted values. At faster flow rates, similar low inactivation levels observed were shown (**Figure 2**), but no significant differences were found between observed and predicted values, because of the broader dispersion of the microbial counts at these faster flow rates. However, at slower flow rates, enhanced inactivation was observed, and significant differences between predicted and observed values increased at low flow rates. Hence, at a flow rate of 70 mL/min, the observed inactivation was more than 10 times greater than the predicted (4.07 log cycles inactivation observed; 0.34 log cycles inactivation predicted; **Table 1**). At slower flow rates, where low inactivation was also predicted, complete microbial inactivation was reached (data not shown).

**Table 2** presents the results obtained for a constant temperature inside the vessel of 70◦C. Again, slower flow rates led to higher levels of inactivation, although in this case, the differences between observed and predicted values were not as big as for a constant temperature inside the vessel of 65◦C. At flow rates slower than 95 mL/min, complete inactivation was reached (data not shown). The maximum difference between observed and predicted values was of about 9 times, with a flow of 133 mL/min (2.78 log cycles inactivation observed; 0.31 log cycles inactivation predicted; **Table 2**). Anyhow, and similarly to the results obtained in the heat exchanger (**Figure 1**), all the experimental inactivation values were significantly higher than their corresponding predictions, which means much higher inactivation than what could be expected from isothermal inactivation kinetics data and an additional safety measure because the predictions are on the "fail safe" side.

# DISCUSSION

When applying continuous treatments in the food industry, such as those achieved in heat exchangers, heating and cooling rates are fast, much faster than those provided with the thermoresistometer Mastia (up to 35◦C/min; Conesa et al., 2009). Experiments performed in the pilot plant heat exchanger reached heating rates as high as 50◦C/min (**Figure 1**). Inactivation higher than expected from the isothermal data was achieved under all the experimental conditions (**Figure 1**). Deviations were particularly important at the late heating times. At these heating times, the temperatures reached (about 60◦C in all cases) were lethal, but the previous thermal profile with an initial high heating rate, enhanced the lethality, leading to several extra log cycles of inactivation. These results prove that previous hypothesis regarding the more lethal effect of high heating rates than slow heating rates on the thermal resistance of this microorganism were correct (Conesa et al., 2009). These results are also in agreement with those obtained for other vegetative microorganisms, such as Salmonella or Staphylococcus aureus in this heat exchanger under similar treatment conditions (Huertas et al., 2015). Still, this pilot plant heat exchanger was not able to achieve the almost instantaneous heating rates that can be obtained under HTST and UHT treatments currently applied in the food industry. In order to overcome this limitation, the thermoresistometer Mastia was used in a continuous mode, using the coil as a heat exchanger, which enables to reach heating rates as high as 22.5◦C/s. Using the instrument in this mode, the effect of very fast heating on the heat resistance of E. coli vegetative cells was investigated. The results of these experiments further confirmed the effect of the heating rates obtained in the heat exchanger.

In all cases, both for experiments performed in the heat exchanger and in the coil of the thermoresistometer, observed inactivation values were higher than predicted. When trying to look for a correlation between the initial heating rate of these experiments and the "over-inactivation" reached, several difficulties raised: the higher initial heating rates are linked to the shorter mean residence times, and consequently to the lower predicted inactivation values (see **Tables 1**, **2**), so there were no similar treatments (in terms of predicted inactivation) with different heating rates. Also, the variability associated with bacterial counts, which is shown through the standard deviation values of **Tables 1**, **2** or through the data points shown in **Figures 1**, **2**, hampers this comparison. Still, the only case with a similar predicted inactivation value (0.17 log cycles) leads to 1.36 log cycles inactivation when the initial heating rate was 21.5◦C/s and the target temperature was 70◦C (**Table 2**), and to only 0.31 log cycles inactivation when the initial heating rate is of 17.1◦C/s and a target temperature of 65◦C (**Table 1**), which would point out to this correlation between the heating rate and the inactivation reached.

These results reveal clearly that fast heating is much more efficient than isothermal treatments in killing E. coli vegetative cells. Previous research on the effect of heating rates on this same strain of E. coli showed that when cells were exposed to nonisothermal treatments at slow heating rates (2◦C /min), their heat resistance was increased (Conesa et al., 2009), leading to less inactivation than expected. In this same research, it was shown that heating rates as high as 10◦C/min led to opposite results, showing more inactivation than expected. It was hypothesized that at 2◦C/min, some heat stress response could be induced, leading to an adaptation to heat, which would be absent at 10◦C/min (Conesa et al., 2009).

The literature on the effect of heating rates on the heat resistance of vegetative cells is scarce and only explores the effect of heating rates as high as 10◦C/min. The observed effects depend on the bacterial genus and on the heating rate value. Some bacterial genera became more heat resistant under high heating rates. For example, Hassani et al. (2006) showed that S. aureus exhibited higher thermal resistance at higher heating rates (up to 9◦C/min), while others turned out to be more resistant at slow heating rates (Humphrey et al., 1993; Stephens et al., 1994; Morozov et al., 1997; Hassani et al., 2005; Valdramidis et al., 2006; Hassani et al., 2007). Also, Mañas et al. (2003) found no influence of the heating rate (between 0.5 and 4◦C/min) on the thermal resistance of Salmonella Senftenberg 775 W, which, on the other hand, is an exceptionally heat resistant strain. These different behaviors could be explained in terms of genus, species or even strain variability. Even our research was performed with one only strain of E. coli. Still, all these researches were performed under heating rates much lower than the ones used in the present manuscript. During the first part of non-isothermal heating bacterial cells are exposed to non-lethal temperatures and it has been suggested that this exposure could entail an enhance on their heat resistance, similar to that observed when cells are exposed to isothermal treatments at sub-lethal temperatures, which act as a heat shock (Stephens et al., 1994; Mañas et al., 2003; Hassani et al., 2005; Valdramidis et al., 2006; Corradini and Peleg, 2007; Sergelidis and Abrahim, 2009; Van Derlinden et al., 2010), probably through the induction of heat shock protein (HSP) expression (Periago et al., 2002). These HSPs may be induced very rapidly (Allan et al., 1988; Yura et al., 2002). However, heating rates as fast as those that take place in heat exchangers are probably too fast to allow HSP synthesis.

Actually, the fastest heating rates are probably reached in the so-called isothermal heat resistance determination experiments performed in the microbiology labs, in which the microbial suspension is suddenly heated up to the treatment temperature. These experiments are then used to set the heat resistance of the microorganisms. For example, the thermal resistance of E. coli (D<sup>60</sup> = 0.38 min; z = 4.7◦C) was calculated inoculating 0.2 mL of the microbial suspension kept at room temperature into approx. 400 mL of the heating medium preheated at different treatment temperatures (Conesa et al., 2009), so, if this instantaneous heating (from room to treatment temperature) has any effect on the thermal resistance of the microorganism, it would be masked by the whole isothermal experiment and would be already taken into account when calculating microbial heat resistance. If this is the case, heat sensitization observed under the very high heating rates of about 20◦C/s reached in the coil of the thermoresistometer in this research would be somehow unexpected, unless there is a difference between these very high heating rates and the instantaneous heating of isothermal experiments. If such difference exists, under very high heating rates there would be an effect of the heating rate and under instantaneous heating there would be no effect. Further research on high heating rates and instantaneous heating should be performed to unravel this hypothetic effect.

During the non-isothermal phases of treatment different phenomena may take place, which may affect heat resistance in some way. These phenomena and their effect on heat resistance should be considered when calculating the heat treatments to be applied in the food industry. The continuous operating mode of the thermoresistometer Mastia enables to determine the effect of high heating rates on bacterial cells, helping to understand the behavior and response of microorganisms to thermal treatments currently applied in the food industry. The results obtained in this research, as well as other studies from the scientific literature on other microorganisms of interest, could help to set more accurately the thermal treatment parameters. These proper settings would lead food industry to provide foods of better nutritional and sensorial quality and to save energy costs, while maintaining high standards of food safety.

# CONCLUSIONS

The heating rate plays an important role on the heat inactivation of microorganisms, when they are exposed to non-isothermal heat treatments. This factor has been usually omitted by authors when estimating microbial heat resistance. Heat resistance of E. coli vegetative cells was much lower than expected under high heating rates. Therefore, estimation of heat treatments based on isothermal D and z values may not provide a realistic estimation (although it falls on the fail-safe area) of the level of inactivation achieved when applying processing technologies,

# REFERENCES


such as heat exchangers. Further research is required to quantify and understand this effect.

# AUTHOR CONTRIBUTIONS

Conceived and designed the experiments: AE, PF, AP. Performed the experiments: JH. Analyzed and interpreted the data: JH, AA, AE, AI, PP, AP. Drafted and revised the manuscript: JH, AA, AE, PF, AI, PP, AP.

# FUNDING

This research was financially supported by the Ministry of Economy and Competitiveness of the Spanish Government and European Regional Development Fund (ERDF) through project AGL2013-48993-C2-1-R.

# ACKNOWLEDGMENTS

Authors are grateful to Raquel Conesa and Santiago Andreu for technical assistance.

and a resistant strain of Staphylococcus aureus in media of different pH. Lett. Appl. Microbiol. 43, 619–624. doi: 10.1111/j.1472-765X.2006. 02014.x


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

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

# Identification of Genes Required for Growth of Escherichia coli MG1655 at Moderately Low pH

#### Bram Vivijs† , Abram Aertsen and Chris W. Michiels\*

Laboratory of Food Microbiology, Department of Microbial and Molecular Systems, and Leuven Food Science and Nutrition Research Centre (LFoRCe), KU Leuven, Leuven, Belgium

The survival of some pathotypes of Escherichia coli in very low pH environments like highly acidic foods and the stomach has been well documented and contributes to their success as foodborne pathogens. In contrast, the ability of E. coli to grow at moderately low pH has received less attention, although this property can be anticipated to be also very important for the safety of mildly acidic foods. Therefore, the objective of this study was to identify cellular functions required for growth of the non-pathogenic strain E. coli MG1655 at low pH. First, the role of the four E. coli amino acid decarboxylase systems, which are the major cellular mechanisms allowing extreme acid survival, was investigated using mutants defective in each of the systems. Only the lysine decarboxylase (CadA) was required for low pH growth. Secondly, a screening of 8544 random transposon insertion mutants resulted in the identification of six genes affecting growth in LB broth acidified to pH 4.50 with HCl. Two of the genes, encoding the transcriptional regulator LeuO and the elongation factor P-β-lysine ligase EpmA, can be linked to CadA production. Two other genes, encoding the diadenosine tetraphosphatase ApaH and the tRNA modification GTPase MnmE, have been previously implicated in the bacterial response to stresses other than low pH. A fifth gene encodes the LPS heptosyltransferase WaaC, and its mutant has a deep rough colony phenotype, which has been linked to reduced acid tolerance in earlier work. Finally, tatC encodes a secA-independent protein translocase that exports a few dozen proteins and thus is likely to have a pleiotropic phenotype. For mnmE, apaH, epmA, and waaC, de novo in frame deletion and genetic complementation confirmed their role in low pH growth, and these deletion mutants were also affected in growth in apple juice and tomato juice. However, the mutants were not affected in survival in gastric simulation medium at pH 2.5, indicating that growth at moderately low pH and survival of extremely low pH depend mostly on different cellular functions.

Keywords: Escherichia coli, acid stress, stress tolerance genes, genetic analysis, acidic foods

# INTRODUCTION

In the food production chain, bacteria encounter various environmental stresses, such as heat, cold, desiccation, oxidants, or acids. Acid stress, in particular, can have various sources, such as the natural acidity of fruits and fruit-based foods, the formation of organic acids in fermented foods and the addition of organic acids to acidify foods or as food preservatives. Moreover, after food ingestion, bacteria become exposed to the extremely acidic environment of the stomach, with a pH typically ranging from 1.0 to 2.5. The ability to adapt to environmental stresses allows bacteria to withstand the hostile conditions experienced in the food chain within certain limits.

#### Edited by:

Abd El-Latif Hesham, Assiut University, Egypt

#### Reviewed by:

Swaine Chen, Genome Institute of Singapore, Singapore Francisco Diez-Gonzalez, University of Georgia, USA

> \*Correspondence: Chris W. Michiels chris.michiels@kuleuven.be

†Present address: Bram Vivijs, Sanofi-Genzyme, Geel, Belgium

#### Specialty section:

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

Received: 18 July 2016 Accepted: 06 October 2016 Published: 25 October 2016

#### Citation:

Vivijs B, Aertsen A and Michiels CW (2016) Identification of Genes Required for Growth of Escherichia coli MG1655 at Moderately Low pH. Front. Microbiol. 7:1672. doi: 10.3389/fmicb.2016.01672

Adaptation to acid stress is especially relevant for infective foodborne pathogens. Indeed, in order to cause foodborne illness, these bacteria may not only become exposed to acid stress during food processing and storage, they also have to survive the harsh conditions in the stomach before they can infect the small intestine or the colon. Therefore, most studies related to acid stress in enteropathogens have mainly focused on the survival of important pathogens like Salmonella and enterohemorrhagic Escherichia coli (EHEC) upon extreme acid challenge for a few hours. These studies have revealed diverse mechanisms to survive extreme acid stress in these bacteria. One group of mechanisms is called acid resistance (AR), and is defined as the ability to survive an extreme acid challenge at pH 1.5–2.5 (Beales, 2004). AR mechanisms are active in E. coli but seem to be mostly lacking in Salmonella, which explains the lower level of AR in the latter. The other group of mechanisms is the acid tolerance response (ATR), and is defined as the ability to survive a normally lethal acid challenge at pH 3.0 – 3.5 after acid adaptation at moderately low pH. These mechanisms have been mainly studied in Salmonella (Álvarez-Ordóñez et al., 2012).

Besides passive cytoplasmic buffering, five active stationaryphase AR systems have been described in E. coli (Kanjee and Houry, 2013). The AR1 system is referred to as the oxidative or glucose-repressed system and is only poorly understood. This system is activated under oxidative growth conditions and allows E. coli cells to survive a subsequent acid challenge. This is usually achieved by growing cells to stationary phase in a glucose-free complex medium at pH 5.5, and by subsequently challenging them in a minimal medium, without an external supply of amino acid substrates, at pH 2.5. The other four AR systems (AR2-5) all rely on the activity of amino acid-dependent decarboxylase/antiporter systems and therefore require the presence of specific amino acids in the acid challenge medium. Each of these systems decarboxylates a specific amino acid (glutamate, arginine, lysine, or ornithine) to its corresponding amine (γ-amino butyric acid (GABA), agmatine, cadaverine, or putrescine, respectively) by a cytoplasmic pyridoxal-5<sup>0</sup> -phosphate (PLP)-dependent decarboxylase, and subsequently exchanges the decarboxylated product for new amino acid substrate via a cognate inner membrane-bound antiporter (Kanjee and Houry, 2013). The glutamate-dependent AR system (AR2) is by far the most potent AR system in E. coli and consists of the homologous inducible glutamate decarboxylases GadA and GadB, and the glutamate/GABA antiporter GadC. In addition to the amino acid decarboxylase systems, the formate hydrogen lyase (FHL) complex, the adenosine deaminase Add and the glutaminase YbaS were also shown to contribute to AR in E. coli by promoting proton consumption (Noguchi et al., 2010; Sun et al., 2012; Lu et al., 2013). The ATR in Salmonella Typhimurium induces the expression of a large number of genes and prepares cells for lethal acid challenge by the activation of homeostatic mechanisms that maintain or raise the intracellular pH, the production of acid shock proteins with a role in protecting or repairing damaged proteins and DNA, and modifying membrane composition to increase the barrier properties for protons (Álvarez-Ordóñez et al., 2012).

Compared to the survival of extremely low pH, less is known about the mechanisms supporting growth of E. coli and other Enterobacteriaceae at moderately low pH (pH 4.0–5.0), although these are at least equally important for food safety and food stability as the acid survival mechanisms. Several studies have reported growth of various members of the Enterobacteriaceae at surprisingly low pH values. For example, the pHmin of 188 E. coli strains in lysogeny broth (LB) acidified with hydrochloric acid ranged from 3.8 to 4.3 (Haberbeck et al., 2015). Acidification of media with organic acids rather than hydrochloric acid increases the pHmin, but even in acidic foods containing organic acids, growth in this pH range has been reported. For example, strains of E. coli O157:H7 increased by 3.5 log cfu/ml in apple cider of pH 3.6 in 16 h at 23◦C (Ukuku et al., 2009), and by 2.5 log cfu/ml in apple plugs of pH 3.3 in 24 h at 25◦C (Alegre et al., 2010). In the context of food safety, it is important to note that the pH boundaries supporting growth are temperature dependent. From an extensive study of the interaction of temperature and pH on growth of E. coli it was concluded that at suboptimal pH values (down to pH 5.0) Tmax decreased strongly, Tmin decreased slightly and Topt showed no clear trend (Baka et al., 2013). However, at pH values close to pHmin, Tmin was found to increase (Haberbeck et al., 2015). In an earlier work, it was found that pHmin of E. coli was constant between 25 and 37◦C (Presser et al., 1998).

E. coli and other enterobacteria exhibit a transcriptional and translational response during growth at moderately low pH that induces mechanisms contributing to improved pH homeostasis, some of which overlap with the AR mechanisms. Several components of the electron transport chain which are coupled to proton translocation are upregulated during aerobic growth under mild acid stress (Maurer et al., 2005). The membrane potential that builds up as a result and that would impede sustained proton expulsion is dissipated by simultaneous uptake of K+. Further, the expression of enzymes that catalyze reactions which consume cytoplasmic protons, such as hydrogenases or amino acid decarboxylases, is increased under acidic conditions (Hayes et al., 2006). Interestingly, the products of lysine and ornithine decarboxylation (cadaverine and putrescine, respectively) also inhibit the OmpF and OmpC porins, thereby decreasing the outer membrane proton permeability (Dela Vega and Delcour, 1996). Growth at moderately low pH also enhances transport and metabolism of secondary carbon sources, such as sugars (e.g., ribose, arabinose, and fuculose) and sugar derivatives (e.g., galactitol, sorbitol, and gluconate) that produce fewer acids upon fermentation than glucose (Hayes et al., 2006). Finally, E. coli incorporates higher proportions of cyclopropane fatty acids during growth at low pH, which reduces the proton permeability of the inner membrane under mild acid stress and has also be linked with a slightly increased ability to extrude protons upon acid stress (Shabala and Ross, 2008).

The objective of the current work was to conduct a comprehensive genetic analysis to identify (novel) genes required for growth of E. coli at moderately low pH, and to determine whether these genes play a role in survival of extreme AR. First, the role of the known decarboxylase AR systems in growth at low pH was investigated, and subsequently a genome-wide screening for mutants deficient in low pH growth was conducted.

# MATERIALS AND METHODS

fmicb-07-01672 October 22, 2016 Time: 14:37 # 3

# Bacterial Strains, Plasmids, and Oligonucleotides

Bacterial strains and plasmids used in this chapter are listed in **Table 1**. E. coli MG1655 gadA/B and adiA were obtained via P1 transduction from E. coli EF522 and E. coli EF1022, respectively (Castanie-Cornet et al., 1999). Gene deletions were de novo constructed via a protocol described by Baba et al. (2006) using primers listed in **Table 2**. In this way, in-frame deletions encompassing the central region of the open reading frame from the second codon through and including the eighth codon before the C-terminus were created, leaving the start codon and translational signal for possible downstream genes intact. All deletions were verified by PCR using inwardly oriented primers complementary to the region left and right of the deleted genes (**Table 2**), and sequence analysis of the PCR products. Complementation plasmids were constructed by ligation of open reading frames into a NcoI and XbaI cut pTrc99A plasmid under control of the inducible Ptrc promoter, and correct insertion was verified using primers pTrc99A\_seq\_FW and pTrc99A\_seq\_REV.



# Screening for Mutants of E. coli MG1655 Affected in Growth at Moderately Low pH

A random knockout library of E. coli MG1655 was constructed using λNK1324, which carries a mini-Tn10 transposon (1.4 kb) with a Cm resistance gene and a mutant transposase with relaxed target specificity, according to the protocol described by Kleckner et al. (1991). To maximize randomness and genome coverage, the mutant collection was assembled from 89 independent transposon mutagenesis experiments from each of which 96 mutants were isolated (i.e., a total of 8544 mutants). The mutants were stored in microplates at −20◦C. To screen for growth at moderately low pH, the mutant library was first inoculated into microplates containing 300 µl LB broth (10 g/l tryptone, 5 g/l yeast extract, 5 g/l NaCl). These plates were covered with a foil and incubated overnight at 37◦C. Then, the stationary-phase cultures were thousandfold diluted in microplates containing 300 µl LB acidified to pH 4.50 with HCl, and the plates were covered with a foil and incubated at 30◦C for 24 h, when the OD<sup>600</sup> was measured. Transposon insertion sites were determined using the method of Kwon and Ricke (2000). Briefly, genomic DNA of the mutants was isolated, digested with NlaIII and ligated with a Y-shaped linker, composed of oligonucleotides linker 1 and linker 2. Next, a PCR amplification was carried out using a transposon-specific primer (NK\_Cm\_DWN) and a primer specific to the Y-shaped linker (Y linker primer). The PCR product was subsequently sequenced using the transposonspecific primer and the insertion site was determined based on the known genome sequence of E. coli MG1655.

# Determination of Growth Curves Via OD Measurements

Escherichia coli strains were grown overnight to stationary phase in 4 ml LB at 37◦C and thousandfold diluted in fresh LB acidified with HCl to the required pH. Culture volumes of 300 µl were transferred to the wells of a microplate, which was sealed with a cover foil and incubated in a Multiskan Ascent plate reader (Thermo Labsystems, Helsinki, Finland) at 30 or 37◦C, and the OD<sup>630</sup> was automatically measured every 15 min. The growth curves were fitted by the model of Baranyi and Roberts (1994), using the Excel add-in package DMFit (Institute of Food Research, Norwich, UK). For the decarboxylase mutants, growth curves were also recorded in a similar way in M9 medium (Sambrook et al., 1989) supplemented with 0.2% glucose, 0.1% casamino acids, and 1 mg/l thiamine, and adjusted to pH 4.80 with HCl, and to which 5 mM L-glutamic acid, L-arginine, Llysine, or L-ornithine were added when required. In this case, wells were covered with 50 µl paraffin oil to create anoxic conditions required for decarboxylase activity.

# Determination of Growth Curves Via Plate Counts

Stationary-phase cultures, grown overnight in 4 ml LB at 37◦C, were 100,000-fold diluted in test tubes containing 10 ml LB, apple or tomato juice that had been first adjusted to specific pH values with HCl or NaOH. The suspensions in LB were transferred to microplates (multiple 300 µl samples) and sealed with a

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

fmicb-07-01672 October 22, 2016 Time: 14:37 # 4


cover foil. The test tubes with acidic juices and the microplates were incubated without shaking at 20 and 30◦C, respectively. At regular time points, 300 µl samples were taken, serially diluted in potassium phosphate buffer (10 mM; pH 7.00) and subsequently spotted (5 µl) on LB agar. After 24 h of incubation at 37◦C, the colony-forming units were determined.

# Acid Challenge (Survival) Assay

Stationary-phase cultures grown overnight at 37◦C in 4 ml LB were hundredfold diluted in test tubes containing 4 ml of simulated gastric fluid (Beumer et al., 1992). The composition of this medium was 8.3 g/l bacteriological peptone, 3.5 g/l glucose, 2.05 g/l NaCl, 0.6 g/l KH2PO4, 0.11 g/l CaCl2, 0.37 g/l KCl, 0.05 g/l bile salts (Oxoid, Basingstoke, UK), 0.1 g/l lysozyme (66200 U/mg, Fluka, Buchs, Switzerland), and 13.3 mg/l pepsin (47 U/g, Fluka, Buchs, Switzerland), and the pH was adjusted to 2.50 with HCl. The suspensions were incubated at 37◦C for 6 h. Every hour, 100 µl samples were taken, serially diluted in potassium phosphate buffer (10 mM; pH 7.00) and subsequently plated on LB agar. D-values (time required for a 10-fold reduction in viable cells) were determined by identification of the log-linear Bigelow model (for the 1waaC mutant) or the log-linear model with shoulder (for the wild-type and other deletion mutants) using GInaFiT (Geeraerd et al., 2005).

# Statistical Analysis

fmicb-07-01672 October 22, 2016 Time: 14:37 # 5

All experiments were carried out in triplicate using cultures grown from different colonies on a single agar plate. Mean values of different strains or treatments were compared by Student's t-test and differences were considered significant when a p-value of <0.05 was obtained.

# RESULTS

# Role of E. coli Amino Acid Decarboxylase Systems in Growth at Moderately Low pH

First, we investigated whether the addition of the four amino acid substrates of the AR systems 2, 3, 4, and 5 (glutamic acid, arginine, lysine, and ornithine, respectively) improves growth of E. coli MG1655 at a low initial pH. This was done in M9 minimal medium to reduce the basal levels of these amino acids compared to rich media like LB. However, 0.1% casamino acids were added to stimulate growth. The initial pH of the medium was set at 4.80, and final pH values were recorded as an indication of decarboxylase activity (**Figure 1**). The experiments with the decarboxylase mutants were conducted at 37◦C. Although this temperature is not very relevant to most conditions at which foods are stored, it is more challenging for growth of E. coli at low pH than 30◦C, and therefore expected to allow a more sensitive detection of any deficiencies in the mutants' capacity to grow at low pH (see below).

The addition of arginine or ornithine to the acidified M9 medium did not result in growth improvement of E. coli MG1655 at initial pH 4.80 and also did not affect the final pH values of the cultures. On the other hand, the addition of lysine, and to a smaller extent glutamate, enhanced growth and also resulted in a significantly higher final pH value. To further investigate whether the corresponding decarboxylase enzymes are involved in this growth improvement, the experiment was repeated with E. coli MG1655 strains deficient in the decarboxylases of the AR systems 2, 3, 4, and 5 (gadA/B, adiA, 1cadA, and 1speF, respectively) (**Figure 2**). The growth curves and final pH values of the adiA and 1speF mutants, with or without addition of arginine or ornithine, respectively, were not significantly different from those of the wild-type strain (data not shown). Further, a small growth improvement and pH increase was observed when glutamate was supplied to cultures of the gadA/B mutant, but this was not different from the wild-type strain, indicating that the stimulating effect of glutamate on growth at pH 4.80 was not due to its decarboxylation. On the other hand, supplementation with lysine did no longer improve growth and increase final pH in the 1cadA mutant (**Figure 2**), suggesting that the low pH growth enhancing effect of lysine on E. coli is due to lysine decarboxylation by CadA.

Subsequently, we investigated whether the decarboxylase enzymes play a role to support growth of E. coli MG1655 in a complex medium (LB) at low pH (**Figure 3**). Knockout of gadA/B, adiA, or speF did not result in diminished growth in LB at pH 4.40 or 4.60. However, the final pH values for the gadA/B and adiA mutants were significantly lower than those of the wild-type strain, indicating that the glutamate and arginine decarboxylases are active during growth of MG1655 under these conditions. Deletion of cadA clearly impaired growth in acidified LB and resulted in the lowest final pH values. Finally, deletion of speF did not influence E. coli growth or final pH.

# Genome-Wide Screening for Mutants of E. coli MG1655 Affected in Growth in LB at Low pH

As an open approach to identify novel genes involved in moderately low pH growth, a random transposon insertion mutant library of E. coli MG1655 was constructed and screened. To identify suitable screening conditions, the wild-type strain was first grown in LB acidified with HCl to different initial pH values (data not shown). This screening was done at 30◦C, because, as mentioned before, this makes the selection more relaxed than 37◦C and thus ensures that only the most sensitive mutants are picked up.

The lowest pH at which growth was observed in this experiment was 4.20, and the pH for this culture increased to 5.10 after 72 h. From pH 4.20 to 4.50, the growth rates and final OD<sup>630</sup> increased rapidly, while a further increase of pH above 4.50 had comparatively less effect on growth. Based on these observations, a pH value of 4.50 was chosen for the screening in combination with an incubation time of 24 h. Although this pH value is only slightly above the pHmin under these conditions, the wild-type strain can grow relatively well at this pH value and therefore this combination of pH and time should be suited to select mutants affected in growth at low pH.

Eleven out of the 8544 mutants tested remained below an OD<sup>600</sup> of 0.100 after 24 h incubation in LB at pH 4.50. Determination of transposon insertion sites revealed six different genes: mnmE (1x), leuO (5x), apaH (1x), waaC (1x), epmA (1x), and tatC (2x). The functions of these genes and the exact position of the transposons are listed in **Table 3**. In all cases except for leuO, the transposons were inserted into the open reading frame of the corresponding genes. In all five leuO mutants, which were retrieved from independent mutagenesis experiments, the transposon was inserted 26 bp upstream of the leuO open reading frame. Although the transposase used in this study has an altered target specificity, exhibiting a much lower degree of insertion specificity than the wild-type transposase (Kleckner et al., 1991), the occurrence of five inserts at exactly the same insertion site may indicate a so-called hotspot at this position. On the other

standard deviations of three replicate cultures. The asterisks indicate when the final pH value was significantly different (p < 0.05) from the final pH value reached with no specific amino acid addition.

bars represent standard deviations of three replicate cultures. The asterisks indicate that the final pH value is significantly different (p < 0.05) from the final pH reached with no specific amino acid addition for the same strain.

#### TABLE 3 | Overview of genes affected in the transposon insertion mutants of which the growth was impaired in LB pH 4.50.


The position of the transposon gives the nucleatide after which the rtansposon was inserted, starting from the first base of the start codon. In the case of leuO, the insertion was 26 bp upstream of the leuO open reading frame.

hand, the two insertions in tatC were at different positions in the tatC gene.

# Construction of Deletion Mutants and Complementation

To exclude the possibility that these acid-sensitive phenotypes were due to polar genetic effects of the transposon or due to unrelated secondary mutations, de novo deletion mutants were made in E. coli MG1655. However, the tatC mutant was excluded from further analysis, since the E. coli K-12 chromosome encodes at least 36 polypeptides that are known or predicted substrates for export by the twin-arginine translocation (Tat) system (TatABCE) (Berks et al., 2005), and inactivation of this system probably has pleiotropic effects on the E. coli physiology. The OD<sup>630</sup> growth curves of these deletion mutants confirmed their reduced growth in LB at pH 4.50, except for the 1leuO mutant, which showed WT growth (data not shown). Since the transposon of the originally isolated mutant was inserted 26 bases upstream of the start codon, we assume it did not completely knock out leuO gene function, but may have modified its expression level or regulation. Subsequently, expression constructs derived from the pTrc99A plasmid were made for complementation and introduced into the corresponding deletion mutants (except for leuO). The deletion mutants harboring the empty pTrc99A plasmid or the complementation plasmid were then grown in LB at pH 4.50 at 30 or 37◦C (**Figure 4**), and the OD<sup>630</sup> growth curves show that complementation effectively restored moderately low pH growth to wild-type levels.

Next, growth of the mutants in acidified LB was evaluated in more detail by determining plate counts and medium pH every

3 h (**Figure 5**). The growth parameter estimates for the initial (y0) and final (ymax) bacterial cell density, the maximum specific growth rate (µmax), and the lag time (tlag) are shown in **Table 4**.

The results show that the cell numbers and the pH values for the wild-type and the 1leuO mutant evolved in a similar way, and that the growth parameters for both strains were not significantly different. The 1apaH mutant was not able to grow at this pH (as already evident from **Figure 4**) and its cell numbers decreased slightly, reaching around 3.1 log(cfu/ml) after 48 h. The remaining three mutants had a significantly slower growth rate (µmax), and, in addition, the 1mnmE and 1epmA mutants had slightly lower maximal cell densities (ymax), while the 1waaC and 1mnmE mutants exhibited a significantly longer lag phase (tlag) than the wild-type (**Table 4**). Interestingly, all strains that initiated growth did so without increasing medium pH initially. Only when cell numbers exceeded around 7.0 log(cfu/ml), medium pH started increasing until early stationary phase. The maximum pH value reached by all strains that started growing, except for the 1mnmE mutant, was around 5.10 and this pH value remained unchanged during stationary phase (**Figure 5**).

# Growth of Mutants in Acidic Foods

In a subsequent step, we investigated growth of the deletion mutants (except the 1leuO mutant) in apple juice (pH 4.60) and tomato juice (pH 4.80) (**Figure 6**). While previous experiments in LB were conducted at 30 or 37◦C, the experiments in juice were done at 20◦C to simulate possible food storage conditions. All the mutants also showed growth defects in these acidic foods but the nature and relative magnitude of the defects was different in these foods than in acidified LB (**Figure 5**) for some mutants. For example, while the 1waaC mutant showed an extended lag phase in acidified LB and the 1epmA mutant did not, this was reversed in the juices. Another example is the 1mnmE mutant, which was able to grow in acidified LB but was slowly inactivated in the juices. The 1apaH mutant showed similar behavior in acidified LB and in the juices, being inactivated in all situations. Also in contrast to growth in acidified LB, bacterial growth in acidic juices did not result in an extracellular pH increase. Instead, the juice pH remained almost constant (maximal variation was 0.04 pH units) for all strains during the entire experiment (data not shown).

# Survival of Mutants in Acid Challenge Assay in Simulated Gastric Fluid

Finally, we investigated whether the mutations rendered E. coli MG1655 also more sensitive to inactivation at extremely low pH. Therefore, the AR of the deletion mutants was examined in a simulated gastric fluid (pH 2.50) at 37◦C (**Figure 7**). The survival data were fitted and D-values were calculated from the linear part. All curves, except that of the 1waaC mutant, displayed a shoulder before a substantial decrease was apparent. Unexpectedly, the 1epmA and 1waaC mutants were more acid resistant than the wild-type strain, showing significantly higher D-values under these conditions, while the other genes did not influence survival of E. coli MG1655 in the simulated gastric fluid.

# DISCUSSION

# Amino Acid Decarboxylase Systems

The best described AR mechanisms in Enterobacteriaceae rely on proton consumption during amino acid decarboxylation. In E. coli, four AR systems (AR 2, 3, 4, and 5) have been described in which a particular amino acid (glutamate, arginine, lysine, or ornithine, respectively) is converted to its corresponding amine (GABA, agmatine, cadaverine, or putrescine, respectively) by a cytoplasmic decarboxylase (the GadA and GadB isozymes, AdiA,

TABLE 4 | Parameter estimates for the initial (y0) and final (ymax) bacterial cell density, the maximum specific growth rate (µmax), and the lag time (tlag) during growth of E. coli MG1655 WT and deletion mutants in acidified LB (initial pH around 4.4) at 30◦C for 48 h (growth curves in Figure 4).


Values are means ± standard deviations (n = 3). The asterisks indicate that the parameter value is significantly different (p < 0.05) from that of the wild-type strain.

CadA, and SpeF, respectively), followed by the exchange of the decarboxylated products for new amino acid substrates via a cognate inner membrane-bound antiporter (GadC, AdiC, CadB, and PotE, respectively) (Zhao and Houry, 2010). The strength of these systems to protect against extreme acid stress has been correlated to the pH optima of the decarboxylase enzymes, being 3.7−3.8, 4.9−5.2, 5.7, and 7.0, respectively (Kanjee and Houry, 2013). Thus, the glutamate-dependent AR system is the most

FIGURE 6 | Growth of E. coli MG1655 WT and deletion mutants in apple juice pH 4.60 (A) and tomato juice pH 4.80 (B) at 20◦C. Error bars represent standard deviations of three indemendent replicate cultures. The line at 2.3 log(cfu/ml) represents the lower detection limit.

potent AR system in E. coli, followed by the arginine- and lysinedependent AR systems (Iyer et al., 2003; Diez-Gonzalez and Karaibrahimoglu, 2004). The ornithine-dependent AR system has been less well studied and plays only a minor role during survival of an extreme acid challenge. For example, in E. coli, only glutamate, arginine, and lysine, but not ornithine, supported robust survival at pH 2.5 in a minimal medium (Iyer et al., 2003). Also survival of Salmonella was only modestly improved by ornithine in a minimal medium at pH 2.3 (Viala et al., 2011).

In our study, the ability of these systems to enhance growth under moderate acid stress was investigated. We showed that only lysine considerably improved growth of E. coli MG1655 at moderately low pH in a minimal medium, and that this effect relies on lysine decarboxylation since it was not observed in a 1cadA mutant (**Figures 1** and **2**). Similarly, only the lysine decarboxylase CadA enhanced growth of E. coli MG1655 at moderately low pH in a complex medium (**Figure 3**). Although also the glutamate decarboxylase isozymes GadA and GadB and the arginine decarboxylase AdiA provoked a slight pH increase during growth in complex medium at moderately low pH, this effect was insufficient to confer a growth advantage. The relative effectiveness of these decarboxylases to support low pH growth may reflect their pH optima. Since E. coli cannot maintain an internal pH of more than two units higher than the external pH (Foster, 2004), the internal pH during growth at an external pH of 4.5 is expected to be around 6.5, which is close to the pH optimum of 5.7 of CadA. However, other factors are also involved, since the ornithine decarboxylase SpeF (pH optimum 7.0) did not contribute to growth under these conditions. This is in contrast to Salmonella, in which both lysine and ornithine decarboxylases significantly improved growth at moderately acidic pH under anoxic conditions in a minimal medium supplemented with these amino acids (Viala et al., 2011).

A mutant that can be indirectly linked to the lysine decarboxylase system is the mutant which had the transposon inserted just 26 bp upstream of the leuO open reading frame, leaving the putative ribosomal binding site fully intact. Remarkably, insertion of a Tn10 transposon at exactly the same site has been reported previously, and was shown to result in overexpression of LeuO (Klauck et al., 1997). This might also

Vivijs et al. E. coli Growth at Low pH

be the case in our study since deletion of leuO did not affect the growth of E. coli MG1655 at moderately low pH. Moreover, it has been shown that LeuO overexpression drastically reduces production of CadC, the essential activator for cadA induction (Shi and Bennett, 1995), and this might explain its role in growth at low pH. Furthermore, LeuO has also indirectly been linked to AR since it represses the small non-coding RNA dsrA, which plays a regulatory role in AR in E. coli (Lease et al., 2004). Finally, also the epmA gene, discussed in the next section, can be linked to CadA activity.

In conclusion, our results show that the decarboxylase enzymes not only produce better survival at extremely low pH by counteracting intracellular acidification (Richard and Foster, 2004), they also trigger deacidification of the extracellular medium, thereby potentially enhancing growth at moderately low pH.

# Aminoacyl-tRNA Synthetases and Translation Elongation

The mutant with the strongest growth defect at low pH was an apaH mutant. This mutant was not able to grow in LB at pH 4.50, while its growth rate was only slightly less than that of the wild-type at neutral pH (data not shown). The apaH gene encodes diadenosine tetraphosphatase, which hydrolyzes 5',5"'- P 1 , P<sup>4</sup> -diadenosine tetraphosphate (AppppA) to two molecules of adenosine diphosphate (ADP) (Guranowski et al., 1983). AppppA is rapidly synthesized by aminoacyl-tRNA synthetases when E. coli is exposed to heat shock or oxidative stress and may serve as a modulator of these stress responses (Johnstone and Farr, 1991). It has been shown that apaH mutants, which have high basal levels of AppppA, are sensitive to killing by heat and oxidative stress, and are unable to grow at 43◦C (Johnstone and Farr, 1991). AppppA binds to several proteins in E. coli, including DnaK, GroEL, and ClpB, and AppppA binding may inhibit DnaK and/or GroEL functions that are required for survival of thermal stress. Thus, the regulation of the AppppA pool is important for the bacterial response to conditions of stress and AppppA may assist the return of cells to normal growth conditions following stress, but constitutively high levels of AppppA may inhibit the function of stress proteins such as DnaK before they have fulfilled their role under that stress (Johnstone and Farr, 1991; McLennan et al., 2001). Our results suggest that AppppA is also involved in cellular adaptation to acid stress. Since the cytoplasmic chaperones DnaK and GroEL have also been linked to counteracting acid stress in several bacteria, such as S. enterica and H. pylori, a similar mechanism may also account for the role of AppppA under acid stress (Bearson et al., 2006).

Another protein that was found in our work to play a role during growth at low pH was EpmA. EpmA specifically aminoacylates the translation elongation factor P (EF-P) at a conserved lysine residue with β-lysine. This posttranslational modification activates EF-P to function in translation elongation of a particular subset of mRNAs (Bullwinkle et al., 2013). It has been demonstrated that EF-P enhances the translation of polyproline-containing proteins by alleviating ribosome stalling at polyproline stretches (Ude et al., 2013). One such protein is CadC, a membrane-integrated transcriptional regulator that both senses external pH and activates expression of the cadBA operon at low external pH. EF-P was shown to be required for translation of CadC and a deletion mutant of epmA lacked CadA activity (Ude et al., 2013), which may account for the diminished growth at moderately low pH in our experiments. Interestingly, EpmA is an aminoacyl-tRNA synthetase paralog, showing sequence similarity to the lysyl-tRNA synthetases LysS and LysU. Whereas LysS is constitutively expressed, LysU is overexpressed under extreme physiological conditions, such as heat shock, and is the most effective AppppA synthetase, producing 80% of total AppppA in E. coli cell extracts (Charlier and Sanchez, 1987). In addition to high temperature, LysU has also been shown to be upregulated at moderately low pH (Hayes et al., 2006). Interestingly, the activity of LysU has been shown to be less sensitive than that of LysS to the competitive inhibitor cadaverine, the decarboxylation product of lysine, suggesting that LysU plays a more important role under physiological conditions causing cadaverine accumulation, such as acid stress (Brevet et al., 1995). Although EpmA is homologous to the catalytic core of the lysyl-tRNA synthetases, it lacks the tRNA aminoacylation activity due to absence of the tRNA anticodon-binding domain. Nevertheless, EpmA retains the ability to activate L-lysine by formation of the lysyl-adenylate intermediate (Ambrogelly et al., 2010), which is also the first step in AppppA synthesis. However, it is not known whether EpmA can also contribute to AppppA production.

The screening also yielded a gene that is involved in tRNA modification, mnmE. Together with MnmG, MnmE forms a complex that adds an aminomethyl or carboxymethylaminomethyl group to position 5 of the anticodon wobble uridine (U34) using methylene-tetrahydrofolate and ammonium or glycine as donors, respectively (Moukadiri et al., 2014). Approximately 85 different modifications of tRNA molecules have been documented, and they are thought to be important for maintaining tRNA structure, and for specific recognition of tRNA molecules by their cognate aminoacyl-tRNA synthetases. Besides, the idea that tRNA modifications can take on second-order regulatory functions, especially in response to stress conditions, has recently emerged. Some recent studies have linked tRNA modification to the control of gene expression at the level of translation in response to environmental stresses. For example, Begley et al. (2007) have shown that the tRNA methyltransferase 9 (Trm9) of Saccharomyces cerevisiae, which catalyzes the last step (methylation) in the methylcarbonylmethyl (mcm) modification on position 5 of the uridine wobble base (U34) of tRNAArg mcm5UCU and tRNAGlu mcm5s2UUC, prevents cell death during methyl methanesulfonate exposure via translational enhancement of DNA damage response proteins whose genes are overrepresented with the AGA (Arg) and GAA (Glu) codons. This indicates that tRNA modifications indirectly help coordinate DNA repair. Also in S. cerevisiae, it has been demonstrated that the modifications 2'-O-methylcytosine (Cm), 5-methylcytosine (m5C), and N 2 ,N 2 -dimethylguanosine (m<sup>2</sup> <sup>2</sup>G) increase following exposure to hydrogen peroxide and that loss of the methyltransferase enzymes catalyzing the formation of these modified nucleosides causes hypersensitivity to hydrogen peroxide (Chan et al., 2010). Thus, tRNA modifications

can dynamically change in response to stress and may be critical features of the cellular stress response. Furthermore, exposing S. cerevisiae to hydrogen peroxide resulted in a Trm4 methyltransferase-dependent increase in the incorporation of m5C in tRNALeu m5CAA, which causes selective translation of mRNA from genes that are enriched in the TTG codon for leucine (Chan et al., 2012). Cells may thus respond and adapt to environmental stresses by reprogramming of tRNA modifications, thereby promoting the selective translation of codon-biased mRNAs for critical stress response proteins (Dedon and Begley, 2014). Our work extends these observations by linking t-RNA modification by MnmE to growth at low pH in E. coli.

# Cell Envelope

The waaC gene which was picked up in our screening is involved in the synthesis of lipolpolysaccharide (LPS). WaaC adds ADP-L-glycero-D-manno-heptose on the inner 3-deoxy-Dmanno-octulosonic acid (KDO) residue of the LPS inner core with the release of ADP (Kadrmas and Raetz, 1998). Knockout of waaC results in a heptoseless truncation of the LPS and a so-called deep-rough phenotype, characterized by mucoid colonies phenotype due to production of a colanic acid capsular polysaccharide (Joloba et al., 2004). This phenotype was clearly visible when these mutants were grown on agar plates in our experiments. The production of this polysaccharide might also be the reason for their higher final OD values compared to the wild-type in some growth experiments (**Figure 4B**). Bielecki et al. (1982) used chemical mutagenesis to isolate mutants of E. coli K-12 that were not able to grow on LB agar plates acidified with HCl to pH 5.4. Four mutants were picked up and a number of altered phenotypes (phage and detergent sensitivities, leakage of periplasmic proteins) suggested that these mutants probably belonged to a group of deep-rough mutants defective in their LPS. Thus, the capacity of the outer membrane to form a proton barrier is probably reduced by lesions in the LPS structure (Rowbury, 2004). Furthermore, Barua et al. (2002) isolated five transposon insertion mutants of E. coli O157:H7 that showed poor or no growth on LB-MES agar with 12 mM acetic acid (pH 5.4). Two of them had an inactivated waaG gene, also causing a deep-rough phenotype. The other three genes (fcl, wecA, and wecB) were involved in the biosynthesis of the surface O-polysaccharide and/or enterobacterial common antigen.

Interestingly, a waaC mutant has been shown to be unable to express the outer membrane protein A (OmpA) (Beher and Schnaitman, 1981). OmpA is one of the most abundant proteins in the outer membrane of E. coli and is believed to be a nonspecific diffusion channel, allowing the passage of various small solutes, but its physiological function is unclear. OmpA is induced by acid in E. coli and has previously been linked with AR in E. coli since an ompA mutant was more readily killed by lethal acid stress (pH 3.8; 60 mM acetic acid) than its parent strain (Wang, 2002). OmpA is also important for the structural integrity of the cell envelope, and loss of OmpA probably leads to increased penetration of protons or undissociated acids through the outer membrane due to reduced barrier properties (Rowbury, 2004).

# Efficiency of Mutant Screening

The discussion above suggests that besides the six mutants isolated in this work, several additional mutants are predicted to have a low pH growth phenotype, like cadA, cadC, ompA, mnmG, fcl, wecA, wecB. Assuming random transposition of the mini-Tn10 transposon, the probability P that the library contains at least one insertion in each (non-essential) gene can be calculated from N = ln(1−P)/ln(1−a/b), with N = number of mutants in the library, a = average size of a gene (1000 bp), and b = genome size (excluding all essential genes). For our library of 8544 mutants, this gives a probability of 86% that any particular non-essential gene is represented. It seems therefore that the screen picked up less mutants than would be expected. This may be partly explained by non-randomness of the tranposon and the fact that inserts in the 3<sup>0</sup> end of a gene do not necessarily knock out protein function. However, probably a more important explanation is that threshold set for isolation of mutants (OD<sup>600</sup> < 0.100 after 24 h incubation in LB at pH 4.50) filtered out only the most sensitive mutants. A screen based on recording full growth curves (as opposed to end point measurement) would most likely reveal additional, more subtle mutations.

# CONCLUSION

By identifying genes required for growth at moderately low pH, this work has yielded new insights in the cellular mechanisms used in E. coli to cope with mild acid stress. The lysine decarboxylase system, but not other amino acid decarboxylases known to contribute to survival of extreme acid stress, was demonstrated to support growth at moderately low pH. In line with previous findings, the integrity of the outer membrane was also shown to be important for growth at low pH. In addition, tRNA modification and diadenosine tetraphosphate hydrolysis were identified for the first time to be required for low pH growth. Except for the lysine decarboxylase, the cellular functions supporting low pH growth did not support survival of extreme low pH. This work will contribute to a better understanding of microbial survival and growth in mildly acidic foods.

# AUTHOR CONTRIBUTIONS

CM: Conception of research project, research strategy, interpretation of results, final stages of writing of manuscript. BV: Research strategy, experimental work, interpretation of results, writing of manuscript. AA: Conception of research project, research strategy, interpretation of results

# FUNDING

This work was supported by research grants from the Fonds voor Wetenschappelijk Onderzoek Vlaanderen (G.A061.11) and the Research Council of KU Leuven (METH/14/03). Author BV was the recipient of a doctoral fellowship from the Agentschap voor Innovatie door Wetenschap en Technologie (IWT).

# REFERENCES

fmicb-07-01672 October 22, 2016 Time: 14:37 # 13


oxidative stress in Escherichia coli K-12. J. Bacteriol. 187, 304–319. doi: 10.1128/JB.187.1.304-319.2005


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

# Effect of the Food Additives Sodium Citrate and Disodium Phosphate on Shiga Toxin-Producing Escherichia coli and Production of stx-Phages and Shiga toxin

Lucas J. Lenzi<sup>1</sup> , Paula M. A. Lucchesi1,2 \*, Lucía Medico<sup>1</sup> , Julia Burgán1,2 and Alejandra Krüger1,2

<sup>1</sup> Laboratorio de Inmunoquímica y Biotecnología, Facultad de Ciencias Veterinarias, Universidad Nacional del Centro de la Provincia de Buenos Aires, Tandil, Argentina, <sup>2</sup> Centro de Investigación Veterinaria de Tandil, CONICET-CIC-UNCPBA, Tandil, Argentina

#### Edited by:

Avelino Alvarez-Ordónez, ´ Teagasc Food Research Centre, Ireland

#### Reviewed by:

Panagiotis Skandamis, Agricultural University of Athens, Greece Jinshui Zheng, Huazhong Agricultural University, China

\*Correspondence:

Paula M. A. Lucchesi paulaluc@vet.unicen.edu.ar

#### Specialty section:

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

Received: 24 February 2016 Accepted: 09 June 2016 Published: 23 June 2016

#### Citation:

Lenzi LJ, Lucchesi PMA, Medico L, Burgán J and Krüger A (2016) Effect of the Food Additives Sodium Citrate and Disodium Phosphate on Shiga Toxin-Producing Escherichia coli and Production of stx-Phages and Shiga toxin. Front. Microbiol. 7:992. doi: 10.3389/fmicb.2016.00992 Induction and propagation of bacteriophages along the food production chain can represent a significant risk when bacteriophages carry genes for potent toxins. The aim of this study was to evaluate the effect of different compounds used in the food industry on the growth of Shiga toxin-producing Escherichia coli (STEC) and the production of stx-phage particles and Shiga toxin. We tested the in vitro effect of lactic acid, acetic acid, citric acid, disodium phosphate, and sodium citrate on STEC growth. A bacteriostatic effect was observed in most of treated cultures. The exceptions were those treated with sodium citrate and disodium phosphate in which similar growth curves to the untreated control were observed, but with reduced OD<sup>600</sup> values. Evaluation of phage production by plaque-based assays showed that cultures treated with sodium citrate and disodium phosphate released phages in similar o lower levels than untreated cultures. However, semi-quantification of Stx revealed higher levels of extracellular Stx in STEC cultures treated with 2.5% sodium citrate than in untreated cultures. Our results reinforce the importance to evaluate if additives and other treatments used to decrease bacterial contamination in food induce stx-phage and Stx production.

Keywords: STEC, Shiga toxin, bacteriophages, food additives, sodium citrate, disodium phosphate

# INTRODUCTION

Shiga toxin-producing Escherichia coli (STEC) are important pathogens that can cause human diseases, like diarrhoea, haemorrhagic colitis, and haemolytic uraemic syndrome (HUS) (Karmali et al., 1985). STEC strains are characterized by their capacity to produce Shiga toxins, which are encoded by bacteriophages, usually named stx-phages. These phages influence the pathogenicity of STEC strains, since the expression of these toxins is upregulated when the lytic cycle of these phages is induced. They also play a role in the spread of stx among E. coli as well as to other bacteria (Schmidt et al., 1999; Muniesa et al., 2004; Probert et al., 2014). Several factors have been shown to induce the lytic pathway of stx-phages [revised by Krüger and Lucchesi (2015)]. Particularly, some antibiotics have been reported to increase phage induction

and Stx production and therefore treatments of human STEC infections with some antibiotics may have adverse clinical consequences (Wong et al., 2000; Zhang et al., 2000; McGannon et al., 2010).

Meat consumption has been identified as one of the risk factors strongly associated with HUS (Bentancor et al., 2012) and several studies have shown that meat is frequently contaminated with STEC strains (Parma et al., 2000; Bosilevac and Koohmaraie, 2011). Food producing animals have been recognized as the most important source for the entry of STEC in the food chain (Arthur et al., 2002; Martin and Beutin, 2011), and ruminants, especially cattle, have been identified as the major reservoir of STEC strains (Caprioli et al., 2005; Mainil and Daube, 2005). In addition to meat and dairy products, other vehicles and transmission routes of STEC strains for human infection have been reported, like person-to-person contact and drinking and swimming water [reviewed by Doyle et al. (2006) and Kaspar et al. (2010)]. Several intervention strategies have been proposed to minimize meat contamination by STEC during slaughtering and meat processing, like hide washing (Arthur et al., 2007), hot water treatment (Bosilevac et al., 2006), high hydrostatic pressure (HHP) treatment (Gola et al., 2000), and treatments with organic acids and/or their salts (Eswaranandam et al., 2004; Over et al., 2009), among others.

Infectious stx-phages have been detected in minced beef and salad samples (Imamovic and Muniesa, 2011). Moreover, stxencoding phages can be more resistant than their host bacteria to chlorination and heat treatments (Muniesa et al., 1999) and also have a high ability to tolerate exposure to several disinfectants (Rode et al., 2011). There is scarce information about the risk of stx-phage induction and stx gene dissemination to other bacteria in foods. Imamovic et al. (2009) showed that stx transduction in food matrices is possible under appropriate conditions. Studies are required to evaluate if antimicrobial interventions, as well as other farming practices and food processing technologies, may increase the rate of induction and propagation of stx-phages.

Several substances are used in the production, processing, treatment, packaging, transportation, and storage of food. In the meat industry, additives like citric acid and sodium citrate are widely applied for pH control, metal chelating, and preservation (Sammel et al., 2006). Phosphates are used in meat and meat products to adjust pH, sequester cations, change the ionic charges distributions, change the ionic strength, and/or to function as a bacteriostatic agent (Long et al., 2011). The aim of this study was to evaluate the in vitro effect of some additives on STEC growth, and on the production of stx-phage and Stx.

# MATERIAL AND METHODS

# Bacterial Strains

STEC strains FB3 (serotype O157:H7, stx2-positive) and FB5 (serotype O145:H-, stx2-positive) isolated from feedlot cattle (Padola et al., 2004), and the reference strain E. coli EDL933 (serotype O157:H7, positive for stx<sup>1</sup> and stx2) were selected to evaluate the effect of additives on STEC growth. In addition, supernatants from E. coli EDL933 cultures were used to evaluate phage and Stx production. E. coli laboratory strain DH5α was used as host strain for stx phages in double-agar-layer plaque assays.

# Additives

The following compounds were prepared in stock solutions and used at indicated final concentrations: lactic acid (0.5 and 2.5% v/v), acetic acid (2.5% v/v), citric acid (0.5 and 2.5% w/v), disodium phosphate (0.1, 0.5, and 1% w/v), and sodium citrate (0.5, 1.0, and 2.5% w/v).

treatments (not shown).

# Bacterial Growth/Lysis Curves

fmicb-07-00992 June 21, 2016 Time: 13:53 # 3

STEC strains were cultivated overnight in Luria Bertani (LB) medium at 37◦C with shaking at 120 rpm. Aliquots (300 µl) were inoculated into 100 ml flasks containing 15 ml of fresh LB medium. The new cultures were incubated at 37◦C and 120 rpm ∼45 min. up to an optical density at 600 nm (OD600) ≈ 0.2−0.3 when each flask was added with the respective additive, or water (untreated control), or mitomycin C (final concentration of 0.5 µg/ml; positive control of phage induction), reaching a final culture volume of 15.5 ml. This moment was identified as 0 h of the assay. The incubation was continued at 37◦C and 180 rpm for 18 h, and spectrophotometrically monitored every hour for the first 5 h. When it was necessary, the aliquots measured were previously diluted. In addition, viable bacterial count at 2 h was

conducted by plating appropriate dilutions on LB agar plates. These assays were repeated at least two times.

# Evaluation of Phage Production

For phage quantification, aliquots of EDL933 cultures under different treatments were assayed. This strain was selected because phages induced from it showed lysis plaques easier to be counted by visual inspection than those from FB3 and FB5 strains. Aliquots from EDL933 cultures were taken at 3 h and centrifuged for 10 min at 10,000 × g, at 4◦C. Supernatants were collected, filtered through low-protein-binding 0.22 µm membrane filters (Millex-GV, Millipore) and tenfold serially diluted for titration assays using the double-agar-layer method as follows. One hundred microliters of each dilution were mixed with 500 µl of an exponential phase culture of E. coli DH5α (OD<sup>600</sup> ≈ 0.6–0.8) and incubated for 30 min at 37◦C with shaking (120 rpm) (phage adsorption step). This suspension was then mixed with 3 ml of LB soft agar (0.75% w/v) supplemented with 9 mM CaCl<sup>2</sup> and 1.5 µg/ml ampicillin and poured onto LB agar plates supplemented with 0.5–1 µg/ml ampicillin. After 18 h incubation at 37◦C, lysis plaques were examined and enumerated. Ten lysis plaques were picked out from each plate

FIGURE 3 | Evaluation of the effect of sodium citrate and disodiun phosphate on E. coli EDL933 growth (CFU/ml). The values and error bars on the graph are averages of three independent experiments.



Supernatants of cultures at 18 h were analyzed after a 1/10 dilution in LB. Experiments were performed twice (assays 1 and 4).

and individually evaluated by PCR for stx carriage using a multiplex assay that detects stx1, stx2, and eae genes (Paton and Paton, 1998). This last gene was used to check the absence of chromosomal DNA from the STEC strain which could otherwise lead to false positive results in stx-phage PCR detection.

Phage quantification assays were repeated at least two times. To avoid possible interferences in the double-agar-layer method related with a possible chelating effect of disodium phosphate and sodium citrate, two more assays (named 3 and 4), were performed with addition of 100 µl CaCl<sup>2</sup> 0.1 M in the phage adsorption step.

# Evaluation of Extracellular Shiga toxin

Stx production was semi-quantified by using an enzyme immunoassay (EIA, Ridascreen <sup>R</sup> Verotoxin, R-Biopharm, Germany). Supernatants of cultures at 18 h were obtained by centrifugation at 12,000 × g for 10 min, diluted 1/10 in LB and then analyzed according to manufacturer instructions. Regarding supernatants from cultures treated with acids, they were neutralized before performing the ELISA.

The results were spectrophotometrically measured at 450 nm and classified as weak positive (1+) if the extinction was >0.1– 0.5 above the negative control, moderate (2+) (>0.5–1.0) and strongly positive 3+ (>1.0–2.0) to 4+ (>2.0). The assays were done twice.

# RESULTS AND DISCUSSION

Some practices used along the food chain, like addition of substances, may influence bacterial growth and could increase the rate of induction and propagation of bacteriophages. This could represent a risk when the bacteriophages carry genes for potent toxins such as Stx. In this study, we evaluated the effect of different compounds used in the meat industry on the growth of STEC strains, and on the production of stx-phages and Stx.

Cultures of FB3, FB5, and EDL933 strains (control cultures and cultures exposed to different concentration of the compounds) were incubated and monitored spectrophotometrically. For each treatment, similar OD<sup>600</sup> patterns were observed for each of the three strains (results for EDL933 are shown). Cultures added with mitomycin C showed an increase in the OD<sup>600</sup> during the first hours, reaching a maximum 2 h after induction, followed by a significant decrease, corresponding to a typical pattern of host cell lysis subsequent to induction of phage lytic cycle. This growth/lysis pattern was not observed in any of the cultures treated with the additives.

Cultures treated with lactic acid, acetic acid, or citric acid, at the concentrations tested, did not show an increase in OD<sup>600</sup> over time (**Figure 1** shows the results for treatments with 2.5% citric acid, 2.5% lactic acid, and 2.5% acetic acid). Phage titration assays showed that phages in supernatants of cultures treated with acids were below the accepted range for countable phage plaques (around 3–5 plaques per plate could be observed from undiluted supernatants, which represented a ∼2-log reduction in the plaque counts in comparison to the untreated control). In addition, extracellular Stx was not detected in the supernatants. Taking into account the results, those acids at tested concentrations inhibited the bacterial growth, without evidence of production of stx-phages and Stx.

Cultures treated with disodium phosphate or sodium citrate showed growth patterns similar to untreated cultures but with a reduction in the OD<sup>600</sup> values with increasing salt concentrations (**Figures 2A,B**). Besides, viable bacterial counts showed that EDL933 cultures exposed to these salts had titers similar or slightly lower than untreated control cultures, and markedly higher than cultures with mitomycin C (**Figure 3**). In the first two phage titration assays, the supernatants of the cultures treated with disodium phosphate or sodium citrate contained phages below the accepted range for countable phage plaques, while titers of 3 × 10<sup>2</sup> and 4 × 10<sup>5</sup> plaque forming units (pfu)/ml were observed in untreated cultures and cultures with mitomycin C, respectively. The supplementation of the supernatants with CaCl<sup>2</sup> in the adsorption step allowed quantification of phages from most of the cultures treated with salts, and also increased the number of pfu for both untreated and mitomycin C added controls. In all cases, the plaques analyzed by PCR were confirmed to correspond to stx2-phages.

Phage titers observed for the supernatants of cultures treated with disodium phosphate and those with 0.5 and 1% sodium citrate were lower than those observed for untreated cultures (**Figure 4**). The exception was the culture treated with 2.5 % sodium citrate, which showed titers similar to the untreated control (**Figure 4**). Regarding extracellular Stx production, cultures treated with disodium phosphate presented similar results to the untreated control (**Table 1**). The cultures supplemented with 0.5% and 1.0% sodium citrate showed similar or slightly higher levels of Stx than the water control, respectively. Interestingly, cultures with 2.5% sodium citrate showed Stx levels considerable higher than those of untreated cultures.

Altogether, the assays showed that sodium citrate and disodium phosphate at the tested concentrations slightly diminished the growth rate of the analyzed STEC strains without evidence of an increment of phage production. However, induction of stx-phage production cannot be ruled out, as phage can be produced but not released, or produced as non-infectious particles. Imamovic and Muniesa (2012) showed that chelation and the increase in stx2-phage induction are linked. In that study, culture treatment with 0.2 M sodium citrate showed effect on stx<sup>2</sup>

phage induction. Interestingly, our results showed an increase in Stx production in cultures treated with 2.5% sodium citrate in relation to the untreated control. The fact that production of Stx was increased in that condition could reinforce the previous arguments that stx phage production could be induced but not detected, or it could suggest Stx production independent of a complete phage production. It is important to note that EDL933 carries stx<sup>1</sup> and stx<sup>2</sup> genes, and both Stx1 and Stx2 toxin types can be detected by the Ridascreen kit. In consequence, we were not able to discriminate between Stx1 and Stx2 production, which are known to have differences in transcription control (Wagner and Waldor, 2002).

In the present study, we have shown that an additive, such as sodium citrate, although not having a strong effect on bacterial growth and phage production, can induce the production of Shiga toxin. Although we evaluated few STEC strains, and these results may not accurately represent the behavior of other strains, the present study alerts for a possible increase of Stx production by STEC in presence of some food additives.

Therefore, we consider that when testing the use of additives or other treatments applied in the food industry to decrease bacterial contamination, it is important to take into account the kind of bacteria that can be present. Particularly, in cases in which the bacteria can harbor phages encoding toxins that could be induced with the treatment.

# REFERENCES


# AUTHOR CONTRIBUTIONS

LL, LM, JB: performed the experiments, participated in the acquisition, analysis and interpretation of the data, approved the final version of the paper. PL, AK: supervised the laboratory work, participated in the analysis and interpretation of the data, drafted the manuscript, and approved the final version of the paper.

# FUNDING

This work was supported by Fondo para la Investigación Científica y Tecnológica (FONCYT) under Grant PICT N◦ 2012-2438; and Consejo Nacional de Investigaciones Científicas y Técnicas (CONICET) under Grant PIP 939. LL and LM were fellowship holders from Comisión de Investigaciones Científicas-Provincia de Buenos Aires (CICPBA) and Consejo Interuniversitario Nacional (CIN), respectively. JB is a fellowship holder from CONICET. PL and AK are members of the Research Career of CONICET.

# ACKNOWLEDGMENT

Authors thank M. R. Ortiz and Vet. G. H. Arroyo for their technical assistance.



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

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

# Whole-Transcriptome Analysis of Verocytotoxigenic Escherichia coli O157:H7 (Sakai) Suggests Plant-Species-Specific Metabolic Responses on Exposure to Spinach and Lettuce Extracts

#### Edited by:

Christophe Nguyen-The, Institut National de la Recherche Agronomique, France

#### Reviewed by:

Ana Allende, Spanish National Research Council, Spain Pascal Delaquis, Agriculture and Agri-Food Canada, Canada

\*Correspondence:

Nicola J. Holden nicola.holden@hutton.ac.uk

#### Specialty section:

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

> Received: 13 May 2016 Accepted: 29 June 2016 Published: 12 July 2016

#### Citation:

Crozier L, Hedley PE, Morris J, Wagstaff C, Andrews SC, Toth I, Jackson RW and Holden NJ (2016) Whole-Transcriptome Analysis of Verocytotoxigenic Escherichia coli O157:H7 (Sakai) Suggests Plant-Species-Specific Metabolic Responses on Exposure to Spinach and Lettuce Extracts. Front. Microbiol. 7:1088. doi: 10.3389/fmicb.2016.01088 Louise Crozier<sup>1</sup> , Pete E. Hedley<sup>1</sup> , Jenny Morris<sup>1</sup> , Carol Wagstaff<sup>2</sup> , Simon C. Andrews<sup>3</sup> , Ian Toth<sup>1</sup> , Robert W. Jackson<sup>3</sup> and Nicola J. Holden<sup>1</sup> \*

<sup>1</sup> Cell and Molecular Sciences, The James Hutton Institute, Dundee, UK, <sup>2</sup> School of Chemistry, Food and Pharmacy, The University of Reading, Reading, UK, <sup>3</sup> School of Biological Sciences, The University of Reading, Reading, UK

Verocytotoxigenic Escherichia coli (VTEC) can contaminate crop plants, potentially using them as secondary hosts, which can lead to food-borne infection. Currently, little is known about the influence of the specific plant species on the success of bacterial colonization. As such, we compared the ability of the VTEC strain, E. coli O157:H7 'Sakai,' to colonize the roots and leaves of four leafy vegetables: spinach (Spinacia oleracea), lettuce (Lactuca sativa), vining green pea (Pisum sativum), and prickly lettuce (Lactuca serriola), a wild relative of domesticated lettuce. Also, to determine the drivers of the initial response on interaction with plant tissue, the whole transcriptome of E. coli O157:H7 Sakai was analyzed following exposure to plant extracts of varying complexity (spinach leaf lysates or root exudates, and leaf cell wall polysaccharides from spinach or lettuce). Plant extracts were used to reduce heterogeneity inherent in plant–microbe interactions and remove the effect of plant immunity. This dual approach provided information on the initial adaptive response of E. coli O157:H7 Sakai to the plant environment together with the influence of the living plant during bacterial establishment and colonization. Results showed that both the plant tissue type and the plant species strongly influence the short-term (1 h) transcriptional response to extracts as well as longer-term (10 days) plant colonization or persistence. We show that propagation temperature (37 vs. 18◦C) has a major impact on the expression profile and therefore pre-adaptation of bacteria to a plant-relevant temperature is necessary to avoid misleading temperature-dependent wholescale gene-expression changes in response to plant material. For each of the plant extracts tested, the largest group of (annotated) differentially regulated genes were associated with metabolism. However, large-scale differences in the metabolic and biosynthetic pathways between treatment

types indicate specificity in substrate utilization. Induction of stress-response genes reflected the apparent physiological status of the bacterial genes in each extract, as a result of glutamate-dependent acid resistance, nutrient stress, or translational stalling. A large proportion of differentially regulated genes are uncharacterized (annotated as hypothetical), which could indicate yet to be described functional roles associated with plant interaction for E. coli O157:H7 Sakai.

Keywords: DNA microarray, stress response, E. coli O157:H7, vegetables, leaves, roots, biological, adaptation

# INTRODUCTION

Verocytotoxigenic Escherichia coli (VTEC) comprise an important group of food-borne pathogens that can enter the human food chain from contaminated plant as well as meat products. It is estimated that ∼20–25% of food-borne VTEC outbreaks worldwide arise from contaminated crop plants, based on publicly available reports (Greig and Ravel, 2009). Plant-based foods that carry the highest risk are leafy greens eaten raw as salads, and include foodstuff consumed raw or lightly cooked, i.e., fruits, vegetables, and sprouted seeds (EFSA Panel on Biological Hazards (BIOHAZ), 2013). It is now established that pathogenic E. coli can interact with plants and use them as secondary hosts (Holden et al., 2015). However, there are still many questions over the mechanism of plant adaptation and, in particular, the role of bacterial-stress responses in plant colonization. The main reservoir for VTEC is ruminants where regular fecal-shedding leads to bacterial dispersal into the environment, necessitating adaptation for survival and persistence and the prevailing view is that exposure to environments outwith the primary reservoir induces metabolic and physio-chemical stresses. However, the prevalence of certain E. coli isolates in the wider environment (Ishii et al., 2009; Brennan et al., 2010), including on plants, suggests that these bacteria do not simply survive and persist on plants, but instead have evolved into semi-specialized plant colonizers to facilitate persistence in the environment. Mesophilic species such as E. coli are adapted to proliferate over the range of temperatures encountered in the wider environment (Ratkowsky et al., 1982) given sufficient nutrients. It appears that VTEC belongs to a group of E. coli isolates that have evolved to adapt to a lifestyle that at least partly involves association with plants, and so can use them as secondary hosts (Holden et al., 2009). Therefore, a better understanding of the bacterial response to plants as hosts will help to improve our perspective of VTEC as a plant-borne human pathogen and thus inform on risk analysis and mitigation strategies.

Global-transcriptomic analysis has identified a range of responses (e.g., induction of stress-resistance) of pathogenic and non-pathogenic E. coli to various plant-associated environments (Kyle et al., 2010; Fink et al., 2012; Hou et al., 2012, 2013; Landstorfer et al., 2014; Linden et al., 2016). However, in many reports on plant-colonization transcriptomics the bacteria were initially cultured at body temperature (37◦C) and were subsequently exposed to plant (or plant extracts) at environmental temperature (∼18◦C); such experimental regimes result in a considerable temperature shift, in addition to the exposure to plant or plant extracts (Thilmony et al., 2006; Kyle et al., 2010; Hou et al., 2012, 2013; Jayaraman et al., 2014). In other reports, the entire experiment was performed at 37◦C (Bergholz et al., 2009; Fink et al., 2012; Visvalingam et al., 2013; Landstorfer et al., 2014) rather than at a temperature (i.e., ∼18◦C) relevant to plants growing in temperate zones. Temperature-dependent control of gene expression in E. coli and other bacteria is wellcharacterized (Phadtare and Inouye, 2008) and it is clear that temperature-induced global expression changes can obscure or complicate responses to other stimuli (Polissi et al., 2003; King et al., 2014). Thus, the specific reaction to the plant might not be accurately distinguished in previous reports where inappropriate temperature regimes were imposed.

Here, we investigate adaptation to and colonization of leafy salad plants by the predominant VTEC serotype O157:H7, using techniques for cultivable bacteria. We assess changes in gene expression profile of E. coli O157:H7 (isolate Sakai) at an environmentally relevant temperature to negate any temperature-dependent responses. Expression responses to a range of plant extracts of varying complexity were tested to avoid any host-defense influences, allowing a clearer identification of the other drivers of the bacterial response. In addition, use of extracts is expected to reduce the heterogeneity imposed on bacterial population by propagation on living plants (as observed for individual gene expression in planta (Rossez et al., 2014a). Spinach was selected as the focus for the response analysis because there have been a number of reported VTEC outbreaks from spinach. Lettuce was included as a comparison for the response to cell wall polysaccharides (CWPSs) as there have also been lettuce-associated VTEC outbreaks and our previous data showed differences in the adherence interactions (Rossez et al., 2014a). We focus on early expression responses (prior to proliferation), to minimize cell-division-dependent gene expression changes. This approach thus considers expression change during the initial, adaptive interactions that occur before establishment. The hypothesis tested is that E. coli O157:H7 undergoes adaptive changes in gene expression upon exposure to the plant that affects the outcome of colonization and persistence. We expect gene expression changes to be quite distinct from those reported during ruminant colonization (Dahan et al., 2004). Whole transcriptome analysis was coupled with investigation of E. coli O157:H7 growth potential over short-time scales in plant extracts and longer-term on plant hosts. The findings relayed here support the notion that plants are genuine secondary hosts for VTEC, rather than incidental habitats.

# RESULTS

fmicb-07-01088 July 9, 2016 Time: 13:1 # 3

# E. coli O157:H7 Exhibits Major Differences in Global Expression in Response to Growth at 37 or 18◦C

We hypothesized that some VTEC isolates undergo adaptive gene expression changes that enable them to colonize plants. In order to gain insight into the mechanisms of adaptation to the plant and to discern any tissue or plant species-associated differences that may occur, transcriptional changes exhibited by E. coli O157:H7 (Sakai) were examined following exposure to plant extracts. The initial stages of the plant–bacterium interaction were examined by whole-transcriptome analysis, using established E. coli DNA microarray technology. For this purpose, E. coli O157:H7 (Sakai) was cultured at a plant-relevant temperature (18◦C) prior to, and during, exposure to plant extract. However, in order to determine the impact of incubation temperature on the transcriptome, it was necessary to firstly compare global gene expression for cultures maintained in minimal M9 glycerol medium at 18◦C (both pre- and post-culture) and 37◦C. Both cultures were transferred to fresh medium at their respective temperatures for 1 h prior to sampling, representative of late lag to early exponential phase. The regime employed ensured assessment of temperature-dependent growth, avoiding any temperature shift or shock effects.

As expected, gene expression of E. coli O157:H7 (Sakai) grown for 1 h at 18◦C was markedly different from that of the culture grown at 37◦C (**Figure 1**). A total of 1,127 genes were differentially expressed in response to incubation temperature, representing 20.6% of E. coli O157:H7 Sakai ORFs. Of these, 500 genes were induced and 627 genes (9.16 and 11.48% of Sakai ORFs) were downregulated (Supplementary Table S1). Notable changes in expression of specific genes at 18◦C (cf. 37◦C) included repression of a subset of genes in the locus of enterocycte effacement (LEE). These included ler (130-fold repression; which encodes the master regulator of the lee genes), several type III secretion (T3SS) genes (ECs4583, escC, escJ, escS, and espF: repressed by 15-, 10-, 12-, 20-, and 10-fold respectively; Supplementary Table S1). The control of ler expression by low temperature is likely caused by H-NS silencing, which is known to suppress A/E lesion formation below 37◦C (Umanski et al., 2002). Motility genes were also repressed, particularly in the flg and fli loci (e.g., flgBCDE, 26–59-fold repressed; fliE, 26-fold repressed. Three hypothetical genes in an apparent operon of unknown function (ECs2623-2625) were amongst those most strongly repressed (∼200-fold), as were a series of prophage CP-933T genes (coxT, Z2971-4; 46–276-fold) possibly in response to QseA control (Kendall et al., 2010). The major class of genes subject to induction at 18◦C were those involved in various aspects of stress resistance: acid resistance (e.g., ECs2098, gadABCE; 38–121 fold induced), heavy-metal resistance (e.g., cusBX; 56–81-fold induced), putrescine metabolism (e.g., ygjG, ECs3955; 52–65 fold induced), multidrug efflux (e.g., sugE; 33-fold induced) and osmotic stress (proVW; ∼25-fold). In addition, a cluster of genes (ECs1653-1655; 14–28-fold) of unknown function was strongly induced as were several genes involved in biofilm formation (Z2229, ECs2085, bdm, c\_1914; 51–59-fold; Supplementary Table S1). In summary, the expression data suggest that growth at ambient rather than body temperature causes reduces motility and increases sessile behavior, reduced ability to colonize the mammalian gut and suppresses some prophage, but raises ability to resist a range of environmental stresses (a possible adaptation to slower growth at lower temperature). Such temperaturedependent changes would be expected to confound interpretation of expression data obtained in previous studies on bacterial plant colonization where a temperature change was included along with plant exposure – a complication that was avoided within the research reported below.

# Exposure of E. coli O157:H7 to Different Plant Extracts Elicits Distinct, Major Alterations in Global-Gene Expression

The whole transcriptome of E. coli O157:H7 (Sakai) was subsequently examined during the early stages of the plant interaction, under the conditions (18◦C, 1 h) employed above. Extracts of spinach (Spinacia oleracea) and roundhead lettuce (Lactuca sativa) were used as these have been associated with large-scale food borne outbreaks of VTEC previously (Cooley et al., 2007; Friesema et al., 2008). Leaf lysates (spinach) represent the combined cellular material and apoplast; root exudates (spinach) represent plant root-derived substrates; and leaf CWPSs (derived from spinach or lettuce) represent the cell wall components that include molecules involved in plant– microbe interactions. To provide an indication of any speciesspecific expression differences, a leaf CWPS extract from lettuce (L. sativa) was used to compare with that from spinach. CWPSs are the least complex of the plant extracts employed here and so are expected to induce more modest expression changes than the other plant samples, which should facilitate identification of any species differences that might occur.

Wholescale changes in E. coli O157:H7 (Sakai) gene expression occurred following 1 h of exposure to the different extracts at 18◦C (**Figure 1**). Exposure to spinach leaf lysate resulted in differential expression of 27% of the Sakai genome, 745 genes were induced and 738 were repressed, while 35% of the Sakai genome was differentially expressed on exposure to spinach root exudates: 981 induced and 972 repressed. In general, there appeared to be an inverse correlation in differential gene expression between exposure to spinach root exudates and spinach leaf lysates (**Figure 1**). The response to leaf CWPSs was examined to exclude the effects of other leaf components (e.g., apoplastic fluid and intracellular contents). Gene expression for E. coli O157:H7 (Sakai) exposed to spinach and lettuce CWPS for 1 h showed marked differences between species, with 460 and 97 genes displaying differential expression in response to lettuce and spinach CWPS, respectively, when compared to the response to the negative, no-plant control (an extract prepared from vermiculite, the inert plant growth substrate; **Figure 1**). Thus, the extent of expression change was far less with the CWPS (average of 3.2%) than with the leaf and root samples (average of 13%), as anticipated. Comparison between the species or tissue types showed little commonality in differentially expressed genes

(**Figure 2**). This is well-illustrated by the observations that only 13 genes were subject to regulation by all three spinach extracts, and only 23 of the 586 CWPS-regulated genes were also regulated by both the spinach and lettuce extracts.

To determine which groups of genes were affected by exposure to the plant extracts, analysis was performed for genes annotated with GO terms. GO-term enrichment enabled identification of over- or under-represented groups of genes that were differentially expressed by each treatment. Analysis of significantly enriched (p < 0.05) 'Biological Processes' were performed on a broad-scale level using GO-Slim terms, and supplemented with GO-Complete for a more a detailed breakdown of smaller classes of genes (**Figure 3**; Supplementary Table S2). When all of the plant extract treatments are considered together, Metabolic Processes is the category with the largest number (41–510) of affected genes, although there are clear differences between treatments in the ratio and number of up and down-regulated genes, and for different specific metabolic classes (**Figure 3**). Exposure to spinach leaf lysate resulted in differential expression of 364 genes in metabolic processes (163 induced, 201 repressed). The highest level of enrichment was for induced genes associated with lipid transport (six induced; sevenfold enriched) and there was significant positive enrichment for translation-related processes (translation, rRNA metabolism, regulation of translation) and protein metabolism (48, and 45 induced, respectively). Genes involved in primary metabolism were subject to a high degree of control with 127 genes induced and 162 repressed. Following exposure to spinach root exudates, a large number of genes associated with metabolic processes were down-regulated genes (340), with half as many (170) induced. In general, more down-regulated genes were enriched for the different GO term categories compared to those induced (67 vs. 33% in **Figure 3B**). The highest level of enrichment for repressed genes was seen for those associated with translationrelated processes (translation, rRNA metabolism, regulation of translation), of which 84 were down-regulated and just 12 induced; this pattern is the reverse of that seen above for spinach leaf lysate. The highest enrichment for induced genes (14) was in response to stress. These expression effects thus suggest that exposure to spinach root exudates caused increased stress combined with reduced translation capacity.

Upon exposure to spinach CWPS, the enriched gene GO terms categories were almost entirely represented by induced genes (159/165; **Figure 3C**). The largest group was in metabolic processes (39 induced, two repressed) and the highest level of enrichment was seen for a small class associated with maltodextrin transport (three genes induced; 51-fold enrichment). Exposure to lettuce CWPS resulted in a similar pattern of enrichment as for spinach CWPS, with the majority of the groups associated with different categories of metabolism, e.g., 105 induced, 35 repressed in metabolic processes (**Figure 3D**). Furthermore, most genes in each category were induced (as above for spinach CWPS), with only one category (response to stimulus) having a greater level of enrichment for repressed genes (seven repressed; fivefold enrichment).

In general, exposure to the different plant extracts generated distinct patterns of GO term enrichment (**Figure 3**), which was also distinct from that for growth at 18◦C (**Supplementary Figure S1**). However, some although some commonalities in enrichment occurred for the more specific categories. Furthermore, a large proportion of genes for each treatment type fell outside the GO annotations that are not considered by the enrichment analysis. Therefore, to examine the regulatory response in more detail, individual genes, or groups of related genes, are compared for each treatment type in more detail below.

# Metabolism

As indicated above, metabolism encompasses the largest number of differentially expressed genes for all of the extracts tested, although there were major differences between extracts. Components of glycolysis and the Krebs cycle that were induced on exposure to lettuce CWPS included enzymes required for the conversion of oxoglutarate to succinyl-CoA (sucAB, 17 and 20-fold), succinate (sucCD, 8- and 29-fold) and fumarate

differentially expressed in plant extracts. The number of genes was compared in a Venn diagram of all four treatments (A) and for the spinach extracts only (B). Key: 'Spin,' spinach (S. olercera); 'Lett,' lettuce (L. sativa); 'LL,' leaf lysates; 'RE,' root exudates; 'PS,' cell wall polysaccharides (CWPSs). Images were generated using the Venny program (Oliveros, 2007).

FIGURE 3 | GO term enrichment for response to plant extracts. GO terms for E. coli O157:H7 (Sakai) genes that were significantly differentially expressed following growth in spinach leaf lysates (A), spinach root exudates (B), CWPS extracts from spinach (C) or lettuce (D), relative to their respective controls. Data was obtained from the Gene Ontology Consortium website. Significantly enriched Biological Processes are shown for induced (blue) and repressed genes (red), using GO-Slim and selected GO-complete categories (indicated by '<sup>∗</sup> '; full list in Supplementary Table S2). The numbers of individual genes are adjacent to each bar on the charts.

(sdhABCD, 11–38-fold), and for malate oxidation (mqo, sixfold; **Figure 4**). This was coupled with a 20-fold induction of the gene encoding the DctA symporter, required for aerobic uptake of C4-dicarboxylates such as succinate (Davies et al., 1999; **Supplementary Figure S2**). The main gene associated with central metabolism that was induced on exposure to root exudates was acetyl-CoA synthetase (acs, fivefold). The gene (pdhR) encoding the pyruvate-dehydrogenase complex regulator (PdhR, an autoregulatory repressor responding to pyruvate) was induced on exposure to both spinach leaf lysate (fourfold) and lettuce CWPS (ninefold), as were the three genes in the PdhR-controlled aceEF-lpd operon (Supplementary Table S1). However, induction of lpdA was higher than that of aceEF in lettuce CWPS (23- cf. 4- to 6-fold) which reflects lpdA expression from an independent promoter and the involvement of lipoamide dehydrogenase (E3) component in both the pyruvate dehydrogenase and 2-oxoglutarate dehydrogenase multienzyme complexes (Cunningham et al., 1998). In contrast, pdhR was 15-fold repressed in root exudates. These findings indicate low cellular pyruvate levels upon exposure to root exudates, suggestive of low carbon source availability (see below).

Exposure to spinach root exudates or lettuce CWPS resulted in up-regulation of the methylgalactose uptake operon (mglABC) by 6–8-fold, but this was subject to eightfold repression by spinach leaf lysate (Supplementary Table S1). Lactose utilization genes (lacZY) were induced (ninefold) only in lettuce CWPS extracts, while genes required for utilization of sorbitol (srlAEBDMRQD) were 10-fold induced in root exudates but fourfold repressed in leaf extract. Genes for xylose metabolism (xylAB) were also induced, fivefold, in root exudates. Fatty acid degradation (fadABDEHIJKL) genes were strongly induced (56-fold) by spinach root exudates while fatty acid synthesis genes (fabHDGacpP-fadF,fadABIZ) were twofold repressed. This reciprocal regulation of the fatty acid systems is likely explained, in part, by the threefold repression of fadR, encoding the fatty-acid responsive fad gene repressor, and the sixfold induction of fabR specifying a repressor of fab genes. The regulatory response observed suggests enhanced availability of fatty acids in the root exudates. The reverse response was seen in lettuce CWPS: a threefold repression of the fad genes and a fourfold induction for the fad genes, indicating low fatty acid availability under this condition. Genes involved in purine and pyrimidine biosynthesis (purABCDEFHKLMNTU,carAB, pyrDFI) were the most strongly induced (average of 36-fold) genes on exposure to lettuce CWPS, but were 18- and 4-fold repressed by spinach root exudates and leaf extract. This indicates availability of nucleotide precursors in the root and leaf samples, but not in lettuce CWPS. Similarly for arginine, since carbamyl phosphate is regulated jointly by arginine and pyrimidines through transcriptional repression of carbamoyl phosphate synthase carAB (Caldara et al., 2006), which was evident in spinach leaf lysates and root exudates coupled with repression of arg and art genes (average 13-fold reduction), whereas in contrast arginine biosynthesis genes were induced in lettuce CWPS by 3- to 48-fold for argC,E,G,S and artIJ.

Exposure to the plant extracts induced changed in global regulators that play a functional role in control of growth. Expression of the gene encoding the factor-for-inversion stimulation protein (fis) was induced on exposure to spinach leaf lysates (threefold) and repressed (28-fold) in spinach root exudates (Supplementary Table S1). CsrA, a glycolysis activator and a gluconeogenesis repressor was induced fivefold in the presence of spinach root exudates. In addition, genes encoding the RNA polymerase subunits for the core enzyme, α, β, β 0 , and ω and sigma subunit 70 were all induced in leaf lysates (3- to 15 fold), whereas only the alternative sigma subunits for sigma E, sigma H and sigma S were marginally induced in spinach root exudates (twofold).

Iron acquisition is often linked to growth and division (Kohler and Dobrindt, 2011), and the extracts induced markedly different responses in associated systems. The ent genes encoding synthesis of the siderophore enterobactin were upregulated on exposure to spinach root exudates (2–11-fold), but not the leaf lysates (Supplementary Table S1), which might be partly explained by the ∼threefold reduced expression of the global iron-responsive repressor, Fur, in the root exudates. Similarly, expression of the haem-transporter (chu) genes were induced in root exudates (3– 20-fold compared), but not in spinach leaf lysates. These results suggest iron restriction is imposed by the root exudates, but not by the leaf extract. In contrast, the ferrous-iron-transport system (feoABC) genes were repressed for E. coli O157:H7 (Sakai) in spinach root exudates (∼3-fold). The iron-storage proteins were induced in two of the extracts: ftnA (fivefold) in lettuce CWPS; and ftnA and bfr (both ∼threefold) in spinach leaf lysates. The IscR-regulated gene cluster (iscRSUA-hscBA-fdx-iscX), associated with Fe–S cluster assembly was induced (average of 8.2-fold) in lettuce CWPS, but was fourfold repressed in spinach root extract.

# Stress Responses

The genes most strongly affected by exposure to leaf and root extracts were those associated with response to various stresses. The asr gene (acid-shock inducible periplasmic protein) was the most strongly induced gene on exposure to spinach leaf lysates (240-fold) and root exudates (637-fold), but not significantly affected in either of the CWPS extracts. Regulators and functional enzymes involved in glutamate-dependent acid resistance included the acid fitness island regulators, gadWX (Tramonti et al., 2008), repressed sixfold in spinach leaf lysates and 15- and 26-fold, respectively in lettuce CWPS, and but induced eightfold in spinach root exudates. Induction of gadAB and gadC encoding the glutamate decarboxylase and glutamate:gamma-aminobutyric acid antiporter occurred in root exudates (2–8-fold) in contrast to gadA repression in spinach leaf lysates or lettuce CWPS (12- or 21-fold, respectively), which supports regulatory control and response of the glutamatedependent acid resistance system. However, it was notable that gadE, a central activator of the response (Ma et al., 2003), was not differentially affected on response to root exudates.

Many of the genes encoding the cold shock proteins (cspA-I) were subject to regulatory change by the plant extracts (Supplementary Table S1). This was particularly clear for the spinach leaf lysates and root exudates where there appeared to be a reciprocal response: cspA and cspF-I were 12-fold induced in leaf lysate, but fivefold repressed in root lysate; whereas cspD was eightfold repressed or 12-fold induced, respectively. Genes

FIGURE 4 | Glycolysis superpathway gene expression profiles. Expression data for E. coli O157:H7 (Sakai) in response to different plant extracts was overlaid onto the metabolic pathway in EcoCyc (Keseler et al., 2013) to generate a color scale of expression from orange for induction to blue for repression and white for no change < ± twofold. Expression is provided for relevant genes in the pathways that were changed in at least one of the four plant conditions. Gene names are in italics and placed adjacent or close to their relevant substrates. The data for all four conditions are arranged in a grid, ordered as indicated in the Key: LL, leaf lysates and RE, root exudates for spinach; Spin\_PS, spinach CWPSs; Lett\_PS, lettuce CWPSs.

encoding the universal stress proteins (uspB,C,D,E,F, and G) were induced in root exudates (average of 10-fold), although three of these genes (uspB,D, and F) were 12–29-fold repressed in lettuce CWPS. spoT, associated with the stringent response, was moderately induced in response to spinach leaf lysates (twofold), but repressed sixfold in root exudates. Stress-response genes, e.g., spoT and cold shock genes play a functional role in response to metabolic-related changes and may reflect translational stalling (discussed below).

# Motility and Adherence

Gene associated with motility and biofilm formation are often associated with successful colonization of plants (Cooley et al., 2003; Van Houdt and Michiels, 2010). Both groups were strongly repressed in the baseline condition of growth in minimal medium at 18◦C compared to 37◦C, as indicated above. However, upon exposure to spinach whole-leaf lysates or root exudates, the genes encoding the master motility regulator FlhDC were repressed 7–28-fold, but were induced 23- and 52-fold (respectively) on exposure to lettuce CWPS (**Figure 5**). In lettuce CWPS, this increase in motility-gene regulator expression was coupled with repression of the biofilm-related gene, ECs2085 (bdm; 50–55-fold repressed) encoding the biofilm-dependent modulation protein, and a modest effect on the genes encoding curli fibers (csgA,B: both threefold). In contrast, curli genes were induced on exposure to spinach root exudates (also by threefold; Supplementary Table S1), indicative of a switch between sessility vs. motility. Some of the genes encoding fimbriae were induced, but only to moderate levels. For example, multiple signals for loc2 were induced in response to root exudates, including ECs0142 (yadM, a putative structural subunit) and yadK (also a structural subunit), by 2- and 3-fold, respectively.

# Hypothetical Genes

Genes annotated as hypothetical accounted for a large number of differentially expressed genes for all four treatments: 432, 603, 7, and 119 genes for spinach leaf lysates, root exudates, spinach CWPS and lettuce CWPS, respectively (Supplementary Table S1). They also accounted for high levels of differential expression: e.g., in spinach leaf lysates two hypothetical genes (b3238, b1722) were ranked as #2 and 3 for level of induction, at ∼50-fold. Probes corresponding to Z5022 and ECs4474 were induced 270- to 300-fold in spinach root exudates, but repressed in spinach leaf lysates and lettuce CWPS (3–92-fold). Some of these genes are unique to the O157:H7 serotype (**Table 1**) and not present in the closely related O157:H7 isolate EDL933. Four of these were differentially expressed in spinach leaf lysates or lettuce CWPSs: ECs1375, ECs2713, ECs4970 and ECs4976, ranging between threefold repressed and sevenfold induced. It is possible that some of these genes play a distinct role in plant colonization that has not yet been investigated.

# Colonization Potential Is a Reflection of Adaptive Gene Expression

To determine the extent to which the global-gene-expression changes reflect the colonization potential of the bacteria in

approximate change (indicated by '∼') is provided for flagella genes (data in Supplementary Table S2B). The data for all four conditions is arranged in a grid, ordered as indicated in the Key: LL, leaf lysates and RE, root exudates for spinach; Spin\_PS, spinach CWPSs; Lett\_PS, lettuce CWPSs. Regulatory connections, both direct and indirect (Pesavento et al., 2008; Pesavento and Hengge, 2012; Guttenplan and Kearns, 2013), with either positive (black arrow) or inhibitory (red bar) effects are shown.

different plant tissue extracts, the ability of the plant tissue extracts to support in vitro growth was assessed. For these assays, minimal M9 medium was used as a basal medium (without carbon source) supplemented with spinach leaf lysate or root exudates (normalized on the basis of protein content), or with 0.2% glycerol as a 'no-plant' control. Bacterial growth could not

#### TABLE 1 | Expression of selected genes encoding for hypothetical proteins of E. coli O157:H7 (Sakai).


Homologs in E. coli K-12 strain MG1655 and O157:H7 isolate EDL933 are indicated by 'X' or 'x' for presence/absence and the top BLASTn hit provided with the percentage of nucleotide identity. The gene expression data from microarray analysis is provided as fold change (relative to the appropriate control; NS is not significant), for each of the four plant extracts: spinach leaf lysates (LL) and root exudates (RE); and cell wall polysaccharides (CWPSs) from spinach and lettuce.

be assessed in medium containing the (insoluble) CWPS extract and as such, is not considered here. E. coli O157:H7 (Sakai) grew well in medium supplemented with spinach leaf lysate at 18◦C, reaching an OD<sup>600</sup> of 0.7 at 48 h, which was just-under 50% of that (1.7) achieved in M9 medium plus 0.2% glycerol (**Figure 6**). In addition, growth with the leaf lysate exhibited a very short lag phase, unlike that with glycerol where a ∼24 h delay in rapid growth was observed. This suggests that the bacteria acclimatized more rapidly to the medium with leaf extract than that with glycerol. In contrast, no growth was evident with spinach root exudate suggesting that carbon was at least one of the limiting energy sources. Indeed, when the root exudate and glycerol were used in combination strong growth was obtained that was similar to that with glycerol alone, suggesting that the weak growth in spinach root exudates was not due to the presence of factors that supress growth (**Figure 6**). No significant difference was found between the growth of E. coli O157:H7 (Sakai) in the glycerol only media compared to the glycerol plus root exudates media. To test whether the root exudate was deficient in suitable carbon sources, the composition of mono- and disaccharides in the extracts was examined by HPLC. The analysis showed ∼200-fold less glucose, fructose, and sucrose in the root exudate compared to the leaf lysate, supporting the suggestion that the root exudate provides limited levels of carbohydrate (**Table 2**).

Although root exudates were collected from plants grown under aseptic hydroponics conditions, and germinated from surfacesterilized seeds, it was apparent that there were native bacteria associated with the spinach plants. Cultivable bacteria were tentatively identified as Pseudomonas azotoformans (with 99.90% nt identity) and Pantoea agglomerans (99.95% nt identity). In our hands, the contaminating bacteria were repeatedly associated with spinach grown under these conditions indicating that they were seed-borne.

# E. coli O157:H7 (Sakai) Colonization Potential of Roots and Leaves of Spinach, Lettuce, and Vining Pea Plants

To examine the longer-term outcome of bacterial adaptation to the plant environment, the colonization potential of E. coli O157:H7 (Sakai) was determined on living plants over 10 days. Here, 'colonization potential' is defined as a measure of the ability



of the bacteria to survive and/or grow. Colonization potential was tested on the leaves and the roots of both spinach and lettuce, as above, and also on vining green pea (Pisum sativum), which is eaten raw as pea shoots, and wild prickly lettuce (Lactuca serriola), an ancestral relation of lettuce. In all cases, the whole E. coli O157:H7 population was enumerated with no attempt made to distinguish epiphytes from endophytes.

An E. coli O157:H7 (Sakai) inoculum of 6.3 log<sup>10</sup> CFU was applied to the adaxial (upper) and abaxial (lower) surface of the leaves of four different plant species and the bacteria enumerated over 10 days. There was a decrease in bacterial numbers compared to the starting inoculum for all four species, on both leaf surfaces. However, in each case, a higher average number of E. coli O157:H7 (Sakai) was recovered from the abaxial than adaxial surface after 10 days (**Figure 7**), although the difference was not significant at the 95% confidence level. The average number of E. coli O157:H7 (Sakai) on both leaf surfaces of both species of lettuce (L. sativa and L. serriola) decreased over the time tested, although the numbers recovered at d10 were significantly different: 1.66/2.84 (adaxial/abaxial) log<sup>10</sup> CFU for L. sativa and at the limit of detection (0.15/0.63 log<sup>10</sup> CFU, adaxial/abaxial) for L. serriola (p < 0.05), with bacteria only recovered from 22% of the samples for L. serriola for this time point. The number of E. coli O157:H7 (Sakai) on spinach also decreased from the starting inoculum and although higher counts were obtained from the abaxial side of the leaf at d2, by d10 they had reached similar levels, stabilizing at 0.69 adaxial and 1.99 abaxial log<sup>10</sup> CFU. Pea was the only plant where the numbers increased between d2 and d10, from 1.05 to 3.08 log<sup>10</sup> CFU (abaxial). By d10, significantly higher numbers were recovered found pea than L. serriola (adaxial, p < 0.01; abaxial, p < 0.05).

Colonization of roots was compared for plants grown in compost or hydroponics medium, to partly account for any potential effect from native compost-associated microbiota. Inoculation of compost-grown plants was achieved by partially immersing the plant pots in a bacterial suspension at 7.3 log<sup>10</sup> CFU/ml, which resulted in the recovery of between 2.0 and 4.0 log<sup>10</sup> CFU/g E. coli O157:H7 (Sakai) from the roots at the initial time point (1 h post inoculation; **Figure 8**). Despite some variation between plant species, the bacterial populations remained relatively stable and did not decrease as observed on leaves. E. coli O157:H7 (Sakai) recovered from P. sativum roots decreased marginally at day 2 but increased again by d10. Highest recovery at d10 occurred from spinach, followed by L. serriola, L. sativa and pea (3.4, 3.35, 2.76, and 2.24 log<sup>10</sup> CFU, respectively). For the colonization potential of E. coli O157:H7 (Sakai) on roots of plants grown under hydroponics (liquid) conditions, the inoculum (7 log<sup>10</sup> CFU/ml) was introduced into the medium adjacent to the roots. The number of E. coli O157:H7 (Sakai) recovered at the first time point was ∼2 orders of magnitude higher than that for compost-grown plants. The levels of E. coli O157:H7 (Sakai) recovered after 10 days were at least as high, or higher, than the initial inoculum (**Figure 2**). Greater recovery of bacteria occurred from L. serriola and spinach than L. sativa at d10 (7.04, 6.36, and 5.88 log<sup>10</sup> CFU, respectively). No proliferation of E. coli O157:H7 (Sakai) occurred in the hydroponics medium in the absence of plant roots, with the

the time points is expressed as the number of CFU recovered per gram of fresh tissue (Log10). Data is generated from nine replicate samples and was analyzed by

population at 4.46 log<sup>10</sup> CFU at d10, significantly different to E. coli O157:H7 (Sakai) from the three plants (p < 0.001). In our hands, it was not possible to remove surface-associated fungi from P. sativum seeds sufficiently well to allow its growth under aseptic hydroponics conditions; therefore, this combination was not tested. These experiments demonstrate that E. coli O157:H7 (Sakai) was able to either stabilize or increase its population on leaf and root, but that there were plant, tissue and growth media specific differences that affected colonization potential.

one-way ANOVA with the Tukey multiple comparison test.

# DISCUSSION

The aim of the experiments reported here was to examine adaptation to and colonization of a key crop-plant-associated pathogen (E. coli O157:H7 Sakai) to the leaves and roots of four distinct leafy vegetables. Examination of the initial expression response of the pathogen upon exposure to the plant allowed for assessment of the physiological changes that facilitate adaptation to the plant niche. E. coli O157:H7 (Sakai) was found to survive on the leaves of all four plants (two lettuce species, spinach and pea) over a 10-days period, although the numbers of cultivable bacterial declined from a high starting inoculum over the first 1– 2 days. Differences in degree of survival and the effect of time were observed suggesting that the bacteria experienced distinct leaf environments during their colonization of each of the four plants tested, which affected their recovery. Survival of E. coli O157 was superior in the root environment, with little decline in bacterial number observed over a 10-days period. However, again there were differences in bacterial recovery between the four plant species indicative of a species distinct impact on bacterial adaptation and survival.

The physiological response of E. coli O157 in response to plant extracts was examined in relation to persistence of E. coli O157 on leaves and roots. To facilitate this, extracts from spinach leaves and root exudates were used. Plant CWPS extracts from lettuce and spinach were included in an attempt to identify species-specific differences in response to plant factors. Extracts,

spinach (S. oleracea; B), wild prickly lettuce (L. serriola; C), and vining green pea (P. sativum; D), following inoculation of the compost (triangles) or hydroponics liquid media (circles). P. sativum was not grown under hydroponics conditions. Bacteria were not recovered from the hydroponic media-only control (squares). The average and the standard error of the mean for each of the time points is expressed as the number recovered per gram of fresh tissue (Log10). Data is generated from nine replicate samples and was analyzed by one-way ANOVA with the Tukey multiple comparison test.

rather than the live plant, were used to ensure sufficient bacterial recovery for expression analysis, to eliminate plant defense effects and to strictly control expression conditions to achieve good reproducibility. Such an approach has been used successfully by others previously (Kyle et al., 2010). However, use of extracts removes the plant host-dependent dynamic that could affect the bacterial response in comparison to the situation on live plants. The time of exposure was limited to just 1 h, which represents the period of initial adaptation. The four plant extracts induced marked differences in the transcript profiles for E. coli O157:H7 (Sakai) during the short (1 h) exposure at 18◦C, reflective of adaptation toward active metabolism and growth. The spinachleaf extract was shown to support growth of E. coli O157:H7 (Sakai) and although root exudate failed to enable such growth (due to an apparent lack of carbon source), it did not significantly inhibit growth when a suitable carbohydrate was provided. These observations suggest that the bacteria remain metabolically active and capable of mounting a regulatory response to their new environment during the 1 h exposure to the plant extracts.

Temperature is a major factor in differential gene expression (Phadtare and Inouye, 2008) likely to have influenced data obtained in many previous global-expression studies on bacterial colonization. Thus, the conditions employed here were controlled to ensure that the only change influencing E. coli O157:H7 (Sakai) gene expression was the introduction of plant extract to the culture medium. Indeed, this approach was vindicated by large-scale changes in gene expression (more than 20% of the genome) induced by growth at 18◦C (plant-relevant

temperature) instead of 37◦C (mammal-relevant temperature). Since a cold shock from 37 to 14◦C has been shown to result in induction of fli and flg genes (Phadtare and Inouye, 2004), the observed repression of these genes at 18◦C compared to 37◦C supports the lack of any cold shock imposed on E. coli O157:H7 (Sakai) under the conditions tested here. Furthermore, repression of genes associated with the type 3 secretion system (T3SS), in particular substantial down-regulation of the master regulator ler, support previous reported data on thermoregulatory control of T3SS in pathogenic E. coli at sub-mammalian temperatures (Umanski et al., 2002).

Other laboratories have investigated various aspects of the transcriptional response of E. coli to fresh produce (Kyle et al., 2010; Fink et al., 2012; Hou et al., 2012, 2013; Landstorfer et al., 2014; Linden et al., 2016) and alternative approaches have investigated genes required for plant-associated bacteria to colonize plant hosts (e.g., Silby et al., 2009). One of the most directly comparable studies examined early expression profiles of E. coli O157:H7 strain EDL933 to lettuce leaf lysates (Kyle et al., 2010), to mimic the bacterial response to damaged plant tissue. There are some parallels with these studies, such as up-regulation of genes involved in transport of metabolites (Kyle et al., 2010), but important specific differences occurred that are likely to have arisen from differences in the experimental approach.

In general, the transcriptome analysis paints a picture of E. coli O157:H7 (Sakai) undergoing a transition toward attempts at active growth, captured at different stages for the different extracts. Each of the plant extracts induced distinct transcriptional profiles for E. coli O157:H7 (Sakai), although metabolism was a common category. Growth phase transitions are known to induce significant changes in metabolite gene expression and production (Jozefczuk et al., 2010), which was reflected here by expression of genes involved in glycolysis and the Krebs cycle, e.g., induction of the genes for succinate and fumarate conversion in the presence of CWPS.

Several pieces of evidence show that E. coli O157:H7 (Sakai) was in a lag phase and in transition to growth following a 1 h exposure to spinach leaf lysates. The factor-for-inversion stimulation protein (FIS) was one of the most strongly induced global regulators in spinach leaf lysates. FIS is DNA binding protein that modulates chromosome dynamics and is highly induced during lag phase as the cells are preparing to divide (Schneider et al., 1997). Induction of MQO in leaf lysates, and repression of malate dehydrogenase (mdh), supports the idea that MQO can sustain low levels of TCA-cycle activity independent of MDH activity (van der Rest et al., 2000), and may also indicate that E. coli O157:H7 was undergoing transition to exponential phase. Induction of the pyruvate dehydrogenase system (phdR,ace,aceF,lpd) indicated the presence of pyruvate on exposure to both spinach leaf lysates and lettuce CWPS, since the operon is de-repressed in the presence of the carbohydrate (Quail et al., 1994). The pyruvate dehydrogenase complex is central to metabolism where PdhR is a master regulator of the genes involved for the transfer of pyruvate, the final product of glycolysis, into the Krebs/TCA cycle (Ogasawara et al., 2007).

The experimental set-up to investigate the response to plant extracts was designed not to incur a temperature shift, yet cspA and cspG were highly induced on exposure to spinach leaf lysates and lettuce CWPS. Cold shock proteins function as RNA chaperones, either re-folding misfolded transcripts or presenting them for degradation by RNases (Yamanaka et al., 1998), and are induced following translational stalling, e.g., on a shift to low temperatures or other 'stress-response' conditions. CspA and CspG RNA chaperones are highly expressed during antibiotic-driven translation inhibition (Etchegaray and Inouye, 1999) and their induction from spinach leaf lysates and lettuce CWPS coupled with the induction of spoT, a marker of the stringent response, supports the idea of a pause in translation during adaptation to the new environment. This may also explain induction of two E. coli O157:H7 (Sakai) cold shock genes on exposure to lettuce leaves of living plants (Linden et al., 2016). In contrast, E. coli K-12 csp genes were shown to be repressed on exposure to lettuce leaves elsewhere (Fink et al., 2012), although differences in the experimental set-up and baseline comparison may explain the observations.

Expression of cspD is indicative of nutrient stress (Yamanaka et al., 1998), and it was repressed in leaf lysates and lettuce CWPS extracts but induced following exposure to spinach root exudates, supporting the inability of E. coli O157:H7 (Sakai) to grow in this extract (**Figure 6**). This was further supported by induction of the usp family of genes, related to a variety of environmental assaults including DNA damage, oxidative stress and iron limitation (Nachin et al., 2005). Induction of the glutamate acid stress response system in root exudates was indicative of a response to acidic conditions in root exudates. The opposite response in spinach leafy lysates and lettuce CWPS indicated the presence of polyamines (spermidine and putrescine) that are known to repress the glutamate decarboxylase dependent acid response in E. coli (Chattopadhyay and Tabor, 2013).

There was evidence for catabolite control in response to CWPS and root exudates, from induction of high-affinity transport systems for malate and galactose normally seen under glucoselimiting conditions (Franchini and Egli, 2006) and induction of lacZY, in lettuce CWPS. There was evidence for degradation of carbohydrates (xylose and sorbitol) in spinach root exudates. A similar scenario of glucose-limitation was reported for E. coli O157:H7 (EDL933) in response to lettuce leaf lysates, e.g., with high levels of induction of genes for malate and sorbose uptake and metabolism (Kyle et al., 2010). Further evidence for use of alternative metabolites was from induction of acetyl-CoA synthetase (acs), which converts acetate to acetyl-CoA and is central to several metabolic pathways including the TCA cycle (Pietrocola et al., 2015). Changes in metabolic flux were also indicated by the presence of CsrA (and CsrD; Romeo, 1998). Fatty acid degradation (fad genes) and fatty acid synthesis (fab genes) is tightly balanced in the cell and co-regulated by FadR, a master regulator that represses fad genes and activates fab genes (My et al., 2015). In spinach root exudates the balance was tipped strongly toward fatty acid degradation, while the opposite occurred in spinach leaf lysates, indicative of membrane biogenesis required for active growth. Fatty acid degradation was also observed for colonization of Pseudomonas fluorescens (isolate SWB25) on sugar beet seedlings (Silby et al., 2009).

Iron scavenging is linked to growth and can also be associated with successful colonization of hosts and progression of disease (Kohler and Dobrindt, 2011). Iron limitation of Fe3++ was apparent from exposure to spinach root exudates, resulting in induction of systems for ferric iron and haem transport, via the enterobactin siderophore and Chu transport system respectively, while the ferrous iron transport system (feo) was repressed. The same limitation was not obvious in the other extracts, although there was some evidence for enterobactin production and transport on exposure to lettuce CWPS. Differences in access to extracellular and intracellular iron were evident, from induction of iron storage systems in spinach leaf lysates, in particular the ferritin protein FtnA (Andrews et al., 2003). Induction of the IscR Fe–S cluster assembly and repair system in the presence of lettuce CWPS supports previous data for exposure to lettuce (Kyle et al., 2010) or leaves of living plants (Linden et al., 2016), whereas the system was either un-induced or repressed in spinach extracts.

The biochemical analysis of the extracts coupled with the whole transcriptome analysis support a scenario in which E. coli O157:H7 (Sakai) adapted toward vegetative growth in the spinach leaf lysates, but could not grow and underwent multiple stress responses in the spinach root exudates. A likely possibility for the lack of available carbohydrates in the spinach root exudate preparations was depletion by native, 'contaminating' bacteria (Kuijken et al., 2015). Despite multiple attempts, it was not possible to remove these bacteria, and in our hands at least, they continue to be associated with spinach (see Materials and Methods).

Motility and adherence are important phenotypes that mark the initial stages of interaction with host tissue (Holden and Gally, 2004; Rossez et al., 2015). Induction of flagella genes in response to lettuce CWPS suggests that there is a signal for induction in the plant cell walls. This is consistent with the observation of flagella-mediated binding of E. coli to ionic lipids in the plasma membrane underlying the cell wall (Rossez et al., 2014b). Curli fibers are associated with biofilm formation and a switch from a motile to a sessile lifestyle is normally indicated by down-regulation of flagella genes and upregulation of curli genes (Pesavento et al., 2008). Such cross-regulation was evident from gene expression on exposure to root exudates consistent with a switch to sessility, i.e., repression of FlhDC, induction of curlin and Bdm. On the other hand, E. coli O157:H7 (Sakai) cells in CWPS were either motile or in transition, with some expression of flhD and csgE. Production of curlin fibers has been linked to colonization of fresh produce (Patel et al., 2011; Macarisin et al., 2012) and starvation conditions have been shown to induce a shift to a curli+ phenotype in plant-associated E. coli O157:H7 isolates (Carter et al., 2011). Bdm, a biofilm modulatory protein, is also linked to control of flagella genes, although the mechanism is as yet unclear (Kim et al., 2015).

The growth potential of E. coli O157:H7 (Sakai) on living plants differed to that seen in the plant extracts, which may reflect a contribution of host-derived factors. Whereas adaptation to the plant environment was only assessed during the initial stages of the plant–microbe interaction and E. coli O157:H7 (Sakai) was capable of growth in the extracts (with sufficient C-source), colonization potential reflects the capacity for the bacteria to become established on the plant. Growth in extracts to similar levels was demonstrated for E. coli O157:H7 (EDL933) in leaf lysates (Kyle et al., 2010). However, multiple factors are likely to impact the interaction including the complexity of the environment (i.e., how the plants are grown); accessibility of plant-derived metabolites; and the presence of an active host defense response, as had been reported for a number of plantassociated bacteria (Rosenblueth and Martinez-Romero, 2006; Holden et al., 2009; Hunter et al., 2010; Bulgarelli et al., 2012; Gutiérrez-Rodríguez et al., 2012; Hol et al., 2013; Turner et al., 2013).

Higher numbers of E. coli O157:H7 (Sakai) were recovered from the roots compared to the phyllosphere since the rhizosphere is a more hospitable environment protected from desiccation and UV irradiation that occur above ground, and has been reported to support substantially higher levels of other human pathogens (Brandl et al., 2004; Kroupitski et al., 2011). In general, higher levels of persistence were observed on the abaxial surfaces of leaves, which is also likely due to differences in UV irradiation and desiccation (Brandl, 2006). E. coli O157:H7 (Sakai) has previously been shown to have a propensity to bind to guard cells (Rossez et al., 2014b) and it is possible that differences in stomata density and distribution (Willmer and Fricker, 1996) may also influence the differences in the number of bacteria recovered. The reduction in numbers of E. coli O157:H7 (Sakai) on either surface of L. serriola may be due to the levels of polyphenols (Chadwick et al., 2016), which are associated with antimicrobial activity (Bach et al., 2011). E. coli O157:H7 (Sakai) was recovered in the highest numbers from the roots of plants grown under hydroponics conditions, which have a substantially reduced or absent native microbiota, suggesting that microbial competition is also an important factor in successful colonization.

Together, the data illustrate a complex interaction between RTE crop plants and E. coli O157:H7 that is dependent on 'system'-specific differences. Metabolism was found to be an important bacterial driver of the initial stages of the interaction. It is possible that some of the uncharacterized genes (annotated as hypothetical) that were strongly regulated on exposure to plant extracts play an important role in bacterial colonization of plants. Furthermore, the differences in bacterial growth in extracts compared to longer-term persistence on live plants indicate that plant and/or environmental factors also influence the interaction. The fact that the plant species and tissue type have a strong influence on the initial bacterial response as well as the potential for colonization provides information that can contribute to predictive modeling or risk-based analysis of the potential for microbial contamination of horticultural crops.

# MATERIALS AND METHODS

# Bacterial Strains and Growth Conditions

Escherichia coli O157:H7 strain Sakai (RIMD 0509952; Dahan et al., 2004) stx<sup>−</sup> Kan<sup>R</sup> was used for all experiments. The bacteria

were grown overnight in Luria-Bertani broth (LB broth) at 37◦C, 200 rpm supplemented with 25 µg/ml kanamycin. For growth curve experiments and colonization assays, the bacterial overnight culture was sub-cultured in a 1:100 dilution into MOPS medium (10x MOPS solution: 0.4 M MOPS, pH 7.4; 0.04 M tricine; 0.1 mM FeSO4; 95 mM NH4Cl; 2.76 mM K2SO4; 5 mM CaCl2; 5.28 mM MgCl2; 0.5 M NaCl; 10 ml micronutrients [3 µM (NH4)6MO7O24H2O; 0.4 mM H3BO3; 0.03 mM CoCl2; 0.01 mM CuSO4; 0.08 mM MnCl2; 0.01 mM ZnSO4]; 0.2% glycerol; 132 mM K2HPO4; 0.02 M thiamine HCl; 50x essential amino acids and 100x non-essential amino acids (Sigma-Aldrich, St. Louis, MO, USA) at 18◦C, 200 rpm until stationary phase. For all microarray experiments, the bacteria were subsequently sub-cultured into M9 minimal medium [20 ml 5x M9 salts (5x M9 salts): 64 g of Na2HPO4.7H2O; 15 g of KH2PO4; 2.5 g of NaCl; 5 g of NH4Cl; dissolved in 1 L sterile distilled water (SDW); 2 mM MgSO4; 0.1 mM CaCl2; 0.2% glycerol; pH 7.0] at 18◦C and 200 rpm, unless otherwise stated.

# Growth Curves

Escherichia coli O157:H7 (Sakai) cultures were grown to saturation (∼18 h) at 18◦C in M9 medium (as above) and diluted to an optical density of 0.02 (OD600) in M9 medium supplemented with 40% plant extracts, at 18◦C and 200 rpm. Extracts were normalized for protein content to a concentration of 1 µg/ml total protein using a Bradford assay using the Micro BCATM Protein Assay kit (Thermo Scientific, Waltham, MA, USA) according to the manufacturer's instructions. 1 ml of culture was taken at each time point and measured in a spectrophotometer at OD600. Samples were set up in triplicate and triplicate readings were taken for each. BSA standards were used to generate a standard curve for comparison.

# Colonization Assays

#### Leaves

Plants were grown in compost (containing peat, sand, limestone, perlite, calcite, Sincrostart, and Multicote 4) at 75% humidity, light intensity of 150 µmol m<sup>2</sup> s −1 (16 h photoperiod: day temperature of 26◦C, night temperature of 22◦C) for three to 4 weeks. The bacterial culture was washed and re-suspended in phosphate buffered saline (PBS) at an OD<sup>600</sup> of 1.0 (equivalent to ∼1x 10<sup>8</sup> CFU/ml). A soft marker pen with indelible ink was used to mark 1 mm spots on to the adaxial and abaxial sides of the leaves (separate leaves were used for each). Two leaves were taken per plant, with three technical replicates of each taken in total. 2 µl of the bacterial culture was applied to the spot and left to dry for 1 h. Two µl of sterile PBS was pipetted onto the spots for un-inoculated, control plants. At each time point, leaves were excised, weighed and macerated in 1 ml PBS. The samples were diluted to 10−<sup>3</sup> and plated onto Sorbitol-MacConkey (SMAC) agar containing 25 µg/ml kanamycin, incubated at 37◦C (∼20 h) and the colonies counted the following day. The microbiological count data was calculated based on the fresh weight of each leaf and standardized as CFU per gram fresh tissue. Three biological repeats of the experiment were carried out. The data was transformed (log10) and analyzed by ANOVA using the Tukey multiple correction test (GraphPad Prism, version 5.0).

### Roots of Compost-Grown Plants

Plants were grown as for the leaf colonization assay. The bacterial culture was diluted to an OD<sup>600</sup> of 0.02 (∼1.6 × 10<sup>7</sup> CFU ml−<sup>1</sup> ) in 1 L of SDW. The plants were not watered for the preceding 24 h and were inoculated by partly immersing their pots in the bacterial suspension for a period of 1 h. Uninoculated negative control plant pots were immersed in 1 L SDW. At each time point, the roots were detached, washed gently in 20 ml PBS to remove the compost and weighed. The roots were then macerated and processed as for the leaves. Three biological repeats of the experiment were carried out. Data was analyzed as for the leaf colonization assay.

### Roots of Hydroponics-Grown Plants

Seeds were surface sterilized with 2% (w/v) calcium hypochlorite (CaCl2O2) and germinated on distilled water agar. Seedlings were grown under aseptic conditions in 300 ml hydroponic pots containing 10 g of sterilized perlite with 10 ml of 0.5x Murashige and Skoog (MS) media with no added sucrose. The bacterial culture was washed and re-suspended in fresh 0.5x MS at an OD<sup>600</sup> of 0.02. The 10 ml of 0.5x MS was removed from the hydroponic pots and replaced with 10 ml of bacterial suspension. The plants were left for 1 h before the first time-point. Uninoculated negative control hydroponic pots had only 0.5x MS solution added. At each time point, the roots were excised and processed as for the leaf colonization assay. A bacteria-only control in 0.5x MS, with no plant, did not show any growth of bacteria. Data was analyzed as for the leaf colonization assay.

#### Plant Extract Preparation

Spinach cv. Amazon (Spinacia oleracea), lettuce cv. Salinas (L. sativa), prickly lettuce (Lactuca serriola), and vining pea (Pisum sativum) were used in this study. Plants were grown in compost for 3–4 weeks for leaf lysate extract preparation. The leaves were removed, snap frozen in liquid nitrogen and ground to a fine powder. 10 g of the leaf powder was re-suspended in 40 ml SDW and centrifuged for 15 min at 5,000 × g. The supernatant was heated at 50◦C for 30 min and clarified by centrifugation at 5,000 × g for 20 min and the final supernatant passed through a 0.22 µm sterile filter.

For root exudate extracts, seeds were first surface sterilized using 2% (w/v) calcium hypochlorite (CaCl2O2) for 15 min and germinated on distilled water agar. Seedlings were transferred to hydroponic pots containing 10 g rockwool and sterile 0.5x MS (no sucrose). After 3 weeks growth, the exudates were removed from 24 plants by three successive aqueous extractions with 50 ml SDW and clarification through a 0.22 µM filter. Spinachassociated bacteria were isolated on LB agar at room temperature, crude whole cell lysates prepared and subject to PCR for the 16 rRNA genes. The variable 2, 3, and 6 regions were sequenced and the isolates tentatively identified from BLAST analysis of the DNA sequence.

To prepare leaf CWPS extracts, plants were grown in vermiculite (William Sinclair Holdings, Lincoln, UK) containing

Osmocote Start 6-weeks short-term base fertilizer for 3–4 weeks. The leaves were excised and macerated to a fine powder in liquid nitrogen. 10 g of the leaf powder was re-suspended in 40 ml SDW and the debris pelleted by centrifugation for 15 min at 5,000 × g. The plant powder was processed to obtain the alcohol insoluble residue (Popper, 2011). Briefly, 70% ethanol was added to the plant powder in a 5:1 ratio and mixed for 10 min at 80 rpm. The samples were pelleted by centrifugation at 5,000 × g for 10 min and the supernatant discarded. Ethanol extraction was repeated five times. 100% acetone was then added to the powder and mixed at 80 rpm for 10 min. The acetone wash step was repeated twice. Following this, the supernatant was discarded and the polysaccharide powder was left to air dry for 48 h. A no-plant vermiculite-only negative control was prepared using the same method to account for any residual carry-over from the vermiculite and serve as a base-line to assess gene expression.

# Plant Extract Inoculation for Whole Transcriptome Analysis

#### Temperature

Escherichia coli O157:H7 (Sakai) was grown in M9 minimal media at either 37 or 18◦C until early stationary phase (OD<sup>600</sup> of ∼1). Each culture was washed in M9 once and sub-inoculated to an OD<sup>600</sup> of 0.5 in fresh M9 media with 0.2% glycerol, which had been preheated to 37 or 18◦C. The cultures were incubated for 1 h at 37 or at 18◦C, with aeration (200 rpm). After 1 h, the cultures were harvested for RNA isolation by mixed with RNA Protect (Qiagen).

### Leaf Lysates/Root Exudates

Escherichia coli O157:H7 (Sakai) diluted to an OD<sup>600</sup> of 0.5 into fresh M9 medium supplemented with 40% (v/v) spinach leaf lysate or root exudate extract (normalized to 1 µg/ml total protein content). Cultures were incubated at 18◦C with aeration (200 rpm) for 1 h, harvested and mixed 1:1 with RNA Protect (Qiagen). E. coli O157:H7 (Sakai) grown at 18◦C in M9 media with 0.2% glycerol without any plant extracts was used as the in vitro control and served as a base-line for gene expression.

### Leaf Cell Wall Polysaccharides

Escherichia coli O157:H7 (Sakai) was diluted at an OD<sup>600</sup> of 0.5 into fresh M9 medium with one of three supplements: 1% (w/v) spinach S. oleracea leaf CWPSs; or 1% (w/v) lettuce (L. sativa) leaf CWPSs (normalized to 1 µg/ml total protein content); or 1% (w/v) vermiculite no-plant control extract, and incubated at 18◦C, 200 rpm for 1 h and processed as for the leaf lysates experiment.

#### RNA Extraction

Total RNA was extracted from samples stored in RNA Protect using the RNeasy Plant Mini kit RNA extraction protocol (Qiagen). The concentration of total RNA was estimated using a NanoDrop (Wilmington, MA, USA) spectrophotometer and visualized for quality using a Bioanalyzer 2100 (Agilent Technologies, Santa Clara, CA, USA). Genomic DNA carryover was removed using the TURBO DNA-free kit (Ambion, Life Tech) and verified as DNA-free from a negative PCR reaction using gyrB primers, compared to a positive control.

#### Microarray Processing and Analysis

The complete microarray experimental plan and datasets are available at ArrayExpress (<sup>1</sup> accessions #E-MTAB-3249 and E-MTAB-4120). Microarray processing was essentially performed as described for other prokaryotic species (Venkatesh et al., 2006). Briefly, cDNA synthesis was performed using Superscript reverse transcriptase (Invitrogen) and labeled with either Cy3 or Cy5 dye according to the microarray plan. The Agilent microarray used (Agilent #G4813A-020097; accession # A-GEOD-8701) contains 15,208 probes representing transcripts from a total of four genomes: E. coli MG1655; E. coli CFT073; E. coli O157:H7 EDL933; and E. coli O157:H7 Sakai. A single color approach was used for the temperature, leaf lysate and root exudate conditions. Four replicate samples of each of the four conditions [E. coli (Sakai) in: (i) M9 media at 18◦C; (ii) M9 media at 37◦C; (iii) M9 media plus spinach leaf lysate at 18◦C, and; (iv) M9 media plus spinach root exudate at 18◦C] were run. For the polysaccharide conditions, a two-color approach was used. Eight replicate samples of the control condition [E. coli (Sakai)] in M9 media with vermiculite extract at 18◦C were labeled as detailed, along with four replicates of the two test conditions [E. coli (Sakai)] in M9 media with spinach/lettuce leaf CWPSs. Labeled cDNA was hybridized to microarrays as recommended by the manufacturer. Microarrays were scanned using a G2505B scanner (Agilent) and data extracted from images using Feature Extraction software (Agilent v. 10.7.3.1) with default parameters. Data were subsequently imported into GeneSpring GX 7.3 (Agilent, USA). Quality control was applied to remove those probes with no consistent signal in any of the conditions tested, whereby data was filtered on flags being present or marginal in two out of the three replicate samples. Principal component analysis was performed to identify any outliers. For all microarray experiments, statistical analysis of the datasets was carried out by performing a Volcano plot on each condition with a twofold minimum cut off for fold change and a Student's t-test with multiple testing correction (Benjamini and Hochberg; p ≤ 0.005 for temperature, spinach leaf lysates and spinach root exudate conditions; p ≤ 0.01 for lettuce polysaccharide; p ≤ 0.05 for spinach polysaccharide). Filtering was carried out in Microsoft Excel on raw values from the array pixel density (>50), and where multiple probes represented the same gene: as a consequence of the array design genes are represented with 1–4 probes for the four strains MG1655 ('b' accession number prefix), CFT073 ('c'), TUV93-0 ('Z'), and Sakai ('ECs'). Data for duplicate probes were removed to provide data preferentially for ECs or Z, followed by b accession numbers. Metabolic pathway analysis was performed using EcoCyc<sup>2</sup> (Keseler et al., 2013). GO enrichment analysis was performed from the Gene Ontology Consortorium website (The Gene Ontology Consortium, 2015), using the PANTHER classification system (Mi et al., 2016) for Biological Processes

<sup>1</sup>https://www.ebi.ac.uk/arrayexpress/

<sup>2</sup>http://ecocyc.org

(GO-Slim and GO-Complete), and only classes with significant enrichment (p < 0.05) were analyzed. Blastn analysis was carried out at the NCBI database (Altschul et al., 1990).

#### HPLC Analysis

Leaf lysate and root exudate extracts were prepared for HPLC by ethanol extraction. 10 ml of samples were freeze dried and re-suspended in 80% ethanol. The mixture was centrifuged at 5,000 × g for 30 min. The supernatant was collected, and freeze dried once more after ethanol evaporation before being resuspended in 2 ml molecular biology grade water. Leaf CWPS samples were prepared by TFA hydrolysis. Briefly, 10 mg of polysaccharide samples was incubated with 2 M trifluoracetic acid and boiled at 100◦C for 1 h. The TFA was removed by evaporation and the sample freeze dried before re-suspending in 1 ml of molecular biology grade water. Samples were run on a Dionex chromatography machine with the Chromeleon software using a PA100 column for glucose, fructose, sucrose, arabinose, and rhamnose.

## Quantitative Reverse Transcriptase (qRT) PCR Analysis and Microarray Data Validation

All qRT-PCR reactions were set up with iTaqTM Universal SYBR© Green Supermix (Bio-Rad) according to manufacturer's instructions, with 300 nm of primer and run in a Step-One Plus machine (Applied Biosystems) using the 11Ct method with an additional melt-curve analysis. All primers were validated as having 95–100% efficiency prior to 11Ct analysis, similar to that of the reference gene. Reference genes were validated using the GeNorm kit and software (Primer Design, Southampton, UK), for which gyrB was used as it was stably expressed under all microarray conditions (M > 0.1). qRT-PCR data was analyzed by averaging three technical and three biological replicates and applying the formula 2−11Ct, with the data normalized to the calibrator sample and to the validated reference gene. Microarray expression data was validated by examining the expression of 18 genes by qRT-PCR and measuring the correlation coefficient between both datasets for relevant subsets of these genes (i.e., significantly up or down-regulated). This was done for the microarrays samples and for an independent set of samples. The correlation coefficients (R 2 ) were 0.9994; 0.9851; 0.9160; 0.9730; 0.9201 for the temperature; spinach leaf lysate; spinach root exudate; spinach CWPS; and lettuce CWPS treatments, respectively.

# REFERENCES


# AUTHOR CONTRIBUTIONS

LC: acquisition, analysis, and interpretation of the data; drafting and revision of the m/s; PH and JM: design, acquisition, and analysis of microarray data; drafting the m/s; CW: provision of L. serriola; design of the colonization experiments; drafting the m/; SA and IT: design of microarray and colonization experiments; drafting and revising the m/s; data interpretation; RJ: conception and design of the work; drafting and revision of the m/s; NH: conception and design of the work; analysis and interpretation of the data; drafting and revision of m/s; all: final approval; agreement for accountability.

# ACKNOWLEDGMENTS

NH, IT, PH, and JM receive funding from the Scottish Government's Rural and Environment Science and Analytical Services Division (RESAS); RJ, CW, and SA receive funding from the University of Reading; LC was funded by a joint studentship awarded to RJ and NH, funded by RESAS and the University of Reading. We acknowledge laboratory support from Jacqueline Marshall and Rob Hancock.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2016.01088

FIGURE S1 | GO term enrichment for response to temperature. GO terms for E. coli O157:H7 (Sakai) genes that were significantly differentially expressed following growth in M9 medium at 18◦C relative to 37◦C. Data was obtained from the Gene Ontology Consortium website. Significantly enriched Biological Processes are shown for induced (blue) and repressed genes (red), using GO-Slim and selected GO-complete categories (indicated by '<sup>∗</sup> '; full list in Supplementary Table S2). The numbers of individual genes are adjacent to each bar on the charts.

FIGURE S2 | Validation of E. coli O157:H7 (Sakai) gene expression from the microarray results by qRT-PCR. Expression of malE (A) and dctA (B) in response to CWPSs was compared for the microarray probes and from two separate sets of RNA extractions for the qPCR analysis (qPCR.1, qPCR.2). The no-plant control is designated 'Control' and either spinach or lettuce CWPSs by 'PS.' Numbers represent the average of nine technical replicates for each of the qPCR datasets; data was analyzed using the 11CT method by comparing to a validated housekeeping gene (GeNorm), using primers of equal (95–100%) efficiency.



environmental conditions including radish sprouts and cattle feces. BMC Genomics 15:353. doi: 10.1186/1471-2164-15-353


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

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

# Corrigendum: Whole-Transcriptome Analysis of Verocytotoxigenic *Escherichia coli* O157:H7 (Sakai) Suggests Plant-Species-Specific Metabolic Responses on Exposure to Spinach and Lettuce Extracts

Louise Crozier <sup>1</sup> , Pete E. Hedley <sup>1</sup> , Jenny Morris <sup>1</sup> , Carol Wagstaff <sup>2</sup> , Simon C. Andrews <sup>3</sup> , Ian Toth<sup>1</sup> , Robert W. Jackson<sup>3</sup> and Nicola J. Holden<sup>1</sup> \*

*<sup>1</sup> Cell and Molecular Sciences, The James Hutton Institute, Dundee, UK, <sup>2</sup> School of Chemistry, Food and Pharmacy, The University of Reading, Reading, UK, <sup>3</sup> School of Biological Sciences, The University of Reading, Reading, UK*

Keywords: DNA microarray, stress response, *E. coli* O157:H7, vegetables, leaves, roots, adaptation, biological

#### *Edited by:*

**A corrigendum on**

*Christophe Nguyen-The, Institut National de la Recherche Agronomique, France*

*Reviewed by: Maria T. Brandl, Agricultural Research Service, USA*

> *\*Correspondence: Nicola J. Holden nicola.holden@hutton.ac.uk*

#### *Specialty section:*

*This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology*

*Received: 02 September 2016 Accepted: 08 September 2016 Published: 21 September 2016*

#### *Citation:*

*Crozier L, Hedley PE, Morris J, Wagstaff C, Andrews SC, Toth I, Jackson RW and Holden NJ (2016) Corrigendum: Whole-Transcriptome Analysis of Verocytotoxigenic Escherichia coli O157:H7 (Sakai) Suggests Plant-Species-Specific Metabolic Responses on Exposure to Spinach and Lettuce Extracts. Front. Microbiol. 7:1506. doi: 10.3389/fmicb.2016.01506* **Whole-Transcriptome Analysis of Verocytotoxigenic Escherichia coli O157:H7 (Sakai) Suggests Plant-Species-Specific Metabolic Responses on Exposure to Spinach and Lettuce Extracts** by Crozier, L., Hedley, P. E., Morris, J., Wagstaff, C., Andrews, S. C., Toth, I., et al. (2016). Front. Microbiol. 7:1088. doi: 10.3389/fmicb.2016.01088

In the Introduction to the article, one of the references, Kyle et al. (2010) was included in error in the statement "However, in many reports on plant-colonization transcriptomics the bacteria were initially cultured at body temperature (37◦C) and were subsequently exposed to plant (or plant extracts) at environmental temperature (∼18◦C); such experimental regimes result in a considerable temperature shift, in addition to the exposure to plant or plant extracts (Thilmony et al., 2006; Kyle et al., 2010; Hou et al., 2012, 2013; Jayaraman et al., 2014)". Instead, the Kyle et al. study used a temperature of 28◦C throughout the experiment, for initial culturing of the inoculum and for subsequent bacteria-plant interactions, so that the bacteria did not encounter a temperature shift. An extract from the Methods section is provided below. Erroneous inclusion of this reference in the statement has no impact on the scientific validity of the results presented.

The correct statement is: "However, in many reports on plant-colonization transcriptomics the bacteria were initially cultured at body temperature (37◦C) and were subsequently exposed to plant (or plant extracts) at environmental temperature (∼18◦C); such experimental regimes result in a considerable temperature shift, in addition to the exposure to plant or plant extracts (Thilmony et al., 2006; Jayaraman et al., 2014; Hou et al., 2012, 2013)"

Extract of the Methods section from Kyle et al. (2010):

"For experiments measuring early gene expression in lettuce leaf lysate (by microarray and QRT-PCR) or on shredded lettuce (by QRT-PCR), overnight cultures were transferred to fresh M9-glucose and grown for several hours into the mid-log phase of growth at 28◦C and 150 rpm and then washed twice with KP buffer before inoculation. The lysates were inoculated with EcO157 cells in the mid-log phase of growth in minimal medium in order to isolate the bacterial responses to romaine lettuce lysates from changes in gene expression solely caused by the transition out of stationary phase (6). Lysates were inoculated at 5×10<sup>6</sup> CFU/ml for growth experiments and at 10<sup>8</sup> CFU/ml for microarray analysis and QRT-PCR and then incubated at 28◦C with shaking at 150 rpm. In order to evaluate gene expression in EcO157 in lettuce lysates, samples for RNA extraction and subsequent microarray or QRT-PCR analysis were taken at 15 or 30 min after exposure of mid-log-phase EcO157 cells to freshly prepared lysates. Short incubation periods in the lysates at 28◦C were used in order to characterize the early response of the pathogen to

# REFERENCES


fluids leaking out of leaf cells after injury occurred, at an ambient daytime temperature that would be present in the field during growth and harvesting, or during processing under conditions that would fail to maintain cool temperatures."

# AUTHOR CONTRIBUTIONS

NH: wrote the correction statement. LC, PH, JM, CW, SA, IT, RJ, and NH: approved statement.

Thilmony, R., Underwood, W., and He, S. Y. (2006). Genome-wide transcriptional analysis of the Arabidopsis thaliana interaction with the plant pathogen Pseudomonas syringae pv. tomato DC3000 and the human pathogen Escherichia coli O157:H7. Plant J. 46, 34–53. doi: 10.1111/j.1365-313X.2006.02725.x

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

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

# Diversity of Survival Patterns among Escherichia coli O157:H7 Genotypes Subjected to Food-Related Stress Conditions

#### Mohamed Elhadidy<sup>1</sup> \* and Avelino Álvarez-Ordóñez<sup>2</sup>

<sup>1</sup> Department of Bacteriology, Mycology and Immunology, Faculty of Veterinary Medicine, Mansoura University, Mansoura, Egypt, <sup>2</sup> Teagasc Food Research Centre, Fermoy, Ireland

### Edited by:

Paula Teixeira, Universidade Católica Portuguesa, Portugal

#### Reviewed by:

Odile Tresse, French National Institute for Agricultural Research/Nantes-Atlantic National College of Veterinary Medicine, Food Science and Engineering, France Alejandra Krüger, CIVETAN– CONICET and Universidad Nacional del Centro de la Provincia de Buenos Aires, Argentina

> \*Correspondence: Mohamed Elhadidy mm\_elhadidy@mans.edu.eg

#### Specialty section:

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

Received: 03 December 2015 Accepted: 29 February 2016 Published: 15 March 2016

#### Citation:

Elhadidy M and Álvarez-Ordóñez A (2016) Diversity of Survival Patterns among Escherichia coli O157:H7 Genotypes Subjected to Food-Related Stress Conditions. Front. Microbiol. 7:322. doi: 10.3389/fmicb.2016.00322 The purpose of this study was to evaluate the resistance patterns to food-related stresses of Shiga toxin producing Escherichia coli O157:H7 strains belonging to specific genotypes. A total of 33 E. coli O157:H7 strains were exposed to seven different stress conditions acting as potential selective pressures affecting the transmission of E. coli O157:H7 to humans through the food chain. These stress conditions included cold, oxidative, osmotic, acid, heat, freeze-thaw, and starvation stresses. The genotypes used for comparison included lineage-specific polymorphism, Shiga-toxin-encoding bacteriophage insertion sites, clade type, tir (A255T) polymorphism, Shiga toxin 2 subtype, and antiterminator Q gene allele. Bacterial resistance to different stressors was calculated by determining D-values (times required for inactivation of 90% of the bacterial population), which were then subjected to univariate and multivariate analyses. In addition, a relative stress resistance value, integrating resistance values to all tested stressors, was calculated for each bacterial strain and allowed for a ranking-type classification of E. coli O157:H7 strains according to their environmental robustness. Lineage I/II strains were found to be significantly more resistant to acid, cold, and starvation stress than lineage II strains. Similarly, tir (255T) and clade 8 encoding strains were significantly more resistant to acid, heat, cold, and starvation stress than tir (255A) and non-clade 8 strains. Principal component analysis, which allows grouping of strains with similar stress survival characteristics, separated strains of lineage I and I/II from strains of lineage II, which in general showed reduced survival abilities. Results obtained suggest that lineage I/II, tir (255T), and clade 8 strains, which have been previously reported to be more frequently associated with human disease cases, have greater multiple stress resistance than strains of other genotypes. The results from this study provide a better insight into how selective pressures encountered through the food chain may play a role in the epidemiology of STEC O157:H7 through controlling the transmission of highly adapted strains to humans.

Keywords: E. coli O157:H7, genotypes, food, stress, survival

# INTRODUCTION

Shiga toxin producing Escherichia coli (STEC) O157:H7 is a food-borne zoonotic pathogen that represents a major public health concern worldwide (Lee et al., 2011). Cattle are the primary reservoir of E. coli O157:H7 and the food chain is the predominant transmission route for outbreaks caused by this pathogen (Rangel et al., 2005; Grant et al., 2008). Outbreaks are commonly attributed to the consumption of contaminated meat, milk, and dairy products, particularly those derived from cattle (Griffin, 1995). Symptoms of infection include bloody diarrhea, vomiting, haemorrhagic colitis, and life-threatening sequelae, such as haemolytic uremic syndrome (HUS) (Rangel et al., 2005). Because of the potential complications of the infection by this pathogen and its low infective dose it is important to reduce the contamination throughout the food chain to low levels (Teunis et al., 2004).

During food processing (particularly in minimally processed foods or those processed using a hurdles technology approach), STEC O157:H7 encounter different stress conditions that might affect their fate along the food chain and therefore their transmission to humans. Moreover, STEC O157:H7 may develop adaptive responses to stress that may enable survival under more harsh conditions, enhancing resistance to subsequent processing conditions, and even impacting the disease-causing potential of bacterial strains and therefore the final outcome of the foodborne disease (Samelis and Sofos, 2003; Alvarez-Ordóñez et al., 2015). These stress conditions include (i) cold stress that occurs during food marketing and storage; (ii) oxidative stress that is induced in food systems by agents added to aid processing due to their powerful bactericidal effect; (iii) osmotic stress, mainly due to the use of salt as a common food preservative to control the growth of food spoilage and pathogenic bacteria; (iv) acid stress imposed by organic acids used to reduce the microbial load in foods or by the gastric acidity that represents the first line of the host innate defense following ingestion of contaminated food; (v) heat induced stress in food pasteurization and sterilization regimes that causes damage to bacterial proteins; (vi) Freeze-thaw cycles that disturb bacterial cells and cell aggregates through strong fluctuations in temperature; and (vii) starvation stress that occurs in the environment following nutrient deprivation.

Several population genetic studies have reported that, among E. coli O157:H7 strains, some bacterial genotypes exhibit significant variation in their relative frequency of isolation between the human and bovine host; in which a significantly lessdiverse group of E. coli O157:H7 genotypes has been recovered from human clinical specimens. This was originally attributed to the possibility that only a subset of genotypes was involved in human infection and this subset would represent a minor subpopulation of the strains of bovine origin (Franz et al., 2012). This divergence is suggested to be crucial and advices a close tracking of clinical-biased strains that are more closely associated with human disease, likely have a high risk for virulence and transmission potential to humans (Franz et al., 2012; Mellor et al., 2013; Elhadidy et al., 2015a) and/or are more correlated with severe clinical symptoms (Manning et al., 2008; Elhadidy et al., 2015b). Whether the intermediate habitat (in particular the farm to fork chain) of E. coli O157:H7 plays a significant role in the shaping of clinical populations remains obscure. The lineagespecific polymorphism assay (LSPA-6) uses six genetic markers identified by octamer-based genome scanning to differentiate E. coli O157:H7 into three lineages (LI, LI/II, and LII) that exhibit phenotypic differences based on pathogenic potential and host specificity. Lineages I and II are recovered mainly from humans and bovines, respectively, while the intermediate lineage I/II has been less characterized regarding its host distribution (Ziebell et al., 2008; Zhang et al., 2010; Lee et al., 2011). Shiga toxin bacteriophage insertion (SBI) site analysis relies on amplification of the stx toxin genes (stx<sup>1</sup> and stx2) and the insertion site junctions of their encoding bacteriophages, and discriminates E. coli O157:H7 strains based on their distribution, gene expression and virulence potential (Shaikh and Tarr, 2003; Besser et al., 2007). Clade typing is a typing method that uses 32 single nucleotide polymorphisms (SNPs) that can distinguish E. coli O157:H7 strains into nine distinct evolutionary clades, with clade 8 strains exhibiting more virulence and a closer association with clinical illness than strains of other clades (Manning et al., 2008). Moreover, tir (A255T) polymorphisms, Shiga toxin 2 subtype, and stx2−-specific Q antiterminator gene allele (located upstream of the prophage stx<sup>2</sup> region and responsible for expression levels of the stx<sup>2</sup> gene) have been also suggested as clinically relevant genetic markers among E. coli O157:H7 (Ahmad and Zurek, 2006; Bono et al., 2007; Persson et al., 2007).

The aim of this study was to evaluate whether the variations in transmission and/or virulence potential among E. coli O157:H7 strains might be attributed, at least in part, to variations in their resistance to adverse stress conditions encountered throughout the food chain through selecting well-adapted strains belonging to clinically relevant genotypes. In order to investigate this hypothesis, the behavior of 33 E. coli O157:H7 strains, isolated from a range of meat and dairy samples in Egypt and belonging to different specific genotypes, was monitored following their exposure to cold, oxidative, osmotic, acid, heat, freeze-thaw, and starvation stresses, and the relationships between stress resistance patterns and genotypes were further analyzed by univariate and multivariate methods.

# MATERIALS AND METHODS

# Bacterial Strains and Genetic Characterization

A total of 33 E. coli O157:H7 strains isolated during a previous study (Elhadidy and Elkhatib, 2015) from various food sources, including different meat and dairy samples, were used in this study. The meat samples included retail minced beef, hamburgers, and fresh beef samples. The dairy samples included raw milk and raw milk cheese samples. All bacterial strains were characterized using different genotyping methods as described in the previous study by Elhadidy and Elkhatib (2015). Genetic characterization of the strains included: lineagespecific polymorphism assay (LSPA-6) (Ziebell et al., 2008; Zhang

et al., 2010), Shiga-toxin-encoding bacteriophage insertion site assay (SBI) (Shaikh and Tarr, 2003), clade typing (Manning et al., 2008), tir (A255T) polymorphism analysis (Bono et al., 2007), stx<sup>2</sup> subtyping (stx2<sup>a</sup> and stx2c) (Persson et al., 2007), and antiterminator Q gene allele (Q<sup>933</sup> and Q21) analysis (Ahmad and Zurek, 2006).

# Bacterial Culture Conditions and Stress Treatments

Bacterial strains were stored at −80◦C using Pro-Lab Microbank cryovials (Pro-Lab, Richmond Hill, ON, Canada) according to the manufacturer's instructions. Strains were cultured from frozen stocks onto Tryptone Soy Agar (TSA; Oxoid Ltd, UK) and incubated aerobically at 37◦C for 24 h for their recovery before use in stress challenge experiments. All strains were subjected to seven different stress conditions commonly encountered during food processing and storage. One isolated colony from each tested E. coli O157:H7 strain was aseptically inoculated into 10 ml of TSB and incubated for 16 h at 37◦C. This suspension was aseptically inoculated to 40 ml of sterile TSB (1:5 dilution), followed by incubation at 37◦C for 24 h, which results in a stationary phase culture with approximately 10<sup>9</sup> cells/ml, as described by Alvarez-Ordóñez et al. (2013). Actual starting bacterial numbers were confirmed by plating serial dilutions on TSA before applying the stressor. For all stress conditions, each stationary-phase culture was centrifuged at 7,500 × g for 7 min. The supernatant liquid was removed and cellular pellets were resuspended in 5 ml of sterile TSB (except for starvation stress in which the pellets were suspended in 0.85% saline). These bacterial suspensions were subjected to different stress conditions including: chilling to 5◦C (cold stress) for up to 7 days, heating in a water bath at 55◦C (heat stress) for up to 6 h, exposure to TSB with 1 mM H2O<sup>2</sup> prewarmed to 37◦C (oxidative stress) for up to 6 h, exposure to TSB with 5% (wt/vol) NaCl prewarmed at 37◦C (osmotic stress) for up to 7 days, and exposure to TSB at pH 2.5 (adjusted with hydrochloric acid) prewarmed to 37◦C (acid stress) for up to 6 h. For freeze-thaw stress, bacterial suspensions were subjected to seven cycles of freezing at −20◦C for 22.5 h followed by thawing at 37◦C for 1.5 h (each cycle lasted for 1 and 7 days were required to complete seven cycles). Finally, for starvation stress, bacterial pellets were resuspended in saline solution (0.85% NaCl, pH 6.6) and incubated at 37◦C for up to 7 days. During each stress treatment, 0.1 ml of bacterial suspension was removed at set time intervals, serially 10-fold diluted in 0.1% peptone water and plated in triplicate on TSA to estimate the mean number of CFU/ml. The set time intervals for heat, oxidative, and acid stress were 0, 2, 4, and 6 h post-treatment. For freeze-thaw, cold, osmotic, and starvation stresses, the set time intervals were days 0, 2, 4, and 7. Three independent trials were carried out for each strain and stress condition.

# Determination of Stress Resistance Parameters

D-values, defined as the time required (in h for heat, oxidative, and acid stress or in days for freeze-thaw, cold, osmotic, and starvation stresses) for inactivation of 90% of the bacterial population, were determined by plotting the log<sup>10</sup> number of survivors against time. The line that best fitted survivor plots was determined by linear regression (GraphPad Prism version 4.00 for Windows. GraphPad Software. San Diego, CA, USA) and the negative reciprocal of the slope was used for D-value determinations.

# Assignement of Relative Stress Resistance Units

An arbitrary unit was assigned for each E. coli O157:H7 strain under each stress condition by calculating the ratio between the D-value estimated for the particular strain subjected to a given stress and the maximum D-value observed among the collection of E. coli O157:H7 strains for that particular stress condition. Arbitrary units calculated for each stress condition were added to calculate a relative stress resistance value for each bacterial strain (ranging from 0 to 7), which was considered indicative of the general environmental robustness of the strain.

# Univariate Analyses of STEC O157:H7 Stress Resistance

D-values of strains from different genotypes [resulting form analysis of lineage-specific polymorphisms, Shiga-toxinencoding bacteriophage insertion sites, clade typing, tir (A255T) polymorphisms, Shiga toxin 2 subtype, and antiterminator Q gene allele] were compared using Student's t-test and the Statistica for Windows v 7.0. program (Statsoft, Inc., Tulsa, OK, USA).

# Multivariate Analysis of E. coli O157:H7 Stress Resistance

To gain insight into patterns of stress resistance, D-values from all of the seven stress resistance assays were analyzed by principal component analysis (PCA) as described elsewhere (Lee et al., 2012). PCA is useful for identifying a trend in a multivariate data set or a correlation between variables. The transformation of PCA consolidates the information of a data set into a few new variables or principal components (PCs). All of the analyses (calculation of coefficients, joining of variables, canonical analysis, and graphical display) were carried out with the Statistica for Windows, v 7.0. program.

# RESULTS

# Inactivation of E. coli O157:H7 Strains by Different Stressors

The resistance to seven stressors (cold, oxidative, osmotic, acid, heat, freeze-thaw, and starvation) of 33 STEC O157:H7 strains was monitored over time, what allowed the study of the inactivation kinetics. Survival curves obtained for all stress conditions fitted properly into a first order kinetic. The goodness of fit was determined both by visual inspection and R 2 value which ranged from 0.83 to 0.99 (in most cases R 2 values of over

0.95 were observed; data not shown). D-values calculated are shown in **Table 1** as mean values and standard deviation.

# Univariate Analysis of Differences in Resistance to Different Stressors among E. coli O157:H7 Genotypes

Associations between estimated D-values and singular STEC O157:H7 genotypes were evaluated individually for each stress condition by using the Student's t-test. Mean D-values obtained for each particular genotype and stress condition are represented in **Figure 1**, while significant differences (P < 0.05) in stress resistance among different genotypes are shown in **Figure 2**. Statistically significant differences among genotypes shown in **Figure 2** refer in all cases to genotypes for which at least five representative strains were available in the strain collection. Lineage I/II strains, with mean D-values of 9.62 h, 2.22 and 2.33 days, respectively, were significantly more resistant to acid, cold, and starvation stressors than lineage II strains, with mean D-values of 4.96 h, 2.09 and 1.81 days. On the other hand they were significantly less resistant to oxidative stress, with a mean D-value of 11.62 h (vs. 18.12 h for lineage II strains). Only one lineage I strain was included in the study and showed D-values of 6.69 h, 10.77 h, 2.18 days and 2.16 days for acid, oxidative, cold and starvation treatments, respectively. Q<sup>933</sup> strains were significantly more resistant to acid and starvation stress (D-values of 10.41 h and 2.35 days) than Q<sup>21</sup> strains (D-values of 6.81 h and 2.09 days). Strains carrying stx2<sup>a</sup> subtype were significantly more resistant to heat stress than strains carrying both stx2<sup>a</sup> and stx2<sup>c</sup> subtypes (mean D-value of 13.29 h vs. 9.32 h). SBI genotype 1 strains were significantly more resistant to acid, heat, cold, and starvation stress than SBI genotype 5 strains (SBI genotype 6 strains were also more resistant than SBI genotype 5 strains for starvation stress). On the contrary, both SBI genotypes 1 and

TABLE 1 | D-values for each strain and stress condition expressed in hours (acid, heat, and oxidative stress) and in days (freeze/thaw, cold, osmotic, and starvation stress).


\*h, in hours; d, in days.

insertion sites; (E) tir (A255T) polymorphisms; (F) clade typing.

6 were significantly more sensitive to oxidative stress than SBI genotype 5 strains. Regarding tir (A255T) polymorphisms, tir (255T) encoding strains, with mean D-values of 9.44 h, 13.03 h, 2.22 days and 2.33 days, respectively, were significantly more resistant to acid, heat, cold, and starvation stressors than tir (255A) encoding strains, with mean D-values of 6.14 h, 8.42 h, 2.11 days and 1.92 days. On the other hand they were significantly less resistant to oxidative stress, with a mean D-value of 11.18 h [vs. 17.76 h for tir (255A) encoding strains]. The same trends described fortir (A255T) polymorphisms were observed for clade typing, with clade 8 strains being significantly more resistant to acid, heat, cold, and starvation stresses and more sensitive to oxidative stress than non-clade 8 strains. For osmotic and freezethaw stresses, no genotype-associated statistically significant differences in stress resistance patterns were found.

# Multivariate Analysis of E. coli O157:H7 Stress Resistance

In order to carry out a global analysis of STEC O157:H7 stress resistance, in first place, a relative stress resistance value that was considered indicative of the general robustness of the strain was calculated for each strain (**Table 2**). The fourteen more robust strains, according to their relative stress resistance value, were of lineage I/II and clade 8. On the other hand the five more stress sensitive strains were non-clade 8 strains. In second place, resistance parameters (D-values) were included in a multivariate PCA. The first and second PCs accounted for 33.97 and 21.13%, respectively, of the variance within the multivariate data set (**Figure 3**). Since most of the variation (55.1%) was contained in the first and second PC, we focused on these PCs. PCA allowed the identification of genotype-associated patterns of stress resistance. Thus, it clustered together all strains of lineage II [all of them Q21, tir (255A), non-clade 8 strains], that therefore had similar stress resistance patterns, and differentiated them from lineage I/II and lineage I strains. The factor loadings on the first PC (Supplementary Table S1) were negative values for acid, heat, cold, osmotic, and starvation stresses. Higher negative scores on the first PC were associated with greater resistance to these five stresses, and positive scores on the first PC were associated with susceptibility to them. Thus, resistance to these stresses was to some extent positively co-related and associated with sensitivity to oxidative and freeze-thaw stresses, which showed positive values for the factor loadings on the first PC.

Pairwise comparisons of PC scores in the first PC and STEC O157:H7 genotypes evidenced that strains of lineages I and I/II, Q933, stx2a, SBI genotypes 1, 3, 6, and 21, tir (255T), and clade 8, which showed negative PC scores in the first PC, were in general more resistant to acid, heat, cold, and starvation stresses, and more sensitive to oxidative stress than their counterparts (Supplementary Table S2).

# DISCUSSION

The high virulence of STEC has stimulated interest in their determinants of survival in food and the environment. STEC encounter various environmental stresses in their ecological niches, in the environment of food-processing industries, on foods and in the host after their ingestion. These include fluctuations of pH, osmolarity, temperature, and oxygen availability, among others. In addition, industrial food preservation regimes commonly rely upon imposing extreme physical and chemical stresses with the aim to inactivate or limit the growth of pathogenic bacteria. Thus, a variety of preservation technologies (including thermal processing) impose a challenge to bacterial cells and can determine their fate along the food chain (Alvarez-Ordóñez et al., 2015).

In this study, the relationship between specific genotypes of STEC O157:H7 and stress resistance was assessed by using univariate and multivariate analyses. Different patterns of stress resistance in STEC O157:H7 tested isolates were observed after exposure to multiple food-related stressors: cold, oxidative, osmotic, acid, heat, freeze-thaw, and starvation stresses. When stress resistance parameters (D-values) were compared among STEC O157:H7 genotypes, significant differences were observed for acid, heat, oxidative, cold, and starvation stresses. Interestingly, lineage I/II, tir (255T), and clade 8 encoding strains, which showed a significantly higher resistance to acid, cold, and starvation stresses, have been shown in different studies to be more virulent and more frequently associated with human infection than their counterparts (Bono et al., 2007; Manning et al., 2008; Elhadidy et al., 2015b). On the contrary, some of these genotypes more commonly associated


Strains are ordered by their relative stress resistance.

with human disease were significantly less resistant to oxidative stress. Although there is a controversy on whether bacterial responses to environmental stresses can modulate virulence (Archer, 1996; Gahan and Hill, 1999), it is documented that some genetic systems involved in specific stress responses are the same as those associated with virulence during infection. Indeed, some stress response regulators (e.g., rpoS in Gram-negatives, σ B in Gram-positives) are known to be involved in the regulation of pathogenicity traits of certain food-borne pathogens, which suggests that stress responses may be an important factor in potentiating the expression of particular virulence factors in vivo (Hengge-Aronis, 2000). Thus, the ability to deal successfully with environmental stresses would indirectly help bacterial virulence (Alvarez-Ordóñez et al., 2015). Genes needed to withstand stress conditions in the environment would help bacteria to access the gastrointestinal interface and eventually provoke virulence in the host, with both stress-related and virulence-related genes being expressed in response to surrounding signals.

Several research groups have dedicated efforts in the last decade to evaluate the heterogeneity in STEC stress resistance, using both laboratory domesticated strains and field isolates (Saridakis et al., 2004; Bhagwat et al., 2006; Malone et al., 2007; Vanaja et al., 2009; Lee et al., 2012; Alvarez-Ordóñez et al., 2013; Elhadidy and Mohammed, 2013). Information available in the literature so far suggests that the ability to survive in stressful conditions varies substantially among isolates within a given genotype, whereas there is controversy on the fact of whether some genotypes are better equipped to face the challenge of a changing environment. In a study that was undertaken to compare resistance to different processing treatments (highpressure processing, heat, ultraviolet, and gamma radiation) among different E. coli O157:H7 isolates encoding different antiterminator Q gene alleles present upstream of the Shiga toxin gene, Malone et al. (2007) showed that isolates encoding the Q<sup>933</sup> allele were more sensitive to all processing treatments than were isolates encoding the Q<sup>21</sup> allele. Moreover, the stx-negative

 Lineage I/II, Q933+Q21, stx2a+c, SBI 1, tir(255T), non-clade 8 strains; Lineage I/II, Q933+Q21, stx2a+c, SBI 1, tir(255T), clade 8 strains; Lineage II, Q21, stx2c, SBI 5, tir(255A), non-clade 8 strains; Lineage II, Q21, stx2a+c, SBI 5, tir(255A), non-clade 8 strains; Lineage II, Q21, stx2a, SBI 1, tir(255A), non-clade 8 strains.

isolates were more resistant to UV and gamma radiation (but not to heat or pressure) than were isolates encoding either the Q<sup>933</sup> or Q<sup>21</sup> allelles. Comparing the resistance to different acidic conditions of E. coli O157:H7 strains belonging to lineage I and lineage II, Saridakis et al. (2004) revealed that lineage I strains were more resistant to volatile fatty acids than lineage II strains after 6 h of challenge. On the other hand, lineage II strains were more resistant to HCl treatment than lineage I strains. A recent study by Lee et al., 2012 evaluated the association between different bacterial genotypes (LSPS-6 and stx) and resistance patterns to different stressors (acid, freeze-thaw, heat, osmotic, oxidative, and starvation) and showed that lineage II strains exhibited lower resistance to heat and starvation than lineage I strains.

Microbial adaptation to a certain stress condition is often associated with enhanced protection against other subsequent stress exposures, which is referred to as "cross protection" (Johnson, 2002). Moreover, and in practice, multiple stress resistance is important in the food industry that applies multiple hurdles as a control measure to decrease pathogen survival (Jay et al., 2005). Although in several occasions it has been proposed that STEC strains more resistant to a given stress tend to be more resistant to various other types of inactivation agents (Humphrey et al., 1995; Benito et al., 1999), there are some reported exceptions to this general trend (Hauben et al., 1997; Uhlich et al., 2008). To characterize global stress resistance patterns and identify robust strains/genotypes with high resistance to multiple stresses, D-values were subjected to multivariate analysis (PCA) and were also used to calculate a relative stress resistance value. Although it was not possible to find any STEC O157:H7 strain consistently showing the highest resistance to all the different inactivation treatments tested, it was possible to rank STEC O157:H7 strains according to their robustness. PCA analysis basically confirmed the prior findings in that it clustered together strains of lineage II [all of them of tir (255A) and non-clade 8], with a multiple stress-sensitive phenotype (sensitive to acid, heat, cold, and starvation stresses) but significantly more resistant to oxidative stress, and separated them from strains of lineages I and I/II. Lineage I/II is an intermediate lineage that has been reported to be more commonly associated with human illness than lineage II and includes strains of the hyper-virulent STEC group responsible for a multistate outbreak linked to spinach (Ziebell et al., 2008; Zhang et al., 2010). Similarly, clade 8 strains have demonstrated higher virulence and association with severe disease outcome than other clades. Also, some investigators have previously suggested that tir (255T) harboring strains are more virulent for humans than tir (255A) harboring strains (Bono et al., 2007; Franz et al., 2012; Mellor et al., 2013). In fact, the tir (255T) genotype has been suggested to be the most distinctive genotype for the detection of bacterial clones with potential risk for human illness from food sources (Elhadidy et al., 2015a) and has been proposed as an attractive candidate

to be used as a surrogate marker for tracking highly severe STEC infections (Elhadidy et al., 2015b). Overall, these results suggest that the increased prevalence of E. coli O157:H7 illness observed among lineage I/II, tir (255T), and clade 8 genotypes might be related to the greater stress robustness of these genotypes, a characteristic that likely facilitates transmission of E. coli O157:H7 throughout the food chain and influences the disease causing potential of the pathogen. Consistent with our results, Lee et al. (2012) examined the behavior of various clinical and bovine strains of E. coli O157:H7 against six stressors commonly found in the food chain and the environment and concluded that some genotypes of STEC O157 associated with human illness had greater multiple stress resistance than did strains of other genotypes. Nonetheless, the study by Lee et al. (2012) was limited to two assays of genotypic characterization (LSPA6 and stx genotypes) without accounting for other genotypes used for comparison in the current study [Shiga-toxin-encoding bacteriophage insertion site assay, clade typing, tir (A255T) polymorphism, and antiterminator Q gene allele analysis].

This study focused on genotypic markers that have been suggested as being clinically relevant among E. coli O157:H7 with the aim of evaluating whether variations in transmission and/or virulence potential among E. coli O157:H7 strains could be attributed, at least in part, to variations in their resistance to stress. Nevertheless, other different genetic features, such as the presence or functionality of genes involved in the general stress response (e.g., rpoS) can also impact the phenotypic behavior of strains. Indeed, previous studies have shown that interstrain variability in stress resistance within and between genotypes/serotypes may be also influenced by other genetic traits, such as the status of the general stress response regulator RpoS (Alvarez-Ordóñez et al., 2013).

# CONCLUSION

This study assessed the resistance of 33 STEC O157:H7 strains to various food-related stresses. Our results showed that strains

# REFERENCES


more commonly associated with human disease were more resistant to food-related stresses highlighting the influence of stressors in the transmission of this human pathogen. Our findings also contribute to increase the knowledge on the resistance of this pathogen to stressors commonly encountered in the food chain, which can lead to the development of new strategies to control the risk of food-borne illness by implementing different decontamination measures in the food processing industry.

# AUTHOR CONTRIBUTIONS

Concieved and Designed the Experiments: ME, AA-O. Performed the experiments: ME, AA-O. Analyzed the data: ME, AA-O. Wrote the manuscript: ME, AA-O.

# ACKNOWLEDGMENTS

The authors are grateful to Dr. Edward Dudley at Penn State University and Dr. Victor Gannon at Laboratory of Food-Borne Zoonosis, Public Health Agency of Canada for providing control strains for the LSPS-6 assay. The authors would like to thank Dr. Marc Heyndrickx and Dr. Koen De Reu at Technology and Food Science Unit, Institute for Agricultural and Fisheries Research, Belgium for their technical help with some of the primers and reagents used in this study. Dr. ME is currently a visiting research fellow at Scientific Institute of Public Health in Brussels funded by Federal Science Policy Office (BELSPO). Dr. AA-O is a Starting Investigator Research Fellow funded by Science Foundation Ireland (SFI) under Grant Number 13/SIRG/2157.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2016.00322


and their potential association with clinical outcome in human infections. Diagn. Microbiol. Infect. Dis. 83, 198–202. doi: 10.1016/j.diagmicrobio.2015. 06.016


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

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

# High-Level Heat Resistance of Spores of *Bacillus amyloliquefaciens* and *Bacillus licheniformis* Results from the Presence of a *spoVA* Operon in a Tn*1546* Transposon

Erwin M. Berendsen1, 2, 3, Rosella A. Koning1, 3, Jos Boekhorst 1, 3, Anne de Jong1, 2 , Oscar P. Kuipers 1, 2 and Marjon H. J. Wells-Bennik 1, 3 \*

#### *Edited by:*

*Christophe Nguyen-The, Institut National de la Recherche Agronomique (INRA), France*

#### *Reviewed by:*

*Louis Coroller, University of Western Brittany, France Sandra Caroline Stringer, Institute of Food Research, UK*

*\*Correspondence:*

*Marjon H. J. Wells-Bennik marjon.wells-bennik@nizo.com*

#### *Specialty section:*

*This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology*

*Received: 01 July 2016 Accepted: 15 November 2016 Published: 02 December 2016*

#### *Citation:*

*Berendsen EM, Koning RA, Boekhorst J, de Jong A, Kuipers OP and Wells-Bennik MHJ (2016) High-Level Heat Resistance of Spores of Bacillus amyloliquefaciens and Bacillus licheniformis Results from the Presence of a spoVA Operon in a Tn1546 Transposon. Front. Microbiol. 7:1912. doi: 10.3389/fmicb.2016.01912*

*<sup>1</sup> Top Institute Food and Nutrition, Wageningen, Netherlands, <sup>2</sup> Laboratory of Molecular Genetics, University of Groningen, Groningen, Netherlands, <sup>3</sup> NIZO Food Research, Ede, Netherlands*

Bacterial endospore formers can produce spores that are resistant to many food processing conditions, including heat. Some spores may survive heating processes aimed at production of commercially sterile foods. Recently, it was shown that a *spoVA* operon, designated *spoVA*2mob, present on a Tn*1546* transposon in *Bacillus subtilis,* leads to profoundly increased wet heat resistance of *B. subtilis* spores. Such Tn*1546* transposon elements including the *spoVA*2mob operon were also found in several strains of *Bacillus amyloliquefaciens* and *Bacillus licheniformis*, and these strains were shown to produce spores with significantly higher resistances to wet heat than their counterparts lacking this transposon. In this study, the locations and compositions of Tn*1546* transposons encompassing the *spoVA*2mob operons in *B. amyloliquefaciens* and *B. licheniformis* were analyzed. Introduction of these *spoVA*2mob operons into *B. subtilis* 168 (producing spores that are not highly heat resistant) rendered mutant 168 strains that produced high-level heat resistant spores, demonstrating that these elements in *B. amyloliquefaciens* and *B. licheniformis* are responsible for high level heat resistance of spores. Assessment of growth of the nine strains of each species between 5.2◦C and 57.7◦C showed some differences between strains, especially at lower temperatures, but all strains were able to grow at 57.7◦C. Strains of *B. amyloliquefaciens* and *B. licheniformis* that contain the Tn*1546* elements (and produce high-level heat resistant spores) grew at temperatures similar to those of their Tn*1546*-negative counterparts that produce low-level heat resistant spores. The findings presented in this study allow for detection of *B. amyloliquefaciens* and *B. licheniformis* strains that produce highly heat resistant spores in the food chain.

Keywords: spores, heat resistance, Tn*1546* transposon, *spoVA* operon, genome analysis

# INTRODUCTION

The ubiquitous presence of bacterial spore formers in nature can be largely attributed to their ability to produce endospores (spores) that can survive harsh environmental conditions (Nicholson et al., 2000; Setlow, 2006). Bacterial spores can enter the food chain from many different sources, for example via soil, dust, and biofilms (Heyndrickx, 2011). The intrinsic resistance properties of spores may result in survival during food processing, in which heating is one of the most commonly applied treatments to reduce bacterial loads. Such treatments put selective pressure on the microflora that is present, allowing for survival of those strains that produce spores with high heat resistance (Postollec et al., 2012). Surviving spores may germinate upon exposure to certain environmental triggers, and can subsequently resume vegetative growth, potentially resulting in food pathogenicity or food spoilage, depending on the species (Scheldeman et al., 2006; Wells-Bennik et al., 2016).

Spores of mesophilic species belonging to the B. subtilis group are commonly found in various food ingredients and food products. The B. subtilis group encompasses the species B. subtilis, B. amyloliquefaciens, B. licheniformis, B. vallismortis, B. mojavensis, B. atropheus, and B. sonorensis, which are phylogenetically close, yet distinguishable (Logan and Vos, 2009). These species can generally grow between temperatures of 30–50◦C, with reported growth temperatures of B. licheniformis up to 58◦C (Warth, 1978). The spores of B. subtilis, B. amyloliquefaciens and B. licheniformis are commonly found in various food ingredients and food products including cocoa, herbs, spices, bread, soups, milk, and milk powders (te Giffel et al., 1996; Oomes et al., 2007; Lima et al., 2011; Lücking et al., 2013; Miller et al., 2015). These species are for instance well-known contaminants of raw materials used in bread making (Rosenkvist and Hansen, 1995; Sorokulova et al., 2003), and the spores can potentially even survive the bread baking process (Valerio et al., 2015). After spore survival, germination, and outgrowth, vegetative cells of B. amyloliquefaciens, B. subtilis or B. licheniformis can result in spoiled food products. B. subtilis, for instance has been reported to be present in cocoa (Lima et al., 2011) leading to spoiled chocolate drinks, B. licheniformis may be present in milk and milk powders leading to spoilage of heat treated dairy products (Gopal et al., 2015), and B. amyloliquefaciens may spoil bread, resulting in ropy bread by degradation of starch and the formation of extracellular polysaccharides (Sorokulova et al., 2003; Valerio et al., 2012, 2015). Certain strains of B. licheniformis can produce a toxin, lichenisyn A, that can cause foodborne illness (Salkinoja-Salonen et al., 1999; Nieminen et al., 2007; Logan, 2012). Lichenisyn is a non-ribosomally synthesized lipo-peptide that is heat-stable (Konz et al., 1999). Due to the pathogenic potential of strains of B. licheniformis, it is critical to control these spores in the food chain (Madslien et al., 2013).

Notable differences have been observed with respect to the spore wet heat resistance properties of strains within the B. subtilis group (Kort et al., 2005; Oomes et al., 2007; Lima et al., 2011; Berendsen et al., 2015). Following a detailed analysis of the heat resistance of spores of 14 strains belonging to the B. subtilis group, strains could be divided in two groups based on spore heat resistance (Berendsen et al., 2015). For B. subtilis strains, it was recently demonstrated that spores with high-level heat resistance contain a Tn1546 transposon, encompassing a spoVA operon that is directly responsible for this phenotype (designated spoVA2mob , where mob indicates the presence on a mobile genetic element; Berendsen et al., 2016a). In addition, we observed high-level heat resistance of spores of B. licheniformis and B. amyloliquefaciens strains that carried the Tn1546 transposon; the spores of these strains showed heat resistance levels similar to as those of spores of B. subtilis strains with a Tn1546 transposon (Berendsen et al., 2015, 2016a).

In this study, we report the presence and composition of Tn1546 transposon homologs of B. subtilis which were found in strains of B. amyloliquefaciens and strains of B. licheniformis that produced highly heat resistant spores. This was performed by genome analysis or PCR detection. The spoVA2mob operons found in B. amyloliquefaciens and B. licheniformis were introduced into B. subtilis to assess their role in spore heat resistance. In addition, the growth temperatures of all B. amyloliquefaciens and B. licheniformis strains (nine each) with or without the Tn1546 transposons were assessed.

# MATERIALS AND METHODS

# Bacterial Strains Used in This Study

The strains used in this study for genomic and phenotypic analyses are listed in **Table 1**. This included nine strains of B. amyloliquefaciens isolated from natural and food environments, nine strains of B. licheniformis from food environments, and two strains of B. subtilis. The genome sequences were available for the two B. subtilis strains, all B. amyloliquefaciens strains, and four B. licheniformis strains (Berendsen et al., 2016a,b; Krawczyk et al., 2016). For B. amyloliquefaciens strains B425 and B4140, the heat inactivation kinetics of spores were described previously (Berendsen et al., 2015) and for all other B. amyloliquefaciens and B. licheniformis strains, the heat inactivation kinetics of spores were previously determined (Berendsen et al., 2016a). Strains of B. subtilis 168 and B. subtilis B4146 were included in the genome analysis as reference strains, and the heat resistances of the spores of these strains were previously analyzed (Berendsen et al., 2015).

# Genome Analysis

Multiple sequence alignments were made for protein sequences of conserved genes that were present in single copy in all genomes using MUSCLE (Edgar, 2004). The core genome phylogenetic tree was constructed using PHYML (Guindon and Gascuel, 2003). To investigate the presence of the Tn1546 transposon and the encoded spoVA (designated spoVA2mob) operon that mediates high-level heat resistance, an orthology matrix was constructed using Ortho-MCL (Li et al., 2003) with the genomes of the four B. amyloliquefaciens strains, the nine B. licheniformis strains, B. subtilis B4146 as a strain that produces spores with high heat resistance, and B. subtilis 168 as a reference strain that produces spores with low-level heat resistance (**Supplementary Data Sheet 1**). The genomic organization of

#### TABLE 1 | Strains used in this study.


the Tn1546 transposon was visualized using the Artemis comparison tool (ACT), and using microbial genomic context viewer (MGcV) (Carver et al., 2005, 2012; Overmars et al., 2013). For the identified Tn1546 transposons, operon predictions were performed using FGENESB (http://www.softberry.com). Additionally, manual sequence comparisons and searches for pseudogenes were performed for all genes in the transposon.

Generic PCR primers were used, as described previously, for the detection of the Tn1546 encoded genes tnpA, spoVAC2mob and cls in the five strains of B. amyloliquefaciens for which no genome sequence information was available (i.e., strains 10A5, FZB42, 10A18, 101 and SB42; Berendsen et al., 2016a).

# Comparison of Heat Inactivation Kinetics of Spores

For all strains the heat resistances of their spores (produced at 37◦C) was previously determined using capillary tubes, at three different temperatures, using at least five time points, as described in Berendsen et al. (2015) and Berendsen et al. (2016a). For comparison of strain specific heat resistance of spores, the z-values (i.e., the increase in temperature required to achieve an additional log unit reduction) were calculated as described previously (van Asselt and Zwietering, 2006). Furthermore, the D-values (i.e., the time needed at that temperature to decrease the spore count 10-fold) reported previously, were used to calculate reference D-values (Dref) at the reference temperature of 110◦C with the corresponding 95% prediction intervals, as presented in **Supplementary Table 1** (van Asselt and Zwietering, 2006).

# Cloning of the *spoVA*2mob Operon

The spoVA2mob operon, including the predicted promotor region as present in strain B. licheniformis B4090 and B. amyloliquefaciens DSM7 was cloned into plasmid pDG1730, using a procedure as previously described (Berendsen et al., 2016a). The obtained constructs were transformed to B. subtilis 168 and integrated in the amyE locus as described in Berendsen et al. (2016a), to verify the role of this operon in increased heat resistance of spores. Spores were prepared for strains 168 amyE::spoVA 2mob (DSM7), 168 amyE::spoVA2mob (B4090) and 168 as described above, and the heat resistance of spores of these strains was assessed by heating at 100◦C for 1 h, followed by plating, incubation and enumeration of survivors.

# Determination of the Growth Temperature

All strains were cultured from the −80◦C glycerol stocks in brain heart infusion broth (BHI, Oxoid). For assessing growth at different temperatures, 24-well plates were filled with 1.5 ml of BHI agar and surface inoculated with 20 µl of turbid individual overnight cultures (>10<sup>7</sup> cfu mL−<sup>1</sup> , i.e., > 2 × 10<sup>5</sup> cells on the surface of the agar in each well). Growth or no growth on the agar surface was assessed on day 2, 6, 13, 23, and 30 at the following temperatures: 5.2, 11.5, 14.5, 30.5, 36.3, 45.9, 55.1, and 57.7◦C. Growth was scored visually, based on the formation of colonies or a lawn on the agar surface in the wells. The day on which growth was first observed was noted. Two biological replicates were performed, and for each biological replicate, two technical replicates were performed.

# RESULTS

# Genome Mining for the Tn*1546* Transposon

For B. subtilis strains, it has been demonstrated that the presence of a Tn1546 transposon is responsible for high-level heat resistance of the spores (Berendsen et al., 2016a). The presence of the Tn1546 transposon was assessed in nine strains of B. amyloliquefaciens and nine strains of B. licheniformis. The genome sequences of all nine strains of B. licheniformis were available and in three of these strains the Tn1546 transposon was found (namely, strains B4090, B4092, and B4094; **Figure 1**). The genome sequences of four strains of B. amyloliquefaciens were available, and the Tn1546 transposon was found in two of these strains of B. amyloliquefaciens (namely in B425 and DSM7; **Figure 1**). The predicted protein for the transposase TnpA, which

FIGURE 1 | (A) Overview of the Tn*1546* transposons found in *B. subtilis* B4146, *B. amyloliquefaciens* B425 and DSM7, and *B. licheniformis* B4090, B4092, and B4090. The predicted gene functions are indicated for the transposon of *B. subtilis* B4146: a transposase gene (*tnpA*), a resolvase gene (*res*), an operon of N-acetylmuramoyl-L-alanine amidase, *gerKA* and *ger*(X)*C* (Operon 1), an operon of a gene with unknown function and a manganese catalase (Operon2), an operon of two genes with unknown functions, *spoVAC, spoVAD, spoVAEb* and two genes with unknown functions (Operon 3, *spoVA*2mob), a fragmented *yetF* gene (Gene 4), and a cardiolipin synthase gene (Gene 5). The transposons in *B. amyloliquefaciens* are smaller, probably due to site specific recombination events, whereby operon 1 and operon 2 were lost. In *B. licheniformis* B4092 and B4094 the resolvase gene was not present, which is most likely a result of gene loss. (B) Maximum likelihood phylogenetic tree based on core genome of single genes of 15 strains of the *B. subtilis* group. The presence and copy number of the Tn*1546* transposon was indicated behind the corresponding strains. *B. subtilis* B4146, *B. licheniformis* B4090, B4092, and B4094 carry a single transposon. The *B. amyloliquefaciens* strain DSM7 carries three copies of the transposon, while for strain B425 the copy number could not be determined.

is part of the Tn1546 transposon, was found in orthologous group OG3133 for all of the strains that carry the Tn1546 transposon. Genome sequences were not available for the other five strains of B. amyloliquefaciens. PCR-based detection of the genes tnpA, spoVAC and cls (that are present on the transposon and very well conserved) using primers for these three target genes showed positive results for B. amyloliquefaciens strains B425 and DSM7 but did not reveal the transposon in the other five strains. In short, two out of nine strains of B. amyloliquefaciens and three out of nine strains of B. licheniformis contained the Tn1546 transposon.

# Heat Resistance of Spores is Related to the Presence of the Tn*1546* Transposon

The heat resistances of spores of B. amyloliquefaciens and B. licheniformis were assessed in relation to the presence or absence of the Tn1546 transposon (**Supplementary Table 1**). The heat resistance of spores was displayed as a reference decimal reduction time of spores per strain at the reference temperature of 110◦C (D110◦C-values) in relation to the presence or absence of the Tn1546 transposon (**Supplementary Table 1**).

Two strains of B. amyloliquefaciens, namely B425 and DSM7, contained the Tn1546 transposon. These strains produced spores that required significantly longer heating times (unpaired t-test p < 0.001) at 110◦C, i.e., approximately 15 times, to achieve one decimal reduction, than spores of the other seven strains without this transposon. Strains B. licheniformis B4090, B4092, and B4094 contained the Tn1546 transposon, and the spores of these strains all required longer heating times (unpaired t-test p < 0.001) (2.5 times) to reach a decimal reduction than the spores of the six B. licheniformis strains that did not possess the Tn1546 transposon.

# The *spoVA*2mob Operon Is Responsible for Increased Heat Resistance of Spores

We previously showed that the introduction of the spoVA2mob operon originating from B. subtilis strain B4067 into laboratory strain B. subtilis 168 resulted in the formation of high-level heat resistant spores by this strain (Berendsen et al., 2016a). To establish whether the spoVA2mob operons present in the Tn1546 transposons of B. licheniformis B4090 and B. amyloliquefaciens DSM7 have a functional role, these operons were also introduced individually into B. subtilis 168. The B. subtilis 168 mutants carrying the spoVA2mob genes of B. licheniformis B4090 and the spoVA2mob genes of B. amyloliquefaciens produced spores with significantly higher heat resistances than the parent strains after heating at 100◦C for 60 min (**Figure 2**). The spores of B. subtilis 168 were inactivated to a level below the detection limit, i.e., more than 8 log units reduction, indicative of a low level of heat resistance of spores. The spores of B. subtilis 168 amyE::spoVA2mob (containing the spoVA2mob operon of B. licheniformis B4090) and spores of B. subtilis 168 amyE::spoVA2mob (containing the spoVA2mob operon of B. amyloliquefaciens DSM7) showed survival of 4.0 log<sup>10</sup> unit (±0.4), and 1.7 log<sup>10</sup> unit (±0.5), respectively, indicating a high level of heat resistance of spores. The control strain B. subtilis 168 amyE::spoVA2mob producing high level heat resistant spores (containing the spoVA2mob operon of B. subtilis B4067) (Berendsen et al., 2016a), showed survival of 2.8 log<sup>10</sup> unit (±0.05).

# Detailed Analysis of the Tn*1546* Transposon

The composition of the Tn1546 transposon in B. licheniformis strains B4090, B4092 and B4094 is shown in **Figure 1A**. In these strains, the Tn1546 transposon is highly similar to the one found in B. subtilis B4146 (**Figure 1A**). The transposons found in B. subtilis and B. licheniformis consist of genes that are required for transposition, and furthermore contain three operons and two single genes. The Tn1546 transposon found in the B. amyloliquefaciens strains DSM7 and B425 was smaller than the transposon found in B. subtilis and B. licheniformis. In both B. amyloliquefaciens strains, the first two operons were absent, possibly due to a site-specific recombination event, as a recombinase gene and a hypothetical gene were present at that genomic location.

The evolutionary relatedness of the different strains and species was visualized in a maximum likelihood core genome phylogenetic tree, based on concatenated protein sequences of conserved genes present in single copy in all genomes (**Figure 1B**). The species B. amyloliquefaciens, B. licheniformis and B. subtilis clustered in separate branches of the phylogenetic tree. For B. amyloliquefaciens, the strains with the Tn1546 transposon clustered together, whereas this was not the case for B. licheniformis strains carrying the Tn1546 transposon.

The genomic locations of the Tn1546 transposons were different for B. subtilis, B. amyloliquefaciens and B. licheniformis. In B. subtilis, the transposon was found at two genomic locations, namely inserted in yitF and between yxjA and yxjB (Berendsen et al., 2016a). In B. amyloliquefaciens, three Tn1546 transposons were found in strain DSM7 at three different genomic locations, namely between a gene encoding for a fructose-1,6-bisphophatase and a hypothetical gene, between two hypothetical genes, and between a hypothetical gene and rapK. For B. amyloliquefaciens strain B425, it was not possible to determine the genomic location(s) and copy number of the Tn1546 transposon, since contig breaks were present on both sides of the Tn1546 transposon. In B. licheniformis strains B4090, B4092, and B4094, a single Tn1546 transposon was found integrated in a gene that encodes a D-alanyl-D-alanine carboxypeptidase.

Detailed analysis revealed that some genes in the Tn1546 transposon were mutated and present as pseudogenes in the transposon of some strains. The genes tnpA and tnpR in the Tn1546 transposon (which are required for active transposition) were intact and present in B. amyloliquefaciens strains B425 and DSM7 and in B. licheniformis strain B4090.

# Determination of Vegetative Growth at Different Temperatures

None of the 18 B. amyloliquefaciens and B. licheniformis strains were able to grow at 5.2◦C within 30 days. At 11.5◦C, six

out of nine B. amyloliquefaciens strain and five out of nine B. licheniformis strains showed growth. At 14.5◦C, all but one strain showed growth. All strains grew at the highest temperature tested (57.7◦C), and at all other temperatures, growth of strains was observed except for strain 101 at 55.1◦C. In the case of the nine B. amyloliquefaciens and nine B. licheniformis strains, high level heat resistance of spores due to the presence of the Tn1546 transposon was not correlated with the ability to grow at different temperatures (**Table 2**).

# DISCUSSION

B. licheniformis strains B4090, B4092, and B4094 contained a single copy of the Tn1546 transposon with a single spoVA2mob operon. The heat resistances of spores of B. licheniformis with or without this operon were significantly different, but relatively modest. For B. amyloliquefaciens, spores of strains B425 and DSM7 showed comparable high-levels of heat resistance, which were significantly higher than those of the spores of other B. amyloliquefaciens strains. Strain DSM7 contains three Tn1546 transposable elements, and it is likely that strain B425 also contains multiple copies, however this remains to be established. The number of spoVA2mob operons have previously been found to correlate with the level of heat resistance of spores in B. subtilis; strains carrying three copies produced spores with the highest level of heat resistance (Berendsen et al., 2016a). For B. subtilis, it has been shown that the Tn1546 transposon was found at different locations in the genome, all leading to highlevel heat resistance of spores (Berendsen et al., 2016a). It remains to be established whether the location of the insertion of the transposon in the genome of B. amyloliquefaciens and B. licheniformis plays a role in the level of heat resistance of spores.

The presence of intact tnpA and tnpR genes in B. amyloliquefaciens DSM7 and B425, and B. licheniformis B4090 suggests that active transposition of the Tn1546 element may be possible for these strains, although active transposition of the Tn1546 transposon is believed to require a plasmid intermediate, as has been described in Enterococcus faecium (Arthur et al., 1993). Interestingly, B. amyloliquefaciens DSM7, containing the intact tnpA and tnpR genes, contained three Tn1546 transposons. The encoded proteins required for transposition potentially allowed for internal transposition within the chromosome of strain DSM7. In B. subtilis strain B4146 and in B. licheniformis strains B4092 and B4094, the transposition genes are absent or not intact, suggesting that active transposition of the Tn1546 transposon is not likely to occur in these strains. This does


TABLE 2 | Determination of ability to grow at different temperatures for nine strains of *B. amyloliquefaciens* and nine strains of *B. licheniformis*.

*<sup>a</sup>Growth in: 2 days (*+ + ++*), 6 days (*+ + +*), 13 days (*++*), 23 days (*+*), and no growth within 30 days(*−*).*

not mean that the transposons cannot be transferred; transfer of genetic material including the Tn1546 transposon can be mediated by other transfer mechanisms, such as phage transduction, as described previously for B. subtilis (Berendsen et al., 2016a), or via the uptake of external DNA via natural competence (Kovacs et al., 2009).

The spoVA2mob operons present on the Tn1546 transposons differ from the spoVA operons (designated spoVA<sup>1</sup> ) that are encoded on the chromosomes of B. subtilis (Tovar-Rojo et al., 2002), B. amyloliquefaciens and B. licheniformis. The SpoVA proteins encoded in the spoVA<sup>1</sup> operon are required for uptake of DPA during the sporulation process, and sporulation cannot be completed upon deletion or disruption of these genes in B. subtilis 168 (Tovar-Rojo et al., 2002). In addition, the SpoVA proteins are involved in the release of Ca-DPA during the germination process (Vepachedu and Setlow, 2004, 2007), with the SpoVAC protein functioning as a mechano sensitive channel during germination (Velasquez et al., 2014). The SpoVAD protein has a binding pocket, whereby it can directly bind DPA (Li et al., 2012). Both the spoVA<sup>1</sup> and the spoVA2mob operons contain genes encoding SpoVAC, SpoVAD and SpoVAEb, while the other genes in the operons are different. Presumably the presence of the spoVA2mob encoded proteins results in the increased uptake of DPA into the spore core, as was previously found for B. subtilis (Berendsen et al., 2016a).

It is known that environmental conditions during sporulation, such as temperature, matrix and medium composition, can influence the heat resistance of spores of the B. subtilis group (Cazemier et al., 2001; Melly et al., 2002; Rose et al., 2007). To allow for a direct comparison of the resistance properties of spores of the different strains, the applied sporulation conditions of all B. licheniformis and B. amyloliquefaciens strains were the same (Berendsen et al., 2015, 2016a). Factors that are known to influence spore heat resistance include the composition of the sporulation medium, and it is known that the addition of salts (Ca2+, Mn2+, Mg2<sup>+</sup> and K+) results in higher resistances of spores to heat (Cazemier et al., 2001). Furthermore, the temperature during sporulation can influence the heat resistance of B. subtilis spores (Melly et al., 2002). In line with these findings, the heat resistances of spores of a B. licheniformis strain have been reported to be higher upon sporulation at 45◦C, with a modeled optimum at 49◦C, than at lower temperatures such as 20◦C (Baril et al., 2012). Overall, the environmental conditions during sporulation, the presence or absence of genetic elements such as the spoVA2mob operon, and the storage conditions of spores will ultimately determine the heat resistance properties of spores. It is therefore conceivable that spores produced under laboratory conditions do not necessarily reach the same levels of heat resistance as spores found in foods (van Zuijlen et al., 2010; Lima et al., 2011).

It is generally assumed for bacterial spore formers that higher optimal growth temperatures of vegetative cells correlates positively with spore heat resistance (Nicholson et al., 2000). In this study, some differences were seen between strains with respect to their abilities to grow at different temperatures, but no consistent pattern was seen for strains that contain the Tn1546 element and produce high level heat resistant spores versus the ones that do not harbor the element and produce low-level heat resistant spores. All strains were able to grow at 57.7◦C. All genes present on the Tn1546 transposon are under the control of sporulation-specific sigma factors K or G (Berendsen et al., 2016a) and are thus only expressed during late stages of sporulation. It is therefore in line with expectations that the genes on the Tn1546 transposon, which determine spore heat resistance, do not influence the ability of vegetative cells to grow at high temperatures as these genes are not expressed during vegetative growth.

# CONCLUSIONS

Variation in heat resistance of spores exists between strains of different spore forming species but also within species (Oomes et al., 2007; Orsburn et al., 2008; Lima et al., 2011; Berendsen et al., 2015). In this study, a genomic analysis revealed the presence of Tn1546 transposons in two strains of B. amyloliquefaciens and in three strains of B. licheniformis. The presence of this transposon, containing the spoVA2mob operon, correlated with high-level heat resistance of spores. Strains producing low or high level heat resistant spores showed similar temperature ranges for growth. A functional role of the spoVA2mob operon in increasing the heat resistance of spores was demonstrated by cloning these operons in B. subtilis 168, resulting in spores with high-level heat resistance. Clearly, mere identification of the species of spores in food products does not provide information on the heat resistance levels of these spores. The knowledge obtained in this study on the role of the spoVA2mob operon in spore heat resistance can be used for specific detection of strains of the B. subtilis group that produce high-level heat resistant spores. Multiple DNA based methods can be used for the detection of such genetic elements, such as whole genome sequencing and specific PCR detection, among others (Caspers et al., 2011; Alkema et al., 2016). The ability to detect certain strains of B. subtilis group that have the ability to produce high level heat resistant spores can aid control of these sporeformers in the food chain.

# REFERENCES


# AUTHOR CONTRIBUTIONS

EB and RK collected the data. EB, RK, AD, and JB analyzed the data. EB, OK, and MW wrote the manuscript.

# ACKNOWLEDGMENTS

The authors would like to thank Patrick Janssen for technical assistance. The authors have declared that no competing interests exist. The research was funded by TI Food and Nutrition, a public-private partnership on pre-competitive research in food and nutrition. The funding organization had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2016.01912/full#supplementary-material

Supplementary Datasheet 1 | Orthology matrix constructed using Ortho-MCL with the genomes of the four *B. amyloliquefaciens* strains, the nine *B. licheniformis* strains, *B. subtilis* B4146 as a strain that produces spores with high heat resistance and *B. subtilis* 168 as a reference strain that produces spores with normal heat resistance.

Supplementary Table 1 | Calculated reference *D*-value (*D*ref) at a reference temperature (Tref) of 110◦C, and *<sup>z</sup>*-value, of spores of nine strains of *B. amyloliquefaciens* and spores of nine strains of *B. licheniformis* (Berendsen et al., 2015, 2016a). In these studies, all strains were sporulated on the same media using the same incubation, namely using Nutrient Agar (Difco, 23g/L) plates supplemented with CaCl2 (1 mM), KCl (13 mM), MgSO4 (1 mM) and MnSO<sup>4</sup> (0.13 mM), with a pH of 7.0, and incubation for 7 days at 37◦C to allow for sporulation (Berendsen et al., 2015). Subsequently spores were harvested from the plates and washed three times with sterile water (5000 *g*, 10 min, 4◦C) (Berendsen et al., 2015).


mechanosensitive channel. Mol. Microbiol. 92, 813–823. doi: 10.1111/mmi. 12591


Zahler, S. A., Korman, R. Z., Thomas, C., Fink, P. S., Weiner, M. P., and Odebralski, J. M. (1987). H2, a temperate bacteriophage isolated from Bacillus amyloliquefaciens strain H. J. Gen. Microbiol. 133, 2937–2944. doi: 10.1099/ 00221287-133-10-2937

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

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

# Recovery of Heat Treated Bacillus cereus Spores Is Affected by Matrix Composition and Factors with Putative Functions in Damage Repair

Alicja K. Warda1,2,3, Marcel H. Tempelaars<sup>2</sup> , Tjakko Abee1,2 \* and Masja N. Nierop Groot1,3

<sup>1</sup> TI Food and Nutrition, Wageningen, Netherlands, <sup>2</sup> Laboratory of Food Microbiology, Wageningen University, Wageningen, Netherlands, <sup>3</sup> Wageningen UR Food & Biobased Research, Wageningen, Netherlands

The ability of spores to recover and grow out after food processing is affected by cellular factors and by the outgrowth conditions. In the current communication we studied the recovery and outgrowth of individually sorted spores in BHI and rice broth media and on agar plates using flow cytometry. We show that recovery of wet heat treated Bacillus cereus ATCC 14579 spores is affected by matrix composition with highest recovery in BHI broth or on rice agar plates, compared to BHI agar plates and rice broth. Data show that not only media composition but also its liquid or solid state affect the recovery of heat treated spores. To determine the impact of factors with putative roles in recovery of heat treated spores, specific genes previously shown to be highly expressed in outgrowing heat-treated spores were selected for mutant construction. Spores of nine B. cereus ATCC 14579 deletion mutants were obtained and their recovery from wet heat treatment was evaluated using BHI and rice broth and agar plates. Deletion mutant spores showed different capacity to recover from heat treatment compared to wild type with the most pronounced effect for a mutant lacking BC5242, a gene encoding a membrane protein with C2C2 zinc finger which resulted in over 95% reduction in recovery compared to the wild type in BHI broth. Notably, similar relative performance of wild type and mutants was observed using the other recovery conditions. We obtained insights on the impact of matrix composition and state on recovery of individually sorted heat treated spores and identified cellular factors with putative roles in this process. These results may provide leads for future developments in design of more efficient combined preservation treatments.

Keywords: germination and outgrowth, thermal treatment, food matrix, spore former, rice, recovery of spores

# INTRODUCTION

An increased demand for food with improved freshness, sensorial, and nutritional values has directed food processing toward the use of milder heat treatments that require secondary mild preservation hurdles to assure stability and safety of the products (Pasha et al., 2014). As a result, these products are challenged by resistant microbial spores, that survive heat and other preservation hurdles used in food processing (Postollec et al., 2012; Stecchini et al., 2013; Checinska et al., 2015). Reduction of the heat treatment intensity may lead to subpopulations of spores that are

#### Edited by:

Avelino Alvarez-Ordóñez, Teagasc Food Research Centre, Ireland

#### Reviewed by:

Elrike Frenzel, University of Groningen, Netherlands Frédéric Carlin, National Institute for Agricultural Research, France

> \*Correspondence: Tjakko Abee tjakko.abee@wur.nl

#### Specialty section:

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

Received: 12 May 2016 Accepted: 30 June 2016 Published: 18 July 2016

#### Citation:

Warda AK, Tempelaars MH, Abee T and Nierop Groot MN (2016) Recovery of Heat Treated Bacillus cereus Spores Is Affected by Matrix Composition and Factors with Putative Functions in Damage Repair. Front. Microbiol. 7:1096. doi: 10.3389/fmicb.2016.01096

sublethally damaged rather than inactivated resulting in increased heterogeneity in the population (Cazemier et al., 2001; Warda et al., 2015). Damaged spores retain the capacity to germinate, repair, and eventually grow out leading to spoilage and safety issues (Smelt et al., 2008; Mafart et al., 2010; Samapundo et al., 2014; Warda et al., 2015). Heterogeneity in spore populations can originate from differences in sensitivity of individual spores to inactivating treatments (Stringer et al., 2011) and/or from differences in repair capacity of individual damaged spores. In addition, the presence of superdormant spores may further increase heterogeneity (Ghosh and Setlow, 2010), and this conceivably results in less accurate prediction of spore outgrowth behavior.

Wet heat treatment is a common practice in food processing intended to reduce the microbial load of food products. Thermal pasteurization processes aim for inactivation of vegetative cells but are insufficient to kill spores. Sterilization processes aim for spore inactivation but may result in spore damage when target process conditions are not reached or when products contain highly heat resistant spores. The exact mechanism of wet heat killing of the spores and concomitant wet heat damage are not yet fully understood. Wet heat resistance of spores, mainly investigated in Bacillus subtilis, is determined by a number of factors including the spore structural components [small acid-soluble proteins (SASP), dipicolinic acid (DPA), metal ions, low core water content] but also the sporulation conditions (temperature, liquid or solid state of medium) affect its resistance (Setlow, 2014; Berendsen et al., 2016; Wells-Bennik et al., 2016). Wet heat treatment is thought to kill spores by damaging one or more key spore proteins, however, the identity of those proteins remains to be determined (Setlow, 2014). Analysis of single wet heat treated spores of Clostridium botulinum (Stringer et al., 2011) and Bacillus species (Wang et al., 2011) revealed a delayed initiation of germination and/or reduced rate of germination, but also the subsequent outgrowth was delayed indicating not only damage to the germination system but also to other spore components affecting outgrowth. The time required for germination and outgrowth of spores was shown to correlate with the wet heat treatment intensity (Aguirre et al., 2012). Heterogeneity in germination and outgrowth of surviving C. botulinum and Bacillus cereus spore populations is more pronounced in the presence of a secondary mild stress factor such as low pH without and with sorbic acid, and increasing levels of salt (Stringer et al., 2011; van Melis et al., 2014; Warda et al., 2015) or the natural components of food media (Warda et al., 2015). In general, damaged spores were shown to be more sensitive to secondary stresses including sodium chloride, pH, sorbic acid compared to undamaged spores (Feeherry et al., 1987; Faille et al., 1997; Cazemier et al., 2001; Cortezzo et al., 2004; Samapundo et al., 2014; van Melis et al., 2014; Warda et al., 2015, 2016). Some studies suggest a pre-plating recovery step in optimal (perhaps strain and treatment specific) conditions to allow recovery of injured cells (Wu, 2008) or spores (Rodriguez-Palacios and Lejeune, 2011). For example, a 7 h incubation step of heat treated Clostridium difficile spores in BHI broth prior to plating resulted in increased recovery of ethanol resistant

In this study we focus on B. cereus, a spore former of concern in processed foods. Its spores are widely present in the environment and are common contaminants in the food chain. B. cereus has been associated with food spoilage (Andersson et al., 1995) and food-borne disease (Stenfors Arnesen et al., 2008). The vegetative cells of B. cereus can cause disease either by secretion of enterotoxins in the small intestine, causing the diarrheic syndrome or by the production of a heatstable toxin (cereulide) in food before ingestion resulting in an emetic syndrome. B. cereus associated diseases are usually mild and self-limiting but in rare instances they can lead to fatal outcomes (Ehling-Schulz et al., 2004; Dierick et al., 2005; Granum, 2005; Schoeni and Wong, 2005; Stenfors Arnesen et al., 2008).

Using a transcriptome approach, we previously identified 21 genes putatively involved in heat damage repair in B. cereus. For one of these candidate genes, cdnL (now referred as cdnL1), a role in spore damage repair was further confirmed using a targeted deletion mutant (Warda et al., 2016). Here we report on behavior of eight newly and one previously (Warda et al., 2016) constructed mutant to assess respective putative roles in recovery efficiency of heat treated B. cereus spores.

To this end, B. cereus ATCC 14579 wild type and its mutant derivative spores were exposed to a wet heat treatment resulting in over 95% of damaged spores in the surviving fractions. The recovery and outgrowth of spores was followed using flow cytometry (FCM) in combination with single spore sorting and a Most Probable Number (MPN) approach. To quantify the effect of matrix conditions on recovery capacity of wild type and mutant spores BHI and rice media, both in solid and liquid form were included. This approach allows for identification of candidate genes that may contribute to recovery capacity of heat treated B. cereus spores.

# MATERIALS AND METHODS

# Strains and Sporulation Conditions

Bacillus cereus ATCC 14579 obtained from the American Type Culture Collection (ATCC), and its mutant derivatives used in this study (**Table 1**) were cultured in Bacto Brain Hart Infusion broth (BHI, Beckton Dickinson) at 30◦C with aeration at 200 rpm. A nutrient-rich, chemically defined sporulation medium (MSM medium) described previously (Garcia et al., 2010) was used to obtain spores. Sporulation and spore handling were performed as described previously (Warda et al., 2015), briefly one ml of an overnight-grown pre-culture was used to inoculate 100 ml of MSM media in 500 ml flasks and incubated at 30◦C with aeration at 200 rpm. Sporulation was monitored by phase contrast microscopy until over 99% of the spores were released from the mother cell (typically after 2– 3 days). Released spores were harvested by centrifugation at 5,000 rpm at 4◦C (5804R, Eppendorf, Germany) for 15 min



<sup>a</sup>no sporulation, <sup>b</sup>poor spore quality, Cm<sup>r</sup> chloramphenicol resistance.

and washed with chilled phosphate buffer (100 mM, pH 7.4) containing 0.1% Tween80 to prevent spore clumping. Spores were washed twice a day for 2 weeks with a phosphate buffer that was gradually decreased in Tween80 concentration until a final concentration of 0.01% (further referred as suspension buffer). Spores free of vegetative cells, debris and mother cells residues were stored at 4◦C and used within 6 months. A single spore crop per strain was used for all the experiments.

# Construction of Deletion Mutants

Deletion mutants (**Table 1**) were constructed using the temperature-sensitive suicide plasmid pAUL-A (Chakraborty et al., 1992). Flanking regions of the individual genes were amplified using KAPA HiFi Hotstart ReadyMix (KAPA Biosystems, USA) and the primers UP\_enzyme\_F/UP\_NotI\_R and DOWN\_NotI\_F/DOWN\_enzyme\_R (Supplementary Table S1) for upstream and downstream flanking regions, respectively. The resulting fragments were fused in frame via a NotI digestion site introduced with the indicated primers. The resulting plasmid was transferred via electroporation (400 , 25 µF, 1.2 kV, 0.2 cm Gene Pulser Cuvette: BIORAD) in B. cereus ATCC 14579 cells, and plated on BHI agar at 30◦C with 10 µg/ml erythromycin (E10) to select for the desired transformants. Two erythromycin resistant colonies were selected and grown overnight in BHI at 30◦C in the presence of E10. The resulting culture was diluted (1:200) in fresh LB with E10 and grown o/n at 42◦C to select for plasmid integration. Selected strains resulting from a single cross-over integration event were grown overnight in BHI at 30◦C to induce double crossover events and subsequently plated and grown at 30◦C. Resulting colonies were replica plated on BHI with and without E10 and incubated at 37◦C. Colonies sensitive to E10 were selected. PCR analyses (using primers UPFlank\_F, DOWNFlank\_R, checkINTERNAL\_R, check\_F, and check\_R) (Supplementary Table S1) and DNA sequencing of erythromycin sensitive colonies confirmed the correct internal in-frame deletion of the gene and lack of other mutations in the targeted region.

A double deletion mutant (1cdnL1/1cdnL2) was obtained as described above with the exception that the cdnL2 knock out plasmid was transformed into a B. cereus 1cdnL1 (BC4714) mutant strain constructed previously (Warda et al., 2016) and 5 µg/ml chloramphenicol was included as selective pressure preventing excision of the chloramphenicol resistance cassette that disrupted the cdnL1.

# Heat Treatment

One hundred micro liter of spore suspension containing approximately 10<sup>8</sup> spores/ml in suspension buffer was transferred, in duplicate, to thin-walled PCR tubes (VWR, The Netherlands). The PCR tubes were kept for 1 min at 4◦C followed by a step at 95◦C for 45 s and finally cooled for 1 min at 4◦C in a thermal cycler (Veriti, Applied Biosystems). The duplicates were pooled and 100 µl of this pooled fraction was used as sample for spore sorting experiments. The same pooled fraction was diluted decimally in suspension buffer and 50 µl samples were used for spore enumeration on BHI plates (in duplicate) and incubated at 30◦C up to 3 days with daily enumeration of resulting colonies. For each strain, from the same spore preparation, at least four independent heating experiments were performed.

### Quantification of Spore Damage

To evaluate the degree of spore damage, the method previously reported by Warda et al. (2016) was used. Briefly, 50 µl of decimally diluted heat treated samples were enumerated in duplicate on BHI plates and BHI plates supplemented with 1.5 and 5.5% salt following incubation at 30◦C. To evaluate possible delay in colony formation, colonies were counted after 1, 2, and 7 days (further extension did not affect colony counts). Obtained colony forming units (CFUs) were used to calculate the total damage and fractions of mildly and severely damaged spores as described

previously (Warda et al., 2016) according to the following formulas:

% Total damage = (Number of cfu's BHI) – (Number of cfu's BHI 5.5% NaCl) (Number of cfu's BHI) <sup>×</sup> <sup>100</sup> % Mild damage = (Number of cfu's BHI 1.5% NaCl) – (Number of cfu's BHI 5.5% NaCl) (Number of cfu's BHI) <sup>×</sup> <sup>100</sup> % Severe damage =

(Number of cfu's BHI) – (Number of cfu's BHI 1.5% NaCl) (Number of cfu's BHI) <sup>×</sup> <sup>100</sup>

# Flow Cytometry and Cell/Spore Sorting

Flow cytometry was performed with a FACSAriaIII cell sorter (BD Biosciences) using a fiber launched solid state air-cooled laser operating at 488 nm. Only forward scatter (FCS) and side scatter (SSC) functionality was used. The machine was calibrated using standard Cytometer Setup & Tracking beads and Accudrop beads (BD Biosciences). All parameters were measured using logarithmic amplification. During the procedures a 85 micron nozzle (drop driving frequency was ∼45 kHz/s) was used with flow rate one and during sorting a maximum event rate of 2000 events/s was used. Cells and spores were discriminated from electronic noise using both SSC and FSC. Sorting criteria and gating strategy were based on FSC and SSC populations (data collection equals 50.000 events) excluding remaining doublets. In order to achieve high purity and recovery, the "Single Cell" precision mode (Purity mask 32 and Phase mask 16) was used for sorting. Cells or spores were sorted on solid and in liquid media.

### Cell Sorting

Five micro liter of an overnight grown culture was diluted in 3 ml of HEPES buffer and loaded into the flow cytometer. Individual vegetative cells of B. cereus ATCC 14579 and the mutant derivatives were spotted in duplicate on a single BHI and rice agar plate (according to the scheme in Supplementary Figure S1C) and incubated at 30◦C up to 3 days to confirm that growth was not affected in the deletion mutants.

### Spore Sorting

One hundred micro liter of unheated or heat treated spore suspension (containing non-damaged, damaged and dead spores) was diluted in 1.5 ml of HEPES buffer (pH 7.4) in 5 ml polystyrene falcon tube (BD, USA) and loaded into the flow cytometer. For heat treated spores, a series of 1, 10, and 100 individual spores were sorted either into wells of 384-well plates (Greiner Bio-One, USA) containing 50 µl of BHI or rice broth or on one of the 52 available locations on standard BHI or rice agar plates. The resulting growth data representing three consecutive decimal dilutions were used as input for the MPN quantification method (Oblinger and Koburger, 1975; Jarvis, 2012). For heat treated spores, for each sorting series of 1, 10, or 100 spores approximately 754 replicates were performed for liquid media and at least 520 replicates for solid media (Supplementary Table S2). A single replicate is defined as one well or location on agar plate to which either 1, 10, or 100 spores were sorted. For untreated spores, only single spores were spotted on 188 and 104 locations (Supplementary Table S2) for liquid or solid media, respectively. The resulting plates were incubated at 30◦C up to 3 days with daily visual scoring for growth, i.e., colony formation on solid media or appearance of turbidity for liquid media. Wells that were positive for turbidity ranged from OD<sup>600</sup> 0.2 to 0.3 for rice media (OD<sup>600</sup> of fresh media 0.16), and in case of BHI values from OD<sup>600</sup> 0.2 to 0.8 (OD<sup>600</sup> of fresh media 0.1). The MPN values and their upper and lower limits were calculated using MPN Calculator<sup>1</sup> .

# Model Food Media used in This Study

A rice based medium was prepared according to the method reported previously (Warda et al., 2015) by boiling ready-to-cook pouches filled with 125 g rice produced by the manufacturer (Lassie B.V, The Netherlands) in demineralized water (5:32 w/v) for 45 min. The rice bags were removed and the remaining liquid was allowed to cool down. The method was modified by addition of a centrifugation step [(AVANTI J-25, Beckman Coulter, USA) for 5 min at 16,000 rpm at 22◦C] and filtering of the resulting supernatant (Filter paper, Whatman, England) to remove the big particles and improve the clarity of the solution. Finally, the suspension was pooled and autoclaved. Sterile rice broth was stored in the dark until use. For preparation of rice agar plates, 1.5% (w/v) Bacteriological Agar was added prior to a second autoclaving step. The final pH of rice broth was 6.7, while the pH of rice agar plates was 7.

# RESULTS

# Impact of Matrix on the Growth of B. cereus Spores

The impact of the liquid and solid media composition on the growth of B. cereus spores was evaluated using FCM in combination with single spore sorting. The single untreated B. cereus spores were sorted into four different media namely BHI broth, rice broth, BHI agar plates and rice agar plates. Besides BHI, a rice media was selected as this food matrix was shown previously to support growth from B. cereus spores on agar plates (Warda et al., 2015) and on Anopore strips (Warda et al., 2015) that resemble growth in broth (den Besten et al., 2007). All four media allowed outgrowth of 94 up to 99% of the sorted spores within 3 days (**Figure 1**; Supplementary Table S2). In rice broth, growth was delayed compared to BHI broth, which was not observed for the corresponding agar media (**Figure 1**). In general, smaller colonies were formed on rice agar plates compared to BHI agar plates for both outgrowing untreated spores (data not shown) and vegetative cells (Supplementary Figure S1). However, after 3 days, the percentage of outgrowing spores on rice agar

<sup>1</sup>http://www.i2workout.com/mcuriale/mpn/

plates reached 99.1% while on BHI agar plates 94.4% was reached (**Figure 1**).

# Impact of Matrix on the Recovery of Heat Treated B. cereus Spores

To allow for high throughput heat treatment of spores, spores were treated in thin-wall tubes in a PCR machine. Using this approach, a 45 s holding time at 95◦C resulted in approximately 2 log inactivation and 99% of damaged spores in the surviving population of wild type spores (Supplementary Figure S2). This number is comparable to previous results (91 to above 95%) obtained with capillary tubes in an oil bath (Warda et al., 2015, 2016). In the surviving population, 13% of spores were mildly damaged, whereas 86% were severely damaged. In previous findings, these numbers were 46 and 45%, for mildly and severely damaged spores, respectively (Warda et al., 2016). We previously showed that a cdnL1 mutant was affected in the ratio between mildly and severely damaged spores (Warda et al., 2016). However, the slightly different heating conditions in the high throughput method resulted in a higher fraction of severely damaged spores in the wild type spores. Using shorter holding times, an increased survival was obtained but again a relatively high percentage of severely damaged spores was observed (data not shown), therefore further experiments were performed using a holding time of 45 s.

To evaluate recovery of sorted heat treated spores in a high throughput format, a combination of single spore sorting with MPN method was applied. Sorting of spores in series of 1, 10, and 100 of heat treated spores at individual locations (well or spot) increased the resolution of the measurements allowing to observe significant differences within expected 2 log inactivation range. Interestingly, for the heat treated spores a comparable recovery in BHI broth and on rice agar plates was observed while rice broth and BHI agar plate supported recovery of approximately 50% of the surviving spores compared to BHI broth (**Figure 2**). This indicated that not only the composition of the media but also its liquid or solid state has an effect on the recovery of the spores.

# Role of Spore Damage Repair Associated Genes in Recovery of Heat Treated B. cereus Spores

Previously, genes expressed during germination and outgrowth of heat treated B. cereus spores were studied in a transcriptome study resulting in a set of 21 genes that were highly expressed

FIGURE 2 | Recovery of heat treated B. cereus ATCC 14579 spores in BHI broth (black), rice broth (light gray), BHI agar plates (dark gray), and rice agar plates (white) after 3 days of incubation. Values represent the percentage recovery relative to BHI broth (100% corresponds to 6.4% survival of wild type). Error bars represent the lower and upper limits of the MPN values expressed in percentage relative to BHI broth.

in heat treated spores relative to the reference time point at 10 min but either temporally or not expressed in untreated spores. Further evaluation with qPCR (Warda et al., 2016) to confirm the microarray data resulted in selection of 13 target genes that were downregulated in untreated spores and/or upregulated in heat treated spores with expression ratio below minus two or above two. This selection included the eight genes previously shown to be specifically upregulated during germination and outgrowth of heat damaged B. cereus spores, namely BC1312, BC3437, BC3438, BC3921, cdnL1 (BC4714), BC4834, BC5038, and BC5242 (Warda et al., 2016). A mutant strain in one of those candidate genes (cdnL1 (BC4714)), a putative transcriptional regulator, was slightly but significantly affected in repair and outgrowth of heat treated B. cereus spores (Warda et al., 2016). A paralog of cdnL1, cdnL2 (BC3648) is encoded on the B. cereus ATCC 14579 genome and it was hypothesized that its gene product masked effects on spore damage recovery in the cdnL1 deletion mutant. Therefore, a cdnL2 (BC3648) mutant and a combined cdnL1/cdnL2 mutant were included in the present study.

Attempts to construct mutants in BC3438 and BC5038 were unsuccessful. Of the 13 successfully constructed mutants (**Table 1**), four displayed various sporulation defects, mutants either did not sporulate (1BC4834), displayed an incomplete sporulation process (1BC1312) or the resulting spores were not fully released form the mother cell (1BC3437 and 1BC3921). Therefore, these mutants were excluded from further analysis.

Spores of B. cereus ATCC 14579 and its mutant derivatives, were exposed to wet heat treatment for 45 s at 95◦C. The reduction in survival of deletion mutants ranged from one up to two log with over 95% of surviving spores being damaged. The fractions of mildly and severely damaged spores were comparable to the wild type (Supplementary Figure S2). The high fraction of damaged spores allows for the assessment of the roles of candidate genes in recovery of heat treated B. cereus in different outgrowth conditions, i.e., liquid and solid forms of rice and BHI media.

1, 10, and 100 of heat treated spores were sorted either in individual wells of a 384 well plate or onto agar plates resulting in four recovery conditions, namely BHI broth, rice broth, BHI agar plates, or rice agar plates. Deletion mutants 1BC5242 and 1BC0853 were highly affected reaching only 3.6 and 9.4% recovery in BHI broth compared to that of wild type spores, respectively (**Figure 3A**). Deletion of 1BC5242 and 1BC0853 led to the highest reduction in recovery for all tested media (**Figure 3**) suggesting that effects of these genes on recovery and possibly damage repair were media independent. In contrast, deletion of BC0690 resulted in higher recovery compared to the wild type in both BHI broth (50% increase) and on BHI agar plates (150% increase) (**Figures 3A,C**). Deletion of cdnL1 (BC4714) resulted

in a recovery in BHI broth comparable to that of the wild type, albeit that time to growth was delayed in BHI broth and to a lesser extent on BHI agar plates (Supplementary Figure S3, data not shown). Deletion ofcdnL2 (BC3648) resulted in a slight reduction in recovery compared to wild type in BHI broth, while recovery of the cdnL1/cdnL2 double mutant (1BC4714/1BC3648) was reduced by approximately 50% in all tested media compared to wild type (**Figure 3C**). BC0852 and BC0460 mutants displayed a comparable reduction in recovery as the cdnL1/cdnL2 double mutant. Finally, the recovery of 1BC1314 depended on the recovery media, in rice broth the recovery was comparable to wild type (**Figure 3B**) while in BHI broth deletion led to over 50% reduction in recovery compared to wild type. Recovery of heat treated spores of all but two (1BC0460 and 1BC0690) deletion mutants was higher on rice agar plates, compared to BHI broth (Supplementary Figure S4). The recovery of those mutant spores improved also in rice broth and on BHI agar plates when compared to relative recovery of the wild type. This suggests that conditions supporting slower growth favor recovery of spores possibly by providing additional time for damage repair.

# DISCUSSION

The capacity of spores to repair damage and grow out is not only affected by the processing conditions, but also by spore history and recovery conditions. Although, several studies report on impact of food components on spore survival and cell growth (Carlin et al., 2000; Choma et al., 2000; Moussa-Boudjemaa et al., 2006), mainly plate counting methods that do not allow for analysis of individual spores have been applied. Moreover, the standard plate counting methods are generally not sensitive enough to show changes within the 10-fold range. In practice, product spoilage may result from a single surviving spore and knowledge on behavior of individual spores can assist in risk evaluation. Here we applied a FCM supported single spore sorting approach in combination with MPN methodology, allowing for evaluation of behavior of individually sorted spores with high resolution for both untreated as well as heat treated spores.

The 45 s heat treatment at 95◦C of B. cereus wild type and mutant spores resulted in approximately 2 log inactivation, and above 95% damaged spores in the surviving population, which is comparable to previously reported survival and total damage at this temperature (Warda et al., 2015, 2016). Limited information is available on the effect of the recovery media on outgrowth of single damaged spores. In the present study, we focused on the effect of media composition, either liquid or solid state, on the combined process of germination, outgrowth and vegetative growth of individually sorted untreated and heat treated B. cereus ATCC 14579 wild type and mutant spores. Firstly, we showed for wild type spores that rice broth was least supporting the growth and recovery of heat treated spores while rice agar plates provided comparable recovery as BHI broth, indicating that not only the composition but also the liquid or solid state of media effects the recovery of heat treated spores. Both heat treated and untreated B. cereus spores showed similar recovery when plated on BHI and rice agar plates (Warda et al., 2015). However, the formation of microcolonies from individual spores on Anopore (a porous membrane allowing nutrient transfer that provides surface for spore/cell growth) conditions, which is more close to conditions in a broth (den Besten et al., 2007) were found different for BHI and rice (Warda et al., 2015). More specifically, rice media increased heterogeneity and delayed outgrowth of untreated spores compared to BHI, and also a heat treatment had a limited additional effect on the behavior of surviving spores (Warda et al., 2015). Now we show that outgrowth from untreated single spores was slower in rice based media compared to BHI, but final counts for untreated single sorted spores on rice plates were 99.1% while on BHI plates 94.4%. In line with our previous observations, the time required for colony formation from untreated B. cereus spores on rice media was extended compared to BHI, indicating that rice media may contain additional factors delaying germination and/or outgrowth or contain suboptimal concentrations of required components (Warda et al., 2015).

Comparative analysis of wild type and selected mutants lacking genes with putative roles in damage repair, showed different capacity to recover from heat stress compared to wild type. The most pronounced effect was observed for a deletion mutant, lacking a membrane protein with C2C2 zinc finger (BC5242). This mutation resulted in reduction in recovery down to 3.6% of the wild type recovery in BHI broth. The function of BC5242 is unknown, but orthologs of its gene product can be found in many B. cereus group strains though not in B. subtilis 168. In eukaryotes, zinc finger containing proteins function in gene transcription, translation, mRNA trafficking, cytoskeleton organization, epithelial development, cell adhesion, protein folding, chromatin remodeling, and zinc sensing (Laity et al., 2001; Gamsjaeger et al., 2007). In prokaryotes, zinc finger motifs (C4 superfamily) are found in proteins involved in DNA damage recognition, i.e., UvrA, Ada, RecR (Ayora et al., 1997), however, the diversity in functionality of zinc finger carrying proteins and the zinc finger domains does not allow for prediction of a role for BC5242 in B. cereus. Notably, BC5242 was not upregulated in vegetative cells of B. cereus ATCC 14579 in response to different stresses including cold, ethanol, some disinfectants, and mild acid (Abee et al., 2011).

BC1314 was found to be highly upregulated during germination and outgrowth of heat damaged B. cereus spores (Warda et al., 2016). The recovery of 1BC1314 spores after a heat treatment was decreased with 50% compared to wild type in BHI broth, and on BHI and rice agar plates, albeit less severe for the latter two media, thus suggesting a role of BC1314 in the recovery of heat treated spores. Analysis of the B. cereus ATCC 14579 genome sequence suggested that BC1314 (and BC1315) result from a frame-shift mutation in the phaQ gene (Supplementary Figure S5). The B. cereus phaQ gene is part of a poly-β-hydroxybutyrate (PHB) synthesis cluster, and PHB was previously shown to be accumulated in cells in the form of granules that serve as a carbon and energy source during the late sporulation process in B. cereus (Kominek and Halvorson, 1965) and B. megaterium (Slepecky and Law, 1961). In B. megaterium, PHB accumulation involves five genes, namely phaP (encoding a phasin protein), phaQ (encoding a repressor of phaP expression),

phaB (acetoacetyl-CoA reductase), phaR and phaC (subunits of PHB synthase) (Supplementary Figure S5) (McCool and Cannon, 1999; Lee et al., 2004). Furthermore, in B. thuringiensis accumulation of PHB via phaPQRBC was shown to be under the control of the sporulation transcription factors sigH and Spo0A (Chen et al., 2010). In strains belonging to the B. cereus group, orthologs of the phaPQRBC system are commonly present, while being absent in B. subtilis 168, pointing to a special role for this system in the indicated group.

The cdnL1/cdnL2 double deletion mutant (1BC4714/1BC3648), lacking genes encoding both CdnL transcriptional regulators present in B. cereus ATCC 14579 showed 60% reduction in recovery in BHI broth compared to the wild type. Deletion of cdnL1 (BC4714) was shown previously to increase the fraction of severely damaged spores in the surviving population after a heat treatment of 1 min at 95◦C (Warda et al., 2016). Since the heat treatments applied in the present study led to a dominant fraction of severely damaged spores already in the wild type, we could not observe the increase in percentage of severely damaged spores in 1cdnL1 (BC4714). However, outgrowth from heat treated 1cdnL1 spores was delayed in BHI broth compared to the wild type spores, eventually reaching comparable recovery efficiency. The 1cdnL2 mutant (BC3648) showed lower recovery compared to 1cdnL1 (BC4714), and this was most pronounced in liquid media. Nevertheless, both 1cdnL1 and 1cdnL2 in media other than BHI broth show improved recovery compared to wild type. It remains to be determined whether the observed increase in recovery of the individual cdnL mutants could be explained by cross regulation of the counterpart. Both cdnL1 and cdnL2 genes are induced in vegetative cells in response to various environmental stresses, including salt and cold stress, whereas acid and oxidative stress specifically induced expression of cdnL1 and not cdnL2 (Abee et al., 2011). Our findings suggest partly overlapping functionalities of cdnL1 and cdnL2 in recovery and possibly repair of heat damage.

Spores of the 1BC0690 mutant, lacking a putative PbsX family transcriptional regulator of unknown function, showed higher recovery compared to wild type spores in all tested conditions, with increase of up to 150% on BHI agar plates. Orthologs of BC0690 are commonly found among B. cereus group strains, but absent in B. subtilis 168, pointing possibly to a unique, but up to now unknown role in heat stress survival in B. cereus group members.

Deletion of BC0852 and BC0853, both encoding putative quaternary ammonium compound resistance proteins annotated as sugE, resulted in reduction in recovery of spores to 9.4 and 28.1% of the wild type in BHI broth, respectively. Orthologs of BC0852 and BC0853 are present in B. cereus group strains, while being absent in B. subtilis 168. Besides BC0852 and BC0853, the B. cereus ATCC 14579 genome encodes a second orthologs pair of small multidrug resistance proteins (BC4213 and BC4214) orthologs to ykkC (BSU13090) and ykkD (BSU13100) of B. subtilis 168. ykkC and ykkD are a paired small multidrug resistance (PSMR) members, and their co-expression in Escherichia coli led to a multidrug-resistant phenotype (Jack et al., 2000). Still, not all PSMR members have demonstrated drug resistance, e.g., B. subtilis YvaD/YvaE and YvdR/YvdS, and small multidrug resistance homologs were suggested to be involved in transport of yet unidentified compounds (Bay and Turner, 2009).

In the current study, the applied heat treatment resulted in at least 95% of damaged spores in the surviving wild type and deletion mutant spore populations, based on the fact that these spores were not able to grow out on salt supplemented plates (compared to BHI agar plates). At the moment it cannot be excluded that differences in spore recovery in BHI broth are due to lack of one or more specific proteins in spores of tested deletion mutants that makes them more or less resistant and/or susceptible to heat damage. However, application of rice media and BHI agar plates compared to BHI broth for sorted spores also revealed differences in recovery between media suggesting different requirements for recovery. Particularly deletion of BC0460 or BC0690 resulted in reduced recovery on rice plates while spores of remaining seven deletion mutants showed improved recovery on rice plates compared to BHI broth (Supplementary Figure S4). As the recovery of the various deletion mutants spores appears matrix dependent, this suggests that mutations conceivably affected different type of damage and/or repair targets as was suggested previously by Adams (Adams, 1973). Apparently, high numbers of damaged spores were present in the surviving wild type and mutant spore population, but nevertheless, subtle effects of mutations in putative repair genes were noted, resulting in a shift from the fraction of mildly damaged to the fraction of severely damaged spores (Warda et al., 2016) and in differences in recovery between different media (this study). Still, recovery of heat treated spores is a complex process conceivably involving many different systems, and more studies are required to elucidate the full repertoire of repair systems and the impact of matrix composition and its solid or liquid state on this process.

# CONCLUSION

We have shown that recovery of heat treated B. cereus spores is affected by the matrix composition with highest recovery of wild type spores in BHI broth or on rice agar plates, followed by BHI agar plates and rice broth. The comparative analysis of the wild type and newly constructed deletion mutants provided new insights in the putative role of the deleted genes in the recovery of heat treated B. cereus spores.

# AUTHOR CONTRIBUTIONS

Conceived and designed the experiments: AW, MT, TA, MNG. Performed the experiments: AW, MT. Analyzed the data: AW, MT. Wrote the paper: AW, MT, TA, MNG.

# FUNDING

TIFN provided support in the form of salaries for authors [AW, MT, MNG, TA], but did not have any additional role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript. The specific roles of these authors are articulated in the 'author contributions' section. Food & Biobased Research (FBR) is a contract research organization, part of Wageningen University and Research centre. FBR was paid by TIFN to provide input on experimental design, data generation, data interpretation, preparation of the manuscript (salary of MNG, AW). FBR has no financial interest in any specific outcome of the study.

# ACKNOWLEDGMENTS

fmicb-07-01096 July 14, 2016 Time: 14:13 # 9

The project is funded by TI Food and Nutrition, a public–private partnership on precompetitive research in food and nutrition.

# REFERENCES


The public partners are responsible for the study design, data collection and analysis, decision to publish, and preparation of the manuscript. The private partners have contributed to the project through regular discussion. We would like to thank Clint van Melis and Yao Lu for their contribution in preliminary experiments.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2016.01096

and sensitizes spores to subsequent stress. J. Appl. Microbiol. 97, 838–852. doi: 10.1111/j.1365-2672.2004.02370.x



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

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

# The Synergistic Effect of High Pressure CO<sup>2</sup> and Nisin on Inactivation of Bacillus subtilis Spores in Aqueous Solutions

Lei Rao1,2, Yongtao Wang<sup>2</sup> , Fang Chen<sup>2</sup> and Xiaojun Liao1,2 \*

<sup>1</sup> Beijing Advanced Innovation Center for Food Nutrition and Human Health, College of Food Science and Nutritional Engineering, China Agricultural University, Beijing, China, <sup>2</sup> Key Laboratory of Fruit and Vegetable Processing, Ministry of Agriculture, Beijing, China

The inactivation effects of high pressure CO<sup>2</sup> + nisin (simultaneous treatment of HPCD and nisin, HPCD + nisin), HPCD→nisin (HPCD was followed by nisin), and nisin→HPCD (nisin was followed by HPCD) treatments on Bacillus subtilis spores in aqueous solutions were compared. The spores were treated by HPCD at 6.5 or 20 MPa, 84–86◦C and 0–30 min, and the concentration of nisin was 0.02%. Treated spores were examined for the viability, the permeability of inner membrane (IM) using flow cytometry method and pyridine-2, 6-dicarboxylic acid (DPA) release, and structural damage by transmission electron microscopy. A synergistic effect of HPCD + nisin treatment on inactivation of the spores was found, and the inactivation efficiency of the spores was HPCD + nisin > HPCD→nisin or nisin→HPCD. Moreover, HPCD + nisin caused higher IM permeability and DPA release of the spores than HPCD. A possible action mode of nisin-enhanced inactivation of the spores was suggested as that HPCD firstly damaged the coat and cortex of spores, and nisin penetrated into and acted on the IM of spores, which increased the damage to the IM of spores, and resulted in higher inactivation of the spores.

#### Edited by:

Michael Gänzle, University of Alberta, Canada

#### Reviewed by:

Sergio I. Martinez-Monteagudo, South Dakota State University, USA Michael A. Matthews, University of South Carolina, USA

> \*Correspondence: Xiaojun Liao liaoxjun@hotmail.com

#### Specialty section:

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

Received: 02 June 2016 Accepted: 08 September 2016 Published: 21 September 2016

#### Citation:

Rao L, Wang Y, Chen F and Liao X (2016) The Synergistic Effect of High Pressure CO<sup>2</sup> and Nisin on Inactivation of Bacillus subtilis Spores in Aqueous Solutions. Front. Microbiol. 7:1507. doi: 10.3389/fmicb.2016.01507 Keywords: high pressure CO2, nisin, synergistic inactivation, Bacillus subtilis spores, inner membrane damage

# INTRODUCTION

Bacterial endospores are metabolically dormant, and extremely resistant to the treatments such as heat, desiccation, UV, and γ-radiation, and some bactericidal chemicals because of their unique structures (Setlow, 1995, 2006; Setlow and Johnson, 2012). As spores of a number of Bacillus and Clostridium species are agents of food spoilage and food borne diseases (Brown, 2000; Logan, 2012; Setlow and Johnson, 2012), inactivation of spores has been receiving great attention in the food industry. Traditionally, thermal processing at relative high temperature (121◦C or higher) is an efficient way to eliminate spores. However, the high temperature compromises organoleptic properties and causes some detrimental effects to the nutritional quality of heat-sensitive food. Consequently, there is a requirement for new ways of mild processing procedures to inactivate spores.

The inactivation effect of high pressure CO<sup>2</sup> (HPCD) was firstly shown on Escherichia coli in 1951 (Fraser, 1951). In the recent years, a number of reports indicate that HPCD at

pressure < 30 MPa and temperature of 20–40◦C can effectively inactivate the vegetative forms of pathogenic and spoilage bacteria, yeasts, and molds (Spilimbergo and Bertucco, 2003; Zhang et al., 2006c; Perrut, 2012), and simultaneously maintain or improve the nutrient quality of liquid food (Damar and Balaban, 2006; Garcia-Gonzalez et al., 2007). Considering the efficient bactericidal effect of HPCD as well as its ability of maintaining or improving food quality, HPCD has been proposed as an alternative non-thermal pasteurization technique for foods. However, a problem of application HPCD in food processing is that HPCD at 20–40◦C cannot inactivate bacterial spores (Rao et al., 2015a), which could be a potential risk for food safety (Brown, 2000). Generally high temperatures (>60◦C) are needed for HPCD treatment to effectively inactivate spores (Enomoto et al., 1997; Ballestra and Cuq, 1998; Spilimbergo et al., 2003; Watanabe et al., 2003; Bae et al., 2009; Rao et al., 2015b). A variety of chemicals are reported to be combined with HPCD to increase the inactivation effect of spores, and hydrogen peroxide, tert-butyl hydroperoxide, peracetic acid, or trifluoroacetic acid could enhance the inactivation efficiency of spores by HPCD treatment at mild conditions (35–60◦C) (White et al., 2006; Zhang et al., 2006a,b, 2007; Hemmer et al., 2007; Shieh et al., 2009; Tarafa et al., 2010; Checinska et al., 2011; Setlow et al., 2016). However, as the addition of these chemicals into food are prohibited, they cannot be applied in the sterilization procedures of food processing.

Nisin is an antimicrobial peptide produced by certain strains of Lactococcus (Delves-Broughton, 1990), and inhibits the growth of gram-positive bacteria and their endospores (de Arauz et al., 2009). It is generally regard as safe (GRAS) and its application in food has been approved by United States Food and Drug Administration (USFDA). Nisin acts on Gram-positive bacteria by forming pores in cell membranes (Ruhr and Sahl, 1985) or inhibits cell wall biosynthesis by disrupting of transglycosylation via binding to and mislocation of lipid II, a precursor for cell wall biogenesis (Wiedemann et al., 2001; Hasper et al., 2006). However, nisin shows no activity on Gram-negative bacteria because its access into the cytoplasmic membrane is blocked by the outer membrane (Stevens et al., 1991). In our previous study, Li et al. (2016) reported that HPCD + 0.02% nisin at 10 MPa and 32◦C for 15 min showed an enhanced inactivation of E. coli in aqueous solutions compared with HPCD alone at the same condition, and this enhancement was due to the damage to the outer membrane of E. coli by HPCD, and then nisin penetrated into and act on the cytoplasmic membrane. Similar to the gram-negative bacteria, nisin also cannot act on the bacterial spores of dormant form because its access into the inner membrane (IM) is blocked by the coat and thick cortex of spores (Black et al., 2007; Gut et al., 2008). But after spore germination, the cortex and coat are degraded, and nisin can act on the IM by forming pores via binding to the lipid II, inhibiting the germinated spores outgrowing into vegetative cells (Gut et al., 2011). Given these findings, we assumed that nisin can theoretically act on the IM of the spores and inhibit their outgrowth if the out layers of spores including cortex and coat were damaged. Zhang et al. (2006b) reported that HPCD could damage the envelope of Bacillus atrophaeus spores inoculated on filter paper including exosporium, coat, cortex and IM through transmission electronic microscopy (TEM) images and DPA analysis. Meanwhile, our previous results also showed that HPCD at 20 MPa and 84–86◦C kill the spores of B. subtilis in aqueous solutions most likely by destroying the structure of spores (Rao et al., 2015b). Thus, we assumed that nisin may act on HPCD-treated spores, and increase the inactivation of spores. In fact, a recent report indicated that combined treatment of HPCD at 30 MPa and 60◦C for 120 min (4 cycles of 30 min each) and 0.01–0.5% nisin showed synergistic effect on inactivating B. subtilis and Geobacillus stearothermophilus spores inoculated on the surface of metal plates (Da Silva et al., 2016), but it did not show how this synergistic effect happened.

In this study, we investigated the effect of different combined treatments of HPCD and nisin on inactivating of B. subtilis spores in aqueous solutions, and figured out the role of the nisin in the inactivation of spores by HPCD. The IM damage of spores was analyzed by detecting the IM permeability using flow cytometry method (FCM) and DPA release, and the structural damage of spores was observed by TEM.

# MATERIALS AND METHODS

# Strain and Spore Preparation

Bacillus subtilis 168 was obtained from China General Microbiological Culture Collection Center (Beijing, China). Overnight cultures of Bacillus strain grown in nutrient broth (Beijing Aoboxing Biological Technology Co. Ltd., Beijing, China) were transferred to sporulation agar plates, nutrient agar (Beijing Aoboxing Biological Technology Co. Ltd., Beijing, China) containing 50 µg/mL Mn2+. After 1 week incubation at 37◦C, the spores were harvested in a sterile flask by flushing the surface of the culture with sterile distilled water and scrapping the surface with sterile glass microscope slide. The spores collected were washed three times by centrifugation at 7000 × g and 4◦C for 15 min using a CF16RXII centrifuge (Hitachi, Japan), resuspended in sterile distilled water with a concentration of approximate 10<sup>9</sup> CFU/mL, and stored at 4◦C until they were used. All spores (>99%) used in this work were free of growing and sporulating cells, germinated spores and cell debris as determined with a BX45-72P15phase contrast microscope (Olympus, Japan). The concentration of the spore suspension was adjusted to approximate 10<sup>7</sup> CFU/mL before treatments.

# HPCD and Nisin Combined Treatment

One gram of nisin (10<sup>6</sup> IU/g) was dissolved in 50 mL distilled water and filter sterilized using 0.22 µm sterile filter. Then, the nisin solution was added to the spore suspensions before immediate HPCD treatment. The final concentration of the nisin was 0.002% (20 IU/mL), 0.01% (100 IU/mL), 0.02% (200 IU/mL), 0.04% (400 IU/mL), respectively. HPCD was performed with a batch HPCD system (Liao et al., 2007). For each experiment, 20 mL of the spore suspension with

pH 6.5 were transferred to a 50 mL sterile glass tube and the tube was covered with a plastic film with a 0.22 µm membrane filter in the center of aeration to prevent microbial contamination. As the pressure vessel of the HPCD system reached the experimental temperature (84–86◦C), the sample tubes were placed in the pressure vessel. Then, the vessel was pressurized by the plunger pump to 6.5 or 20 MPa within 0.1 min or 2.5 min, respectively. After holding for required treatment time, the depressurization was performed by opening the pressure relief valve at CO<sup>2</sup> outlet on the pressure vessel. The depressurization time was 0.5 and 2.5 min for 6.5 and 20 MPa, respectively. After HPCD, the sample tubes were taken out from the vessel and analyzed immediately. The CO<sup>2</sup> purity was 99.5% in all the experiment treatments. The combined treatments of HPCD and nisin was carried out as follows. (i) HPCD→nisin treatment: the spore suspensions without nisin were treated by HPCD at 20 MPa and 84–86◦C for 30 min, then cooled down to ambient temperature, and 0.02% nisin was added into the HPCD-treated spores for 30 min, then centrifuged at 4◦C and 7000 × g for 10 min and resuspended in sterile distilled water. (ii) nisin→HPCD treatment: the spore suspensions were treated with 0.02% nisin at ambient temperature for 30 min, centrifuged at 4◦C and 7000 × g for 10 min and resuspended in sterile distilled water, then nisintreated spores were treated by HPCD at 20 MPa and 84–86◦C for 30 min. (iii) HPCD + nisin treatment: the spore suspensions were treated by HPCD and 0.02% nisin at 20 MPa and 84–86◦C for 30 min, then cooled down to ambient temperature, centrifuged at 4 ◦C and 7000 × g for 10 min and resuspended in sterile distilled water.

The inactivation of the spore suspensions by heat treatment with or without nisin at 86◦C was carried out at 0.1 MPa without CO<sup>2</sup> using a water bath. Twenty microliter of the spores suspended in sterile distilled water with pH 6.5 or pH 3.0 (HCl was used to adjust the acidity) was transferred to a 50 mL sterile glass tube, which was then immersed in a water bath equilibrated at 86◦C for 0–30 min. After treatments, the sample tubes were taken out and analyzed immediately.

# Enumeration of Surviving Spores

The number of surviving spores was determined by the viable plate count method. For the spores treated with nisin, the samples were centrifuged at 4◦C and 7000 × g for 10 min and resuspended in sterile water to eliminate the nisin. Then, each sample was serially (1:10) diluted with sterile distilled water and pour-plated on nutrient agar (Beijing Aoboxing Biological Technology Co. Ltd., Beijing, China) in duplicate. The plates were incubated at 37◦C for 24 h. After incubation, the colonies were counted.

# Measurement of DPA

The DPA release was measured using the fluorescence method (Hindle and Hall, 1999). The treated spores were centrifuged at 7000 × g and 4◦C for 10 min (CF16RXII, Hitachi, Japan), and assaying DPA in the supernatant solution was carried out by its fluorescence with Tb3<sup>+</sup> in a 96-well plate. One hundred µL of supernatant solution were added to 100 µL 20 µmol/L terbium (III) chloride hexahydrate (99.9%, Aladdin Industrial Corporation, Shanghai, China) buffered with 1 mol/L acetic acid (99.8%, Beijing Chemical Works, Beijing, China) at pH 5.6. All the samples were analyzed with a Multiskan MK3 microplate reader (Thermo, MA, USA). Samples were excited at 270 nm, and emission spectra were collected at 545 nm. The total amount of DPA in each individual batch was determined after autoclaving at 121◦C for 20 min (Zhang et al., 2006a), which was used as a positive control while the one in untreated spores was used as a negative control. HPCD-induced DPA release ratio was calculated by the equation as follows:

$$\text{DPA\%} = \frac{F\_1 - F\_0}{F\_2 - F\_0} \tag{1}$$

Where F0, F1, and F<sup>2</sup> were the fluorescence intensity of untreated spores, HPCD-treated spores and autoclaved-spores, respectively.

# Flow Cytometry Analysis

Samples for flow cytometry were prepared with propidium iodide (PI), a DNA staining dye according to a reported method (Reineke et al., 2013). PI is membrane-impermeable and can be used to indicate the IM damage (Mathys et al., 2007). The treated spore suspensions were adjusted to concentrations of about 10<sup>7</sup> spores/mL in sterile distilled water. The concentration of the PI was 15 µmol/L in the spore suspensions. Afterward, the samples were stored in the dark at room temperature for 45 min (Pappas et al., 2015).

Stained samples were then analyzed with an Accuri C6 (BD Accuri Cytometer Inc., USA) flow cytometry equipped with a 488 nm, 50 mWlaster. PI fluorescence was quantified with the FL2 detector at 585 ± 20 nm. The forward scatter threshold was set at 5000 to ensure that the small spores were not omitted as events. Spores were analyzed at a nominal flow rate of 14 µL/min, with a stream core diameter of 10 µm. All samples were evaluated after 30000 events had been recorded.

# Spores Preparation for TEM

Samples were prefixed in 2.5% glutaraldehyde (Sigma–Aldrich) overnight at room temperature, rinsed three times in 0.1 mol/L phosphate-buffered saline (PBS) for 15 min, then postfixed in 1% osmium tetroxide (Sigma–Aldrich) for 90 min and rinsed three times in 0.1 mol/L PBS for 15 min, and subsequently dehydrated with ethanol series. Afterward, the samples were embedded in epoxy resin and kept at 37◦C overnight followed by 60◦C for 24 h. The resin blocks were cut into ultrathin sections of 70 nm with a Lecia EM UC6 ultramicrotome (Leica, Germany) and stained with 3% aqueous lead citrate and uranyl acetate. Finally, the samples were examined by a H-7650B TEM (Hitachi, Japan).

# Data Analysis

Flow cytometry data were analyzed using the FlowJo version 7.6.1 software (FlowJo, OR, USA). Analysis of variance (ANOVA) was carried out by using software PASW statistic 18 (SPSS, USA). ANOVA tests were carried out for all experimental runs to determine significance at α = 0.05 level. All experiments were carried out in triplicate.

FIGURE 1 | Inactivation of Bacillus sutilis spores by (A) heat without or with nisin, or ordered sequential treatment of HPCD and nisin. <sup>a</sup>Spores in aqueous solutions with pH 6.5 were treated by heat at 0.1 MPa and 86◦C for 30 min; <sup>b</sup>Spores in aqueous solutions with pH 3.0 were treated by heat at 0.1 MPa and 86◦C for 30 min; <sup>c</sup>Spores in aqueous solutions with pH 6.5 were treated by 0.02% nisin at 0.1 MPa and ambient temperature for 30 min; <sup>d</sup>Spores in aqueous solutions with pH 6.5 were treated by heat at 0.1 MPa and 86◦C for 30 min with 0.02% nisin; <sup>e</sup>Spores in aqueous solutions with pH 3.0 were treated by heat at 0.1 MPa and 86◦C for 30 min with 0.02% nisin; <sup>f</sup>Spores were treated by HPCD at 20 MPa and 84–86◦C for 30 min; <sup>g</sup>Spores in aqueous solutions with pH 6.5 were treated by HPCD at 20 MPa and 84–86◦C for 30 min followed by 0.02% nisin treatment at 0.1 MPa and ambient temperature for 30 min; <sup>h</sup>Spores in aqueous solutions with pH 6.5 were treated by 0.02% nisin at 0.1 MPa and ambient temperature for 30 min followed by HPCD treatment at 20 MPa and 84–86◦C for 30 min; <sup>i</sup>Spores in aqueous solutions with pH 6.5 were treated by HPCD at 20 MPa and 84–86◦C for 30 min with 0.02% nisin; (B) heat or HPCD at different pressures without () or with 0.02% nisin ( ). Heat: 0.1 MPa, 86◦C, 30 min; HPCD: 6.5 MPa or 20 MPa, 84–86◦C, 30 min; (C) heat or HPCD without (, 4) or with (, N) 0.02% nisin at different times. (, ) Heat: 0.1 MPa, 86◦C, 30 min; (4, N) HPCD: 20 MPa, 84–86◦C, 30 min.

# RESULTS

# Inactivation of Spores by Combined Treatment of HPCD and Nisin

In a preliminary trial of screening the concentrations of nisin (Supplementary Figure S1), 0.02% nisin was chosen to combine with HPCD to inactivate the spores in this study. As shown in **Figure 1A**, both heat treatment at 86◦C for 30 min and 0.02% nisin at ambient temperature for 30 min exhibited small inactivation (≤ 0.2 log reduction), while HPCD at 20 MPa and 84–86◦C for 30 min displayed high inactivation (2.1 log reduction), indicating that HPCD exhibited significantly higher

inactivation than heat or nisin, which was similar to our previous results (Rao et al., 2015b). When nisin was added, the inactivation of the spores by heat treatment at pH 6.5 (0.99 log reduction) was enhanced, but there was no enhancement of inactivation for the heat treatment at pH 3.0 (0.13 log reduction), indicating that acidity was not enough to effectively inactivate spores. For HPCD treatment, the combined treatments of HPCD and nisin achieved a synergistic inactivation effect, and HPCD + nisin achieved higher inactivation (4.1 log reduction) than HPCD→nisin (2.7 log reduction) or nisin→HPCD (2.9 log reduction). As the HPCD + nisin was most efficiency to inactivate spores, it was employed in the following study. The inactivation of spores by HPCD + nisin as a function of the pressure was shown in **Figure 1B**. The addition of nisin increased 1.4 and 1.9 log more reduction of the spores than HPCD at 6.5 and 20 MPa, respectively, indicating that higher pressure achieved stronger synergistic inactivation effect of the spores. Inactivation kinetics of the spores by HPCD + nisin was shown in **Figure 1C**. For the spores treated by HPCD at 20 MPa and 84–86◦C, the inactivation showed no difference from 0 to 10 min, and then increased with increasing the time, exhibiting a slow to fast inactivation pattern. When nisin was added, the synergistic inactivation effect also increased with increasing the time, and the inactivation kinetics of the spores also exhibited a slow to fast inactivation pattern.

# FCM Analysis

The IM permeability of the spores treated by HPCD + nisin was examined by FCM with membrane-impermeable PI, and untreated spores were used as negative control. As shown in **Figure 2**, the FCM histograms of red fluorescence distribution of the spores stained by PI were divided into M1 and M2 regions. M2 was the negative area in which the spores were not stained by PI and had intact IM, while M1 was the positive area indicating that the spores were stained by PI and the IM TABLE 1 | Inactivation of wet or dry bacterial spores by HPCD treatment at different conditions.


of spores were damaged. Compared with the untreated spores (**Figure 2A**), the fluorescence distribution of the HPCD- and HPCD + nisin-treated spores were moved toward M1, but the HPCD + nisin-treated spores showed a stronger move than the HPCD-treated ones, indicating that HPCD + nisin achieved

more severe damage to the IM of spores than HPCD. It was also evidenced by higher PI-positive percentage of the HPCD + nisintreated spores, and the percentage was increased with increasing the treatment time (**Figure 2G**).

# DPA Release

The DPA release of the spores was another indicator of the damage to the IM of spores. As shown in **Figure 3**, the DPA release of the spores treated by heat at 86◦C for 0–30 min was less than 14%, indicating that the IM of mostly heat-treated spores was intact. The DPA release of the spores was increased and less than 21% when the spores were subjected to heat + nisin, suggesting that the IM of many spores was still intact. The DPA release of the HPCD-treated spores for > 5 min was far higher than that by heat or heat + nisin, the highest DPA release was 80%, indicating the IM of mostly spores was damaged. Moreover, the DPA release of the HPCD + nisin-treated spores was significantly higher than that of the HPCD-treated spores, confirming that HPCD + nisin achieved more severe damage to the IM of spores.

# DISCUSSION

In this study, the inactivation of the combined treatments of HPCD and nisin on B. subitlis spores in aqueous solutions was investigated, and a synergistic effect of HPCD + nisin, HPCD→nisin and nisin→HPCD on inactivation of spores was

fmicb-07-01507 September 21, 2016 Time: 12:57 # 6

inactivation of spores by HPCD.

found (**Figure 1**). Similarly, Da Silva et al. (2016) reported that HPCD at 30 MPa and 60◦C for 120 min was not able to efficiently inactivate (0.41 log reduction) B. subtilis spores inoculated on metal surface, while HPCD + 0.05% nisin achieved 3.2 log inactivation, showing an enhanced inactivation effect (**Table 1**). However, it not clear how nisin enhanced the inactivation of the spores by HPCD, and it was necessary to figure out possible action mode of nisin in the synergistic inactivation effect.

It is reported that nisin cannot act on intact spores because its access to the IM was blocked by the coat and cortex, but after the spores germinated and degraded their coat and cortex, nisin could act on them and inhibit their outgrowth by forming pores in the IM of spores (Gut et al., 2011). Our previous work indicated that the inactivation of the spores in aqueous solutions by HPCD was likely attributed to the structural damage of the spores, rather than the germination (Rao et al., 2015b). Bae et al. (2009) observed damages to the surface and internal structures of Alicyclobacillus acidoterrestris spores in apple juice treated by HPCD at 10 MPa and 70◦C for 30 min using scanning electron microscopy (SEM) and TEM (**Table 1**). Zhang et al. (2007) also evidenced that HPCD damaged the envelope of B. atrophaeus spores inoculated in filter paper including exosporium, coat, cortex, and IM (**Table 1**). In this study, the HPCD-treated spores showed visible structure changes with increasing the treatment time (**Figure 4**), especially after HPCD treatment of 10, 20, and 30 min, the damage to the coat (**Figures 4D–F**), cortex (**Figures 4D–F**), IM (**Figures 4E,F**), and core (**Figure 4F**) was manifested. These HPCD caused damages to the barrier (the coat and cortex of spores) blocking the access of nisin to the IM allowed nisin penetrated into and acted on the IM, and resulted in more damage to the IM of spores. This reasoning was evidenced by FCM (**Figure 2**) and DPA analysis (**Figure 3**), which suggested that the HPCD + nisintreated spores exhibit higher IM permeability than the HPCDtreated spores. Therefore, we proposed that HPCD promoted the penetration of nisin into the spore cells by damaging the coat and cortex of spores, and then nisin acted on the IM by binding to the lipid II and forming pores in the IM, inhibiting the outgrowth of spores (Gut et al., 2011). As nisin-increased damage to the IM of spores (**Figures 2** and **3**) was coincident with the higher inactivation of the HPCD + nisin-treated spores (**Figure 1C**), the action mode of nisin in the synergistic inactivation effect of the spores was explained as follows (**Figure 5**). Firstly, HPCD damaged the coat and cortex of spores and increased their permeability, then nisin penetrated into the spores and acted on the IM by binding to the lipid II and forming pores, then increased more IM damage of the spores and resulted in higher inactivation of the spores. Similarly, the synergistic inactivation effect at higher pressures (**Figure 1B**) and prolonged times (**Figure 1C**) was also due to increased damage to the structure of spores, which benefited the penetration of nisin into the spores and increased the damage to the IM of spores and inactivation. Moreover, the enhanced inactivation efficiency of these HPCD and nisin treatments was HPCD + nisin > HPCD→nisin or nisin→HPCD (**Figure 1A**). Comparatively, HPCD→nisin generated lower synergistic inactivation effect of spores, and this was most likely attributed to the lower temperature during nisin treatment compared to HPCD + nisin treatment since higher temperature increased the efficiency of inactivating spores by nisin (**Figure 1A**). Meanwhile, nisin→HPCD also achieved lower synergistic inactivation effect of spores. Theoretically, nisin→HPCD should not enhance the inactivation of the spores since nisin at room temperature cannot act on the spores. Possible explanation for nisin→HPCD enhanced inactivation was probably due to the adherence of nisin to the surface of spores after nisin treatment (Chilton et al., 2013; Kraus et al., 2015), and the remaining nisin played an enhanced inactivation of the spores. Its action mode was similar to that of HPCD + nisin.

This study showed that HPCD + nisin achieved a synergistic inactivation effect of B. subtilis spores compared with HPCD or nisin alone, and this synergistic effect was due to nisin-increased damage to the IM of spores. Moreover, nisin could access into the IM of spores because HPCD damaged the barrier of spores including the coat and cortex. However, how HPCD damaged the barrier of spores is not clear, and needed to be further studied.

# AUTHOR CONTRIBUTIONS

LR: carrying out the experiments and writing the manuscript. YW: giving advice and assistance for the experiment. FC: reviewing the manuscript and giving advice. XL: designing the experiment and revising the manuscript.

# ACKNOWLEDGMENTS

This study was supported by Collaborative Research on Early Warning and Control of Microbiological Hazards in Minimally Processed Fruits and Vegetables (No. 2013DFA31450) of International Science & Technology Cooperation Program of China, Key Project of Chinese Ministry of Education (No. 113011A), Key Project of National Natural Science Foundation of China (NSFC) (No. 31530058), Chinese Universities Scientific Fund (No. 2015SP004), Beijing Training Project For The Leading Talents in S & T (No. Z151100000315032), and Beijing Advanced Innovation Center for Food Nutrition and Human Health, Beijing, China.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2016.01507

# REFERENCES

fmicb-07-01507 September 21, 2016 Time: 12:57 # 8


spores using supercritical carbon dioxide. J. Supercrit. Fluids 38, 268–273. doi: 10.1016/j.supflu.2006.02.015


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

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

# Investigating the Inactivation Mechanism of Bacillus subtilis Spores by High Pressure CO<sup>2</sup>

Lei Rao1,2,3, Feng Zhao2,3, Yongtao Wang2,3, Fang Chen2,3, Xiaosong Hu1,2,3 and Xiaojun Liao1,2,3 \*

<sup>1</sup> Beijing Advanced Innovation Center for Food Nutrition and Human Health, College of Food Science and Nutritional Engineering, China Agricultural University, Beijing, China, <sup>2</sup> National Engineering Research Center for Fruit and Vegetable Processing, Beijing, China, <sup>3</sup> Key Lab of Fruit and Vegetable Processing, Ministry of Agriculture, Beijing, China

The objective of this study was to investigate the inactivation mechanism of Bacillus subtilis spores by high pressure CO<sup>2</sup> (HPCD) processing. The spores of B. subtilis were subjected to heat at 0.1 MPa or HPCD at 6.5−20 MPa, and 64−86◦C for 0−120 min. The germination, the permeability of inner membrane (IM) and cortex, the release of pyridine-2, 6-dicarboxylic acid (DPA), and changes in the morphological and internal structures of spores were investigated. The HPCD-treated spores did not lose heat resistance and their DPA release was lower than the inactivation, suggesting that spores did not germinate during HPCD. The flow cytometry analysis suggested that the permeability of the IM and cortex of HPCD-treated spores was increased. Furthermore, the DPA of the HPCD-treated spores were released in parallel with their inactivation and the fluorescence photomicrographs showed that these treated spores were stained by propidium iodide, ensuring that the permeability of IM of spores was increased by HPCD. The scanning electron microscopy photomicrographs showed that spores were crushed into debris or exhibited a hollowness on the surface, and the transmission electron microscopy photomicrographs exhibited an enlarged core, ruptured and indistinguishable IM and a loss of core materials in the HPCD-treated spores, indicating that HPCD damaged the structures of the spores. These findings suggested that HPCD inactivated B. subtilis spores by directly damaging the structure of the spores, rather than inducing germination of the spores.

#### Edited by:

Michael Gänzle, University of Alberta, Canada

#### Reviewed by:

Zuzana Hruska, Mississippi State University, USA Sara Spilimbergo, University of Padova, Italy

> \*Correspondence: Xiaojun Liao liaoxjun@hotmail.com

#### Specialty section:

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

Received: 26 June 2016 Accepted: 25 August 2016 Published: 07 September 2016

#### Citation:

Rao L, Zhao F, Wang Y, Chen F, Hu X and Liao X (2016) Investigating the Inactivation Mechanism of Bacillus subtilis Spores by High Pressure CO2. Front. Microbiol. 7:1411. doi: 10.3389/fmicb.2016.01411 Keywords: high pressure CO2, inactivation, Bacillus subtilis spores, mechanism, inner membrane damage

# INTRODUCTION

Spores of a number of Bacillus and Clostridium species are extremely resistant to a variety of severe stresses including extreme temperatures (steam at 121◦C), desiccation, chemicals and radiation because of their unique structures (Setlow, 1995, 2006). These spores are common agents of food spoilage, foodborne illnesses, and detrimental changes to the organoleptic quality of food (Brown, 2000; Logan, 2012), which make them a significant problem in the food industry. Consequently, there is much interest in methods that inactivate these spores as well as the inactivation mechanisms.

Traditionally, spores are inactivated by heat at extremely high temperature (121◦C or higher) (Block, 2001). It is known that heat inactivates spores by damaging one or more proteins, most

likely some enzymes involved in metabolism (Coleman et al., 2007, 2010). However, the identity of this key protein or proteins is not known. Although high temperature can effectively inactivate spores, it also can impart undesirable organoleptic changes and cause some detrimental effects to the nutritional quality of heat-sensitive food. Consequently, nonthermal technologies such as irradiation, pulsed electric fields, pulsed magnetic fields, high hydrostatic pressure (HHP), and high pressure CO<sup>2</sup> (HPCD), have been proposed as foodprocessing methods. Among these technologies, the HHP is the most studied, and shows the potential to inactivate bacterial spores when combined with mild temperatures (Black et al., 2007a; Reineke et al., 2013b). However, the large investment cost due to the extremely high processing pressure and the non-continuous nature of the process hamper the industrial applications and commercialization of the HHP (Devlieghere et al., 2004; Estrada-Girón et al., 2005; Garcia-Gonzalez et al., 2007; Perrut, 2012).

The inactivation effect of HPCD was first shown in 1951 on Escherichia coli (Fraser, 1951). In recent years, HPCD treatment has been proposed as an alternative non-thermal pasteurization technique for foods because of its environmentally benign nature (CO<sup>2</sup> is nontoxic), as well as the much lower pressure (generally lower than 30 MPa) compared with the high pressure (100−600 MPa) employed in the HHP processing (Garcia-Gonzalez et al., 2007). Previous studies indicate that HPCD at less than 30 MPa and at 20 to 40◦C can effectively inactivate the vegetative forms of pathogenic and spoilage bacteria, yeasts, and molds, but has no effect on bacterial spores (Spilimbergo and Bertucco, 2003; Damar and Balaban, 2006; Zhang et al., 2006b; Garcia-Gonzalez et al., 2007; Perrut, 2012; Rao et al., 2015a). Several studies suggested that cycled-pressure HPCD or HPCD at temperature ≥60◦C can effectively inactivate bacterial spores, and different inactivation mechanisms have been proposed (Spilimbergo et al., 2002; Spilimbergo and Bertucco, 2003; Spilimbergo et al., 2003; Bae et al., 2009).

One possible inactivation mechanism is that the spores are first activated and germinated, and then inactivated during the HPCD treatment. As reported in a previous study (Spilimbergo et al., 2002), a tyndallization effect (approximately 3.5-log reduction) is observed in Bacillus subtilis spores as a result of cycled-pressure (30 cycles/h, 1P = 8 MPa) HPCD treatment at 15 MPa and 36◦C for 30 min, and the inactivation mechanism is explained as follows: the initial pressure cycles induce spore activation such that germination takes place during the holding time between two different cycles. The germinated spores are inactivated during the cycles that follow (Spilimbergo et al., 2002). It is hypothesized that a combined treatment of temperature (at least 60◦C) and CO<sup>2</sup> induces shock in the spore structure, which leads to their activation. The spores start their germination during a long contact time of HPCD treatment. The geminated spores became more sensitive to the antimicrobial effect of CO2, which ultimately results in their inactivation (Spilimbergo and Bertucco, 2003; Spilimbergo et al., 2003). However, there is no data to support the claim that the spores are truly germinated during HPCD treatment in these studies. Although it has been reported that 40% of Bacillus coagulans spores and 70% of Bacillus licheniformis spores are germinated by HPCD at 35◦C for 120 min (Furukawa et al., 2004), it is not clear whether the spores can be induced to germinate by HPCD at temperature ≥60◦C.

Another possible inactivation mechanism is that HPCD inactivates spores by directly damaging the spore structures. A previous study used scanning electron microscopy (SEM) and energy-filtering transmission electron microscopy (EF-TEM) methods to investigate the morphological changes of Alicyclobacillus acidoterrestris spores treated by HPCD at 10 MPa and 70◦C for 30 min (Bae et al., 2009). The SEM photomicrographs revealed that the treated spores were crushed and exhibit a high degree of hollowness on the surface. The EF-TEM photomicrographs showed an enlarged periplasmic space and a loss of cytoplasm in the treated spores. Based on these images, the authors concluded that HPCD directly affected and inactivated the A. acidoterrestris spores (Bae et al., 2009). However, the germination of spores was not examined in the study, and there are no data to exclude the possibility that spores may first germinate, and later be inactivated by damage to their structures. Moreover, it is also not clear how HPCD damages the spore structures, which needs further research. Recently, the flow cytometry method (FCM) has been used to assess the structural changes of high pressure treated spores stained with SYTO 16 and propidium iodide (PI) (Mathys et al., 2007; Reineke et al., 2013a). It is reported that the dormant and decoated spores are not stained by SYTO16. However, after hydrolysis of the spore cortex, the membrane-permeable SYTO 16 can permeate into the spore core and exhibit green fluorescence by binding to the nucleic acid (Black et al., 2005; Black et al., 2007b). Therefore, SYTO 16 can be designed as an indicator of the damage of the spores' cortex. Meanwhile, PI red fluorescence nucleic acid dye is membrane impermeable, and used to identify the damage of the spores' inner membrane (IM) (Mathys et al., 2007).

In our previous work, B. subtilis spores were inactivated by HPCD at temperatures higher than 82◦C, and exhibited nonlinear inactivation curves with a shoulder and a log-linear region (Rao et al., 2015b). The inactivation included two steps as follows: the spores first lost their resistance at the shoulder regions, and then were inactivated at the log-linear regions. The loss of resistance of the spores at the shoulder regions was explained as follows: (i) spores were induced to germinate by HPCD and lost resistance; (ii) spore structures including cortex and IM as well as some important proteins crucial to spore germination and outgrowth were damaged by HPCD. Therefore, more work is needed to determine the real reasons. In the current work, we examined the spore germination by determining the loss of heat resistance (Furukawa et al., 2004; Shah et al., 2008) and the pyridine-2, 6-dicarboxylic acid (DPA) release of HPCDtreated spores (Yi and Setlow, 2010; Setlow et al., 2016). We have also investigated the permeability of the cortex and IM of spores by the flow cytometry method (FCM) and confocal laser scanning microscopy (CLSM), as well as the release of the core material of DPA. Moreover, we have visually demonstrated changes to the surface and internal structure of spores by SEM and TEM.

# MATERIALS AND METHODS

fmicb-07-01411 September 2, 2016 Time: 12:59 # 3

# Strain and Spore Preparation

Bacillus subtilis 168 was obtained from the China General Microbiological Culture Collection Center (Beijing, China), and the sporulation was carried out as previously described (Rao et al., 2015b). Overnight cultures of Bacillus strain grown in nutrient broth were transferred to sporulation agar plates, nutrient agar containing 50 µg/mL Mn2+. After 1 week incubation at 37◦C, the spores were harvested in a sterile flask by flushing the surface of the culture with sterile distilled water and scrapping the surface with sterile glass microscope slide. The spores collected were washed three times by centrifugation at 7,000 × g and 4◦C for 15 min, resuspended in sterile distilled water with a concentration of approximately 10<sup>9</sup> CFU/ml, and stored at 4◦C until they were used. All spores (>99%) used in this work were free of growing and sporulating cells, germinated spores, and cell debris, as determined with a BX45-72P15 phase contrast microscope (Olympus, Japan). The concentration of the spore suspension was adjusted to approximately 10<sup>7</sup> CFU/mL before treatments.

# HPCD Treatment

The treatment conditions were shown in **Table 1**. HPCD treatment was performed with a batch HPCD system (Liao et al., 2007). For each experiment, 20 mL of the spores suspended in sterile distilled water was transferred to a 50 mL sterile glass tube and the tube was covered with a plastic film with a 0.22 µm membrane filter in the center of aeration to prevent microbial contamination. As the pressure vessel of the HPCD system reached the experimental temperature (64−66◦C or 84−86◦C), the sample tubes were placed in the pressure vessel. Next, the vessel was pressurized by the plunger pump to 6.5 or 20 MPa within 0.1 or 2.5 min, respectively. After holding for required treatment time, the depressurization was performed by opening the pressure relief valve at the CO<sup>2</sup> outlet on the pressure vessel. The depressurization time was 0.5 min or 2.5 min for 6.5 or 20 MPa, respectively. After HPCD, the sample tubes were taken out from the vessel and analyzed immediately. The CO<sup>2</sup> purity was 99.5% in all the experiment treatments.

The inactivation of the spore suspensions by heat at 86◦C was carried out at 0.1 MPa without the addition of CO<sup>2</sup> using a water bath. Similar to HPCD treatment, 20 mL of the spore suspension was transferred to a 50 mL sterile glass tube and immersed in a water bath equilibrated at 86◦C for 0−30 min. The experiment was done in triplicate. After treatment, the sample tubes were taken out and immediately analyzed.

# Enumeration of Surviving Spores

The number of surviving spores was determined by the viable plate count method. Each sample was serially (1:10) diluted with sterile distilled water and pour-plated on nutrient agar in duplicate. The plates were incubated at 37◦C for 24 h. After incubation, the colonies were counted.

# Measurement of Germination

As it was known that the spores would lose heat resistance and release almost all the DPA after germination (Setlow, 2003), the germination of spores during HPCD treatment was investigated by determining the loss of heat resistance and DPA release (Shah et al., 2008; Setlow et al., 2016). The spores treated by HPCD were subjected to wet heat at 80◦C for 20 min, then diluted and pour-plated on nutrient agar. Following incubation at 37◦C for 24 h, the colonies were counted. Germination was expressed as the change in colonies before and after exposure to heat.

The DPA release was measured using the fluorescence method (Hindle and Hall, 1999). The treated spores were centrifuged at 7,000 × g and 4◦C for 10 min, and assaying DPA in the supernatant fluid was carried out by its fluorescence with Tb3<sup>+</sup> in a 96-well plate. One hundred microliter of supernatant fluid were added to 100 µL 20 µmol/L terbium (III) chloride hexahydrate buffered with 1 mol/L acetic acid at pH 5.6. All the samples were analyzed with a microplate reader (Multiskan MK3, Thermo, USA). Samples were excited at 270 nm, and emission spectra were collected at 545 nm. The total amount of DPA in each individual batch was determined after autoclaving at 121◦C for 20 min (Zhang et al., 2006a), which was used as a positive control while the one in untreated spores was used as a negative control. HPCD-induced DPA release was calculated by the following equation:

$$\text{DPA\%} = \frac{\text{F}\_1 - \text{F}\_0}{\text{F}\_2 - \text{F}\_0} \tag{1}$$

Where F0, F1, and F<sup>2</sup> were the fluorescence intensity of untreated spores, HPCD treated spores, and autoclaved spores, respectively.

# Flow Cytometry Analysis

Samples for flow cytometry were prepared with two DNA staining dyes according to a reported method (Reineke et al., 2013a). The dyes propidium iodide (PI) (Sigma−Aldrich) and SYTO 16 (Invitrogen) are both able to stain DNA. The membrane-permeable SYTO 16 acts as an indicator for cortex damage (Black et al., 2005), and the membrane-impermeable PI indicates the IM damage (Mathys et al., 2007). The treated spore suspensions were adjusted to concentrations of about 10<sup>7</sup>


spores/mL in sterile distilled water. The concentration of the fluorescent dyes in the spore suspensions were 15 µmol/L PI and 0.5 µmol/L SYTO 16. Afterward, the samples were stored in the dark at room temperature for 45 min (Pappas et al., 2015).

Stained samples were then analyzed with an Accuri C6 (BD Accuri Cytometer Inc., USA) flow cytometry equipped with a 488 nm, 50 mW laster. SYTO 16 fluorescence was quantified with the FL1 detector at 530 ± 15 nm. PI fluorescence was quantified with the FL2 detector at 585 ± 20 nm. The forward scatter threshold was set at 5000 to ensure that the small spores were not omitted as events. Spores were analyzed at a nominal flow rate of 14 µL/min, with a stream core diameter of 10 µm. All samples were evaluated after 30000 events had been recorded. The live gate, in which spores are all alive and cannot be stained with PI or SYTO16, was based on untreated spores as negative controls, while the dead gate, where the spores are all dead and can be stained with PI or SYTO16, was based on the spores exposed to 121◦C for 20 min as positive controls.

# Fluorescence Analysis

The untreated, autoclaved (121◦C, 20 min), heat treated (86◦C, 20 min), and HPCD treated, spore suspensions were adjusted to about 10<sup>7</sup> spores/mL in sterile distilled water and stained with 15 µmol/L PI for 45 min at room temperature. The stained spore samples were imaged with a Zeiss LSM710 confocal laser scanning microscope (Zeiss, Germany) with 100× oil lens. The fluorescent photomicrographs were acquired with the Zeiss AIM image browser software (Zeiss, Germany).

FIGURE 1 | Loss of heat resistance and pyridine-2, 6-dicarboxylic acid (DPA) release of Bacillus subtilis spores by high pressure CO<sup>2</sup> (HPCD) at different conditions. (A) 6.5 MPa, 64−66◦C, 0−60 min; (B) 20 MPa, 64−66◦C, 0−60 min; (C) 6.5 MPa, 84−86◦C, 0−30 min; (D) 20 MPa, 84−86◦C, 0−30 min. ( ) spore inactivation by HPCD; (#)spore inactivation by HPCD <sup>+</sup> <sup>80</sup>◦C, 20 min; (N) DPA release of spores treated by HPCD.

# Spore Preparation for SEM and TEM

fmicb-07-01411 September 2, 2016 Time: 12:59 # 5

For both SEM and TEM analysis, spore suspensions were centrifuged (10,000 × g) and prefixed in 2.5% glutaraldehyde (Sigma−Aldrich) overnight at room temperature, rinsed three times in 0.1 mol/L phosphate-buffered saline (PBS) for 15 min and centrifuged again (10,000 × g). The spore pellets were postfixed in 1% osmium tetroxide (Sigma-Aldrich) for 90 min and rinsed three times in 0.1 mol/L PBS for 15 min, and subsequently dehydrated with ethanol series (50, 70, 80, 90 and 100%). For SEM, the dehydrated spore samples were stored at −20◦C for 20 min and subjected to critical point drying. The spore samples were sputter coated in about 12 nm of gold and palladium under vacuum, and subsequently analyzed by SEM (FEI Quanta 200, FEI, Czech Republic). For TEM, the dehydrated spore samples were embedded in epoxy resin and kept at 37◦C overnight followed by 60◦C for 24 h. The resin blocks were cut into ultrathin sections of 70 nm with an ultramicrotome (Lecia EM UC6, Leica, Germany) and stained with 3% aqueous lead citrate and 3% aqueous uranyl acetate. Finally, the spore samples were examined by TEM (H-7650B, Hitachi, Japan).

# Data Analysis

Flow cytometry data were analyzed using the FlowJo version 7.6.1 software (FlowJo). Analysis of variance (ANOVA) was carried out by using software PASW statistic 18 (SPSS, USA). ANOVA tests were carried out for statistical significance of group differences at α = 0.05 level. All experiments were carried out in triplicate.

# RESULTS

# Germination Detection

**Figures 1A–D** show the germination of the HPCD-treated spores estimated by the loss of heat resistance. The inactivation of the HPCD-treated spores increased with increasing time. After HPCD treatment at the temperature of 64−66◦C for 30 min (**Figures 1A,B**), the inactivation was 23.3% at 6.5 MPa and 44.0% at 20 MPa. When the temperature was increased to 84- 86◦C (**Figures 1C,D**), the inactivation was increased to 66.6% at 6.5 MPa and 95.2% at 20 MPa. After all the HPCD-treated spores were subjected to heat treatment at 80◦C for 20 min, there were no changes in the inactivation of the treated spores (P > 0.05), indicating that there were no germinated spores in the HPCDtreated spores. The DPA release of the HPCD-treated spores was also determined. Similarly, the DPA release was increased with increasing the time. After HPCD treatment at 64−66◦C for 30 min (**Figures 1A,B**), the DPA release was 4.3% at 6.5 MPa and 10.0% at 20 MPa. When the temperature was increased to

84−86◦C (**Figures 1C,D**), the DPA release was 27.2% at 6.5 MPa and 70.3% at 20 MPa. Obviously, the DPA release was much lower than the corresponding inactivation (P < 0.05), suggesting that a portion of the inactivated spores, rather than all of them, released the DPA.

# FCM Analysis

fmicb-07-01411 September 2, 2016 Time: 12:59 # 6

As the spores were not germinated during the inactivation process by HPCD, the spores were presumed to be inactivated by damage to their structures. The permeability of the IM (**Figure 2**) and cortex (**Figure 3**) of spores were tested by FCM. The untreated spores were used as negative control, and the heat-treated spores at 121◦C for 20 min (>7 log-reduction) were used as positive control. Meanwhile, the heat-treated spores at 86◦C for 20 min (0.067 log-reduction) were used to do the comparison analysis with the HPCD-treated spores at 20 MPa and 84-86◦C for 0 min (0.15 log-reduction), 10 min (0.45 log-reduction), and 20 min (0.85 log-reduction). As shown in **Figure 2**, the FCM histograms of red fluorescence distribution of the spores stained by PI were divided into two areas, M1 and M2. According to the histogram of the untreated spores (**Figure 2A**), M2 was the negative area in which the spores had intact IM and were not stained by PI, while M1 was the positive area, indicating that the IM was damaged, and the spores were stained by PI. Compared with the untreated spores, the fluorescence distribution of the heat-treated spores at 121◦C for 20 min moved increasingly towards M1 (**Figure 2B**), indicating that the IM of the spores were damaged. Meanwhile, there was only a very slight increase of the fluorescence in M1 for the heat-treated spores at 86◦C for 20 min (**Figure 2C**), suggesting that these heattreated spores have intact IM. For the HPCD-treated spores at 20 MPa and 84−86◦C (**Figures 2D–F**), the fluorescence in M1 was increased, and the fluorescence distribution moved towards M1 with increased the time, indicating that the IM of spores was damaged during the HPCD treatment. Similar results are shown in **Figure 3**. According to the FCM histogram of green fluorescence distribution of the untreated spores (**Figure 3A**), M2 was the negative area in which the spores had intact cortex, and were not stained by SYTO 16, while M1 was the positive area, indicating that the cortex was damaged, and the spores were stained by SYTO 16. Compared with the untreated spores, the heat-treated spores at 121◦C for 20 min showed a marked increase of the fluorescence in the M1 (**Figure 3B**), indicating damage to the cortex of spores. For the heat-treated spores at 86◦C for 20 min (**Figure 3C**), there was no change of the fluorescence in M1, suggesting that heat treatment at 86◦C for

FIGURE 5 | Fluorescent photomicrographs of B. subtilis spores treated by HPCD at 20 MPa, 84−86◦C for 0−20 min and stained by PI. (a) untreated; (b) autoclaved at 121◦C for 20 min; (c) heat treated at 86◦C for 20 min; HPCD treated at 20 MPa, 84−86◦C for 0 min (d), 10 min (e), 20 min (f). The samples were imaged with 100× oil lens as detailed in the Section "Materials and Methods".

20 min did not damage the cortex of spores. For the HPCDtreated spores at 20 MPa and 84−86◦C (**Figures 3D–F**), the fluorescence distribution moved towards M1 with increased time, indicating that the cortex of spores was gradually damaged during HPCD treatment.

# DPA Release

Given the results of the FCM analysis, further tests were needed to ensure the damage of the IM of spores. Thus, the release of DPA, main core material in spores, was determined after HPCD treatment at 20 MPa and 84−86◦C for 0−30 min. Meanwhile, the heat-treated spores treated at 86◦C for 0−30 min were tested and compared with the HPCD-treated spores. With increased time, the inactivation of the HPCD-treated spores was increased, and the maximum value was 99.0% (**Figure 4A**). For the heat-treated spores, the inactivation was slowly increased with increased time, and the maximum value was 29.8% (**Figure 4A**). Similar to the results of the inactivation, the release of DPA (**Figure 4B**) for the HPCD-treated spores was also increased with increased time, and the maximum value was 86.8%, which indicated that the IM of most spores were damaged during the inactivation process by HPCD. For the heat-treated spores, the release of DPA (**Figure 4B**) was slightly increased with increased time, and the maximum value was 15.2%, suggesting that most spores maintained an intact IM.

# CLSM Photomicrographs

In addition to testing the DPA release, CLSM was used to image the PI stained spores and examine the permeability of the IM of spores. The untreated spores with intact IM were not stained by PI (**Figure 5a**) while the autoclaved spores were all dead and stained red by PI (**Figure 5b**). The heat-treated (86◦C, 20 min) spores (0.067 log-reduction) were not stained (**Figure 5c**), indicating these spores retained an intact IM. For the spores treated by HPCD at 20 MPa and 86◦C, most of the spores were not stained or stained peripherally at 0 min (0.15 log-reduction) (**Figure 5d**) and 10 min (0.45 log-reduction) (**Figure 5e**), while completely stained at 20 min (0.85 logreduction) (**Figure 5f**), further ensuring that the IM of the spores was gradually damaged during the HPCD treatment.

# SEM and TEM Photomicrographs

To visually confirm damage to the spore structures, SEM and TEM were used to examine the morphological changes of the spores during the inactivation process by HPCD treatment at 20 MPa and 84−86◦C for 10 min (**Figures 6B** and **7B**), 20 min (85.9% inactivation) (**Figures 6C** and **7C**) and 30 min (99.0% inactivation) (**Figures 6D** and **7D**). The SEM photomicrographs showed that the HPCD-treated spores were crushed into debris, or exhibited a high degree of hollowness on the surface with increased time (**Figures 6B–D**), whereas the untreated

spores were intact planiform ellipsoids (**Figure 6A**). In the TEM photomicrographs, significant changes of the core area and the morphological structures of the HPCD-treated spores (**Figures 7B–D**) were observed compared with the untreated spores (**Figure 7A**). The treated spores showed an enlarged core and a loss of the core materials, and the IM of spores was ruptured (**Figures 7C,D**). Moreover, the coat of the treated spores was deformed (**Figures 7B–D**). These SEM and TEM photomicrographs further confirmed that HPCD damaged the spore structures.

# DISCUSSION

In this study, the germination of spores was examined during the inactivation process by HPCD at different conditions (**Figures 1A–D**). As it is known that spores will lose heat resistance and release almost all the DPA after germination (Setlow, 2003; Yi and Setlow, 2010; Setlow et al., 2016), the loss of heat resistance and DPA release were used as indicators for spore germination (Shah et al., 2008; Setlow et al., 2016). Our results showed that there was no change in the population of the HPCDtreated spores after exposure to heat treatment at 80◦C for 20 min (**Figures 1A–D**), suggesting that no germinated spores existed in the HPCD-treated spore population. Moreover, the DPA release was always lower than the inactivation of spores (**Figures 1A–D**), indicating that only some of the inactivated spores released DPA rather than all (Coleman et al., 2007). Thus, the inactivated spores did not undergo the germination process during the inactivation process by HPCD treatment. Indeed, there was evidence to support the premise that the spore germination may be suppressed under HPCD conditions. A previous report indicated that the germination of Bacillus cereus, Clostridium sporogenes, and Clostridium perfringens spores was inhibited completely by carbon dioxide at atmospheric pressure, 1.0 and 2.5 MPa, respectively (Enfors and Molin, 1978), and this inhibition of spore germination might be caused by low pH under the HPCD conditions. It was also reported that the optimum pH for pressure-induced germination at moderate pressure of 100 MPa, just like the nutrient-induced germination, was close to neutral (pH 7.0), and lowering the pH would strongly inhibit the germination (Sale et al., 1970; Bender and Marquis, 1982; Wuytack and Michiels, 2001). Another study found that reduction of pH from 7 to 5.5 completely inhibited spore germination of Clostridium botulinum 12885A (Blocher and Busta, 1985a). Similar results were reported for the C. botulinum and B. cereus spores at pH 4.5 (Blocher and Busta, 1985b), B. cereus and B. subtilis spores at pH 3.6 (Yi and Setlow, 2010). It was assumed that the inhibition of the spore germination in an acidic environment was induced by the inhibition of the germinant binding to the germinant receptors (GRs), which would block the commitment step of the germination (Blocher and Busta, 1985b; Yi and Setlow, 2010). This inhibition of the germinant binding was probably due to the protonation of a

function group in or near the GRs (Blocher and Busta, 1985b). In our study, the HPCD treatment was conducted in a pure water−CO<sup>2</sup> system, in which the pH was a strong function of pressure, but a weak function of temperature (Meyssami et al., 1992). With increased pressure, the pH decreased to approximately 3.0 at 5.5 MPa and then remained constant (Meyssami et al., 1992). Obviously, spore germination would be inhibited at such a low pH value. Given our results and those of others, it was concluded that the inactivation of spores by HPCD was not attributed to spore germination.

As the inactivation of spores by HPCD was not due to the spore germination, the question of how HPCD inactivated spores, still remains. In fact, as the CO<sup>2</sup> had smaller molecular weight and higher penetrability under the HPCD state than did the spore germinants (e g., L-valine, L-alanine, D-glucose), which were able to penetrate into to the spore cells and bind to the germinant receptors (GRs) in the IM (Setlow, 2003), the CO<sup>2</sup> of HPCD could theoretically penetrate into the spores cells and act on the IM. In this study, the structural changes of spores during HPCD treatment were examined. The FCM analysis indicate that the permeability of the IM of HPCD-treated spores and cortex increased (**Figures 2** and **3**), which was due to the damage to these two structures by HPCD. This damage was confirmed by the large release of the main core material of DPA (**Figure 4**) and the CLSM photomicrographs (**Figure 5**) of spores after HPCD treatment. Moreover, the SEM and TEM photomicrographs further visually confirm the damage of the morphological structures of the HPCD-treated spores (**Figures 6** and **7**). Although our results indicated that the HPCD-inactivated spores exhibited structural damage, it was necessary to determine which specific structure was damaged to the point where it caused the spore inactivation. In the FCM results for the HPCD-treated spores at 20 MPa and 84−86◦C for 0 min (29.4% inactivation) (**Figures 2D** and **3D**), the red fluorescence in M1 was significantly increased (**Figure 2D**) compared with the untreated spores (**Figure 2A**), while the green fluorescence in M1 was not changed (**Figure 3D**) compared with the untreated ones (**Figure 3A**). This indicates that the damage to the spore IM occurred prior to inactivation and the damage to the spore cortex in large portion of the spores. Therefore, the damage of the IM of spores would be the reason for spore inactivation by HPCD. Similar results were reported in previous studies. It has been reported that HPCD + 0.02% H2O<sup>2</sup> treatment at 27.5 MPa and 40◦C for 240 min effectively inactivated Bacillus atrophaeus (Zhang et al., 2006a) and Bacillus anthracis (Zhang et al., 2007) spores by >6 log-reduction. The TEM photomicrographs, DPA analysis and BacLight fluorescence results in these studies indicated that HPCD damaged the IM of spores, which allowed the penetration of H2O<sup>2</sup> into the core and subsequent oxidation of the vital structures that caused spore death (Zhang et al., 2006a, 2007). A recent study achieved more than 6 log inactivation of B. subtilis spores by HPCD + 0.0035−0.0055% peracetic acid (PAA) at 9.8 MPa and 35◦C for 25 min, and the authors suggested that the inactivation of the spores was induced by damaging the IM of spores (Setlow et al., 2016). It is known that HPCD has the ability to extract constituents from the cells and cell membrane, modify the structure of cell membrane and damage the proteins, especially enzymes (Garcia-Gonzalez et al., 2007; Hu et al., 2010). Thus, the lethal effect of the IM damage by HPCD to the spores could be explained as follows: (i) HPCD treatment modified and increased the permeability of the IM of spores. Then, during the spore germination (if the spores are able to germinate) and outgrowth process, the germinated spores plasma membrane, which is derived from the IM of spores, became leaky and were unable to carry out proper energy metabolism (Setlow et al., 2016); (ii) the crucial proteins related to the spore germination and outgrowth in the IM, including the GRs that recognize nutrient germinants, the SpoVA proteins essential for DPA release, and the cortex-lytic enzymes that degrade cortex peptidoglycan (Setlow, 2013), were damaged by HPCD treatment, which in turn, may have induced lethal effects in the spores (Setlow et al., 2016).

Based on our results, HPCD inactivates spores by directly damaging the structures, especially the IM, rather than inducing spore germination. However, it is still not clear how the HPCD precisely acts on and damages the IM of spores. Further studies are required to examine the changes of the properties of the IM, and to identify the specific proteins whose damage induced spore inactivation by the HPCD treatment.

# AUTHOR CONTRIBUTIONS

LR carrying out the experiments and writing the manuscript. FZ giving advice and assistance during the experiments. YW giving advice and assistance during the experiments. FC reviewing the manuscript and giving advice on the manuscript. XH reviewing the manuscript and giving advice on the manuscript. XL designing the experiments, reviewing and revising the manuscript.

# FUNDING

This work was supported by Key Project of Chinese Ministry of Education (No. 113011A), Key Project of National Natural Science Foundation of China (NSFC) (No. 31530058 and 21025099), Chinese Universities Scientific Fund (No. 2015SP004), and Beijing Training Project for the Leading Talents in S & T (No. Z151100000315032).

# ACKNOWLEDGMENT

We are grateful to Prof. Peter Setlow for the advice on the manuscript.

# REFERENCES

fmicb-07-01411 September 2, 2016 Time: 12:59 # 11


carbon dioxide. Langmuir 27, 909–916. doi: 10.1021/la10 3482x


Zhang, J., Dalal, N., Gleason, C., Matthews, M. A., Waller, L. N., Fox, K. F., et al. (2006a). On the mechanisms of deactivation of Bacillus atrophaeus spores using supercritical carbon dioxide. J. Supercrit. Fluids 38, 268–273. doi: 10.1016/j.supflu.2006.02.015

Zhang, J., Dalal, N., Matthews, M. A., Waller, L. N., Saunders, C., Fox, K. F., et al. (2007). Supercritical carbon dioxide and hydrogen peroxide cause mild changes in spore structures associated with high killing rate of Bacillus anthracis. J. Microbiol. Methods 70, 442–451. doi: 10.1016/j.mimet.2007.05.019

Zhang, J., Davis, T. A., Matthews, M. A., Drews, M. J., LaBerge, M., and An, Y. H. (2006b). Sterilization using high-pressure carbon dioxide. J. Supercrit. Fluids 38, 354–372. doi: 10.1016/j.supflu.2005.05.005

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

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

#### Edited by:

Abram Aertsen, Katholieke Universiteit Leuven, Belgium

#### Reviewed by:

Raivo Vilu, Tallinn University of Technology, Estonia Marc Heyndrickx, Instituut voor Landbouw en Visserijonderzoek, Belgium

#### \*Correspondence:

Stanley Brul s.brul@uva.nl

#### †Present address:

Christoph J. Blohmke, Oxford Vaccine Group, Centre for Clinical Vaccinology and Tropical Medicine, The Churchill Hospital, Oxford, UK Hendrik Folkerts, Department of Experimental Hematology, University Medical Center Groningen, University of Groningen, Groningen, Netherlands Anna Zakrzewska, Crucell Vaccine Institute, Janssen Centre of Excellence for Immunoprophylaxis, Leiden, Netherlands Alexander Ter Beek, Dutch DNA Biotech, Utrecht, Netherlands. ‡These authors have contributed equally to this work.

#### Specialty section:

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

Received: 16 May 2016 Accepted: 30 September 2016 Published: 21 October 2016

#### Citation:

van Beilen J, Blohmke CJ, Folkerts H, de Boer R, Zakrzewska A, Kulik W, Vaz FM, Brul S and Ter Beek AT (2016) RodZ and PgsA Play Intertwined Roles in Membrane Homeostasis of Bacillus subtilis and Resistance to Weak Organic Acid Stress. Front. Microbiol. 7:1633. doi: 10.3389/fmicb.2016.01633

# RodZ and PgsA Play Intertwined Roles in Membrane Homeostasis of Bacillus subtilis and Resistance to Weak Organic Acid Stress

Johan van Beilen<sup>1</sup> , Christoph J. Blohmke<sup>1</sup>† , Hendrik Folkerts<sup>1</sup>† , Richard de Boer<sup>1</sup> , Anna Zakrzewska<sup>1</sup>† , Wim Kulik<sup>2</sup> , Fred M. Vaz<sup>2</sup> , Stanley Brul<sup>1</sup> \* ‡ and Alexander Ter Beek<sup>1</sup>†‡

<sup>1</sup> Laboratory for Molecular Biology and Microbial Food Safety, Swammerdam Institute for Life Sciences, University of Amsterdam, Amsterdam, Netherlands, <sup>2</sup> Laboratory Genetic Metabolic Diseases, Academic Medical Center, University of Amsterdam, Amsterdam, Netherlands

Weak organic acids like sorbic and acetic acid are widely used to prevent growth of spoilage organisms such as Bacilli. To identify genes involved in weak acid stress tolerance we screened a transposon mutant library of Bacillus subtilis for sorbic acid sensitivity. Mutants of the rodZ (ymfM) gene were found to be hypersensitive to the lipophilic weak organic acid. RodZ is involved in determining the cell's rod-shape and believed to interact with the bacterial actin-like MreB cytoskeleton. Since rodZ lies upstream in the genome of the essential gene pgsA (phosphatidylglycerol phosphate synthase) we hypothesized that expression of the latter might also be affected in rodZ mutants and hence contribute to the phenotype observed. We show that both genes are co-transcribed and that both the rodZ::mini-Tn10 mutant and a conditional pgsA mutant, under conditions of minimal pgsA expression, were sensitive to sorbic and acetic acid. Both strains displayed a severely altered membrane composition. Compared to the wild-type strain, phosphatidylglycerol and cardiolipin levels were lowered and the average acyl chain length was elongated. Induction of rodZ expression from a plasmid in our transposon mutant led to no recovery of weak acid susceptibility comparable to wild-type levels. However, pgsA overexpression in the same mutant partly restored sorbic acid susceptibility and fully restored acetic acid sensitivity. A construct containing both rodZ and pgsA as on the genome led to some restored growth as well. We propose that RodZ and PgsA play intertwined roles in membrane homeostasis and tolerance to weak organic acid stress.

Keywords: weak organic acids, acetic acid, sorbic acid, Bacillus subtilis, rod Z, pgs A, membrane compositional fluctuations

# INTRODUCTION

Weak organic acids (e.g., sorbic-, acetic-, and benzoic- acid) are commonly used preservatives in the food industry since they inhibit the growth of spoilage bacteria, yeasts, and molds (Brul and Coote, 1999; Davidson and Harrison, 2002; Beales, 2004; Brul and Ter Beek, 2010). The acids are most effective at pH conditions close to or below their pK<sup>a</sup> value. Depending on the lipophilic nature of the compound, the neutral undissociated form of the molecule is able to dissolve in and

diffuse over the membrane. The hydrophobic tail of, e.g., lipophilic sorbic acid, can also more permanently insert into the membrane perturbing its structure and interfering with the function of proteins (Sheu and Freese, 1972; Stratford and Anslow, 1998; Chu et al., 2009). Inside the cell the acid dissociates and releases protons to a large extend, since most microorganisms exhibit an intracellular pH (pHi) near neutrality. Consequently, the proton gradient dissipates and, depending on the buffering capacity of the cell, the cytosol may acidify, affecting oxidative phosphorylation, the transport of nutrients, and a number of other metabolic functions (Bauer et al., 2003; Cotter and Hill, 2003; Brul and Ter Beek, 2010; van Beilen et al., 2014). The generation of reactive oxygen species has been described in Saccharomyces cerevisiae and recently has been detected in Bacillus cereus upon weak acid stress (Piper, 1999; Mols et al., 2010), which could damage iron-sulfur clusters, proteins, and DNA. Finally, it has been shown that the accumulation of the anion in the cell can cause a rise in osmolarity and affect cytosolic enzymes (Azukas et al., 1961; York and Vaughn, 1964; Russell, 1992).

Bacillus subtilis is one of the organisms that causes food spoilage and its growth is inhibited by weak organic acids (Eklund, 1983). Previously, we preformed time-resolved transcriptome analysis of B. subtilis sub-lethally stressed with potassium sorbate (KS) to elucidate the sorbic acid adaptive responses of this organism at the molecular level (Ter Beek et al., 2008). The results indicated that sorbic acid induces responses normally seen upon nutrient limitation and alters the expression of many cell envelope-related genes. Upregulation of fatty acid biosynthesis (fab) genes and BkdR-regulated genes indicated the synthesis of longer and more branched lipids. We proposed an adaptation in the fatty acid composition of the B. subtilis plasma membrane as a stress response mechanism, since the sensitivity of cells toward the fab inhibitor cerulenin was reduced in the presence of sorbic acid (Ter Beek et al., 2008). Gene groups regulated by extracytoplasmic function sigma factors SigW and SigX (controlling functions associated with the cell surface and transport) (Mascher et al., 2007) were downregulated, indicating a reduction in cell envelope remodeling and, consequently, a change in cell envelope composition. In addition, similar analyses have been performed by us in later experiments for acetic acid stressed cells in comparison to sorbic acid and the classical uncoupler carbonyl cyanide-m-chlorophenyl hydrazone (CCCP). It was observed that the inhibitory effect of sorbic acid seems to be more focussed on the cell membrane than that of acetic acid and that sorbic acid has an effect on cell physiology that is more akin to a classical uncoupler (Ter Beek et al., 2014; van Beilen et al., 2014). In B. subtilis, PgsA is an essential protein committed to the synthesis of phosphatidylglycerol (PG), which besides being the only essential phospholipid is also the precursor for cardiolipin (CL) and lysyl-phosphatidylglycerol (L-PG) (Matsumoto et al., 2006; Salzberg and Helmann, 2008). Lopez and co-workers have shown that salt-stressed cells increase their CL phospholipid levels and decrease both PG and L-PG levels (López et al., 1998). Sorbic acid was shown to interact with the phospholipid headgroups (Chu et al., 2009) and mutants resistant to uncouplers were found to contain mutations in desaturase, resulting in more unsaturated fatty acids (Krulwich et al., 1990). This further corroborates the notion that the membrane is likely to play a crucial role in weak organic acid sensitivity, especially so since this may help to maintain the proton motive force.

A few years ago, RodZ was discovered as new player in bacterial cell morphogenesis (Gerdes, 2009). RodZ has been described in, among others, Escherichia coli, Caulobacter crescentus and B. subtilis and is reported to interact with the bacterial actin-like MreB cytoskeleton controlling cell shape and cell wall synthesis (Shiomi et al., 2008; Alyahya et al., 2009; Bendezú et al., 2009; White et al., 2010; Dempwolff et al., 2011). Interestingly, in several Bacillus species rodZ (ymfM) lies upstream in the genome of pgsA, encoding phosphatidylglycerol phosphate synthase. Until now, no functional link between RodZ and phospholipid synthesis has been reported. But, since RodZ is also associated with the cell elongation complex (den Blaauwen et al., 2008; Alyahya et al., 2009) a link between the only essential phospholipid and the cell envelop seems obvious as they would need to keep track of each other within a growing cell. Moreover, weak organic acids can act as uncouplers, they may interfere with correct localization of several components of the cytoskeleton (Strahl and Hamoen, 2010; van Beilen et al., 2014).

To elucidate novel weak acid resistance mechanisms in B. subtilis we created a mini-Tn10 transposon library and screened for sorbic acid hypersensitive mutants. We show that inactivation of rodZ by a transposon insertion and reduction in pgsA expression using a conditional mutant lead to a weak acid hypersensitive phenotype. We demonstrate that PgsA depletion contributes primarily to the weak acid sensitivity observed and speculate on a possible link between membrane and cell wall homeostasis through RodZ and PgsA.

# MATERIALS AND METHODS

# Strains, Growth Conditions, and Genetic Manipulation

All bacterial strains and plasmids used in this study are listed in **Table 1**. The B. subtilisstrains are derivatives of the laboratory 168 wild-type (WT) lab-strains PB2 (trp2C) or 1A700 (trp2C). E. coli strains XL1-Blue and MC1061 were grown in lysogeny broth (LB) at 37◦C. B. subtilis strains were grown in LB, buffered with 80 mM 3-(N-morpholino)propanesulfonic acid (MOPS) at pH 7.4 or 6.4 at 37◦C. When required for selection, the following antibiotics were added to the medium at given concentrations: 100 µg/ml ampicillin, 1 µg/ml erythromycin, 100 µg/ml spectinomycin, 10 µg/ml kanamycin.

Standard molecular genetics techniques were used as described by Sambrook et al. (1989). The pDG148 vector was used to overexpress rodZ and pgsA (Stragier et al., 1988). The rodZ gene was PCR amplified from B. subtilis PB2 genomic DNA using the rodZ\_FW and rodZ\_RV primers (see Supplementary Table S1 of the Supporting Information for the sequences of all used primers). All cloning PCR reactions were performed with Pfu polymerase (Fermentas, Thermo Fisher Scientific). The forward primers for each construct contain a ribosome

#### TABLE 1 | Plasmids and strains used in this study.

fmicb-07-01633 October 19, 2016 Time: 18:35 # 3


<sup>a</sup>SpR, spectinomycin resistance; EmR, erythromycin resistance; ApR, ampicillin resistance; PhR, phleomycin resistance; KmR, kanamycin resistance; MCS, multiple cloning site.

<sup>b</sup>Bacillus Genetic Stock Center (http://www.bgsc.org/).

binding site, which is not present on the plasmid when it is cut with HindIII. The PCR product and vector were digested with HindIII and SalI and ligated with T4 ligase (Fermentas), thus creating pDG-rodZ. For pDG-pgsA, the pgsA gene was amplified using the pgsA\_FW and pgsA\_RV primers and introduced between the HindIII and SalI sites of pDG148. The combined rodZ-pgsA construct was amplified using the rodZ\_FW and pgsA\_RV primers and inserted as described above to create pDG-rodZ-pgsA. All constructs were first transformed to chemically competent E. coli MC1061 cells before plasmids were isolated and transformed to competent B. subtilis strains. Plasmids were isolated using a QIAprep Spin Miniprep Kit (Qiagen). Competent B. subtillis cells were obtained and their transformations were performed as described previously (Kunst and Rapoport, 1995). The nucleotide sequence of all newly constructed plasmids was verified by sequencing.

# Identification of Sorbic Acid-Susceptible Genes

Two independent transposon mutant libraries in B. subtilis WT strain PB2 were created as previously described using the mini-Tn10 delivery vector pIC333 (Steinmetz and Richter, 1994). The resulting libraries were validated for correct transposition efficiency and randomness of transposition (data not shown). Both constructed mutant libraries have the expected statistical properties according to Petit et al. (1990) and Maguin et al. (1992). The two transposon mutant libraries were subjected to a screen to identify mutants hypersensitive to the presence of sorbic acid. The mutant libraries were plated on 80 mM MOPSbuffered LB agar (pH 7.4). Following overnight incubation at 37◦C, cells were transferred by replica plating using sterile velvets to 80 mM MOPS-buffered LB agar of pH 6.4, containing 30 mM potassium sorbate (KS). Next, the same sterile velvets were used to replicate cells on LB plates without KS to verify the successful transfer of cells. The mutants that showed no or minimal growth on KS containing LB agar compared to the plates without sorbate were stored at −80◦C until further analysis. Chromosomal DNA from the hypersensitive mutants was isolated using the DNeasy Blood & Tissue Kit (Qiagen). The purified DNA was digested with HindIII. DNA from the restriction reaction was self-ligated using a Ready-To-Go DNA T4 Ligase (Amersham Bioscience). Next, ligated DNA was used directly for transformation of E. coli XL1-BLUE cells (Stratagene). Isolated plasmid DNA was used as a template for sequencing with specific to either end of the transposon and enabling sequencing of the cloned chromosomal DNA. The primers were purchased from Isogen Life Science (see Supplementary Table S1 of the Supporting Information for the sequences) and the sequencing was performed by BaseClear. The sequencing results were aligned to the re-annotated genome sequence of B. subtilis 168 using BLAST at the SubtiList database (Barbe et al., 2009) 1 in order to identify the affected gene.

# Phase-Contrast Microscopy

Bacillus subtilis strains PB2 (WT) and rodZ transposon mutant ATB012 were grown into the exponential phase in 80 mM MOPSbuffered LB medium of pH 6.4. Cells were immobilized on 1% agarose as described previously by Koppelman et al. (2004), and photographed with a CoolSnap fx (Photometrics) CCD camera mounted on an Olympus BX-60 fluorescence microscope through an UPLANFl 100x/1.3 oil objective (Japan).

# Characterization of Stress Sensitivity of Various Mutant Strains

To further characterize the acid sensitivity of the (transposon) mutants the cells were grown on both solid agar plates and in liquid medium containing different concentrations of KS, potassium acetate (KAc) or NaCl. For the conditional pgsA mutant strain (Pspac-pgsA) (Hashimoto et al., 2013), the medium contained additional 0.1 mM isopropyl-β-D-1 thiogalactopyranoside (IPTG). Overexpression of rodZ and/or pgsA from the pDG148 vector was induced with 1 mM IPTG in exponentially growing cells 3 h before the start of a stress experiment. For the plating assay on solid medium, the cells were first grown exponentially in 80 mM MOPS-buffered liquid LB medium of pH 6.4, containing the appropriate antibiotics and 0.1 mM IPTG, at 37◦C. At an OD<sup>600</sup> of 0.2, 10-fold serial dilutions of the cultures were spotted on 80 mM MOPS-buffered LB plates of pH 6.4, containing 1% agar and IPTG if required, together with the indicated stress factors. After 24 h of incubation at 37◦C pictures of the plates were taken. For the plating assays, the following end-concentrations of chemicals were used: KS: (15, 30,

<sup>1</sup>http://genolist.pasteur.fr/GenoList/Bacillus

or 40 mM), KAc (80, 125, or 200 mM), and NaCl (0.7 or 1.4 M). Biologically independent experiments were performed at least three times. To monitor the growth of mutants in liquid media, exponentially growing cultures were grown to an OD<sup>600</sup> of 0.4. Then, the cells were twofold diluted in a 96-well micro titer plate containing different concentrations of weak organic acid or NaCl (see below), and IPTG if required. Cells were further cultivated in a FluoStar Optima microtiter plate reader (BMG Labtech) under rigorous shaking at 37◦C for 12 h. Cells were exposed to the following stresses: KS (15 mM or 40 mM), KAc (40 mM or 125 mM), and NaCl (0.7 or 1.4 M). All conditions were tested in the microtiter plate reader at least in duplicate, and biologically independent experiments were performed at least three times.

# Reverse Transcriptase PCR (RT-PCR)

Cells of B. subtilis wild-type strain PB2 and rodZ transposon mutant strain ATB012 were grown exponentially in 80 mM MOPS-buffered LB medium of pH 6.4. At an OD<sup>600</sup> of 0.2, half of the cells were exposed to 3 mM of KS and samples were withdrawn from the stressed and control cultures at 0, 10, 20, 30, 45, and 60 min after the addition of KS. Cultures of PB2 harboring empty pDG148 and pDG148-rodZ were grown exponentially in MOPS-buffered LB medium of pH 6.4 and 1 mM IPTG was added 3 h before sampling. At an OD<sup>600</sup> of 0.2 samples were taken every 30 min for 2 h. All samples were snap-frozen in liquid nitrogen and stored at −80◦C prior to RNA extraction. Biological independent experiments were performed twice.

Total RNA was isolated using the RNeasy Kit (Qiagen), as described by the manufacturer's instructions. Total RNA was eluted in sterile RNase/DNase free water (Ambion), concentrations were determined with Nanodrop UV spectroscopy (Ocean Optics), and integrity was analyzed by agarose gel electrophoresis. RNA samples were treated with Turbo DNase (Ambion) to remove genomic DNA as described by the manufacturer.

To see if rodZ and pgsA are part of the same transcriptional unit and possibly the same operon, a reverse transcriptase reaction using the Superscript First Strand kit (Invitrogen) was performed with RNA isolated from the WT strain and a single reverse primer annealing to the sequence of pgsA, named pgsA\_Q1\_RV. cDNA was amplified for 40 cycles in a Biometra T3000 Thermocycler using the rodZ\_Q1\_FW and pgsA\_Q1\_RV primers. The PCR product was analyzed on a 1% agarose gel. The visible band was isolated from the gel using the QIAquick gel extraction kit (Qiagen) and the DNA was sequenced.

To determine the expression levels ofrodZ and pgsA in the WT and mutant strains cDNA was synthesized using the Superscript First Strand kit (Invitrogen), using random hexamers. Semiquantitative real-time PCR analysis was carried out on a 7300 Real-Time PCR system (Applied Biosystems). Primer Express 3.0 software (Applied Biosystems) was used to design specific primers (purchased from Isogen Life Science) for real-time PCR (see Supplementary Table S1 of the Supporting Information). Reactions were carried out in a 20 µl mixture consisting of 1 µl of 3 µM specific primers, 5 µl of 100-fold-diluted cDNA template and SYBR green PCR master mix (Applied Biosystems). The cycling conditions were as follows: 1 cycle at 50◦C for 2 min, 1 cycle at 95◦C for 10 min, and 40 cycles at 95◦C for 15 s and at 60◦C for 1 min. Melting curves were used to monitor the specificity of the reaction. RNA of all time points and independent experiments were analyzed with real-time PCR in triplicate. Because the amplification efficiencies of the target and reference genes was tested and found to be approximately equal (not shown), the 11C<sup>T</sup> method could be used to calculate relative gene expressions(Livak and Schmittgen, 2001). The accA (α subunit of acetyl-CoA carboxylase) and rpsM (small subunit ribosomal protein S13) genes were used as two independent internal controls, since their expression levels were stable during exponential growth, irrespective of sorbic acid stress.

# Phospholipid Analysis

Cultures of the different B. subtilis strains were grown exponentially in MOPS-buffered LB medium of pH 6.4. At an OD<sup>600</sup> of 0.2, half of each culture was stressed with 5 mM KS and 45 min later 50 ml of each cell culture was harvested by centrifugation (5 min, 4000 rpm) and the pellets were frozen in liquid nitrogen. Three independent biological replicates were assessed.

For HPLC-MS/MS analysis, samples were lyophilized overnight. The total protein concentration of the samples was measured using a BCA Protein assay kit (Thermo Scientific). Cell material corresponding to 1 mg protein was resuspended in 300 µl water and 300 µl 1:1 chloroform-methanol (v/v). The following internal standards (obtained from Avanti Polar Lipids, Inc.) were added: phosphatidylglycerol (PG), lysylphosphatidylglycerol (L-PG), phosphatidylethanolamine (PE), and cardiolipin (CL). Each sample including internal standards was shaken vigorously and placed on ice for 15 min, after which it was centrifuged at 1000 × g for 10 min. The organic layer was transferred into another tube, and the aqueous layer was re-extracted with 3 ml of 2:1 chloroform-methanol (v/v). The combined organic layers were evaporated under a stream of nitrogen at 45◦C. The residue was dissolved in 150 µl of 50:45:5 chloroform-methanol-water (v/v/v) (Houtkooper et al., 2006).

The relative abundances of the species in the sampleextracts were determined, using HPLC-MS/MS. The liquidchromatographic separation was achieved on a modular HPLC system (Surveyor; Thermo Finnigan) consisting of a cooled autosampler set at 12◦C, a low-flow quaternary MS pump, and an analytical HPLC column: 2.1 × 250 mm silica column, 5 µm particle diameter (Merck). Phospholipids were separated and eluted with a programmed linear gradient between solution B (chloroform-methanol, 97:3, v/v) and solution A (methanolwater, 85:15, v/v) as described previously (Houtkooper et al., 2006). MS/MS analyses were performed on a TSQ Quantum AM (Thermo Finnigan Corporation, San Jose, CA, USA) operated alternating in the negative- and positive ion electrospray ionization (ESI) mode in consecutive runs. The surface induced collision was set at 10 V; spray voltage was 3600 V and the capillary temperature was 300◦C. In the MS/MS experiments argon was used as collision gas at a pressure of 0.5 mtorr; collision energy ranged between 20 and 40 eV for the different optimized transitions. In the negative mode mass spectra of CL molecular

species were obtained by continuous scanning between m/z 400– m/z 1000 (2 s/scan). In the positive mode characteristic constant neutral loss (CNL) scans were used to selectively detect specific phospholipids in their corresponding retention time windows: CNL(141) for PE, CNL(172) for PG, and CNL(300.1) for L-PG.

# RESULTS

# Identification of rodZ::mini-Tn10 as a Sorbic Acid-Hypersensitive Clone

In order to identify stress resistance mechanisms against weak organic acids and potential new antimicrobial targets we screened two independent transposon mutant libraries of B. subtilis 168 lab-strain PB2 for sorbic acid-hypersensitive clones on LB plates of pH 6.4 containing 30 mM potassium sorbate (KS). Around 10.000 clones were screened for sorbic acid sensitivity and resulted in 132 candidates in an initial evaluation. After thorough investigation of these mutants both on solid and in liquid medium, we ended up with four verified clones displaying a hypersensitive phenotype toward KS. All four clones were identified to have a transposon insertion in the rodZ (ymfM) gene (**Figure 1A**). The mutants containing an interrupted rodZ gene were isolated from both independently constructed transposon libraries and the location of insertion (at basepair 51) was found to be the same for all four mutants.

of the rodZ::mini-Tn10 cells. Scale bar: 10 µm.

Sorbic acid hypersensitivity in liquid and on solid media of one of the identified rodZ::mini-Tn10 mutants (strain ATB012) is shown in **Figures 1B,C**, respectively. In non-stressed liquid cultures of pH 6.4 the transposon mutant lagged a bit behind the WT strain PB2, yet it reached a higher yield than the WT strain after 5 h. However, when stressed with 15 mM KS the rodZ mutant, unlike the WT strain, displayed only a very slow increase in optical density. Although WT strain PB2 was able to grow, the mutant did not form colonies on LB plates of pH 6.4 containing 30 mM KS after overnight incubation at 37◦C (**Figure 1C**, lower). It can be noted that the colonies of mutant strain ATB012 formed on plates are more round when compared to PB2 (**Figure 1C**, upper). Additionally, the rodZ::Tn10 mutant was found to display an irregular cell morphology, having curved endings and being broader when grown in the exponential phase and examined under the microscope, with wild-type cells being 1.03 ± 0.06 µm and rodZ::Tn10 cells 1.13 ± 0.07 µm (**Figure 1D**).

# rodZ and pgsA Are Co-transcribed

In B. subtilis and in most Gram-positive bacterial genomes that have rodZ, essential gene pgsA lies next to rodZ (Alyahya et al., 2009) (**Figure 1A**). We hypothesized that rodZ and pgsA are part of the same operon, although the two genes are separated by an inverted repeat which, in spite of the lack of a T-rich tail, may act as a termination sequence (Rasmussen et al., 2009). While we cannot rule out their presence, we did not find indications for additional regulatory elements of pgsA expression. To confirm that rodZ and pgsA are indeed co-transcribed, we performed a reverse transcriptase reaction using one single reverse primer annealing to pgsA and purified RNA of WT strain PB2 as the template. A PCR reaction using specific primers for rodZ and pgsA resulted in one specific product. The sequence of the fragment corresponded exactly with the region between rodZ and pgsA (**Supplementary Figure S1** of the Supporting Information).

Additionally we tested whether the mRNA levels of rodZ and/or pgsA were affected by sorbic acid stress. Stressing exponentially growing cells of the WT strain in LB medium of pH 6.4 with 3 mM KS did not significantly change the expression levels of both genes during the first 60 min after KS stress (See Supplementary Excel Data Sheet S1). This observation was in line with our previous studies performing microarrays with 3 mM KSstressed cells in defined minimal medium of pH 6.4 (Ter Beek et al., 2008).

# Stress Profiles of the rodZ Transposon Mutant and a Conditional pgsA Mutant

RT-PCR results indicated a transcriptional link between rodZ and pgsA. Therefore, we decided to further characterize the identified sorbic acid-sensitive rodZ::mini-Tn10 strain, as well as a conditional pgsA mutant (pgsA::Pspac-pgsA) strain by testing their susceptibility to other stresses. Since pgsA is essential, this strain (MHB001), kindly provided by Kouji Matsumoto (Hashimoto et al., 2013), needs a minimal amount of IPTG to be able to grow. The lowest IPTG concentration which sustained exponential growth similar to that of the WT strain in liquid cultures was 0.1 mM in our exponential conditions. Hereby we wanted to identify whether these genes are involved specifically in lipophilic sorbic acid stress, in weak organic acid stress more in general or in generic stress tolerance. Thus we tested their stress sensitivity for the highly water soluble weak organic acid: acetic acid. Although the side chains of sorbic- and acetic acid differ significantly, both have a similar pK<sup>a</sup> of 4.76. Additionally we tested osmotic stress by the addition of NaCl, thus investigating a form of stress unrelated to weak organic acid stress.

Next to the previously described sorbic acid-hypersensitivity (**Figure 1**) the results clearly showed also hypersensitivity of the rodZ mutant strain for acetic acid stress on plates when compared to the WT strain (**Figure 2A**). No colonies were formed on LB plates of pH 6.4 containing 200 mM potassium acetate (KAc). Additionally, the ability of rodZ::mini-Tn10 cells to form colonies was clearly more inhibited by 1.4 M NaCl than of cells from the parent B. subtilis strain. In liquid cultures, the rodZ mutant also displayed clear sensitivity toward KAc and NaCl stress (see **Supplementary Figure S2** of the Supporting Information). Interestingly, when grown on solid plates representing the various stress conditions and supplemented with 0.1 mM IPTG, the pgsA mutant strain showed a similar stress profile as that of the rodZ mutant (**Figure 2B**). Hypersensitivity to sorbic- and acetic acid was observed, and in accordance with previous data, a clear stress sensitivity was detected on plates containing 0.7 M NaCl. The conditional pgsA mutant grown in liquid cultures and supplemented with 0.1 mM IPTG also revealed a clear sensitive phenotype toward KS, KAc, and NaCl (see **Supplementary Figure S3** of the Supporting Information). However, the observed phenotypes disappeared when liquid cultures contained 10-fold more (1 mM) IPTG (data not shown).

The strong overlap in the phenotypes of both the rodZ::mini-Tn10 mutant and the conditional pgsA mutant and the observation that both genes are co-transcribed suggests that the inactivation of rodZ in the transposon mutant gives rise to polar effects. In support of this inference, our RT-PCR data showed 11C<sup>T</sup> values for pgsA of up to 5 upon comparing rodZ::mini-Tn10 with control cells using accA as reference gene and 3.6 with rpsM as reference (See Supplementary Excel Data Sheet S1). Given the observed phenotypes of both mutants and the involvement of PgsA in crucial membrane phospholipid biosynthesis we next assessed the membrane composition in the rodZ::mini-Tn10 and pgsA::Pspac-pgsA strains.

# Phospholipid Composition and Structure in rodZ::mini-Tn10 and pgsA::Pspac-pgsA

Our earlier findings suggested that adaptation to sorbic acid would (in part) be by remodeling of the plasma membrane (Ter Beek et al., 2008). We therefore investigated the effect of sorbic acid on the membrane composition of the WT strain, the rodZ::mini-Tn10 mutant, and the pgsA conditional mutant (pgsA::Pspac-pgsA), as these two mutant strains displayed a similar sensitivity profile. In both the rodZ mutant and in the conditional pgsA mutant strain (under conditions of minimal pgsA expression), the average acyl-chain length of phospholipids

was already higher than that of the WT strain when grown without weak organic acid stress. **Figure 3** shows the acyl tail length distribution of PG. Similar results were observed for PG-derived phospholipids (CL and L-PG) and PE (see **Supplementary Figure S4** of the Supporting Information). Upon KS stress, the most prominently observed phospholipid tail length (2·16 = 32 carbon atoms) seemed to increase further in the WT (**Figure 3**), corroborating our earlier results (Ter Beek et al., 2008). For the conditional pgsA mutant this is less clear, and was not observed in the rodZ mutant.

In terms of the presence of different phospholipid classes the main difference between the WT strain and the rodZ::mini-Tn10 strain under non-stressed conditions is a significant decrease in CL and PG phospholipids in the mutant strain (**Figure 4**). In the conditional pgsA mutant similarly lowered levels of CL and PG phospholipids were observed. Consequently, in both mutant strains the relative PE levels were increased when compared to the levels in the WT strain under control conditions. When the WT strain was exposed to sorbic acid, there was a significant drop observed in the relative L-PG levels (**Figure 4**), while the relative content of the other phospholipids was hardly affected. This trend was also observed in the conditional pgsA mutant, however, not in the rodZ mutant. Compared to the WT strain, the mutants showed a significant shift in the distribution of the acyl chain lengths (longer) (**Figure 3**). Also the mutants seems to have a less negative membrane charge due to the significant reduction of PG (−1) and CL (−2) when compared to the WT (**Figure 5**). Sorbic acid stress does seem to increase the negative net charge on the plasma membrane of WT cells (reduction in L-PG), which may reduce the entry of the sorbate anion. From the data presented it is clear that the phosholipid composition between the two mutants is quite similar, suggesting that the inactivation of rodZ may give rise to polar effects and the hypersensitivity of this mutant strain is perhaps caused by lower expression levels of pgsA.

# Complementation of the rodZ::mini-Tn10 Mutant

We tried to rescue the stress sensitivity of the rodZ transposon mutant by overexpression of rodZ and/or pgsA by introducing the IPTG-inducible plasmid pDG148 (Stragier et al., 1988) containing rodZ and/or pgsA under control of a Pspac promotor. Under control conditions (LB-M, pH = 6.4), the wild-type strain, and mutant strains with pDG148 or pDG-rodZ performed equally during early exponential growth. Mutants containing pDG-pgsA or pDG-rodZ-pgsA displayed a longer lag phase, but ended up with a higher end OD<sup>600</sup> (**Figure 5A**). Weak

acid sensitivity of the rodZ::mini-Tn10 mutant with pDG-rodZ (Pspac-rodZ) was not restored (**Figure 5B**). Complementation with pgsA via pDG-pgsA (Pspac-pgsA) in the rodZ mutant reduced KS stress almost completely. Complementation with the pDG-rodZ-pgsA plasmid (Pspac-rodZ-pgsA) restored some KS sensitivity, but not as much as overexpressing pgsA alone (**Figure 5B**). When cultured in the presence of 125 mM KAc stress, the overexpression of RodZ and PgsA from pDG-rodZpgsA or PgsA alone significantly increased the growth rate of the rodZ::mini-Tn10 mutant strain (**Figure 5C**). Induction of pDG-rodZ had no effect (**Figure 5C**).

# DISCUSSION

In order to have a direct measure of functional importance for B. subtilis weak organic acid stress resistance we decided to opt for the construction and screening of a transposon mutant library for sorbic acid hypersensitive mutants. Sorbic acid (trans-trans-2,4-hexadienoic acid) is a six-carbon unsaturated fatty acid with a pKa of 4.76 and the acid, or its anionic salt, is commonly utilized by the food industry. We chose to screen for sorbic acidsusceptible genes after pre-growing the transposon library first on LB plates for 24 h. We decided to use rich LB medium in our experiments, so that as many as possible mutants were able to grow with relatively normal rate prior to the screen. Longer incubations revealed many more small colonies emerging on the

plates. These slow growers on plates were not used in the screen because they evidently already have severe problems growing in control conditions.

Screening of the library for sorbic acid hypersensitivity led to the identification of a uniquely stress sensitive phenotype in which the transposon was inserted into the rodZ gene (**Figures 1B,C**). One and the same insertion site was identified in the four discovered clones coming from two independent constructed mutant libraries (**Figure 1A**). The library was made in B. subtilis WT strain PB2 using the mini-Tn10 delivery vector pIC333 and was designed to increase randomness of the sites of insertion (Steinmetz and Richter, 1994). Perhaps that only a specific insertion into the rodZ gene of B. subtilis resulted in a viable clone with comparable growth rate to the WT strain in control conditions.

Significantly,rodZ lies immediately upstream in the genome of the essential pgsA gene encoding phosphatidylglycerophosphate synthase (**Figure 1A**). Interestingly, in many other Grampositive bacteria (e.g., Bacilli, Streptococci, and Stapylococci) rodZ and pgsA are predicted to be part of the same operon (Ermolaeva et al., 2001; Alm et al., 2005; Alyahya et al., 2009). Our data, showing that a transcript containing both genes can be amplified from B. subtilis, corroborates this notion (See **Supplementary Figure S1** of the Supporting Information). Moreover, according to several prediction tools, B. subtilis RodZ has one transmembrane domain (amino acids 89 – 113) (Cserzö et al., 1997; Hirokawa et al., 1998) and an Xre-like helix-turnhelix (HTH) motive, commonly seen in DNA binding proteins, in its N-terminal side. Together, these observations initially

led us to believe that the RodZ protein might be involved in regulating pgsA expression. However, RT-PCR experiments with overexpression of RodZ in the WT strain failed to show any changes in pgsA mRNA levels (see **Supplementary Figure S5** of the Supporting Information). HTH motifs also have been shown to function in DNA replication, RNA metabolism, and proteinprotein interactions (Aravind et al., 2005). RodZ was shown to co-localize with components of the cytoskeleton and depend on MreB for its localization (Alyahya et al., 2009) and interaction of RodZ with MreB was shown to be specifically with this HTH motif in Thermotoga maritima (van den Ent et al., 2010). The mutual functional dependence of RodZ and MreB was reinforced by the observation that loss of RodZ, or at least its N-terminal domain, resulted in aberrant localization of MreB and cessation of its movement (van den Ent et al., 2010; Garner et al., 2011).

Since MreB organization also depends on the membrane potential (Strahl and Hamoen, 2010) and weak organic acids lower the proton gradient (by releasing protons in the cell) and may act in certain cases as uncouplers of the membrane potential (van Beilen et al., 2014), the impact of weak acid stress on cell growth may be partially mediated through membrane perturbation effects on the correct localization of MreB and RodZ containing cell shape determining protein complexes. This in turn could lead to a perturbed localization and functioning of PgsA, exacerbating the sorbic acid sensitivity. PgsA normally localizes primarily to the septal membranes in conjunction with cardiolipin and plays an essential role in cell division (Nishibori et al., 2005).

The observation that rodZ and pgsA are part of the same transcript as well as our earlier studies (Ter Beek et al., 2008; van Beilen et al., 2014) suggested to us that the membrane plays a crucial role in weak organic acid stress tolerance. Hence we studied the phospholipid composition of the various strains. The rodZ::mini-Tn10 mutant was shown to contain severely lowered levels of PG and CL (**Figure 4**). The average acyl chain length of the remaining phospholipids was increased in the rodZ mutant when compared to the WT (**Figure 3**; **Supplementary Figure S4** of the Supporting Information). The same phenomena were observed in the conditional pgsA::Pspac- pgsA mutant that displayed similar sorbic acid-, acetic acid-, and salt-stress sensitivities (**Figure 2**; **Supplementary Figures S2** and **S3** of the Supporting Information). Alterations of the cell membrane lipid composition were also found in response to sorbic acid stress where a pronounced increase in acyl chain length and lowering of L-PG was seen most prominently in WT B. subtilis (**Figures 3** and **4**). An increase in chain length stiffens the membrane and lowers its permeability. On the other hand, the induction of the BkdRregulated genes (involved in the synthesis of precursor molecules for branched-chain fatty acids (Debarbouille et al., 1999) in sorbic acid-stressed cells (Ter Beek et al., 2008) may indicate increased branching in phospholipids and thereby balancing membrane fluidity levels. Interestingly, Lopez and co-workers have shown that salt-stressed cells increase their CL phospholipid levels and decrease both PG and L-PG levels (López et al., 1998), a similar trend observed in KS-stressed cells (**Figure 4**). However, they also measured a clear decrease in branched chain fatty lipids in cultures grown in NaCl. Noteworthy is the observation that both

to induce overexpression.

and with 15 mM KS (B) or 125 mM KAc (C). 1 mM IPTG was added to the pre-cultures of all strains 3 h before the start of the experiments at t = 0 min.

mutants hardly seem to change their membrane composition when stressed with KS (**Figures 3** and **4**). The possible differences in net charge of the membrane between the mutants (significantly reduced PG and CL levels) and the WT may be one explanation for the observed hypersensitivity toward weak organic acid stress. Another explanation might be the inability of the mutants to modify their membrane composition further upon stress, e.g., they cannot adapt further. Besides a potential role for L-PG in weak acid sensitivity, it has strongly been implicated in resistance against cationic antimicrobial peptides (Dunkley et al., 1988; Sohlenkamp et al., 2007; Samant et al., 2009). Alterations in the membrane composition likely pertain to changes in the rate of net proton influx. That is: the rate limiting step for protonophoric uncouplers is the rate with which the anion traverses the liquidlipid interface (Spycher et al., 2008; Chu et al., 2009; Ter Beek et al., 2014; van Beilen et al., 2014). It is therefore likely that an increase in CL levels in combination with increased acylchain length confers increased resistance to uncouplers. On the other hand, this adaptation may conflict with growth at elevated temperatures, but B. subtilis may ultimately strive toward homeo-proton permeability (van de Vossenberg et al., 1999). Alternatively, anion build-up has also been implicated as a stress factor caused by weak organic acid preservatives (Russell, 1992). Both of these mechanisms may interfere with proper cytoskeleton function and put stress on the total cell envelop. If one of these players fails to work in concert, the cell is weakend.

Our data presented here on the rodZ and the conditional pgsA mutant suggest that the phenotypes observed in the rodZ transposon mutant might be primarily the result of polar effects on pgsA expression. The RT-PCR data on pgsA expression levels in the rodZ::mini-Tn10 strain compared to WT PB2 corroborate this conclusion. Finally, overexpression of pgsA in the rodZ::mini-Tn10 background stimulated growth of cells cultured in liquid media in the presence of sorbic and acetic acid. However, overexpression of rodZ in the rodZ transposon mutant did not restore weak acid stress sensitivity to WT levels. These results also indicate that the phenotypes observed in the rodZ transposon mutant can almost solely be ascribed to polar effects on pgsA.

We present here a direct link between phospholipid synthesis and weak acid sensitivity and propose that PgsA plays an important role in membrane homeostasis and tolerance to weak organic acid stress. Future studies are aimed at assessing the membrane permeation efficacy in Bacillus strains with different phospholipid perturbations (Salzberg and Helmann, 2008), by measuring intracellular acidification rates upon addition of various weak organic acids. This can be done with the aid of the pH-sensitive fluorescent protein pHluorin expressed in the lumen of the bacterium. The protocol for this has recently also been established by us in B. subtilis (van Beilen and Brul, 2013; van Beilen et al., 2014; Ter Beek et al., 2014).

# AUTHOR CONTRIBUTIONS

Conceived and designed the experiments: JB, AZ, SB, AB. Performed the experiments: JB, CB, HF, RB, AB. Analyzed the data: JB, CB, AZ, AB. Contributed reagents/materials/analysis tools: WK, FV. Wrote the paper: JB, SB, AB.

# FUNDING

AB was supported by a grant from the Dutch Foundation for Applied Sciences (STW 10431).

# ACKNOWLEDGMENTS

Kouji Matsumoto (Saitama University, Japan) is acknowledged for his gift of the MHB001 strain before publication. We are grateful to Tanneke den Blaauwen of the Bacterial Cell Biology group for use of their phase-contrast microscope. We thank Rachna Pandey, Ingar Seeman, Alicia Prats Tur, and Henk van Lenthe for their technical assistance, and Bart J. F. Keijser for initial suggestions on the experiments.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2016.01633

FIGURE S1 | rodZ and pgsA are co-transcribed. cDNA synthesized from RNA of PB2 using the reverse primer for pgsA (pgsA\_Q1\_RV) was PCR amplified with the ymfM\_Q1\_FW and pgsA\_Q1\_RV primers and run on an agarose gel. M: GeneRuler DNA ladder mix (Fermentas); 1: Genomic DNA of B. subtilis WT strain PB2 used as the template (positive control); 2: cDNA made with pgsA\_Q1\_RV used as the template; 3: No template (primer and RNA control); 4: RNA incubated with RNase A (10 µg/ml) for 20 min at 37◦C prior to cDNA reaction (genomic DNA contamination control); 5: No reverse transcriptase control.

FIGURE S2 | Inactivation of rodZ leads to increased sensitivity for acetic acid and salt stress in liquid cultures. Growth curves of WT strain PB2 and the rodZ::mini-Tn10 mutant grown in liquid LB medium of pH 6.4 in control conditions (closed and open circles, respectively) and in the presence of 125 mM KAc (A) or 0.7 M NaCl (B) (closed downward triangles and open triangles, respectively).

FIGURE S3 | Conditional pgsA mutant displays increased sensitivity for sorbic acid, acetic acid, and salt. Growth curves of WT strain 1A700 and the pgsA::Pspac-pgsA conditional mutant grown in liquid LB medium of pH 6.4 in control conditions (closed and open circles, respectively) and in the presence of 15 mM KS (A), 125 mM KAc (B) or 0.7 M NaCl (C) (closed downward triangles and open triangles, respectively).

FIGURE S4 | Acyl tail length distributions of phospholipids measured in WT strain PB2, conditional pgsA mutant, and the rodZ transposon mutant, in the absence and presence of 5 mM potassium sorbate (KS). Acyl chain length distributions of the following phospholipid classes were measured: (A) cardiolipin (CL), (B) lysyl-phosphatidyl glycerol (L-PG), and (C) phosphatidyl glycerol (PE). The number behind the abbreviation indicates the total number of carbon atoms in the acyl chains per molecule.

FIGURE S5 | Overexpression of rodZ did not have an effect on pgsA expression. Relative quantification of gene expression using real-time RT-PCR was determined between B. subtilis WT strain PB2 carrying pDG148 and pDG-rodZ every 30 min until 2 h after induction. Expression of rodZ was induced with 1 mM IPTG at t = 0 min in exponentially growing cells. Data were normalized to accA expression. The error bars indicate the standard deviation.

# REFERENCES

fmicb-07-01633 October 19, 2016 Time: 18:35 # 11



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

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

# Temperature Exerts Control of Bacillus cereus Emetic Toxin Production on Post-transcriptional Levels

Markus Kranzler<sup>1</sup> , Katharina Stollewerk<sup>1</sup> , Katia Rouzeau-Szynalski<sup>2</sup> , Laurence Blayo<sup>2</sup> , Michael Sulyok<sup>3</sup> and Monika Ehling-Schulz<sup>1</sup> \*

<sup>1</sup> Functional Microbiology, Institute of Microbiology, Department of Pathobiology, University of Veterinary Medicine Vienna, Vienna, Austria, <sup>2</sup> Food Safety Microbiology, Nestec Ltd, Nestlé Research Center, Lausanne, Switzerland, <sup>3</sup> Center for Analytical Chemistry, Department of Agrobiotechnology, IFA Tulln, University of Natural Resources and Life Sciences Vienna (BOKU), Vienna, Austria

In recent years, the emetic toxin cereulide, produced by Bacillus cereus, has gained high

#### Edited by:

Avelino Alvarez-Ordóñez, Teagasc Food Research Centre, Ireland

#### Reviewed by:

Anne-Brit Kolstø, University of Oslo, Norway Heidy Den Besten, Wageningen University and Research Centre, Netherlands

\*Correspondence:

Monika Ehling-Schulz monika.ehlingschulz@vetmeduni.ac.at

#### Specialty section:

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

Received: 26 June 2016 Accepted: 03 October 2016 Published: 25 October 2016

#### Citation:

Kranzler M, Stollewerk K, Rouzeau-Szynalski K, Blayo L, Sulyok M and Ehling-Schulz M (2016) Temperature Exerts Control of Bacillus cereus Emetic Toxin Production on Post-transcriptional Levels. Front. Microbiol. 7:1640. doi: 10.3389/fmicb.2016.01640 relevance in food production and food safety. Cereulide is synthesized non-ribosomal by the multi-enzyme complex Ces-NRPS, which is encoded on a megaplasmid that shares its backbone with the Bacillus anthracis pX01 toxin plasmid. Due to its resistance against heat, proteolysis and extreme pH conditions, the formation of this highly potent depsipeptide toxin is of serious concern in food processing procedures including slow cooling procedures and/or storage of intermediate products at ambient temperatures. So far, systematic data on the effect of extrinsic factors on cereulide synthesis has been lacking. Thus, we investigated the influence of temperature, a central extrinsic parameter in food processing, on the regulation of cereulide synthesis on transcriptional, translational and post-translational levels over the growth temperature range of emetic B. cereus. Bacteria were grown in 3◦C interval steps from 12 to 46◦C and cereulide synthesis was followed from ces gene transcription to cereulide toxin production. This systematic study revealed that temperature is a cardinal parameter, which primarily impacts cereulide synthesis on post-transcriptional levels, thereby altering the composition of cereulide isoforms. Our work also highlights that the risk of cereulide production could not be predicted from growth parameters or sole cell numbers. Furthermore, for the first time we could show that the formation of the recently identified cereulide isoforms is highly temperature dependent, which may have great importance in terms of food safety and predictive microbiology. Notably the production of isocereulide A, which is about 10-fold more cytotoxic than cereulide, was specifically supported at low temperatures.

Keywords: Bacillus cereus, cereulide, isocereulide, temperature, food safety

# INTRODUCTION

The incidence of foodborne intoxications caused by bacterial toxins has been steadily increasing over the last decade. Especially, the number of reported food poisonings related to Bacillus cereus toxins has shown a steep increase from 2006 onward (Anonymous, 2009, 2015). The rising seriousness of B. cereus is also reflected in the growing number of reports on severe intoxications

related to the emetic B. cereus toxin cereulide, requiring hospitalization or even leading to death (Dierick et al., 2005; Posfay-Barbe et al., 2008; Ichikawa et al., 2010; Naranjo et al., 2011; Ehling-Schulz and Messelhäusser, 2012). Because of the high toxicity of cereulide and the high incidence rates up to 100%, usually observed in connection with outbreaks, accurate data on contamination sources and factors promoting toxin formation are urgently needed to prevent contamination and toxin production in food production processes. Especially comprehensive prevalence data on potential hazardous sources from the food processing chain are missing, and critical downstream processing steps at risk for cereulide production are hitherto unknown. Since B. cereus is a ubiquitous sporeformer, it cannot be totally avoided in many raw material and food products. Toxin formation is of serious concern in food processing procedures that include slow cooling procedures and/or storage of intermediate products at ambient temperatures. Because of its chemical structure, the emetic depsipeptide toxin cereulide shows an extreme pH and heat stability. Once preformed in food ingredients or matrices, this toxin will most likely not be destroyed or inactivated throughout the food production processes. Due to its size (1.2 kDa), the toxin cannot be removed by conventional filtration and survives subsequent thermal processing (Agata et al., 1995; Rajkovic et al., 2008). For instance, contaminations with cereulide or cereulide-producing bacteria have been reported from infant foods in Finland, various food products in Belgium and Bavaria as well as from ice creams in Germany (Shaheen et al., 2006; Rajkovic et al., 2007; Messelhäusser et al., 2010, 2014).

Cereulide is a dodecadepsipeptide, composed of alternating α-amino and α-hydroxy acids (D-O-Leu–D-Ala–L-O-Val–L-Val)3, that is produced non-ribosomally by an enzyme complex with an unusual modular structure, named cereulide synthetase (Ces NRPS; Ehling-Schulz et al., 2005b; Magarvey et al., 2006). Very recently, it has been shown that, in parallel to the known cereulide, at least 17 different isoforms are synthesized by the Ces NRPS, which represents a novel type of a NRPS (Marxen et al., 2015a,b). The cereulide isoforms significantly differ in their cytotoxic properties. Isocereulide A shows an 8- to 10-fold higher cytotoxicity than cereulide, while some other isoforms are almost non-toxic (Marxen et al., 2015a). All emetic B. cereus strains so far analyzed (n = 80) produced all the same isoforms although at different ratios (Marxen et al., 2015a,c).

The ces gene locus responsible for the cereulide synthesis is encoded on a 270kb mega virulence plasmid related to the Bacillus anthracis toxin plasmid pXO1 (Ehling-Schulz et al., 2006; Rasko et al., 2007), and the ces genes have been shown to be polycistronically transcribed from a central promoter (Dommel et al., 2010). The cereulide synthesis is intrinsically controlled by a complex and tightly regulated transcriptional network that ensures correct timing of ces gene expression within a short period of the cell cycle (Lücking et al., 2009, 2015; Dommel et al., 2011; Frenzel et al., 2012). Although the exact mechanisms triggering cereulide production are far from being understood, latest research clearly illustrates that cereulide synthetase gene expression and cereulide toxin production is influenced by complex interactions of various intrinsic as well as extrinsic factors (for review see Ehling-Schulz et al., 2015). While significant progress has been made on the understanding of the intrinsic factors embedding cereulide synthesis in the bacterial life cycle, systematic studies on the influence of external factors on cereulide synthesis are missing. This study therefore aimed to gain new insights into the regulation of cereulide production by examining the influence of temperature, a key external parameter in food processing, on cereulide synthesis on a transcriptional, translational and post-translational level.

# MATERIALS AND METHODS

# Bacterial Strains

For this study, four emetic B. cereus strains were used: two wellcharacterized strains isolated in frame of food borne outbreaks, the emetic reference strain F4810/72, isolated from vomit (also designated AH187; NCBI Reference Sequence: NC\_011658) and the strain F5881, isolated from Chinese takeaway fried rice, (detail for these strains can be found in Ehling-Schulz et al., 2005a), and, further two strains from food industrial environments, one strain isolated from wheat flour (B626) and one strain isolated from dehydrated puree with onions (AC01).

# Growth Conditions and Determination of Sampling Points

Bacterial overnight cultures, pre-grown in 3 ml LB broth under laboratory standard conditions (pH 7.0, 30◦C, 120 rpm), were used for kinetic inoculation (final inoculum 10<sup>3</sup> cfu/ml) of the main cultures (100 ml LB broth, pH 7.0, and 120 rpm) in baffled flasks as described previously (Dommel et al., 2011). Cultures were incubated under rotary shaking at different temperatures in the range from 12 to 46◦C in 3◦C intervals. At least two independent growth experiments were carried out for each temperature.

According to preliminary tests, all strains showed almost identical growth and comparable values of optical (OD600) and cell (cfu/ml) densities at 24◦C (**Figure 1**, inset). Under these standard conditions, maximal rates of transcription and translation of cesB could be determined for all strains at approx. OD<sup>600</sup> of 10, which is in line of results from a previous study carried out at 30◦C (Frenzel et al., 2012). Thus, growth of bacteria was investigated photometrically and samples were taken at OD<sup>600</sup> of 10 from bacterial cultures grown at different temperatures for further processing as outlined below. For transcriptional analyses, additional samples were taken at OD<sup>600</sup> of 1.

# Transcriptional Analyses of cesB

Transcription of the ces gene cesB was analyzed by qRT-PCR as described previously (Dommel et al., 2011). In brief, RNA was isolated from frozen cell pellets via TRIzol Reagent (Invitrogen) and bead beating. Phase separation was carried out with chloroform and nucleic acid was precipitated with 75% ethanol. Subsequently, DNA was digested with RQ1 DNAse I (Promega) and a total amount of 1 µg RNA was used for cDNA

calibrator for transcriptional analysis (Figure 2A). Arrows indicate the OD<sup>600</sup> sampling points at all tested temperatures. Data represent means and error bars

synthesis (cDNA qScript Supermix, Quanta Biosciences). qRT-PCR for each sample was performed in triplicates on a C1000 Touch Thermal Cycler CFX96 Real-Time System (BioRad). In order to normalize and compare the relative expression the REST method was employed (Pfaffl et al., 2002), which has been shown previously to be a quite suitable tool for analysis of ces transcription (see, e.g., Lücking et al., 2009, 2015; Dommel et al., 2010, 2011; Frenzel et al., 2012). As internal calibrator, (with a relative expression value of 1.00) ces gene expression at an OD<sup>600</sup> of 1.0 was chosen since it has been shown previously to be consistent throughout experiments (Dommel et al., 2011). Samples at OD<sup>600</sup> of 1 and 10 were always taken from the same cultures. Sample to sample variation was corrected by using the 16S rDNA gene as a reference (rrn). Mean values and standard deviations were calculated from two independent experiments.

indicating standard deviations of at least two independent growth experiments.

# Translational Analyses of CesB

Ces translation was investigated by immunoblotting using a recently generated monoclonal antibody specific for CesB (Lücking et al., 2015) and a peroxidase-conjugated secondary antibody (AffiniPure Goat Anti-Mouse IgG, Dianova). Western blot analysis was carried out according to Lücking et al. (2015). Briefly, SDS-PAGE was performed according to Laemmli (1970) and immunoblotting was performed on a BioRad Transblot SD Semi Dry Transfer Cell. Determination of total protein load was carried out by staining with Ponceau S Solution (Applichem). For the chemiluminescent reaction, Super Signal West Pico Chemiluminescent Substrate (Thermo Scientific) was applied and quantification was performed with the ImageQuant TL Software (GE Healthcare Life Sciences). At least two technical replicates per sample were performed. Relative translation of CesB was determined by calculating all CesB Western Blot

signals in relation to the CesB blot signal with highest intensity (AC01, 21◦C), which was set to 100%. Blot to blot variation was corrected by using CesB protein as a reference included in all blots, including at least two technical replicates per sample. Mean values and standard deviations were calculated from two independent experiments.

# Cereulide and Isocereulide Quantitation by Means of Ultra Performance Liquid Chromatography (UPLC) Tandem Mass Spectrometry (MS/MS)

For the extraction of cereulide, bacteria grown in LB medium were pelletized by centrifugation (8.000 g, 23◦C, 10 min) and 50 mg of bacterial biomass was resuspended in 1 ml acetonitrile (99%, HPLC grade, Carl Roth). After incubation for 16 h on a rocking table, the pellet was centrifuged and the supernatant was directly transferred to a HPLC vial and stored at room temperature. MS measurements were performed using a QTRAP <sup>R</sup> 5500 MS/MS system (Applied Biosystems, Foster City, CA, USA) equipped with a TurboV electrospray ionization (ESI) source and a 1290 series UHPLC system (Agilent Technologies, Waldbronn, Germany). Chromatographic separation was performed at 25◦C on a Gemini <sup>R</sup> C<sup>18</sup> – column, 150 × 4.6 mm i.d., 5 µm particle size, equipped with a C<sup>18</sup> security guard cartridge, 4× 3 mm i.d. (all from Phenomenex, Torrance, CA, USA). Stock solutions of 500 ppm cereulide (Chiralix, Nijmegen, The Netherlands) and 100 ppm <sup>13</sup>C6- cereulide (Chiralix) were prepared for the standard curve and internal standard, respectively. MSanalysis was performed as described by Bauer et al. (2010) and quantitation of isocereulides was carried out as reported previously (Marxen et al., 2015c). All samples were measured in two different dilutions as duplicates. Mean values and standard deviations were calculated from two independent experiments.

# Protease Activity Measurement

Determination of protease activity was carried out with Pierce Fluorescent Protease Assay kit (Thermo Scientific), according to manufacturer's instructions and spectrophotometric measurement was carried out as described above.

# Statistical Analysis

Data were analyzed with R (version 3.3.1) software (Hornik, 2016) using the Welch's t-test to evaluate the null hypothesis that temperature does not affect the tested parameter. Data from experiments carried out at 24◦C served as reference for each strain. A P-value of 0.01 or less was considered statistically significant. Data represent means and error bars showing standard deviations from at least two independent experiments.

# RESULTS

The influence of temperature on cereulide synthesis of B. cereus grown in LB medium was investigated photometrically at a wavelength of 600 nm (OD600). Preliminary kinetic analyses carried out at the reference temperature 24◦C, revealed a maximum of cesB transcription and CesB translation around OD<sup>600</sup> of 10 for all four strains included in this study (data not shown). Thus, OD<sup>600</sup> of 10 was chosen as sampling point. To cover the full growth range of emetic strains, bacterial cultures were grown from 12◦C to 43◦C in 3◦C intervals to OD<sup>600</sup> of 10 and at 46◦C, a temperature at which OD<sup>600</sup> of 10 was not reached, up to the maximum OD600). From cultures of all four strains grown at the different temperatures to an OD<sup>600</sup> of 10 (or to the maximum OD<sup>600</sup> at 46◦C), samples were taken in parallel for transcriptional, translational and post-translational analysis of cereulide synthesis. Samples were processed as described previously (Frenzel et al., 2012; for details see also material and method section). For transcriptional analysis, additional samples were taken from all cultures at OD<sup>600</sup> of 1 to be used as internal calibrator (Dommel et al., 2011).

# Effect of Temperature on Growth of Emetic B. cereus

An overview on the time needed to reach OD<sup>600</sup> of 1 and OD<sup>600</sup> of 10, respectively, is provided in **Figure 1**. The inset shows the growth curves for all strains at the reference temperature of 24◦C. The fastest start of growth was observed at temperatures >33◦C, reaching an OD<sup>600</sup> of 1 within 4–4.5 h (37◦C: 4.5 h, 40◦C: 4.0 h, 43◦C: 4.25 h, and 46◦C: 4.25 h). However, at temperatures exceeding 37◦C, the time needed to reach OD<sup>600</sup> of 10 was slightly longer, compared to the time needed at 37◦C. All strains reached OD<sup>600</sup> of 10 between 33 and 37◦C after 10–12 h, while at 46◦C none of the strains reached OD<sup>600</sup> of 10. At the latter temperature, a maximal OD<sup>600</sup> of 5–7 was observed for all strains after 12 h, thereafter the OD<sup>600</sup> values dropped (data not shown), pointing toward severe growth temperature stress around the upper growth temperature boundary. Generally, below 18◦C, all strains showed slow growth and the cell division rate was significantly reduced. Around the lower growth temperature boundary at 12◦C, OD<sup>600</sup> of 1 was reached after 4.5–6 days and the time needed to reach OD<sup>600</sup> of 10 required 7–9 days, respectively. At room temperature (24◦C), all strains showed the same growth behavior while at the growth boundaries some strain-specific growth behavior was observed. Strain AC01 tends to be the slowest growing strain, especially at low temperatures. In contrast, strain B626 was growing faster at low temperatures than the other strains while it appeared to be the slowest growing strain at temperatures >40◦C, raising the question whether this strain is more adapted to low temperatures than the other emetic strains included in this study.

# Effect of Temperature on cesB Transcription and Translation

The influence of growth temperature on transcription and translation of the cereulide synthetase was determined by quantitative RT-PCR detecting cesB mRNA levels and immunoblotting, using a CesB specific monoclonal antibody (Lücking et al., 2015). The ces genes cesP,T,A,B,C,D are transcribed polycistronically from a main promoter upstream of cesP and qRT-PCR revealed the same transcriptional kinetics

(Lücking et al., 2015). For relative expression analyses of cesB at OD<sup>600</sup> of 10, the cesB transcript level at an OD<sup>600</sup> of 1 was used as the calibrator (RE = 1.0) for each strain at all tested temperatures. Relative translation of the CesB protein at different temperatures was determined by calculation CesB blot signal intensity of each sample relative to CesB showing highest blot signal intensity (RE = 100%). Data represent means and error bars indicate standard deviations of at least two independent growth experiments. <sup>∗</sup>N.D., not determined.

for cesP,T,A,B,C,D (Dommel et al., 2010, 2011; Frenzel et al., 2012). Currently, only for CesB a specific antibody (CesB mAB) is available but not for any of the other Ces proteins encoded by the ces operon. Thus cesB/CesB was chosen for our survey to use the same target on transcriptional and translational level. Unexpectedly, no significant differences on transcript levels of cesB (P < 0.01) were observed over a broad growth temperature range from 12◦C or 15–33◦C, depending on the strain. A decrease of the temperature even to the lower growth boundary (12◦C) did not reveal a significant difference in cesB transcription in strain AC01 compared to the cesB transcription level at 24◦C, which served as reference temperature, while a significant decrease of cesB transcription was found for the other strains at 12◦C (P < 0.01). An increase of the temperature from

24 to 37◦C resulted in significant decrease of cesB transcription in all four strains (P < 0.01). At higher temperatures, transcription was declining but still detectable up to the upper growth boundary (46◦C; **Figure 2A**). Ces translation was also observed over the complete growth range. However, in contrast to cesB transcription, translation turned out to be highly temperature dependent (**Figure 2B**). A steep increase of the CesB protein signal in the immunoassay was observed from 15 to 18◦C. Between 21 and 24◦C the plateau of maximal CesB translation was reached in all strains. No significant differences were found in CesB levels between 21 and 24◦C (P < 0.01). At higher temperatures, the CesB signal was dropping in a strain-specific manner. For instance, the signal in the CesB-specific immune assay for AC01 and B626 at 27◦C dropped to approximately 50 and 25%, respectively, of the signals detected at 24◦C. The CesB signal for the emetic reference strain F4810/72 sharply declined at 30◦C, while the decline of the CesB signal of the strain F5881 at high temperatures was not as pronounced as for the other strains. Remarkably, strain B626, which appeared to be the best adapted strain to low temperatures among the strains used in this study, showed the lowest CesB levels in most cases, except at 15◦C, at which the highest signal of all strains was observed. CesB protein signals were also detectable for all strains at 12◦C, but could not be quantified due to the severe growth impairments and cell damages at the lower growth boundary (**Figure 1**).

Additionally, a protease assay was employed to test if the drastic drop of CesB signals from 18 to 15◦C is caused by increased protease activity at the lower temperature. However, results from a comparative study of protease activity from bacterial cultures grown at 24 and 15◦C to OD<sup>600</sup> of 10 and accompanying CesB stability test did not reveal a decreased stability or specific degradation of CesB at 15◦C (data not shown). Thus, it is more likely that the differences observed at 18 and 15◦C are the result of temperature dependent CesB translation rather than the consequence of CesB degradation.

# Effect of Temperature on Cereulide Toxin Production

Whereas the cesB transcript and CesB protein were detectable over the whole temperature range, the toxin was not detectable at the upper growth boundary (46 and 43◦C) and only at very low levels at the lower growth boundary (12◦C; **Figure 3A**). For instance, at 12◦C toxin amounts between 2 and 8 µg/g bacterial wet weight were found in the analyzed strains, while the highest toxin amounts—ranging from 31–194 µg/g bacterial wet weight—were detected at 33 and 37◦C, respectively. Paralleling the results from the translation analyses, a steep increase in cereulide toxin amounts was observed when the growth temperature was raised from 15 to 18◦C. A second and even more pronounced increase in cereulide amounts was found when temperature was shifted from either 30 to 33◦C (AC01 and B626) or from 33 to 37◦C (F4810/72 and F5881), indicating a strain-specific response to growth temperature. For instance, at 33◦C strain F4810/72 produced about 50 µg/g and AC01 produced about 130 µg/g, whereat the cereulide levels detected at 37◦C were completely reverse in these two strains (**Figure 3A**). A maximum of cereulide toxin was detected for one out of the four strains at 33◦C (AC01: 147 µg/g bacterial wet weight) and for the other three strains at 37◦C (F4810/72: 131 µg/g bacterial wet weight, F5881: 149 µg/g bacterial wet weight, B626: 172 µg/g bacterial wet weight). After the maximum of toxin accumulation was reached, further increase of the temperature to 40◦C caused a sharp decline in toxin amounts in all strains tested, down to levels comparable to those between 18 and 24◦C. A further increase of the temperature to 43◦C resulted in an almost complete stop of cereulide production, although all strains showed fast growth. The maximum toxin levels at 43◦C ranged from 0.003 to 0.09 µg/g wet weight. Thus, the risk of cereulide toxin formation cannot be predicted from sole cell numbers or growth parameters. Interestingly, cereulide levels found at 40◦C were similar to that of 18–30◦C, but cesB transcription and CesB expression were significantly lower at 40◦C.

# Effect of Temperature on Isocereulide Production

Recently, 17 isoforms of cereulide—apart from the known cereulide—have been discovered and a novel method for multiparametric quantitation of isocereulides A–G was developed (Marxen et al., 2015a,c), which provided us the possibility to examine the influence of temperature on the composition of isocereulides. Our study showed that isocereulide A and isocereulide B were produced in LB-grown bacterial cultures up to levels of 14% of the ones observed for cereulide, while isocereulide C and D were produced at levels equaling 1–2% of cereulide (**Figures 3B,C**). Most interestingly, highest amounts (9–14%) of the highly cytotoxic isocereulide A were found at low temperatures (12 and 15◦C), while relative low levels of isocereulide B (0.8–1.8%) were detected at these temperatures. However, the latter isoform was produced to levels up to 10% between 18 and 27◦C, corresponding to the temperature range in which comparable low amounts of isocereulide A were detected. Isoforms C and D were below 2%, whereas isoform D was not detectable at low temperatures (12 or 15◦C). Therefore, it is tempting to speculate that emetic strains are shifting the production of isocereulide A and isocereulide B in a temperature dependent manner, which must be taken into account during risk assessments in the food industry.

# DISCUSSION

# Temperature: A Cardinal Environmental Effector of Post-transcriptional Cereulide Synthesis Regulation

Cereulide synthesis in emetic B. cereus is controlled by a network of bacterial intrinsic factors, involving different realms of regulation (for review see Ehling-Schulz et al., 2015). These bacterial intrinsic factors include key factors, which act on the transcriptional level (Lücking et al., 2009; Frenzel et al., 2012), thereby embedding cereulide production in the bacterial lifecycle (Dommel et al., 2011). Transcription of ces genes is highly dynamic and restricted to a distinct growth phase (Dommel et al., 2011; Frenzel et al., 2012; this work). Very recently

FIGURE 3 | Cereulide toxin and isocereulide production in dependence of temperature. Cultures were grown to an OD<sup>600</sup> of 10 in LB broth, 120 rpm, at different temperatures in a range from 12 to 46 ◦C, using 3◦C intervals. Cereulide and isocereulide extraction was carried out with acetonitrile (99%) and quantitation was performed by ESI-HPLC MS/MS analysis. Cereulide (A) and isocereulide (B) levels are depicted as µg cereulide per g bacterial wet weight as well as relative amount of cereulide isoforms, referred to the amount of cereulide values, are shown (C). Data represent means and error bars indicating standard deviations of at least two independent growth experiments.

it has been shown that in addition to these chromosomally encoded central transcription factors, proteins embedded in the cereulide synthetase (ces) locus exert control of the nonribosomal synthesis of cereulide on transcriptional, translational and post-translational levels (Lücking et al., 2015). We therefore aimed in the current work to assess the influence of temperature as a key extrinsic factor in food processing environments on the different levels of cereulide synthesis control. In contrast to intrinsic factors, which are tightly regulating cereulide synthesis on a transcriptional level, temperature—as an extrinsic signal influenced toxin expression mainly on a post-transcriptional level. No significant differences on transcript levels of cesB were observed over a broad growth temperature range (**Figure 2A**). At higher temperatures, transcription was declining but still detectable up to the upper growth boundary. In comparison, studies on the effect of temperature on the botulinum toxin complex L-TC have shown higher mRNA levels at lower temperatures and lower transcript levels as well as less stability at high temperatures (Couesnon et al., 2006; Chen et al., 2008). Therefore, considering the differential regulation of CesB protein and cereulide, it is tempting to speculate that temperature sensing of the cell triggers a response of cereulide synthesis on a translational and/or post-translational level rather than on a transcriptional level.

It was also noticeable that at the upper growth boundary the CesB protein is still formed but apparently not the toxin. Diminished cereulide levels at 21 and 24◦C are not in coherence to the high expression of CesB at these temperatures. Thus, it can be assumed that temperature has rather a post-translational regulatory effect, and may orchestrate the production of the toxin out of the dipeptides by the cereulide synthetase—but not the expression of the single NRPS modules—or may influence the activity of Ces NRPS in another, hitherto unknown way. Very recently, a novel mechanism for the action of cereulide synthetase has been proposed, highlighting dipeptides rather than single amino or hydroxy acids as the basic modules in toxin assembly (Marxen et al., 2015b). Further studies are needed to fully understand the mechanism of cereulide toxin synthesis and export as well as the exact role of external factors, such as temperature, on this highly complex process.

# Temperature: A Key Parameter for Cereulide Production

Although temperature abuse is one of the key factors in emetic food poisoning caused by B. cereus (Ehling-Schulz et al., 2004; Dierick et al., 2005; Ehling-Schulz and Messelhäusser, 2012), information on the influence of temperature on cereulide synthesis in emetic B. cereus is still limited. Generally, emetic strains show a shift of their growth temperature limits toward higher temperatures compared to non-emetic strains (Carlin et al., 2006). These results are in line with those from our current study (**Figure 1**). Temperatures below 21◦C led to strongly decelerated growth, with long pre-exponential and exponential phases. Generally, the time needed to reach OD<sup>600</sup> of 1 was decreasing with increase of growth at temperatures up to 43◦C, pinpointing the temperature shift of emetic strains toward higher temperatures compared to non-emetic B. cereus strains. All strains reached OD<sup>600</sup> of 1 at 40 and 43◦C after 4 h. Around the temperature growth boundaries, a 3◦C change significantly impacted the growth behavior of all strains, although, in a different way at the lower than at the upper temperature growth boundary. At 12◦C, growth was generally retarded, while at 46◦C the strains reached a maximum OD<sup>600</sup> of 5–7 already after 12 h, thereafter the OD<sup>600</sup> values were dropping (data not shown). Thus, it is tempting to speculate that bacteria successfully start growing at high temperatures but are rapidly accumulating cell damages caused by heat stress (e.g. radical formation, malfunction of chaperons, etc.). In contrast, at the lower growth boundary OD<sup>600</sup> of 10 was reached after seven to 9 days, which may rather reflect general growth impairments than acute cell damages. The latter must be taken in consideration when interpreting results of cereulide expression studies.

Various studies have analyzed the production of cereulide under different temperature conditions in laboratory broth, on agar plates or in different food matrices (e.g., Finlay et al., 2000; Häggblom et al., 2002; Rajkovic et al., 2006; Shaheen et al., 2006; for review see Ehling-Schulz et al., 2004). However, results are difficult to compare due to the lack of standardization for cultivation (pre-culture conditions, inoculation levels, and media) and methods used for cereulide quantitation (bioassays, cytotoxicity assays, and mass spectrometry). For instance, it has been shown that history of a strain could affect strain capacity for cereulide production (Thorsen et al., 2009; Dommel et al., 2011). Furthermore, to our knowledge, hitherto no systematic analysis on the effect of temperature on cereulide production in emetic strains has been performed. Thus, we systematically investigated cereulide production over the complete growth temperature range using narrow temperature intervals. As shown by this approach, an increase of only 3◦C at certain temperatures resulted in a twofold (e.g., for F4810/72: from 5 µg/g at 12◦C to 11 µg/g at 15◦C) or even more than fivefold (e.g., for F4810/72: from 7.5 µg/g at 27◦C to 43.9 µg/g at 30◦C) increase of cereulide levels, while a further increase of 3◦C, after the maximum of cereulide production was reached, resulted in a drastic drop of cereulide levels (**Figure 3A**), pinpointing the importance of systematic and in-depth analyses, such as carried out in the present work. Corroborating results from previous studies, we found detectable, although low cereulide levels at the lower growth temperature boundary. Generally, 12◦C seems to be the minimum temperature limit for cereulide production (for review see Ehling-Schulz et al., 2004). Finlay et al. (2000), who studied cereulide production in skim milk medium at various temperatures, detected toxin in a temperature range between 12 (after 4 days) and 37◦C (after 24 h), which is in principle in line with our results. However, in contrast to the findings of Finlay et al. (2000), we found the highest levels of cereulide at 33–37◦C and not at low temperatures. Probably these discrepancies might be explained by the different approaches used for quantitation of cereulide. Finlay et al. (2000) used a cell culture assay for cereulide quantitation, which cannot discriminate the different isocereulides, while in the current study mass spectrometry based systems for accurate quantitation of cereulide and isocereulides, were employed (Bauer et al., 2010; Marxen et al., 2015c).

# Temperature: The First Extrinsic Parameter Shown to Impact Isocereulide Composition

The recently discovered isocereulides, which are characterized by alternating positions of the dipeptides within the cyclic dodecadepsipeptide ring structure, significantly differ in their cytotoxic properties (Marxen et al., 2015a,b). For instance, isocereulide A shows an 8- to 10-fold higher cytotoxicity than the classical cereulide while isocereulide B is almost non-toxic (Marxen et al., 2015a). If or how external parameters, such as temperature, are influencing the composition of isocereulides, was hitherto unknown. We therefore aimed in this study to assess the effect of temperature on the occurrence and quantity of cereulide isoforms, using a novel method for simultaneous detection and quantitation of isocereulide A–G (Marxen et al., 2015c). To our knowledge, this is the first study investigating the influence of an extrinsic factor on the production of isocereulides. The isocereulide levels found in our study were in the same %-range as those from samples of recent foodborne outbreaks (Marxen et al., 2015c). Isocereulide A was detected in all samples tested positive for cereulide (**Figure 3**). Our results show a dramatic shift of the relative amount of isocereulide A and isocereulide B at the boundary of room temperature, at 18◦C, a temperature that also represents a threshold for significantly slowed growth. Thus, low temperatures turned out to specifically support the production of this highly toxic isoform. This finding makes low temperatures a new critical parameter in terms of food hygiene and storage, although it must be kept in mind that the absolute amount of isocereulide A is still higher at 33 and 37◦C, due to generally elevated cereulide levels at the latter temperatures.

# CONCLUSION

Our systematic investigation of temperature, a cardinal parameter in cereulide toxin synthesis and critical factor in food processing environments, showed that temperature exerts

# REFERENCES


control of cereulide synthesis in emetic B. cereus primarily on post-translational levels. How these external signals are embedded in the bacterial lifecycle and internal signaling is largely unknown and warrants further investigation. The cereulide toxin production ceased at temperatures at which fastest growth was observed, highlighting that the risk of toxin formation cannot be deduced from sole cell numbers or growth parameters. To our knowledge, this study is the first one investigating the effect of temperature on the formation of isocereulides. Importantly, low temperatures turned out to specifically support the production of the highly toxic isocereulide A. The amounts of isocereulides found in our study paralleled those from recent food-borne outbreaks. Thus, further studies are needed to systematically investigate the influence of environmental factors on the production of isocereulides and to assess the consequences of these novel, unexpected findings for B. cereus-related risk assessments.

# AUTHOR CONTRIBUTIONS

Performed the experiments: MK, KS, and MS; Analyzed the data: MK, KS, MS, and ME-S. Conceived and designed the experiments: ME-S, KR-S, and LB. Wrote and revised the paper: MK, ME-S, KR-S, and LB; ME-S has conceptualized, supervised, and acted as overall study director.

# FUNDING

This work was financially supported by Nestec Ltd, Vevey, Switzerland.

# ACKNOWLEDGMENTS

We thank Elisabeth Mader and Tatjana Svoboda for excellent technical assistance and Benedikt Schulz for R software based statistical analysis.

at refrigeration temperature. Appl. Environ. Microbiol. 74, 6132–6137. doi: 10.1128/AEM.00469-08



of cereulide, the emetic toxin of Bacillus cereus. Sci. Rep. 5:10637. doi: 10.1038/srep10637


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

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

# Adaptation in Bacillus cereus: From Stress to Disease

### Catherine Duport\*, Michel Jobin and Philippe Schmitt

Sécurité et Qualité des Produits d'Origine Végétale, UMR0408, Avignon Université, Institut National de la Recherche Agronomique, Avignon, France

Bacillus cereus is a food-borne pathogen that causes diarrheal disease in humans. After ingestion, B. cereus experiences in the human gastro-intestinal tract abiotic physical variables encountered in food, such as acidic pH in the stomach and changing oxygen conditions in the human intestine. B. cereus responds to environmental changing conditions (stress) by reversibly adjusting its physiology to maximize resource utilization while maintaining structural and genetic integrity by repairing and minimizing damage to cellular infrastructure. As reviewed in this article, B. cereus adapts to acidic pH and changing oxygen conditions through diverse regulatory mechanisms and then exploits its metabolic flexibility to grow and produce enterotoxins. We then focus on the intricate link between metabolism, redox homeostasis, and enterotoxins, which are recognized as important contributors of food-borne disease.

#### Edited by:

Lorena Ruiz, Universidad Complutense de Madrid, Spain

#### Reviewed by:

George-John Nychas, Agricultural University of Athens, Greece Folarin Anthony Oguntoyinbo, University of Lagos, Nigeria

\*Correspondence: Catherine Duport catherine.duport@univ-avignon.fr

#### Specialty section:

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

Received: 28 July 2016 Accepted: 15 September 2016 Published: 04 October 2016

#### Citation:

Duport C, Jobin M and Schmitt P (2016) Adaptation in Bacillus cereus: From Stress to Disease. Front. Microbiol. 7:1550. doi: 10.3389/fmicb.2016.01550 Keywords: Bacillus cereus, redox homeostasis, oxygen sensing, acidic pH, metabolism

# INTRODUCTION

Micro-organisms display astonishing abilities to survive and grow in hostile environments. Some have adapted their life cycles to extreme conditions (thermophiles, psychrophiles, halophiles, acidophiles, etc.), but all can marshal temporary adaptation mechanisms to help them survive until conditions have become more favorable. When environmental conditions change, a cell must modify its physiology accordingly to cope with them and survive or reproduce. However, adaptation has limits, which depend on the microorganism and on the environmental variables that have become extreme, upset the cell's equilibrium, and caused stress. Limits to adaptation depend on the micro-organism's intrinsic capacity to cope; the physiological responses the cell can call on to address environmental variations include changes in metabolism and/or mechanisms of adaptation and resistance.

Bacillus cereus is a Gram-positive, facultative anaerobe, rod-shaped endospore-forming bacterium that is known to inhabit primarily the soil, where it can complete its saprophytic life cycle (Vilain et al., 2006). As a result of its saprophytic soil life cycle, B. cereus is found in water, vegetables, and many other food ingredients, resulting in the contamination of a wide variety of finished food products. Ingestion of contaminated foods by humans can lead to two types of gastrointestinal infections, both damaging the host epithelium. The emetic type of food poisoning is caused by the ingestion of food containing the toxin cereulide, whereas the diarrheal type depends on the ingestion of B. cereus cells followed by the production of virulence factors in the human small intestine (Stenfors Arnesen et al., 2008; Senesi and Ghelardi, 2010; Ceuppens et al., 2012). B. cereus cells originate from ingested vegetative cells that survive gastric passage (Wijnands et al., 2009) and/or from ingested spores which first adhere to the intestinal mucosa and then germinate. Recent evidence suggests that B. cereus-induced diarrhea is not caused by massive

B. cereus proliferation and virulence factor production in the intestinal lumen but by localized growth and virulence factor production at the hosts mucus layer or epithelial surface (Ceuppens et al., 2012). These virulence factors then damage the nearby epithelial cells by pore formation, resulting in microvilli damage and osmotic lysis of the host's epithelial cells and eventually diarrhea (Beecher et al., 1995; Hardy et al., 2001; Minnaard et al., 2001; Lindback et al., 2004; Ramarao and Lereclus, 2006; Fagerlund et al., 2008). In the vicinity of the epithelial layer, B. cereus is exposed to different oxygen concentrations and different oxidoreduction potentials (ORP; Moriarty-Craige and Jones, 2004; Marteyn et al., 2010a,b) that induce compensatory metabolic pathways in an attempt to maintain the intracellular redox state. The cellular redox status governs the status of redox-sensitive macromolecules and protects against endogenous oxidative stress. Recent studies suggested that virulence factor production by B. cereus is dynamic and shaped by cellular oxidation (Madeira et al., 2015). Thus, there is an intricate link between metabolism, redox homeostasis, and virulence factor production. In the first part of this review, we focus on how microorganisms and B. cereus detect and respond to acid stress, and review the different behavioral, physiological and molecular mechanisms underpinning acid stress adaptation. In the second part, we will begin by describing the basics of B. cereus physiology and will then discuss how metabolism and redox global regulators influence the production of virulence factors under changing oxygen conditions.

# ACID STRESS RESISTANCE

Acid resistance is especially important for B. cereus that must survive the acidic pH of the stomach – which is 1.5 in the fasting state (Van de Guchte et al., 2002) and rises to 3–5 after ingestion of food (Cotter and Hill, 2003) – before entering and colonizing the small intestines or colon (Stenfors Arnesen et al., 2008). Acid stress is also frequently encountered naturally in many foods, as a result of the use of weak organic acids or short-chain (volatile) fatty acids (FA; e.g., acetic acid, citric acid, and propionic acids) as food preservatives (Alvarez-Ordonez et al., 2010a). Thus, the ability to adapt to an acidified environment is crucial to the virulence of a food-borne pathogen such as B. cereus.

Neutrophilic bacteria have evolved multiple tolerance or resistance mechanisms to prevent cell damage due to acid stress; these are generally referred to as acid tolerance responses (ATRs) and acid resistance mechanisms, respectively. Which system(s) plays the dominant role(s) depends on: (i) the phase of growth of the cells when the ATR is elicited [exponential phase- versus stationary phase- ATR]; and/or (ii) whether certain amino acids are present during exposure to the acidic pH; and/or (iii) whether acidification of the environment results from inorganic or organic acids (Van de Guchte et al., 2002; Alvarez-Ordonez et al., 2010b). B. cereus vegetative cells, like many other bacteria, are able to induce an ATR (Jobin et al., 2002; Thomassin et al., 2006). In addition, it has been shown that B. cereus cells preadapted at pH 6.3 coped better with both ethanol stress (12%) and heat stress (49◦C) (Browne and Dowds, 2002), suggesting that ATR and/or induction of acid-resistance mechanisms confer cross-protection for other stresses. In this article, we review some model acid-resistant and their related acid resistance mechanisms to understand how B. cereus can adapt to acid stress.

# Common Mechanisms of Acid Resistance

Micro-organisms deploy various mechanisms and strategies to address the hostile conditions of low-pH environment, e.g., modification of the architecture and composition of the membrane, change in metabolism and production of alkaline substances, and homeostasis of internal pH (**Figure 1**).

## Modification of the Membrane

The cell membrane of bacteria is in direct contact with external media. It is therefore the first one to be affected by harmful environmental conditions, e.g., an acidic medium. The fluidity of the membrane is important for cells, as it can affect membrane functions such as biochemical reactions, transport systems, and protein secretion. The membrane FA composition is responsible for the maintenance of membrane fluidity, and a number of studies have suggested a relationship between membrane fluidity and stress adaptation. Acid adaptation generally decreased membrane fluidity, and this is likely linked to the overall increase in short-chain saturated FA as observed in Escherichia coli, Salmonella, and Listeria (Kwon and Ricke, 1998a,b; Yuk and Marshall, 2004; Moorman et al., 2008; Alonso-Hernando et al., 2010). However, in some oral bacteria, exposure to acidic condition resulted in increased level of long-chain monounsaturated FA, and fluidity (Fozo and Quivey, 2004; Fozo et al., 2004; Papadimitriou et al., 2007). In E. coli, acid adaptation causes the conversion of a significant proportion of the unsaturated FA to their cyclic derivatives, known as cyclopropane fatty acids (CFA) during the transition from exponential to stationary phase (Merrell and Camilli, 2002; Merrell et al., 2002). The CFAs are formed by CFA synthase, which is encoded by cfa. Defective cfa mutants are unable to produce CFA and are sensitive to low pH (Booth, 2002). Phospholipid composition is also important in E. coli: stationary phase E. coli cells are much more sensitive to acid shock at pH 3 in the absence of the main phospholipid, phosphatidylethanolamine (Canet et al., 2003). Membrane adaptation in response to acid stress has not been yet investigated in B. cereus. However, it has been shown that B. cereus cells increase membrane fluidity by altering membrane FA composition in response to cold and saline stresses (Ultee et al., 2000; De Sarrau et al., 2012).

# Production of Alkali

Some bacteria produce alkaline compounds, and specifically ammonia, to neutralize internal pH when exposed to an acidic environment. Ammonia is generated by two systems, the urease and the arginine deiminase systems (ADI; Van de Guchte et al., 2002; Cotter and Hill, 2003). In the urease system, urea is hydrolysed to two molecules of ammonia and one of CO<sup>2</sup> by urease. In the arginine deiminase pathway, arginine is catabolized to ornithine with the release of ammonia and CO2. The urease system is more widespread and can protect some oral bacteria

against acid-induced changes (Cotter and Hill, 2003; Wilson et al., 2014). Although the catalytic reaction is relatively simple, biogenesis of a functional urease is a highly complex process requiring at least seven genes, which are generally organized in operons [ureABCEFGD, (Marquis and Hager, 2000)]. The expression of bacterial ureases can be constitutive, but more often it is regulated by environmental conditions. Commonly, in enteric bacteria, the presence of urea or limitation for nitrogen can induce urease gene transcription, generally through activation of transcription. Specifically, urease expression is almost completely repressed at neutral pH values, regardless of the limiting nutrient or growth rate. In acidic conditions, the urease genes become rapidly derepressed, and expression then becomes sensitive to carbohydrate availability and rate of growth, with the highest levels of expression under conditions of carbohydrate excess and fast growth rate (Chen and Burne, 1996; Chen et al., 1996; Cotter and Hill, 2003). B. cereus is in contact with the urea present in its diverse habitats, such as soil, human urine, human saliva (2.3–4.1 mM) (Mobley, 2000; Dawes and Dibdin, 2001), stomach (4.8 mM) (Neithercut et al., 1993), blood (1.7–8.3 mM) (Mackay and Mackay, 1927), and some animal foods (e.g., milk), which contain 4.4–6.4 mM urea (Carlsson et al., 1995). Although the activity of urease can play an important role in the life cycle of B. cereus, little information is available on its role in nitrogen metabolism and in acid stress survival in this bacterium. The ADI has also been identified in a broad variety of bacteria, including B. cereus (Van de Guchte et al., 2002; Cotter and Hill, 2003; Senouci-Rezkallah et al., 2011). This system is a three-enzyme pathway that initially converts arginine to citrulline and ammonia via arginine deiminase (encoded by arcA). The citrulline is then transformed into ornithine and carbamyl phosphate by ornithine transcarbamylase (encoded by arcB). The third enzyme in the pathway, carbamate kinase (encoded by arcC), cleaves carbamyl phosphate to ammonia and CO2, concomitantly donating the phosphate to ADP to produce ATP (Griswold et al., 2004). Similar to ureolysis, the net reaction yields two molecules of ammonia and one of CO2, but also provides ATP for growth. Thus, many ADI-positive bacteria can grow with arginine as the sole source of energy. ADI-positive organisms often coordinately regulate the synthesis of an arginine:ornithine antiporter (encoded by arcD) (Cotter and Hill, 2003; Budin-Verneuil et al., 2006). The expression of arcABC operon is induced by low pH or arginine and is suppressed by excess oxygen pressure in Listeria monocytogenes (Ryan et al., 2009). The arcABC operon also contributes to the growth and survival of Lactobacillus plantarum in lowpH environment (De Angelis et al., 2002; Spano et al., 2004). In B. cereus, arginine deiminase gene arcA showed significant up-regulation upon exposure to non-lethal acid shock at pH 5.4– 5.5 (Mols et al., 2010a; Senouci-Rezkallah et al., 2011), suggesting that ADI may be of great importance for B. cereus survival in low pH environments.

## Homeostasis of Internal pH

fmicb-07-01550 October 1, 2016 Time: 13:47 # 4

Internal pH (pHi) is an important factor in bacterial physiology, and cells regulate its value precisely. The regulation of pHi implies a heightened control of membrane permeability to protons, which can take place via the ion transporters that facilitate proton entry. In general, growth studies have shown that pH<sup>i</sup> falls with culture pH (Russell, 1991; O'Sullivan and Condon, 1999; Siegumfeldt et al., 2000). In B. cereus, the pH<sup>i</sup> in continuously cultured cells sampled at equilibrium fell with growth rate, showing that growth rate has an effect on pH<sup>i</sup> (Thomassin et al., 2006). pHi is disturbed by two main factors: (i) passive movement of protons through the cytoplasmic membrane, and (ii) the production of acids and/or bases in the cytoplasm. Thus, in fermenting bacteria, the accumulation of acidic fermentation products in the cytoplasm can decrease pH<sup>i</sup> , despite continuous efflux of protons (Russell and Diez-Gonzalez, 1998). The organic acids present in the culture medium can also impede pH<sup>i</sup> homeostasis, especially when pH<sup>i</sup> > external pH (pHe). Bacteria that can allow their pHi to decrease, so as to keep a low 1pH (pH<sup>i</sup> – pHe), may be more acid-resistant than those that keep their pH<sup>i</sup> neutral (Siegumfeldt et al., 2000). Non-dissociated organic acids freely cross the permeable membrane lipid bilayer, limiting their accumulation in the cell. In bacteria that maintain a higher pH<sup>i</sup> , organic acids dissociate in the more alkaline cytoplasm, and accumulate in the cell, halting growth (Mercade et al., 2000). Their dissociation depends on their pKa value, generally less than 5.0. The accumulation of these organic acids in anionic form depends on the pH gradient across the membrane (Russell, 1991). To keep pH<sup>i</sup> at a value that will conserve the cell's physiological integrity, bacteria can use many different strategies to control proton flux. These include (i) active transport of protons across the membrane (via the F1F0-ATPase activity), (ii) decarboxylation systems [glutamate decarboxylase (GAD), arginine decarboxylase (AD), and lysine decarboxylase], and (iii) the buffering ability of their cytoplasm.

### **F1F0-ATPase**

A number of studies have demonstrated a role for F1F0-H<sup>+</sup> translocating ATPase in pH<sup>i</sup> homeostasis (Miwa et al., 1997; Kullen and Klaenhammer, 1999; Cotter and Hill, 2003; Fortier et al., 2003). The F0F1-ATPase is a well-established mover of protons across the cell membrane (**Figure 2**). This complex couples the energy released as protons move into the cell to the generation of ATP from ADP and Pi. The ATPase can also function in the opposite direction, hydrolyzing ATP to pump protons out of the cell. F1F0-ATPase is composed of two protein complexes (Negrin et al., 1980; Takemoto et al., 1981; Cotter and Hill, 2003). The F0 complex, which is integrated in the membrane, allows protons to cross the membrane. The F<sup>1</sup> complex bears the catalytic site for the synthesis of ATP. The locus atp codes for the five sub-units α, β, δ, γ, and ε that form the complex F1, and the three subunits a, b and c that compose F0. It has been shown that the expression of atp was induced by an acid pH (Kullen and Klaenhammer, 1999; Quivey et al., 2001; Fortier et al., 2003). In E. coli, F1F0-ATPase acts as a proton pump during acid stress, driving protons out of the cell with a parallel hydrolysis of ATP (Richard and Foster, 2003). The F1F0-ATPase of B. cereus has been isolated from cytoplasmic membranes, purified and characterized. Structurally, B. cereus F1F0-ATPase resembles the enzyme isolated from E. coli and B. subtilis (Voelz, 1964; Banfalvi et al., 1980a,b). However, its enzymatic activity is insensitive to pH, and to N,N'-dicyclohexylcarbodiimide (DCCD), which is known to inhibit the activity of membranebound ATPase (Higuti et al., 1992). Recent study showed that B. cereus ATPase activity led to an increase in pH<sup>i</sup> when cells were exposed to acid stress. Indeed, DCCD had a negative effect on the ability of B. cereus cells to survive and maintain their pH<sup>i</sup> during acid shock. Furthermore, transcriptional analysis revealed that expression of atpB (encoding b subunit of F1F0-ATPase) was increased in acid-adapted cells compared to non-adapted cells before and after acid shock. These data demonstrate that B. cereus is able to induce an ATR during growth at low pH, depending on the ATPase activity induction and pH<sup>i</sup> homeostasis (Senouci-Rezkallah et al., 2015).

# **Decarboxylation of amino acids**

The presence of amino acids (glutamate, lysine, and arginine) in food such as vegetables and dairy products can enhance the ability of bacteria to adapt to and survive acid stress. These amino acids can be decarboxylated by systems composed of one or more decarboxylases, which convert their substrates into amine derivatives and carbon dioxide or bicarbonate, and an antiporter that then exchanges each amino acid for its decarboxylated amine. Decarboxylation of amino acids controls bacterial pH by consuming hydrogen through the decarboxylation reaction. The lysine, arginine and GAD systems predominate in acid tolerance (Booth, 2002). The GAD system has been identified in a variety of bacteria such as E. coli, L. monocytogenes, Shigella flexneri, and Lactococcus lactis (Cotter and Hill, 2003; Cotter et al., 2005; De Biase and Pennacchietti, 2012; Kanjee and Houry, 2013). In E. coli, known components of glutamate-dependent acid resistance include two isoforms of GAD (GadA and GadB) and a putative glutamate: γ-aminobutyric acid (GABA) antiporter called GadC. GadA/GadB is assumed to catalyze the conversion of protonated glutamate to GABA, whereas GadC exports GABA in exchange for a new extracellular glutamate molecule. This process consumes protons in cells, which eventually increases pHi, protecting the cell from the damage caused by acid shock (Richard and Foster, 2003). Three decarboxylases (GadD1, GadD2, and GadD3) and two antiporters (GadD1T1 and GadD2T2) were identified in L. monocytogenes. The GadD2/T2 system was found to be responsible for the survival of the cells in acidic conditions at pH 2.8. GadD1T1 plays a role in growth at moderately acidic pH values (5.1) (Cotter et al., 2005). Gene encoding GAD was identified in B. cereus ATCC 10987, while no gene encoding GABA glutamate exchanger was found. As a result, the gad gene was not up-regulated under low-pH exposure (Mols et al., 2007, 2010a,b). Unlike ATCC 10987, B. cereus ATCC 14579 genome does not contain gad gene. However, glutamate enhanced the resistance of B. cereus ATCC 14579 cells to pH 4.0 acid shock (Senouci-Rezkallah et al., 2011).

The system AD system is composed of a cytoplasmic arginine decarboxylase (AdiA) and an arginine/agmatine antiporter (AdiC). After proton-consuming decarboxylation of arginine by AD to give agmatine in the cell, the agmatine is carried out of the cell by the antiporter in exchange for arginine. The consumption of protons during decarboxylation reduces acidity in the cytoplasm. In E. coli, the presence of arginine during acid shock raised pH<sup>i</sup> from 3.7 to 4.7, showing that AD could enable pH<sup>i</sup> homeostasis during acid stress (Richard and Foster, 2004). In S. typhimurium, the AD system is active only under acid growth conditions (Kieboom and Abee, 2006; Alvarez-Ordonez et al., 2010c). In B. cereus, two genes have been annotated as encoding ADs [speA and yaaO, (Ivanova et al., 2003)], but no gene has been annotated as encoding an arginine/agmatine antiporter. However, it has been shown that the presence of arginine improved acid stress resistance of B. cereus cells. It is thus probable that B. cereus utilizes the AD system for surviving in acidic environment (Senouci-Rezkallah et al., 2011).

The lysine decarboxylase system is composed of a decarboxylase (CadA) and a lysine/cadaverine antiporter (CadB). After decarboxylation of lysine to cadaverine in the cell by proton-consuming lysine decarboxylase, the cadaverine is carried out of the cell by the antiporter in exchange for lysine. The consumption of protons during the decarboxylation reaction again lowers acidity in the cytoplasm, allowing the homeostasis of pH<sup>i</sup> during acid shock. In E. coli and Vibrio vulnificus, the lysine decarboxylase system is encoded by the cadBA operon. This operon is activated by CadC, and repressed by LysP (Neely and Olson, 1996; Rhee et al., 2005). In Vibrio parahaemolyticus, the expression of cadBA has been shown to increase in the presence of lysine. The mutation of the gene cadA also impaired survival in acid shock conditions relative to the wild-type strain, showing the role of this system in resistance to acid shock by maintenance of pH<sup>i</sup> (Tanaka et al., 2008). In B. cereus ATCC 14579, the gene encoding the enzyme lysine decarboxylase is yvdD (Ivanova et al., 2003), but the gene encoding the antiporter has not yet been identified. Like glutamate and arginine, the addition of lysine improves B. cereus resistance to acid stress, suggesting a role of the lysine decarboxylase system in this bacterium (Senouci-Rezkallah et al., 2011).

## **Buffering ability of cytoplasm**

fmicb-07-01550 October 1, 2016 Time: 13:47 # 6

Independently of the involvement of the decarboxylation systems, the pHi of a cell can be stabilized by the relatively high buffering ability of the cytoplasm. When acid or alkaline compounds enter the cell, the buffering action of the cytoplasm tends to offset these variations and keep the pHi neutral. Rius and Loren (1998) compared the buffering ability of Bacillus alcalophilus, an alkaliphilic bacterium, with those of B. subtilis and S. aureus, both neutrophilic bacteria. They showed that the buffering ability of B. alcalophilus was influenced by the culture pH and other conditions. In addition, this buffering ability was greater in anaerobic growth conditions than in aerobic growth conditions, clearly showing that the cells have a significantly greater buffering activity in a fermenting medium. In B. cereus, a study showed that cells grown in unregulated batches at pH<sup>e</sup> 7.0 showed a pH<sup>i</sup> of 9.0, which fell to 7.9 and 6.2 when the cell growth at pH 7.0 was followed by 1 h incubation at pH 6.3 and 4.6, respectively. It has also been observed that the pH<sup>i</sup> of cells adapted at pH 6.3 was higher (pH<sup>i</sup> 6.6) than that of nonadapted cells (pH<sup>i</sup> 6.1) (Browne and Dowds, 2002). However, in unregulated batch culture, the cell environment is not controlled and several factors, such as growth phase, growth rate, and carbon and oxygen resource availability can therefore also influence pH<sup>i</sup> .

### Cell Density

In addition to responses to several stresses, bacteria are known to regulate diverse physiological processes in a cell density-dependent manner. Cell density-dependent regulation appears to follow a common theme, in which a small, selfgenerated molecule is exported as the signal for intercellular communication, commonly called quorum sensing. Cell density was found to modulate acid adaptation in Streptococcus mutans log-phase cells, since pre-adapted cells at a higher cell density or from a dense biofilm displayed significantly higher resistance to the killing pH than the cells at a lower cell density (Li et al., 2001). The authors also showed that mutants defective in the comC, comD, or comE genes, which encode a quorum sensing system essential for cell density-dependent induction of genetic competence, had a diminished log-phase ATR. They concluded that optimal development of acid adaptation in S. mutans involves both low pH induction and cell–cell communication. Also in this strain, the gene luxS involved in the signal synthesis in quorum sensing plays a role in the regulation of tolerance to acid stress (Wen and Burne, 2004). The synthesis of LuxS is induced in E. coli on exposure to acetic acid, suggesting that its expression in this organism is induced at low pH (Frees et al., 2003b). Four quorum sensing systems operate in B. cereus and, one of them controls the oxidative stress response (Slamti et al., 2014). Exposure to acid stress induces secondary oxidative stress in B. cereus (Mols and Abee, 2011a). Therefore, the quorum sensing system that controls oxidative stress could be involved in acid adaptation.

# Protection and Repair of Proteins and DNA

In an acid stress situation, proteins can undergo modifications, from changes in conformation to complete denaturing, all of which will significantly affect their activity. Denatured proteins are dealt with chaperone proteins, such as DnaK/DnaJ and GroES/EL (Susin et al., 2006). In L. lactis, these chaperones are induced at pH 4.5, and not at pH 5.5, and therefore respond at a certain level of acidity in the medium, or a certain concentration of denatured proteins in the cytoplasm (Frees et al., 2003b). In S. mutans, the expression of dnaK and the quantity of DnaK are higher in acid-adapted cells than in non-adapted cells, and increase in response to an acid shock (Jayaraman et al., 1997), suggesting that the regulation of dnaK by pH is transcriptional in this bacterium. In S. typhimurium DT104, an increase level of DnaK and GroEL was observed in acid-adapted cells (Berk et al., 2005). Exposure of L. lactis to low pH revealed that, in addition to DnaK and GroEL, several heat shock proteins are part of the acid shock response. Among them are ClpE and ClpP (Frees et al., 2003a). The Clp proteins are ATPase-dependent proteases involved in the turn-over of denatured proteins: they are responsible for the rapid degradation of damaged proteins and the regulation of the levels of some proteins in the cell (enzymes and regulators). In B. cereus, DnaK is overproduced in response to acid shock (Browne and Dowds, 2002). Proteins of 66 and 59 kDa, which could be, respectively, DnaK and GroEL, have also been found expressed in stationary phase cells irrespective of pH (Jobin et al., 2002). More recent study showed that chaperoneencoding genes dnaK and groES and protease-encoding gene clpC were up-regulated upon exposure to sublethal acid shocks (Mols et al., 2010b), suggesting that these proteins are involved in acid stress response in this bacterium.

When pH<sup>i</sup> becomes too acidic, DNA loses purine and pyrimidine units (Cotter and Hill, 2003). First the bases are protonated and then, the glucoside bonds are cleaved. The ultraviolet (UV) excinulease system UvrA-UvrB which is known to repair damage caused to DNA by UV radiation or exposure to many chemical agents (Lage et al., 2003) is associated with low-pH adaptation. The UvrA–UvrB complex identifies the changes in conformation or structure in the damaged DNA. UvrA then dissociates from UvrB, which remains bound to the DNA. UvrC in turn binds to the UvrB-ADN complex and effects the incision of seven nucleotides. UvrD intervenes to remove UvrC and the damaged nucleotides. UvrB remains bound to the DNA until DNA polymerase I synthesizes the excised sequence (Skorvaga et al., 2004). In S. mutans, uvrA is induced by acidity. In addition, a 1uvrA mutant is more sensitive to growth at pH<sup>e</sup> 5.0 than the wild-type strain, and when it is pre-incubated at a non-lethal acidic pH, it is unable to survive at pH 3.0. This indicates that UvrA is involved in adaptive response to low pH (Hanna et al., 2001). In Methylobacterium dichloromethanicum, a 1uvrA mutation significantly limited viability and growth on dichloromethane, when intracellular hydrochloric acid is produced. This dehalogenation produced a genotoxic intermediate compound. However, it is not yet known whether DNA lesions are caused by this reaction product or by the acidity (Kayser et al., 2002). Mutations in the genes recA (coding for a protein involved in homologous recombination and which is a regulator of the SOS response) and uvrB caused a significant decrease in tolerance to low pH in H. pylori (Thompson and Blaser, 1995; Thompson et al., 1998), showing the importance of these DNA repair mechanisms for survival in acid conditions. In B. cereus, the uvrA and uvrB genes are upregulated upon exposure to pH 5.5 set with lactic acid, suggesting a role of the UvrA–UvrB system in stress acid resistance (Mols et al., 2010b).

# Concluding Remarks

fmicb-07-01550 October 1, 2016 Time: 13:47 # 7

Although much has been learned about other bacteria respond to acid stresses, there remains a great deal to discover, making B. cereus acid stress responses a fruitful and important area of future research. In addition, the majority of studies that have analyzed the response of food-borne pathogens to acid stress were performed in presence of oxygen. Anoxia may thus provide attractive condition to study the acid resistance mechanisms in B. cereus and other food-borne pathogens.

# OXYGEN SENSING AND ADAPTATION TO ANOXIA

Bacillus cereus is a facultative anaerobic microorganism, i.e., it can survive at various levels of oxygenation. The common terminology to describe the various oxygen conditions, such encountered by B. cereus in food and GI tract is based on a comparison with the atmospheric level, which is 21% (v/v), a level referred to as normoxia. Hypoxia is a condition where oxygen concentration is lower than 21%. No free oxygen is available under anoxia.

# B. cereus Central Metabolism and Redox Homeostasis

Whatever the oxygenation conditions, B. cereus catabolizes glucose through glycolysis and, to a lesser extend, through the pentose phosphate pathway (PPP; Zigha et al., 2006). Glycolysis and PPP are fuelled by the phosphoenolpyruvate (PEP)-dependent phosphotransferase system (PTS), a transport system that cells use to bring glucose into the cytoplasm using energy transferred by PEP (Deutscher et al., 2006; Cao et al., 2011; Warda et al., 2016). Glycolysis produces two molecules of pyruvate per molecule of activated glucose (glucose-6-phosphate, G-6P) and, in the process, reduces two molecules of NAD<sup>+</sup> to NADH and produces a net gain of two ATP molecules (**Figure 2**). In addition to providing ATP by substrate-level phosphorylation, glycolysis is a major source of metabolic intermediates for biosynthetic pathways. The processing of activated glucose (G-6P) through the PPP also produces biosynthetic precursors. Two of these biosynthetic precursors, ribose-5-phosphate, and erythrose-4-phosphate, are essential for the synthesis of nucleotides, histidine, and aromatic amino acids. In addition to providing biosynthetic intermediates, the PPP generates two molecules of NADPH per molecule of G-6P. NADPH provides the reducing power that drives numerous anabolic reactions, including those responsible for the biosynthesis of all major cell components (Spaans et al., 2015). NADPH is also required to maintain and regenerate the cellular detoxifying and antioxidative defense systems (Agledal et al., 2010). The antioxidant defense system of B. cereus is constituted by an elaborate, often overlapping network of enzymes, such as superoxide dismutase (SOD), catalase, flavohemoglobin, peroxiredoxins, thioredoxins, among others, and low molecular mass (LMW) thiols such bacillithiol (BSH), coenzyme A (CoASH) and cysteines (Newton et al., 1996; Fang et al., 2013). In addition to their roles in detoxification of ROS, LMW thiols, which are present in millimolar concentrations in the cytoplasm, function as major thiol-redox buffers to maintain the redox state of the cytoplasm (Newton et al., 2009; Loi et al., 2015). The PPP and glycolysis are linked by transketolase and transaldolase that convert ribose-5-phosphate into glyceraldehyde-3-phosphate and fructose-6-phosphate. Fructose-6-phosphate is a precursor for N-acetylglucosamine, which is required for BSH (Helmann, 2011). Increased carbon flow through the PPP is often associated with stressful conditions or infections in Grampositive pathogens (Eisenreich et al., 2006; Bergman et al., 2007).

In the presence of oxygen, pyruvate is converted to acetyl coenzyme A (Acetyl-CoA) by the pyruvate dehydrogenase complex (Pdh) (Duport et al., 2006). Acetyl-CoA can enter into the tricarboxylic acid cycle (TCA) via a condensation reaction with oxaloacetate that is catalyzed by citrate synthase. Then two carbons are lost as CO<sup>2</sup> for every two carbons (i.e., Acetyl-CoA) that enter the TCA. TCA provides three biosynthetic intermediates that are critical for the novo synthesis of many amino acids and porphyrins: oxaloacetate, α-ketoglutarate, and succinate/succinyl-CoA. The reducing equivalents generated by the glycolysis and TCA (NADH and FADH) are reoxidized through the aerobic respiratory chain, resulting in the build-up of a proton motrice force and the subsequent synthesis of ATP by ATP synthase (38 ATP per molecule of consumed glucose). Although it has not been thoroughly studied, the aerobic respiratory chain of B. cereus resembles the respiratory chain of B. subtilis (Contreras-Zentella et al., 2003; Rosenfeld et al., 2005; Garcia et al., 2008; Melo and Teixeira, 2016). It contains two major branches, one quinol oxidase branch (with cytochrome bd or cytochrome aa3 as its terminal oxidase) and one cytochrome oxidase branch (with cytochrome caa3 as its terminal oxidase) (**Figure 2**). The terminal oxidases catalyze the four-electron reduction of dioxygen to two water molecules. The aerobic respiratory chain is not only a main source of ATP but it also a major source of reactive oxygen species (ROS; Gonzalez-Flecha and Demple, 1995; Messner and Imlay, 2002; Mailloux et al., 2011). ROS generation starts with the formation of a superoxide anion (O•− 2 ). Within the respiratory chain, NADH: menaquinone oxidoreductase (complex I) and bc complex (complex III) are generally considered as the main producers of O•− 2 (**Figure 2**). Dismutation of O•− 2 (either spontaneously or through a reaction catalyzed by SODs) produces hydrogen peroxide (H2O2), which in turn may be fully reduced to water or partially reduced to hydroxyl radical (OH• ), one of the strongest oxidants in nature (Imlay, 2013). ROS production serves as a metabolic signal and under normal conditions are quenched by the antioxidant defense system to maintain them to non-toxic levels. However, when released in excess under certain stress conditions such as hypoxia and change in pH that abruptly affect the electron transport chain, ROS can also directly damage cells (Mols et al., 2009, 2010b, 2011).

When the oxygen concentration drops to a level at which oxygen becomes limiting as a substrate for cytochrome c oxidase, ATP production via oxidative phosphorylation is no longer able to meet cellular demands for ATP. This can be compensated by the activation of glycolytic activity to increase ATP production by substrate-level phosphorylation. However, compared with oxidative phosphorylation, the ATP production through glycolysis alone is much lower. Therefore, glycolytic activity must be strongly up-regulated under hypoxic conditions to generate sufficient ATP. This phenomenon, namely the Pasteur effect requires the efficient recycling of NAD<sup>+</sup> from NADH, otherwise glycolysis will become limited by the availability of NAD+. Therefore, the fermentative pathways are induced by activating the expression of key enzymes.

Under anoxia and in absence of external electron acceptor, B. cereus carries out mixed acid-butanediol fermentation (**Figure 2**). Lactic acid is the major by-product of fermentation (more than 60% of total production) both at neutral and acidic pH (Duport et al., 2004, 2006; Messaoudi et al., 2010; Le Lay et al., 2015). The pyruvate-to-lactate pathway involves three Llactate dehydrogenases, LdhA, B, and C. It has been shown that LdhA exerted a major control on both B. cereus fermentative growth and enterotoxin production (Laouami et al., 2011). The conversion of pyruvate to acetyl-CoA and formate requires high levels of pyruvate formate lyase (Pfl) compared to Pdh complex, probably to avoid excessive NADH formation under fermentative conditions (Duport et al., 2006). The acetyl-CoA is converted to acetate through the ATP-producing acetate pathway and to ethanol through the NADH-recycling ethanol pathway. At neutral pH, acetate and formate are produced in similar amounts (each accounting for ∼15% of total production), while ethanol, and to higher extend succinate and 2,3 butanediol are minor fermentation products (Duport et al., 2004; Rosenfeld et al., 2005). However, the relative rate of formation of all these glucose by-products is influenced by the ORP of the growth medium (Zigha et al., 2006). At acidic pH, production of 2,3-butanediol highly increased at the expense of acetic acid and succinic acid and to lesser extend lactic acid (Le Lay et al., 2015). During fermentative growth, TCA functions only to supply biosynthetic precursors and is transformed from a cyclic pathway to two oppositely oriented half cycles (**Figure 2**). In fermenting cells, the direct formation of ROS is abolished by the absence of oxygen. However, several anoxia-specific alterations can promote the oxidative response (Lumppio et al., 2001; Rusnak et al., 2002). The occurrence of an oxidative component in response to oxygen deprivation has been confirmed in B. cereus by microarray studies on the whole genome level and by proteomic studies (Mols and Abee, 2011b; Clair et al., 2012; Madeira et al., 2015).

Under anoxia, B. cereus can growth via nitrate ammonification (Zigha et al., 2006). Nitrate in the human intestine originates both from endogenous synthesis and dietary products rich in nitrate (Tannenbaum et al., 1978; Lidder and Webb, 2013). During nitrate respiration, nitrate is reduced by the respiratory nitrate reductase (NarGHI) to nitrite in B. cereus cells. Nitrite is further reduced to ammonia by a general nitrite reductase (NasDE). Nitrate reduction is coupled to ATP generation through proton motrice force (Rosenfeld et al., 2005). Due to the drastically different ATP yields of respiratory and fermentative processes, B. cereus uses a fine-tuned regulatory system to maintain the most efficient mode of ATP generation under anoxia (Zigha et al., 2006). Nitrate-respiring B. cereus cells produce nitric oxide (NO) as an intermediate product of nitrate reduction to N2O (Kalkowski and Conrad, 1991). The chemical properties of NO make this gas a good candidate for a signaling molecule (Gusarov and Nudler, 2012). Gram-positive bacteria, including B. cereus can also generate NO through a bacterial analog of mammalian NO synthase (bNOS) in presence of oxygen. It has been shown that bNOS from B. subtilis, B. anthracis displayed NO-forming activity dependent on arginine (Adak et al., 2002; Gusarov et al., 2008). The bNOS-mediated NO was implicated in the protection of bacteria against oxidative stress, a variety of antibiotics and other stresses such as acid stress (Tan et al., 2010; Gusarov and Nudler, 2012). NO also regulates growth and pathogenicity of B. anthracis (Popova et al., 2015).

# Exoproteins and Virulence Factors Supporting Pathogenesis

Bacillus cereus excretes high level of proteins into the extracellular medium. However, the level of excreted proteins is lower during anaerobic fermentative growth than under aerobic respiratory growth (Madeira et al., 2016a,b). Like transcription and translation, exportation of proteins is an energetically expensive process. Therefore, the decrease of protein excretion in fermentative cells may be attributed to decreased energy availability. Most exoproteins are secreted as precursors with a cleavable N-terminal signal sequence, but a significant fraction is secreted by non-classical pathways, i.e., without signaling peptides and sequence motifs for surface anchoring. Six signal peptide-dependent pathways are currently recognized in Grampositive bacteria (Schneewind and Missiakas, 2014): the general secretory (Sec) pathway, the twin arginine targeting (Tat) pathway, the fimbrillin-protein exporter, the flagellar export apparatus, the holins, and the ESAT-6/WXG100 secretion system. The Sec pathway is considered the general housekeeping protein translocation system and is essential in B. cereus (Fagerlund et al., 2010; Senesi and Ghelardi, 2010; Senesi et al., 2010). The proteins arising from cellular secretion and other protein export mechanisms are components of the B. cereus exoproteome (Armengaud et al., 2012). B. cereus exoproteome has been the focus of several shotgun proteomic studies (Clair et al., 2010, 2013; Laouami et al., 2014; Madeira et al., 2015, 2016b). These proteomic studies identified up to 377 different exoproteins. Among them 65 putative virulence factors were identified, including 15 toxin-related proteins, 12 motility-related proteins and 36 adhesins and degradative enzymes (Madeira et al., 2015, 2016a,b). These virulence-related proteins represent more than 85% of exoproteins whatever the growth condition (Clair et al., 2010; Madeira et al., 2015). B. cereus exoproteome includes numerous cytoplasmic proteins involved in metabolic pathways (mainly glycolysis) and oxidative stress response. Many of these proteins are conserved in the exoproteome of pathogens. A significant number of these extracellular cytoplasmic proteins have been found to serve two or more functions and are

referred as "moonlighting" proteins (Henderson and Martin, 2011; Gotz et al., 2015). Moonlight proteins have been shown to localize at the cell surface and participate in adhesion, colonization and virulence (Henderson and Martin, 2011; Ebner et al., 2016a,b). Surface-associated moonlight proteins have been reported to be reversible and pH dependent (Nelson et al., 2001; Antikainen et al., 2007). Extracellular cytoplasmic proteins are mainly excreted during the stationary growth phase (Yang et al., 2011) and in B. cereus, their time dynamic is negatively correlated to the dynamic of toxin-related proteins, indicating that a specific selection process has to occur (Madeira et al., 2015).

The toxin-related proteins found in B. cereus exoproteome include the lytic components (L1 and L2) and the binding component (B) of Hemolysin BL (Hbl). Hbl requires all three components for full activity (Beecher et al., 1995). Hbl may form a pore similar to other soluble channel-forming proteins in host cell membranes (Madegowda et al., 2008; Stenfors Arnesen et al., 2008). The tripartite Hbl complex is encoded by genes clustered into a polycistronic operon with the transcriptional order hblC, hblD, and hblA (Ryan et al., 1997). An ORF, name hblB, is located immediately downstream of hblCDA in the B. cereus ATCC 14579 genome and is transcribed independently (Clair et al., 2010). hblB encodes HblB', which is structurally related to the B component of the Hbl complex. Its activity is currently unknown (Clair et al., 2010). NheA, NheB, and NheC are the three components of the non-hemolytic enterotoxin (Nhe) and are encoded by the nheABC operon (Lindback et al., 2004). All three components NheA, NheB, and NheC are required for full toxic activity, although NheC is only expressed in small amounts due to translational repression (Lindback et al., 2004). Nhe is a poreforming toxin (Didier et al., 2012; Phung et al., 2012). Cytotoxin K (CytK) is a single-component protein toxin and belongs to the family of β-barrel pore-forming toxins (Baida et al., 1999; Lund et al., 2000). CytK possesses dermonecrotic, cytotoxic, and hemolytic activities (Lund et al., 2000; Hardy et al., 2001). HlyI is a thiol-activated cholesterol-binding cytolysin (Kreft et al., 1983; Minnaard et al., 2001; Ramarao and Sanchis, 2013). All the genes encoding Hbl, Nhe, CytK, and HlyI belong to the PlcR virulence regulon (Gohar et al., 2008). Hemolysin II (HlyII) is cytotoxic due to its ability to disrupt cellular and artificial membranes by pore formation (Andreeva et al., 2006). EntFM exhibits three protein– protein interaction SH3 domains and a NlpC/P60 domain that shares similarities with cell wall peptidase. There is controversy over the role of EntFM in B. cereus cytotoxicity, but wide consensus on its role in pathogenicity (Boonchai et al., 2008; Tran et al., 2010). The structurally related proteins EntA, EntB, EntC, and EntD were annotated as "enterotoxin/cell wall-binding proteins" because they possess, in addition to two SH3 domains, an extracellular cell wall-binding 3D domain (Clair et al., 2010). Recent study showed that EntD plays a crucial role in maintaining cell wall structure and that, in the absence of EntD, B. cereus cells are able to reoriente their metabolism to maintain cell wall integrity. Such adaptation program leads to decreased virulence factor production, specifically Nhe and Hbl production (Omer et al., 2015). The functions of EntA, EntB and EntC are currently unknown.

Taken together, the 15 toxin-related proteins found in B. cereus exoproteome represent more than 30% of exoproteins during the exponential growth phase, suggesting that they may have an important cellular function for the producer bacterium (Clair et al., 2010). Toxin-related proteins contain methionine residues that are susceptible to intracellular oxidation in both respiratory and fermenting cells (Madeira et al., 2015). Methionine residues of proteins are known to act as ROS scavengers (Luo and Levine, 2009). High level secretion of toxin-related protein during active growth may thus contribute to the protection of B. cereus cells against cellular oxidation and maintain redox homeostasis by keeping endogenous ROS at bay whatever the oxygen condition.

The regulation of B. cereus toxin gene expression mobilizes a complex machinery (Ceuppens et al., 2011; Jessberger et al., 2015) that includes the virulence regulator PlcR (Salamitou et al., 2000; Gohar et al., 2008) and several transcriptional regulators that coordinately control metabolic and virulence genes such as the CodY repressor (Lindback et al., 2012; Bohm et al., 2016), the ferric uptake regulator Fur (Sineva et al., 2012), the catabolite control protein A [CcpA, (Van der Voort et al., 2008)] and the redox regulators, Fnr, ResD, Rex, and OhrR. All these four redox regulators are able to regulate directly the expression of hbl and nhe after binding to the promoter region of hblCDA and nheABC (Esbelin et al., 2008, 2009; Clair et al., 2013; Laouami et al., 2014). In addition, it was shown that ResD and Fnr form a ternary complex with the virulence regulator PlcR (Esbelin et al., 2012).

# Redox Regulators that Coordinate Central Metabolism and Production of Toxin-Related Proteins

A number of classical sensors/regulators are employed by various species of bacteria to sense oxygen changing conditions (Bueno et al., 2012). These sensors/regulators include Fnr, ResDE, Rex, and OhrR (**Figure 3**). All of them could orchestrate the expression of virulence determinants in B. cereus, both directly, and indirectly, by impacting key metabolic and regulatory circuits.

### B. cereus Fnr

The B. cereus Fumarate and nitrate reductase regulator Fnr is a member of the Crp/Fnr (cyclic AMP-binding protein/fumarate nitrate reduction regulatory protein) family of helix-turn-helix transcriptional regulators (Korner et al., 2003). Like all the members of the Crp/Fnr family, B. cereus Fnr contains an N-terminal region made up of antiparallel β-strands able to accommodate a nucleotide, and a C-terminal extension with four cysteine residues that coordinate a [4Fe-4S]2<sup>+</sup> cluster (Esbelin et al., 2009). Fnr is essentially present in the apo-form (clusterless) in aerobically grown cells, and in the holo-form in anaerobically grown cells (Esbelin et al., 2012). Under aerobiosis, Fnr is able to sense oxygen concentration changing and probably ROS and NO (Jiang et al., 2016) through the oxidation of Cys thiol groups, which link two monomers in an inactive dimer. The inactivation of Fnr by oxygen is reversible. In the current model for Fnr function, the active dimeric forms of Fnr bind to the promoter region of target operons/genes to activate or repress

Binding of ResD to DNA is not dependent on ResD phosphorylation status, which is regulated by the kinase and phosphatase activities of ResE. Rex senses the intracellular redox status through changes in the NADH/NAD<sup>+</sup> ratio. Unlike the NAD+-bound Rex form, the NADH-bound Rex form is incapable of binding DNA. OhrR senses oxygen concentration or ROS through cysteine residues (see text). OhrR is a non-covalent dimer in its reduced form and a covalent dimer in its oxidized form. Dihydrolipoate (DHLA) and low molecular mass thiols (LMW) participate in the recovery of reduced OhrR from the oxidized form. B. cereus OhrR can bind DNA both under its reduced and oxidized form.

their transcription. Fnr plays a key role within the regulatory cascade governing fermentative pathways in B. cereus because (i) its transcription is strongly induced in fermenting cells and (ii) the inactivation of its gene abolishes fermentative growth. The role of Fnr under anaerobic and aerobic respiratory growth is more moderated (Zigha et al., 2007).

### B. cereus ResDE

The ResDE two-component signal transduction system consists of a membrane-bound histidine sensor kinase (ResE) and a cytoplasmic response regulator (ResD). The resD and resE genes compose a transcriptional unit included into a larger operon that comprises resABC; these three genes encode proteins similar to those involved in cytochrome c biogenesis (Duport et al., 2006). The B. cereus resABCDE locus is organized similarly to that in B. subtilis and B. anthracis (Sun et al., 1996; Wilson et al., 2008). The ResDE two-component system regulates the expression of several genes of the fermentative and respiratory pathways in B. cereus. However, it appeared to exert a more important role in anaerobic fermentative pathways (with a more pronounced effect under high reductive conditions) than in aerobic respiratory pathways. Unlike fnr, the resDE mutation did not abolish the fermentative growth of B. cereus, indicating that although it plays an important role, it is not indispensable for B. cereus. The ResDE system is modulated primarily by the autophosphorylation activity of ResE at a conserved histidine residue. The redox signal activating ResE has not been identified in B. cereus but have been postulated to be the redox state of menaquinones in B. subtilis under aerobiosis (Geng et al., 2007). Upon activation, ResE donates a phosphate to its cognate regulator, ResD. The phosphatase activity of ResE controls the level of phosphorylated ResD (ResD∼P). In Bacillus subtilis, phosphatase activity of ResE is regulated by oxygen availability and anaerobic induction of the ResDE regulon is partly due to a reduction of the ResE phosphatase activity during anaerobiosis (Nakano and Zhu, 2001). ResE is the only relevant kinase able to phosphorylate ResD. However, it was proposed that acetyl-phosphate (produced through the acetate pathway) could transfer its phosphate to ResD (Gueriri et al., 2008), linking its phosphorylation state to the metabolic status of the cell. Both phosphorylated and unphosphorylated forms of dimeric B. cereus ResD are able to bind DNA but their DNA binding affinity depends on promoter architecture of the target genes. For example, phosphorylation of ResD, which is higher under anaerobiosis than under aerobiosis enhances its ability to bind to its own promoter and fnr promoter but not to the enterotoxin gene promoters. Both ResD and ResD∼P physically interacts with Fnr and simultaneously bind their target promoter. A model was proposed on which ResD∼P may act as a Fnr co-activator and ResD as a Fnr anti-activator (Esbelin et al., 2012).

### B. cereus Rex

Changes in oxygen availability and ORP influence the relative level of dinucleotides NAD<sup>+</sup> and NADH in the cells, and such changes are sensed by the transcriptional regulator Rex. The crystal structure of Rex from Thermus aquaticus, Thermus thermophilus in complex with NADH and of B. subtilis Rex without cofactor has been determined (Sickmier et al., 2005; Wang et al., 2008). Rex is composed of two structural domains, an N-terminal domain that adopts a winged helix-turn-helix fold that most likely interacts with DNA, and a C-terminal NADH binding domain. In the complex with NADH, the N-terminal domains pack close to each other in a compact dimer. This conformation of Rex is unable to bind DNA. Rex is thus active as a repressor only when the NAD+/NADH ratio indicates adequate NAD+. By monitoring the NAD+/NADH ratio, Rex helps the cells to regulate pathways that regenerate NAD+. Typically, B. cereus Rex regulates the carbon flow distribution at the pyruvate node by favoring and limiting the carbon flow entry into the NADH-recycling lactate pathway under anoxic and oxic conditions, respectively (Laouami et al., 2014). By controlling this carbon flow, Rex also controls the availability of glycolytic intermediates for macromolecular synthesis as well as supporting NADPH production through different enzymes located in the TCA cycle, glycolysis and PPP (Laouami et al., 2014). In addition to fine-tune the levels of the prooxidant NADH and the antioxidant NADP, Rex regulates directly the synthesis of antioxidant systems like the OhrRA system (Laouami et al., 2014). Rex also regulates directly the expression of fnr, resD, and ldhA. LdhA activity is a critical factor in B. cereus virulence (Laouami et al., 2011).

### B. cereus OhrRA

The B. cereus OhrRA (organic hydroperoxide resistance) system comprises a thiol-dependent peroxidase protein (OhrA), which functions as a low ORP sensor under anoxic conditions and a redox-sensing transcriptional regulator (OhrR), which belongs to the MarR family of winged helix-turn-helix DNA binding protein (Panmanee et al., 2006; Dubbs and Mongkolsuk, 2007). The genes encoding OhrR and OhrA form a bicistronic transcriptional unit (Clair et al., 2012). OhrA is usually reported as a protein that detoxifies the organic hydroperoxides (OHP). OHP could result from oxidation of unsaturated FA by molecular oxygen or ROS (Mongkolsuk et al., 1998). This may explain why OhrA is induced under normoxia and low-ORP anoxia where ROS are the by-product of secondary oxidative stress. B. cereus OhrR is an atypical OhrR protein because it contains four cysteine residues at its N-terminal domain while most OhrR regulators contain two cysteines. Like its orthologs in other bacteria, B. cereus OhrR functions as a transcriptional regulator that binds to its own promoter under its dimeric reduced form. However, unlike most of its orthologs, it may function as a repressor and an activator of several metabolism-related genes. OhrR is mainly a non-covalent dimer in its reduced form and a covalent dimer in its oxidized form. Dihydrolipoate and LMW participate in the recovery of reduced OhrR from the oxidized form and thus may control the activity of the redox-sensing OhrR regulator (Clair et al., 2013). Besides to regulate the antioxidant system, which includes OhrA, B. cereus OhrR controls the abundance level of key enzymes of central metabolism. In this way, (i) it modulates the glycolytic flux and restricts B. cereus growth under low ORP fermentative conditions and (ii) it sustains high TCA capacity and limits energy spilling through an overflow metabolism under aerobic respiratory growth.

### Interactions among the Redox Regulators

fmicb-07-01550 October 1, 2016 Time: 13:47 # 12

Although each of the global redox sensors senses different signals, they interact with each other and with operon-specific promoters. Transcriptional activation by Fnr is probably the first response to changing oxygen availability. Fnr regulates its own transcription and the transcription of resDE and rex, thereby making ResD and Rex more active/inactive as regulators of both catabolic pathways that supply the metabolic intermediates necessary to synthetize enterotoxins, and enterotoxin gene expression. Because Fnr and ResD interact with each other, the change of ResD level generated by Fnr accentuates or reduces this interaction. In addition, the changes of menaquinone redox state and NAD+/NADH ratio induced by changing oxygen concentrations also affect the activities of ResD and Rex, respectively; this impacts the expression of fnr and their own expression. By regulating the synthesis of the OhrRA system, which is indirectly stimulated by ROS under both normoxia and low-ORP anoxia, Rex also impacts indirectly carbon flow, and enterotoxin synthesis. In conclusion, the regulatory network involving redox sensors is undoubtedly complex but permits the microorganism to coordinate efficiently its central metabolism with enterotoxin production (**Figure 4**).

# Concluding Remarks

Maintaining an appropriate redox balance is essential for B. cereus adaptation to changing oxygen availability, and probably for resistance to acid stress (Liu et al., 2016). Redox homeostasis depends of the antioxidant system, which enables bacterial cells to maintain proteins and other cellular components in active state for metabolism. However, to date, we lack knowledge on the intracellular B. cereus environment, the behavior of redox couples under different environmental conditions, and the mechanisms of sustained redox homeostasis in B. cereus. In particular, a fundamental challenge is to understand how antioxidant-oxidant interactions modulate B. cereus pathogenesis.

# CONCLUSION

Bacillus cereus adaptation to acid and low oxygen environments follows pathways that look quite similar to those that have been examined and described in great detail for other bacteria. However, there are also several clear differences that warrant the further examination of the adaptation mechanisms of B. cereus to its challenging environments from both a fundamental and industrial point of view.

What is clearly lagging in B. cereus research is knowledge of the possible roles that small RNAs (sRNAs) may play in stress responses. Indeed, it is now evident that sRNAs play an essential role in gene regulation under various stress conditions (Babu et al., 2011; Hoe et al., 2013; Miller et al., 2014). Thus, it is crucial that we uncover this important regulatory layer in B. cereus, as this will certainly lead to new insights and refine our understanding of the way in which B. cereus withstands environmental stress. Another issue that requires attention is the importance of cell individuality in responding to stressors. Cells in a bacterial population, even in a very uniform environment, may differ considerably with respect to the genetic program that is operative under these conditions. Such flexibility occurs in B. cereus spores (Van Melis et al., 2014). It would be interesting and important to find out whether culture heterogeneity also plays a role in the responses of B. cereus toward acid stress and low oxygen tension (Pandey et al., 2016). Overall, it is

expected that investigations of the stress physiology of B. cereus will continue to be central to understanding its behavior in challenging environments. Omics approaches available today will lead to important discoveries that can be applied in food safety.

# REFERENCES

fmicb-07-01550 October 1, 2016 Time: 13:47 # 13


# AUTHOR CONTRIBUTIONS

CD wrote the paper. MJ and PS contribute to the writing of the part one of the manuscript.

the impact of CodY on virulence heterogeneity. Front. Microbiol. 7:768. doi: 10.3389/fmicb.2016.00768


with an ability to grow at low pH. Appl. Environ. Microbiol. 71, 2832–2839. doi: 10.1128/AEM.71.6.2832-2839.2005




and cloning of the gene encoding this enzyme. J. Dent. Res. 80, 371–377. doi: 10.1177/00220345010800011301



phase is not due to cell lysis. J. Bacteriol. 193, 5607–5615. doi: 10.1128/JB. 05897-11


in Bacillus cereus F4430/73. J. Bacteriol. 189, 2813–2824. doi: 10.1128/JB. 01701-06

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

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

# The Biofilm Lifestyle Involves an Increase in Bacterial Membrane Saturated Fatty Acids

Florence Dubois-Brissonnet\*, Elsa Trotier and Romain Briandet

Micalis Institute, Institut National de la Recherche Agronomique, AgroParisTech, Université Paris-Saclay, Jouy-en-Josas, France

Biofilm formation on contact surfaces contributes to persistence of foodborne

#### Edited by:

Lorena Ruiz, Universidad Complutense de Madrid, Spain

#### Reviewed by:

Chiara Montanari, University of Bologna, Italy Bernadette Dora Gombossy De Melo Franco, University of São Paulo, Brazil

#### \*Correspondence:

Florence Dubois-Brissonnet florence.duboisbrissonnet@agroparistech.fr

#### Specialty section:

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

Received: 12 July 2016 Accepted: 06 October 2016 Published: 28 October 2016

#### Citation:

Dubois-Brissonnet F, Trotier E and Briandet R (2016) The Biofilm Lifestyle Involves an Increase in Bacterial Membrane Saturated Fatty Acids. Front. Microbiol. 7:1673. doi: 10.3389/fmicb.2016.01673 pathogens all along the food and feed chain. The specific physiological features of bacterial cells embedded in biofilms contribute to their high tolerance to environmental stresses, including the action of antimicrobial compounds. As membrane lipid adaptation is a vital facet of bacterial response when cells are submitted to harsh or unstable conditions, we focused here on membrane fatty acid composition of biofilm cells as compared to their free-growing counterparts. Pathogenic bacteria (Staphylococcus aureus, Listeria monocytogenes, Pseudomonas aeruginosa, Salmonella Typhimurium) were cultivated in planktonic or biofilm states and membrane fatty acid analyses were performed on whole cells in both conditions. The percentage of saturated fatty acids increases in biofilm cells in all cases, with a concomitant decrease of branched-chain fatty acids for Gram-positive bacteria, or with a decrease in the sum of other fatty acids for Gram-negative bacteria. We propose that increased membrane saturation in biofilm cells is an adaptive stress response that allows bacteria to limit exchanges, save energy, and survive. Reprogramming of membrane fluidity in biofilm cells might explain specific biofilm behavior including bacterial recalcitrance to biocide action.

#### Keywords: biofilm, membrane lipids, fatty acids, membrane fluidity, lipidomics

# INTRODUCTION

Biofilms are surface-associated communities embedded in a self-produced extracellular polymeric substances and organized in a three-dimensional structure (Costerton et al., 1987; Giaouris et al., 2015). Foodborne pathogens such as Listeria monocytogenes (Smith et al., 2009; Carpentier and Cerf, 2011; Srey et al., 2013), Salmonella (Smith et al., 2009; Dubois-Brissonnet, 2012; Giaouris et al., 2014; Yang et al., 2016), Yersinia enterocolitica (Mosteller and Bishop, 1993), Staphylococcus aureus (Di Ciccio et al., 2015), Escherichia coli EHEC (Smith et al., 2009), or Campylobacter (Smith et al., 2009) form biofilms on food contact surfaces. Microbial deposits on wet surfaces, in particular floors and surfaces of equipment, are now recognized to be the main cause of pathogen persistence in food environments. They can be a periodic source of bacterial pathogens contaminating food

products during their transformation and can thus lead to foodborne intoxications (Stocki et al., 2007; Dubois-Brissonnet, 2012; Giaouris et al., 2012; Srey et al., 2013). As an example, surface contamination was identified as a cause in 34% of outbreaks occurring in collective catering in France (INVS, 2014).

The essential contributing factor explaining pathogen persistence in food-processing environments is the lack of disinfection efficacy. Despite extensive use of disinfectants on food-contact surfaces, these procedures are less effective when applied on biofilms compared to their effects on freeliving bacteria (Bridier et al., 2011). Depending on the species and the biocide considered, biofilms cells can be 1–1000 times more tolerant than their planktonic counterparts. The mechanisms involved in biofilm tolerance to antimicrobial treatments are multifaceted. They are in particular associated with heterogeneous metabolic activity and cell adaptive responses that are specific to physical and chemical microenvironments within the biofilm (e.g., varied conditions of pH, osmotic strength, nutrients or exposure to sublethal concentrations of biocide; Bridier et al., 2011; Giaouris and Nesse, 2015). Biofilm cell transcriptomic or proteomic profiles are studied since 2000' (Sauer, 2003; Resch et al., 2006; Vilain and Brozel, 2006), revealing up- or down- regulated functions at different stages of biofilm formation compared to their free cell counterparts. Surprisingly, few studies focused on biofilm lipidomics despite the involvement of membrane fatty acid composition in bacterial adaptation to fluctuating environments (Denich et al., 2003). Lipids are the main constituents of the cytoplasmic membrane and are essential for bacterial integrity, survival and growth, by allowing passive permeability of hydrophobic compounds and by modulating the function of membrane-associated proteins. Fluidity and permeability of the membrane rely on lipid acyl chain composition (Parsons and Rock, 2013). To survive despite environmental disturbances such as sub-optimal temperatures or the presence of toxic compounds at sub-lethal doses and to sustain optimum membrane fluidity, bacterial cells can alter the acyl chain structure of membrane glycerophospholipids by changing the ratios of: (1) saturation to unsaturation, (2) cis to trans unsaturation, (3) branched to non-branched structures, (4) acyl chain length, and (5) the synthesis of cyclopropane fatty acids (CFA; Denich et al., 2003; Loffhagen et al., 2007). It has also been shown that free exogenous fatty acids (EFA) available in the growth environment can alter the bacterial fatty acids (FA) composition (Brinster et al., 2009, 2010).

The objective of this study was to investigate how membrane fatty acid composition is adjusted when pathogenic bacteria are grown in the biofilm state and further, how this adjustment is done when some free exogenous FA are available in the environment. To this end, we first compared FA profiles of two Gram-positive bacteria, S. aureus and L. monocytogenes, and two Gram-negative bacteria, Pseudomonas aeruginosa and Salmonella Typhimurium, when grown in the planktonic state (harvested in exponential or stationary phase) or in the biofilm state (on polystyrene plates). Afterward, we investigated the impact of medium supplementation with free saturated or unsaturated exogenous FA on S. aureus profiles in both planktonic or biofilm states.

# MATERIALS AND METHODS

# Strain and Growth Conditions

Staphylococcus aureus RN 4220, S. aureus HG003, L. monocytogenes Scott A, S. Typhimurium ATCC 13311, and P. aeruginosa ATCC 11442 were used in this study. They were inoculated in Tryptone Soya Broth (TSB; Biomérieux, Marcy l'Etoile, France) or in Brain Heart Infusion (BHI; Oxoid, Basingstoke, UK) at 1% v/v with a standardized inoculum (∼10<sup>8</sup> cells/mL) obtained after two subcultures in the same broth. Cultures were incubated and harvested in exponential phase or stationary phase for fatty acid analysis.

When indicated, an exogenous fatty acid (EFA) is added to the culture medium: myristic acid (C14) or palmitic acid (C16) [saturated fatty acids (SFA)], oleic acid (C18:1cis9; monounsaturated fatty acid), linoleic acid (C18:2w6), linolenic acid (C18:3w6), or arachidonic acid (C20:4w6) [polyunsaturated fatty acids]. EFA were first dispersed in a bovine serum albumin (BSA) solution before addition to the growth medium. Final EFA concentration in BHI was 0.9 mM.

# Biofilm Formation

Biofilms were grown on the polystyrene base of 96-well polystyrene microtiter plates (Greiner Bio-One 655090, France) as previously described (Bridier et al., 2010). 250 µL of bacterial subculture (∼10<sup>6</sup> cells/mL) were poured into the wells and adhesion was done by sedimentation for 2 h. Subsequently, the planktonic bacterial suspension was removed and 250 µL of medium were added in each well. Microtiter plates were incubated for 48 h without shaking to allow biofilm development.

# Structural Biofilm Properties Observed by Confocal Laser Scanning Microscopy (CLSM)

Following incubation, biofilms were rinsed with 150 mM NaCl and refilled with TSB or BHI containing 5 µM Syto 9 (1:1000 dilution from a Syto 9 stock solution at 5 mM in DMSO; Invitrogen, France), a cell-permeable green fluorescent nucleic acid marker. The plate was then incubated in the dark at 30◦C for 20 min to enable fluorescent labeling. Images were acquired using a Leica SP2 AOBS confocal laser scanning microscope (Leica Microsystems, France) at the MIMA2 microscopy platform<sup>1</sup> . The excitation laser wavelength used for Syto 9 was 488 nm, and emitted fluorescence was recorded within the range 500– 600 nm. Images (512 × 512 pixels) were acquired through a 63 × Leica oil immersion objective (numerical aperture, 1.4) with a z step of 1 µm and a frequency of 400 Hz. 3D projections were generated with the Easy 3D IMARIS function (Bitplane, Zurich, Switzerland). Biofilms structural parameters [biovolume (µm<sup>3</sup> ), mean thickness (µm), and roughness (µm)] were extracted from image series using the PHLIP Matlab routine<sup>2</sup> . Each value presented is the average of six image series acquired in three independent experiments.

<sup>1</sup>http://www6.jouy.inra.fr/mima2

<sup>2</sup>https://sourceforge.net/projects/phlip/

# Membrane Fatty Acid Analysis

Bacterial planktonic cultures grown as described above were harvested by centrifugation (7000 g, 20◦C, 10 min) in exponential or stationary phase according to OD growth curves. Pellets were washed twice with 0.1% triton X-100 in order to remove unincorporated EFA when present. Biofilm cells for each species were collected from ten 96-plate wells. Biofilms were first rinsed once in 0.1% triton X-100 before biofilm cells were removed from polystyrene wells by scratching and re-suspended and washed again in 0.1% triton X-100. Extraction and methylation of fatty acids were carried out directly on bacterial pellets. Fatty acids of whole cells were first saponified and esterified by methanolic NaOH and methanolic HCl (first step: 1 mL NaOH 3.75 mol/l in 50% v/v methanol solution for 30 min at 100◦C; second step: addition of 2 mL HCl 3.25 mol/l in 45% v/v methanol solution for 10 min at 80◦C; Méchin et al., 1999). Fatty acid methyl esters were extracted with a diethyl ether/cyclohexane solution (1:1 v/v). The organic phase was at the end washed with a dilute base (NaOH 0.3 mol/l). Analytical gas chromatography of fatty acid methyl esters was carried out on a 6890HP system (Agilent Technologies, Santa Clara, CA, USA) equipped with a DB5 capillary column (Agilent Technologies, Santa Clara, CA, USA) and a flame-ionization detector. Column temperature was set at 150◦C for 4 min and then increased to 250◦C at the rate of 4◦C/min. Data were acquired using a HPCORE ChemStation system (Agilent Technologies, Santa Clara, CA, USA) and expressed as a percentage of the total area. Fatty acids were identified using fatty acid methyl ester standards and grouped in classes: saturated fatty acids (SFA), unsaturated fatty acids (UFA), hydroxylated fatty acids (HFA), cyclopropane fatty acids (CFA), branched-chain fatty acids (BCFA). Results are the average of at least eight profiles (two injections of four to nine extractions from independent cultures) for each condition.

# Statistics

ANOVA variance analyses were performed using Statgraphics software (ManugisticTM, Rockville, MD, USA). Evaluated factors were considered as statistically significant when P-values associated with the Fischer test were below 0.05.

# RESULTS

# Biofilm Cells Are Rich in Saturated Fatty Acids

Fatty acid profiles of S. aureus HG003, L. monocytogenes Scott A, S. Typhimurium ATCC 13311 and P. aeruginosa ATCC 11442 were compared when grown in TSB at 30◦C in planktonic state (exponential or stationary phases) or biofilm state. The fatty acid profiles of stationary phase cultures grown in planktonic conditions are shown as reference in Supplementary Table S1.

In planktonic cultures, SFA content significantly increased (P < 0.05) between exponential and stationary phases for S. aureus (+4.3%; **Figure 1A**), L. monocytogenes (+5.3%; **Figure 1B**), and P. aeruginosa (**Figure 1C**). Concomitant decreases are seen in amounts of anteiso-BCFA for both Gram-positive bacteria, and in cis-UFA at the expense of trans-UFA for P. aeruginosa (P < 0.05). Interestingly, decreases in anteiso-BCFA and UFA may lead to decreased membrane fluidity in stationary phase cells.

Biofilm cells exhibit different FA profiles compared to planktonic cells in both exponential and stationary phases. SFA content in biofilm cells is significantly higher than in planktonic cells for the four tested bacteria (P < 0.05) (**Figures 1A–D**). This increase is considerable for both Gram-positive bacteria, with, respectively, +12 and +12.7% compared to stationary phase for S. aureus and L. monocytogenes. This effect is mainly due to the increase of C16 content in S. aureus (+11.7%) and of C16 and C18 contents for L. monocytogenes (+4.4 and 6.4%, respectively). Both iso-C15 and anteiso-C15 BCFA contents significantly decrease in biofilm cells (P < 0.05) compared to their planktonic counterparts in stationary phase (respectively, −7.1 or −5.5% iso-C15 and −8.5 or −13.1% anteiso-C15 for S. aureus or L. monocytogenes). As before the observed increases in SFA and/or decreases in anteiso-BCFA species might indicate a more rigid membrane in biofilm than planktonic cells.

Biofilms observed by CLSM exhibit a variety of spatial organization according to the species. S. aureus forms flat and dense biofilms characterized by a high biovolume, low thickness and roughness (**Figures 2A,E–G**). L. monocytogenes also forms flat structures (lower biovolume, higher thickness and roughness; **Figures 2B,E–G**). The two Gram-negative bacteria trigger typical mushroom-like structures with high thickness and roughness. P. aeruginosa produces sparse but larger mushrooms (**Figure 2C**), in contrast to S. Typhimurium that produces numerous smaller clusters (**Figure 2D**).

# The Biofilm Lifestyle Alters S. aureus Selectivity of Exogenous FA Incorporation

Fatty acid profiles of S. aureus RN 4220 were analyzed when grown in BHI at 37◦C in the planktonic state (stationary phase) versus biofilm state in the absence or presence of exogenous FA (saturated or unsaturated FA dispersed in BSA; **Figures 3A,B**). In control conditions (without supplementation), the S. aureus RN 4220 qualitative fatty acid profile was similar to that of HG003 strain (Supplementary Table S1). Moreover, cells grown as biofilms also display a significantly higher SFA content (+13.3%) compared to cells in the planktonic state, due to the high increases of C18 and C20 SFA (respectively, +5.5 and +8.3%). At the same time, we detected a high decrease of C15-BCFA (−13.4%) together with a slight increase of anteiso- C17 (+2.1%).

Supplementation of saturated or unsaturated FA dispersed in BSA does not modify the S. aureus growth curve (data not shown). When exogenous C14 or C16 SFA are available in the culture medium, both planktonic and biofilm S. aureus membranes contain more of the added FA (C14 or C16) and more C20 FA. But, as in control conditions, biofilms cells contain more SFA (+18.5 and +15.9% with exogenous C14 and C16, respectively), namely C18 and C20 SFA, and less C15 BCFA than planktonic cells (**Figure 3A**). UFA addition also significantly affected fatty acid profiles of both cell populations and among

UFA, C18:1 and C18:2 were more incorporated than C18:3 or C20:4. In contrast to SFA, UFA are less incorporate in biofilm cells as compared to planktonic cells (−2.1, −10.5, −1.5, −5.4% with exogenous C18:1, C18:2, C18:3, C20:4, respectively; **Figure 3B**).

Overall, SFA are significantly higher and iso-BCFA are significantly lower in biofilm cells than in planktonic ones (P < 0.05). No significant difference is observed for anteiso-BCFA. When present, UFA are significantly less incorporated in biofilm cells (P < 0.05).

Staphylococcus aureus RN 4220 forms flat biofilm structures, as does strain HG003 (**Figure 4**). Supplementation with SFA or UFA does not significantly impact biofilm biovolume and thickness after 48 h (P > 0.05). An increase of roughness can only be observed with C18:3 and C20:4 supplementation in comparison of control (P < 0.05).

# DISCUSSION

Numerous studies invoke physiologic state of biofilm cells to explain their increased tolerance to antibiotics and disinfectants compared to their free counterparts. Although membrane is the first integrity barrier of the cell, few studies focused on phospholipidic membrane adaptation when bacteria are grown in biofilm state. In this study, we investigated the impact of the biofilm lifestyle on bacterial fatty acid composition and have shown that FA profiles of biofilm cells differed significantly from those of planktonic cells for four different bacterial species that show various spatial structuration of the biofilm (flat or mushroom-like structures). In all cases, FA profiles of biofilm cells contain significantly higher proportions of SFA compared to planktonic ones. In Gram-positive bacteria, increases in SFA (in particular long chain FA) are concomitant to decreases in BCFA content. This FA shift in biofilm cells was previously described in L. monocytogenes on simply adhered cells (Gianotti et al., 2008). It was also described that Rhodococcus erythropolis produced in general 5% more SFA than their planktonic counterparts in Muller-Hinton broth, even if lipid composition mainly depends on the type of surface and medium composition (Rodrigues and de Carvalho, 2015). The increase of SFA leads to a higher phase transition temperature, density of packing, and bilayer stability (Denich et al., 2003). Moreover, increases in long chain SFA may increase penetration of FA into the bilayer, favor interactions between acyl chains, and increase bilayer rigidity (Denich et al., 2003). Bacteria living within a biofilm consortium are surrounded by a polymeric matrix that triggers a heterogeneous environment where nutrients and oxygen are less available than in a liquid medium (Bridier et al., 2011). Nutrient stress was previously

shown to increase SFA content in planktonic cells (Guckert et al., 1986). But high SFA levels in biofilm cells could also reflect a specific physiological state induced during the early phases of attachment, as shown for L. monocytogenes (Gianotti et al., 2008).

Besides well-known bacterial ability to adapt FA synthesis under environmental conditions, it was demonstrated that bacteria can incorporate free exogenous FA from its environment to its membrane. Incorporation of serum FA into membranes allows S. aureus survival and growth in the presence of antibiotics targeting the FA synthesis pathway (Brinster et al., 2010). Vibrio cholerae was also shown to incorporate long chain PUFA present in bile into its membrane phospholipids (Giles et al., 2011). In this context, we also investigated the impact of free exogenous FA on FA membrane profiles of biofilm S. aureus cells in comparison to planktonic counterparts to evaluate how ubiquitous is the phenomenon of membrane saturation in biofilm cells. Planktonic cells and biofilm cells can both incorporate FA. Supplementation with SFA, namely C14 or C16, or with UFA, namely C18:1 or C18:2, leads to FA elongation with, respectively, a high increase in C20 for SFA and in C20:1 or C20:2 for UFA. It was recently shown that S. aureus can incorporate exogenous FA via a fatty acid kinase-dependent pathway (Parsons et al., 2014). SFA and UFA diffuse passively through the cytoplasmic membrane and bind, respectively, to FakB1 and FakB2 fatty acid binding proteins before they are converted to acyl carrier protein. In biofilm cells, as in planktonic cells, SFA and UFA can be incorporated and elongated. FA profiles of biofilm cells always showed higher SFA content compared to their planktonic counterparts (**Figure 3**). Biofilm cells thus display a specific physiological behavior which tends to decrease membrane fluidity, leading to fewer exchanges between bacteria and their environment, and likely improving

survival in a harsh environment (MacGarrity and Armstrong, 1975).

For Gram-negative bacteria, such as P. aeruginosa or S. Typhimurium, the increase of SFA content in biofilm cells compared to planktonic ones is lower than for Gram-positive bacteria. Membrane composition of Gram-negative bacteria is more complex, containing CFA and cis- and trans-UFA and membrane fluidity also depends on ratios of cis- to trans-UFA and of UFA to CFA. In the literature, results about P. aeruginosa are contradictory as some authors demonstrated a decrease of membrane fluidity in biofilm cells due to increased long chain FA and decreased BCFA (Benamara et al., 2011), whereas others showed a less rigid membrane by decrease in linear SFA content and decrease in fatty acid length chains (Chao et al., 2010).

## CONCLUSION

This study demonstrates that the FA content of biofilm cells can be statistically differentiated from those of free cells (exponential or stationary phase) and that SFA content in biofilm cells is always higher than in planktonic cells in similar conditions. But this global approach obviously does not evaluate the level of physiological heterogeneity within the population. Physiological diversity in biofilms has been shown to promote adaptation to stressful conditions and thus to enhance bacterial survival and resistance to antimicrobial agents (Boles et al., 2004). An interesting way to evaluate heterogeneity of FA modifications within the biofilm will be to use fluorescent reporter tools for monitoring

# REFERENCES


gene expression and membrane status on the single cell level.

# AUTHOR CONTRIBUTIONS

FD-B provided the general concept, designed and supervised the experiments, and wrote the manuscript. ET carried out most experiments. RB participated in designing the study. All authors contributed to the discussion of the research and approved the final manuscript.

# FUNDING

The study received support from the French research agency FattyBact project ANR-132101.

# ACKNOWLEDGMENTS

Authors would like to thank Alexandra Gruss and Gilles Lamberet for helpful discussion. They also thank Virginie Thiry and Julien Deschamps for their technical assistance.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2016.01673/full#supplementary-material


processing environments: causes, implications, role of bacterial interactions and control by alternative novel methods. Meat Sci. 97, 298–309. doi: 10.1016/j.meatsci.2013.05.023


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

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

# The Role of Stress and Stress Adaptations in Determining the Fate of the Bacterial Pathogen Listeria monocytogenes in the Food Chain

Kerrie NicAogáin and Conor P. O'Byrne\*

Bacterial Stress Response Group, Microbiology, School of Natural Sciences, College of Science, National University of Ireland, Galway, Ireland

The foodborne pathogen Listeria monocytogenes is a highly adaptable organism that can persist in a wide range of environmental and food-related niches. The consumption of contaminated ready-to-eat foods can cause infections, termed listeriosis, in vulnerable humans, particularly those with weakened immune systems. Although these infections are comparatively rare they are associated with high mortality rates and therefore this pathogen has a significant impact on food safety. L. monocytogenes can adapt to and survive a wide range of stress conditions including low pH, low water activity, and low temperature, which makes it problematic for food producers who rely on these stresses for preservation. Stress tolerance in L. monocytogenes can be explained partially by the presence of the general stress response (GSR), a transcriptional response under the control of the alternative sigma factor sigma B ( <sup>B</sup>σ ) that reconfigures gene transcription to provide homeostatic and protective functions to cope with the stress. Within the host <sup>B</sup>σ also plays a key role in surviving the harsh conditions found in the gastrointestinal tract. As the infection progresses beyond the GI tract L. monocytogenes uses an intracellular infectious cycle to propagate, spread and remain protected from the host's humoral immunity. Many of the virulence genes that facilitate this infectious cycle are under the control of a master transcriptional regulator called PrfA. In this review we consider the environmental reservoirs that enable L. monocytogenes to gain access to the food chain and discuss the stresses that the pathogen must overcome to survive and grow in these environments. The overlap that exists between stress tolerance and virulence is described. We review the principal measures that are used to control the pathogen and point to exciting new approaches that might provide improved means of control in the future.

Keywords: Listeria monocytogenes, general stress response, σ <sup>B</sup>, visible light, PrfA, virulence, RTE food safety

# INTRODUCTION

Listeria monocytogenes is a robust bacterial pathogen that is widely found in the environment. Its ability to persist in a diverse range of niches is underpinned by a sophisticated ability to sense and respond to the physicochemical stresses it encounters (Gandhi and Chikindas, 2007; O'Byrne and Karatzas, 2008). The term "stress" in this context is intended to mean any environmental

#### Edited by:

Avelino Alvarez-Ordóñez, Teagasc Food Research Centre, Ireland

#### Reviewed by:

Laurent Guillier, French Agency for Food, Environmental and Occupational Health & Safety, France Qingping Zhong, South China Agricultural University, China

> \*Correspondence: Conor P. O'Byrne conor.obyrne@nuigalway.ie

#### Specialty section:

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

Received: 05 July 2016 Accepted: 04 November 2016 Published: 23 November 2016

#### Citation:

NicAogáin K and O'Byrne CP (2016) The Role of Stress and Stress Adaptations in Determining the Fate of the Bacterial Pathogen Listeria monocytogenes in the Food Chain. Front. Microbiol. 7:1865. doi: 10.3389/fmicb.2016.01865

perturbation that reduces the growth rate (a mild stress) or negatively impacts cell survival (a more severe stress). In general stress imposes an energy cost on cells because they have to invest resources in protection (e.g., homeostasis, synthesis of new macromolecules, repair and replacement of damaged components) if they are to continue to survive and grow. The stress responses deployed when stress is encountered confer on L. monocytogenes the ability to persist in soil environments, water, mammalian and avian feces as well as in food and food processing environments. They also allow it to make a successful transition from food into the gastrointestinal tract of mammalian hosts, which is a prerequisite for establishing infections in immunocompromised individuals. The stress tolerance mechanisms at its disposal allow L. monocytogenes to withstand acidic conditions, environments with low water activity, desiccation, low temperatures and bile. Many of these stress tolerance mechanisms are under the control of an alternative sigma factor called sigma B (σ B ) whose role is to associate with RNA polymerase directing it to SigB promoters, which in turn leads to the reprogramming of the transcriptional profile of cells enabling the expression of protective functions (van Schaik and Abee, 2005; Chaturongakul et al., 2008; O'Byrne and Karatzas, 2008). The genes under the control of σ B , collectively known as the General Stress Response (GSR) regulon, are now well defined and many contribute to specific stress protective functions. Once within the host, an additional set of genes that allow cell invasion and systemic spread are expressed and these are regulated by a master transcriptional regulator called PrfA (Scortti et al., 2007). The roles of most of the virulence genes under PrfA control have well defined roles in the intracellular life cycle of the pathogen and indeed their study has fuelled the development of new areas of cell biology (Cossart and Toledo-Arana, 2008).

Although food-borne infections caused by L. monocytogenes are comparatively rare they are associated with unusually high mortality rates; typically 20–30% of clinical cases result in mortality. Immunocompromised individuals are most at risk, especially those with reduced T-cell immunity including elderly or very young patients, pregnant women, and individuals infected with HIV or on immunosuppressive treatment regimens (Lecuit, 2007). The organism is readily killed by normal cooking regimes including food processing treatments that use high temperatures (e.g., pasteurization). Therefore the main at-risk foods are the so-called ready-to-eat (RTE) foods, foods eaten without prior heating that have physio-chemical properties that can sustain the growth of L. monocytogenes (Chan and Wiedmann, 2008). Some of these foods include raw fruit and vegetables, dairy produce made with unpasteurised milk, minimally processed seafood, cold meats and pates (Farber and Peterkin, 1991; Lecuit, 2007). Although most countries enforce strict regulations on the tolerance for this pathogen in RTE foods its prevalence in the environment means that it is very difficult, if not impossible, to eradicate it from the food chain. Within Europe, if a product is capable of supporting growth, the producer must be able to demonstrate that levels of L. monocytogenes will not increase higher than 100 CFU/g over the course of the shelf life by means of a challenge study. However, if a RTE product is not capable of supporting growth, levels must not exceed 100 CFU/g during shelf life (European Commision, 2005). This differs from regulation in the US, where absence of L. monocytogenes is required in all RTE products (FSIS, 2014).

In this review, we discuss the route that L. monocytogenes can use to enter the food supply chain and discuss the behavior of the pathogen in foods. We outline the key stresses that L. monocytogenes must overcome to survive and grow in RTE foods and discuss the main protective systems that this pathogen uses to defend itself. The involvement of σ B and the GSR regulon in these responses is a particular focus of the review. Finally, we discuss traditional control measures used to reduce the risk of L. monocytogenes contamination of foods as well as some more innovative approaches that are currently being developed.

# ENTRY OF L. monocytogenes INTO THE FOOD CHAIN

# Soil

During the 1970's it was suggested that soil was a natural environment for L. monocytogenes (Welshimer and Donker-Voet, 1971; Weis and Seeliger, 1975). However, more recent studies have suggested that soil contamination by the organism may come from other sources such as sewage, animal manure and decaying plant vegetation (Fenlon et al., 1996). Many studies have investigated the survival of L. monocytogenes in soil and have observed that the foodborne pathogen can survive over a period of time, although, soil type, water content, pH, and temperature can all have an influence on the rate of survival (**Figure 1**; Ivanek et al., 2009; McLaughlin et al., 2011). For example, Locatelli et al. (2013) found that survival of L. monocytogenes was higher in fine soil with high clay content, which they suggest has a higher number of pores for protection against predation by protists and also has a cation content that is more compatible with long term survival.

Microflora within the soil can highly affect the survival of L. monocytogenes (**Figure 1**). Interactions between L. monocytogenes and different types of protozoa have previously been demonstrated (Ly and Muller, 1990; Zhou et al., 2007; Pushkareva and Ermolaeva, 2010). Sterilization of soil can lead to an increase in growth of L. monocytogenes suggesting that the microflora of the soil such as bacteriophage or protozoa have an effect on persistence of the bacterium, although this effect has not yet been fully explained. McLaughlin et al. (2011) confirmed that the microbiota of the soil plays an important role on survival. In their study, they partially reconstituted sterile soil with culturable aerobic components of the soil microbiota and observed that this lead to a decrease in survival at later time points in the experiment. They discuss the possibility that this decrease may be due to competition by different microflora for nutrients within the soil. Other factors which may affect the survival of the organism in soil include chemical properties as well as geographical and meteorological influences (Ivanek et al., 2009; Strawn et al., 2013). For example, Weller et al. (2015b) examined temporal factors (irrigation and rainfall) leading to

contamination of pre-harvest spinach (**Figure 1**). There was a greater chance of isolating L. monocytogenes after irrigation than rainfall and this chance was highest within 24 h of the event (Weller et al., 2015b). Other studies have confirmed similarly that irrigation is a risk factor for contamination of pre-harvest foods (Genereux et al., 2015; Weller et al., 2015a). This is often due to the contamination of the water source used for irrigation of the fields (Strawn et al., 2013; Genereux et al., 2015). Along with irrigation, the use of manure as a fertilizer can increase isolation of L. monocytogenes from produce production sites (Watkins and Sleath, 1981; Fenlon et al., 1996; Garrec et al., 2003). This is not surprising as animals are known reservoirs of the bacterium (Fenlon et al., 1996; Esteban et al., 2009; Mohammed et al., 2010).

Spatial factors such as proximity to urban areas, farms or water sources can lead to higher detection of L. monocytogenes (Sauders et al., 2012; Strawn et al., 2013; Weller et al., 2015b). One study conducted in New York State found that incidence of L. monocytogenes was much higher in samples taken from farms compared to a natural environment (an undeveloped area with minimal human presence) suggesting that the presence of humans and animals is highly associated with isolation of L. monocytogenes (Chapin et al., 2014).

# Seawater

Many studies have shown that water sources such as rivers, ponds and creeks can act as reservoirs for L. monocytogenes (Schaffter and Parriaux, 2002; Lyautey et al., 2007; Linke et al., 2014). However, one environment which has been considered to a much lesser extent is seawater. As isolation of this foodborne pathogen has been associated with seafood (Colburn et al., 1990; Johansson et al., 1999; Gonzalez et al., 2013; Leong et al., 2015), it may be a source of contamination worth considering. As with rivers, it is possible that effluent and land run off may increase levels of contamination by this microorganism in coastal waters (Watkins and Sleath, 1981; Fenlon et al., 1996). Some studies have shown isolation of L. monocytogenes from marine environments (Colburn et al., 1990; Motes, 1991; Rorvik et al., 2000; Rodas-Suarez et al., 2006). Motes (1991) found that Listeria spp. including L. monocytogenes could be isolated from fresh seafood and their harvest waters, suggesting

that L. monocytogenes can survive in seawater for a period of time. Colburn et al. (1990) also isolated L. monocytogenes from samples taken from an estuary and a bay in California. However, while some studies don't dispute that L. monocytogenes can be isolated from water, they do disregard it as an important source of contamination within fish farms (Jemmi and Keusch, 1994; Rorvik et al., 2000). More recent studies have also shown the survival of the pathogen in seawater, although many of these report that the survival of L. monocytogenes is strain and temperature dependent with lower temperatures correlating with higher survival (Bremer et al., 1998; Hsu et al., 2005; Hansen et al., 2006). However, besides temperature, other factors must also be considered for the survival of L. monocytogenes in seawater. Those factors include osmotic stress, predation by protozoa, nutrient availability, and solar irradiation (Smith et al., 1994; Tedetti and Sempere, 2006).

# Food Processing Environments

Within food production facilities, it is known that L. monocytogenes can survive over long periods of time; however, the source of contamination is often unknown (**Figure 1**). Persistence is often defined as a particular subtype re-isolated from the same environment over an extended period of time (Carpentier and Cerf, 2011; Ferreira et al., 2014). However, it is often difficult to determine whether a particular strain is persisting within an environment such as a food processing environment or if the strain is being reintroduced into the facility at different times. It is also disputed as to whether a genotype associated with persistence exists or whether L. monocytogenes can colonize specific favorable niches within a processing environment and therefore "persist" over a longer period of time (Carpentier and Cerf, 2011; Ferreira et al., 2014). Studies have compared phenotypic characteristics that cause strains to persist compared to non-persistent strains (Lunden et al., 2008; Ringus et al., 2012; Magalhaes et al., 2016). One inherent limitation of this sampling process is that only a subset of the population is sampled, and that persistent clones may be missed on multiple sampling occasions. Therefore, categorizing strains as nonpersistent can be difficult as it may happen that a persistent strain was only isolated sporadically in a study (Ferreira et al., 2014). Another challenge is that apparent persistence could be caused by the repeated introduction of the same strain to a food production facility, which could happen if contaminated personnel, equipment or product serve as a vector to continually introduce the same strain from some reservoir outside the plant.

Different studies have been conducted to investigate the main sources of contamination within food processing facilities (Johansson et al., 1999; Hansen et al., 2006; Leite et al., 2006; Ho et al., 2007; Chen et al., 2010a; Rivoal et al., 2010). Hansen et al. (2006) found that there was evidence of strains isolated from the outside environment also being identified within fish slaughterhouses. Other studies have shown that operators within a facility or different pieces of equipment may also be considered sources of contamination (Leite et al., 2006; Lomonaco et al., 2009; Chen et al., 2010a). Lomonaco et al. (2009) isolated L. monocytogenes from locker rooms, hallways and toilets within a gorgonzola producing facility suggesting the possibility that personnel within the factory contributed to the problem of contamination. Chen et al. (2010b) found that water used to chill fish products along with a weighing table were important sources of contamination within their facility.

It is still disputed as to whether seasonal variation has a contributing role in the isolation of L. monocytogenes from food processing environments. Many studies show no correlation between seasonal variation and occurrence of L. monocytogenes (Garrec et al., 2003; Ho et al., 2007; Esteban et al., 2009; Mohammed et al., 2010; Leong et al., 2014) but others have disputed these findings by showing a link between the two (Rivoal et al., 2010).

# STRESSES ENCOUNTERED IN FOOD

# Osmotic Shock

As salt is widely used in the preservation of food, osmotic stress is an important stress that L. monocytogenes must overcome to survive within many foods. This foodborne pathogen can survive salt concentrations as high as 3 M NaCl (Cole et al., 1990). It has been suggested that L. monocytogenes has a socalled primary and secondary response to osmotic shock. The primary response involves the influx of K<sup>+</sup> and glutamate into the cell, while the secondary response involves the uptake of small molecules known as compatible solutes (Kallipolitis and Ingmer, 2001; Brøndsted et al., 2003). These methods of combating osmotic shock play a role in helping the bacterium to restore turgor pressure, cell volume and also help to stabilize cell protein structure and function (O'Byrne and Fraser, 2000; Sleator et al., 2003).

Listeria monocytogenes accumulates the compatible solutes, glycine betaine, and carnitine, in hyperosmotic environments (Fraser et al., 2000; Sleator and Hill, 2001; Wood et al., 2001; Sleator et al., 2003). These solutes can often be found in different foods, with glycine betaine commonly found in foods of plant origin and carnitine from foods of animal origin (Sleator et al., 2003). The presence of these osmolytes in foods can help to enhance the growth of L. monocytogenes in the presence of hyperosmotic conditions. Besides these main osmolytes, other compatible solutes including proline, proline betaine, acetylcarnitine, gamma-butyrobetaine and 3 dimethylsulfonioporpionate have also been found to help the growth of L. monocytogenes in osmotic stress conditions (Bayles and Wilkinson, 2000). Uptake of compatible solutes occurs via three main transporters, Gbu, BetL, and OpuC (Sleator et al., 1999; Fraser et al., 2000; Gerhardt et al., 2000; Angelidis et al., 2002; Angelidis and Smith, 2003b). BetL or Betaine Porter I is one of two systems involved in the transport of glycine betaine into the cell (Gerhardt et al., 1996; Sleator et al., 1999) and is dependent on the presence of Na<sup>+</sup> (Gerhardt et al., 1996). Gbu, the second system involved in betaine uptake is an ATP dependent transporter which can be activated independently of Na<sup>+</sup> in response to osmotic shock by excess sucrose or KCl (Ko and Smith, 1999; Gerhardt et al., 2000). Finally OpuCA has been characterized as a carnitine transporter (Fraser et al., 2000,

2003). Deletion of genes encoding these transporters leads to an increase in generation time of the bacteria in the presence of hyperosmotic stress when incubated with glycine betaine, and carnitine (Angelidis and Smith, 2003a). Interestingly, SigB promoter sites have been identified upstream of each of these genes and deletion of σ B leads to reduced survival in response to high salt concentrations (Sleator et al., 1999; Fraser et al., 2003; Cetin et al., 2004). Further studies have shown that opuC and gbuA are under the control of σ B , but despite the presence of the putative SigB dependent promoter site upstream of betL, this gene does not appear to be under SigB control. Utratna et al. (2011) showed that transcription of opuC in response to osmotic shock occurred in a transient manner and the level of σ B activity observed also appeared to be proportional to the level of osmotic stress encountered.

Along with overcoming osmotic upshock, some bacteria have mechanisms to deal with hypoosmotic conditions. Mechanosensitive channels can allow the controlled release of osmolytes and water from the cell to aid the survival of a rapid increase in turgor pressure that occurs during osmotic downshock (Wood et al., 2001). Not much information is known about the existence of these channels in L. monocytogenes but two genes, lmo1013 and lmo2064, have been identified as having homology to genes encoding mechanosensitive channels in Escherichia coli and Streptococcus pneumoniae (Sleator et al., 2003). Rapid efflux of osmolytes, glycine betaine, and carnitine, has also been observed in L. monocytogenes cultures exposed to hypoosmotic conditions providing evidence for the presence of systems involved in downshock survival (Verheul et al., 1997).

Listeria monocytogenes also exhibits an adaptive response to NaCl known as osmoadaptation, where treatment of cells with a sub-lethal level of NaCl can offer increased survival following further exposure to lethal salt concentrations (Faleiro et al., 2003). A cross protection between osmotolerance and other stresses has also been confirmed. Schmid et al. (2009) found that csp genes are upregulated in the presence of either cold shock or osmotic shock. Deletion of some of these genes can lead to stunted growth when treated with low temperatures or high salt concentrations leading this group to hypothesize that the use of the CSP proteins may help to offer cross protection between osmotolerance and cold shock or vice versa depending on the condition encountered first by the bacterium (Schmid et al., 2009).

# Cold Shock

Listeria monocytogenes is capable of growth at temperatures as low as −0.4◦C (Walker et al., 1990). Various studies have demonstrated growth of this foodborne pathogen in different foods at refrigeration temperatures. However, at these temperatures the doubling time of the bacterium can be up to 50 h or more (Angelidis and Smith, 2003a). During an encounter with cold temperatures, bacterial membranes become more rigid, the rate of enzymatic reactions reduces and the level of uptake and transport of molecules is also decreased (Graumann and Maraheil, 1996). The bacterium must modulate its gene expression to mitigate the effect of these physical changes. Changes in expression usually occur for genes involved in cell membrane function, lipid, carbohydrate and amino acid synthesis, ribosomal structure and biogenesis and motility (Chan et al., 2007b; Cordero et al., 2016).

During exposure to cold temperature, one of the methods used by L. monocytogenes to combat cold shock is the accumulation of low molecular weight solutes such as glycine betaine, and carnitine. High amounts of these solutes are found in various foods (Zeisel et al., 2003; Demarquoy et al., 2004), which may help to promote the survival and growth of this pathogen in foods at refrigeration temperatures. The generation time of L. monocytogenes reduces by more than 20 h at 4◦C when cells are incubated in the presence of compatible solutes (Angelidis and Smith, 2003a). The BetL glycine betaine transporter (see Osmotic Shock) does not seem to be involved in cryotolerance (Sleator et al., 2003). Chan et al. (2007b) identified increased expression of both Gbu and OpuC but not BetL in response to cold shock, while a metabolomics study also showed increased quantities of glycine betaine, and carnitine present within L. monocytogenes when grown at 8◦C compared to 37◦C (Chan et al., 2007b; Singh et al., 2011). The increase in solute levels within the cell may help to decrease loss of intracellular water from the cell when temperatures drop.

A number of studies have investigated the role of σ B in adaptation to cold stress, but the data show conflicting results. It seems likely that survival, during exposure to cold temperatures, is controlled in a manner that is at least partly σ B -dependent. For example, Chan et al. (2007a) demonstrated that while some coldinduced genes were under σ B control (opuCA) or were preceded by a σ <sup>B</sup> dependent promoter site, they could be activated in a σ B independent manner at 4◦C indicating that genes responding to cold shock may be partially under σ B control. They also showed that a mutant lacking sigB did not have reduced growth at 4◦C compared to the wildtype (Chan et al., 2007a). Utratna et al. (2014) showed σ <sup>B</sup> does not play a large role in survival at low temperatures. They also showed that σ B could be activated at 4 ◦C in a manner that was independent of RsbV without levels of RsbW being affected (Utratna et al., 2014). Other systems that have been suggested to play a role in adaptation to cold stress include the two component regulatory systems, YycGF and LisRK (Pontinen et al., 2015). Transcript levels of the yycF gene were shown to be increased at 4◦C (Chan et al., 2007b) and Pontinen et al. (2015) suggested that YycF was more involved in survival of initial cold stress than acclimation over time, whilst LisRK seems to be more involved in acclimation.

# Low pH

Listeria monocytogenes can often encounter acidic conditions either in food matrices or within the gut of the host. These acidic conditions can arise from either weak organic acids such as lactate, benzoate, acetate or sorbate, or by strong acids like hydrochloric acid. Once L. monocytogenes enters the host following the ingestion of contaminated food, it encounters acidic conditions firstly within the stomach but also within the vacuole of the macrophage phagosome after intracellular uptake. The bacterium possesses a variety of different mechanisms including the adaptive acid tolerance response (ATR), the Glutamate Decarboxylase (GAD) system and the Arginine Deaminase (ADI) system to help it overcome these acidic environments (Davis et al., 1996; Cotter et al., 2001a; Ryan et al., 2009).

## Acid Tolerance Response (ATR)

fmicb-07-01865 November 21, 2016 Time: 15:57 # 6

Davis et al. (1996) first confirmed the presence of the Adaptive ATR in L. monocytogenes. This study showed that when exponential cells were pre-exposed to a sub-lethal pH (pH 5.0) for 1 h prior to exposure to a lethal pH (pH 3.0), cells exhibited a much higher survival rate compared to unexposed cells (Davis et al., 1996). The ATR results from pre-exposure to cells at a sublethal pH, typically between pH 5-6, before exposure to more lethal acids (Davis et al., 1996; Skandamis et al., 2012). Some studies have shown how this response can help L. monocytogenes survival on low pH foods (Gahan et al., 1996; Skandamis et al., 2012) while other studies also showed that this protective effect could be extended to other stresses such as heat and osmotic shock (O'Driscoll et al., 1996; Skandamis et al., 2009). Ferreira et al. (2003) have investigated whether the GSR has a role in the ATR. They suggest that while an isogenic 1sigB mutant survives less than the parent strain after being pre-exposed to sub-lethal pH, survival increases after pre-exposure suggesting that there are other σ B -independent mechanisms working on survival against acid (Ferreira et al., 2003).

# Arginine Deaminase (ADI) System

The ADI system is involved in enhanced survival at low pH in a variety of Gram-positive microorganisms including L. monocytogenes (Cunin et al., 1986). The system works by converting molecules of arginine into ornithine using three enzymes encoded for by the arcABC operon. A membrane antiporter ArcD, transports a molecule of arginine into the cell which is then converted to ornithine, CO2, ammonia and ATP. Ornithine is then transported back out of the cell in exchange for another molecule of arginine. During this process, the by-product ammonia can associate with intracellular protons to form NH<sup>4</sup> + and this leads to an increase of the cytoplasmic pH (Cunin et al., 1986). Many studies have investigated the role of the ADI system in acid survival in L. monocytogenes (Ryan et al., 2009; Chen et al., 2011; Cheng et al., 2013). Ryan et al. (2009) first showed the presence of a functional ADI system within L. monocytogenes and demonstrated that it is implicated in survival at low pH and virulence in vivo. They also identified ArgR as a regulator of the ADI system. Another study showing the role of Lmo0036 (ArcB) in acid tolerance also confirms the role of the ADI system (Chen et al., 2011). Interestingly, the transcription of arcA and argR, have been shown to be both SigB and PrfA-dependent (Ryan et al., 2009). Hain et al. (2008) and Bowman et al. (2010) also identified arcA as being under SigB control, whilst Milohanic et al. (2003) identified it as being controlled by PrfA suggesting a role for the ADI system in both stress response and virulence.

### Glutamate Decarboxylase (GAD) System

Another system identified in L. monocytogenes to help to maintain pH homeostasis within the cell is the GAD system, known to be important for survival both within synthetic gastric fluid, infection in mouse models and also in acidic foods (Cotter et al., 2001a,b; Feehily et al., 2014). It works to increase the internal pH of the organism in the presence of extracellular acidic conditions. However, it has been suggested that the GAD system is only responsible for survival of strong acidic conditions (below pH 4.5) and does not have a role to play in tolerance to weak acids (Heavin et al., 2009; Karatzas et al., 2010). The process works by the utilization of glutamate, which is present in all foods and living organisms. Under acidic conditions an extracellular molecule of glutamate is taken up by an antiporter (GadT) and then converted into γ-aminobutyric acid (GABA) by a decarboxylase enzyme, GadD. GABA is then exported back out of the cell in exchange for another molecule of glutamate. The decarboxylation process consumes one proton, thereby leading to an increase in intracellular pH (Lund et al., 2014). In L. monocytogenes, there are five genes involved in the Gad system, three genes encoding glutamate decarboxylases (GadD1, GadD2, and GadD3) and two encoding antiporters (GadT1 and GadT2). These genes are arranged into three operons, gadD1T1, gadD2T2, and gadD3 (Cotter et al., 2005). GadD1T1 seems to be required for growth at mild pH conditions while GadD2T2 is important in more severe acidic conditions. GadD2T2 and GadD3 have been shown to be at least partially under σ B regulation (Kazmierczak et al., 2003; Wemekamp-Kamphuis et al., 2004), but little more is known about the regulation of the GAD system in L. monocytogenes.

# Nisin

Different antimicrobial compounds including bacteriocins have been studied extensively over the past decades as a method of controlling bacterial contamination within food products. Some examples of antimicrobials which have been proven to be active against L. monocytogenes include lauric arginate, chitosan, pediocin, and nisin (Kaur et al., 2013; Kang et al., 2015). Nisin is one of the most common antimicrobials used in the food industry especially within dairy products and acidic foods (Delves-Broughton et al., 1996). It is a bacteriocin that is produced by the lactic acid bacterium, Lactococcus lactis. Compared to other bacteriocins, nisin has been shown to be most effective at reducing numbers of L. monocytogenes (Kaur et al., 2013). However, when used in combination with other antimicrobials, levels of inhibition increase further (Tokarskyy and Marshall, 2008; Kaur et al., 2013). It has also been observed that L. monocytogenes isolates can develop resistance to nisin, which is potentially a worrying prospect for the food industry (Gravesen et al., 2002). Cross resistance can also develop between bacteriocins meaning that combinations of different bacteriocins may not always be feasible (Kaur et al., 2013).

The antimicrobial effect of nisin involves interference with cell wall biosynthesis, disruption of the cell membrane by the formation of pores and consequent disruption of cell membrane associated processes (Bruno et al., 1992; Abee et al., 1994). It has been suggested that resistance of L. monocytogenes to nisin may arise due to changes within the cell wall composition which stops the bacteriocin from gaining access to the cell and therefore increasing survival (Kaur et al., 2012). Different systems including two component regulatory systems and the GSR have been implicated in L. monocytogenes resistance to nisin. Kang et al. (2015) showed that a mutant deficient in the

response regulator VirR had a greater loss of membrane integrity compared to the wild-type strain, while Begley et al. (2006) found that a sigB mutant had decreased growth and survival in response to nisin. However, Palmer et al. (2009) reported data which conflicted these results. They suggested that σ B contributes to nisin resistance in L. monocytogenes but only when it is deleted in a background lacking another alternative sigma factor, SigL (σ L ). When 1sigB is solely deleted, growth and survival actually increases in response to nisin. These data suggest that both σ B and σ <sup>L</sup> have a role to play in nisin resistance in L. monocytogenes (Palmer et al., 2009). Thus the actual role of σ B in the response of L. monocytogenes to nisin has yet to be determined.

# Light

Within environments where L. monocytogenes can persist, the bacterium can encounter varying amounts of light. Light has previously been used as a method of bacterial decontamination both within clinical environments and on food products (Ozer and Demirci, 2006; Maclean et al., 2010; Hosein et al., 2016; Xu and Wu, 2016) and therefore may be useful as a means of controlling L. monocytogenes contamination within the food industry. Recently it has been shown that blue light triggers the activation of the GSR within L. monocytogenes and therefore should be considered as a stress for the bacterium (Ondrusch and Kreft, 2011; Tiensuu et al., 2013). It is known that many bacteria have light sensing mechanisms which help them to overcome this stress. Within Bacillus subtilis, a light sensing protein YtvA has been discovered (Losi et al., 2002; Avila-Perez et al., 2006, 2009). This protein is present in a stress sensing complex known as the stressosome which is composed of the proteins RsbR and its paralogs as well as RsbS and RsbT (**Figure 2**; Gaidenko et al., 1999; Kim et al., 2004b; Hecker et al., 2007; Marles-Wright et al., 2008; Jurk et al., 2013). The stress signals are thought to be sensed by the protruding N- termini of these sensory proteins and are transduced into the core of the stressosome (Marles-Wright et al., 2008). This leads to a signaling cascade downstream of the stressosome which ultimately leads to the activation of σ B in response to the stress (**Figure 2**; Avila-Perez et al., 2009). While it has not yet been confirmed in L. monocytogenes, it is hypothesized that a similar stress-sensing complex exists. Within L. monocytogenes, the paralogs of RsbR include, Lmo0799, Lmo0161, Lmo1642, and Lmo1842 (Ondrusch and Kreft, 2011; Heavin and O'Byrne, 2012). Although it is not clear what stress signals most of these proteins sense, Lmo0799, a homolog of YtvA in B. subtilis, has been confirmed as a blue light photoreceptor (Ondrusch and Kreft, 2011). Mutants lacking Lmo0799 have been shown to have similar phenotypes to a 1sigB mutant with higher levels of motility in the presence of blue light and have lost the ability to form rings in response to cycles of light and dark (Ondrusch and Kreft, 2011; Tiensuu et al., 2013). The protein consists of an LOV (Light, Oxygen, and Voltage) domain at its N-terminus and a STAS domain at its C-terminal region. LOV domain proteins belong to the Per-Arnt-Sim (PAS) superfamily and can bind a flavin cofactor such as FMN, to facilitate light sensing (Christie et al., 1999; Herrou and Crosson, 2011). During light exposure it is thought that a covalent bond forms between a thiol residue of a conserved cysteine residue at position 56 of the Lmo0799 protein and the FMN molecule found in the binding pocket of the protein (Chan et al., 2013). Recently, O'Donoghue et al. (2016) constructed a mutant with a missense mutation, changing Cys56 to an alanine. When tested in response to light this mutant showed similar phenotypes to both 1sigB and 1lmo0799 suggesting that this residue is indeed required for light sensing by this protein (O'Donoghue et al., 2016). Interestingly, while it has been shown that σ B is activated in response to light, it has also been demonstrated that virulence genes have also been activated. Ondrusch and Kreft (2011) investigated the transcription levels of the internalin genes (inlA and inlB) that are involved the in invasion of L. monocytogenes into epithelial cells. Transcription of both inlA and inlB was increased in response to blue light in combination with 0.3 M NaCl and invasion into Caco-2 enterocyte-like human cells was also increased under these conditions. These data suggest that along with activating the stress response, blue light may also play a role in activation of virulence genes within L. monocytogenes.

# IMPLICATIONS FOR FOOD SAFETY

As RTE foods are of major concern for contamination by L. monocytogenes, it is beneficial to investigate how the bacterium behaves in such foods. This section looks at the response of stress related and virulence genes within different food matrices and how these could prime the bacterium for survival within the host. We also discuss the relationship between σ B and PrfA within the host and how this may aid survival and pathogenesis.

# Behavior of L. monocytogenes within Food Matrices

Studies have been conducted to investigate whether the transcriptional response of stress related and virulence genes of L. monocytogenes differ within various food matrices (Olesen et al., 2010; Bae et al., 2011; Alessandria et al., 2013). Within RTE foods, the bacterium encounters many of the stresses discussed in this review. Therefore, transcriptional studies can provide information on genes involved in allowing the survival of L. monocytogenes in the presence of these stresses in situ. Importantly, virulence of L. monocytogenes has been shown to be heterogeneous between strains and between food matrices (Duodu et al., 2010; Olesen et al., 2010; Rantsiou et al., 2012a,b; Hadjilouka et al., 2016). Virulence genes have been reported to be more highly induced under laboratory conditions than in food matrices, (Olesen et al., 2010; Rieu et al., 2010) and this was confirmed by a study which tested the effects on mice fed with broth cultures compared to contaminated food (Mahoney and Henriksson, 2003). The mice that were fed with fermented salami batter spiked with L. monocytogenes had a lower rate of infection than mice intragastrically challenged with a broth culture (Mahoney and Henriksson, 2003). However, mixed results have been found for the levels of sigB transcription when comparing broth cultures to food matrices. When grown on RTE deli turkey meat, transcriptional levels of sigB and related genes remained unchanged when compared to cultures grown in BHI broth (Bae et al., 2011). Rantsiou et al. (2012b)

observed that sigB transcript levels are generally upregulated in food matrices incubated at low temperature compared to BHI broth at 37◦C, while in contrast Olesen et al. (2010) showed that the level of sigB transcription is increased in BHI broth compared to standard liver pâté. Somewhat surprisingly when NaCl concentration was reduced in the pâté compared to the standard pâté, which contained 3.66% (w/v) NaCl in the water phase, sigB transcript levels were significantly increased in some strains (Olesen et al., 2010). Rantsiou et al. (2012b), suggest that temperature is the main variable contributing to the differences in sigB transcription that they observe. Differences in upregulation of expression of stress related genes and virulence genes have also been observed in other studies but often no particular pattern can be established (Duodu et al., 2010; Hadjilouka et al., 2016). Overall these studies show that different stresses encountered within foods can influence the induction of stress related genes and therefore have the potential to influence the gastrointestinal stages of a food-borne infection by L. monocytogenes.

# Overlap between Stress Response and Virulence

Tolerance to environmental stress and virulence can be considered to be overlapping facets of the biology of L. monocytogenes (O'Byrne and Karatzas, 2008). Firstly, without a robust stress response this pathogen would not be able to survive and persist in the food chain sufficiently well to allow it to gain access to a mammalian host. Secondly, the stresses encountered within the host, especially in the upper gastrointestinal tract, represent a significant barrier that must be overcome in order for L. monocytogenes to establish an infection. Particular challenges are presented by the acidic pH of the stomach, the osmolality and presence of bile in the ileum. As discussed earlier, (see Osmotic Shock and Low pH) L. monocytogenes has specific mechanisms for coping to acid and osmotic stress, some of which are under the control of σ B . This pathogen is also remarkably tolerant to bile. It can colonize the murine gall bladder (Hardy et al., 2004), aided by its bile salt hydrolase (BSH; Sue et al., 2004; Begley et al., 2005), a bile exclusion system called BilE (Sleator et al., 2005) and two efflux pumps (MdrM and MdrT; Quillin et al., 2011). The bsh gene and the bilE operon are both under σ B control (Fraser et al., 2003; Sue et al., 2004; Begley et al., 2005), while the efflux pumps are under the control of BrtA, a TetR-type transcriptional regulator (Quillin et al., 2011).

Having survived the stresses imposed by the GI tract the next step in establishing an infection is the invasion of epithelial cells in the intestinal villi (Cossart and Toledo-Arana, 2008). Invasion of epithelial cells is dependent on a surface protein called internalin (encoded by the inlA gene) whose expression is dependent on σ B (Kim et al., 2004a, 2005). It is interesting that the regulator of the GSR has been co-opted to participate in regulating the expression of a dedicated virulence gene and perhaps suggests that escape from the harsh conditions in the lumen of the gastrointestinal tract can be partly viewed as a response to stress (O'Byrne and Karatzas, 2008). The transcriptional regulator PrfA, a member of the Crp/Fnr family of regulators, is the master regulator controlling expression of virulence genes required for the intracellular stages of the infection caused by L. monocytogenes (reviewed in Scortti et al., 2007). PrfA expression is activated at 37◦C by a thermal sensing switch in the 50UTR region of the prfA transcript

(Johansson et al., 2002) and is also influenced by the CodY transcriptional regulator under conditions where branched chain amino acid levels are low (Lobel et al., 2015). The activity of PrfA is also modulated post-translationally by an association with a ligand whose identity has been elusive for many years. Recently, however, glutathione was identified as an allosteric modulator of PrfA activity (Reniere et al., 2015).

A number of lines of evidence indicate that there is regulatory cross talk between PrfA and SigB but the precise nature of this link has been difficult to define (O'Byrne and Karatzas, 2008). A number of transcriptomic studies have identified sets of genes whose regulation is influenced both by PrfA and by σ B (Kazmierczak et al., 2003; Milohanic et al., 2003; Ollinger et al., 2009; Toledo-Arana et al., 2009; Chaturongakul et al., 2011). σ B contributes directly to the regulation of a number of virulence genes including the inlAB operon (Kim et al., 2004a, 2005) and prfA itself (Nadon et al., 2002; Rauch et al., 2005; Schwab et al., 2005). Although prfA is preceded by a σ B -dependent promoter (designated prfAP2) the significance of this promoter in vivo remains unclear since it overlaps with a σ <sup>A</sup> promoter and it can be deleted without an obvious effect on haemolysis (Nadon et al., 2002). Overall it appears that the dominant role for σ B is during the gastrointestinal stage of the infection (Garner et al., 2006) whereas PrfA dominates after the intestinal barrier has been breached (Toledo-Arana et al., 2009). But the multiple and complex regulatory inputs that exist to control PrfA expression and activity probably allow σ B -mediated fine tuning of the PrfA regulon under certain conditions.

# METHODS OF CONTROLLING L. monocytogenes

Measures to control L. monocytogenes in the food chain mainly focus on the food processing environment, including personnel, and the formulation and processing of the product itself. Here we review some of the sanitizers that are in common use to control L. monocytogenes in food processing environments and consider some novel control strategies that are beginning to show promise and that might find application at different points in the food chain in the future.

# Sanitizers

Different sanitizers such as quaternary ammonium compounds (QACs), hydrogen peroxide, peracetic acid and sodium hypochlorite are often used for cleaning within food processing environments. It is known that these sanitizers are effective at killing planktonic L. monocytogenes cells (Kastbjerg and Gram, 2012; Ruckerl et al., 2014) and their effectiveness does not seem to differ between persistent and non-persistent strains of L. monocytogenes isolated from food environments (Magalhaes et al., 2016). Development of resistance against different sanitizers has also been investigated but the overall conclusion is that resistance does not seem to occur (Kastbjerg and Gram, 2012). Therefore no correlation between persistence and resistance to sanitizers has been discovered (Ruckerl et al., 2014). Different sanitizers have different mechanisms of inhibition. For example QACs such as benzalkonium chloride attack the cell membrane of cells, leading to cell leakage, while peracetic acid and sodium hypochlorite tend to act as oxidizing agents, creating reactive oxygen species (ROS) which lead to damage of cellular components (McDonnell and Russell, 1999).

To date, very little is known about the role of σ B in the mechanism of resistance of L. monocytogenes to sanitizers. However, it has been observed that σ <sup>B</sup> does play a role in the resistance of both planktonic and biofilm cells to benzalkonium chloride and peracetic acid over short periods of time (Ryan et al., 2009; van der Veen and Abee, 2010). Deletion of sigB reduces the levels of resistance against these sanitizers but does not affect growth in sub-lethal concentrations, while complementation of the mutation restores or even increases the resistance compared to the wild-type (van der Veen and Abee, 2010). While no studies have shown a correlation between σ B and resistance to sodium hypochlorite, it is considered that it may have a role to play as genes involved in oxidative stress are under σ B control (Ferreira et al., 2001; Boura et al., 2016). It is important to note that other systems controlled independently of σ B (e.g., the efflux pumps QacH, MdrL, and Lde) have also been observed to impact survival in the presence of sanitizers such as benzalkonium chloride (Romanova et al., 2006; Muller et al., 2013, 2014).

# Photodynamic Inactivation

Alongside the discovery that several bacterial strains respond to light as a stress agent, interest has developed in the possible use of light as a bacterial containment method. Specifically, photodynamic inactivation (PDI) has been shown to be effective in the treatment of different bacteria, including antimicrobial resistant strains of bacteria (Maclean et al., 2010; Luksiene and Paskeviciute, 2011; Endarko et al., 2012; Murdoch et al., 2012; Bumah et al., 2015; Hosein et al., 2016). In the case of L. monocytogenes, light can decrease cell numbers in liquid culture, on surfaces and decrease its biofilm production meaning that PDI could be a very useful way of treating L. monocytogenes contamination in the food production environment (Murdoch et al., 2012; McKenzie et al., 2013; O'Donoghue et al., 2016). This treatment involves the use of a photosensitizer in combination with light and oxygen. The photosensitizer can be added to the medium or can be found naturally within cells in the form of endogenous molecules like porphyrins (Hamblin et al., 2005; Buchovec et al., 2010; Luksiene et al., 2010). The mechanism of PDI involves a photosensitizer becoming activated by the absorption of photons and this leads to the creation of a singlet state of the photosensitizer which can decay and omit fluorescence as it returns to the ground state, or it can form an excited triplet state. From this triplet state, photooxidation can occur via two different pathways leading to the formation of ROS or singlet oxygen (**Figure 2**; Sibata et al., 2001; Luksiene, 2003; Luksiene and Zukauskas, 2009; Robertson et al., 2009). The generation of ROS in response to light can lead to interactions with lipids and proteins within the cell membrane and also lead to DNA damage which can result in cell death (**Figure 2**). Addition of reactive oxygen scavengers to quench the effects of ROS has been shown to increase growth and survival of L. monocytogenes in the presence of blue light suggesting that ROS contribute to

inhibition by visible light (Endarko et al., 2012; O'Donoghue et al., 2016). Interestingly, Tiensuu et al. (2013) found that many genes activated by Lmo0799 and σ B in response to blue light, are involved in combating oxidative stress.

Gram-positive bacteria have been shown to be more susceptible to PDI than Gram-negative possibly due to differences in cell wall composition or due to different amounts of endogenous porphyrins being produced within the cell (Nitzan et al., 2004; Maclean et al., 2009). Many different bacteria including foodborne pathogens such as L. monocytogenes, Bacillus cereus and Salmonella enterica have been inhibited in various studies using visible light (Luksiene and Paskeviciute, 2011; Endarko et al., 2012; Murdoch et al., 2012; O'Donoghue et al., 2016). Endogenous porphyrins are produced through the heme biosynthetic pathway in bacteria and act as natural photosensitzers within the cell. Some studies have proposed boosting porphyrin production within cells by adding increased amounts of 5-aminolevulinic acid, a precursor of the heme biosynthetic pathway (Nitzan et al., 2004; Buchovec et al., 2010). Other studies have sought to test whether the addition of exogenous photosensitizers to the medium could increase the sensitivity of Listeria to PDI (Romanova et al., 2003; Luksiene et al., 2010; Lin et al., 2012). Luksiene and Paskeviciute (2011) successfully used light in combination with Na-chlorophyllin to reduce levels of contamination by L. monocytogenes on strawberries, proving that PDI could also be used in combination with approved food additives to control the growth and survival of L. monocytogenes on food products.

# Innovative Strategies for Reducing the Risk of L. monocytogenes

Strategies aimed at reducing the risk of listeriosis usually focus on the elimination of the organism at the stage of food processing as well as designing food preservation regimes that don't support the growth of L. monocytogenes. Food preservation systems generally employ generic stress "hurdles" that act synergistically to inhibit microbial growth (Leistner, 2000), (e.g., reduced water activity combined with acidic pH), but the next generation of food preservatives might usefully target specific protective mechanisms and thereby prevent food pathogens from protecting themselves. As discussed in section "Nisin" is increasingly being used to prevent the growth of L. monocytogenes in food (reviewed in Cleveland et al., 2001). Its inhibitory mode of action is twofold; it interferes with cell wall biosynthesis and also disrupts the cytoplasmic membrane (McAuliffe et al., 2001). This dual action renders the cell vulnerable particularly when additional preservation-related stresses are also present in the food matrix. Lytic bacteriophages that target L. monocytogenes have also been considered for biocontrol of this pathogen. For example, broad host-range phages such as A511 and P100 have been shown to be effective at reducing viable L. monocytogenes cells to undetectable levels in some RTE foods (Carlton et al., 2005; Guenther et al., 2009; Bigot et al., 2011). In the future it might also be possible to target the regulators that control stress tolerance (σ B ) and virulence (PrfA). A small molecule that blocks σ B activity and reduces host cell invasion has recently been described (Palmer et al., 2011). The compound, 2 fluoro-phenyl-styrenesulfonamide (FPSS), apparently blocks the release of σ B from its anti-sigma factor RsbW, thereby preventing it from participating in transcription (Ringus et al., 2013), but the precise mode of action has not yet been established. Blocking the expression of virulence functions might also be a viable means of reducing the risk to food consumers. Recently a class of ring-fused 2-pyridone molecules have been identified that bind to PrfA and decrease its affinity for its consensus binding site on DNA (Good et al., 2016). A structural analysis of the interaction of PrfA with one of these molecules revealed that it interacts at two different sites on the protein that could prevent both allosteric activation of PrfA and also the correct alignment of the DNA binding helixturn-helix domain, thereby interfering with its ability to stimulate virulence gene expression. Additional work will be needed to develop these molecules further as potential therapeutic agents or even as designer food-preservatives.

# CONCLUDING REMARKS

While L. monocytogenes continues to present a very real risk to human health we now have a greatly improved understanding of its ecology, genetics and physiology. The ability to rapidly identify the sources of contamination using the latest genetic typing methods (including whole genome sequencing) means that food producers will better know where to target their efforts at reducing the occurrence in foods. Our understanding of the biology of L. monocytogenes, including a detailed knowledge of the protective strategies it uses to defend itself against harsh conditions, should better equip us to design food processing and preservation strategies that target the organisms Achilles' heel. There are still significant gaps in our knowledge, not least of which concerns the precise mechanisms that L. monocytogenes uses to sense its environment and how it couples its stress response to its pathogenicity, but the impressive research activity in these fields is likely to produce answers to these questions in the near future. The prevalence of this pathogen in the natural environment means that eliminating it from the food chain is almost impossible so reducing its occurrence in food through sound processing practices and carefully designed preservation strategies is the best approach for reducing the risk to food consumers. Novel approaches like inactivation with visible light or using inhibitors designed to target the regulatory machinery of this pathogen show great promise and are likely to be adopted in the years ahead.

# AUTHOR CONTRIBUTIONS

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

# FUNDING

The Bacterial Stress Response Group is supported by research grants from Science Foundation Ireland (11/RFP.1/GEN/3267),

the Department of Agriculture, Food and the Marine (FIRM # 11F008), Horizon 2020 (MSCA-ITN List\_MAPS 641984) and the Higher Education Authority of Ireland under Cycle 5 of the Programme for Research in Third Level Institutions (PRTLI).

# REFERENCES


# ACKNOWLEDGMENT

The authors thank their colleagues in Microbiology for helpful discussions and especially Tara Vollmerhausen for constructive comments on the manuscript.

at high osmolarity and low temperature. FEMS Microbiol. Lett. 219, 233–239. doi: 10.1016/S0378-1097(03)00052-1


mediates acid tolerance. Microbiology 157, 3150–3161. doi: 10.1099/mic.0. 049619-0


regulatory protein PrfA. Cell Chem. Biol. 23, 404–414. doi: 10.1016/j.chembiol. 2016.02.013





monocytogenes in spinach fields in New York state. Appl. Environ. Microbiol. 81, 6059–6069. doi: 10.1128/AEM.01286-15


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

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

# Homology-Based Modeling of Universal Stress Protein from *Listeria innocua* Up-Regulated under Acid Stress Conditions

Patrizio Tremonte<sup>1</sup> , Mariantonietta Succi <sup>1</sup> , Raffaele Coppola<sup>1</sup> , Elena Sorrentino<sup>1</sup> , Luca Tipaldi <sup>1</sup> , Gianluca Picariello<sup>2</sup> , Gianfranco Pannella<sup>1</sup> \* and Franca Fraternali <sup>3</sup>

*<sup>1</sup> Department of Agricultural, Environmental and Food Sciences (DiAAA), University of Molise, Campobasso, Italy, <sup>2</sup> Institute of Food Science, National Research Council (ISA-CNR), Avellino, Italy, <sup>3</sup> Randall Division of Cellular and Molecular Biophysics, New Hunt's House King's College, London, UK*

#### *Edited by:*

*Christophe Nguyen-The, Institut National de la Recherche Agronomique, France*

#### *Reviewed by:*

*David Rodriguez-Lazaro, University of Burgos, Spain Sandhya Srikant Visweswariah, Indian Institute of Science, India*

#### *\*Correspondence:*

*Gianfranco Pannella gianfranco.pannella@unimol.it*

#### *Specialty section:*

*This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology*

*Received: 10 June 2016 Accepted: 29 November 2016 Published: 20 December 2016*

#### *Citation:*

*Tremonte P, Succi M, Coppola R, Sorrentino E, Tipaldi L, Picariello G, Pannella G and Fraternali F (2016) Homology-Based Modeling of Universal Stress Protein from Listeria innocua Up-Regulated under Acid Stress Conditions. Front. Microbiol. 7:1998. doi: 10.3389/fmicb.2016.01998* An Universal Stress Protein (USP) expressed under acid stress condition by *Listeria innocua* ATCC 33090 was investigated. The USP was up-regulated not only in the stationary phase but also during the exponential growth phase. The three dimensional (3D) structure of USP was predicted using a combined proteomic and bioinformatics approach. Phylogenetic analysis showed that the USP from *Listeria* detected in our study was distant from the USPs of other bacteria (such as *Pseudomonas* spp., *Escherichia coli*, *Salmonella* spp.) and clustered in a separate and heterogeneous class including several USPs from *Listeria* spp. and *Lactobacillus* spp. An important information on the studied USP was obtained from the 3D-structure established through the homology modeling procedure. In detail, the Model\_USP-691 suggested that the investigated USP had a homo-tetrameric quaternary structure. Each monomer presented an architecture analogous to the Rossmann-like α/β-fold with five parallel β-strands, and four α-helices. The analysis of monomer-monomer interfaces and quality of the structure alignments confirmed the model reliability. In fact, the structurally and sequentially conserved hydrophobic residues of the -strand 5 (in particular the residues V<sup>146</sup> and V<sup>148</sup> β ) were involved in the inter-chains contact. Moreover, the highly conserved residues I<sup>139</sup> and H<sup>141</sup> in the region α4 were involved in the dimer association and functioned as hot spots into monomer–monomer interface assembly. The hypothetical assembly of dimers was also supported by the large interface area and by the negative value of solvation free energy gain upon interface interaction. Finally, the structurally conserved ATP-binding motif G-2X-G-9X-G(S/T-N) suggested for a putative role of ATP in stabilizing the tetrameric assembly of the USP. Therefore, the results obtained from a multiple approach, consisting in the application of kinetic, proteomic, phylogenetic and modeling analyses, suggest that *Listeria* USP could be considered a new type of ATP-binding USP involved in the response to acid stress condition during the exponential growth phase.

Keywords: universal stress protein, acid stress, *Listeria*, exponential growth phase, homology modeling, 2-D Electrophoresis, ATP-binding motif

# INTRODUCTION

Listeria innocua is regarded as a non-pathogenic indicator for the presence of Listeria monocytogenes in foods (Rosimin et al., 2016). Numerous ecological and genomic comparative studies highlighted a high similarity between the two species (Glaser et al., 2001; Girardin et al., 2005; den Bakker et al., 2010). Similarly to L. innocua, pathogenic L. monocytogenes is frequently found in various foodstuffs (Kovacevic et al., 2012; Jami et al., 2014; Ebner et al., 2015; Melo et al., 2015), especially those characterized by pH values higher than 4.4 (CAC, 2009). In the food industry a number of strategies are used to inhibit the growth of undesirable microorganisms, like technological processing (Tremonte et al., 2014), addition of natural substances (Tipaldi et al., 2011; Tremonte et al., 2016), protective microbial cultures (Sorrentino et al., 2013) as well as organic acids (Davidson et al., 2013). Unfortunately, in the case of several food types, sub-lethal pH values may induce resistance mechanisms to acid stress, which make the cells more resistant to severe acid conditions (Gandhi and Chikindas, 2007).

Acid stress response in L. monocytogenes has been the subject of several investigations, which documented the induction of a number of molecular mechanisms involving the F1F0-ATPase complex, the arginine deaminase (ADI) and the glutamate decarboxylase (GAD) pathways (Cotter et al., 2001; Ryan et al., 2009; Karatzas et al., 2012). On the other hand, little information is available on the role of Universal Stress Proteins (USPs) in the stress response of Listeria spp., although the expression of USPs was already studied in numerous other microorganisms (Tkaczuk et al., 2013).

USPs are cytoplasmic proteins found in Bacteria and Archea, as well as in fungi and plants. Their production is stimulated by several types of environmental stress or by specific physiological cell conditions (Kvint et al., 2003). Recently, a genomic approach demonstrated the involvement of usp encoding genes in the survival of L. monocytogenes in acid or oxidative stress conditions (Seifart Gomes et al., 2011), but the key role of USPs into cellular mechanisms remains generally unclear. To understand the molecular basis of possible USPs functions, the knowledge of their three-dimensional (3D) structures is essential. Amongst the ca. 110,000 structure deposited in the Protein Databank (PDB) only a few USPs structures have been determined so far. Indeed, the 3D-structures for USPs of Listeria are not available as of today. This gap may be filled by bioinformatics approaches such as homology modeling, on the condition that the sequence identity with known related structures is above 30% (Marti-Renom et al., 2000). Homology or "comparative" modeling, use an experimentally determined structure of a related protein as a template to model the structure of a target protein, and is the method of choice in the case of close sequence relationship (Petry and Honing, 2005). This approach is based on the observation that evolutionary and functionally related proteins generally share similar 3D structures. In this work, we used complementary proteomic and bioinformatics approaches in order to characterize USP proteins up-regulated in L. innocua ATCC 33090 under acid stress conditions and to predict in silico their 3D structure by referring to available template homologs in PDB.

# MATERIALS AND METHODS

# Bacterial Strain and Growth Condition

L. innocua ATCC 33090, obtained from the Leibniz Institute DSMZ-German Collection of Microorganisms and Cell cultures, was revitalized in Brain Heart Infusion broth (BHI; Oxoid, Milan, Italy) at 37◦C and then stored in cryovials (Pro-Lab Diagnostics, Richmond Hill, Canada) at −80◦C. Prior to use, cells were propagated twice in the same medium and incubation conditions, and collected in the middle of exponential phase. The growth was assessed for 30 h in 500 mL of BHI (conventional condition, batch C) and in BHI adjusted at pH 4.5 (acid stress condition, batch AS). Moreover, a further trial was performed using preacid-adapted cells. For this purpose, cells cultured under acid stress condition were collected in the middle of exponential phase and inoculated in BHI (batch pa-C) and in BHI at pH 4.5 (batch pa-AS). In all cases an initial inoculum of about 10<sup>7</sup> CFU/mL was used.

Plate counts were performed in BHI agar at different intervals and the growth kinetic parameters were estimated with the Dmodel of Baranyi and Roberts (1994) using the excel add-in DMFit v.3 (Baranyi and Le Marc, 1996). Three independent experiments were performed and the results were reported as average.

# Protein Extraction

Cells of L. innocua ATCC 33090 cultivated as described above (batches C, AS, pa-C, and pa-AS) were collected by centrifugation (7500 rcf at 4◦C for 15 min, Centrifuge Eppendorf, 5804R) in the middle of the exponential phase and during the stationary phase. Cells were washed three times with Tris-HCl (50 mM, pH 7.5), standardized at an OD600-value of 1.0 and re-suspended in a lysis buffer (Tris-HCl 50 mM, lysozyme 2 mg/mL, mutanolysin 50 U, protease inhibitor cocktail 1X, pH 7.5). Eight glass beads (Ø 0.4 mm) were added to each cellular suspension (140 µL), then suspensions were vortexed (3 min), incubated for 2 h at 37◦C, vortexed (3 min), and sonicated for 5 min with an ultrasonic homogenizer (100 W power, 100% amplitude, 0.8 cycle; Labsonic M, Sartorius, Italy) using a probe of 0.5 mm diameter. After centrifugation at 17,500 rcf for 30 min at 4◦C, the pellet was discarded and the supernatant (lysis buffer), containing the protein extract, was subjected to the Bradford-based protein assay kit (Bradford protein assay, Bio-Rad, Italy) to determine the protein concentration. Bovine serum albumin was used as a standard.

# Two-Dimensional Gel Electrophoresis (2-DE)

Proteins of L. innocua ATCC 33090, cultivated as above (batches C, AS, pa-C, and pa-AS), were collected from the lysis buffer with methanol/chloroform according to the method described by Wessel and Flügge (1984). Isoelectrofocusing (IEF) was performed using precast immobilized pH gradient (IPG) 24 cm strips with a pI 4–7 linear gradient. IPG strips were passively rehydrated for 21 h with a buffer containing urea 8 M, CHAPS 2%, DTT 50 mM, 2% of ampholine 3/10, bromophenol blue 0.002%, and 800 µg of proteins. IEF was performed at 65,000 Vh using the Ettan IPGphor apparatus (GE Healthcare Bio). Prior to 2-DE, the strips were equilibrated for 25 min with 50mM Tris-HCl, pH 8.8, urea 6M, glycerol 30% (v/v), SDS 2% (w/v), bromophenol blue 0.002% (w/v) containing alternatively 65 mM of dithiothreitol (DTT) or 70 mM iodoacetamide. The SDS-PAGE separation was performed at constant current (18 mA/gel) and temperature (15◦C) using the Ettan DALTsix Electrophoresis System apparatus (Amersham, Bio-sciences). Gels were stained for 2 h with the Bio-Safe colloidal Coomassie Blue G-250 (Bio-Rad) and digitalized using a GS-800 calibrated densitometer (Bio-Rad).

The 2-DE protein patterns were recorded as digitalized images using the Densitometer Calibrate GS–800 (Bio-Rad, Hercules, CA, USA). Spot detection, quantification, and analysis, were performed using the PDQuestTM 2-D gel analysis software, Version-8 (Bio-Rad, Hercules, CA, USA).

# Protein Identification by Mass Spectrometry

Proteins with modified expression level were identified by matrix assisted laser desorption ionization-time of flight (MALDI-TOF) mass spectrometry (MS)-based peptide mass fingerprinting (PMF). Briefly, protein spots were manually excised, destained, and digested overnight at 37◦C with 12 ng/mL proteomic grade trypsin (Promega, Madison, WI, USA). Afterwards, peptides were extracted in 1% formic acid/acetonitrile (1:1), vacuum dried, and analyzed using α-cyano-4-hydroxycinnamic acid (10 mg/mL in 50% acetonitrile/0.1 TFA) as the matrix. Mass spectra were obtained in the reflectron positive ion mode on a MALDI-TOF Voyager-DETM PRO mass spectrometer (Applied Biosystems) equipped with a 337 nm N<sup>2</sup> laser, acquiring at least 400 laser shots from each sample. A peptide mixture (Sigma-Aldrich Co.) was used as external standard. Proteins were identified with the MS-Fit Protein Prospector software (website: http:// prospector.ucsf.edu), searching the Uniprot database (2015.3.5 vers.). C-carbamidomethylation was set as a fix modification, while pyroglutamic acid at N-terminal glutamine and methionine oxidation were variable modifications. Up to one trypsin missed cleavage was allowed and peptide mass tolerance was 40 ppm. Search was taxonomically limited to microorganisms and then refined, with taxonomical restriction to Listeria spp. Only protein hits with at least 15 matching peptides, MOWSE score higher than 10<sup>8</sup> , and coverage higher than 25% were considered as positive identifications. Protein identification was validated by a software-assisted comparison of experimental and expected Mw and pI, as inferred from the UniProtKB/Swis-Prot database through the TagIdent tool (http://web.expasy.org/tagident/).

# Target and Template Selection for USPs Sequences Alignment

Based on the mass spectrometry results, the FASTA sequence (160 aa) of the USP (Uniprot accession no. A0A0E1Y4Z4) from L. monocytogenes FSL F2-208 was used as target. Homolog sequences and the USP template sequence were searched in the non-redundant protein sequence database (NR) and Protein Data Bank (PDB) respectively, using the protein-protein Basic Local Alignment Search Tool (BLASTp). The USP structure (PDB ID: 3S3T) from Lactobacillus plantarum was chosen as template for the model building due to its protein sequence showing the highest alignment score (31% of identity and E-value of 1−21). In detail, the four homolog chains A, B, C, and D of the complete USP structure from Lb. plantarum were aligned with the target sequence of Listeria (**Figure 1**). Sequence alignments were performed with the PRALINE program server (Simossis and Heringa, 2005) and edited with Jalview (Waterhouse et al., 2009).

# USPs Phylogenetic Analysis

The USP target from L. monocytogenes FSL F2-208 and other 54 bacterial domains of USPs previously characterized in other studies were used in the phylogenetic analysis. For this purpose, a multiple sequence alignment was constructed using ClustalW algorithm with Gonnet substitution matrix, gap open penalty of 10, and gap extension penalty of 0.2 (Thompson et al., 1994). Phylogenetic tree calculation was performed with the Neighbor-Joining method (Saitou and Nei, 1987) using MEGA7 software (Kumar et al., 2016). The statistical significance of the phylogenetic tree was tested by using bootstrap analysis (Felsenstein, 1985), with each bootstrap value reflecting the confidence of each branch.

# Model Building and Validation

Comparative homology modeling (Sali and Blundell, 1993) was used to build a model of L. innocua USP by means of the MODELLER software (Marti-Renom et al., 2000; Webb and Sali, 2014) and generating 1000 models. The model with the lower Discrete Optimized Potential Energy (DOPE) score was submitted to a Procheck (Laskowski et al., 1993) analysis for a preliminary investigation of the model structure stereochemical quality. The model was thereafter refined by the use of the KoBaMIN web server, which consist applying a protein structure refinement protocol based on the minimization of a knowledgebased potential of mean force (Rodrigues et al., 2012). The final model was validated via the Procheck and QMEAN Server (Benkert et al., 2009). The QMEAN Z-score was used in the evaluation, providing an estimation of the absolute quality of a model by relating it to reference structures solved by X-ray crystallography (Benkert et al., 2011). The Root Mean Squared Deviation (RMSD) was calculated to evaluate the similarity between the 3D structure of the template and the model. PyMol software was used for the RMSD calculation and to generate model images.

# Interfaces Analysis and Structure Alignments

The interfaces were explored using the Webservers PISA (Krissinel and Henrick, 2007) and POPSCOMP (Kleinjung and Fraternali, 2005). Moreover, the electrostatic potential distribution was calculated using the APBS (Adaptive Poisson-Boltzman Solver) and PDB2PQR software packages (Baker

et al., 2001; Dolinsky et al., 2004, 2007) and mapped onto a molecular surface of protein model using the PyMol software. Additional template structures were searched with the DALI sever (Holm and Rosenström, 2010). The structural alignment program MUSTANG was used to perform multiple structural alignments (Konagurthu et al., 2006, 2010). Sequence alignments were displayed with ESPript (Robert and Gouet, 2014).

# Statistical Analysis

Statistical analysis was performed on three independent experiments through ANOVA followed by the Tuckey's mulptiple comparison. For this purpose, the software GraphPad Prism version 6.0 was used.

# RESULTS

# Microbial Growth of *Listeria innocua* in Acid Condition

The effect of conventional (C) and acid stress (AS) conditions on the growth kinetics of non-pre-acid-adapted cells of L. innocua are reported in **Figure 2A**. The figure also highlights the kinetic curves of pre-acid-adapted cells cultivated in conventional (pa-C) and in acid stress (pa-AS) conditions (**Figure 2B**). The results evidenced that the acid stress condition strongly affected the kinetic parameters (**Table 1**) and the effect was particularly noticeable considering the maximum specific growth rate (µmax). In fact, the non-pre-acid-adapted cells (**Figure 2A**) showed a lower growth rate when cultivated in acid condition, as evidenced by µmax-values three-fold lower in AS (µmax of about 0.43 h−<sup>1</sup> ) than that exhibited in C (µmax of about 0.14 h−<sup>1</sup> ). Moreover, nonpre-acid-adapted cells cultured in conventional conditions (C) reached the stationary phase (yend ≈1.9<sup>9</sup> CFU/mL) in about 12 h, whereas in acid stress conditions (AS) the maximum population (yend ≈5.2<sup>8</sup> CFU/mL) was reached in about 20 h (**Figure 2A** and **Table 1**). Conversely, low or no differences were detected between the batches pa-C and pa-AS considering all the kinetic parameters. In this case, the µmax was about 0.42 h−<sup>1</sup> for pa-C condition and about 0.30 h−<sup>1</sup> for pa-AS. Moreover, no significant differences (p > 0.05) between pa-C and pa-AS were found in the final microbial levels (yend, **Table 1**).

# Protein Expression in Exponential Phase under Acid Stress Conditions

Proteins were extracted from cells of L. innocua ATCC 33090 in the exponential growth phase and the proteome of cells cultivated in conventional (C, pa-C) and acid stress (AS, pa-AS) conditions was compared by 2-DE. More than 500 spots were detected in each gel apart from the cultural conditions. The image analysis highlighted significant differences between conventional and acidic stress conditions. In detail, an up-regulation of several spot proteins was detected in the proteome of cells

TABLE 1 | Growth kinetic parameters of *L. innocua* ATCC 33090 cultivated in conventional and acid stress conditions after non-pre-acid-adapted or pre-acid-adaptation.


*Mean* ± *standard deviation of three independent experiments. Means in the same column with different superscript small letters are significantly different (p* < *0.05).*

cultivated in acid stress conditions (AS, pa-AS). In detail, after normalization over a number of protein spots with unmodified expression, densitometric analysis of the spot intensity evidenced 19 differentially expressed gene products at a threshold ≥2 (p < 0.05), which were selected for the MS-based identification (**Figure 3** and **Table 2**).

Four protein spots (a, b, c, and d) were highly upregulated, with expression ratio ranging from 3 to 8, namely succinate-semialdehyde dehydrogenase, NADP-dependent arylalcohol dehydrogenase, general stress protein (CTC), and (USP). Using the MS-based approach, the USP was identified in the proteome of L. innocua (Uniprot No H1GE00) but the primary structure was identical to the USP from L. monocytogenes (Uniprot accession number A0A0E1Y4Z4), sharing each other 100% of structural homology (**Table 2**).

# USP Expression As a Function of Growth Phase and Acid Conditions

A targeted 2-DE analysis was performed to investigate the expression of the USP in L. innocua in different conditions (**Figure 4**). The densitometric quantification of the protein spots evidenced that USP expression was significantly dependent on the growth phase, the cultivation in acidified medium and the pre-acid-adaptation conditions. The high magnification of 2- DE gels (**Figure 4B**) and the statistical analysis of spot area (**Figure 4A**) highlighted that when non-pre-acid-adapted cells were cultured in conventional conditions, the USP was upregulated more than four-fold in the stationary phase (C\_stat) compared to the exponential phase (C\_exp). The cultivation in acid condition significantly affected the USP expression not only in the stationary phase (AS\_stat) but also in the exponential phase (AS\_exp). In fact, when non-pre-acid-adapted cells were cultured in acid conditions, the USP was already up-regulated in the exponential phase (AS\_exp).

Similar results were obtained for pre-acid-adpted cells (**Figures 4C,D**). However, when cells were reinoculated in the conventional conditions (pa-C\_exp), a down-regulation of the USP expression was observed during the exponential phase compared to that revealed during the same phase in acid condition.

# A Functional Model for the USP Protein

USPs proteins are highly conserved and all deemed important for the stress response. The FASTA sequence (160 aa) of USP from L. monocytogenes FSL F2-208 was obtained from the Uniprot database (Uniprot No A0A0E1Y4Z4). Similar sequences were searched into non-redundant protein sequences database using BLASTp. About 100 protein sequences, with 140–161 residues, showed a high sequence identity (56–100%) with a score ranging from 181 to 314 and a E-value from 9e−<sup>55</sup> to 6e−107. The multi-sequences alignment (Figure S1) showed highly conserved regions especially between the residues 14 and 54 and between the residues 95 and 150. The conserved regions contain functionally relevant residues (Nachin et al., 2008). Moreover, phylogenetic analysis of the refined structural models of USPs could be exploited for further important functional information. To this purpose, a phylogenetic tree was constructed using 55 bacterial USP domain sequences and was divided into

four clusters (**Figure 5**). In agreement with previous results (Nachin et al., 2008; Gury et al., 2009), UspA, UspC, and UspD were grouped in the same cluster (cluster green); while in a separate cluster (cluster blue) were grouped the UspE1; finally, the UspF and UspG were grouped into a third cluster (cluster red). Interestingly, phylogenetic analyses showed that the USPs


from Listeria and Lactobacillus strains do not fit into any class described by Nachin et al. (2008) and formed, together the USPs of Halomonas elongata, Mycobacterium tuberculosis, and Thermus thermophilus, a distinct class that we have arbitrarily labeled UspL (cluster cyan).

This last class, in addition to the template and target sequences, includes also USPs from Lb. plantarum and L. monocytogenes strains, such as USP EHS84548 USP1 and USP Q8Y6V1\_LSMO, involved in acid stress resistance.

The modeled 3D structures of the USP belonging to L. innocua ATCC 33090 were stored as PDB output file and the best model (Model\_USP-691) with a lower DOPE score (−0.247) was refined and used for both validation and interface analysis. Model\_USP-691 (**Figure 6A**) was selected as the best to represent the quaternary structure of USPs with a homotetrameric conformation. Four monomers (chain A, B, C, and D) with an architecture similar to the Rossmann-like α/β-fold have five parallel β-strands and four α-helices (**Figure 6B**).

The Model\_USP-691 showed a good accuracy in both the stereochemical properties and the absolute quality of the structure. The main chain-conformations for 95.9% aminoacid residues were allocated within the most favored region of Ramachandran plot, only two residues (GLN 169B and ASN 378C) were found in disallowed regions of the plot (Figure S2). The reliability of the selected Model\_USP-691 was also confirmed by the good distribution of normalized QMEAN Z-score (−1.01) represented in Figure S3. Superposition of the Cα trace of the USP model (Model\_USP-691) from L. innocua ATCC 33090 (magenta) and the template 3s3t (green) from Lb. plantarum was, as expected, very close, and small differences were observed in the N-terminal and C-terminal of helix α2.

# Interface Analysis and Structural Alignments

To understand if the quaternary Model\_USP-691 structure may have relevance in explaining some of the biological observations, the macromolecular interfaces of the predicted quaternary structures were explored in detail. We focused on the interfaces involved in both dimer and tetramer formation of model structure. Results obtained with PISA and POPSCOMP showed the presence of six interfaces (**Table 3**), but only four of them were thermodynamically favored, having a negative value in the solvation free energy gain (1G) from PISA. The interface area in the formation of dimers between chains A and B or C and D (**Figure 7**) was of about 1400 Å<sup>2</sup> (PISA) or 1000 Å 2 (POPSCOMP) involving a total of ∼25% (PISA) or ∼30% (POPSCOMP) of residues (Table S1) belonging to β5, α1, and α4 (**Figure 6B**). The tetrameric contacts between chains A and C or between B and D (**Figure 7**) covered an interface area of about 1000 Å<sup>2</sup> (PISA) or 800 Å<sup>2</sup> (POPSCOMP) including the 20% (PISA) or 22% (POPSCOMP) of residues (Table S1) dislocates in the regions α2, α3 and α4, and β2 (**Figure 6B**).

**Figure 8** shows the surface charge density and electrostatic potential distribution at the interface between chains A and B (**Figure 8A**), C and D (**Figure 8B**), C and A (**Figure 8C**), and between chains D and B (**Figure 8D**). The complementary interfaces were covered by positive and negative charges as well as by hydrophobic patches. Moreover, the surface charges density highlighted that the dimer interfaces (A–B and C–D) were characterized by a hydrophobic core surrounded by a ring of polar residues. While, in the interfaces A–C and B–D a concave pocket positively charged can be noted.

The Model\_USP-691-A (-A, chain A) was compared with the 3D structures of USPs available in the PDB by means of the

DALI server (**Table 4**). Results showed that the Model\_USP-691- A shared up to 34% of sequences identity overall 3D structure with the other USPs belonging both Bacteria and Archaebacteria. The overall folds were highly conserved in all USPs, some difference was detectable in the poorly conserved region α2 (**Figure 8**). The regions (α1, α4, and β5), involved into monomermonomer interaction were highly conserved in all USPs. In addition, except for the fold α2, also the regions (α3, α4, and β5) of tetrameric contacts were conserved. Furthermore, the 31% (12/39) of residues involved in the formation of dimers Model\_USP-691-A were conserved in at least 60% of the overall structures and 18% (6/33) of the residues involved into tetrameric association interfaces were conserved in 60% of the overall structures (**Figure 9**). Interestingly, the quaternary structure evidenced the occurrence of a loop containing the ATP-binding motif G-2X-G-9X-G-(S/T/N) characterized by three residues of glycine interspersed with two and nine amino acid residues between the first and the second glycine residues, respectively and with a serine/threonine/asparagine following the third glycine. The ATP-binding motif was also detectable into other 11 USP structures (**Table 4**, **Figure 9**).

# DISCUSSION

In this study, L. innocua ATCC 33090 was used as a surrogate of the pathogen species L. monocytogenes, to investigate the effect of sub-lethal acid pH on the growth and protein expression. In agreement with results obtained for the pathogenic microorganism (Begley and Hill, 2015), L. innocua was able to quickly adapt to metabolic pathways in response to acid stress, modifying the expression of a protein subset.

In fact, the results highlighted that under acid stress conditions, 19 gene products were at least two-fold up-regulated in L. innocua during the growth (exponential phase).

Some of the upregulated proteins identified in the current study have been already associated to acid stress response as well as to other stress factors. The SSDH is the second enzyme of the γ-aminobutyrate (GABA) shunt pathway (Zhu et al., 2010). In the GABA shunt pathway, GABA is firstly converted to succinate-semialdehyde (SSA) by means of GABA aminotransferase (GABA-AT) and then oxidized by SSDH to succinate with formation of CO2. The GABA shunt can operate as an alternative pathway to provide succinate in some steps of the tricarboxylic acid (TCA) cycle in bacteria, including L. monocytogenes, that lacks of a complete set of genes necessary for the TCA cycle (Glaser et al., 2001). Moreover, GABA shunt may be an important source of nitrogen in certain bacteria (Schneider et al., 2002), as well as may play a role in acid tolerance in L. monocytogenes (Abram et al., 2008).

The CTC protein belongs to the L25 ribosomal protein family and is involved in the adaptation of L. monocytogenes to osmotic stress in the absence of osmoprotectants (Duché et al., 2002; Gardan et al., 2003). Moreover, the results in our study also highlighted the up-regulation of a specific USP during the growth (exponential phase) in acid conditions. It is likely to suspect that this specific protein -together with other acid stress proteins was involved in the response to acid stress during the growth phase of L. innocua. A link between the USP up-regulation and the acid stress condition was clearly revealed. In fact, the specific USP was up-regulated or down-regulated as response to acid stress condition or to the restoration of conventional conditions, respectively.

To the best of our knowledge, only Seifart Gomes et al. (2011) highlighted the importance of USP in response to the acid stress in pathogenic L. monocytogenes. These authors revealed a clear role of USP in the survival of cells showing that the resistance of usp-deleted mutants was significantly reduced compared to the wild stains. The role of USPs in response to several stress conditions was better elucidated in other bacteria, including Escherichia coli (Gustavsson et al., 2002; Nachin et al., 2005) and Salmonella typhimurium (Liu et al., 2007; Bangera et al., 2015), Haemophilus influenzae (Fleischmann et al., 1995; Sousa and McKay, 2001), Mycobacterium tuberculosis (O'Toole and Williams, 2003; Drumm et al., 2009), Pseudomonas aeruginosa (Boes et al., 2008), and Lactobacillus plantarum (Licandro-Seraut et al., 2008; Gury et al., 2009). In all cases, the expression of USP has been associated to the arrest of cellular growth in response to prolonged stress (Hingley-Wilson et al., 2010). Nyström and Neidhardt (1992, 1994) showed that the

β-strands and four α-helices.


#### TABLE 3 | Interfaces analysis of Model\_USP-691.

*PC, POPSCOMP server; N\_res, number of residues;* ∆*G, solvation free energy gain; N\_HB, number of hydrogen bonds; N\_SB, number of salt bridges.*

survival of E. coli was reduced in the uspA-mutated strains. Likewise, the mutation of uspA gene reduced the survival of S. thyphimurium to carbon or phosphorous (Liu et al., 2007). Moreover, the USP PA3309 and PA4352 were essential for survival of Pseudomonas aeruginosa under anaerobic conditions (Boes et al., 2006; Schreiber et al., 2006). Usp-deleted mutants of Burkholderia glumae showed a significant reduction of survival when compared to wild-type strains after the heat-shock stress (Kim et al., 2012).

Herein, USP over-expression was associated with the cellular growth arrest of L. innocua, both in the presence and in the absence of acid. However, when the strain was cultivated in acid conditions, unexpectedly USP was over-expressed during the exponential phase. This finding suggests that USP in Listeria could play a crucial role in the response to acid stress during the exponential growth, and represents an important advance in the knowledge of the functional role of the USP family. For this purpose, the phylogenetic analysis offered an interesting information, showing that USPs of Listeria were distant from other previously characterized USPs (Nachin et al., 2008) belonging to E. coli, Salmonella, and H. influeanzae. In detail, all the USPs from Lactobacillus spp. and Listeria spp. (including template and target) clustered in a separate and heterogeneous class, arbitrarily called USPL. Therefore, we can assume that the USP from the USPL class could play a different role in the stress response than USPs grouped in other classes.

Structural and biochemical studies suggest a wide array of functions of USPs. Anyway, few USP structures are available in PDB and there is little structural information is available for most of them. We found that of the 20 proteins with a

sequence similar (Z-score significance ≥15.0) to the Model\_USP-691-A, only for about 50% of them there is available structural information (**Table 4**). Generally, USPs have a structure typical of a Rossmann-like α/β-fold having five-stranded parallel βsheet surrounded by four α-helices that homo-dimerize, in an antiparallel fashion via the fifth β-strand on each subunit (Zarembinski et al., 1998; Schweikhard et al., 2010; Tkaczuk et al., 2013). Moreover, proteins such as TTHA0350, Rv2623, and YdaA were found to display two USP domains (Drumm et al., 2009; Iino et al., 2011; Bangera et al., 2015). Both single and double USP domain-containing proteins may assembly to form either tetrameric structure characterized by four USP domains or a homo-dimeric of two chains with two USP domains. In this work, the homology modeling technique used supports a homotetrameric structure of USP (**Figure 6B**). Bioinformatics analyses were addressed to the comparison of the modeled structures and of the interface regions with previously characterized USP members. The Model\_USP-691-A shared a sequences identity of 17–34% with a Z-score ranging from 15.0 to 26.2 (**Table 4**) when compared to resolved USP structures deposited in PDB. Moreover, the model reliability was confirmed by the presence of highly conserved folding regions (**Figure 9**), especially those represented by the β-strands. The β-strands 1–5 contained residues with hydrophobic properties playing an important role in the protein folding. These regions are involved in the formation of a stable β-sheet, described by Iino et al. (2011) as a typical USP molecular core. Interestingly, the hydrophobic properties of high structurally and sequentially conserved residues (V146-L147-V148-V149) of the β-strand 5 seem to have an essential role into monomer-monomer interface formation (**Figures 6B, 7, 9**). In particular the residue V146, which is conserved in all the USP structures and sequences observed so far (**Figure 9** and Figure S1), could play a key role into monomer-monomer contacts. Schweikhard et al. (2010) found that the residues V<sup>144</sup> and V<sup>146</sup> of TeaD protein, corresponding to V<sup>146</sup> and V<sup>148</sup> of our Model\_USP-691-A, were involved into monomer-monomer contacts through hydrogen bonds. We also predicted the presence of hydrogen bonds between the residues V<sup>146</sup> and V<sup>148</sup> of chain A with the residues V<sup>148</sup> and V <sup>146</sup> of chain B respectively. The region α4 took part in the formation of dimers with the involvement of 4 residues (E<sup>136</sup> - 2X-I139-R140-H141; where X indicate the number of residues

between) as well, but only the residues I<sup>139</sup> and H<sup>141</sup> resulted conserved in the sequence. Probably these residues, together with the others found conserved in the structure and sequences of USPs, may function as hot spots in monomer–monomer interface assembly (**Figure 9**). On the base of the herewithcollected structural and sequence information, it would seem that the dimeric arrangement of Model\_USP-691 is a plausibly correct assembly. This hypothesis is also supported by the analysis of the composition of interfaces. We found the presence of a large interface area (∼1400 Å<sup>2</sup> ) between the chains A and B or C and D of model, compatible in size with the interfaces (740– 1900 Å<sup>2</sup> ) of other crystallographic resolved USPs. In addition, a complementary electrostatic charge distribution was found at the interface. Moreover, the negative value of solvation free energy gain upon formation of the interface, as well as the high number of hydrogen bounds found at the interface, suggest a favorable interaction between the USP monomers. The proposed USP assembly (**Figure 7**) suggests that the dimer could have a crucial functional role. Although the biological function of these dimers is still unknown, this observation is supported by the recovery of proteins (TTHA0350, Rv2623, YdaA) with two tandem USP domains. In the work conducted by Schweikhard et al. (2010) the USP protein TeaD showed a dimeric state as assessed by SEC (Size exclusion chromatography) and Blue Native PAGE analyses. The same authors, showed that when ATP was added to TeaD, a tetrameric state was also observable, concluding that ATP significantly contributed to the stabilization of the molecular tetrameric conformation. Other authors highlighted the putative role of ATP in the contact between tetrameric assembly of USPs (Sousa and McKay, 2001). Recently, Bangera et al. (2015) showed that both single domain USP YnaF and two domains USP YdaA from S. typhimurium had a tetrameric or dimeric organization, respectively, and that together they could bind ATP. In fact, ATP or nucleotide binding USPs display a conserved ATP-binding motif G-2X-G-9X-G(S/T-N). We found this motif present in the structure of Model\_USP-691-A, and structurally conserved compared to the more closely related USPs (**Figure 9**). Notably, the presence of a positively charged pocket indicated an electrostatic compatibility with the ATP molecule. The residues H141, S133, L128, P99, A67, I<sup>48</sup> were structurally and sequentially conserved, pointing a role into tetrameric conformation of model. Overall, this information supports the tetrameric assembly shown by Model\_USP-691. The possible ATP-binding property would be fundamental for the USP assembly and consequently for the protein function. According to a previous biochemical study, USPs from


*bTotal number of residues.*


other bacterial species have been shown to display ATPase activity (Zarembinski et al., 1998). More recently, Bangera et al. (2015) reported the USP YdaA (PDB ID: 4r2j) of S. typhimurium showing ATPase activity and contain an ATP binding motif; in contrast, an additional USP (YnaF, PDB ID: 4r2l) protein did not show any ATPase activity, but was able to bind ATP though lacking the specific ATP binding motif. Therefore, the biochemical and biological function of the USPs could rely on an ATP dependent-factor, which is likely to be linked to a specific energetic state of the cell.

In conclusion, based on the current structural prediction, Listeria USP might be deemed as a new type of ATP-binding USP. Contrarily to USP-types involved into growth arrests, the USP of L. innocua could have a key role in the response to acid stress during the exponential growth phase.

# AUTHOR CONTRIBUTIONS

PT: design of the work, analysis and interpretation of the microbial and proteomic data, drafting the work; MS: interpretation of the data, drafting the work, and revising it

# REFERENCES


critically; RC: conception of the work, drafting the work, and revising it critically; ES: involved in experimental designing; LT: analysis of the microbial growth; GPi: analysis and interpretation of mass spectrometry data; GPa: design of the work, execution of 2-DE and bioinformatic analyses, interpretation of data, drafting and revising the work. FF: design of the bioinformatics approaches, agreement to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

## SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2016.01998/full#supplementary-material


Rev. Biophys. Biomol. Struct. 29, 291–325. doi: 10.1146/annurev.biophys.29. 1.291


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

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

# Insights into the Mechanism of Homeoviscous Adaptation to Low Temperature in Branched-Chain Fatty Acid-Containing Bacteria through Modeling FabH Kinetics from the Foodborne Pathogen Listeria monocytogenes

#### Lauren P. Saunders, Suranjana Sen, Brian J. Wilkinson and Craig Gatto\*

School of Biological Sciences, Illinois State University, Normal, IL, USA

#### Edited by:

Christophe Nguyen-The, Institut National de la Recherche Agronomique, France

#### Reviewed by:

Hélène Simonin, Agrosup Dijon, France Haihong Wang, South China Agricultural University, China Diego De Mendoza, Universidad Nacional de Rosario, Argentina

> \*Correspondence: Craig Gatto cgatto@ilstu.edu

#### Specialty section:

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

Received: 12 May 2016 Accepted: 22 August 2016 Published: 07 September 2016

#### Citation:

Saunders LP, Sen S, Wilkinson BJ and Gatto C (2016) Insights into the Mechanism of Homeoviscous Adaptation to Low Temperature in Branched-Chain Fatty Acid-Containing Bacteria through Modeling FabH Kinetics from the Foodborne Pathogen Listeria monocytogenes. Front. Microbiol. 7:1386. doi: 10.3389/fmicb.2016.01386 The psychrotolerant foodborne pathogen Listeria monocytogenes withstands the stress of low temperatures and can proliferate in refrigerated food. Bacteria adapt to growth at low temperatures by increasing the production of fatty acids that increase membrane fluidity. The mechanism of homeoviscous increases in unsaturated fatty acid amounts in bacteria that predominantly contain straight-chain fatty acids is relatively well understood. By contrast the analogous mechanism in branched-chain fatty acid-containing bacteria, such as L. monocytogenes, is poorly understood. L. monocytogenes grows at low temperatures by altering its membrane composition to increase membrane fluidity, primarily by decreasing the length of fatty acid chains and increasing the anteiso to iso fatty acid ratio. FabH, the initiator of fatty acid biosynthesis, has been identified as the primary determinant of membrane fatty acid composition, but the extent of this effect has not been quantified. In this study, previously determined FabH steady-state parameters and substrate concentrations were used to calculate expected fatty acid compositions at 30◦C and 10◦C. FabH substrates 2-methylbutyryl-CoA, isobutyryl-CoA, and isovaleryl-CoA produce the primary fatty acids in L. monocytogenes, i.e., anteiso-odd, iso-even, and iso-odd fatty acids, respectively. In vivo concentrations of CoA derivatives were measured, but not all were resolved completely. In this case, estimates were calculated from overall fatty acid composition and FabH steady-state parameters. These relative substrate concentrations were used to calculate the expected fatty acid compositions at 10◦C. Our model predicted a higher level of anteiso lipids at 10◦C than was observed, indicative of an additional step beyond FabH influencing fatty acid composition at low temperatures. The potential for control of low temperature growth by feeding compounds that result in the production of butyryl-CoA, the precursor of SCFAs that rigidify the membrane and are incompatible with growth at low temperatures, is recognized.

Keywords: FabH, psychrotolerance, fatty acid biosynthesis, membrane fluidity, kinetic modeling, listeriosis, branched-chain carboxylic acids

# INTRODUCTION

The foodborne Gram-positive bacterial pathogen Listeria monocytogenes is the cause of the potentially serious disease listeriosis that is characterized by a high fatality rate. Detection of food contamination with L. monocytogenes continues to lead to large scale food product recalls. Recently, there has been a multistate outbreak of listeriosis linked to frozen vegetables (http://www.cdc.gov/listeria/outbreaks) that has resulted in an extensive recall of 358 consumer frozen vegetable and fruit products sold under 42 separate brands. Such outbreaks are very expensive, the costs of a 2008 outbreak in Canada being estimated at \$242 million Canadian dollars (Thomas et al., 2015).

A critical aspect of L. monocytogenes in its role as a foodborne pathogen is its ability to grow at refrigeration temperatures and below to temperatures as low as −0.1◦C (Walker et al., 1990). We have been interested in understanding the mechanisms underlying how L. monocytogenes copes with the stress of low temperatures over several years. A critical aspect of the psychrotolerance of the organism is to adjust its membrane fatty acid composition to maintain membrane fluidity. The fatty acids of L. monocytogenes are composed almost entirely of branchedchain fatty acids (BCFAs), with the three major fatty acids being the odd-numbered anteiso fatty acids anteiso C15:0 and anteiso C17:0, and odd-numbered iso fatty acid iso C15:0 (Annous et al., 1997; Nichols et al., 2002; Zhu et al., 2005). When the bacterium is grown at low temperatures the content of anteiso C15:0 rises markedly through a combination of reduction in fatty acid chain length and branching switching from iso to anteiso fatty acids (Annous et al., 1997; Nichols et al., 2002; Zhu et al., 2005). Anteiso fatty acids, which have a methyl branch on the antepenultimate carbon atom, disrupt the close packing of fatty acyl chains (Willecke and Pardee, 1971; Poger et al., 2014), resulting in increased membrane fluidity (Edgcomb et al., 2000), in what is termed homeoviscous adaptation (Sinensky, 1974).

The mechanisms involved in changes in fatty acid composition resulting in increased membrane fluidity are very different depending on whether the fatty acids of the bacterium are a mixture of straight-chain saturated fatty acids (SCFAs) and straight-chain unsaturated fatty acids (SCUFAs), or are predominately BCFAs. Species that contain primarily SCFAs and SCUFAs, which include many Gram-negative and some Gram-positive species, predominately alter fluidity via chain length and the ratio of SCFA to SCUFA (Zhang and Rock, 2008). In contrast, species with a high proportion of BCFAs, which are predominately Gram-positive species, alter chain length and the ratio of anteiso to iso fatty acids (Suutari and Laakso, 1994). Our work is involved in attempting to further understand the mechanisms underlying changes in fatty acid composition in L. monocytogenes that allow growth at low temperatures that may have applicability to BCFA-containing bacteria in general.

The mechanism by which SCFA- and SCUFA-containing bacteria increase the proportion of SCUFAs is understood in considerable detail in Escherichia coli. The major fatty acids in E. coli grown at 37◦C are C16:0 (45%), C16:119 (35%), and C18:1111 (18%) (Cronan and Rock, 2013). As growth temperature drops the percentage of C16:0 drops and C18:119 increases, thereby increasing the proportion of unsaturated fatty acids and membrane fluidity. The 1 position of E. coli phospholipids is primarily occupied by C16:0 and C18:1111 and the C2 position by C16:119. At lower growth temperatures, C18:119 amounts increase at position 1 and C16:0 decreases (**Figure 1**). SCUFAs are produced by the activities of FabA and FabB. FabA dehydrates 3-hydroxyacyl-ACP to

trans-2-enoyl-ACP during fatty acid elongation and at the 10-carbon stage trans-2-deconyl-ACP is isomerized to cis-3 decenoyl-ACP by FabA (Heath and Rock, 1996). cis-3-decenoyl-ACP is elongated by FabB rather than the FabF condensing enzyme to form C16:1-ACP, which is then elongated to 18:1- ACP prior to its incorporation into phospholipids by FabF. FabF is subject to thermal regulation and is responsible for increased C18:1111 in cells grown at low temperatures (de Mendoza et al., 1983).

Homeoviscous adaptation in BCFA-containing bacteria appears to revolve around increasing the content of fatty acid anteiso C15:0 as exemplified by L. monocytogenes (Annous et al., 1997), Bacillus subtilis (Klein et al., 1999) and other BCFAcontaining bacteria (Suutari and Laakso, 1994). Unsaturated fatty acids do not play a major role in homeoviscous lipid adaptation in BCFA-containing bacteria, which lack the fabA and fabB genes required for their synthesis (Lu et al., 2004). A cold-inducible system that introduces double bonds into existing phospholipids in B. subtilis is probably of minor significance (Cybulski et al., 2002) in long term cold adaptation. In BCFA-containing bacteria including S. aureus, anteiso C15:0 preferentially occupies position 2 in phospholipid molecules and anteiso C17:0 position 1 (Kaneda, 1991; Parsons et al., 2011). Fatty acid anteiso C15:0 rises to 65% or more of the total fatty acids in low temperature grown L. monocytogenes (Annous et al., 1997). It is likely that some of the anteiso C17:0 fatty acid on position 1 is replaced by anteiso C15:0 (**Figure 1**). Anteiso C15:0 then plays a similar role to the SCUFAs in fluidizing the membrane at low temperatures. We have much less knowledge of the mechanisms involved in increasing the proportion of anteiso fatty acids than SCUFAs.

The first condensation reaction in fatty acid biosynthesis is catalyzed by FabH, which plays a major role in determining the fatty acids produced by bacteria. The major fatty acids in L. monocytogenes, which contains almost exclusively BCFAs, are anteiso odd, iso odd, and iso even fatty acids, which are biosynthesized from 2-methylbutyryl-CoA (2MB-CoA), isovaleryl-CoA (IV-CoA), and isobutyryl-CoA (IB-CoA), respectively, produced through the activities of branched-chain amino acid transaminase and branched-chain α–keto acid dehydrogenase on the branched-chain amino acids isoleucine, leucine and valine, respectively. Interestingly, a non-native FabH will switch an organism's fatty acid composition to reflect that of the organism from which it originated (Choi et al., 2000; Li et al., 2005), showing that FabH substrate specificity plays a major role in fatty acid composition. We have examined the kinetics of LmFabH at high and low temperatures and have provided evidence that the enzyme shows an increased preference for 2MB-CoA at 10◦C compared to 30◦C, which likely contributes to increased production of anteiso fatty acids at low temperatures (Singh et al., 2009).

Singh et al. (2009) measured LmFabH kinetics and substrate concentrations at 30◦C and 10◦C to determine the role of FabH in the changes in relative fatty acid proportions at low temperatures in L. monocytogenes. These kinetic data describe the mechanism of FabH when only one substrate is present; however, endogenous substrate concentrations measured directly from L. monocytogenes show that all three branched-chain acyl Co-A derivatives are present under physiological conditions. Thus, competition exists between these substrates and therefore the in vivo kinetics must differ from the in vitro determined kinetic values obtained from individual substrates. In this paper, we extend the analysis in Singh et al. (2009) via calculations and simulations to estimate the in vivo kinetic parameters and preferences of LmFabH. These in vivo substrate preferences can then be compared to the overall fatty acid composition of L. monocytogenes, which has been measured under a variety of conditions.

# MATERIALS AND METHODS

# Data Modeling

Data were modeled according to the Briggs-Haldane mechanism for all substrates simultaneously using Scheme 1:

$$\begin{array}{ccccc} & E\_1 & \xrightarrow{ES\_1} & E + P\_1\\ & k\_1 \uparrow\downarrow k\_{-1} & & \\ \downarrow E + S\_1 + S\_2 + S\_3 & \xrightarrow{k\_3} & E S\_2 \xrightarrow{k\_4} & E + P\_2\\ & k\_{-5} \uparrow\downarrow k\_{5} & & \\ & & E S\_3 & \xrightarrow{k\_6} & E + P\_3 \end{array} \qquad \text{Scheme 1}$$

in which FabH is E; S and P refer to the different substrates and products, respectively, and k1/k−1, k3/k−3, k5/k−<sup>5</sup> refer to the forward and reverse rates associated with the KM. The values k2, k4, and k<sup>6</sup> are the kcat rate constants for chemical conversion of the various substrates to their corresponding products. As the concentration of malonyl-acyl carrier protein, i.e., the second substrate, was constant for all experiments (Singh et al., 2009), it was incorporated into the K<sup>M</sup> and kcat values, reducing the kinetic mechanism to the Briggs-Haldane equation. Reaction rates were determined from Singh et al. (2009) with a reverse catalytic rate of zero. Rates for the initial equilibrium were modeled from the K<sup>M</sup> by setting the forward reaction rate to a constant for all substrates and varying the reverse reaction rate to account for K<sup>M</sup> value. Changes to the forward reaction rate did not significantly affect results (**Figure 2B**). Not all acyl-CoA substrates could be separated via HPLC in Singh et al. (2009); thus, temperatures were compared both under equal substrate conditions and under the values reported for in vivo conditions.

Composite steady-state parameters were calculated from the steady-state parameters of the individual substrates in Singh et al. (2009) using the equation:

$$\frac{V\_{S1}}{I[E]} = \frac{V\_{\text{max}}\,\text{[S\_1]}}{[\text{S\_1}] + K\_{M1}(1 + \frac{[\text{S\_2}]}{K\_{M2}} + \frac{[\text{S\_3}]}{K\_{M3}})} \tag{1}$$

in which VS<sup>1</sup> is the velocity of hydrolysis of substrate 1 and S and K<sup>M</sup> correspond to the concentration and measured K<sup>M</sup> of the species denoted in the subscript, respectively. This equation is modified from the equation for two competitive inhibitors (Segel, 1975) by replacing the inhibitors with substrates 2 and 3

and using the Briggs-Haldane rather than the Michaelis-Menten constants.

. LmFabH concentration was 1µM.

IB-CoA is orange and its product purple; IV-CoA is black and its product red; 2MB-CoA is green and its product blue. Forward rates for the pre-kcat

# Determination of Fatty Acid Composition

L. monocytogenes strain 10403S was grown in Brain Heart Infusion (BHI) media at 30◦C, under conditions identical to those in Sen et al. (2015). L. monocytogenes was grown concurrently in unsupplemented BHI and BHI supplemented with increasing amounts of the lipid precursors 2-methylbutyrate (2MB), isovalerate (IV), isobutyrate (IB), and butyrate (B), that act as precursors for biosynthesis of odd-numbered anteiso, oddnumbered iso, even-numbered iso BCFAs and even-numbered SCFAs, respectively (Julotok et al., 2010). L. monocytogenes cultures were harvested in exponential phase and the fatty acid compositions were determined as described by Zhu et al. (2005).

# RESULTS

FabH can utilize both straight- (acetyl-CoA; Ac-CoA) and branched- (IV-CoA, 2MB-CoA, and IB-CoA) chain substrates; however, LmFabH greatly prefers branched-chain substrates. The specific activity of FabH with Ac-CoA is greater than 10-fold lower than with the three branched substrates (Singh et al., 2009). Simulations based on Scheme 1 show that negligible amounts of Ac-CoA are utilized when all four substrates are present in equal concentrations (**Figure 2**). The presence of Ac-CoA therefore has nearly no impact on FabH substrate utilization and was not analyzed in subsequent simulations. FabH's low affinity for Ac-CoA agrees well with the native fatty acid composition of L. monocytogenes, as there are negligible amounts of SCFAs in its membrane. As Ac-CoA is present at relatively high concentrations (18.0µM at 30◦C), lack of this precursor is not preventing L. monocytogenes from making SCFAs. Rather, our data suggest that FabH substrate selectivity results in the lack of SCFAs.

Simulations show FabH prefers 2MB-CoA as a substrate at both 30◦C and 10◦C with equal substrate concentrations of 2MB-CoA, IV-CoA, IB-CoA, and Ac-CoA (**Figure 2**, **Table 1**). 2MB-CoA produces anteiso fatty acids, the predominant membrane fatty acids in L. monocytogenes. The next best substrates only produce ∼75% (IV-CoA) and ∼50% (IB-CoA) of the 2MB-CoA product at 30◦C and 10◦C, respectively (**Figure 2**). Interestingly, IV-CoA is the second best utilized substrate at 30◦C, but is not preferred over IB-CoA at 10◦C. As IV-CoA is converted into isoodd fatty acids, this modeled decrease in preference likely reveals the reason for observed lower proportion of these fatty acids at low temperatures (Annous et al., 1997; Nichols et al., 2002; Zhu et al., 2005).

In the simplest model for the determination of fatty acid composition, only FabH contributes to fatty acid composition and that composition reflects the substrate concentrations and FabH's enzymatic properties. This hypothesis is described by Scheme 1. To determine the validity of this hypothesis, substrate concentrations are needed for FabH's three main substrates. While Singh et al. (2009) measured endogenous substrate concentrations in vivo, not all substrates were resolved. For instance, butyryl-CoA (B-CoA) could not be separated from IB-CoA, and 2MB-CoA could not be separated from IV-CoA. Substrate concentrations were therefore varied within the range of values measured in Singh et al. (2009) to see if the data from the substrate concentrations, enzymatic parameters, and fatty acid composition could be described by the model (Scheme 1). As SCFAs (i.e., the products from B-CoA) are present only in low amounts in L. monocytogenes, B-CoA is not expected to contribute significantly to the fatty acid profile. Thus, the in vivo amount of IB-CoA can be estimated directly from the membrane fatty acid composition.

**Figure 3** shows a model depicting the expected relative concentrations of fatty acids for varying IV-CoA and 2MB-CoA concentrations. As IB-CoA concentration is not changing, the amount of iso-even fatty acids produced changes much less than either anteiso- or iso-odd fatty acids. An IB-CoA

equilibrium were set to 10 s−<sup>1</sup>


TABLE 1 | Normalized amount of product produced at various temperatures and substrate concentrations by FabH activity.

Calculations were done using Equation (1), with no substrate inhibition, equal amounts of 2MB-CoA and IV-CoA (i.e., 2.85µM 2MB-CoA and 2.85µM IV-CoA at 30◦C), and the extreme case in which only 2MB-CoA is present with no IV-CoA.

0.9µM. (B) Fraction of product based on concentrations and kinetic parameters at 10◦C. Product amounts are calculated from Equation (1). Total IV and 2MB concentration is 0.8µM, thus the [2MB-CoA] = 0.8µM – [IV-CoA]. IB-CoA concentration is 0.149µM.

concentration of 0.9µM would result in the observed amount of ∼5% even-numbered iso-fatty acids at 30◦C (**Figure 3A**). At 10◦C, an equivalent ratio of IB-CoA:B-CoA was used to model lipid composition which resulted in slightly higher iso-even fatty acid concentrations of ∼10% (**Figure 3B**). This is a higher concentration of iso-even fatty acids than is observed experimentally (Annous et al., 1997; Nichols et al., 2002; Zhu et al., 2005). Therefore, while the model can describe experimental data obtained at 30◦C, the model cannot describe the fatty acid composition experimentally observed at both temperatures without changes in the relative substrate concentrations. Specifically, iso-even fatty acid levels have never been reported as high as 10% as predicted by the model at 10◦C (**Figure 3B**); thus, the easiest resolution to this discrepancy is that in vivo an additional step regulates either the supply of acyl-CoA precursors to FabH or the incorporation of iso-even fatty acids into the lipid bilayer.

As for the 2MB-CoA and IV-CoA concentrations, in BHI media at 30◦C, the proportions of anteiso and iso odd fatty acids are 81.5 and 13.8%, respectively, giving a product ratio of 5.9 (Zhu et al., 2005). Modeling at 30◦C suggests this product ratio of fatty acids occurs at 1.4µM IV-CoA and 4.3µM 2MB-CoA, a substrate ratio of 3.0. Thus, there is good agreement between the model and measured fatty acid content at 30◦C. If changes in FabH's substrate preferences alone are responsible for the differences in the lipid profile, then the 10◦C fatty acid ratio (i.e., 7.5) should correspond to the same substrate ratio (3.0). However, at 10◦C a substrate ratio of 3.0 (i.e., 0.2µM IV-CoA, 0.6µM 2MB-CoA) produces a calculated fatty acid product ratio of 19.5, which greatly exceeds what is observed (Annous et al., 1997; Nichols et al., 2002; Zhu et al., 2005). This dichotomy between the model and the actual fatty acid composition measured suggests that FabH, although critical in determining final membrane fatty acid content, may not be the sole regulatory step involved. The simplest explanation for the substantial overestimated anteiso fatty acid content predicted by the model is that an additional regulatory step alters either the concentration of acyl-CoA precursors, or a later regulatory step regulates the amount of anteiso fatty acids that get incorporated into the lipid bilayer. As with IB-CoA, the model describes the data at 30◦C, but cannot describe the data at both temperatures, and thus another step must be involved in determining fatty acid composition.

To further investigate the effects of FabH, L. monocytogenes was cultured in medium supplemented with fatty acid precursors and the effects on fatty acid composition were observed. Concentrations of fatty acid precursors were small enough that no effect on bacterial growth was observed (data not shown). The amount of exogenously added short-chain carboxylic acid made available to FabH as the corresponding CoA derivative is uncertain which hampers our calculations. Thus, we necessarily assume that the partitioning of the substrate is a consistent percentage of the exogenously added precursor which becomes available to FabH. This percentage of available substrate was added to the model as an additional variable in our calculations (i.e., 2MB + x%<sup>∗</sup> [exogenous 2MB]) and simulations of fatty acid composition were run using this equation.

The effect of precursor addition (i.e., 2MB, IV, IB, and B) on the fatty acid composition of L. monocytogenes was measured, and the amount of available substrate was calculated from fatty acid composition. Based on our model, 0.2% of the exogenously added substrate is available to FabH under steadystate conditions, and substrate identity does not affect substrate availability (**Figure 4**). The increase in the associated product was measured for each fatty acid precursor added, and the data are well described by the model (**Figure 4**). As the kinetic parameters for B were not measured in Singh et al. (2009), the Ac-CoA parameters were instead used in the fitting. B-CoA appears to be a better FabH substrate than Ac-CoA, as more of it is incorporated into the membrane fatty acids than predicted from Ac-CoA

precursor addition. 0, 25, 75, 250, and 500 uM of 2MB (green), IV (black), IB (red), and B (blue) were added to L. monocytogenes cultures and the fatty acid compositions were determined. The increase in the fatty acid precursor's fatty acid product was measured and modeled using FabH substrate preferences (triangles, lighter colors).

kinetic parameters. 2MB-CoA, IV-CoA, and IB-CoA percentages are well described by FabH preference.

# DISCUSSION

FabH plays an integral role in determining membrane fatty acid composition in L. monocytogenes, as it does for other bacteria (Choi et al., 2000; Cronan, 2003). In this report, we modeled the activity of FabH at both 30◦C and 10◦C and found a qualitative explanation for the fatty acid profile change that occurs in this organism between high and low temperatures. This provides confidence that FabH is one of the enzymes that promotes growth at low temperatures for L. monocytogenes via increases in membrane fluidity. However, our model does not quantitatively mimic the exact distribution fractions of fatty acids at both temperatures, suggesting that another enzyme combines with FabH to produce the final membrane composition in L. monocytogenes.

In order to model fatty acid composition with additional substrates in the growth media, one must have an estimate of the amount of available cellular substrate. Millimolar quantities of fatty acid precursors are usually added to produce fatty acid composition changes (Julotok et al., 2010). This implies that only low levels of fatty acid precursors are converted into fatty acids, consistent with our finding that only ∼0.2% of exogenously supplemented precursors was available to FabH. It is not yet known whether this limitation is due to slow precursor uptake by the cell or another slow enzymatic step in producing the CoA derivatives of the various precursors. All four exogenous precursors appear to be equally available to FabH, as the percentage of available substrate was the same for all four substrates. This percentage would be expected to change if the substrate concentrations were tightly regulated, so it appears that there is little regulation of substrate concentrations at 30◦C.

An increase in anteiso fatty acids at the expense of isoodd fatty acids at lower temperatures is clearly shown in our model; however, the extent of replacement predicted by the model exceeds that seen in vivo. Our model assumes no change in the relative amounts of IV-CoA and 2MB-CoA under different temperature conditions; however, the concentrations of IV-CoA and 2MB-CoA were not separable in Singh et al. (2009). Thus, the deviation from the model may reflect alterations in relative substrate amounts, which would be compatible both with the data and our model. Alternatively, an upstream step from FabH that leads to the production of the CoA substrates may also be temperature dependent. The IV-CoA pool must increase from ∼1/3 at 30◦C to ∼2/3 of the IV-CoA / 2MB-CoA pool at 10◦C to fit the ratio seen in the fatty acid composition data. This requires a significant change in an earlier step in fatty acid production, and thus an additional enzyme that helps control fatty acid composition in L. monocytogenes. If substrate concentrations are not temperature dependent, a downstream step from FabH could selectively prefer iso-odd fatty acids at lower temperatures to regulate a possible overproduction of anteiso fatty acid precursors by the enzyme. The lower than predicted incorporation of anteiso fatty acids could be an adaptation for a rapid response to temperature change; when the temperature decreases, e.g., the cells need additional anteiso fatty acids, and FabH's preference for 2MB-CoA could be modulated via a feedback mechanism once the appropriate level of anteiso fatty acids are synthesized.

## CONCLUSION

No other studies have measured endogenous FabH substrate concentrations, so there is no basis for comparison to determine how well FabH substrate preference in other organisms compares to final fatty acid composition. In this study, fatty acid composition was calculated from FabH substrate preferences and composition, and the differences between this model and the actual fatty acid compositions of L. monocytogenes were compared. The data presented here suggest that FabH is the primary controller of fatty acid composition in L. monocytogenes. FabH preference can be used to predict fatty acid composition at 30◦C with and without added substrates, and qualitatively predict temperature induced changes in fatty acid composition. However, an additional control step beyond FabH alone is required to adequately predict fatty acid composition changes at lower temperatures.

Further in the type II fatty acid biosynthesis pathway is the rate limiting enzyme FabI (enoyl-ACP reductase). Schiebel et al. (2012) reported the ratio of specificity constants of FabI from S. aureus (a BCFA-containing gram-positive bacterium) was 1:24:1, straight: iso: anteiso. This is a potential candidate for a further control point that could be temperature

## REFERENCES


regulated. FabH substrate preference plays a significant role in L. monocytogenes survival at low temperatures, and methods can be devised to exploit this step to control L. monocytogenes growth. Similarly, identifying the enzymatic processes in addition to FabH that control the levels of lipid bilayer BCFA content may also reveal potential targets for controlling growth of this organism at low temperatures. For example, fatty acid composition can be modified by manipulation of precursor concentrations (Julotok et al., 2010) and encouraging the biosynthesis of SCFAs that are derived from B-CoA would decrease the growth of L. monocytogenes at low temperatures.

# AUTHOR CONTRIBUTIONS

CG: Designed experiments, evaluated and interpreted data, contributed extensively to the writing of the manuscript, supported the science with grant funds. BW: Designed experiments, evaluated and interpreted data, contributed extensively to the writing of the manuscript, supported the science with grant funds. LS: Developed the kinetic model, ran simulations, evaluated and interpreted data, contributed extensively to the writing of the manuscript. SS: Grew strains of L. monocytogenes, carried out fatty acid analyses, helped with data interpretation.

# FUNDING

This work was supported by grant R15-AI099977 from the National Institute of Health to BW and R15-GM61583 to CG.

acid composition of Listeria monocytogenes at 37 and 10◦C. Appl. Environ. Microbiol. 76, 1423–1432. doi: 10.1128/AEM.01592-09


potential implications for in vivo essentiality. Cell Struct. 20, 802–813. doi: 10.1016/j.str.2012.03.013


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

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

# Cold Shock Proteins: A Minireview with Special Emphasis on Csp-family of Enteropathogenic Yersinia

Riikka Keto-Timonen\*, Nina Hietala, Eveliina Palonen, Anna Hakakorpi, Miia Lindström and Hannu Korkeala

Department of Food Hygiene and Environmental Health, Faculty of Veterinary Medicine, University of Helsinki, Helsinki, Finland

Bacteria have evolved a number of mechanisms for coping with stress and adapting to changing environmental conditions. Many bacteria produce small cold shock proteins (Csp) as a response to rapid temperature downshift (cold shock). During cold shock, the cell membrane fluidity and enzyme activity decrease, and the efficiency of transcription and translation is reduced due to stabilization of nucleic acid secondary structures. Moreover, protein folding is inefficient and ribosome function is hampered. Csps are thought to counteract these harmful effects by serving as nucleic acid chaperons that may prevent the formation of secondary structures in mRNA at low temperature and thus facilitate the initiation of translation. However, some Csps are non-cold inducible and they are reported to be involved in various cellular processes to promote normal growth and stress adaptation responses. Csps have been shown to contribute to osmotic, oxidative, starvation, pH and ethanol stress tolerance as well as to host cell invasion. Therefore, Csps seem to have a wider role in stress tolerance of bacteria than previously assumed. Yersinia enterocolitica and Yersinia pseudotuberculosis are enteropathogens that can spread through foodstuffs and cause an enteric infection called yersiniosis. Enteropathogenic Yersinia are psychrotrophs that are able to grow at temperatures close to 0◦C and thus they set great challenges for the modern food industry. To be able to efficiently control psychrotrophic Yersinia during food production and storage, it is essential to understand the functions and roles of Csps in stress response of enteropathogenic Yersinia.

Keywords: adaptation, Csp, cold stress, stress response, stress tolerance, Yersinia enterocolitica, Yersinia pseudotuberculosis

# INTRODUCTION

Bacteria encounter changing environmental conditions during food production and storage and they have evolved a number of mechanisms for coping with stress and adapting to changing environments. In modern food production refrigeration is one of the key elements in maintaining food safety. At cell level temperature downshift decreases the fluidity of cell membranes, which affects active transport and protein secretion (Phadtare and Severinov, 2010). In addition, the efficiency of transcription and translation is reduced due to stabilization of the secondary structures of DNA and RNA, protein folding is inefficient, and ribosomes need to be adapted to cold before they can function properly (Phadtare, 2004). The aim of cold shock response is to help bacterial cells to overcome these changes (Phadtare and Severinov, 2010).

#### Edited by:

Avelino Alvarez-Ordóñez, Teagasc Food Research Centre, Ireland

#### Reviewed by:

Diego Garcia-Gonzalo, University of Zaragoza, Spain Karen LeGrand, University of California, Davis, USA

> \*Correspondence: Riikka Keto-Timonen riikka.keto-timonen@helsinki.fi

#### Specialty section:

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

Received: 12 May 2016 Accepted: 11 July 2016 Published: 22 July 2016

#### Citation:

Keto-Timonen R, Hietala N, Palonen E, Hakakorpi A, Lindström M and Korkeala H (2016) Cold Shock Proteins: A Minireview with Special Emphasis on Csp-family of Enteropathogenic Yersinia. Front. Microbiol. 7:1151. doi: 10.3389/fmicb.2016.01151

Bacteria respond to a rapid temperature drop (cold shock) by a transient induction of cold induced proteins (Cips) (Graumann and Marahiel, 1996; Phadtare, 2004) and the production of Cips increases with the severity of the cold shock (Hébraud and Potier, 1999). In Escherichia coli numerous Cips have been identified so far including, e.g., cold shock protein (Csp) family (Yamanaka et al., 1998), RNA helicase csdA (Charollais et al., 2004), exoribonucleases PNPase and RNaseR (Phadtare, 2012), initiation factors 2α and 2β, NusA and RecA (Jones et al., 1987). This minireview focuses on a subgroup of Cips, the small homologous Csps that are classified together in the Cspfamily.

Cold shock proteins counteract some harmful effects of temperature downshift and thus help the cells to adapt (Phadtare, 2004). After the immediate cold shock response, the synthesis of Csps declines and synthesis of other proteins increases. This enables the cells to grow at low temperature, although at a slower rate (Ermolenko and Makhatadze, 2002). Csps are known to be important during cold shock response, however, recent studies have shown that Csps might have a wider role in stress tolerance of bacteria (Schmid et al., 2009; Duval et al., 2010; Loepfe et al., 2010; Michaux et al., 2012; Schärer et al., 2013; Wang et al., 2014; Derman et al., 2015).

Yersinia enterocolitica and Yersinia pseudotuberculosis are enteropathogens that can cause a foodborne enteric infection called yersiniosis. In 2013, yersiniosis was the third most frequently reported zoonosis in the European Union (EFSA and ECDC, 2015). Food products containing pork are considered to be the main vehicles of Y. enterocolitica infection (Fredriksson-Ahomaa et al., 2006), whereas large Y. pseudotuberculosis outbreaks have been linked to contaminated vegetables (Nuorti et al., 2004; Jalava et al., 2006), drinking water (Fukushima et al., 1988), and raw milk (Pärn et al., 2015).

Yersinia enterocolitica and Y. pseudotuberculosis are psychrotrophic bacteria that can grow even at temperatures close to 0◦C (Fredriksson-Ahomaa et al., 2010). The control of Yersinia in the modern food chain involving refrigeration as the sole means to increase the shelf life of food products is thus challenging. To be able to efficiently control enteropathogenic Yersinia in refrigerated foods, a thorough knowledge of their cellular adaptation mechanisms to changing environments during food production and storage is needed. This minireview summarizes the current understanding of the functions of Csp-family proteins and highlights what is currently known about their role in stress response of enteropathogenic Yersinia.

# COLD SHOCK PROTEINS

Cold shock proteins are small nucleic acid-binding proteins ranging from 65 to 75 amino acids in length (Graumann and Marahiel, 1996; Czapski and Trun, 2014) and they have been found in psychrophilic, mesophilic, thermophilic and even hyperthermophilic bacteria (Phadtare, 2004; Jin et al., 2014). CspA was first described in E. coli (Goldstein et al., 1990). Later it was found that E. coli CspA family consists of nine homologous proteins named CspA through CspI that share a 46–91% amino acid sequence similarity (Yamanaka et al., 1998). Similar naming has been used in other bacteria as well, but genes with identical names do not necessarily share identical structure or function in different bacteria.

Structurally Csps are highly conserved, however, their thermostability varies (Lee et al., 2013; Jin et al., 2014). CspA of the psychrotrophic Listeria monocytogenes has a melting temperature of 40◦C (Lee et al., 2013), whereas Csp of thermophilic Thermus aquaticus has a more rigid structure and a melting temperature as high as 76◦C, suggesting that psychrophilic Csps need higher structural flexibility to accommodate nucleic acids upon cold shock (Jin et al., 2014). At 37◦C, the cspA mRNA of E. coli is very unstable, its half-life being only 12 s, but after cold shock its stability increases producing a half-life of more than 20 min (Mitta et al., 1997). Transient cspA mRNA stabilization due to low temperature is probably an important factor in the induction of CspA during cold shock (Phadtare and Severinov, 2005).

# FUNCTION OF COLD SHOCK PROTEINS DURING COLD SHOCK

Csps have a highly conserved nucleic acid binding domain, called the cold shock domain (CSD) (Graumann and Marahiel, 1996). CSD contains two nucleic acid binding motifs, ribonucleoprotein 1 and 2 (Lee et al., 2013) that facilitate binding to target RNA and DNA (Chaikam and Karlson, 2010). Jiang et al. (1997) demonstrated that CspA of E. coli binds weakly and with low sequence specificity to ssRNA. Csps function as RNA chaperones by destabilizing secondary structures in target RNA at low temperature so that the single-stranded state of target RNA is maintained. This enables efficient transcription and translation (Jiang et al., 1997; Phadtare, 2004). The weak and low sequence specific binding of RNA is important for the chaperone function of CspA and its homologues. Otherwise the movement of ribosomes on target mRNA would be hindered (Yamanaka, 1999). Due to the chaperone function, Csps can also act as transcription antiterminators by preventing formation of hairpin structures, which can act as transcriptional termination or pause sites in target RNA at low temperatures (Bae et al., 2000; Phadtare et al., 2002).

The cspA mRNA of E. coli can sense temperature changes and adapt to these changes by adopting different functional structures (Giuliodori et al., 2010). Giuliodori et al. (2010) showed that at low temperature the cspA mRNA derived from E. coli undergoes a temperature dependent structural change and as a result the cspA mRNA is translated more efficiently and is less prone to degradation than the cspA mRNA structure at 37◦C. Similarly, Mega et al. (2010) discovered that cold-induced ttcsp2 of thermophilic Thermus thermophilus acts as a thermosensor by adopting a more stable secondary structure due to a temperature drop.

Only CspA, B, CspE, G and CspI of E. coli are induced by cold shock (Etchegaray et al., 1996; Nakashima et al., 1996; Wang et al., 1999; Uppal et al., 2008) (**Table 1**). Xia et al. (2001) discovered that in E. coli four out of nine csp genes (cspA, cspB,

TABLE 1 | Cold shock protein genes of Escherichia coli, their reported functions and the predicted csp genes of Yersinia enterocolitica 8081 and Yersinia pseudotuberculosis IP32953 that share the highest amino acid sequence similarity (in percentage)<sup>a</sup> with csp genes of Escherichia coli K-12 W3110.


<sup>a</sup>Alignments were calculated using Clustal Omega v.1.2.1. Nucleotide and amino acid sequences were derived from The Kyoto Encyclopedia of Genes and Genomes (Kanehisa et al., 2016).

cspE, and cspG) had to be deleted until cold sensitive phenotype was obtained. In addition, deletion of one or two csp genes increased and prolonged the expression of the remaining cold induced csp genes. This indicates that the functions of the CspA family members overlap and they can compensate for each other (Xia et al., 2001). Similarly in Bacillus subtilis, the loss of one or two csp genes increased the production of remaining Csps after cold shock (Graumann et al., 1997). In E. coli, temperature fluctuation between 37◦C and 8◦C increased the cspA and cspB transcription during each temperature downshift whereas transcription decreased when temperature was raised (Ivancic et al., 2013). However, the CspA and CspB protein concentrations were only increased during the first temperature downshift and the relative protein levels remained constant during temperature fluctuations suggesting that the proteins are rather stable and not degraded at higher temperature (Ivancic et al., 2013). Csps may aid bacteria to survive in extremely cold polar environment. Jung et al. (2010) discovered that when the CspA of psychrophilic Psychromonas arctica isolated from Artic sea sediments was overexpressed in E. coli, the survival of cells after repeated freezing and thawing increased over 10-fold.

Mesophilic Clostridium botulinum Group I and III strains have two or three csp genes (cspA, cspB and cspC), but of the psychrotrophic Clostridium botulinum Group II strains, only one type B toxic strain carries one csp gene, whereas none of type E toxic strains hascsp genes (Söderholm et al., 2013). The lack ofcsp genes in psychrotrophic Group II C. botulinum indicates that the cold tolerance of these strains is due to some other mechanism. In C. botulinum ATCC3502 inactivation of just one of the three csp genes, cspB, resulted in cold sensitive phenotype, suggesting

that cspB is the major Csp of Group I C. botulinum (Söderholm et al., 2011). This is also supported by the fact that all C. botulinum Group I and III strains have homologues for cspB (Söderholm et al., 2013).

# COLD SHOCK PROTEINS OF ENTEROPATHOGENIC Yersinia

The published genome sequences of enteropathogenic Yersinia reveal that they carry several putative Csp encoding genes. Y. enterocolitica and Y. pseudotuberculosis strains have 6–10, and 7–9 csp genes, respectively. The Csps of Yersinia are highly homologous, and the cold-shock proteins encoding genes of, e.g., Y. pseudotuberculosis IP32953 share 32.8–100% amino acid and 44.6–99.5% nucleotide sequence similarity (**Table 2**). When the csp genes of E. coli K-12 W3110 are compared to the most closely similar csp genes of Y. enterocolitica 8081 and Y. pseudotuberculosis IP32953, the amino acid sequence similarity varies between 48.6 and 98.6% (**Table 1**). Since some of the Csp encoding genes of enteropathogenic Yersinia share a very high amino acid sequence similarity (74.3–98.6%) with cspA, cspB, cspC, cspD, cspE, cspG, and cspI of E. coli it can be assumed that they also have common functions. Lower amino acid sequence similarity is observed with E. coli genes cspF and cspH which function in E. coli is unknown (Czapski and Trun, 2014).

Yersinia enterocolitica, Y. pseudotuberculosis, Yersinia pestis, and Yersinia ruckeri carry a locus containing a tandem cspA duplication (cspA1 and cspA2) which produces both monocistronic (CspA1) and bicistronic (CspA1/A2) mRNA templates (Neuhaus et al., 1999). At high temperatures monocistronic mRNA predominates. When the temperature decreases, more bicistronic mRNA is produced but the longer the cold shock lasts, the more monocistronic mRNA is produced (Neuhaus et al., 1999). Compared to E. coli, these monoand bicistronic mRNA templates give Y. enterocolitica a better transcriptional capacity during the cold shock. Synthesis of proteins is more efficient when the transcript contains two copies of the protein (Neuhaus et al., 1999).

Annamalai and Venkitanarayanan (2005) studied the Csp expression of Y. enterocolitica in Luria-Bertani (LB) broth, milk, and on pork meat after temperature drop from 30◦C to 4◦C. CspA1 and CspA2 were first detected 2 h after cold shock from LB and milk cultures, but it took as long as 8 h after cold shock to observe detectable levels of Csps on pork meat. On pork meat the delayed expression of Csps and genes might be due to the extended lag phase of Y. enterocolitica on solid meat surface compared to liquid medium. Both in LB broth and on pork meat the expression of Csps and genes continued until 24 h of cold shock suggesting that in Y. enterocolitica Csps are not only needed during the immediate cold shock response but also during the cold acclimation. At 30◦C, CspA1 and CspA2 were not observed at all (Annamalai and Venkitanarayanan, 2005).

After cold adaptation, Csp mRNAs must be degraded to liberate ribosomes for translation of mRNAs of non-Csps and thereby to enable growth to resume at low temperature (Neuhaus et al., 2000). Neuhaus et al. (2000) observed correlation between beginning of exponential growth of Y. enterocolitica cells and degradation of cspA1/A2 mRNA. After adaptation to cold, Y. enterocolitica cspA1/A2 transcripts were smaller than the original transcript, and all cut off at the same sequence (AGUAAA) that was later named the cold shock cut box (CSCbox) (Neuhaus et al., 2003). Mutation of the CSC-box caused a delayed growth resume after cold shock. The CSC-box facilitates the degradation of Csp mRNA. Endonuclease RNaseE first cuts the cspA1/A2 transcript to smaller pieces which are further degraded by PNPase (Neuhaus et al., 2003). PNPase is necessary for Y. enterocolitica growth at low temperature (Goverde et al., 1998) and it regulates gene expression by selectively degrading Csp mRNAs (Yamanaka and Inouye, 2001b).

# ROLE OF COLD SHOCK PROTEINS DURING NORMAL GROWTH AND IN STRESS RESPONSES UNRELATED TO COLD

Cold shock proteins are not only produced during cold stress. A minimum of one csp gene is essential for viability of B. subtilis,

TABLE 2 | Nucleotide and amino acid sequence similarity (in percentage)<sup>a</sup> of the putative cold shock proteins of Yersinia pseudotuberculosis strain IP32953.


<sup>a</sup>Translated amino acid sequence comparisons above the diagonal and nucleotide sequence comparisons below the diagonal. Alignments were calculated using Clustal Omega v.1.2.1. Nucleotide and amino acid sequences were derived from The Kyoto Encyclopedia of Genes and Genomes (Kanehisa et al., 2016). <sup>b</sup>Gene length in parentheses.

suggesting that Csps are needed during non-shock growth (Graumann et al., 1997). In E. coli CspA forms 1% of all soluble proteins at the early exponential growth phase at 37◦C, suggesting that CspA has functions also at optimal growth temperature (Brandi et al., 1999). In E. coli, both growth medium and growth phase have been shown to affect the transcription of csp genes (Czapski and Trun, 2014). In MOPS defined minimal glucose medium cspE mRNA levels were higher than mRNA levels of other csp genes, whereas in MOPS defined rich glucose medium transcripts for cspA, cspB, and cspE predominated (Czapski and Trun, 2014). In E. coli transcripts of cspA predominate during the lag phase and first stages of logarithmic growth whereas only small levels of cspA mRNA can be detected at other stages of growth (Brandi and Pon, 2012). Also nutritional upshift has been shown to induce CspA in E. coli, although the related level of induction was only one-sixth of that caused by cold shock (Yamanaka and Inouye, 2001a).

Functions of non-cold-inducible Csps are yet poorly understood (Tanaka et al., 2012). DNA microarray analysis revealed that deletion of ttcsp1, which is the only non-cold inducible csp gene of T. thermophilus, did not alter the gene expression profile compared to wild type at optimal growth conditions. Nevertheless, expression levels of some proteins were significantly upregulated or downregulated in 1ttcsp1, suggesting that ttCsp1 contributes to translational control (Tanaka et al., 2012). Tanaka et al. (2012) also suggested that different stress factors might alter the nucleotide binding affinities of some Csps. On the contrary, the whole transcriptome of a cspA mutant of Brucella melitensis revealed that a total of 446 genes were differentially expressed compared to wild type. Differences in expression were especially observed in genes associated with virulence and metabolism and the cspA mutant also showed reduced growth in minimal medium. Results suggest that cspA of B. melitensis plays a role in virulence and metabolism regulation, but the CspA-mediated regulatory mechanisms are not understood (Wang et al., 2016). These contrary results observed in gene expression of different csp mutants warrant that further research is needed to understand by which mechanism Csps function during normal growth and in stress responses unrelated to cold.

In E. coli, cspD inhibits DNA replication and is induced during stationary phase to resign growth. Overproduction of this Csp is thus toxic to the cells (Yamanaka et al., 2001; Uppal et al., 2014). CspD has also been linked to biofilms and persister cell formation (Kim and Wood, 2010; Kim et al., 2010). Recently it was also shown that Csps might have a role in buffering deleterious mutations since overexpression of CspA improved fitness of E. coli strains that had accumulated deleterious mutations during long-term laboratory experiments (Rudan et al., 2015). Rudan et al. (2015) suggested that CspA and other RNA chaperones could aid misfolded RNA to adopt functional conformation and thus suppress harmful mutations that affect RNA structure. Since mutation rate can be elevated in response to stress (Hersh et al., 2004; Foster, 2007), it would of interest to investigate whether Csps could also counteract harmful stress-induced mutations.

Derman et al. (2015) revealed that cspB and cspC, but not cspA, play an important role in NaCl, pH and ethanol stress in C. botulinum ATCC3502. Mutation ofcspA and cspC also reduced motility and hampered flagella formation. In L. monocytogenes CspA is the major cold-shock-responsive Csp, whereas CspB and CspD are involved in host cell invasion (Schmid et al., 2009; Loepfe et al., 2010). Deletion of one or both of cspB and cspD resulted in severely impaired Caco-2 cell invasion, whereas deletion of cspA had no effect on cell invasion. Simultaneous deletion of cspB and cspD or deletion of all three csp genes of L. monocytogenes resulted in most severe impairment of host cell invasion and also caused reduced intracellular growth in macrophages (Loepfe et al., 2010). Schärer et al. (2013) suggested that Csps contribute to regulation of listeriolysin O, which is a pore-forming cytolysin needed in intracellular survival. In addition, cspA contributed to oxidative stress tolerance (Loepfe et al., 2010), and single deletion of cspD, but not of cspA and cspB, resulted in impaired osmotic stress adaptation in L. monocytogenes (Schmid et al., 2009). Since Csps promote L. monocotygenes adaptation against different stress conditions, exposure to one type of stress, e.g., during food production might induce cross-protection against others (Schmid et al., 2009).

# CONCLUSION

Production of Csps is one of the most prominent responses of bacteria to cold shock and recent studies have shown that Csps have a much wider role in stress response of bacteria than previously assumed. Enteropathogenic Yersinia encode several Csps which show a high homology to those of E. coli. Some of these E. coli csp genes are linked to stresses other than cold. Y. enterocolitica CspA1 and CspA2 are known to be involved in cold shock response, however, it is not known what the role of other csp genes is in stress response of enteropathogenic Yersinia. Therefore, knowledge about the function of Csps is needed to be able to develop measures that limit the growth and survival of enteropathogenic Yersinia in various foods during food production and storage.

# AUTHOR CONTRIBUTIONS

RK-T, NH, EP, ML, and HK designed the minireview, RK-T and AH performed sequence comparisons, RK-T and NH drafted the manuscript, EP, ML, and HK critically revised the manuscript.

# FUNDING

This project was supported by the Finnish Veterinary Foundation and the Walter Ehrström Foundation.

# REFERENCES

fmicb-07-01151 July 20, 2016 Time: 11:12 # 6


gene for polynucleotide phosphorylase, for growth at low temperature (5◦C). Mol. Microbiol. 28, 555–569.



Clostridium botulinum ATCC 3502. Int. J. Food Microbiol. 146, 23–30. doi: 10.1016/j.ijfoodmicro.2011.01.033


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

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

# Proline-Based Cyclic Dipeptides from Korean Fermented Vegetable Kimchi and from *Leuconostoc mesenteroides* LBP-K06 Have Activities against Multidrug-Resistant Bacteria

*Edited by:* Michael Gänzle, University of Alberta, Canada

#### *Reviewed by:*

Sergio I. Martinez-Monteagudo, South Dakota State University, USA Emanuele Zannini, University College Cork, Ireland

#### *\*Correspondence:*

Min-Kyu Kwak genie6@snu.ac.kr Sa-Ouk Kang kangsaou@snu.ac.kr

#### *† Present Address:*

Andrew H. Kim, Department of Clinical Pharmacology and Therapeutics, College of Medicine and Hospital, Seoul National University, Seoul, South Korea Sa-Ouk Kang, Irwee Institute, Seoul National University, Seoul, South Korea

#### *Specialty section:*

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

*Received:* 15 June 2016 *Accepted:* 12 April 2017 *Published:* 02 May 2017

#### *Citation:*

Liu R, Kim AH, Kwak M-K and Kang S-O (2017) Proline-Based Cyclic Dipeptides from Korean Fermented Vegetable Kimchi and from Leuconostoc mesenteroides LBP-K06 Have Activities against Multidrug-Resistant Bacteria. Front. Microbiol. 8:761. doi: 10.3389/fmicb.2017.00761 Rui Liu, Andrew H. Kim † , Min-Kyu Kwak \* and Sa-Ouk Kang\* †

Laboratory of Biophysics, School of Biological Sciences, and Institute of Microbiology, Seoul National University, Seoul, South Korea

Lactobacillus plantarum and Leuconostoc mesenteroides play a prominent role as functional starters and predominant isolates in the production of various types of antimicrobial compound-containing fermented foods, especially including kimchi. In the case of the bioactive cyclic dipeptides, their racemic diastereomers inhibitory to bacteria and fungi have been suggested to come solely from Lactobacillus spp. of these strains. We previously demonstrated the antifungal and antiviral activities of proline-based cyclic dipeptides, which were fractionated from culture filtrates of Lb. plantarum LBP-K10 originated from kimchi. However, cyclic dipeptides have not been identified in the filtrates, either from cultures or fermented subject matter, driven by Ln. mesenteroides, which have been widely used as starter cultures for kimchi fermentation. Most importantly, the experimental verification of cyclic dipeptide-content changes during kimchi fermentation have also not been elucidated. Herein, the antibacterial fractions, including cyclo(Leu-Pro) and cyclo(Phe-Pro), from Ln. mesenteroides LBP-K06 culture filtrates, which exhibited a typical chromatographic retention behavior (tR), were identified by using semi-preparative high-performance liquid chromatography and gas chromatography-mass spectrometry. Based on this finding, the proline-based cyclic dipeptides, including cyclo(Ser-Pro), cyclo(Tyr-Pro), and cyclo(Leu-Pro), were additionally identified in the filtrates only when fermenting Chinese cabbage produced with Ln. mesenteroides LBP-K06 starter cultures. The detection and isolation of cyclic dipeptides solely in controlled fermented cabbage were conducted under the control of fermentation-process parameters concomitantly with strong CDP selectivity by using a two-consecutive-purification strategy. Interestingly, cyclic dipeptides in the filtrates, when using this strain as a starter, increased with fermentation time. However, no cyclic dipeptides were observed in the filtrates of other fermented products, including other types of kimchi and fermented materials of plant and animal origin. This is the first report to conclusively demonstrate evidence for the existence of antimicrobial cyclic dipeptides produced by Ln. mesenteroides in kimchi. Through filtrates from lactic acid bacterial cultures and from fermented foods, we have also proved a method of combining chromatographic fractionation and mass spectrometry-based analysis for screening cyclic dipeptide profiling, which may allow evaluation of the fermented dairy foods from a new perspective.

Keywords: cyclic dipeptides, kimchi, Korean fermented foods, lactic acid bacteria, *Leuconostoc mesenteroides* LBP-K06

# INTRODUCTION

The beneficial effects of lactic acid bacteria (LAB) have been steadily demonstrated, such as competitive exclusion of enteric pathogens, tumor suppression through cell-mediated immunity, and especially host defense enhancement by antimicrobial productions (Naidu et al., 1999). Among these pivotal behaviors, antimicrobial activities have closely associated with secretory metabolic products found in LAB themselves and in their culture filtrates (Naidu et al., 1999; Ross et al., 2002; Rouse and Van Sinderen, 2008). They are proficient regulators, governing the influx of the environmental and pathogenic or spoilage-causing microbes in fermented foods: Gram-positive and -negative bacteria—e.g., Arthrobacter sp., Acinetobacter sp., Bacillus subtilis, Escherichia coli, Listeria monocytogenes, Pseudomonas aeruginosa, Staphylococcus aureus—(Sarika et al., 2012), and fungi—e.g., Aspergillus nidulans, Penicillium commune, Fusarium sporotrichioides, Rhodotorula mucilaginosa—(Magnusson et al., 2003; Digaitiene et al., 2012). Additionally, Lactobcillus casei DN-114 001 culture supernatants and Lb. paracasei ST284BZ bacteriocins bacST284BZ have antiviral activities against infections by changing glycosylation or galactosylation of rotavirus receptors in HT-29 cells (Freitas et al., 2003a,b) and herpes simplex virus Type 1, despite its elusive mechanism of actions (Todorov et al., 2008). Thus, the significance of LAB-produced secondary metabolites has been focused, so far, mainly on low-molecularweight molecules seemingly representative of bacteriocins or bacteriocin-like substances and peptidyl or non-peptidyl small molecules (Messens and De, 2002). For this reason, amino acid (AA) metabolism renders bulk contribution to physiological events sufficiently amicable to permit small peptide and protein biosynthesis, pH modulation, metabolic energy/redox balance alteration, and stress resistance by various types of intermediate metabolites, and most importantly, whether differently in raw material fermentation or in culture media fermentation (Fernádez and Zúñiga, 2006; Rouse and Van Sinderen, 2008). For example, when using complex media, Lb. plantarum displays absolute requirements for typical types of essential proteogenic (i.e., isoleucine, leucine, valine, lysine, tryptophan, and threonine) and non-essential proteogenic (i.e., glutamic acid and cysteine) AAs (Saguir and Manca De Nadra, 2007). However, the combination of essential AAs with a simple composition allows for Ln. mesenteroides growth regulated by several types of branched-chain AA transport systems (Foucaud et al., 1997, 2001).

Non-bacteriocin molecules, commonly including organic acids, AAs, fatty acids, diacetyl, and hydrogen peroxide, have been generally characterized as antimicrobial effectors against bacteria, fungi, and viruses (Vandenbergh, 1993; Naidu et al., 1999; Rouse and Van Sinderen, 2008). Particularly, the antimicrobial small compounds from Lactobacillus spp. associated with fermented products have been steadily emphasized. The reutericyclin of Lb. reuteri LTH2584 has been shown to have antibacterial activity against Grampositive bacteria (Gänzle et al., 2000). In culture filtrates from Lactobacillus plantarum VTT E-78076, benzoic acid, methylhydantoin, mevalonolactone, and cyclo(glycyl-l-leucyl) have been found to be active against bacteria and fungi (Niku-Paavola et al., 1999). Furthermore, significant antifungal properties have been demonstrated in Lb. casei AST18 producing compounds—e.g., cyclo(Leu-Pro), 2,6-diphenyl-piperidine, and 5,10-diethoxy-2,3,7,8-tetrahydro-1H,6H-dipyrrolo[1,2 a;1′ ,2′ -d]pyrazine—(Li et al., 2012), Lb. plantarum MiLAB 393 generating substrates—e.g., cyclo(L-Phe-L-Pro), cyclo(L-Phetrans-4-OH-L-Pro), and 3-phenyllactic acid—(Ström et al., 2002), and Lb. plantarum IMAU10014 secreting metabolites e.g., benzoic acid and benzeneacetic acid—(Wang et al., 2012).

Considering these small substances, several types of 2,5-diketopiperazines, cyclic dipeptides (CDPs), and their stereoisomers have been suggested to exhibit potent antimicrobial activities (Witiak and Wei, 1990; Prasad, 1995; Dinsmore and Beshore, 2002; Huang et al., 2010, 2014). Bioactive CDPs function due to their structural chirality and varied side chains; thus, they serve as attractive scaffolds for drug design (Borthwick, 2012). These dipeptidyl cyclic ring closures have been suggested for decades as signal molecules, reducing virulence-factor production and strongly inhibiting microbial growth (Campbell et al., 2009; Huang et al., 2010; Sauguet et al., 2011). For example, cyclo(1Ala-L-Val), cyclo(L-Pro-L-Tyr), and cyclo(L-Phe-L-Pro) activate a LuxR-based N-acyl homoserine lactones biosensor, one of the most intensively investigated families of intercellular signal molecules (Holden et al., 1999). Cyclo(His-Pro), which have also shown potential therapeutic utility in an array of neurological and peripheral inflammatory diseases, are also known as a group of hormone-like molecules appearing in organisms from bacteria to humans (Bellezza et al., 2014).

**Abbreviations:** AA, amino acid; CCK, Chinese cabbage kimchi; CDP, cyclic dipeptide; CI, chemical ionization; EI, electron ionization; GC-MS, gas chromatography–mass spectrometry; HPLC, high-performance liquid chromatography; MIC, minimum inhibitory concentration; MRS, de Man, Rogosa and Sharpe agar; ODS, octadecyl silica; SRK, sliced radish kimchi; WRK, water-based radish kimchi; LAB, lactic acid bacteria; YRK, young radish kimchi.

Our recent reports have demonstrated the bioactivity of proline-based CDPs, including cis-cyclo(L-Leu-L-Pro), cis-cyclo(L-Phe-L-Pro), and cis-cyclo(L-Val-L-Pro). These compounds, fractionated from culture filtrates of Lb. plantarum LBP-K10, were inhibitory to the proliferation of influenza A virus (H3N2) (Kwak et al., 2013) and plant and human pathogenic fungi (Kwak et al., 2014a). Lb. plantarum LBP-K10 was observed to be a key antimicrobial isolate from kimchi, together with other potent Leuconostoc spp., Lactobacillus spp., Weissella spp., and a Lactococcus lactis (Kwak et al., 2013). All these strains have already been recognized to be relevant as the predominant LAB during kimchi fermentation (Lee et al., 1999; Yang and Chang, 2010; Kwak et al., 2014a,b). Particularly, Ln. mesenteroides is an important starter culture strain capable of performing uniform high quality commercial kimchi, which has been generally made from Chinese cabbage (Brassica rapa subsp. pekinensis) (Lee et al., 1992; Jung et al., 2014).

However, despite the importance of exactly which bioactive metabolites appear in starter culture strains and vegetables for kimchi fermentation, no available studies examined CDPs particularly, or their derivatives with potent and selective antimicrobial activity associated with the specific bacterial fermentation products. Most prominently, the experimental evidence for CDP production by Ln. mesenteroides is completely elusive in culture filtrates and fermented materials, as with kimchi. Therefore, this study considers the novel possibility that antimicrobial CDPs in the filtrates from Leuconostoc cultures and from fermented kimchi produced with or without Ln. mesenteroides LBP-K06 starter cultures. Interestingly, we failed to find CDPs in the whole filtrates from non-starter kimchi and other fermented foods of plant or animal origin. We only detected CDPs in Chinese cabbage kimchi (CCK) produced with Ln. mesenteroides starter cultures. Consequently, we identified two CDPs—cyclo(Leu-Pro) and cyclo(Phe-Pro)—in the culture filtrates from Ln. mesenteroides LBP-K06 and three proline-based CDPs—cyclo(Ser-Pro), cyclo(Tyr-Pro), and cyclo(Leu-Pro)—in starter kimchi using high-performance liquid chromatography (HPLC) followed by gas chromatography–mass spectrometry (GC-MS). For the first time, this study demonstrates isolated antimicrobial CDPs and their relative amounts in filtrates from bacterial cultures and from kimchi fermented with Ln. mesenteroides.

# MATERIALS AND METHODS

# Strains and Culture Conditions

All bacterial strains used in this study are listed in **Table 1**. Ln. mesenteroides LBP-K06 and Lb. plantarum LBP-K10 were routinely cultured on modified de Man, Rogosa, and Sharpe (MRS without beef extract) agar (De Man et al., 1960) in a broth at 30◦C for 3 days. The culture was stored anaerobically on 1.0% MRS agar plates at 5◦C. Ln. mesenteroides LBP-K06 was used for fermenting Chinese cabbage, as described previously, with some modifications (Cheigh and Park, 1994).

To observe the antibacterial activity of the isolated substances and filtrate fractions from the bacterial cultures and the fermented kimchi, respectively, we used Gram-positive and negative bacterial indicators and multidrug-resistant bacteria in this study (**Table 1**). All bacterial pathogens were supplied by the Korea National Institute of Health.

# Preparation of Fermented Foods and Their Filtrates

The filtrates from Chinese cabbage were obtained by controlled and spontaneous fermentation with and without inoculation by Ln. mesenteroides LBP-K06 as a starter strain, respectively, as proposed previously (Cheigh and Park, 1994; Jung et al., 2011, 2012, 2014), with some modifications. The spontaneously fermented Chinese cabbage was obtained at the late kimchi fermentation stage. In the case of the controlled fermentation for the production of kimchi made from Chinese cabbage, after it was fermented until the initial stage, an additional fermentation was performed at 25◦C for 72 h until the middle kimchi fermentation stage. Other types of kimchi, including young radish (YRK), water-based radish (WRK), and sliced radish (SRK), were spontaneously fermented without any inoculum to the late kimchi fermentation phase described previously (Cheigh and Park, 1994). Kimchi filtrates were prepared from all types of fermented products by freeze drying, grounding to a powder, and filtering through a #80-mesh (180 micron) sieve. The resulting powder of fermentation filtrates, which was dissolved in sterilized distilled water and filtered with a 0.22µm-cellulose acetate membrane, was subjected to methylene chloride (HClO4) extraction for further HPLC fractionation.

All other fermented products were prepared as follows. The Korean traditional fermented foods of plant origin, including cheonggukjang (fast-fermented bean paste), doenjang (soybean paste), B. subtilis natto, and soy sauce were prepared ¯ according to the proposed methods (Shin and Jeong, 2015). Additionally, pickled or salted shrimp and clams, fermented products of animal origin, were purchased from the traditional market, Togulsaeujeot-gil (underground tunnel in Gwangcheon province), in Chungcheongnam-do in Korea. The fermentation filtrates from the above fermented products were obtained by centrifugation, lyophilization, HClO<sup>4</sup> extraction, and filtration, respectively, as described earlier.

# HPLC Fractionation

The filtrate fractionation was performed by HPLC as proposed previously (Kwak et al., 2013). The resulting primary fractions were consecutively lyophilized, extracted with HClO4, and further fractionated by changing the mobile-phase compositions so as to give an improved quantitative chromatographic fractionation. The filtered samples were separated using a semi-preparative HPLC system (Agilent, USA) with a semipreparative Hypersil octadecyl silica (ODS) C18 reverse-phase column (9.4 × 250 mm, Agilent, USA) and ChemStation HPLC software (Kwak et al., 2013). The initial mobile phase was 67% water, 3% acetonitrile, and 30% methanol for 45 min, and the wavelengths for observing the corresponding chromatograms were 210, 260, and 280 nm, respectively. Each fraction was collected and concentrated by lyophilization to obtain a powder for antibacterial activity determination and GC-MS analysis.


TABLE 1 | LAB strains isolated from three types of Korean traditional fermented vegetables and Gram-positive, Gram-negative, and multidrug-resistant bacteria used in this study.

<sup>a</sup>This was supported by the Korea National Institute of Health.

# Antibacterial Assays

The antibacterial activity against Gram-positive indicators, Gram-negative bacterial indicators, and multidrug-resistant bacteria was examined every 24 h after seed inoculation. The dilution method was used to determine the minimum inhibitory concentration (MIC) of antimicrobial substances (Huys et al., 2002; Paulo et al., 2010). To evaluate antimicrobial activity, we prepared the identified CDPs and filtrate fractions from LAB cultures and from various types of fermented products and tested them against certain pathogenic bacteria.

## Mass Analysis

To perform electron ionization (EI) and chemical ionization (CI) of each fraction, we used GC-MS (Agilent, Germany), as described previously (Kwak et al., 2013). A chromatographic system consisting of an Agilent 6890 series GC equipped with a 7679 series automatic liquid sampler was used. Mass analysis was conducted using a high-resolution mass spectrometer (JEOL JMS-700, Japan).

## Statistical Analysis

Results are presented as means ± standard deviation (SD). The statistical significance of the differences was tested using Student's t-test in Microsoft Office Excel (2013). For all comparisons, values of p < 0.05 (<sup>∗</sup> ) were considered statistically significant.

# RESULTS

# Identification of Proline-Containing CDPs Found in the Culture Filtrate from *Ln. mesenteroides* LBP-K06 Isolated from Kimchi

Ln. mesenteroides LBP-K06, as one of the predominant (45.4%) Leuconostoc spp. among 205 isolates, has exhibited outstanding antibacterial performance on bacterial indicators, E. coli and B. subtilis (Supplementary Tables S1, S2). Inspired by the antibacterial activity of the tested material and to verify that CDPs are produced by Ln. mesenteroides LBP-K06, we examined HPLC retention times (tR) of every fraction that might directly relate to CDP peaks (**Figure 1**). The chromatographic separation of the culture filtrates displayed the particular Leuconostoc peaks distinguished by HPLC elution order. Interestingly, the two isolates showed very similar retention characteristics (**Figure 1**). However, the broad shoulder in the curved shape of the peak area from F13 to F15 of Lb. plantarum LBP-K10 differed from that of the completely sharp peak of N13 of Ln. mesenteroides LBP-K06.

**Table 2** shows the antibacterial activity of each fraction of the cell-free filtrates from the growth of Ln. mesenteroides LBP-K06 in mMRS broth after primary HPLC fractionation on C18, strongly indicating a hydrophobic nature of antimicrobial compounds. As shown in the MIC assay, further separation of the fractions by HPLC and activity against B. subtilis and E. coli highlighted the two fractions, N13 and N15, of the 15 fractions collected. N13 and N15 were repeatedly subjected to semi-prep HPLC, developed with HClO<sup>4</sup> extraction, and were recovered in amounts of 7.09 and 9.51 mg/L, respectively. We obtained the molecular ion [M+1]<sup>+</sup> of fractions N13 and N15 at [M+1]<sup>+</sup> 211 and 245, respectively. Strong evidence of identity was previously established by F13 and F17, including cis-cyclo(L-Leu-L-Pro) and cis-cyclo(L-Phe-L-Pro) (Kwak et al., 2013). As obtained from the distinctive chromatographic fractionation and GC-MS analysis, the EI values of N13 and N15 and their fragmentation patterns were assigned to be C11H18N2O<sup>2</sup> and C14H16N2O2, the proline-based cyclo(Leu-Pro) and cyclo(Phe-Pro) (**Figure 2**, **Table 3**, and Supplementary Figure S1), respectively, coinciding with those produced by Lb. plantarum LBP-K10 (Kwak et al., 2013). These isolated CDPs showed significant antibacterial activity against Gram-positive (i.e., B. subtilis and S. aureus)

TABLE 2 | Relative antibacterial activity of each fraction in the culture filtrate from *Ln. mesenteroides* LBP-K06.


<sup>1</sup> Symbol: +, < 15 mm; ++, < 22 mm; + + +, > 22 mm (Indicator strains: B. subtilis<sup>a</sup> , E. coli<sup>b</sup> ).

<sup>2</sup> MIC: Minimum inhibitory concentration (Indicator strains: B. subtilis<sup>c</sup> , E. coli<sup>d</sup> ).

<sup>3</sup> mg/L.

\*The values represent the average ± SD of three independent experiments (\*p < 0.05).

and Gram-negative (i.e., S. Typhimurium and E. coli) reference bacteria and multidrug-resistant strains (i.e., S. aureus, 11471, and S. Typhimurium, 12219) (**Table 4**), indicating that the resulting MIC values of these compounds gave near identical status to those from Lb. plantarum LBP-K10 (data not shown).

Considering all of these data, we first found that Ln. mesenteroides secretes homologous antimicrobial CDPs and selected a potent antibacterial strain Ln. mesenteroides LBP-K06 as a principal fermenter to isolate CDPs in further study, based on the antibacterial activities of CDPs.

# The Filtrate Fractionation of Kimchi Fermented from Chinese Cabbage and from Other Plant Materials

To further investigate the dynamic CDP production ability of Ln. mesenteroides LBP-K06 during kimchi fermentation, we preliminarily fractionated the filtrates to detect CDPs after spontaneously fermenting uninoculated CCK to the late kimchi fermentation stage (**Figure 3**) as proposed previously (Cheigh and Park, 1994). Even though we hypothesized that the observation of CDPs depends on achieving a particular fermented state by controlling the fermentation time of Chinese cabbage, other types of kimchi, including YRK, WRK, and SRK, also spontaneously fermented as a reference experiment along with bacterial culture filtrates. The HPLC fractionation was then subjected to GC-MS to validate purity of filtrate fractions, including CCK 1–6, YRK 1–5, WRK 1–2, and SRK 1–5. All kimchi filtrates typically seemed to have an increased amount of each fraction in common, particularly at retention times from ∼15 to 25 min (**Figure 3**), consistent with Lb. plantarum LBP-K10 fractions from F8 to F15 as antimicrobial CDP-rich fractions (Kwak et al., 2013, 2014a).

However, a primary HPLC separation of filtrates demonstrated asymmetric and overlapping peaks absorbing at 210 nm (**Figure 3**) with retention times that did not coincide

FIGURE 2 | The identified CDPs from fractions N113 and N15 from *Ln. mesenteroides* LBP-K06. EI fragmentation patterns are indicated. EI-values of N13 and N15 and their fragmentation patterns have been assigned to be the proline-based (A) cyclo(Leu-Pro) (C11H18N2O2) and (B) cyclo(Phe-Pro) (C14H16N2O2), respectively.



<sup>a</sup>Retention time (min).

#### TABLE 4 | Antibacterial activity of N13 and N15 isolated from *Ln. mesenteroides* LBP-K06 filtrate.


<sup>a</sup>MIC: Minimum inhibitory concentration. The values represent the average ± SD (bars) of triplicate determinations as indicated (\*p < 0.05).

<sup>b</sup>Multidrug-resistant Gram-positive bacteria.

<sup>c</sup>Multidrug-resistant Gram-negative bacteria.

with those of any other CDP fraction from previously tested Lactobacillus CDPs (Kwak et al., 2013, 2014a). For example, although the naturally fermented WRK displayed a similar peak profile to that of SRK, and all peaks from SRK were slightly lower than those of WRK, these filtrates did not seem to contain any CDP peak, in contrast to the prediction from the bacterial-isolate culture filtrates such as N13 (F13) and N15 (F17) (**Figures 1**, **3**). Additionally, no compound similar to CDP was observed in any of the fractions, including three types of undetectable YRK peaks (i.e., YRK1, YRK2, and YRK5) and all SRK peaks (i.e., SRK 1–5; **Figure 3**). These fractions were consecutively re-chromatographed on an ODS C18 column under a wide range of different isocratic mobile-phase composition combinations (i.e., 3, 5, 10, 15, and 20% methanol, 3, 5, and 10% acetonitrile, and 67–94% HPLC grade water). Despite elaborate experimental trials to obtain pure components through the recursive HClO<sup>4</sup> extraction and further individual fraction separations, it was difficult to identify peak-area normalization for quantitative impurity content determination. We also failed to verify the relative retention peak area proportion relevant to the fraction cyclo(Phe-Pro) in all types of filtrates in contrast to the isolated bacterial culture filtrates (**Figure 1**). Nevertheless, CCK showed a completely different fractionated profile from any other kimchi and bacterial isolates (**Figure 3**), indicating that the different chromatographic profile of CCK compared to other types of materials might be due to kimchi species-specific characteristics during fermentation. Additionally, in non-starter kimchi, we observed no molecular ions in their poor EI mass spectra, and also fragment ions did not clearly appear from uncertain molecular ions and high-resolution mass measurements (data not shown).

Specifically, because several peaks in radish kimchi fused together, depending on the degree of overlap, the overlapping peaks that might impact each other were more separated by using different isocratic mobile-phase compositions at various pHs to distinguish primary peaks before making accurate estimates of any parameter. Despite efforts to isolate single peaks from complex chromatograms (**Figure 3**), we could not isolate even a single compound in all types of kimchi, despite controlling experimental parameters of elution volume at peak maximum and peak height.

# Identification of the Increased Fractions of Filtrates from CCK Produced with *Ln. mesenteroides* LBP-K06 Starter Cultures by Fermentation Time

Considering the experimentally confirmed unique peak shape and distribution derived from chromatographic resolution characteristics by fermentation source (**Figure 3**), the use of defined starter strains in fermented foods might be pivotal in maintaining starter predominance to control fermentation by microbial growth. We thus had to use Chinese cabbage fermentation filtrates obtained from starter kimchi inoculated with Ln. mesenteroides LBP-K06 in the early kimchi fermentation stage. Moreover, we hypothesized additional fermentation to monitor whether significant peak-pattern change would occur by rapid fermentation at 25◦C for 72 h to the middle kimchi fermentation stage. In contrast to non-starter kimchi, we observed distinct CDP-like peaks designated KF1 to KF6 and their content changed in chromatograms by fermentation time (**Figure 4**). Surprisingly, HPLC peaks constructed with the retention time–peak area data matrices ranging from 15 to 25 min were almost consistent with those of bacterial culture filtrates, including F9 (KF1), F11 (KF2), F12 (KF3, N12), F13 (N13, KF4), and F16 (N14, KF5; **Figure 4**). We then structurally investigated the isolated fractions by GC-MS after conversion to lyophilized compounds. Molecular ion [M+1]<sup>+</sup> of fractions KF1 and KF2 were obtained at [M+1]<sup>+</sup> 185 and 261, respectively. Compared to other fractions, the dramatically more increased fraction KF4 was predicted to be cyclo(Leu-Pro) by CI [M+1]+, 211, similar to cis-cyclo(L-Leu-L-Pro) in our previous work (**Table 5**) (Kwak et al., 2013).

Together with the mass fragmentation pattern under EI, we determined these compounds to be C8H12N2O3, C14H16N2O3, and C11H18N2O2, corresponding to cyclo(Ser-Pro), cyclo(Tyr-Pro), and cyclo(Leu-Pro), respectively (**Table 5**, **Figure 5**), and Supplementary Figure S2). However, fraction KF5, which showed accordance with N14 and F16, increased at 24 h during further rapid fermentation, showing the same retention characteristics (**Figure 4**). The established molecular ion peak was not observed in this fraction. Although predicted from chromatographic peak similarity between N12 and F12 in bacterial isolates (**Figure 1**), fraction KF3 showed unchanging peak-to-peak amplitudes and line widths for 72 h and did not contain any compound identical to the non-starter CCK (**Figure 3**). Therefore, we confirmed changes in the identified CDP contents in the starter kimchi by fermentation time (**Table 6**), as well as different antibacterial activity of the starter kimchi fractions against bacterial indicators following primary separation of filtrates by HPLC (Supplementary Table S3).

# Antibacterial Activity Comparison between Isolated CDPs from Bacterial Culture Filtrates and from CCK Produced with *Ln. mesenteroides* LBP-K06 Starter Cultures

Next, the antibacterial activities of the identified CDP fractions from the starter kimchi were determined through MIC dilution assays against bacterial indicators and multidrug-resistant bacteria. We compared the MIC values of each fraction to determine whether the antimicrobial activities of starter-kimchi CDPs might be consistent with those of Ln. mesenteroides LBP-K06 culture filtrates (**Tables 4**, **7**). The CDP fractions from the Leuconostoc culture filtrates did have significant antibacterial activity against reference strains and multidrug-resistant bacteria (**Tables 4**). We found that the kimchi-filtrate CDP fractions had an MIC of ∼11.0–14.0 and 17.0–19.5 mg/L for cyclo(Leu-Pro) against bacterial indicators and multidrug-resistant bacteria, respectively, corresponding to 10.0–14.0 mg/L for fraction N13 cyclo(Leu-Pro) from bacterial culture filtrates.

Specifically, for the Gram-positive and Gram-negative indicator strains, including B. subtilis, S. aureus, S. Typhimurium, and E. coli, the total growth-inhibition MIC were 13.96, 11.45, 11.78, and 11.3 mg/L, similar to the MIC for cyclo(Leu-Pro) from the Leuconostoc filtrate fraction N13, 13.55, 12.06, 12.98, and 10.41 mg/L (**Table 4**). Additionally, the MIC of 17.98 and 19.33 mg/L for cyclo(Leu-Pro) against multidrug-resistant bacteria showed a similar MIC pattern to a bacterial isolate, 17.28 and 18.19 mg/L. However, the MIC of cyclo(Leu-Pro) against the multidrug-resistant Gram-positive and Gram-negative bacteria were significantly higher, approximately 1.51–1.86 fold, than those against the Gram-positive and Gram-negative indicators (**Tables 4, 7**). The antibacterial activity of cyclo(Tyr-Pro) against bacterial reference strains was lower than that of cyclo(Leu-Pro), showing that Gram-positive or Gram-negative bacterial status was confirmed by the cyclo(Tyr-Pro) MIC of 26.1 and 11.3 mg/L or above. However, the isolated cyclo(Ser-Pro) was not bioactive, as determined by MIC using bacterial indicator strains.

# The Absence of CDPs in Other Fermented Foods of Plant or Animal Origin

To support our hypothesis about CDP production in the CCK starter in contrast to spontaneously fermented kimchi (**Figure 4**, **Tables 5**, **6**), several types of non-starter fermented products, plant and animal, were analyzed for CDPs and their analogs. We screened the filtrates derived from fermented plant materials including soybean paste, B. subtilis natto, soy sauce, and fast- ¯ fermented bean paste—for CDPs (Supplementary Figure S3). We tried to verify fractions from soy sauce and fast-fermented bean paste filtrates (i.e., SS 1-6 and BP 1-2). However, the resulting peaks in HPLC could not be used to separate and collect fractions in all tested fermented products because the HPLC profiles of the corresponding fractions displayed severely overlapping peaks and further fractionated peaks also showed increasing retention volume with unseparated peaks. Thus, the HPLC could not separate pure compounds and also no activity was observed when tested against bacterial indicators. Moreover, the EI/CI values could not be used to calculate the exact mass of molecules as they displayed very poor ionization patterns with few signals in these fermentation filtrates (data not shown). This phenomenon was likely due to the naturally fermented kimchi (**Figure 3**), indicating both the limitation of requirements for amount of samples and the decisive quantification difficulty of all analytes due to the overlapping

peaks from poor chromatographic fractionation. This result also suggests that overlapping peaks might consist of various types of inseparable compounds. In the case of fermented materials of animal origin, two types of Korean traditional pickled or salted shrimp and clams also conveyed no peaks containing CDPs (Supplementary Figure S4), and the subsequent GC-MS analysis revealed nothing of value. Interestingly, a predominant strain in pickled or salted shrimp was Staphylococcus equorum using 16S rRNA sequencing (Supplementary Figure S5) among isolates, coinciding with previous investigation (Jeong et al., 2014). Therefore, we concluded that Ln. mesenteroides could synthesize CDPs in the culture medium and in starter kimchi,


<sup>a</sup>Retention time (min).

but not in non-starter fermented materials. Furthermore, non-LAB species in some types of traditional Korean fermented foods did not seem to be involved in producing CDPs, although many types of Gram-positive or -negative bacteria were elucidated to produce and excrete CDPs, as previously reported (Kwon et al., 2000; Rhee, 2002; Yan et al., 2004; Lind et al., 2007; Huang et al., 2010, 2014).

# DISCUSSION

Based on our previous findings (Kwak et al., 2013, 2014a), we used bacterial filtrates from the supernatant and freezedried powder of all kinds of kimchi, readily available for CDP purification. This strategy prominently comes from repeated methodological development applying 10-fold HClO<sup>4</sup> extraction with remarkable selectivity for CDP isolation and several time change of mobile phase, strongly rendering CDPs sufficiently pure to be analyzed and confirmed by GC-MS for EI/CI mass spectrometry as a single compound without any other peptidyl or non-peptidyl compounds. By using this CDP characterization model, we exploited specific-strain starter systems because use of filtrates from non-starter kimchi have been proven to have limitations for collecting any CDP or single fraction directly due to poor chromatographic separation (**Figure 3**). This result strongly suggests the possible interference of a hydrophobic nature by poor chromatographic-driven polar-component content changes commonly observed in foodstuffs (Young, 2016), crucially arising from secondary metabolite profile alteration during spontaneous fermentation by necessarily complex heterofermentative microflora (Jung et al., 2012). This phenomenon can be reliably and meaningfully evidenced by secondary byproduct changes of starter and non-starter kimchi affected by the controlled microflora (Jung et al., 2012, 2014; Jeong et al., 2013), along with more poor chromatographic retention behaviors of the analytes caused by polar-compound increases resulting from metabolite changes of foodstuffs (Young, 2016).

Other studies also support our current work regarding starter and non-starter kimchi, indicating that Ln. mesenteroides starter normally maintains its predominance ∼88% during fermentation (Eom et al., 2008) as an absolute dominant over Lb. plantarum and other microflora during the early kimchi fermentation stage (Jung et al., 2012). To prevent an influx of other kimchi microflora (e.g., Leuconostoc spp., Lactobacillus spp., and Weissella spp.) governing fermentation rate, microbial


<sup>a</sup>Data presented as means with standard deviation (SD). Asterisks represent statistically significant values (\*p < 0.05).

#### TABLE 7 | Antibacterial activity of identified CDPs from starter kimchi against Gram-positive and Gram-negative bacterial indicators and multidrug-resistant bacteria.


<sup>a</sup>MIC: Minimum inhibitory concentration.

<sup>b</sup>Multidrug-resistant Gram-positive bacteria.

<sup>c</sup>Multidrug-resistant Gram-negative bacteria.

<sup>d</sup>The values represent the average ± SD (bars) of triplicate determinations as indicated (\*p < 0.05).

community, and extra metabolite production (Lee et al., 2008; Chang and Chang, 2010), we thereby modeled the kimchi starter using Chinese cabbage, which has been widely used in Korea for making ordinary kimchi (Jung et al., 2011, 2012, 2014), inoculated with Ln. mesenteroides LBP-K06 as a strong candidate for antimicrobial CDP-biosynthesizing isolate (**Figure 4**).

Underlying principles of this fermentation control by starter cultures primarily come from the evaluation of poor chromatographic separation of non-starter cultures with no detection of CDP peaks in our experiments. Simultaneously, different from a naturally fermented kimchi for dairy use, our factorial design-based study controlling fermentation process parameters (Panda et al., 2007) including temperature, pH, time, and inoculum amount, has been employed as a close-ended system in every screening round when and whether testing the primary CDP detection according to the type of raw materials and thereafter with the secondary use of starter cultures, to efficiently produce/characterize bioactive CDPs. The effect of designed fermentation strategies herein remarkably enhanced the yield of detectable CDPs that are most importantly dependent on parameters using the starter strain Ln. mesenteroides LBP-K06 (**Figure 3**) and the specific raw material Chinese cabbage (**Figure 4**). These two-type parameter-controlled fermentations accompanying fermentation time-course modifications adjusted by fermentation temperature experimentally facilitated the monitoring of different HPLC chromatographic separation patterns, showing and not showing distinctly changing CDP peaks between the controlled and the spontaneously fermented products.

Although the production of kimchi CDPs in a close-ended system is evidently proven by the microbial community from starter cultures used as a predominant fermentation parameter, the entire byproduct contents resulting from metabolite changes that undergo fermentation also seem to closely align with the type of kimchi material from our results. This assumption lies in that a different dietary composition and ratio in raw materials of plant (or animal) origin may drive different fermentation metabolite production coupled with defined starter-culture regulation. Results of the content analysis by a previous study (Jung et al., 2012) convincingly supported our hypothesis, suggesting the significant quantitative difference of CCK and YRK carbon sources, such as glucose and fructose. The significant different levels of carbon source profiles in raw materials before fermentation maintain constantly until the middle stage of fermentation and thereby fundamentally can influence LAB metabolism in the production of various fermentation products. Specifically, free sugars in Chinese cabbage are relatively higher than those in radish commonly found in raw materials and fermented kimchi.

Interestingly, other evidence of differences in foodcomposition content of raw materials affecting metabolite production during fermentation is explained by free AA concentrations in raw vegetables (Kim et al., 2009). As illustrated in the different content of carbon sources between Chinese cabbage and radish, the verified content of various AAs in these raw materials also show an almost similar pattern to carbon sources. The corresponding content of all kinds of AAs, such as branched AAs, sulfur-containing AAs (i.e., methionine and cysteine), and aromatic AAs (i.e., phenylealanine and tyrosine), represent the relative quantitative difference among the raw vegetables. Thus, the relatively lower level of nutrient availability driven by the raw vegetables seems to be responsible for affecting the metabolite profile in fermented products, coinciding with a non-starter WRK study showing remarkably lowered amounts of metabolites (Jeong et al., 2013), including AAs, organic acids, glucose, and fructose, in contrast to those in starter and nonstarter CCK and YRK at the same fermentation stage (Jung et al., 2012). Similarly, our HPLC data and further CDP purification study from several types of radish kimchi (SRK, WRK, and YRK) also do not correlate with CDP production, presumably due to the relatively and remarkably lower levels of AAs and carbon sources of raw materials than those of Chinese cabbage. These phenomena imply that kimchi metabolites are important dietary components and their composition might be partly applied to predicting, estimating, or evaluating CDP-producing behaviors along with kimchi tastes or flavors. Hence, changes in metabolite compositions, including organic acids (i.e., lactic and acetic acids) and flavoring compounds (i.e., mannitol and amino acids) (Ha et al., 1989), can give a partial possibility of CDP action during fermentation according to type of kimchi material. Therefore, for example, we suggest here Chinese cabbage with starter Ln. mesenteroides to make CDP-rich kimchi to meet CDP functions. Our data convincingly demonstrate, for the first time, CDP production in the controlled fermented cabbage—under the control of specific fermentation-process parameters, including starter dominance, time-course modification, and temperature adjustment—concomitantly coupled with strong CDP selectivity by using a two-consecutive purification strategy.

Although negligible differences emerged during the entire fermentation period in the production of organic acids and mannitol in fermented vegetables (Jung et al., 2012), which is a naturally occurring, non-carcinogenic, and diabetic polyol at high levels in kimchi, the use of Ln. mesenteroides as a starter culture inspires commercial production of fermented kimchi (Grobben et al., 2001). In the present study, 85.7% of fermented kimchi by obligately heterofermentative microflora is occupied by Chinese cabbage (Ji et al., 2009). This commercial trend is also thought to facilitate monitoring and assessment of starter Leuconostoc-driven CDPs.

In contrast, in the case of culture filtrates of Lactobacillus and Leuconostoc cells, CDPs significantly showed the time point with the highest total amount at 72 h simultaneously with a decline in cell numbers (Kwak et al., 2013) (**Figure 1**). Similar relationships are observed in agr-mediated dual-channel quorum-sensing signaling engaged in CDP production of Lb. reuteri RC-14 cyclo(l-Phe-l-Pro) and cyclo(l-Tyr-l-Pro) (Li et al., 2011). These growths of Lactobacillus or Leuconostoc species in culture media have been commonly recognized to require complex nitrogen sources (Amoroso et al., 1993; Elli et al., 1999). Growth behaviors entirely derived from complex medium in laboratory experiments seem to be very attractive for use in fermentation standardization, but completely different from Chinese cabbage or other raw-material fermentation when seen as part of the AA use pattern, whether used in growth in media or raw materials of plant or animal origin. Chinese cabbage normally contains only 663 mg of total AA in 100 g of raw vegetables (Ji et al., 2009) completely different from rich MRS medium. Additionally, Leuconostoc strains used in the cheesemaking process show that AA use is limited to branchedchain AAs in the case of milk, because L-leucine and L-valine commonly act as competitive inhibitors occupied by branched AA transport systems for the uptake of AAs in Ln. mesenteroides (Mayshak et al., 1966; Winters et al., 1991; Foucaud et al., 2001). Hence, these properties of having species-specific metabolism presumably affecting AA concentrations and consecutively other metabolite changes are also thought to align with aspects of a key feature of CDP production in raw material fermentation affected by fermentation control parameters.

Specifically, CDPs containing 2, 5-diketopiperazines come from their rigid backbone that can mimic preferential peptide conformations and contain highly constrained AAs essentially resulting from the double condensation of two α-AAs (Ciarkowski, 1984). Because starter or non-starter ripening flora uses AAs to a greater extent as part of their primary metabolic activities in fermenting kimchi, AA consumption influenced by CDP biosynthesis seems to be in line with a previous finding of gradually decreased AA content after the middle fermentation stage (Jeong et al., 2013). As evidence for this hypothesis, proline-based CDP fractions, including KF1, KF2, and KF4, showed remarkably higher content proportional to fermentation time compared to those of starter kimchi at the early fermentation period (0 h) (**Figure 4** and **Table 6**). Moreover, the absence of CDP, especially cyclo(Phe-Pro), in the non-starter kimchi (**Figure 3**) and in other spontaneously fermented products (Supplementary Figures S3, S4), in contrast to the presence of cyclo(Ser-Pro), cyclo(Tyr-Pro), and cyclo(Leu-Pro) in the controlled fermented cabbage (**Table 4**), also coincidentally correspond to previously established experiments on the remarkably lower content of specific AAs, including tyrosine, histidine, threonine, alanine, valine, phenylalanine, isoleucine, leucine, and methionine, in spontaneously fermented cabbage, compared to controlled fermented cabbage using Ln. mesenteroides NCIM 2073 as a starter after the middle kimchi fermentation stage (Jagannath et al., 2012). Additionally, finding cyclo(Tyr-Pro) containing one type of aromatic AA, tyrosine (**Table 5**), also seems to be caused by the use of the defined starter Ln. mesenteroides LBP-K06 because hydroxy phenyllactic acid produced from tyrosine has an antimicrobial spectrum as phenyllactic acid, particularly in starter kimchi (Crowleya et al., 2013; Naz et al., 2013). These results strongly imply that spontaneously fermented cabbage is likely to have a varied microflora, as seen in the very different chromatographic separation pattern and CDP production compared to starter kimchi (**Figures 3**, **4**).

Considering the antibacterial activity of cyclo(Leu-Pro) against multidrug-resistant S. aureus 11471 and S. Typhimurium 12219, we observed MIC values (**Table 7**) similar to 17.28 and 18.19 mg/L for cyclo(L-Leu-L-Pro) from Ln. mesenteroides LBP-K06 culture filtrates. In the case of cyclo(Tyr-Pro), its active concentration against bacterial strains seems to be very similar to the MIC of 31.25 mg/L from previous investigations of Streptomyces sp. strain 22-4 cyclo(L-Pro-L-Tyr) and cyclo(D-Pro-L-Tyr), which displayed antibacterial activity against Xanthomonas axonopodis pv. Citri and Ralstonia solanacearum (Wattana-Amorn et al., 2015), although different pathogenic strains were used to test antimicrobial activity. These results also coincide with previous reports that cyclo(Leu-Pro) (Yang et al., 2011), cyclo(Tyr-Pro) (Kwak et al., 2014a), and cyclo(Phe-Pro) have antibacterial or antifungal activities (Ström et al., 2002). Additionally, maturation of gastrointestinal cells was enhanced by synthetic cyclo(Phe-Pro) and cyclo(Tyr-Pro) (Graz et al., 1999). Interestingly, the antimicrobial CDP fractions from kimchi filtrates (**Table 5**) from Ln. mesenteroides LBP-K06 and Lb. plantarum LBP-K10 were mainly found in the latter part of HPLC chromatogram from ∼15 to 32 min from N9 (F9) to N15 (F17).

Additionally, non-antimicrobial fractions of Leuconostoc and kimchi filtrate were displayed in the forepart of the chromatograms (**Table 2**). These results coincide with our preliminary experiments showing that the significantly higher amounts of CDPs only from F11 to F15 were produced by L-proline-supplemented Lactobacillus cells in the presence of D-glucose under buffer conditions; glucose-depleted or supplemented buffer conditions made no change in the amount of other fractions irrelevant to the presence or absence of Lproline (data not shown). Similarly, the relative content of cyclo(Leu-Pro) and cyclo(Tyr-Pro) from starter kimchi were significantly more affected than other fractions by fermentation time (**Figure 4** and **Table 5**), suggesting that fermentingcondition optimization (ripening time and temperature) could change CDP content in kimchi. Also, despite differences in the amount of antimicrobial substances among isolates (**Table 6**), Ln. mesenteroides excreted CDPs that acted as bioactive mediators, found in common with specific kimchi types. Therefore, the time-dependent fermentation control strategy for screening CDPs in CCK suggests that Ln. mesenteroides LBP-K06 inoculation as a predominant starter might facilitate detection of CDPs, whereas food fermentation, usually relying on naturally inoculated (inherent) microbial flora, resulted in variable and uncontrolled product quality. This result also corresponds with the result that naturally fermented materials might be insufficient to evaluate the starter and its predominance in such complex and mixed-strain microbial variety in foods (Giraffa and Rossetti, 2004).

# CONCLUSION

Our current study aimed to demonstrate CDPs in Korean fermented products with antimicrobial activity against pathogenic microbes for the first time. Our results focused

# REFERENCES


on controlled Chinese-cabbage fermentation with a single-strain starter culture, presumably affecting the growth of LAB and their metabolites. This hypothesis was reflected in the significant proline-based CDP production of starter kimchi in contrast to that of other fermented products. Moreover, CDP production during kimchi fermentation was observed to be different from culture filtrates of Ln. mesenteroides LBP-K06. The unique chromatographic profile of Chinese cabbage filtrates of starter kimchi inoculated with Ln. mesenteroides LBP-K06 can be used to isolate unknown compounds using other fermented materials in further study.

This is the first report showing that antibiotic CDPs in CCK filtrates might be applied for antimicrobial preservation and other purposes. Based on experimental clues from MIC determination, we demonstrated the active concentration of each CDP in kimchi filtrates to be similar to that of CDPs in previously established bacterial filtrate fractions. Our experiments confirm beneficial substances in kimchi. These findings can provide a framework for future research or industry regarding the fermenting process or of making kimchi. Additionally, antibiotic CDPs in these fermented products might provide the capability to assess the antibiotic effects of CDPs and other possible applications.

# AUTHOR CONTRIBUTIONS

MK, RL, AK, and SK designed the research. MK, RL, and AK performed the research. MK, RL, AK, and SK analyzed the data. MK and RL contributed new reagents/analytic tools. MK, RL, and SK wrote the manuscript with significant input from the others.

# ACKNOWLEDGMENTS

The authors thank Dr. Chun Kang and Dr. Gi-eun Rhie (Center for Infectious Diseases, National Institute of Health, Korea Centers for Disease Control and Prevention, Cheongwongun, Chungcheongbukdo, South Korea) for helpful materials regarding multidrug-resistant bacteria. We also thank NCIRF at Seoul National University for supporting the GC-MS experiments. This work was supported by the Research Fellowship of the BK21plus project.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2017.00761/full#supplementary-material


strain, Leuconostoc citreum GJ7 as a starter. J. Food Sci. 75, M103–M110. doi: 10.1111/j.1750-3841.2009.01486.x


isolates of lactic acid bacteria. FEMS Microbiol. Lett. 219, 129−135. doi: 10.1016/S0378-1097(02)01207-7


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

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

# Environmental Factors Affecting Microbiota Dynamics during Traditional Solid-state Fermentation of Chinese Daqu Starter

Pan Li<sup>1</sup> , Weifeng Lin<sup>2</sup> , Xiong Liu<sup>1</sup> , Xiaowen Wang<sup>1</sup> and Lixin Luo<sup>1</sup> \*

<sup>1</sup> Guangdong Key Laboratory of Fermentation and Enzyme Engineering, School of Bioscience and Bioengineering, South China University of Technology, Guangzhou, China, <sup>2</sup> College of Light Industry and Food Sciences, South China University of Technology, Guangzhou, China

Edited by:

Michael Gänzle, University of Alberta, Canada

#### Reviewed by:

Jinshui Zheng, Huazhong Agricultural University, China Jussi Loponen, Fazer, Finland

> \*Correspondence: Lixin Luo btlxluo@scut.edu.cn

#### Specialty section:

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

Received: 23 May 2016 Accepted: 25 July 2016 Published: 04 August 2016

#### Citation:

Li P, Lin W, Liu X, Wang X and Luo L (2016) Environmental Factors Affecting Microbiota Dynamics during Traditional Solid-state Fermentation of Chinese Daqu Starter. Front. Microbiol. 7:1237. doi: 10.3389/fmicb.2016.01237 In this study, we investigated the microbiota dynamics during two industrial-scale traditional solid-state fermentation (SSF) processes of Daqu starters. Similar evolution profiles of environmental parameters, enzymatic activities, microbial amounts, and communities were observed during the medium temperature SSF (MTSSF) and low temperature SSF (LTSSF) processes. Orders of Rickettsiales and Streptophyta only dominated the initial 2 days, and Eurotiales only predominated from days 10 to 24, however, phylotypes of Enterobacteriales, Lactobacillales, Bacillales, Saccharomycetales, and Mucorales both prevailed throughout the MTSSF and LTSSF processes. Nevertheless, the pH in MTSSF process on day 5 were 5.28, while in LTSSF process (4.87) significantly lower (P < 0.05). The glucoamylase activities in MTSSF process dropped from 902.71 to 394.33 mg glucose g−<sup>1</sup> h <sup>−</sup><sup>1</sup> on days 5 to 24, while significantly lower (P < 0.05) in LTSSF process and decreased from 512.25 to 268.69 mg glucose g−<sup>1</sup> h −1 . The relative abundance of Enterobacteriales and Lactobacillales in MTSSF process constituted from 10.30 to 71.73% and 2.34 to 16.68%, while in LTSSF process ranged from 3.16 to 41.06% and 8.43 to 57.39%, respectively. The relative abundance of Eurotiales in MTSSF process on days 10 to 24 decreased from 36.10 to 28.63%, while obviously higher in LTSSF process and increased from 52.00 to 72.97%. Furthermore, lower bacterial richness but higher fungal richness were displayed, markedly differences in bacterial communities but highly similarities in fungal communities were exhibited, during MTSSF process comparatively to the LTSSF process. Canonical correspondence analysis revealed microbial structure transition happened at thermophilic stages under environmental stress of moisture, pH, acidity, and pile temperature. These profound understanding might help to effectively control the traditional Daqu SSF process by adjusting relevant environmental parameters.

Keywords: solid-state fermentation, Daqu, microbiota dynamics, environmental factors, temperature, relationship

# INTRODUCTION

fmicb-07-01237 August 2, 2016 Time: 13:18 # 2

Highly complex microbial communities play a critical role for ecosystems by supporting the main global biogeochemical cycles, however, microorganisms are greatly affected by environmental factors. Over recent decades, extensive studies were performed to provide profound insights into the intricate relationship between the environmental parameters and the microbial dynamics in various ecosystems including marine ecosystems (Gilbert et al., 2012; Tinta et al., 2015), hot springs (Song et al., 2013; Coman et al., 2015), soils (De Gannes et al., 2015; Žifcáková et al., 2016 ˇ ) and composting systems (Lu et al., 2015). These studies have shown that environmental conditions dictate the structuring of microbial composition. Among these environmental factors, temperature was shown to largely affect the microbial dynamics, and the microbial diversity generally decreased with temperature increased (Miller et al., 2009). Consequently, the understanding on how highly variable environmental parameters affect microbial structure might lead to predictable patterns of microbial assemblages.

From an ecological point of view, traditional spontaneous solid-state fermentation (SSF) of cereal starters, which locally called "Daqu" and anciently used as starters to produce Chinese liquor and vinegar, is a dynamic process due to the combined activity of varieties of microbial populations, which were linked to consecutive environmental conditions (Zheng et al., 2012). Various factors, such as temperature, moisture content, pH and acidity, related to one another determine the succession of the different environmental conditions appearing throughout the SSF process. Generally, Daqu is produced in an open-work environment with non-autoclaved raw materials of mixtures of barley, wheat and peas, the preparation process of Daqu mainly involved three stages (Zheng et al., 2012): (i) material grinding, mixing, and shaping; (ii) spontaneous SSF process with temperature controlled; and (iii) drying and ripening. According to the highest inoculation temperature, three typical types of Daqu starters can be distinguished, low temperature Daqu (45–50◦C), medium temperature Daqu (50–60◦C) and high temperature Daqu (60–65◦C). Accordingly, the moisture content obviously decreased from approximately 35–40% to 8– 12% during Daqu SSF process (Zheng et al., 2014; Li et al., 2015). In responses to these environmental variations, microbial communities must undergo complex changes during the Daqu SSF process.

Unfortunately, despite a growing understanding of the microbial structure and dynamics during the SSF process of various types of Daqu starters (Nie et al., 2013; Li et al., 2015), however, only two previous studies have addressed the potential links between microbial diversity and environmental factors, moisture content and temperature were found to be strongly correlated with the composition of Bacillus sp. and thermophilic fungi (Zheng et al., 2014; Wang and Xu, 2015). Thus, a wide variety of thermophilic and drought-resistant communities, such as Bacillales, Eurotiales, and Mucorales, have been detected in various types of Daqu starters (Li et al., 2015; Wang and Xu, 2015). Likewise, our recent study also indicated that the composition of the microbial communities was significantly corrected with temperature, acidity and moisture content during Daqu SSF process (Li et al., 2015).

Up to date, the knowledge on how highly variable environmental parameters affect microbial community structure in Daqu SSF process is still scarce, no comparative studies have been performed to clear the relationship of microbial community composition, abundance, and diversity with different environmental factors gradient. In present study, we analyzed the microbiota dynamics and compared the difference in microbial communities during a medium temperature SSF (MTSSF) process and a low temperature SSF (LTSSF) process of Daqu starters by Illumina-based high-throughput sequencing and quantitative PCR (qPCR) analyses, and indicated whether these differences in microbial assemblages were due to environmental factors changes. To our knowledge, this is the first report to perform comparative studies to investigate the influence of environmental factors on microbiota dynamics during Daqu SSF process.

# MATERIALS AND METHODS

# Solid-state Fermentation of Daqu and Sampling

Spontaneous SSF of Daqu starters at an industrial scale was performed in the fermentation room of a traditional vinegar production factory in Shanxi province, China. Initially, cereal materials of approximately 4200 kg barley and 1800 kg wheat were ground and mixed. Then the mixtures were stirred with the addition of 36–37% water, and were shaped into bricks (28 cm × 18 cm × 5 cm) and layer-by-layer piled in the fermentation rooms. The stacked layers of Daqu blocks were incubated for 24 days with strict temperature control. According to traditional SSF techniques, the variation of room temperature and core temperature of Daqu SSF process was controlled by forced ventilation, and the piles of Daqu blocks were manual turned every 2 days at the thermophilic and cooling stages of SSF process to allow adequate aeration and to decrease the inoculation temperature. To investigate the influence of environmental factors on microbiota dynamics during Daqu SSF process, a MTSSF process and a LTSSF process of Daqu starters were conducted with the highest inoculation temperature reached 48–53◦C and 53–55◦C (**Figure 1**, Stage B), respectively. The MTSSF and LTSSF processes of Daqu starters were both conducted through two independent experiments. Daqu samples were separately collected at days 1, 2, 5, 10, 14, and 24 according to temperature evolution during the SSF process (**Figure 1**). To obtain adequate information and representation, Daqu blocks from each stage were randomly selected from the upper, middle, and lower locations of two individual processes in triplicates, which were then ground, mixed, and pooled into sterile Stomacher bags (Stomacher Lab System, London, United Kingdom) to provide an experimental Daqu powder sample (approximately 500 g). All of the samples were stored at −20◦C for further analysis.

# Physicochemical and Enzymatic Analysis

The room temperature was monitored every day by a thermohygrometer (GEMlead TH339, Fuzhou, China) stuck on the wall of fermentation room, and the core temperature of Daqu (pile temperature) was determined every day using a thermograph (Neo Thermo TVS-700, Nippon Avionics, Tokyo, Japan) inserted into Daqu blocks with a depth of 1 cm. The moisture of Daqu was determined by dry/wet weight measurement method at 105◦C. The pH was measured with a pH meter (Sartorius PB-10, Germany). The total titratable acidity was determined by titration with 0.02 M NaOH exhibiting a titration endpoint of pH 8.2. The reducing sugar content was determined by DNS method (Miller, 1959). The amino acid nitrogen content and protease activity were determined according to the national professional standard methods (SB/T 10317-1999, 1999; QB/T 4257-2011, 2011). One unit of protease activity was defined as the amount of amino acid nitrogen liberated per hour by 1 g Daqu under the assay conditions. Amylase and glucoamylase activities were determined as our previously described (Li et al., 2015). One unit of amylase activity was defined as the amount of starch liquefied per hour by 1 g Daqu in sodium acetate buffer (50 mM, pH 4.6) at 35◦C. One unit of glucoamylase activity was defined as the amount of glucose liberated per hour by 1 g Daqu in sodium acetate buffer (50 mM, pH 4.6) at 40◦C.

# DNA Extraction and qPCR Analysis

DNA extraction from Daqu samples was performed using the Soil DNA Kit (Omega Bio-Tek, Norcross, GA, USA) according to the manufacturer's instructions. qPCR analysis was performed in quadruplicate using the commercial kit (SYBR <sup>R</sup> Premix Ex TaqTM II, Takara, Dalian, China) with an ABI 7500 Real Time PCR System (Applied Biosystems). Primers pairs P1/P2, Lac1/Lac2, B1/B2, and Y1/Y2 were used to quantify the specific gene abundance of total bacteria, LAB, Bacillus, and fungi (Muyzer et al., 1993; Xu et al., 2011), respectively. Each reaction was performed in a 25 µL volume containing 12.5 µL SYBR Premix Ex Taq (Takara, Dalian, China), 0.5 µL of each primer (10 mM) and 2 µL of 10-fold (Bacillus) or 100-fold (bacteria, LAB and fungi) dilution DNA template. The qPCR thermocycling steps were as follows: 95◦C for 30 s, 40 cycles of 95◦C for 5 s, 55◦C for 34 s, 72◦C for 30 s. The fluorescent products were detected at the annealing step of each cycle. Melting curve analysis and agarose gel electrophoresis were performed to confirm the specificity of the amplification. The amplification efficiency and correlation coefficient (R 2 ) for the amplification of specific gene of bacteria, Bacillus, LAB and fungi were 99.3% and 0.992, 97.1% and 0.990, 94.4% and 0.991, 98.1% and 0.991, respectively.

# Illumina HiSeq Sequencing

fmicb-07-01237 August 2, 2016 Time: 13:18 # 4

The V4 regions of bacterial 16S rRNA gene and ITS1 regions of fungal rRNA genes were amplified used the specific primers 515f/806r (Peiffer et al., 2013) and ITS5-1737F/ITS2-2043R (Huang et al., 2016) with the barcodes, respectively. All PCR reactions were carried out in triplicate 30 µL reactions with 15 µL of Phusion <sup>R</sup> High-Fidelity PCR Master Mix (New England BioLabs), 0.2 µM of each primer and 10 ng DNA templates. Thermal cycling consisted of initial activation at 98◦C for 1 min, followed by 30 cycles of denaturation at 98◦C for 10 s, annealing at 50◦C for 30 s, and elongation at 72◦C for 60 s and finally elongation at 72◦C for 5 min. Negative control were treated similarly with the exclusion of template DNA and failed to produce visible PCR products. PCR products were mixed in equimolar ratios and mixture PCR products were purified with QIAquick Gel Extraction Kit (QIAGEN, Dusseldorf, Germany). Sequencing libraries were generated using TruSeq <sup>R</sup> DNA PCR-Free Sample Preparation Kit (Illumina, USA) following manufacturer's recommendations and index adaptors were added. The library quality was assessed on the Qubit@ 2.0 Fluorometer (Thermo Scientific) and Agilent Bioanalyzer 2100 system. Finally, the library was sequenced on an Illumina HiSeq2500 platform, at Novogene, Beijing, China.

# Data Processing and Bioinformatics Processing

Paired-end reads from the original DNA fragments were merged using FLASH (Zhang et al., 2014) and assigned to each sample with the unique barcodes. UPARSE software package (Uparse v7.0.1001) with the UPARSE-OTU and UPARSE-OTUref algorithms was used to pick operational taxonomic units (OTUs) at the 97% similarity (Edgar, 2013). Representative sequences were picked for each OTU, and RDP classifier (Version 2.2) was used to annotate the taxonomic information for each representative sequence. Alpha diversity indices Chao1, Shannon, Simpson, ACE and Goods coverage were performed in QIIME (Version 1.7.0) to reflect the diversity and richness of microbial community in different samples (Caporaso et al., 2010). For beta diversity, QIIME calculated both the unweighted and weighted UniFrac distances (Lozupone et al., 2011). Sequencing results are available through the GenBank sequence read archive database under accession number PRJNA316566.

# Statistical Analysis

The statistical significance (P ≤ 0.05) of the difference among different batches were identified using a one-way analysis of variance (ANOVA). Pearson's test was performed to reveal the correlations between environmental variables and abundant classes using SPSS Statistics 19.0. Principal coordinate analysis (PCoA) was performed with the weighted UniFrac distance. Paired t-test and wilcoxon tests within the stats R package (Version 2.15.3) were performed to test whether there was a significant difference in the alpha and beta diversity indices among the two batches. In addition, Analysis of similarities (ANOSIM) (Clarke, 1993) and multi-response permutation procedure (MRPP) (He et al., 2010) analyses were further employed to examine the community difference among the two batches. Canonical correspondence analysis (CCA) between Daqu microbial community and measured variables was performed with Canoco 5.0 software.

# RESULTS

# Dynamics of Physicochemical Characteristics

Temperature was a universal indicator used to monitor and control the Daqu SSF process. Dynamics of this parameter and other physicochemical characteristics throughout the SSF process are shown in **Figure 1**. The thermal profile was followed the typical evolution of Daqu SSF process, and allowed to distinguish four stages during the MTSSF and LTSSF processes: the mesophilic stage (days 1 to 2), the thermophilic stage (days 2 to 10), the cooling stage (days 10 to 14), and the maturation stage (days 14 to 24) (**Figure 1A**). However, the temperature of MTSSF process was generally 3–7◦C and 5–12◦C higher than that in LTSSF process at thermophilic stage and cooling stage (**Figure 1A**), respectively. The temperature increased from 48 to 53◦C and 44 to 50◦C (Phase A of thermophilic stage), and maintained around 55◦C and 50◦C (Phase B of thermophilic stage) at thermophilic stage of MTSSF and LTSSF processes, respectively.

Generally, the evolution profiles of other physicochemical parameters were similar during the MTSSF and LTSSF processes. The moisture slightly declined, and then markedly decreased after 2 days and finally fell to 12.18 and 11.38% (**Figure 1B**), respectively. The reducing sugar contents shown a substantially decreased tendency during the MTSSF process with the exception of cooling stage, however, a noteworthy increase of reducing sugar contents (32.94 mg g−<sup>1</sup> dry starter) could be observed at phase A and followed by a quick depletion (dropped to 15.33 mg g−<sup>1</sup> dry starter) at phase B of the thermophilic stage during the LTSSF process (**Figure 1B**). Despite the variation in initial pH and titratable acidity, the pH firstly decreased to about 4.22 and the titratable acidity increased to approximately 0.20 mmol g−<sup>1</sup> at mesophilic stage, but afterward the pH obviously increased and the titratable acidity strongly decreased at the thermophilic stage, and finally the pH and titratable acidity both maintained near 6.40 and 0.06 mmol g−<sup>1</sup> until the end of MTSSF and LTSSF processes (**Figure 1C**). Nevertheless, significantly (P < 0.05) higher in pH and lower in titratable acidity were displayed at the mesophilic and thermophilic stages of the MTSSF process compare with the LTSSF process (**Figure 1C**). The pH and titratable acidity on day 5 were 5.28 and 4.87, 0.11, and 0.17 mmol g−<sup>1</sup> during MTSSF and LTSSF processes, respectively.

# Dynamics of Enzymatic Activities

fmicb-07-01237 August 2, 2016 Time: 13:18 # 5

Dynamics of enzymatic activities throughout the SSF process are shown in **Figure 2**. Overall, the evolution profiles of enzymatic activities were similar during the MTSSF and LTSSF processes. The protease activities increased from 0.09 to 0.82 mg amino acid nitrogen g−<sup>1</sup> h <sup>−</sup><sup>1</sup> on days 1 to 14 and fell to 0.66 mg amino acid nitrogen g−<sup>1</sup> h −1 at the end of MTSS process, however, the protease activities shown a gradual increased tendency from initial 0.11 to 0.72 mg amino acid nitrogen g−<sup>1</sup> h −1 at the end of the LTSS process (**Figure 2A**). The contents of amino acid nitrogen slightly increased within the first 2 days, and significantly (P < 0.05) increased from days 2 to 5 and fluctuated around a peak on day 5. Moreover, significantly (P < 0.05) higher in protease activity and amino acid nitrogen content were displayed on day 14 during MTSSF process compare with the LTSSF process (**Figure 2A**). The amylase activities slightly declined within the first 2 days, and then increased rapidly from days 2 to 5, and maintained near 1.37 g liquefied starch g−<sup>1</sup> h −1 until the end of MTSSF and LTSSF processes (**Figure 2B**). However, the glucoamylase activity slightly decreased within the first 2 days, and then obviously declined throughout the MTSSF and LTSSF processes (**Figure 2B**). The glucoamylase activities in MTSSF process dropped steadily from 902.71 to 394.33 mg glucose g−<sup>1</sup> h <sup>−</sup><sup>1</sup> on days 5 to 24, in LTSSF process from 512.25 to 268.69 mg glucose g−<sup>1</sup> h −1 . Significantly (P < 0.05) higher in glucoamylase activities were noted after the mesophilic stage of the MTSSF process compare with the LTSSF process (**Figure 2B**).

# Biomass Dynamics by qPCR

Dynamics of biomass of total bacteria, LAB, Bacillus and fungi by qPCR throughout the SSF process are shown in **Figure 3**. Despite the variation in initial quantities, the levels of total bacteria, LAB, Bacillus and fungi firstly decreased significantly

(P < 0.05) at phase A of thermophilic stages during the SSF process (**Figure 3**). Afterward, their quantities increased significantly (P < 0.05) at phase B of thermophilic stages. Notably, higher levels of total bacteria, LAB, Bacillus and fungi were displayed at the thermophilic stages of the MTSSF process compare with the LTSSF process. Afterward, the microbial quantity displayed decreased and increased trends at cooling stages during MTSSF and LTSSF processes, respectively. The levels of total bacteria, LAB, Bacillus and fungi during the MTSSF and LTSSF processes ranged from 9.40 to 11.24 Log copies g−<sup>1</sup> , 8.6 to 10.30 Log copies g−<sup>1</sup> , 7.58 to 9.90 Log copies g−<sup>1</sup> and 8.69 to 10.10 Log copies g−<sup>1</sup> , respectively. Overall, microbial evolution throughout SSF process is a much more dynamic process, and the quantities of total bacteria, LAB, Bacillus and fungi exhibited similar trends during the MTSSF and LTSSF processes.

# Microbial Composition and Dynamics by Illumina HiSeq Sequencing

Illumina HiSeq sequencing was used to investigate the microbial composition and dynamics during the MTSSF and LTSSF processes. After filtering the low-quality reads and chimeras, for bacterial communities, a total of 633,751 effective tags with an average length of 248 bp from all samples were obtained, and each sample contained 34,386 to 66,238 effective tags with different phylogenetic OTUs ranging from 175 to 485 were generated (Supplementary Table S1). For fungal communities, a total of 714,806 effective tags with an average length of 270 bp from all samples were obtained, and each sample contained 26,404 to 73,805 effective tags with different phylogenetic OTUs ranging from 32 to 134 were generated (Supplementary Table S2), via 97% sequence identity cutoff. Moreover, the rarefaction curves approached the saturation plateau (**Supplementary Figure S1**), which indicated that almost all bacterial and fungal communities could be well represented.

In general, highly similar and successional dynamics of microbial communities were exhibited during the MTSSF and LTSSF processes at order level (**Figures 4A,B**) and genus level (**Supplementary Figure S2**). Groups of Rickettsiales and Streptophyta only dominated the initial 2 days but were retrieved at low frequencies (<1% of total sequences) until the end of SSF process (**Figure 4A**). Meanwhile, thermophilic Eurotiales first became prominent order on day 10 and predominated until the end of MTSSF and LTSSF processes. However, phylotypes of

Enterobacteriales, Lactobacillales, Bacillales, Saccharomycetales, and Mucorales both prevailed the whole MTSSF and LTSSF processes (**Figures 4A,B**). Among these prevailing orders, despite some variation at some moment (**Figures 4A,B**), (i) the relative abundance of Lactobacillales and Saccharomycetales first increased from days 1 to 5, then decreased from days 5 to 10, and finally increased from days 10 to 24; (iii) the relative abundance of Enterobacteriales and Bacillales first increased from days 1 to 5, then decreased from days 5 to 14, and finally increased from days 14 to 24; (iii) the relative abundance of Mucorales first increased from days 1 to 10, then decreased from days 10 to 14, and finally increased from days 14 to 24. Nevertheless, several facts should be highlighted (**Figures 4A,B**): (i) the relative abundance of Enterobacteriales in MTSSF process constituted from 10.30% to 71.73%, while obviously lower in LTSSF process and the percentages ranged from 3.16 to 41.06%; (ii) the relative abundance of Lactobacillales in MTSSF process constituted from 2.34 to 16.68%, while obviously higher in LTSSF process and the percentages ranged from 8.43 to 57.39%; (iii) the relative abundance of Eurotiales in MTSSF process decreased steadily from 36.10 to 28.63% on days 10 to 24, while obviously higher in LTSSF process and the percentages gradual increased from 52.00 to 72.97% on days 10 to 24.

Variations in some small proportions of the top 10 bacterial groups of Xanthomonadales (0.10 to 3.94%), Pseudomonadales (0.17 to 1.85%), Clostridiales (0.02 to 0.98%), Actinomycetales (0.03 to 0.68%), and Burkholderiales (0.03 to 0.57%), and fungal groups of Hypocreales (0.00 to 0.42%), Pleosporales (0.00 to 0.41%), Sordariales (0.00 to 0.25%), and Ascomycota sp. (0.01 to 0.10%) were observed during the MTSSF and LTSSF processes. Other bacterial orders of Rhizobiales (0.01 to 0.14%), Rhodobacterales (0.00 to 0.29%), Bacteroidales (0.03 to 0.22%), Flavobacteriales (0.00 to 0.17%), iii1–15 (0.00 to 0.16%), Desulfobacterales (0.00 to 0.16%), Rhodospirillales (0.00 to 0.14%), and Campylobacterales (0.00 to 0.14%), and other fungal orders of Capnodiales (0.00 to 0.17%), Pezizales (0.00 to 0.07%), and Mortierellales (0.00 to 0.05%) were also observed during the MTSSF and LTSSF processes.

FIGURE 5 | Box and whisker plots of the variance in observed species (A), shannon (B), beta diversity based on weighted unifrac (C) of bacterial communities, and variance in observed species (D), shannon (E), beta diversity based on weighted unifrac (F) of fungal communities, as obtained by Illumina HiSeq sequencing analysis, respectively.

# Statistical Analysis of Microbial Diversity and Richness

Alpha diversity indexes were conducted to evaluate the microbial richness and diversity varied during the MTSSF and LTSSF processes (Supplementary Tables S1 and S2). For bacterial richness and diversity (Supplementary Table S1), the observed OTUs indexes during the MTSSF process generally declined from 261 to 148, in contrast, the observed OTUs indexes during the LTSSF process markedly increased from 147 to 407 during the LTSSF process. Opposite evolution patterns of Chao 1 and ACE values were also observed. However, for fungal richness and diversity (Supplementary Table S2), similar evolution patterns of alpha diversity indexes were observed during the MTSSF and LTSSF processes, for example, the observed OTUs indexes decreased during the MTSSF and LTSSF processes from days 1 to 10 and days 1 to 5, and afterward increased. Notably, obviously lower in bacterial alpha diversity indexes of Shannon, observed OTUs, Chao 1, and ACE indexes during the MTSSF process compare with LTSSF process were observed after days 2, 5, and 10, respectively.

In addition, the box and whisker plots showed that obviously lower (P > 0.05) and significantly lower (P < 0.05) in bacterial alpha diversity indexes of observed OTUs and Shannon indexes were observed during the MTSSF process compare with the LTSSF process, respectively (**Figures 5A,B**). However, commonly higher in fungal alpha diversity indexes were shown during the MTSSF process compare with LTSSF process. Furthermore, the box and whisker plots showed that dramatically higher (P > 0.05) in fungal alpha diversity TABLE 1 | Analysis of similarities (ANOSIM) and MRPP to test for differences in microbial communities of MTSSF and LTSSF processes.


R < 0 means non-significantly differences among the two batches, P > 0.05 means statistically non-significant.

indexes of observed OTUs and Shannon indexes were observed during the MTSSF process compare with the LTSSF process (**Figures 5A,B**).

Principal coordinate analysis was conducted to evaluate the similarities in microbial communities using the weighted UniFrac distance during the MTSSF and LTSSF processes. For bacterial communities, samples from MTSSF process and LTSSF process tended to form clusters (**Figure 4C**), the weighted UniFrac distance in MTSSF process were significantly lower (P < 0.01) than that in LTSSF process (**Figure 5C**). However, for fungal communities, clusters separately tended to form days 1 and 2, days 5, and days 10, 14, 24 both during the MTSSF and LTSSF processes (**Figure 4D**), no significant difference in the weighted UniFrac distance in the MTSSF and LTSSF processes (**Figure 5F**). Moreover, both ANOSIM and MRPP tests indicated that obviously differences (R > 0, P > 0.05) in bacterial communities but highly similarities (R < 0, P > 0.05) in fungal communities were exhibited in the MTSSF and LTSSF processes (**Table 1**).

# Relationships between Environmental Variables and Microbial Communities

The effect of environmental factors on the bacterial (**Figure 6A**) and fungal (**Figure 6B**) communities distribution was evaluated by CCA. Overall, the two axes explained 90.24 and 77.43% of the variation in bacterial and fungal community differentiation, respectively, suggesting the remarkable correlation between microbial structure and environmental factors. Moisture and acidity showed strongly positively correlation with bacterial and fungal composition at the mesophilic stages, but negative correlation with bacterial and fungal composition at cooling and maturation stages (**Figure 6**). However, core temperature (pile temperature) and pH showed negative correlation with bacterial and fungal composition at the mesophilic stages, but positively correlation with bacterial and fungal composition at cooling and maturation stages (**Figure 6**). Furthermore, Daqu samples from days 5 were distributed between days 1 to 2 and days 10 to 24 (**Figure 6**). Pearson's correlation analysis revealed that moisture, acidity and glucose were mainly positively correlated with the relative abundances of Enterobacteriales, Lactobacillales, Streptophyta, Rickettsiales, Bacillales, and Saccharomycetales but negatively correlated with Eurotiales and Mucorales (Supplementary Table S3). However, pH showed contrasting correlations (Supplementary Table S3). Temperature were mainly positively correlated with the relative abundances of Enterobacteriales, Lactobacillales, Bacillales, Saccharomycetales, and Mucorales but negatively correlated with Streptophyta, Rickettsiales, and Eurotiales (Supplementary Table S3).

# DISCUSSION

Traditional Chinese Daqu starters contain highly complex microbial communities, the microbial composition correlates with environmental parameters prevailing during the SSF process. Temperature, moisture, pH, and acidity are traditionally considered the crucial environmental variables used to evaluate Daqu SSF process. Nowadays, increasing attention is likely to be poured to the microbial composition of Daqu correlates with environmental factors prevailing during the fermentation process. This study combined molecular techniques to reveal microbiota dynamics during the MTSSF and LTSSF processes of Daqu starters differing in temperature regimes with consideration for potential relation to environmental variables.

As expected, the thermal profile was followed the quintessential evolution of Daqu SSF process, the temperature quickly rose to thermophilic temperature at thermophilic stages of MTSSF and LTSSF processes (**Figure 1A**), these might derived from heat generation due to the degradation of sugars and proteins by microbial activity (Troy et al., 2012). However, a noteworthy increase of reducing sugar contents was observed at phase A of the thermophilic stage during the LTSSF process (**Figure 1B**), this fact highlighted the intense microbial decomposition activity and the even faster decomposition raw materials to reducing sugar than the utilization of reducing sugar. Similar tends have also been observed by previously reported (Sulaiman et al., 2014). Proteases activity was appropriate indicator of decomposition of organic matter. In this study, the proteases activities and amino acid nitrogen contents slightly increased at the mesophilic stages, and significantly increased at thermophilic stages of MTSSF and LTSSF processes. Furthermore, pH first decreased at the mesophilic stages, and thereafter markedly increased at thermophilic stages of MTSSF and LTSSF processes (**Figure 1C**). This first decline in pH was attributed to the production of organic acids from the breakdown of sugars by the microbes (Liu et al., 2011), and the thereafter increase in pH was attributed to the decomposition of organic acids and to the production of amino acids and

peptides fractions associated with protein degradation in the raw materials (Awasthi et al., 2014). Nevertheless, (i) in case of MT process relatively higher temperature profile was observed than LT process at the mesophilic, thermophilic and cooling stages (**Figure 1A**); (ii) significantly (P < 0.05) higher in pH and lower in titratable acidity were displayed at the mesophilic and thermophilic stages of the MTSSF process compare with the LTSSF process (**Figure 1C**); (iii) significantly (P < 0.05) higher in protease activity and amino acid nitrogen content were displayed on day 14 during MTSSF process compare with the LTSSF process (**Figure 2A**); all these facts demonstrated that the MTSSF process had a higher degradation rates of organic matter than LTSSF process at the mesophilic, thermophilic, and cooling stages, which were coincided with previously reported (Awasthi et al., 2015).

Daqu starters are primary liquefying and saccharifying agents that are used to initiate fermentation in the production of Chinese liquor and vinegar. Amylase and glucoamylase, the main mediators of SSF process, are the major contributors to the liquefaction and saccharification abilities. The amylase activity increased rapidly from days 2 to 5, however, the glucoamylase activity obviously declined throughout the SSF process (**Figure 2**), these can be attributed to the fact that the thermal stability of amylase, which was much better than glucoamylase at high incubation temperature of approximately 60◦C (Soni et al., 2003). Nevertheless, significantly (P < 0.05) higher in glucoamylase activity were noted after the mesophilic stage of the MTSSF process compare with the LTSSF process (**Figure 2**), which was also indicated with its higher temperature profile and degradation rates of organic matter at the mesophilic, thermophilic and cooling stages (**Figure 1**), these results further suggested that the microbial activity in MTSSF process was higher comparatively to the LTSSF process.

Generally, most microorganisms are supposed to disappear and the microbial population was obviously declined when temperature reaches the thermophilic range (Chroni et al., 2009). Consequently, microbial quantities decreased significantly (P < 0.05) at phase A of thermophilic stages during the SSF process (**Figure 3**). Nevertheless, their quantities increased significantly (P < 0.05) at phase B of thermophilic stages, which suggested that some microbes showed the ability to survive as thermotolerant and an intensive thermophilic recolonization could proceed at phase B of thermophilic stages during the SSF process (**Figure 3**). However, higher levels of total bacteria, LAB, Bacillus, and fungi were displayed at the thermophilic stages of the MTSSF process compare with the LTSSF process, this might conditioned by higher temperature and lower titratable acidity during the MTSSF process compare with the LTSSF process. Production of enzyme depends on microbial biomass. Thus, this could provide a reason that the microbial activity in MTSSF process was higher comparatively to the LTSSF process. The microbial quantities showed decreased and increased trends at cooling stages during MTSSF and LTSSF processes, respectively. Theses contradictory tendencies might provide a direct evidence that persistent high temperature might result as deleterious at cooling stage (Insam and De Bertoldi, 2007), in consideration of higher in temperature but no obviously difference in moisture, pH and acidity in MTSSF process than that in LTSSF process (**Figure 1**).

In general, highly similar and successional dynamics of microbial communities were exhibited during the MTSSF and LTSSF processes at order level (**Figures 4A,B**). Phylotypes of Enterobacteriales, Lactobacillales, Bacillales, Saccharomycetales, and Mucorales both predominated the whole MTSSF and LTSSF processes, these members possessed a remarkable capacity to adapt to a wide range of temperatures and moisture levels and were widespread in various Daqu starters and other high-temperature ecosystems (Mouchacca, 2007; Adams et al., 2010; Wang and Xu, 2015). However, groups of Rickettsiales and Streptophyta only dominated the mesophilic stages, these groups were reported as obligate endosymbionts, and low water availability and high temperature could have a deleterious effect on these communities (Botella et al., 2010; Warner et al., 2010). Meanwhile, thermophilic Eurotiales first became prominent order on day 10 and dominated until the end of MTSSF and LTSSF processes. These microbiota dynamics coincided with Fen-Daqu fermentation process (Zheng et al., 2014) and our previously reports (Li et al., 2015), and can be explained by "systematic robustness" principle (Brenner et al., 2008; Freedman and Zak, 2015).

Interestingly, the relative abundance of Enterobacteriales in MTSSF process was obviously higher comparatively to the LTSSF process, but the relative abundance of Lactobacillales and Eurotiales in MTSSF process was obviously lower comparatively to the LTSSF process (**Figure 4B**). These contradictory discrepancies could be due to: (i) the restriction from the higher temperature and lower titratable acidity during the MTSSF process compare with the LTSSF process (**Figure 6**, Supplementary Table S3), for example, the growth of major genus of Thermoascus within Eurotiales (**Supplementary Figure S2**) were restrained when the temperature was >50◦C (Kalogeris et al., 2003); (ii) the difference in initial relative abundance originated from the non-autoclaved raw materials and the open and uncontrolled industrial production environment (Zheng et al., 2012). CCA and Pearson's correlation analyses further clarified that the lower in acidity correlated with the lower in relative abundance of Lactobacillales in MTSSF process (Zheng et al., 2014), and higher in temperature correlated with lower but higher in relative abundance of Eurotiales and Enterobacteriales in MTSSF process, respectively (**Figures 4** and **6**; Supplementary Table S3). These results might provide a direct evidence that persistent high temperature at cooling stage might have a deleterious effect on Eurotiales communities, in consideration of higher in temperature but no obviously difference in moisture, pH and acidity in MTSSF process than that in LTSSF process (**Figure 1**).

Moreover, CCA results revealed that moisture and acidity were the most important factors influencing bacterial and fungal composition at the mesophilic stages (**Figure 6**), however, core temperature (pile temperature) and pH were the most important factors influencing bacterial and fungal composition at cooling and maturation stages (**Figure 6**). In addition, CCA results indicated that Daqu samples from days 5 were distributed between days1 to 2 and days 10 to 24 (**Figure 6**), revealing

that microbial structure transition happened at thermophilic stages under environmental stress of moisture, pH, acidity and pile temperature, significant microbial composition transition was also reported in fermentation pit mud of Chinese strongflavored liquor (Tao et al., 2014; Hu et al., 2016). Meanwhile, our results suggested that lower bacterial richness and diversity but higher fungal richness and diversity were observed during the MTSSF process compare with the LTSSF process (**Figure 5**). Similarly, both PCoA, ANOSIM and MRPP analyses indicated that obviously differences (R > 0, P > 0.05) in bacterial communities but highly similarities (R < 0, P > 0.05) in fungal communities were exhibited in the MTSSF and LTSSF processes (**Figure 4**; **Table 1**). These results indicated that bacterial community and diversity was likely to be more sensitive to environmental variables adjustments than fungal community and diversity. Therefore, it could be more appropriate to consider that a significant proportion of microorganisms growing at thermophilic stages during the SSF process were actually thermotolerant and drought-resistant community, and the microbial communities can adapt to variations in environmental conditions by changes in the thermotolerant and droughtresistant community structure during the SSF process.

# CONCLUSION

There was considerable consistency of the microbial composition during the MTSSF and LTSSF processes. The microbial communities can adapt to variations in environmental conditions by changes in the thermotolerant and drought-resistant community structure during the SSF process. However, different environmental variables affect microbial composition during the MTSSF and LTSSF processes. The microbial activity in MTSSF process was higher comparatively to the LTSSF process. Obviously differences (R > 0, P > 0.05) in bacterial

# REFERENCES


communities but highly similarities (R < 0, P > 0.05) in fungal communities were exhibited in the MTSSF and LTSSF processes. These profound understanding might help to effectively control the traditional Daqu SSF process by adjusting relevant environmental parameters.

# AUTHOR CONTRIBUTIONS

Conceived and designed the experiments: PL, LL. Performed the experiments: PL, XL, XW. Generated and analyzed the data: PL, WL, LL. Wrote the paper: PL.

# ACKNOWLEDGMENT

This project was supported by the National Natural Foundation of China (grant 31271924).

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2016.01237

FIGURE S1 | Rarefaction curves based on the OTUs at the cutoff of 97% 16S rRNA (A) and ITS1 regions (B) of fungal rRNA genes sequence similarity obtained by Illumina HiSeq Sequencing.

FIGURE S2 | Dynamics of relative abundances of the major bacterial (A) and fungal (B) genus during the MTSSF and LTSSF processes, as obtained by Illumina HiSeq sequencing analysis. The abundance was presented as of percentage of total effective bacterial sequences. The abundances of bacterial "other" genera were <0.40%. The abundances of fungal "other" genera were <0.20%. The taxonomy: <sup>∗</sup> , IS–s-Mucorales sp.-Mucorales; ††, Un–s-Hypocreales sp.


communities at elevated CO2. Ecol. Lett. 13, 564–575. doi: 10.1111/j.1461- 0248.2010.01453.x


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

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

fmicb-07-01237 August 2, 2016 Time: 13:18 # 12

# Detection of Thermal Sublethal Injury in Escherichia coli via the Selective Medium Plating Technique: Mechanisms and Improvements

### Laura Espina, Diego García-Gonzalo and Rafael Pagán\*

Departamento de Producción Animal y Ciencia de los Alimentos, Facultad de Veterinaria, Instituto Agroalimentario de Aragón – IA2, CITA-Universidad de Zaragoza, Zaragoza, Spain

#### Edited by:

Avelino Alvarez-Ordóñez, Teagasc Food Research Centre, Ireland

#### Reviewed by:

Gonzalo García De Fernando, Complutense University of Madrid, Spain Hélène Simonin, Agrosup Dijon, France Paula María Periago, Universidad Politécnica de Cartagena, Spain

> \*Correspondence: Rafael Pagán pagan@unizar.es

#### Specialty section:

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

Received: 03 May 2016 Accepted: 19 August 2016 Published: 30 August 2016

#### Citation:

Espina L, García-Gonzalo D and Pagán R (2016) Detection of Thermal Sublethal Injury in Escherichia coli via the Selective Medium Plating Technique: Mechanisms and Improvements. Front. Microbiol. 7:1376. doi: 10.3389/fmicb.2016.01376 In food preservation, the synergistic combination of different technologies aims to maximize the total lethality of the process and minimize the intensity of each hurdle. This is especially the case when at least one of the treatments can cause sublethal (reparable) injury in a great proportion of the population, so that sublethally injured cells can end up being entirely inactivated by the other hurdle(s). The selective medium plating technique (SMPT) is extensively used to enumerate bacterial sublethal injury after inimical treatments, being sodium chloride added to the recovery medium to detect damaged bacterial envelopes. However, little work has been done to explain the reasons for the inability of sublethally injured cells to outgrow in selective agar media, whereas they are able to grow in non-selective agar. In the present paper, the performance of SMPT on Escherichia coli cells after heat treatments is explored by applying different selective agents in the recovery media, using mutants lacking factors involved in osmoregulation, and also by examining the integrity of the cytoplasmic membrane. In view of the results, the possibility of a specific toxic effect of Na<sup>+</sup> as the main mechanism under SMPT was discarded, since the same level of sublethal injury was detected using KCl instead of NaCl. The synthesis of the osmoprotectant trehalose determined the maximum osmotolerance of intact cells to the selective agents, but was not crucial in the quantification of sublethal injury. Moreover, for the first time, the extent of sublethal injury detected via SMPT was directly correlated with the physical loss of integrity of the cell membrane in 99.999% of the initial population. This was achieved through statistical analysis of flow cytometry data using propidium iodide-exclusion technique when that dye was added before thermal treatments. The present work confirms the adequacy of SMPT as a tool for detecting the occurrence and quantity of sublethally injured cells after thermal treatments and thus, for efficiently designing the combination of heat with other preservation techniques. We also propose the study of statistical analysis from flow cytometry data for a more rapid quantification of bacterial sublethal injury in a broad detection range.

Keywords: sublethal injury, osmoregulation, selective media, Escherichia coli, flow cytometry

# INTRODUCTION

fmicb-07-01376 August 27, 2016 Time: 12:1 # 2

In bacteriology, viability has been traditionally defined and measured as the ability of organisms to self-replicate in culture media (Bogosian and Bourneuf, 2001; Nyström, 2001). However, it has long been known that the failure of a bacterial cell to produce a colony on a standard nutrient plate may not necessarily mean that the cell was dead at the time of sampling (Nyström, 2001). For instance, microorganisms that are metabolically active despite their inability to grow in laboratory culture media are said to be in a "viable but non-culturable" (VBNC) state, which, under harsh environmental conditions, can be triggered as a survival mechanism (Bogosian and Bourneuf, 2001). On other occasions, exposure to chemical or physical processes can lead to the sublethal injury of bacterial cells: this state is considered to be transient, since cells are able to repair their damages and resume growth if suitable environmental conditions emerge (Mackey, 2000).

In food preservation it has been demonstrated that, once one has applied preservation treatments to control bacterial food contamination, a considerable proportion of the population may become sublethally injured in addition to both the surviving (non-injured) and the inactivated populations (Wesche et al., 2009). The adequate identification and quantification of the sublethally injured population plays a key role in food safety. Since damaged cells are not generally able to grow on the conventional selective enrichment media used in the food industry (Restaino et al., 2001), they can remain undetected, subsequently repair their damages and reach infective concentrations (Mackey, 2000). On the other hand, according to the "hurdle effect" (Leistner and Gorris, 1995), repair of sublethally injured cells after a preservation treatment can be adequately prevented by the combination of additional preservation agents (hurdles) that interfere with cellular homeostasis maintenance, thereby synergistically increasing the combined process's global lethality (Mackey, 2000).

In this regard, although new methods are being developed for the detection of sublethally injured bacteria (Back et al., 2012; Gelaw et al., 2014), the most widely used strategy among microbiologists is still differential enumeration on nonselective and selective agar, following the so-called selective medium plating technique (SMPT; Mackey, 2000). For this purpose, out of all possible selective agents, subinhibitory concentrations of sodium chloride have been consistently incorporated in the recovery medium (Chilton et al., 2001; Ulmer et al., 2002; Miller et al., 2006). It is believed that the increase in osmotic pressure caused by the addition of sodium chloride explains the selective outgrowth of only those cells whose cytoplasmic membrane remains intact (Mackey, 2000).

Despite the observed selective effect of the osmolyte NaCl on bacterial growth, little research has been done to study the osmoregulatory mechanisms of sublethally injured cells and, therefore, to find out more about their ability to maintain selective permeability after different stresses. In intact bacteria, an osmotic upshock unleashes a cascade of events intended to maintain turgor pressure within limits by regulating the total osmotic solute pool in the cytoplasm (and in the periplasm in Gram-negative bacteria; Wood, 2011). As the osmolality of the surrounding environment increases, turgor pressure drops and growth slows or halts (Wood, 2011). The most rapid response to this osmotic upshock is an increase in potassium ion influx that increases cytosolic osmolality (Wood, 2011). Since high intracellular concentrations of K<sup>+</sup> interfere with many important cellular functions, the cell starts to accumulate large quantities of so-called compatible solutes, which are more congruous with its physiology (Wood et al., 2001). The compatible solute trehalose is synthesized (via Ots system) and accumulated up to levels that may comprise as much as 20% of cytoplasmic osmolality under conditions of high osmolality (Wood, 1999). Other compounds, when present externally (such as glycine betaine), can be incorporated via transporters such as BetT or ProP (Lucht and Bremer, 1994; Wood, 2011), leading to a decrease in trehalose levels and stimulating bacterial growth rates under hyperosmotic conditions. It has been estimated that after 1 h of osmotic stress, a cell's physiology and structure are largely restored via these osmoregulatory systems (Wood, 1999). A better knowledge of the interaction between cellular osmoregulatory mechanisms and the permeabilization of the cytoplasmic membrane and how they influence bacterial ability to outgrow in selective media could facilitate the estimation of sublethal injury and, therefore, help us improve the design of food preservation processes.

In the present study, SMPT is applied as the primary technique to detect and quantify the proportion of sublethally injured cells in their cytoplasmic membrane after exposure to a lethal stress. Thermal treatment was selected as the lethal stress, since it is the most studied and best understood treatment known to sublethally injure microorganisms (Wesche et al., 2009). It should be noted that mild thermal treatments applied in fluid environments have been demonstrated to disturb the permeability of the outer membrane earlier and more intensely than the permeability of the cytoplasmic membrane (Mackey, 2000; Shigapova, 2004); thus, the outer membrane does not interfere with the detection of sublethal injury in the cytoplasmic membrane. The microorganism Escherichia coli was also selected, since it is the model microorganism for studying bacterial osmoregulation (Shabala et al., 2009). Besides, the availability of a great variety of E. coli mutants lacking factors involved in the osmoregulatory system (Baba et al., 2006) can be used to determine those factors' role in SMPT.

The primary objective of this study was (i) to gain a better understanding of the mechanisms underlying SMPT by trying to identify which bacterial osmoregulatory mechanisms or physical structures are modified by heat and are thus responsible for the prevention of bacterial growth in selective media. Additionally, we aimed (ii) to improve traditional SMPT by testing the effect of different variations in the composition of the recovery media, and also (iii) to explore the possible use of flow cytometry as a complementary technique to assess sublethal injury.

# MATERIALS AND METHODS

# Preparation of Media

fmicb-07-01376 August 27, 2016 Time: 12:1 # 3

Minimal medium M9 was chosen as the broth and treatment medium, since it is commonly used for the culture of E. coli (Neidhardt et al., 1974), and because its minimal composition reduces the presence of osmolytes or osmoprotectants influencing the osmoregulation processes. M9 minimal broth was prepared following the steps indicated in Maniatis et al. (1982): its composition is of 38 mM Na2HPO4, 20 mM KH2PO4, 7.7 mM NaCl, 17 mM NH4Cl, 1 mM MgSO4, 0.1 mM CaCl2, and 0.2% glucose.

Regarding the recovery media, both minimal and rich agar plates were prepared to cover a whole range of culture conditions, as both types are commonly used in the study of sublethal injury (Wesche et al., 2009). In addition to the ingredients in M9 minimal broth, the M9 minimal agar medium contained 15 g/L of Agar Technical No. 3 (Oxoid, Basingstoke, UK).

Tryptic soy agar (Biolife, Milan, Italy) plus 0.6% of yeast extract (Biolife; TSAYE) was selected as the rich recovery medium, given its widespread use in the enumeration of bacterial injury (Miller et al., 2006; Noriega et al., 2013). Preliminary experiments showed that recovery in M9 minimal agar medium after different thermal treatments yielded similar counts than in TSAYE (data not shown).

Although, NaCl is the solute most commonly used to inhibit growth in selective agar media when evaluating sublethal injury in the cytoplasmic membrane, we also tested the osmolytes KCl and saccharose. With the objective of determining the influence of the type of osmolyte in the detection of sublethal injury, each solute was added in the concentration required to achieve the same osmolality values in the agar medium. For this purpose, the osmolality values of the agar (Os/kg of M9 agar medium) were chosen to correspond with those created by the addition of 1–6% of NaCl, and resulted in a range of 0.34–2.05 Os/kg of agar medium. The KCl and saccharose concentrations required to achieve such osmolality values were 1.27–7.68% KCl or 11.63–70.17% saccharose.

Betaine was added as osmoprotectant at 1 mM, following the lines of previous research (Le Rudulier et al., 1984; McLaggan et al., 2002). Higher concentrations were not proven more effective to osmotically protect cells (data not shown).

# Micro-Organisms and Growth Conditions

The strains used were E. coli BW25113 and its deleterious mutants E. coli 1otsA, 1proP, 1nhaA, 1nhaB, and 1nhaR. While factors OtsA and ProP are involved in the synthesis of trehalose and the uptake of betaine respectively, the different subunits of the factor Nha are involved in the excretion of Na+. All strains were obtained from the KEIO collection (Baba et al., 2006).

The cultures were maintained in cryovials at −80◦C prior to use. Broth subcultures were prepared by inoculating one single colony from a plate in a 50-mL flask containing 10 mL of sterile M9 minimal medium. After inoculation, the flasks were incubated overnight at 37◦C. With these subcultures, 250 mL Erlenmeyer flasks containing 50 mL of M9 medium were inoculated into a final concentration of 3 × 10<sup>6</sup> CFU/mL. These flasks were incubated with agitation (130 rpm; Selecta, mod. Rotabit, Barcelona, Spain) at 37◦C until stationary growth phase was reached (24 h).

# Thermal Treatments

Before inoculation, cultures were centrifuged at 6000 × g for 5 min and resuspended in the treatment medium (M9 medium).

For the preparation of heat-treated samples, 0.1 mL of culture at 10<sup>9</sup> CFU/mL was added to a tube containing 0.9 mL of M9 medium tempered at 55 ± 0.2◦C or at 53, 57, or 59 ± 0.2◦C (FX Incubator, A. F. Ingeniería S. L., Valencia, Spain). The actual temperature was controlled with a thermocouple wire introduced in a 0.9 mL M9 broth test tube inside the incubator. After each individual treatment interval, samples were taken, immediately placed on ice, and adequately diluted in 0.1% w/v peptone water (Biolife). Survivors were evaluated as explained below.

Exceptionally for an experiment aimed to compare inactivation kinetics in the absence and presence of the osmoprotectant betaine, survival curves to heat treatments were obtained in a specially designed thermo-resistometer, as previously described (Condón et al., 1993). This device has a thermocouple (Pt 100) to monitor the temperature during heat treatment and for the injection of inoculum. Once the temperature had stabilized (at 58, 61, 64, 67, or 70◦C), 0.2 mL of culture was injected via a solenoid-valve-operated automatic syringe into the 400-mL treatment chamber containing the treatment medium under constant agitation. Samples were taken at regular intervals and survivors were evaluated as explained below.

# Collection of Samples, Counts of Culturable Cells and Quantification of Sublethally Injured Cells

In order to quantify bacterial cell injury, in a first step the maximum non-inhibitory concentration (MNIC) of each osmolyte in M9 agar medium was determined. To achieve this, untreated cells were spread plated onto M9 agar media with different concentrations of each solute (NaCl, KCl, or saccharose), and plates were incubated at 37◦C for 48 h. According to previous work (Cebrián et al., 2014), the MNIC was defined as the highest concentration which inhibited less than 20% of the initial untreated bacterial population.

After treatments, 0.02 mL volumes of adequately diluted samples (using M9 broth as the dilution medium) were spread on the surface of prepared M9 agar and/or TSA plates, in both nonselective and selective plates. Exceptionally, samples treated with the thermo-resistometer were poured either directly on plates (for treatment temperatures of 58, 61, and 64◦C) or were pourplated after having been collected in agar-medium-containing tubes placed on a rotating carousel (for experiments performed at 67 and 70◦C). This sample-collection device allowed for the characterization of survival curves despite the high inactivation rates at these treatment temperatures.

In all cases, plates were incubated at 37◦C for 48 h. Previous experiments showed that longer incubation times did not influence the amount of surviving cells regardless of the added osmolyte. For each dilution, 10–200 colonies were counted on the surface of the agar medium in spread-plated samples. For pourplated samples, colonies were counted with an improved Image Analyzer Automatic Counter (Protos; Analytical Measuring Systems, Cambridge, UK) as described in earlier work (Condón et al., 1993). Taking into account the initial cell concentration in the thermoresistance experiments (10<sup>8</sup> CFU/mL), the detection limit was of 5 log<sup>10</sup> cycles.

Inactivation was expressed in terms of the extent of reduction in log<sup>10</sup> counts (CFU) after any treatment. Survival curves were obtained by plotting the decimal log<sup>10</sup> fraction of survivors versus the treatment time for each independent experiment. The extent of sublethal injury was expressed as the difference between the log<sup>10</sup> count (CFU) on nonselective medium (M9) and the log<sup>10</sup> count on selective media. Likewise, the percentage of injured cells at each treatment time corresponded to the following equation (Busch and Donnelly, 1992):

$$\% \text{Injured cells} = 1 - \left(\frac{\text{CFU}/mL\_{\text{selective}}}{\text{CFU}/mL\_{\text{nonselective}}} \times 100\right) \tag{1}$$

According to this representation, "2 log<sup>10</sup> cycles of injured cells" means a 2-log<sup>10</sup> difference in the count on selective and non-selective media, or that 99% of survivors were sublethally injured.

Experimental data were obtained from at least three independent experiments performed on different days.

## Thermotolerance Parameters

When appropriate, survival curves were fitted by a model based on a Weibull-like distribution, which was chosen based on their linear and concave upward profiles. For this investigation we used the equation proposed by Mafart et al. (2002) (Eq. 1):

$$L \log\_{10} \frac{N\_l}{N\_0} = -\left(\frac{t}{\delta}\right)^{\rho} \tag{2}$$

where t is the treatment time (min); N<sup>t</sup> and N<sup>0</sup> are the population densities (CFU/mL) at time t and time 0, respectively; and δ and ρ are two characteristic parameters of the equation. The δ value is called the time to the first decimal reduction (time necessary to inactivate the first 1 log<sup>10</sup> CFU of the microbial population). The ρ value is the shape parameter.

# Determination of the State of Cells Grown in Agar Media Containing NaCl

To determine the state of cells (viable, inhibited, or inactivated) when grown in agar media with different concentrations of NaCl, wild type (WT) or 1proP untreated or heat-treated cells (10 min at 55◦C) were carefully sampled onto plates with M9 agar medium added with 0–10% NaCl. The initial sampled concentration of cells was 5 × 10<sup>6</sup> CFU/plate, and plates were incubated for 48 h. After that first incubation, a method was developed to recover colonies from colony-lacking plates in a highly reproducible way. For this, 4 g of agar of each plate from the first incubation were carefully extracted, placed in sterile plastic bags with peptone water 0.1%, and homogenized for 20 s at 230 rpm in a stomacher laboratory blender (model 400, Tekmar, Co., Cincinnati, OH, USA). Next, 1 mL-aliquots were spread plated onto non-selective M9 agar plates and incubated once more for 48 h. After that second incubation, the surface of the plates was visually inspected and classified into positive growth (presenting a high enumerable concentration of CFU/plate) or negative growth (with less than 5 CFU/plate).

For each degree of NaCl concentration in the first agar medium, the state of intact or heat-treated cells was classified as viable (when colonies were observed after the first incubation at the expected concentration of 4 × 106–5 × 10<sup>6</sup> CFU/plate), inhibited (colony-lacking plates after the first incubation but with positive growth after the second incubation) or inactivated (colony-lacking plates after the first incubation and with negative growth after the second incubation). Data shown are results from a representative experiment repeated twice with similar results.

# Measurement of Cell Permeabilization via Propidium Iodide (PI) Uptake

For the evaluation of cell permeabilization, PI at a concentration of 0.08 mM (Pagán and Mackey, 2000) was added to the treatment medium prior to the thermal treatment. Alternatively, PI was not added before treatments and was incorporated immediately after each treatment in order to obtain additional information. Cell permeabilization was analyzed by fluorescence microscopy and by flow cytometry.

For the analysis under the fluorescence microscope, treatments were applied at 55◦C for 0–5 min. For the flow cytometry analysis, the treatment temperature was 53◦C in order to achieve longer intervals between samples. After each treatment, samples were immediately placed on ice, subsequently incubated for 15 min at 20◦C, centrifuged at 6000 × g for 5 min, and washed three times. For the flow cytometry analysis, samples were also immediately fixated with a preparation of 4% paraformaldehyde in PBS, washed three times and diluted to a concentration of 10<sup>5</sup> CFU/mL in PBS.

The measurement of cell permeabilization with the fluorescence microscope (Nikon, Mod. L-Kc, Nippon Kogaku KK, Japan) was performed by direct counting of non-fluorescent and fluorescent bacteria at 1000× magnification. About 200 bacteria were visible in a field of vision, and bacteria from five fields of vision were counted per sample and replicate.

For each sample analyzed by flow cytometry, 10,000 events were counted using a MACSQuant Analyzer (Miltenyi Biotec, Cologne, Germany) flow cytometer. Fluorescence data were collected using the 488 nm excitation laser and the 614–650 nm filter, corresponding to the B2 channel in the MACSQuant Analyzer.

The evaluation of PI uptake by each of those two techniques was run in triplicate on separate days.

# Statistical Analyses and Management of Flow Cytometry Data

For kinetics analysis of the data from survival curves, the leastsquares criterion of the GraphPad PRISM program (GraphPad Software, San Diego, CA, USA) was used. This program was also used to perform ANOVA and t-test; differences were considered significant if p ≤ 0.05.

Data from flow cytometry was analyzed with FCS Express 5 (De Novo Software, Los Angeles, CA, USA). For the measurement of fluorescence intensity, the parameter "area under the curve" was chosen over "pulse height" in order to consider not only the maximum fluorescence of each event, but also the time required to collect data. No gates were created to obtain histograms or statistical data thereof. Before running the actual samples, unstained and stained cells were analyzed in the flow cytometer to establish the adequate threshold levels for the identification of "events" as "cells" and for the sensitivity of the fluorescence signals.

# RESULTS AND DISCUSSION

The SMPT allows for the estimation of the occurrence of sublethal injury after each treatment by measuring the difference between the inactivation level achieved in a selective medium and the inactivation level achieved in a non-selective medium (Mackey, 2000). In order to assess the damage in the cytoplasmic membrane, sodium chloride is added at its MNIC, so that only non-damaged cells are able to multiply.

In the present study we primarily intended to offer a simple example of the performance of SMPT after thermal treatments on E. coli. Cells were recovered in M9 agar with 1, 2, o 3% NaCl. Concentrations over 3% NaCl (MNIC) in the agar inhibited the growth of untreated cells. The results, depicted in **Figure 1**, show that after 10 min of treatment less than 0.2 log<sup>10</sup> cycles of the initial population failed to grow in non-selective agar medium. However, when recovered in agar medium containing 1, 2, or 3% NaCl, the population of cells unable to grow increased in 0.2, 1.6, or 4.8 log<sup>10</sup> cycles respectively. This graph demonstrates that even a very mild thermal treatment can result in an increased sensitivity to NaCl in the agar media in the majority of the initial bacterial population, corresponding to sublethally injured cells.

On the other hand, there was a gradual inverse relationship between the osmolality of the recovery medium and the proportion of growing cells. Therefore, the more severely injured cells are, the lower the NaCl concentration required to prevent their growth – which fits perfectly with the previously stated hypothesis of the coexistence of different levels of damage, from minor to eventually lethal (Wesche et al., 2009; Noriega et al., 2013).

# Insights into the Failure of Sublethally Injured Cells to Grow on Osmotically Selective Media

The increased sensitivity of cells to NaCl after thermal treatments does not have a clear origin, although it has been traditionally

ascribed to the loss of permeability control, leading to their irreversible inactivation (Mackey, 2000). However, little research has been done to identify mechanisms or structures that are damaged by heat and therefore prevent bacterial growth in the presence of osmotically selective agents. As different factors could be involved, in the present study we decided to investigate the mechanisms underlying SMPT by individually considering (i) the osmoregulatory mechanisms aimed to upregulate the solute pool, (ii) the possible toxicity of the selective agent in the agar media, and (iii) the selective permeability of the cytoplasmic membrane.

# Role of the Upregulation of the Solute Pool in SMPT

The synthesis and accumulation of trehalose, or the influx of other osmoprotectants when present in the media, are the result of a cascade of osmoregulatory events triggered in living bacterial cells by osmotic upshocks and intended to maintain their correct turgor pressure (Wood, 2011).

In the present work we explored the osmoregulatory response of WT and mutant cells impaired in trehalose synthesis or in the influx of osmoprotectants, with the objective of determining the involvement of those osmoregulatory mechanisms in SMPT.

## **Upregulation of the solute pool through the accumulation of trehalose**

High osmolarity stimulates the transcription of Ots system to synthesize trehalose in media devoid of osmoprotectants, and mutants impaired in otsA are osmotically sensitive due to their inability to synthetize trehalose (Lucht and Bremer, 1994). For the present work, we decided to compare the state of untreated or thermally treated WT cells with that of 1otsA cells when plated onto agar with different NaCl concentrations. Also, the proportions of sublethally injured cells were calculated, for each strain and treatment time, by calculating the difference between

the survival level in the presence of its MNIC of NaCl and in the absence of NaCl.

**Table 1** shows that, as expected, E. coli 1otsA presented a lower NaCl MNIC value (2%) than the WT; the fact that untreated 1otsA cells are unable to grow in agar medium with 3% NaCl is probably due to the absence of the osmoprotectant effect of accumulated trehalose. Furthermore, the reduced osmotolerance of 1otsA cells was also detected in the finding that NaCl concentrations above 8% were capable of inactivating untreated cells (instead of only inhibiting their growth, as observed for the WT cells).

The application of a prior thermal treatment resulted in the inactivation of otherwise inhibited cells when plated with 6–10% NaCl (**Table 1**). Therefore, we were able to confirm that thermally treated cells of both strains lost their ability to survive in media containing high NaCl concentrations. This observation could be related to the increase in the intracellular accumulation of Na<sup>+</sup> in cells when plated onto agar with an external osmolality of 2 Os/kg, corresponding to 6% NaCl (Shabala et al., 2009). When considering the proportion of sublethal injury at their respective MNICs, both strains behaved similarly (2,5 ± 0,5 and more than 5 log cycles of sublethal injury after 5 and 20 min of heat treatment respectively, data not shown). The higher osmosensitivity of the mutant lacking the complete trehalose synthesis pathway in comparison with the WT exposes the relevance of trehalose synthesis in SMPT. This finding also agrees with a previously observed reduction in MNIC values of several osmolytes in E. coli mutants in the Ots-controlling sigma factor RpoS (Cebrián et al., 2015).

Regarding the specific role of trehalose synthesis or accumulation in the detection of sublethal injury by SMPT, the similar proportions of sublethal injury detected in both strains seem to suggest that, once cells have been thermally damaged, other mechanisms or cellular structures are responsible for their difficulty to outgrow in selective agar media. Further research should be done on the thermosensitivity of Ots as a key factor in the way trehalose and its synthesis pathway are involved in the inhibition and inactivation of sublethally injured cells.

TABLE 1 | State of untreated or thermally treated cells after the incubation in M9 agar medium added with each NaCl concentration.


Additionally, an unexpected discovery was made in the results in **Table 1**. Whereas untreated WT cells were inhibited when grown in the presence of concentrations above the MNIC, thermally treated cells remained inhibited when recovered onto agar medium with 3% NaCl, which corresponds to their MNIC and therefore is commonly used to determine the degree of sublethal injury (García et al., 2005). These results contradict, for the first time, the previously accepted hypothesis that sublethally injured cells are inactivated when plated at the MNIC determined for untreated cells (Mackey, 2000): the explanation is that the cells are not being actually inactivated but inhibited in hyperosmotic agar media. For simplicity, throughout the present study we continue to use the term "inactivation" to describe the lack of growth in the recovery medium. On the other hand, this discovery can turn out to be of great relevance, from an applicative point of view, in helping us correctly interpret the lethality of each treatment in combined preservation processes. This is especially true when low water activity is considered as one of the hurdles: since inhibited cells can resume growth under favorable conditions, the error of considering them as inactivated cells would imply that one would underestimate the bacterial content in food and thereby incur in possible health risks for the consumers.

### **Upregulation of the solute pool through the influx of external osmoprotectants**

Nutritionally rich agars containing osmoprotectants are commonly used for the detection of sublethal injury in food preservation (Wu, 2008; Wesche et al., 2009). Bacteria take up osmoprotectants from surrounding media via membrane transporters such as ProP or BetT (Haardt et al., 1995; Wood et al., 2001), and their stability could be impaired after thermal treatments and therefore influence the outcome of SMPT. Among osmoprotectants, betaine has been demonstrated to increase growth of E. coli cells in hyperosmotic media (Le Rudulier et al., 1984), so E. coli mutants lacking the betaine transporter ProP (1proP) were selected to help determine the role of osmoprotectant transporters in SMPT.

The addition of betaine to the recovery agar medium resulted in an increase in the NaCl MNIC value from 3 to 5% in untreated WT cells, as represented in **Figure 2A** by comparing the black bars (showing inactivation of more than 80% of the initial cell population when plated in the presence of 4% NaCl) with blue bars (showing that inactivation of more than 80% of the initial cell population is only achieved when plated in the presence of 6% NaCl). When WT cells were treated at 55◦C for 10 min, a significantly higher proportion of cells were recovered at each % NaCl when recovered in media with betaine, than when betaine was absent (**Figure 2B**; p < 0.05). In contrast, **Figure 2B** shows that thermally treated mutants lacking ProP were unable to incorporate betaine: the proportion of growing cells was the same (p > 0.05) regardless of the osmoprotectant. Therefore, ProP was still active after thermal treatment (as indicated by the difference between treated WT and 1proP cells when recovered in the presence of betaine), while other transporter(s) possibly responsible for the uptake of betaine in untreated 1proP cells were inactive after the thermal treatment.

The observation of the remaining activity of ProP after heat prompted us to attempt to ascertain whether more intense thermal treatments could impair ProP and therefore interfere with the detection of sublethal injury. For this purpose, the thermosensitivity of ProP was analyzed by culturing thermally treated WT cells in agar medium containing 3% NaCl, with and without betaine added. Survival curves were modelized so that the time required to inactivate 1 log<sup>10</sup> cycle of the initial cell population could be compared between the two treatments. According to the results (**Figure 3**), no statistically significant differences were observed between the slopes of the two TDT curves (p > 0.05). This implies that the osmoprotectant effect of betaine was maintained throughout the whole range of assayed temperatures (57–70◦C), demonstrating the functionality of ProP in the assayed conditions. As a consequence, the possibility that cells might be unable to incorporate osmoprotectants in order to repair their sublethal injury was generally discarded for thermal treatments at temperatures up to 70◦C.

On the other hand, the evident influence of added betaine on the osmoregulatory response of E. coli and in the results obtained with SMPT using M9 agar medium showed the relevance of the composition of the recovery medium in the interpretation of the sublethally injured fraction via SMPT. Moreover, previous results have demonstrated that the presence of betaine in the recovery medium can compensate for defective phenotypes in

their osmoregulatory systems (Cebrián et al., 2015). However, the evaluation of the occurrence of sublethal injury incurred in E. coli after inimical treatments via SMPT usually employs complete and nutritionally rich agar media (such as tryptic soy yeast extract agar [TSAYE] or plate count agar) as the non-selective medium (Wuytack et al., 2002; Noriega et al., 2013). In contrast with the controlled and osmoprotectant-free composition of the M9 agar medium, TSAYE contains the osmoprotectant betaine (Dulaney et al., 1968). In our experiments, the presence of betaine in TSAYE was demonstrated by the determination of a MNIC of NaCl at 5% (as in M9 agar medium with betaine), and by the repetition of the treatments applied to obtain **Figure 2** but with TSAYE as the recovery medium: both E. coli WT and 1proP behaved in TSAYE similarly as in M9 agar medium with betaine (p > 0.05; data not shown).

In order to control the adequacy of the SMPT using TSAYE as the recovery medium, several thermal treatments of different durations and at different temperatures causing less than 0.5 log<sup>10</sup> cycles of inactivation in non-selective M9 agar medium were applied to E. coli WT cells. The number of log<sup>10</sup> cycles of inactivation was measured in M9 agar medium, M9 agar medium with betaine and TSAYE, having added their respective MNIC of NaCl (3, 5, and 5%; **Figure 4**). The good correlations obtained between the measurements in the M9 + 3% NaCl agar medium with those in the M9 + 5% NaCl + betaine agar (R <sup>2</sup> > 0.95) or with those in the TSAYE + 5% NaCl agar (R <sup>2</sup> > 0.92) suggest that, despite the presence of osmoprotectants like betaine, the estimation of the amount of sublethal injury remains constant because of the corresponding increase in the MNIC. Therefore, these results lead to the conclusion that the selection of TSAYE, a recovery agar with osmoprotectants, does not lead to an underestimation of the proportion of sublethally injured cells.

**Figures 3** and **4**, when read together, can lead to a further conclusion. It is easily understandable that only those cells

which take up betaine are able to outgrow in agar medium with 5% NaCl, since that osmoprotectant prevents them from being inhibited or inactivated at concentrations above 3% NaCl. However, considering that the applied thermal treatments do not affect ProP (**Figure 3**), the ability to introduce betaine in the cytoplasm is not a limiting factor for heat-treated cells to grow. This would mean that only cells with functional osmoregulatory mechanisms (without considering osmoprotectants) continue to grow after thermal treatments. Given the good correlation between the inactivation detected in treated cells growing in the presence of 5% NaCl with betaine and those growing in the presence of 3% NaCl, it was demonstrated that the latter selective medium is correctly preventing the growth of those cells whose osmoregulation is not completely functional, as previously assumed (Mackey, 2000).

# Possible Toxicity of Na<sup>+</sup> in the Cell

Sublethal injury to microbial cell membranes caused by inimical treatments has been linked to the cell's ability to exclude toxic materials (Gilbert, 1984). Sodium can be considered one of those toxic materials, since E. coli cells have to maintain an intracellular Na<sup>+</sup> concentration lower than the extracellular concentration via the active extrusion systems NhaA and NhaB, regulated by NhaR (Padan and Krulwich, 2000). Moreover, Na<sup>+</sup> stress is enhanced under conditions in which membrane integrity is compromised, and it has been suggested that in E. coli high osmolarity may lead to the induction of specific Na<sup>+</sup> efflux pathways (Padan and Krulwich, 2000).

In order to investigate the possible toxic effect of the presence of Na<sup>+</sup> in the selective recovery medium, we studied the MNICs of different solutes on untreated cells, as well as their effect on the survival kinetics of thermally treated cells. This way, equivalent osmotic values were achieved in the agar media by incorporation of different concentrations of the ionic osmolytes Na<sup>+</sup> or K<sup>+</sup> (as NaCl or KCl), or the non-ionic osmolyte saccharose. In order to facilitate the comparison among osmolytes, the level of inactivation achieved after a very mild thermal treatment (which caused no inactivation in non-selective agar) was obtained in the presence of 25, 50, 75, and 100% MNIC of each osmolyte.

The MNIC values of NaCl and KCl were obtained at the same osmolality value (1.02 Os/kg), while the MNIC of saccharose was determined at a greater osmolality value (1.70 Os/kg). In this regard, it has been observed that ionic and non-ionic osmotica trigger different osmoregulatory responses (Shabala et al., 2009). However, this distinction did not seem relevant in SMPT, since similar levels of thermal inactivation (p > 0.05) were detected in the presence of the MNIC or lower concentrations of either NaCl, KCl or saccharose (**Figure 5**). Besides, mutants lacking the Na<sup>+</sup> extrusion systems NhaA, NhaB, or NhaR showed the same MNIC of NaCl than E. coli WT cells (3%; data not shown). All these observations suggest that thermal treatments impairing the Na<sup>+</sup> efflux systems could be dismissed as one of the factors intervening in the detection of the sublethal injury under the conditions assayed. This hypothesis agrees with previous observations by Cebrián et al. (2014), who concluded that no specific inhibition mechanisms could be attributed to the ionic osmolytes NaCl or KCl other than the same hyperosmotic stress as imposed by saccharose.

In our attempt to update SMPT, we also noted that not only the MNIC of NaCl and KCl were obtained at the same osmolality value, but also similar E. coli survival curves after thermal treatments of 5, 10, or 20 min were obtained for each osmolality value (data not shown). Therefore, although NaCl is commonly used as the selective agent at its MNIC in SMPT (García et al., 2006; Miller et al., 2006; Arroyo et al., 2009), its substitution with KCl could be an alternative possibility.

#### Impairment of Cytoplasmic Membrane Integrity

After the exploration of specific osmoregulatory mechanisms triggered by osmotic upshocks, research into the reason for the inability of sublethally injured cells to outgrow in selective agar medium in SMPT should further explore the physical integrity of the cytoplasmic membrane. Not only is membrane integrity considered to be a key for the maintenance of osmoregulation (Wood et al., 2001), but the inability of cells to overcome the action of the selective agent is considered to reveal structural damage in the cytoplasmic membrane (Wesche et al., 2009). However, little research has been carried out to prove the relationship between the extent of sublethal injury and the physical integrity of the cytoplasmic membrane.

For the study of membrane integrity, measurement of its degree of permeabilization with the membrane-impermeant dye propidium iodide (PI) has been extensively used to quantify cell damage by penetrating membranes with pores larger than 660 Da (Pagán and Mackey, 2000; García et al., 2005; Kennedy et al., 2011). **Figure 6** shows the correlation between the percentage of permeabilized cells and the level of inactivation measured when thermally treated cells were recovered in agar media with NaCl or KCl at their MNIC. The percentage of permeabilization corresponds to the fraction of cells in which PI had entered through membrane pores during treatment, while the "level of inactivation" factor refers to the percentage of cells – out of the total initial sample population – which were unable to outgrow in the selective agar media (comprising both dead and sublethally

injured cells). The total percentage of inactivation – determined by the proportion of cells unable to outgrow in non-selective agar medium – was under 5% even after the longer treatment times of 3 and 5 min (data not shown).

The good correlation between both factors (no significant differences between their slopes, p > 0.05), confirms the hypothesis that damage is due to the impairment of membrane permeability (Mackey, 2000; Wesche et al., 2009). **Figure 6** also demonstrates that, independently from the functionality of the osmoregulatory mechanisms, a direct relationship exists between the extent of sublethal injury detected via SMPT and the physical state of the cytoplasmic membrane. It is noteworthy that this good correlation was obtained when PI had been added before the treatment, corresponding to the creation of pores throughout the whole treatment (Pagán and Mackey, 2000). In contrast, the incorporation of PI immediately after each thermal treatment required more than 5 min of thermal treatment in order to lead to the permeabilization of the majority of cells as evidenced by microscopy, and the resulting staining intensity measured by flow cytometry was much lower at any treatment time than when PI was added beforehand (data not shown). This would agree with previous observations which determined that 20 min treatments at 60◦C were unable to permeabilize more than 80% of the E. coli population via staining with post-treatment PI (Shigapova, 2004; Kennedy et al., 2011). Furthermore, these results agree with the view that PI is a sensitive marker of cell damage, but a poor indicator of cell death (Amor et al., 2002).

# Exploration of the Possible Use of Flow Cytometry as a Complementary Technique to Assess Sublethal Injury

Counting the number of PI-positive cells in terms of percentage only allows for the accurate evaluation of ca. 1 log<sup>10</sup> cycle of the initial population, as opposed to the 5-log<sup>10</sup> scale obtained in the

previous results by viable plate count. In an attempt to overcome this major limitation, we decided to use flow cytometry as a more appropriate methodology to assess membrane permeabilization through PI uptake. Not only is flow cytometry highly convenient in view of this goal (Amor et al., 2002; Kennedy et al., 2011), but there is also increasing interest in its possible use to complement or even substitute plate count techniques (Nebe-von-Caron et al., 2000; Aronsson et al., 2005).

In view of this objective, flow cytometer data of PI-stained samples were subjected to statistical analysis and complemented with the measurement of sublethal injury via SMPT. Each overlay in the histogram of **Figure 7** depicts the frequency distribution of the fluorescence intensity's logarithmic value for a specific treatment time. The position of the overlays alongside the x-axis shows a clear tendency toward increasing fluorescence with longer treatments. Given this observation and the Gaussian aspect of the histogram overlays, we decided to obtain, for each treatment time, a simple parameter characterizing the average fluorescence intensity. For this purpose, the median value of the total of fluorescence-area values of all events was calculated and divided by the maximum average fluorescence achieved by any sample for that assay (**Figure 8**). The inactivation levels detected in M9 agar medium containing 3% NaCl for different treatment times were also plotted (**Figure 8**).

Firstly, we wanted to distinguish between the proportion of stained cells and the total fluorescence value for each sample. In order to calculate the proportion of stained cells, in the data grid obtained through flow cytometry we selected and counted only those events that surpassed the fluorescence threshold. Results showed that the proportion of stained cells increased in parallel with treatment duration for the first 5 min (data

Frontiers in Microbiology | www.frontiersin.org August 2016 | Volume 7 | Article 1376 |

between both factors, should be performed in order to establish a reference data matrix for further studies. In addition, simultaneous staining with other fluorochromes could provide a better description of the composition of each bacterial sample (Nebe-von-Caron et al., 2000), and therefore help us understand the evolution of treated cells from the viable to dead conditions. From a practical point of view, the rapid detection of the extent of sublethal injury via flow cytometry (and not only the extent of inactivation, as commonly performed) could significantly help in the design of food preservation processes by determining which treatment conditions could be more favorable in the synergistic combination of different hurdles.

# Conclusion on the Evidence of Sublethal Injury through SMPT

According to the results, in SMPT only cells with intact osmoregulatory properties can overcome the osmotic pressure in the selective agar medium (**Figures 3** and **4**). In contrast, those which are considered sublethally injured remain inhibited at the MNIC of the selective agent (**Table 1**). Therefore, cells whose osmoregulatory mechanisms or physical structures become nonfunctional after thermal treatments are unable to outgrow in osmotically challenging agar media, although they can outgrow in the absence of the selective agent. The identification of such mechanisms or structures, as well as their relationship with the extent of sublethal injury detected, are a key in understanding the performance of SMPT.

In the present study, two of the hypothesized osmoregulatory mechanisms have been discarded as key factors in the performance of SMPT in detecting sublethal injury after heat in E. coli: the exclusion of Na<sup>+</sup> from the cytoplasm and the uptake of osmoprotectants from the agar media. The toxicity of Na<sup>+</sup> as a cause of sublethal injury had been previously proposed (Gilbert, 1984; Padan and Krulwich, 2000), but we have found evidence

not shown). At this point, about 90% of the cells were already stained independently of flourescence intensity (data not shown). Therefore, we confirmed the previously observed correlation (assessed by optical microscopy) between the proportion of permeabilized cells and the proportion of cells unable to grow on the selective medium, since after 4 min of treatment around 85% of cells (nearly 1 log<sup>10</sup> cycle) had been unable to outgrow in the selective agar medium.

In contrast with the proportion of fluorescent cells, the median value of fluorescence intensities refers to the total fluorescence emitted by the entire bacterial population of each sample, i.e., it assesses the number of PI-stained cells and the fluorescence emitted by each of them. At this point, it should be noted that different staining intensities often occur (Shi et al., 2007), as we have observed from previous experiments using the fluorescence microscope. As can be seen in **Figure 8**, the median values followed a linear evolution throughout treatment time until reaching a maximum value when treatments were 20 min or longer. Therefore, although nearly all the cells had been stained after 5 min of treatment, they were mostly weakly stained, and the fluorescence intensity of the whole population went on increasing throughout treatment at a constant rate. The linearity in the average fluorescence intensity of different samples is a promising concept that has been barely approached. In this regard, Berney et al. (2007) correlated the geometric mean of fluorescence intensity with the amount of nucleic acids, but research could be expanded to different fluorescent probes in order to reveal different grades of a high variety of metabolic processes.

A new methodology for the determination of the occurrence of sublethal injury at a broad detection range (at least 5 log<sup>10</sup> cycles, depending on sample size) could be developed following the results depicted in **Figure 8**. The meticulous determination of cell plate counts and fluorescence measurements after inimical treatments, as well as calculations of the correlation

fmicb-07-01376 August 27, 2016 Time: 12:1 # 10

neither of Na<sup>+</sup> toxicity, nor of thermal treatments affecting the Na<sup>+</sup> extrusion systems. Regarding osmoprotectants, their uptake is absolutely necessary for cells to outgrow in rich media added with NaCl at its MNIC (**Figure 4**). Since the transporter ProP remained active after intense thermal treatments (**Figure 3**), the inability to uptake betaine should not be hypothesized as the reason for the non-survival of sublethally injured cells in selective agar media.

In the absence of osmoprotectants, the main osmoregulatory mechanisms accumulate trehalose. Its absence leads to an increased osmosensitivity and thermosensitivity in untreated and treated cells, and impairs the correct quantification of sublethal injury via SMPT (**Table 1**). However, no direct relationship between the impairment of trehalose synthesis and accumulation systems and the extent of sublethal injury could be established. In contrast, we found a direct relationship between the structural damage of the cell membrane and SMPT via the PI-exclusion technique when PI was added before thermal treatments. In this way, the extent of sublethal injury detected via SMPT could be ascribed to the physical loss of integrity of the cell membrane independently of specific functional osmoregulatory processes. The detection of sublethal injury of E. coli after thermal stress has been previously ascribed to the physical loss of integrity of the cell membrane (Mackey, 2000; Ukuku et al., 2008; Wesche et al., 2009). However, to the best of our knowledge, this is the first time that a direct correlation between both factors has been demonstrated, especially at such a high proportion of the bacterial population.

Furthermore, some of the results of the present study can result in the improvement of SMPT. For instance, variations in the composition of the selective media without affecting the outcome of the technique are being proposed: M9 agar + 3% NaCl, M9 agar + 3.88% KCl, M9 agar + betaine + 5% NaCl, and TSAYE + 5% NaCl yielded similar levels of sublethal injury. On the other hand, the possibility of complementing SMPT with flow cytometry to detect bacterial inactivation and injury at a detection range of 5 log<sup>10</sup> cycles is presented here, since the extent of cell permeabilization (measured simply and rapidly thanks to flow cytometry) was found to be an indicator of the extent of sublethal injury detected with SMPT.

# REFERENCES


# CONCLUSION

This work demonstrates, for the first time, that the incorporation in the recovery agar of selective agents which increase its osmotic pressure (such as sodium chloride or potassium chloride) inhibits the growth of E. coli cells whose envelopes are physically impaired by mild thermal treatments. Previous hypotheses regarding the implication of two different factors on the performance of SMPT (the toxicity of Na<sup>+</sup> in the agar and the destruction of transporters of osmoprotectans) were discarded. Moreover, the extent of this physical damage was found to be correlated with the proportion of treated cells unable to grow in selective agar, confirming the adequacy of the SMPT to assess thermal sublethal injury. Further investigation aimed to improve the performance of the SMPT or its combination with flow cytometry could help to maximize its usefulness in food preservation.

# AUTHOR CONTRIBUTIONS

Conceived and designed the experiments: LE, DG-G, and RP. Performed the experiments: LE. Analyzed the data: LE, DG-G, and RP. Wrote the paper: LE, DG-G, and RP.

# FUNDING

This work was supported by the Spanish Ministerio de Economía y Competitividad (CICYT Projects AGL2012–32165 and AGL2015–69565–P).

# ACKNOWLEDGMENTS

Spanish Ministerio de Educación, Cultura y Deporte that provided LE with a grant to carry out this investigation. The authors would like to thank Dr. Santiago Condón for his assistance during the research. The authors also wish to thank Stanley Hanks (translator) for having revised and proofread the final version of this manuscript.

This manuscript is dedicated to the loving memory of Bernard Mackey.



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

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

# Quantification of Growth of Campylobacter and Extended Spectrum β-Lactamase Producing Bacteria Sheds Light on Black Box of Enrichment Procedures

#### Wilma C. Hazeleger<sup>1</sup> \*, Wilma F. Jacobs-Reitsma<sup>2</sup> and Heidy M. W. den Besten<sup>1</sup>

<sup>1</sup> Laboratory of Food Microbiology, Wageningen University, Wageningen, Netherlands, <sup>2</sup> National Institute for Public Health and the Environment, Bilthoven, Netherlands

Campylobacter is well recognized as the leading cause of bacterial foodborne diarrheal disease worldwide, and is routinely found in meat originating from poultry, sheep, pigs, and cattle. Effective monitoring of Campylobacter contamination is dependent on the availability of reliable detection methods. The method of the International Organization for Standardization for the detection of Campylobacter spp. in food (ISO 10272-1:2006) recommends the use of Bolton broth (BB) as selective enrichment medium, including a pre-enrichment step of 4–6 h at 37◦C to revive sublethally damaged cells prior to incubation for 2 days at 41.5◦C. Recently the presence of abundantly growing extended spectrum β-lactamase producing Enterobacteriaceae (ESBL bacteria) has become one of the most important factors that interfere with the isolation of Campylobacter, resulting in false-negative detection. However, detailed growth dynamics of Campylobacter and its competitors remain unclear, where these would provide a solid base for further improvement of the enrichment procedure for Campylobacter. Other enrichment broths, such as Preston broth (PB) and BB plus clavulanic acid (BBc) have been suggested to inhibit competitive flora. Therefore, these different broths were used as enrichments to measure the growth kinetics of several strains of Campylobacter jejuni and ESBL bacteria separately, in co-culture and of strains in chicken samples. The maximum cell numbers and often the growth rates of Campylobacter in mixed culture with ESBL bacteria were significantly lower than in single cultures, indicating severe suppression of Campylobacter by ESBL bacteria, also in naturally contaminated samples. PB and BBc successfully diminished ESBL bacteria and might therefore be a better choice as enrichment medium in possibly ESBL-bacteria contaminated samples. The efficacy of a pre-enrichment step in the BB ISO-procedure was not supported for cold-stressed and non-stressed cells. Therefore, omission of this step (4–6 h at 37◦C) might be advised to obtain a less troublesome protocol.

Keywords: ESBL, pre-enrichment, ISO 10272-1, Bolton broth, Preston broth, clavulanic acid, competition, inhibition

#### Edited by:

Avelino Alvarez-Ordóñez, Teagasc Food Research Centre, Ireland

#### Reviewed by:

M. Luisa De Garnica, University of León, Spain Nabila Haddad, Oniris, France Inge Van Damme, Ghent University, Belgium

\*Correspondence:

Wilma C. Hazeleger wilma.hazeleger@wur.nl

#### Specialty section:

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

Received: 26 April 2016 Accepted: 29 August 2016 Published: 12 September 2016

#### Citation:

Hazeleger WC, Jacobs-Reitsma WF and den Besten HMW (2016) Quantification of Growth of Campylobacter and Extended Spectrum β-Lactamase Producing Bacteria Sheds Light on Black Box of Enrichment Procedures. Front. Microbiol. 7:1430. doi: 10.3389/fmicb.2016.01430

**Abbreviations:** BB, Bolton broth; BBc, Bolton broth supplemented with clavulanic acid; BHI, brain heart infusion broth; CAB, Columbia agar base; CFA, Campyfood agar plates; CFU, colony forming units; ESBLs, extended spectrum β-lactamase producing bacteria; HI, heart infusion broth; ISO, International Organization for Standardization; mCCDA, modified charcoal-cefoperazone-deoxycholate agar; PB, Preston broth; RCA, Rapid' Campylobacter agar plates.

# INTRODUCTION

fmicb-07-01430 September 8, 2016 Time: 12:29 # 2

Campylobacteriosis is the most commonly reported zoonosis in the European Union showing an increasing trend over the period of 2008–2014, and the occurrence of thermotolerant Campylobacter in broiler meat remains high at 38.4% in 2014 (European Food Safety Authority [EFSA] and European Centre for Disease Prevention and Control [ECDC], 2015). Concomitantly, broilers are often contaminated with extended spectrum β-lactamase producing Enterobacteriaceae (ESBL bacteria; Bortolaia et al., 2010; Depoorter et al., 2012; Dierikx et al., 2013; Kawamura et al., 2014; Olsen et al., 2014). The presence of these ESBL bacteria in food has become an important factor interfering with the isolation of Campylobacter, resulting in false-negative detection, since abundantly growing ESBL bacteria hamper the isolation of Campylobacter colonies (Jasson et al., 2009).

The protocol of the International Organization for Standardization (ISO) for detection of thermotolerant Campylobacter spp. in food and animal feeding stuffs (ISO, 2006), describes the use of Bolton broth (BB) which is mixed 10:1 with the food sample including a 4–6 h pre-enrichment step at 37◦C to resuscitate sublethally damaged cells before further enrichment is done at 41.5◦C for 2 days. After enrichment, campylobacters are isolated on modified charcoal-cefoperazonedeoxycholate agar (mCCDA) and a second selective medium, with a principle different from mCCDA. The antibiotics in BB and mCCDA do not inhibit the growth of ESBL bacteria (Jasson et al., 2009), therefore the selectivity of the media is diminished. Nonetheless, detailed growth dynamics of Campylobacter and its competitors during enrichment remain unclear, and these would provide a good starting point for developing a proper enrichment procedure for Campylobacter. Furthermore, strong scientific evidence for better isolation of the pathogen using a pre-enrichment step at 37◦C is scarce (Humphrey, 1986, 1989).

Therefore, in this study, growth kinetics of Campylobacter and ESBL bacteria were determined during the enrichment procedure in BB and also in previously suggested alternative enrichment broths, such as Preston broth (PB; Scotter et al., 1993; Uyttendaele and Debevere, 1996) and BB plus clavulanic acid (BBc; Moran et al., 2011). Single strains and mixed cultures of Campylobacter and ESBL bacteria were tested and also naturally contaminated samples were examined. To create sublethally damaged cells, naturally contaminated chicken samples and inoculated chicken samples were cooled and/or frozen previously to the enrichment procedures to determine the effect of the pre-enrichment step at 37◦C on the recovery of Campylobacter.

# MATERIALS AND METHODS

# Bacterial Strains and Preparation of Stationary Phase Cultures

Campylobacter jejuni ATCC 33560 (= NCTC 11351), which is indicated as suitable control strain (ISO, 2006), C. jejuni LU 160891 (Wageningen University; isolate from chicken filet), Campylobacter coli WCDM 00004, a strain advised for performance testing (ISO, 2015), and Escherichia coli ESBL strains RIVM 2 and RIVM 3 (National Institute for Public Health and the Environment; isolates from chicken filet) were used as single cultures and as Campylobacter and ESBL–E. coli mixed cultures. Campylobacter stock cultures were grown in Heart Infusion broth (HI, Becton Dickinson) for 48 h at 37◦C, then supplemented with 15% glycerol and stored at −80◦C. E. coli stocks were cultured in Brain Heart Infusion broth (BHI, Becton Dickinson) for 24 h at 37◦C, then supplemented with 15% glycerol and stored at −80◦C as well. To obtain precultures for the growth experiments, C. jejuni was plated from the −80◦C vials onto Columbia agar base (CAB, Oxoid, supplemented with 5% (v/v) lysed sheep blood (BioTrading Benelux B.V. Mijdrecht, Netherlands)) and grown for 48 h at 37◦C, whereas ESBL E. coli was plated onto BHI agar or tryptone soya agar (TSA, Oxoid) and grown for 24 h at 37◦C. Subsequently, single colonies were resuspended in HI and BHI for C. jejuni and E. coli, respectively, and cultured at 37◦C for, respectively, 48 and 24 h to obtain stationary phase cultures. Cell concentrations were determined by plating appropriate dilutions on CAB for Campylobacter and on TSA for ESBL E. coli. Campylobacter was cultured under micro-aerobic conditions (5% O2, 10% CO2, 85% N2) in flushed jars (Anoxomat WS9000, Mart Microbiology, Drachten, Netherlands) unless stated otherwise.

# Pretreatment of Chicken Samples

To determine the effect of a pre-enrichment step for 4–6 h at 37◦C on the growth of sublethally damaged cells, inoculated chicken samples were stored at 4◦C or −20◦C, to mimic the situation in practice where chicken samples are purchased in cooled, respectively, frozen state. For that, chicken skin samples (5 g, confirmed previously to be Campylobacter-free, using ISO 10271-1 (ISO, 2015), were kindly provided by Johan Roelofs, Plukon B.V., Wezep, Netherlands). The samples were stored at −20◦C, thawed before use at 21◦C for a maximum of 30 min and then inoculated with about 0.5 mL of diluted 48 h cultures [10<sup>2</sup> to 10<sup>4</sup> colony forming units (CFU) per 5 g chicken skin sample] of C. jejuni or C. coli and/or 24 h cultures of ESBL E. coli and subsequently stored in stomacher bags (Antonides, Oosterzee, Netherlands) for 60 h at 4◦C or −20◦C. Choices of inoculation levels were aiming for similar starting levels of Campylobacter and ESBL bacteria after the cooling or freezing treatment which were determined in separate experiments, where the reduction in cell numbers on chicken stored at 4◦C and −20◦C was quantified (data not shown). Frozen samples were thawed for 30 min at 21◦C and cooled samples were allowed to reach room temperature for 5 min before chicken juice was prepared (see Preparation of Chicken Juice) and then enrichment procedures were started.

# Preparation of Chicken Juice

To be able to take regular samples in time without disturbing the micro-aerobic conditions, growth curves were made in infusion bottles (100 mL), sealed with a thick (1 cm) rubber stopper and secured by an aluminum cap. Chicken juice of inoculated chicken skin (n = 12), was prepared to allow adding of chicken product to the broth using a syringe and at the same time to recover as

much of the present micro flora as possible to meet the ISOprocedure (ISO, 2006). Chicken skin juices were made by adding chicken skin at a 1:1 ratio to peptone physiological salt [PPS; 0.9% NaCl and 0.1 % peptone (Oxoid)] in a filter stomacher bag. The bag was massaged by hand for 2 min, and further homogenized for 30 s in a Pulsifier 100E (Microgen Bioproducts, Camberley, UK). Juice from the inoculated chicken skin was then immediately used in the enrichment procedures (see Measuring Growth Dynamics).

To examine naturally contaminated chicken liver, chicken juices were prepared similarly to the inoculated chicken skin as described above. For chicken wings, the same procedure was followed, except the samples were mixed in PPS at a 2:1 ratio. To confirm and quantify presence of Campylobacter and ESBL bacteria, 1 mL of juice was spread onto three Campyfood agar plates (CFA, bioMérieux) or Rapid' Campylobacter agar plates (RCA, Bio-Rad) and onto Brilliance ESBL plates (Oxoid), respectively. Campylobacter plates were incubated for 48 h at 41.5◦C (micro-aerobic conditions) after which confirmation was done microscopically and using a Latex agglutination test for Campylobacter (Microgen Bioproducts). Brilliance ESBL plates were incubated for 24 h at 37◦C. All chicken juices were stored for 2–3 days in 50 mL Greiner tubes at 4◦C until the results of the plating showed presence of Campylobacter and ESBL bacteria and these juices (n = 26) were then directly used in the enrichment procedures (see Measuring Growth Dynamics). For ESBL bacteria all colonies were counted on the Brilliance ESBL plates and no distinction was made between E. coli or other ESBL containing bacteria.

# Measuring Growth Dynamics

Infusion bottles were filled with 45 mL of enrichment medium. After that, either 5 mL of diluted stationary phase cultures (to mimic a C. jejuni concentration of 2–3 log CFU g−<sup>1</sup> chicken (European Food Safety Authority [EFSA], 2011), resulting in a starting concentration in the enrichment broth of about 1–2 log<sup>10</sup> CFU mL−<sup>1</sup> ), or 5 mL of chicken juice were added and the head space was flushed for 2 min with a gas-mixture of 5% O2, 10% CO2, and 85% N<sup>2</sup> by a home-made gasflushing device using syringes to puncture the rubber stopper of the bottles. BB (Oxoid) with selective supplement (Oxoid SR0208E) and 5% of sterile lysed horse blood (Oxoid) or sheep blood (Biotrading) was used as enrichment medium. Also alternative enrichment broths were used; to obtain BBc, 2 mg L−<sup>1</sup> (end concentration) potassium clavulanate (Sigma-Aldrich) was added to BB (Moran et al., 2011). PB was prepared as Nutrient Broth No. 2 (Oxoid) with Preston Campylobacter Selective Supplement (Oxoid SR0204), Campylobacter Growth Supplement (Oxoid), and 5% of lysed horse blood. Inoculated infusion bottles were incubated in water baths set at 37◦C (4–6 h) and 41.5◦C (up to 48 h).

At regular time intervals, 1 mL samples were taken from the bottles using a syringe and after every second sample, bottles were flushed again with the appropriate gas. Samples were immediately diluted and plated onto CFA, RCA, or mCCDA (Oxoid, supplemented with Oxoid SR155E) for counting Campylobacter and onto Brilliance ESBL agar for ESBL bacteria, and incubated as described in Section "Preparation of Chicken Juice." In order to prevent swarming of Campylobacter colonies, all plates for enumeration were dried in a 41.5◦C incubator for 15–20 min before plating. At least two biologically independent reproductions per strain or strain combination were performed on different days in all enrichment media.

# Data Analysis

Plate counts were transformed to log<sup>10</sup> CFU mL−<sup>1</sup> and growth curves were constructed using Microsoft Excel 2010 and counts were fitted with the modified Gompertz Model (Zwietering et al., 1990) using the Solver add-in of Excel. The analysis was verified using TableCurve 2D V5.01 and a t-test was used to examine statistical significance in the growth parameters λ (lag phase; h) and µ (maximum growth rate; log<sup>10</sup> h −1 ) of the microorganisms at the different conditions (P < 0.05).

# RESULTS

# Single and Mixed Cultures

Growth of single C. jejuni strains in BB showed that using a start inoculum of 2 log<sup>10</sup> CFU mL−<sup>1</sup> , the maximum level (8 log<sup>10</sup> CFU mL−<sup>1</sup> ) was reached after about 20 h (**Figure 1A**). ESBL E. coli showed higher growth rate, and reached stationary growth phase already after 10–12 h of incubation (**Figure 1A**). When both strains were cultured together, growth of ESBL E. coli was comparable to single culture conditions. C. jejuni, however, showed severe growth reduction in mixed culture with ESBL E. coli (**Figure 1B**). When ESBL E. coli reached the stationary phase, or just before that, growth of C. jejuni ceased and remained at a maximum level of 4–6 log<sup>10</sup> CFU mL−<sup>1</sup> . Due to the limited number of data points in the beginning of these curves, lag phases could not be determined accurately but seemed to be non-significant in most cases. In general, growth rates of C. jejuni in BBc (**Figures 2A,B**) and PB (**Figures 2C,D**) were similar to growth in BB (P > 0.05), although some lag time was observed in BBc. However, BBc and PB successfully inhibited growth of ESBL E. coli, both in single cultures (data not shown) and in co-cultures with C. jejuni (**Figures 2B,D**). Similar results were found for all tested C. jejuni and ESBL E. coli strains. Since no difference was observed in data with and without preenrichment at 37◦C, only graphs without pre-enrichment are shown (see Effect of Pre-enrichment) for the results described in **Figures 1–3**.

# Naturally Contaminated Samples

Growth characteristics of Campylobacter and ESBL bacteria from most naturally contaminated samples show similar trends as mixed cultures (**Figure 3**), even though the naturally contaminated samples were stored refrigerated before enrichment. In BB, Campylobacter was mostly outcompeted (**Figure 3A**) by ESBL bacteria (n = 23). However, in 12% of the cases (n = 3), the pathogen was able to grow to similar levels

FIGURE 1 | Growth of Campylobacter jejuni is severely inhibited when cocultured with ESBL E. coli. Representative growth curves (n = 2) of C. jejuni ATCC 33560 (blue circles) and ESBL E. coli RIVM 2 (purple squares) in Bolton broth (48 h 41.5◦C) when cultured as single cultures (A) and in co-cultures (B). Strains were precultured to stationary phase and subsequently inoculated in enrichment broth. The lines are curves fitted with the modified Gompertz model (Zwietering et al., 1990). Detection limit is 1 CFU mL−<sup>1</sup> .

exhibits good growth in all media. Representative growth curves (n = 2) of C. jejuni ATCC 33560 as single cultures in Bolton broth (48 h 41.5◦C; blue circles, A,C), in Bolton broth plus clavulanic acid (orange circles, A), and in Preston broth (green circles, C). Representative growth curves (n = 2) of mixed cultures of C. jejuni ATCC 33560 (orange, respectively, green) and ESBL E. coli RIVM 2 (purple squares) in Bolton broth plus clavulanic acid (B) and Preston broth (D). Strains were precultured to stationary phase and subsequently inoculated in the respective enrichment broths. The lines are curves fitted with the modified Gompertz model (Zwietering et al., 1990). Detection limit is 1 CFU mL−<sup>1</sup> .

as ESBLs (**Figure 3B**). The alternative enrichment broths PB (n = 7) and BBc (n = 11) always completely inhibited growth of ESBL bacteria (**Figures 3C,D**), comparable to the situation in laboratory strains (**Figure 2**).

# Effect of Pre-enrichment

No significant difference (P > 0.05) was observed in growth kinetics of C. jejuni or ESBL E. coli grown with and without a pre-enrichment incubation step for 4–6 h at 37◦C, in single

cultures or mixed cultures in BB (**Figures 4A,B**). For C. jejuni no effect of pre-enrichment was found in BBc (**Figure 4C**) or PB (**Figure 4D**) either. Similar results were found for all tested Campylobacter and ESBL E. coli strains. Growth characteristics of the cold-stressed bacteria from naturally contaminated chicken samples showed similar results for BB, where no effect of the pre-enrichment step at 37◦C was observed either (**Figure 5A**). Cooling (60 h at 4◦C) and freezing (60 h at −20◦C) of artificially contaminated chicken neck skin samples resulted in, respectively, a 0.3 log<sup>10</sup> and 1.5 log<sup>10</sup> average reduction in Campylobacter numbers (calculated in CFU mL−<sup>1</sup> BB). ESBL bacteria were not reduced (max 0.1 log<sup>10</sup> reduction) at 4◦C and 0.1–0.3 log<sup>10</sup> reduced after freezing. After applying these stress conditions, a pre-enrichment step of 6 h at 37◦C did not lead to better growth of C. jejuni, C. coli or ESBL bacteria compared to enrichment starting immediately at 41.5◦C for cold stored samples or frozen samples (**Figures 5B,C**, respectively).

# DISCUSSION

Selectivity of BB is based on the addition of four antibiotics; vancomycin, trimethoprim, amphotericin B, and cefoperazone, where the latter two are also used in the isolation plate (mCCDA) in the ISO-protocol for detection of Campylobacter (ISO, 2006). Cefoperazone is an antibiotic belonging to the group of third generation cephalosporins, and the β-lactam ring in this antibiotic is easily hydrolyzed by ESBL-containing organisms, rendering them insensitive to the selective compound (Chong et al., 2011). This fact, in combination with the increasing numbers of ESBL bacteria in chicken products (Costa et al., 2010; Overdevest et al., 2011; Dierikx et al., 2013), results in reduced isolation efficacy of Campylobacter, due to overgrowth of ESBL bacteria (Jasson et al., 2009). In the present study, the proliferation during enrichment was followed for both Campylobacter and ESBL bacteria. ESBL bacteria showed significantly higher growth rates in BB than the target organism (P < 0.05). In mixed cultures with Campylobacter, this resulted also in higher maximum cell numbers for ESBL bacteria, where the ESBL bacteria probably profit from their higher growth rate and the growth of Campylobacter ceased just before, or when the ESBL bacteria entered the stationary phase in growth. The observed differences in maximum cell numbers in mixed growth cultures, showed a 2–4 log<sup>10</sup> CFU reduction of the target organism compared to single cultures. This clearly explains the difficulties in recognizing Campylobacter colonies on an mCCDA plate if 100- to 10,000-fold ESBL bacterial colonies are present, keeping in mind that growth of ESBL bacteria on mCCDA is not

clavulanic acid, and Preston broth. Representative growth curves (n = 2) of C. jejuni ATCC 33560 (circles and triangles) in Bolton broth (A,B), Bolton broth plus clavulanic acid (C), and Preston broth (D) as single culture (A,C,D) with (4 h 37◦C, 44 h 41.5◦C; red triangles) and without (48 h 41.5◦C; circles) pre-enrichment step. Growth of ESBL E. coli RIVM 2 (squares) with (4 h 37◦C, 44 h 41.5◦C; red squares) and without (48 h 41.5◦C; purple squares) pre-enrichment step in Bolton broth as single culture (A) and in mixed culture (B) with C. jejuni ATCC 33560. Strains were precultured to stationary phase and subsequently inoculated in the respective enrichment broths. Detection limit is 1 CFU mL−<sup>1</sup> .

hindered either (data not shown). In a limited number of the cases, however, Campylobacter did grow to similar levels as ESBL bacteria, apparently these strains were not too much affected by ESBL bacteria, which can be partly explained by a very low initial number of ESBL bacteria in some of the naturally contaminated samples or ESBL bacteria with different growth characteristics. The ratio in numbers of Campylobacter and ESBL bacteria and species or type of both microorganisms may also be contributing factors to the growth dynamics during enrichment procedures. In this study, starting levels of around 10–100 CFU mL−<sup>1</sup> were chosen in the enrichment broths, to aim for realistic levels reported in chicken of about 10–50 CFU g−<sup>1</sup> product for ESBL bacteria (Cohen Stuart et al., 2012) and 10–1000 CFU g−<sup>1</sup> for Campylobacter (European Food Safety Authority [EFSA], 2011).

To reduce growth of ESBL bacteria, β-lactamase inhibitors have been suggested (Payne et al., 1994), for instance tazobactam in the isolation medium mCCDA, showing good repression of ESBL bacteria (Smith et al., 2015). To increase selectivity of the enrichment in BB, addition of potassium clavulanate, another β-lactamase inhibitor, was suggested and examined (Moran et al., 2011; Chon et al., 2013a). In the current study, potassium clavulanate indeed inhibited the growth of ESBL bacteria in single and mixed cultures and also in naturally contaminated samples. The growth was not only inhibited, numbers of ESBL bacteria were even reduced to below the detection limit mostly within 4 h showing that selectivity of the medium was efficiently restored.

As alternative enrichment broth, PB has previously been described providing good selectivity against non-target flora in the enrichment procedure of Campylobacter (Bolton and Robertson, 1982; Uyttendaele and Debevere, 1996; Jasson et al., 2009; Habib et al., 2011; Ugarte-Ruiz et al., 2012) with selective components polymyxin B, rifampicin, trimethoprim, and cycloheximide/amphotericin B. Polymyxin B is probably the component that inhibits the ESBL bacteria since it has been shown to be active against most Gram-negative bacteria (Bolton and Robertson, 1982). Chon et al. (2013b) used polymyxin B in enrichment broth with cefoperazone, and restored selectivity in that way. Some studies have, however, shown that PB may inhibit growth of some Campylobacter strains as well, resulting in false negative outcomes (Baylis et al., 2000; Paulsen et al., 2005), especially for C. coli (Goossens et al., 1986). In this study, growth curves in PB showed good growth of Campylobacter for both

C. jejuni lab strains and for naturally occurring campylobacters on chicken; although in some cases the growth rate or maximum cell number were lower compared to growth in BB and BBc but this was not significant (P > 0.05). However, the medium successfully reduced growth of ESBL bacteria in all cases and their numbers dipped below the detection limit within 1–2 h. Levels of Campylobacter are yet high enough (6–9.5 log CFU mL−<sup>1</sup> ) to produce colonies on the isolation medium if 10 µL is streaked.

Currently, the ISO-protocol for detection of Campylobacter is revised (ISO, 2015) and in this protocol, a distinction is made between different food samples, where PB is advised for samples in which high background flora such as ESBL bacteria is suspected. BB is still recommended for samples with low numbers of non-target organisms and low numbers of potentially stressed or sublethally damaged Campylobacter. The revision of the ISO-protocol is supported by the results described in this paper and using the new protocol will probably improve Campylobacter detection from food samples, provided that labs choose the correct enrichment broths for their specific samples. Furthermore, the advice to use two isolation media with different selective principles will also reduce the risk of overgrowth of Campylobacter by non-target flora in at least one of the agars thereby lowering the number of false negative detection results.

Thermotolerant Campylobacter, such as C. jejuni, cannot grow below 30◦C, however, metabolic activity in the cells has been described even at 4◦C, showing that not all processes in the cells have stopped (Hazeleger et al., 1998) and cooling and freezing are considered to be stressful for Campylobacter in broth (Ray and Johnson, 1984; Jasson et al., 2007) and in raw milk and river water (Humphrey, 1986). Therefore, subsequent growth in detection procedures may be impeded. To overcome cold stress, a preenrichment step of 2–4 h at 37◦C was described by Humphrey (1986, 1989) in water, milk, and poultry samples to resuscitate sublethally damaged cells at a less selective temperature. In this paper, cooling and freezing, being common factors in the production chain of poultry, were used to induce sublethal damage in the naturally present or inoculated Campylobacter on chicken before the advised BB enrichment procedure was started. However, the results do not support the efficacy of a pre-enrichment step, using BB at 37◦C prior to incubation at 41.5◦C as recommended in the ISO-protocol (ISO, 2006). The different findings may be explained by the use of different enrichment broths and a more selective temperature of 43◦C used by Humphrey (1986, 1989). The isolation temperature for Campylobacter has since been reduced to 41.5◦C for practical reasons, to match with enrichment temperatures for detection of Salmonella and E. coli O157 (Corry and Atabay, 2012). Taking also into account that changing the incubation temperature from 37 to 41.5◦C after 4–6 h of pre-enrichment is troublesome in routine laboratory practice, omission of the pre-enrichment step in the current procedure of enrichment in BB could be considered.

# AUTHOR CONTRIBUTIONS

WH, WJ-R, and HB contributed to conception and design. WH and HB contributed to acquisition, analysis, and interpretation of data. WH, WJ-R, and HB drafted and/or revised the article.

## ACKNOWLEDGMENTS

The authors thank Maren Lanzl, Paula P. Bonilla, Jing Zang, Ourania Brella, and Jiayun Lu for excellent lab work.

# REFERENCES

fmicb-07-01430 September 8, 2016 Time: 12:29 # 8


chicken meat. Food Microbiol. 28, 1117–1123. doi: 10.1016/j.fm.2011. 03.001


detection and isolation of thermophilic Campylobacter from different matrices. J. Appl. Microbiol. 113, 200–208. doi: 10.1111/j.1365-2672.2012. 05323.x


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

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

# Application of Impedance Microbiology for Evaluating Potential Acidifying Performances of Starter Lactic Acid Bacteria to Employ in Milk Transformation

Elena Bancalari <sup>1</sup> , Valentina Bernini 1, 2, Benedetta Bottari 1, 2, Erasmo Neviani 1, 2 and Monica Gatti 1, 2 \*

*<sup>1</sup> Laboratory of Food Microbiology, Department of Food Science, University of Parma, Parma, Italy, <sup>2</sup> Multidisciplinary Interdepartmental Dairy Cente, University of Parma, Parma, Italy*

### Edited by:

*Lorena Ruiz, Universidad Complutense de Madrid, Spain*

#### Reviewed by:

*Leandro Lorenzelli, Fondazione Bruno Kessler, Italy Giulia Tabanelli, University of Bologna, Italy*

> \*Correspondence: *Monica Gatti monica.gatti@unipr.it*

#### Specialty section:

*This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology*

Received: *08 June 2016* Accepted: *29 September 2016* Published: *17 October 2016*

#### Citation:

*Bancalari E, Bernini V, Bottari B, Neviani E and Gatti M (2016) Application of Impedance Microbiology for Evaluating Potential Acidifying Performances of Starter Lactic Acid Bacteria to Employ in Milk Transformation. Front. Microbiol. 7:1628. doi: 10.3389/fmicb.2016.01628* Impedance microbiology is a method that enables tracing microbial growth by measuring the change in the electrical conductivity. Different systems, able to perform this measurement, are available in commerce and are commonly used for food control analysis by mean of measuring a point of the impedance curve, defined "time of detection." With this work we wanted to find an objective way to interpret the metabolic significance of impedance curves and propose it as a valid approach to evaluate the potential acidifying performances of starter lactic acid bacteria to be employed in milk transformation. To do this it was firstly investigated the possibility to use the Gompertz equation to describe the data coming from the impedance curve obtained by mean of BacTrac 4300®. Lag time (λ), maximum specific M% rate (µmax), and maximum value of M% (Yend) have been calculated and, given the similarity of the impedance fitted curve to the bacterial growth curve, their meaning has been interpreted. Potential acidifying performances of eighty strains belonging to *Lactobacillus helveticus, Lactobacillus delbrueckii* subsp. *bulgaricus, Lactococcus lactis,* and *Streptococcus thermophilus* species have been evaluated by using the kinetics parameters, obtained from Excel add-in DMFit version 2.1. The novelty and importance of our findings, obtained by means of BacTrac 4300®, is that they can also be applied to data obtained from other devices. Moreover, the meaning of λ, µmax, and Yend that we have extrapolated from Modified Gompertz equation and discussed for lactic acid bacteria in milk, can be exploited also to other food environment or other bacteria, assuming that they can give a curve and that curve is properly fitted with Gompertz equation.

Keywords: impedance micorbiology, lactic acid bacteria, starter activity, BacTrac, Gompertz model

# INTRODUCTION

Impedance microbiology is a rapid method that enables qualitative and quantitative tracing of microorganisms by measuring the change in the electrical conductivity. It is based on a principle that dates back to 1899 (Stewart, 1899) but its application to food microbiology field is most recent and mainly associated to rapid detection of foodborne pathogenic bacteria (Yang and Bashir, 2007).

Impedance, applied to microbiology, can be defined as the resistance to flow of an alternating current that passes through a conducting microbial growth medium. During microbial growth, metabolic processes produce electrically measurable changes in the growth medium due to the metabolism of high-molecular weight nutrients into smaller charged ionic components that increase the electrical conductivity of the medium. Variation in electrical conductivity, monitored during time, is proportional to the change in the number of microorganisms and therefore the microbial growth can be measured (Batrinou et al., 2005).

Different systems, able to perform this measurement, are available. In the past, the most common commercial equipments used for impedance microbiology were, RABIT systemTM (Don Whitley Scientific, Shipley, UK), BactometerTM (bioMerieux, Marci l'Etolie, France) and MalthusTM (Malthus Instrument, Crawley, England). A more recent equipment is the BacTracTM microorganism growth analyser (SyLab, Purkersdorf-Vienna, Austria). Common to all the systems is the measurement of an electronic signal that quantify the movement of ions between two electrodes (conductance) while, in some devices, the storage of charge at the electrodes medium interface (capacitance) is also measured (Noble et al., 1999). Plotting of the continuous measurement of cumulative increase in conductance, or capacitance, graphically results in an impedance curve (Rediers et al., 2012). The most common way to use this curve in microbiological analysis is fixing a point, generally defined as "time of detection." However, fixed the microorganism, medium and temperature, this point varies between one device and another. Time to detection (TTD) for RABIT corresponds to the point where the cumulative change in conductivity from the baseline meets or exceeds a set value over a defined time interval (Rediers et al., 2012). Detection Time (DT) for Malthus is obtained when a change in conductance over a threshold reference value set by the operator is observed (Lanzanova et al., 1993). DT of Bactometer is the amount of time required to cause a series of significant deviations from baseline impedance values (Noble et al., 1999). DT for Backtrack is the time when the impedance curve meets the threshold level of 5% (Curda ˇ and Plocková, 1995). Indeed, the "time of detection" is the principal parameter measured by all the devices and it coincides with the reaching of a cells concentration of about 106–10<sup>7</sup> cells per ml (Noble et al., 1999) thus, it is strongly affected by bacterial cells physiological state (Lanzanova et al., 1993). this parameter is largely used to monitoring pathogens or spoiling bacteria in food (Gracias and McKillip, 2004), and also antimicrobial activity (Marino et al., 2001; Silva et al., 2003; Kunicka-Styczynska ´ and Gibka, 2010) included lytic infections by bacteriophages (Amorim et al., 2009). Recently, an intriguing unconventional approach to impedance microbiology was considered to detect bacteriophages responsible for cell lysis (Mortari et al., 2015). However, to the author's knowledge, the significance of the whole impedance curve have never been objectively related to microbial behavior.

The responses of microorganisms to specific environmental conditions, such as temperature, pH and aw, can be described by predictive microbiology, a sub-discipline of food microbiology dealing with the development of mathematical models (Baranyi and Roberts, 1995). Several models have been developed to represent and predict microbial growth or inactivation in food and, nowadays, such models can be very useful in food technology and processing since they are applied to predict the outcome of fermentation processes under particular circumstances and to assess the effects of environmental conditions on microbial growth. Examples of primary models, widely applied to describe the growth of lactic acid bacteria, include sigmoidal equations, such as Logistic and Modified Gompertz models (Chowdhury et al., 2007; Slongo et al., 2009). This describes the changes of the microbial population density as a function of time using a limited number of kinetic parameters (e.g., lag time, growth or inactivation rate and maximum population density) while it is not taken into account the stage of death. The Gompertz model provides a convenient mathematical tool that approximates the way in which microbiologists have traditionally estimated the graph of the growth kinetics (Buchanan et al., 1997).

Aim of this work was firstly to investigate the possibility to use the Gompertz equation to describe the data coming from the impedance curve obtained by mean of BacTrac 4300 <sup>R</sup> and, secondly, to use the so described kinetics parameters, to evaluate the potential acidifying performances of several lactic acid bacteria strains for their possible use as starters in milk transformation.

# MATERIALS AND METHODS

# Strains, Media, and Growth Conditions

Eighty strains representing four starter lactic acid bacteria species, Lactobacillus helveticus, Lactobacillus delbrueckii subsp. bulgaricus, Lactococcus lactis, and Streptococcus thermophilus (**Table 1**), were analyzed by impedance measurements. The strains, belonging to the collection of the Laboratory of Food Microbiology of the Department of Food Science of University of Parma, have been previously isolated from dairy matrixes and identified by16S rRNA sequencing.

Strains, maintained as frozen stocks cultures in MRS (Oxoid, Ltd., Basingstoke, United Kingdom) (Lactobacillus), or M17 (Oxoid Ltd.) (Lactococcus and Streptococcus) broth containing 20% (v/v) glycerol at −80◦C, were recovered in MRS or M17 broth by two overnight sub-culturing (5% v/v) at 42◦C for Lactobacillus and Streptococcus, and 30◦C for Lc. lactis. Then, other 28 h sub-culturing (5% v/v) of each strain in skim milk powder (Oxoid Ltd.), reconstituted to 10% (w/v) and sterilized at 110◦C for 30 min (SSM), were performed before use.

# Impedance Measurement

A BacTrac 4300 <sup>R</sup> Microbiological Analyzer (Sylab, Austria) system, consisted of two incubators allowing four different temperatures simultaneous setting, was used. The strains L. helveticus 5, L. delbrueckii subsp. bulgaricus 202, Lc. lactis 4068, and S. thermophilus 547 were 10-fold (first dilution), 100 fold (second dilution), 1000-fold (third dilution), 10,000-fold (fourth dilution), 100,000-fold (fifth dilution) diluted in ringer solution (Oxoid Ltd.). Not diluted colture and each dilution were

#### TABLE 1 | Lactic acid bacteria strains used in this study.


*(Continued)*

#### TABLE 1 | Continued


inoculated (2% v/v) into previously sterilized measuring cells filled with 6 ml of SSM.

The impedance measurement was performed at 42◦C for Lactobacillus and Streptococcus strains, and 30◦C for Lactococcus strains. Subsequently 100µl of the second dilution was used as inoculum for the analysis of all the 80 strains at their optimum growth temperature.

Moreover, three strains for each species (L. helveticus 3, 9, 23; L. delbrueckii subsp. bulgaricus 260, 265, 3436; Lc. lactis 664, 4064, 4067, and S. thermophilus 192, 160, 526) were also tested at different temperatures: 32◦ , 37◦ , 42◦ , and 47◦C for Lactobacillus and Streptococcus strains and 20◦ , 25◦ , 30◦ , and 35◦C for Lactococcus strains For each test, impedance measurement was recorded every 10 min for 80 h. All the analysis were carried out in duplicated. One negative sample, consisting of non-inoculated SSM, was also incubated for each temperature tested.

# Statistical Analysis

The means and standard deviations of impedance changes in the medium (M%) data were calculated using SPSS (Version 21.0, SPSS Inc., Chicago, IL, USA) statistical software.

# RESULTS AND DISCUSSION

# Impedance Curve Interpretation

Impedance measurement is based on the principle that during microbial growth, metabolic processes produce electrically measurable changes in the growth medium. Milk has itself conductive properties because it is rich in charged compounds, especially minerals and salts (Mucchetti et al., 1994). During lactic acid fermentation, the decrease of lactose and the subsequent increase of lactic acid lower the medium pH and, at the same time, enhance its electrical conductivity as a result of the accumulation of lactate ions during fermentation (Carvalho et al., 2003).

Moreover, acidification of milk changes equilibria of buffer system and solubilizes casein-bound calcium and phosphorous salts. This phenomenon increases conductivity sharply, so there is a positive correlation between increased conductivity and milk acidification due to lactic acid bacteria activity.

This variation of electrical conductivity of milk is proportional to the change in microorganisms number and their metabolic activity and, therefore, microbial growth in milk can be measured (Mucchetti et al., 1994). The BacTrac 4300 <sup>R</sup> system measures two specific impedance values, the E-value which is referred to as the impedance change at the electrode surface, and the M-Value which is the change in conductivity in medium, SSM in this case (Batrinou et al., 2005).

The system enables a separate registration of impedance changes in the SSM (M-value) and at the electrode (E-value). For the experiments carried out in the present study, the impedance change (M-value) of the SSM was used. This value, recorded every 10 min, is revealed as a relative change in the measurement signal and shown as M% percentage in function of time (80 h) in an impedance curve (continues line in **Figure 1**).

With the aim of translating the metabolic significance of the impedance curve into objective parameters, M% data were fitted to the Modified Gompertz equation (Gibson et al., 1988) using DMfit version 2.1 Excel add-in (http://www.combase.cc/index.php/en/tools). DMfit is part of the system used in-house at the Institute of Food Research to model the time-variation of the logarithm of cell concentration of bacterial batch cultures (www.ifr.ac.uk). Particularly, MS Excel adding DMfit is a free software application for predictive microbiology modeling developed by the Computational Biology Group at Institue of Food Research (Norwik, UK; Perez-Rodriguez and Valero, 2013). Among the primary models available, modified Gompertz equation was used to describe the microbial evolution with time (Swinnen et al., 2004). In this research, the equation was used instead to describe M% in function of time. The fitted data are represented by a sigmoidal curve (shown as dotted line **Figure 1**) with two inflection points and generate 3 parameters: (i) lag time (λ), (ii) maximum specific M% rate (µmax), and (iii) maximum value of M% (Yend) (**Figure 1**). The possibility to fit the original data to the Modified Gompertz equation is tied to the necessity that the two curves overlap. All the curves obtained in this study have respected this rule (data not shown).

Lag phase is an adjustment period during which bacterial cells modify themselves in order to take advantage from the substrate, milk in this case, and initiate exponential growth, so the cells are assumed to be non-replicating (Swinnen et al., 2004). The duration of the Lag phase depends on the strain, temperature and the substrate in which bacteria grow. Many hypotheses have been proposed to describe the formation and duration of the bacterial Lag phase in a growth curve. One of this hypotheses is the individual cell lag time theory (Huang, 2016). Based on this theory, the formation of Lag phase in a bacterial culture is determined by each cell and each cell may leave its lag state individually. Each cell would need to accumulate critical substance before it can grow and start dividing. Once a cell leaves its Lag phase, it enters the exponential phase, starting to grow and divide immediately (Huang, 2016). Based on this concept, lag time (λ) of an impedance curve can be considered as the time that the inoculated cells need to adapt to the condition of the analysis. In the same medium (SSM) at the same temperature (42◦ and 30◦C depending on the species), as expected, for all the species the lower was the inoculum, the greater was the lag time and thus, this parameter is inoculum dependent (**Figure 2**, **Table 2**). It has not been possible to register λ value for the inocula of the first and second dilutions of S. thermophilus because the time was incompatible with the minimum time of registration of the system that needs 1 h to start recording data. During this time, λ values of the first and second dilutions are reached but not recorded.

The second parameter, maximum specific M% rate (µmax) is comparable to the exponential phase and can be used to define LAB fermentation or acidification rate in SSM, which is an important parameter in technological processes, since the greater is the rate, the faster is the acidification. This parameter was inoculum independent as evidenced by the coefficient of variation lower than 10% (**Table 2**). However, due to the limit

µmax, maximum specific M% rate; Yend, maximum value of M%.

of this system that needs 1 h to start recording data, it is better not to use the inocula with highest cell concentrations, such as the undiluted inoculum for L. helveticus and S. thermophilus because the exponential phase of these cells starts during the BacTrac stabilization. For other devices, which need less time to start recording data, also undiluted inoculum would be used. Considering that the cells divide at a constant rate depending on the composition of the growth medium and the conditions of incubation, the M% rate (µmax) parameter could also be used to determine the time of duplication or generation time. However, as generation time is the time required for microbial cells to double in number (Madigan et al., 2009), to extrapolate the value of generation time from impedance value, a correlation with µmax and number of cells should be carried out.

The third parameter (Yend), is the highest point of the fitted curve, very close to the maximum variation of impedance recorded (**Figure 1**). This value can be interpreted as the maximum capability of each strain to modify the impedance in SSM and thus depends mainly on its capability to accumulate lactate ions during growth. The capability to accumulate lactate ions can be measured as total amount of lactic acid, as for example, those produced in yogurt, by means of different chemical methods (De Noni et al., 2004). The amount of lactate ions accumulated during growth depends on different aspects, among which, the initial amount of lactose, and acidity tolerance of the strains. In the same medium, with the same amount of initial lactose, Yend can be associated to the acidifying capability and to the resistance of the bacteria to the produced acidity. Of the three considered parameters, Yend is the more independent from the amount of inoculated cells (**Table 1**).

Considering what has been observed with this first part of the work, if the purpose is to know acidification rate and the amount of produced acid, different inoculum concentrations can be used, getting the same results. However, also considering the minimum time of registration of BacTrac 4300 <sup>R</sup> , the use of a high bacterial concentration, corresponding to the undiluted inoculum or to first dilution, has to be excluded, because it does not allow the visualization of the Lag phase. In this study, we wanted to consider also the λ-value and thus we decide to carry out the analysis with the same inoculum concentration. The highest inoculum that allows the best description of the microbial growth performance in SSM was found to be the second dilution, that has been therefore used for the following determination.

# Impedance Analysis of Starter LAB at Optimal Growth Temperature

Aiming at evaluating the metabolic significance of the three kinetics parameters λ, µmax, and Yend, 100µl of the second dilution of 20 strains for each considered LAB species were analyzed in duplicate at their optimal growth temperature, 42◦C, for Lactobacillus and Streptococcus, and 30◦C for Lactococcus.

λ was variable among species and L. helveticus showed, on average, the highest values of this parameter. This can be translated into a longer transition period during which the specific growth rate increases to the maximum value characteristic of the culture environment (Swinnen et al., 2004) and thus it can be interpreted as a slower adaptability of the species to the growth condition (**Figure 3A**). However λ

#### TABLE 2 | Values of Lag, Rate, and yEnd obtained from the serial dilutions of one strain for species.


\**Dilution.*

§*Standard Deviation.*

†*Coefficient of variation.*

‡*Not determined.*

of 20 strains for each species, (C) Yend (maximum %M) mean value of 20 strains for each species. Error bars show standard deviation (SD) for each species.

was also highly variable within the species: L. helveticus and L. delbrueckii subsp. bulgaricus were the most heterogeneous species, as revealed by standard deviation values (SD) in **Figure 3**, while Lc. lactis and S. thermophilus strains showed the lowest and less variable values. Variability of L. helveticus in acidifying activity is well known and already measured in different way (Gatti et al., 2003). However, results coming from impedometric analysis and expressed as time of detection, are prone to the variability of the used system. By using the λ value instead, measurements can be made and compared independently from the systems used for the analysis.

Less used in food impedance microbiology but more interesting, if we consider its metabolic significance, is the parameter µmax. Thanks to the approach followed in this work, that is the elaboration of impedometric data by Modified Gompertz equation it is possible to define and compare the different µmax features of the four species. Considering impedance curve of LAB in SSM, high µmax means high acidification rate and S. thermophilus and Lc. lactis showed higher value in milk respect to Lactobacillus species. To one side, this behavior of species is not new (Michel and Martley, 2001; Leroy and Vuyst, 2004) but new and of great interest, is the possibility to easily compare acidification rates among the species and above all among the strains. In this regards, L. delbrueckii subsp. bulgaricus showed the highest intra-species variability (**Figure 3B**). This variability could be of great technological interest because among the same species it is possible to choose the strain with higher or lower acidification rate depending on their possible application. For instance, fast acidifying ability can be required for a fermented milk production. On the other hand, lower rate could be desirable for mixed cultures where LAB coexist during changing environmental stimuli and stresses, which can affect their cellular physiology (Arioli et al., 2016).

Variability of maximum acidification rate, calculated measuring pH changing after defined time intervals, for 40 S. thermophilus strains, was already observed by Zanatta and Basso (Zanatta and Basso, 1992) using the system Micros (Conegliano, Italy). They found that the maximum acidification rate was the main variable discriminating strains in fast, medium and slow acidifying group (Zanatta and Basso, 1992). More similarly to our approach, a Don Whitley RABIT system (Sherry et al., 2006) was used for a qualitative study of Salmonella. µmax for 14 Salmonella serovars was determined in less than 7 h, in respect to 24 h needed by conventional method (Sherry et al., 2006).

Different values of Yend, thus different final acidification capability in milk, were found for the four species (**Figure 3C**). The highest value, with the lowest SD was found, on average, for L. helveticus. On the contrary, the lowest value, with the highest SD, was found for L. delbrueckii subsp. bulgaricus. This means that the first species has the best acidification capability, while the latter has the worst. However, L. delbrueckii subsp. bulgaricus also showed, as already observed for the others parameters, the more heterogeneous behavior within the species. This can be due to different acid resistance, but it can also be associated with the different capability of the species to metabolize the galactose moiety after lactose uptake. L. helveticus is able to ferment the glucose and galactose moieties of lactose (Mollet and Pilloud, 1991), while consumption of galactose by L. delbrueckii species depends on the subspecies, and the inability of the subspecies delbrueckii and bulgaricus to metabolize galactose could be due to the loss of the galT gene (Germond et al., 2003). The low average value of Yend and high level of SD measured for L. delbrueckii subsp. bulgaricus, were due to the presence of at least 6 strains characterized by Yend values lower than 20 (data not shown), possibily linked to the absence of the galT gene (Germond et al., 2003). Incapacity to metabolize galactose may be also the reason for lower levels of Yend found for Lc. lactis (**Figure 3C**), confirming that during the metabolism of lactose by Lc. lactis, part of the galactose 6P is dephosphorylated and excreted into the growth medium, while the glucose moiety is readily used (Neves et al., 2010). High and homogeneous level of Yend in S. thermophilus could be due to galactose positive strains. The existence of galactose negative strains has been reported, but only as a mutation of recent past (de Vin et al., 2005).

Correlation between the two parameters µmax and Yend was not found (data not shown), indicating that fastest strains were not always the greatest acidifying ones. Thus, this method of characterization allow to choose the best strains considering which parameter is the most important for the desired technological application. For example, L. helveticus 35 was the best acidifying strain among all studied strains (Yend 31.4) but it was the slowest of its species (data not shown). On the other hand, one of the best acidifying S. thermophilus strain, 410 (Yend 29.6), was the faster (rate 5.8) among all studied strains (data not shown).

# Impedance Analysis of Starter LAB at Different Growth Temperature

Mesophilic bacteria, such as Lc. lactis, have an optimum growth temperature of 30◦C, while thermophilic species, such as L. helveticus, L. delbrueckiisubsp. Bulgaricus, and S. thermophilus, have an optimum growth temperature of 42◦C. However, starter LAB employed in dairy fermentations can grow over a wide temperature range varying from 4 to 50◦C (Hickey et al., 2015). This aspect is of particular importance because the milk transformations, such as microbial fermentation for the production of fermented milks and acidification of the curd in cheeses production, may involve temperatures quite far from the optimal for bacterial growth.

Considered this, in order to see how starter strains change their performances depending on temperatures, three strains for each species, chosen among the 20 previously evaluated, were tested through impedance analysis under temperatures 5◦ and 10◦C lower and 5◦C higher than the optimal for thermophilic species and 5◦C lower and 5◦ and 10◦C higher than the optimal for mesophilic species.

Varying temperature, Lag was the parameter that changed greatly. The differences were relevant for L. delbrueckii subsp. bulgaricus, L. helveticus and Lc. lactis and to a lesser extent for S. thermophilus, although it was clear that differences were strains dependent (**Figures 4A–D**). The time of adaptation to the different temperatures was longer when temperature was

higher than the optimal both for the Lactobacillus species and Lc. lactis. On the contrary, it was shorter, when temperatures were lower than optimal. However, the strains L. helveticus 3, L. delbrueckii subsp. bulgaricus 3436, and Lc. lactis 664 and 4067 showed to adaptable more easily to the higher temperatures (**Figures 4B–D**). Of particular interest was the thermal tolerance observed for L. delbrueckii subsp. bulgaricus 3436 (**Figure 4B**).

The acidification rate, measured as M% (µmax), was variable for all species in function of the variation of the temperature; at the optimum, it was higher than rates at lower or higher temperatures for all strains (**Figures 4A–D**). This data is in agreement with the effect of changing temperature on the specific growth rate µ, evaluated by a pH-auxostat study for one strain of S. thermophilus, one of L. delbrueckii subsp. bulgaricus and one of Lc. lactis (Adamberg et al., 2003). These authors observed that S. thermophilus had the highest specific growth rate at 44◦C and a slight decrease at 45◦C; 43◦C was the temperature at which L. delbrueckii subsp. bulgaricus reached the maximum rate level, while Lc. lactis reached the maximum at 35◦C and then, slightly decreased (Adamberg et al., 2003). Interestingly, in the present work, we found exceptions for L. delbrueckii subsp. bulgaricus 221, that at 47◦C acidified faster than at its optimum (42◦C; **Figure 4B**).

All Lc. lactis strains, and in particular strain 4067, slowed consistently the rate when incubated at 40◦C, while they tolerated the oscillation of 5◦ , higher and lower than their optimum 30◦C (**Figure 4D**).

The acidifying capacity was not greatly affected by the temperature for the thermophilic species even when incubated at 10◦C below the optimum (42◦C. **Figures 4A–C**). In particular, all the three S. thermophilus strains have maintained comparable acidification capacity values at each considered temperatures (**Figure 4A**). Instead, the response of Lc. lactis was strictly strain specific at all the temperatures. In particular, Lc. lactis 4067 and Lc. lactis 4064 showed similar acidification capacity at respectively 40◦ and 30◦C (optimum, **Figure 4D**).

In an intriguing experiment, it was demonstrated that one Lc. lactis strain, mutant TM29, after a long adaptation of 860 generation, was able to grow well up to 39◦C due to mutations accumulated, most of which were shown to affect thermal tolerance (Chen et al., 2015). The goal of that research was to demonstrate a simple approach to obtain non-GMO derivatives of Lc. lactis that possess properties desirable by the industry, such as thermal robustness and increased rate of acidification. In fact, Chen et al. (2015) report that in the same cheese production, during the curdling process, the temperature is often raised to around 40◦C, or even beyond, and in those condition Lc. lactis stops growing dramatically, reducing curd acidification. In this perspective, Lc. lactis 4067, which by the way, is a wild strain isolated from raw cow milk used for Grana Padano cheese production, could have a great potential industrial interest.

# CONCLUSION

The impedance microbiology is used since the seventies, but, besides the food control analysis to which it is commonly applied, only few researches had the purpose to study its different potential applications. With this work we wanted to find an objective way to interpret the metabolic significance of impedance curves and propose it as a valid approach to evaluate the potential acidifying performances of starter lactic acid bacteria to employ in milk transformation. The novelty and importance of our findings, obtained by means of BacTrac 4300 <sup>R</sup> , are that they can also be applied to data obtained from other impedometric devices. Moreover, the meaning of Lag, µmax and Yend that we have extrapolated from modified Gompertz equation and discussed for LAB in milk, can be exploited also to other food environment or other bacteria, assuming that they can give a curve and that curve is properly fitted with Gompertz equation Through this study, it was possible to highlight that the LAB species with the highest

# REFERENCES


acidification rate were S. thermophilus and Lc. lactis, while L. helveticus and S. thermophilus showed the greatest acidification capacity. Among the 80 studied strains, 20 for each species, the widest heterogeneity was observed within L. delbrueckii subsp. bulgaricus subspecies. This intraspecific diversity was particularly evident when temperature was far from the optimal. Results obtained for some strains may be of interest for fermented milk and cheese production, particularly for cooked or semi-cooked cheeses.

# AUTHOR CONTRIBUTIONS

EB Substantial contributions to the design of the work; the acquisition, analysis and interpretation of data for the work. VB Substantial contributions to the design of the work and interpretation of data for the work. BB Drafting the work and revising it critically for important intellectual content. EN revising the work critically for important intellectual content and final approval of the version to be published. MG Substantial contributions to the conception and design of the work, interpretation of data for the work, agreement to be accountable for all aspects of the work in ensuring that questions related to the accuracy and integrity of any part of the work are appropriately investigated and resolved.

# ACKNOWLEDGMENTS

The authors are grateful to Sacco Clerici Group (22071 Cadorago, Italy) for making available some strains used in this research.

acidilactici H during the production of bacteriocin pediocin AcH. J. Food Eng. 80, 1171–1175. doi: 10.1016/j.jfoodeng.2006.08.019


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

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

# Recent Advances on Multi-Parameter Flow Cytometry to Characterize Antimicrobial Treatments

Lucie Léonard, Lynda Bouarab Chibane, Balkis Ouled Bouhedda, Pascal Degraeve and Nadia Oulahal\*

Univ Lyon, Université Claude Bernard Lyon 1, ISARA Lyon, BioDyMIA (Bioingénierie et Dynamique Microbienne aux Interfaces Alimentaires), Equipe Mixte d'Accueil n◦3733, IUT Lyon 1, Bourg en Bresse, France

The investigation on antimicrobial mechanisms is a challenging and crucial issue in the fields of food or clinical microbiology, as it constitutes a prerequisite to the development of new antimicrobial processes or compounds, as well as to anticipate phenomenon of microbial resistance. Nowadays it is accepted that a cells population exposed to a stress can cause the appearance of different cell populations and in particular sub-lethally compromised cells which could be defined as viable but non-culturable (VBNC). Recent advances on flow cytometry (FCM) and especially on multi-parameter flow cytometry (MP-FCM) provide the opportunity to obtain high-speed information at real time on damage at single-cell level. This review gathers MP-FCM methodologies based on individual and simultaneous staining of microbial cells employed to investigate their physiological state following different physical and chemical antimicrobial treatments. Special attention will be paid to recent studies exploiting the possibility to corroborate MP-FCM results with additional techniques (plate counting, microscopy, spectroscopy, molecular biology techniques, membrane modeling) in order to elucidate the antimicrobial mechanism of action of a given antimicrobial treatment or compound. The combination of MP-FCM methodologies with these additional methods is namely a promising and increasingly used approach to give further insight in differences in microbial sub-population evolutions in response to antimicrobial treatments.

Keywords: multi-parameter flow cytometry, microorganisms, antimicrobial treatment, double-staining, antimicrobial mechanism, viability, culturability

# INTRODUCTION

Inactivation of microorganisms by physical treatments [heat, Ultra-Violet (UV) light irradiation, supercritical carbon dioxide, high hydrostatic pressure,. . . ] or by the action of antimicrobial compounds (biocides, organic acids, peptides, essential oils,. . . ) can result from several mechanisms: inhibition of cell wall synthesis, disruption of the cytoplasmic membrane, binding to DNA, inhibition of protein synthesis, or anti-metabolite activity (Lee et al., 2015). To develop new antimicrobial processes or compounds in food or medical microbiology, an understanding of their mechanisms of action is vital in order to apply them best and to anticipate microbial resistance phenomenon.

Over the years, several methods have been developed to measure viability and vitality of microbes under various stresses: plating, slide culture, vital stains, metabolic activity monitoring,

#### Edited by:

Christophe Nguyen-The, Institut National de la Recherche Agronomique, France

#### Reviewed by:

Kiiyukia Matthews Ciira, Mount Kenya University, Kenya Marielle Bouix, Agro ParisTech, France

> \*Correspondence: Nadia Oulahal nadia.oulahal@univ-lyon1.fr

#### Specialty section:

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

Received: 17 May 2016 Accepted: 22 July 2016 Published: 08 August 2016

#### Citation:

Léonard L, Bouarab Chibane L, Ouled Bouhedda B, Degraeve P and Oulahal N (2016) Recent Advances on Multi-Parameter Flow Cytometry to Characterize Antimicrobial Treatments. Front. Microbiol. 7:1225. doi: 10.3389/fmicb.2016.01225 cell components monitoring, fermentation capacity, acidification potential, or oxygen uptake ability (Hayouni et al., 2008). Such methods are time-consuming and labor intensive (Wilkinson, 2015). Besides, when classical plate count method is used to assess survival of a bacterial population after exposure to an environmental stress, the viability is determined by counting live cells which are those that managed to replicate under the particular experimental conditions, while all the others will be presumed dead (Hayouni et al., 2008). Nowadays, it is well-documented that, under stress conditions a population will exhibit cell subpopulations with phenotypes that most likely escape this logic. Environmental stresses can trigger the occurrence of certain cell populations, called viable but nonculturable cells (VBNC), which were stressed and lost their ability to grow on agar medium, but still showed metabolic activity (Ananta et al., 2004). Quantification of injured cells is a great concern for microbiologists, as this subpopulation might be critical if cells can recover and revert to their physiologically active condition (Ayari et al., 2013). For these reasons, fluorescence techniques combined with direct optical detection methods for the rapid assessment of bacterial viability have been increasingly favored for about 10 years.

Among these techniques, multi-parameter flow cytometry (MP-FCM) has been shown to be a powerful tool for rapidly analyzing populations on a cell-by-cell basis and can be applied in many areas of food or medical microbiology (detection of pathogenic bacteria, monitoring lactic acid bacteria fermentation, rapid microbiological analysis of drinking water; Schenk et al., 2011). A flow cytometer can be described as an "automatic microscope" with the advantages of objectivity, high analysis rate, precision and sensitivity (Díaz et al., 2010). The principle is that particles in suspension are pumped into a narrow flow stream intersected by one or more laser beams. Single particles, such as microbial cells, are illuminated individually with the resulting light scatter and fluorescence emission detected at appropriate wavelengths (Bridier et al., 2015). A very large number of particles can be measured, 5000 cells per second in common and even up to 100,000 in specialized instruments, measuring multiple cellular parameters on each cell simultaneously (Díaz et al., 2010). Each individual cell can be characterized based on its fluorescence color, the intensity of the fluorescence signal, as well as the size, shape and granularity of the particles (Bridier et al., 2015). This method is highly compatible with a broad range of fluorescent stains and cell labeling methods. Díaz et al. (2010) detailed principles and instrumentation through schematic descriptions in their review about application of flow cytometry (FCM) to monitor industrial microbial bioprocess.

Firstly FCM developments occurred in human clinical applications especially for immunological analysis (Wilkinson, 2015). For less than 20 years, FCM has become an indispensable tool to the complex area of microbiology. In 2000, a set of publications in the Journal of Microbiological Methods presented cytometry for bacteria (Nebe-von-Caron et al., 2000; Shapiro, 2000; Steen, 2000). More recently, these first data were supplemented by a review of Tracy et al. (2010). The ability to use FCM to visualize, enumerate and analyze a population of cells into subpopulations of varying physiological status is a valuable aid to understanding this intricate area for the microbiologists. Bridier et al. (2015) reviewed the applications in food microbiology such as study of food bacteria function, detection of food microbial communities or detection and persistence of food pathogens. Moreover FCM can be used to elucidate antimicrobial mechanism in food or health domain. Recently, Mathur et al. (2016) published a review about FCM as a tool to study the effects of bacteriocins on prokaryotic and eukaryotic cells. The advancement of FCM and the introduction of novel fluorochromes allow to study the viability of cells, the membrane structure and its integrity, and the membrane potential at a single-cell level. In the perspective of elucidation of the antimicrobial mechanism, FCM should be a very interesting tool. In this review, we describe some recent studies that use FCM as a tool to evaluate the effect of antimicrobial treatment on microbial cells. More specifically, we focus on the use of MP-FCM with individual and simultaneous staining describing the advantages and the limitations. FCM is by definition a multiparametric technique: cells are gated on at least size or complexity parameters and one parameter of fluorescence. Nevertheless, MP-FCM was defined here, as described in the literature, as FCM using several fluorochromes in the same study (Tracy et al., 2010). Based on recently published results, the complementarity with other methods and particularly plate counts methods is discussed.

# INFORMATIONS RESULTING FROM DIRECT ANALYSIS OF MICROBIAL CELLS BY FLOW CYTOMETRY

Although staining of microbial cells with dyes prior to flow cytometry analysis is dominating (and will be presented thereafter), direct analysis of cells without staining cells already gives information regarding the morphology of analyzed cells.

Even without staining the sample, a cell immersed in the injected solution already produces signals through the orthogonal-to-flow laser focused beam. The fraction of light scattered collected in the same direction as the incident light is known as Forward Angle Light Scatter (FALS) or Forward Scatter (FS or FSC). This fraction allows an estimation of the cell size: indeed, the quantity of the scattered light increases with the cell size (Díaz et al., 2010; Tracy et al., 2010). The fraction of light scattered laterally and fluorescence are collected and divided by a lens at 90◦ from the incidence axis of the laser. The fraction of light scattered in right angle is known as Right Angle Light Scatter (RALS) or Side Scatter (SS or SSC). This signal is related to cell complexity described by morphological characteristics such as cell surface roughness, cell membrane, nucleus and internal granular material, number of organelles (Díaz et al., 2010). For example, this first information concerning SSC vs. FSC allowed Kramer and Thielmann (2016) detecting that bacterial cells aggregated during heat exposure by change of the scatter signals. The strong increase of SSC and FSC indicated that cells formed large agglomerates, exhibiting up to 100-times higher scatter signals than single cells. Booyens and Thantsha (2014) also detected differences in shape and density of the populations' scatter patterns after exposure to garlic clove extract compared to control population for all the tested Bifidobacterium strains. Their hypothesis was that this change in size and external morphology was a change from rod to coccoid shape. Schenk et al. (2011) highlighted the same modification for Escherichia coli cells after UV-C light treatment.

# ANALYSIS OF MICROBIAL CELLS BY FLOW CYTOMETRY AFTER THEIR STAINING WITH DYES

Additional information can be obtained provided that samples are stained using fluorochromes. Scattering and fluorescence signals provide information about intrinsic and extrinsic cell parameters, respectively.

# Multi-Parameter Flow Cytometric Analysis: Individual Use of Dyes

A way to use the flow cytometry to characterize antimicrobial treatments is to perform a multi-parameter analysis with different stains in combination. This approach will be presented thereafter when these stains are used simultaneously (in 3.2. part). However, several researchers have chosen to use dyes separately (**Table 1**).

The following paragraphs briefly discuss microbial FCM dyes commonly employed in recent works to characterize antimicrobial treatments.

### Membrane Integrity

To interrogate membrane integrity, the nucleic acids (NA) content of individual cells is analyzed and dye exclusion methods are favored. NA dyes can stain DNA, RNA, or both (Tracy et al., 2010). Cells showing intact membranes are impermeable to multiple charges dyes such as dyes of the SytoxTM family or to cyanines such as TO-PRO <sup>R</sup> 3. If cells lose membrane integrity, these dyes enter into the cells emitting fluorescence upon NA binding. Propidium iodide (PI) is the most commonly used dye (Díaz et al., 2010). This dye is usually employed for dead cells detection and it is suitable for multi-parametric analysis along with green fluorochromes such as SYTO9 <sup>R</sup> . It contains two positive charges and is normally excluded from cells due to its divalence (Kim et al., 2009). Therefore, PI can only enter permeabilized cytoplasmic membranes. For instance, the commercial available LIVE/DEAD <sup>R</sup> BacLightTM kit from Molecular Probes is the most used (Possemiers et al., 2005; Kim et al., 2009; Muñoz et al., 2009; Martínez-Abad et al., 2012; Choi et al., 2013; Booyens and Thantsha, 2014; Fernandes et al., 2014; Manoil et al., 2014; Pal and Srivastava, 2014; Boda et al., 2015; Freire et al., 2015; Li H. et al., 2015; Li W. et al., 2015).

# Pump Activity

Ethidium bromide (EB) is a positively-charged monovalent compound that is used to evaluate the efflux pump system of bacteria. It is a membrane-permeant and it enters into intact cell membranes, but it is actively pumped out of the cell via a non-specific proton anti-port transport system (Kim et al., 2009; Díaz et al., 2010). When the membrane is damaged and the efflux pump also malfunctions, EB can stain the intracellular DNA of the cell (Kim et al., 2009).

## Membrane Potential

Membrane potential is generated due to the different ions content inside and outside the cell and it varies from 100 to 200 mV. Only living cells are able to maintain membrane potential: therefore, it is one of the most used parameters to assess cell viability (Díaz et al., 2010). When this difference decreases to zero, the membrane is structurally damaged, and ions go across the membrane freely, but if it means a decrease in cell activity, it does not necessarily imply death. Measurements are carried out by using lipophilic dyes which go through the cell membrane and accumulate according to their charge. The fluorescence signal can be directly related to cell energy levels and to test the reliability of staining, it is recommended to observe if the dye uptake is sensitive to ionophores such as carbonyl cyanide mchlorophenylhydrazone (CCCP; Pianetti et al., 2008; Díaz et al., 2010; Hammer and Heel, 2012; Li W. et al., 2015).

Cationic dyes, such as carbocyanins, DiOCn(3), or Rhodamine 123, accumulate inside polarized cells because viable cells are permeable to those dyes (Díaz et al., 2010; Tracy et al., 2010). Nevertheless, Gram negative bacteria outer membranes can present a barrier to lipophilic dyes uptake. However, a mild treatment with a chelating agent such as Tris-EDTA (ethylenediamine tetraacetic acid) can overcome this limitation (Tracy et al., 2010; Boda et al., 2015).

Anionic and lipophilic dyes, such as those belonging to oxonols family accumulate inside non-viable cells and concentrate by association with intracellular compounds. Without permeabilization protocols, oxonols uptake is more related to membrane integrity rather than to membrane potential and depolarization (Díaz et al., 2010). DiBAC4(3) (bis-(1,3-dibarbituric acid)-trimethine oxonol) or BOX seems to be the most used recently to detect depolarized cells of numerous species after antimicrobial treatment (Novo et al., 2000; Wu et al., 2010a,b; Silva et al., 2011; Caldeira et al., 2013; Kramer and Muranyi, 2014; Duarte et al., 2015; Lee et al., 2015; Coronel-León et al., 2016; Grau-Campistany et al., 2016; Kramer and Thielmann, 2016).

# Metabolic Activity

Metabolic activity detection suggests the absence of cell death, but giving the conclusion of alive cell or dead cell is difficult in the case of cell damage, dormancy, and starvation (Nebe-von-Caron et al., 2000; Díaz et al., 2010). In general, a non-fluorescent permeant substrate is taken up by the cell by diffusion and converted inside the cell by intracellular enzymes to a fluorescent substance which is ideally retained in cells with intact membranes (Tracy et al., 2010).

# **Respiratory activity**

Bacterial cells with electron transport system activity or respiratory activity are able to reduce 5-cyano-2, 3-ditolyl tetrazolium chloride (CTC) to an insoluble fluorescent CTCformazan product that accumulates inside the cells (Caldeira et al., 2013; Ferreira et al., 2014; Duarte et al., 2015). For


#### TABLE 1 | Examples of individual uses of dyes to perform a multi-parameter flow cytometry analysis in order to characterize antimicrobial mechanisms.

PI, Propidium Iodide; cFDA, carboxyFluorescein DiAcetate; DiBAC4, Bis-(1,3-DibutylBarbituric ACid) Trimethine Oxonol = bis-oxonol = BOX; CTC, 5-Cyano-2,3-ditolyl Tetrazolium Chloride; EB, Ethidium Bromide; 2-NBGD, 2-[N-(7-nitrobenz-2-oxa-1,3-Díazol-4-yl) amino]-2-deoxy-Dglucose; EdU: 5-ethynyl-2-deoxyuridine; FITC, fluorescein isothiocyanate; DRAQ5, 1,5-bis{[2-(di-methylamino) ethyl]amino}-4, 8-dihydroxyanthracene-9,10-dione; DiSC3, 3,3′ -Dipropylthiadicarbocyanine Iodide, DCFH-DA, 2′ ,7′ -dichlorofluorescein-diacetate.

CTC-formazan fluorescence analysis, two regions of CTCformazan relative fluorescence were analyzed, depending on the fluorescence intensity of positive and negative controls (Ferreira et al., 2014; Morishige et al., 2015). Ferreira et al. (2014) and Díaz et al. (2010) expressed the limit of this stain: CTC staining allows the detection of the most metabolically active bacteria, cells with low respiratory activity may not be detected as CTCpositive, probably due to the relative toxicity of cellular CTC accumulation.

# **Enzymatic activity**

Esterase activity is the most common way to evaluate enzymatic activity (Díaz et al., 2010), particularly in studies about damages after antimicrobial treatment (Ananta et al., 2004; Hayouni et al., 2008; Ananta and Knorr, 2009; Schenk et al., 2011; Ayari et al., 2013; Thabet et al., 2013; Antolinos et al., 2014; Kramer and Muranyi, 2014; Surowsky et al., 2014; Hong et al., 2015; Combarros et al., 2016; Kramer and Thielmann, 2016; Meng et al., 2016). Fluorescein and fluorescein derivatives have been used for a wide range of microorganisms as probes for enzymatic activity measurement (Díaz et al., 2010). Among these, carboxyfluorescein diacetate (cFDA) is used primarily for the evaluation of esterase cellular activity (Ananta et al., 2004). It is a lipophilic non-fluorescent precursor that rapidly diffuses across the cell membranes. In the intracellular compartment, diacetate groups of cFDA are hydrolyzed by unspecific esterases into carboxyfluorescein (cF) which is a polar membrane-permeant fluorescent compound. The cells only remain fluorescent if their membranes are intact, thus for cells to be associated as viable, this probe requires both active intracellular enzymes and intact membranes (Hoefel et al., 2003). Moreover, efflux of cF upon glucose addition could also be used as an additional indicator of metabolic performance of the cell (Ananta et al., 2004). Schenk et al. (2011) showed how carefully this staining must be used. Untreated E. coli, Listeria innocua, and Saccharomyces cerevisiae stained with cFDA showed a heterogeneous behavior in their fluorescence labeling properties. Not all cells yielded high green fluorescence and appeared in expected quadrant in the cytogram. They formulated two hypotheses: (i) the presence of the outer membrane with lipopolysaccharides in Gram negative bacteria or the thick wall of peptidoglycan in Gram positive bacteria which does not allow cFDA freely diffusing across cytoplasmic membrane; (ii) the active expulsion of cF outside the cell by bacteria pumps and consequently the lack of green fluorescence despite the existence of metabolic activity.

Anyway, concentrations of probes, incubation periods, cytometric setting, and control should be assessed for each studied bacteria individually (Kramer and Thielmann, 2016), as performed by Nexmann Jacobsen et al. (1997) for Listeria monocytogenes. They compared the use of LIVE/DEAD <sup>R</sup> BacLightTM kit, Rhodamine 123, 2′ ,7′ -bis(2-carboxyethyl)- 5(6)-carboxyfluorescein acetoxymethylester (BCECF-AM), Chemchrome B and cFDA. In conclusion, they determined that only cFDA and Chemchrome B were suitable for rapid, almost real time counting of pure cultures of L. monocytogenes by flow cytometry, after 6−24 h incubation in selective enrichment media.

FCM with fluorescent dyes can thus be considered as a suitable tool for the assessment of structural and/or functional microbial cell properties such as metabolic activity, membrane potential, and integrity. The combination of dyes is a pertinent multi-method approach to reveal the presence of intermediate physiological states between life and cell death, showing the heterogeneities of microbial populations. Nebe-von-Caron et al. (2000) classified cells according to some active functions or the integrity of cell structures. These authors distinguished among reproductively viable, metabolically active, intact, and permeabilized cells.

A multi-method approach is a suitable tool to monitor the impact of inactivation treatments on bacteria, providing information about the mode of action, the heterogeneity of populations, species-specific differences to stressors and valuable insight in vital functions beyond pure culturability (Kramer and Thielmann, 2016).

The results obtained by FCM are in the form of graphical visualization of scattering and fluorescence cell parameters, which are being analyzed and stored for further analysis (Díaz et al., 2010). Acquired data are identified as events, as the number of cells showing a desired physical property or probe. In the case of single-staining cells, two types of graphical results could be obtained (e.g., **Figure 1**). A monoparametric histogram presents the number of cells (y-axis) vs. the scattering or fluorescence intensity (x-axis; e.g., **Figure 1-I**). Lee et al. (2015) detected membrane permeabilization of Candida albicans by Sytox Green fluorescence. Cells treated by scolopendin 2 and melittin indicated a significant increase in fluorescence intensity. These results show that the fungal cell membranes were permeabilized by the peptides. A biparametric histogram represents the intensity of the signals corresponding to different parameters in each axis (e.g., **Figure 1-II**: scattering intensity (y-axis) vs. CTC-fluorescence activity). Each dot represents a single cell and different regions can be defined into the cytogram to describe cells physiological state.

The publication of Silva et al. (2011) showed well the value of this approach in determining the mechanism of coriander essential oil against Candida spp. with three dyes: PI (membrane permeability), DiBAC<sup>4</sup> (membrane potential), DRAQ5 (DNA staining). Firstly, the percentage of PI-positive cells seemed to depend on essential oil concentration: higher essential oil concentration caused higher membrane permeability. The structure of the cell membrane was disrupted by the essential oil. Permeation to PI, particularly following short incubation period, such as 30 min, indicated that the mode of action of the essential oil involved a lesion of the cell membrane that resulted from direct damage to the cell membrane instead of a metabolic impairment leading to secondary membrane damage. Secondly, the percentage of depolarized cells was also essential oil concentration dependent. Thirdly, DNA distribution histograms were very similar in the control and in essential oiltreated cells, which could indicate that coriander essential oil did not interfere with DNA synthesis. However, at ½Minimal Inhibitory Concentration (MIC), fluorescence intensity values were slightly different: essential oil-treated cells showed higher fluorescence intensity values than control cells. This could indicate that in response to cell damage by coriander essential oil, cells were synthesizing more DNA in order to repair damage functions. Moreover, at 1 MIC, lowest fluorescence intensity values were observed, probably indicating DNA leakage from the cells. In conclusion, these three staining procedures

permitted Silva et al. (2011) to propose an antimicrobial mechanism: coriander essential oil killed Candida spp. by damaging the cytoplasmic membrane, leading to an impairment of all cellular functions. In the same way, Teng et al. (2014) used two dyes (PI and FITC-labeled AvBD103b) to demonstrate that defensin AvBD103b destroyed the membrane integrity. The antimicrobial target of the defensin was the Salmonella enteritidis CVCC3377 cell membrane. A strong permeation occurred in just 5 min treatment. Nevertheless, they made Transmission Electron Microscopy (TEM) observations and followed cellular DNA by spectroscopy to elucidate further the antimicrobial mechanism. Membrane injury was not the only mechanism of microbial killing. In the same way, by using PI-DiBAC4-EB-cFDA individual staining, Kramer and Muranyi (2014) demonstrated that pulsed light exposition induced oxidative stress with concomitant damage to the DNA molecule which were directly responsible for the loss of culturability of L. innocua and E. coli cells rather than a direct rupture of cell membranes or inactivation of intracellular enzymes.

Multi-parameter FCM analysis allowed highlighting VBNC cells populations such as in the study of Morishige et al. (2015). They treated an environmental isolate of Salmonella enteritidis clone (SE C1#15–1) with different concentrations of H2O<sup>2</sup> and analyzed by MP–FCM the respiratory activity, the glucose-uptake activity and DNA synthesis activity individually. Compared with plate count results, they defined VBNC populations as "metabolically active but non-culturable." H2O2 treated S. enteritidis cells lost their respiratory activity in a dose-dependent manner. Besides, they showed that H2O<sup>2</sup> treatment did not severely decrease the glucose-uptake activity. These results suggested that metabolic mechanism of glucoseuptake might have been more tolerant to H2O2treatment than aerobic respiration. Finally, the subpopulation of cells with DNA synthesis disappeared almost entirely after treatment with more than 3 mM H2O2. Therefore, the higher the concentration of H2O<sup>2</sup> was, the lower the population of cells showing activities of viable cells. Nevertheless, the size of the population varied considerably among the biological activities: glucoseuptake being the largest, CTC-reduction and DNA synthesis, the smallest. In this manner, Morishige et al. (2015) showed that VBNC state might be divided differently according to the metabolic activities by which they were estimated. This illustrated the value of MP-FCM analysis to screen several physiological responses in order to describe thoroughly the cellular state. Moreover, Kramer and Thielmann (2016) revealed the difficulty to discuss with only one staining and the importance of the complementary results of the multi-staining approach. They developed this idea especially for the use of cFDA. Indeed cF, a fluorescent molecule released by esterase activity on cFDA, is usually retained in intact bacteria but leaks from damaged cells, leading to a lower fluorescence intensity. cFDA, esterase substrate may not only serve as an indicator for enzymatic activity but also for membrane integrity. The discrimination of cells with compromised membrane from cells with inactive esterase is difficult. Kramer and Thielmann (2016) recommended to check for a time dependent change in the fluorescence properties of cF-stained cells and also to include a membrane impermeable dye like PI in the characterization of bacterial cells to check membrane integrity. This limitation encouraged many authors to develop a different MP-FCM approach, namely the use of two dyes simultaneously.

# Multi-Parameter Flow Cytometric Analysis: Simultaneous Use of Dyes

Simultaneous double-staining could allow characterizing an intermediate state where cells show fluorescence with both probes (**Figure 2**). The analysis with several dyes depends on the

pressure (100,400,600 MPa) and thermal (95◦C)treatment on esterase activity (cFDA) and membrane integrity (P1): quadrant A1, cF−/P1+; quadrant A2, cF+/P1+;quadrant A3, cF−/P1−; quadrant A4, cF+/P1− (Ananta et al., 2004). (II). Dual parameter dot plots with one discriminated dye. II. 1 Two different behaviors detected: initial state/Final state. II. 2.Syto9-CTC stained Arcobacter cryaeophilus LMG 10829 with resveratrol at concentration of 0 (control), 1xMIC and 4xMIC: initial state: Syto9+/CTC; Final state: Syto9+/CTC−(Ferreira et al., 2014).

compatibility of the dyes namely spectral emission and staining conditions. If the two dyes are discriminating, four quadrants can be delimited corresponding to four different physiological states (**Figure 2-I**). If only one dye is discriminating, such as Syto9 <sup>R</sup> which is able to enter all cells, two states are described: an initial and a final state (**Figure 2-II**).

Different dye couples can be used depending on the targeted cell functions (**Table 2**).

# Membrane Integrity

Using the combination between a Syto family compound and PI, four populations could be distinguished corresponding to live, dead, injured cells, or unstained debris (Possemiers et al., 2005; Kim et al., 2009; Muñoz et al., 2009; Martínez-Abad et al., 2012; Choi et al., 2013; Booyens and Thantsha, 2014; Fernandes et al., 2014; Manoil et al., 2014; Pal and Srivastava, 2014; Boda et al., 2015; Freire et al., 2015; Li H. et al., 2015; Li W. et al., 2015). The quantification of intact cells by FCM was based on membrane integrity, which is a more general property that demonstrates the protection of cell constituents and potentially capable of metabolic activity/repair and reproductive growth (Spilimbergo et al., 2010).

During an antimicrobial treatment, an increase of unstained population could appear as reported by Booyens and Thantsha (2014). They formulated two hypotheses: (i) a correspondence to the cells that have undergone severe lysis and thus lost their nucleic acids, thereby rendering them unstainable called "ghost cells"; (ii) cells that clumped together or formed interlaced chains which may decrease staining accuracy, correlated to a change in light scatter signals.

Thiazole orange (TO) is another dye for membrane integrity, which is very interesting because it is able to bind both to DNA and RNA. Since TO-DNA complex has a higher fluorescent intensity than TO-RNA complex, this allows a differentiation. Based on this approach, Surowsky et al. (2014) defined thus the bactericidal action of cold plasma against Citrobacter freundii as a mechanism based on cell permeabilization and RNA damage.

### Membrane Integrity/Physiological State

Enzymatic activity is principally studied by using cFDA dye (**Table 2**). After treatment, four kinds of subpopulations were observed with different staining characteristics: (i) lower left quadrant, unstained debris; (ii) upper left quadrant, PI-labeled cells considered as dead; (iii) lower right quadrant, cFlabeled cells considered as viable; (iv) upper right quadrant, double stained population considered as injured/compromised (Hayouni et al., 2008; Ayari et al., 2013). In Hayouni et al. (2008) study, the percentage of the unstained subpopulation after treatment with essential oils did not exceed 8%. Lactobacilli are well-known to exhibit a planar division with cells remaining attached to each other, thus producing chains characteristic for the genus and misleading FCM results. This phenomenon was avoided in many works by using sonication treatment of the cells or gentle agitation (Hayouni et al., 2008). With a PI/cFDA staining, Hong et al. (2015) explored the antimicrobial mechanism of an antimicrobial peptide, Tachyplesin I, against E. coli and Staphylococcus aureus cells. The fraction of E. coli cells with PI fluorescence increased proportionally with increase in the concentration of Tachyplesin I, as well as the fraction with cF fluorescence decreased. Nevertheless, intracellular esterase inactivation and membrane damage did not occur simultaneously after exposure to a low concentration of Tachyplesin I (5µg.mL−<sup>1</sup> ), whereas bacteria died instantaneously at concentrations exceeding 5µg.mL−<sup>1</sup> . They detected injured bacteria which cannot form visible colonies on agar plates, but which exhibited detectable metabolic activity. Schenk et al. (2011) highlighted difference of strains behavior after exposure to thermal treatment. For E. coli and S. cerevisiae, inactivation coincided with compromise of membrane, whereas for L. innocua, four subpopulations appeared. For this latter strain, thermal treatment induced cell death and was achieved in presence or absence of membrane degradation and in cells with or without enzymatic activity depending on the intensity of the treatment.

The dual staining with a cell-permeant dye and a dye detecting loss in membrane potential is another interesting possibility to distinguish between different degrees of damage. Coronel-León et al. (2016) used PI-DiBAC<sup>4</sup> double-staining to study the antimicrobial mechanism of N<sup>α</sup> -lauroyl arginate ethylester (LAE <sup>R</sup> ). The reduction in viability of Yersinia enterocolitica was evidenced for sub-MIC values of LAE <sup>R</sup> , the bactericidal action of LAE <sup>R</sup> increased with its concentration. Even if a similar effect was observed on Lactobacillus plantarum, a high proportion of unstained population was found despite the high reduction in viability. This suggested that the effect of LAE <sup>R</sup> on L. plantarum not only caused membrane depolarization and permeability but also some other non-specific effects inside the cell, such as a collapse of the cytoplasmic material, could be observed. Novo et al. (2000) reported a discrepancy between the drop in membrane potential of the cytoplasmic material and cell viability, suggesting a recovery in the bacterial population after the initial antimicrobial impact. Nucleic acid dyes should be used with caution as an indicator of cell death (Novo et al., 2000). This was the case of LAE <sup>R</sup> -treated L. plantarum: 72% of the population remained unstained, while the reduction in viability was 98.6%. This difficulty could be due to the multitarget sites of biocides, or to a reversion to an impermeable state (Novo et al., 2000). This illustrated well the need to analyze several parameters to have a comprehensive understanding of the mechanism (Coronel-León et al., 2016).

Novo et al. (2000) compared two couples of dyes, PI/DiBAC<sup>4</sup> and TO-PRO <sup>R</sup> 3/DiOC2, to analyze antibiotic effects on membrane potential, membrane permeability, and bacterial counts of S. aureus and Micrococcus luteus. The results for membrane permeability were comparable with PI and TO-PRO <sup>R</sup> 3 staining, whereas discrepancies were measured between the results obtained with DiBAC<sup>4</sup> and DiOC2for measurement of membrane permeability. Novo et al. (2000) expressed reservation about DiBAC4staining which was strongly influenced by cell size.

Another element to take into account is the nature of microorganisms: the behaviors of yeast and bacterial cells may differ. Thabet et al. (2013) treated S. cerevisiae cells by UV-A photocatalysis. Damaged cells, PI- and cF-fluorescent cells, were culturable. This was probably due to the greater


#### TABLE 2 | Multi-fluorescence properties of double-stained cells collected in a same area with the explanation of the underlying cellular mechanism.



PI, Propidium Iodide; TO, Thiazole Orange; cFDA, carboxyFluorescein DiAcetate; DiBAC4, Bis-(1,3-DibutylBarbituric ACid) Trimethine Oxonol = bis-oxonol = BOX; DiCO2, 3,3-DiethyloxaCarbocyanine Oxide; CTC, 5-Cyano-2,3-ditolyl Tetrazolium Chloride; EB, Ethidium Bromide.

complexity of cell organization compared to bacteria. Damages to the membrane do not imply an immediate enzyme activity break and cell death. For example, contrary to bacteria, respiration functions are not supported by plasma membrane, but in yeast and other eukaryotic organisms, they are compartmented to mitochondria. Thus, considering exclusively the membrane status to characterize photocatalysis effects on yeast cell viability appeared to be insufficient (Thabet et al., 2013).

Following analysis by combining dyes, a single staining may also be done in addition to not excessively complicate the analysis. This is often the case for the screening of depolarization by DiBAC<sup>4</sup> or DiOC2. In several works, these dyes were used to study the polarization of cell membrane in parallel of a double-staining experimentation (Pianetti et al., 2008; Surowsky et al., 2014; Tamburini et al., 2014; Li W. et al., 2015). (Li W. et al., 2015) performed a Syto9/PI doublestaining and a single DiOC<sup>2</sup> staining of E. coli cells to study the antimicrobial mechanisms of proline-rich peptides (monomers, dimers, and tetramers). The shift in the green fluorescent population as peptide concentration increased, indicated a shift from a mixed hyperpolarized and depolarized cell population to a more depolarized membrane population. With all MP-FCM results, they concluded that multimerization of the Chex-Arg20 monomer to dimer and tetramer altered the mode of action from non-lytic to a membrane disruptive capacity.

### DNA/Antibody

Huang et al. (2016) reported the development of a high sensitive FCM method to probe minority population of antibioticresistant bacteria. Nucleic acid dye Syto62 was used to stain all the bacteria red. Then, monoclonal antibody against TEM-1 βlactamase and Alexa Fluor 488-conjugated secondary antibody were used to selectively label resistant bacteria which retained β-lactamase activity green. This immunofluorescent-staining method remains marginal as it requires specific antibodies to be available.

Fernandes et al. (2014) expressed reservations about the use of MP-FCM. Intermediate states are generally misclassified or simply identified as "unknown" and poorly characterized population. According to them, a way to circumvent this limitation is through a well-defined gating strategy that is highly dependent on how accurate the positive and negative controls are in order to define cell population (viable, compromised, and dead cells). Moreover, that is also why other methods could be applied in parallel for a better description of the cell populations identified by MP-FCM (**Table 3**).

# CORRELATION WITH PLATE COUNTS AND OTHER TECHNIQUES USED SIMULTANEOUSLY

# Viability vs. Culturability

Nowadays it is accepted that a cell population exposed to a stress can cause the appearance of different cell populations and, in particular sub-lethally stressed/injured cells which could be defined as "active but non-culturable" or "viable but nonculturable." Nevertheless, the definition of these states is still controversial (Antolinos et al., 2014). The presence of such injured bacteria in food or in clinical applications might be critical in terms of their potential activity on excreting toxic or food spoiling metabolites, on transferring of genes (Ananta et al., 2004; Schenk et al., 2011; Ayari et al., 2013; Hong et al., 2015). Even if FCM analysis provides high-speed information at real time on damage at single cell level, plate count method also gives an indication of cells able to grow at a certain time.

First of all, Kramer and Thielmann (2016) described bacterial cell aggregation after a heat treatment. They discussed that this phenomenon could lead to an erroneous result and an overestimation of the inactivation. Indeed a single colony on the plate could contain a cell aggregate, namely more than one culturable cell (Hayouni et al., 2008). Agglomeration effect remains undetected with classical culture-based methods, then FCM acquisition was in this case a very useful method (Manoil et al., 2014).

Most studies included fluorescence staining FCM analysis and plate counting to monitor antimicrobial effects. Nevertheless, the consistency of results obtained by these two methods varied greatly across studies. On one hand, some revealed a good correlation; while on the other hand, others showed clearly a significant discrepancy between conventional plate counts and different viability staining parameters (**Table 4**). If the comparison between Colonies Forming Units (CFU) and FCM results shows a significant difference, this suggests the presence of sub-lethally stressed subpopulations, not able to form colonies on agar plates (Ayari et al., 2013). Such observations depended on the nature of antimicrobial treatments applied, and thus likely on underlying mechanisms involved. These studies also revealed differences of behavior between strains for the same treatment. The question whether injured cells are able or not to retain their TABLE 3 | Additional methods to complete MP-FCM analysis in order to better describe cell populations after exposure to an antimicrobial treatment.


proliferation capacity when returned to a favorable environment is still under debate (Manoil et al., 2014).

# Viability-Culturability Correlation

On one hand, some researchers showed a strong correlation between MP-FCM and plate counting method results (Possemiers et al., 2005; Hayouni et al., 2008; Muñoz et al., 2009; Ayari et al., 2013; Thabet et al., 2013; Boda et al., 2015; Combarros et al., 2016; Grau-Campistany et al., 2016). For example, Muñoz et al. (2009) used MP-FCM and plate counting to evaluate the viability of L. monocytogenes exposed to several essential oils with antibacterial properties. By comparing MP-FCM results and plate counts, they assumed that compromised cells may recover and grow in appropriate media. As well, Live/Dead analysis proved to be a good alternative to plate counts to monitor the antimicrobial effect of D-erythro-sphingosine in saline toward three intestinal pathogens (Possemiers et al., 2005). According to Grau-Campistany et al. (2016), the viability reduction obtained by plate count is generally in good correlation with membrane damage caused by antimicrobial peptides.

# Absence of Correlation between Viability and Culturability

One the other hand, for example, carbon dioxide treatment seemed to be a treatment causing the appearance of VBNC cells (Kim et al., 2009; Spilimbergo et al., 2010; Tamburini et al., 2013; Ferrentino et al., 2015; Li H. et al., 2015). After 5 min Super-Critical carbon dioxide (SC-CO2) pasteurization treatment at 45 and 50◦C of Listeria cells, about 4 and 6 log reductions were measured by plate counts, respectively, while the reductions measured by MP-FCM were 2 log at 45◦C and 4 log at 50◦C (Ferrentino et al., 2015). These results indicated that a substantial amount of Listeria cells of about 10<sup>2</sup> cells.g−<sup>1</sup> remained intact but non-culturable after treatment. Spilimbergo et al. (2010) performed a study on S. cerevisiaie cells in saline solutions treated by SC-CO<sup>2</sup> at 36◦C and 10 mPa and they also indicated that conventional cultivation-based methods did not allow an accurate quantification of intact cells. For a treatment at 5000 rpm-100 bar-36◦C during 10 min, the difference was particularly evident with a difference in the percentage of cell survival exceeding 55%. These data indicated a high presence of non-culturable but still intact yeast cells in the treated samples. Nevertheless, after a treatment under the same operative conditions but during 20 min, the difference decreased down to 10% between the two analytical methods. This experimental evidence was probably due to the mechanical irreversible stresses to the cells membranes induced by a longer time. Similar findings were reported on FCM and plate counts analysis to evaluate the effect of SC-CO<sup>2</sup> (Tamburini et al., 2013) and high-pressure carbon dioxide on food borne bacteria L. monocytogenes, Salmonella enterica, and E. coli (Garcia-Gonzalez et al., 2010). Conversely, the efflux pump was completely damaged or malfunctioning before the effect of SC-CO<sup>2</sup> on the culturability of S. enterica serotype Typhimurium cells became apparent (Kim et al., 2009).

Kramer and Muranyi (2014) also observed a significant discrepancy between conventional plate counts and different viability staining parameters, which showed that a pulsed light treatment against L. innocua DSM 20649 and E. coli DSM 498 did not cause an immediate shutdown of vitality functions even when the number of colony-forming units already decreased for more than 6 log<sup>10</sup> sample−<sup>1</sup> . Even if L. innocua DSM 20649 and S. aureus DSM 346 showed a same trend of concomitant depolarization and loss of respiration activity after a thermal treatment, colony counts were decreased only by less than 0.5 log for L. innocua whereas by more than 3 log for S. aureus (Kramer and Thielmann, 2016). A part of the heat induced de-energized L. innocua population was still able to recover and multiply, and an opposite trend was observed for S. aureus. These results highlighted residual activity of non-culturable bacteria and the strain-dependence of the response to an antimicrobial treatment. In the same study, these authors also observed a strong difference for the uptake of PI after thermal treatment for Gram negative bacteria, which were much more susceptible than Gram positive

#### TABLE 4 | CFU and MP-FCM results correlation and non-correlation.


PI, Propidium Iodide; cFDA, carboxyFluorescein DiAcetate; cFDA-AM, carboxyFluorescein DiAcetate-acetoxymethyl; DiBAC4, Bis-(1,3-DibutylBarbituric ACid) Trimethine Oxonol = bis-oxonol = BOX; EB, Ethidium Bromide.

ones. However, this difference was not reflected by the plate count results: this suggested thus that membrane disruption is not the primary mechanism of mild thermal inactivation of bacterial cells (Kramer and Thielmann, 2016). Even if MP-FCM is a very valuable rapid microbiological method for the assessment of functional properties of bacteria exposed to external stressors, classical plate count method can bring crucial additional information. In fact, reproductive growth requires stringent conditions including both metabolic activity and membrane integrity (Spilimbergo et al., 2010). Manoil et al. (2014) isolated by fluorescent-activated cell sorting an unknown injured population of Streptococcus mutans cells which were exposed to blue light-activated curcumin. Then they cultured this subpopulation on agar plates. They indicated that only 0.6% of this unknown injured population was able to grow, thus supporting the idea of injured cells losing their proliferation capacity in their case. Ananta et al. (2004) also showed that the occurrence of esterase activity did not correlate with viability according to plate enumeration. High hydrostratic pressure treated Lactobacillus rhamnosus GG cells still possessed residual esterase activity. This metabolic activity was not critical for the maintenance of viability.

The existence of differences between culturability and viability with silver as antimicrobial has already been reported (Martínez-Abad et al., 2012). But, conversely, with silver as antimicrobial, it was found that after a period where viable counts were not detected, bacterial populations recovered, and were able to proliferate in most cases. The resuscitation of the cultures was explained by both the existence of a resilient fraction of bacteria in a compromised state and the parallel inactivation of the silver species (Martínez-Abad et al., 2012).

Ayari et al. (2013) described the two cases (correlation and non-correlation) in the same study. For an intensive treatment, such as Bacillus cereus cells pre-treated with carvacrol combined with nisin and then exposed to sub-lethal radiation treatment (1 kGy), CFU results were in accordance with FCM analysis, whereas for single treatment with carvacrol and nisin, good agreement was not found. These results showed that the culturability of subpopulations with identical fluorescence characteristics depends on the treatments imposed to the cells. Therefore, a rapid loss of culturability is not necessarily correlated to the complete cell death (Ayari et al., 2013). In the same way, Hayouni et al. (2008) showed a high correlation between CFU and FCM results, except for Lactobacillus pentosus and Lactobacillus rhamnosus. This could be due to the distribution of dead cells in chains.

The presence of VBNC subpopulations was often as a function of the intensity of the treatment. For example, thermal treatments at different temperatures against L. rhamnosus resulted in different responses of the cell to PI/cFDA labeling (Ananta and Knorr, 2009). Exposure to 60◦C reduced considerably the cFaccumulation activity, however, there was no significant loss of membrane integrity. In contrast, when cells were subjected to 75◦C, PI uptake already occurred in the first 90 s. From this staining behavior, they concluded that at higher temperatures the primary target of lethal effect of heat was the bacterial cytoplasmic membrane.

Moreover the correlation seemed to depend on the physiological mechanism which is analyzed. Duarte et al. (2015) showed a lack of direct correlation between the percentage of depolarized cells and the time-kill curves results for Campylobacter spp. and Arcobacter butzleri after exposure to resveratrol inclusion complexes. The cell membrane depolarization could be only a transition state which would occur before membrane permeabilization (Hammer and Heel, 2012) and might be caused by several factors or antimicrobial agents. The cells in this state would have the capacity to regain culturability. Metabolic activity reduction may not be directly related to a decrease in the number of culturable cells (Ferreira et al., 2014).

# Microscopy

Transmission Electron Microscopy (TEM; Wu et al., 2010a; Ayari et al., 2013; Choi et al., 2013; Teng et al., 2014; Hong et al., 2015; Li H. et al., 2015; Coronel-León et al., 2016), Scanning Electron Microscopy (SEM; Spilimbergo et al., 2010; Ferreira et al., 2014; Surowsky et al., 2014; Hong et al., 2015; Li H. et al., 2015; Muriel-Galet et al., 2015) and fluorescence microscopy (Tamburini et al., 2013; Thabet et al., 2013; Fernandes et al., 2014; Hong et al., 2015; Li W. et al., 2015) could be used to observe cell interior structure, cell surface morphology and localize fluorescence compounds or components, respectively.

Boda et al. (2015) confirmed Syto9/PI double-staining FCM results with TEM micrographs showing the ruptures of E. coli cell walls following their exposure to a 4 Tesla pulsed magnetic field.

Spilimbergo et al. (2010) used scanning electron microscopy (SEM) to confirm a clear modification of S. cerevisiae cell structure due to CO<sup>2</sup> treatment. Untreated sample showed cells having a round shape with a smooth surface and no spot, while almost the totality of the cells of the treated sample presented an irregular shape with some dark points/zones, indicating a modification of the cell wall. As well, Muriel-Galet et al. (2015) observed L. monocytogenes and E. coli cell membranes by SEM: they concluded that the membrane of these bacteria is the main target of LAE <sup>R</sup> . Microscopic observations allow thus to gain further insights into the inactivation mechanism.

If the studied antimicrobial molecule can be labeled, microscopy methods can also be applied to visualize these molecules on bacterial cells as recently performed by (Li W. et al., 2015). Together with MP-FCM results, these authors could conclude that multimerization of the Chex-Arg20 antimicrobial monomer peptide to dimer and tetramer altered its mode of action against E. coli cells from non-lytic to a membrane disruptive capacity. To localize peptides in E. coli cells, Alexa-Fluor 647-labeled peptides were assembled and E. coli membranes were labeled with FM lipophilic styryl dye (FM, 4-64 FX). With high-resolution fluorescence microscopy, they observed that peptides localized primarily in the cytosol, while at higher peptide concentrations, peptides associated with both the cytosol, and membrane. It was clear that membrane interaction of the dimer and the tetramer induced membrane lysis, whereas the monomer did not.

# Other Techniques Spectroscopy

Fourier transform infrared spectroscopy (FTIR) permits to study the entire molecular composition of microbial cells, to reveal the biochemical composition of cellular constituents such as cell wall, membrane (phospholipid bilayer, peptidoglycan, lipopolysaccharides), and cytoplasm (fatty acids, water, nucleic acids, proteins, polysaccharides). Booyens and Thantsha (2014) observed a change in size and granularity of Bifidobacterium populations exposed to garlic clove extract by MP-FCM and completed this approach by using FTIR spectroscopy. In the presence of garlic, there was a decrease in lipid content of the membrane, which was an additional element that could best describe the state of the cells and the antimicrobial mechanism. Meng et al. (2016) revealed a change of the membrane phospholipid molecules of B. subtilis after an ultrahigh hydrostatic pressure and a mild heat (HPMH) treatment by FTIR spectroscopy. They passed from a liquid crystalline state to a gel state with a decrease in membrane fluidity. HPMH treatment decreased the α-helix content, while it increased the random coil content of the cellular proteins, which resulted in protein denaturation.

Tamburini et al. (2014) used the Nuclear Magnetic Resonance spectroscopy (NMR) to observe the changes in cell membrane lipid composition. Firstly, they established the phospholipid profile of E. coli K12 membranes and after SC-CO<sup>2</sup> treatment, they revealed that there were strong perturbations of membrane architecture namely on the two dominant phospholipid species (phosphatidyglycerol and phosphatidylethanolamine). This treatment had no detectable effect on cell density or granularity, whereas the cellular volume changed.

Circular Dichroism spectroscopy (CD) is used extensively to study chiral molecules particularly secondary structure or conformation of proteins sensitive to the environment, temperature, pH modifications. CD can be used to observe how secondary structure of proteins changes with environmental conditions or on interaction with other molecules. Teng et al. (2014) investigated change of the secondary structure of S. enteritidis genomic DNA after exposure to the avian defensin AvBD103b. S. enteritidis genomic DNA showed a typical negative peak and a positive peak around 240 and 270 nm, respectively. When the cells were treated with AvBD103b, the intensity of the DNA ellipticity became weaker. These results suggested that AvBD103b interacted with the S. enteritidis genomic DNA by changing the DNA conformation. With a gel retardation assay, they showed the insertion of the base pairs.

#### Vesicles as Membrane Models

Lee et al. (2015) elucidated the scolopendin 2 antimicrobial mechanism as a membrane-active mechanism leading to the formation of pores in microbial plasma membrane. Using giant unilamellar vesicles encapsulating calcein and large unilamellar vesicles containing fluorescein isothiocyanatedextran, which were similar in composition to typical E. coli O157:H7 and Candida albicans membranes, they demonstrated that scolopendin 2 disrupted membranes, resulting in a pore size between 4.8 and 5.0 nm. Membrane modeling appeared thus as a useful method to corroborate and complete the results obtained by flow cytometry. Grau-Campistany et al. (2016) and Wu et al. (2010a) also used lipid vesicles models of POPG (1-palmitoyl-2-oleoyl-glycero-sn-glycero-3-phospho-(1′ -rac-

glycerol)) or POPE/POPG (1-palmitoyl-2-oleoyl-sn-glycero-3 phosphoethanolamine/POPG) to mimic the Gram positive and Gram negative bacterial membrane. Grau-Campistany et al. (2016) correlated MP-FCM data with biophysical experiments in model membranes, with more permeabilization for E. coli than for S. aureus, corresponding to more leakage from POPE/POPG vesicles than pure anionic POPG vesicles at the same antimicrobial lipopeptide antibiotics concentration. Wu et al. (2010a) described S-thanantin antimicrobial mechanism against E. coli and Bacillus subtilis: first, it disrupted the membrane permeability, and then depolarized the cell membrane.

## Propidium MonoAzide Quantitative-Polymerase Chain Reaction (PMA-qPCR)

Propidium monoazide quantitative-Polymerase Chain Reaction (PMA-qPCR) is another viability test method that Ferrentino et al. (2015) used in parallel of a SYBR-PI double-staining. The apparent minor inactivation efficiency evaluated by PMAqPCR (about 2 log reduction) compared to MP-FCM could be due to incomplete exclusion of dead cells signals leading to false-positive signals, a known drawback of this technique (Ferrentino et al., 2015). The suppression of dead cells signals depends on the complexity of the sample matrix and on the length of DNA amplicon. Besides, PMA-qPCR is also the most sensitive quantification technique compared to MP-FCM and plate counts (Ferrentino et al., 2015). Tamburini et al. (2013) also used PMA-qPCR to assess the effect of SC-CO<sup>2</sup> treatment on L. monocytogenes, E. coli, and S. enterica cells, and their PMAqPCR and FCM results strongly correlated, even if they detected VBNC population. Indeed, a vast majority of cells remained in the partially permeabilized state.

# CONCLUSION

To investigate the efficacy of antimicrobial treatments against microbial cells and to elucidate their mechanism of action,

# REFERENCES

Ananta, E., Heinz, V., and Knorr, D. (2004). Assessment of high pressure induced damage on Lactobacillus rhamnosus GG by flow cytometry. Food Microbiol. 21, 567–577. doi: 10.1016/j.fm.2003.11.008

it is interesting to conduct MP-FCM which provides a rapid acquisition of information related to the physiological state of cells and to interpret precisely their survival modes. Nonetheless it has to be kept in mind, that the cellular vital properties are not measured directly, but through the distribution of fluorescent probes or the conversion of substrates. Therefore, the interpretation of data could be different between Gram negative and Gram positive bacteria as well as between bacteria and yeast. Controls must be performed with caution and conclusions should always be drawn carefully.

Additional methods such as microscopy, spectroscopy, membrane modeling, or molecular biology techniques could complete and corroborate MP-FCM results. Viability and culturability potential of microbial cells could be determined by MP-FCM and plate counts, respectively. Depending on the antimicrobial treatment, its intensity, the strains, and the target cell physiological state, viability and culturability can correlate or not. If there was no correlation, this revealed the presence of a compromised VBNC subpopulation which appeared as viable with MP-FCM results but unable to grow on agar plates. The characterization by MP-FCM of such injured bacteria in food or in clinical applications might be critical in terms of their potential capacity to excrete toxic or food spoiling metabolites, to transfer genes (Ananta et al., 2004; Schenk et al., 2011; Ayari et al., 2013; Hong et al., 2015). In food associated microbial communities, gene transfer can have direct implications for human health via the acquisition of new metabolic traits: substrate utilization, bacteriocin, exopolysaccharide and biogenic amine production, immunity to bacteriophages and antibiotic resistance (Kelly et al., 2009; Rossi et al., 2014). This phenomenon could contribute to the distribution of acquired traits to intestinal bacteria (Rossi et al., 2014).

This review of MP-FCM methodologies to assess antimicrobial mechanism gives access to a set of protocols and antimicrobial mechanisms of actions following different treatments descriptions. This could allow the development of such methodology and thus give further information about impact of antimicrobial compounds or physical treatments on microbial cells.

# AUTHOR CONTRIBUTIONS

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

# ACKNOWLEDGMENTS

The authors gratefully acknowledge the French National Agency for Research (ANR-14-CE20-0005-01 ACTIPHEN) for the financial support of this work.

Ananta, E., and Knorr, D. (2009). Comparison of inactivation pathways of thermal or high pressure inactivated Lactobacillus rhamnosus ATCC 53103 by flow cytometry analysis. Food Microbiol. 26, 542–546. doi: 10.1016/j.fm.2009.01.008

Antolinos, V., Esteban, M.-D., Ros-Chumillas, M., Huertas, J.-P., Periago, P. M., Palop, A., et al. (2014). Assessment of the of acid shock effect on viability of Bacillus cereus and Bacillus weihenstephanensis using flow cytometry. Food Res. Int. 66, 306–312. doi: 10.1016/j.foodres.2014.09.029


activity against Gram positive and Gram negative bacteria. Biochim. Biophys. Acta 1858, 333–343. doi: 10.1016/j.bbamem.2015.11.011


cerevisiae evaluated by flow cytometry. Process Biochem. 45, 647–654. doi: 10.1016/j.procbio.2009.12.013


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

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