# INNATE IMMUNITY IN AQUATIC VERTEBRATES

EDITED BY : Leon Grayfer, Stephanie DeWitte-Orr and Eva-Stina Isabella Edholm PUBLISHED IN : Frontiers in Immunology

#### Frontiers eBook Copyright Statement

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

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

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

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

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

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

ISSN 1664-8714 ISBN 978-2-88963-427-9 DOI 10.3389/978-2-88963-427-9

### 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

# INNATE IMMUNITY IN AQUATIC VERTEBRATES

Topic Editors:

Leon Grayfer, George Washington University, United States Stephanie DeWitte-Orr, Wilfrid Laurier University, Canada Eva-Stina Isabella Edholm, Arctic University of Norway, Norway

Citation: Grayfer, L., DeWitte-Orr, S., Edholm, E.-S. I., eds. (2020). Innate Immunity in Aquatic Vertebrates. Lausanne: Frontiers Media SA. doi: 10.3389/978-2-88963-427-9

# Table of Contents

	- *134 Studies Into ß-Glucan Recognition in Fish Suggests a Key Role for the C-Type Lectin Pathway*

Jules Petit, Erin C. Bailey, Robert T. Wheeler, Carlos A. F. de Oliveira, Maria Forlenza and Geert F. Wiegertjes

*146 Characterization and Transcript Expression Analyses of Atlantic Cod*  Viperin

Khalil Eslamloo, Atefeh Ghorbani, Xi Xue, Sabrina M. Inkpen, Mani Larijani and Matthew L. Rise


Chen Li, Jiaxin Liu, Xin Zhang, Shina Wei, Xiaohong Huang, Youhua Huang, Jingguang Wei and Qiwei Qin


Jassy Mary S. Lazarte, Kim D. Thompson and Tae Sung Jung

*223 Functional Characterization of Dark Sleeper (*Odontobutis obscura*) TBK1 on IFN Regulation*

Jian Chen, Zhuo Cong Li, Long Feng Lu, Pei Li, Xi-Yin Li and Shun Li


Xiaoxue Yin, Liangliang Mu, Jun Li and Jianmin Ye

*262 Effects of Live Attenuated Vaccine and Wild Type Strains of* Edwardsiella ictaluri *on Phagocytosis, Bacterial Killing, and Survival of Catfish B Cells* Adef O. Kordon, Safak Kalindamar, Kara Majors, Hossam Abdelhamed, Wei Tan, Attila Karsi and Lesya M. Pinchuk

# Editorial: Innate Immunity in Aquatic Vertebrates

#### Stephanie DeWitte-Orr <sup>1</sup> , Eva-Stina Edholm<sup>2</sup> and Leon Grayfer <sup>3</sup> \*

*<sup>1</sup> Department of Biology, Wilfrid Laurier University, Waterloo, ON, Canada, <sup>2</sup> Faculty of Biosciences, Fisheries and Economics, Norwegian College of Fishery Science, University of Tromsø-The Arctic University of Norway, Tromsø, Norway, <sup>3</sup> Department of Biological Sciences, George Washington University, Washington, DC, United States*

Keywords: innate immunity, fish, amphibia, antiviral, Research Topic

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

#### **Innate Immunity in Aquatic Vertebrates**

The alarming growth of the world's population is putting ever-greater demand on agricultural industries and is manifesting in environmentally detrimental consequences. While the development of better aquaculture practices presents a promising means of meeting the needs of this population growth, overcrowding in aquaculture, climate change, and habitat destruction are resulting in the emergence of new and opportunistic infections within farmed and wild aquatic vertebrate species, often to the detriment of these animals. The prevention and circumvention of these infections and die-offs requires much greater understanding of the mechanisms by which these animals' immune systems develop, recognize, and respond to distinct pathogens. Indeed, we already know that while animals like cartilaginous/bony fish and amphibians exhibit hallmark components associated with mammalian immunity, they also provide examples of novel strategies for immune cell development and antimicrobial defenses. As these organisms possess arguably less developed adaptive immune responses, they rely more heavily on their innate immunity to control infiltrating pathogens. In turn, these animals reside in vastly distinct environments to those within which (the much more extensively characterized) mammalian immune system has evolved, so it is not surprising that aquatic vertebrates possess many intriguing immunological differences from terrestrial animals.

It is by gaining greater insight into these immune processes that we may hope to better our aquacultural practices and combat the devastating effects of human activities on aquatic animal communities around the globe. Toward this end and through this collection of 17 articles, which include both original research as well as comprehensive reviews, we coalesce recent advances in the current understanding of the innate immune responses of aquatic vertebrates.

### ANTIFUNGAL DEFENSES

Aquatic vertebrates reside within pathogen-rich environments, with their skin mucosa representing an important barrier to these pathogens, but also a means of pathogen entry. The global amphibian declines resulting from the Batrachochytrium dendrobatidis and Batrachochytrium salamandrivorans chytrid fungi infections of amphibian skins is an important example of this skin mucosa-pathogen interface. The comprehensive review by Varga et al. underlines the importance of the amphibian skin as an innate immune barrier to aquatic pathogens, discusses the anatomy and cell (immune and non-immune) composition to of the amphibian skin, and focuses on the skin pattern recognition receptors (PRR) and antimicrobial peptide responses therein.

#### Edited and reviewed by:

*Geert Wiegertjes, Wageningen University & Research, Netherlands*

\*Correspondence: *Leon Grayfer leon\_grayfer@gwu.edu*

#### Specialty section:

*This article was submitted to Comparative Immunology, a section of the journal Frontiers in Immunology*

Received: *21 November 2019* Accepted: *03 December 2019* Published: *19 December 2019*

#### Citation:

*DeWitte-Orr S, Edholm E-S and Grayfer L (2019) Editorial: Innate Immunity in Aquatic Vertebrates. Front. Immunol. 10:2959. doi: 10.3389/fimmu.2019.02959*

**5**

In turn, Grogan et al. provide an extensive overview of the documented and anticipated amphibian immune responses against chytrid pathogens, covering topics such as the determinants of skin anti-fungal protection, constitutive skin immune defenses, innate immune recognition, and the ensuing innate immune and adaptive immune responses to fungal pathogens. Grogan et al. evaluate and discuss the presumed and potential roles of pathogen detection, immune suppression, fungal immune evasion, immunological successes, and possible failures as well as immunopathology in the context of chytridiomycosis.

### PATHOGEN RECOGNITION RESPONSES

Aquatic animals are subject to very different pathogen pressures to those that have shaped the terrestrial immune response, and yet many aspects of their innate immune armamentarium are conserved. While these animals possess many of the same PRR genes as terrestrial mammals, they also encode species-specific pathogen receptors and may well-utilize the mammalian PRR homologs in distinct ways.

As an example of the above and unlike mammals, aquatic animals are notoriously insensitive to the lipopolysaccharide (LPS) and presumably have evolved distinct/complementary means for LPS detection. In this respect, Bi et al. demonstrate that the nucleotide-binding oligomerization domain-containing protein 1 (NOD-1), which is best known as a receptor for intact bacteria-derived peptidoglycan; in fish may also serve as a means for recognizing intracellular LPS, resulting in the canonical activation of NF-κB signaling pathway and the ensuing proinflammatory response.

Across vertebrates, β-glucan carbohydrates present on the surfaces of an array of pathogens also represent important PRR ligands and therefore a means of pathogen recognition. While the mammalian Dectin-1 receptor (member of C-type lectin receptor family; CLR) is the best characterized β-glucan PRR, this gene has to date not been clearly annotated in fish genomes, although fish such as carp have been shown to recognize this pathogen associated molecular pattern (PAMP). Petit et al. demonstrate that in response to β-glucan stimuli, common carp macrophages undergo cell signaling pathway that are characteristic of CLR activation. Moreover, using a number of bioinformatics approaches, this study identifies several putative carp CLR- β-glucan receptors, some of which possess gene synteny and structural similarities to the mammalian Dectin-1. This presumably highlights both the convergence and the diverged evolution of the fish and terrestrial mammal innate immune pathogen recognition.

### GRANULOCYTE DEVELOPMENT AND RECRUITMENT

In terrestrial mammals, granulocytes are amongst the first cells to respond to infiltrating pathogens as well as the most represented immune populations in circulating blood. While the kinetics of the aquatic vertebrate immune infiltration of infected tissues appear to correspond to those of mammals, the mechanisms by which fish and frogs generate and recruit their granulocyte populations differ from what is seen in mammals.

Where the granulocyte colony-stimulating factor (G-CSF) is the principal driver of granulopoiesis, it is interesting to consider that while mammals possess a single G-CSF, teleosts encode multiple G-CSF isoforms. Intriguingly, Katakura et al. demonstrate the presence in the common carp genome of four G-CSF paralogs (g-csfa1 and g-csfa2; g-csfb1 and g-csfb2), which exhibit distinct expression across fish tissues, leukocytes, and following immune stimulation. Moreover, while the common carp G-CSFa1 and G-CSFb1 both elicit neutrophil chemotaxis and proliferation of kidney cells, only G-CSFb1 promotes neutrophil-lineage differentiation of head kidney cells.

In turn, while mammals possess a single CXCL8 chemokine bearing the ELR motif, characteristic of pro-inflammatory granulocyte chemokines, Koubourli et al. demonstrate that the amphibian Xenopus laevis encode two CXCL8s, one of which possesses the ELR motif and appears to be involved in inflammatory responses, and the other lacking this motif and being involved in the recruitment of healing/immunosuppressive granulocytes.

### ANTIVIRAL IMMUNITY

Aquatic animals are important models for the study the converged and divergent evolution of vertebrate innate and antiviral immunity. As the interferon (IFN) cytokine responses represents the cornerstone of vertebrate antiviral defenses, it is exciting to consider that while the emergence of type III IFN responses was thought to emerge with tetrapods, Redmond et al. show that cartilaginous fish encode both type I and type III IFNs, thus instead suggesting the loss of this cytokine family in bony fish and its reemergence in amphibians.

Aquatic habitats teem with viral pathogens so it is perhaps not surprising that aquatic vertebrates have evolved elaborate antiviral defenses, several of which are discussed here. Amongst these, Lazarte et al. comprehensively review the current understanding of the fish Mda5 antiviral PRR and its roles in fish recognition of intracellular viral and bacterial pathogens, the initiation of the fish type I IFN response and the consequences of the activation of this receptor to bony fish immunity. Chen et al. report on the characterization of a fish TANKbinding kinase 1, which appears to be an important regulator of fish IFN response. Xu et al. report on a fish-specific PKR analog, protein kinase Z, which activates a number of hallmark antiviral signaling components and elicits the expression of IFN. Eslamloo et al. characterize the cod Viperin antiviral effector gene, model its protein architecture in comparison to mammalian Viperins and examine cod Viperin expression during cod development, following immune stimulation of cod macrophages and in conjunction with a panel of immune inhibitors, thereby elucidating possible regulatory pathways for this gene. Zhang et al. report on the characterization of the grouper cholesterol 25-hydroxylase (CH25H) IFN-induced gene including in silico, expression, subcellular localization, and functional analyses of the grouper CH25H in the context of Singapore grouper iridovirus and red-spotted grouper nervous necrosis virus infections. Lastly, Li et al. describe the antiviral roles of the orange-spotted grouper autophagy-related gene-5 (Atg5) in the context of red-spotted grouper nervous necrosis virus and Singapore grouper iridovirus infections.

### INNATE-LIKE B CELLS

Teleost fish appear to possess greater numbers of innate-like phagocytic B cells than mammals and thus, understanding the roles of these cells during immune responses and how they are affected by vaccination is key to better fish vaccine development. To this end Wu et al. show that while the Nile tilapia IgMlo B cells (resembling plasma-like cells) possess decreased phagocytic activity compared to the naïve/maturelike IgMhi B cells, suggesting that B cell differentiation may cause the decrease in phagocytic capacities of bony fish B cells. In turn, this work may indicate that the evolution of more specialized (further differentiated) B cell responses in mammals compared to bony fish, may have come at the expense of decreased phagocytic capacities of these cells, in favor of antibody-production. Concurrently, Kordon et al. challenge catfish kidney-derived B cells with wild type and vaccine strains of Edwardsiella ictaluri, showing that both bacterial strains are phagocytosed by the B cells, eliciting antimicrobial activity but also inducing apoptosis in these fish B cells.

### FISH HEALTH AND VACCINATION

While nanoparticles are being increasingly utilized in many industries, the consequences of their bioaccumulation within aquatic environments remains poorly addressed. Accordingly, Torrealba et al. review the use of fish innate immune markers as an auspicious means for assessing fish health following nanoparticle exposure.

Greater insights into aquatic animal immune responses lead to the development of better vaccination strategies for these animals and several of the manuscripts in this Research Topic exemplify this notion. For example, Braden et al. have utilized the previously documented salmon immune responses to Aeromonas salmonicida spp. salmonicida (Asal) and several vaccines to demonstrate in the Arctic charr (an emerging aquacultural species) that efficacies of vaccine-based protection against Asal depend on the upregulation and control of fish baseline humoral responses, including factors such as complement and coagulation factors, acute phase-proteins, and iron hemostasis proteins.

### CONCLUDING REMARKS

The primary articles and reviews featured in this Research Topic are great examples of the exciting new research being conducted on innate immunity of aquatic vertebrates. With every new article, we gain greater understanding of the interesting and often unique mechanisms governing these animals' antimicrobial defenses. In turn, these studies will pave the way toward the development of better aquacultural practices, aquatic habitat preservation and remediation as well as a deeper understanding of the evolution of vertebrate immune responses.

## AUTHOR CONTRIBUTIONS

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

### FUNDING

SD-O acknowledges support from the Natural Sciences and Engineering Research Council of Canada. E-SE acknowledges support in the way of a Tromsø Research Foundation starting grant. LG acknowledges support from the National Science Foundation (NSF) (IOS: 1749427).

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

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

# Recognition of Lipopolysaccharide and Activation of NF-**κ**B by Cytosolic Sensor NOD1 in Teleost Fish

*Dekun Bi 1,2,3, Yue Wang4 , Yunhang Gao5 , Xincang Li <sup>4</sup> , Qing Chu1,2,6, Junxia Cui 1,2,3 and Tianjun Xu1,2,3,6\**

*1Key Laboratory of Exploration and Utilization of Aquatic Genetic Resources, Shanghai Ocean University, Ministry of Education, Shanghai, China, 2National Pathogen Collection Center for Aquatic Animals, Shanghai Ocean University, Shanghai, China, 3 Laboratory of Fish Biogenetics & Immune Evolution, College of Marine Science, Zhejiang Ocean University, Zhoushan, China, 4East China Sea Fisheries Research Institute, Chinese Academy of Fishery Sciences, Shanghai, China, 5College of Animal Science and Veterinary Medicine, Jilin Agriculture University, Changchun, China, 6 International Research Center for Marine Biosciences at Shanghai Ocean University, Ministry of Science and Technology, Shanghai, China*

Lipopolysaccharide (LPS) is the major component of the outer membrane of Gramnegative bacteria. This molecule can induce strong immune response and various biological effects. In mammals, TLR4 can recognize LPS and induce inflammatory response. However, the innate receptor in fish for recognizing LPS remains ambiguous. LPS can invade the cytoplasm *via* outer membrane vesicles produced by Gram-negative bacteria and could be detected by intracellular receptor caspase-11 in mammals, so, there may also exist the intracellular receptors that can recognize LPS in fish. NOD1 is a member of NOD-like receptors family and can recognize the iE-DAP in the cytoplasm in mammals. In fish, NOD1 can also respond to infection of Gram-negative bacteria and may play an important role in the identification of bacterial components. In this study, to study whether NOD1 is a recognition receptor for LPS, we detected the expression of NOD1 and several cytokines at transcript levels to determine whether LPS can induce inflammatory response in teleost fish and NOD1 can respond to LPS. Then, we perform the binding analysis between NOD1 and ultrapure LPS by using Streptavidin pulldown assay and enzyme-linked immunosorbent assay to prove that NOD1 can be combined with LPS, and using dual luciferase reporter gene assay to verify the signal pathways activated by NOD1. Next, through cell viability analysis, we proved that LPS-induced cytotoxicity can be mediated by NOD1 in fish. The results showed that NOD1 can identify LPS and activate the NF-κB signal pathway by recruiting RIPK2 and then promoting the expression of inflammatory cytokines to induce the resistance of organism against bacterial infection.

Keywords: NOD1, lipopolysaccharide, RIPK2, NF-**κ**B, teleost fish

### INTRODUCTION

Lipopolysaccharide (LPS) is a heat-stable endotoxin and is main constituent of the outer membrane of Gram-negative bacteria. This molecule has long been considered as a significant factor in septic shock (septicemia) in humans and can induce strong response from normal animal immune systems (1, 2). In mammals, LPS has been exclusively identified by TLR4 under the participation of myeloid differentiation protein 2 (MD2), LPS binding protein (LBP), and cluster of differentiation 14 (CD14) at the cell surface (3). After distinguishing LPS, TLR4 can activate certain signaling pathways, such as

#### *Edited by:*

*Leon Grayfer, George Washington University, United States*

#### *Reviewed by:*

*Victoriano Mulero, Universidad de Murcia, Spain Katherine Buckley, Carnegie Mellon University, United States*

> *\*Correspondence: Tianjun Xu tianjunxu@163.com*

#### *Specialty section:*

*This article was submitted to Comparative Immunology, a section of the journal Frontiers in Immunology*

*Received: 18 March 2018 Accepted: 06 June 2018 Published: 26 June 2018*

#### *Citation:*

*Bi D, Wang Y, Gao Y, Li X, Chu Q, Cui J and Xu T (2018) Recognition of Lipopolysaccharide and Activation of NF-κB by Cytosolic Sensor NOD1 in Teleost Fish. Front. Immunol. 9:1413. doi: 10.3389/fimmu.2018.01413*

**8**

MyD88-dependent or MyD88-independent pathways. In MyD88 dependent pathways, TLR4 recruits MyD88 to transfer the signal and activate the NF-κB signaling pathway (4, 5). However, in MyD88-independent pathways, MD-2 (the polymer of LPS) and TLR4 forms an endosome to enter the cytoplasm and then recruits TRIF to transfer the signal induce inflammatory response (6, 7). Recent study has shown that Gram-negative bacteria can secrete outer membrane vesicles (OMVs) as an intermediary to deliver LPS into the cytosol. During Gram-negative bacteria infection, OMVs can be internalized *via* endocytosis, which then release the LPS into the cytosol from early endocytic compartments (8). Furthermore, previous studies showed that the receptors that can distinguish the LPS and activate innate immunity also exist in cytoplasm, such as caspase-4/5/11. These receptors are considered as intracellular receptors for cytosolic LPS (8–10).

Frequent outbreak of bacterial diseases in fish aquaculture results in great economic loss and is a major factor that restrains aquaculture development. Among these microbes, Gramnegative bacteria are the main pathogenic bacteria that can cause various diseases (11). As a major component of the outer membrane of Gram-negative bacteria, LPS also exhibits various biological effects. For example, LPS can induce the expression of cytokines and acute-phase proteins, and it can play an important role in pathological, neuro-immunological, and immunoendocrinological activities in a variety of fish (12). Thus, receptors that can identify LPS should be studied extensively. TLR4 is the specific receptor in mammals for identifying LPS. However, previous studies have found that except zebrafish and several other cyprinidae, for example, rare minnow and common carp, most fish species including miiuy croaker do not exist TLR4 orthologs (13, 14). In addition, in mammals, TLR4 needs to form a tripolymer along with MD2 and the leucine-rich repeat (LRR) protein CD14 to identify LPS; however, all fish genomes lack these costimulatory molecules, MD2 and CD14 (15, 16). So, TLR4 in fish does not recognize the stimulation of LPS. And later studies on zebrafish TLR4 have showed that zeTLR4 cannot recognize LPS (15, 16), which directly confirms the above viewpoint. Multiple TLRs have been identified in fish and thorough analysis of the role of TLRs showed that other TLRs could also not identify LPS (15). Because of the limitations in research methods, the role of immune genes in fish was rarely investigated, thus the receptors that can recognize LPS in fish are also not clear.

NOD1 is a member of the NOD-like receptor (NLR) family, which was encoded by *CARD4* genes. This receptor is composed of an N-terminal effector binding domain (CARD), a central nucleotide oligomerization NACHT domain, and a C-terminal LRR domain (17). NOD1 is widely distributed in various tissues (18) and exists in a wide range of species. Previous studies have shown that NOD1 exists in a variety of bony fishes, such as zebrafish (19), goldfish (20), grouper (21), and miiuy croaker (22). And it also showed that miiuy croaker NOD1 is highly homologous in structure and sequence compare with many fish and mammalian species. In miiuy croaker, NOD1 can expression in all tissues, especially the high expression of NOD1 was observed in liver and skin (22). Intensive studies in mammals and fish showed that NOD1 can recognize the G-d-glutamyl-meso-diaminopimelic acid (iE-DAP) moieties that are derived from Gram-negative bacteria and then activates NF-κB signaling pathway by recruiting RIPK2 to induce inflammatory reaction (23–26). Furthermore, studies have also found that both PGN and LPS, the pathogenic components of Gram-negative bacteria, can induce significant expression of inflammatory cytokines in teleosts (27, 28). iE-DAP as the composition of PGN, it can be recognized by NOD1, so LPS may also be recognized by NOD1 in teleost fish.

Gram-negative bacteria are the main pathogenic bacteria in the majority of fish, and LPS is an important pathogenic component of these bacteria. Thus, analyzing the receptors that can detect LPS is important to study fish diseases. In this study, we use miiuy croaker (*Miichthys miiuy*) as a model fish species to study the recognition between NOD1 and LPS because of the extensive background of this species in immunology research (29–32). The results showed that both Gram-negative bacterial infection and LPS stimulation can induce inflammatory response in teleost fish, NOD1 can also respond to LPS stimulation and Gram-negative bacterial infection. The expression of inflammatory cytokines will be markedly inhibited after knockdown of NOD1 gene. Overexpression of NOD1 can activate NF-κB signal pathway, and stimulation of cells with LPS, which were overexpression of NOD1 can significantly promote the expression of NF-κB. The results of streptavidin pulldown assay and enzyme-linked immunosorbent assay (ELISA) showed that LPS can bind with NOD1 protein. Overexpression of mutant NOD1 plasmid could not activate NF-κB; simultaneous overexpression of NOD1 and RIPK2 plasmids resulted in more evident activation of NF-κB, and immunoprecipitation analysis showed that NOD1 can interact with RIPK2. These results illustrate that NOD1 can identify LPS and activate NF-κB signal pathway by recruiting RIPK2 to promote the expression of inflammatory cytokines.

### MATERIALS AND METHODS

### Preparation of Tissue and Macrophage

Healthy miiuy croakers (750 ± 20 g) were obtained from Zhoushan Fisheries Research Institute (Zhejiang, China) and cultured in aerated seawater tanks at 25°C for a week. To obtain the infected tissues, the healthy fish were randomly divided into two groups: those in the experimental group were intraperitoneally injected with 1 ml *Vibrio anguillarum* (1.5 × 108 CFU/ml), *Vibrio harveyi* (1.5 × 108 CFU/ml), *Staphylococcus aureus* (1.5 × 108 CFU/ml), and LPSs derived from *Escherichia coli* 055:B5 (1 mg/ml) and those in the control group were injected with 1 ml of physiological water. Then, the fish were dissected at different times, and the liver tissues were collected from three individual at each times.

To separate and obtain the macrophages, head kidney tissues from healthy miiuy croakers were collected and chopped, next, conduct sterile filtration by using cell filter with 100 µm pore size in L-15 medium, which was contained 2% FBS, penicillin (100 IU/ml), streptomycin (100 µg/ml), and heparin (20 U/ ml). Then, the cell suspension was added into 51% Percoll (Pharmacia, USA) separating medium and centrifuged at the condition of 400 *g* at 4°C for 40 min. Next, the supernatant was removed and the cells were collected at interface, washed the cells twice with L-15 medium, and seeded in a 6-well plate at a density of about 4 × 107 per well, the cells were then cultured in the incubator at 26°C with 4% CO2. After overnight culture, replace the medium with fresh L-15 medium, which contained 20% FBS. The cells were treated with ultrapure LPS-B5 (3 µg/ ml, tlrl-pb5lps, InvivoGen), ultrapure LPS-EK (3 µg/ml, tlrlpeklps, InvivoGen), Lipoteichoic acid (LTA, 1 µg/ml, L3265, Sigma), Zymosan A (25 µg/ml, Z4250, Sigma), poly(I:C) (10 µg/ml, tlrl-picw, InvivoGen), and infected with SCRV at a multiplicity of infection (MOI) of 5, and then the cells were collected at different times. The cells without treated with any pathogenic component as the control, and each experiment will perform three biological replicates. This study was carried out in accordance with the recommendations of National Institutes of Health's Guide for the Care and Use of Laboratory Animals. The study protocol was approved by the Research Ethics Committee of the Shanghai Ocean University (SHOU-DW-2018-047).

### Real-Time Quantitative PCR Analysis

To perform Real-time Quantitative PCR analysis, firstly, TRIzol reagent (Invitrogen) was used to extract the total RNA from macrophages and tissues, and then FastQuant RT Kit (Tiangen) was used to perform reverse transcription to avoid genomic contamination. Next, we designed the specific primers to detect the expression of NOD1, TNFα, IL-1β, IL-6, IL-8, and IFNβ in miiuy croaker, the expression of β-actin as an internal control. Using SYBR® Premix Ex Taq™ (Takara) and 7300 real-time PCR system (Applied Biosystems, USA) to perform real-time quantitative PCR, the mixture of amplification containing 10 µl SYBR Premix (2×), 0.4 µl ROX Dey (50×), 0.8 µl of each primer (10 µM), 2 µl cDNA template, and 6 µl ddH2O. The conditions of cycle were 30 s at 95°C, followed by 40 cycles at 95°C for 5 s, and at 60°C for 34 s. Then, the dissociation curve was performed to determine the target specificity after each analysis. The triplicate experiments were performed for each sample and all the primers are listed in **Table 1**.

### Plasmid Construction

Miiuy croaker NOD1 (GenBank accession No. KP715094.1) was amplified from total cDNA by using a pair of primers with the HA tag, which were then digested with *Kpn* I and *Xba* I restriction


endonucleases (Takara) and the products were connected to the vector of pcDNA3.1 (Invitrogen) between the endonucleases sites of *Kpn* I and *Xba* I. The plasmids encoding miiuy croaker TLR5s in the pFLAG-CMV-3 vector was constructed as described in Ref. (33). We designed specific primers to amplify miiuy croaker RIPK2 and NOD1 from miiuy croaker cDNA, then digest the DNA fragments, and insert into pcDNA3.1-flag and pEGFP-N1 vectors, respectively. The methods of double enzyme digestion and sequencing were used to validate the recombinant plasmids. Based on the recombinant plasmid, we designed several pairs of primers to amplify the miiuy croaker NOD1, which were deleted in different domains to construct the mutant plasmids, and named as NOD1ΔCARD, NOD1ΔLRR1, NOD1ΔLRR2-4, NOD1ΔLRR5-7, and NOD1ΔLRR. For expression of NOD1-LRR protein in bacteria, an expression system harboring the desired expression vector was constructed. We have designed a pair of primers to amplify the encoding LRRs domain of NOD1. The harvested DNA fragment was digested and inserted into pET-32a expression vector. To carry on the promoter activity analysis, NF-κB luciferase reporter plasmid was purchased from Promega and ISRE luciferase reporter plasmid was purchased from Stratagene. Endotoxin-Free Plasmid DNA Miniprep Kit (Tiangen) was used to extract the plasmids. All of the primers are listed in **Table 1**.

### Cell Culture, Transfection, and Luciferase Reporter Assays

Miiuy croaker kidney cell lines (MKC) were cultured in L-15 medium supplemented with 15% FBS (Gibco), 100 U/ml penicillin, 100 µg/ml streptomycin at 26°C. Epithelioma papulosum cyprini (EPC) cell lines were cultured in 199 medium that contain 10% FBS, 100 U/ml penicillin, and 100 µg/ml streptomycin under the humidified condition, at 26°C, and 5% CO2. HEK293 cell lines were cultured in DMEM high glucose medium that contain 100 U/ml penicillin, 100 µg/ml streptomycin, 10% FBS, and 2 mM l-glutamine under humidified condition, at 37°C, with 5% CO2.

Before transient transfection, cells were seeded into 24- or 12-well plates and incubated overnight. When the cell density reached about 80% of the cell culture plate, the plasmids were transfected into cells. NOD1 expression plasmid and NF-κB or ISRE luciferase reporter plasmid were co-transfected into cells to verify the role of NOD1 through luciferase reporter gene assay. After co-transfection of NOD1 expression plasmid and NF-κB reporter plasmid, the cells were stimulated with ultrapure LPS (2 µg/ml) to validate whether LPS can be detected by NOD1. The NOD1 plasmid and 100 nM NOD1-siRNA were co-transfected into cells to perform NOD1 knockdown experiment, and the NOD1 plasmid and RIPK2 plasmid along with NF-κB reporter plasmid were co-transfected to verify the interaction between NOD1 and RIPK2. The mutant plasmids of NOD1 and NF-κB reporter plasmid were co-transfected to check the role of different domain. Renilla luciferase reporter (pRL-TK, Promega) plasmid was used as the internal control and lipofectamine 2000™ reagent (Invitrogen) was used as the transfection reagent. The concentration of plasmid solution was tested by Nanodrop 2000 spectrophotometer (Thermo scientific), and all the experiments were repeated three times.

### RNA Interference

Miiuy croaker NOD1-specific siRNAs, RIPK2-specific siRNA, and miiuy croaker NOD2-specific siRNA were designed by RNAi Target Sequence Selector website (Clontech). The sequence of NOD1 siRNA1 was 5′-GAGAAAGGUGAUCAGGAAGTT-3′ (sense) and 5′-CUUCCUGAUCACCUUUCUCTT-3′ (antisense); the sequence of NOD1-siRNA2 was 5′-GGUUAACACAGAUCCCA UCTT-3′ (sense) and 5′-GAUGGGAUCUGUGUUAACCTT-3′ (antisense); the sequence of NOD1-siRNA3 was 5′-ACGAAAG UCUGGGCUUCUUTT-3′ (sense), and 5′-ACGAAAGUCUG GGCUUCUUTT-3′ (antisense). The sequence of RIPK2-siRNA was 5′-CCAUCAAGUGCCUGAAACUTT-3′ (sense) and 5′-AG UUUCAGGCACUUGAUGGTT-3′ (antisense). The sequence of NOD2-siRNA was 5′-GCUCGACCUGGUUUAUACATT-3′ (sense) and 5′-UGUAUAAACCAGGUCGAGCTT-3′ (antisense). The negative control-siRNA sequence was 5′-UUCUCCGAA CGUGUCACGUTT-3′ (sense) and 5′-ACGUGACACGUUCG GAGAATT-3′ (antisense). Miiuy croaker macrophages were inoculated to 24-well plates and cultivated overnight, then transfected with 100 nM-specific siRNA into cells by using Lipofectamine 2000™ for 24 h, the cells that transfected with negative control-siRNA were used as the mock control, and then the cells were stimulated with ultrapure LPS.

### Western Blotting

To detect the expression of target gene, HEK293 cells were transfected with NOD1 expression plasmid or other expression plasmids, and after transfection of 48 h, the cells were collected using 1 × SDS loading buffer and gel electrophoresis was performed, then the semi-dry process (Bio-Rad Trans Blot Turbo System) was used to transfer the protein from gel to PVDF membrane. After blocked by 5% evaporated milk, the membrane was incubated overnight in anti-HA monoclonal antibody (Sigma) and incubated 90 min in secondary antibody that conjugated with horseradish peroxidase (Beyotime). Then, immunoreactive proteins were detected using the BeyoECL Plus (Beyotime). The digital imaging was performed by using the cold CCD camera.

### Purification of Recombinant Proteins

For NOD1-LRR protein expression in bacteria, the recombinant plasmid of NOD1-LRR was transformed into competent *E. coli* Rosetta (DE3) cells for overexpression. After induction expression with a final concentration of 0.4 mM isopropyl-β-d-thiogalactoside at 28°C for 12 h, the bacteria pellets were collected by centrifugation and resuspended in PBS containing 1% Triton X-100 for probe sonication lysis. The recombinant NOD1-LRR protein was purified using Ni-NTA His Bind Resin (QIAGEN) according to previously performed methods (34). Besides, thioredoxin (TRX) with 6 × His-tag encoded by parent vector pET-32a in *E. coli* was also expressed and used as the control.

### Streptavidin Pulldown Assay

To certify that LPS can bind to NOD1 in cells, the cells were transfected of indicated HA tagged NOD1 expression plasmids or flag-tagged TLR5s expression plasmids. After 36 h of transfection, cells were stimulated with 2 µg biotinylated ultrapure LPS that was derived from *E. coli*, O111:B4 strain (LPS-EB Biotin, tlrl-3blps, InvivoGen), and after 12 h of stimulation, cells were collected and lysed in a buffer that contain 20 mM Tris (pH7.5), 150 mM NaCl, 1% Triton X-100, and multiple protease inhibitors for 15 min. Protein extracts were incubated with 50 µl Streptavidin MagneSphere® Paramagnetic Particles (SA-PMPs, Promega) for 60 min at 4°C. Unconjugated ligands were eliminated by washing the SA-PMPs-ligands complexes three times. And then, the precipitates were treated with 50 µl 1× SDS loading buffer and boiled at 95°C for 5 min followed by immunoblotting analyses.

### Binding Activity With LPS and iE-DAP

Enzyme-linked immunosorbent assay was conducted to evaluate the binding ability of NOD1-LRR to ultrapure LPS-B5, ultrapure LPS-EK, Synthetic Lipid A, iE-DAP (γ-d-Glu-mDAP, tlrl-dap, Invivogen), and β-1,3-Glucan (89862, Sigma). Each well of a microtiter plate was coated with 100 µl of 20 µg/ml ultrapure LPS, Lipid A, iE-DAP, or β-1,3-Glucan, and then incubated overnight at 37°C following a previously described protocol (35). Wells incubated with 100 µl of 50 mM Tris–HCl were used as negative control. Each well was blocked with BSA (2 mg/ml, 100 µl) for 2 h at 37°C, and then washed four times with TBST (0.05% Tween 20 in TBS). Subsequently, a series of diluted NOD1-LRR or TRX protein (0–0.6 µM in TBS containing 0.1 mg/ml BSA) were added. After incubation with recombinant protein for 3 h at room temperature, plates were rinsed four times with TBST and incubated with peroxidase-conjugated mouse monoclonal anti-His Tag antibody (1:5,000 dilution in TBS with 0.1 mg/ml BSA) at 37°C for 2 h. After rewashing four times with TBS, the plate was developed with 0.01% 3,3′,5,5′-tetramethylbenzidine (Sigma). The reaction was stopped with 2 M H2SO4, and absorbance was read at 450 nm wavelength. All assays were conducted in quintuplicate.

### Cell Viability Analysis

Adenosine triphosphate (ATP) is an important energy source for all organisms, and it can be used as an indicator of cell activity. To verify whether NOD1 can mediate cytotoxicity induced by LPS, EPC or HEK293 cells were transfected with miiuy croaker NOD1 expression plasmids or specific siRNAs and stimulated with 1 µg LPS-B5 or LPS-EK, and after stimulation of 12 h, the cells were collected to detect the concentration of ATP in the cells, then the concentration of ATP was used to represent the activity of cells. ATP bioluminescence assay kit (Promega) was used to collect the cells and GloMax 20/20 Luminometer (Promega) was used to detect the fluorescence intensity.

### Immunoprecipitation

To validate the interaction between NOD1 and RIPK2, HA-tagged NOD1 expression plasmids and flag-tagged RIPK2 expression plasmids were co-transfected into HEK293 cells. After 48 h of transfection, the cells were lysed and the protein extracts were immunoprecipitated with anti-flag antibody and Protein A + G Agarose beads overnight at 4°C on a rocker. Then, the complexes were washed three times, and 50 µl 1× SDS loading buffer was added to boil at 95°C for 5 min followed by immunoblotting analyses.

### Immunostaining and Confocal Imaging

Hela cell lines were purchased from the ATCC (Manassas, VA, USA) and cultured in DMEM high glucose medium. Hela cells were plated onto coverslips and incubated overnight, then NOD1-GFP expression plasmids and flag tagged RIPK2 expression plasmids were co-transfected into the cells. After 48 h of transfection, the cells were fixed with immunostaining fixative (Beyotime) for 30 min, washed three times with PBS, and blocked with immunostaining blocking buffer (Beyotime) for 60 min. Then, the cells were incubated in anti-flag antibody (Sigma) overnight at 4°C, followed by incubation with Cy3-conjugated anti-mouse IgG secondary antibody (Sigma) and added with the anti-fluorescence quenching reagent (Beyotime). The images were obtained with Leica TCS SP5 confocal system (Leica) equipped with 63× objective.

### Statistical Analysis

The data on relative gene expression were obtained by using the 2−ΔΔCt method, and comparisons between different groups were made by one-way ANOVA Kruskal–Wallis test with Dunn's multiple comparison test (36). All the data were represented in the form of mean ± SE, and the significant differences between different experimental groups were testified by using two-tailed Student's *t*-test, and it was significant when the *P* < 0.05 or *P* < 0.01 versus the control groups (*n* = 3).

## RESULTS

### LPS Can Induce the Inflammatory Response

To prove whether LPS can induce the innate immune response in teleost, the expression of TNFα, IL-1β, and IFNβ in macrophages that challenged with ultrapure LPS, poly(I:C), LTA, and Zymoscan A was detected by qRT-PCR. These results showed that both LPS-B5 and LPS-EK can apparently promote the expression of TNFα and IL-1β (**Figures 1A,B**), but showed little effect on the expression of IFNβ (**Figure 1C**). Correspondingly, poly(I:C) more evidently induced the expression of IFNβ compared with TNFα and IL-1β. Compared with LPS and poly(I:C), the increase in expression of these cytokines was relatively few, which was activated by LTA and Zymosan A. These results indicated that LPS could be recognized and induce the inflammatory response in miiuy croaker.

### NOD1 Is Sensitive to the Stimulation of LPS

To determine whether NOD1 plays a role in the signaling pathway induced by different pathogens. The expression of NOD1 was detected in liver tissues infected with Gram-negative bacteria (*V. anguillarum* and *V. harveyi*), LPS-B5, Gram-positive bacteria (*S. aureus*), and in the macrophages challenged by SCRV, with 5 MOI which can frequently infect fish and challenged by poly(I:C), which was a synthetic analog of dsRNA. The expression of NOD1 increased remarkably in the liver after infection with *V. anguillarum*, *V. harveyi,* and LPS-B5 (**Figures 2A–C**). However, no marked increase was observed in the liver tissues infected by

Figure 1 | Lipopolysaccharide (LPS) can induce the inflammatory response. The expression profiles of TNFα (A), IL-1β (B), and IFNβ (C) in miiuy croaker macrophages that treated with LPS-B5, LPS-EK, poly(I:C), LTA, and Zymosan A, respectively, which were detected by using qRT-PCR. The expression of β-actin was used as the internal control, and the data were represented in the form of mean ± SE. The statistically significant differences between control and experience groups were indicated with asterisks (\**P* < 0.05 and \*\**P* < 0.01).

*S. aureus* and in the macrophages treated with SCRV or poly(I:C) (**Figures 2D–F**). These results demonstrate that NOD1 may be involved in Gram-negative bacteria and LPS-induced signaling pathways, but does not considerably function in the immune response induced by Gram-positive bacteria and viruses. Because NOD1 can recognize iE-DAP (23, 26) in both fish and mammal; furthermore, similar to iE-DAP, LPS is also a pathogenic component of Gram-negative bacteria. So, we can guess that NOD1 may be a recognition receptor that can identify the pathogenic components of the Gram-negative bacteria in cytoplasm.

### LPS May Be Recognized by NOD1

The expression of NOD1, TNFα, IL-8, and IL-1β was detected in LPS-treated macrophages to further verify whether LPS can be identified by NOD1 and induce inflammatory response. Stimulation of macrophages with LPS-B5 and LPS-EK resulted in the marked increase in the expression of NOD1 and several inflammatory cytokines (**Figure 3A**). This result explains that LPS can probably promote the expression of NOD1 and induce inflammatory response in fish. Then, we transfected NOD1 siRNA into the macrophages and detected the expression of NOD1, TNFα, IL-6, and IL-8 before and after LPS stimulation to further verify the role of NOD1 in the identification of LPS. The result showed that NOD1-siRNA can efficiently inhibit the expression of NOD1. Moreover, after knockdown of the expression of NOD1 gene, the expression of TNFα, IL-6, and IL-8 was also decreased obviously whether or not stimulated with LPS-B5 (**Figure 3B**). These results can also prove that NOD1 plays an important role in the inflammatory response induced by LPS. After inhibition of the expression of NOD1 resulted in decrease in the production of inflammatory cytokines.

### NOD1 Is Sensitive to LPS and Activate the NF-**κ**B Signal Pathway

The signal pathway that can be activated by NOD1 was determined by co-transfecting NOD1 expression plasmids and NF-κB or ISRE reporter plasmids into HEK293 cells, pRL-TK plasmids as the internal control, and then dual luciferase reporter gene assay was performed (**Figure 4A**). The result showed that compared with the negative control, NOD1 can significantly activate the expression of NF-κB. Then, a concentration gradient

experiment was performed to further verify the result. The findings indicated that NOD1 may induce the immune response by activating the NF-κB signal pathway. To confirm whether NOD1 can recognize LPS, HEK293 cells were co-transfected with NOD1 expression plasmid and NF-κB reporter plasmid. Then, the cells were stimulated with different ligands, and the luciferase activity was detected after 12 h of stimulation (**Figure 4B**). The result showed that after stimulating the cells with various pathogens, only LPS-B5 and LPS-EK could not markedly promote the NF-κB luciferase activity without overexpression of NOD1 but can significantly activate the NF-κB luciferase activity after overexpression of NOD1. To further verify this result, a concentration gradient experiment of ultrapure LPS-B5 and LPS-EK in HEK293 cells (**Figure 4C**) and the dual luciferase reporter assay in EPC cells (**Figure 4D**) were performed. Results showed that after stimulated with LPS-B5 and LPS-EK, over-expression of NOD1 can substantially promote the NF-κB luciferase activity both in HEK293 cells and in EPC cells. To identify the location of NOD1 distinguish LPS, HEK293 cells were transfected with NOD1 expression plasmid and NF-κB reporter plasmids, pRL-TK plasmids as the internal control. Then perform intracellular and extracellular stimulation by using LPS-B5 and LPS-EK (Figure S7 in Supplementary Material). Collect the cells to detect Luciferase activity. Result showed that NOD1 can identify the stimulation of LPS in the cytoplasm. Therefore, from the above results, we believe that NOD1 may be the intracellular recognition receptor for LPS.

### Specific Binding Activity to LPS and iE-DAP

To further reveal the likely physiological function of NOD1 in fish, streptavidin pulldown assay was performed. As shown in **Figure 5A** and Figure S2A in Supplementary Material, miiuy croaker NOD1 proteins can be pulled down by SA-PMPs (Streptavidin MagneSphere® Paramagnetic Particles, Z548, Promega) through biotinylated ultrapure LPS. On the contrary, miiuy croaker TLR5s protein, which was a recognition receptor of the TLR family (33), cannot be pulled down by SA-PMPs. So, it can be proved that miiuy croaker NOD1 can combine with LPS in cells.

Then, to further validate the binding effect between NOD1 and LPS, ELISA was performed. Although we used ultrapure LPS-B5and LPS-EK in experiments, in order to further eliminate the interference from other components, we also did the binding experiments between NOD1 and Lipid A. Lipid A is the innermost of the three regions of the LPS, and many immuneactivating abilities of LPS can be actually attributed to lipid A. Thus, the binding between NOD1 and Lipid A was detected to further confirm the binding between LPS and NOD1 (Figure S2B in Supplementary Material). In addition, iE-DAP is recognized as a ligand for NOD1 in mammal and fish (23, 26), considering that the LRR domain of NOD1 is the binding domain for iE-DAP (25). We detected the binding activity of NOD1–LRR with iE-DAP as the positive control to confirm the binding activity of NOD1-LRR with LPS. At the same time, because β-1,3-Glucan

NOD1 and TLR5s. The cropped HA and Flag blot are shown for pulldown and input, a cropped Tubulin blot is shown for input. For um-cropped blot, see Figures S3A,B in Supplementary Material. (B) Binding activity analysis of miiuy croaker NOD1-LRR and TRX to ultrapure LPS-B5, LPS-EK, iE-DAP, and β-1,3-Glucan by using enzyme-linked immunosorbent assay. For un-cropped NOD1-LRR and TRX blot, see Figure S3C in Supplementary Material.

was the ligand that cannot be recognized by NOD1, the binding activity of NOD1–LRR with β-1,3-Glucan was also detected as a negative control. These results revealed that NOD1–LRR could strongly bind to ultrapure LPS, Lipid A, and iE-DAP in a concentration-dependent manner, but could not bind to β-1,3-Glucan. Moreover, NOD1–LRR possessed stronger binding ability to ultrapure LPS and Lipid A than to iE-DAP, because NOD1–LRR showed apparent binding activity to LPS and Lipid A at a low concentration (less than 0.1 nM). By contrast, significant binding activity with iE-DAP required higher NOD1–LRR concentration (more than 1 nM) (**Figure 5B**). Additionally, NOD1–LRR harvested higher optical density values for the binding to ultrapure LPS and Lipid A than to iE-DAP at the same NOD1–LRR concentration. These results also demonstrated that NOD1-LRR possessed more potent binding activity with LPS. By contrast, TRX exhibited very weak binding activity with ultrapure LPS, Lipid A, iE-DAP, or β-1,3-Glucan. These findings suggest that NOD1 is the receptor for iE-DAP (25), and NOD1 could be a potential receptor for LPS.

### Knockdown of NOD1 Reduces NF-**κ**B Activity

Three siRNAs were designed and transfected into cells to knockdown NOD1 to further examine the role of NOD1 in the inflammatory response induced by LPS. Moreover, different experimental techniques were used to validate the efficiency of these siRNAs. Western blot result and the relative optical density value clearly showed the inhibitory effect of NOD1-siRNA on the expression of NOD1 (**Figure 6A**). Then, NOD1 expression plasmids and NF-κB reporter plasmids were co-transfected with

Figure 6 | NOD1-siRNA inhibits the role of NOD1 in recognizing lipopolysaccharide (LPS). (A) Miiuy croaker NOD1 expression plasmid was co-transfected into HEK293 cells along with three NOD1-siRNA or control siRNA, respectively, after 48 h of transfection, the cells were collected to perform western blotting analysis and the relative optical density value assay were performed according to the western blotting analysis results. A cropped NOD1 and β-actin blot are shown. For un-cropped NOD1 and β-actin blot, see Figure S4 in Supplementary Material. (B) HEK293 cells were co-transfected with NOD1 expression plasmid and NF-κB reporter plasmid along with three NOD1-siRNA, then the luciferase activity was detected. Next, co-transfection of NOD1 expression plasmid and NF-κB reporter plasmid along with different quantity of NOD1-siRNA3 to perform a concentration gradient experiment. (C) HEK293 cells were co-transfected with miiuy croaker NOD1-GFP plasmid and NOD1-siRNA3 or control siRNA to detect the fluorescence quantity. (D) Co-transfection of NOD1 expression plasmid and NF-κB reporter plasmid along with NOD1-siRNA3 into HEK293 cells, then stimulate the cells with LPS-B5 and LPS-EK to perform concentration gradient experiment. (E) Co-transfection of NOD1 expression plasmid and NF-κB reporter plasmid along with NOD1-siRNA3 into EPC cells, then stimulate the cells with LPS-B5 and LPS-EK to perform dual-luciferase reporter gene assay. The pRL-TK plasmid was used as the internal control, and data are expressed as mean ± SE. The statistically significant differences were indicated by asterisks (\**P* < 0.05 and \*\**P* < 0.01).

the three NOD1-siRNAs into HEK293 cells to further test the role of NOD1-siRNA by detecting the luciferase activity. **Figure 6B** showed that all of the three NOD1-siRNAs can play a significant inhibitory role on the expression of NOD1. However, the role of NOD1-siRNA3 was more obvious compared with the other two siRNAs. NOD1-GFP recombinant plasmid can express the NOD1 protein with green fluorescence protein. Thus, transfection of NOD1-GFP plasmid along with NOD1-siRNA3 or control siRNA into cells was performed, then, the intensity of fluorescence was detected to validate the efficiency of NOD1-siRNA3 again. As shown in **Figure 6C**, result showed that NOD1-siRNA3 could effectively inhibit the expression of NOD1. Finally, a concentration gradient of LPS-B5 and LPS-EK and the dual luciferase reporter gene assay in EPC cells were performed to demonstrate the role of NOD1 in the recognition of LPS (**Figures 6D,E**). After inhibiting the expression of NOD1, the activity of NF-κB induced by LPS was also greatly reduced. These results indicate that NOD1 may be a receptor, which can recognize LPS in cells.

### NOD1 Can Identify LPS and Mediates Cytotoxicity

To further confirm that miiuy croaker NOD1 can identify LPS, cell viability was analyzed by detecting the concentration of ATP in HEK293 cells and EPC cells. Results showed that after stimulation with LPS-B5 and LPS-EK, the cell viability will decrease

significantly in both HEK293 and EPC cells with over-expression NOD1, but the cell activity only have little change after being stimulated with LTA (**Figure 7A**). Moreover, as shown in **Figure 7B**, we can find that this change is concentration dependent. Then, cotransfected of NOD1 expression plasmid and NOD1-siRNA into cells to perform knockdown analysis, results showed that both in HEK293 and EPC cells, after stimulated with LPS, knockdown of NOD1, the cell activity will raise substantially (**Figures 7C,D**), this indicated that knockdown of NOD1, the cytotoxicity induced by LPS will be unable to affect the cell viability. So, from the above results, we can consider that NOD1 can identify LPS and mediate LPS-induced cytotoxicity in fish cells.

### NOD1 Activates NF-**κ**B Signaling Pathway by Recruiting RIPK2

RIPK2 is the receptor-interacting protein of NOD1 and NOD2, and this protein has been confirmed in mammals. To determine whether NOD1 was also needed to recruit RIPK2 through protein–protein interaction to transfer the signal in fish, RIPK2 expression plasmids were constructed and co-transfected with NOD1 plasmids into cells for luciferase activity assay. The results showed that the activation of NF-κB was significantly increased when the cells were co-transfected with RIPK2 and NOD1 plasmids compared with the experiment that singlely transfected with NOD1 or RIPK2 plasmid. Then, we performed a concentration gradient experiment of NOD1 to further verified this result (**Figure 8A**). Next, the cells were transfected with NOD1 and RIPK2 plasmids along with NF-κB reporter plasmids and then stimulated with LPS-B5 and LPS-EK, after 12 h of stimulation, the cells were lysed and the fluorescence activity was detected. As shown in **Figure 8B**, LPS can induce significant increase in NF-κB activity after co-transfection of NOD1 and RIPK2 plasmids. To validate the protein–protein interaction between NOD1 and RIPK2, immunoprecipitation and confocal imaging were performed, the results can be unambiguously proved that NOD1 and RIPK2 can be combined directly to transmit the signal and activate the downstream signaling pathway (**Figures 8C,D**).

Then, to further verify the influence of RIPK2 for signal transduction in fish, the macrophages were transfected with RIPK2-siRNA to silence the expression of RIPK2. Then, the cells were stimulated with LPS-B5 and the expression of RIPK2, TNFα, IL-6, and IL-8 were detected by using qRT-PCR (**Figure 8E**). Result showed that after inhibition of the expression of RIPK2, the expression of TNFα, IL-6, and IL-8 were also suppressed. These results illustrate that similar to mammals, NOD1 was also needed to recruit RIPK2 to transfer the signal to induce inflammatory response in fish.

Figure 8 | NOD1 recruits RIPK2 to activate the NF-κB signaling pathway. (A) Co-transfection of miiuy croaker RIPK2 and miiuy croaker NOD1 expression plasmids into HEK293 cells, and after 48 h of transfection, the luciferase activity was detected. Next, co-transfection of RIPK2 and different amount of NOD1 expression plasmids into HEK293 cells to perform a concentration gradient experiment, pRL-TK plasmids was used as the internal control. (B) HEK293 cells, which have been transfected with NOD1 and RIPK2 plasmids, were stimulated with lipopolysaccharide (LPS)-B5 and LPS-EK, pRL-TK plasmids were used as the internal control, and after 12 h of stimulation, cells were collected and luciferase activity was detected. (C) Immunoprecipitation analysis of NOD1 and RIPK2. A cropped HA blot is shown for IP and input, a cropped Flag blot is shown for input. For un-cropped HA and Flag blot, see Figure S5 in Supplementary Material. (D) Immunostaining and confocal imaging experiment of NOD1 and RIPK2. (E) Transfection of RIPK2-siRNA into miiuy croaker macrophages, which were treated with LPS-B5, and the expression of RIPK2, TNFα, IL-6, and IL-8 were detected. The data were represented as mean ± SE, and the statistically significant differences were expressed with asterisks (\**P* < 0.05 and \*\**P* < 0.01).

## NOD1 Recognition of LPS Rely on LRR Domain

We constructed different mutant plasmids to verify the function of the different domains of NOD1. The schematic of NOD1 mutant plasmids was displayed in Figure S1 in Supplementary Material. First, the cells were transfected with the wild-type or mutant NOD1 plasmids to examine the MW and confirm the expression (**Figure 9A**). Given that NOD1 was structurally homologous with the mammalian NOD1, the same domains may have the same role in the identification of ligands. In general, the LRRs domain was used to identify ligands, and a study has shown that a conserved motif of "LxxLxLxxNxL" exists in the LRR sequence of miiuy croaker NOD1 (**Figure 9B**) (29). Accordingly, LRRs domain may also play the role of recognizing ligands in teleost fish. Then, we determine the role of different domains of miiuy croaker NOD1 by transfecting the wild-type or mutant NOD1 plasmids into cells along with NF-κB reporter plasmids and perform dual luciferase reporter gene assay. Results showed that the expression of NF-κB that was activated by mutant NOD1 plasmids was far less than that activated by wild-type NOD1 plasmid. These results imply that all the domains will play an essential role in activating the NF-κB signal pathway (**Figure 9C**). Then, we transfected the mutant plasmids, which were mutated in the LRR structure, along with NF-κB reporter plasmids and RIPK2 plasmid into cells, and treated the cells with LPS-B5 and LPS-EK to further detect the function of LRR domains in the

identification of LPS. **Figures 9D,E** shows that, whether or not existence of RIPK2, after recognizing LPS, the expression of NF-κB that activated by mutant NOD1 plasmids which lack one or several LRR structures was obviously reduced compared with activated by wild-type NOD1 plasmids. These results mean that LRR domains were the major areas of recognition, and deletion of any area can cause NOD1 to lose its recognition function.

Based on the above results, we conceive a possible signal pathway to explain the process of NOD1 recognition of LPS and induction of inflammatory response (**Figure 10**). First, LPS was secreted from Gram-negative bacteria and then invaded to the cytosol, and after secretion into the cytoplasm, LPS could be identified by NOD1 through the LRR domain, and then through CARD domain to recruit RIPK2 to transfer the signal. This process was followed by activation of the NF-κB signal pathway to induce the expression of inflammatory cytokines.

### DISCUSSION

Fish being the lower vertebrate is a favorable animal material to study the function evolution of the immune system. Comparison of the immune genes between fish and mammals shows certain difference in both quantity and function. For example, in mammals, 13 TLRs have been characterized and showed to identify the specific pathogen-associated molecular pattern (37). However, at least 18 TLR types have been identified in various species of fish (13, 38), and TLR18–20 and TLR23–28 are only found in fish (39, 40). Furthermore, TLR6 and TLR10 were not found in fish, but TLR14 plays functions similar to TLR6 and TLR10 in identifying various pathogens and activating immune response (41). Thus, the function of the same gene in fish may not play the same role compared with the genes in mammals. In mammals, TLR4 is a central protein that distinguishes LPS. By contrast, in fish, TLR4 is only found in several kinds of fish and cannot identify LPS (15). Thus, other methods to identify LPS should be explored.

In the mammal immune system, the complex of TLR4, CD14, and MD2 has been proved to be the receptor for LPS at the cell surface (3). And several intracellular LPS receptors, such as caspase-4/5/11, are also present in the cytoplasm (9). These receptors can play an important role against invading bacteria. For example, epithelial cells were insensitive to extracellular LPS. However, LPS can also activate the NF-κB signal in the cytoplasm of these cells (42). This characteristic indicates that, in some epithelial cells which are located in particular tissues, such as the

gut, because these cells are often exposed to bacteria or bacterial products; however, the expression of cell surface receptors in these cells is very low, the intracellular receptors will play an important role in resistance against bacterial infection (43, 44). In this study, we demonstrated that NOD1, as an intracellular receptor, can identify LPS and activate the NF-κB signal pathway to induce inflammatory response in teleost fish. In aquaculture, fish gill and gut are always exposed to large amounts of bacteria or bacterial products, especially Gram-negative bacteria. To date, the receptors on the cell surface of fish that can identify LPS remain unclear. Thus, NOD1 shows unparalleled importance in the resistance against bacterial infections in fish. In addition, from **Figure 3B**, we can find that after silencing the expression

### REFERENCES


of NOD1 by using NOD1 siRNA, the stimulation of LPS can still significantly increase the expression of inflammatory cytokines, this shows that other receptors still exist in fish similar to in mammals to recognize LPS and activate the inflammatory response. Further research is required to determine whether these receptors also play an important role in fish.

NOD-like receptor is a promising intracellular recognition receptor family. Recent reports have indicated that several members of this family exist in the lower vertebrates (45) and in invertebrates (46). Many studies on this gene family have been conducted in mammals. However, few related studies have been performed in other vertebrates, such as birds, amphibians, and fish (22), resulting in little to no information about NLRs in these species (47). NOD1 is one of the representative members of the NLR family. This molecule has been studied extensively in mammals, and it can detect a unique muropeptide of iE-DAP (23). However, the NOD1 function in fish remains ambiguous because of the limited available experimental materials for studying the function of genes. NOD1 is structurally homologous with mammalian NOD1, which can detect a wide array of microbial components (48). Therefore, in this study, we have demonstrated that NOD1 can distinguish LPS through LRR domain and recruit RIPK2 to transfer the signal to induce inflammatory response in fish. This study was first to report that NOD1 can recognize LPS in teleost fish. These results elucidate the resistance of fish against bacterial infections. A theoretical basis is provided for future studies on the treatment of fish diseases.

### AUTHOR CONTRIBUTIONS

Conceived and designed the experiments: TX and DB. Performed the experiments: DB, YW, TX, XL, YG, QC, and JC. Analyzed the data: TX and DB. Wrote the paper: TX, DB, and XL.

### FUNDING

This study was supported by National Natural Science Foundation of China (31672682).

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at https://www.frontiersin.org/articles/10.3389/fimmu.2018.01413/ full#supplementary-material.

humans. *Arch Immunol Ther Exp (Warsz)* (2013) 61:427–43. doi:10.1007/ s00005-013-0243-0


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

*Copyright © 2018 Bi, Wang, Gao, Li, Chu, Cui and Xu. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Amphibian (*Xenopus laevis*) Interleukin-8 (CXCL8): A Perspective on the Evolutionary Divergence of Granulocyte Chemotaxis

### Daphne V. Koubourli, Amulya Yaparla, Milan Popovic and Leon Grayfer\*

*Department of Biological Sciences, George Washington University, Washington, DC, United States*

The glutamic acid-leucine-arginine (ELR) motif is a hallmark feature shared by mammalian inflammatory CXC chemokines such the granulocyte chemo-attractant CXCL8 (interleukin-8, IL-8). By contrast, most teleost fish inflammatory chemokines lack this motif. Interestingly, the amphibian *Xenopus laevis* encodes multiple isoforms of CXCL8, one of which (CXCL8a) possesses an ELR motif, while another (CXCL8b) does not. These CXCL8 isoforms exhibit distinct expression patterns during frog development and following immune challenge of animals and primary myeloid cultures. To define potential functional differences between these *X. laevis* CXCL8 chemokines, we produced them in recombinant form (rCXCL8a and rCXCL8b) and performed dose-response chemotaxis assays. Our results indicate that compared to rCXCL8b, rCXCL8a is a significantly more potent chemo-attractant of *in vivo*-derived tadpole granulocytes and of *in vitro*-differentiated frog bone marrow granulocytes. The mammalian CXCL8 mediates its effects through two distinct chemokine receptors, CXCR1 and CXCR2 and our pharmacological inhibition of these receptors in frog granulocytes indicates that the *X. laevis* CXCL8a and CXCL8b both chemoattract tadpole and adult frog granulocytes by engaging CXCR1 and CXCR2. To delineate which frog cells are recruited by CXCL8a and CXCL8b *in vivo*, we injected tadpoles and adult frogs intraperitoneally with rCXCL8a or rCXCL8b and recovered the accumulated cells by lavage. Our transcriptional and cytological analyses of these tadpole and adult frog peritoneal exudates indicate that they are comprised predominantly of granulocytes. Interestingly, the granulocytes recruited into the tadpole, but not adult frog peritonea by rCXCL8b, express significantly greater levels of several pan immunosuppressive genes.

Keywords: interleukin-8, amphibian, granulocyte, chemotaxis, CXCL8, FV3

### INTRODUCTION

The mammalian cysteine-X-cysteine (CXC) chemokines can be subdivided into two groups depending on whether they posses the Glu-Leu-Arg (ELR) motif at their N-termini (1, 2). This ELR motif is responsible for binding of these chemokines to their cognate receptors on neutrophils, resulting in the neutrophil chemotaxis. By contrast, the CXC chemokines lacking this motif do not attract neutrophils and instead target mononuclear phagocytes and different lymphocyte subsets (1, 2). With the exception of Gadiformes (cod, haddock), the teleost fish CXC

#### *Edited by:*

*Geert Wiegertjes, Wageningen University & Research, Netherlands*

#### *Reviewed by:*

*Miki Nakao, Kyushu University, Japan Magdalena Chadzinska, Jagiellonian University, Poland*

> *\*Correspondence: Leon Grayfer leon\_grayfer@gwu.edu*

#### *Specialty section:*

*This article was submitted to Comparative Immunology, a section of the journal Frontiers in Immunology*

*Received: 04 May 2018 Accepted: 21 August 2018 Published: 12 September 2018*

#### *Citation:*

*Koubourli DV, Yaparla A, Popovic M and Grayfer L (2018) Amphibian (Xenopus laevis) Interleukin-8 (CXCL8): A Perspective on the Evolutionary Divergence of Granulocyte Chemotaxis. Front. Immunol. 9:2058. doi: 10.3389/fimmu.2018.02058*

**23**

chemokines lack ELR motifs and instead generally possess X(other residue)-Leu-Arg (XLR, DLR in salmonids) motifs, which are presently thought not to contribute to the function of these respective CXC chemokines or their recruitment of fish neutrophils (3). Because the mutation of the mammalian CXC chemokine ELR motifs to DLR did not abolish their neutrophil recruitment, it was originally thought that the fish DLR (and XLR) motifs function as the mammalian CXC chemokine ELR counterparts (3). However, more recent work has demonstrated that the fish CXC chemokine DLR/XLR motifs are dispensable to the fish neutrophil chemotaxis (4).

CXCL8 (interleukin-8, IL-8) is an important inflammatory CXC chemokine, first discovered in mammals for its role in the chemotaxis of neutrophils (5). CXCL8-mediated neutrophil recruitment occurs as the result of this chemokine binding to the G protein-coupled CXC chemokine receptor 1 (CXCR1, CXCL8Rα) or CXCR2 (CXCL8Rβ) (6, 7). Of these CXC chemokine receptors, CXCR1 is only ligated by CXCL8 and CXCL6, whereas CXCR2 is bound by several chemokines including CXCL8, CXCL1, and CXCL2 (6, 7). CXCL8 is important to both the initiation as well as the resolution of inflammatory responses. By recruiting neutrophils to sites of injury and/or infection, this chemokine promotes the resolution of tissue damage and clean up of infiltrating pathogens through neutrophil-mediated phagocytosis, respiratory burst, and the release of neutrophil extracellular traps (8). In turn, CXCL8 facilitates wound repair by activating the angiogenic response and by eliciting endothelial cell proliferation, survival, and recruitment (9), resulting in the formation of new blood vessels (10), thereby contributing to the resolution of inflammatory stimuli and promoting healing.

Cxcl8 genes have been identified across a range of bony fish species, with many species encoding multiple CXCL8 isoforms. Cyprinid fish such as zebrafish and carp encode two distinct Cxcl8 genes (11, 12) corresponding to two Cxcl8 homologs that have been designated as Cxcl8\_L1 and Cxcl8\_L2. The expression of both fish Cxcl8\_L1 and Cxcl8\_L2 genes is upregulated in response to bacterial infection (13) and wound-associated inflammation (12). Interestingly, under certain immune conditions, these two genes are differentially regulated (14, 15). Despite lacking ELR motifs, both of these cyprinid CXCL8 homologs chemoattract fish neutrophils (11, 12). While salmonid fish such as trout encode several Cxcl8 genes (16), these all share close homology to the cyprinid Cxcl8\_L1 [76] and the trout CXCL8 likewise chemoattracts fish neutrophils (17). Indeed, while Cxcl8 chemokines have been identified across multiple teleost fish species, as mentioned above, all of these Cxcl8 genes lack the ELR motif, with the exception of Gadiformes such as Atlantic cod (18) and haddock (19).

Here we report on the amphibian (Xenopus laevis) Cxcl8 isoforms (designated Cxcl8a and Cxcl8b), only one of which possesses an ELR motif. We show that these frog Cxcl8a and Cxcl8b genes are differentially expressed in healthy animals and following immunological challenge. Moreover, we demonstrate that these frog CXCL8 chemokines possess distinct chemoattractive capacities and may have functionally diverged to recruit distinct tadpole granulocyte populations.

### MATERIALS AND METHODS

### Animals, Culture Media, and Conditions

Outbred tadpole and adult X. laevis were purchased from the Xenopus 1 facility, housed and handled under strict laboratory and IACUC regulations (Approval number 15-024).

The cell culture media and conditions have been previously described (20).

### Production of Recombinant Frog Cytokines and Chemokines

The production of recombinant G-CSF, M-CSF, and IL-34 has been previously described (20, 21). The recombinant CXCL8a and CXCL8b were generated by PCR-amplifying the respective sequences, corresponding to the signal peptidecleaved Cxcl8a and Cxcl8b transcripts, ligating these into the pMIB/V5 His A insect expression vectors (Invitrogen) and introducing positive clones into Sf9 insect cells (Cellfectin II, Invitrogen). The production of recombinant (r)CXCL8a and rCXCL8b by the transfected Sf9 cells was confirmed by western blot against the V5 epitopes on the recombinants and the positive transfectants were selected using 10µg/mL blasticidin. These protein expressing cultures were scaled up to 500 ml, grown for 5 days, pelleted and the supernatants collected. The supernatants were dialyzed overnight at 4◦C against 150 mM sodium phosphate, concentrated against polyethylene glycol flakes (8 kDa) at 4◦C, dialyzed overnight at 4◦C against 150 mM sodium phosphate and passed through Ni-NTA agarose columns (Qiagen). Columns were washed with 2 × 10 volumes of high stringency wash buffer (0.5% Tween 20; 50 mM Sodium Phosphate; 500 mM Sodium Chloride; 100 mM Imidazole) and 5 × 10 volumes of low stringency wash buffer (as above, but with 40 mM Imidazole). Recombinant cytokines were eluted using 250 mM imidazole. The eluted recombinant (r)CXCL8a and rCXCL8b were resolved by SDS PAGE, transferred onto nitrocellulose membranes and western blots were performed using an HRP-conjugated mouse anti-V5 (Sigma) to determine which elution fractions contained rCXCL8a (15 kDa) and rCXCL8b (16 kDa). The fractions containing the respective recombinants (**Supplemental Figure 1**) were pooled, concentrated against polyethylene glycol flakes (8 kDa) at 4 ◦C, dialyzed overnight against saline at 4◦C and the protein concentrations were determined by Bradford protein assays (BioRad). Halt protease inhibitor cocktail (containing AEBSF, aprotinin, bestatin, E-64, leupeptin, and pepstatin A; Thermo Scientific) was added to the purified proteins, which were then stored at −20◦C in aliquots until use.

### *X. laevis* Myeloid Cell Isolation and Culture

Tadpole granulocytes were generated as previously described (20). Briefly, tadpoles (stage NF 54) were injected ip with 1 µg total of rG-CSF using finely pulled glass needles. One day following injection, peritoneal leukocytes were lavaged with saline, enumerated via a hemocytometer and using trypan blue (Sigma) exclusion.

Tadpole and adult frog rCXCL8a- and rCXCL8b-elicited leukocytes were derived by injecting tadpoles (stage NF 54, N = 6) and adult frogs (1 year old N = 6) with 1µg/g of body weight of either chemokine. After 4 h of injection, tadpoles, and adults were lavaged with saline and the recovered cells were enumerated by hemocytometer counts, using trypan blue (Sigma) exclusion. The results depicted in **Figure 6**, corresponding to the transcriptional analysis of these cell populations, are representative of 3 independent experiments, with each iteration performed with cells from 6 tadpoles or adult frogs.

The generation of adult frog M-CSF- and IL-34- macrophages and G-CSF-granulocytes has been previously described (22, 23). Briefly, X. laevis adult frogs were sacrificed and their femur bone marrow cells were isolated and incubated with 250 ng/ml of the respective recombinant growth factors at 27◦C and 5% CO2. After 3 days of culture, the cells were again treated with the respective cytokines and after 5 days of culture, the cells were enumerated and used in the gene expression or chemotaxis assays.

### Frog Virus 3 Stocks and Infections

Frog Virus 3 (FV3) production has been described previously (24). In brief, baby hamster kidney (BHK-21) cells were infected with FV3 (multiplicity of infection; MOI: 0.1), grown at 30◦C and 5% CO<sup>2</sup> for 5 days. The FV3-containing supernatants were collected over 30% sucrose by ultracetrifugation, re-suspended in saline and the viral titers were determined by plaque assay over BHK-21 cells.

Tadpoles (N = 5) and adult frogs (N = 5) were infected with FV3 by intraperitoneal (ip) injection with 1 × 10<sup>4</sup> and 5 × 10<sup>6</sup> PFU of FV3, respectively or mock infected with saline (not containing FV3). Animals were euthanized by tricaine mesylate overdose (tadpoles: 1%; adult frogs: 5%), kidney tissues excised, immediately flash-frozen in Trizol reagent (Invitrogen) over dry ice and stored at −20◦C until RNA isolation.

For all in vitro infection studies, leukocytes were infected with a multiplicity of infection (MOI) of 0.5 plaque forming units (PFU) of FV3 for 16 h, incubated in the medium described above at 27◦C with 5% CO2. Subsequently, the cells were trypsinized to remove attached but not internalized virus and washed with saline and processed for RNA isolation and cDNA synthesis. Alternatively, leukocytes were incubated with heat-killed E. coli for 16 h prior to RNA isolation and cDNA synthesis.

### Tadpole and Adult Frog Wounding and Tissue Repair Studies

For the tadpole tissue inflammation and repair/regeneration study, 15 tadpoles were anesthetized with tricaine mesylate and the furthermost thirds of their tails were amputated using clean razor blades. Approximately 1 mm sections were cut from the amputated tails (portions closest to the cut site), and used as controls for these studies. After 5 h, 1, 3, and 6 days of the initial amputation, 5 tadpoles (per time point) were anesthetized and ∼1 mm sections of their amputated, regenerating tails were excised using clean razor blades. These 1 mm section were used for RNA isolation, as described below.

For the adult frog, wounding and repair experiments, 20 frogs (1 year old) were anesthetized with tricaine mesylate and pieces of skin ∼1 mm<sup>2</sup> were excised from their hind right legs. These sections were used as the expression controls for this study. After 5 h and 1, 3, and 12 days of the initial incisions, 5 of the frogs were again anesthetized and the skin around the initial incision was removed in a 1 mm perimeter, and subjected to RNA isolation.

### RNA Isolation, cDNA Synthesis, and Quantitative Gene Expression Analyses

For all experiments, tadpole and adult frog cells or kidney tissues from FV3-infected animals were homogenized in Trizol reagent (Invitrogen), flash frozen on dry ice and stored at −80◦C until RNA isolation in accordance to manufacturer's directions. The isolated RNAs (500 ng total) were reverse transcribed into cDNAs using cDNA qscript supermix (Quanta), according to manufacturer's instructions.

All quantitative analysis of X. laevis gene expression was performed using the CFX96 Real-Time System and iTaq Universal SYBR Green Supermix. The BioRad CFX Manager software (SDS) was employed for all expression analysis. All primers were validated prior to use and the sequences of all employed primers are listed in the **Supplemental Table 1**.

All expression analyses were conducted relative to the Gapdh endogenous control gene. The expression of the Cxcl8a and Cxcl8b genes was directly compared by calculating the delta∧delta CT values for these two genes relative to the highest CT value (lowest mRNA levels) across all of the derived Cxcl8a and Cxcl8b CT values. Likewise, the expression of Cxcr1 and Cxcr2, as presented in **Figures 2**, **3**, **4C** and **Supplemental Figure 2B**, was directly compared by calculating the delta∧delta CT values, relative to the lowest observed CT value across the derived Cxcr1 and Cxcr2 CT values within the respective experiments. For all other gene expression analyses, the delta∧delta CT values were derived relative to the lowest expressing tissue/cell type within the given data set (highest CT) and the derived relative quantification data were normalized against that lowest sample.

### Chemotaxis Assays

All chemotaxis assays were performed using blind well chemotaxis (Boyden) chambers (Neuro Probe), with medium alone or 10<sup>3</sup> -10−<sup>7</sup> ng/ml of rCXCL8a or rCXCL8b (in culture medium) loaded into bottom wells of these chambers. The bottom wells were overlaid with 13 mm chemotaxis filters (5µm pore size; Neuro Probe) and tadpole or adult frog granulocytes (10<sup>5</sup> cells/well) were added to the top wells. After 3 h of incubation at 27◦C with 5% CO2, the top layers were aspirated, the top sides of the filters were wiped with cotton swabs. The filters were then removed, washed, and the bottom faces of the filters were stained with Giemsa stain and the numbers of migrating cells was determined by counting ten random fields of view per filter (40x objective). For the chemokinesis experiments, both the bottom and the top wells of the chemotaxis chambers were loaded with the most potent chemoattractive concentrations of either chemokine; tadpole granulocyte: 10−<sup>3</sup> ng/ml of rCXCL8a and 10<sup>1</sup> ng/ml of rCXCL8b; adult granulocytes: 10−<sup>5</sup> ng/ml of rCXCL8a and 10−<sup>3</sup> ng/ml of rCXCL8b. The assays were performed as above. The reparixin (CXCR1/2 inhibitor, 1 and 100 nM, MCE) and SB265610 (CXCR2 inhibitor, 5 and 100 nM, Sigma) inhibition studies were carried out by using the respective optimal doses of rCXCL8a or rCXCL8b to examine tadpole and adult granulocyte migration in the presence of 0, 1, or 100 nM final concentrations of reparixin across both lower and upper chambers. Cells from three individual animals (N = 3) were used to test each concentration of either rCXCL8. The tadpole and adult frog granulocyte chemotaxis toward the most chemo-attractive concentrations of rCXCL8a and rCXCL8b was confirmed twice, independently, using cells from three individuals for each experiment. The results from the 3 independent studies (N = 9) were combined and are presented in **Figure 4**. The pharmacological inhibition experiments were performed using cells from 4 tadpoles (per treatment group) and 4 adult frogs (N = 4).

### Anti-*X. laevis* G-CSFR Polyclonal Antibody

A recombinant form of a fragment of the extracellular portion of the X. laevis G-CSFR was produced by PCRamplifying the corresponding signal peptide-cleaved Gcsfr transcript of 702 nucleotides and cloning it into the pET SUMO prokaryotic expression vector (ThermoFisher). This construct was introduced into One Shot Mach-T1R Chemically Competent E. coli, (Invitrogen), plated onto kanamycin (50µg/ml) LB plates and the resulting colonies were screened by colony-PCR for Gcsfr-positive constructs. Positive colonies were grown in LB + kanamycin (50µg/ml), the plasmids were isolated and sequenced to confirm the presence of in frame 702 nt sequences corresponding to the X. laevis Gcsfr fragment. Several of these positive clones were introduced into BL21(DE3) One Shot Chemically Competent E. coli (Invitrogen),and pilot IPTG induction studies were performed to determine which clones resulted in the highest protein production and to deduce the optimal recombinant protein induction time. According to these pilot studies, the best rG-CSFR-expressing E. coli culture was scaled up into several 500 ml LB + kanamycin (50µg/ml) cultures, grown for 2 h and induced with IPTG (1 mM final concentration) for an additional 4 h. The cultures were then collected by centrifugation and lysed by 3 repeated freeze-thaw cycles in the presence of B-PER bacterial protein extraction reagent (ThermoScientific). The lysates were then incubated with Ni-NTA agarose (Qiagen) to isolate the (His<sup>6</sup> -tagged) rG-CSFR. The isolation of the rG-CSFR using the Ni-NTA beads was conducted using the same procedure as described above for the rCXCL8a/b isolation. The rG-CSFR elution fractions containing the protein were pooled, concentrated against polyethylene glycol flakes (8 kDa) at 4◦C, dialyzed overnight against saline at 4◦C and the protein concentration was determined by the Bradford protein assay (BioRad). The isolated rG-CSFR (1 mg total) was submitted for rabbit immunization protocols (ProSci Inc). The resulting rabbit immune sera (2 rabbits) were examined by western blot against the rG-CSFR to determine which of the two sera had greater detection of the recombinant, using secondary HRP-conjugated goat anti-rabbit IgG (ThermoScientific) to for the detection. The corresponding serum was then applied to a HiTrap Protein A HP column (GE Health) to isolate the IgG fraction from the rabbit serum. A Sulfo-Link Protein Kit (ThermoScientific; according to manufacturer's instructions) was then used to purify the IgG fraction that cross-reacted with the rG-CSFR. To confirm the specificity of this reagent, we pre-absorbed this purified anti-rG-CSFR IgG against the rG-CSFR prior to western blot analysis of rG-CSFR. Whereas the non-pre-absorbed antibody detected the rG-CSFR at the expected 40 kDa molecular weight (**Supplemental Figure 2A**, lane 1), the rGCSFR-pre-absorbed antibody did not (**Supplemental Figure 2A**, lane 2), confirming the specificity of this reagent.

For staining of rCXCL8a- and rCXCL8b-elicited granulocytes, the recovered cells were cyto-centrifuged onto glass slides, stained with the HiTrap Proetin A HP column and A Sulfo-Link Protein Kit-purified anti-rG-CSFR primary rabbit antibody (2.5 µg total) overnight at 4◦C, and goat anti-rabbit IgG Dylight 488 (ThermoScientific) secondary antibodies (1 h) and counterstained with Hoechst nuclear stain (ThermoScientific).

### Statistical Analysis

Statistical analysis was conducted using a one-way analysis of variance (ANOVA) and post-hoc t-test, using Vassar Stat (http:// vassarstats.net/anova1u.html). Probability level of P < 0.05 was considered significant.

### RESULTS

### *In silico* Analyses of *X. laevis* CXCL8a and CXCL8b

While all of the vertebrate CXCL8 proteins exhibit the CXC motif, with the exception of haddock (Gadiformes), teleost and cartilaginous fish CXCL8 proteins lack the characteristic ELR motif (**Figure 1A**). Interestingly, while the frog (X. laevis) CXCL8a protein possesses an ELR motif, the frog CXCL8b lacks this motif (**Figure 1A**). Moreover, these X. laevis CXCL8a and CXCL8b are fairly different in their respective protein sequences (**Figure 1A**).

We performed phylogenetic analysis to discern the evolutionary relationships between the amphibian CXCL8 isoforms and the other vertebrate CXCL8 proteins (**Figure 1B**). The mammalian CXCL8 protein sequences branched as a separate clade from all other vertebrate CXCL8 sequences (**Figure 1B**). The cyprinid fish CXCL8\_L2 sequences formed a separate clade and the avian and reptile CXCL8 sequences together formed a distinct clade (**Figure 1B**). Notably, the amphibian CXCL8b proteins formed a distinct clade and branched ancestrally to the avian and reptilian CXCL8 clade as well as to the independent amphibian CXCL8a clade. The bony (including the cyprinid CXCL8\_L1s) and cartilaginous fish CXCL8 sequences also split into respective clades that branched ancestrally to all other vertebrate CXCL8s (**Figure 1B**).

### *X. laevis* Tadpoles and Adult Frogs Exhibit Distinct Expression Patterns of *Cxcl8a* and *Cxcl8b*

We examined the expression of Cxcl8a and Cxcl8b genes in the tissues of X. laevis tadpoles and adult frogs to determine

FIGURE 1 | *In silico* analyses of CXCL8 phylogeny and protein sequence conservation.(A) The protein alignment was performed using ClustalW2 server. Fully conserved residues are indicated by an asterisk (\*), partially conserved and semi-conserved substitutions are represented by ":" and ".", respectively. Putative signal peptides are bolded, the ELR motif is boxed in gray and the conserved CXC motif is highlighted. (B) The phylogenetic tree was *(Continued)* FIGURE 1 | constructed using the neighbor joining method and bootstrapped 10,000 times (denoted as %s). The accession numbers for the respective protein sequences are: guinea pig CXCL8: NP\_001166870.1; naked mole rat CXCL8: XP\_004833980; rhesus monkey CXCL8: NP\_001028137.1; chimpanzee CXCL8: NP\_000575.1; human CXCL8: NP\_000575.1; horse CXCL8: NP\_001077420.2; hare CXCL8: ALG04568.1; pig CXCL8: NP\_999032.1; cow CXCL8: NP\_776350.1; zebrafish CXCL8\_L2: XP\_009305130.1; carp CXCL8\_L2: XP\_018936341.1; chicken CXCL8: NP\_990349.1; painted turtle CXCL8: XP\_005304195.1; soft shelled turtle CXCL8: ACP28489.1; alligator CXCL8: XP\_006018817.1; kiwi CXCL8: XP\_013807042.1; barn owl CXCL8: XP\_009963343.1; chimney swift CXCL8: XP\_010001929.1; Himalayan frog CXCL8: XP\_018421489.1; Western clawed frog CXCL8a: XP\_002942578.2; African clawed frog CXCL8a: OCU00045.1; Western clawed frog CXCL8b: XP\_002942578.2; African clawed frog CXCL8b: NP\_0010912223.1; zebrafish CXCL8\_L1: XP\_001342606.2; carp CXCL8\_L1: XP\_016375461; trout CXCL8: XP\_020330727.1; salmon CXCL8: NP\_001134182.1; haddock CXCL8: CAD97422.2; gilthead seabream CXCL8: AGS55343.1; fugu CXCL8: NP\_001027759.1; snapper CXCL8: AGV99968.1; banded houndshark CXCL8: BAB79448.1; whale shark CXCL8: XP\_020370926.1; elephant shark CXCL8: NP\_001279539.1.

if these two genes are under similar or distinct transcriptional regulation (**Figure 2A**). Tadpoles possessed significantly greater levels of Cxcl8a than Cxcl8b mRNAs in their kidney and skin tissues (**Figure 2A**). Adult frogs possessed significantly greater Cxcl8b transcripts in their liver and spleen tissues but greater Cxcl8a mRNA levels in their skin and intestines (**Figure 2A**). Compared to adult frogs, tadpoles possessed significantly greater levels of Cxcl8a transcripts in their kidney, skin and intestine tissues (**Figure 2A**). By contrast, adult frogs exhibited greater splenic expression of Cxcl8b than tadpoles (**Figure 2A**).

### FV3-Challenged Tadpoles and Adult Frogs Exhibit Distinct *Cxcl8a* and *Cxcl8b* Gene Expression

Anuran (frogs and toads) tadpoles are substantially more susceptible to the Frog Virus 3 ranavirus than the respective adult frogs (25–29). Notably, we recently demonstrated that this susceptibility stems at least in part from the inability of FV3-challenged tadpoles to recruit granulocytes into their kidneys (20), which are a central site of FV3 replication (24). Here we examined the expression of Cxcl8a and Cxcl8b in the kidneys of tadpoles and adults 3 days after FV3 infection, to discern the potential roles of these chemokines in this granulocyte recruitment (**Figure 2B**). After 3 days of FV3 infection, tadpoles did not exhibit significant changes in their expression of Cxcl8a and possessed significantly decreased gene expression of Cxcl8b and the granulocyte colony stimulating factor receptor (Gcsfr) granulocyte marker (**Figure 2B**). By contrast, FV3-infected adult frogs possessed significantly elevated Cxcl8a (but not Cxcl8b) mRNA levels, concomitant with increased kidney Gcsfr gene expression (**Figure 2B**), suggesting that Cxcl8a (but not Cxcl8b) may be involved in the adult frog granulocyte recruitment to this FV3 infection site.

kidney (kid), liver (liv), muscle (mus), spleen (spl), skin (sk), bone marrow (bm), and intestine (int) tissues (*N* = 6). (B) *Cxcl8a*, *Cxcl8b, and Gcsfr* gene expression in tadpole and adult kidneys 3 days post FV3 infection (10<sup>4</sup> PFU/tadpole; 5 × 10<sup>6</sup> PFU/adult) (*N* = 5). (C) *Cxcl8a* and *Cxcl8b* gene expression in amputated tadpole tails (*N* = 5). (D) *Cxcl8a* and *Cxcl8b* gene expression in adult frog hind leg skin wounds (*N* = 5). All gene expression was compared relative to *Gapdh* control and all results are presented as means + SEM. Above-head letters denote statistical designations: experimental groups described by distinct letters are statistically different (*P* < 0.05) while those marked by the same letters are not.

### The Frog *Cxcl8a* and *Cxcl8b* Genes Are Differentially Expressed During Wounding and Repair

To further define the potential roles of the frog CXCL8a and CXCL8b under inflammatory and wound repair settings, we examined the expression of Cxcl8a and Cxcl8b during the amputation and regeneration of tadpole tails and following wounding and repair of adult frog skin tissues (**Figures 2C,D**, respectively). Tadpoles exhibited elevated Cxcl8a but not Cxcl8b gene expression 5 h and 1 day post tail clipping (**Figure 2C**) and adult frogs likewise possessed significantly increased Cxcl8a but not Cxcl8b mRNA levels 1 day post skin wounding. Six days after the tail clipping, tadpoles had regenerated a substantial proportion of their tails, and this corresponded to significantly elevated Cxcl8b (but not Cxcl8a) gene expression (**Figure 2C**). By 12 days post injury, adult frogs skins were completely healed but we did not see changes in the expression of the adult skin Cxcl8b at any examined time during the skin wounding/repair study (**Figure 2D**).

### Frog Myeloid Cells Respond to Immune Challenge by Upregulating *Cxcl8a*

Our in vivo expression studies indicated that frogs increased their Cxcl8a but not Cxcl8b gene expression following immune challenge and under inflammatory settings. We previously demonstrated that the tadpole and the adult frog granulocytecolony stimulating factor (G-CSF)-differentiated granulocytes (Grn) and the adult frog macrophages (Mφs) differentiated by the macrophage-colony stimulating factor (M-CSF) and interleukin-34 (IL-34) cytokines are important innate immune effectors of these animals (20, 21). Accordingly, here we examined the expression of Cxcl8a and Cxcl8b in tadpole rG-CSF-derived peritoneal granulocytes and adult frog bone marrow-derived, M-CSF- or IL-34-Mφs and G-CSFgranulocytes (**Figure 3**) We previously demonstrated that these respective cultures comprise predominantly (85–95%) of cells that morphologically represent the respective populations (**Figure 3A**) and express distinct myeloid markers and immune genes (20, 23).

At steady state, the adult frog G-CSF-granulocytes had the greatest expression of Cxcl8a, with comparable Cxcl8a transcripts detected in the adult frog M-CSF- and IL-34-Mφs and much lower Cxcl8a mRNA in the tadpole G-CSF-granulocyte (**Figure 3B**). The adult frog G-CSF-granulocytes, M-CSF- and IL-34-Mφs exhibited similar Cxcl8b expression levels, which were higher than the Cxcl8b transcript levels detected in the tadpole G-CSF-granulocytes (**Figure 3B**). While all of the examined adult frog myeloid populations possessed greater expression of Cxcl8a than Cxcl8b, the tadpole granulocytes expressed similar levels of both chemokine isoforms (**Figure 3B**).

Since the tadpole cells exhibited such negligible baseline Cxcl8a and Cxcl8b expression, we focused on the adult myeloid cell populations to examine their expression of these chemokine genes following in vitro challenge with either heatkilled E. coli (**Figure 3C**) or FV3 (**Figure 3D**). Notably, after challenge with either heat-killed E. coli or FV3, all three immune populations upregulated their Cxcl8a gene expression (**Figures 3A,B**). While the E. coli and FV3-stimulated G-CSFgranulocytes also upregulated their Cxcl8b transcript levels (albeit to a much lesser extent than Cxcl8a), the M-CSF- and IL-34-Mφs did not (**Figures 3A,B**).

### The Frog CXCL8a and CXCL8b Possess Distinct Chemotactic Capacities

The distinct Cxcl8a and Cxcl8b gene expression patterns suggested that these two chemokine isoforms could have nonoverlapping functional roles. To discern this possibility, we produced both CXCL8a and CXCL8b in recombinant form (rCXCL8a and rCXCL8b) and examined the dose-dependent capacities of these respective proteins to chemoattract tadpole and adult frog granulocytes (**Figures 4A,B**, respectively), using blind well chemotaxis chambers. While both rCXCL8a and rCXCL8b elicited characteristic bell shaped dose-dependent chemotaxis of tadpole granulocytes, the rCXCL8b-induced chemotaxis peaked at higher concentrations of the recombinant chemokine (10<sup>1</sup> ng/mL) and decreased at subsequently lower doses **(Figure 4A**). By contrast, the rCXCL8a-mediated tadpole granulocyte chemotaxis peaked at much lower concentration (10−<sup>4</sup> ng/mL; **Figure 4A**). Similarly, adult frog granulocyte chemotaxis toward rCXCL8a peaked at a higher concentration of the chemokine (10−<sup>6</sup> ng/ml) than chemotaxis toward rCXCL8b (10−<sup>4</sup> ng/ml; **Figure 4B**), together indicating that CXCL8b is a more potent chemoattractant of both tadpole and adult granulocytes than CXCL8a.

To confirm that the observed cell migration was gradient dependent (chemotaxis) rather than increased random cell motility (chemokinesis), we performed chemokinesis experiments using tadpole and adult granulocytes. To this end, we abolished rCXCL8a and rCXCL8b gradients by adding the respective optimal doses of either chemokine to both upper and lower chemotaxis chambers and measured tadpole and adult frog granulocyte migration (**Figures 4A,B**). Tadpole granulocyte chemotaxis toward rCXCL8a and rCXCL8b, and the adult frog granulocyte chemotaxis toward rCXCL8a (but not to rCXCL8b) was substantially reduced, but not completely abolished under these conditions, indicating that some of the observed granulocyte migration was due to chemokinesis (**Figure 4A,B**). Conversely, the adult frog granulocyte migration toward rCXCL8b was abolished in the chemokinesis experiments, indicating that this migration was entirely gradient dependent chemotaxis and not chemokinesis (**Figure 4B**).

Because the tadpole and adult frog granulocytes differed in their chemotactic activity toward the rCXCL8a and rCXCL8b (**Figures 4A,B**), we examined their gene expression of the putative CXCL8 receptors, Cxcr1 and Cxcr2 (30) (**Figure 4C**). The tadpole G-CSF-granulocytes exhibited greater expression of both Cxcr1 and Cxcr2 genes than the adult G-CSF granulocytes (**Figure 4C**). Interestingly, while the tadpole granulocytes exhibited significantly greater transcript levels of Cxcr2 than Cxcr1, the adult frog granulocytes possessed similar mRNA levels for both receptors (**Figure 4C**).

### The Frog CXCL8a and CXCL8b Signal Through CXCR1 and CXCR2

The mammalian CXCR1 and CXCR2 receptors may be pharmacologically inhibited by 1 and 100 nM of reparixin, respectively (30). To discern whether the tadpole and adult frog granulocyte chemotaxis to rCXCL8a and rCXCL8b was mediated by CXCR1 and/or CXCR2, we performed chemotaxis experiments using optimal concentrations of the respective chemokines in the absence or presence of 1 or 100 nM of reparixin (**Figures 5A,B**). At 1 nM, reparixin reduced the tadpole and adult frog granulocyte chemotaxis toward rCXCL8a to background levels and significantly reduces these cells' migration toward rCXCL8b, albeit not to background levels (**Figures 5A,B**). The 100 nM reparixin treatment was less effective at inhibiting the tadpole and adult frog granulocyte chemotaxis toward rCXCL8a, abolished the tadpole cell migration toward rCXCL8b to background levels, and decreased the adult frog granulocyte migration toward rCXCL8b to levels comparable to those seen at the 1 nM dose of the antagonist (**Figures 5A,B**).

To reevaluate the roles of CXCR2 in the rCXCL8a and rCXCL8b chemotaxis of tadpole and adult frog granulocytes, we repeated the above experiments, this time utilizing a specific CXCR2 inhibitor, SB265610 (**Figures 5C,D**). This compound inhibits the mammalian neutrophil chemoattractantinduced calcium mobilization and neutrophil chemotaxis at 3.7 and 70 nM, respectively (31). Accordingly, we used final SB265610 concentrations of 5 and 100 nM for our tadpole and adult frog granulocyte CXCR2 inhibition assays. At 5 nM,

SB265610 had a modest but significant inhibitory effect on the tadpole granulocyte chemotaxis toward rCXCL8a and rCXCL8b (**Figure 5C**). The 100 nM dose of SB265610 did not further decrease the tadpole granulocyte chemotaxis toward rCXCL8a (**Figure 5C**). Interestingly, the 100 nM dose of SB265610 resulted in significantly greater inhibition of the rCXCL8bmediated chemotaxis than the inhibition of the rCXCL8amediated migration (**Figure 5C**). Concurrently, this inhibition of rCXCL8b-mediated activity was significantly greater than that seen at 5 nM of the compound (**Figure 5C**).

The 5 nM dose of SB265610 resulted in modest, but significant inhibition of the rCXCL8b-, but not the rCXCL8a-mediated chemotaxis of adult frog granulocytes (**Figure 5D**). Conversely, at the 100 nM, this inhibitor resulted in comparable significant abrogation of the adult frog granulocyte chemotaxis toward both rCXCL8a and rCXCL8b, (**Figure 5D**).

### The Tadpole CXCL8a and CXCL8b Chemoattract Distinct Granulocyte Populations

In consideration of the potentially disparate roles of the frog CXCL8a and CXCL8b, we intraperitoneally administered rCXCL8a and rCXCL8b to tadpoles and adult frogs, recovered the resulting peritoneal exudates, and examined these cells for their expression of a panel of immune genes (**Figure 6**). The rCXCL8a and rCXCL8b elicited similar numbers of adult cells bearing granulocyte morphology whereas rCXCL8a chemoattracted more cells into tadpole peritonea than rCXCL8b (**Figure 6A**). The adult rCXCL8a- and rCXCL8b-elicited cells did not exhibit significant transcriptional differences (data not shown). The tadpole rCXCL8a and rCXCL8b-chemoattracted cells also expressed comparable levels of Cxcr1, Cxcr2, NADPH oxidase catalytic subunit p67phox and myeloperoxidase (Mpo; **Figure 6C**). Conversely and in comparison to rCXCL8arecruited cells, the rCXCL8b-elicited population possessed significantly greater mRNA levels of lysozyme (**Figure 6C**) as well as genes associated with immune suppression and wound repair, including suppressor of cytokine signaling 3 (Socs3), arginase-1 (Arg1), interleukin-10 (IL-10), vascular endothelial growth factor (Vegf), and indoleamine 2,3 dioxygenase (Ido; **Figure 6B**).

While our results indicated that the tadpole G-CSF granulocytes express greater levels of Cxcr2 than Cxcr1 (**Figure 4C**), we did not see such expression patterns in the tadpole rCXCL8a- and rCXCL8b-elicited cells (**Supplemental Figure 2B**). Moreover, while the tadpole G-CSF granulocytes expressed greater levels of both of the Cxcr1 and Cxcr2 genes than the adult G-CSF granulocytes (**Figure 4C**), in fact the adult frog rCXCL8a- and rCXCL8b-elicited cells expressed greater levels of both receptors than their tadpole counterparts (**Supplemental Figure 2B**).

To confirm that the rCXCL8a- and rCXCL8b-recruited cells were in fact granulocytes, we generated a polyclonal antibody against a recombinant form of the frog G-CSFR and stained tadpole chemokine-derived cells with this reagent. Both rCXCL8a- and rCXCL8b-elicited exudates were predominantly composed of G-CSFR-positive cells bearing characteristic polymorphonuclear granulocyte morphology (**Figure 6D**).

### DISCUSSION

Vertebrate chemokine genes are believed to have diverged more rapidly with evolutionary time than most other components of the vertebrate immune system (32, 33), presumably reflecting the distinct physiological and evolutionary pressures of these diverging species. This notion is largely supported by the vast and highly distinct chemokine ligand and receptor repertoires seen across different species (34). It is interesting to consider that despite this apparent diverging evolutionary pressure on the vertebrate chemokine genes, interleukin-8 and its cognate receptors (CXCR1 and CXCR2) are retained in most of these species, albeit in multiple isoforms across some. This suggests that CXCL8 plays important biological roles that cannot be as easily amended to the evolutionary pressures faced with species' divergence. Conversely, the presence of multiple CXCL8 isoforms presumably permits the neo-functionalization of CXCL8 isoforms without compromising the indispensable, evolutionarily conserved roles of CXCL8. This is exemplified in the cyprinid fish CXCL8\_L1 and CXCL8\_L2 lineages (15, 34) as well as in the X. laevis frog CXCL8a and CXCL8b, as presented here. Indeed, the frog CXCL8a and CXCL8b possess fairly distinct protein sequences, suggesting that they have diverged with evolutionary time. In the case of the X. laevis CXCL8a and CXCL8b gene expression and function, our results suggest that the frog CXCL8a is serving the inflammatory roles attributed to other vertebrate CXCL8 molecules while the frog CXCL8b may have adopted unique biological roles, at least within the tadpole life of this animal. It is notable that adult frog granulocytes upregulated their Cxcl8b gene expression following bacterial and viral stimulation, suggesting that CXCL8b may also play a role during the adult frog immune responses to certain pathogens. Further research will revel the exact role of CXCL8b during frog immune responses.

Here we report that, Xenopodinae frogs possess both an ELR motif-containing CXCL8 (CXCL8a) as well as an CXCL8 (CXCL8b) that lacks this motif. Moreover, these chemokine protein sequences are somewhat distinct, supporting the notion that they may have diverged in their respective functions. Indeed, the ELR motif-lacking CXCL8b branches ancestrally to the frog CXCL8a proteins as well as to all higher vertebrate CXCL8s. Xenopodinae frogs are presently

known to possess three additional putative CXCR1/CXCR2 ligands; CXCL2, CXCL5, CXCL6, of which only the CXCL2 possesses an ELR motif (sequences available on GenBank). By contrast, while chickens also possess three putative CXCR1/CXCR2 ligands, including two CXCL8 isoforms,

(*P* < 0.05) while those marked by the same letters are not.

all of these proteins possess ELR motifs (35). This suggests that emergence of ELR motif-containing CXCL8 proteins as well as other CXCR1/CXCR2 ligands occurred in tetrapods. Despite lacking an ELR motif, the fish CXCL8 proteins are chemoattractive to neutrophils (11, 12), bringing to question the functional significance of the emergence of ELR-bearing CXCL8 chemokines. Further work using animal models like Xenopus, which possess both ELR-containing and ELR-lacking CXCL8 chemokines will be invaluable to addressing this question.

Our chemotaxis experiments using recombinant forms of the X. laevis CXCL8a and CXCL8b indicate that these proteins have distinct capacities to chemoattract tadpole and adult frog granulocytes. In general, when performing in vitro chemotaxis assays, increasing chemokine concentrations results in the loss of the chemokine gradient across the chemotaxis chambers, leading to chemokine receptor saturation and preventing further migration, which results the characteristic bell shape of the doseresponse curves (36), akin to those reported here for rCXCL8a and rCXCL8b. In this respect, it is notable that in the case of both

FIGURE 6 | Analysis of immune gene expression and cytology of rCXCL8aand rCXCL8b-elicited tadpole cells. Tadpoles and adult frogs were injected intraperitoneally with rCXCL8a or rCXCL8b (1µg/g of body weight) in 10 µl of saline or with an equal volume of the vector control (supernatants from empty vector-transfected Sf9II cell, processed in parallel to rCXCL8a and rCXCL8b production). After 4 h, animals were lavaged with saline and the cells were enumerated (A). The tadpole (B,C) and adult frog (not presented) rCXCL8a and rCXCL8b-elicited cell immune gene expression was examined relative to the *Gapdh* control (*N* = 6). All results are presented as means + SEM. Above-head letters denote statistical designations: experimental groups described by distinct letters are statistically different (*P* < 0.05) while those marked by the same letters are not. (D) Tadpole rCXCL8a and rCXCL8b-elicited cells were stained with a rabbit anti-frog rG-CSFR (or saline) Ab and secondary goat anti-rabbit Ab and examined by confocal microscopy.

tadpole and adult granulocyte chemotaxis, the migration peaked at higher doses of rCXCL8a than rCXCL8b. This indicates that rCXCL8a-mediated receptor saturation occurs at higher doses than with rCXCL8b, in turn suggesting that the receptor-ligand interactions of these two chemokines are distinct.

While 1 and 100 nM concentrations of reparixin block the signaling through the mammalian CXCR1 and CXCR2, respectively (30), our results indicated that the 1 nM reparixin effectively blocked the rCXCL8a- and rCXCL8b-mediated chemotaxis while the 100 nM dose did not confer further inhibitory effects. Notably, our pharmacological inhibition of CXCR2 resulted in decreased tadpole and adult frog granulocyte chemotaxis toward rCXCL8a and rCXCL8b, together indicating that both CXCR1 and CXCR2 are engaged by the CXCL8a and CXCL8b. These results may reflect the fact that the downstream signaling through the frog CXCR1 and CXCR2 is significantly more sensitive to reparixin inhibition than the counterpart mammalian receptors. Interestingly, the tadpole but not the adult frog granulocytes expressed greater levels of Cxcr2 than Cxcr1 while the 100 nM dose of the CXCR2 antagonist resulted in significantly greater reduction in tadpole (but not adult frog) granulocyte chemotaxis toward rCXCL8b, than rCXCL8a. Possibly, the frog CXCL8b relies more heavily on signaling through the CXCR2 for the recruitment of tadpole granulocytes. In this respect, we were surprised to find that the tadpole granulocytic cells recruited by rCXCL8a and rCXCL8b did not exhibit the greater Cxcr2 expression seen in the tadpole G-CSF-granulocytes. Presumably, the G-CSF-granulocytes, the rCXCL8a- and rCXCL8b-elicited granulocytes represent distinct tadpole immune populations and/or differentially activated granulocyte subsets. The mammalian granulocyte gene expression of Cxcr1 and Cxcr2 is tenuous and subject to their activation states (37), so it is possible that the downstream signaling elicited by CXCL8b in tadpole granulocytes may be dampening the Cxcr2 gene expression or enhancing the gene expression of Cxcr1. Moreover, X. laevis are presently thought to encode single copies of Cxcr1 and Cxcr2 genes, ruling out the possibility that CXCL8a and CXCL8b function through distinct receptor isoforms. In consideration of our findings, we speculate that the frog CXCL8a and CXCL8b may have distinct affinities for CXCR1 and CXCR2, culminating in their disparate chemotactic properties. It is also notable that tadpole and adult granulocyte chemotaxis peaked at distinct concentrations of either chemokine, suggesting differences between tadpole and adult responsiveness to CXCL8a and CXCL8b. These differences may be explained by the differences in the Cxcr1 and Cxcr2 gene expression by the tadpole and adult frog cells.

Amphibian tadpoles differ from adult frogs in their ability to regenerate amputated limbs (38) and tadpoles experience a refractory period during which they temporarily lose this regenerative capacity (39). Interestingly, this refractory period may be circumvented by treating tadpoles with certain antiinflammatory, immunosuppressive, or antioxidant agents (39). This is particularly notable considering that our results indicate that tadpole Cxcl8b gene expression is increased during wound repair while tadpole (but not adult) rCXCL8brecruited granulocytes express immunosuppressive and antiinflammatory genes (IL-10, Socs3, Ido). Indeed, IL-10 is a hallmark anti-inflammatory mediator (40), SOCS3 is associated with immunosuppressive immune states (41), and the IDO enzyme is associated with immune modulation and the induction of immunological tolerance (42). Moreover, the

tadpole rCXCL8b-elicited granulocytes also exhibited elevated expression of vascular endothelial growth factor and arginase-1, which are very important in angiogenesis (43) and to tissue repair (44), respectively. Together, our findings strongly suggest that in tadpoles, CXCL8b is involved in recruiting a subset of granulocytes that are functionally polarized akin to alternatively-polarized (M2) macrophages (45), as immunosuppressive, healing effectors that may be playing a role during tadpole wound-healing and tissue regeneration.

The duplication of vertebrate genes and their subsequent neo-functionalization is widely believed to being a major driving force behind speciation and species-specific physiological diversification (46). Considering the unique physiological and pathogenic pressures that have molded amphibian physiology, together with their highly distinct tadpole and adult life stages; it is intuitive that they would have developed distinct mechanisms for dealing with their physiological demands. Amphibian tadpole development is closely linked to their temporal regulation of inflammatory genes (47) and it is intriguing to consider that their use of CXCL8b as a means for immune suppression and tissue repair may have evolved out of the inflammatory functions associated with CXCL8.

### ETHICS STATEMENT

This study was carried out in accordance with the recommendations of and following approval by the Institutional Animal Care and Use Committee (approval number 15-024).

## AUTHOR CONTRIBUTIONS

DK and LG designed and planned the studies. DK, AY, and MP performed the experiments. DK and LG analyzed the data, wrote the manuscript, and prepared the figures.

### REFERENCES


### FUNDING

This work was funded by George Washington University beginning investigator start-up support.

### ACKNOWLEDGMENTS

DK thanks the GWU Harlan Research Program. AY thanks the GWU, Dept. Biological Sciences for GTA support and support from the GWU Harlan Research Program. LG thanks the GWU for research support at the early investigator stage.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2018.02058/full#supplementary-material

Supplemental Figure 1 | Western blot of the produced, purified, and eluted rCXCL8a and rCXCL8b fractions. The eluted recombinant (r)CXCL8a and rCXCL8b were resolved by SDS PAGE, transferred onto nitrocellulose membranes and western blots were performed using an HRP-conjugated mouse anti-V5 (Sigma) to determine which elution fractions contained rCXCL8a (15 kDa) and rCXCL8b (16 kDa).

Supplemental Figure 2 | Analyses of the anti-G-CSFR polyclonal antibody and the rCXCL8a- and rCXCL8b-elicited granulocyte expression of the *Cxcr1* and *Cxcr2* genes. (A) Rabbits were immunized with rG-CSFR and the resulting serum was applied to a HiTrap Proetin A HP column (GE Health) to isolate the IgG fraction, and to a rG-CSFR-bound Sulfo-Link Protein column to purify the IgG fraction that cross-reacted with the rG-CSFR. To confirm the specificity of this reagent, this anti-rG-CSFR IgG farction was used to perform a western blot of the rG-CSFR before (lane 1) or after (lane 2) pre-absorbing the Ab against the rG-CSFR. (B) Tadpoles and adult frogs were injected intraperitoneally with rCXCL8a or rCXCL8b (1µg/g of body weight) in 10 µl of saline or with an equal volume of the vector control. After 4 h, animals were lavaged with saline and the cells were enumerated and examined for their expression of *Cxcr1* and *Cxcr2.* Above-head letters denote statistical designations: experimental groups described by distinct letters are statistically different (*P* < 0.05) while those marked by the same letters are not.


defense against Salmonella Typhimurium. Dev Comp Immunol. (2015) 49:44– 8. doi: 10.1016/j.dci.2014.11.004


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

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

# Review of the Amphibian Immune Response to Chytridiomycosis, and Future Directions

Laura F. Grogan<sup>1</sup> \*, Jacques Robert <sup>2</sup> , Lee Berger 3,4, Lee F. Skerratt 3,4 , Benjamin C. Scheele5,6, J. Guy Castley <sup>1</sup> , David A. Newell <sup>7</sup> and Hamish I. McCallum<sup>1</sup>

*<sup>1</sup> Environmental Futures Research Institute and School of Environment and Science, Griffith University, Nathan, QLD, Australia, <sup>2</sup> University of Rochester Medical Center, Rochester, NY, United States, <sup>3</sup> One Health Research Group, College of Public Health, Medical and Veterinary Sciences, James Cook University, Townsville, QLD, Australia, <sup>4</sup> Faculty of Veterinary and Agricultural Sciences, Melbourne Veterinary School, University of Melbourne, Werribee, VIC, Australia, <sup>5</sup> Fenner School of Environment and Society, The Australian National University, Canberra, ACT, Australia, <sup>6</sup> Threatened Species Recovery Hub, National Environmental Science Program, Fenner School of Environment and Society, The Australian National University, Canberra, ACT, Australia, <sup>7</sup> Forest Research Centre, School of Environment, Science and Engineering, Southern Cross University, Lismore, NSW, Australia*

#### Edited by:

*Leon Grayfer, George Washington University, United States*

#### Reviewed by:

*Louise A. Rollins-Smith, Vanderbilt University, United States Carly Rae Muletz-Wolz, National Zoological Park (SI), United States*

> \*Correspondence: *Laura F. Grogan l.grogan@griffith.edu.au*

#### Specialty section:

*This article was submitted to Comparative Immunology, a section of the journal Frontiers in Immunology*

Received: *29 June 2018* Accepted: *15 October 2018* Published: *09 November 2018*

#### Citation:

*Grogan LF, Robert J, Berger L, Skerratt LF, Scheele BC, Castley JG, Newell DA and McCallum HI (2018) Review of the Amphibian Immune Response to Chytridiomycosis, and Future Directions. Front. Immunol. 9:2536. doi: 10.3389/fimmu.2018.02536* The fungal skin disease, chytridiomycosis (caused by *Batrachochytrium dendrobatidis* and *B. salamandrivorans*), has caused amphibian declines and extinctions globally since its emergence. Characterizing the host immune response to chytridiomycosis has been a focus of study with the aim of disease mitigation. However, many aspects of the innate and adaptive arms of this response are still poorly understood, likely due to the wide range of species' responses to infection. In this paper we provide an overview of expected immunological responses (with inference based on amphibian and mammalian immunology), together with a synthesis of current knowledge about these responses for the amphibian-chytridiomycosis system. We structure our review around four key immune stages: (1) the naïve immunocompetent state, (2) immune defenses that are always present (constitutive defenses), (3) mechanisms for recognition of a pathogen threat and innate immune defenses, and (4) adaptive immune responses. We also evaluate the current hot topics of immunosuppression and immunopathology in chytridiomycosis, and discuss their respective roles in pathogenesis. Our synthesis reveals that susceptibility to chytridiomycosis is likely to be multifactorial. Susceptible amphibians appear to have ineffective constitutive and innate defenses, and a late-stage response characterized by immunopathology and Bd-induced suppression of lymphocyte responses. Overall, we identify substantial gaps in current knowledge, particularly concerning the entire innate immune response (mechanisms of initial pathogen detection and possible immunoevasion by Bd, degree of activation and efficacy of the innate immune response, the unexpected absence of innate leukocyte infiltration, and the cause and role of late-stage immunopathology in pathogenesis). There are also gaps concerning most of the adaptive immune system (the relative importance of B and T cell responses for pathogen clearance, the capacity and extent of immunological memory, and specific mechanisms of pathogen-induced immunosuppression). Improving our capacity for amphibian immunological research will require selection of an appropriate Bd-susceptible model species, the development of taxon-specific affinity reagents and cell lines for functional assays, and the application of a suite of conventional and emerging immunological methods. Despite current knowledge gaps, immunological research remains a promising avenue for amphibian conservation management.

Keywords: chytridiomycosis, immune, innate, adaptive, frogs, declines, amphibian, Batrachochytrium dendrobatidis

### INTRODUCTION

Two decades have passed since the discovery and characterization of the devastating amphibian skin disease, chytridiomycosis, caused by two unusual fungal species that likely originated in Asia (1, 2). Batrachochytrium dendrobatidis (hereafter Bd) was first detected in the 1990s and is now widespread globally (3, 4), whereas B. salamandrivorans (Bsal) primarily affects salamanders (5) and was recently described after arriving in Europe in 2010. Chytridiomycosis is reported to affect >350 amphibian species and has had a dramatic worldwide impact on amphibian biodiversity, having caused the decline and possible extinction of greater than 200 species (6, 7). Chytridiomycosis has thus been the subject of intense study, with research focused on understanding fungal virulence, pathogenesis, immunology, treatment, and epidemiology [reviewed in (8–11)]. A central aim of this research has been finding ways to mitigate disease in the field to reduce or prevent further species declines and extinctions (12, 13).

The evolution of resistance and/or tolerance to infection is a key long-term goal for managing in situ amphibian populations in regions where Bd is now enzootic (14), and immunological research is central to this goal. A recent study demonstrated that the fungus can maintain high virulence post-emergence (15), which may be a result of its broad host range (where fungal persistence may not be affected by the loss of highly susceptible host species). However, many amphibian species are recovering in the wild (10), and some have increased survival rates consistent with improved immunity (16). A study comparing skin secretion inhibitory activity against Bd pre- and post- emergence suggests that the evolution of natural immunity may be occurring in some species in situ (15). Several studies have made progress uncovering other putative mechanisms for improved immunity, including directional selection of major histocompatibility complex (MHC) alleles (17–21). Unfortunately, many endangered frog species appear to be running out of time. Without sufficient genetic, phenotypic, or behavioral evolution of the host, many susceptible populations remain threatened by chytridiomycosis and are experiencing ongoing declines, sometimes decades post-initial chytridiomycosis outbreaks (10, 22–24). Other susceptible species may persist despite chytridiomycosis-associated mortality due to high reproductive capacity. However, compensatory recruitment may be reducing selection pressure for the evolution of immunity (25), and these populations remain highly vulnerable to other threats (26). Furthermore, animals repatriated from captivity continue to succumb to disease in the field (27, 28).

While the amphibian immune response to chytridiomycosis has been the subject of some research to date, many aspects remain poorly understood, likely owing to the complexity of the system and the vast range in species' responses to infection. Indeed, Bd and Bsal are the main fungi from their phylum found to cause disease in vertebrates, and the observed host immune response to these pathogens appears to depart from an expected "normal" immune response to an intracellular or fungal pathogen. Previous reviews [e.g., (11, 29–31) have covered (1) components of innate immune defenses such as secretion of skin antimicrobial peptides, and maintenance of symbiotic skin bacteria and their antifungal metabolites (29, 32), and (2) adaptive immune components such as MHC allele selection, antibody production, and lymphocyte responses (33, 34). However, the field is overdue for an update that incorporates the results of recent transcriptomic and immunogenetic studies, as well as to provide a more thorough overview of the role of key immune components. Concerning the innate arm of the immune system, virtually nothing is known about the role of pattern recognition receptors (PRRs), complement, cytokines and chemokines, macrophages and dendritic cells, other phagocytes, and natural killer cells. For the adaptive arm of the immune system, besides the possible inhibition of lymphocyte proliferation response by Bd and importance of antibodies in the skin of infected frogs, very little is known about B and T cell responses, immunological memory and antigen detection. Improving our capacity for amphibian immunological research and our understanding of the host immune response to chytridiomycosis may result in numerous applied benefits. These may include: (1) identifying targets for further research, treatment, and marker-assisted evolution, (2) identifying immunologic management strategies including environmental manipulation, vaccine design, selective breeding, genetic engineering and pathogen virulence attenuation, and (3) predicting species at continued risk of decline and implementing timely mitigation measures.

In this review, we present an integrated synthesis of current understanding of the amphibian host immune response to chytridiomycosis within the classical scaffold of innate and adaptive immunological mechanisms [reviewed in (35)]. We have targeted this review for amphibian chytridiomycosis researchers, but we expect it will also be of interest for researchers in the broader fields of fungal immunology and amphibian conservation. We focus specifically on host mechanisms; predominantly in response to Bd [host responses to Bsal are likely similar but are currently poorly understood; reviewed in (11)]. We do not attempt to review the vast range of factors contributing to variations in susceptibility to infection between individuals and species. For a broad introductory overview of chytridiomycosis, see **Box 1**. For convenience, we provide a glossary of terms and abbreviations in **Box 2**. Throughout this review, where amphibian-specific immune knowledge is lacking, we instead refer to the better characterized immune system of mammals. Please see **Box 3** for a brief comparison between amphibian and mammalian immune systems. We also focus primarily on post-metamorphic and adult amphibians (especially anurans) because larval amphibians (tadpoles) usually survive Bd infections that localize to their keratinized mouthparts (see **Box 4** for a brief overview of tadpole vs. post-metamorphic immune systems). We start by outlining several key (non-mutually exclusive) immune stages: (1) the naïve immunocompetent state, (2) immune defenses that are always present (constitutive defenses), (3) mechanisms for pathogen recognition and induction of innate immune defenses, and (4) adaptive immune defenses. For each stage, we briefly describe the expected immune response to an invading infectious organism such as Bd (see **Figure 1**), then compare it with current knowledge of chytridiomycosis, highlighting research gaps. We then examine and discuss evidence for the role of immunosuppression and immunopathology in chytridiomycosis. We conclude by suggesting future research directions that will contribute to improving mitigation strategies for chytridiomycosis.

### THE IMMUNOCOMPETENT UNINFECTED STATE

Normal uninfected integument of an immunocompetent amphibian host consists of epidermal and deeper dermal layers (**Figure 1A**). The epidermis constitutes an immediate innate



physical defense barrier against pathogen invasion and consists of cell layers that mature as they migrate toward the skin surface. Epidermal layers include the basal lamina (basement membrane), then the roughly cuboidal or columnar-shaped proliferative cells of the stratum germinativum followed by the stratum spinosum and the stratum granulosum, through to the highly differentiated keratinized squamous epithelial cells found at the surface of the skin, the stratum corneum. These superficial cells are joined by tight junctions, which help maintain the skin barrier (35, 72). Intermittent sloughing of the outer layer of the epidermis, the stratum corneum, may assist in the physical removal of skin microorganisms (73). On the surface of this uppermost stratum sits a layer of mucus produced by mucous glands. This mucus contains a number of defensive molecules, including (1) lysozyme and other enzymes produced by phagocytes and keratinocytes, (2) antimicrobial peptides secreted via serous glands, (3) mucosal antibodies, as well as (4) commensal symbiotic bacterial communities together with their secreted antimicrobial compounds [reviewed in (11, 35)]. A number of peripheral immune surveillance cells are typically present in the epidermis, particularly epidermal dendritic cells including putative Langerhans cells (74), which in Xenopus spp. express high levels of MHC class II molecules (75, 76). These various dendritic cells are likely to serve as efficient antigen presenting cells, although this remains to be demonstrated. Dendritic epidermal T cells (DETCs), and gamma/delta T cells have also been described in Xenopus spp. (76).

The highly collagenous dermis underlies the epidermis. It consists of deeper stratum compactum and thicker and more superficial stratum spongiosum, separated in some species by the substantia amorpha granular calcified layer (77–79). The dermis provides nutrition and sensory integration to the epidermis via a network of capillaries and nerves that course through the dermis. Serous (granular or poison) glands and mucous glands are also present within the dermis, along with pigment-bearing chromatophores and smooth muscle fibers. Serous glands are often widely distributed throughout the skin and are able to discharge their contents in response to noxious stimuli. In the uninfected state, the repertoire of naïve B and T lymphocytes lie mainly quiescent within the spleen and blood, as well as in other secondary lymphoid organs such as the liver and intestine (**Figure 1A**). In amphibians, the spleen functions as the primary and major secondary lymphoid organ. Naïve B and T lymphocytes each possess a unique and specific antigen receptor combination.

### EARLY NAÏVE INFECTION AND CONSTITUTIVE DEFENSES

The earliest stage of the infection process with Bd involves the likely chemotaxis of infectious zoospores toward the skin surface [**Figure 1B,** (11, 80, 81)], whereupon they encounter mucus and any associated constitutive defenses of the skin. We discuss these defenses in the context of chytridiomycosis below.

#### Box 2 | Abbreviations and glossary of terms.


### Skin Sloughing Increases With Chytridiomycosis and May Physically Reduce Microbial Burdens

Ecdysis or skin sloughing in ectotherms can function as a constitutive innate immune defense mechanism by physically removing skin microbiota (73, 82). Sloughing usually occurs soon after dark, approximately every 3 or 4 days depending on species, and the frequency of sloughing increases with ambient temperature (73). Abnormal sloughing is a clinical sign of chytridiomycosis and corresponds with an increased frequency of sloughing and the production of smaller shed pieces (3, 83, 84). This may be the result of sporangia initiating premature keratinization and cell death in infected epidermal cells, in concert with hyperplasia and stimulated epidermal turnover as observed by electron microscopy (77). Using infrared video recordings, Ohmer et al. (84) found that while sloughing rate increased with infection, chytridiomycosis did not affect sloughing behavior, duration or rhythmicity, although diseased frogs typically did not eat the sloughed skin as they normally would. Independent of temperature, the extent to which increased sloughing reduces Bd infection loads varies depending on species' intrinsic susceptibility, and sometimes results in clearance of infection (85). As such, Ohmer et al. (84) hypothesized that skin sloughing may be both beneficial and detrimental in the face of chytridiomycosis (for example, by removing pathogens or symbiotic bacteria, and disrupting skin homeostasis). Of interest, the most resistant species they studied, Limnodynastes tasmaniensis, demonstrated increased sloughing rates at lower infectious burdens, which Ohmer et al. (85) suggested may indicate an effective induced defense response. Increased rates of sloughing in warmer environments, induced as an immune defense, or in association with behavioral fever (86, 87) may partly explain the improved clearance of Bd infection at higher temperatures (88).

### Natural Mucosal Antibodies (Generated by Innate-Like B Cells) May Inhibit Zoospores

Naïve frogs (not previously exposed to Bd) are unlikely to express specific mucosal antibodies that bind to and inhibit zoospores (89). However, as occurs in other vertebrates, natural antibodies may be present and may limit initial pathogen burdens. Natural antibodies are polyreactive against highly conserved pathogen epitopes. They are typically encoded by germline genes, and in mammals, they are produced by innate-like B cells (90–92). Little is currently known about their efficacy against Bd in amphibians.

### Lysozyme and Other Defensive Enzymes May Limit Zoospore Invasion

Lysozyme is a constitutively expressed antimicrobial enzyme found in body fluids and mucosal linings. Lysozyme from amphibian skin secretions has potent bactericidal activity (93, 94). Although typically considered an antibacterial enzyme, lysozyme has also been reported to possess antifungal properties (95, 96). Thus, amphibian lysozymes may have similar activity against pathogens such as Bd (11, 31, 97). To date there is little evidence for the role of lysozyme and other phagocyte-

#### Box 3 | The immune system of amphibians is similar to other vertebrates.

The larval and adult immune system of amphibians, in particular *Xenopus* spp. (South African clawed frog), has been subject to extensive investigation as a transitional non-mammalian model organism for comparative and evolutionary immunology and studies of immune ontogeny (46). The adult anuran immune system is fundamentally similar to other jawed vertebrates (47), and responds similarly to antigenic stimulation (48). However, there are some differences.


Although many aspects are still poorly characterized in amphibians (particularly CD4 T helper cell function), adaptive B and T cell biology is fundamentally conserved between mammals and amphibians. However, there are several notable differences.


or keratinocyte-derived constitutive enzymes (e.g., antimicrobial lectins, secretory phospholipase A2) in defense against Bd. However, Rosenblum et al. (98) reported an increase in Box 4 | Comparison between tadpole and post-metamorphic amphibian immune system.


expression of lysozyme genes in the skin of infected Xenopus tropicalis at 21 days post exposure in a transcriptomic study. Grogan et al. (18) similarly identified upregulation of lysozyme C genes in the skin of Litoria verreauxii alpina throughout infection, although the efficacy of this response for limiting Bd infection is currently unknown.

### Amphibians Produce a Range of Antimicrobial Peptides With Activity Against Bd, and These Likely Play a Role in Defense

The antimicrobial peptides (AMPs) of vertebrate skin are typically small hydrophobic cationic peptides produced by serous glands that provide non-specific defense against pathogenic organisms (99, 100). Clarke (99) describes four main categories of defensive molecules including alkaloids, steroids, biogenic amines, and other peptides and proteins. The range and quantity of AMPs produced by amphibians is remarkable among vertebrates and has been the target of medical studies for decades, particularly for use in pharmaceutical applications [e.g., (101–103)]. The production of antimicrobial peptides can be induced and modulated by the presence of microbial flora (104) and chronic corticosteroid administration (105). Serous glands release antimicrobial peptides to the skin surface at a low continuous rate, however, mild activation of the sympathetic nervous system (such as alarm caused by a predator cue) is sufficient to stimulate the contraction of gland-associated muscle fibers and the release of larger quantities of AMPs to the skin surface (89, 106). It is unlikely that zoospore invasion alone would be sufficient stimulus to produce this response.

Through in vitro growth inhibition assays, many amphibian peptides and peptide mixtures (at concentrations likely to occur in vivo) have been found to inhibit the growth of various pathogens including Bd as well as other fungi [reviewed in (31, 107)]. Antimicrobial peptide defenses are considered reliable predictors of natural resistance of amphibians to chytridiomycosis. However, the efficacy of AMPs in defense against Bd appears to vary substantially by species and other factors, which may limit the value of AMP data for predicting and mitigating amphibian declines (108–113). These factors may include: (1) intrinsic peptide efficacy against Bd as demonstrated in vitro, (2) concentration, number and type of peptides produced, (3) rate and location of release to the skin surface in response to microbial pathogens, and (4) presence of host- and Bd-secreted proteases that may degrade AMPs (106, 114, 115). For example, depletion of AMPs led to increased infection probability in resistant amphibian species Xenopus laevis (89) and Rana pipiens (116), but not in Pelophylax esculentus and P. lessonae (117). Validation studies involving the functional modification of key AMP molecules are still needed to confirm these associations. Woodhams et al. (118) found that AMP expression differed between infected and uninfected wild-caught Litoria serrata, with infected frogs demonstrating reduced expression. However, they did not identify whether this was a cause or consequence of infection. Ribas et al. (119), Rosenblum et al. (120) identified AMP genes or precursors (including preprocareulein and cathelicidin) via microarray studies of the spleen and skin of frogs. More recently, comparing the anti-Bd activity of skin secretions collected from frogs before and after Bd emergence in Panama, Voyles et al. (15) found a significant increase in inhibitory efficacy post-emergence, consistent with an evolutionary shift in the host immune response. The mechanisms underlying this change in efficacy are unknown but may involve altered concentration and diversity of peptides or improved inhibitory function.

### The Skin Microbiome May Inhibit Zoospore Colonization, and Bioaugmentation May be an Effective Management Strategy

Commensal microbial communities are present in the mucus layer on amphibian skin and may provide another mechanism of constitutive innate immunity against Bd via several mechanisms (32, 61, 121). For example, Meyer et al. (82) reported the cultivation of approximately 0.5–1.7 × 10<sup>6</sup> bacterial colonies and 1.6–2.6 × 10<sup>4</sup> fungal colonies per square cm of dorsal skin of Rhinella marina. Interestingly, detected microbial loads were much lower on ventral skin, despite being in more frequent contact with moist substrates than the dorsum. Many bacteria and fungi secrete antimicrobial compounds with repellant and growth-inhibitive properties against pathogenic microbes (122– 124). Furthermore, microbiota may also compete directly with Bd, and may functionally change host immune responses (121).

Numerous epibiotic bacterial species isolated from amphibian skin are growth inhibitive for Bd in vitro (125, 126). For example, the bacterial species Janthinobacterium lividum has shown particular promise for the anti-Bd properties of its secreted metabolite violacein, at concentrations greater than 15µM (127). In clinical Bd exposure experiments, both frogs (Rana muscosa) and salamanders (Plethodon cinereus) inoculated with J. lividum did not become infected (128), whereas depletion of bacteria resulted in high Bd infection intensities (129). This effect also extended to soil augmentation and environmental

FIGURE 1 | Amphibian host immunity schematic, depicting a histological section through the skin and progressive infection stages with Bd. The inference of the main cellular components is based on mammalian immunology and an expected "normal" immune response (including the expected response to vaccination). (A) Normal skin: Layers of uninfected frog skin epidermis including, from deepest to most superficial, the *basal lamina*, *stratum germinativum*, *stratum spinosum*, *stratum granulosum*, *stratum corneum* and the superficial mucus layer. Two immune surveillance cells are illustrated within the epidermis, an immune dendritic cell (homologous to Langerhans cell), and a dendritic epidermal T cell (dETC). Within the dermis is a capillary with the nucleated red blood cells of amphibians. An example complement of naïve B and T lymphocytes are depicted waiting quiescent in the spleen, illustrated schematically as the lower band on the figure (please note that the spleen is a separate organ and does not lie adjacent to the dermis in living amphibians). (B) Early infection: Expected immune mechanisms upon initial exposure to Bd, assuming constitutive defenses (such as AMPs and bacteria) are insufficient. Zoospores are illustrated penetrating the mucus layer, and early thalli with zoosporangia developing are illustrated inside deeper host cells. In a normal immune response, pathogen recognition should lead to the infiltration of innate immune cells, illustrated here to include macrophages and granulocytes (such as neutrophils). (C) Intermediate infection: Expected response at an intermediate stage of infection includes the recognition of antigens by dendritic cells that then differentiate into antigen presenting cells and migrate to the spleen enabling antigen-specific selection of lymphocytes. Simultaneously, membrane-bound immunoglobulin on naïve B lymphocytes is exposed to extracellular Bd antigens (transported via the blood circulatory and lymphatic systems). With the assistance of T helper cells, these B cells are activated to respond to infection. (D) Late infection: The late adaptive response involves lymphocyte clonal expansion, differentiation into plasma cells and activated T cells (including cytotoxic and helper T cells), as well as the production of antibodies by plasma cells. (E) Recovery: If the frog is cleared of infection (perhaps by topical antifungals or heat), the skin might be expected to return to normal, however, a cohort of selected memory lymphocytes should remain. (F) Re-exposure: If the frog is then later re-exposed to Bd, the memory lymphocytes (produced during the previous clonal expansion) are then activated and induced to replicate and differentiate, leading to a much more rapid and effective adaptive immune response on re-exposure. This is the concept of immunization (vaccination).

transfer of bacterial species to some amphibian hosts, thereby inhibiting Bd infection, at least temporarily (130, 131). Despite these promising results, outcomes in other species are mixed. For example, inoculation of Panamanian Golden frogs (Atelopus zeteki; extinct in the wild) with probiotic bacteria was not associated with improved survival (132, 133).

There is some evidence for population-level correlation between the presence of (and proportion of individuals harboring) anti-Bd bacterial species and population declines (134, 135), although not all studies corroborate these findings (136). The presence of diverse bacterial communities in addition to anti-Bd bacterial species on amphibians, and synergy (or antagonism) between bacteria and AMPs, will affect the degree of Bd growth inhibition as demonstrated in vitro (137–139). Other field studies have shown that skin microbial communities differ consistently between Bd-susceptible and resistant/tolerant species, as well as between sites with differing infection histories (121, 140). It is unclear from these studies whether identified bacterial communities promote Bd tolerance or resistance, or are the result of it (121). Further applied work should target probiotic candidate selection and trial bioaugmentation approaches (32, 141, 142).

### INNATE IMMUNE DEFENSES

If the invading Bd zoospores are not contained with constitutive host defenses, then they encyst upon the keratinized skin surface and attach via adhesive molecules (11, 77). Germination tubes are then sent through one or more cell layers (143, 144), injecting the contents of the zoospore cyst into the cytoplasm of deeper cells of the host epidermis, including the stratum spinosum and stratum granulosum [**Figure 1B**, (3)]. The intracellular location and process of injecting zoospore contents into deeper epidermal cells may permit Bd evasion of host immune surveillance, as has been described with other fungal pathogens (98, 145–147).

### Pattern Recognition Receptors Detect Common Fungal Structures and Initiate Inflammation, Although There Is Currently Limited Data to Assess Their Role in Chytridiomycosis

In the absence of targeted immune evasion, an invading microbe should prompt host recognition. Mechanisms responsible for pathogen recognition induce innate then adaptive immune responses in the host via antigens either secreted, expressed on the pathogen cell surface, or processed after phagocytosis. These antigens often contain widely recognized structural moieties known as pathogen-associated molecular patterns (PAMPs) that are common among different groups of microorganisms. These PAMPs bind to host germline-encoded pattern recognition receptors (PRRs) expressed within or on cells of the innate immune system (macrophages, dendritic Langerhans cells) and nonprofessional immune cells (keratinocytes, fibroblasts) [reviewed in (35)]. In mammals, four classes of PRRs include: (1) Toll-like receptors (TLRs) and (2) C-type lectin receptors (CLRs) within cell membranes, (3) NOD-like receptors (NLRs), and (4) Retinoic acid-inducible gene (RIG)-I-like receptors (RLRs) within the cytoplasm of host cells (148). Binding of PAMPs by PRRs leads to an innate amplifying inflammatory cascade that varies depending on the initial signaling pathways involved. Binding of PRRs may induce signaling pathways and the release of cytokines [pathways such as nuclear factor kappa-light-chain-enhancer of activated B cells [NFκB], and mitogen-activated protein [MAP] kinase]. Binding may also stimulate phagocytosis and destruction of extracellular microorganisms, or cell-mediated cytotoxicity and apoptosis of infected host cells [reviewed in (35)]. Patin et al. (149) presents an extensive review of existing knowledge about fungal recognition in mammals via PRRs, and lists common fungal ligands including β-glucans, zymosan, mannose, phospholipomannan, unmethylated DNA, and chitin.

Little is known about PRRs and their interactions with fungal pathogens in amphibians, although genes homologous to mammalian PRRs are present in model amphibian genomes (e.g., Xenopus spp.). The expression pattern and inducibility of TLRs has been studied to some extent in Xenopus spp. and Rana catesbeiana (51, 150). To date there is little evidence for early upregulation of PRR-coding genes in frogs infected with any pathogens (including Bd), although this may be the result of a lack of early immune cell infiltration rather than a lack of constitutive recognition by nonprofessional immune cells per se. However, studies with detailed kinetic analysis appropriate for examining early innate immune responses in amphibians are currently lacking. For example, Grogan et al. (18) did not find any evidence of differentially expressed Dectin genes (key members of the CLR family for detecting common fungal pathogens) at any point during infection, compared with uninfected control frogs. However, several other putative PRRs (TLRs and NLRs) and their downstream signaling pathways were found upregulated in the skin of frogs during late stage infections (18, 98, 151–153). Representatives of these pathways included mannose-associated genes, fc receptor 5 genes, NFκB subunit genes, caspase 6, 7, and 10 analog genes, and MyD88 pathway genes (18). These results from across the several species studied suggest that putative PRR-encoding genes are present and are likely activated. Although, it is possible that the pathogen actively interferes with these pathways, such as the inhibition of NFκB detected by Rosenblum et al. (120). We currently lack sufficient evidence to determine the efficacy of the early innate immune response to Bd. The detected late gene activation of these downstream signaling pathways may alternatively be associated with cellular stress and trauma or secondary bacterial infections (151), producing damage associated molecular patterns (DAMPs) that are similarly recognized by PRRs and protease-activated receptors (PARs) (154). These findings are consistent with the results of Brem and Parris (155) who showed that toads were less likely to be become infected if the epidermis was scraped (causing erosions) prior to Bd exposure. Priming of the innate immune system in response to trauma (Bd-associated epidermal erosions) may contribute to amplification of the innate response, leading to an exacerbated late-stage response. Similar immune priming may have applications for infection mitigation strategies. The precise characteristics of pathogen recognition and signaling may be further elucidated by detailed kinetic studies.

### The Alternative Complement Cascade May be Important for Early Defense in Resistant Individuals

The complement system constitutes a set of receptors and soluble plasma proteins and enzymes with an important role in defense against pathogens. Complement activation stimulates a rapid cascade of molecular interactions triggered by bound antibodies and PAMPs that results in formation of the membrane attack complex (MAC) and plays a central role in innate defense against fungal infections [reviewed in (156)]. Bound complement components function to agglutinate extracellular pathogens and lyse their cell membranes, as well as attract phagocytes to the locality and enhance their phagocytosis of pathogens via opsonization. Although considered most capable at neutralizing extracellular pathogens, complement C3 binding prior to pathogen intracellularization can activate autonomous immunity within infected cells (via NFκB signaling), ultimately leading to cell destruction via apoptosis (157). The complement cascade may be activated via three mechanisms. The first mechanism is the classical pathway (triggered by antigenantibody complexes, bacterial lipopolysaccharide, pentraxins such as C-reactive protein [CRP] and serum amyloid, etc.). A second mechanism is the alternative pathway (recognizes pathogen associated patterns or PAMPs). The third mechanism of activation is via the lectin pathway (recognizes carbohydrate structures via mannan-binding lectin [MBL], MBL-associated serine proteases [MASPs], and ficolins) [reviewed in (156)]. It is noteworthy that all elements of the complement system are well conserved across jawed vertebrates (158), such that with Xenopus spp. antiserum, it is possible to use purified mammalian complement to lyse amphibian red blood cells (159).

As an early and rapid defense response, examining the extent of complement activation may be critical for assessing the efficacy of the innate immune response. Several studies thus far have indicated early downregulation of complement pathway genes (98, 119, 120) in the infected skin of several susceptible species. In contrast, Grogan et al. (18) found that gene analogs associated with the alternative complement pathway (venom factor 1 and complement factor B) were upregulated from the early infection stage in a phenotypically more resistant population of Litoria verreauxii alpina frogs, but only upregulated at later infection stages in more susceptible populations (18). This is consistent with studies in a variety of species where complement pathway genes were predominantly upregulated in late-stage infection (151–153). These findings may indicate that the alternative complement pathway plays an important role in defense against Bd in more resistant individuals. However, Bd may downregulate or fail to activate the complement cascade in susceptible individuals, at least until the late infection stage (98). Further research on activation of the alternative complement pathway may provide genetic markers for resistance and opportunities for selective breeding or genetic engineering.

### There Is Limited Data to Assess Cytokine Upregulation in the Crucial Early-Stage Infection Period

Cytokines are endogenous inflammatory mediators and include lymphokines (such as macrophage activating factor [MAF]), interleukins (ILs), tumor necrosis factors (TNFs), interferons (IFNs), transforming growth factors (TGFs), chemokines, colony stimulating factors (CSFs), polypeptide growth factors (GFs), and stress proteins [including heat shock proteins (HSPs)]. Proinflammatory cytokines can act on adjacent cells or distant cells via the systemic circulation to amplify the inflammatory cascade, attract leukocytes to the site of infection, activate pathways involved in blood coagulation, and promote tissue repair [reviewed in (35)].

Experimental studies performed on chytridiomycosis thus far have lacked sufficiently early time-points (i.e., 6–24 h postexposure) and infection-targeted tissue sampling strategies to evaluate expression of putative cytokines or their encoding genes. However, gene expression studies sampling tissues between 3 and 8 days post exposure have reported limited evidence for upregulation of putative cytokine-encoding genes. These included IL-17A/F-like gene, calcineurin IL-2 inducible gene, HSPs, TNF associated factor (TRAF) and guanylate binding protein IFN inducible gene in spleens of X. tropicalis (98, 119), IFN and IL-associated genes in Rana spp. (120), and IL-1B, IL-17C, and IL-17E, TNFα, IFN and IFN-induced genes, granulocyte colony-stimulating factor, and several chemokineassociated genes in Litoria verreauxii alpina (18). In contrast, studies of late-stage infections demonstrated changes across the spectrum of cytokines (numerous IFNs, ILs, TNFs, and chemokines), with the most dramatic responses observed in skin tissues from the most susceptible individuals (18, 151, 152). Of particular interest, several gene expression studies on multiple species identified significant upregulation of numerous IFNinduced very large GTPase gene analogs throughout infection (18, 151, 160). Interferon-induced GTPase signaling is important for eliminating intracellular pathogens in epithelial cells via their sequestration and destruction within inflammasomes [thought to be especially important for defense against skin fungal pathogens (161)], and thus could be a key mechanism of cellautonomous immunity in chytridiomycosis (162). From these results, it appears that putative cytokine pathways are active in host response to Bd, although we currently have insufficient data to evaluate their relevance in the immediate post-exposure period. The reported upregulated late-stage cytokine responses may instead represent immunopathology from a dysregulated and non-protective immune response (18, 151).

### There Is Limited Evidence for Innate Leukocyte Recruitment and Infiltration Throughout Infection

Recruitment of leukocytes (immune effector cells) to the site of infection is a central component of the host immune response. Leukocytes of the innate immune system include circulating monocytes that differentiate into macrophages at the site of infection, polymorphonuclear phagocytes including neutrophils, eosinophils, and basophils, as well as natural killer cells and mast cells. These leukocytes contribute to recruit lymphocytes at the site of infection, amplify the inflammatory cascade, destroy extracellular pathogens via phagocytosis, and trigger apoptosis of damaged or infected host cells [reviewed in (35)].

The cellular immune response in chytridiomycosis appears inconsistent and is generally mild or decreased across species (48, 163). These studies largely compared skin and blood of Bd-infected and uninfected control frogs. For example, Woodhams et al. (111) found decreased circulating neutrophils and eosinophils, and increased numbers of basophils in infected adult Litoria chloris frogs. Davis et al. (164) and Peterson et al. (165) found increased circulating neutrophils and fewer eosinophils in infected Rana catesbeiana tadpoles and Litoria caerulea adults respectively [Peterson et al. (165) also found low circulating lymphocyte numbers [lymphopaenia]]. These results are consistent with a classic mammalian acute stress response [with the exception of the absence of lymphopaenia in the former study, (166)]. Young et al. (48) found lower circulating total white blood cell numbers in chronically infected adult L. caerulea, with overall impairment of responses on immune stimulation, and relatively higher neutrophil to lymphocyte ratios in infected frogs. These variable results may indicate that other unaccounted factors are playing a role (such as corticosteroid levels), or that species' responses differ.

The cellular immune response within skin tissue appears inconsistent from studies thus far performed. For example, histopathology has revealed a variable mild inflammatory response in 10–40% of skin sites, involving foci with macrophages and few neutrophils. This mild response is also often present near areas of ulceration suggesting a possible association with secondary bacterial infections (77, 151, 167, 168). No evidence of specific leukocyte-associated gene upregulation has been reported during early infections in gene expression studies performed thus far. However, during late-stage infection, Rosenblum et al. (98) found a mild increase in neutrophil-associated genes in the skin and liver of infected Xenopus tropicalis frogs, while Ellison et al. (151) and Grogan et al. (18) found predominantly increased expression of several macrophage and neutrophil associated genes. Taken together, these results indicate that an early leukocyte response is weak or lacking with Bd infection, and furthermore, that the late-stage response is inconsistent and likely insufficient to limit Bd infection (and may alternatively be associated with epidermal damage or secondary bacterial infection). This overall observed poor inflammatory response with Bd infection could be the result of inadequacy of innate immune activation with minimal cytokine-mediated leukocyte recruitment toward Bd. This could possibly be associated with immunoevasion and/or active suppression of immune responses by Bd. However, a limited innate immune response, particularly in late stage infections, may also be symptomatic of an inadequate or impaired adaptive immune response to Bd (discussed below).

### ADAPTIVE IMMUNE RESPONSE

The adaptive immune system provides a more specific defense against invading pathogens compared with the innate immune system, although it is slower to manifest initially. Amphibian pathogen-specific antibodies (IgY) are undetectable after initial ranaviral infection as is the case in mammalian response to primary infection with large DNA viruses (169). However, in mammals, antibody responses typically improve in efficacy upon subsequent exposures to the same pathogen [**Figures 1C–F**; reviewed in (35)], and studies performed in amphibians with Frog Virus 3 and Bd support this finding. Between two and three exposures to a pathogen over 4–6 weeks resulted in a detectable pathogen-specific IgY antibody response (89, 169, 170). These antibodies were detectable at 1 week after the last exposure, and peaked between 2 and 3 weeks. This means that the adaptive immune response may be most critical for infections that fail elimination by non-specific mechanisms of the innate immune response. In comparison with mammalian immune responses, the amphibian adaptive immune response is typically slower to manifest, and of lesser magnitude and efficacy. The adaptive immune system is also dependent on initial activation and costimulation by receptors and mediators of the innate immune system [reviewed in (35)]. However, as we have detailed in the previous sections, the innate immune response to Bd-infection appears somewhat inadequate, at least in the critical early stages of infection (first few days post exposure), and this may reduce the efficacy of the adaptive immune response.

### Key Components of the Adaptive Immune Response, Pathogen Specificity, and Immunological Memory

The primary components of the adaptive immune system include lymphocytes (T and B cells) and their respective mature effector forms responsible for enacting pathogen-specific cell-mediated immunity (cytotoxic or helper T cells) and humoral immunity (antibody-secreting plasma cells) together with secreted or membrane-bound antibodies (immunoglobulin). Unlike the germline-encoded components of the innate immune system, the adaptive immune system is characterized by unique antigen receptors. T and B cell receptors are generated when segments of immunoglobulin genes are rearranged by a unique process called somatic recombination. This can occur via V, D, and J (variable, diversity, joining) mechanisms, which require products from recombination-activating genes (RAGs). This gives rise to millions of naïve T and B cells during the development of an individual, with numerous distinct cell clones bearing unique surface receptors that together constitute the unique lymphocyte receptor repertoire of the host. Lymphocytes activated by antigen binding in combination with co-stimulatory molecules (**Figure 1C**) undergo clonal expansion and differentiation into their effector forms (**Figure 1D**). B cell receptors additionally undergo further changes after activation. This involves somatic hypermutation mediated by activationinduced cytidine deaminase (AICDA), which is accompanied by class switching during affinity maturation. During this clonal expansion process, large numbers of antigen-specific long-lived memory lymphocytes are produced (**Figure 1E**), and upon reexposure to the pathogen or antigen, these memory lymphocytes respond more rapidly and effectively than the initial response (**Figure 1F**). Thus, not only is the adaptive immune system able to respond in a specific way to novel pathogens, but it adapts to those pathogens during the course of an infection. Assuming that the individual survives initial infection, a functioning adaptive immune response should increase in efficacy with subsequent exposures to a pathogen (or antigen) during an individual's lifetime (unlike the innate immune response), leading to the concept of immunization [or vaccination; reviewed in (35)].

### The Role of PRRs, MHC and Dendritic Cells for Activation of the Adaptive Immune Response

Initial activation of the adaptive immune system involves the detection of pathogen-derived antigens by binding to PRRs expressed by host cells. Antigen binding stimulates endocytosis, degradation of the pathogen, and subsequent presentation of the antigen peptide on the cell surface via major histocompatibility complex (MHC) proteins. The two classes of MHC molecules, classes I and II, are expressed differently by cells of the body, and interact with different subsets of lymphocytes. MHC class I molecules are expressed by most nucleated somatic cells (for example, non-immune epithelial cells) and interact with CD8 cytotoxic T cells, leading to cytotoxicity and death of the host cell expressing antigen with the MHC class I molecule. MHC class I molecules are particularly important for recognition and elimination of intracellular pathogens via cell killing. MHC class II molecules are mainly expressed by professional immune cells such as dendritic cells and macrophages. These cells also recognize pathogens, but can present the antigen via the MHC class II molecules at the cell surface [reviewed in (35)]. Antigen binding then promotes differentiation of dendritic cells into antigen presenting cells (APC), and TNF-α stimulates APC migration to the spleen via the lymphatic or circulatory system where they contact lymphocytes with a variety of antigen-specific receptors (**Figure 1C**). MHC class II molecules bound to antigen on the surface of APCs interact with CD4 T helper cells in the presence of other co-stimulatory molecules, and their main function is to activate other immune effector cells (such as B cells; **Figure 1C**). Co-stimulatory molecules are expressed on APCs in response to mediators of the innate immune system (such as TLRs and NFκB), and they are essential for the activation of the naïve CD4 T helper cells.

### MHC Genes Associate With Survival and Are Under Positive Selection, Supporting the Rapid Evolution of Resistance or Tolerance to Bd

The diversity of MHC proteins expressed by cells is generated by polygeny (the presence of multiple interacting genes), allele codominance, and gene polymorphism [reviewed in (35)]. The evolution of MHC genes has been widely demonstrated to occur under selection by infectious diseases (171). Their inter-generational heritability (unlike T and B cell receptors) makes them important bridging elements between the innate and adaptive immune systems, and potential markers for selection of either resistance or tolerance to infection (14, 34, 172). Genes encoding MHC classes I and II molecules have been found to be upregulated throughout Bd infection, particularly within skin tissues (18, 34, 120, 151, 153). A variety of studies have demonstrated associations between characteristics of MHC alleles (allelic diversity, degree of heterozygosity, presence of certain alleles, presence of certain protein conformational elements) and degree of Bd susceptibility as it differs between species, populations and individuals (17, 20, 21, 173). Furthermore, several studies have demonstrated signals of positive selection at certain MHC loci in populations persisting with enzootic Bd, when compared with background levels of neutral genetic change (17, 20, 21). These findings suggest that certain MHC genes and alleles may play an important role in determining degree of Bd susceptibility, and that these are under directional selection for resistance or tolerance to Bd. However, recent recognition of the expansion of MHC class I-like genes in Xenopus spp. and presumably other ectothermic vertebrates [see **Box 3**; 64, 65] may require a revisit of some of the reported studies, especially transcriptomics. Indeed, MHC class I-like genes encode molecules with typical MHC primary structures but are polygenic and not or minimally polymorphic (66).

### Both Cell-Mediated and Humoral Immunity Are Likely Important for Defense Against Bd

When lymphocytes are activated in the presence of peptidic antigen bound to MHC with appropriate co-stimulation, they proliferate by clonal expansion, differentiate into their effector type and migrate to the site of infection (**Figure 1D**). CD8 T lymphocytes stimulated by MHC class I-bound antigens differentiate into cytotoxic T cells that recognize and kill infected host cells. This form of cell-mediated immunity is likely to be especially important for intracellular pathogens such as Bd, as the most efficient means to eliminate the reproductive stage of the pathogen (zoosporangium) is to destroy infected host cells (31). However, B cells and their differentiated effector type (plasma cells) are likely to be similarly important for eliminating Bd. Unlike in mammals, amphibian B cells demonstrate phagocytic capabilities (58). B cells may also act as antigen-presenting cells for T helper cells, and their effector plasma cells produce antibodies (immunoglobulin, either membrane-bound or secreted) that may be capable of targeting and destroying extracellular pathogen stages such as zoospores and secreted toxins, as well as killing infected host cells. Antibodies target pathogens and kill infected host cells via a suite of mechanisms including (1) binding specifically with the epitope of the antigen and causing the antigens to agglutinate, inactivating them, (2) activating the classical complement cascade, leading to the membrane attack complex to lyse pathogens directly, and (3) antibody-dependent cellmediated cytotoxicity (ADCC) involving tagging antigens for destruction by natural killer cells or phagocytes [reviewed in (35)]. Thus, both cell-mediated adaptive immunity (T cell elimination of the intracellular reproductive stage of Bd and B cell phagocytosis) and antibody-dependent humoral immunity [destroying zoospores and secreted products) are likely to be important for controlling Bd burdens (31)].

### There Is Limited Evidence for an Effective Adaptive Immune Response to Bd Infection

Although it has been suggested that herd immunity may protect populations if > 80% of frogs are immune (or resistant, through the effects of symbiotic bacteria) to Bd (134, 135, 174), there is currently little evidence to suggest that herd immunity operates in wild amphibian populations. Indeed, the existence of a herd immunity threshold relies on infection transmission being density-dependent, rather than frequency-dependent, as is expected for amphibian breeding aggregations (175). Similarly, herd immunity thresholds are unlikely to occur where the force of infection is unaffected by the presence of resistant individuals, as is the case for indirectly transmitted pathogens and those with multiple host species (or life-stages) with differing tolerance and susceptibility to Bd infection. Instead, the temporal patterns of enzootic Bd infection appear regulated by season and temperature rather than adaptive immunity in field populations (22). The few laboratory studies performed to date support these findings, suggesting limited activation of a protective adaptive immune response to chytridiomycosis. Young et al. (48) reported a decrease in total and IgY serum antibody responses [via anti-sheep red blood cells [SRBC] haemagglutination assay] in Bd-infected L. caerulea compared with uninfected frogs, while circulating numbers of lymphocytes were greatly reduced in infected frogs (48, 165). Results from histopathology of the skin showed only a mild response with foci of lymphocytes associated with regions of ulceration, or no evidence of lymphocytes (77, 163, 167, 168). In terms of gene expression results, Rosenblum et al. (98) found no change in lymphocyte markers or MHC genes in X. tropicalis. Results were similar in their second study on Rana spp. (120) despite mild upregulation of MHC class II genes in the skin during late-stage infection. Ribas et al. (119) found that adaptive immune genes were generally downregulated in the spleen of X. tropicalis throughout infection. In contrast, other studies demonstrated upregulation of numerous adaptive immune genes associated with B and T lymphocytes, immunoglobulins and MHC genes, particularly in skin tissues at late stages of infection (18, 151, 160, 176). However, a countering signal of downregulated T cell associated genes was also detected in several studies (18, 151, 152). These conflicting results indicate a more complex set of interactions operating within the adaptive immune system, which may be associated with the different temperatures at which animals were exposed or housed as well as the timing of collection of samples. The latter finding of downregulated T cell responses is particularly important and will be discussed in detail in its own section below.

Immunization against Bd was suggested early on as a management strategy for chytridiomycosis (177) given the highly successful examples from humans and domestic animals (178). Studies reported by Rollins-Smith et al. (31) and Ramsey et al. (89) attempted to immunize Xenopus laevis frogs against chytridiomycosis via an intraperitoneal injection with heat-killed Bd. They found promising results with a high-titer pathogenspecific IgM and IgY serum antibody response in the immunized frogs at 14 days post final immunization. In another experiment, Bd-binding mucosal antibodies (IgM, IgY, and IgX) were demonstrated after repeated exposure to Bd (89). These results are supported by the finding of upregulated immunoglobulin genes in re-exposed Atelopus zeteki frogs, suggesting production of memory lymphocytes (151). Furthermore, X-irradiation of frogs to impair T cells increased Bd infection loads in X. laevis (89). In contrast, a repeat experiment with killed-Bd injections into the dorsal lymph sac (days 0 and 14) and peritoneum (day 28) of X. laevis followed by splenocyte culture with Bd showed generally weak lymphocyte proliferation in comparison with samples cultured with phytohaemagglutinin (PHA) alone (31). In another experiment, young boreal toads (Bufo boreas) were immunized following a similar protocol and then exposed to Bd, however there was no evidence for a difference in survival between the immunized and sham-injected exposed frogs, suggesting that the immunization had not been successful in stimulating protective adaptive immunity in the young toads (31). Stice and Briggs (179) immunized Rana muscosa with formalin-killed Bd in combination with adjuvants [saline, Freunds Complete [FCA], and Incomplete Adjuvant [FIA]] by injection into the dorsal lymph sac and found no differences in the proportion of frogs infected nor time to infection. A study by Cashins et al. (180) did not detect any evidence for a protective effect of prior infection on re-exposure in Litoria booroolongensis. However, a study in B. boreas by Murphy et al. (181) found that previously exposed frogs survived slightly longer if they had a dry habitat option upon re-exposure. A study by McMahon et al. (182) found that multiple prior exposures to Bd slowed the rate of progression of chytridiomycosis, although this finding may instead be associated with repeated innate immune priming through trauma (155). The variable results of these studies may be associated with differing routes of immunization or dose-rates of Bd exposure. Furthermore, these results suggest that although the adaptive immune system may be activated during Bd infections in some species, the capacity for a robust and protective adaptive response appears limited, which may be associated with Bd-induced suppression (discussed below).

### HOT TOPICS: STRESS, IMMUNOSUPPRESSION AND IMMUNOPATHOLOGY IN CHYTRIDIOMYCOSIS

### Limited Evidence That Stress Predisposes Hosts to Chytridiomycosis via Corticosterone Responses

There is no evidence to suggest that immunosuppression is necessary to predispose amphibians to chytridiomycosis epizootics, particularly with numerous observations of disease emergence in abundant species in undisturbed naïve localities (3, 163, 183). Furthermore, signs indicative of generalized immunosuppression, such as secondary bacterial infections, appear to be largely lacking (8, 48, 88). However, stress-induced immunosuppression may play a role particularly in the infection of more resistant individuals and species [reviewed in (30)]. The extent to which environmental stressors and corticosterone mediate chytridiomycosis and its effects on amphibians is currently unclear. Environmental stressors (poor nutritional status, high densities and exposure to predator cues) have been putatively linked with elevated corticosterone and reduced immune capacity in some tadpole studies (184, 185), although it is unknown whether corticosterone is a direct mediator of these effects. Elevations in corticosteroids have been demonstrated to have a range of detrimental effects on the immune system of frogs, including inhibiting the humoral response, and reducing both numbers and viability of circulating lymphocytes [reviewed in (30)]. Indeed, exogenous application of corticosterone was found to increase Bd infection abundance in adult amphibians (105, 186), but only had sublethal effects on tadpoles (187). Gabor et al. (188) inhibited corticosterone synthesis (using metyrapone) and found that this did not prevent Bd-associated reductions in mass, although it did increase Bd loads. They concluded that the adverse effects of Bd on growth were not mediated by corticosterone.

In the field, non-invasive measures of corticosterone in free-living populations of tadpoles revealed that corticosterone levels correlated both with Bd infection and altitude, and that infections with a more virulent strain of Bd (BdGPL) led to higher corticosterone release (189, 190). Measuring urinary corticosterone, Graham et al. (191) and Kindermann et al. (192) similarly found higher levels in infected frogs as well as frog populations at higher altitudes. Furthermore, Peterson et al. (165) measured plasma corticosterone and found that diseased frogs (showing clinical signs of chytridiomycosis) demonstrated higher corticosterone levels than subclinically infected frogs. Thus, from current evidence it appears that elevated corticosterone correlates with infection in situ, and both predisposes to chytridiomycosis, and is a result of infection. However, the link between putative environmental stressors and elevated corticosterone is less robust, and elevated corticosterone does not appear to mediate the sublethal effects of chytridiomycosis (growth and mass).

### Batrachochytrium dendrobatidis Suppresses Lymphocyte Responses in Susceptible Individuals

Throughout this review, we have synthesized the results of numerous studies and highlighted the lack of a generally robust and protective immune response to Bd infection. We considered the epidemiology, general degree of inflammation, as well as markers of the innate immune response during early infection stages, and the adaptive immune response during late infection stages. This observed apparent lack of immune response may be the result of either (1) the failure of the host to recognize Bd as a pathogen, through low inherent antigenicity (possibly due to intracellular localization), immunoevasion, or masking antigens, or (2) Bd-induced immunosuppression or downregulation of key immune responses necessary for a protective immune response (145). Both of these mechanisms may occur in parallel in chytridiomycosis. For example, PRRs are generally not upregulated in early infection, suggesting a possible lack of pathogen recognition, whereas T cell responses appear actively suppressed or inhibited, as are complement-associated pathways.

Current evidence supports a specific role for Bd-induced immunosuppression, detected first via skin histopathology (77) and general immune function measures (31, 48), and corroborated via gene expression data (18, 98, 119, 120, 151, 153). Further experimental work has characterized at least one mechanism by which this might occur, via soluble Bd-secreted factors. Fites et al. (193) demonstrated that soluble factors released by Bd zoosporangia inhibited proliferation and/or caused apoptosis of T cells in vitro. For this work, they used in vitro immune experiments involving the proliferation of splenic lymphocytes (from X. laevis and R. pipiens) in culture. Interestingly, they found that macrophage phagocytosis was not similarly affected. Another study investigated apoptosis (via TUNEL and caspase assays) and found that programmed cell death was positively associated with infection load and morbidity (194). They speculated that apoptosis may thus be a pathogen virulence mechanism. In vivo studies also revealed immune inhibition activity associated with Bd supernatants by measuring delayed-type-hypersensitivity responses (33). Rollins-Smith et al. (195) then went on to characterize two metabolites (methylthioadenosine and kynurenine) produced by Bd that are capable of inhibiting lymphocyte proliferation and survival in vitro.

### Late Stage Immunopathology Characterizes Infections in Susceptible Individuals

Despite relative Bd-associated immunosuppression, several gene expression studies on a variety of amphibian species have demonstrated that susceptible individuals express both greater number and variety of dysregulated immune genes during late stage infections than more resistant individuals (18, 151, 152). This negative correlation between extent of immune response and degree of phenotypic susceptibility suggests that susceptible individuals may mount massively dysregulated and non-protective immune responses. This immunopathology is likely associated with DAMPs induced late in infection as the pathogen damages skin cells in order to release subsequent generations of zoospores. Such a dysregulated response may rapidly disrupt cellular homeostatic mechanisms. Indeed, recent metabolomics findings support this hypothesis by demonstrating the significant depletion of the "immune nutrient factor," alphaketoglutarate and its associated metabolite glutamate in severely infected animals (196, 197). This metabolic dysregulation has carry-on effects on numerous other aspects of cell homeostasis, particularly cellular energy metabolism (alpha-ketoglutarate is a key intermediate of the Krebs cycle). Furthermore, in vitro and gene expression studies suggest massive disruption of homeostatic mechanisms involved in epithelial stability, water and ion transport and musculoskeletal functions in susceptible individuals (18, 41, 98, 120, 151–153, 176). Therefore, immunopathology within susceptible amphibian species may not only cause their immune responses to be ineffective at eliminating the pathogen, but it may contribute to host morbidity and mortality due to the extensive disruption of cellular homeostasis and consumption of energy resources.

### RECOMMENDATIONS FOR FUTURE WORK

Despite two decades of research on chytridiomycosis, we still have only a limited understanding of the amphibian immune response to chytridiomycosis, and there is much to be discovered that may assist with disease mitigation. While continued support for existing approaches is essential, improving our capacity for amphibian immunological research will require: (1) the selection of an appropriate Bd-susceptible model species that could be bred to a MHC defined inbred strain (traditional amphibian models, Xenopus spp., are not sufficiently susceptible), (2) the development of a suite of taxon-specific affinity reagents (such as antibodies) for detection and imaging of pathogen-associated or host immune molecules of interest, and (3) the isolation or transgenic development of cell lines (including immune cells and skin explants) for in vitro functional assays.

A suite of conventional and emerging immunological methods from the fields of human and comparative immunology may be adapted for further study of amphibian chytridiomycosis. These methods enable the detection, quantification, isolation, functional evaluation, examination of signaling pathways, and localization of specific molecules of interest from homogenates, subcellular compartments (via biochemical fractionation), cells (separated by fluorescent-activated cell sorting [FACS] and flow cytometry), blood or tissues. For example, bioassays such as enzyme-linked immunosorbent assay (ELISA), measure the presence of various molecules (such as antibodies or antigens) via enzyme or ligand binding. Other bioassays may detect the presence and quantity of specific DNA (southern blot or qPCR), RNA (northern blot or RT-PCR), proteins (western blot) or their post-translational modifications (eastern blot). Flow cytometry enables the analysis of immune cells and their products, and cell sorting for proliferation and viability studies. Modern high-resolution imaging technologies include light microscopy combined with flow cytometry or standard labeling techniques with antibodies (immunohistochemistry and immunocytochemistry) or other stains, as well as electronmicroscopy.

Several preliminary exploratory systems biology studies have been reported in this review, for example, employing transcriptomics for gene expression and metabolomics for metabolite accumulation (152, 196). However, there are emerging approaches using high-throughput technologies such as next generation sequencing and mass spectrometry that still have unrealized potential for the study of amphibian chytridiomycosis (such as whole exome sequencing, proteomics, secretomics, and fluxomics). Importantly, further research using these emerging technologies would benefit from considering a broader temporal range in samples from experimental animals. In particular, experiments that compare the immune response in the very early infection period immediately post-exposure with the response later during infection would shed important light on initial susceptibility and within-host pathogen recognition and signaling dynamics. There is also potential for the use of microfluidics for single-cell-targeted approaches. Mass spectrometry and associated technologies [high performance liquid chromatography (HPLC) and nuclear magnetic resonance (NMR)] permit the high-throughput separation, identification and quantification of molecules of interest in a mixture. Combination techniques may permit high-dimensional data from these high-throughput technologies, such as the singlecell resolution of numerous cellular parameters over millions of cells via mass cytometry. Chromosome conformation capture may permit the identification of regulatory elements for immune genes of interest, and transgenic technologies may enable improved functional validation of the role of such genes and their translated protein products. These approaches include gene knock-in and knock-out on the pathogen and other organism cell lines (such as host immune cells), and include gene silencing (RNA interference), lentivectors, transposons and CRISPR/Cas9 genome editing. Indeed, some of these approaches may be used to advance therapeutic outcomes also. For example, recombinant Bd proteins or Bd genes introduced via vector may improve results in immunization trials compared with techniques already tried.

### CONCLUSIONS

In summary, we have provided an overview of the major aspects of the amphibian host immune response to chytridiomycosis, and how they differ from an expected efficacious immune response. Importantly, we highlighted an observed discord between the extent and efficacy of the response to chytridiomycosis comparing resistant and susceptible individuals. These findings suggest that resistant individuals likely possess more effective constitutive defenses (such as AMPs and symbiotic bacteria), and/or may mount a more effective innate immune response early in infection, combined with avoiding Bd-induced immunosuppression of their adaptive responses. Conversely, constitutive and innate defenses of individuals that succumb to chytridiomycosis are likely limited in their overall efficacy. Although their late-stage immune response may be characterized by exacerbated immune gene transcription, these responses likely constitute immunopathology, and may be ineffective due to pathogen-suppression of lymphocyte pathways. Indeed, severe immune dysregulation may contribute to a mortality outcome. Hence a combination of factors likely contributes to amphibian susceptibility to chytridiomycosis, rather than the presence or absence of any one immune mechanism or gene. This is particularly important when comparing potential factors conferring resistance or tolerance between distantly related amphibian taxa.

Our review has highlighted numerous gaps in current knowledge, particularly concerning: (1) mechanisms of initial pathogen detection and possible immunoevasion by Bd, (2) degree of activation and efficacy of the innate immune response, (3) the unexpected absence of innate leukocyte infiltration, (4) the relative importance of B and T cell responses for pathogen clearance, (5) the capacity and extent of immunological memory, (6) specific mechanisms of pathogen-induced immunosuppression, and (7) the role of immunopathology in pathogenesis. These aspects would benefit from further empirical study using the techniques we have discussed above. This also leaves us with an unanswered question for amphibian conservation management: can we manipulate the immune machinery of the host to improve resistance or tolerance both within individuals (immunization), and across populations through generations (evolution or assisted selection)? It is important to recognize that management approaches should be considered on two time-scales; (1) securing species in the short-term, and (2) developing long-term sustainable solutions (12). As we have reported earlier in this review, evidence is emerging that evolution of resistance and tolerance may be leading to recovery of some affected frog populations and communities (15). There is, as yet, limited proof of concept for strategies that might accelerate these evolutionary processes. However, immunological research remains a promising avenue

### REFERENCES


for amphibian conservation management, in light of the dramatic advances achieved in the human medical field in recent years.

### AUTHOR CONTRIBUTIONS

LG reviewed the literature and wrote the first draft. All authors contributed to revisions of the manuscript drafts.

### FUNDING

LG was supported by Australian Research Council (ARC) grants DP180101415 and LP110200240. JR was supported by grants from the National Institute of Allergy and Infectious Diseases (NIH/NIAID, R24-AI-059830) and from the National Science Foundation (IOS-1456213). LB was supported by ARC grants FT100100375, LP110200240, and DP120100811. LS was supported by the ARC grants LP110200240 and DP120100811. HM was supported by ARC grant DP180101415. The funders had no role in study design, data collection and analysis, decision to publish or preparation of the manuscript.


15 years after introduction of Batrachochytrium dendrobatidis. Conserv Biol. (2013) 27:1058–68. doi: 10.1111/cobi.12073


fungus Batrachochytrium dendrobatidis. Conserv Biol. (2005) 19:1460–8. doi: 10.1111/j.1523-1739.2005.00195.x


cell transfer in xenopus. Transplantation (2004) 78:1415–21. doi: 10.1097/01.TP.0000140846.73210.91


In: Bruce RC, Jaeger RG, Houck LD, editors. The Biology of Plethodontid Salamanders. New York, NY: Kluwer Academic/Plenum Publishers (2000). pp. 451–62. doi: 10.1007/978-1-4615-4255-1\_25


dendrobatidis: the combined effect of symbiotic bacterial metabolites and antimicrobial peptides of the frog rana muscosa. J Chem Ecol. (2012) 38:958– 65. doi: 10.1007/s10886-012-0170-2


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

Copyright © 2018 Grogan, Robert, Berger, Skerratt, Scheele, Castley, Newell and McCallum. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Innate Immunity Provides Biomarkers of Health for Teleosts Exposed to Nanoparticles

#### Débora Torrealba1,2, Juan A. More-Bayona<sup>1</sup> , Jeremy Wakaruk <sup>2</sup> and Daniel R. Barreda1,2 \*

*1 Immunology and Animal Health Laboratory, Department of Biological Sciences, University of Alberta, Edmonton, AB, Canada, <sup>2</sup> Department of Agricultural, Food and Nutritional Science, University of Alberta, Edmonton, AB, Canada*

In recent years, the unique properties of nanoparticles have fostered novel applications in various fields such as biology, pharmaceuticals, agriculture, and others. Unfortunately, their rapid integration into daily life has also led to environmental concerns due to uncontrolled release of nanoparticles into the aquatic environment. Despite increasing awareness of nanoparticle bioaccumulation in the aquatic environment, much remains to be learned about their impact on aquatic organisms and how to best monitor these effects. Herein, we provide the first review of innate immunity as an emerging tool to assess the health of fish following nanoparticle exposure. Fish are widely used as sentinels for aquatic ecosystem pollution and innate immune parameters offer sensitive and reliable tools that can be harnessed for evaluation of contamination events. The most frequent biomarkers highlighted in literature to date include, but are not limited to, parameters associated with leukocyte dynamics, oxidative stress, and cytokine production. Taken together, innate immunity offers finite and sensitive biomarkers for assessment of the impact of nanoparticles on fish health.

### Edited by:

*Eva-Stina Isabella Edholm, UiT The Arctic University of Norway, Norway*

### Reviewed by:

*Annalisa Pinsino, Istituto di Biomedicina e di Immunologia Molecolare Alberto Monroy (IBIM), Italy Sylvia Brugman, Wageningen University & Research, Netherlands*

#### \*Correspondence:

*Daniel R. Barreda d.barreda@ualberta.ca*

#### Specialty section:

*This article was submitted to Comparative Immunology, a section of the journal Frontiers in Immunology*

Received: *14 September 2018* Accepted: *12 December 2018* Published: *09 January 2019*

#### Citation:

*Torrealba D, More-Bayona JA, Wakaruk J and Barreda DR (2019) Innate Immunity Provides Biomarkers of Health for Teleosts Exposed to Nanoparticles. Front. Immunol. 9:3074. doi: 10.3389/fimmu.2018.03074* Keywords: innate immunity system, nanoparticles, teleosts, oxidative stress, leukocytes

### INTRODUCTION

Nanoparticles (NPs) are an emerging technology, currently being applied in various fields including medicine, cosmetics, electronics, space science, chemical manufacturing, cellular and molecular biology, agricultural and animal science (1–4). Nanomaterials are structures of 1–1,000 nm in size; however, stricter definitions restrict them to those in the 1–100 nm range (National Nanotechnology Initiative). NPs have been developed in many different inorganic and organic forms but most current NPs are classified into four material-based categories: (i) carbon based NPs such as fullerenes and carbon nanotubes; (ii) inorganic-based NPs including metal (e.g., gold NPs, silver NPs) and metal oxide NPs (e.g., titanium dioxide NPs, zinc oxide NPs); (iii) organicbased NPs such as dendrimers, liposomes, polymers; and (iv) composite-based NPs: combinations of different material-based NPs (1, 2).

The rising use of NPs in modern technologies has led to unregulated release and accumulation into the aquatic environment, contributing to further pollution (5, 6). For example, aquatic release of titanium dioxide NP (TiO2) has been estimated at 17 % of their total production per year (7). Based on a production scale of 88 kt/year, this translates into a significant aquatic release of 15.6 kt/year. Releases for other nanomaterials have been recently summarized (7). NPs enter the aquatic environment mainly through wastewater and effluents from industrial sources, as well as via atmospheric deposition, leaching from soil, accidental spillages, and agricultural drainage water (1, 5). Naturally-occurring NPs also enter aquatic

environments through waterways across several landscapes (8). Despite our growing knowledge about NP bioaccumulation within aquatic environments, little is known about the detrimental effects of nanoparticles on animal and human health. Various features such as size, chemical makeup, biodegradability, and the physicochemical environment all affect the extent to which NPs are contributors of biotoxicity (5, 9).

Fish are widely used as sentinels for aquatic ecosystem pollution stemming from chemical exposure and are the preferred model for development of chemical testing guidelines (10, 11). Among others, their practical relevance are based on: (i) wide distribution in aquatic environments; (ii) high ecological relevance due to position within food web structure, nutrient cycling and energy transfer; and (iii) expanding models and tools that researchers can use (12). In addition, testing on native species under natural housing conditions (e.g., Oncorhynchus mykiss to test freshwater pollution and aquatic toxicity in Canadian cool waters) adds relevance to these datasets (13). Consequently, fish species have now become a preferred model to study NP toxicity (14–16). A summary for recent applications of innate immune parameters as biomarkers of impact on animal health is provided in **Tables 1**, **2**.

A biomarker can be any measurable biological response that reproducibly changes upon exposure to environmental pollutants (108). Recently, innate immunity has yielded a wide set of biomarkers for immunotoxicity against multiple xenobiotics including metal ions, pesticides, oil products, and chlorinated hydrocarbons (4, 9, 109–111). Among the most commonly used we find: leukocytes dynamics, phagocytic activity, lysozyme production, production of antimicrobial peptides, cytokines expression, and reactive oxygen species (ROS) production (110). Multiple studies have shown that innate immune parameters display high sensitivity to NPs. The conservation of innate parameters across animal models further makes them amenable for complementary studies, for instance, when toxicity to aquatic, and terrestrial ecosystems is of interest. Finally, early kinetics of innate parameters induction coupled to their sensitivity to capture additive or synergistic effects from environmental contaminants makes them a powerful alternative to evaluate ecotoxicity. In this review, we have drawn on recent literature to highlight those innate immunity biomarkers most often used in the assessment of teleost fish health exposed to NPs.

### EXTERNAL BARRIERS

External barriers to microbes infecting fish encompass mucous secretion produced in multiple tissues such as gills, skin, and intestine. These surfaces constitute physical barriers supported by immunologically-active agents such as lysozyme, complement, lectins, proteolytic enzymes, antimicrobial peptides, and reactive chemical species (112–116). While some of these defense molecules have been used as biomarkers in fish ecotoxicology (117), just a small proportion of these have been applied to the assessment of fish health after NPs exposure. Here, we discuss their use to date and how researchers can take advantage of them to assess NP toxicity.

TABLE 1 | Innate immune defenses used as biomarkers to assess fish health post exposure to nanoparticles.


*(Continued)*

### TABLE 1 | Continued


#### TABLE 1 | Continued


*(Continued)*

*(Continued)*

TABLE 1 | Continued


### Mucus

The amount and biochemical content of mucus have been used as biomarkers for the evaluation of teleost fish health exposed to NPs. Mucus on the fish skin is a first line of defense, forming a barrier protecting tissues from the surrounding environment. This barrier is secreted by goblet cells and composed mainly of water, mucopolysaccharides, mucoproteins, and other soluble materials (114). Mucus, as part of the innate immune system, plays an essential role in protecting fish from xenobiotic exposure (114, 118). When fish health is compromised, the mucosal matrix may also be affected. Therefore, it can be a useful tool to assess the effect of NPs in fish. Nevertheless, only two studies have evaluated skin mucus after exposure to NPs (**Table 1**). In the first study, Oliveira et al. presented a non-invasive method to assess the effects of gold NPs (Au) in Sparus aurata by analyzing skin mucus (17). Measurements of total antioxidant capacity and esterase activity showed the sensitivity of skin mucus exposed to NPs, even at low concentrations (17). The second study in Pimephales promelas showed an increase in mucus production between 4 and 24 h after exposure to silver nitrate NPs (AgNO3) (18). By day 3, fish had considerably reduced their ability to produce mucus (18). The lack of additional studies employing mucus might be related to the limited knowledge concerning the repertoire of immune factors present in skin mucus and their precise protective role to study fish health fitness (118). Additionally, there is an inherent complexity in obtaining enough amount of mucus to analyze. Other potential biomarkers in mucus skin could be the quantification of activity of well-characterized enzymes such as protease and lysozyme (119). These prospective biomarkers may contribute to assess fish responses against a pathogen after NPs exposure. In addition, other immune-related parameters such as mucosal IgT secretion levels and the capacity for development of mucosal memory will allow the evaluation of fish health following NP exposure. These non-invasive biomarkers can be meaningful to study the fish health fitness while avoiding fish slaughter, thus, reducing the number of individuals used in environment pollution monitoring. Nevertheless, more studies are needed to define whether these parameters of skin mucus maintain consistent results in multiple experimental conditions such as NP type, concentration and exposure time.

### Gills

In fish, gills are the main organs for gas exchange and have a relevant role in ionic and osmoregulatory function (120). Moreover, gills are considered to be the most sensitive organ to a majority of xenobiotics because, like the mucus on the skin, gills are in direct contact with the environment (121). However, that may not apply to NPs. In gills, evaluation of NPs effects have been performed mainly by histopathology (19– 23, 25, 28, 30–33, 39, 40), NP bioaccumulation (21, 23, 24, 27, 32, 34, 37, 39) and measurement of oxidative stress biomarkers (19, 22, 24, 25, 29, 30, 34, 35, 37, 39, 41) (refer to Oxidative stress section and **Table 1**). Regarding histopathology, some inaccurate histopathology-based results have been published due to the high degree of expertise necessary (122). For instance, controversial results have been published regarding the effects of silver NPs (Ag) in Danio rerio. Some studies report histopathological lesions TABLE 2 | Species used in the evaluation of nanoparticles' effect on fish health.




*(Continued)*

in gills such as hyperplasia and inflammation after exposure of Ag NPs for 4 and 21 days at different concentrations (1.5–15 µg/L) (31, 36, 123). Other studies showed lack of pathology in gills after 28 days at Ag NPs concentrations between 10 µg/L and 1 mg/L (21, 32). These gill lesions would affect oxygen intake, osmoregulation, acid-base balance, and excretion of nitrogenous waste that in turn would likely produce acute toxicity (117). As a first line of contact between the host and its environment, one would expect gills to offer an optimal site to assess NP bioaccumulation. However, different studies showed that gills are not the main target for NP accumulation. For example, Oreochromis niloticus exposed to iron oxide NPs (Fe2O3) for 60 days showed greatest bioaccumulation in the spleen whereas gills displayed much lower levels of bioaccumulation (spleen > intestine > kidney > liver > gills > brain > and muscle) (27). Another study also using O. niloticus but exposed to copper NPs (Cu) for 30 days yielded a similar conclusion (bioaccumulation levels: liver > kidney > gills > skin > and muscle) (23). Bruneau et al. also described a higher bioaccumulation in liver than in gills in O. mykiss after an exposure of Ag NPs for 4 days (34). Thus, although NPs can be absorbed by the gills, they are not the preferred uptake route in teleosts (124). In general, histopathology, and NP bioaccumulation in gills might not be the best biomarker of NPs toxicity. They must be complemented with others potential biomarkers for NPs. For example, the number of goblet cells or mucus production are relevant biomarkers in aquatic toxicology that can be tested to evaluate NPs effects on fish health (117).

### CELLULAR RESPONSE

### Leukocytes

Immune cells play a pivotal role in the clearance of pathogens or other foreign elements like NPs (125). Thus, assessment of leukocyte engagement may provide insights into NP toxicity. Indeed, kidney NP bioaccumulation is linked to a reduction in neutrophils function and a possible reduction in their ability to control bacterial infections (48). In this section, we review recent reports on the effects of NPs on the dynamics and functionality of fish leukocytes (also summarized in **Table 1**). Additional effects on leukocyte responses will be addressed in the following sections (refer to Internalization of nanoparticles, Oxidative stress, and Cytokines sections).

### Macrophages

Resident macrophages offer early detection of insults (e.g., pathogen infiltration, tissue damage, toxicant exposure) at various tissue sites (126, 127). Changes of macrophage viability and function have long been implicated with toxicity resulting from exogenous pollutants (e.g., metals, sewage, hormones disrupting compounds, pharmaceuticals chemicals) (12, 110, 128). Thus, macrophages and their proper function can be used as biomarkers for immunotoxicity (12). Most recently, this modulation in macrophage function has been used as biomarkers to evaluate NPs toxicity (**Table 1**). For instance, TiO<sup>2</sup> NPs induced upregulation of macrophage colony-stimulating factor 1 (MCSF-1) in multiple pooled tissues of P. promelas (anterior kidney, liver, spleen and gills) (48). Changes in MCSF-1 gene expression have been associated with impairment of macrophage function. This is consistent with its important role in viability, differentiation, mobilization, and activation of macrophages and their precursors (48). Using an in vitro approach, the effect of carbon nanotubes on kidney mix population of macrophages in O. mykiss revealed that SWCNT induces upregulation of pro-inflammatory cytokine expression such as IL-1β but not in IFNα expression, indicating a selective pathway (89). These results suggest a different effect in pro-inflammatory soluble mediators. Duan et al. evaluated the effect of sublethal increasing concentrations of silica NPs (Si) on macrophage function using zebrafish embryos (50). Results showed a downregulation of gene expression in macrophage inhibitory factor (MIF) and vascular endothelial growth factor receptor 2 (VEGFR2). The higher Si NP concentration, the stronger the down regulatory effect resulting in a decrease in macrophage activity (50). These results showed that even at sublethal doses of NPs, macrophage gene expression may provide useful read-outs for evaluation of fish immunity. Melanomacrophages are highly pigmented macrophage type that possess phagocytic function and play an important role in the immune response (129). Melanomacrophage centers (MMCs) have been used as biomarkers to assess fish health and aquatic environmental pollution (130–132). In this context, using the P. reticulata model, it has been shown that citrate-functionalized maghemite (γ-Fe2O3) induces significant changes in MMCs in liver. Acute (3–7 days) and chronic exposure (14–21 days) to γ-Fe2O<sup>3</sup> NPs increase number, cellular content and size in MMCs, suggesting that γ- Fe2O<sup>3</sup> might be involved in the regulation of innate immune responses (49). Altogether, these demonstrates that NPs are capable of interfering with multiple aspect of macrophage function (4). Due to the limited number of observations on the responses of MCSF-1, gene expression of pro-inflammatory cytokines, and immune-related genes, and MMCs, these parameters cannot yet be considered as valid biomarkers for NPs toxicity purposes. However, the key role of these parameters in macrophage function motivates continued research on its feasibility as a biomarker.

### Neutrophils

Neutrophils are polymorphonuclear leukocytes that are central to the induction and regulation of acute inflammation (133, 134). Neutrophils act as the predominant phagocytic cells for first-line defense to be recruited to an inflammatory site against diverse xenobiotics (135, 136). Thus, neutrophils play an important role in recognizing and eliminating foreign agents, including some NPs (137). In this regard, NPs have been shown to not only impair neutrophil functionality but also affect a diverse array of biochemical responses. These effect in neutrophils features have been used as biomarkers to evaluated NPs toxicity (**Table 1**). However, based on these studies, there is little clarity to date on the effects of NPs on neutrophil biology. For example, fullerenes have shown to inhibits some function such as ROS production and NETosis in mixed neutrophil populations from kidney of P. promelas (44). Interestingly, this effect did not induce significant changes in total neutrophil counts, suggesting that neutrophil antimicrobial functions are the primarily affected following fullerene internalization rather than neutrophil development per se (44). On the other hand, Jovanovic et al. revealed that TiO<sup>2</sup> NPs injections in D. rerio increased neutrophil migration, oxidative burst and phagocytic activity, associated to higher tissue damage (48, 95). In this context, increased neutrophil activation can be associated to phosphorylation of p38 mitogen-activated protein kinase (MAPK) and extracellular signal-regulated kinases. Conversely, TiO<sup>2</sup> NPs downregulated matrix metalloproteinase 9 (neutrophil elastase) expression, which further decreased IL-8 neutrophil mobilization from hematopoietic tissues (95). Małaczewska and Siwicki et al. found that even when Ag NPs did not induce an effect on peripheral neutrophils, spleen phagocytic activity was inhibited only at high concentrations of Ag NPs using the O. mykiss model (43). These results might be attributed to differences in concentration and time of exposure (43). Ortega et al. used PAA functionalized NPs [cerium oxide (CeO2), TiO2, Fe2O<sup>3</sup> and zinc oxide (ZnO)] to evaluate their effect on function of a mixed neutrophil population of kidney in Carassius auratus (42). Here, it was shown that NPs decreased neutrophil viability in a concentration and time-dependent manner. Additionally, lower concentration of NPs decreased neutrophil degranulation, and increased ROS production along with an increase of gene expression of pro- and anti-inflammatory cytokines in neutrophils such as IL-1β and IFNα. Hence, these sub-lethal doses of NPs might be linked to a higher susceptibility against pathogens and might thus impair fish health (42). Altogether, these studies reveal that several neutrophil functions can be modulated following NPs exposure. However, depending upon the type of NPs evaluated, different results can be observed. Based on the small number of studies to date, it is difficult to estimate the reproducibility of these findings and the reliability of each neutrophil function as biomarkers. Nevertheless, any disruption in neutrophil function will represent major implication of fish immune defenses. Therefore, more studies are needed to clarify if these neutrophil parameters could be used as biomarker to monitor NPs toxicity in fish.

### Lymphocytes

Lymphocytes offer a diverse array of effector and regulatory functions in multiple tissues. They are typically associated to long-term immune responses; however, a wide group of lymphocytes has relevance in innate defenses. For instance, phagocytic B lymphocytes participate in initial responses following pathogen invasion (138). Furthermore, innate lymphoid cells (ILCs) like cells, a recently described group in zebrafish (139), have a role in inflammation, tissue remodeling, and homeostasis regulation (140, 141). Thus, lymphocyte development, viability and proper function have a novel potential as biomarker for NP toxicity. For example, O. mykiss exposed to Ag NPs, Au NPs and Cu NPs revealed a differential effect on lymphocytes isolated from blood and spleen by density gradient separation. Lymphocyte viability was decreased only by Ag NPs following two-day incubation. However, no effect on lymphocyte proliferation was observed at the same concentration while enhanced proliferation rates were stimulated with lower concentration of Ag NPs. Conversely, Au NPs had the strongest effect on lymphocyte proliferation within the NPs tested (43). In addition to the effect of NPs on lymphocyte viability and proliferation, there are descriptions of interaction with other pollutants that potentiate their overall effect. For example, in vivo exposure of Dicentrarchus labrax to TiO<sup>2</sup> NPs, showed a cytotoxic effect when it interacted with dioxins in splenocytes, suggesting that the spleen might be a possible biomarker for NP toxicity due to NP bioaccumulation in this tissue (53, 91). These changes in lymphocyte development will further promote changes in the overall total lymphocyte counts. This was shown by Khabbazi et al. who found significant differences in lymphocytes counts in blood in O. mykiss exposed to copper oxide NPs (CuO) (52). On the other hand, P. promelas exposed to hydroxylated fullerenes or TiO<sup>2</sup> NPs did not produce differences in lymphocytes counts in blood samples (44, 48). Due to the limited number of studies and notable difference between results in viability and lymphocyte counts after an exposure with NPs, these parameters cannot yet be considered as biomarkers to assess NPs toxicity. Although most studies discussed here used morphological approaches for lymphocyte definition, further studies are needed for deeper and more accurate characterization of lymphocyte subsets in fish and their specific contribution in NP toxicity.

### Internalization of Nanoparticles

Internalization of invading pathogens or threats is the paramount to immune defense in several organisms. In the case of NP, their size, shape, surface chemistry, and mechanical properties influences the mechanism of cellular internalization (142, 143). For example, following entry, small NPs (<100 nm) are rapidly internalized by professional cells through endocytosis whereas large NPs (larger than 500 nm) are uptaken by micropinocytosis or phagocytosis (34, 144). Thus, despite their origin and internalization processes, NPs have been shown to modulate cellular responses in a variety of professional phagocytes such as neutrophils, monocyte, and macrophages (43). For instance, Malaczewska and Siwicki evaluated the effect of Ag NPs, Au NPs and Cu NPs on the phagocytic activity of spleen leukocytes from O. mykiss, showing an increase in phagocytic capacity in fish exposed to Ag NPs, but not observed in animals exposed to Au NPs nor Cu NPs (43). Through intraperitoneal injections, TiO<sup>2</sup> NPs were administrated to Trachinotus carolinus and cellular uptake was evaluated after 3 days. Results demonstrated NP uptake in the liver, kidney, lung, and spleen following a similar process that is also observed in mice and human cells (145, 146). However, the authors suggested that other internalization processes might occur in fish other than phagocytosis since they detected TiO<sup>2</sup> NPs reaching the nucleus compartment. Bruneau et al. have shown that cadmium tellurium quantum dots (Cd-TeQDs) induce an increase in phagocytosis rates associated to the size and form of those NPs in O. mykiss (51). In a different experiment, Greven et al. demonstrated that aggregation of polystyrene and polycarbonate nanoplastic particles promotes NPs internalization in P. promelas (45). As described, NP induced significant changes in the capacity of leukocytes to perform internalization of NPs. This suggests a regulatory effect, occurring in a time, and concentration-dependent manner, that is associated to the type of NP. Studies summarized here suggest that internalization of NPs by phagocytic cells, is a potential biomarker that can offer relevant information about the effect of the NPs at the intracellular level and should be complemented with other biomarkers to gain an integrative view of altered cellular function.

### Oxidative Stress

Once NPs have gained access to the organism, distributed to multiple tissues, and internalized into cells, they are capable to promote intracellular responses such as oxidative stress. Oxidative stress refers to the imbalance between the production of free radicals and the protective antioxidant defense system (109, 147). Oxidative stress has become a relevant biomarker for aquatic toxicology (148), because this phenomenon in fish can be triggered by many chemicals including NPs, metal ions, pesticides, oil products, and chlorinated hydrocarbons (9, 95, 109–111). These environmental pollutants can induce oxidative stress in fish via two ways: directly by affecting the animals or indirectly by modifying the environmental conditions (111). Oxidative stress can be detected and measured giving a quantitative indication of fish health status and it can be evaluated using different parameters such as free radicals, antioxidant defenses and biomarkers of oxidative stress (**Table 1**).

### Reactive Oxygen Species and Nitric Oxide

ROS are free radicals formed as a natural by-product of the normal metabolism of oxygen. ROS have important roles in cell signaling and homeostasis (149). However, ROS may significantly increase under stressful scenarios such as infection, inflammation, and exposure to environmental pollutants (111, 150). This increase may disrupt homeostasis, producing damage at the cellular level, and disease or death at the organism level (109, 148). ROS has been one of the most common parameters used for the evaluation of fish health after exposure to different types of NPs (**Table 1**). The preference of ROS as a biomarker is attributed to: (i) its high sensitivity to different NP exposure; (ii) fast-triggered response compared to other free radicals; (iii) remarkable consistency of results between studies; and (iv) increasing development of novel highly-sensitive reagents that allow the detection of slight changes of ROS production. NPs that have been shown to induce ROS production in fish include carbon-based, metal-based, plastic-based, and polymeric carrier NPs.

Levels of ROS have been successfully measured after exposure to NPs in different fish species such as D. rerio (25, 37, 46, 56, 60, 66, 68, 73, 74, 94), P. promelas (44, 45, 48), Oryzias latipes (70, 72), Poeciliopsis lucida (71, 106), O. niloticus (27, 40), C. auratus (42), O. mykiss (43), Oncorhynchus tshawytscha (67), Epinephelus coioides (65), Ictalurus punctatus (64), Anguilla anguilla (29), Apistogramma agassizii (22), Paracheirodon axelrodi (22), and Prochilodus lineatus (39). Interestingly, these studies showed a high versatility of ROS that can be measured in different stage of fish development, tissues, and cells. Most studies using embryos (46, 47, 56, 66, 73, 74) and larvae (70) observed an increase of ROS levels. However, an exception were those studies using Si NPs where no effect was observed in embryos (46, 47). These authors claimed that the lack of increase in ROS levels was related to the type of NP used (47). The use of embryo/larvae models to analyze ROS levels confer an integrative view of whole organism response. Even when some studies showed that no effect was observed in ROS levels, use of embryos/larvae confer a low cost and easy handling tool that still make them valuable as biomarkers. In the case of adult fish exposed to NPs, there are studies that assess ROS production in whole organs. ROS have been measured in different tissues such as gills (22, 25, 37, 39), liver (37), intestine (37), and brain (37), and fluids as serum (27, 40). Using gills, different studies showed a persistent ROS activation after NP exposure. For instance, A. agassizii exposed to Cu NPs showed greatest increase in ROS production at 3 days of exposure (22). Souza Khabbazi et al. observed a significant increase in ROS levels at day 1 of exposure with graphene oxide (GO) in gills of zebrafish (25). As first line of contact between the host and NPs, one would expect gills to offer a faster biomarker than internal organs. However, a study also using zebrafish but exposed to SWCNT showed a significant increase in ROS production in gills at 3 days of exposure comparatively slower than in liver and brain, where this increase was at 2 days (37). No modulatory effect was observed in the intestine (37). The available information reveals that measurement of ROS in gills is a valid biomarker to assess NPs toxicity, even though time of response can be slower than other organs. Furthermore, ROS production have been evaluated in vitro using both cell lines and primary cell cultures. These studies took advantage of a large variety of cell types such as: zebrafish liver (ZFL) (60, 68, 94); PLHC-1 derived from hepatocellular carcinoma in P. lucida (71, 106); and CHSE-214 derived from O. tshawytscha embryo (67); and primary cell cultures such as: neutrophils (42, 44, 45, 48); hepatocytes (64, 65, 69); splenocytes (43) and phagocytes (29). In cell lines, different NPs, concentrations, and exposure time, increased ROS levels. In contrast, not all NPs successfully modulate ROS production in primary cell cultures (42, 69). These results reveal that ROS production in cell lines, as per their homogeneous nature, is more consistent than in primary cell-based assays. This might be attributed to factors such as heterogeneity, cellular viability, and activation state of primary cells. Diverse studies have been done on a wide range of fish species and targets that can be used to test NPs effect on ROS production. These studies point out ROS as a reliable, sensitive, and valid biomarker of NPs toxicity in fish.

NO is the second free radical used as a biomarker to assess NPs toxic effects in fish described in literature. NO is produced by inducible nitric oxide synthase (iNOS). iNOS converts L-arginine and oxygen into L-citrulline and NO (151). NO may produce a wide range of physiological and pathophysiological effects (152). In contrast to ROS, there are only two studies which have evaluated NO expression in fish after exposure to NPs (**Table 1**). Based on these studies, it appears that NO may also serve as a useful biomarker. In the first study, zebrafish embryos exposed to CuO NPs showed a significant increase in NO production. Interestingly, the same increasing trends were observed for ROS production (66). In a second study, also using zebrafish embryos two NPs were analyzed, CuO NPs, and CeO<sup>2</sup> NPs. CuO NPs increased NO levels in contrast to CeO<sup>2</sup> NPs that decreased them (75). The authors explained that this difference was due to the high toxicity of CuO and the ability of Ce to scavenge NO (75). The low number of studies that focus on NO production may be due to the response rate of this free radical, where ROS levels have been shown to respond faster than NO. As mentioned above, NO levels need to be produced by iNOS expression and this specific pathway is slower than ROS production (153). As showed above, since only two studies analyze the effect on NO more research is required to determine its potential use as a biomarker in the context of NPs toxicity. Among other, evaluation of diverse NPs at different concentrations that can confirm that NO reproducibility of the results.

### Antioxidant Defenses

Free radical levels can be balanced by the antioxidant system which scavenges free radicals and delays or inhibits cellular damage (154, 155). There is a wide range of antioxidant defenses as diverse as free radicals themselves (156). Antioxidant defenses have been widely used as biomarkers of environmental toxicology (11, 109). Different antioxidants have been analyzed as tools to measure the effects of NPs on fish health (**Table 1**). The three main antioxidants and direct free radical scavengers include superoxide dismutase (SOD), catalase (CAT), and glutathione peroxidase (GPx). Other antioxidant enzymes that inactivate secondary metabolites have been measured to evaluate the effects of NPs on fish health such as glutathione sulfotransferase (GST), total glutathione (GSH), glutathione in its reduced state (GR), and its oxidized state (GSSG) (157, 158) (**Table 1**). Antioxidant levels have been measured mainly by biochemical assays but also by gene expression studies (65, 70, 79–81). As a homeostasis system, we would expect that antioxidants defenses levels increase after NPs exposure. However, some studies showed that antioxidants defenses do not always showed these patterns. For instance, Srikanth et al. used chinook salmon cells (CHSE-214) exposed to CuO NPs, they found a significant increase in ROS, SOD, CAT and GPx levels as expected (67). In contrast, Ganesan et al. observed a marked decrease in antioxidant defenses (SOD, CAT and GPx) in zebrafish embryos after exposure to CuO NPs, however, ROS levels showed significant increases (66). The authors claimed that the antioxidant system was overpowered by increased ROS production (66). Despite that multiple studies have used only antioxidant defenses to assess fish health following NPs exposure (30, 34, 41, 59, 76, 77, 79–83), we point out that antioxidant defenses should be measured as a complement of ROS and not by themselves.

### Biomarkers of Oxidative Stress

NPs are potent triggers of oxidative stress in fish and this can be detected through the measurement of molecular biomarkers. However, no single biomarker has been identified as sensitive and specific enough to detect oxidative stress alone (159). Generally, products of cells with oxidative stress or tissue damage are observed after exposure to NPs. Examples of these are malondialdehyde (MDA), lipid peroxidation (LPO), myeloperoxidase activity (MPO) and protein carbonyl (PC) (**Table 1**). As expected, these biomarkers display similar patterns of ROS production. For example, zebrafish larvae exposed to ZnO NPs showed a significant increase in ROS and MDA levels at different concentrations (73). Ganesan et al. reported similar results: an increase in ROS, NO, and biomarkers such as LPO and PC in zebrafish embryos after exposure to CuO NPs (66). Using an in vivo approach with ZnO NPs, plasma in O. niloticus revealed that both small and large NPs incremented both ROS and MPO levels, showing a clear activation of oxidative stress (40). Biomarkers of oxidative stress have lower specificity for NP toxicity (159), thus, they should be used to complement ROS and antioxidant defenses results following NP exposure.

### Cytokines

Cytokines are small proteins produced by immune cells that act as signaling molecules within the immune system (160). Thus, cytokines regulate inflammatory signals against pathogens or any external agent such as NPs. This modulation of cytokine expression has been used to assess fish health following NPs exposure and mainly through molecular assays such as qPCR (**Table 1**). The rising number of studies evaluating cytokine gene expression in NPs toxicity context is likely related to the increasing cost effectiveness and availability of molecular techniques such as qPCR, and the rapid progress in the sequencing of fish cytokines (161–164). In this regard, exposure of S. aurata to Au NPs causes altered gene expression in head kidney (80), when pro-inflammatory cytokines such as IL-1β and TNFα were upregulated after 4 days exposure. Interestingly, through the same technique, authors observed an increase in oxidative stress in head kidney (80). Picchietti et al. observed an increase in IL-8 and TGF-β expression, and internalization of TiO<sup>2</sup> NPs in DLEC cells (a cell line established from D. labrax) after 1 day of exposure (62). Another study observed upregulated expression of IL-1β and TNF-α in intestine of E. coioides after exposure to Cu NPs for 25 days (77). Together, these results showed that cytokine gene expression is sensitive to NP exposure. Hence, cytokines emerge as a possible biomarker for monitoring the impact of NPs on overall fish health, since current methodologies provide sensitive platforms to assess variations at the molecular levels.

### Lysozyme

Lysozyme is a relevant defense component of the innate immune system through its antibacterial activity. Furthermore, lysozyme can also act as an opsonin and activate complement system and phagocytes (165). It is widely distributed in mucus, plasma, kidney, spleen, intestine, and gills (166–168). Lysozyme activity in plasma or serum is a standard ecotoxicological biomarker in fish (169). However, to date there are only three studies which have evaluated lysozyme activity in fish after exposure to NPs (**Table 1**). For example, E. coioides exposed to Cu NPs and copper sulfate NPs (CuSO4) revealed diminished lysozyme activity in the intestine after 25 days (77). This suppressive effect was also observed in blood samples in O. niloticus exposed to Fe2O<sup>3</sup> NPs after 60 days (27). In another study, this decrease in lysozyme activity in the serum was only observed for large ZnO NPs at the lowest concentration on day 14 (40). Some studies have described that lysozyme are able to bind metal oxide NPs (170, 171) producing a decrease of lysozyme activity (171). Thus, NPs have

a suppressor effect on lysozyme activity in fish. Hence, lysozyme activity represents a good indicator of immunosuppression that NPs can cause in fish. However, more research is needed to show the effectiveness of lysozyme as a biomarker due to the limited number of existing studies until today. Among others, analysis using different species, tissues (e.g., mucus on the skin), NPs and concentration of NP concurrently are needed to support consistent results. Ideally, performing a pathogen challenge after NP exposure would be best to assess the complete effect of NPmediated lysozyme suppression on fish response to pathogen exposure.

### INDUSTRIAL POINT OF VIEW

The industrial production of engineered NPs has grown at a considerable rate with an increasing number of commercial products utilizing them, which includes; paints, fabrics, cosmetics, treated wood, electronics, and sunscreen (172). Established in 2005, the Nanotechnology Consumer Products Inventory (CPI) listed 54 consumer products containing nanomaterials. Over 1,800 products from 622 companies in 32 countries are currently inventoried (173).

The rate of development for environmental exposure limits and monitoring policies has been surpassed by the growth rate of this emerging class of pollutants. The resulting gap in regulations for engineered NP has been identified as a key focus area in Europe and the United States through the European Commission and the National Institute for Occupational Safety and Health, respectively (174). Monitoring methodology able to detect NPs present in the environment is very limited, combined with inherent challenges for sampling, creating barriers in the ability to distinguish adverse effects (175).

To complement biological read-outs as those described above, analytical techniques used to measure regulated compounds typically rely on reactive potential within a closed system (i.e., conductivity, polarity, bond with coloring agent) thus allowing for cost effective high throughput production of samples. Required detection limits for NP's reduces options to highly sensitive instrumentation i.e., field-flow fractionation, sizeexclusion chromatography, liquid chromatography, transmission electron microscopy (TEM), and atomic force microscopy (AFM) (176). Cost restrictions and equipment infrastructure are limitations preventing methodology viability.

Alternative methods based on bioanalytics are not bound by individual chemical structures, which enables them to assess the net effect on whole biological systems (177). The European Environment Agency (EEA) and the Australian Governments National Water Initiative (AU NWI) have directed significant resources to the evaluation of biology-based monitoring tools to ensure water quality. The European Union Water Framework Directive (EU WFD) and the Australian Governments National Urban Water and Desalination Plan have both found that bioanalytics offered a distinct advantage as a monitoring tool offering the only read-out that integrates the effects of complex mixtures during evaluation of water quality on important ecosystem functioning (177). Interactions of NPs with the innate immune system of fish elicit a number of several, quantifiable, and reproducible responses such as ROS, antioxidants defenses, internalization of NPs, and cytokine production. Therefore, biomarkers based on innate immune responses offers significant opportunities for the development of robust methodologies that can provide functional biological outputs assessing the health of an aquatic ecosystem following exposure to a variety of NPs.

### CONCLUDING REMARKS

The literature summarized here explores a range of possible innate immune biomarkers as tools for the assessment of fish health following NP exposure. These biomarkers reveal multiple alterations produced by NPs into diverse organs and tissues in multiple fish species (**Figure 1**). To date, oxidative stress is the most widely innate immune biomarker studied to demonstrate NPs toxicity in fish, particularly using the measurement of ROS, and antioxidant defenses. This is attributed to its high sensitivity to NPs exposure and the remarkable consistency of results among studies on a wide range of fish species and targets. These studies pointed out ROS and antioxidants defenses as a reliable, sensitive, and valid biomarkers of NPs toxicity in fish. Additionally, this review reveals other potential biomarkers that can be used to evaluated fish health following NPs exposure. For example, analysis of mucus on the skin, leukocyte functionality (macrophages and neutrophils), internalization of NPs, cytokine expression, and lysozyme levels. These potential biomarkers showed significantly promising results, although, more research is needed to determine their consistency and reliability. Furthermore, some innate immune parameters used on current literature are not valid options as biomarkers to evaluated NPs toxicity. For example, histopathology, and NP bioaccumulation in gills and biomarkers of oxidative stress. These parameters

### REFERENCES


do not provide accurate and relevant information in NPs toxicity context. Despite these considerations, industry needs biomarkers as tracking tools to evaluate NPs toxicity, because analytical techniques present cost restrictions and equipment infrastructure limitations that constraint methodologies viability. In conclusion, it can be stated that innate immune biomarkers are promising tools to assess fish health following NPs exposure. However, much work must be done in order to test and interpret some of biomarkers responses present in this review. Based on the diversity of these parameters, the number of studies analyzing NP effects in fish is yet limited. Further studies on the impact of NPs will provide a better understanding of detrimental effects of NPs to fish health.

### AUTHOR CONTRIBUTIONS

The conceptualization of the review was performed by DT and DB. DT was the primary writing contributor to this review. JM-B and JW wrote and edited parts of the manuscript. DT and DB edited the final manuscript. All authors approved the final version of the manuscript.

### ACKNOWLEDGMENTS

DT was supported by a CONICYT-Chile postdoctoral fellowship (Becas Chile N◦ 74170029). JM-B was supported by a National Fund for Innovation in Science and Technology (FINCyT/Innovate-Peru) scholarship and a Graduate Teaching Assistantship by the Department of Biological Sciences at the University of Alberta. This work was supported by an NSERC Discovery grant RGPIN-2018-05768 (Canada) to DB.


reticulata exposed to iron oxide (maghemite) nanoparticles. Aquat Toxicol. (2018) 198:63–72. doi: 10.1016/j.aquatox.2018.02.014


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

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

# Frog Skin Innate Immune Defences: Sensing and Surviving Pathogens

Joseph F. A. Varga, Maxwell P. Bui-Marinos and Barbara A. Katzenback\*

*Department of Biology, University of Waterloo, Waterloo, ON, Canada*

Amphibian skin is a mucosal surface in direct and continuous contact with a microbially diverse and laden aquatic and/or terrestrial environment. As such, frog skin is an important innate immune organ and first line of defence against pathogens in the environment. Critical to the innate immune functions of frog skin are the maintenance of physical, chemical, cellular, and microbiological barriers and the complex network of interactions that occur across all the barriers. Despite the global decline in amphibian populations, largely as a result of emerging infectious diseases, we understand little regarding the cellular and molecular mechanisms that underlie the innate immune function of amphibian skin and defence against pathogens. In this review, we discuss the structure, cell composition and cellular junctions that contribute to the skin physical barrier, the antimicrobial peptide arsenal that, in part, comprises the chemical barrier, the pattern recognition receptors involved in recognizing pathogens and initiating innate immune responses in the skin, and the contribution of commensal microbes on the skin to pathogen defence. We briefly discuss the influence of environmental abiotic factors (natural and anthropogenic) and pathogens on the immunocompetency of frog skin defences. Although some aspects of frog innate immunity, such as antimicrobial peptides are well-studied; other components and how they contribute to the skin innate immune barrier, are lacking. Elucidating the complex network of interactions occurring at the interface of the frog's external and internal environments will yield insight into the crucial role amphibian skin plays in host defence and the environmental factors leading to compromised barrier integrity, disease, and host mortality.

Keywords: amphibian, anuran, epithelial cells, mucosal tissue, antimicrobial peptides (AMPs), pattern recognition receptors (PRRs), skin microbiome, skin immunology

### INTRODUCTION

Nearly 8,000 amphibian species have been discovered to date (88% belonging to order Anura–frogs and toads) and approximately 150 new species are discovered each year (1). Collectively, frogs have evolved unique skin adaptations to live in aquatic and terrestrial environments (2, 3), while exhibiting common elements in their skin composition and structure (4–6). Skin is an integral interface between an organism's internal and external environment and undergoes routine exposure to a myriad of environmental factors, including pathogen challenge. Frog skin is no exception; it acts as a critical immune organ constituting a complex network of physical, chemical, immunological, and microbiological barriers to pathogen insult. Striking commonalities exist between frog, fish, and mammalian skin and exemplify the importance of endeavours in comparative vertebrate skin biology to address numerous research areas (7, 8). As a consequence of their reliance

#### *Edited by:*

*Stephanie DeWitte-Orr, Wilfrid Laurier University, Canada*

#### *Reviewed by:*

*Jonathan P. Rast, Emory University School of Medicine, United States Serge Morand, Center for the National Scientific Research (CNRS), France*

#### *\*Correspondence:*

*Barbara A. Katzenback barb.katzenback@uwaterloo.ca*

#### *Specialty section:*

*This article was submitted to Comparative Immunology, a section of the journal Frontiers in Immunology*

*Received: 31 October 2018 Accepted: 18 December 2018 Published: 14 January 2019*

#### *Citation:*

*Varga JFA, Bui-Marinos MP and Katzenback BA (2019) Frog Skin Innate Immune Defences: Sensing and Surviving Pathogens. Front. Immunol. 9:3128. doi: 10.3389/fimmu.2018.03128* Varga et al. Frog Skin Innate Immune Defences

on terrestrial or aquatic habitats, or a combination thereof, amphibian skin is a sophisticated mucosal organ with specialized adaptations required to perform various critical physiological functions (e.g., ion transport, respiration, water uptake, etc.), while still maintaining a selective barrier to the external environment (2, 3, 9). Other than the presence of a sophisticated glandular system, a miraculous feature of amphibian skin that sets frogs apart from other vertebrates is their ability to rapidly heal deep wounds which protrude through the dermal layers without scar formation, including complete regeneration of any glands affected by the injury (8). Despite extensive studies showing that amphibian skin is vital to survival, and apart from antimicrobial peptides (AMPs) (10, 11), relatively little focus has been placed on examining the role of frog skin epithelium to pathogen defence. This focus is paramount since mucosal epithelia are more prone to pathogen attack, such as that seen in mammalian lung and gut epithelium (12–15). With the rise of declining amphibian populations globally (16), wherein emerging infectious diseases such as frog virus-3 (FV3), the type species of the genus Ranavirus (family Iridoviridae), and the fungal pathogen Batrachochytrium dendrobatidis (Bd) (17, 18) are believed to be the proximal cause (19, 20). It is therefore important to understand the interplay between frog skin, pathogens and contributing environmental factors.

### AMPHIBIAN SKIN—THE FIRST BARRIER OF DEFENCE

Maintenance of amphibian skin integrity is important for overall frog health—both in terms of conducting essential physiological processes and for defence against invading pathogens. Depending on the species, amphibian skin contributes to water uptake, ion transport, respiration, heat transfer, camouflage, and predator deterrence (9). Yet frog skin is particularly vulnerable to cutaneous injury due to the relatively thin and permeable nature of the organ—characteristics necessary to support many of the aforementioned physiological processes. Thus, frog skin is an important first line of defence against harmful agents in the environment that may disrupt skin function and/or cause cutaneous or systemic diseases, leading to interruption of essential physiological functions and ultimately amphibian death. In addition to common innate immune elements, amphibians have evolved specialized features to enhance innate immune defences to protect the vulnerable skin barrier, including a glandular network beneath the skin surface that are capable of producing a plethora of antimicrobial and toxic substances, thus aiding in the defence against pathogens and predators (6). While much remains to be elucidated, the holism between amphibian skin, host physiology and immunity is apparent.

### Skin Layer Organization and Composition

Frog skin is composed of epidermal and dermal layers, with each layer predominantly consisting of epithelial and fibroblastic cells, respectively. While mammalian epidermal strata layers are welldefined due to its thickness, frog epidermis is relatively thin and thus often limited to the stratum corneum (outermost layer), central stratum spinosum, and stratum germinativum (basal layer) (**Figure 1**) (7). Frog epidermis is composed of stratified squamous epithelium, wherein the stratum corneum is composed of a very thin layer of keratinized cells (**Figure 1**) (7, 21). Cells in the epidermis of tadpoles are ciliated in most of the frog species studied and cilia regress leading up to metamorphosis. In general, this is characterized by a global loss of ciliated skin cells at Gosner stages 25–30 with the exception of the retention of cilia around the eye and nasal areas (22, 23). To date, there are no studies on the importance of the mucociliary epithelium in adult frogs. We presume the mucociliary function in amphibians is similar to that of other organisms, where the cilia play an important role in sweeping trapped microbes away from mucosal surfaces (24, 25). The stratum spinosum is composed of terminally differentiating cells, acting as an intermediate layer between the stratum corneum and the regenerative stratum germinativum layer (7). The stratum germinativum, which directly connects to the dermis, contains a mixture of cell types including epithelial cells, immune cells (described in the paragraph immediately below) and chromatophores that provide frogs with dynamic pigmentation patterns (26). The dermal layer can be divided into two distinct layers: the upper spongious dermis and lower compact dermis (**Figure 1**). The spongious dermal layer is composed of loose connective tissue, while the compact dermal layer is formed by a series of interweaving collagenous fibre bundles, with fibronectin situated between breaks in the collagenous layers (**Figure 1**) (27, 28). Fibroblastic cells, which produce collagenous fibres to form connective tissue, are integral in anchoring the epidermal and dermal layers to the hypodermis particularly through collagenous columns (**Figure 1**) (27). A unique feature to select, mainly terrestrial, adult amphibian dermis is the separation of spongious and compact dermis by the Eberth-Katschenko (EK) layer (**Figure 1**) (5). This noncellular layer is composed entirely of glycosaminoglycans and glycoconjugates, wherein hyaluronan and dermatan sulphate have been reported as key constituents in various species (29, 30). Hyaluronan and other hyaluronan-like molecules in the EK layers are predominantly found on the dorsal side of amphibian skin. Hyaluronan molecules are proposed to reduce water evaporation thereby aiding in the prevention of desiccation, particularly in basking amphibians, since the molecules are highly water retentive (30). In addition to the EK divide, thick

**Abbreviations:** AMP, antimicrobial peptide; Bd, bactrachochytrium dendrobatidis; CCRs, chemokine receptors; CLRs, C-type-lectin like receptors; CPF, caerulein precursor fragment; DAMPs, damage-associated molecular patterns; dsRNA, double stranded RNA; EGFR, epidermal growth factor receptor; EK, Eberth-Katschenko; FPLR1, formyl peptide receptor-like-1; FV3, frog virus 3; GLP-1, glucagon-like peptide 1; GPCR, G-protein coupled receptor; hBD, human beta defensin; HDP, host defence peptide; IC50, inhibitory concentration 50; LGP2, laboratory of genetics and physiology 2; LPS, lipopolysaccharide; MAMPs, microbe-associated molecular patterns; MDA5, melanoma differentiationassociated gene 5; MIC, minimum inhibitory concentration; NF-κB, nuclear factor kappa-beta; NLR, nucleotide-binding oligomerization domain- (NOD-) like receptor; NOD, nucleotide-binding oligomerization domain; P2X7, purinergic receptors; PAMP, pathogen-associated molecular pattern; PGLa, peptide glycineleucine-amide; PRR, pattern recognition receptor; RIG-I, retinoic acid inducible gene-I; RLR, RIG-I-like receptor; STAT3, signal transducer and activator of transcription 3; TLR, TOLL-like receptor; TLR, TOLL-like receptor; XPF, xenopsin precursor fragments.

microbiological barrier. The image was, in part, created with the aid of BioRender.

collagenous columns extend upwards from the hypodermis to the spongious dermis layer, without impacting the compact dermis integrity, and functions to anchor the layers of skin (27). This anchoring is completed by hemidesmosomes that connect epidermal cytoskeletal filaments to dermal collagenous fibrils (31). The epidermal and dermal layers are essential to the overall integrity of amphibian skin.

Though not largely studied, and thus not often described in studies examining frog integument, resident immune cells responsible for detecting and responding to pathogen exposure have been identified in frog skin (**Figure 1**). Intertwined amongst the skin epidermal cells of American bullfrogs, (Rana catesbeiana syn. Lithobates catesbeiana), Northern leopard frogs, (Rana pipiens syn. Lithobates pipiens), and African clawed frogs, (Xenopus laevis) are dendritic-like or Langerhans-like cells (32–34). Mast cells play an important role in inflammatory and anti-parasitic responses via degranulation of biologicallyactive compounds, such as histamine, (35) and have also been identified in histological preparations of R. catesbeiana skin tissues (29). Less common in the literature are sporadic reports on the presence of macrophages and lymphocytes in healthy skin tissue (7). However, during epithelial wounding or pathogen insult, recruitment of circulating immune cells to the site can occur. Although T cells do not appear to be resident in the skin tissues of frog species studied thus far, it is clear through skin allograft studies that cytotoxic T cells can infiltrate the frog skin tissue and mediate rejection of non-self-tissue and exemplifies the conservation of adaptive immunity and allograft rejection akin to mammalian studies (32, 36, 37). In addition, B cells were also found capable of infiltrating frog skin in response to transplantation of Western clawed frog (Xenopus tropicalis) skin onto X. laevis (38).

### Glands

A hallmark of amphibian skin is the presence of varied glands located in the spongious dermal layer (**Figure 1**) that support the vital physiological functions performed by frog skin including, but not limited to, respiration, ion regulation, water transport, immune function and predator defence (2, 6, 9). The most ubiquitous and prominent glands in amphibian skin are mucosal glands and granular glands. Both types of glands are established in a sac-like formation surrounded by secretory cells that release granular contents and, myoepithelial cells that contract in the presence of appropriate stimuli (**Figure 1**) (39– 43). While the precise molecular mechanisms have not yet been elucidated, whole frog studies have demonstrated that electrical stimulation, injection with norepinephrine to the dorsal lymph sacs, or chasing a frog in a bucket for 5–10 min, stimulates the release of mucosal and granular gland contents (17, 39, 40, 42, 43).

Mucosal glands secrete mucus to maintain the moisture, permeability and elasticity of the skin, all of which are necessary for amphibian homeostasis (2, 9, 44). Though species dependent, mucosal glands are generally widely distributed across frog dorsal and ventral skin, with a higher density existing on the dorsal surface (45). The pattern of mucus discharge also varies across species. In general, terrestrial and basking frogs appear to secrete mucus at a more constant rate to aid in heat exchange and water balance. Since aquatic, arboreal and nocturnal frogs do not experience the same level of evaporative water loss, the maintenance of skin moisture is more dependent on the environment (44, 46). Notable exceptions to this general observation are X. laevis and X. tropicalis; although they are largely aquatic in nature, these Sub-Saharan native frogs appear to maintain continuous mucus coverage (47–49). In accordance with this, the observation of skin in Telmatobius aquatic frogs showed a more even density of granular and mucosal glands between the dorsal and ventral skin, although the mucosal glands are relatively small (50). The study of mucus production in the skin remains challenging. There is difficulty in determining natural physiological parameters of mucus production such as the volume of mucus on the skin, rate of mucus production and discharge, and the ability to determine exact concentrations of skin-secreted compounds in the mucus.

Granular glands, which include small mixed glands and other types of specialized granular glands (**Figure 1**), have been identified in frog skin and contain bioactive molecules involved in host defence and predator defence. Granular glands, and their contents, are arguably the best studied amphibian skin gland due to the rich diversity of biomolecules they secrete– notably antimicrobial peptides and toxic alkaloids. Although commonly referred to as granular glands, these glands have the potential to secrete serous fluid, or toxic substances, and are therefore also known as serous or parotoid/venom glands (6, 41). Granular (serous) glands contain bioactive molecules with demonstrated broad-spectrum antimicrobial activity, many of which are classified as antimicrobial peptides (AMPs, discussed below; **Figure 1**) (6, 51, 52). Granular (parotoid/venom) glands may also sequester and release toxic alkaloid biomolecules that function in predator deterrence and/or defence (6, 53, 54). While relatively less abundant, granular glands maintain a similar distribution, and density pattern compared to mucosal glands wherein there is a higher density of granular glands on the dorsal side than the ventral side (45, 55). Granular glands appear to be further concentrated in specific regions of the skin, such as the central region of the skin vs. head or leg regions (45, 55). Similar to granular glands, small mixed glands host a reservoir of biologically active molecules or mucus and appear more evenly spread across the skin surface (29, 56). Other types of specialized glands have been identified in certain frog species with apparent functions ranging from greater granular content storage capacity, lipid secretion, and odorous secretion for predator deterrence (40, 57, 58). Skin gland diversity, both in type and chemical composition within the glands, varies with frog species and developmental stage (59–61).

### SKIN AS A PHYSICAL BARRIER

### Cellular Junctions and Importance in Barrier Integrity

Across all vertebrates, skin is, undoubtedly, an important physical barrier between an organism and its environment. Skin barrier integrity and permeability is maintained by cellular junctions primarily between epithelial cells and include tight junctions, gap junctions, adherens junctions, and desmosomes (4, 62– 64). The key proteins which comprise these junctions (and those present in mammalian skin epidermis) include claudins (claudin-1) and occludins that form tight junctions, cadherins (Ecadherin) that form adherens junctions, connexins (connexin-43) that comprise gap junctions, and desmogleins (desmoglein-3) that comprise desmosomes (62). All cellular junctions are pertinent to overall skin integrity: tight junctions connect neighbouring cells at the apical membrane, adherens junctions, and desmosomes aid in further stabilizing cell-cell adhesion, and gap junctions form channels between adjoining cells necessary for cell-cell communication (62). Tight junction proteins are detected as early as the gastrulation stage and persist until full development (4). The presence of tight junction claudin-1 proteins is crucial during gastrulation in X. laevis embryos (65), but general observations of tight junction proteins in adult frogs are lacking. In early larvae, tripartite junctional complexes of tight junction, adherens junction, and desmosomes are observed, wherein these complexes appear to lose significant contribution from adherens junctions in larvae approaching metamorphosis and in adult frogs (4). Nonetheless, strong expression of adherens-dependent cadherin protein has been detected in adult X. laevis skin (66). While the presence of gap junctions and desmosomes have been reported in the skin of other vertebrates, the observation of these junctions in frogs has been limited to observations in frog embryos undergoing development, or simple presence identification in adult frogs (67–70). Though all cellular junctions have been identified in frogs at different developmental stages, it is important to note that these studies have been limited to X. laevis and R. pipiens species and thus may not necessarily be representative of all frogs. Collectively, epithelial cell junctions allow for a continuous epithelial network that is relatively closed to the external environment while remaining open to the basal collagenousrich dermal layers. As such, maintenance of epithelial cellular junctions is important for barrier integrity, and thus pathogen defence, particularly considering the relatively thin epidermal layer in frogs.

Skin permeability, and thus barrier integrity, is a feat made possible by cellular junctions, wherein changes in junction proteins in response to environmental conditions regulate permeability. Tight junctions are specially known to contribute to paracellular transport of molecules (i.e., through the intercellular space and across epithelium) and thus integral to epithelial permeability in mammals, fish and frogs (64, 71, 72). A plethora of studies in mammalian models describe the impact on barrier integrity, and thus barrier function, in response to various skin diseases or environmental factors (62, 71, 73). In general, downregulation of tight junction associated proteins is widely observed among an array of human skin diseases, relating to a weakening of barrier integrity (71, 73). Presence of microbes, whether commensal or pathogenic, triggers an initial upregulation of genes encoding for tight junction proteins, and thus skin barrier strengthening (71, 73). However, persistence of pathogens leads to downregulation of gene expression for junction proteins and eventual weakening of the skin barrier (71, 73). In addition to this, studies on mammalian and fish mucosal tissue, such as the gastrointestinal tract, have defined the importance of barrier integrity in response to pathogen invasion (12, 13, 15, 74). In adult frog skin, the interplay between cellular junctions and influx/efflux of water and ions demonstrates the participation of tight junctions in acting as a selective permeable interface between the frog and its environment (2, 64, 75). While the importance of the skin barrier and of the cellular junctions necessary for maintaining barrier integrity is well-reported in vertebrates (63, 73, 76), the investigations on the regulation of skin barrier integrity in adult frogs in response to environmental stimuli is lacking.

Skin sloughing, a normal process in the maintenance of amphibian skin (77, 78), may function as an innate immune barrier. Skin sloughing may serve to remove skin-associated microbes, including pathogens (79), and the rate of skin sloughing increases with certain infections, perhaps as a mechanism to limit pathogen numbers on the skin (77). However, sloughing also exposes the underlying non-keratinized layers of the skin barrier (77). The underlying mechanism controlling the rate of skin sloughing is unclear and requires further investigation.

### Mucus

Mucus plays a critical role in physical and chemical defence against pathogen invasion (12, 13, 24, 74). Recent studies observing the epidermis of X. tropicalis tadpoles have identified the development of multiciliated cells, ionocytes, goblet cells, and small secretory cells as integral to establishing a mucosal barrier (49, 80, 81). Manipulation of the mucus barrier composition in X. tropicalis tadpoles has demonstrated the key role of the Otogelin-like structural mucin glycoprotein, that provides a 6µm thick mucosal surface barrier on tadpoles, towards conferring protection to infection of tadpoles with Aeromonas hydrophila (49). Presumably in conjunction with mucous, ciliated cells within the epidermal layer aid in removing trapped pathogens from the skin surface (24, 25, 49, 82). Current observation of the contribution of the skin mucus macromolecule composition in adult frogs to pathogen defence is lacking. In this regard, the mucus functions as a physical barrier. Yet, mucus also provides a framework for the various secretions from granular and small mixed glands, thereby contributing to the establishment of a formidable chemical barrier (2, 6, 44).

## CHEMICAL BARRIERS

### Antimicrobial Peptides

In general, antimicrobial peptides (AMPs) from metazoans are gene-encoded cationic and hydrophobic molecules ranging from 12 to 50 amino acids in length (83). AMPs have been shown to aid in the direct defence against pathogens and recent investigation has uncovered the role of AMPs in modulating immune responses in human and mouse systems (84–86). Frog skin is the most abundant natural source of AMPs found on earth (87, 88). The diversity of AMPs in frogs may not be surprising considering the biphasic life cycles of many frog species; residing in an aquatic environment during tadpole development and transitioning to a terrestrial environment post-metamorphosis. Exposure of frogs to aquatic and terrestrial pathogens, or contact with other animals that serve as pathogen reservoirs, can enhance the incidence of disease and host mortality (89, 90), necessitating the evolution of a broad arsenal of antimicrobial defence. AMPs in human skin have been extensively characterized and it is generally accepted that disruption of AMP expression may lead to cutaneous disease (91–94). Similarly, a lack of AMPs on frog skin has been shown to be detrimental to adult X. laevis defence against the fungal pathogen Bd (17). It is evident that AMPs serve a significant role in the defence of frog skin against pathogens, however our understanding of the ability of frog AMPs to exert antimicrobial activity on frog pathogens is limited and knowledge surrounding their potential immunomodulatory activity in frogs is completely lacking.

### Structure and Diversity of Frog Skin Derived Antimicrobial Peptides

To date, 1,078 unique AMPs have been identified from amphibians (95). Collectively, amphibian AMPs are slightly shorter than mammalian AMPs, ranging from 12 to 46 amino acids (96), with no two AMPs identical in amino acid composition. Although metazoan AMPs can be classified into one of four groups based on structure alone, including alpha-helical, beta-sheet, mixed and linear, most amphibian AMPs belong to the alpha-helical and linear groups of peptides (97, 98). The major classes of frog AMPs include: brevinins, cathelicidans, dermaseptins, esculentins, japonicins, magainins, nigrocins, palustrins, ranatuerins, ranalexins, temporins, and tigerinins (99–103), although not all AMP classes are expressed in the skin of any given frog species. For example, X. laevis harbours four distinct families of AMPs: caerulein precursor fragment (CPF), peptide glycine-leucine-amide (PGLa), xenopsin precursor fragments (XPF) and magainins (104). In fact, the most well-characterized frog AMPs to date are of the magainin family, magainin-1 and magainin-2 (105–107). Magainin-1 and magainin-2 are both 23 amino acids and differ in the composition of 2 amino acids. Both magainins possess an alpha-helical structure and, like most AMPs, are amphipathic (105). The native structure and biochemistry of frog AMPs is particularly important as it dictates AMP function (108) and allows for intrinsic interactions with anionic membranes, such as those found on bacteria, fungi, viruses, and parasites (109). The association of the frog skin AMPs with anionic membranes, and the mechanisms by which they disrupt membrane integrity, are well-studied (110–114). The mechanisms responsible for disrupting membrane integrity are heavily influenced by lipid composition (111, 115, 116), and include lipid flip-flop, leakage, or transmembrane integration (111, 115, 117).

The distribution of AMPs across frog species is sporadic and some do not appear to synthesize AMPs at all (118). The ability to synthesize AMPs has been suggested to confer an evolutionary advantage to frogs but is not required for the survival of a species (118). For example, Coqui frogs (Eleutherodactylus coqui) have been shown to survive with a lack of AMPs, even when the deadly chytrid fungus, Bd, is detected on their skin (119). However, discovery of AMPs has traditionally relied on the isolation of active fractions from amphibian skin or amphibian skin secretions and in vitro testing on microbes of human importance (120, 121). Thus, there may exist additional AMPs present in amphibian skin that have previously gone unidentified (122). The use of transcriptomic approaches to investigate immune function of frog skin has yielded an effective strategy to identify AMP peptide diversity across frog species, developmental stage, and environmental factors (e.g., abiotic and biotic elements) (123–125). Recent transcriptomic approaches applied to frog skin tissues have illustrated the power of untargeted approaches to identify AMPs in frog skin and suggests the existence of a greater number and diversity of AMPs produced in individual frog species (126, 127).

AMP secretion from granular glands is constitutive and can be inducible in response to stress, injury or infection (99, 128). Although difficult to quantify the rate at which peptides are secreted, concentrations of peptides in the skin mucus of X. laevis has been reported at 3,256µg/ml (constitutive secretion) whereas the average amount of AMPs found in the skin mucosal secretions of chase-stressed or norepinephrine injected X. laevis (inducible secretion) was 19,581 and 41,646µg/ml, respectively (17). Both transcription and translation are likely responsible for the low levels of AMPs found on the skin of resting animals (17). However, few studies have examined the molecular mechanisms that lead to the inducible transcription of frog skin derived AMPs (129). In other organisms, such as humans, bovine and insects, the promoter regions of AMP genes have been found to harbour nuclear factor kappa-beta (NF-κB) transcription factor binding motifs and were identified as important regulatory elements for AMP gene expression (130–132). Nuclear factor kappa-beta (NF-κB) may also stimulate the transcription of AMP genes in frog skin as NF-κB has been shown to immunolocalize with the glandular cells of Chinese brown frogs (Rana dybowskii) (129, 133). However, nuclear localization was not apparent from these studies. In addition, NF-κB, nuclear factor NF-IL6, or cisregulatory element 2 (CRE2) transcription binding sites have been identified in the promoter regions of several AMP genes in wrinkled frogs (Rana rugosa) (134), oriental fire-bellied toads (Bombina orientalis) (135), X. laevis, and X. tropicalis (136). Future investigation is required to dissect the potential role of NF-κB-mediated frog skin AMP gene expression, and/or other putative transcription factors, in the maintenance of frog skin homeostasis and rapid AMP production and secretion during stress, wounding, or pathogen insult.

### Functions of Frog Skin Derived Antimicrobial Peptides

### **Direct antimicrobial activity towards frog pathogens**

Extensive investigation has demonstrated frog AMPs to exert broad-spectrum antimicrobial activity against human pathogens, including bacteria, viruses, fungi and parasites, reviewed in (120, 121, 137, 138). Only recently, however, has there been a shift in focus towards understanding whether frog skin derived AMPs are antimicrobial to frog pathogens. Emerging infectious diseases continue to decimate worldwide amphibian populations and, pathogens, such as ranaviruses and Bd, are implicated as proximal causes in frog declines (139). It is critical to gain a further understanding of how to mitigate these diseases in order to conserve dwindling frog populations.

Frog skin derived peptides that have been tested for antipathogen activity span a diverse range of peptide families from several frog species and collectively have anti-bacterial (**Tables 1**, **2**), anti-viral (**Table 3**), anti-fungal (**Table 4**), and anti-parasite (**Table 5**) activities. Frog AMPs are effective antimicrobial agents against Aeromonas sp., the causative agents of red-leg, a polymicrobial disease that is characterized by congestion of the skin, ulceration, haemorrhage, bloating, failure to respond to stimuli, and tetanic seizures (150). Differences in susceptibility to AMPs exist across Aeromonas sp. and illustrate that there exists some microbial selectivity to antimicrobial action. For example, Aeromonas caviae are highly susceptible to dermaseptin-S1 from the waxy monkey tree frog (Phyllomedusa sauvagii) with minimal inhibitory concentrations (MICs) as low as 0.5–1µM, while other Aeromonas strains such as A. hydrophila have been reported to be resistant to dermaseptin-S1 (**Table 1**). In addition, some bacteria appear to be completely refractory to antimicrobial peptide families. For example, AMPs (either single AMPs or mixed preparations) from X. laevis failed to inhibit A. hydrophila growth (**Tables 1**, **2**). However, peptides from X. laevis were very effective against Citrobacter fruendii, another causative agent of red-leg, either alone or in combination (**Tables 1**, **2**). In addition, magainin-2 alone was not effective against Chryseobacterium meningiosepticum but when the natural mixture of X. laevis skin secretions was applied to this pathogen, it was effective at reducing its growth (**Tables 1**, **2**). This evidence suggests that some peptides may require synergy to work against select pathogens.

Currently, the only viral pathogen of frogs that skin AMPs have been tested on is FV3. Frog skin AMPs have mixed antiviral efficacy on FV3. While dermaseptin-S1 from the waxy monkey tree frog, (P. sauvagii) and temporin A from the common frog (R. temporaria) were capable of inactivating FV3, magainin-2 from X. laevis was not able to inhibit FV3 infectivity at the AMP concentrations tested (10). The synergistic activity of AMPs towards FV3 is unknown.

Not surprisingly, the majority of frog skin derived AMPs tested against fungal pathogens of frogs have focused on Bd (**Table 4**). Based on the MICs reported, the most effective antifungal frog skin AMPs belong to X. laevis and Ranid species, the foothill yellow-legged frog (Rana boylii) and the Oregon spotted frog (Rana pretiosa) (**Table 4**). The promising effects of frog skin AMPs have been shown be effective against Bd zoospores in vitro



*ND indicates not determined (i.e., at the level tested, the AMP was found to have no effect on the pathogen). MIC values are reported in* µ*M unless indicated otherwise.*

(151, 152) and important in X. laevis skin defence against Bd in in vivo infection studies (17). Although magainin-2 and PGLa applied individually to Bd and Basidiobolus ranarum, (another fungus that infects the skin of amphibians) were quite effective at reducing fungal growth, the combination highly reduced the MIC required to inhibit the fungi (i.e., was more potent) (**Table 4**). This is strong evidence to support that synergistic mechanisms may be more beneficial in combating particular pathogens than individual peptides. In general, the minimal inhibitory concentration of frog skin AMPs required to inhibit fungal pathogens is much higher than the amount required to inhibit bacteria or viruses (**Table 4**).

There is also some evidence to support the anti-parasitic role of frog skin AMPs (11). A native mixture of peptides obtained from adult R. catesbaeiana skin was effective at inhibiting trematode cercariae viability at all AMP mixture concentrations tested (**Table 5**). However, the peptide composition was not determined. To date, 12 different AMPs have been identified in R. catesbeiana skin secretions (95). Albeit limited in number, these studies demonstrate frog skin AMPs to be direct antimicrobial agents in innate immune defence against frog pathogens.

### **Wound healing**

Research in murine models demonstrate that mammalian AMPs such as cathelicidan-related antimicrobial peptide are beneficial in combating skin infections in mice where they clear invading bacteria, activate immune cells and promote wound closure (78, 153–155). In mammalian systems, AMPs can bind cell surface receptors such as formyl peptide receptor-like-1 (FPLR1), purinergic receptors (P2X7), Toll-like receptors (TLRs), chemokine receptors (CCRs), G-protein coupled receptors (GPCRs), and epidermal growth factor receptor (EGFR) to activate downstream signalling pathways to promote wound healing (156). A few studies have examined the ability of frog skin AMPs to promote wound healing in mammalian models. The application of cathelicidan-NV from the skin of a plateau frog (Nanorana ventripunctata) onto wounded mouse skin resulted in the acceleration of wound re-epithelization by direct stimulation of keratinocyte motility and proliferation (157). Cathelicidan-NV treatment also upregulated numerous genes involved in migration, proliferation and differentiation in wounded mouse skin tissue (157). Another frog skin AMP, esculentin-1a(1-21) from the common European frog (Rana


TABLE 2 | Minimal inhibitory concentration (MIC) of skin secretions containing frog skin-derived antimicrobial peptides against amphibian bacterial pathogens.

*ND indicates not determined (i.e., at the level tested, the AMP was found to have no effect on the pathogen). MIC values are reported in* µ*M unless indicated otherwise. For peptide mixtures, the antimicrobial peptides identified through mass spectrometry are summarized.*

esculenta), also promoted wound healing by stimulating human keratinocyte migration (158). Although frog skin AMPs are capable of promoting wound healing in mouse systems, and frogs are known for their remarkable wound healing ability, the function of frog skin AMPs in frog skin wound healing and the underlying cellular and molecular mechanisms are still unclear.

### **Immunomodulation of innate immunity**

The most well-characterized human AMPs belong to the cathelicidan (LL-37) and defensin (hBD-1, hBD-2, hBD-3, hBD-4, HD-5, and HD-6) families (156, 159, 160) and are also considered host defence peptides (HDPs) since they have been shown to modulate innate and adaptive immune responses of homologous host cells (156, 161). Treatment of mammalian cells with frog skin AMPs (i.e., heterologous system) revealed mammalian cells to be responsive to frog skin AMPs (157, 158, 162, 163). For example, Esculentin-1a(1-21) treatment of human keratinocytes resulted in increased phosphorylation of signal transducer and activator of transcription 3 (STAT3), activating the transcription of downstream genes involved in wound healing (158).


TABLE 3 | Inhibitory concentration 50 (IC50) of frog skin-derived antimicrobial peptides against amphibian viral pathogens.

*ND indicates not determined (i.e., at the level tested, the AMP or skin peptide mixture was found to have no effect on the pathogen). Values are reported in* µ*M where available.*

Cathelicidan-NV induced fibroblast-to-myofibroblast transition and also significantly increased collagen production in the wound (157). Another frog skin AMP, brevinins-1Pa, from R. pipiens, stimulated the release of insulin from rat pancreatic islet cells (162). Insulin is known to play a role in keratinocyte function by inducing migration through the PI3-Akt-RhoA network (164). Several AMPs from X. laevis (CPF, magainin-1, magainin-2, PGLa) and the Taiwanese frog (Holobatrachus rugulosas) tigerinin-1R have been shown to stimulate the secretion of glucagon-like peptide 1 (GLP-1) from GLUTag cells (163). GLP-1 is an immunomodulatory molecule and decreases the inflammatory response during allergen and infection-induced inflammation (165). The experimental evidence suggests that various frog skin AMPs have a substantial effect on mammalian cells processes such as cell migration, inflammation, immunity and repair (158, 164, 166). Unfortunately, the functions of frog skin AMPs on frog cells (i.e., homologous system) have not been explored and whether frog skin AMPs act as HDPs in frogs remains unknown.

### Alkaloids

Lipid-soluble alkaloid compounds are believed to originate from the amphibian diet, largely from insects (167). Identification of these alkaloid compounds has mainly focused on those excreted by frogs in the Dendrobatidae family (poison dart frogs) with observation from over 150 species (168, 169). Nonetheless, toxic alkaloid substances have also been observed in Eleutherodactylidae, Leptodactylidae, Mantella, Myobatrachidae, and Ranidae frogs (54, 169– 171). Toxic alkaloids are primarily involved in predation avoidance, however, a few also participate in defence against microbes (167, 172). Readers interested in alkaloid compound diversity are referred to reviews on this topic (169, 173).

### EPITHELIAL CELLS AS MICROBIAL SENSORS AND INITIATORS OF INNATE IMMUNE RESPONSES

Epithelial cells are emerging as crucial contributors to innate immune responses through the detection of microorganisms– both commensal and pathogenic—in the external environment (174, 175) through the use of pattern recognition molecules. A relatively limited number of germ-line encoded pattern recognition receptors (PRRs) detect non-self and damage signals and these recognition events are crucial to initiating innate immune response. Classes of PRRs are generally divided into transmembrane and cytosolic PRRs. Transmembrane receptors include TLRs, C-type-lectin like receptors (CLRs), and scavenger receptors, while cytosolic PRRs include retinoic acid inducible gene- (RIG-) I-like receptors (RLRs), nucleotidebinding oligomerization domain- (NOD-) like receptors (NLRs), and various cytosolic DNA sensors (176). Collectively, PRRs recognize a variety of pathogen-associated molecular patterns (PAMPs), also known as microbial-associated molecular patterns (MAMPs), including lipopolysaccharide, peptidoglycan, lipopeptides, flagellin, single stranded RNA, double stranded RNA, double stranded DNA, carbohydrate structures, as well as other PAMPs (176). PRRs also recognize damageassociated molecular patterns (DAMPs) released upon cellular stress (177). Ligand sensing by PPRs leads to intracellular signalling cascades that regulate the transcription of genes encoding for pro-inflammatory, chemoattractive and antiviral functions (176). In accordance with the location of epithelial cells at the host-environment interface, epithelial cells in mammalian models have been shown to express diverse PRRs including TLRs (178), RLRs (179), and NLRs (180) to sense invading microorganisms and initiate innate immune responses.

Few studies have focused on the characterization of amphibian PRRs (e.g., ligands, signalling pathways, downstream gene targets), let alone their role in amphibian skin epithelial cell biology. Yet, it is evident that cells within frog skin tissue are capable of sensing bacterial, viral and fungal pathogens, including commercially available mimics of PAMPs, and initiate innate immune responses through the upregulation genes encoding for pro-inflammatory cytokines, anti-viral cytokines, antimicrobial peptides, and other immune proteins (181–183). The cell type(s) and receptors involved in microbial recognition by amphibian skin tissues are largely unknown. Thus, much of our basis for understanding the role of frog skin epithelial cells to microbial detection is limited to the identification of key pattern recognition molecules in the frog genome and implied conservation of their function based on limited expression data in frog skin tissues. In the below subsections, we summarize the current state of knowledge surrounding the presence of genes encoding for pattern recognition molecules identified in frog genomes and the expressions of these genes in frog skin tissues.


*MIC values are reported in* µ*M unless indicated otherwise. For peptide mixtures, the antimicrobial peptides identified through mass spectrometry are summarized.*

### Toll-Like Receptors (TLRs)

The first glimpse into the frog TLR multigene family came about through a bioinformatics approach to study the evolution of vertebrate TLRs and was spearheaded as a result of the influx of draft genome sequences of fish (e.g., Takifugu rubripes) and frog (X. tropicalis) (184). Molecular evolutionary analysis demonstrated that TLRs are evolving at approximately the same, slow rate and are under strong purifying selection, presumably to ensure maintenance of TLR function both in terms of ligand recognition and initiation of intracellular signalling cascades (184, 185). Through the construction of molecular trees, six major TLR families emerged (**Table 6**), each encompassing subfamilies of TLRs that that recognized a general set of PAMPs/MAMPs (184, 186). At least 19 TLR genes were identified in the X. tropicalis genome (JGI 4.1) and included orthologues of both mammalian and fish specific (e.g., TLR21, TLR22) TLRs. Characteristic of the mammalian TLR2 family is the ability of TLR2 family members to form heterodimeric pairs with TLR2 Varga et al. Frog Skin Innate Immune Defences



(176) to recognize a diverse set of ligands, and is presumed to also occurin frogs (184). In X. tropicalis the TLR2 family encompasses one TLR1, two TLR2, two TLR6, and four TLR14 subfamily members and appears to lack the TLR10 subfamily (**Table 6**) (184, 186). The TLR14 subfamily appears to have expanded in X. tropicalis, and possibly in other frogs, to four TLR14 subfamily members that are hypothesized to form heterodimeric pairs with TLR2, similar to other subfamily members of the TLR2 family (184). One member of each of the TLR3 (senses dsRNA), TLR4 (senses LPS) and TLR5 (senses flagellin) families were identified in X. tropicalis (**Table 6**) (184, 186). However, the putative X. tropicalis tlr4 gene does not appear to encode for a transmembrane region based on in silico structural prediction (186). Genes for cd14 or md-2, involved in TLR4 function in mammals (176), have not been identified in the X. tropicalis genome and thus the function of the putative X. tropicalis TLR4 as an LPS sensor is uncertain (186). Another interesting deviation from the mammalian system is the predicted presence of a soluble TLR5, termed tlr5s (184), similar to the soluble TLRs found in fish species (188). The tlr5s gene is predicted to encode for the extracellular leucine rich repeat (LRR) region and is lacking the transmembrane and intracellular TIR signalling domains suggesting it may act as a soluble receptor to potentially regulate TLR5 signalling (184). The TLR7 family is crucial for sensing endosomal PAMPs in mammals (189) and a single orthologue of tlr7 and tlr9, and two orthologues of tlr8 were identified in X. tropicalis (**Table 6**) (184, 186). Lastly, a single orthologue of TLR12, TLR13, TLR21 and TLR22 subfamilies were identified in X. tropicalis (**Table 6**) (184, 186). In silico prediction of X. tropicalis TLRs protein structures revealed overall similar X. tropicalis TLR structure to corresponding human TLR orthologues, including a similar size and number of LRR domains, transmembrane region and an intracellular TIR domain (186).

Aside from the identification of TLR genes in few frog species (125, 183, 184, 186), little investigation has focused on characterization of frog TLRs, and their role in frog skin innate immunity. In X. laevis, the TLR genes, including the putative tlr4, are expressed in the skin of tadpoles and adults (181, 186). Transcriptomic studies from skin of healthy Japanese brown frogs (Rana japonica), Montane brown frog (Rana TABLE 6 | Toll-like receptor genes identified in frog species.


*X. tropicalis tlr genes shown in bold are predicted to contain introns, non-bolded genes are predicted to be intronless.* \**Soluble short form lacks the transmembrane and TIR domains. \$Sequences were detected by RT-PCR with cDNA as a template, gene sequence structure not reported.*

ornativentris), Tago frog (Rana tagoi) (187), and the Yunnan firebelly toad (Bombina maxima) (183) have demonstrated the presence of tlr transcripts in skin tissue and further support the important role of anuran skin and the cells within as important sensors of microbes and regulators of innate immunity. Indeed, several transcriptomic studies of anuran skin tissues, including Ranidae, Megophryidae, Rhacophoridae, and Bufonidae families, revealed the enrichment of transcripts involved in processes reflected in the gene ontology terms "immune system process," "immune system," and "signal transduction," further supporting anuran skin as an immune organ (123–125). However, only a single study has examined the potential sensing of a PAMP by a frog TLR; LPS (10µg/ml) treatment of R. temporaria frog urinary bladder epithelial cells positive for TLR4 (albeit demonstrated through the use of non-homologous anti-TLR4 antibody) triggered epithelial cell activation through an NF-κB dependent mechanism (190). Although these urinary bladder epithelial cells appear to be LPS responsive, unequivocal evidence that TLR4 is responsible for LPS sensing is lacking.

### Cytosolic Pattern Recognition Sensors

RLRs, NLRs, and cytoplasmic DNA sensors are vital cytosolic pattern recognition molecules involved in initiating pro-inflammatory and anti-viral responses (191). RLR family members include retinoic acid-inducible gene-I (RIG-I), melanoma differentiation-associated gene 5 (MDA5), and laboratory of genetics and physiology 2 (LGP2) (191). In mammals, RIG-I and MDA5 bind viral RNA via the common RNA helicase domain and ligand recognition results in activation of interferon regulatory factor 3 and NF-kB transcription factors to initiate transcription of an anti-viral interferon response (191). LGP2 is known to interfere with viral RNA binding to RIG-I and MDA5 (192). While rig-i, mda5, and lgp2 genes have been identified in the X. tropicalis genome (193) and rig-i and mda5 found expressed in frog skin (181, 183), little else is known about the role of RLRs in anurans.

In mammals, NLRs are organized into five subfamilies (NLRA, NLRB, NLRC, NLRX, NLRP) based on the N-terminal effector domain and collectively sense a wide range of MAMPs (194). NLR activation leads to receptor oligomerization and formation of the inflammasome and activation of downstream inflammatory caspases that cleave interleukin 1 cytokine family members (IL-1, IL-18) (194). Seven NLR genes were identified in the X. tropicalis genome, including NLRA/CIITA, NLRC1/NOD1, NLRC3, NLRC4, NLRC5, and NLRX1, while NLRC2/NOD2 appears to be absent (195, 196). Members of all five NLR subfamily were identified in the B. maxima skin transcriptome including NLRA/CIITA, NLRB/NAIP, NLRC1/NOD1, NLRC3, NLRC5, NLRP1, NLRP3, NLRP5, and NLRX1 (183).

In addition to RLRs and NLRs, cytosolic DNA sensors are also expressed in frog skin. Amphibian skin transcriptomes from the Chinese giant salamander (Andrias davidianus), Asiatic toad (Bufo gargarizans), and black-spotted frog (Rana nigromaculata) revealed the presence of transcripts in the "cytosolic DNAsensing pathway" and the expression of a DNA-dependent RNA polymerase III that functions as a cytosolic DNA sensor by transcribing an RNA copy for recognition by RIG-I (124), suggesting a conserved evolutionary anti-microbial mechanism. However, an AIM-2-like receptor, another cytosolic DNA sensor that can lead to inflammasome activation, is seemingly absent in X. tropicalis (195).

### IMPACT OF ENVIRONMENT ON HOST BARRIERS

### Abiotic Factors

Frogs are to the environment as canaries were to coal mines. They are an important indicator species and their physiology is heavily influenced by the environment (197– 199). Most studies examining the impact of abiotic factors on amphibian skin have focused on AMPs. Temperature, dehydration, shade, acidification, oxygen and altitude (200, 201) have been documented to influence frog skin AMPs. For example, increased environmental temperatures (from 5 to 30◦C) triggered brevinin-1SY AMP production in R. sylvatica skin tissue (200). Albeit, the underlying mechanism for the production of brevinin-1SY at higher temperatures is unclear, increased microbial colonization of the skin at the higher temperature or increased transcriptional/translational kinetics may be involved. Microbes on the skin surface may stimulate PRRs on the membrane of epidermal cells, leading to downstream signalling that potentially induces transcription of AMP genes with NFκB in the promoter region (202). Cold stress in mammals (203) and cultured amphibian primary epidermal cells (204) has been shown to reduce the rates of transcription and translation, leading to decreased global protein synthesis. Thus, low body temperatures of R. sylvatica may have led to a near halt in AMP synthesis. In R. catesbeiana tadpoles, shade and acidification of the environment have been shown to modulate the production and bioactivity of AMPs (201, 205). Another environmental factor that has an effect on AMPs is hydration status. Dehydration in R. sylvatica increased the expression of brevinin-1SY in the skin (201). In addition to dehydration, other environmental stressors such as anoxia or freezing, also enhances the antimicrobial activity of R. sylvatica brevinin-1SY against select microbial strains (201). Decreased oxygen availability or hypoxia, has been associated with an increased number of granular glands in Tibetan frog (Nanorana parkeri) middorsal skin (199). The biological significance of increased granular glands found in hypoxic conditions is unknown. It is evident from these findings that the regulation of AMPs and the diversity among the AMP secretome is complex but is shaped by the environment.

### Chemical Contaminants

Anthropogenic factors, such as pesticides, also impair immunity and can reduce chemical skin defences (146, 206). Compared to mammalian skin, frog skin has significantly greater uptake potential of xenobiotics that can bioconcentrate and may be detrimental to frog health (207–209). In some instances, the chemicals exert a direct effect on the skin epidermal cells. For example, short-term exposure of Italian pool frog (Pelophylax bergeri) skin cultures to cadmium resulted in alteration and disorganization of the skin epidermal layers, and ultimately induced cellular and molecular stress responses (210). In addition, exposure to environmental contaminants has been documented to directly affect the paracellular transport of ions across frog skin (211, 212), wherein cellular junctions play an important role in ion transport (2, 64, 75). Chemical contaminants can also impact host immune function resulting in altered host resistance to pathogens. Pesticide exposure has been shown to influence antiviral immunity in larval and adult frogs that led to increased susceptibility to pathogen invasion (213–215). It is then proposed that the potential for chemical contaminants to impact epidermal organization and alter frog skin permeability leads to increased pathogen susceptibility and host mortality. In general, while these studies are comprehensive at analysing either the impact of contaminants on amphibian skin or effect on ion permeability and pathogen susceptibility, none appear to directly report the regulation of cellular junctions in combination with pathogen susceptibility. Besides the effects of pesticides on skin permeability and pathogen susceptibility, specific pesticides such as carbaryl, have also been shown to significantly reduce frog skin peptide levels, but not bioactivity (146, 216).

### UV Radiation

Overexposure of frogs to UV-B radiation, in part due to deforestation and habitat loss, results in damage to the epidermal layer of larval and adult frogs (217, 218). Skin damage is characterized by epidermal shedding and sore formation, causing pronounced detrimental effects to maintenance of skin integrity and to physiological processes such as water and ion transportation (217, 218). Though largely unexplored in frogs, it is suggested that UV radiation breaches the skin barrier and induces host immunosuppression, causing the frog to be more susceptible to both pathogen invasion and exposure to chemical contaminants, leading to host mortality (218). Simultaneous exposure of larval X. laevis to pesticides and UV-B radiation resulted in higher mortality and instances of malformations, including those of the skin (208, 219). The interplay between UV and chemical exposure on frog skin immunocompetence, however, is not well-studied. While extensive research has been conducted in mammalian and fish models to elucidate the impact of irradiation on skin barrier integrity (220, 221), this is largely lacking in amphibian models.

### Pathogens

Much of our understanding of frog skin-pathogen interactions with FV3 and Bd derives from studies using X. laevis as a model (17). FV3 is transmitted through the environment, either through direct contact, indirect contact or consumption of infected carcasses (222, 223) and therefore must cross either the skin epithelial barrier or the gut epithelial barrier. Adult X. laevis are relatively resistant to FV3 and generally recover from mild symptoms 3–4 weeks after infection (18, 224, 225), whereas tadpoles are highly susceptible to FV3 infection (226). While the majority of X. laevis-FV3 research has bypassed the skin barrier via intraperitoneal injection of virus into the host (18, 227–229), water-bath exposure of healthy tadpole and adult X. laevis to FV3-infected frogs in the same tank revealed that healthy individuals become infected with FV3 within 3 h of exposure (230). A key symptom of FV3 infection in susceptible developmental stages or frog species is the formation of skin lesions, skin shedding, and epidermal cell necrosis (231, 232). It is proposed that loss of the skin barrier during FV3 infection allows for increased pathogen entry and ultimately leads to mortality in susceptible hosts, stressing the overall importance of the skin barrier and barrier integrity. While the precise contribution of frog skin innate immunity to FV3 resistance is unclear, initial studies suggest the initiation of a type I interferon response in the skin tissue of adults, compared to a type III interferon response in the skin of susceptible tadpoles, is important in conferring protection against FV3 viral entry and replication, and host mortality outcomes (181, 233).

Infection of susceptible frogs with Bd results in the disruption and cellular death of epidermal layers, resulting in host mortality (77, 234, 235). Comprehensive transcriptomic analyses on the skin of frogs infected with Bd revealed significant transcriptional regulation in the skin with generalized decreases in collagen, fibrinogen, elastin and keratin pathway transcript abundance, which corroborates with the observed disruption in epidermal skin integrity and loss of osmotic balance (236). Furthermore, a generalized lack of gene upregulation for key pro-inflammatory genes was observed, and instead an increase in transcripts for anti-inflammatory markers such as NF-κB inhibitors were seen (236) suggesting Bd may possess immunosuppressive capacity to limit frog skin innate immune defences and activation of underlying immune cells. Overall, these studies somewhat parallel observations in skin from FV3-infected frogs and suggests the loss of skin structural integrity may allow for increased pathogen entry and host mortality.

With the new era of transcriptomics approaches, untargeted transcriptomic molecular approaches have unveiled new insights into the impressive array of physiological functions performed by amphibian mucosal skin epithelium. Recent studies have analysed and compared the transcriptome of 3 anuran families to unveil genes involved in biosynthesis, metabolism, immunity, defence processes, and identification of antimicrobial peptides (123). In addition, transcriptomic studies have been performed on Ranidae and Centrolenidae frogs, species that are largely susceptible to pathogens plaguing amphibian populations, and included skin-specific immune gene expression analysis (125, 237, 238). However, the underlying molecular basis and mechanisms governing resistance and susceptibility of frog species are not well-understood. Further comparisons of frog skin transcriptomes from resistant and susceptible frogs will aid in elucidating the contribution of amphibian skin to resistance against lethal amphibian pathogens.

### MICROBIOME

In mammals, the skin microbiome plays a significant role in the defence against pathogens, injury and infection (239). Recently, attention has turned to elucidating the contribution of the frog skin microbiome in innate immune defences to emerging infectious diseases of amphibians, and in particular to Bd. The frog skin microbiome is seeded by microbes in the external environment (240–242) and shaped by the selective skin microenvironment (241, 243, 244). External contributors to the frog microbiome are the aquatic and soil environments (107) that are believed to serve as a reservoir for the frog skin microbiota (245–248), though horizontal transmission (e.g., during mating) (249), or vertical transmission (i.e., parent to offspring, although not common) (244) are also potential sources. In general, the main bacterial phyla found on frog skin consists mainly of Proteobacteria and Actinobacteria, however, this may vary across frog species, habitat and environmental factors (250–252). Not surprisingly, the frog skin microbiome is influenced by life stage (253), body region (254, 255), diet (254), capture site (256), habitat, captivity (254, 257), exposure to anthropogenic contaminants (258, 259), and treatment with antibiotics (260). While some of these factors may directly influence commensal skin microbes, it is possible that these same factors influence AMP gene expression, secretion of AMPs onto the skin, and AMP bioactivity. Initial studies have shown that the presence of commensal frog skin microbes is important for AMP synthesis (129). Thus, in light of the documented antimicrobial activity of many frog AMPs (**Tables 1**–**4**), the altered levels and activities of frog AMPs on the skin may also contribute to alteration of frog skin microbial communities. Depending on the conditions, skin microbiome dysbiosis may contribute to disease susceptibility in frogs, as observed in other vertebrates (252).

As in other vertebrates (239), symbiotic bacteria on frog skin appear to play an important role in defence against invading pathogens. Investigations of frog skin commensal microbes have revealed certain commensal bacteria to produce metabolites with anti-Bd activity (128, 241, 242, 261, 262). The frog skin commensal bacteria that produce anti-fungal metabolites are documented in the Antifungal Isolates Database (120, 263). Interestingly, metabolites produced by bacteria present on frog skin can also synergize with AMPs on the skin to inhibit Bd (264). Despite the exciting advances in the contribution of frog skin microbial communities to innate immune functions of frog skin, much remains to be elucidated in terms of host-microbiomeenvironment interplay.

### CONCLUDING REMARKS AND FUTURE PERSPECTIVES

Research on the innate immune functions of amphibian skin is emerging and beginning to shift from silos (e.g., investigating skin structure, AMPs or microbiomes) to integrative studies in which multiple facets of skin innate immunity are considered. This approach is critical to elucidating the complex hostpathogen-environment interactions at the skin interface that are participating in amphibian susceptibility to emerging infectious diseases and underpin the global decline in amphibian populations. However, it is evident from the literature that large knowledge gaps exist within each of the skin innate immune barrier silos and in understanding the intricate web of cellular and molecular mechanisms that function to maintain skin homeostasis and rapidly fend against pathogen insult and/or mediate wound healing. We believe there exists an imminent need to unravel the contribution of physical, chemical, cellular and microbiological barriers, to the innate immune function of amphibian skin and the abiotic and biotic environmental factors that regulate skin immunocompetency. Research on the presence and regulation of skin epithelial cell junction proteins under normal and stress conditions would provide vital information on which junction proteins are involved in skin epithelial cell junctions and under what conditions these junction proteins may be controlled to regulate skin permeability. The involvement of the diversity of junction proteins in amphibian skin barrier function is unknown. Little is known of the epithelial cells themselves in terms of the expressions of pattern recognition receptors, the localization of surface receptors (e.g., presence on apical or basal membrane), the role of epithelial cells in the direct sensing of non-self (and distinguishing commensal vs. pathogenic microbes) and in the initiation of innate immune responses leading to the direction of adaptive immune responses. Scrutiny of the literature yielded little information on amphibian PRRs themselves, save for their presence in the frog genome and apparent overall conservation of the signalling pathways as determined by molecular evolutionary analyses. The functional identification of PRR ligands, signalling pathways and downstream gene targets remains untouched. While the identification of amphibian AMPs and the characterization of their antimicrobial activity to human pathogens has been a topic of extensive investigation, comparatively little has been done to examine the antimicrobial activity of frog AMPs on frog pathogens. Virtually nothing is known of their contribution to amphibian skin wound healing or putative innate immune modulation functions, and if present, the receptors through which they bind, the signalling pathways they activate or the gene targets they regulate the expression of. An increasing number of researchers are surveying the commensal microbes present on frog skin, how frog skin microbial communities change with species, life stage, environment and presence of pathogens, yielding insight into the role of these microbes in defending against pathogenic insult. Yet much remains to be uncovered regarding how the frog host creates a permissive niche for certain microbial species while restricting others. Microbemicrobe interactions may also contribute to establishment of "healthy microbiomes" and deeper investigation into the metabolic capacities of commensal microbes will likely yield insight into the maintenance of certain microbial communities. Perhaps further characterization of the skin microbiome may foster the development of deployable "environmental probiotics" to habitats in which threatened or endangered amphibians reside as a way to seed the amphibian skin microbiome, thereby aiding in commensal microbe-mediated defence against frog pathogens. Achieving a complete understanding of skin innate immune function and the factors that affect skin barrier homeostasis may inform environmental policies aimed at conservation of amphibians to mitigate detrimental stressors that alter skin integrity and innate immune competency, or to develop strategies to safeguard threatened amphibians from further disease and population declines.

### AUTHOR CONTRIBUTIONS

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

### FUNDING

Natural Sciences and Engineering Research Council (NSERC) Discovery Grant (RGPIN-2017-04218) awarded to BK. JV and MB-M received financial support in the form of Graduate Teaching Assistantships, Science Graduate Experience Awards, and Science Graduate Student Awards from the Department of Biology and University of Waterloo.

### ACKNOWLEDGMENTS

The authors would like to thank Marie-Claire Wasson for their assistance in finding references for the manuscript and the two reviewers for their thoughtful comments that aided in the improvement of the manuscript.

### REFERENCES


responses. Immunol Lett. (2014) 158:14–24. doi: 10.1016/j.imlet.201 3.11.013


hosts from infection by Batrachochytrium dendrobatidis. Biol Cons. (2012) 152:119–26. doi: 10.1016/j.biocon.2012.03.022


the skin-associated bacterial community. J Herp. (2017) 51:402–10. doi: 10.1670/16-092


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

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

# Vaccine-Induced Protection Against Furunculosis Involves Pre-emptive Priming of Humoral Immunity in Arctic Charr

Laura M. Braden1†, Shona K. Whyte<sup>1</sup> , Alyson B. J. Brown<sup>1</sup> , Carter Van Iderstine<sup>1</sup> , Corinne Letendre<sup>2</sup> , David Groman<sup>1</sup> , Jeff Lewis <sup>1</sup> , Sara L. Purcell <sup>1</sup> , Tiago Hori <sup>3</sup> and Mark D. Fast <sup>1</sup> \*

#### Edited by:

*Stephanie DeWitte-Orr, Wilfrid Laurier University, Canada*

#### Reviewed by:

*Irene Salinas, University of New Mexico, United States Daniel Barreda, University of Alberta, Canada*

> \*Correspondence: *Mark D. Fast mfast@upei.ca*

#### †Present Address:

*Laura M. Braden, AquaBounty Canada, Inc., Souris, PE, Canada*

#### Specialty section:

*This article was submitted to Comparative Immunology, a section of the journal Frontiers in Immunology*

Received: *15 September 2018* Accepted: *15 January 2019* Published: *04 February 2019*

#### Citation:

*Braden LM, Whyte SK, Brown ABJ, Van Iderstine C, Letendre C, Groman D, Lewis J, Purcell SL, Hori T and Fast MD (2019) Vaccine-Induced Protection Against Furunculosis Involves Pre-emptive Priming of Humoral Immunity in Arctic Charr. Front. Immunol. 10:120. doi: 10.3389/fimmu.2019.00120* *<sup>1</sup> Hoplite Laboratory, Department of Pathology and Microbiology, Atlantic Veterinary College, University of Prince Edward Island, Charlottetown, PE, Canada, <sup>2</sup> Department of Veterinary Sciences, Universite de Montreal, Montreal, QC, Canada, <sup>3</sup> Centre for Aquaculture Technologies Canada, Souris, PE, Canada*

With respect to salmonid aquaculture, one of the most important bacterial pathogens due to high mortality and antibiotic usage is the causative agent of typical furunculosis, *Aeromonas salmonicida* spp. *salmonicida* (*Asal*). In Atlantic salmon, *Salmo salar,* the host response during infections with *Asal* is well-documented, with furunculosis outbreaks resulting in significant mortality in commercial settings. However, less is known about the host-pathogen interactions in the emerging aquaculture species, Arctic charr *Salvelinus alpinus.* Furthermore, there is no data on the efficacy or response of this species after vaccination with commonly administered vaccines against furunculosis. To this end, we examined the immunological response of *S. alpinus* during infection with *Asal*, with or without administration of vaccines (Forte Micro®, Forte Micro® + Renogen®, Elanco Animal Health). Artic charr (vaccinated or unvaccinated) were i.p.-injected with a virulent strain of *Asal* (10<sup>6</sup> CFUs/mL) and tissues were collected pre-infection/post-vaccination, 8, and 29 days post-infection. Unvaccinated Arctic charr were susceptible to *Asal* with 72% mortalities observed after 31 days. However, there was 72–82% protection in fish vaccinated with either the single or dual-vaccine, respectively. Protection in vaccinated fish was concordant with significantly higher serum IgM concentrations, and following RNA sequencing and transcriptome assembly, differential expression analysis revealed several patterns and pathways associated with the improved survival of vaccinated fish. Most striking was the dramatically higher basal expression of complement/coagulation factors, acute phase-proteins, and iron hemostasis proteins in pre-challenged, vaccinated fish. Remarkably, following *Asal* infection, this response was abrogated and instead the transcriptome was characterized by a lack of immune-stimulation compared to that of unvaccinated fish. Furthermore, where pathways of actin assembly and FcγR-mediated phagocytosis were significantly differentially regulated in unvaccinated fish, vaccinated fish showed either the opposite regulation (ForteMicro®), or no impact at all (ForteMicro®Renogen®). The present data indicates that vaccine-induced protection against *Asal* relies on the pre-activation and immediate control of humoral immune parameters that is coincident with reduced activation of apoptotic (e.g., NF-κB) and actin-associated pathways.

Keywords: furunculosis, RNAseq, Arctic charr, Aeromonas salmonicida, complement, vaccine, aquaculture

### INTRODUCTION

Outbreaks of disease caused by parasitic, viral, and bacterial pathogens are a critical factor impeding sustainable growth of global finfish aquaculture. Vaccine development for viral and parasitic aquatic pathogens have lagged behind the growth of the industry due to a paucity of information regarding vaccine efficacy, host-pathogen interactions, and host biology (1). In contrast, there are many vaccines currently available in commercial applications for the control of bacterial pathogens [reviewed in (1)]. However, the basic mechanisms involved in the vaccine-associated protection against these pathogens are not well-understood.

Furunculosis is a systemic disease of salmonid and nonsalmonid fishes caused by the Gram-negative bacillus Aeromonas salmonicida subspecies salmonicida (hereafter referred to as Asal). Originally described in 1890 (2), furunculosis has been a significant source of mortality of cultured fish worldwide (3) and is a major cause of antibiotic usage in commercial aquaculture. Furunculosis has several presentations, from chronic infections that result in pathological symptoms including lethargy, darkened skin, loss of appetite, and the development of boils or furuncles on the skin and musculature, to acute infections that are usually associated with juvenile fish and results in rapid septicemia and necrotic lesions of the epidermis (4). This latter form of the disease is accompanied by significant mortality [reviewed by (5)]. Pathological symptoms are due in part to actions of the type III secretion system (T3SS), which provides the bacteria with a mechanism to inject effector proteins into host cells, resulting in immune evasion through inhibition of intracellular killing and phagocytosis (6, 7), as well as by glycerophospholipid cholesterol acyltransferase complexed with lipopolysaccharide (GCAT/LPS) (8, 9).

There are intra- and interspecific differences in susceptibility to Asal among salmonids. For example, chum (Oncorhynchus keta), coho (Oncorhynchus kisutch), and chinook (Oncorhynchus tshawytscha) salmon exhibit variability in furunculosis susceptibility (10); brown trout (Salmo trutta), Atlantic salmon (Salmo salar), and brook trout (Salvelinus fontinalis) are more susceptible to Asal infection in comparison with other species; and, intra-specific resistance to infection exists among different populations of steelhead salmon (Oncorhynchus mykiss), brown trout, and Atlantic salmon (11, 12). Variable genetic host responses have been correlated with improved survival following Asal-infection. For instance, proteomic and transcriptomic analysis of Asal-infected rainbow trout spleen revealed significant induction of iron-regulating proteins (e.g., ferritin), pathogen-recognition receptors (e.g., CD209), and anti-inflammatory cytokines (e.g., IL-13/4) (13). Hepatic transcriptome analysis correlated survival in vaccinated Atlantic salmon with decreased expression of a number of transcripts involved in recruitment and motility of immune cells, including leukocyte cell-derived chemotaxin, annexins, and integrin binding proteins (14). Furthermore, resistant Atlantic salmon have significantly higher haemolytic activity pre-challenge than susceptible salmon, and survival is correlated with Th2-type responses (15). Moreover, there are several examples of higher survival correlated with specific allelic variants or genotypes in Atlantic salmon (16, 17).

Aeromonas salmonicida has also been isolated from Arctic charr (Salvelinus alpinus), and outbreaks of furunculosis are known to occur in farming operations (18). Arctic charr is a salmonid that exhibits the most northern distribution among species in this family and has evolved to tolerate extremely cold temperatures, likely due to elevated plasma electrolyte concentrations and altered epidermal characteristics (19). Due to their remarkable phenotypic and life history variation, Arctic charr have been proposed as the most variable vertebrate species (20). There is an existing commercial fishery for Arctic charr in Canada, and over the last decade, a modest effort to grow Arctic charr commercially resulted in ∼10,000 metric ton production globally. Commercial strains are based on three wild populations—Nauyuk Lake and Tree River strains from Nunavut, and the Fraser River strain from Labrador—which are genetically differentiated from each other (21, 22).

For the last 20 years, incidence of furunculosis has been reduced in food fish using vaccines (23, 24); however, associated protection has been inconsistent, the nature of protection is unclear, and outbreaks of Asal infection persistently occur. It is thought that high individual variation of responses to vaccination in Atlantic salmon, together with high diversity of Asal strains, and limited knowledge of mechanisms of pathogenicity are likely contributors to limited vaccine success. Earlier work demonstrated significant but comparable efficacy of Furogen-2 R (Aqua Health, USA) in protection against furunculosis in two different strains of Arctic charr (24); however, the immune response responsible for associated protection was not determined. There is no comparable report for currently administered vaccines against furunculosis. Thus, improving current vaccines or development of new vaccines are contingent on a more comprehensive understanding of the molecular mechanisms underlying these host-pathogen interactions.

Over the last several years, leveraging significant advances in genetic analysis such as high-throughput RNA sequencing has become a popular avenue for understanding pathogenicity and host responses in aquaculture. For example, there have been several studies utilizing such approaches to assess the host response during Asal infection (15, 25). However, to date, there are no such studies reporting whole transcriptome responses of Arctic charr to Asal. And further, there is no data on the efficacy of vaccines currently available for the control of Asal (i.e., ForteMicro <sup>R</sup> , Elanco Animal Health) in this species. To this end, we performed a comparative analysis on head kidney from vaccinated and unvaccinated Arctic charr during an experimental challenge with Asal using high-throughput mRNA sequencing. Our analysis demonstrates that vaccination significantly improves survival of the Fraser River strain of Arctic charr during infection through marked pre-activation of innate and adaptive humoral immune factors.

### MATERIALS AND METHODS

### Fish Husbandry and Vaccination

All procedures involving the handling and treatment of fish in this study were approved by the University of Prince Edward Island Animal Care and Use Committee prior to initiation and performed under the animal use permit #13-044. Arctic Charr (Salvelinus alpinus) juveniles (n = 1,500, Fraser River strain; >10 g) were obtained from a commercial supplier (Coastal Zone Research Institute) were housed in 1200 L holding tanks at 11 ± 1.0◦C in a flow-through fresh well water system. Fish were fed twice daily to satiation with a commercial feed (EWOS Transfer, St. George, New Brunswick Canada) and maintained on a 14 h light:10 h dark photoperiod. Once all fish reached appropriate size (>20 g) they were sedated with tricaine methanosulfonate (TMS-222; 100 mg/L) and individually tagged with passive integrated transponder (PIT) tags. Prior to initiation of the study, fish were randomly assigned to a treatment group and separated into triplicate tanks (300 m<sup>3</sup> ) per group (n = 30 per tank): Phosphate-buffered saline (PBS)-injected fish (Sham controls), ForteMicro <sup>R</sup> -vaccinated fish (FM-vaccinates), and ForteMicro <sup>R</sup> +Renogen <sup>R</sup> -vaccinated fish (FM+R-vaccinates). Fish were sedated with TMS-222 (100 mg/L) prior to being intraperitoneally (i.p.)-injected with either PBS or vaccine.

### Bacterial Culture and Infection Challenges

U11545-99, a virulent isolate of Aeromonas salmonicida spp. salmonicida (Asal), recovered from Artic charr with typical symptoms of furunculosis in 1999 (Aquatic Diagnostic Services, AVC) was used in this study. A sample from frozen stock was cultured in tryptic soy broth (TSB; BactoTM, Becton, Dickinson and Company, Sparks, MD, USA) at 22◦C with constant shaking at 200 rpm until the cell density reached ∼1 × 10<sup>6</sup> cells mL−<sup>1</sup> based on optical density at 600 nm (OD600). The suspension was centrifuged and resuspended following washing and dilution in PBS.

A pre-vaccine challenge dose titration was performed to determine a challenge dose that would result in ∼60–70% mortality. Ninety (90) unvaccinated, naive fish were separated into three experimental groups, each group consisting of duplicate tanks, holding 15 fish each. Fish were maintained as previously described. Three doses of the bacterium, 10<sup>5</sup> ,10<sup>3</sup> , and 10<sup>1</sup> CFU per 0.1 mL were selected. Fish were anesthetized as previously described and i.p.-injected with 0.1 mL/fish of the appropriate bacterial dose. Following recovery and return to their tank of origin fish were observed at least twice daily for ∼4 weeks.

Vaccinated fish were transferred to study tanks (160 L) by treatment group (n = 30/tank in triplicate); treatment was randomized to tank. Fish were acclimated for 2 weeks before challenge. At 517 degree-days post-vaccination (ddpv), fish were anesthetized (TMS-222; 100 mg/L), challenged i.p. with 100 µl (10<sup>5</sup> CFUs/fish) bacterial suspension or sterile PBS, allowed to recover, and returned to their tank of origin.

In both the dose determination and vaccine efficacy, fish were monitored for morbidity (inability to maintain normal dorsoventral position in the water column, lack of response to stimuli, etc.) and mortalities a minimum of twice daily. Upon detection, mortalities were promptly removed, and infection established by re-isolation of the bacterium from the head kidney on Blood Agar (BA). Plates were incubated at 13◦C and monitored at least once weekly for growth; plates were monitored for 10 days.

### Sampling Design

Sampling was conducted at three time points: (1) prior to infection but after vaccination (517 degree-days post-vaccination [ddpv]), (2) after mortalities were observed in all triplicate tanks of the PBS-injected Sham control (8 days post-infection [dpi]; 605 ddpv), and (3) after cessation of mortalities in all three groups (29 dpi; 836 ddpv) (**Figure 1A**). The experiment was terminated at 31 dpi. At each sampling time fish (n = 4) were opportunistically removed from each triplicate tank (n = 12 fish per treatment per time) and euthanized by overdose in TMS-222 (200 mg/L). Blood was sampled from the caudal vein, allowed to clot at room temperature for 30 min, centrifuged at 4,000 × g/4◦C for 20 min, and resulting plasma was frozen at −80◦C until further use. Sections of the anterior kidney, spleen, gill and proximal intestines were aseptically removed and immediately frozen on dry ice before transferring to −80◦C until RNA or DNA isolation or preserved in 10% neutral buffered formalin (10% NBF).

### Enzyme-Linked Immunosorbent Assay

Antibodies directed against Asal were quantified using a direct ELISA. Briefly, serum samples from PBS-injected, Forte Micro <sup>R</sup> - and Forte Micro <sup>R</sup> +Renogen <sup>R</sup> -vaccinated Arctic charr underwent a serial dilution of pooled sera from each group to determine ideal concentrations for ELISA quantification. PBSinjected fish were measured at 1/128 dilution, whereas vaccinated groups were measured using a 1:5,000 dilution with 1% bovine serum albumin (BSA) in PBS-T (phosphate buffered saline with Tween 20 <sup>R</sup> : 137 mM NaCl; 1 mM KH2PO<sup>4</sup> monobasic; 8 mM Na2HPO<sup>4</sup> dibasic; 3 mM KCl; 0.05% Tween 20 (v/v); pH 7.4) and optimized against the same strain of typical Asal. Flat bottomed 1 × 12 Immulon <sup>R</sup> 2HB ELISA strips (Thermo Labsystems), arranged in 96-well strip-holders were coated with Asal bacteria (grown to Optical Density (OD) 550 = 1.0) and coating buffer (15 mM Na2CO3, 35 mM NaHCO3, pH 9.6), and incubated at 4◦C overnight to allow the antigen to bind. The antigen suspension was removed, and subsequently saturated with 200 µl

ForteMicro®, and dual administration with Renogen® further enhanced protection.

of 3% BSA/PBS-T and incubated at room temperature for 1 h to block the remaining binding sites. Unless otherwise stated, after each step, the plate was incubated for one h at 37◦C and then washed three times with PBS-T. To each well, 100 µl of diluted serum was added and allowed to incubate at 13◦C for 1.5 h. The primary antibody (mouse monoclonal α-trout IgM) was diluted 1:100 with 1% BSA-PBST, and 100 µl was added to each well and allowed to incubate for 1.5 h at 37◦C. The secondary antibody (1:1,000 in 1%BSA-PBST goat α-mouse IgG+M) was then added (100 µl) and incubated at 37◦C for 1 h. Lastly, 100 µl of 2,2′ -azino-bis(3-ethylbenzothiazoline-6-sulphonic acid) solution (ABTS)-H2O<sup>2</sup> substrate was added per well, and incubated in the dark for 30 min at 37◦C. Optical densities were measured at 405 and 490 nm on a SpectraMAX 340 plate reader (Molecular Devices).

### Nucleic Acid Isolation RNA Isolation

Forty-eight samples of head kidney from three time points (0 dpi, 517 ddpv; 8 dpi, 605 ddpv; 29 dpi, 836 ddpv) consisting of three experimental conditions (Sham control, FM-vaccinates, FM+Rvaccinates; n = 5–7 individuals per condition) were selected for library construction, RNA-seq analysis, and RT-qPCR analysis.

RNA was extracted from frozen head kidney samples using Tri-Reagent as previously described (26), and genomic DNA contamination was eliminated after DNase treatment (Ambion <sup>R</sup>

TURBO DNA-freeTM). Quality of resulting purified RNA was determined using ExperionTM Automated Electrophoresis Station (BioRad) and Nanodrop 3,000 (Thermo Fisher) was used to test both purity and quantity. RNA was stored at −80◦C until subsequent use.

### DNA Isolation

DNA was isolated from 81 individual head kidney tissues (3 fish per group pre-challenge, and 4 fish per tank [triplicate tanks × 3 treatment groups × 2 post-challenge samplings]) using a Qiagen <sup>R</sup> DNeasy Blood and Tissue Kit following manufacturer's instructions with the following changes: samples (∼25 mg) were homogenized for 20 min at 50 Hz (TissueLyser, Qiagen) and incubated at 56◦C for 4 h. Resulting purified DNA was quantified spectrophotometrically (Nanodrop 3000) and diluted with nuclease-free water to 50 ng/ml. Integrity of gDNA was assessed on a 1% agarose gel before downstream enzymatic assays.

### Bacteriology, Histopathology, and Immunohistochemistry (IHC) Bacteriology

Posterior kidneys were cultured from fish collected on designated sampling days and mortalities occurring throughout the trial by aseptically plating culture swabs on BA media at 13◦C. The plates were checked at least once weekly for bacterial growth. The presence of bacterial growth, and the characteristic brown colonies on tyrosine-rich media was considered a positive criterion for typical Asal. Plates showing uncharacteristic growth (possible contamination) or no growth were considered negative for the presence of the bacteria.

### Histopathology and IHC

Serial sections of paraffin-embedded formalin-fixed tissues were sectioned at 7µm and either stained with haematoxylin and eosin (H&E) or Wright's Giemsa following standard histological techniques or probed with anti-trout IgM monoclonal antibodies using the Expose Mouse-Rabbit Specific HRP/DAB kit (Abcam) following manufacturer's instructions. Briefly, after deparaffinization and rehydration, sections were heated to 100◦C in antigen retrieval buffer (pH 9) cooled to room temperature for 10 min in phosphate-buffered saline (PBS), and then washed twice in Tris-buffered saline plus 0.2% Tween 20 (TBS-T; pH 8.0) for 5 min with gentle agitation. Sections were blocked in protein blocker for 10 min before gentle rinsing with TBS-T. The sections were incubated with α-trout mIgM (1:50) in TBS-T and 1% bovine serum albumin (BSA; Sigma) overnight at 4◦C in a humid chamber. After incubation, the sections were washed in TBS-T (two times for 5 min each) and incubated in a mouse-specifying reagent (Expose mouse/rabbit specific horseradish peroxidase [HRP]/3,3=-diaminobenzidine [DAB] kit; Abcam) for 10 min, followed by a 10-min incubation in hydrogen peroxide blocker. Labeled cells were detected after a 15-min incubation in a goat anti-rabbit HRP conjugate followed by 10 min with DAB in PBS with 0.015% H2O2. All sections were counterstained in 1% Alcian blue (3 min) and Mayer's hematoxylin (diluted 1/20, 30 s), dehydrated in graded ethanol, cleared in xylene, and mounted (Permount). Sections treated with irrelevant antibodies served as negative controls, while sections known to contain IgM-labeled cells served as positive controls. After staining or immunolabeling, sections were visualized with a ZEISS AxioCam IC.

### RNA-Sequencing

### Sequencing and de novo Library Construction

Total RNA was submitted for library construction and sequencing services by the McGill University and Génome Québec Innovation Center, Montréal, Canada.

RNA quality and purity were assessed using an Experion Bioanalyzer (BioRad), and only samples with a minimum RIN of 6 proceeded to library construction. Forty-eight TRUseq 100 bp pair-end stranded mRNA libraries were generated from the experimental conditions listed in **Supplemental Table 1** and were sequenced in 4 lanes (12 samples per lane) on an Illumina HiSeq 2000 platform. De novo transcriptome assembly was performed on resulting reads following the pipeline described by Haas et al. (27) based on the Trinity assembly software suite v 2.1 (28). Briefly, reads were trimmed with Trimmomatic software (29) with a minimal Phred score of 33 and a minimal length of 32 bp. A normalized metric of reads was generated using Trinity normalization utility and surviving paired reads were assembled using the Trinity assembler (27). The final assembly quality was checked using R-correct and contigs with poor read support were removed. Trinotate v2.0.2 was used to identify putative coding transcripts and all putative transcripts were aligned against the UniProt protein database (downloaded March 2016) using the blastx program from the NCBI BLAST family as implemented in the blast+ package. Annotation was assigned to each longest putative coding transcript derived from each de Bruijn graph component (unigene) based on highest blastx score with an Expected value (E) cut-off <1e-10.

### Mapping of Sequence Reads and Differential Expression Analysis

Read alignment to the reference transcriptome was performed using RNA-Seq by Expectation Maximization Method (RSEM) described by Li et al. (30) using the de novo assembled transcriptome as a reference. This method was selected as it allows for the inclusion of non-uniquely mapping reads in abundance estimates through the use of a likelihood method that accounts for read mapping uncertainty.

RSEM quantities were used as the input for the programs edgeR (31) and DESeq2 (32) using scripts provided with the Trinity pipeline and a list of DE transcripts representing the mean expression values from samples for every treatment and time combination. Lists of DE transcripts generated using DESeq2 were generally larger and were almost always contained within the edgeR DE lists, and for this reason we chose to use the more conservative edgeR-generated DE lists going forward with analysis. Differential expression was tested at a significance level of α = 0.05. Transcripts were considered differentially expressed at a p < 0.05, and log2fold-change > 2, following a false discovery rate (FDR) adjustment of 5% (0.05) as implemented by the BH algorithm in the p.adjust function of the R stats package.

### Gene Ontology

Functional gene ontology was conducted using GOrilla (33) by comparing groups of differentially expressed genes to the de novo reference assembly as the background with an FDR-corrected p < 0.001. Using this open-access software, enriched GO terms are identified from ranked lists of DEGs as well as from target lists of DEGs compared to a background. Resulting enriched lists were summarized and visualized using REViGO (34) and Cytoscape.

### Conventional and Quantitative PCR

### RNAseq Validation

To validate sequencing data, gene-specific primers were designed using Primer3 (35) and the contig sequences for nine significantly expressed transcripts, and analyzed through qPCR (**Supplemental Table 2**). Total RNA was extracted from samples of head kidney as described above. For every sample, 750 ng of DNase-treated total RNA was reverse transcribed using a High Capacity cDNA Reverse Transcription Kit (Thermo Fisher) following manufacturer's instructions in a reaction volume of 20 µl. qPCR was performed in triplicate reactions containing 2X Universal SYBR Green Master Mix (BioRad), 100 nm of each forward and reverse primer, and 2 µl template (diluted 1/20) in a total reaction volume of 11 µl. Cycling was performed on a CFX96 thermal cycler (BioRad) with the following thermal profile: 95◦C for 30 s (1 cycle), 95◦C for 15 s then 60◦C for 30 s (40 cycles), followed by melt curve analyses from 65 to 95◦C with fluorescence being read every 0.5 s with a ramp rate of 0.5◦C to ensure amplification specificity. An equal volume of every cDNA sample was pooled and diluted 10-fold in fivesteps to determine amplification efficiency and linearity for every primer pair. To confirm an absence of gDNA contamination, noRTs (no reverse transcriptase control) were performed for each gene.

Target gene expression was normalized to the two most stable of nine reference genes: 60S ribosomal protein L7 (60S), beta-2-microglobulin (β2m), elongation factor 1-alpha (efIα), glyceraldehyde-3-phosphate dehydrogenase (gapdh), hypoxanthine-guanine phosphoribosyltransferase (hprt), ribosomal protein S20 (rps20), beta-actin (ß-actin), eukaryotic initiation factor 5A(eif5), and tubulin alpha chain (tubulin-α). Elongation factor 1-alpha and rps20 showed the highest stability, with geNorm M value and coefficient of variation of 0.993 and 0.37, respectively, using the geNorm algorithm found in qBASE<sup>+</sup> software (36). Calibrated normalized relative quantities (CNRQ) were calculated with sample-specific normalization factors and gene-specific amplification efficiencies, and internal positive controls were included to calibrate run-to-run variation among plates.

Amplicons for each gene were gel purified using QIAquick Gel Extraction Kit (Qiagen) as per the manufacturer's instructions. Purified products were sent to The Center for Applied Genomics (TCAG) at the Hospital for Sick Children (Toronto, Canada) for primer verification (**Supplemental Table 3**).

### Quantification of Aeromonas salmonicida

Specific primers for the Asal A449 aopO gene located in the low-copy-number pAsa5 plasmid (GenBank accession no. DQ386862.1) have been published previously (37). Extracted DNA from posterior kidneys of moribund and sampled fish was used to confirm Asal infection by conventional and realtime PCR. PCR products were synthesized using 12.5 µl GoTaq (ProMega), 2 µl of forward aopO primer (5′ -AGCTCATCCAAT GTTCGGTATT-3′ ), 2 µl of reverse aopO primer (5′ -AAGTTC ATCGTGCTGTTCCA-3′ ), 5 µl of DNA template (50 ng/µl) and 3.5 µl of nuclease-free water. The samples were run on an Eppendorf MasterCycler thermal cycler with the following thermal profile: initial heating at 95◦C for 2 min, denaturation at 95◦C for 1 min, annealing at 60◦C for 30 s, extension at 72◦C for 30 s, repeat for 34 cycles, final extension at 72◦C for 5 min, and hold at 4◦C. DNA from pure cultures of Asal was included as a positive control, whereas reactions lacking DNA template were used as negative controls. To confirm if samples were positive or negative for the aopO gene, PCR products were separated on a 2% agarose gel, and visualized using an ultraviolet light transilluminator (UV Transilluminator 2000, Bio Rad), with the presence of a 119-bp band considered as a positive result.

Real-time PCR was performed on a CFX96 thermal cycler (BioRad) and the CFX ManagerTM software detection system (BioRad). The reaction mix contained 12.5 µl of 2X Universal SYBR Green Master Mix, 400 nM of each primer, 2 µl of template, and nuclease-free water to reach a final volume of 25 µl. The thermal cycling profile was comprised of an initial incubation of 95◦C for 30 s, 35 cycles of 95◦C for 15 s then 62◦C for 30 s. Melt curve analysis was performed to ensure amplification specificity using a temperature gradient from 65 to 95◦C with fluorescence being read every 0.5 s with a ramp rate of 0.5◦C/s. Confirmed positive template were used as a positive control, while DNA extracted from fish prior to Asal-challenge were used as negative controls. All real-time PCR assays were performed in triplicate, and an NTC was included on every plate.

## Data Analysis

### Survival Probabilities

The Kaplan-Meier (KM) survival curves were calculated in R using the survival package (R Core Developmental Team, V3.4.2).

### Relative Proportional Survival

The relative proportion survival (RPS) was calculated for the average of the replicates for each combination of treatment factors following the formula (38):

$$RPS = \left[1 - \left(\frac{a}{b}\right)\right] \times 100\%,$$

where a = cumulative mortalities in vaccinates

b = cumulative mortalities in non-vaccinates

ELISA data were analyzed using a two-way ANOVA followed by Tukey HSD. Correlations between mortality and antibody titer were detected using the Pearson method and performed in R (V3.4.2).



*Values represent the average of three replicate tanks per treatment* ± *standard deviation of the mean.*

### RESULTS

### Arctic Charr Are Susceptible to Aeromonas salmonicida by i.p.-injection

The challenge with A. salmonicida spp. salmonicida (Asal) resulted in widespread mortality of Sham controls by 8 days post-infection (dpi). At 29 dpi, mortalities were observed in all groups; however, onset of mortality was delayed in both the FM– and FM+R-vaccinated groups, with the latter displaying signs of increased protection as evidenced by a 12% increase in survival probability (**Figure 1B**). Dying fish presented typical signs of acute furunculosis including apathetic behavior, external hemorrhage, ascites accumulation in the peritoneal cavity, hemorrhage of the liver and body wall, hemorrhage of the stomach and pyloric caeca, enlarged spleen and liver, and swollen intestine. At the study end-date (31 dpi), there were 28.8% survivors in the Sham control group, 72.2% survivors in the FM-vaccinated group, and 83.3% survivors in the FM+R-vaccinated group (**Table 1**). Relative Percentage Survival (RPS) was calculated for each vaccine treatment at each time point. ForteMicro <sup>R</sup> resulted in 69.2% protection at 8 dpi and 58.3% protection at 29 dpi, while the combination of ForteMicro <sup>R</sup> +Renogen <sup>R</sup> resulted in 92.3 and 75% protection at 8 and 29 dpi, respectively (**Table 1**).

### Detection of Asal

To establish presence of Asal in sampled tissues, we incubated posterior kidney swabs on blood agar plates, as well as using conventional and real-time PCR to quantify the apoO gene as previously described (37). The kidney of all naïve Arctic charr became infected with Asal following i.p.-injection as determined by bacteriology, PCR, and qPCR (**Figure 2A**; **Supplemental Table 4**). Interestingly, there was incongruence in the results of the diagnostic tests, with the molecular assays appearing to be more sensitive in detecting presence of the bacteria. For example, while there was no bacterial growth detected in FM-vaccinates at 8 dpi, PCR indicated 9 of 12 fish were positive. However, quantification by qPCR revealed a significant upregulation of apoO in the Sham control group, but there was no difference in either FM– or FM+R-vaccinates at 8 dpi compared to 0 dpi (**Figure 2**). At 29 dpi, there was no bacterial growth detected in any group; however, PCR determined that all three groups were positive for Asal, while qPCR showed a significant downregulation of apoO in Sham controls. As expected, there were more positive fish in the unvaccinated Sham group at 8 dpi (83.3%), as confirmed by PCR, while the FM– and FM+R-vaccinated groups were 75 and 50% positive for Asal by PCR, respectively. At 29 dpi, only 50% of non-vaccinated controls were positive by PCR, while 75% of FMvaccinates were positive. Only 16.7% of FM+R-vaccinates tested positive for Asal at 29 dpi.

### Humoral and Cellular Response

The results from the ELISA indicated a significant positive correlation between vaccination and the presence of Asalspecific antibodies (Pearson's r = 0.998, p = 0.031; **Figure 2B**). There were significantly higher concentrations of Asal-specific antibodies in both groups of vaccinated Arctic charr compared to Sham controls at all time-points in the experiment (8 dpi → 29 dpi). There was an increase in antibody concentration over time (0 dpi → 29 dpi) in vaccinated fish, however, this was only significant in the dual-vaccine group (p = 0.020). We did not detect a significant difference between vaccine groups.

Sections of gill, intestine, kidney, and spleen from FM– and FM+R-vaccinated and PBS-injected Asal-infected Arctic charr were stained with H&E and Wright's Giemsa. At 8 dpi, we observed evidence of pathological changes in Sham controls in all tissues examined, with bacteria present in gill, kidney, and spleen, presenting as large distinct colonies or disseminated throughout the tissue section (**Figure 3**). We detected several large lymphoid cells in blood vessels that appeared to harbor intracellular Asal bacteria (**Figures 3C,D**).

Sacciform cells were observed in the gills of both vaccinated (FM– and FM+R) and unvaccinated charr, which were present as strong eosinophilic-staining cells, often associated with a blebbing from the gill epithelium (**Figures 4A–C**). In Asalinfected fish, we observed pathological changes (e.g., filament clubbing, hypertrophy, fusion) in the gills, however, there was no apparent association between sacciform cells and gill pathology and/or the presence of Asal colonies. Eosinophilic granular cells (ECGs) were also observed in the gills of vaccinated Asal-infected charr (**Figures 4D–G**). These cells were more prominent in areas of lamellar thickening/fusion, in the intrabranchial lymphoid tissue (ILT) or in the gill epithelium. We were unable to quantify EGCs due to a low number of histological replicates; however, they were a conspicuous feature of FM- and FM+R-vaccinated charr and appeared to increase in number from 8 to 29 dpi, while in unvaccinated charr we failed to detect EGCs in gill tissue at any time. Eosinophilic granular cells were also observed in the intestine of Asal-infected charr in high numbers, both in vaccinated and unvaccinated groups (data not shown).

At 8 dpi, in the spleen of FM- and FM+R-vaccinated and Sham control groups, numerous melanomacrophages (MMs) were observed scattered throughout the red and white pulp (**Figure 5**). Melanomacrophages were present in arterioles and surrounding ellipsoids, and there was also evidence of apoptotic cells (pyknotic nuclei) (**Figure 3I**). In FM- and FM+R-vaccinated charr specifically, MMs were often associated with aggregates of IgM<sup>+</sup> lymphocytes clustered around ellipsoids or inside

FIGURE 2 | (A) Transcript abundance of *Asal apoO* gene was quantified over time in head kidney of vaccinated and unvaccinated Arctic charr prior to and after i.p.-injection (8 and 29 dpi). There was significantly higher expression in PBS-injected controls compared to either vaccine group at 8 dpi. At 29 dpi, expression decreased in controls to the same levels as vaccinated fish. Scatterplots are showing all samples with the mean denoted by the hashmark (*n* = 12) and the error bars indicating the standard deviation. Significance was determined by 2-way ANOVA followed by *post-hoc* Tukey's HSD (*p* < 0.05). Lower-cased letters denote significant differences among groups at each time point, while asterisk denotes differences over time within the same group. (B) Concentration of *Asal*-specific antibodies in charr plasma were measured by ELISA. Data shown are from 1:125 (PBS-control) and 1:5,000 (FM- and FM+R-vaccinate groups) dilution of serum samples (*n* = 9, per group). Statistical significance was determined using a two-way ANOVA followed by Tukey *post-hoc* HSD (*P* < 0.05). There was a significant difference between control and either FM- or FM+R-vaccinated charr at all-time points (*p* < 0.001).

arterioles, with higher numbers of IgM<sup>+</sup> cells in the spleen of FM-vaccinates (**Figures 5B,D**). At 29 dpi, the spleen of FM+Rvaccinated charr was densely populated with MMs, with a smaller number of IgM<sup>+</sup> clusters (**Figure 5G**). In contrast, at 29 dpi, MMs in the spleen of FM-vaccinates were lower in number and there were fewer IgM<sup>+</sup> lymphocytes detected (**Figure 5H**). We did observe IgM<sup>+</sup> lymphocytes in the spleen of Sham controls, however, these were very sparse and did not appear to associate with MMs or with colonies of Asal (**Figure 5**). In contrast, the kidney of Sham controls was populated with IgM<sup>+</sup> cells that were associated with MMs (**Figure 5**).

Kidney of both Sham controls and vaccinated charr were densely populated with MMs, and in controls we observed areas of necrosis associated with bacterial colonies in the Sham group (**Figures 3E,F**).

Sham control intestine was characterized by numerous cells with pyknotic nuclei, mucocyte hyperplasia, hemorrhage and necrosis of the lamina propria, all of which were absent in FM- and FM+R-vaccinated fish intestine (**Supplemental Figures 1A,B**). Positively immunolabeled cells were observed in the lamina propria and migrating throughout intestinal mucosa (**Supplemental Figure 1C**).

### RNA Sequencing

### Raw Sequencing Data and Quality Statistics

An average of 15.7 million reads per library was generated, with 73% of all reads mapping to the reference transcriptome (**Supplemental Table 5**). All contigs from the de novo assembly were analyzed using Transdecoder to identify the best putative ORF. Resulting protein sequences were annotated using the Trinotate v2.0.2 using the uniport database retrieved on March 2016.

### Transcriptomic Response of Arctic Charr to A. salmonicida

RNA-seq libraries were generated from 48 samples of Arctic charr arbitrarily chosen from 9 discrete treatment conditions (**Supplemental Table 1**). Raw sequence data from all libraries has been deposited in the NCBI Sequence Read Archive as accession PRJNA507334.

Transcriptome de novo assembly yielded a total of 664,663 contigs with an average transcript length of 786 bp. These sequences represent 423,594 trinity genes, of which, ∼181,321 (42.8%) were suggested as being protein-coding following Trinotate and Uniprot blastx annotation analysis (**Supplemental Data File 1**). Of these, 69,019 were annotated by UniProt. Abundance estimates (RSEM expected counts) for all transcripts identified in each RNA-seq library is provided in **Supplemental Data File 2**. To provide concordance in differential expression we first used two different pipelines (edgeR, DESeq2). Lists of DE transcripts generated from DESeq2 were generally larger and fully encompassed edgeR DE transcript lists (**Supplemental Figure 2**). We decided to consider data obtained by the more conservative edgeR pipeline for subsequent analysis. Lists of differentially expressed transcripts identified from abundance estimates using edgeR are provided in **Supplemental Data File 3**.

The transcriptomic profile of both groups of vaccinated Arctic charr followed a similar pattern to that of Sham controls, characterized by low numbers of differentially regulated transcripts prior to bacterial challenge, followed by an increase

in over-expressed transcripts during the bacterial infection, ending with a decrease in over-expressed transcripts in postchallenge, surviving fish. However, in contrast to Sham controls, the degree and magnitude of the response in vaccinated fish was significantly reduced. Furthermore, the transcriptomic response was observed to differ depending on the vaccine administered with a much stronger transcriptomic response in FM-vaccinates compared to FM+R-vaccinates (**Figure 6**; **Supplemental Tables 6, 7**).

### Vaccination Induces a Primed Humoral Immune Response

To evaluate the effect of vaccination on the transcriptome, we first compared the transcriptomic profiles of vaccinated (FM or FM+R) to unvaccinated (Sham) fish prior to infection with Asal (517 ddpv). There were 145 and 160 differential expressed genes (DEGs) in fish vaccinated with FM+R and FM, respectively, with 19 transcripts concordantly overexpressed by both vaccinate groups (**Table 2**). Three of these transcripts (40S ribosomal protein S12, ubiquitin carboxyl-terminal hydrolase 3, and 7SK snRNA methylphosphate capping enzyme) were expressed in opposing directions (i.e., up-regulated in one group while downregulated in the other).

Compared to Sham controls, the transcriptomic response of FM-vaccinates at 517 ddpv was characterized mainly by over-expression of genes involved in the acute phase response (e.g., alpha-2-HS-glycoprotein, serum albumin 1, ladderlectin), complement and coagulation cascade (e.g., complement factor H-related protein 1, complement C1r subcomponent, complement C3, alpha-2-macroglobulin; **Figure 7**), and iron homeostasis (e.g., hemopexin, serotransferrin-2, ceruloplasmin). In contrast, FM+R-vaccinates were characterized by up-regulation of genes involved in protein synthesis (e.g., 60S ribosomal protein L13a, 40S ribosomal protein S12), cellular transport (e.g., choline transporter-like protein 4, phospholipid-transporting ATPase ID) and transcription (e.g., eukaryotic translation initiation factor 3 subunit L, nuclear receptor coactivator 4, CREB-regulated transcription coactivator 2).

There was a high number of over-expressed transcripts associated with complement and coagulation in the FMvaccinated charr (n = 13), compared to the group receiving the combined vaccination of FM+R (n = 3; **Supplemental Table 8**). Additionally, we only detected significant up-regulation of transcripts involved in the acute-phase response in FMvaccinates prior to infection with Asal, such as alpha-2-HSglycoprotein (Fetuin-A, FC = 17.3), serum albumin 1 (FC = 13.6), ladderlectin (FC = 13.1), apolipoprotein B-100 (FC = 12.5), and apolipoprotein A-I-1 (FC = 12.4).

### Transcriptomic Response During Asal Infection in Vaccinated Charr

Compared to PBS-injected controls, at 8 dpi/605 ddpv, the number of differentially expressed transcripts in FM-vaccinated fish was greater than the response in FM+R-vaccinated fish (1,163 and 145, respectively; **Figure 6C**, **Supplemental Table 7**). Of these, only 52 were concordantly over-expressed in both groups, including tissue remodeling enzymes (matrix metalloproteinase-9), hemostasis-associated genes (hemebinding protein 2, hemoglobin subunit alpha), mitochondrial enzymes ([3-methyl-2-oxobutanoate dehydrogenase [lipoamide]] kinase), cytoskeleton-associated transcripts (keratin type I cytoskeletal 18, WAS/WASL-interacting protein family member 1), and actin-binding enzymes (gelsolin, dematin) (**Supplemental Table 9**).

Transcripts specifically upregulated in FM-vaccinated fish were characterized by enrichment in "RNA metabolic process" (GO:0007275), "regulation of gene expression" (GO:0010468), and "multicellular organism development" (GO:0035987). In contrast, in FM+R-vaccinated fish there was enrichment of "granulocyte chemotaxis" (GO:0071621), and "dynamin family protein polymerization involved in mitochondrial fission" (GO:0003374) (**Figure 8**; **Supplemental Data File 4**).

There was no enrichment in down-regulated transcripts in FM-vaccinated Arctic charr, however, in FM+R-vaccinates there was enrichment in "respiratory electron transport chain" (GO:0022904), and "regulation of transcription of nucleolar large rRNA by RNA polymerase I" (GO:1901836).

We observed over-expression of several adaptive immunityassociated transcripts in immunized fish at 8 dpi, with a higher number of these transcripts over-expressed in FM-vaccinates. These included BCL-6 corepressor-like protein 1 (FC = 8.54), Ig heavy chain V region (FC = 5.44), Ig heavy chain V region BCL1 (FC = 8.32), and Ig kappa chain V-I region DEE (FC = 7.44). One transcript associated with adaptive immunity (Ig lambda-3 chain C region) was concordantly over-expressed as multiple isoforms in both FM- and FM+R-vaccinated fish (FC = 6.52–8.36 and FC = 3.6–8.95, respectively).

Interestingly, compared to unvaccinated controls, there was significant upregulation of a hemolytic protein, neoverrucotoxin subunit alpha, only in the FM-vaccinated group (FC = 8.9–10.3).

### Transcriptomic Response in Surviving Vaccinated Fish

At 29 dpi/836 ddpv, the number of over-expressed transcripts in the FM-vaccinates decreased to a comparable level to that of FM+R-vaccinates (129 and 90, respectively; **Figure 6C**, **Supplemental Table 7**). Fifteen transcripts were concordantly over-expressed between vaccinated groups, with functions in RNA-binding (RNA-binding protein 10), DNA-binding (coiledcoil and C2 domain-containing protein 1A), DNA-repair enzymes (tyrosyl-DNA phosphodiesterase 2), transferases (probable E3 ubiquitin-protein ligase HERC1), helicases (fanconi anemia group J protein homolog), nuclear protein transport (importin-7), microtubule organization (centrosomal protein of 170 kDa), signal transduction (leucine-rich repeat-containing protein 28), and cytoskeletal organization (bromodomain and WD repeatcontaining protein 3) (**Table 3**). Within these genes, three were differentially expressed between vaccine groups (transposon TX1 uncharacterized 149 kDa protein, synergin gamma, and membralin).

Transcripts specifically up-regulated in FM-vaccinated fish included those involved in DNA repair (e.g., DNA cross-link repair 1A protein, fanconi anemia group J protein homolog, tyrosyl-DNA phosphodiesterase 2), protein phosphatase/kinase activity (e.g., protein phosphatase

1H, calmodulin), transcriptional regulation (e.g., coiledcoil and C2 domain-containing protein 1A, breast cancer metastasis-suppressor 1-like protein-A, polyhomeotic-like protein 1), cell cycle regulation (e.g., aurora kinase), and cellular trafficking (e.g., probable E3 ubiquitin-protein ligase HERC1, importin-7). GO analysis detected enrichment in processes such as "negative regulation of phospholipase activity" (GO:0010519), and "epithelial fluid transport" (GO:0042045) (**Figure 8**; **Supplemental Data File 4**).

The transcriptional profile of FM+R-vaccinates also included many genes involved in the same pathways; however, in addition, we observed significant up-regulation in cellular respiration (e.g., NADH-ubiquinone oxidoreductase 75 kDa subunit, mitochondrial, adenylosuccinate lyase, glucokinase, COX assembly mitochondrial protein 2 homolog), adaptive immune regulation (e.g., butyrophilin subfamily 1 member A1, nuclear factor of activated T-cells 5), and proinflammatory regulation (e.g., leukotriene A-4 hydrolase).

There was significant downregulation at 29 dpi in FM-vaccinates, with enrichment in processes such as "protein activation cascade" (GO:0072376), "acutephase response" (GO:0006953), "platelet degranulation"



*Multiple ranges indicate transcript isoforms.* \**Differential expression between vaccine groups.*

(GO:0002576), and "complement activation" (GO:0006956) (**Supplemental Data File 4**). In contrast, down-regulated genes in FM+R-vaccinates were only enriched for one biological process, "cellular senescence" (GO:0090398).

### Infection With Asal Results in Massive Over-expression in Non-vaccinated Arctic Charr

There were 1,334 transcripts up- and 511 transcripts downregulated in Asal-infected Sham controls at 8 dpi compared to uninfected Sham controls. Functional Gene Ontology (GO) of upregulated transcripts revealed enrichment of biological processes involved in a number of physiological functions including "cytokine-mediated signaling pathway" (GO:0019221), "regulation of cell proliferation" (GO:0042127), and "response to lipopolysaccharide" (GO:0071222) (**Figure 8**; **Supplemental Data File 3**). In contrast, enrichment in downregulated transcripts included processes such as "secretion" (GO:0046903), "exocytosis" (GO:0006887), and "regulation of reactive oxygen species metabolic process" (GO:20000377).

As the complement and coagulation cascade was highly upregulated in vaccinated fish prior to Asal-challenge, we were interested in comparing the expression profiles of these transcripts in unvaccinated fish over the experimental time period. Interestingly, there was high concordance in expression of many transcripts between non-vaccinated surviving charr and infected vaccinated charr (**Table 4**).

Compared to uninfected controls, the transcriptomic response decreased in surviving unvaccinated charr at 29 dpi, with 123 up-regulated transcripts and 130 down-regulated transcripts. Gene ontology analysis of up-regulated transcripts indicated enrichment in biological processes involved in cell-cycle process and proliferation, including "cell cycle phase transition" (GO:0044770), "mitotic cell cycle process" (GO:1903047), and "meiotic spindle organization" (GO:0000212) (**Supplemental Data File 4**). Enrichment of down-regulated transcripts resulted in one biological process of "negative regulation of erythrocyte differentiation" (GO:0045647) (**Supplemental Data File 4**).

### Comparison Between ForteMicro®- and ForteMicro®+Renogen®-Vaccinated Charr

We then compared the transcriptomic response between the two vaccine-groups in attempts to identify differences in the host response that might account for the increased protection in the FM+R-vaccinates.

Compared to Sham controls prior to bacterial challenge (517 ddpv), there were 125 over-expressed transcripts, with 47 up-regulated in FM-vaccinated charr compared to 78 up-regulated in the FM+R-vaccinated group. Gene Ontology analysis of up-regulated transcripts in FM-vaccinates revealed enrichment in biological processes such as "acute phase response" (GO:0006953), "regulation of proteolysis" (GO:0030162), and "platelet degranulation" (GO:0002576). In contrast, FM+R-vaccinates showed no enrichment of GO terms. Enrichment of downregulated transcripts at 517 ddpv in FMvaccinates revealed enrichment in processes including "positive regulation of lamellipodium organization" (GO:1902743), and "cellular response to biotic stimulus" (GO:0071216). In FM+R-vaccinates there was one category enriched in downregulated transcripts, "negative regulation of chromatin binding" (GO:0035562).

During the proliferative growth stage of Asal infection (8 dpi, 605 ddpv), 108 transcripts were over-expressed, with 62 up-regulated in FM-vaccinated fish, compared to 46 upregulated transcripts in the FM+R-vaccinated fish. Functional GO of upregulated transcripts revealed FM-vaccinates had enrichment in "defense response" (GO:0006952), "positive regulation of establishment of T cell polarity" (GO:1903905), and "inflammatory response" (GO:0006954). In contrast, FM+Rvaccinates had enrichment of "androgen receptor signaling pathway" (GO:0030521), "positive regulation of actin nucleation" (GO0051127) (**Supplemental Data File 4**). In contrast, GO enrichment of downregulated transcripts of FM-vaccinated fish included processes of "chemotaxis" (GO:0006935), "regulation of signaling" (GO:0023051), and "leukocyte migration" (GO:0050900). In FM+R-vaccinates, there was much fewer enriched categories in downregulated transcripts, including "positive regulation of phagocytosis" (GO:0050766), "positive regulation of endocytosis" (GO:0045807), and "PERK-mediated unfolded protein response" (GO:0036499).

pathway while in PBS-injected controls (orange symbol), these genes were up-regulated.

The response in surviving vaccinated fish (29 dpi, 836 ddpv) reduced to 85 over-expressed transcripts comprising 39 upregulated in FM-vaccinates, and 46 up-regulated transcripts in FM+R-vaccinates. In FM-vaccinated fish, there was enrichment in upregulated transcripts for processes like "histone H3-K9 modification" (GO:0006413:000061647), "regulation of catabolic process" (GO:0009894), and "peptidyl-lysine monomethylation" (GO:0018026). There was no enrichment of GO terms FM+Rvaccinates in upregulated transcripts, and there was no enrichment in downregulated transcripts for either FM- or FM+R-vaccine group.

Comparison between overexpressed transcripts during the experimental time course showed that there was a similar profile between 517 ddpv and 8 dpi (∼86% similarity). There were 16 transcripts specific to the pre-challenge FM-vaccinated group with significant activation in several humoral immune effectors (**Table 5**). In contrast, the profile of overexpressed transcripts in surviving vaccinates was discrete.

frequency are shown.

### RNAseq Validation and Exploratory Gene Expression by qPCR

Nine genes of interest (B-cell linker protein, mx2, il6rb, c7, ladd, fibronectin, fibrinogen gamma chain, hsp90b) were chosen for validation with RT-qPCR on the RNA-seq samples, as well as on an expanded number of samples to increase the power of testing. Some of these genes were chosen to cross validate the DE results from RNA-sequencing, and others were of immunological importance. Fold-change comparisons between both methods revealed a significant positive correlation (Pearson's r = 0.60 p = 0.024) (**Supplemental Data File 5**).

We then correlated the expression of these immune associated genes with the expression of apoO from Asal. There was significant negative correlation in both blnk and finc, indicating that presence of Asal negatively affected expression of these genes (**Supplemental Table 10**).

Acute phase response genes ladderlectin, complement component 3, and complement component 7 were expressed higher in FM-vaccinates than in PBS-injected controls as identified by RNA-seq (**Figure 9A**). Genes involved in immunological processes such as B-cell differentiation and cytokine signaling were also confirmed to be differentially regulated through vaccination. ForteMicro <sup>R</sup> -immunized fish had a significantly higher expression of finc and blnk compared to Sham controls after infection with Asal (8 dpi, 605 ddpv; **Figure 9B**), with expression increasing in the Sham and FM+R group by the final sampling time. Other genes examined using RT-qPCR, as mentioned above, had concordant expression with the RNA-seq data, but were not all significant over time and treatment comparisons (**Supplemental Figure 3**).

## DISCUSSION

Transcriptomic analyses are an informative approach to provide insight on mechanisms of host responses during infection, or to provide information on vaccine efficacy. In the present study, we used Illumina sequencing to quantify the response of Arctic charr S. alpinus during infection with Aeromonas salmonicida subsp. salmonicida (Asal) with or without administration of vaccines commonly used in salmonid aquaculture: ForteMicro <sup>R</sup> or combination of ForteMicro <sup>R</sup> + Renogen <sup>R</sup> . ForteMicro <sup>R</sup> (Elanco Animal Health) is a vaccine that contains formalin-inactivated cultures of Aeromonas salmonicida, Vibrio anguillarum serotypes I & II, Vibrio ordalii and Vibrio salmonicida serotypes I & II in liquid emulsion with an oil-based adjuvant. In contrast, Renogen <sup>R</sup> (Elanco Animal Health) is a live culture of a non-virulent Arthrobacter spp. that shares common antigenic determinants with Renibacterium salmoninarum and is indicated to protect salmonids from bacterial kidney disease (BKD; https://www.drugs.com/vet/ renogen-can.html). However, in commercial settings, Renogen <sup>R</sup> is only administered in conjunction with anti-furunculosis vaccines such as ForteMicro <sup>R</sup> , and only in situations where BKD is a risk (39).

Both vaccine regimes resulted in significant protection to Arctic charr from Asal-associated mortality; however, the TABLE 3 | Transcripts that were concordantly over-expressed in both FM- and FM+R-vaccinated groups after bacterial challenge (29 dpi, 836 ddpv), showing the log2-transformed fold-change compared to unvaccinated fish.


\**Differential expression between vaccine groups*

dual-vaccination resulted in significant improvement on the administration of ForteMicro <sup>R</sup> alone. Protection was positively correlated with elevated levels of Asal-specific antibodies in both vaccine groups, similar to previous reports for Atlantic salmon (40) and rainbow trout (41). Notably, significant levels of Asal-specific antibodies in both vaccine groups prior to bacterial challenge (over 50X > sham) indicated protective immunity was in part due to circulating antibodies. It should be stated that although dual-vaccination resulted in enhanced protection above single administration against furunculosis, this approach may come with extra cost and the current study did not examine other non-target effects of this approach.

There were patterns of differential expression between unvaccinated and vaccinated Arctic charr, as well as between the two vaccinated groups, that provides information about the effects of vaccination on the host immune response, as well as the molecular pathways involved in protection. The most remarkable pattern in this study was the significant over-expression of innate humoral molecules in Arctic charr immunized with ForteMicro <sup>R</sup> . Notably, there was significant upregulation of key regulatory components (e.g., complement factor H, vitronectin, anti-thrombin), as well as the genes necessary for induction of complement and coagulation (e.g., C1q, C3, C3b, fibrinogen gamma chain, cobra venom factor). Additionally, many genes associated with metal homeostasis (e.g., ceruloplasmin, serotransferrin-2, hemopexin), and the acute phase response (e.g., fetuin-A, inter-alpha-trypsin inhibitor TABLE 4 | Expression profiles of complement and coagulation-associated genes in Sham controls surviving infection (29 dpi) and FM-vaccinated fish following infection with *Asal* at 28 dpi/836 ddpv.


*Values indicate the log*2*transformed fold change. Multiple values indicate isoforms of the same transcript.*

heavy chain H2, serum albumin 1, apolipoprotein B-100) were significantly upregulated in this group prior to infection with Asal.

Surprisingly, despite elevated cumulative survival, we did not observe the same degree of activation of the complement cascade in pre-challenge fish vaccinated with the combination of ForteMicro <sup>R</sup> and Renogen <sup>R</sup> vaccines. Instead the transcriptomic response was characterized by upregulation of a single acute phase protein (serum amyloid A-1 protein), several cellular effector markers (e.g., eosinophil peroxidase), and only two complement molecules (cobra venom factor, complement factor D). The expression profile also included transcription factors (e.g., eukaryotic translation initiation factor 3 subunit L, nuclear receptor coactivator 4, transcription factor E2F2), transporters (e.g., choline transporter-like protein 4, intermediate conductance calcium-activated potassium channel protein 4), activators of signaling cascades (e.g., WD repeat domain-containing protein 83), and actin binding pathways (e.g., alpha-adducin, Rho-related GTP-binding protein RhoE). Interestingly, there was upregulation of a hemolytic toxin (stonustoxin subunit alpha) specific to pre-challenge fish vaccinated with ForteMicro <sup>R</sup> and Renogen <sup>R</sup> . Transcripts with high homology to venom factors have also been detected in the head kidney transcriptome of Atlantic salmon infected with Asal (42), and are known ancestral forms of perforin-like



genes with pore-forming abilities and immune recognition domains (43).

After challenge with Asal, the transcriptome of Sham controls was characterized by a substantive and apparent unrestrained increase in expression. This observation is in agreement with reports for Atlantic salmon (44), rainbow trout (13), and turbot (45). Namely, there was exaggerated expression profiles of genes involved in innate immunity such as acute phase, inflammation, antigen presentation, cell differentiation, complement and coagulation, and wound repair. Additionally, there was an abundance of overexpressed transcripts associated with fatty-acid synthesis, hemoglobin synthesis, cell-cycle pathways and protein folding responses. Others have suggested that a strong immune response is critical for Atlantic salmon surviving Asal infections (15); however, this present work suggests that for the Fraser River strain of Arctic charr, a strong immune response after infection is not desirable. Overall, the present data demonstrates significant differential regulation of immune parameters in susceptible nonvaccinated charr that is otherwise absent in vaccinated fish shortly after bacterial infection, and furthermore, that protection is associated with elevated basal levels of these same parameters pre-challenge.

### Asal Interferes With Proinflammatory Pathways in Unvaccinated Charr

Bacterial pathogens establish within their host by circumventing immune responses and/or dysregulating apoptotic pathways (46). Type-3 secretion system (T3SS)-harboring pathogens such as Asal achieve this by injecting host cells with a number of virulence factors that act to interfere with critical signal transduction pathways (i.e., NFκB signaling) (47). In unvaccinated charr, we detected over-expression in several regulators of NFκB activity, including the inhibitor of κB (IκB) B-cell lymphoma 3 protein (bcl-3) which is thought to limit transcription of pro-inflammatory mediators and regulate cellular effectors (48). Significant up-regulation of several bcl-3 isoforms was a striking transcriptomic feature of non-vaccinated charr. Thus, along with concomitant upregulation of nfκb inhibitor α and nuclear factor interleukin-3-regulated protein, Asal-induced differential regulation of NFκB signaling appeared to be occurring in the susceptible PBS-injected group. Further to this, Asal AopP is known to induce apoptosis in affected cells (46, 49), resulting in the formation of subcutaneous wounds (i.e., furuncles) in host tissue and septicaemia (5), as is found with its closest ortholog, YopJ (46). We observed the formation of furuncles in unvaccinated charr at 8 dpi that was concomitant with over-expression of several transcripts involved in apoptosis (e.g., tumor necrosis factor receptor superfamily member 11B, apoptosis-enhancing nuclease, BCL2/adenovirus E1B 19 kDa protein-interacting protein 3, DNA damage-inducible transcript 4 protein). In contrast, over-expression of apoptotic-related or NFκB-associated transcripts was not evident in vaccinated charr during Asal infection.

Despite their important role in anti-microbial responses (50), previous work demonstrated that activation of toll-like receptors (TLRs) does not appear to be a feature of the host response against Asal (51). The present study reveals a similar finding, as we only detected tlr2, tlr5, and tlr13 overexpression in non-vaccinated fish, while tlr7 over-expression was observed in vaccinated fish. However, there was some interesting patterns in expression of these TLRs. For example, significant over-expression of tlr5 was observed in non-vaccinated Arctic charr head kidney during the proliferative growth of Asal. TLR5 binds bacterial flagellin and activates production of inflammatory mediators in MyD88-dependent pathways (50). The activation of TLR5 during Asal infection is unexpected, as the bacterium is non-motile. However, the T3SS is proposed to have evolved from the flagellar apparatus (52), and thus a common antigenic determinant present in the injection apparatus may be responsible for activation of TLR5. Indeed, the transcriptomic profile of non-vaccinated charr included pronounced upregulation of genes involved in TLR5-MyD88 pathways such as interleukins (il3, il6, il8, and il10) and chemokines (ccl2, and ccl19). Earlier work demonstrated a relationship between expression profiles of tlr5 and resistance to Asal in Atlantic salmon (15). Here, only susceptible non-vaccinated Arctic charr overexpressed tlr5 during the proliferative phase of Asalinfection; however, it is possible that due to the nature of our sampling design (i.e., after 8 dpi), we may have missed immediate activation of TLRs (or other immune regulators) in vaccinated fish (or unvaccinated fish), as expression of several TLRs was observed in head kidney of another salmonid species in the first 72 h post-infection with Asal (53). Thus, subsequent studies including more sampling events earlier during infection (i.e., 3 dpi) are necessary to capture the dynamics of immune activation during bacterial incubation. Others have suggested that during incubation, the T3SS may act to promote immunoregulatory

pathways characteristic of tolerogenic dendritic cells (tDCs) and regulatory T cells (Treg) (54). Our data supports this hypothesis, as a prominent feature of non-vaccinated charr was over-expression of il10, a cytokine produced by tDCs which is responsible for potentiating an immunosuppressive Treg environment (55). Orthologs to the Asal T3SS effector AcrV in Yersinia promote tDC differentiation, with resulting pathogenesis reliant upon functional IL10 (56). Further to this, Fast et al. (7) showed that T3SS deletion mutants significantly reduced production of il10 in Atlantic salmon head kidney leukocytes. We failed to detect over-expression of il10 in either vaccinated group, suggesting that the observed protection may be related to inhibition of Asal-induced immunosuppression. The mechanism for this inhibition is not fully understood, however, a novel feature of FM-vaccinated Arctic charr was the over-expression of tlr7, an endosomal TLR that binds to small synthetic compounds and ssRNA (57, 58), and of all the TLRs overexpressed in this study, is the only one known to activate IFN-α production by B cells and DCs (59, 60). The ability to activate DCs and thus elicit Th1 and CD8<sup>+</sup> T cell responses has been exploited to enhance the efficacy of vaccination (61, 62), and this may explain the failure of Asal to promote a tolerogenic immune response. Significant upregulation of tlr7 might imply that there was a vaccine-induced activation of DCs occurring, either by potentiating antigen stimulation or increased antigen uptake by DCs as suggested elsewhere (62). Further to this, significant upregulation of pore-forming perforin-like toxins (e.g., stonustoxin subunit alpha, neoverrucotoxin subunit alpha) in both groups of vaccinated fish supports an activation of cytotoxic activity (e.g., by CD8<sup>+</sup> T cells), which would be

critical for killing host cells infected with immune-evading Asal. Thus, the present transcriptomic data suggests that vaccination may act first through a substantive induction of antibodymediated complement activity targeting extracellular bacteria, and secondly through cytotoxic-mediated intracellular targeting of infected self-cells. However, subsequent studies investigating the relative contributions of DCs and cytotoxic cellular effectors in Asal-mediated killing are required to fully understand this mechanism.

### Vaccination Primes Humoral Immunity in Arctic Charr

Activation of the coagulation cascade plays a critical role in fighting infections by entrapment and prevention of systemic dissemination of bacteria (63) and is highly conserved among vertebrates as a general immune defense mechanism against bacterial pathogens (64, 65). Our results suggest that ForteMicro <sup>R</sup> primes this response, thus preparing the fish for rapid production of proteins involved in fibrin-production. During systemic inflammation, activation of the coagulation cascade is accompanied by activation of complement (66). However, detrimental uncontrolled and simultaneous activation of complement and coagulation systems can occur during severe infections, reinforcing the necessity of tight regulation. This is achieved by inhibitors (e.g., plasminogen activating inhibitor-1, α-2-antiplasmin) that act to avoid excessive localized and systemic plasmin generation (67). In the present study, we observed significant upregulation of both activators (e.g., cobra venom factor) and inhibitors (e.g., α-2-antiplasmin) prior to bacterial infection in fish vaccinated with ForteMicro <sup>R</sup> , while there was a delay in this response in unvaccinated fish. In addition to the over-expression of coagulation components, there was also significant induction of fibrinolytic enzymes in ForteMicro <sup>R</sup> vaccinates prior to bacterial infection. Interestingly, components of the fibrinolytic cascade were highly over-expressed in unvaccinated Arctic charr after infection with Asal, suggesting that uncontrolled fibrinolysis may play a role in pathogenesis, while a protective mechanism of ForteMicro <sup>R</sup> -vaccination appears to be the capacity for immediate coagulation. Earlier work indicated that a major cause of death in acute furunculosis in Atlantic salmon is circulatory failure caused by bacterialassociated coagulation (68). In agreement with Salte et al. (69), who demonstrated that prior administration of exogenous antithrombin and α-2-macroglobulin improve survival of Atlantic salmon with acute furunculosis, the present data indicates that priming this response also contributes to improved survival in Arctic charr.

The acute phase response was also activated in pre-challenge Arctic charr vaccinated with ForteMicro <sup>R</sup> (e.g., serum albumin 1, C-reactive protein, fetuin-A, haptoglobin, inter-α-trypsin inhibitor, apolipoprotein A1, apolipoprotein B100), which may have helped suppress systemic inflammation. For example, fetuin-A is protective against lethal septicaemia and end toxemia in mice by directly inhibiting production of inflammatory high mobility group proteins (70), and the presence of apolipoprotein B100 was shown to inhibit growth of Staphylococcus aurous in mice (71). Furthermore, high-density lipoprotein apolipoprotein A1 is an inhibitor of cytokine production and may play a role as an anti-inflammatory mediator [reviewed in (72)]. We observed a negative correlation between expression of high mobility group proteins (TOX high mobility group box family member 2 and high mobility group protein B3) and fetuin-A in non-vaccinated Asal-infected charr. Moreover, there was significant over-expression of serum lipoproteins apob100 and apoa1 (<5,000-fold) in the transcriptome of pre-challenge ForteMicro <sup>R</sup> -vaccinates. Thus, the present data suggests an effect of ForteMicro <sup>R</sup> may include suppressing mediators of lethal, systemic inflammation. Interestingly, this pattern was not present in ForteMicro <sup>R</sup> +Renogen <sup>R</sup> -vaccinates, suggesting that either the addition of the live Arthobacter spp. or the different formulation of the vaccine somehow abrogated this response.

Activation of metal homeostasis mechanisms was also a prominent feature in the transcriptome of ForteMicro <sup>R</sup> vaccinated Arctic charr prior to Asal challenge. Controlling availability of essential transitional metals such as copper, zinc, iron and manganese to bacterial pathogens (i.e., nutritional immunity) is a critical component of successful immune responses across vertebrates, and is accomplished by transferrin's, lactoferrins, ferritins and hem proteins (73–76). An increase in transferrin and hemopexin is highly correlative with anti-microbial activity and protection against plague in mice immediately after attenuated Yersinia pasties vaccine administration (77), with protection involving biological activities of host iron and heme-binding proteins. Acquisition of iron by Asal includes siderophore-dependent and -independent mechanisms and interference of these pathways severely compromises virulence (78). In the present study, we observed evidence of nutritional immunity in both vaccinated and nonvaccinated Arctic charr. However, the pattern of expression was markedly different among the groups, with significant overexpression in pre-challenge vaccinated fish, but in unvaccinated fish, expression was only detected after infection with Asal. Thus, a consequence of ForteMicro <sup>R</sup> appears to include early over-expression of genes involved in nutritional immunity.

### Disruption of Host Actin Pathways in Vaccinated Fish

The bacterium is thought to enter the fish host at multiple sites, including the skin, gill, and intestine (79). The typical incubation period of Asal is 3–4 days, where the bacterium rapidly disseminates in kidneys, followed by the spleen, liver, and muscles (80). Virulence during this time is attributed to the T3SS, which functionally impairs cytoskeletal functions and neutralizes immune defenses [reviewed by (81)]. The T3SS is composed of Ext, Apo, Ati2, AopP, AopO, Apo (54). Of these, Ext, AopP, and Ati2 have shown functional homology with analogs of Salmonella and Yersinia spp. (81).

One important virulence mechanism of Asal is the effect on host cell cytoskeletal integrity as evidenced by the T3SS effector protein Ext ATPase-activating domain that acts on small monomeric Gases of the Rho family (Rho, Race, and Cdc42) and the ADP ribosylating domain that depolymerizes actin (82), causes rounding of host cells (83). Furthermore, Ati2 is involved in detachment of actin binding proteins from the plasma membrane leading to the destabilization of the host cell and its cytolysis (5). Resistance to Asal pathology has been associated with downregulation of cytoskeleton-associated genes such as profiling, coiling, and actin-binding proteins in Atlantic salmon (14). In a similar fashion the present data suggests that a consequence of vaccination with ForteMicro <sup>R</sup> alone or in combination with Renogen <sup>R</sup> appears to include down-regulation of Rho-associated Gases prior to and after infection with Asal. The Rho family of Gases is critical for successful phagocytosis in macrophages by functioning to reorganize filamentous actin, assist in NF-dB and MAPK kinase transcriptional pathways, and facilitate respiratory burst (84). Thus, T3SS-associated inhibition of actin reorganization and effector cell phagocytosis permits bacterial escapement of antibody-mediated complement. We hypothesize that by giving the host an early boost, vaccination provides the immune system time to respond effectively so that downstream toxic effects of T3SS are minimized or prevented. This same mechanism has been suggested in experimental vaccinations of channel catfish Ictalarus punctuates against Flavobacterium columnare (85). In order to demonstrate this in Asal-infections, subsequent studies should aim to investigate the effect of vaccination on the specific virulence factors and their impacts on host cell actin pathways.

### CONCLUSIONS

The present data suggests that the protective effect of vaccination with ForteMicro <sup>R</sup> alone or in combination with Renogen <sup>R</sup> in the Fraser River strain of Arctic charr involves immunological priming of innate humoral components that are able to combat infection immediately upon bacterial challenge. We propose that this priming of humoral innate (complement, coagulation, and metal homeostasis effectors) and adaptive (Asal-specific antibodies) molecules allows for a rapid aggressive response during bacterial invasion, followed by swift regulation of these same genes to help circumvent immunopathology. This twopronged approach of enhancing extracellular binding and killing along with reduced intracellular uptake and survival resulted in significant protection from host mortality that was specific to Asal, but which would likely offer some level of protection against other pathogens. Future studies should investigate the contribution of individual components and particular adjuvant preparations to the host-pathways activated. It is important to note that host immune responses are likely to differ depending on the target tissue (86), and thus interpretation of the present data should consider this potential limitation. Furthermore, in order to initiate disease our infection model included an i.p.-injection

### REFERENCES


of Asal, which differs from the natural route of infection (e.g., penetration of damaged epithelium), and which may induce a more immediate and non-specific response due to accumulation of cellular effectors at the injection site (87). Therefore, to help fully elucidate mechanisms responsible for protection against Asal, examination of other routes of exposure (88), differing environmental conditions (87), and potential mucosal immune induction (i.e., systemic induction of IgT) are necessary.

### AUTHOR CONTRIBUTIONS

LB performed histological and immunohistological assays, transcriptomic data analysis and statistical analysis, and wrote the manuscript. MF and SW contributed to the conception and design of the study, performed the vaccinations, bacterial challenges, fish sampling, and contributed to writing of the manuscript. TH performed bioinformatic analysis on the RNAseq dataset. SP prepared bacterial cultures and vaccinations, helped sample fish and performed ELISA, and contributed to writing of the manuscript. AB performed qPCR assays and contributed to writing of the manuscript. All authors contributed to manuscript revision, read and approved the submitted version.

### FUNDING

This study was funded by an Atlantic Canada Opportunities Agency—Atlantic Innovation Fund grant between the Coastal Zone Research Institute (CZRI, Shippigan, NB) and the University of Prince Edward Island—Atlantic Veterinary College. Funding was also provided through the Innovation PEI graduate student scholarship program (AB) the Ocean Frontiers Institute, NSERC PGS program (AB), Novartis/Elanco Research Chair in Fish Health, and Elanco Animal Health Limited kindly donated the vaccines used in this study. LB held an NSERC post-doctoral fellowship during this work and graciously acknowledges the Centre for Excellence in Research for grants assisting in bioinformatics discovery.

### ACKNOWLEDGMENTS

The authors are grateful to the staff of the aquatic facilities at the AVC. The authors acknowledge two reviewers for their thorough and insightful comments of an earlier version of manuscript.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2019.00120/full#supplementary-material


salmonicida, the causative agent of fish furunculosis. Vet Microbiol. (2014) 169:1–7. doi: 10.1016/j.vetmic.2013.06.025


that targets the host cytoskeleton. J Biol Chem. (2007) 282:28843–52. doi: 10.1074/jbc.M704797200


88. Nordmo R, Ramstad A. Comparison of different challenge methods to evaluate the efficacy of furunculosis vaccines in Atlantic salmon, Salmo salar L. J Fish Dis. (1997) 20:119–26. doi: 10.1046/j.1365-2761.1997.d01-114.x

**Conflict of Interest Statement:** MF reports grants from Elanco Animal Health Canada Inc., personal fees from Elanco Animal Health Canada Inc., outside the submitted work.

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

Copyright © 2019 Braden, Whyte, Brown, Van Iderstine, Letendre, Groman, Lewis, Purcell, Hori and Fast. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Paralogs of Common Carp Granulocyte Colony-Stimulating Factor (G-CSF) Have Different Functions Regarding Development, Trafficking and Activation of Neutrophils

Fumihiko Katakura<sup>1</sup> \* † , Kohei Nishiya1†, Annelieke S. Wentzel <sup>2</sup> , Erika Hino<sup>1</sup> , Jiro Miyamae<sup>1</sup> , Masaharu Okano<sup>1</sup> , Geert F. Wiegertjes 2,3 and Tadaaki Moritomo<sup>1</sup>

*<sup>1</sup> Laboratory of Comparative Immunology, Department of Veterinary Medicine, Nihon University, Fujisawa, Japan, <sup>2</sup> Cell Biology and Immunology Group, Wageningen University & Research, Wageningen, Netherlands, <sup>3</sup> Aquaculture and Fisheries Group, Wageningen Institute of Animal Science, Wageningen University & Research, Wageningen, Netherlands*

### Edited by:

*Leon Grayfer, George Washington University, United States*

#### Reviewed by:

*James L. Stafford, University of Alberta, Canada Victoriano Mulero, University of Murcia, Spain*

#### \*Correspondence:

*Fumihiko Katakura katakura.fumihiko@nihon-u.ac.jp*

*†These authors have contributed equally to this work*

#### Specialty section:

*This article was submitted to Comparative Immunology, a section of the journal Frontiers in Immunology*

Received: *07 September 2018* Accepted: *29 January 2019* Published: *19 February 2019*

#### Citation:

*Katakura F, Nishiya K, Wentzel AS, Hino E, Miyamae J, Okano M, Wiegertjes GF and Moritomo T (2019) Paralogs of Common Carp Granulocyte Colony-Stimulating Factor (G-CSF) Have Different Functions Regarding Development, Trafficking and Activation of Neutrophils. Front. Immunol. 10:255. doi: 10.3389/fimmu.2019.00255* Mammalian granulocyte colony-stimulating factor (G-CSF; CSF3) is a primary cytokine that promotes the development, mobilization, and activation of neutrophils and their precursors. Teleosts have been reported to possess two paralogs as a likely result of the teleost-wide whole genome duplication (WGD) event, but functional divergence of G-CSF paralogs remains poorly understood. Common carp are an allotetraploid species owing to an additional WGD event in the carp lineage and here, we report on genomic synteny, sequence similarity, and phylogeny of four common carp G-CSF paralogs (*g-csfa1* and *g-csfa2*; *g-csfb1* and *g-csfb2*). *G-csfa1* and *g-csfa2* show differential and relatively high gene expression levels, while *g-csfb1* and *g-csfb2* show low basal gene expression levels in most tissues. All paralogs are expressed higher in macrophages than in other leukocyte sub-types and are highly up-regulated by treatment of macrophages with mitogens. Recombinant G-CSFa1 and G-CSFb1 both promoted the proliferation of kidney hematopoietic cells, while only G-CSFb1 induced the differentiation of kidney cells along the neutrophil-lineage. Colony-forming unit assays revealed that G-CSFb1 alone stimulates the formation of CFU-G colonies from head- and trunk-kidney whereas the combination of G-CSFa1 and G-CSFb1 stimulates the formation of both CFU-G and CFU-GM colonies. Recombinant G-CSFa1 and G-CSFb1 also exhibit chemotactic activity against kidney neutrophils and up-regulation of *cxcr1* mRNA expression was highest in neutrophils after G-CSFb1 stimulation. Furthermore, G-CSFb1 more than G-CSFa1 induced priming of kidney neutrophils through up-regulation of a NADPH-oxidase component p47*phox* . *In vivo* administration of G-CSF paralogs increased the number of circulating blood neutrophils of carp. Our findings demonstrate that gene duplications in teleosts can lead to functional divergence between paralogs and shed light on the sub-functionalization of G-CSF paralogs in cyprinid fish.

Keywords: teleost, granulocyte colony-stimulating factor, neutrophil, hematopoiesis, cell migration, respiratory burst

### INTRODUCTION

Granulocyte colony-stimulating factor (G-CSF), also called colony-stimulating factor 3 (CSF3), is a primary cytokine that promotes the proliferation, differentiation and survival of neutrophil progenitors and enhances trafficking and immunological functions of mature neutrophils in mammals (1). Human G-CSF is produced mainly by monocytes and macrophages (2), but is also produced by fibroblasts (3), endothelial cells (4), and bone marrow stromal cells (5). Although healthy individuals express low G-CSF protein levels in serum, remarkable elevations of G-CSF production can be induced by several inflammatory stimuli, including increased presence of pro-inflammatory cytokines and LPS during infections (6–8). Effects are mediated by the binding of G-CSF with its cognate receptor G-CSFR on neutrophils and their progenitors, activating downstream signaling cascades and thereby influencing subsequent gene expression and cellular immune responses [reviewed in (1)]. Mice lacking G-CSF/G-CSFR signaling (G-csf-deficient or G-csfr-deficient mice) exhibit a reduction in myeloid progenitors and impaired neutrophil mobilization into the circulation, resulting in chronic neutropenia (9, 10). This suggests that G-CSF is a major regulator of neutrophil development and contributes to the regulation of multipotent hematopoietic progenitors. At the same time, G-CSF also influences the phenotype and function of mature neutrophils and does so by modulating expression of for example chemokine receptors, up-regulating phagocytosis and production of reactive oxygen species (ROS) and enhancing bactericidal activity of neutrophils (11).

G-CSF was first purified and characterized in mice (12), only later followed by studies in non-mammalian vertebrates such as chicken (13), African-clawed frog (14), and a number of teleost fish species including flounder, fugu, green-spotted pufferfish (13), black rockfish (15), and zebrafish (16). Owing to a teleost-specific whole genome duplication (WGD) event (17), teleost can generally be expected to express two G-CSF paralogs, type A (G-CSFa) and type B (G-CSFb), which may not necessarily have the same function. Indeed, zebrafish express an A and B paralog and earlier studies suggest that both G-CSFa and G-CSFb are required for hematopoietic stem cell (HSC) emergence and expansion of primitive and mature neutrophils and macrophages in vivo and in vitro (16). G-csfr morphants were affected on early myeloid cell migration and development, but had functionally normal myeloid cells (18). Zebrafish G-CSFb was involved in neutrophil mobilization toward an injury site (19), but the contribution of G-CSFa remained unclear. Therefore, the exact role of teleost G-CSF paralogs as regulators of diverse markers of neutrophil activation and/or regulators of multipotent hematopoietic progenitor development has remained unresolved.

In this study, we report on the molecular and functional characterization of G-CSF paralogs from the common carp. The close kinship of zebrafish and carp (20) allows for comparative use of genetic information from the well-described zebrafish genome whereas the large size of carp allowed us to perform cell type specific gene expression and ex vivo functional studies on large number of cells. Because common carp is an allotetraploid species owing to an additional WGD event in the carp lineage (21), we report on the cloning and molecular characterization of two type A copies (g-csfa1 and g-csfa2) and two type B copies (g-csfb1 and g-csfb2). For functional characterization we chose to produce recombinant proteins for two G-CSF paralogs particularly highly expressed in macrophages (G-CSFa1 and G-CSFb1) and examined their ability to drive proliferation, differentiation and colony-formation of carp hematopoietic kidney cells along the neutrophil lineage. We also studied the effect of these G-CSF paralogs on neutrophil function including phenotype, chemotaxis and production of ROS. In vivo effects of G-CSF paralogs on circulating blood neutrophils were further investigated. We discuss the functions of teleost G-CSF regarding development, trafficking and activation of neutrophils and discuss the importance of studying paralogs of granulocyte colony-stimulating factor.

### MATERIALS AND METHODS

### Animals

Common Carp (Cyprinus carpio L.) were kept at Nihon University (NU) and at Wageningen University (WU). Carp weighing 40–100 g (10 to 15 cm in length) were purchased from commercial farms and reared at NU, Japan. Fish were kept at 23–25◦C in a recirculation system with filtered water disinfected by ultraviolet light, fed with pelleted dry food (Hikari, Kyorin CO., LTD., Japan) daily and acclimated to this environment for at least 3 weeks prior to use for all experiments except **Figures 2**– **4**. Carp were also bred and reared in the Aquatic Research Facility of WU, the Netherlands. Here, carp were raised at 23◦C in recirculating UV-treated tap water, fed pelleted dry food daily (Skretting, Nutreco) and utilized for experiments in **Figures 2**–**4**. Since G-CSF paralogs of Asian and European common carp show very high sequence identity (98 to 100%), we combined data from NU and WU. Experiments were performed in accordance with the guidelines of NU and WU and with approval of the animal experimental committee of WU.

### Isolation of Carp Tissues and Leukocytes and Purification of Leukocyte Sub-types Such as B Cells, Granulocytes, Macrophages, Thymocytes and Thrombocytes

For tissue and cell isolation, carp were anesthetized with 0.01% Benzocaine (Sigma-Aldrich) or Tricaine Methane Sulfonate (TMS, Crescent Research Chemicals, Phoenix, USA), bled from the caudal vein and euthanized.

**Abbreviations:** WGD, whole genome duplication; LPS, lipopolysaccharide; ConA, concanavaline A; PMA, phorbol 12-myristate 13-acetate; polyI:C, polyinosinicpolycytidylic acid; IPTG, isopropyl β-D-L-thiogalactopyranoside; MTT, 3-(4, 5-dimethyl-2-thiazolyl)-2, 5-diphenyltetrazolium bromide; CFU-G, granulocyte colony-forming unit; CFU-GM, granulocyte/macrophage colony-forming unit; fMLP, N-formyl-methionyl-leucyl-phenylalanine; DHR123, dihydrorhodamine 123; FITC, fluorescein isothiocyanate.

Leukocytes were obtained from kidney (head and/or trunk kidney) and spleen. Cell suspensions were obtained by macerating tissues on a sterile mesh in 10 mL of Eagle's minimal essential medium (MEM, Nissui, Tokyo, Japan). Cells were collected by centrifugation at 250 × g for 5 min at 4◦C, re-suspended in 5 mL of MEM, layered onto a Percoll (1,075 g/cm<sup>3</sup> , GE healthcare) and centrifuged at 430 × g for 20 min at 4 ◦C. Cells at the medium/Percoll interface (mononuclear cells) were harvested, washed twice with MEM by centrifugation, re-suspended with E-RDF medium (Kyokuto Pharm. Ind. Co.,Ltd., Tokyo, Japan) containing 20% fetal bovine serum and 2.5% carp serum (E-RDF20/2.5) and passed through 40µm filter to remove aggregate.

Peripheral blood leukocytes (PBL) were obtained from carp blood. In short, 1 mL of blood was withdrawn from the caudal vein from fish with heparinized syringe, transferred to 9 mL of ice-cold MEM, layered onto a Percoll (1,075) and centrifuged at 430 × g for 20 min at 4◦C without brakes. Cells at the medium/Percoll interface were harvested, washed twice with MEM by centrifugation and re-suspended with E-RDF20/2.5.

Kidney neutrophils were isolated as described previously (22) with minor modifications. Briefly, trunk kidney cells were layered onto a Percoll discontinuous gradient (1,080 and 1,100 g/cm<sup>3</sup> ) and centrifuged at 430 × g for 20 min at 4◦C. Cells at the 1,080/1,100 interface (neutrophils and erythrocytes) were harvested after complete removal of cells at the upper phase and then washed once. The neutrophil/erythrocyte pellet was treated with 1× red blood cell lysis buffer (150 mM NH4Cl, 10 mM KHCO3, 0.1 mM EDTA). Cells were washed twice with MEM by centrifugation and re-suspended with appropriate medium for each experiment. The purity of the neutrophils was verified to be >92% by a flow cytometry using a BD FACS Canto (BD Biosciences) and a peroxidase staining according to a DAB oxidization.

Thymocytes (23), macrophages (24), neutrophilic granulocytes (25) were obtained as previously described. Magnetic activated cell sorting (MACS) was used to isolate B cells and thrombocytes from peripheral blood leukocytes (PBLs) using anti-carp IgM [WCI12, (26)] and anti-carp thrombocytes [WCL6, (27)] antibodies and neutrophilic granulocytes from trunk kidney [using monoclonal antibody TCL-BE8; (25)]. The purity of the sorted cells was verified to be >98% by flow cytometry using a BD FACS Canto A (BD Biosciences).

### Identification and in silico Analysis of Carp G-CSFs

Genomic loci of carp G-CSF were predicted by the Augustus gene prediction server using information on genes (med24, psmd3, and kpnb1) known to be neighboring G-CSF in several other species. Primers were designed against carp genomic sequences encoding putative carp G-CSFs. The complete list of primers used for PCR, RACE PCR, qRT-PCR and recombinant protein expression are listed in **Supplementary Tables S2–S4**. PCR reactions were performed using cDNA from carp kidney and heart with PrimeSTAR HS DNA polymerase (Takara, Shiga, Japan). Generated amplicons were gel purified using the FastGene Gel/PCR Extraction kit (Nippon genetics, Tokyo, Japan), ligated into the pMD20-T vector (Takara) using the 10×A-attachment Mix (Toyobo, Osaka, Japan) and the Ligation high ver. 2 (Toyobo) and transformed into the competent Escherichia coli DH5α. Positive colonies were identified by colony PCR using the vector specific M13 RV and M4 primers, plasmids isolated using the FastGene Plasmid Mini kit (Nippon genetics) and inserts sequenced using a BigDye terminator v3.1 cycle sequencing kit (Applied Biosystems) and an ABI PRISM 3130 Genetic Analyzer (Applied Biosystems). Sequence analysis was performed using the Genetyx version 11 (Genetic Information Processing Software) and sequences aligned and analyzed using BLAST programs.

G-CSF protein sequences from fish, amphibian and mammals were aligned using Clustal Omega software (EMBL-European Bioinformatics Institute). Signal peptide regions of respective G-CSF proteins were estimated using the SignalP 4.0 server (http://www.cbs.dtu.dk/services/SignalP/) and conserved motifs were predicted using the SMART server (http://smart.emblheidelberg.de/). Phylogenetic analysis was conducted using the MEGA version 6 software using the neighbor-joining (NJ) method and the Poisson method, and bootstrapped 1,000 times, with values expressed as percentages. The full-length sequences of carp G-CSFa1, G-CSFa2, G-CSFb1, and G-CSFb2 (accession number: MG882495, MG882496, MG882497, and MG882498, respectively) have been submitted to GenBank.

### RT-qPCR Analysis of Gene Expression of Carp G-CSF Paralogs in Healthy Carp Tissues, Different Cell Types and Macrophages Stimulated With Mitogens

To investigate basal gene expression levels of G-CSF paralogs, tissues and leukocytes were collected from healthy carp (detailed in the figure legends), then total RNA was isolated using the RNeasy kit (QIAGEN) including on-column DNase treatment according to the manufacturer's instruction and stored at −80◦C. cDNA was synthesized with SuperScript III First Strand Synthesis System (Invitrogen) according to manufacturer's instructions. Real-time quantitative PCR (RT-qPCR) was performed with a Rotor-Gene 6000 (Corbett Research) using ABsolute QPCR SYBR Green Mix (Thermo Scientific). Fluorescence data from RT-qPCR experiments were analyzed using Rotor-Gene software v1.7. The take-off value for each sample and the average reaction efficiencies (E) for each primer set were obtained upon Comparative Quantitation Analysis from Rotor Gene Software. The relative expression ratio (R) of target genes were calculated based on the average E and relative to the s11 protein of the 40 s subunit as a reference gene. Take-off values of samples in which genes of interest were non-detectable were given an arbitrary take-off value of 32.

### Generation of Recombinant Carp G-CSFa1 and G-CSFb1

The portions of the carp G-CSFa1 and G-CSFb1 sequences corresponding to the mature, signal sequence-cleaved peptides were PCR amplified from carp kidney and heart cDNA using primers designed to meet the requirements of the pET-16b expression vector (Novagen). The resulting PCR products were ligated into pMD20-T vector, digested with two restriction enzymes, NdeI and BamHI, isolated by gel electrophoresis, ligated into the pET-16b which carry an N-terminal 10x His-tag, transformed into competent E. coli DH5α. Plasmids containing the in frame insert of carp G-CSFa1 or G-CSFb1 were transformed into the T7 polymerase expressing Rossetta-gami B (DE3) pLysS E. coli (Novagen), induced with appropriate IPTG and the optimal induction times and temperatures deduced in pilot runs. The bacteria were scaled up accordingly.

Recombinant carp G-CSFa1 and G-CSFb1 were purified from bacterial cells using Ni-NTA agarose (Qiagen, Hilden, Germany) according to the manufacturer's procedure. Briefly, transformed E. coli cells were grown in 100 mL LB medium containing 50 mg/mL ampicillin and 30 mg/mL chloramphenicol to density of OD<sup>600</sup> = 0.5 at 37◦C, and then cooled on ice. Expression of recombinant G-CSFa1 was induced by addition of 0.5 mM IPTG at 37◦C for 4 h, and expression of recombinant G-CSFb1 was induced by addition of 0.25 mM IPTG at 25◦C for 8 h. After shaking the cultures, cells were harvested, lysed in the lysis buffer (20 mM sodium phosphate, 500 mM NaCl, 50 mM imidazole, pH 7.4, 0.1% Triton-X and protease inhibitor cocktail) and sonicated. The insoluble materials were removed by centrifugation at 9,600 × g for 20 min and the supernatants were incubated with 500 µL of Ni-NTA agarose slurry at 4◦C for 1–2 h with gentle rotation. The resin was then washed with 30 mL of the wash buffer (20 mM sodium phosphate, 500 mM NaCl, 50 mM imidazole, pH 7.4) on columns. Proteins were eluted from the resin using the elution buffer (20 mM sodium phosphate, 500 mM NaCl, 500 mM imidazole, pH 7.4). Subsequently, recombinant G-CSFa1 and G-CSFb1 were purified with gel-filtration to further clarify and simultaneously to determine their molecular weight under a native condition. Gel-filtration was performed using a Sephacryl S-100 column (HR 16/160, GE Healthcare) and the proteins were eluted with 20 mM sodium phosphate buffer containing 300 mM NaCl, pH 7.4 in 0.5 mL/min flow rate. The purified proteins were then passed through Pierce High Capacity Endotoxin Removal Columns (Thermo Fisher Scientific) to remove potential traces of endotoxin, buffer exchanged to 1x PBS with 0.05% BSA, filter sterilized (0.22µm) and stored at 4◦C or −80◦C until use. Residual endotoxin was checked to be less than 5 pg/mL according to a Limulus ES-II Single Test (Wako, Osaka, Japan). A Bradford assay (Thermo Fisher Scientific) was performed according to manufactures' directions to determine protein concentration.

Recombinant carp erythropoietin (EPO), kit ligand A (KITLA) and thrombopoietin (TPO) were produced and purified as described previously (28, 29).

### Cell Proliferation Assay

Freshly isolated carp head and trunk kidney leukocytes from 4 individuals were adjusted to a concentration of 4–8 × 10<sup>5</sup> cells/mL in E-RDF20/2.5 medium. Fifty microlitre of this cell suspension was added to each well of a 96-well plate to which 50 µL of treatment in E-RDF20/2.5 medium was added. Treatments consisted of the E-RDF medium (negative control), 25% cell conditioned medium (CCM) derived from the carp kidney leukocyte culture in which macrophages develop [positive control, (30)], a combination of 100 ng/mL TPO and 100 ng/mL KITLA (positive control), recombinant G-CSFa1 or G-CSFb1 at final concentrations of 500, 100, 20, 4, 0.8, 0.16 ng/mL and heat-inactivated (98◦C for 30 min) recombinant G-CSFa1 and G-CSFb1. Cell proliferation was determined using the colorimetric MTT assay (Nacalai Tesque, Kyoto, Japan) which was first shown to provide comparable data for different leukocyte cell types (data not shown). Briefly, 10 µL of MTT reagent was added to each well and plates were incubated at 30◦C for 5 h to develop a coloration reaction depend on live cell number. One hundred microlitre of solubilization solution (acid-isopropanol) was then added to each well and plates were sealed and kept at 30◦C for 12 h. Cell proliferation was determined on days 0, 3, 6, and 9, and plates were read at absorbance of 570 nm and 650 nm as a reference using Multiskan GO microplate reader (Thermo Fisher Scientific). Values obtained at absorbance of 650 nm from each well were subtracted from values obtained at absorbance of 570 nm from each well.

### RT-qPCR Analysis of Gene Expressions in Cells Treated With Recombinant G-CSFa1 and/or G-CSFb1

Freshly isolated carp kidney leukocytes were seeded into 24-well plates in 0.5 mL of E-RDF20/2.5 at a concentration of 4 × 10<sup>5</sup> cells/mL. Cells were either treated with the medium (untreated control), recombinant G-CSFa1 (100 ng/mL final concentration), recombinant G-CSFb1 (100 ng/mL final concentration) or the combination of 100 ng/mL G-CSFa1 and 100 ng/mL G-CSFb1 in the E-RDF20/2.5 for 12 h and 4 days at 30◦C. Following incubation, cells were collected, total RNA was isolated using the NucleoSpin RNA kit (Macherey-Nagel, Düren, Germany) and reverse transcribed into cDNA using the Omniscript RT kit (Qiagen) according to manufacturer's instructions. Quantitative PCR was performed for carp transcription factors known to be involved in early hematopoiesis (gata2) (31), myelopoiesis (pu.1, cebpa and irf8) (32–35), erythropoiesis (gata1) (36) and lymphopoiesis(gata3 and pax5) (37, 38) and myeloid cell markers (gcsfr1, gcsfr2, csf1r and mpx) (22, 35) using a Thermal Cycler Dice Real Time System II (Takara). Beta-actin (β-actin) was employed as an endogenous control. Quantitative PCR cycling conditions were 95◦C for 30 s followed by 40 cycles of 95◦C for 5 s and 60◦C for 30 s. Data were analyzed using the Thermal Cycler Dice Real Time System software (Takara) and is represented as the average of the four fish (n = 4) with standard deviation.

Likewise, kidney neutrophils from carp were treated with the medium, G-CSFa1 and G-CSFb1 in the E-RDF20/2.5 for 6 h at 30◦C, RNA isolated, and cDNA synthesis as described above. Q-PCR was performed for carp chemokine receptors (cxcr1, cxcr2 and cxcr4) (39, 40) and NADPH oxidase components (gp91phox , p22phox , p40phox , p67phox , and p47phox) (41). Beta-actin (β-actin) was employed as an endogenous control. Quantitative PCR cycling conditions were 95◦C for 30 s followed by 40 cycles of 95◦C for 5 s and 60◦C for 30 s. Data were analyzed using the Thermal Cycler Dice Real Time System software (Takara) and is represented as the average of the three fish (n = 3) with standard deviation.

### Semi-solid Colony-Forming Unit Assay

Freshly isolated leukocytes from carp kidney, spleen or peripheral blood were re-suspended to 2 × 10<sup>5</sup> cells/mL in E-RDF medium containing 40% FBS and 5% carp serum. Colony-forming unit assay using semi-solid media was carried out as previously described (28) with minor modifications. Briefly, a complete methylcellulose medium was prepared by mixing a 2.0% methylcellulose stock solution with an equal amount of the cell suspension. In some cases (for experiments of colony counting based on the morphology), a complete 0.45% agarose medium was prepared by mixing an 1.8% agarose solution; 2× E-RDF medium; and the cell suspension in the volume ratio of 1:1:2. Then, 2.4 mL of the cell suspension/complete semi-solid medium was added to a 5 mL tube with a 2.5 mL syringe and 16-gauge bluntend needle, along with cytokines or PBS. Tubes were tightly capped and the solution gently mixed. One milliliter of the solution (in duplicate) was aliquoted onto a solid E-RDF medium containing 0.45% agarose (Lonza), 20% FBS and 2.5% carp serum in a 35 mm dish or a 6-well plate. Dishes and Plates were incubated at 30◦C with an additional 5% CO<sup>2</sup> atmosphere and 100% humidity for 7–13 days, followed by examination for colony formation.

The number of progenitor cells in each organ was estimated according to the formula described below.

No. of progenitors = (Total No. of leukocytes from each organ) × (No. of colonies forming per plate)/(No. of leukocytes plated).

### Characterization of Colony Cells

Cell colonies formed in the methylcellulose media were aspirated by micropipette under a microscope and characterized by morphology, cytochemistry and RT-PCR analyses as previously described (28). In short, the colony cells were re-suspended in MEM, and cyto-centrifuged with a Cytospin (Shandon). Slides were dried, fixed and stained with May-Grünwald Giemsa (MGG, Wako Pure Chemicals, Osaka, Japan) or Peroxidase stain based on the DAB oxidization. For RT-PCR analyses, total RNAs were extracted from each colony cell type using RNeasy Micro Kit (Qiagen) and cDNA was synthesized using Omniscript RT kit (Qiagen). Expression of hematopoietic marker genes was analyzed by PCR using carp specific primers and EmeraldAmp PCR Master Mix (Takara). PCR was conducted as follows: one cycle of 94◦C for 1 min, followed by 23 to 38 cycles of denaturation at 98◦C for 30 sec, annealing at 58◦C for 30 s and elongation at 72◦C for 30 s. Colonies treated with recombinant carp erythropoietin (EPO) was utilized for the control group of the erythroid lineage.

## Neutrophil Chemotactic Response to Recombinant Carp G-CSFa1 and G-CSFb1

Neutrophils obtained from carp trunk kidney were washed twice in MEM and adjusted to a final concentration of 1 × 10<sup>6</sup> cells/mL. The chemotaxis assay was performed using blind well chemotaxis chambers (Neuro Probe, Inc.). Two hundred microliters of different concentrations of recombinant carp G-CSFa1 or G-CSFb1 (1, 10, and 100 ng/mL, final concentrations) in the serum free MEM were added to the bottom well of the chemotaxis chambers, and the bottom chamber was separated from the top chamber using 5µm pore size polycarbonate membrane filters (Neuro Probe, Inc.). To the top chamber, 200 µL of neutrophil suspension was added. Negative controls consisted of medium alone and the positive control was 10 ng/mL of fMLP (Sigma-Aldrich). The chemokinesis control consisted of 100 ng/mL of G-CSFa1 or 100 ng/mL of G-CSFb1 in both the upper and lower chambers of the chemotaxis apparatus.

The incubation period was 1 h after which the cell suspensions were carefully aspirated from the top chamber and the filters removed and applied bottom side up on a slide glass. Filters were stained with MGG. Chemotactic activity was determined by counting the total number of cells found on the underside of the polycarbonate filters in 20 random fields of view (40× magnification). Technical duplicates were conducted for all treatments (n = 4, biological replicates).

### Respiratory Burst Assay

Respiratory burst assay was performed as previously described (42) with minor modifications. The neutrophils harvested from carp kidney were re-suspended in E-RDF20/2.5 medium at a concentration of 2.5 × 10<sup>6</sup> cells/mL. Four hundred microliters of the cell suspension was added to each 1.5 mL tubes, and cells were treated or untreated with recombinant carp G-CSFa1 (100 ng/mL) or G-CSFb1 (100 ng/mL) at 25◦C for 6 h. Following the incubation, DHR123 (Molecular Probes) was added to the cells at a final concentration of 10µM and incubated for 5 min to allow the cells to take up the DHR123. PMA (Abcam, Cambridge, UK) was then added at a final concentration of 100 ng/ml. The cells were further incubated for 30 min to allow oxidation of the DHR. All samples were appropriately staggered with respect to timing to accommodate the transient state of oxidized DHR fluorescence. Live cells were gated according to the forward scatter and side scatter parameters. DHR fluorescence was detected in the FITC channel, and the mean values of the FITC fluorescence in neutrophils were normalized to untreated controls.

### In vivo Effects of G-CSFa1 and G-CSFb1

Carp were bled 200 µL from the caudal vein using a heparinized syringe 2 days before administration of G-CSF paralogs and blood was centrifuged in a capillary glass tube at 1,500× g for 5 min. Leukocytes on top of the erythrocyte layer were obtained, treated with the 1× red blood cell lysis buffer, washed twice with Hanks' balanced salt solution and then used to measure the ratio of neutrophils per total number of leukocytes by flow cytometry analysis, based on forward scatter vs. side scatter parameters. Subsequently, carp were injected intraperitoneal (i.p.) with 100 ng/g body weight of recombinant proteins in 200 µL of 1×PBS, or were injected with 1×PBS only. After 6, 24 and 48 h, 200 µL of peripheral blood was collected and analyzed as described above. Three fish for each group were examined.

### Statistical Analysis

Raw data of technical replicates were first averaged per individual before statistical analysis was performed. Statistical analysis was performed after log-transformation of datasets that were not normally distributed. Subsequently, normality was assumed and statistical significance was tested using an un-paired Student's t-test (independent observations, **Figure 3**) for one-toone comparisons and a (repeated-measures) one-way analysis of variance (ANOVA) followed by a Dunnet's post-hoc test (in **Figures 4**, **5A,D,E**, **7**, **8**). Or, Tukey's HSD (**Figure 2**) and Dunnet's T3 (in case of unequal variances) (**Figure 2**) were used for multiple comparisons. A two-way ANOVA was performed followed by Tukey's post-hoctest for multiple comparisons shown in **Figures 9**, **10**. Prism 7 software (GraphPad Software, La Jolla, CA, USA) was used. In absence of sphericity, the Geisser-Greenhouse correction was applied. A value of p ≤ 0.05 was accepted as significant.

### RESULTS

### Identification of Four Carp G-CSF Paralogs

The presence of four G-CSF paralogs in the genome of common carp was expected as common carp has undergone a WGD event compared to the zebrafish (21), in which two G-CSF paralogs are already present. Referring to carp genome and transcriptome databases (project no. PRJEB7241 and PRJNA73579) at the National Center for Biotechnology Information (NCBI), four putative loci encoding carp G-CSF homologs were found next to conserved genes PSMD3, MED24, or LRRC3 (**Figure 1A**). Synteny of each paralog is highly conserved with either the zebrafish G-CSFa locus on chromosome 12 or the GCSFb locus on chromosome 19.

The complete open reading frames (672, 675, 588, and 582 bp) of four carp paralogs' cDNA transcripts, respectively, encoding 224, 225, 196, and 194 amino acids with 5 exons were obtained (**Figure 1B** and **Supplementary Figures S1A–D**). Despite quite low sequence identity and similarity (**Supplementary Table S1**), carp G-CSF paralogs share a similar predicted structure and one of the copies (G-CSFa1) was predicted to have an additional helical region from Ser<sup>160</sup> to Ser<sup>164</sup> which is acidic amino acid residue (Asp and Glu)-abundant (**Figure 1B** and **Supplementary Figure S2**). Carp G-CSF paralogs all possess the consensus domain of Pfam IL6/GCSF/MGF protein family, whereas the four cysteine residues involved in two disulfide bonds are not conserved. Carp G-CSF copies also share conserved acidic amino acids involved in major ligand-receptor binding demonstrated in human G-CSF, while there is no acidic amino acid residue near the α-helix E in carp G-CSFa2 (**Figure 1B**). Phylogenetic analysis revealed that all the G-CSFs were found to form a single evolutionary clade outside a related cytokine interleukin-6, suggesting that the G-CSFs are orthologous. Taking into account there may be G-CSF paralogs present in the teleost species shown that have not yet been reported, each of the four carp paralogs did cluster with either teleost G-CSFa or G-CSFb paralogs (**Figure 1C**). Hence, based on clustering and the conserved synteny, we named the four carp G-CSF paralogs as G-CSFa1, G-CSFa2, and G-CSFb1, G-CSFb2.

### Carp g-csf Paralogs Show Differential Expression in Immune Tissues and Cells

Assessment of basal g-csfa1 expression in tissues from healthy carp revealed generally very low expression of g-csfa1 in most tissues, with significantly higher gene expression in spleen, muscle and gill (**Figure 2**). G-csfa2 was significantly higher expressed in spleen than in gill, brain, thymus, trunk-kidney and head-kidney (**Figure 2**). Basal expression levels of g-csfb1 and g-csfb2 were generally low or non-detectable in most carp tissues examined (**Figure 2**).

At basal levels, g-csfa expression is markedly higher than g-csfb expression in all immune cells examined. Strikingly, all g-csf paralogs were highest expressed at basal level in macrophages, indicating these cells as the major producers of G-CSF, comparable to mammalian macrophages. Within macrophages, g-csfa1 and g-csfb1 were significantly higher expressed compared to their respective counterparts (**Figures 3A,B**). Remarkably, a clear expression of g-csfa1 was observed also in thrombocytes (**Figure 3A**).

### Expression of Carp g-csf Paralogs Are Immediately Enhanced After Stimulation

In order to determine induction of the different G-CSF paralogs upon antigenic stimulation, we investigated expression levels in freshly isolated kidney leukocytes and head kidney-derived macrophages following the stimulation with LPS, ConA, PMA and poly I:C (only freshly isolated kidney leukocytes). In freshly isolated kidney leukocytes, all paralogs were highly up-regulated after stimulation with LPS and the combination of ConA/PMA at 3 and 6 h but not after stimulation with Poly I:C (**Supplementary Figure S3**). Likewise, in the cultured macrophages gene expressions of the four paralogs were clearly enhanced by LPS and PMA stimulations (**Figure 4**). Despite nondetectable g-csfb1 transcripts, its gene expression was induced with LPS stimulation. Interestingly, in macrophages both gcsfa1 and g-csfa2 are relatively high expressed at basal level (**Figure 3**) and appear to show a relatively small increase upon stimulation with LPS (**Figure 4**), whereas g-csfb1 and g-csfb2, which are relatively low or non-detectable at basal level (**Figure 3**), show a large increase in gene expression after LPS stimulation (**Figure 4**). We could also show that expression in macrophages of interleukin-1 beta, which is a proinflammatory cytokine, was significantly up-regulated after LPS stimulation (**Figure 4**).

### Recombinant Carp G-CSFa1 and G-CSFb1 Are Monomeric Forms

Based on expression levels in macrophages and the clear induction in stimulated macrophages, we chose to express two copies, G-CSFa1 and G-CSFb1, to investigate

their function. Recombinant G-CSFa1 and G-CSFb1 purified using Ni-affinity chromatography were passed through a gel filtration column under a non-denaturing condition to calculate their molecular weights. As a result, the molecular weights of G-CSFa1 and G-CSFb1 were estimated to 25,275 and 22,355, similar to the deduced values based on their primary structures and similar to the result of SDS-PAGE under the denaturing condition,

reference gene and are presented as mean + standard deviation (*n* = 3, except thymus, *n* =2). *G-csfb1* expression was non-detectable in all tissues examined (ND indicates "non-detectable"). Significant differences in expression between tissues were determined using one-way ANOVA followed by Tukey's HSD (*g-csfa1* and *g-csfb2*) or Dunnet's T3 *post-hoc* test for unequal variances (*g-csfa2*).

indicating that both recombinants form monomers (**Supplementary Figure S4**).

### Both G-CSFa1 and G-CSFb1 Induce Proliferation of Kidney Leukocytes, but Only G-CSFb1 Induce Differentiation of Cells Along the Neutrophil Lineage

Carp kidney leukocytes treated with the cell conditioned medium containing macrophage growth factor(s) and recombinant TPO plus KITLA exhibited active proliferation, indicating that there are heterogeneous hematopoietic progenitors in the kidney leukocytes (**Figure 5A** and **Supplementary Figure S5**). Treatment of carp kidney leukocytes with 0.8, 4, 20, 100, and 500 ng/mL of G-CSFa1 induced a dose-dependent proliferative response, with the highest proliferation observed in cells treated with more than 20 ng/mL of G-CSFa1, whereas heat-inactivated (98◦C for 30 min) G-CSFa1 had no effect. Likewise, treatment of kidney leukocytes with 4, 20, 100, and 500 ng/mL of G-CSFb1 induced a dose-dependent proliferative response, with the highest proliferation observed in cells treated with more than 100 ng/mL of G-CSFb1, whereas heat-inactivated G-CSFb1 had no effect (**Figure 5A** and **Supplementary Figure S5**). Furthermore, treatment of kidney

followed by Dunnet's *post-hoc* test, (*p* < 0.05) are denoted by asterisks (\*). Non-detectable samples were given an arbitrary value of CT = 32. Hash mark (#) indicate significant differences using these arbitrary values.

leukocytes with a combination of 100 ng/mL of G-CSFa1 and 100 ng/mL of G-CSFb1 enhanced the proliferative response compared with those cells treated with G-CSFa1 alone or G-CSFb1 alone (data not shown).

A lot of growing cells treated with G-CSFa1 adhered onto the plastic and with each other, whereas cells treated with G-CSFb1 exhibited low adhesive property and dispersed (**Figure 5B**). Morphologically, most cells treated with G-CSFa1 for 8 days were blast-like cells, having a basophilic cytoplasm and round to oval nuclei (**Figure 5C**). In contrast, the cells treated with G-CSFb1 for 8 days appeared to be at different developmental stages from myeloblast-like to metamyelocyte-like (**Figure 5C**). Most growing cells with each treatment were ascertained to be myeloid cells by staining with TCL-BE8 monoclonal antibody which mainly binds to carp neutrophils (43) (data not shown).

To characterize the roles of G-CSFa1 and G-CSFb1, we examined the gene expressions of transcription factors (TFs) and cell surface markers involved in the development of various cell lineages in carp kidney leukocytes treated with G-CSFa1, G-CSFb1 and a combination of G-CSFa1 and G-CSFb1. The mRNA levels of the TFs involved in myelopoiesis (pu.1, cebpα and irf8), early hematopoiesis (gata2), erythropoiesis (gata1) and lymphopoiesis (gata3 and pax5) in cells treated with or without G-CSFa1, G-CSFb1 and a combination of them for 12 h were analyzed by quantitative PCR. Kidney cells treated with G-CSFa1 did not undergo any change of TFs mRNA levels. On the other hand, kidney cells treated with G-CSFb1 exhibited a significant up-regulation of cebpα mRNA levels compared to those of the medium-treated controls (**Figure 5D**), while other TFs tested showed no significant change (data not shown). Cells treated with the combination of G-CSFa1 and G-CSFb1 also

were frequently observed (small enclosure). Scale bars indicate 10µm. (D) Quantitative gene expression analysis of carp transcription factors involved in granulopoiesis (*cebp*α) in carp kidney leukocytes treated or untreated with G-CSFa1 (100 ng/mL), G-CSFb1 (100 ng/mL) or a combination of G-CSFa1 (100 ng/mL) and G-CSFb1 (100 ng/mL) for 12 h. The mRNA levels were calculated using β*-actin* as a reference gene. Data were normalized to the control cells (dashed like at y = 1) and mean + standard deviation is shown (*n* = 4). Significant differences compared to unstimulated controls were determined using one-way ANOVA followed by Dunnet's *post-hoc* test, (*p* < 0.05) are denoted by asterisks (\*). (E) Quantitative gene expression analysis of myeloid cytokine receptors and myeloperoxidase in carp kidney leukocytes treated or untreated with G-CSFa1, G-CSFb1 or a combination of G-CSFa1 and G-CSFb1 for 4 days. The mRNA levels were calculated using β*-actin* as a reference gene. Data were normalized to the control cells (dashed line at *y* = 1) and mean + standard deviation is shown (*n* = 4). Significant differences compared to unstimulated controls were determined using one-way ANOVA followed by Dunnet's *post-hoc* test, (*p* < 0.05) are denoted by asterisks (\*).

showed a moderate up-regulation of cebpα levels compared to those of the controls (**Figure 5D**). Next, we examined whether G-CSFa1 and G-CSFb1 modulate expression of myeloid cytokine receptors and neutrophil-specific myeloperoxidase in carp kidney cells. The mRNA levels of csf1r, gcsfr1, gcsfr2, and mpx in cells treated with the same treatments for 4 days were analyzed by quantitative PCR. Expression of csf1r, which is the macrophage colony-stimulating factor receptor gene, in the kidney cells was unaffected with any treatment examined (**Figure 5E**). Gcsfr1 and mpx expression in the kidney leukocytes was up-regulated with the treatment of G-CSFb1 alone and the combination of G-CSFa1 and G-CSFb1, but not with G-CSFa1. On the other hand, gcsfr2 expression in the kidney leukocytes shows a trend toward upregulation with G-CSFa1 treatment but downregulation with G-CSFb1 treatment (**Figure 5E**).

### G-CSFb1 Stimulates Granulocyte Colony Formation and Cooperates With G-CSFa1 to Stimulate Granulocyte/Macrophage Colony Formation

In order to further examine the hematopoietic function G-CSFa1 and G-CSFb1 and identify granulocyte progenitor cells, we used an in vitro methylcellulose/agarose colony assay system. As expected, plating of carp kidney leukocytes (100,000 cells) without addition of cytokine resulted in no colony formation (data not shown). In the presence of G-CSF paralogs, overall two types of colonies appeared (**Figure 6A**). Surprisingly, when carp kidney leukocytes were cultured with G-CSFa1 alone, few colony formations were observed at any dose (**Figure 6B** left). On the other hand, in the presence of G-CSFb1, approximately 25 homogeneous colonies were formed after 7 days of the

FIGURE 6 | Colony formation of kidney cells in response to recombinant carp G-CSFa1 and G-CSFb1. (A) Colony-formation of kidney cells in response to G-CSF paralogs. Overall two types of colonies (type 1 and type 2) were observed. Bars indicate 200µm. (B) Colony counts during semi-solid culture of kidney cells (1 × 10<sup>5</sup> ) in the presence of 100 ng/mL G-CSFa1 alone, 100 ng/mL G-CSFb1 alone, or a combination of 100 ng/mL G-CSFa1 and 100 ng/mL G-CSFb1. Each point indicates mean colony counts from 4 individual fish under each condition. Cultures scored every 2 days between 3 and 13 days of incubation. (C) May-Grünwald Giemsa (MGG) staining of colony cells (type 1; left and type 2; right). Bars indicate 10µm. (D) Peroxidase-staining of cells obtained from type 1 colonies (left) and type 2 colonies (right), counterstained with Mayer's Hematoxylin. Arrow heads indicate myeloperoxidase-positive cells. Bars indicate 10µm. (E) RT-PCR analysis for expression of lineage-associated marker genes in type 1 (lane 1) and type 2 (lane 2) colony cells. cDNA from carp kidney leukocytes was used as a positive control (lane K). cDNA from cells cultured in the presence of 100 ng/mL carp EPO was utilized for the control group of the erythroid lineage (lane E).

TABLE 1 | The number of type 1 and type 2 colonies formed from 100,000 cells in head kidney, trunk kidney, spleen, and PBLs in the semi-solid culture with the combination of 100 ng/mL G-CSFa1 and 100 ng/mL G-CSFb1.


*Data were obtained from duplicate cultures in the presence or absence of both 100 ng/mL G-CSFa1 and 100 ng/mL G-CSFb1 and shown only from the culture with the cytokines (n* = *4). Type 1 colonies were counted at 7 days of cultivation. Type 2 colonies were counted at 10 days of cultivation. ND, not determined.*

incubation (**Figure 6B** middle). These colonies consisted of uniform small round cells scattered (type 1, **Figure 6A** left). When kidney cells were cultured with a combination of G-CSFa1 and G-CSFb1, morphologically two kinds of colonies were observed. One appeared to be similar to the type 1 colonies formed in the presence of G-CSFb1 alone, the other seemed to consist of roughly agminated cells with distinct sizes and shapes (type 2, **Figure 6A** right). Approximately ten type 1 colonies per 100,000 cells plated were formed at day 5 to 7 in the culture and then gradually disappeared. The peak of type 2 colony formation (about 20 colonies per 100,000 cells plated) was observed after 11 days of cultivation (**Figure 6B** right). Both type 1 and type 2 colony cells consisted of morphologically neutrophil lineage cells at distinct developmental stages, which are myeloperoxidase-positive and –negative (**Figures 6C,D**). To characterize colony types, the expression of lineage-associated marker genes was analyzed. **Figure 6E** shows a typical expression patterns in type 1 and type 2 colonies. Type 1 colonies treated with G-CSFb1 alone or G-CSFa1 plus G-CSFb1 highly expressed g-csfr, cebpα, and mpx mRNAs involved in neutrophil development and slightly expressed csf1r which is the macrophage colony-stimulating factor receptor gene, but did not express other genes examined, indicating that type 1 colonies are derived from the progenitor cells corresponding to mammalian granulocyte colony-forming units (CFU-G). Type 2 colonies treated with the combination of G-CSFa1 and G-CSFb1

FIGURE 7 | Recombinant G-CSFa1 and G-CSFb1 induces chemotactic response of kidney neutrophils. Chemotactic response of kidney neutrophils after 1 h of incubation with duplicate filters separating cells and cytokines at the concentrations indicated. Cells were stained with MGG and the total number of cells in 20 random fields of view (40× magnification) was determined. Medium and 10 ng/mL fMLP served as negative and positive controls, respectively. Equal concentrations (100 ng/mL) of cytokines in the upper and lower chemotaxis chambers served as chemokinesis control. The data represent mean + standard deviation (*n* = 4). Significant differences compared to medium control were determined using one-way ANOVA followed by Dunnet's *post-hoc* test, (*p* < 0.05) are denoted by asterisks (\*).

highly transcribed not only neutrophil-specific marker genes but also monocyte/macrophage lineage markers csf1r and irf8, suggesting that type 2 colonies are derived from the progenitors corresponding to mammalian granulocyte/macrophage CFU (CFU-GM) (**Figure 6E**).

### Granulocyte/Macrophage Progenitors and Granulocyte Progenitors Are Localized in the Head Kidney and Trunk Kidney but Not in the Spleen of Carp

To assess the contribution of hematopoietic organs to the neutrophil development in common carp, a myeloid colony forming assay was performed. Leukocytes were harvested from head kidney, trunk kidney, spleen and peripheral blood of adult carp (10 to 15 cm in length). Approximately 1 × 10<sup>5</sup> cells were cultured in the methylcellulose/agarose media in the presence or absence of 100 ng/mL G-CSFa1 plus 100 ng/mL G-CSFb1 and colony counts were performed after 6–11 days in the culture. PBLs and splenocytes did not form any colonies regardless of addition of cytokine or not. Conversely, cells from head kidney and trunk kidney formed about 25 to 40 colonies of each of type 1 and type 2 in the presence of both G-CSFa1 and G-CSFb1. Total number of CFU-G and CFU-GM in each organ was estimated as **Table 1**.

FIGURE 9 | Recombinant G-CSFa1 and G-CSFb1 induces increased respiratory burst capability. Respiratory burst capability of kidney neutrophils after pre-treatment with the medium, 100 ng/mL G-CSFa1 or 100 ng/mL G-CSFb1 for 6 h and subsequently treated with or without 100 ng/mL PMA for 30 min in the presence of DHR123. Mean of DHR123 fluorescence intensity (MFI) in gated neutrophil population was measured by flow cytometry. Data points are presented as mean values of individuals and error bars show standard deviation. Kidney neutrophils were obtained from four fish. Significant differences compared to every other group with two factors of G-CSF pre-treatment and PMA treatment were determined using two-way ANOVA followed by Tukey's *post-hoc* test, (*p* < 0.05) are denoted by asterisks (\*). N.S. represents 'not significant'.

### G-CSFa1 and G-CSFb1 Directly Induce a Chemotactic Response of Kidney Neutrophils and Up-Regulates the Gene Expression of a Chemokine Receptor cxcr1

Following the development of neutrophils at the sites of hematopoiesis, the migration and the recruitment of these cells toward the sites of infection or injury is essential for an efficient inflammatory response. We investigated the chemotactic effect of recombinant G-CSFa1 and G-CSFb1 on kidney neutrophils from normal adult carp employing a blind-well chemotaxis apparatus (**Supplementary Figure S6**). Neutrophils migrated toward fMLP placed in the bottom chamber (**Figure 7**), consistent with previous reports (22). In the presence of high doses of G-CSFa1 or G-CSFb1, kidney neutrophils migrated toward the sources (**Figure 7**). The chemokinesis controls indicated that neutrophil migration was cytokine-gradient dependent, since the migration of neutrophils was similar to the medium control when the recombinants were placed in both upper and lower chemotaxis chamber (**Figure 7**).

To assess the ability of G-CSFa1 and G-CSFb1 to modulate the gene expression of chemokine receptors, kidney neutrophils were treated with medium, G-CSFa1 or G-CSFb1 for 6 h. Teleost CXCR1 and CXCR2 are conserved receptors for interleukin-8 (IL-8, also termed CXCL8) and are important for the regulation of neutrophil recruitment and migration to sites of infection and injury (44, 45). Cxcr1 mRNA levels in neutrophils treated with G-CSFa1 and G-CSFb1 were significantly up-regulated compared to the medium control, indicating that both enhances a chemotactic sensibility of neutrophils toward chemotactic mediators such as IL-8 (**Figure 8A**). Neither cxcr2 mRNA levels in neutrophils treated with G-CSFa1 nor G-CSFb1 were changed compared to the medium control (**Figure 8A**). The mRNA levels of cxcr4 encoding a receptor for stromal cell-derived factor 1 (SDF-1, also termed CXCL12) in neutrophils were not modulated with the treatment of G-CSFa1 and G-CSFb1 (**Figure 8A**).

### G-CSFa1 and G-CSFb1 Enhance the Respiratory Burst Capacity in Kidney Neutrophils Through Up-Regulation of a NADPH Oxidase Component p47phox

The respiratory burst in neutrophils is the result of the formation of superoxide anions, in a process catalyzed by NADPH-oxidase (46, 47). Fish NADPH-oxidase components have been shown to have similar modes of activation and functional activities to mammalian counterparts (41, 48). To assess if the NADPH oxidase is induced by G-CSFa1 and G-CSFb1 treatments, we measured the mRNA levels of NADPH oxidase components (gp91phox , p22phox , p47phox , p67phox , and p40phox) in neutrophils treated with G-CSFa1 and G-CSFb1 for 6 h. mRNA levels of p47phox in neutrophils treated with G-CSFa1 and G-CSFb1 and p40phox in neutrophils treated with G-CSFb1 were significantly increased compared to the medium control. In contrast, mRNA levels of other components were not significantly changed (**Figure 8B**).

Furthermore, in order to investigate whether the treatment of carp kidney neutrophils with G-CSFa1 or G-CSFb1 induces their priming to prepare antimicrobial responses, we measured the respiratory burst in PMA-stimulated neutrophils. Neutrophils were pre-treated with the medium, G-CSFa1 or G-CSFb1 for 6 h. Following these treatments, neutrophils were treated with or without PMA in the presence of DHR123 and then analyzed by flow cytometry. Neither treatment of neutrophils with G-CSFa1 nor G-CSFb1 directly induced the respiratory burst without PMA stimulation (**Figure 9**). Both G-CSFa1 and G-CSFb1 significantly up-regulated the respiratory burst in PMAstimulated neutrophils compared to the medium control, while the enhancement of respiratory burst in neutrophils treated with G-CSFb1 was higher than that of G-CSFa1 treated neutrophils (**Figure 9**), which is consistent with the result of the upregulation of p47phox enhancement (**Figure 8B**).

### In vivo Administration of G-CSFa1 and G-CSFb1 Increases Circulating Neutrophils

Following intraperitoneal (i.p.) injection of PBS and repeated bleeding, the population of peripheral blood neutrophils did not change for 24 h. In contrast, i.p. injection of G-CSFa1 induced a significant increase of peripheral blood neutrophils 6 and 24 h after injection. Likewise, the population of peripheral blood neutrophils was significantly increased after 6 and 24 h of G-CSFb1 injection. At 24 h, G-CSFb1 injection had induced a significantly higher circulating number of neutrophils than injection with G-CSFa1 (**Figure 10**). However, at 48 h after G-CSFa1 injection, neutrophil numbers no longer were higher than those of unhandled or PBS-injected fish, probably due to the repeated bleedings affecting the peripheral blood neutrophil population of the control groups (**Supplementary Figure S7**).

### DISCUSSION

Here we cloned and functionally characterized carp G-CSF. All four carp g-csf genes contain five exons separated by four introns, retaining the gene structure found in human and mouse G-CSF as well as G-CSF of other teleost species (13). Carp and human G-CSF molecules share a similar structure of a signal peptide and a four-plus-one helical Pfam IL6/GCSF/MGF domain. All teleost fish G-CSF molecules share relatively high homology between each other at the helical regions. They also share acidic residues such as Asp and Glu, involved in the ligand-receptor binding, with mammalian G-CSF (49, 50). The four carp G-CSF paralogs identified in carp may have arisen from an ancestral G-CSF molecule through a series of duplications, including the teleostspecific 3rd WGD event and the carp-specific 4th WGD event (16, 21, 51–53). Overall, despite the overall low homology of teleost fish G-CSF sequences with mammalian G-CSF molecules, our in-silico analyses provide clear evidence that all four paralogs identified in carp are indeed orthologs of mammalian G-CSF.

Carp and other teleost fish G-CSF paralogs share only limited conservation of cysteine residues responsible for disulfide bonds with tetrapod G-CSF. Carp G-CSFa1 and G-CSFb1 express two structural differences: (i) an additional helix enriched with acidic residues in G-CSFa1 and (ii) a location of conserved cysteine residues. These structural differences prompted us to further investigate function of the different paralogs. Where g-csfa1 and g-csfa2 were highly expressed at basal level especially in spleen, g-csfb1 and g-csfb2 basal expression levels were very low in all tissues examined, indicating that g-csf transcription is differentially regulated between paralogs. Similarly, basal gcsfa gene expression was markedly higher than g-csfb expression in all immune cells examined. Macrophages are known to be the major cellular source of mammalian G-CSF (2). Strikingly,

G-CSF paralogs were always highest expressed in macrophages of carp.

Basal gene expression levels of g-csfa1 in macrophages were higher than those of g-csfa2 and gene expression levels of g-csfb1 in macrophages were higher than those of g-csfb2, which prompted us to further investigate function of G-CSFa1 and G-CSFb1 by production of these molecules as recombinant proteins. Recombinant proteins were produced in a bacterial expression system with the limitation that proteins are non-glycosylated and could possibly be contaminated with traces derived from bacteria. However, previous studies on mammalian G-CSF reported that glycosylation is not required for its activity and indeed, the non-glycosylated form is utilized in recombinant therapeutics (54). Even though the relative insensitivity to LPS has been reported in fish living in the aquatic environment with high pathogenic pressure (55), the recombinant proteins used in our assays were extensively purified up to the absence of LPS traces. Similar to mammalian G-CSF and zebrafish G-CSF (16, 56), carp G-CSF induced proliferation of hematopoietic cells in a dosedependent manner, albeit with apparent different activities for the two paralogs studied: G-CSFa1 induced proliferation of blast-like cells adhesive to culture dishes, whereas G-CSFb1 induced proliferation of cells with neutrophil characteristics. Indeed, treatment with G-CSFb1 showed up-regulation of the transcription factor cebpα involved in neutrophil development (34). Also, we investigated at least two carp G-CSF receptor genes (data not shown) and found that only G-CSFb1 enhanced gcsfr1 and myeloperoxidase (mpx) gene expression. Our data indicate that G-CSFb1 and G-CSFR1 are the main players involved with neutrophil development in carp. In zebrafish, both G-CSFa and G-CSFb may bind to the G-CSF receptor, expressed in both neutrophils and macrophages, and promote cell proliferation (16). In contrast to the latter study, carp G-CSFa1 alone did not stimulate colony formation in our semisolid culture system, in which an agarose layer prevented natural formation of a stromal and an adherent cell layer. This possibly restricted access to spontaneously secreted growth factors from adherent macrophages (57), which are possibly required for colony formation. Further studies would be required to determine if G-CSFa1 directly induces macrophages to produce autocrine growth factors or that G-CSFa1 synergizes with some factors spontaneously secreted from adherent macrophages to synergistically stimulate macrophage development. Meanwhile, carp G-CSFb1 alone did stimulate CFU-G colony formation, whereas the combination of G-CSFa1 and G-CSFb1 stimulated formation of not only CFU-G but also CFU-GM colonies. Our data indicate that carp G-CSFb1 may drive granulopoiesis restricted to neutrophil-lineage development, whereas carp G-CSFa1 may be a cytokine with proliferative effect stimulating CFU-GM or earlier stem/progenitor cells. The functional differences between the G-CSFa1 and G-CSFb1 cytokine preparations make it highly unlikely that the induced cell responses could be due to traces of bacterial contaminations and thus appear cytokine-specific. No matter the indicative differences in biological function between paralogs, carp G-CSFs appears to act as a hematopoietic growth factors.

Mammalian G-CSF is chemo-attractive to neutrophils (58, 59). In zebrafish, G-CSFb but not G-CSFa could be linked to in vivo trafficking of neutrophils to the site of severe injury (19). Our results indicate that carp kidney neutrophils are strongly attracted to G-CSFb1 and are moderately attracted to G-CSFa1, possibly under influence of IL-8 (or CXCL8) (40, 60, 61). Indeed, treatment of carp kidney neutrophil with G-CSF paralogs showed a significant up-regulation of CXCR1 as the IL-8 receptor required for neutrophil recruitment, but not CXCR2 required for neutrophil reverse migration and resolution (45). Unlike mammalian G-CSF, carp G-CSF paralogs did not mediate transcription of CXCR4, important for retention of neutrophils in the hematopoietic tissue in mammalian models (62). In conclusion, carp G-CSFb appears to be the most important G-CSF paralog to induce neutrophil migration.

Once neutrophils receive inflammatory cytokine signals, they become "primed" and capable of promptly and vigorously exerting antimicrobial responses (63). We could not find a significant change of phagocytic activity in neutrophils against beads and zymosan particles following stimulation of G-CSF paralogs for any period tested (data not shown), indicating that neutrophil phagocytosis is regulated by other signals in fish. Although mammalian G-CSF alone is not able to initiate a respiratory burst in naïve neutrophils, pre-incubation with this cytokine primes the cell for an enhanced superoxide anion production following stimulation with physiological stimuli such as fMLP and PMA (11, 64). In fish, following stimulation of phagocytes with inflammatory cytokines, ROS production is activated through at least three sequential steps: (i) activation of protein kinase C (PKC), (ii) phosphorylation of p47phox , and (iii) the production of ROS catalyzed by the NADPH oxidase complex (48). In our hands, expression analysis of NADPH oxidase components in neutrophils treated with carp G-CSF paralogs exhibited up-regulation of especially p47phox , indicating that the priming effect of carp G-CSF paralogs on neutrophils was regulated through the increase of p47phox. Our study provides the first report in teleost fish on the priming effects of G-CSF on neutrophils and analysis of respiratory burst indicated that G-CSFb1 primed neutrophils more effectively than G-CSFa1.

Previous studies in cyprinids showed that circulating blood neutrophils increased in number 6 to 18 h after i.p. injection with killed E. coli or zymosan and then quickly decreased after 24 h, indicating that an intraperitoneal inflammation in fish induces a temporal mobilization of kidney-derived neutrophils into the circulation (42, 65). In our study, administration of G-CSF paralogs increased the number of circulating blood neutrophils 6 and 24 h after i.p. injection, suggesting that also in vivo G-CSFa1 and G-CSFb1 work as chemoattractants and granulopoietic growth factors, in agreement with the in vitro results. However, it remains unclear whether in vivo excess of G-CSF paralogs induce the expression of other inflammatory cytokines and/or chemokines in immune cells. In human clinical medicine, recombinant G-CSF is used as a biophylatic agent to specifically induce granulopoiesis in patients with chemotherapy- and radiation-induced neutropenia to prevent bacterial and fungal infections (66). Further studies will be required to investigate if fish G-CSF paralogs can act as biophylatic agents against infectious diseases in a similar way. Here, functional analyses were limited to G-CSFa1 and G-CSFb1, and we can only speculate that G-CSFa2 and G-CSFb2 could have similar, different, or combinatorial functions in common carp. Additional biochemical investigations involving all native carp G-CSF paralog proteins will be needed to elucidate the full and complicated picture of immune regulation in this polyploid species.

In summary, we identified four carp G-CSF paralogs, studied their gene expression patterns and characterized the functional differences between A and B types of G-CSF on carp hematopoietic cells and neutrophils. We report important differences in their regulation: A type G-CSFs have a relatively high constitutive gene expression and could thus be involved in maintenance of a homeostatic state, whereas B type G-CSFs have a low gene expression and require induction and could thus have a responsive, immunological role associated with a state of infection. In general, G-CSFa1 alone stimulates proliferation of granulocyte/macrophage progenitors, while G-CSFb1 promotes proliferation, differentiation and colony formation of granulocyte/macrophage progenitors and granulocyte progenitors in kidney of carp, similar to the G-CSF mammalian counterpart. G-CSFa1 and G-CSFb1 act as chemo-attractants to neutrophils modulating the expression of the chemokine receptor CXCR1, suggesting a role for G-CSF paralogs in neutrophil trafficking. Both, G-CSFa1 and G-CSFb1 appear to induce neutrophil "priming." The carp G-CSF paralogs reported herein provide us with valuable tools to further study the immune system of teleost fish.

### REFERENCES


### AUTHOR CONTRIBUTIONS

FK and KN: conceived and designed the experiments; FK, KN, AW, and EH: performed the experiments; FK, KN, AW, JM, and MO: analyzed the data; FK, KN, AW, GW, and TM: wrote and edited the paper.

### FUNDING

This work was supported in part by a Grant-in Aid for Young Scientists (B) (16K18752) from the Japan Society for the Promotion of Science (JSPS) to FK and a grant of International joint research and training of young researchers for zoonosis control in the globalized world (S1491007) from the Ministry of Education, Culture, Sports, Science and Technology (MEXT) Japan. AW was funded by the European Commission under the 8th (H2020) Framework Program for Research and Technological Development of the European Union (PARAFISHCONTROL Grant No. 634429).

### ACKNOWLEDGMENTS

The authors greatly appreciate Dr. Maria Forlenza and Mr. Jules Petit (Wageningen University) for helpful discussion and advice.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2019.00255/full#supplementary-material


cytokines and their roles in hematopoietic development and maintenance. Blood (2013) 122:3918–28. doi: 10.1182/blood2012-12-475392


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

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

# Studies Into β-Glucan Recognition in Fish Suggests a Key Role for the C-Type Lectin Pathway

Jules Petit <sup>1</sup> , Erin C. Bailey 2,3, Robert T. Wheeler 2,3, Carlos A. F. de Oliveira<sup>4</sup> , Maria Forlenza<sup>1</sup> and Geert F. Wiegertjes 1,5 \*

<sup>1</sup> Cell Biology and Immunology Group, Wageningen University & Research, Wageningen, Netherlands, <sup>2</sup> Department of Molecular & Biomedical Sciences, University of Maine, Orono, ME, United States, <sup>3</sup> Graduate School of Biomedical Sciences and Engineering, University of Maine, Orono, ME, United States, <sup>4</sup> Department of Research and Development, Biorigin Company, Lençóis Paulista, Brazil, <sup>5</sup> Aquaculture and Fisheries Group, Wageningen University & Research, Wageningen, Netherlands

#### Edited by:

Leon Grayfer, George Washington University, United States

#### Reviewed by:

Pierre Boudinot, Institut National de la Recherche Agronomique (INRA), France Mark D. Fast, University of Prince Edward Island, Canada

\*Correspondence:

Geert F. Wiegertjes geert.wiegertjes@wur.nl

#### Specialty section:

This article was submitted to Comparative Immunology, a section of the journal Frontiers in Immunology

Received: 25 October 2018 Accepted: 01 February 2019 Published: 26 February 2019

#### Citation:

Petit J, Bailey EC, Wheeler RT, de Oliveira CAF, Forlenza M and Wiegertjes GF (2019) Studies Into β-Glucan Recognition in Fish Suggests a Key Role for the C-Type Lectin Pathway. Front. Immunol. 10:280. doi: 10.3389/fimmu.2019.00280 Immune-modulatory effects of β-glucans are generally considered beneficial to fish health. Despite the frequent application of β-glucans in aquaculture practice, the exact receptors and downstream signalling remains to be described for fish. In mammals, Dectin-1 is a member of the C-type lectin receptor (CLR) family and the best-described receptor for β-glucans. In fish genomes, no clear homologue of Dectin-1 could be identified so far. Yet, in previous studies we could activate carp macrophages with curdlan, considered a Dectin-1-specific β-(1,3)-glucan ligand in mammals. It was therefore proposed that immune-modulatory effects of β-glucan in carp macrophages could be triggered by a member of the CLR family activating the classical CLR signalling pathway, different from Dectin-1. In the current study, we used primary macrophages of common carp to examine immune modulation by β-glucans using transcriptome analysis of RNA isolated 6 h after stimulation with two different β-glucan preparations. Pathway analysis of differentially expressed genes (DEGs) showed that both β-glucans regulate a comparable signalling pathway typical of CLR activation. Carp genome analysis identified 239 genes encoding for proteins with at least one C-type Lectin Domains (CTLD). Narrowing the search for candidate β-glucan receptors, based on the presence of a conserved glucan-binding motif, identified 13 genes encoding a WxH sugar-binding motif in their CTLD. These genes, however, were not expressed in macrophages. Instead, among the β-glucan-stimulated DEGs, a total of six CTLD-encoding genes were significantly regulated, all of which were down-regulated in carp macrophages. Several candidates had a protein architecture similar to Dectin-1, therefore potential conservation of synteny of the mammalian Dectin-1 region was investigated by mining the zebrafish genome. Partial conservation of synteny with a region on the zebrafish chromosome 16 highlighted two genes as candidate β-glucan receptor. Altogether, the regulation of a gene expression profile typical of a signalling pathway associated with CLR activation and, the identification of several candidate β-glucan receptors, suggest that immune-modulatory effects of β-glucan in carp macrophages could be a result of signalling mediated by a member of the CLR family.

Keywords: β-glucan, primary macrophage, transcriptome analysis, RNAseq analysis, CTLD, C-type lectin-like domain, teleost, Cyprinidae

## INTRODUCTION

Immunomodulation by β-glucans has been widely studied in teleost fish. Regardless of the administration route or fish species, β-glucan administration often has an immune stimulatory effect and can result in increased resistance to both viral and bacterial infections [reviewed by (1–3)]. For both, mammalian vertebrates (4, 5) and invertebrates (6, 7), specific mechanisms responsible for β-glucan recognition and/or downstream signalling have been described. Yet for teleost fish, despite the frequent application of β-glucans in aquaculture practice, the exact mechanisms underlying the induced effects are ill described.

In mammals, several non-exclusive pathways play a role in the recognition and down-stream signalling after β-glucan stimulation, however the best-described β-glucan receptor is Dectin-1, also known as C-type Lectin domain Family 7 member A (CLEC7A). Dectin-1 is a C-type lectin super family V, NK cell receptors transmembrane receptor with a single carbohydrate recognition domain (CRD) and a cytoplasmic tail containing one ITAM motif (8–10). Dectin-1 is predominantly expressed on cells from both the monocyte/macrophage and neutrophil lineages, where it acts as the major β-glucan receptor (11). Ligation of β-(1,2)-glucan to Dectin-1 is dependent on two specific amino acid residues in the CRD; tryptophan (W) and histidine (H), separated by a third residue (WxH motif) (12). Additionally, a tyrosine (Y) residue separated from histidine in the CRD by four residues (WxHxxxxY motif) is crucial for shaping the β-glucan binding cleft (13). This β-glucan binding cleft is formed by spatial arrangement of tryptophan, histidine and tyrosine in a triangular fashion resulting in a shallow hydrophobic surface groove, capable of accommodating and binding β-glucan chains through hydrophobic interactions (12–14). Also true for invertebrates, the same three residues are present in the binding domain of β-(1, 3)-glucan recognition protein (GNBP3), although not as a conserved motif (15). Dectin-1 signalling is activated following clustering in synapse-like structures formed within minutes after activation by particulate β-glucans (16). Following the interaction with β-glucan, Dectin-1 signals via Syk kinase and the adaptor protein Card9 to send a downstream signal through Bcl10 and Malt1 to the transcription factor NF-κB (17). Activation of NF-κB leads to an inflammatory profile typical of stimulation with β-glucans.

The presence or absence in fish genomes of Dectin-1, along with the entire superfamily V of CLRs (NK cell receptors), is debated (18). Although initially a cKLR from Paralabidochromis chilotes was described as a member of the superfamily V (19), and 28 distinct KLR loci identified a particular chromosomal region in Nile tilapia (Oreochromis niloticus) (20), a subsequent and thorough phylogenetic analysis suggested these receptors to be members of superfamily II rather than superfamily V (18). At the same time, it is important to realize that this phylogenetic analysis was primarily based on an early genome assembly of a single fish species, Fugu rubripes (18), indicating that the presence of true superfamily V CLR members in fish genomes has not been systematically investigated in other teleosts.

Functional and therefore indirect evidence of the presence of β-glucan receptor(s) in fish exists for at least Atlantic salmon (Salmo salar) macrophages, channel catfish (Ictalurus punctatus) neutrophils and seabream (Sparus aurata) leukocytes, based on observations that pre-treatment of these cell types with βglucans reduces the uptake of yeast (Saccharomyces cerevisiae) glucan particles, zymosan or whole yeast cells (21–23). In carp, injection of β-glucans induced a complement receptor 3 (CR3)-dependent rosette formation of leukocytes and deposition of iC3b and C3d fragments on zymosan (24, 25). β-glucan was also shown to regulate the expression of several pattern recognition receptors including tlr2 in primary macrophages of European eel (Anguilla anguilla) (26), regulation of tlr3 and cxc receptors in common carp (27, 28), the purinergic receptor p2x4 in Japanese flounder (Paralichthys olivaceus) (29) and nod2 and tlr2 in zebrafish (Danio rerio) (30). Although these genes were regulated by β-glucan stimulation, their involvement in the recognition of β-glucans remains to be confirmed. Oral intubation of Atlantic salmon (S. salar) with β-glucan resulted in the up-regulation of syk kinase and three salmon CLRs with one or two ITAMs and a single WxH motif (31, 32), alike the Dectin-1 architecture described above. We reasoned that macrophages could provide an informative starting point for investigating signalling pathways induced upon β-glucan stimulation. Indeed, primary macrophages of common carp have been shown to respond to curdlan and to zymosan depleted of Toll Like Receptor (TLR) stimulating properties; both considered Dectin-1-specific ligands in mammals (33). We therefore hypothesized that immune-modulatory effects of βglucan in carp macrophages could be triggered by an unknown member of the CLR family, different from Dectin-1.

Here, we used primary macrophages of common carp for a whole transcriptome analysis of differentially expressed genes (DEG) induced by two different β-glucans. Analyses of gene ontology revealed comparable profiles and a clear regulation of the CLR pathway by both β-glucans. Subsequent investigation of the common carp genome for candidate β-glucan receptors identified a number of genes based on their architecture or expression profile, all encoding proteins with at least one C-type Lectin Domain (CTLD). Preliminary phylogenetic analysis of the CTLD sequences of candidate proteins showed no clustering with CTLD sequences of known group V members. Synteny analysis of the genome of zebrafish, a close relative of common carp, identified two CTLD-encoding genes with apparent conservation with mammalian CLR group V members, namely clec4c and sclra. Overall, our study identifies several teleost CLRs of interest for future functional studies aimed at further specifying the modulatory effects of β-glucans on the fish immune system.

### METHODS

### Animals

European common carp (Cyprinus carpio carpio L.) of the R3 × R8 strain were used, which originate from a cross between the Hungarian R8 strain and the Polish R3 strain (34). Carp were bred and raised in the aquatic research facility of Wageningen University, Carus, at 20–23◦C in recirculating UV-treated water and fed pelleted dry feed (Skretting, Nutreco) twice daily. All experiments were performed with the approval of the animal experiment committee of Wageningen University.

### In vitro Culture of Head Kidney-Derived Macrophages

Fish were anaesthetized with 0.3 g/l Tricaine Methane Sulfonate (TMS) (Crescent Research Chemicals, Phoenix, USA) in aquarium water buffered with 0.6 g/l sodium bicarbonate and bled via the caudal vein. Carp head kidney-derived macrophages were cultured for 6 days, as described previously (35), and will be referred to as macrophages.

### Macrophage Stimulation

Macrophages were harvested by placing culture flasks on ice for 15 min and by gentle scraping. Cell suspensions were centrifuged at 450× g for 10 min at 4◦C. Macrophages were resuspended in complete NMGFL (incomplete-NMGFL-15 medium supplemented with 2.5% heat-inactivated pooled carp serum and 5% bovine calf serum (Invitrogen Life Technologies) with 100 U/mL of penicillin and 50µg/mL streptomycin) (35). Subsequently, macrophages were seeded in 24-well flatbottom culture plates (CorningTM 3526, FischerScientific) at 1.5 × 10<sup>6</sup> macrophages/300 µL per well. For stimulation of the cells, curdlan (C7821, Sigma Aldrich) (a high molecular weight linear polymer consisting of β-1-3-linked glucose residues from Alcaligenes faecalis) and MacroGard <sup>R</sup> [a cell wall preparation of S. cerevisiae comprising 91% β-glucan (Zilor, São Paulo, Brazil)] were used (36). β-glucans were prepared as previously reported (33). Cells were stimulated with β-glucan preparations at 25µg/mL, a concentration at which both β-glucan preparations were previously shown to induce considerable nitric oxide production in carp macrophages (33). For each stimulus at least three independent cultures were used and each stimulation was performed in technical triplicate.

After 6 h of stimulation, three replicate wells were pooled and 4.5 × 10<sup>6</sup> macrophages were lysed in 350 µL RLT buffer (QIAgen, Netherlands) and stored at −80◦C until RNA extraction. Total RNA was extracted using the RNeasy Mini kit according to the manufacturer's protocol (QIAgen) including on-column DNase treatment with the RNase-free DNase set (QIAgen). RNA was stored at −80◦C until use.

### Illumina Sequencing and Data Analysis

RNA quality and concentration was checked on a Bioanalyzer (Agilent 2100 total RNA Nano series II chip, Agilent). RNAseq libraries were prepared from 0.5 µg total RNA using the TruSeq <sup>R</sup> Stranded mRNA Library Prep kit according to the manufacturer's instructions (Illumina Inc. San Diego, CA, USA). Similar to the previous carp study (37), all RNAseq libraries were sequenced on an Illumina HiSeq2500 sequencer as 1 × 50 nucleotides singleend reads. Image analysis and base calling were performed using the Illumina pipeline. Using TopHat (version 2.0.5) (38), reads were aligned to the latest published genome assembly of common carp (BioProject: PRJNA73579) (37). For each independent sample at least 10 million raw reads were sequenced, on average 65% of the raw reads could be mapped to annotated genes of this carp genome assembly. Secondary alignments of reads were excluded by filtering the files using SAMtools (version 0.1.18) (39). Aligned fragments per predicted gene were counted from SAM alignment files using the Python package HTSeq (version 0.5.3p9) (40).

Differential gene expression was analysed using the bioinformatics package DESeq 2.0 (v1.22.2) (41) or edgeR (v3.24.3) (42, 43) from Bioconductor (v3.8) (44) in R statistical software (3.1.2) (45). Statistical analysis was performed using a paired design with unstimulated cells as control and performed for curdlan and MacroGard <sup>R</sup> independently (n = 3 independent cultures for curdlan and for MacroGard <sup>R</sup> ). The paired design allowed for a better comparison between independent cultures, reducing noise generated by culture to culture differences (41). For DESeq 2.0, p-values were adjusted using Benjamini & Hochberg corrections for controlling false discovery rate and results were considered statistically significant when p-adjusted ≤ 0.05. For edgeR, genes were considered significantly regulated if both p-value ≤ 0.05 and FDR ≤ 0.05. Only genes identified as significantly regulated by both, DESeq 2.0 and edgeR, were used for subsequent analyses (**Supplementary Tables 1, 2** for curdlan- and MacroGard <sup>R</sup> -DEGs, respectively). Venn diagrams were generated with the webtool from the University of Ghent Bioinformatics and Evolutionary Genomics group.<sup>1</sup>

### Gene Ontology Annotation and Enrichment Analysis

The common carp genome has been annotated against Ensembl zebrafish GRCz10 (37). Due to its tetraploid nature because of an additional genome duplication event (46–48), the common carp generally has two copies of each zebrafish gene. However, for gene ontology (GO) and KEGG analysis only single IDs were used, resulting in a dataset with unique Ensembl zebrafish IDs (curdlan DEGs n = 421 unique genes and MacroGard <sup>R</sup> DEGs n = 638 unique genes). Gene Ontology (GO) analysis was performed with GOrilla (49). Using differentially expressed genes as a target list and the entire list of annotated common carp genes as a background list, GO term enrichment was analysed (**Supplementary Tables 3, 4** for curdlan and MacroGard <sup>R</sup> , respectively). FDR q-values were calculated by adjusting pvalues using Benjamini & Hochberg method for controlling false discovery rate, GO terms were considered statistically enriched when FDR ≤ 0.05.

Independent KEGG analysis (50) was performed using the stable Ensembl zebrafish ID's of each differentially expressed gene in KOBAS v3.0 (51), using the well-annotated zebrafish genome as a reference list. KOBAS was run with Chi-square test and for FDR correction the Benjamini and Yekutieli method was used. Pathways were considered significantly over-represented if the corrected p-value was p ≤ 0.05. Recently, a zebrafishspecific KEGG pathway map for the CLR pathway was released (dre04625). As this map was not yet incorporated in the KOBAS analysis, we performed manual mapping using the "userdata mapping" feature on the zebrafish-specific KEGG pathway for C-type lectin receptor signalling.

<sup>1</sup>http://bioinformatics.psb.ugent.be/webtools/Venn/

### Genome Search for CTLD-Encoding Common Carp Sequences

The conceptual translation of all annotated carp gene sequences as submitted to NCBI (from here on referred to as carp proteins) [BioProject: PRJNA73579 (37)], was used to perform a Protein family (Pfam) domain search using CLC Main Workbench v8.0<sup>2</sup> with the Pfam Database v31 (52). Proteins without a C-type lectin domain (CTLD) (PF00059) were filtered out. Subsequently, using the PatmatDB feature from EMBOSS in the public Galaxy server (v5.0.0) at Wageningen University and Research Centre (WUR, The Netherlands<sup>3</sup> , protein sequences containing an immunoreceptor tyrosine-based activation motif (ITAM) were identified using the signature YxxL/I sequence. Presence of a WxH motif within the CTLD was investigated using the PatmatDB feature from EMBOSS in Galaxy (v5.0.0) using the signature WxH. Alternatively, presence of a WxHxxxxY motif within the CTLD was investigated using the signature WxHx(1, 4)Y sequence, allowing for 1–4 random residues between histidine and tyrosine. Transmembrane regions of all proteins with at least one CTLD, were predicted using TMHMM Server v. 2.0<sup>4</sup> . These analyses highlighted a restricted number of proteins with one or more CTLDs and characterized the number of ITAM motifs, WxH motifs and transmembrane regions present in the conceptual translations. Subsets of candidate receptors were selected from the restricted number of proteins based on the following three criteria: (1) presence of a conserved WxH motif in the CTLD; (2) corresponding expression of ≥50 reads per kilobase million (RPKM) in unstimulated macrophages (n = 5); (3) differential regulation of expression in carp macrophages stimulated with β-glucans. Automatic annotation of candidate receptors was manually verified using BLASTx against the nr database from NCBI.

The CTLD sequences of identified candidate receptors were aligned with CTLD sequences from selected fish CLRs, selected mouse CLRs, several chicken CLRs (53) and with CTLDs from human CTLD-encoding genes (PF00059) present in Ensembl (GRCh38.p12) using MUSCLE v3.8 (54). In case of more than one CTLD sequence per protein, the first CTLD was designated 1, the next 2, and so on. Subsequently, Model Selection feature of MEGA-X (55) was used to calculate the most appropriate amino acid substitution model using all sites of the alignment as input data. The evolutionary history was inferred by using the Maximum Likelihood (ML) method based on the Whelan and Goldman model (56) allowing for Gamma distribution (+G) with four Gamma categories and using the rate variation model allowed for some sites to be evolutionarily invariable [(+I), 0.19% sites]. The bootstrap consensus tree inferred from n = 500 replicates was taken to represent the evolutionary history of the taxa analysed (57). Initial tree(s) for the heuristic search were obtained automatically by applying Neighbor-Join and BioNJ algorithms to a matrix of pairwise distances estimated using a JTT model, and then selecting the topology with superior log likelihood value. A discrete Gamma distribution was used to

model evolutionary rate differences among sites (+G, parameter = 1.6778). The final alignment involved a total of n = 223 CTLD sequences.

As a fourth criterion for the identification of candidate receptors, conservation of synteny with the mammalian NK cell receptor cluster was added. Synteny analysis relied on genomic location data from NCBI and Ensembl Gene Summary databases. The GRCh38.p12, GRCm38.p6, and the Zv10 primary genome assemblies were used for human, mouse, and zebrafish genomic location data, respectively. For genomic location of carp genes the latest common carp assembly (37) was used.

### RESULTS

### β-Glucan Stimulation of Macrophages Leads to Regulation of the CLR Signalling Pathway

Transcriptome analysis of macrophages stimulated with curdlan retrieved a total of n = 528 differentially expressed genes (DEGs) (**Supplementary Table 1**) of which almost 85% were upregulated. Transcriptome analysis of macrophages stimulated with MacroGard <sup>R</sup> retrieved a total of n = 781 DEGs (**Supplementary Table 2**), of which almost 80% were upregulated. Subsequent comparison of the expression profile of both DEG datasets revealed a comparable profile, with n = 291 DEGs that followed concordant expression patterns, only in one case discordant regulation was observed as up-regulation by curdlan and down-regulated by MacroGard <sup>R</sup> (**Figure 1**).

Automated gene ontology (GO) analysis of the two DEG datasets using GOrilla could map ∼65% of the DEGs to GO terms. GO term enrichment analysis revealed 9 and 37 GO terms significantly enriched (FDR q-value ≤ 0.05) among curdlan-DEGs and MacroGard <sup>R</sup> -DEGs, respectively. For both datasets, the GO term with smallest FDR q-value was "Immune system process (GO:0002376)." Within the domain Biological Process, all significantly enriched GO terms in the curdlan-DEG dataset

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

<sup>3</sup>http://galaxy.wur.nl)

<sup>4</sup>http://www.cbs.dtu.dk/services/TMHMM/

TABLE 1 | Over-represented KEGG pathways following automated KEGG analysis of differentially expressed genes (DEGs) in macrophages stimulated with curdlan or MacroGard®.


Curdlan-DEGs (n = 356 genes) could be mapped to n = 92 pathways, of which n = 4 were significantly over-represented. MacroGard®-DEGS (n = 541 genes) could be mapped to n = 112 pathways of which n = 13 were significantly over-represented. KEGG analysis was performed with KOBAS v3.0 with Chi-square test and the Benjamini and Yekutieli method for FDR correction. Pathways were considered significantly over-represented at corrected p-value ≤ 0.05.

(n = 7), were also significantly enriched in the MacroGard <sup>R</sup> - DEG dataset (n = 25) (**Supplementary Tables 3, 4**), suggesting that the gene expression profile regulated by curdlan is also regulated by MacroGard <sup>R</sup> .

Automated annotation of the two DEG datasets to zebrafish KEGG pathways resulted in the mapping of the curdlan-DEGs to n = 92 different pathways and of the MacroGard <sup>R</sup> -DEGs to n = 112 different pathways. Four pathways were significantly over-represented in both DEG datasets, and a further nine unique pathways were significantly over-represented only in the MacroGard <sup>R</sup> -DEGs (**Table 1**). The overlap in the overrepresented KEGG pathways in the two datasets supports the notion that in carp macrophages, the profile induced by curdlan is also induced by MacroGard <sup>R</sup> .

Automated annotation did not include the zebrafish-specific KEGG map for the "C-type lectin receptor signalling pathway" (dre04625) which has become available only recently. Manual mapping of curdlan- and MacroGard <sup>R</sup> -DEGs showed regulation of a large number of genes associated with the C-type lectin receptor pathway (**Figure 2**). Further, not only end-products of the CLR signalling pathway, such as cytokines which are often shared among pathways, but also several upstream molecules, such as card9, bcl10, malt1, and calm were regulated (see **Figure 2**). The manual mapping suggests that in carp macrophages, both β-glucans induced expression of genes typically associated with C-type lectin receptor signalling.

### Search for Candidate Receptors

Following the above-described mapping of DEGs to the CLR pathway, we continued with screening all 50,527 conceptually translated carp proteins for the presence of at least one CTL domain (CTLD), narrowing down the search for candidate receptors to a total of n = 239 proteins. These CTLDcontaining proteins were further characterized for the presence of ITAM sequences and transmembrane helices (data not shown). Studies have shown that the presence of a WxH motif or a WxHxxxxY motif in the carbohydrate binding region of Dectin-1 determines the β-glucan binding capacities of Dectin-1 (12–14). This criterion was used to further narrow down the search for candidate receptors, identifying a subset of n = 13 carp proteins with a WxH motif (**Table 2**). We could not identify proteins with a WxHxxxxY motif present specifically in their CTLD. Interestingly, in four WxH-containing proteins, we could also identify a transmembrane region and one or two ITAMs and thus a "complete" Dectin-1-like architecture. However, based on the RNAseq analysis, none of these candidates was significantly regulated by β-glucan stimulation or were constitutively expressed (higher than 10 reads per kilobase million) in unstimulated macrophages (**Table 2**), suggesting that these 13 candidates could not likely explain the functional responses to β-glucans in our current experimental set up.

As the above-mentioned screening for WxH motif did not identify candidate receptors expressed or regulated in carp macrophages, we widened again our search and used constitutive expression or differential regulation as new criteria to narrow down the search for candidate receptors. Screening of all n = 239 CTLD-containing proteins for constitutive expression of their corresponding gene identified a subset of n = 12 genes that were expressed at an arbitrary threshold set at ≥ 50 Reads Per Kilobase Million (RPKM), corresponding to on average 1.5% of β-actin expression. Screening for regulation identified a subset of n = 6 candidate receptors as differentially expressed after stimulation with β-glucans, all of which overlapped with the expressed subset (**Table 3**). Of interest, all six were down-regulated by β-glucan stimulation and two out of the six candidate receptors (both Asialoglycoprotein receptor orthologues) were regulated by both β-glucans. The finding that these candidate receptors were all regulated could suggest involvement in the response to β-glucan simulation of carp macrophages.

Combining the subsets of candidate β-glucan receptors identified based on criteria 1 (WxH motif), 2 (RPKM ≥ 50) and 3 (differential regulation) resulted in a total of n = 25 proteins of specific interest, containing n = 39 CTLD sequences. To investigate the phylogenetic relation between these candidates of interest and known CLR family members, CTLD sequences were aligned. The phylogenetic tree revealed a clear subdivision of the cypCars related to the carp CTLD proteins. None of the CTLD sequences from the identified candidates clustered with family V members, the group containing Dectin-1 (**Figure 3**). Instead, seven CTLD sequences from WxH motif-containing candidates (criterion 1) clustered with human CLEC20A, among human CLR group II members, and seven other WxH-candidates clustered together with human Attractin. Maybe not surprisingly, four of the constitutively expressed CTLD sequences (criterion 2) clustered with CTLD sequences from macrophage mannose receptors, group VI family. Six CTLD sequences from the

differentially regulated subset (criterion 3) also clustered with macrophage mannose receptors, the other six CTLD sequences in this subset followed no specific clustering pattern. Overall, preliminary phylogenetic analysis suggests the absence of group V members in our set of candidate receptors.

### Synteny Analysis of Zebrafish and Carp CTLD-Encoding Genes

Investigation of the carp CTLD containing proteins identified several candidates with architecture similar to mammalian Dectin-1. In mammalian genomes Dectin-1 is located in the NK cell receptor cluster, which contains several CLRs. This NK cell receptor cluster shows conserved synteny between human chromosome 12 (hCHR12) and mouse chromosome 6 (mCHR6) (58). Conservation of synteny describes preservation of colocalization of genes on chromosomes of different species and might therefore serve as an additional criterion (criterion 4) for the identification for candidate receptors involved in the recognition of β-glucans. Synteny analysis can be best performed in a well-assembled genome. The genome assembly of zebrafish, a close relative of common carp, is among the best assembled genomes in teleost fish, including large chromosome scaffolds that can be used for synteny analysis, in contrast to the carp genome which is still largely fragmented into small scaffolds (37, 47). The NK cell receptor cluster in human includes, among others, DECTIN-1, DECTIN-2, and MINCLE. Regions surrounding the NK cell receptor cluster on hCHR12 and mCHR6 showed conserved synteny with regions on zebrafish chromosome 16 (zCHR16), based on co-localization of pex5, clstn3, lpcat3, gnb3, cops7a, and znf384 (**Figure 4**). Intriguingly, this region of zCHR16 also includes two CTLD-encoding genes, clec4c (NCBI Gene ID: 563797), and sclra (NCBI Gene ID: 564061), highlighting them both as genes of interest for our study.

Further investigation of synteny between the zebrafish region of interest on zCHR16 and carp scaffolds revealed partial conservation of synteny with four scaffolds. Two of these scaffolds contained cypCar\_00024225 and cypCar\_00032523, putative paralogues of zebrafish clec4c; of which cypCar\_00024225 is a CTLD-encoding gene, already identified in this study as candidate receptor based on its regulation in macrophages after stimulation with MacroGard <sup>R</sup> (criterion 3). The two other scaffolds contained cypCar\_00016746 and cypCar\_00029396, putative paralogues of zebrafish sclra; both of which are CTLD-encoding carp genes, already identified in this study as candidate receptors based on their conserved WxH-motif (criterion 1). The synteny analysis provided a fourth criterion, additional to the previously formulated criteria (WxH motif, constitutive expression and differential regulation), to identify candidate receptors for β-glucan recognition. A

<sup>5</sup>https://www.genome.jp/kegg-bin/show\_pathway?dre04625

#### TABLE 2 | Candidate receptors with a WxH signature sequence in their CTL domain.


Description refers to the closest BLASTx hit upon blasting the identified genes of interest and accession numbers refer to protein sequences in NCBI. Number of conserved CTL domains (CTLDs), ITAM sequences and transmembrane domains (TM) in the protein sequence are included. RPKM refers to the average (n = 5) reads per kilobase million in unstimulated macrophages and is a measure for gene expression. cypCar codes identify common carp genes (Cyprinus carpio). Names or numbers in brackets refer to genes with identical BLASTx hits (likely owing to the additional genome duplication event in common carp). Proteins with architecture similar to Dectin-1 are highlighted in bold. Candidate receptors are ordered numerically by cypCar code.

TABLE 3 | Candidate receptors expressed in carp macrophages.


Candidates were selected based on a minimal expression of ≥ 50 RPKM in unstimulated carp macrophages or differential expression following stimulation of macrophages with βglucans. Description refers to the closest BLASTx hit, with associated NCBI protein accession code. Number of conserved CTL domains (CTLDs), ITAM sequences and transmembrane domains (TM) in the protein sequence are included. RPKM refers to the average (n = 5) reads per kilobase million in unstimulated macrophages and is a measure for gene expression. CRD and MG refer to differentially expressed genes in macrophages after stimulation with curdlan (CRD) or MacroGard® (MG) based on RNAseq analysis (highlighted in bold). cypCar codes identify common carp genes (Cyprinus carpio). Names or numbers in brackets refer to genes with identical BLASTx hits (duplicated genes). Candidate proteins are ordered based on their expression (RPKM).

Venn diagram, graphically representing the different subsets of candidate β-glucan receptors identified by the four criteria can be found **Supplementary Figure 1**.

### DISCUSSION

Primary macrophages of common carp had previously been shown to respond to prototypical Dectin-1 ligands, which led to the hypothesis that the CLR pathway must play an important role in the recognition of β-glucans in carp macrophages. In our approach, we used head kidney-derived carp macrophages as a starting cell population to test our hypothesis and investigate activation of the CLR pathway upon stimulation with βglucans. Indeed, pathway analysis of differentially expressed genes confirmed our hypothesis that β-glucans regulate a downstream signalling pathway typical of CLR activation. Further, we could identify in the transcriptome of β-glucanstimulated carp macrophages, several differentially expressed genes with a C-type lectin domain. These data are of high interest for further functional studies on the mechanisms underlying β-glucan-induced immunomodulation in teleost fish.

We used two different β-glucans: curdlan, a linear polymer of β-(1,3)-linked glucose and considered a Dectin-1-specific ligand, and MacroGard <sup>R</sup> , a branched polymer of β-(1,3/1,6) glucose widely-applied as feed additive in aquaculture. Overall, MacroGard <sup>R</sup> regulated a higher number of differentially expressed genes than curdlan, possibly owing to differences in purity, source, degree of polymerization, and nature of the glycosidic bonds in the β-glucans (59). Regardless of the extent of gene regulation, manual mapping of the DEGs revealed a clear regulation of the CLR signalling pathway (KEGG)

for both β-glucans. Indeed, up-regulation of homologues of all three players of the card9-Bcl10-Malt1 complex, previously shown to play a crucial role in β-glucan signalling through the CLR pathway (17), strongly supports regulation by the CLR pathway. We continued our study by identifying candidate genes encoding for proteins with one or more C-type Lectin Domains (CTLD), which could be of potential interest with respect to recognition of β-glucans, using a recently published database of RNAseq-validated gene predictions for carp (37). We used four criteria to identify candidate receptors in carp

macrophages: (1) conservation of the glucan binding WxHmotif in the CTLD; (2) constitutive expression higher than 50 RPKM; (3) differential regulation upon stimulation with β-glucans; (4) conservation of synteny with mammalian NK cell receptor cluster.

Based on criterion 1 (conserved WxH motif) we identified two candidates (cypCar\_00016746 and cypCar\_00029396) of which the CTLD clustered together with known CTLD sequences from zebrafish and from Atlantic salmon, known as salmon C-type lectins sclra and sclrb (31). Although the carp sclr paralogues were

scale. Owing to the tetraploid nature of carp, two corresponding scaffolds for each zebrafish gene are shown.

not constitutively expressed in macrophages (criterion 2), both salmon sclrs have been associated with the response to β-glucans (32). Interestingly, the mammalian WxHx[4]Y motif was not conserved, however a motif with five rather than four residues (WxHx[5]Y) separating histidine from tyrosine was conserved between all sequences. All three residues are considered crucial to form the β-glucan binding cleft of mammalian Dectin-1 (13), and also present in invertebrate β-glucan binding proteins (GNBP3), but not as a WxHx[5]Y motif (13, 15). Although not constitutively expressed in macrophages, carp sclr could well play a role in β-glucan binding in other cell types.

Based on criterion 2 (constitutive expression ≥ 50 RKPM), we identified a further 13 candidate receptors, of which six were differentially regulated (criterion 3). Without exception, all six genes were down-regulated upon stimulation with βglucans, which could possibly be explained by a need to restrict de novo protein synthesis and duration of signalling to prevent over-stimulation (60–63). Possibly, analysis of protein and/or gene expression at different time points could show upregulation. In Atlantic salmon, three sclrs were up-regulated 7 days after oral administration of MacroGard <sup>R</sup> , concomitantly with syk kinase (32). We similarly noticed a concomitant regulation of syk, suggesting the expression of CTLD-encoding genes and syk is co-regulated. No matter what, the observed modulation of gene expression strongly suggests involvement of CLR family members upon recognition of β-glucans by carp macrophages.

Based on criterion 3 (regulation of gene expression), we identified several additional candidate receptors. Among these, the CTLD-encoding gene with the highest expression and regulation in carp macrophages (cypCar\_00031274), clustered together with three other genes of interest: (i) a CTLD-encoding gene already identified in this study based on criterion 2 (cypCar\_00044344), (ii) an unknown zebrafish full length cDNA (zgc174904) encoding a protein with a CTLD, a transmembrane helix and a WxHx[5]Y motif just outside the borders of the CTLD and (iii) an Atlantic salmon C-type lectin receptor-c (sclrc), different from the ones previously mentioned in criterion 1. The salmon sclr genes were first identified in (suppression subtractive) EST libraries (31), while a follow up study revealed up-regulation of all three sclr genes after oral intubation of Atlantic salmon with MacroGard <sup>R</sup> (32). The salmon sclrc gene contains a WxH motif within the CTLD while the carp candidate genes (both cypCar\_00031274 and cypCar\_00044344) contain a WxHx[5]Y. Interestingly, the latter motif was found just outside the boundaries predicted for a CTLD, which suggests that further manual scrutiny of predicted domains in fish sequences could expand the list of currently identified domains of interest. Overall, the fact that closely-related sclr genes were identified as candidate receptors in both, salmon and carp, suggests a potential role of this CTL receptor in the response to β-glucans in fish and supports a need for its further characterization of function.

Based on criterion 4 (synteny in the zebrafish genome), we identified two CTLD-encoding genes of interest; putatively named clec4c and sclra. Three out of four corresponding duplicates at syntenic regions of the carp genome were already identified as genes of interest based on the selection criteria discussed above. This means that next to the sclr genes, we could identify clec4c as another gene of interest deserving further attention as candidate receptor for β-glucan. Taken together, our broad NGS approach helped us describe a clear regulation of the CLR pathway and identify a number of CTLD-containing candidate receptors for β-glucan binding. As proteins, these receptors would form a good starting point for future sugar binding assays and for further functional characterization with e.g., glycome microarrays, a high-throughput method devised to analyse β-glucan-binding proteins through an oligosaccharide microarray, followed by mass-spectrometric sequencing (64, 65). Altogether, the candidates discussed in this study should help pave the way to future functional studies that could ultimately lead to the identification of β-glucan receptor(s) in fish.

### ETHICS STATEMENT

This study was carried out in accordance with good animal practice as defined by the European Union guidelines for the handling of laboratory animals (http://ec.europa.eu/environment/chemicals/lab\_animals/home\_ en.htm). Animal work in Wageningen University was approved by the local experimental animal committee (DEC number: 2017.W-0034).

### REFERENCES


### AUTHOR CONTRIBUTIONS

JP, MF, and GW contributed to the design of the experiments, acquisition of samples, and analysis of data. GW acquired funding. RW and EB contributed with the phylogenetic and synteny analysis. CdO contributed with reagents, materials, and analysis tools. JP, RW, MF, and GW wrote the manuscript.

### FUNDING

This research was funded by the Netherlands Organisation for Scientific Research and São Paulo Research Foundation, Brazil (FAPESP) as part of the Joint Research Projects BioBased Economy NWO-FAPESP Programme (Project number 729.004.002). EB was supported by fellowships from the Maine INBRE Program through NIH grant P20GM103423. RW is a Burroughs Wellcome Fund Investigator in the Pathogenesis of Infectious Disease and this work was supported by NIH grant R15AI133415.

### ACKNOWLEDGMENTS

Pierre Boudinot is gratefully acknowledged for his suggestions on the phylogenetic analysis investigating the evolutionary relationships between CTLD sequences.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2019.00280/full#supplementary-material


approach and mass spectrometry. Mol Cell Proteomics (2015) 14:974–88. doi: 10.1074/mcp.M115.048272

65. Liu Y, Palma AS, Ten FZ, Chai WG. Insights into glucan polysaccharide recognition using glucooligosaccharide microarrays with oximelinked neoglycolipid probes. Methods Enzymol. (2018) 598:139–67. doi: 10.1016/bs.mie.2017.09.001

**Conflict of Interest Statement:** CdO is an employee of Biorigin Company.

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

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

# Characterization and Transcript Expression Analyses of Atlantic Cod Viperin

Khalil Eslamloo<sup>1</sup> \*, Atefeh Ghorbani <sup>2</sup> , Xi Xue<sup>1</sup> , Sabrina M. Inkpen<sup>1</sup> , Mani Larijani <sup>2</sup> and Matthew L. Rise<sup>1</sup> \*

*<sup>1</sup> Department of Ocean Sciences, Memorial University of Newfoundland, St. John's, NL, Canada, <sup>2</sup> Division of Biomedical Sciences, Faculty of Medicine, Memorial University of Newfoundland, St. John's, NL, Canada*

Viperin is a key antiviral effector in immune responses of vertebrates including the Atlantic cod (*Gadus morhua*). Using cloning, sequencing and gene expression analyses, we characterized the Atlantic cod *viperin* at the nucleotide and hypothetical amino acid levels, and its regulating factors were investigated. Atlantic cod *viperin* cDNA is 1,342 bp long, and its predicted protein contains 347 amino acids. Using *in silico* analyses, we showed that Atlantic cod *viperin* is composed of 5 exons, as in other vertebrate orthologs. In addition, the radical SAM domain and C-terminal sequences of the predicted Viperin protein are highly conserved among various species. As expected, Atlantic cod Viperin was most closely related to other teleost orthologs. Using computational modeling, we show that the Atlantic cod Viperin forms similar overall protein architecture compared to mammalian Viperins. qPCR revealed that *viperin* is a weakly expressed transcript during embryonic development of Atlantic cod. In adults, the highest constitutive expression of *viperin* transcript was found in blood compared with 18 other tissues. Using isolated macrophages and synthetic dsRNA (pIC) stimulation, we tested various immune inhibitors to determine the possible regulating pathways of Atlantic cod *viperin*. Atlantic cod *viperin* showed a comparable pIC induction to other well-known antiviral genes (e.g., *interferon gamma* and *interferon-stimulated gene 15-1*) in response to various immune inhibitors. The pIC induction of Atlantic cod *viperin* was significantly inhibited with 2-Aminopurine, Chloroquine, SB202190, and Ruxolitinib. Therefore, endosomal-TLR-mediated pIC recognition and signal transducers (i.e., PKR and p38 MAPK) downstream of the TLR-dependent pathway may activate the gene expression response of Atlantic cod *viperin*. Also, these results suggest that antiviral responses of Atlantic cod *viperin* may be transcriptionally regulated through the interferon-activated pathway.

Keywords: Gadus morhua, rsad2, teleost ISGs, qPCR, dsRNA, inhibition of antiviral responses

### INTRODUCTION

Interferon-stimulated genes (ISGs) play crucial roles as immune effectors and regulators in antiviral immune responses of fishes and other vertebrates (1, 2). As in mammals (3–5), the antiviral response of teleosts is triggered by recognizing viruses or "viral mimics" [e.g., synthetic double-stranded RNA (dsRNA): polyriboinosinic polyribocytidylic acid (pIC)] through

#### Edited by:

*Leon Grayfer, George Washington University, United States*

#### Reviewed by:

*Pierre Boudinot, Institut National de la Recherche Agronomique (INRA), France Chia-Ying Chu, National Taiwan University, Taiwan*

> \*Correspondence: *Khalil Eslamloo*

*keslamloo@mun.ca Matthew L. Rise mrise@mun.ca*

#### Specialty section:

*This article was submitted to Comparative Immunology, a section of the journal Frontiers in Immunology*

Received: *13 September 2018* Accepted: *06 February 2019* Published: *06 March 2019*

#### Citation:

*Eslamloo K, Ghorbani A, Xue X, Inkpen SM, Larijani M and Rise ML (2019) Characterization and Transcript Expression Analyses of Atlantic Cod Viperin. Front. Immunol. 10:311. doi: 10.3389/fimmu.2019.00311* intracellular Toll-like receptors (TLRs) or RIG-I-like receptors (RLRs), thereby activating transcription factors and enhancing the production of type I and II Interferons (IFNs) (6–9). Secreted IFNs initiate the Janus kinase-Signal transduction and activator of transcription (JAK-STAT) signaling pathway that enhances the transcription of ISGs (e.g., viperin and isg15) containing IFN gamma (IFNG)-activated sequences (GAS) and/or IFN-sensitive response elements (ISRE) in their promoters (2, 7, 10).

Virus inhibitory protein, endoplasmic reticulum (ER) associated, IFN-inducible (Viperin), also known as Radical Sadenosyl methionine (SAM) domain-containing 2 (RSAD2) or Virus-induced gene 1 (VIG1), is an antiviral protein, highly inducible by pIC, lipopolysaccharide (LPS), viruses, and bacteria (11, 12). The expression of mammalian viperin is induced via IFN-dependent and independent pathways, both of which may be activated by detection of viruses or dsRNA through a member of the RLR family and activation of transcription factors (e.g., IFN regulatory factor 3, IRF3) (12–14). Mammalian Viperin localizes in the ER-derived lipid droplets and inhibits viral replication [e.g., hepatitis C virus (HCV) and influenza] (12– 14). Viperin expression enhances the TLR-mediated production of type I IFN via forming a signaling complex consisting of Interleukin-1 receptor-associated kinase (IRAK1) and Tumor necrosis factor receptor-associated factor 6 (TRAF6) on lipid bodies and facilitating the nuclear translocation of IRF7 (15).

In addition to mammals, antiviral responsiveness of viperin was also observed in teleosts [reviewed by (16)] and an invertebrate species, i.e., Pacific oyster (Crassostrea gigas) (17, 18). Teleost viperin has been characterized and shown to be a pIC and LPS-induced gene in various species, i.e., tilapia (Oreochromis niloticus), annual fish (Nothobranchius guentheri), and red drum (Sciaenops ocellatus) (2, 16, 19–21). In addition, Atlantic salmon (Salmo salar) and crucian carp (Carassius auratus) viperin gene expression was induced in response to infectious salmon anemia virus (ISAV) (22) and grass carp reovirus (GCRV) (23), respectively, and Viperin exhibited antiviral activity against megalocytivirus in rock bream (Oplegnathus fasciatus) (24). As in mammals, viperin was shown to be an IFN-induced gene in zebrafish (Danio rerio) (25), and crucian carp viperin was suggested to be transcriptionally regulated via the RLRactivated IFN pathway (23). Although fish Viperins share some characteristics with their mammalian counterparts, the mechanisms involved in Viperin responses to immunogenic stimuli are not well-understood in fishes. Additionally, this gene/protein is not fully characterized in several teleost models.

In addition to its importance in Atlantic fisheries (26), Atlantic cod (Gadus morhua) exhibits a unique immune system among teleosts (27). Genomic studies have indicated that the Gadiformes lineage, including Atlantic cod, lack Major histocompatibility complex II (MHC II), CD4, Mx, and TLR5 genes, and show a unique expansion of genes including MHC I and TLR22 (28–30). Using transcriptome profiling of antiviral responses, several ISGs including viperin were previously identified in Atlantic cod. viperin displayed a strong induction in the brain of nodavirus carrier fish, the spleen and brain of pICinjected fish, and macrophages stimulated with pIC, but not LPS (31–35). However, the full sequence, developmental and tissue expression profiles, and regulating factors of Atlantic cod viperin remained unknown. In this study, we aimed to fully characterize Atlantic cod viperin, at the nucleotide and hypothetical amino acid (AA) levels, and determine its tissue distribution, developmental expression, and signaling pathways underlying its gene expression regulation during antiviral response.

### MATERIALS AND METHODS

### Gene Cloning, Sequencing, and Sequence Assembly

Gene-specific primers (GSPs) for rapid amplification of cDNA ends (RACE) were designed (see **Table 1**) using Primer3web v4.0.0 (http://primer3.ut.ee/) and the partial sequence of Atlantic cod viperin (obtained from NCBI GenBank). A pool of columnpurified RNA samples from the spleens of 10 Atlantic cod injected with pIC and sampled at 24 h post-injection (HPI) (5 µg RNA per sample) was used as RNA template for the RACE cDNA synthesis [see Inkpen et al. (36) and Hori et al. (33) for experimental design].

All PCR reactions in the present study were conducted on a Bio-Rad Tetrad 2 Thermal Cycler (Bio-Rad, Hercules, CA). Full-length 5′ and 3′ RACE cDNAs were synthesized using 1 µg total RNA and the SMARTer RACE cDNA Amplification Kit according to the manufacturer's instructions (Clontech, Mountain View, CA). The resulting 5′ and 3′ RACE cDNAs (i.e., 10 µl reaction) were diluted by adding 100 µl of Tricine-EDTA buffer and used for RACE PCR. A touch-down PCR [cycling program: 1 min at 95◦C; 5 cycles of (94◦C for 30 s, 72◦C for 3 min); 5 cycles of (94◦C for 30 s, 70◦C for 30 s, 72◦C for 3 min); 25 cycles of (94◦C for 30 s, 68◦C for 30 s, 72◦C for 3 min); and 1 final extension cycle of 72◦C for 10 min] was conducted in 50 µl reactions using the Advantage 2 Polymerase, Advantage 2 PCR buffer, dNTP Mix (Clontech), and GSPs (i.e., viperin-GSP; **Table 1**) as well as Universal Primer Mix provided by the kit (Clontech), following the manufacturer's instructions. Thereafter, 5 µl of the amplified 5′ and 3′ RACE products were diluted 50 times using Tricine-EDTA buffer and used for nested PCR. The nested PCR was performed in 50 µl reactions, using Nested GSPs (i.e., viperin-Nested GSP; **Table 1**) and Nested Universal Primer provided by the kit (Clontech), following the manufacturer's instructions. The cycling parameters for nested RACE PCR consisted of 1 min at 95◦C, followed by 20 cycles of (94◦C for 30 s, 68◦C for 30 s, 72◦C for 3 min), and 1 final extension cycle at 72◦C for 10 min.

PCR products were examined on 1.2% agarose gel, and then were extracted using the QIAquick Gel Extraction kit (Qiagen) according to the manufacturer's recommendations. TA cloning of gel-extracted PCR products was performed using pGEM-T-Easy vector (Promega, Madison, WI) at 4◦C overnight, using the manufacturer's instructions. Recombinant plasmids were transformed into Subcloning Efficiency DH5α Competent Cells (i.e., chemically-competent cells) (Invitrogen, Burlington, Ontario) following the manufacturer's instructions. The transformed cells were incubated in 300 µl of SOC medium

#### TABLE 1 | Primers used for the gene characterization and expression studies.


*GSPs, gene-specific primers; ORF, open reading frame; viperin, virus inhibitory protein, ER-associated, IFN-inducible; ifng, interferon gamma; isg15-1, interferon stimulated gene 15-1; lgp2, RNA helicase lgp2; il1b, interleukin 1, beta; eef1a, eukaryotic elongation factor 1* α*; rpl4a, 60S ribosomal protein l4-a; tubb2, beta-2 tubulin; eif3, eukaryotic translation initiation factor 3; rplp1, 60S acidic ribosomal protein P1; N/A, not applicable.*

(Invitrogen) for 1 h at 37◦C with shaking (∼225 rpm), and then cultured on Luria broth (LB)/agar plates containing 100 µg m1−<sup>1</sup> ampicillin and 40 µl plate−<sup>1</sup> of 40 mg ml−<sup>1</sup> X-gal (Sigma, St. Louis, MO) for 16 h at 37◦C. Thereafter, colonies were taken using blue/white selection and cultured in LB supplemented with ampicillin (100 µg ml−<sup>1</sup> ) at 37◦C overnight. Plasmid DNA was extracted using QIAprep Spin Miniprep Kit (Qiagen), following the manufacturer's instructions. The insert sizes of recombinant plasmids were checked using EcoRI (Invitrogen) digestion and agarose gel (1%) electrophoresis. Four colonies from each 5′ and 3′ RACE products were used for sequencing performed at the Genomics and Proteomics (GaP) facility, Core Research Equipment and Instrument Training (CREAIT) network, Memorial University of Newfoundland. Sequencing was conducted using an ABI 3730 DNA Analyzer with BigDye Terminator v3.1 Cycle Sequencing Kit (Applied Biosystems, Foster City, CA).

Lasergene 7.20 software (DNASTAR, Madison, WI) was used to acquire overlapping sequence from 5′ and 3′ RACE products and assemble the full-length viperin cDNA. GSPs amplifying the open reading frame (ORF) were designed (see **Table 1**) to verify the sequence assembly of full-length viperin. The ORF PCR was performed using the TopTaq polymerase kit (Qiagen, Mississauga, Ontario) and cDNA of pIC-stimulated Atlantic cod macrophages (i.e., pIC sub-group of no-inhibitor treatment in the macrophage stimulation experiment; see section Macrophage Isolation and Stimulation) in a 50 µl reaction. The PCR reaction was composed of 2 µl of cDNA (representing ∼10 ng of input RNA; see section qPCR Assays for the cDNA synthesis method), 0.5µM each of forward and reverse GSP, 5 µl of 10X TopTaq PCR buffer, 1.25 U of TopTaq DNA Polymerase, and 200µM of each dNTP, with the following PCR program: initial denaturation of 3 min at 94◦C, followed by 30 cycles of (94◦C for 30 s, 60◦C for 30 s, 72◦C for 2 min), and 1 final extension cycle at 72◦C for 10 min. The size of the resulting PCR product was determined by gel electrophoresis (1.2% agarose) alongside a 1 kb Plus Ladder (Invitrogen).

### Sequence Characterization and in silico Analyses

The AA sequence of Atlantic cod Viperin was predicted based upon assembled cDNA sequence using the SeqBuilder software of the Lasergene package (DNASTAR). To map the gene structure and determine intronic regions and genomic location, the cDNA sequence of viperin was aligned with genomic DNA sequence of Atlantic cod obtained from the Ensembl (http://www.ensembl. org) and Centre for Ecological and Evolutionary Synthesis (CEES: http://cees-genomes.hpc.uio.no) Genome Browsers. The AA sequences of Viperin of other fish species, as well as other vertebrate and invertebrate species (see **Supplemental Table S1**), were collected from the NCBI GenBank non-redundant (nr) AA database, and used for multiple sequence alignment (MSA) and phylogenetic tree construction. MSA analysis of predicted AA sequences of Atlantic cod Viperin with orthologous sequences in other species was implemented in MEGA6 software using the MUSCLE feature (37, 38). The Radical SAM domain of Atlantic cod Viperin was predicted using the PFAM database (http://pfam. xfam.org/) (39). The deduced AA sequences of Viperin homologs were aligned and used to generate a phylogenetic tree using the Neighbor-joining method in MEGA6 software (bootstrapped 10,000 times).

The neighboring genes of Atlantic cod viperin and its conserved synteny with other species were mapped using the Genomicus database (http://www.genomicus.biologie.ens.fr), powered by the Ensembl database. The neighboring genes and genomic location of Atlantic cod viperin were also confirmed using CEES Genome Browser (http://cees-genomes.hpc.uio.no).

To gain insight into transcriptional regulation of Atlantic cod viperin, putative transcription factor binding sites (TFBSs) were predicted in the proximal promoter region of this gene. The 1,000 bp 5′ -upstream of the transcription start site of Atlantic cod viperin were taken from genomic DNA sequence, available at CEES Genome Browser (http://cees-genomes.hpc. uio.no/). Putative TFBSs were identified using the vertebrates' profile of the TRANSFAC database (http://genexplain.com/ transfac/). This prediction was performed using the default parameters (i.e., Minimize False Positives) suggested by the TRANSFAC database, and the predicted TFBSs (core score > 0.8) with putative functions in immune responses were selected and presented herein.

### Prediction of Viperin Protein Structure

The recently-described partial mouse (Mus musculus) Viperin crystal structures were used as templates for homology modeling (40–43). The templates were retrieved from the Research Collaboratory for Structural Bioinformatics—Protein Data Bank (RCSB-PDB) (https://www.rcsb.org/; PDB ID: 5VSL and 5VSM). PyMOL v1.7.6 (http://www.pymol.org/) was used to visualize templates and models. Atlantic cod Viperin was computationally modeled using default parameters of I-TASSER (http://zhanglab. ccmb.med.umich.edu/I-TASSER/) (41–43). Twenty models were generated of which four models with the highest C-score (>– 1.8) were further analyzed. C-score (range −5 to 2), which indicates the confidence of the quality of the predicted model, was estimated by I-TASSER. The majority of the Atlantic cod Viperin was homology-modeled based on the mouse partial Viperin crystal structure with the exception of the first 56 N-terminal AAs and the last 23 C-terminal AAs which were missing from crystal structure and, therefore, were modeled ab initio by I-TASSER. The same approach was used to model the full-length mouse Viperin as well as zebrafish, Atlantic salmon, and human (Homo sapiens) Viperins. Both [4Fe-4S] cluster and Radical SAM analog (S-adenosylhomocysteine; SAH) were manually placed in the catalytic pocket of the Atlantic cod Viperin. The surface charge and the isoelectric point of Atlantic cod and full-length mouse Viperin were estimated using the PDB2PQR server (http://nbcr-222.ucsd.edu/pdb2pqr\_2.0.0/). Prediction of natively disordered regions of Viperin was conducted using the default parameters of the Protein disorder prediction server (PrDOS) website (http:// prdos.hgc.jp/cgi-bin/top.cgi).

### Animals

The Atlantic cod used in this experiment were kept in two 21 m<sup>3</sup> flow-through tanks (one tank for broodstock fish and one tank for fish used for tissue sampling and macrophage isolation) with optimal conditions (5.2–6.4◦C, 95–110% oxygen saturation and under an ambient photoperiod) in the Dr. Joe Brown Aquatic Research Building (JBARB) of the Ocean Sciences Centre (OSC). The fish [2.29 ± 0.42 kg (mean ± SE)] used for tissue sampling and macrophage isolation were fed 3 days weekly (i.e., 1% body weight per feeding time) with a commercial diet (Skretting, BC, Canada; crude protein 50%, crude fat 18% and crude fiber 1.5%). The broodstock fish were fed mackerel, herring and squid supplemented with vitamins 2 days per week before and during spawning season. Fish (i.e., for tissue sampling and macrophage isolation experiments) were fasted 24 h and euthanized with an overdose of MS222 (400 mg L−<sup>1</sup> ; Syndel Laboratories, Vancouver, BC) prior to the sampling. All procedures applied in the current investigation were approved by the Memorial University of Newfoundland's Institutional Animal Care Committee, according to the guidelines of the Canadian Council on Animal Care.

### Tissue Sampling

To evaluate the constitutive transcript expression of viperin in various tissues of adult Atlantic cod, four individuals (i.e., 2 male and 2 female) were used. Following euthanasia and dissection, samples were collected from 19 different tissues (i.e., blood, eye, brain, gill, heart, head kidney, posterior kidney, spleen, liver, gonad, stomach, pyloric caecum, midgut, hindgut, dorsal skin, ventral skin, dorsal muscle, ventral muscle, and fin) of each individual; the samples were immediately flash-frozen using liquid nitrogen and kept at −80◦C until RNA extraction.

### Sampling for Developmental Series

The floating fertilized eggs were automatically collected after overnight communal spawning. Then, 1.4 L of fertilized eggs (i.e., 2-cell to ∼64-cell embryos) [henceforth referred to as the zero days post-fertilization (DPF), or Day 0] were distributed into three 50 L conical incubator tanks (350 ml of eggs per tank) with 25 L h−<sup>1</sup> flow rate and gentle aeration. The fertilized eggs were incubated and kept in these tanks until yolk-sac absorption stage (i.e., 20 DPF; before active feeding) at 5.5–6.1◦C and under an ambient photoperiod. The developmental stage of embryos was determined daily (44). The embryos were in blastula/gastrula stages from 1 to 6 DPF (34.4 degree-days). The segmentation period began at 7 DPF (40.2 degree-days), and the golden eye stage was observed at 12 DPF (68.5 degree-days). Hatching started at 15 DPF (86.2 degree-days), and 100% of the embryos were hatched by 18 DPF (103.9 degree-days). To study viperin expression during early developmental stages, embryos (∼0.5 ml) or larvae (∼0.4 ml) were sampled daily (i.e., 0–20 DPF) from each incubator tank (i.e., one pooled sample from each replicate tank per day; n = 3), using 500µm Nitex. The collected (∼180 embryos or larvae per sample) samples from each tank were then flash-frozen using liquid nitrogen and kept at −80◦C until RNA extraction.

### Macrophage Isolation and Stimulation

Atlantic cod macrophages were isolated from the head kidneys of 5 individuals as described in Eslamloo et al. (34, 45). All reagents (e.g., culture medium) and equipment used in this experiment were identical to Eslamloo et al. (34). Briefly, following fish dissection, the macrophage-like cells isolated from each fish were seeded into 6-well plates (Corning, Corning, NY) at an equal density of 3 × 10<sup>7</sup> cells (in 2 ml L-15+) per well (16 wells per fish). The cells were cultured in Leibovitz's L-15+ [i.e., L-15 (Gibco, Carlsbad, CA) culture medium supplemented with 2 mM L-glutamine, 4.2 mM NaHCO3, 25 mM HEPES, 1.8 mM glucose, 100 U ml−<sup>1</sup> penicillin, 100 µg ml−<sup>1</sup> streptomycin (Gibco) and 1% fetal bovine serum (FBS; Gibco)] overnight at 10◦C. The nonadherent cells were then removed by washing the culture dishes 3 times with L-15+.

The immune inhibitors used in this study were purchased from InvivoGen (San Diego, CA), and the stock solutions were prepared using the manufacturer's instructions. The effective doses of different immune inhibitors used in this experiment were obtained from previously published in vitro-based studies in fish (46–49), except for Ruxolitinib (RUX) for which dose was based on an in vitro study on mammalian cells (50). 2- Aminopurine (2-AP), dissolved in phosphate-buffered saline (PBS): glacial acetic acid (AcOH) (200:1) was used at 5 mM as an inhibitor of double-stranded RNA-activated protein kinase (PKR) (48). Chloroquine (CHQ), dissolved in water, was used at 80µM as an inhibitor of endosomal TLR (49). In addition, Dimethyl sulfoxide (DMSO)-dissolved Resveratrol (RESV) and SB202190 (S90) were utilized at 50µM for inhibiting Nuclear factor kappa-B (NFKB) and p38 Mitogen-activated protein kinase (p38 MAPK) pathways, respectively (46, 47). To inhibit JAK1/JAK2, DMSO-dissolved RUX was used at 5µM (50). The pIC (Sigma-Aldrich) was dissolved in PBS at 10 mg ml−<sup>1</sup> and used as a stock solution in the present experiment. Starting 24 h after seeding, Atlantic cod macrophages isolated from each fish were exposed to 5 inhibitors (i.e., 2 wells per inhibitor for each fish). Moreover, macrophages from each individual were subjected to PBS, DMSO (0.57 µl ml−<sup>1</sup> of L-15+), or AcOH (0.17 µl ml−<sup>1</sup> of L-15+) as control conditions (i.e., a total of 8 experimental groups; 2 wells per condition for each fish). The cells were incubated with inhibitors for 1 h. Afterward, macrophages from each fish under different inhibitor treatments or controls (2 wells per group) were exposed to either 50 µg ml−<sup>1</sup> pIC (34) or PBS (5 µl of pIC solution or PBS ml−<sup>1</sup> of L-15+). The culture medium of all groups in this study contained an identical level of PBS (i.e., 38.8 µl ml−<sup>1</sup> of L-15+). The macrophages from each individual were incorporated into all experimental groups (i.e., 16 conditions in total; 5 biological replicates per group). Our previous study determined the time-dependent pIC responses of Atlantic cod macrophages (34), and based on these results the 24 h post-stimulation time point was selected for assessment of gene expression responses of macrophages in the current study. The samples were collected 24 h after pIC stimulation by removing the media and adding 800 µl TRIzol (Invitrogen, Burlington, ON) into each culture well plate. TRIzollysed samples were kept at −80◦C until RNA extraction.

### RNA Extraction and Purification

Total RNA was extracted using TRIzol (Invitrogen) following the manufacturer's instructions. RNase-free (i.e., baked at 220◦C for 7 h) ceramic mortars and pestles were used to homogenize the firm tissues (i.e., eye, gill, heart, stomach, pyloric caecum, midgut, hindgut, dorsal skin, ventral skin, dorsal muscle, ventral muscle, and fin), whereas other tissue and developmental samples were TRIzol-lysed using RNase-Free Disposable Pellet Pestles (Fisherbrand). Then, the homogenates of tissue and developmental samples were passed through QIAshredder (Qiagen) spin columns and used for RNA extraction following the manufacturer's instructions. The TRIzol-lysed macrophage samples were also processed for RNA extraction according to the manufacturer's recommendations.

Prior to purification, RNA samples (≤50 µg) in all experiments were treated with 6.8 Kunitz units of DNaseI (Qiagen) with the manufacturer's buffer (1X final concentration) for 10 min at room temperature to remove residual genomic DNA. RNA purification of adult tissue and developmental series samples (see sections Tissue Sampling and Sampling for Developmental Series) was performed using the RNeasy Mini Kit (Qiagen), whereas macrophage samples were purified using the RNeasy MinElute Cleanup Kit (Qiagen) according to the manufacturer's instructions. RNA concentration and quality were assessed using NanoDrop spectrophotometry (ND-1000), and RNA integrity was assessed by agarose gel electrophoresis (1% agarose). All column-purified RNA samples subjected to gene expression analyses in this study showed acceptable purity (i.e., A260/230 and A260/280 ratios>1.8) and integrity (i.e., tight 18S and 28S ribosomal RNA bands).

### qPCR Assays

cDNA synthesis was performed using RNA of each sample (i.e., 1 µg total RNA for macrophage samples or 5 µg total RNA for adult tissue and developmental samples), random primers (250 ng; Invitrogen), 1 µl of dNTPs (10 mM each; Invitrogen) and M-MLV reverse transcriptase (200U; Invitrogen) in the manufacturer's first-strand buffer (1X final concentration) and DTT (10 mM final concentration) at 37◦C for 50 min following the manufacturer's instructions.

The qPCR assays in the current study were designed and performed on the basis of the Minimum Information for Publication of qPCR Experiments (MIQE) guidelines (51). The qPCR analyses (including an inter-plate linker and no-template controls) were conducted using a ViiA7 Real-Time PCR system (Applied Biosystems, Burlington, Ontario) in the 384-well format and a qPCR program consisting of one cycle of 50◦C for 2 min, one cycle of 95◦C for 10 min, and 40 cycles of (95◦C for 15 s and 60◦C for 1 min), followed by a dissociation curve analysis (1 cycle at 60–95◦C in increments of 0.05◦C per second). Fluorescence data detection occurred at the end of each cycle. qPCR assays (13 µl) were comprised of 6.5 µl Power SYBR Green Master Mix (Applied Biosystems), 50 nM of each forward and reverse primer (0.52 µl of forward and 0.52 µl of reverse primers per reaction), 1.46 µl nuclease-free water (Gibco) and 4 µl cDNA (corresponding to 4 ng of input total RNA for macrophage samples or 10 ng of input total RNA for adult tissue and developmental samples).

The primer sequences in this study were obtained from our previous study (34) (see **Table 1**). Prior to the qPCR assays, the primer quality tests were conducted in triplicate using a 5 point, 3-fold serial dilution of the given cDNA template [starting with cDNA representing 10 ng (macrophage experiment) or 20 ng (adult tissue and developmental experiments) of input total RNA] as well as a no-template control. The cDNA template used for primer quality tests in the macrophage experiment was a pool of 3 individual pIC-stimulated samples, whereas the template used for primer quality tests in the adult tissue and developmental studies was a pool of different adult tissues and developmental samples. Each primer pair selected for qPCR assays showed an absence of amplification in the no-template controls, an amplicon with a single melting peak (i.e., no evidence of primer dimers or non-specific products in the dissociation curve), and amplification efficiency (52) ranging from 88 to 100% (**Table 1**).

To identify suitable normalizer genes for qPCR assays, the expression of candidate normalizers in adult tissue [i.e., cyclophilin a (cypa), 60S acidic ribosomal protein P1(rplp1), 60S ribosomal protein l4-a (rpl4a), beta-2 tubulin (tubb2), heat shock cognate 70 kDa (hsc70), eukaryotic translation initiation factor 3 (eif3), and eukaryotic elongation factor 1 α (eef1a)], embryonic and early larval stages (cypa, eif3, eef1a, tubb2, and beta-actin) and macrophage stimulation [protein phosphatase 1, catalytic subunit, gamma isozyme (ppp1cc), cypa, rplp1, rpl4a, tubb2, eif3, and eef1a] experiments was assessed in duplicate using 50% of the samples from each experiment. The expression results were subjected to geNorm analysis using qBase software as in Eslamloo et al. (34). The normalizers expressed comparably (i.e., with lowest M-value, a measure of transcript expression stability) in samples were selected for qPCR assays of adult tissue (eef1a, rpl4a, and rplp1), embryonic and early larval development (eif3 and tubb2), and macrophage stimulation (rplp1 and eef1a) experiments. In addition to viperin, the expression levels of Atlantic cod ifng, interferon stimulated gene 15-1 (isg15-1), RNA helicase lgp2 (lgp2; alias dhx58) and interleukin 1 beta (il1b) were measured in macrophage samples. Since ifng, isg15-1 and lgp2 play important roles in antiviral responses (2, 16), and they showed a strong pIC induction in Atlantic cod macrophages (34), we included them in qPCR assays as positive biomarkers of antiviral responses. On the other hand, since il1b is an antibacterial and pro-inflammatory biomarker in fish macrophages (53), this gene was added to qPCR assays in the macrophage experiment as a negative biomarker for the inhibition of targeted pathways.

The qPCR assays for samples of each experiment were performed in triplicate using 4 ng (macrophage samples) or 10 ng (adult tissue and embryonic and early larval development samples) of input total RNA per reaction. The performance of assays between qPCR plates used in a given experiment was tested using an inter-plate linker sample [C<sup>T</sup> (threshold cycle) value variations < 0.5] as well as no-template controls. ViiA7 software v1.2.2 (Applied Biosystems) was applied to calculate the relative quantity (RQ) values, relative to a calibrator sample (i.e., sample with the lowest normalized expression within each experiment) for the gene of interest, using the C<sup>T</sup> values (i.e., gene of interest and normalizers) and the amplification efficiency of each primer pair (see **Table 1**).

The qPCR results (RQ values) of tissue and macrophage experiments were statistically analyzed using the Prism package v6.0 (GraphPad Software Inc., La Jolla, CA). The Kolmogorov-Smirnov test was performed to check the normality of the data. The transcript expression data in the macrophage experiment were analyzed using a repeated measures two-way ANOVA test designed for randomized-block experiments, whereas the transcript tissue expression data were analyzed by a oneway ANOVA test. These analyses were followed by Sidak multiple comparisons post hoc tests to determine the significant differences (p ≤ 0.05) between adult tissues as well as between and within the groups in the macrophage experiment. The qPCR results from the developmental study were not subjected to statistical analyses.

### RESULTS

### Characterization of Atlantic Cod Viperin Sequence

Assembly of RACE sequencing reads for viperin, validated by ORF PCR, generated a 1342 bp cDNA sequence (excluding poly-A tail) (GenBank accession: MH279971; **Figure 1**). As predicted using SeqBuilder, the Atlantic cod viperin cDNA consisted of a 55 bp 5′ -UTR, 1044 bp (347 AA) ORF and a 243 bp 3′ - UTR. Also, three polyadenylation signal (AAUAAA) sequences were found in the 3′ -UTR (**Figure 1**). As determined by CEES and Ensembl databases, Atlantic cod viperin contained 6 exons (i.e., 1: 359 bp, 2: 162 bp, 3: 230 bp, 4: 150 bp, 5: 33 bp, and 6: 408 bp). This gene is located on Linkage Group (LG) 5 of the Atlantic cod genome (CEES, LG05:12624935- 12621299). Atlantic cod viperin is located downstream of cytidine monophosphate (UMP CMP) kinase 2 (cmpk2) and upstream of ring finger protein 144 (rnf144) and shows synteny similarity to zebrafish and human. **Figure 2** illustrates the genomic organization of the Atlantic cod viperin gene and


and inferred amino acid translation. The nucleotide sequences are numbered on both sides and the inferred amino acids are shown below the coding sequence. The period shows the predicted stop codon. The lower-case letters indicate the non-coding nucleotide sequence, whereas the protein coding sequence of *viperin* is shown in upper-case letters. The intronic sequences were determined using the Ensembl (http://www.ensembl.org) and Centre for Ecological and Evolutionary Synthesis (CEES: http://cees-genomes.hpc.uio.no) Genome Browsers. Three polyadenylation signals (polyA-signal: AAUAAA) were found in the 3′UTR.

its syntenic comparison with the viperin loci of zebrafish and human.

The multiple alignment of putative AA sequences of Atlantic cod Viperin and orthologous sequences from other eukaryotic species is shown in **Figure 3**. This comparison revealed considerable identity (i.e., 61–82%), notably in the radical SAM domain, between Viperin of Atlantic cod and representatives of different invertebrate phyla [Ciliophora: (i.e., Tetrahymena thermophila) and Mollusca: (i.e., C. gigas)] or vertebrate classes (e.g., Actinopteri, Amphibia, Reptilia, and Mammalia). Also, the radical SAM domain of Atlantic cod Viperin contained a conserved SAM binding motif (CXXXCXXC) (see **Figure 3**). However, no discernible conservation in AA sequence was noted in the N-terminus of Viperin putative orthologs. The lowest percentage of similarity (i.e., 61%) was found between Viperin of Atlantic cod and the ciliated protozoan, T. thermophila (see **Supplemental Table S1**), whereas Atlantic cod Viperin showed the highest percentage of similarity to its putative orthologs in teleosts, i.e., rainbow trout (Oncorhynchus mykiss) (82%) and Orange-spotted grouper (Epinephelus coioides) (79%). The phylogenetic tree of Viperin was constructed using a MSA of putative orthologous sequences from various species (i.e., ciliated protozoan, amphioxus, mollusc, fishes, amphibian, reptiles, bird, and mammals) (**Figure 4**). As expected, putative orthologous sequences were grouped and sub-grouped based upon the associated phyla and classes. For example, some fish species

within a given order [e.g., Salmoniformes (O. mykiss and S. salar) and Cypriniformes (Cyprinus carpio, C. auratus, and D. rerio)] were clustered together (**Figure 4**). Moreover, species (O. mykiss, S. salar, and Esox lucius) within the Protacanthopterygii superorder were clustered together. The Viperin of the ciliated protozoan, T. thermophila, formed a separate branch from the other species in the phylogenetic tree. Interestingly, Viperin of amphioxus (Branchiostoma japonicum) and Pacific oyster (C. gigas) were clustered together and separated from other species.

In silico analysis of the 5′ -upstream region was used to predict the immune-related putative TFBSs that may play roles in regulating the expression of Atlantic cod viperin. As shown in **Figure 5**, this analysis predicted IRF3/9, STAT and Activating transcription factor (ATF) motifs located in the proximal promoter of Atlantic cod viperin. Moreover, GAS and ISRE, which are the binding motifs for IFNG activation factor (GAF; STAT1 homodimer) and IFN-stimulated gene factor 3 (ISGF3), respectively, were identified in the proximal promoter region of Atlantic cod viperin.

### Structure Prediction of Viperin Protein

The recently-described partial mouse Viperin crystal structures (PDB: 5VSL, 5VSM) were used as templates for homology modeling prediction of full-length Atlantic cod and mouse Viperins (**Figure 6**). Comparison of the predicted structure of Atlantic cod Viperin to the partial crystal structure of mouse Viperin revealed a nearly identical overall architecture, namely a partial (βα)6-barrel folding (**Figure 6A**). The CXXXCXXC or the radical SAM binding motif within which the cysteine residues ligate three iron atoms of the [4Fe-4S] cluster, and the GEE motif as well as a serine and an arginine which were shown to form hydrogen bonds with SAH, are also found to be conserved in Atlantic cod Viperin (G125G126E<sup>127</sup> and S<sup>180</sup> and R<sup>194</sup> in mouse Viperin; G110G111E<sup>112</sup> and S<sup>165</sup> and R<sup>179</sup> in Atlantic cod Viperin; **Figure 6B**) (40). Further, the aromatic residues adjacent to the third cysteine in the CXXXCXXC motif were also conserved in the Atlantic cod Viperin (F<sup>90</sup> and F<sup>92</sup> in mouse Viperin; F<sup>75</sup> and F<sup>77</sup> in Atlantic cod Viperin; **Figure 3**). To generate a soluble derivative of mouse Viperin for crystallization, the N-terminus region containing a suggested α-helix (residues 1 to 71) was truncated (40). Our computational modeling also suggested the conservation of this N-terminus α-helix in the Viperin of Atlantic cod, zebrafish, Atlantic salmon and human (**Supplementary Figure S1**). As is the case for the mouse protein (40), the Viperins of the Atlantic cod as well as the aforementioned other species were also predicted to have an intrinsically disordered N-terminal region


residues of Viperin putative orthologs. The Radical SAM domain of Viperin was predicted using the PFAM database (http://pfam.xfam.org/), and it includes a conserved SAM binding motif (CXXXCXXC) (i.e., marked with a gray line within the Radical SAM domain). Atlantic cod (*Gadus morhua*), Orange-spotted grouper (*Epinephelus coioides*), zebrafish (*Danio rerio*), Rainbow trout (*Oncorhynchus mykiss*), Pufferfish (*Takifugu rubripes*), Mouse (*Mus musculus*), Human (*Homo sapiens*), Chicken (*Gallus gallus*), Western clawed frog (*Xenopus tropicalis*), Alligator (*Alligator mississippiensis*), Pacific oyster (*Crassostrea gigas*), Tetrahymena (*Tetrahymena thermophila*). See Supplemental Table S1 for GenBank accession numbers and the percentage of sequence similarity to Atlantic cod Viperin.

(**Supplementary Figure S2**), thus allowing for a lower degree of confidence for prediction of this region.

The majority of the AA differences between the Atlantic cod and mouse Viperins are located in the α-helices and loops forming the surface of the protein, while the core of the protein forming the potential catalytic pocket cavity remained remarkably conserved (Identity = 65.1%; homology = 77.7%; **Figures 6C,D**). Strikingly, surface charge analysis of cod Viperin revealed that most AA differences between Atlantic cod and mouse Viperin were replacing negative/neutral AAs with more positivelycharged residues, thus increasing the isoelectric point (pI) of the Atlantic cod Viperin compared to its mouse counterpart (mouse Viperin: charge at pH 7 = −5.61, pI = 5.89; Atlantic cod Viperin: charge at pH 7 = +3.78, pI = 8.61; **Figure 6E**).

species obtained from NCBI protein database were used to infer the evolutionary relationship among Viperin orthologs (see Supplemental Table S1 for GenBank accession numbers and the percentage of sequence similarity to Atlantic cod Viperin). The phylogenetic tree was generated by Neighbor-joining method and bootstrapped 10,000 times using MEGA6 software. The numbers at the branch points represent the bootstrap values. Branch lengths are proportional to calculated evolutionary distances. The scale represents number of substitutions per site. Arrowhead shows the Atlantic cod Viperin sequence.

### Constitutive Expression of Atlantic Cod Viperin During Early and Late Life Stages

qPCR revealed viperin to be a low-expressing gene (i.e., C<sup>T</sup> values above 30) during early developmental stages of Atlantic cod. The expression of the normalizer genes was slightly lower (**Figure 7**) during very early Atlantic cod embryonic development (days 0– 3), and this appeared to influence the RQ values of viperin for these developmental time points. Therefore, we did not subject these results to statistical analyses. However, as illustrated in **Figure 7**, the expression of Atlantic cod viperin was relatively higher from mixed cleavage until the mid-blastula stages (i.e., 0– 2 DPF). Thereupon, viperin levels dropped to a non-detectable level during gastrula and early segmentation stages (days 4–7). In other words, Atlantic cod viperin was not detected at the onset of zygotic gene expression, suggesting that viperin is a maternal transcript. Atlantic cod viperin expression increased during the segmentation stage, and then appeared to decrease after hatching.

The constitutive expression of Atlantic cod viperin was assessed in 19 different adult tissues. As shown in **Figure 8**, the highest expression of viperin transcript was seen in Atlantic cod blood, which was significantly higher than all other tissues. Interestingly, the levels of viperin transcript in immune-related tissues, notably head kidney and spleen, were significantly lower than that in blood. Additionally, viperin transcript had


significantly higher expression in gill and pyloric caecum compared to the dorsal skin and liver. The transcript expression of viperin was relatively, but not significantly, lower in the skin, muscle and fin tissues compared with some digestive tissues (e.g., midgut).

### Pathway Inhibition and Viperin Induction

We used different inhibitors to gain a better understanding of the signaling pathways activating the antiviral response of Atlantic cod viperin. As shown in **Figure 9A**, 2-AP, an inhibitor of the PKR-dependent pathway, significantly repressed the pIC induction of Atlantic cod viperin compared to the AcOHmatched control. There was no significant difference between group-matched PBS and AcOH controls, showing that viperin suppression in the 2-AP group was not influenced by the AcOH in which 2-AP was dissolved. CHQ also inhibited the viperin transcript expression in the pIC-stimulated Atlantic cod macrophages compared to the DMSO-matched vehicle control, indicating that TLR-activated pathways play an important role in the Atlantic cod viperin antiviral response. The expression of viperin transcript in pIC-treated macrophages was likewise significantly reduced in the S90 group compared to its pICtreated DMSO vehicle control, suggesting that inhibition of the MAPK pathway strongly affects Atlantic cod viperin induction. Finally, our results revealed that Atlantic cod viperin may be a JAK1/JAK2-activated gene downstream of the IFN pathway, as there was a strong repression in the RUX-exposed pIC group compared to the pIC-treated DMSO vehicle control. On the other hand, there was no significant difference between the RESV (i.e., NFKB inhibitor) group and its DMSO-matched control. Also, with respect to the role of ifng in innate immune responses and viperin induction, the expression levels of ifng as well as two important IFN-induced genes, i.e., isg15-1 and lgp2, were assessed in the macrophage samples (**Figures 9B–D**). The expression profiles of isg15-1 and lgp2 (**Figures 9C,D**) in response to various immune inhibitors were similar to viperin. Although there was a significant increase in the level of isg15-1 and lgp2 in the pIC sub-group of RESV treatment compared to its PBS control, the pIC induction of these genes in the RESV-treated samples was significantly attenuated compared to the pIC-stimulated DMSO

*(Continued)*

FIGURE 6 | β-strands, and α-helices are labeled in mouse crystal structure. (B) Ribbon model of crystal structure of mouse Viperin (left panel) and predicted structure of Atlantic cod Viperin (right panel) showing the SAH and [4Fe-4S] cluster coordinating residues in magenta. (C) Comparison between the identity and homology of the mouse and Atlantic cod Viperin. Non-identical/homologous residues are shown in yellow. (D) Ribbon model of crystal structure of mouse Viperin (left panel) and predicted structure of Atlantic cod Viperin (right panel) showing the amino acid residues forming the catalytic cavity in cyan. (E) Predicted surface topology of full-length mouse and Atlantic cod Viperins. The positive, neutral, and negative residues are colored blue, white, and red, respectively.

CT values are the average of the geometric mean value of 2 normalizers in 3 pooled samples (see section Materials and Methods for details). These results were not statistically analyzed since the CT-values of the normalizers in the first 4 sampling time points were slightly higher compared with most other sampling time points.

vehicle control (**Figures 9C,D**). Further, Atlantic cod ifng showed a largely comparable expression pattern to viperin, and its pIC induction was significantly suppressed by 2-AP, CHQ, S90 and RUX. However, in contrast to viperin and similar to isg15-1 and lgp2, the pIC induction of ifng was significantly decreased by RESV (i.e., 5.1-fold pIC induction) compared to the DMSO control group (13.5-fold pIC induction). Unlike isg15-1 and lgp2, though, there was no significant difference between the PBS control and pIC sub-groups of RESV treatment for ifng, indicating the suppressed induction of this gene by RESV. These findings suggest that, while the expression of all genes studied herein is regulated through PKR, TLR, MAPK and IFN pathways, only ifng shows a significant NFKB-dependent transcriptional activation. None of the inhibitors used in this study influenced viperin, ifng, isg15-1, or lgp2 expression in non-stimulated macrophages, and the inhibitors only suppressed the expression of these antiviral genes in the pIC group compared with pICstimulated controls (PBS, DMSO, or AcOH). This suggests that the constitutive expression of these antiviral biomarkers in Atlantic cod macrophages is regulated independent of the inhibited pathways. The expression of il1b was measured as an antibacterial and pro-inflammatory biomarker to examine the specificity of the immune inhibitors used in the current study. The transcription of il1b was not induced by pIC stimulation in any of the treatments (**Figure 9E**). The expression of il1b remained unchanged in the RESV group compared to the DMSO-matched control. Unlike the antiviral biomarkers, the constitutive expression of il1b was suppressed by 2-AP and S90 (**Figure 9E**), suggesting that PKR and MAPK pathways may have roles in regulation of il1b basal expression in Atlantic cod macrophages. Interestingly, il1b expression increased in the RUX-exposed group compared to the DMSO-matched control. Also, the expression of il1b was higher in PBS group of CHQ treatment compared to its PBS-matched control group. These results show that the expression profile of il1b is different from those of the antiviral biomarkers subjected to qPCR assays, suggesting that the inhibitors in the present study influenced their specific targets in immune pathways. In the present study, there were no significant differences between DMSO, PBS and AcOH control groups within pIC or PBS treatment, showing that the DMSO vehicle and AcOH used herein did not change the basal or pIC-induced expression of Atlantic cod viperin, ifng, isg15-1, lgp2, and il1b. **Figure 10** summarizes the pathway characterization results of the current study and illustrates

inhibitors, the target molecules, and their effects on the antiviral immune responses of Atlantic cod macrophages.

### DISCUSSION

The sequencing results showed that the Atlantic cod viperin transcript is 1,342-bp long (excluding poly-A tail) and consists of 6 exons. Exon 5 of Atlantic cod viperin is the shortest, whereas exons 1 and 6 are relatively longer compared to the other exons. The transcript size of Atlantic cod viperin is comparable with other fish species such as red drum (21), mandarin fish (Siniperca chuatsi) (54), Chinese perch (S. chuatsi), and Ara (Niphon spinosus) (55). As in Atlantic cod, viperin transcripts of red drum (21) and human (56) were reported to include short 5′ -UTRs. Similar to Atlantic cod, viperin genes of various vertebrates, e.g., mandarin fish (54, 55), chicken (Gallus gallus) (57), and mouse (M. musculus) (58), include 6 exons. The viperin exons in these species show a comparable size distribution to exons of Atlantic cod viperin, suggesting an evolutionarilyconserved exon/intron organization for viperin in vertebrates. The present study found viperin to be flanked by rnf144 and cmpk2 in the genome of Atlantic cod, which showed a conserved synteny to human and zebrafish. Previously published studies established the same gene order for viperin and adjacent genes of various fish (i.e., Elephant shark and tilapia), avians [i.e., chicken and zebra finch (Taeniopygia guttata)] and mammals [i.e., mouse and chimpanzee (Pan troglodytes)] (57, 59). Taken together, it seems that the genomic arrangement of viperin and its flanking genes, notably the opposite orientation of viperin and cmpk2, is conserved in vertebrate evolution.

As expected, the phylogenetic tree showed that the relatedness of Viperin putative orthologs appears to agree with taxonomic classification. Previous studies conducting MSAs and molecular phylogenetic analyses of Viperin in different species obtained comparable results to the current study (21, 24, 54, 59–61). Our MSA analyses revealed high levels of diversity in Nterminal AA sequences of putative orthologous Viperins among the representative species of different phyla and classes; in contrast, the radical SAM domain and the C-terminus are highly conserved.

This is the first study predicting protein structure of a teleost Viperin using crystal structure of a mammalian Viperin, and it suggests an overall structural similarity between Viperin of Atlantic cod and mouse. Specifically, the AA residues involved in coordinating the radical SAM domain as well as the [4Fe-4S] cluster of Atlantic cod were shown to be highly conserved compared to its mammalian ortholog. As in duck (Anas platyrhynchos) (60) and red drum (21), a conserved SAM binding motif (CXXXCXXC) was observed (see **Figure 3**) in the Atlantic cod Viperin. The conserved aromatic residues adjacent to the third cysteine in the CXXXCXXC motif were suggested to modulate the oxidation–reduction midpoint of the [4Fe-4S] cluster (62). The conservation of these residues in Atlantic cod Viperin seen herein suggests that the [4Fe-4S] cluster is likely used for the same function in both species. Moreover, our computational modeling revealed a catalytic cavity that is conserved between the Atlantic cod Viperin and its mammalian ortholog. Also, consistent with Viperin of human, duck, and crucian carp (60, 63, 64), our computational modeling analyses predicted the formation of an α-helix in the N-terminal region of Atlantic cod Viperin, though the N-terminus of all Viperins modeled herein are predicted to be a disordered region. Taken together, these results indicate that Atlantic cod Viperin exhibit the overall conserved structure observed/predicted in

other Viperin orthologs, highly suggestive of a comparable functional role.

The radical SAM Superfamily domain is found in hundreds of proteins that play a wide range of roles (65). Viperin has been well-documented to exhibit antiviral activities against human viruses [e.g., Zika, HCV and human cytomegalovirus (HCMV)] (56, 66–68). The radical SAM domain of Viperin is a pivotal factor in the antiviral roles of this protein (67, 69). The antiviral properties of mammalian Viperin chiefly rely on the interaction of its C-terminus with viruses [HCV and Dengue Virus Type-2 (DENV-2)] (70, 71). Similar to the mammalian Viperin, the overexpression of this protein by intramuscular injection of Viperin plasmid enhanced the resistance of rock bream against megalocytivirus (24). Accordingly, the conserved antiviral activity of mammalian and teleost Viperin may be attributed to conserved SAM domains (e.g., near-identical structures of SAM domains of Atlantic cod and mouse reported herein). While N-terminal residues of mammalian Viperin were not necessary for antiviral activity of this protein, they may modulate Viperin antiviral activity (67). The N-terminal amphipathic α-helix anchors Viperin into the ER membrane and is needed for protein localization in lipid droplets (63, 72);

(receptor-interacting protein 1), IKK (NFKBIA kinase), NFKBIA (NF-kappa-B inhibitor alpha), NFKB1 (nuclear factor kappa-B 1), IFN (interferon), NEMO (NFKB1 essential modulator or IKKG), TLR (Toll-like receptor), TRIF (TIR domain-containing adaptor protein inducing IFNB), TRAF (TNF receptor-associated factor), TANK (TRAF family member-associated NFKB activator), TBK (tank-binding kinase), IRF (IFN regulatory factor), MAPK (mitogen-activated protein kinase), AP1 (transcription factor AP1), Peli1 (pellino E3 ubiquitin protein ligase 1), PKR (IFN-induced, double-stranded RNA-activated protein kinase), TAK1 [transforming growth factor beta (TGFB)-activated kinase 1], TAB (TAK1-binding protein), MyD88 (myeloid differentiation primary response gene 88), IRAK (interleukin-1 receptor-associated kinase), IFNGR (IFN-gamma receptor), JAK (Janus kinase), STAT1 (signal transducer and activator of transcription 1), GAF (IFNG-activated factor), GAS (IFNG-activated sequence), ISG15-1 (IFN-stimulated gene 15-1), LGP2 (RNA helicase LGP2).

correspondingly, the mammalian Viperin N-terminus assists the inhibition of lipid droplet-dependent viral replication [reviewed by (13)]. ER localization of the teleost Viperin was previously shown in rock bream (24) and crucian carp (64). The evolution of fish Viperin involved the positive selection of N-terminal residues (73); therefore, the positively-selected N-terminus and the conserved C-terminus of Viperin may reflect the speciesdependent and ancestral functions, respectively, of this protein in antiviral responses (73). The present protein structure findings suggest that the molecular ability of teleost Viperin for binding to ER-associated lipid droplets, most likely, remained conserved, despite a large diversity between N-terminal residues of Viperin orthologs. Further studies are needed to test the correlation between diversity of N-terminal amphipathic α-helix and lipid binding and antiviral functions of teleost Viperin.

Atlantic cod viperin was found as a weakly-expressed gene during embryonic development. However, the expression of viperin was higher in mixed cleavage stage until mid-blastula (i.e., day 0–2) compared with subsequent embryonic stages (e.g., gastrula, early segmentation), showing the existence of viperin transcript prior to the onset of zygotic gene expression. This embryonic expression profile of viperin alongside its presence in adult fish gonads suggests that viperin is a maternal transcript in Atlantic cod. Maternal molecules (e.g., transcripts and proteins) play an important role in defense responses of fish during early life stages (74, 75). Some maternal (i.e., pre-midblastula expression) transcripts (e.g., irf7, ifngr1, and cathelicidin) involved in innate immune responses were previously identified in Atlantic cod (76–78). There was a considerable decrease in level of Atlantic cod viperin transcript after mid-blastula stage (i.e., in gastrula to early segmentation stages, 4–8 dpf). The transition from maternal to zygotic gene expression occurs at mid-blastula stage (i.e., maternal-embryo transition) (76, 79, 80). Therefore, with respect to the expression and degradation patterns of maternal transcripts (79), the non-detectable levels of Atlantic cod viperin immediately following mid-blastula stage may be attributed to degradation of maternal transcripts. It remains unknown if Viperin has any function in oogenesis or early embryogenesis. Nonetheless, viral hemorrhagic septicemia virus (VHSV)-induced viperin transcript was reported in eyed eggs and hatching fry of rainbow trout (81). Additionally, levels of viperin transcript increased in 48-h post-fertilization (hpf) larvae of zebrafish and 24-hpf D-veliger larvae of oyster infected with herpes simplex virus 1 (HSV-1) and Ostreid herpesvirus (OsHV-1), respectively (18, 82). We observed a steady increase in expression of Atlantic cod viperin transcript from early segmentation until hatch, and a somewhat decreased viperin expression after the hatch event. Therefore, if viperin function is conserved in teleost fish larvae, then its increasing constitutive expression in later stages of Atlantic cod embryonic development may provide information on the ontogeny of antiviral defense in this species. Further investigations are needed to determine whether viperin is a virus-responsive transcript in Atlantic cod larvae. However, similar to the viperin results seen herein, some immune-relevant transcripts (cxc chemokine, interleukin 8, atf3, and gaduscidin-1) of Atlantic cod were reported to increase during hatching; it was suggested that this induction may be involved in preparing cod embryos at the defensome level to combat environmental pathogens that may be encountered posthatch (76).

This study showed that the constitutive expression of Atlantic cod viperin varied among different tissues. Atlantic cod viperin was strongly expressed in blood and, interestingly, viperin levels in immune-related tissues (i.e., head kidney and spleen) were significantly lower than in blood. Likewise, the expression of viperin in the blood of red drum (21), large yellow croaker (Larimichthys crocea) (83) and duck (60) was higher than other tissues, including the immune-related and hematopoietic tissues (i.e., kidney of red drum and bursa of Fabricius of duck). Head kidney is the hematopoietic site in teleost species (84), and it contains a large number of differentiating cells (e.g., myeloid progenitor cells). The tissue-dependent expression of viperin in vertebrates may be associated with the differentiation of immune cells, as its expression is lower in hematopoietic tissues than that in the blood. The expression of Atlantic cod viperin transcript in an intestinal tissue (i.e., pyloric caecum) was higher than some tissues (e.g., fin, skin, and muscle). The intestinal expression of viperin was previously reported in different species (Rock bream, amphioxus, and duck) (24, 59, 60). As in other vertebrates (85), the teleost intestine contains various immune cells (e.g., granulocytes and macrophages) (86), and viperin expression in the digestive system of Atlantic cod may be attributed to the mucosal immunity of this species. Collectively, the results suggest that some aspects of cell- and tissue-dependent expression of viperin may be conserved among vertebrates. However, the function of Viperin in uninfected cells and tissues remains undescribed (13), and further studies are needed to determine if the constitutive expression of viperin is related to its immune or potential non-immune roles.

The present study examined if different pathway inhibitors may change the pIC response of Atlantic cod viperin and other well-known antiviral genes (i.e., ifng, isg15-1, and lgp2). In addition to these genes, the expression of il1b was assessed, as a pro-inflammatory and antibacterial biomarker, to check if the inhibitors used in this study have gene-specific effects. Previous studies have documented the induction of Atlantic cod viperin in pIC-exposed macrophages or larval cell line (ACL cells) as well as the spleen of pIC-injected fish (31, 33, 34, 87). Similarly, pIC-triggered expression of viperin has been confirmed by several in vivo- or in vitro-based studies in Pacific oyster (17), crucian carp (23), red drum (21), tilapia (19), large yellow croaker (83), annual fish (20), amphioxus (59), duck (60), and mice (11). Previous studies showed no induction of Atlantic cod viperin in macrophages or ACL cells stimulated with different LPSs (35), as well as in spleens of Atlantic cod injected with different LPSs or formalin-killed atypical Aeromonas salmonicida (35, 88). We also found similar results in LPS-treated Atlantic cod macrophages (unpublished data). Accordingly, the current study did not examine the effects of inhibition of immune-related pathways on the antibacterial response of Atlantic cod viperin. However, antibacterial induction of vertebrate viperin was reported in LPS-stimulated tilapia (19) and chicken (57) as well as mouse macrophages (11) and dendritic cells (89). LPS injection slightly up-regulated viperin expression in the spleen of orange-spotted grouper, but viperin induction was stronger in response to pIC or viral (i.e., grouper iridovirus, GIV) stimulation (90). It seems that, while the transcriptional regulation of the vertebrate viperin by the antiviral response is conserved among vertebrates, viperin may have species-dependent antibacterial responses.

**Figure 10** depicts the inhibitors, the targeted factors in antiviral immune responses and the summary of pathway characterization results in this study. The current investigation showed that 2-AP, CHQ, S90, and RUX significantly repressed the pIC-triggered expression of viperin in Atlantic cod macrophages, and similar results were seen for ifng, isg15-1, and lgp2 (**Figure 10**). In agreement with the present study, 2-AP was previously reported to block pIC induction of viperin in a monocyte/macrophage-like cell line (RTS11) of rainbow trout (91) and mx promoter of Japanese flounder (Paralichthys olivaceus) embryo cells (48). 2-AP is known as an inhibitor of IFN-induced PKR autophosphorylation (92); however, a study revealed that inhibitory effects of 2-AP on IFNB transcription may occur PKR-independently, through inhibiting Akt and consequently nuclear translocation of activated IRF3 (93). The inhibitory mechanisms of 2-AP in fish species are yet to be determined, and 2-AP-associated repression of viperin and other studied genes in Atlantic cod may be caused by PKR- or IRF3 dependent mechanisms. In contrast to antiviral biomarker genes, 2-AP suppressed the basal expression of il1b in Atlantic cod macrophages. In agreement with this result, 2-AP-dependent inhibition of il1b expression was reported in human (94). Taken together, it seems that PKR-regulated expression of il1b is conserved, and 2-AP suppresses PKR-derived immune responses of Atlantic cod in a gene-specific manner. To determine TLRdependent responses of Atlantic cod viperin, we used CHQ, which blocks the pIC response and TLR signaling by hindering endosomal acidification, thereby impairing PAMP recognition by intracellular TLRs (e.g., TLR3) (95). CHQ can also suppress autophagy by hindering lysosomal acidification, and it can be used as a drug with diverse functions (e.g., in malaria and cancer treatments) (96). However, immunosuppressive and antiviral activities of CHQ are associated with its roles in endosomal pH modulation and blocking nucleic acid binding to TLRs (97, 98). Therefore, the CHQ-mediated immunosuppression seen herein may be attributed to its effects on antiviral responses initiated by endosomal TLRs. CHQ was previously shown to inhibit the antiviral activity of rainbow trout macrophages (99), to decrease pIC induction of il1b in gilthead seabream (Sparus aurata) macrophages (100) and to block R848 (i.e., TLR7 ligand) response [e.g., myeloid differentiation primary response 88 (myd88) and il6] in peripheral blood leukocytes of Japanese flounder (101). The expression of il1b was not influenced by pIC stimulation in the current investigation, but this gene previously showed a slight up-regulation (i.e., 1.4-fold increase at 24 h poststimulation) in response to pIC in Atlantic cod macrophages (34). Slight differences between il1b results of our previous and present studies may be due to biological variability of immune responses among individuals. However, CHQ enhanced the expression of il1b in the control group, and this gene expression profile (i.e., diverged response of antibacterial and antiviral biomarkers) reflects the specific effects of CHQ on the antiviral response of Atlantic cod macrophages. Collectively, it seems that CHQ influences intracellular TLRs of teleosts, and that activation of Atlantic cod viperin, ifng, isg15-1, and lgp2 is highly dependent upon the endosomal recognition of pIC.

S90 is known to inhibit the activity of p38 MAPK (102). S90 was found to be a strong inhibitor of LPS-induced inflammation relevant genes (e.g., il1b and tnfa) in head kidney leukocytes of Atlantic salmon (47). Generally, p38 MAPK is a well-established key regulator of inflammatory responses (103). Therefore, the suppressed constitutive expression of il1b in the current study may be explained by p38 MAPKmediated regulation of inflammatory cytokines, as in a previous study involving murine macrophages (104). Nonetheless, with respect to S90-related inhibition of the virus-responsive IFNs, RIG-I-dependent p38 was suggested to be a pivotal factor in antiviral responses of mammalian dendritic cells (105). In Atlantic salmon, transcriptome profiling of antiviral responses of macrophage-like cells identified several pIC-responsive MAPKs (106). While the association of MAPK activation and antiviral responses of teleosts is not yet fully understood, our findings suggest an indirect or direct role of the p38 pathway in induction of Atlantic cod ifng and the other putative IFN-induced genes (i.e., viperin, isg15-1, and lgp2).

RUX blocks the activation of JAK1/JAK2 following the engagement of induced (e.g., pIC and LPS) type I and II IFNs with the IFN receptors (107, 108). As in Atlantic cod viperin, RUX significantly suppressed the expression of ifng, isg15-1, and lgp2 in the present study, suggesting the activation of these genes downstream of JAK1/JAK2-dependent pathway. There is no report on RUX-based inhibition of IFN-dependent responses of fish species, although RUX has been reported to significantly reduce the production of IFNG in mice (50). In mammalian macrophages, RUX suppressed LPS-induced expression of IFNregulated genes (109) and the IFN-mediated response of genes containing STAT-binding sites in their promoters (107). Similar to crucian carp viperin (23), the proximal promoter region of Atlantic cod viperin contains putative binding sites for GAS and ISRE, suggesting the IFN- or STAT-dependent regulation of this gene. Although the viperin putative TFBSs (i.e., GAS and ISRE sites) identified herein are compatible with the JAK1/JAK2 dependent Atlantic cod viperin transcript expression, these in silico results need to be experimentally validated by future studies. ISRE-regulated activation, as well as IFN receptor (IFNR)- and IRF-dependent induction of viperin have been described for mammalian macrophages (11), and chicken viperin transcript was found to be IFN-responsive (57). Likewise, stimulation of Atlantic salmon TO cells with recombinant IFN (110) or overexpression of IFN in zebrafish embryos (111) up-regulated the expression of teleost viperin. Also, induction of teleost viperin through activating factors downstream of MDA5 and IFN pathways was reported in crucian carp (23). In the present study, we observed a comparable gene expression profile between viperin and ifng in response to different inhibitors (except for RESV). Moreover, two putative IFN-induced genes (i.e., isg15- 1 and lgp2) studied herein showed expression profiles that were similar to that of viperin, suggesting that these genes may share signaling pathways (e.g., IFN-related pathway) activating their pIC response. Conversely, il1b in this study showed a different expression pattern and its expression increased in the RUX group compared to the DMSO control group. Likewise, RUX-mediated JAK inhibition increased the expression of pro-inflammatory cytokines (e.g., il6) in mouse macrophages (112). The qPCR results of the current study showed that RUX-mediated immune inhibition variably influenced the response of genes with different putative functions (e.g., antiviral vs. antibacterial roles) and regulatory pathways. We used RQ values of all samples from all treatments (i.e., pIC and PBS samples of inhibitor and control groups) and Pearson correlation coefficient tests to examine correlations between expression of ifng and other assessed genes. There were significant correlations (p < 0.0001) between the expression of Atlantic cod ifng and other antiviral genes [i.e., viperin (R: 0.71), isg15-1 (R: 0.70), and lgp2 (R: 0.94)], but no correlation was seen between ifng and il1b expression (R: 0.04; p = 0.74). This suggests that the IFN pathway plays important roles in transcriptional regulation of Atlantic cod viperin, isg15- 1 and lgp2. Additionally, the repressed antiviral response of Atlantic cod viperin, isg15-1 and lgp2 by other immune inhibitors (e.g., 2-AP and CHQ) may also be attributed to their influence on IFNG secretion, as ifng expression was significantly influenced by these inhibitors. Taken together, our results, alongside previous studies, suggest that IFN-dependent regulation of viperin may be conserved among vertebrates. However, a previous study suggested that the pIC induction of rainbow trout viperin may be independent of protein synthesis (91). Therefore, future studies using protein synthesis inhibitors and recombinant IFNs are needed to confirm IFN inducibility of Atlantic cod viperin.

RESV can inhibit NFKB through preventing Nuclear factor kappa-B inhibitor alpha (NFKBIA) phosphorylation (113), and it was shown to down-regulate the immune responses of head kidney leucocytes in turbot (Scophthalmus maximus). However, RESV did not significantly change the expression of Atlantic cod viperin transcript in response to pIC herein. Conversely, the pIC response of Atlantic cod ifng was suppressed by RESV. Also, there was a decrease in pIC induction of isg15-1 and lgp2 in the RESV group compared to the pIC-stimulated DMSO vehicle control group. In agreement with these results, RESV suppressed virus-induced IFNG expression in mice (114). These results suggest that the TLR- or RLR-activated NFKB may enhance the expression of Atlantic cod ifng, and may influence the intensity of the pIC response of isg15-1 and lgp2. Although the current findings suggest NFKB-independent stimulation for Atlantic cod viperin, further studies using a wider range of RESV doses and multiple sampling points are needed to evaluate the time- and dose-dependent effects of RESV on Atlantic cod viperin expression.

In conclusion, the present study showed that Atlantic cod viperin is an evolutionarily conserved gene that has similar gene organization and flanking genes to its putative orthologs in other vertebrates. Atlantic cod Viperin exhibits a close phylogenetic relationship with Viperin of other teleosts. A highly-conserved protein structure reported herein suggests a functional role for the Atlantic cod Viperin comparable with that of other Viperin orthologs. Atlantic cod viperin transcript showed a tissue-specific

### REFERENCES


constitutive expression and was most strongly expressed in the blood. The inhibitory effects of 2-AP, CHQ, and S90 on pIC induction of viperin transcript in Atlantic cod macrophages revealed that the expression of this gene may be dependent upon PKR, intracellular TLRs and MAPK, and/or possibly the factors (e.g., IRFs) activated downstream of these pathways. Also, RUX-associated suppression of Atlantic cod viperin, alongside the GAS and ISRE motifs predicted in the proximal promoter of this gene and its significant correlation with ifng expression in response to different immune inhibitors, suggest the IFNmediated regulation of Atlantic cod viperin.

### AUTHOR CONTRIBUTIONS

KE took a lead role in gene characterization, sequence analyses, experimental design, sampling, cell isolation, qPCR assays, data analyses, data interpretation, and the writing of manuscript draft. AG helped with RNA extraction, performed computational modeling of Viperin proteins and took part in manuscript writing. XX and SMI helped with sequence characterization, tissue sampling and RNA extraction. ML took part in data analyses and interpretation as well as manuscript writing. MLR was involved in experimental design, data analyses, and data interpretation, and took an active role in manuscript writing. All authors read and approved the final manuscript.

### FUNDING

This study was funded by Natural Sciences and Engineering Research Council of Canada (NSERC) Discovery Grants to MLR (341304-2012) and ML (047960-2015).

### ACKNOWLEDGMENTS

We would like to thank the Dr. Joe Brown Aquatic Research Building (JBARB) staff for assistance with fish husbandry. We are grateful to Dr. Tiago S. Hori for providing us with the RNA samples used as a template in RACE PCR. We would also like to thank Kathleen S. Parrish and David N. G. Huebert for helping with macrophage isolation and technical assistance in computational modeling analyses, respectively.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2019.00311/full#supplementary-material


double-stranded RNA viruses. Cytokine Growth Factor Rev. (2011) 22:63–72. doi: 10.1016/j.cytogfr.2011.02.001


crocea (Family: Sciaenidae). Dev Comp Immunol. (2018) 79:59–66. doi: 10.1016/j.dci.2017.10.013


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

# Fish Cholesterol 25-Hydroxylase Inhibits Virus Replication via Regulating Interferon Immune Response or Affecting Virus Entry

#### Ya Zhang<sup>1</sup> , Liqun Wang2,3, Xiaohong Huang<sup>1</sup> , Shaowen Wang<sup>1</sup> , Youhua Huang<sup>1</sup> \* and Qiwei Qin1,4 \*

*<sup>1</sup> College of Marine Sciences, South China Agricultural University, Guangzhou, China, <sup>2</sup> Key Laboratory of Tropical Marine Bio-resources and Ecology, South China Sea Institute of Oceanology, Chinese Academy of Sciences, Guangzhou, China, <sup>3</sup> University of Chinese Academy of Sciences, Beijing, China, <sup>4</sup> Laboratory for Marine Biology and Biotechnology, Qingdao National Laboratory for Marine Science and Technology, Qingdao, China*

#### Edited by:

*Stephanie DeWitte-Orr, Wilfrid Laurier University, Canada*

#### Reviewed by:

*Alberto Cuesta, University of Murcia, Spain Nguyen T. Vo, McMaster University, Canada*

#### \*Correspondence:

*Youhua Huang huangyh@scau.edu.cn Qiwei Qin qinqw@scau.edu.cn*

#### Specialty section:

*This article was submitted to Comparative Immunology, a section of the journal Frontiers in Immunology*

Received: *01 December 2018* Accepted: *07 February 2019* Published: *06 March 2019*

#### Citation:

*Zhang Y, Wang L, Huang X, Wang S, Huang Y and Qin Q (2019) Fish Cholesterol 25-Hydroxylase Inhibits Virus Replication via Regulating Interferon Immune Response or Affecting Virus Entry. Front. Immunol. 10:322. doi: 10.3389/fimmu.2019.00322* Cholesterol 25-hydroxylase (CH25H) is an interferon (IFN)-induced gene that catalyzes the oxidation of cholesterol to 25-hydroxycholesterol (25HC), which exerts broad-spectrum antiviral function. To investigate the roles of fish CH25H in Singapore grouper iridovirus (SGIV) and red-spotted grouper nervous necrosis virus (RGNNV) infection, we cloned and characterized a CH25H homolog from orange-spotted grouper (*Epinephelus coioides*) (EcCH25H). EcCH25H encoded a 271-amino-acid polypeptide, with 86 and 59% homology with yellow croaker (*Larimichthys crocea*) and humans, respectively. EcCH25H contained a conserved fatty acid (FA) hydroxylase domain and an ERG3 domain. EcCH25H expression was induced by RGNNV or SGIV infection, lipopolysaccharide (LPS) or poly (I:C) treatment *in vitro*. Subcellular localization showed that EcCH25H and mutant EcCH25H-M were distributed in the cytoplasm and partly colocalized with the endoplasmic reticulum. SGIV and RGNNV replication was decreased by EcCH25H overexpression, which was reflected in the reduced severity of the cytopathic effect and a decrease in viral gene transcription, but replication of both viruses was increased by knockdown of EcCH25H. Besides, the antiviral activity was dependent on its enzymatic activity. Treatment with 25HC significantly inhibited replication of SGIV and RGNNV. EcCH25H overexpression positively regulated the IFN-related molecules and proinflammatory cytokines, and increased both IFN and ISRE promoter activities. Moreover, 25HC treatment significantly suppressed SGIV and RGNNV entry into host cells. The similar inhibitory effect on SGIV entry was observed in EcCH25H overexpression cells. Taken together, our findings demonstrated that EcCH25H inhibited SGIV and RGNNV infection by regulating IFN signaling molecules, and might also influence viral entry via an effect on cholesterol.

Keywords: cholesterol 25-hydroxylase, grouper, Singapore grouper iridovirus, red-spotted grouper nervous necrosis virus, viral replication, interferon-stimulated gene, viral entry

### INTRODUCTION

The innate immune response is the first line of defense against invading pathogens, by which numerous pattern recognition receptors (PRRs) recognize pathogens, triggering the promotion of the interferon (IFN)-mediated innate immune response and leading to transcription of numerous IFN-stimulated genes (ISGs) (1–3). It has been reported that ISGs exert antiviral effects by inhibiting specific stages in the life cycle of viruses (4). For example, IFN-induced transmembrane proteins (IFITMs) block viral entry (5, 6), and protein kinase R (PKR) inhibits translation of some viral proteins through the suppression of eukaryotic translation initiation factor (eIF)2a elongation factors (7), while viperin and tetherin inhibit virus release from host cells (8, 9). Cholesterol 25-hydroxylase (CH25H), an ISG (10, 11), is an endoplasmic reticulum (ER)-associated membrane protein that catalyzes the conversion of cholesterol to 25 hydroxycholesterol (25HC) to reduce cholesterol accumulation (12, 13). 25HC, a natural oxysterol, plays a role in regulating cholesterol homeostasis and sterol biosynthesis through the regulation of nuclear receptors and sterol response element– binding proteins (SREBPs; in particular SREBP2) (12, 14, 15). Recent reports have revealed that 25HC inhibits the replication of various enveloped viruses, including hepatitis C virus (HCV) (16, 17), porcine reproductive and respiratory syndrome virus (PRRSV) (18), murine cytomegalovirus (MCMV) (19), West Nile virus (20), pseudorabies virus (PRV) (21), herpes simplex virus 1 (22), as well as one non-enveloped virus (poliovirus) (23), but it is inactive against another non-enveloped virus (adenovirus) (19). Besides, increased evidence demonstrates that 25HC blocks the entry of some viruses, including vesicular stomatitis virus (VSV) (19), human immunodeficiency virus (HIV) (19) and PRRSV (24), by modifying membrane fusion between viruses and cells.

Groupers, Epinephelus spp., are important commercial fish species in China and Southeast Asian countries. In recent years, outbreaks of infections caused by Singapore grouper iridovirus (SGIV) and red spotted grouper nervous necrosis virus (RGNNV) have caused heavy economic losses in the grouper industry (25–27). SGIV, an enveloped double-stranded large DNA virus, was first isolated from diseased grouper (Epinephelus tauvina) and characterized as a novel member of the genus Ranavirus, family Iridoviridae (26, 28). The SGIV genome consists of 140,131 bp, encoding 162 open reading frames (ORFs) (29). RGNNV is a non-enveloped icosahedral RNA virus, belonging to the genus Betanodavirus, family Nodavirdae, which has been frequently isolated from grouper and European sea bass (Dicenthrarchus labrax). The RGNNV genome is composed of two single-stranded positive-sense RNAs (30); RNA1 (3.1 kb) encodes the RNA-dependent RNA-polymerase (RdRp), while RNA2 (1.4 kb) encodes the capsid protein (CP). In addition, RNA3, a subgenomic transcript of RNA1, contains one ORF that encodes two non-structural proteins (31).

In order to clarify the mode of action of grouper viruses, a large number of immune genes were cloned and their roles in grouper virus infection were studied. For example, some immune-related genes, such as ISG15 (32), mitochondrial antiviral signaling protein (MAVS) (33), and IFN regulatory factor (IRF) 3 (34) exhibited different antiviral activities against SGIV or RGNNV infection. Notably, the transcriptome analysis of virus-infected grouper spleen showed that several ISGs were regulated in response to SGIV (35), or RGNNV (36), while the roles of most ISGs in fish virus replication remained unclear.

In the present study, a CH25H homolog from orangespotted grouper (EcCH25H) was cloned and characterized. We investigated the antiviral roles of EcCH25H during the replication of SGIV and RGNNV. In addition, we found that EcCH25H influenced the entry of SGIV and RGNNV. Our data will provide new insights into the function of fish CH25H genes against infection by fish viruses.

### MATERIALS AND METHODS

### Cells and Viruses

Grouper spleen (GS) cells used in this study were grown in Leibovitz's L15 medium containing 10% fetal bovine serum (FBS; Gibco) at 28◦C (37). The virus stocks of SGIV and purified SGIV were propagated in GS cells, while the RGNNV stocks were propagated in grouper brain (GB) cells, and the titers of the viruses were determined in GS cells and GB cells, respectively, that were both grown in Leibovitz's L15 medium containing 10% FBS (32, 38). Virus stocks were maintained at −80◦C.

### Reagents

25HC was purchased from Santa Cruz Biotechnology (sc-214091), reconstituted in chloroform to a concentration of 12.4 mM, and stored at −20◦C. Poly (I:C) and lipopolysaccharide (LPS) were purchased from Sigma–Aldrich. Poly (I:C) was reconstituted in RNase-free water at a concentration of 5 mg/mL and LPS was reconstituted in phosphate-buffer saline (PBS) at a concentration of 2.5 mg/mL. The lipophilic dye DiO and amine-reactive Cy5 were purchased from Biotium.

### Cloning of EcCH25H and Sequence Analysis

According to the expressed sequence tag (EST) sequences of EcCH25H from the grouper spleen transcriptome (35), the fulllength EcCH25H sequence was cloned using the primers listed in the **Table 1** by PCR amplification. Sequence analysis of EcCH25H was carried out using the BLAST program (http://www.ncbi. nlm.nih.gov/blast), and the conserved domains were predicted using the SMART program (http://smart.embl-heidelberg.de/). Multiple amino-acid sequence alignments were performed using the ClustalX1.83 software and the data were edited using the GeneDoc program. The phylogenetic tree was carried out using the MEGA 6.0 software.

### Expression Patterns of EcCH25H in Innate Immunity

To illustrate the expression changes of EcCH25H in response to fish virus infection, GS cells were infected with SGIV or RGNNV at a multiplicity of infection (MOI) of 2.0, and harvested at 4, 8, 18, 24, 36 h post-infection (h.p.i.) for RNA extraction and quantitative real-time PCR (qPCR) analysis. To examine

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


the expression profiles of EcCH25H in response to pathogenassociated molecular pattern (PAMP) molecules, GS cells were transfected with 200 ng poly (I:C) or treated with 4µg/mL LPS, and collected at 4, 8, 12, 24, 36 h for further qPCR analysis.

### Plasmid Construction and Cell Transfection

To clarify the molecular function of EcCH25H in vitro, the full-length wild-type EcCH25H (aa 1–271) sequence was subcloned into pEGFP-C1 using the primers in **Table 1**. Histidine codons (H) at positions 239 and 240 of wild-type EcCH25H (16) were converted to glutamine codons (Q) by sitedirected mutagenesis to create the mutant form (EcCH25H-M) lacking catalytic activity. The constructed plasmids (pEGFP-EcCH25H and pEGFP-EcCH25H-M) were subsequently verified by DNA sequencing.

Cell transfection was carried out using Lipofectamine 2000 reagent (Invitrogen). GS cells were seeded in 24-well plates at 60–70% confluence for 18–24 h, and cells were transfected with the mixture of 2 µL Lipofectamine 2000 and 800 ng plasmids and incubated for 6 h. After replacing with fresh normal medium, cells were cultured at 28◦C for further study.

### Subcellular Localization

To determine the subcellular localization of EcCH25H, pEGFP-C1, pEGFP-EcCH25H, and pEGFP-EcCH25H-M were cotransfected with the pDsRed2-ER plasmid into GS cells as described above. At 48 h post-transfection, cells were fixed with 4% paraformaldehyde, and stained with 4,6-diamidino-2-phenylindole (DAPI). Fluorescent images were observed by confocal laser scanning microscopy (CLSM).

### Virus Infection Assay

To demonstrate the roles of CH25H during virus infection, GS cells were manipulated using gene overexpression, RNA interference, or treatment with the product of CH25H, and then infected with fish viruses. In detail, GS cells that were transfected with pEGFP-C1, pEGFP-EcCH25H, or pEGFP-EcCH25H-M for 24 h were infected with SGIV or RGNNV at MOI 2.0. Cell morphology was observed and photographed using a phase contrast microscope; meanwhile, mock- and virus-infected cells were harvested at 24 h.p.i. for RNA extraction and qPCR analysis.

To knockdown the expression level of EcCH25H in GS cells, three small interfering RNAs (siRNAs) targeting different sequences of EcCH25H mRNA were commercially synthesized by Invitrogen. GS cells were grown in 24-well plates and then transfected with siRNAs (**Table 1**). The knockdown efficiency and transcription of viral genes were measured by qPCR.

In addition, the antiviral effects of metabolic product of CH25H, GS cells were treated with different concentrations of 25HC (1, 2, 4, or 8µM) and incubated for 72 h. The cell viability was determined by 3-(4,5-dimethylthiazol-2-yl)- 2,5-diphenyltetrazolium bromide (MTT) assay. The whole-cell lysates of treated infected cells were collected at the indicated time points and virus titers were determined. GS cells were seeded in 96-well plates for 18–24 h, and then infected with serial 10 fold dilutions of SGIV samples in eight replicates. After 96– 144 h.p.i., the 50% tissue culture infective dose (TCID50) assay was determined using the Reed–Muench method.

### Immunofluorescence Assay

GS cells were pretreated with different concentrations of 25HC (1, 2, or 4µM) for 12 h, and infected with RGNNV or SGIV at MOI 2.0. At 24 h.p.i., cells were fixed with 4% paraformaldehyde at 4◦C overnight, and permeabilized with PBS containing 0.2% Triton X-100 for 15 min. After washing three times with PBS, the cells were blocked with 2% bovine serum albumin (BSA) (Sigma-Aldrich) for 30 min, and incubated with primary antibodies (anti-CP serum 1:200) diluted in 0.2% BSA for 2 h at room temperature. The cells were washed three times with PBS, and incubated with the secondary antibody, fluorescence isothiocyanate-conjugated goat anti-rabbit IgG (1:200; Pierce) at room temperature. After 2 h, cells were stained with 1 mg/mL DAPI and observed under an inverted fluorescence microscope (Zeiss).

### Reporter Gene Assay

To illustrate the activated patterns of IFN promotion by EcCH25H, luciferase plasmids including ISRE-Luc, IFN1-Luc and IFN3-Luc were used. GS cells were cotransfected with 150 ng ISRE-Luc, IFN1-Luc, and IFN3-Luc, and 400 or 800 ng of either pEGFP-EcCH25H or pEGFP-C1. A total of 40 ng SV40 was included to normalize luciferase activity. At 48 h post-transfection, cells were harvested to measure the luciferase activities using the Dual-Luciferase <sup>R</sup> Reporter Assay System (Promega), as described in the manufacturer's instructions.

### Virus Entry Assay

Whether EcCH25H affected SGIV and RGNNV entry was investigated using two strategies, namely observation under CLSM and qPCR. On the one hand, GS cells were seeded in glass-bottom cell culture dishes and transfected with pEGFP-EcCH25H for 24 h, or pretreated with 25HC for 12 h. After the indicated times, the cells were prechilled at 4◦C for 5 min and infected with Cy5-labeled SGIV at 4◦C for 20 min in serum-free medium at 28◦C for 1 h. After being washed three times with medium, cells were fixed with 4% paraformaldehyde overnight and stained with the lipophilic dye DiO for 30 min (38). The cells were observed under CLSM and photographed. In each sample, 30 cells were selected for analysis and the data were represented as mean ± SEM.

Given that highly purified RGNNV particles (about 25 nm in diameter) was difficult to be obtained for confocal microscopy, the effects of the 25HC on RGNNV entry were determined using qPCR. In brief, GS cells were pretreated with 25HC for 12 h and infected with RGNNV or SGIV in serum-free medium. After 1 h, cells were washed three times with cold serum-free medium to remove unbound virus. Medium with 25HC was added and the cells were cultured for another 4 h (SGIV) or 1 h (RGNNV) at 28◦C. Cells were washed once with medium and treated with 0.2 mL citric acid buffer (citric acid 40 mM, potassium chloride 10 mM, sodium chloride 135 mM, pH 3.0) for 1 min. Cells were harvested for RNA extraction and qPCR analysis to measure the amount of virus that had entered the cells (39).

### RNA Isolation and qPCR

Total RNA was extracted and reversed using the ReverTra Ace qPCR RT Kit (TOYOBO) as described previously (32). qPCR was performed in a Roche 480 Real Time Detection System (Roche). Each assay was carried out under the following cycling conditions: 95◦C for 5 min for activation, followed by 45 cycles at 95◦C for 5 s, 60◦C for 10 s and 72◦C for 15 s. The primers are listed in **Table 1**. The expression level of target genes normalized to β-actin was calculated with the 2−11CT method. The data were indicated as mean ± SD and were shown from one representative experiment carried out in triplicate. Statistics were carried out using SPSS version 20 (IBM, Armonk, NY, USA) by one-way ANOVA. Differences were considered statistically significant when P was <0.05.

### RESULTS

### Sequence Characterization of EcCH25H

On the basis of the EST sequences from grouper spleen transcriptome, the full-length ORF of EcCH25H was obtained by PCR. EcCH25H encoded a 271-amino-acid polypeptide that shared 86% and 59% identity with the yellow croaker, Larimichthys crocea (KKF33056.1) and human, Homo sapiens (NP\_003947.1), respectively. Amino acid (aa) alignment analysis indicated that EcCH25H contained two conserved domains: ERG3 (aa 113–264) and fatty acid (FA) hydroxylase (aa 127– 260) (**Figure 1A**). Phylogenetic analysis revealed that EcCH25H showed the closest relationship to that of L. crocea. All the CH25Hs from different fish species were clustered into one group that was separated from the other groups, including birds, reptiles, mammals, and insects (**Figure 1B**).

## Expression Profiles of EcCH25H in vitro

To assess the changes in expression of EcCH25H in challenged GS cells, the transcription level of EcCH25H was examined by qPCR after infection with RGNNV or SGIV, or treatment with poly (I:C) or LPS. As demonstrated in **Figure 2A**, EcCH25H transcription level increased gradually and reached a peak at up to 72-fold higher than of the mock-infected control cells at 36 h during RGNNV infection. During infection with SGIV, EcCH25H transcription level was up-regulated between 18 and 24 h.p.i., reaching 216-fold higher than the level attained in the mock-infected cells, before decreasing at 36 h.p.i. (**Figure 2B**). After transfection with poly (I:C), EcCH25H expression level was induced gradually and reached a peak of 4.3-fold at 24 h compared to the controls (**Figure 2C**), while EcCH25H mRNA levels could also be induced by LPS (**Figure 2D**). Thus, EcCH25H is induced in response to poly (I:C), LPS or fish virus infection.

## Subcellular Localization of EcCH25H

CH25H is an ER-associated hydroxylase that catalyzes the conversion of cholesterol to 25HC, and CH25H contains tripartite histidine residues, which are critical for their hydroxylase activities (40). To explore whether EcCH25H was colocalized with ER in vitro and whether its localization was dependent on hydroxylase activity, EcCH25H-M was generated by converting histidine codons at positions 239 and 240 to glutamines, using site-directed mutagenesis. For subcellular localization analysis, pEGFP-C1, pEGFP-EcCH25H, or pEGFP-EcCH25H-M, was cotransfected with pDsRed2-ER plasmid into GS cells. The green fluorescence in pEGFP-C1-transfected cells was distributed throughout the cytoplasm and nucleus (**Figure 3**). However, in pEGFP-EcCH25H- and pEGFP-EcCH25H-M-transfected cells, the green fluorescence was distributed in the cytoplasm in two forms: point-like uniform and dot-like aggregation forms,

straight and dotted lines, respectively. Accession numbers were: *Larimichthys crocea*, KKF33056.1; *Danio rerio*, ASU11040.1; *Gallus gallus*, NP\_001264283.1; *Chrysemys picta bellii*, XP\_005292570.1; *Mus musculus*, NP\_034020.1; *Homo sapiens*, NP\_003947.1. (B) Phylogenetic analysis of CH25Hs. A neighbor-joining tree was constructed based on the protein sequences of CH25H-like genes from different species using MEGA 6.0 software. Numbers at the nodes denote the bootstrap values of 1,000 replicates. Scale represents the numbers of substitutions per 1,000 bases.

of EcCH25H by qPCR (*n* = 3, means ± SD). \**P* < 0.05.

which were both colocalized partly with the red fluorescence of the ER. Therefore, we suggest that EcCH25H encoded an ER-localized protein, and that its localization was independent of hydroxylase activity.

### Antiviral Effects of EcCH25H on Fish Virus Replication in vitro

To explore the effects of EcCH25H overexpression on fish virus replication and whether it depends on the enzymatic activity of CH25H, GS cells were transfected with pEGFP-C1, pEGFP-EcCH25H, or pEGFP-EcCH25H-M for 24 h, and then infected with SGIV or RGNNV for a further 24 h. The transcription level of EcCH25H was examined by qPCR to confirm the successful ectopic expression of EcCH25H and EcCH25H-M in transfected cells. mRNA expression of EcCH25H and EcCH25H-M was significantly increased up to 1,322.2- and 1,235.7-fold compared with that of the control vector cells (**Figure 4A**). Compared to the control vector cells, the cytopathic effect (CPE) induced by SGIV and RGNNV was weakened in EcCH25H-overexpressing cells. In contrast, CPE severity was significantly increased in EcCH25H-Moverexpressing cells compared with EcCH25H-overexpressing cells (**Figure 4B**). Consistently, the transcription levels of the SGIV major capsid protein (MCP) and VP19 genes, as well as RGNNV CP and RdRp genes, were significantly inhibited when EcCH25H was overexpressed. However, the mutant significantly decreased the inhibitory effects of EcCH25H on viral gene transcriptions (**Figures 4C,D**). Thus, the results demonstrated that EcCH25H overexpression inhibited replication of RGNNV and SGIV, and that the antiviral effects were dependent on its enzymatic activity.

To further investigate whether knockdown of EcCH25H promoted SGIV and RGNNV replication, we designed three siRNAs targeting EcCH25H, and examined interference efficiency in GS cells using qPCR. Compared with the negative control siRNA, siRNA1 decreased expression of EcCH25H, with 47% knockdown efficiency (**Figure 5A**). After transfection with siRNA-EcCH25H for 24 h, GS cells were infected with SGIV and RGNNV for a further 24 h, and collected to examine the transcription of viral genes by qPCR. Knockdown of EcCH25H by siRNA promoted SGIV and RGNNV replication compared with the cells transfected with the negative control siRNA (**Figures 5B,C**). Thus, it was proposed that EcCH25H exerted the antiviral effects on SGIV and RGNNV infection.

It has been reported that CH25H exerts antiviral activities by producing 25HC (18, 19, 41). To explore whether 25HC plays a role in SGIV or RGNNV infection, GS cells were pretreated with different concentrations of 25HC for 12 h, then infected with SGIV or RGNNV for 24 h. Cell viability was not decreased when cells were exposed to 4µM 25HC for 72 h (**Figure 6A**). Virus titer assay indicated that 25HC significantly inhibited SGIV replication (**Figure 6B**). 25HC significantly inhibited expression of the CP protein of RGNNV in a dose-dependent manner (**Figure 6C**). 25HC significantly decreased the transcriptional level of MCP and VP19 of SGIV and CP and RdRp of RGNNV compared with chloroform (**Figures 6D,E**). Taken together, these results suggested that EcCH25H exhibited antiviral activity by producing 25HC, with the virus inhibition being dependent on the EcCH25H enzymatic activity.

### EcCH25H Overexpression Positively Regulates the IFN Immune and Inflammatory Response

To clarify the effects of EcCH25H on host IFN immune and inflammatory response, we examined expression of IFN signaling molecules, and proinflammatory cytokines by qPCR and the promoter activities of IFN-1, IFN-3 and ISRE in EcCH25H- and

transcription level of EcCH25H by qPCR at 48 h post-transfection. The primers EcCH25H-RT-F/R were designed to detect expression of EcCH25H and EcCH25H-M. (B) EcCH25H overexpression weakened the severity of CPE induced by RGNNV and SGIV in GS cells. GS cells were transfected with pEGFP-C1, pEGFP-EcCH25H, or pEGFP-EcCH25H-M for 24 h, and infected with SGIV or RGNNV, respectively. At 24 h.p.i., GS cells were observed for CPE using microscopy. In SGIV-infected cells, the black arrows indicated the severity of CPE induced by SGIV infection, which was characterized by cell rounding and aggregation of cells. The white arrows showed that vacuoles were induced by RGNNV infection. (C,D) Viral gene transcription of SGIV or RGNNV in EcCH25H- or EcCH25H-M-overexpressing cells. EcCH25H- and EcCH25H-M-overexpressing cells were infected with RGNNV or SGIV, and collected at 24 h.p.i. to determine by qPCR expression of MCP and VP19 of SGIV and CP and RdRp of RGNNV (*n* = 3, means ± SD). \**P* < 0.05.

EcCH25H-M-overexpressing cells. Compared with the controls, overexpression of EcCH25H significantly increased expression of several IFN-related genes, including IRF3, IRF7, ISG15, IFNinduced 35-kDa protein (IFP35), myxovirus resistance gene (MX)I and MXII. In EcCH25H-M-overexpressing cells, the concentration of mRNA transcripts was significantly lower than that in EcCH25H overexpressing cells (**Figure 7A**). Reporter gene analysis showed that EcCH25H overexpression significantly increased the luciferase activity of ISRE, IFN1, and IFN3 promoters in a dose-dependent manner (**Figure 7B**). We propose that EcCH25H positively regulated IFN immune response.

The effects of EcCH25H overexpression on expression of proinflammatory cytokines were also examined. mRNA production by tumor necrosis factor (TNF)-α, interleukin (IL)- 6 and IL-8 genes was significantly increased in EcCH25Hoverexpressing cells compared with control cells, with the stimulatory effects of EcCH25H being dependent on enzymatic activity (**Figure 7C**). Ectopic expression of EcCH25H in vitro upregulated the expression of IFN and the inflammatory response, and was related to enzyme activity.

### EcCH25H Inhibits SGIV and RGNNV Entry

To further investigate whether EcCH25H suppressed SGIV or RGNNV entry, GS cells were transfected with pEGFP-EcCH25H or pretreated with 25HC at 4µM for 12 h, then infected with RGNNV or Cy5-labeled SGIV. EcCH25H overexpression reduced the number of red-fluorescence-labeled SGIV particles in the cytoplasm to 29.7% of that in empty-vector-transfected cells (**Figures 8A,B**). Furthermore, SGIV entry into GS cells pretreated with 25HC was significantly reduced to 48% of that in cells treated with chloroform (**Figures 8C,D**). 25HC decreased the transcriptional level of MCP of SGIV and CP of RGNNV compared with that achieved in cells exposed to chloroform (**Figures 8E,F**). These results suggested that EcCH25H and 25HC might influence the entry of SGIV or RGNNV, by reducing the infectivity of virus particles in the cytoplasm.

### DISCUSSION

An increasing number of studies have shown that CH25H can inhibit various enveloped and non-enveloped viruses by production of 25HC (18, 23, 39). However, few have focused on the function of CH25H in lower vertebrates. Here, we cloned and characterized EcCH25H and analyzed its enzymatic product 25HC as antiviral restriction factors against iridovirus and nodavirus.

Based on bioinformatics analysis, we found that EcCH25H from E. coioides shared 86% identity to that from L. crocea. Amino acid alignment analysis has shown that EcCH25H contains the characteristic FA hydroxylase domain, which is similar to that of the zebrafish CH25H (42), suggesting that CH25H is conserved from lower vertebrates to mammals. Using qPCR analysis, our data showed that the concentration of EcCH25H increased significantly during SGIV and RGNNV infection, a finding which was consistent with the previous studies in which CH25H was induced in response to VSV (19),

using anti-CP antibody (green). The infectivity was quantified as the percentage of treated cells with incorporated viruses relative to that for control cells. The viral infectivity of control cells was set as 100% (C). Meanwhile, GS cells were harvested for TCID50 assay of SGIV (B) and qPCR for expression of SGIV (D) or RGNNV (E) genes at 24 h.p.i. (*n* = 3, means ± SD). \**P* < 0.05.

Zika virus (39), HCV (17), and MCMV (43). In addition, studies had demonstrated that CH25H could be induced by poly (I:C) (17) and LPS (44, 45), which was similar to the results of the present study, suggesting that CH25H is induced by PAMPs and plays a crucial role in the innate immune response against virus infection. In addition, subcellular localization analysis has shown that EcCH25H and the catalytically inactive mutant EcCH25H-M are both localized in the cytoplasm and have diffuse intracellular expression patterns, which are the same as the localization pattern of pig CH25H in vitro (21).

activities of IFN-1, IFN-3, and ISRE promoter induced by EcCH25H using reporter gene assay (B) (*n* = 3, means ± SD). \**P* < 0.05.

As mentioned above, the broad antiviral effects of CH25H have been demonstrated in mammalian cells for some viruses, and these studies have indicated that CH25H exerts an antiviral function by production of 25HC (39, 46). It is suggested that 25HC directly inhibits virus replication by suppressing viral gene expression (19, 43). In our study, ectopic expression of EcCH25H significantly weakened the severity of CPE induced by SGIV and RGNNV infection. Consistently, transcription of viral genes was inhibited by EcCH25H overexpression and by treatment with 25HC, but was

enhanced by silencing EcCH25H. In other words, EcCH25H and 25HC suppressed viral gene expression to inhibit infection of SGIV and RGNNV.

In addition, recent studies have shown that CH25H-M lacking enzymatic activity to produce 25HC still had antiviral activity against HCV (19) and PRRSV (18), but not against other viruses, such as mouse hepatitis virus 68 and PRV (21), suggesting that CH25H can inhibit viral infection through either enzyme-activity-dependent or -independent pathways. In this study, we found that EcCH25H-M was unable to inhibit replication of SGIV and RGNNV, confirming that the antiviral mechanism of EcCH25H against SGIV and RGNNV is dependent on its enzyme activity and similar to that against other viruses, such as HCV (19) and PRRSV (18). Unexpectedly, the EcCH25H-M showed no effect on SGIV, but enhanced RGNNV replication moderately, indicating that the enhancement of virus replication by EcCH25H-M might be dependent on virus type.

Moreover, as a naturally occurring secreted oxysterol, 25HC can block the fusion of the enveloped viruses with cell membranes to inhibit viral entry (47). For example, 25HC can modify cell membranes to inhibit efficient fusion between HIV and cell membranes, without affecting HIV transcription, translation and budding (19). 25HC inhibits VSV entry, by a route which is independent of regulation of SREBP2, mevalonate production, or protein prenylation (19). In addition, 25HC inhibits HCV infection at a postentry stage by suppressing activation of SREBP2 (17). Our results show that EcCH25H overexpression and 25HC treatment inhibit SGIV entry, in agreement with previous observation for VSV (19). As an enveloped virus, SGIV entered host cells by fusion with the plasma membrane, and EcCH25H might inhibit SGIV entry by blocking the fusion of viral envelope with cell membranes. Differently, as a non-enveloped virus, RGNNV entry might be affected by the action of EcCH25H on interferon immune response. The detailed mechanism by which EcCH25H suppresses SGIV or RGNNV entry needed further investigation.

A growing number of studies have revealed that CH25H is involved in IFN immune response in few mammals (17, 21, 24). 25HC promotes the secretion of some inflammatory cytokines, such as IL-6, IL-8, and macrophage colony-stimulating factor in macrophages and epithelial cells (48, 49). To clarify the effects of EcCH25H on host IFN and inflammatory response, we examined the expression of IFN signaling molecules and proinflammatory cytokines in EcCH25H- and EcCH25H-M-overexpressing cells. Expression levels of several IFN-related cytokines, including IRF3, IRF7, ISG15, IFP35, MXI, and MXII, as well as that of inflammatory cytokines, such as TNF-α, IL-6, and IL-8, were all significantly increased. Further analysis showed that EcCH25H-overexpression enhanced IFN and ISRE promoter activities. Thus, we speculate that EcCH25H positively regulates IFN immune and proinflammatory response to inhibit SGIV and RGNNV infection.

In summary, we demonstrated the roles of EcCH25H in RGNNV and SGIV replication and entry. A fish CH25H homolog from orange-spotted grouper (EcCH25H) was cloned and characterized. EcCH25H encoded a cytoplasmic protein and was induced by RGNNV or SGIV infection, or by treatment with either LPS or poly (I:C). Overexpression of EcCH25H in

### REFERENCES


vitro significantly suppressed replication of SGIV and RGNNV due to its positive effect on the host IFN immune response, which is contrary to the results of RNAi. Furthermore, our results indicate that EcCH25H inhibits SGIV and RGNNV entry, and shed more light on the mode-of-action of CH25H against fish virus infection.

### AUTHOR CONTRIBUTIONS

YZ performed the experiments, analyzed the data, and wrote the manuscript. LW and XH prepared virus stocks including SGIV, RGNNV and the purified SGIV. SW participated in confocal microscopy analysis. YH and QQ participated in the design of the study, data analysis and revised the manuscript.

### FUNDING

This work was supported by grants from the National Key Research and Development Program of China (2018YFD0900505), the National Natural Science Foundation of China (31472309, 31772877), National Key R&D Program of China (2017YFC1404504), and Open Fund of Key Laboratory of Experimental Marine Biology, Chinese Academy of Sciences (No. KF2017NO5).


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

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

# Fish Autophagy Protein 5 Exerts Negative Regulation on Antiviral Immune Response Against Iridovirus and Nodavirus

Chen Li <sup>1</sup> , Jiaxin Liu<sup>1</sup> , Xin Zhang<sup>1</sup> , Shina Wei <sup>1</sup> , Xiaohong Huang<sup>1</sup> , Youhua Huang<sup>1</sup> , Jingguang Wei <sup>1</sup> \* and Qiwei Qin1,2 \*

*<sup>1</sup> Joint Laboratory of Guangdong Province and Hong Kong Region on Marine Bioresource Conservation and Exploitation, College of Marine Sciences, South China Agricultural University, Guangzhou, China, <sup>2</sup> Laboratory for Marine Biology and Biotechnology, Qingdao National Laboratory for Marine Science and Technology, Qingdao, China*

Autophagy is an important biological activity that maintains homeostasis in eukaryotic cells. However, little is known about the functions of fish autophagy-related genes (Atgs). In this study, we cloned and characterized Atg5, a key gene in the autophagy gene superfamily, from orange-spotted grouper (*Epinephelus coioides*) (EcAtg5). EcAtg5 encoded a 275-amino acid protein that shared 94 and 81% identity to seabass (*Lates calcarifer*) and humans (*Homo sapiens*), respectively. The transcription level of EcAtg5 was significantly increased in cells infected with red-spotted grouper nervous necrosis virus (RGNNV). In cells infected with Singapore grouper iridovirus (SGIV), EcAtg5 expression declined during the early stage of infection and increased in the late stage. Fluorescence microscopy revealed that EcAtg5 mainly localized with a dot-like pattern in the cytoplasm of grouper cells. Overexpression of EcAtg5 significantly increased the replication of RGNNV and SGIV at different levels of detection, as indicated by increased severity of the cytopathic effect, transcription levels of viral genes, and levels of viral proteins. Knockdown of EcAtg5 decreased the replication of RGNNV and SGIV. Further studies showed that overexpression EcAtg5 activated autophagy, decreased expression levels of interferon related cytokines or effectors and pro-inflammatory factors, and inhibited the activation of nuclear factor κB, IFN-sensitive response element, and IFNs. In addition, ectopic expression of EcAtg5 affected cell cycle progression by hindering the G1/S transition. Taken together, our results demonstrated that fish Atg5 exerted a crucial role in virus replication by promoting autophagy, down-regulating antiviral IFN responses, and affecting the cell cycle.

Keywords: grouper, Atg5, SGIV, RGNNV, interferon

### INTRODUCTION

Autophagy is a conserved cell biological pathway that delivers cytoplasmic components to lysosomes for degradation and elimination of useless or harmful substrates to maintain cell homeostasis in all eukaryotic cells (1). This fundamental process involves formation of double membrane autophagosomes, which fuse with lysosomes to degrade the sequestered cargo (2, 3). Viruses depend on the host cell's machinery to replicate the genome and generate progeny

#### Edited by:

*Stephanie DeWitte-Orr, Wilfrid Laurier University, Canada*

#### Reviewed by:

*Yishan Lu, Guangdong Ocean University, China Hai-peng Liu, Xiamen University, China*

#### \*Correspondence:

*Jingguang Wei weijg@scau.edu.cn Qiwei Qin qinqw@scau.edu.cn*

#### Specialty section:

*This article was submitted to Comparative Immunology, a section of the journal Frontiers in Immunology*

Received: *25 December 2018* Accepted: *26 February 2019* Published: *19 March 2019*

#### Citation:

*Li C, Liu J, Zhang X, Wei S, Huang X, Huang Y, Wei J and Qin Q (2019) Fish Autophagy Protein 5 Exerts Negative Regulation on Antiviral Immune Response Against Iridovirus and Nodavirus. Front. Immunol. 10:517. doi: 10.3389/fimmu.2019.00517* virus particles. Autophagy, as a cell steward, has been reported to play an important role in virus replication. Many studies have reported that viruses cause accumulation of autophagosomes and exploit these membrane structures as "virus factories" (4). In addition, several studies have confirmed that autophagy interacts with the innate antiviral immune response (5–7). Autophagy can amplify the innate immune response that is mediated by nucleic acid-sensing Toll-like receptors and enhance delivery of cytosolic pathogen-associated molecular patterns (8). Additionally, some autophagy factors can downregulate RIG-I (retinoic acidinducible gene I)-like receptors and type I interferon (IFN) signaling and suppress inflammasome activation (9, 10).

The process of autophagosome formation is regulated by several autophagy-related genes (Atgs) (11). In the autophagy gene superfamily, Atg5 is a key gene that plays an important role in early autophagosome formation. Enhanced or reduced Atg5 levels affect the occurrence and alteration of autophagy pathways. On one hand, Atg5 protein can conjugate to Atg12 to form a complex with the multimeric protein Atg16, and the Atg12-Atg5-Atg16 complex facilitates extension of the autophagosome (12, 13). On the other hand, the combination of Atg5 complex and autophagic vesicle membrane can promote recruitment of LC3 (Atg8) to autophagic vesicles (14). Atg5 is involved in various physiological and pathological processes. In lipid metabolism, silencing or knocking out Atg5 can lead to lipid deposition. Atg5 also plays a role in regulating IFN immune and inflammation responses (7).

The grouper (Epinephelus spp.) is a well-known mariculture species that is widely distributed in South China and Southeast Asia. In 2016, the scale of grouper breeding in China was 108,319 t, which was 8.31% higher than that in 2015. However, outbreaks of viral diseases have caused heavy economic losses in the grouper aquaculture industry. Two representative pathogens are Singapore grouper iridovirus (SGIV) and red-spotted grouper nervous necrosis virus (RGNNV) (15, 16). Current research on the prevention and control of viral diseases in grouper is mainly focused on exploration of the anti-virus immune network and key immune genes. Although numerous immune regulatory molecules have been found to play vital roles in the grouper antiviral response (17–21), the roles of Atgs in the replication of SGIV or RGNNV have not been reported. In other studies of aquatic viruses, proliferation of SVCV was significantly reduced in Beclin-1(Atg6) and LC3(Atg8)-depleted endothelial progenitor cells. However, references to Atg5 in aquatic animal viruses are limited (22).

In this study, we cloned a key autophagy related gene (Atg5) from orange-spotted grouper (E. coioides) (EcAtg5) and investigated the roles of EcAtg5 in autophagy, innate immunity, and cell cycle. Our results provide new insights into the roles of fish Atg5 in virus infection.

### MATERIALS AND METHODS

### Cloning of EcAtg5 and Bioinformatic Analysis

Based on several expressed sequence tag sequences of EcAtg5 from the grouper spleen transcriptome (23), primers (**Table 1**) TABLE 1 | Primers used in this study.


were designed to amplify the full-length open reading frame (ORF) of EcAtg5. Identity analysis between EcAtg5 and other Atg5 sequences was performed using BLASTP searches of the NCBI database. Amino acid alignments were conducted using MEGA5.0 software and edited with the GeneDoc program. The phylogenetic analysis was carried out using the boot-strapped neighbor joining method in ClustalX 2.1 software.

### Tissue Distribution Analysis of EcAtg5

Orange-spotted groupers (30–40 g) used in this study were purchased from a local farm in Hainan Province and kept in a laboratory recirculating seawater system as described previously (20). The relative expression level of EcAtg5 was examined using quantitative real-time PCR (qRT-PCR) in selected tissues, including liver, spleen, head kidney, kidney, heart, intestine, brain, gill, stomach, muscle, fin, and skin. All tissues were collected from three fish.

### Cells and Virus

A grouper spleen (GS) cell line was established in our lab (24), and cells were propagated and maintained at 28◦C in Leibovitz's L-15 medium (Gibco, USA) supplemented with 10% fetal bovine serum (Gibco, USA). SGIV and RGNNV were isolated in our laboratory and propagated in GS cells with titer of 10<sup>5</sup> TCID50/ml as described previously (15, 16).

### Plasmid Construction

To clarify the molecular function of EcAtg5 in vitro, EcAtg5 was subcloned into the vectors pEGFP-C1 and pcDNA3.1-3×HA using the primers listed in **Table 1**. All recombinant plasmids were confirmed by DNA sequencing.

### siRNA-Mediated EcAtg5 Knockdown

GS cells were transfected with EcAtg5 siRNA (siEcAtg5:5′ - GAAAGAGAUGUACCCUGCUGCUUUA-3′ ) or same volume negative control for 24 h, and then infected with SGIV or RGNNV for 12, 24, and 36 h. At the end of each incubation period, the total RNA or protein of cells were extracted for detection.

### Cellular Localization Analysis

GS cells were seeded onto cover slips (10 × 10 mm) in 24-well plates. After allowing the cells to adhere for 24 h, pEGFP-C1 and pEGFP-EcAtg5 plasmids were transfected into GS cells using the transfection reagent Lipofectamine 2000 (Invitrogen, Carlsbad, CA, USA). At 24 h post-transfection, cells on the cover slips were washed with phosphate buffered saline (PBS), fixed with 4% paraformaldehyde for 1 h at room temperature, and then stained with 4,6-diamidino-2-pheny-lindole (DAPI) for 10 min. Finally, cells were observed under a fluorescence microscope (Leica, Wetzlar, Germany).

### RNA Extraction and Gene Expression Analysis

Total RNA isolation was performed using the SV Total RNA Isolation System (Promega, USA) according the manufacturer's instructions, and reverse transcription was carried out using ReverTra Ace (Toyobo, Osaka, Japan). qRT-PCR was performed in an ABI Quant studio 5 device (Applied Biosystems, Carlsbad, CA, USA). The expression levels of viral genes and host immune genes were detected. The relative expression ratio of the selected gene vs. β-actin (reference gene) was calculated using the 2 <sup>−</sup>11CT method. Reactions of SYBR Green were performed in a 10 µl volume containing 5 µl of 2 × SYBR <sup>R</sup> Premix Ex TaqTM, 0.3µl of each forward and reverse primer (10µM), 3.4 µl of water, and 1µl of cDNA. All experiments were performed in triplicate, and the cycling parameters were chosen according to the manufacturer's instructions.

### Western Blot Analysis to Measure Protein Levels

Cells were collected and lysed in RIPA buffer. Proteins were separated by 12% SDS-PAGE and transferred onto Immobilon-P polyvinylidene difluoride membranes (Millipore, Temecula, CA, USA). Blots were incubated with the indicated primary antibody: anti-HA (1:1,000 dilution), anti-β-actin (1:1,000 dilution), anti-ATG5 (1:500 dilution), anti-LC3 (1:1,000 dilution), anti-SGIV major capsid protein (MCP) (1:1,000 dilution), anti-RGNNV capsid protein (CP) (1:1,000 dilution). Subsequently they were incubated with horseradish peroxidase (HRP)-conjugated goatanti-rabbit IgG (1:5,000 dilution). Mouse monoclonal anti-HA antibody was purchased from Sigma (USA). Mouse monoclonal anti-β-actin antibody and rabbit polyclonal ATG5 antibody were purchased from Proteintech (Rosemont, IL, USA). Rabbit monoclonal anti-LC3 antibody was purchased from Abcam (USA). HRP-conjugated goat anti-rabbit and anti-mouse antibodies were purchased from KPL (USA). The polyclonal anti-MCP antibody of SGIV and the polyclonal anti-CP antibody of RGNNV were prepared in our lab. Immunoreactive proteins were visualized using an Enhanced HRP-DAB Chromogenic Substrate Kit (Tiangen, Beijing, China).

### Dual-Luciferase Reporter Assays

To evaluate the promoter activity regulated by EcAtg5, luciferase reporter plasmids, including interferon-sensitive response element (ISRE)-Luc, IFN3-Luc, and nuclear factor (NF)-κB (Clontech, USA), were used for co-transfection. Briefly, GS cells were transiently transfected with the luciferase plasmids together with the indicated EcAtg5 expression vectors using Lipofectamine 2000 reagent. The pRL-SV40 Renilla luciferase vector was used as an internal control. Luciferase activity of total cell lysates was measured using a luciferase reporter assay system (Promega).

## Flow Cytometry Analysis of the Cell Cycle

To evaluate the role of EcAtg5 in cell cycle progression, GS cells were transfected with pcDNA3.1-3×HA-EcAtg5 or the empty vector. At 36 h post-transfection, cells were harvested and fixed in 70% ice-cold ethanol overnight at −30◦C. Cells then were washed with PBS and centrifuged for subsequent incubation in PBS containing 50 mg/mL of propidium iodide (PI) and 100 mg/mL of RNaseA for 30 min. The PI fluorescence was measured with a Beckman Coulter flow cytometer (Brea, CA, USA), and 10,000 cells were analyzed for each sample. The data were analyzed using ModFit LT 4.1 software.

### Statistical Analysis

Statistical analysis was performed using SPSS Version 13. Oneway ANOVA was used to evaluate the variability between treatment groups (∗p < 0.05, ∗∗p < 0.01).

### RESULTS

### Characterization of EcAtg5

The full-length ORF of EcAtg5 was obtained using PCR amplification. Sequence analysis indicated that EcAtg5 encoded a 275-amino acid protein that shared 94% and 81% identity to seabass (Lates calcarifer) and humans (Homo sapiens), respectively (**Figure 1A**). Phylogenetic analysis indicated that EcAtg5 was closely related to the fish subgroup, followed by amphibians, birds, and mammals (**Figure 1B**).

### Expression Patterns of EcAtg5

To analyze the tissue distribution, qRT-PCR was conducted in different tissues of healthy juvenile orange-spotted grouper.

NCBI database. The phylogenetic analysis was carried out using the boot-strapped neighbor joining method in ClustalX 2.1 software.

EcAtg5 was constitutively expressed in all the analyzed tissues in healthy grouper, and it was relatively high mRNA levels in the brain, liver, and fin (**Figure 2A**). To analyze the gene expression profiles in response to different viral infections, the transcription levels of EcAtg5 were examined in RGNNV or SGIV infected cells. The transcription levels of EcAtg5 were significantly increased in RGNNV infected cells. In SGIV infected cells, the expression levels of EcAtg5 first decreased within 24 h post-injection and then increased after 36 h (**Figure 2B**).

### EcAtg5 Encodes a Cytoplasmic Protein

To demonstrate the subcellular localization of EcAtg5, pEGFP-EcAtg5 was transfected into grouper cells, and fluorescence was observed under fluorescence microscopy. Green fluorescence was observed in the cytoplasm in EcAtg5 transfected grouper cells, and most of these cells exhibited fluorescence aggregation (**Figure 3**). In pEGFP-C1 transfected cells, fluorescence was distributed both the cytoplasm and nucleus. The results showed that EcAtg5 was a cytoplasmic protein.

### EcAtg5 Triggered Autophagy in GS Cells

To clarify the function of EcAtg5, the eukaryotic expression vector of pcDNA3.1-3×HA-EcAtg5 was constructed, and the recombinant plasmid successfully expressed HA-EcAtg5 protein

FIGURE 3 | Subcellular localization of EcAtg5 in grouper cells. GS cells were transfected with pEGFP-C1 or pEGFP-EcAtg5. After 24 h, fixed cells were stained with DAPI and imaged by fluorescence microscopy.

after being transfected into GS cells (**Figure 4A**). On the contrary, EcAtg5 protein level was decreased after siRNA

Western blot, and β-actin was used as the internal control.

silencing (**Figure 4B**). Autophagy is characterized by the formation of autophagosomes. Conjugation of the essential LC3 to phosphatidylethanolamine is required for autophagosome biogenesis. Therefore, LC3 lipidation is used as a faithful marker of autophagy activation (22, 25). To assess whether EcAtg5 overexpression affected GS autophagy, we investigated the level of LC3 lipidation in cells overexpressing or silencing EcAtg5. The LC3-II (the lipidated form) level was higher in cells transfected with EcAtg5 compared with control cells (**Figure 4A**), and EcAtg5 knockdown reduced the LC3-II level (**Figure 4B**), which suggests that EcAtg5 might activate autophagy by promoting LC3 lipidation in GS cells.

### EcAtg5 Increased SGIV and RGNNV Replication

To clarify the effects of EcAtg5 overexpression on virus infection, EcAtg5 transfected cells were infected with SGIV or RGNNV, and then viral replication was investigated. Severity of the cytopathic effect (CPE) induced by SGIV infection evoked at 24 h (**Figure 5A**). The amount and severity of vacuoles induced by RGNNV infection also increased in EcAtg5 overexpressing cells compared to the empty vector transfected cells. At the transcription level, the expression of SGIV MCP, ICP18, VP19, and LITAF increased in EcAtg5 overexpressing cells after SGIV infection (**Figure 5B**). The transcription of RGNNV CP and RdRp genes also increased compared with control cells after RGNNV infection (**Figure 5C**). Consistently, ectopic expression of EcAtg5 also increased the protein levels of SGIV MCP and RGNNV CP (**Figure 5D**).

Meanwhile, the effects of silencing EcAtg5 on SGIV and RGNNV replication were also studied. Quantitative analysis results showed that the transcription level of SGIV MCP, ICP18, VP19, and LITAF decreased after EcAtg5 knockdown (**Figure 6A**). The transcription of RGNNV CP and RdRp genes also decreased compared with negative control cells (**Figure 6B**). Consistently, EcAtg5 knockdown decreased the protein levels of SGIV MCP and RGNNV CP in GS cells (**Figure 6C**). Together, EcAtg5 was speculated to promote SGIV and RGNNV replication in GS cells.

### Overexpression of EcAtg5 Decreased the Interferon Immune Response and Pro-Inflammatory Cytokines

To explore the potential mechanism involved in the action of EcAtg5 in fish virus infections, the roles of EcAtg5 on the host interferon immune and inflammation response were evaluated. The pcDNA3.1-3×HA-EcAtg5 and the empty vector were transfected into GS cells, and cells were harvested at 24, 36, and 48 h. The transcription levels of host immune factors and pro-inflammatory cytokines were detected using qRT-PCR. As shown in **Figure 7**, expression levels of interferon related cytokines or effectors, including IRF3, IRF7, MDA5, ISG15, LGP2, MXI, IFP35, MyD88, and TRAF6, were all decreased in EcAtg5 overexpressing cells compared with control vector transfected cells. In addition, we also found that the expressions of pro-inflammatory cytokines such as interleukin (IL)-1β, IL-6, IL-8, and tumor necrosis factor alpha, were all significantly decreased in EcAtg5 overexpressing cells (**Figure 8**).

### EcAtg5 Suppressed ISRE and IFN and NF-κB Promoter Activities

To further explore the roles of EcAtg5 during fish virus infection, the promoter activity of reporter genes in typical antiviral pathways, including ISRE, type I IFN, and NF-κB, were measured

using the plasmids ISRE-Luc, INF-Luc, and NF-κB-Luc. As shown in **Figure 9A**, EcAtg5 overexpression suppressed the promoter activity of these genes. In addition, siEcAtg5 was co-transfected with the plasmids ISRE-Luc, INF-Luc, and NFκB-Luc, and the promoter activity of three reporter genes were measured. The results showed that siRNA-mediated Atg5 knockdown increased the promoter activity of three reporter genes (**Figure 9B**). Thus, we proposed that EcAtg5 negatively regulates NF-κB and the IFN immune responses.

### Effects of EcAtg5 Overexpression on Cell Cycle Progression

Mammalian Atg5 plays a causal role in regulating cell cycle progression (26, 27). Whether EcAtg5 has the similar effects on cell cycle remains uncertain. To explore the role of EcAtg5 on cell cycle progression, GS cells were transfected with pcDNA3.1- 3×HA or EcAtg5. Ectopic expression of EcAtg5 clearly inhibited the G1/S transition compared to the empty vector overexpressing cells. The percentages of G1 phase cells in pcDNA3.1-3×HA and EcAtg5 overexpressing cells were 65.34 and 72.28%, respectively (**Figure 10**). Those results indicated that EcAtg5 may affect cell cycle progression from the G1 to the S phase and arrest cells in the G1 phase.

### DISCUSSION

Autophagy is a highly conserved pathway, and it plays an important role in resistance to intracellular viruses and other pathogens (28). Autophagosome formation relies on the Atg family (29), and Atg5 has been studied in detail in mammals. Atg5 is involved in autophagic membrane extension and curvature, LC3 (Atg8) recruitment, and lysosome and late intracellular regeneration. It also has been implicated in the IFN immune response, inflammation response, and lipid metabolism (25, 30). In fish, the roles of Atg5 gene in zebrafish neurogenesis and organogenesis has been reported, and the results showed that the formation of the Atg5-Atg12 conjugate may depend on Atg5 protein generation and its splicing (31). Twelve autophagyrelated genes from yellow catfish Pelteobagrus fulvidraco and their transcriptional responses to waterborne zinc exposure were also characterized (32). However, little has been reported about the roles of Atg5 in virus replication and its relationship with the innate antiviral immune response in aquatic animals by far. In the present study, an Atg5 homolog from orange-spotted grouper (EcAtg5) was cloned and its roles during fish virus infection were investigated. EcAtg5 encoded a 275-amino acid protein that shared 94% and 81% identity to seabass (Lates calcarifer) and humans (Homo sapiens), respectively. Atg5 functions as an E3

ligase-like enzyme (33). EcAtg5 possesses all the characteristic features of canonical ubiquitin ligase, including two ubiquitinlike domains, a helix-rich domain, and the conserved calpain cleavage sites (32, 33).

As a preliminary step to unravel the physiological role of EcAtg5, the mRNA tissue distribution was determined. The present results indicated that the mRNA expression of EcAtg5 was ubiquitous within all the tested tissues. The ubiquitous distribution suggested that autophagy was implicated in many metabolic pathways among the tissues. Originally autophagy was identified as a response to nutrient deficiency, so it is thought to be a receptor of cellular energy and metabolism (34). However, it is now evident that autophagy can be induced by a variety of factors, including starvation, reactive oxygen

β-actin was used as the internal control. Band intensity was calculated using Quantity-one software and ratios of MCPor CP/β-actin was assessed (*p* < 0.05).

species, endoplasmic reticulum stress, microbial invasion and so on (35). Based on this, we detected the expression of EcAtg5 under the stimulation of two viruses. Transcription levels of EcAtg5 increased from the early stage of RGNNV infection, suggesting that RGNNV infection may significantly induce autophagy activity to facilitate its proliferation. In SGIV infected cells, the expression levels of EcAtg5 firstly decreased within 24 h post-infection and then increased after 36 h. This pattern might be caused by the lack of cellular nutrition. With the cell growth, metabolism and virus replication, the nutrient deficiency in cells will increase autophagy activity over time.

Autophagy is an important cellular process by which Atg5 initiates the formation of double membrane vesicles (DMVs). Recently, the contribution of an autophagy protein, Atg5, to viral replication has been demonstrated (36), and Atg5 was identified as an interacting protein for the hepatitis C virus NS5B (37). The altered expression level of EcAtg5 in GS cells infected with SGIV and RGNNV suggested that EcAtg5 might play an essential role in the grouper response to fish virus infection, so the impact of EcAtg5 overexpression on virus proliferation were investigated. Overexpression of EcAtg5 promoted SGIV and RGNNV replication, evidenced by the severity of CPE, the increased transcription levels of viral genes, and the increased levels of viral proteins. Knockdown of EcAtg5 decreased SGIV and RGNNV replication by assessing transcription and protein levels of viral genes. The results suggested that Atg5 might share conserved function to viral replication from fish to mammals.

Studies of mammals suggested that Atg5-Atg12 promotes viral replication by negatively regulating the IFN response (38). The Atg5-Atg12 conjugate interacts directly with the mitochondrial antiviral-signaling protein (MAVS) and retinoic acid-inducible gene I (RIG-I) through the N-terminal caspase recruitment domain (CARD), resulting in inhibition of type I IFN production (38). N-terminal fragments of RIG-I and IFN-α possess great capacity to activate IFN-β and ISRE promoters in Atg5 deficient cells (39). When the formation of autophagosomes was promoted, the activity of the IFNβ promoter was decreased so that autophagy contributed to sustained hepatitis C virus infection (40). Here, overexpression of EcAtg5 in grouper cells not only decreased the expression levels of several interferon related cytokines or effectors, but also negatively regulated the expression of pro-inflammatory factors. Moreover, the ectopic expression of EcAtg5 significantly decreased ISRE, IFN, and NF-κB promoter activities, and

knockdown of EcAtg5 raised promoter activities of these reporter genes. Atg5-Atg12 inhibits the production of IFN in canonical autophagy while playing the opposite role in alternative autophagy (41). Thus, overexpression of EcAtg5 might activate canonical autophagy in GS cells. Overexpression of EcAtg5 up-regulated the level of LC3-II, indicating that EcAtg5 can activate autophagy. Taken together, we speculated that EcAtg5 decreased interferon immune response and activated autophagy might contribute greatly to its promoting effect on SGIV and RGNNV replication.

In mammals, Atg5 can induce cell cycle arrest at the G1/S phase by up-regulating expression of p21 (a cyclindependent kinase inhibitor) at the level of post-transcription in response to challenges such as nutrient deficiency (42, 43). Considering that Atg5 is a key and relatively conserved protein, we speculated that fish Atg5 might play a similar role in cell cycle progression. In the present study, EcAtg5 affected cell cycle progression from the G1 to the S phase and arrested cells in the G1 phase. It was also reported that the replication level and virus titer of RGNNV were greater in cells released from the G1 phase or S phase of the cell cycle compared to cells released from the G2 phase (44). Those results suggested that overexpression of EcAtg5 may facilitate RGNNV replication. However, whether fish Atg5 affects the cell cycle by regulating p21 requires further investigation.

In conclusion, a key autophagy related gene (Atg5) from orange-spotted grouper (E. coioides) (EcAtg5) was cloned, and the roles of EcAtg5 in autophagy, innate immunity, and cell cycle were investigated in this study. The results showed that EcAtg5 plays crucial roles in virus replication via promoting autophagy, down-regulating antiviral IFN responses, and affecting cell cycle. This study identified a link between the autophagic machinery and innate immune signaling against viral infection.

## DATA AVAILABILITY

The datasets for this manuscript are not publicly available because this data has not been published. Requests to access the datasets should be directed to Jingguang Wei, weijg@scau.edu.cn.

### ETHICS STATEMENT

All animal-involving experiments of this study were approved by the Animal Care and Use Committee of College of Marine Sciences, South China Agricultural University, and all efforts were made to minimize suffering.

### AUTHOR CONTRIBUTIONS

QQ and JW designed the experiments. CL performed the majority of the experiments, analyzed data, and wrote the manuscript. JL and XZ contributed experimental suggestions. SW, YH, and XH helped to design the experiments. All authors revised the manuscript.

### FUNDING

This work was supported by grants from the National Key R&D Program of China (2018YFC0311302, 2018YFD0900501, and 2017YFC1404504), China Agriculture Research System (CARS-47-G16), National Natural Science Foundation of China (31572643, 31772882, and 41506176), Open Fund of Key Laboratory of Experimental Marine Biology, Chinese Academy

### REFERENCES


of Sciences (No. KF2018NO3), Science and Technology Planning Project of Guangdong Province, China (2015TQ01N118), and National High Technology Development Program of China (863) (2014AA093507).


44. Mai W, Liu H, Chen H, Zhou Y, Chen Y. RGNNV-induced cell cycle arrest at G1/S phase enhanced viral replication via p53-dependent pathway in GS cells. Virus Res. (2018) 2:142–52. doi: 10.1016/j.virusres.2018.06.011

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

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

# The Fish-Specific Protein Kinase (PKZ) Initiates Innate Immune Responses via IRF3- and ISGF3-Like Mediated Pathways

Xiaowen Xu<sup>1</sup> , Meifeng Li <sup>1</sup> , Chuxin Wu<sup>1</sup> , Dongming Li <sup>2</sup> , Zeyin Jiang<sup>1</sup> , Changxin Liu<sup>1</sup> , Bo Cheng<sup>1</sup> , Huiling Mao<sup>1</sup> and Chengyu Hu<sup>1</sup> \*

<sup>1</sup> College of Life Science, Nanchang University, Nanchang, China, <sup>2</sup> Fuzhou Medical College, Nanchang University, Fuzhou, China

#### Edited by:

Stephanie DeWitte-Orr, Wilfrid Laurier University, Canada

#### Reviewed by:

Maria Del Mar Ortega-Villaizan, Universidad Miguel Hernández de Elche, Spain Eduardo Gomez-Casado, Instituto Nacional de Investigación y Tecnología Agraria y Alimentaria (INIA), Spain

> \*Correspondence: Chengyu Hu hucy2008@163.com

#### Specialty section:

This article was submitted to Comparative Immunology, a section of the journal Frontiers in Immunology

Received: 06 September 2018 Accepted: 04 March 2019 Published: 28 March 2019

#### Citation:

Xu X, Li M, Wu C, Li D, Jiang Z, Liu C, Cheng B, Mao H and Hu C (2019) The Fish-Specific Protein Kinase (PKZ) Initiates Innate Immune Responses via IRF3- and ISGF3-Like Mediated Pathways. Front. Immunol. 10:582. doi: 10.3389/fimmu.2019.00582 PKZ is a fish-specific protein kinase containing Zα domains. PKZ is known to induce apoptosis through phosphorylating eukaryotic initiation factor 2α kinase (eIF2α) in the same way as double-stranded RNA-dependent protein kinase (PKR), but its exact role in detecting pathogens remains to be fully elucidated. Herein, we have found that PKZ acts as a fish-specific DNA sensor by initiating IFN expression through IRF3- or ISGF3-like mediated pathways. The expression pattern of PKZ is similar to those of innate immunity mediators stimulated by poly (dA:dT) and poly (dG:dC). DNA-PKZ interaction can enhance PKZ phosphorylation and dimerization in vitro. These findings indicate that PKZ participates in cytoplasmic DNA-mediated signaling. Subcellular localization assays have also shown that PKZ is located in the cytoplasm, which suggests that PKZ acts as a cytoplasmic PRR. Meanwhile, co-IP assays have shown that PKZ can separately interact with IRF3, STING, ZDHHC1, eIF2α, IRF9, and STAT2. Further investigations have revealed that PKZ can activate IRF3 and STAT2; and that IRF3-dependent and ISGF3-like dependent mediators are critical for PKZ-induced IFN expression. These results demonstrate that PKZ acts as a special DNA pattern-recognition receptor, and that PKZ can trigger immune responses through IRF3-mediated or ISGF3-like mediated pathways in fish.

Keywords: PKZ, PRR, IRF3-mediated and ISGF3-like mediated pathways, innate immunity, teleost

### HIGHLIGHTS


### INTRODUCTION

Innate immunity is the first defense line for vertebrates, especially lower vertebrates, to fight off invading pathogens (1, 2). Patternrecognition receptors (PRRs) that recognize pathogen-associated molecular patterns (PAMPs) play vital roles in initiating IFN signaling pathways (3–5). In recent years, many complicated and effective PRRs have been identified in mammals. A few toll-likereceptors (TLR) and RIG-I-like receptors (RLR) are known to detect bacterial and viral RNAs (6, 7). TLR3 recognizes doublestranded RNA (dsRNA), and then triggers IFN-β expression by activating interferon transcription factor 3 (IRF3) and nuclear factor κB (NF-κB) (8, 9). RLRs (RIG-I and MDA5) detect cytosolic RNA via the adaptor protein IPS-1 and activate IFN-β signaling pathways (10).

In addition to the RNA sensors described above, a number of intracellular DNA sensors have been confirmed in recent times, such as absent in melanoma 2 (AIM2) (11), DNAdependent activator of IFN-regulatory factors (DAI) (12), IFNinducible protein 16 (IFI16) (13), DEAD (Asp-Glu-Ala-Asp) box polypeptide 41 (DDX41) (14), RNA polymerase III (15), and Cyclic GMP-AMP synthase (cGAS) (16). These proteins are important in responding to various DNA infections. For example, IFI16 is a sensor for HIV DNA through interaction with stimulator of IFN gene (STING), and is also important in inducing IFN-β reactions in human macrophages (13). Cyclic GMP-AMP synthase (cGAS) is known to detect cytosolic DNA in various cell types (16). Recognition of viral DNA by cGAS induces the synthesis first of the second message cGAMP from ATP and GTP, and then of cGAMP which binds to STING; and STING enhances IRF3 phosphorylation by TBK1, thus activating downstream genes (17–21).

In recent years, much progress has been made in the study of fish PRRs. Along with the discovery of RNA sensors in mammals, RLRs (RIG-I, MDA5, and LGP2) and TLRs, which are involved in dsRNA recognition, were also identified in teleost fish (22–27). While the findings on RNA-recognizing PRRs in fish are quite clear, those on DNA-recognizing PRRs are ambiguous. A novel member of vertebrate eIF2α kinase, named PKZ (protein kinase containing Zα binding domains), has been identified exclusively in fish (28–30). The Zα binding domains of fish PKZ similar to the DNA sensors in mammals (28–30). This suggests PKZ might be a special DNA sensor in fish (31). Recent studies have demonstrated that fish PKZ can specifically respond to poly (dG:dC), then phosphorylate eIF2α, and finally lead to apoptosis (30, 32, 33).

The present study has demonstrated that the expression profile of PKZ is similar to those of IFN response mediators; that PKZ can specifically recognize poly (dA:dT) and poly (dG:dC) but not poly I:C (dsRNA analog) in vitro; and that PKZ is located in the cytoplasm and is a basic factor in DNA recognition. As well, when cells are stimulated by invading DNA molecules, PKZ can interact with and activate the mediators in the IRF3-dependent or ISGF3-like dependent pathway. Our findings suggest that PKZ performs as a special DNA-sensing receptor and triggers immune responses through the IRF3-dependent or ISGF3-like dependent pathway.

### MATERIALS AND METHODS

### Cells and Virus Analogs

Grass carp (Ctenopharyngodon idellus) ovary cells (CO) and C. idellus kidney cells (CIK) were cultured in medium 199 supplemented with 10% FCS at 28◦C. Human Embryonic Kidney 293T cells (HEK-293T) were maintained at 37◦C under 5.0% CO<sup>2</sup> in DMEM supplemented with 10% FCS. CO cells are superior to CIK cells due to its stronger ability of expression and better cell morphology. In addition, HEK-293T cells were used instead of CIK and CO cells in immunoprecipitation assay because of their super-high transfection efficiency.

ISD-PS (biotin tagged ISD-PS), poly (dA:dT) [biotin tagged poly (dA:dT)] and poly (dG:dC) [biotin tagged poly (dG:dC)] were all synthesized by Sangon Biotech (China) and their sequences were presented in **Table 1**. ISD-PS is an RNA analog carrying a biotin tag (34). Poly I:C was purchased from Sigma (USA). The nucleic acids [poly (dA:dT), poly (dG:dC), and ISD-PS] were synthesized in accordance with the manufacturer's protocols from 10 µl DNA oligo, 2 µl 10 × oligo annealing buffer (100 mM Tris-HCl, pH 8.0, 10 mM EDTA, pH8.0, 1 M NaCl), and 8 µl DNase/RNase-free water. The 20 µl total- reaction mixture was incubated at 95◦C for 4 min, and then kept at room temperature for 5–10 min.

### Plasmids and Recombinant Construction

pCMV-FLAG and pcDNA3.1(+) were both purchased from Invitrogen (USA). pEGFP-C1, pGL3, and E. coli strains DH5α were all bought from Promega (USA). The C. idellus open reading frames (ORFs) of PKZ, IRF3, STING, ZDHHC1, IRF9, STAT2, eIF2α, and ADAR1 were prepared, and each of them was separately inserted into pCMV-FLAG, pEGFP-C1, and pcDNA3.1(+) vectors to construct expression plasmids. They were used for in vitro translation, nucleic acids pull down, sublocalization, and co-IP assays. Each of the C. idellus mutants of PKZ-C (176–513 aa), PKZ-N (1–169 aa), and PKZ-198K (K198R) was separately inserted into pcDNA3.1 (+); meanwhile, each of the mutants mentioned above was also individually inserted into pCMV-FLAG (**Figure 3A**). All of the constructs were confirmed by DNA sequencing. The primers used for plasmid construction were given in **Table 1**. The plasmids, genes, types of experiment, and Genbank ID for this paper were listed in **Table 2**.

### Luciferase Activity and Quantitative RT-PCR Assays

Each plasmid was transfected according to the FuGENE <sup>R</sup> 6 (Promega, USA) protocol. Each well of 6-well plate was transfected with 2 µg of plasmid, and each 10-cm dish was transfected with 5 µg of plasmid. In nucleic acid stimulation, nucleic acid (2 µg) was transfected into cells. In luciferase activity assays, 0.25 µg of PKZ-pcDNA3.1 (or basic-pcDNA3.1 plasmid), 0.25 µg of IFN-pro-pGL3, and 0.025 µg of pRL-TK Renilla luciferase vectors were transfected in CIK or CO cells. IFN promoter was inserted into pGL3 plasmid. As this recombinant plasmid was used in dual-luciferase activity assay, the luciferase activity of the promoter can reflect IFN transcription levels.



In the quantitative RT-PCR assay, CIK cells were seeded in 6 well plates and transfected with 2 µg of poly (dA:dT) or 2 µg of poly (dG:dC). After all RNA was extracted from CIK cells, qRT-PCR was also used to detect the expression levels of IFN, IPS-1, TBK1, and PKZ. Grass carp IFN expression was detected in CO cells transfected with 2 µg of PKZ-pcDNA3.1, 2 µg of IRF3-pcDNA3.1, 2 µg of STING-pcDNA3.1, 2 µg of ZDHHC1 pcDNA3.1, 2 µg of IRF9-pcDNA3.1, 2 µg of STAT2-pcDNA3.1, and 2 µg of empty vectors. Twenty-four hours later, the cells were transfected with 2 µg of poly (dG:dC) for 6 h.

### Subcellular Localization Analysis

In subcellular localization analysis, CO cells were plated on 35 cm<sup>2</sup> microscopic petri dishes. After 12 h they were transfected with PKZ-GFP, ADAR1-GFP, and GFP-C1. Twenty-four hours post transfection, the cells were washed three times with PBS, and fixed with 4% (v/v) paraformaldehyde at room temperature for 15 min. The cells were then dyed with DAPI (0.1µg/ml) and examined under a confocal microscope (Olympus, FV1000).

### Nucleic Acids Pulldown Assays

In nucleic acids pulldown assays, synthetic ISD-PS was used as RNA analog. Two microgram each of biotin-labeled ISD-PS, poly (dA:dT), and poly (dG:dC) were incubated with M-280 streptavidin-coupled Dynabeads (Invitrogen) for 5 h. The unbound biotin-labeled nucleic acids were washed with 1×BW buffer (5 mM Tris-HCl pH 7.5, 0.5 mM EDTA, 1 M NaCl) five times and immobilized on M-280 streptavidin-coupled Dynabeads (Invitrogen) at 4◦C. CO cells were plated on 10 cm diameter dishes and transfected with 5 µg of PKZ-FLAG (or PKZ-C-FLAG or PKZ-N-FLAG) plasmids. Twenty-four hours later, the cells were stimulated with poly (dG:dC) for 12 h. The medium was carefully removed and washed twice with PBS, and lysates were cleared by sonication and centrifugation. Cleared lysates were incubated with Dynabeads which were separately immobilized with biotin-nucleic acids (poly (dA:dT), poly (dG:dC) and ISD-PS at 4◦C on a rotary wheel for 12 h. The experiment was divided in two groups. In non-competitor group, lysates were incubated with Dynabeads immobilized with biotin-nucleic acids; in competitor group, lysates, and competitors were incubated with Dynabeads immobilized with biotin-nucleic acids. Ten micrograms each of non-biotin tagged nucleic acids (poly (dA:dT), poly (dG:dC), and ISD-PS) acted as competitors. Highly concentrated non-biotin tagged nucleic acids in incubations (lysates and Dynabeads immobilized with biotin-nucleic acids) will competitively interact with proteins. The unbound proteins in Dynabeads were removed by five consecutive washes with 1 × BW buffer. Given that the proteins can interact with nucleic acids, PKZ-FLAG can be detected in the non-competitor group and cannot be detected in the competitor group. The beads of what were eluted with 100 µl of 2×SDS sample buffer by boiling for 7 min at 95◦C, before performing Western blot analysis.

### Antibodies, Western Blot, Co-immunoprecipitation (co-IP) Assays, and Analysis of Phosphorylation State

Mouse monoclonal antibodies against FLAG and GFP were purchased from Sigma (USA) and Abmart (USA), respectively. Phospho-IRF3 (Ser 396) rabbit monoclonal antibody was purchased from Cell Signaling Technology (USA). Anti-GAPDH rabbit polyclonal antibody was purchased from CWBIO (China). Rabbit polyclonal anti-grass carp PKZ, IFN, IRF3, STING, and ZDHHC1 antisera were prepared in our lab (33, 35). HEK-293T cells were used in co-IP assays instead of fish cells because of their superior transfection efficiency. Co-IP assays and Western blot were performed as previously described (36, 37).


In the absence of specific phospho-PKZ and phospho-STAT2 antibodies, immunoprecipitation assays were performed to study PKZ and STAT2 phosphorylation. To begin with, CO cells were transfected with 2 µg of PKZ-FLAG/STAT2-GFP plasmids. Twenty-four hours later, the cells were separately stimulated with poly I:C, poly (dA:dT) and poly (dG:dC) for 12 h. All the proteins were then harvested with lysis (Transgene, China). First, the amount of PKZ-FLAG/STAT2-GFP proteins was determined by Western blot. The concentrations of protein in lysates were determined by an Enhanced BCA Protein Assay kit (Beyotime). Then the same amount of PKZ-FLAG/STAT2- GFP was immunoprecipitated from cell lysates. Finally, the phosphorylation of PKZ/STAT2 was detected with an antiphospho-Ser/Thr/Tyr antibody (AnaSpec, CA).

### Native PAGE

CO cells were separately transfected with the negative control (4 µg of basic-pcDNA3.1), the positive control [4 µg of poly I:C, 4 µg of poly (dA:dT), 4 µg of poly (dG:dC)], the experimental group (4 µg of PKZ-pcDNA3.1) plasmids. The lysis of total protein was studied in two groups of experiment. One group was detected in SDS-PAGE; another group detected in native PAGE. Native PAGE assays were conducted with 8% acrylamide gel without SDS. In the same way as described previously, the sample was dissolved in 5 × loading buffer without SDS. The gel was run with 0.5% deoxycholate, 25 mM Tris-HCl (pH 8.4), and 192 mM glycine for 60 min at 35 mA on ice, and then immunoblotted to a nitrocellulose membrane (Millipore, USA).

### RNA Interference-Mediated Gene Knockdown Assays

In the RNA interference-mediated gene knockdown assays, the specific siRNA sequence against C. idellus PKZ, IRF9, and the negative control RNA (N.C.) oligonucleotides were prepared (Shanghai GenePharma Co., Ltd.) (**Table 1**). The siRNA sequence against IRF3 was used as in the previous study (35). The transfection reagent, Hiperfect <sup>R</sup> (Qiagen, Germany), has been widely used in our previous study (35). The transfection was carried out according to the manufacturer's protocol and instructions.

### Data Analysis

Statistical analysis of qRT-PCR was performed and graphs were prepared using Microsoft excel. The Gray value of western blot was analyzed and confocal images were made by image J. The diagram of signaling pathway was constructed with the Portable Pathway Builder Tool 2.0. Data analysis of qRT-PCR and dualluciferase assay were both presented using the unpaired twotailed t-test, and p < 0.05 were considered statistically significant ( <sup>∗</sup>p < 0.05, ∗∗p < 0.01).

### RESULTS

### PKZ and Some Mediators of Immune Responses Are Up-Regulated Under Stimulation With Poly (dA:dT) and Poly (dG:dC)

To delineate the functional link of PKZ with IFN, IRF3, STING, ZDHHC1, IPS-1, and TBK1, their expression patterns in CIK cells were identified under the same conditions. PKZ, IFN, IRF3, STING, and ZDHHC1 were upregulated under stimulation with poly (dA:dT) (**Figure 1A**) and poly (dG:dC) (**Figure 1B**) at mRNA level. It was also found that when the cells were stimulated with poly (dA:dT) and poly (dG:dC), PKZ transcripts showed a significant increase at the early phase of stimulation, and the transcripts of STING, IRF3, and ZDHHC1 showed similar increases. The increase in transcripts of IFN, IPS-1, and TBK1 appeared to be biphasic under stimulation with poly (dA: dT) rather than poly (dG: dC). PKR acted as a negative control in all cases. These results indicate that fish

and then the cells were separately transfected with poly (dA:dT) or poly (dG:dC) (2µg/ml) for 0, 6, 12, 24, 48, and 72 h. Then the cells were sampled, and the mRNAs (Continued)

FIGURE 1 | of IFN, IPS-1, STING, PKZ, TBK1, IRF3, and ZDHHC1 were detected by qRT-PCR. The mRNA expression levels of target genes were normalized relative to the reference gene β-actin. The groups of 0 h were controls, and the others (6, 12, 24, 48, 72 h) were experimental groups. Data represent mean ± SD (n = 3) of three experiments and were tested for statistical significance using unpaired two-tailed t-test \*p < 0.05, \*\*p < 0.01. The asterisk above the bracket indicated statistical significance between the control groups with experimental group.

PKZ responds to poly (dA:dT) and poly (dG:dC) stimulations, just like some well-known mediators of immune response. PKZ is closely related to innate immune activity and may act as an up-stream protein under stimulation with poly (dA:dT) or poly (dG:dC).

### PKZ Is Located in the Cytoplasm

It is necessary to determine the location of PKZ for its recognition of viral nucleic acids in fish cells. The plasmids PKZ-GFP and ADAR1-GFP were separately transfected into CO cells. Confocal microscopic detection suggests that PKZ-GFP is located in the cytoplasm; however, ADAR1-GFP (another Zα protein, used as a control) is located in the nuclear (**Figure 2**).

## PKZ Can Bind to Both Poly (dA:dT) and Poly (dG:dC) in vitro

PKZ has two Zα domains at its N-terminus and 11 conserved eIF2α kinase catalytic sub-domains at its C-terminus (**Figure 3A**). A series of PKZ protein mutants were constructed, including C-terminus of PKZ (PKZ-C-pcDNA3.1/FLAG), Nterminus of PKZ (PKZ-N-pcDNA3.1/FLAG), and Lys<sup>198</sup> mutant of PKZ(PKZ-198K-pcDNA3.1/FLAG) and these were used for subsequent experiments (**Figure 3A**). It is known that the Lys<sup>198</sup> mutant of PKZ lacks kinase activity (31, 33). To investigate the compatibility of PKZ with nucleic acids, nucleic acid pulldown assays were performed by incubating biotin-labeled DNA with PKZ protein. In this assay, non-biotin tagged nucleic acids [poly

FIGURE 3 | PKZ binds to poly (dA:dT) and poly (dG:dC) in vitro. (A) PKZ and the mutants. (B) CO cells were seeded on 10 cm diameter dishes and transfected with PKZ-FLAG, at 24 h post-transfection, cells were harvested, a total of 1% of PKZ-FLAG containing cell lysates was loaded as input, and the rest of lysates incubated with Dynabeads were separately immobilized with 2 µg of biotin-nucleic acids (poly (dA:dT), poly (dG:dC) and ISD-PS), and 10 µg of non-biotin tagged nucleic acids acted as competitors. The DNA in Dynabeads was detected in agarose gel. (C) PKZ mutants (PKZ-C-FLAG and PKZ-N-FLAG) were transfected into CO cells, at 24 h post-transfection, cells were harvested, a total of 1% of PKZ-FLAG containing cell lysates was loaded as input, and the rest of lysates incubated with Dynabeads were individually immobilized with biotin-nucleic acids (poly (dA:dT) and poly (dG:dC), and 10 µg of non-biotin tagged nucleic acids acted as competitors. The DNA in Dynabeads were examined in agarose gel. The molecular masses of PKZ-FLAG, PKZ-C-FLAG, and PKZ-N-FLAG are 67, 45, and 25 KD, respectively. Each Western blot is representative of at least three independent experiments.

(dA:dT), poly (dG:dC), and ISD-PS] acted as competitors. High concentrations of non-biotin tagged nucleic acids in incubations (lysates and Dynabeads immobilized with biotin-nucleic acids) will result in competitive interaction with proteins. Qualitative analysis of nucleic acids-bound protein with ISD-PS being used as a negative control (**Figure 3B**) showed that PKZ-FLAG effectively binds to poly (dA:dT) and poly (dG:dC), but not to ISD-PS (RNA analog). While Zα domains (PKZ-N terminus) can combine with both poly (dA:dT) and poly (dG:dC), eIF2α kinase catalytic sub-domain (PKZ-C terminus) cannot with either of them (**Figure 3C**). These results indicate that fish PKZ can recognize and interact with DNA by its Zα domain.

### PKZ Is Activated by DNA Stimulation

To further investigate whether PKZ can be phosphorylated or not, CO cells were transfected with PKZ-FLAG. Twenty-four hours later, cells were separately transfected with 2 µg of poly I:C, 2 µg of poly (dA:dT), and 2 µg of poly (dG:dC) for 12 h. PKZ-FLAG protein was immunoprecipitated from cell lysates. The phosphorylation of PKZ-FLAG was detected with an antiphospho-Ser/Thr/Tyr antibody. These results show that poly (dA:dT) and poly (dG:dC) can promote PKZ phosphorylation but poly I:C cannot; and poly (dG:dC) had stronger effects on PKZ than poly (dA:dT) (**Figure 4A**). Poly (dG:dC) was therefore selected as a stimulator for PKZ-dependent pathway in subsequent experiments.

Subsequently, two sets of PKZ constructs containing FLAG tag and GFP tag were used in the co-IP assay. Co-IP assay was performed for HEK-293T cells. The plasmids of PKZ-FLAG and PKZ-GFP were co-transfected into HEK-293T cells. Twenty-four hours post-transfection, the cells were divided into two groups, namely the poly (dG:dC)-stimulated group and a non-stimulated group. In the poly (dG:dC)-stimulated group, anti-GFP antibody-immunoprecipitated protein complex was detected by anti-FLAG (PKZ-FLAG) antibody; which was however not detected in the non-stimulated group. Likewise, anti-FLAG antibody-immunoprecipitated protein complex was detected with the anti-GFP (PKZ-GFP) antibody in the poly (dG:dC) stimulated group. IgG-immunoprecipitated protein complex was not detected with anti-GFP (PKZ-GFP) or anti-FLAG (PKZ-FLAG) antibody (**Figure 4B**). Co-IP results suggest that poly (dG:dC) promotes dimerization of PKZ. Overall, these findings demonstrate that fish PKZ can be activated by DNA stimulation.

### PKZ Initiates IFN Expression

To confirm whether or not PKZ can trigger innate immunity, PKZ-pcDNA3.1 was over-expressed in both CIK cells (**Figure 5A**) and CO cells (**Figure 5B**). Twenty-four hours after transfection, cells were treated with equal amount of poly (dG:dC) for 6 h. The promoter activity of grass carp IFN was increasingly upregulated in comparison with that of the cells transfected with basic-pcDNA3.1, and results indicated that IFN transcription level was upregulated in CIK or CO cells overexpressed PKZ. However, after knockdown of PKZ, the mRNA and protein levels of PKZ and IFN were significantly downregulated (**Figures 5C,D**). In addition, IFN transcription

dishes were separately co-transfected with PKZ-FLAG and PKZ-GFP. This experiment was divided into two groups: one group was stimulated with poly (dG:dC), and the other was not. Forty eight hours later, cell lysates were separately immunoprecipitated with anti-FLAG antibody, anti-GFP antibody and IgG (used as a control), then the immunoprecipitants were examined by Western blotting. Each Western blot is representative of at least three independent experiments.

level was upregulated by transfection with PKZ in higher concentrations in CO cells (stimulated with poly (dG:dC) of equal strength for 6 h prior to experiments) (**Figure 5E**). The transfection efficiency was shown in **Supplemental Figure 1**. These results indicate that PKZ plays critical roles in inducing IFN under DNA stimulation.

In subsequent assays, the mutants were used to investigate the relationship between the PKZ structure and IFN expression. The mRNA and protein levels of IFN were upregulated in the overexpressed PKZ-C terminus and PKZ-wt. However, no evidence was found that the PKZ-N terminus and Ser<sup>198</sup> PKZ deficit contributed to IFN expression (**Figures 5F,G**). These data suggest that PKZ can promote IFN expression by PKZ-C terminus.

### PKZ Interacts With Mediators of IRF3- and ISGF3-Like Dependent Pathway

To further explore the mechanism of PKZ upregulating IFN expression, it is necessary to identify the PKZ-dependent pathway. A previous study has found that PKZ can interact with eIF2α (31), so eIF2α was chosen as a positive control in this experiment. Co-IP assays were performed with HEK293T cells, in which PKZ-FLAG were individually co-overexpressed with

FIGURE 5 | PKZ initiates IFN expression. (A,B) CIK and CO cells seeded in 24-well plates were co-transfected with IFN-pro and PKZ-pcDNA3.1. Basic-pcDNA3.1 was used as a control. A total of 0.025 µg of pRL-TK was included to standardize the expression level. Twenty four hours later, the cells were stimulated with poly (dG:dC) for 12 h. The transfected cells were harvested to detect the luciferase activities. Data were expressed mean ± SD (n = 3) and were tested for statistical significance using unpaired two-tailed t-test \*p < 0.05, \*\*p < 0.01. Statistical significance was analyzed between the control group (basic-pcDNA3.1) with experimental group (PKZ-pcDNA3.1). (C,D) CO cells seeded in 6-well plates were transfected with siRNA-PKZ. Twenty four hours later, the cells were stimulated with poly (dG:dC) for 12 h. The whole-cell mRNA and protein were prepared; qRT-PCR and Western blot were used to exam the expression profiles of PKZ and IFN when we knocked down PKZ in CO cells. The data from qRT-PCR represent mean ± SD (n = 3) and were tested for statistical significance using unpaired two-tailed t-test \*p < 0.05, \*\*p < 0.01. Statistical significance was analyzed between the control groups (N.C) and the experimental groups (siRNA-PKZ). (E) CO cells seeded in 6-well (Continued)

FIGURE 5 | plates were individually transfected with PKZ-pcDNA3.1 in concentrations of 0, 1, 2, and 4 µg, then qRT-PCR was used to detect the expression of IFN. qRT-PCR represents mean ± SD (n = 3) and was tested for statistical significance using unpaired two-tailed t-test \*p < 0.05. Fold changes were determined relative to cell transfected with 0 µg of PKZ-pcDNA3.1. (F,G) CO cells seeded in 6-well plates were separately transfected with PKZ-pcDNA3.1, PKZ-C-pcDNA3.1, PKZ-N-pcDNA3.1, PKZ-198K-pcDNA3.1, and basic-pcDNA3.1. Twenty four hours later, the cells were stimulated with poly (dG:dC) for 12 h. qRT-PCR and Western blot were used to exam the induction of IFN. The data from qRT-PCR represent mean ± SD (n = 3) and were tested for statistical significance using unpaired two-tailed t-test \*p < 0.05. Statistical significance was analyzed between the control groups (basic-pcDNA3.1) and the experimental groups (PKZ-pcDNA3.1, PKZ-C-pcDNA3.1, PKZ-N-pcDNA3.1, PKZ-198K-pcDNA3.1). Each Western blot is representative of at least three independent experiments.

the following recombinant plasmids (IRF3-GFP, STING-GFP, ZDHHC1-GFP, IRF9-GFP, STAT1-GFP, STAT2-GFP, STAT6-GFP, and eIF2α-GFP). Forty-eight hours post-transfection, the cells were stimulated with poly (dG:dC) for 6 h before all the proteins were harvested. IgG-immunoprecipitation acts as a negative control, anti-FLAG antibody-immunoprecipitation (**Figure 6A**) and anti-GFP antibody-immunoprecipitation (**Figure 6B**) show significant interaction between PKZ-FLAG and IRF3-GFP, STING-GFP, ZDHHC1-GFP, eIF2α-GFP, IRF9-GFP, and STAT2- GFP. These results show that PKZ can interact with IRF3, STING, ZDHHC1, eIF2α, IRF9, and STAT2.

### PKZ Initiates IFN Expression Through IRF3- and ISGF3-Like Dependent Pathways

Subsequent experiments were conducted to determine the relationships among PKZ and some common mediators of innate immunity. Dimerization and phosphorylation levels of IRF3 were increased when CO cells were transfected with poly I:C, poly (dA:dT), poly (dG:dC) and PKZ-pcDNA3.1, with pcDNA3.1 basic acting as negative control (**Figure 7A**). Interestingly, STAT2 phosphorylation activity was also enhanced in the process (**Figure 7B**). Moreover, IFN expression was upregulated in CO cells in which IRF3, STING, ZDHHC1, IRF9, and STAT2 were overexpressed. However, the knockdown of PKZ blocked the IFN expression (**Figures 7C–E**), whereas knockdown of IRF3 or IRF9 inhibited the IFN expression through PKZ (**Figures 7F,G**). The effect of siRNA-IRF9 was examined in **Supplemental Figure 2**. These data suggest that PKZ initiated immune responses through mediators of the IRF3 dependent pathway (IRF3, STING, and ZDHHC1) as well as the ISGF3-like (IRF9 and STAT2) dependent pathway (**Figure 8**).

### DISCUSSION

In mammals, eIF2α is phosphorylated with one of the eIF2α kinases that consist of PKR, PKR-like ER kinase (PERK), general control non-derepressible-2 (GCN2), and hemeregulated eIF2α kinase (HRI) (38). Among these kinases, PKR is most widely studied. PKR is known to mediate IκBkinase β (IKKβ) phosphorylation and be able to activate the NF-κB pathway (39). PKR also inhibits viral protein synthesis via eIF2α phosphorylation (40) and thus helps resist virus infection (41, 42). Because the fish-specific protein kinase PKZ is homologous with mammalian PKR, it was believed—for a short while—that it is a duplicate of mammalian PKR (28–30). That PKZ possesses some functions similar to those of PKR (31, 33, 43–45) is not in question.

Fish PKZ have been identified in C. idellus, Carassius auratus, Atlantic salmon, Gobiocypris rarus, and Danio rerio (28–30, 32, 44). BLAST homologous research indicates that C. idellus PKZ full-length cDNA has 2,158 bp with a largest open reading frame (ORF) encoding 513 amino acid and shares high-level homology with other fish PKZ (32). All fish PKZ contain two Zα domain in N-terminus and a conserved catalytic domain in C-terminus (28–30, 32, 44).

Ctenopharyngodon idellus PKZ induces the apoptosis through eIF2α phosphorylation (33). In the same way dsRNA induction lead to PKR auto-phosphorylation and auto-dimerization (46– 48), PKZ can interact with the [poly (dA:dT) DNA or poly (dG:dC)], though it cannot interact with dsRNA (**Figure 3**), just as poly (dA:dT) and poly (dG:dC) can facilitate PKZ phosphorylation but poly I:C cannot (**Figure 4A**). Because of this, poly (dG:dC) was chosen as activator for PKZ in our subsequent study. Besides PKZ phosphorylation, poly (dG:dC) promotes PKZ dimerization (**Figure 4B**). Consistently with our results, the N-terminus of crucian carp PKZ was found not to interact with poly I:C and this indicates that dsRNA is unable to activate PKZ (31). It was also observed that PKZ is located in the cytoplasm (**Figure 2**), where the binding with infected DNA takes place. We believe strongly that PKZ may act as an essential cellular DNA receptor.

It is well-known that DNA sensors can promote IFN-β expression, and PKZ facilitates type I IFN expression in different grass carp cell lines, which indicates how so that PKZ has identical functions in various types of fish cells (**Figures 5A–D**). As indicated, IFN expression was gradually upregulated with increasing concentrations of overexpressed PKZ in the experiments (**Figure 5E**), and there is clear evidence of functional differentiation in the kinase domain and Zα domain of PKZ. The N-terminal Zα domain is mainly responsible for recognizing cytoplasmic DNA and also has B-Z transition activity (49, 50). The C-terminal kinase catalytic domain of PKZ plays a significant role in triggering IFN expression (**Figures 5F,G**). A similar observation was also documented in cGAS (16). Liu et al. (31) have also suggested that fish PKZ is a new cytosolic sensor for DNA detection by virtue of the unique N-terminal Zα domains.

DNA-sensed PRRs used to make intrinsic immunestimulating properties of plasmid DNA vaccines can recognize intracellular DNA (51). For example, DAI can be used to make intrinsic immune-stimulating property of plasmid DNA vaccines which promote the transcription of genes encoding type I IFNs, proinflammatory cytokines, and co-stimulatory molecules (52). Our findings may provide a potential applicability for DNA vaccines. This is to say that PKZ also could be used to make

separately co-transfected PKZ-FLAG with one of the following recombinant plasmids (IRF3-GFP, STING-GFP, ZDHHC1-GFP, eIF2α-GFP, IRF9-GFP, STAT1-GFP, STAT2-GFP, and STAT6-GFP), this experiment was divided into two groups: one group was immunoprecipitated with anti-FLAG antibody and the other group was immunoprecipitated with anti-GFP antibody. Forty eight hours later, cell lysates were separately immunoprecipitated with anti-FLAG (anti-GFP) antibody and IgG. Then the immunoprecipitants were examined by Western blot with the anti-GFP (anti-FLAG) antibody. The molecular masses of IRF3-GFP, STING-GFP, ZDHHC1-GFP, eIF2α-GFP, IRF9-GFP, STAT1-GFP, STAT2-GFP, and STAT6-GFP are 85, 70, 85, 60, 80, 90, 98, and 87 kD, respectively. Each Western blot is representative of at least three independent experiments.

DNA vaccines. Meanwhile, PKZ provides mediated basics for poly (dA:dT) and poly (dG:dC) acting as immune inducers in up-regulation of IFN. We believe that PKZ could be used to make DNA vaccines and our findings may be a useful contribution in that regard.

Though PKZ is known to be able to initiate IFN expression, the PKZ-mediated pathways need further study, which was done in our subsequent experiments. The IFN-induction pathway mainly includes IRF3-dependent and ISGF3-dependent pathways (53, 54). The IRF3-dependent pathway is primarily responsible for IFN and ISGs expression at the early stages of viral infection (55), while the ISGF3-dependent pathway is responsible at the later period of infection (56). It is well-known that IRF3 (35), STING (21), and ZDHHC1 (57) are the members

significance using unpaired two-tailed t-test. Asterisks indicate significant differences from control (N.C) \*p < 0.05, \*\*p < 0.01. Fold changes were determined relative to cell transfected with N.C. Each Western blot is representative of at least three independent experiments.

of the IRF3-dependent pathway. Our previous results show that grass carp IRF3, STING, and ZDHHC1 are all upregulated under stimulation with poly (dA:dT) and poly (dG:dC) (35). The expression characteristic of PKZ is similar to those of IRF3, STING, and ZDHHC1 under stimulations with different DNAs (**Figure 1**). In addition, PKZ can directly interact with these mediators under stimulation by poly (dG:dC) (**Figure 6**).

ISGF3 (IRF9/STAT1/STAT2) acts as a transcription regulator induced by IFN (58). However, many studies have shown that STAT2 rather than STAT1 is the catalytic domain for ISGF3 (59, 60). Therefore, IRF9 and STAT2 are the major subunits of the ISGF3 complex. This functional unit is known as ISGF3 like. Our previous studies also found that grass carp IRF9 and STAT2 can form an activated heterodimer (61). In this paper,

PKZ was shown to promote the IRF3 dimerization and phosphorylation (**Figure 7A**) and increase STAT2 phosphorylation (**Figure 7B**). These results indicate that PKZ can respond to DNA stimulation and activate IRF3 and ISGF3-like. IFN expression is activated in IRF3- and ISGF3-likemediated pathways, which are dependent on the presence of PKZ (**Figures 7C–E**).

In conclusion, when cells are invaded by pathogenic DNA, PKZ recognizes it and then binds with it. The activated PKZ first interacts with and activates the mediators of the IRF3-dependent pathway to form the tetramer of PKZ-IRF3- STING-ZDHHC1, and then activates IRF3. In the ISGF3 like mediated pathway, the activated PKZ can activate the dimer of IRF9-STAT2 to form the trimer of PKZ-IRF9-STAT2 (**Figure 8**). Notably, these data indicate that fish PKZ gives antiviral signals in IRF3-dependent or ISGF3-like dependent pathways under stimulation with DNA, and thus helps initiate immune responses.

## AUTHOR CONTRIBUTIONS

CH supervised the research. XX conceived the study, designed, and performed the experiments and wrote the manuscript. ML, XX, DL, BC, HM, and CW analyzed the experiments and data. ZJ, and CL provided reagents, technical assistance and contributed to completion of the study. DL revised the manuscript. All authors reviewed the results and approved the final version of the manuscript.

## FUNDING

This work was supported by Major projects of Natural Science Foundation of Jiangxi Province (20171ACB20004), the National Natural Science Foundation of China (Grant 31472304), the earmarked fund for Jiangxi Agriculture Research System (Grant JXARS-04), the Graduate Innovation Special Fund of Nanchang University (Grant CX2016200), the Science and Technology Project of Education Department of Jiangxi Province (GJJ180112), Science and Technology Innovation Special Fund of National Science and Technology Department (2018ZDF40023), and the Science & Technology Foundations of Education Department of Jiangxi (GJJ161303).

## ACKNOWLEDGMENTS

We are gratitude to Prof. Pin Nie (Institute of Hydrobiology, Chinese Academy of Sciences, Wuhan, China) for providing us with CO cells.

## SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2019.00582/full#supplementary-material

### 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 © 2019 Xu, Li, Wu, Li, Jiang, Liu, Cheng, Mao and Hu. This is an openaccess article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Pattern Recognition by Melanoma Differentiation-Associated Gene 5 (Mda5) in Teleost Fish: A Review

Jassy Mary S. Lazarte<sup>1</sup> , Kim D. Thompson<sup>2</sup> and Tae Sung Jung<sup>1</sup> \*

*<sup>1</sup> Laboratory of Aquatic Animal Diseases, College of Veterinary Medicine, Gyeongsang National University, Jinju, South Korea, <sup>2</sup> Moredun Research Institute, Pentlands Science Park, Penicuik, United Kingdom*

Teleost fish, as with other vertebrates, rely on their innate immune system as a first line of defense against invading pathogens. A very important characteristic of the innate immune response is its ability to recognize conserved molecular structures, such as viral dsRNA and ssRNA. Mda5 is one of the three pattern recognition receptors (PRRs) that recognize cytoplasmic viral ligands. Teleost Mda5 is widely conserved among several fish species and possesses the same structural domains as those seen in their mammalian counterparts. Fish Mda5 has been shown to be capable of initiating an inflammatory response both *in vitro* (in different fish cell lines) and *in vivo* using synthetic viral analogs or virus. The interferon (IFN) pathway is triggered as a result of Mda5 activation, leading to the expression of type I IFNs, IFN- stimulated genes and pro-inflammatory cytokines. Although it is known that Mda5 acts as a receptor for virally-produced ligands, it has been shown more recently that it can also initiate an immune response against bacterial challenges. This review discusses recent advances in the characterization of teleost Mda5 and its potential role in antiviral and antibacterial immunity in teleost fish.

### Edited by:

*Stephanie DeWitte-Orr, Wilfrid Laurier University, Canada*

#### Reviewed by:

*Ryan Noyce, University of Alberta, Canada Sarah J. Poynter, University of Waterloo, Canada*

> \*Correspondence: *Tae Sung Jung jungts@gmail.com*

#### Specialty section:

*This article was submitted to Comparative Immunology, a section of the journal Frontiers in Immunology*

Received: *03 January 2019* Accepted: *09 April 2019* Published: *26 April 2019*

#### Citation:

*Lazarte JMS, Thompson KD and Jung TS (2019) Pattern Recognition by Melanoma Differentiation-Associated Gene 5 (Mda5) in Teleost Fish: A Review. Front. Immunol. 10:906. doi: 10.3389/fimmu.2019.00906* Keywords: teleost fish, innate immune system, pattern recognition receptors, melanoma differentiationassociated gene 5, interferon pathway

### INTRODUCTION

Vertebrates have both innate and adaptive immune systems that help them defend themselves against pathogens, such as viruses and bacteria. The innate immune system acts as the initial line of defense against infection and plays a pivotal role in mediating an immediate immune response, which in turn helps to activate the adaptive immune system (1). While higher vertebrates (i.e., mammals), have a much more complex adaptive immune system compared to lower vertebrates, studies have shown that lower vertebrates (including fish) have an intricate innate immune system that compensates for their less developed adaptive immune system (2).

Early pathogen recognition is paramount for an organism's survival. It is important that the host has a set of "sensors" that can instantly recognize the presence of microbial/viral nucleic acids within its cytoplasm. One of these is the DExD/H-box (DDX) protein family that includes RNA and DNA helicases possessing a DExD/H-box domain. DDX proteins are directly involved in the regulation of gene induction and other important processes including signal transduction, gene promoter regulation, mRNA splicing, translational regulation and most importantly, they have been implicated in innate immunity, acting as RNA sensors or signaling molecules (3). Another group of receptors that recognize the presence of cytoplasmic nucleic acids are the pattern recognition receptors (PRRs), which are more thoroughly studied and well-characterized in most vertebrates. PRRs are innate immunity receptors, defined by their ability to specifically recognize microbes and/or microbial moieties (4). In the presence of invading pathogens, the innate immune response is initiated primarily through the recognition of conserved pathogen-associated molecular patterns (PAMPs) by the PRRs (5). These PRRs serve as a pathogen surveillance system in all eukaryotic organisms, which recognize the conserved molecular pathogen signatures comprised of proteins, lipids and nucleotides (6). The PRRs, therefore, allow the immune system to distinguish self from non-self, while still retaining the capacity to respond effectively during an infection. Ultimately, the recognition of PAMPs by the PRRs trigger the activation of multiple signaling cascades in the host immune cells, including the stimulation of interferons (IFNs) and several other cytokines (6, 7).

PRRs are categorized into three groups depending on their function: (i) soluble bridging PRRs, which facilitate the recognition and elimination of their ligands by phagocytes, (ii) endocytic PRRs, which mediate the recognition and internalization of microbes and/or microbial moieties, and (iii) signaling PRRs, which are involved in cell activation in response to a diverse range of microbial moieties (4, 8). Signaling PRRs are functionally very distinct from the other groups of PRR and are further sub-categorized into three different groups, namely: (i) toll-like receptors (TLRs); (ii) nucleotide oligomerization domain-like receptors (NLRs) and (iii) retinoic acid-inducible gene-I (RIG-1)-like receptors (RLRs) (7, 9–12).

RLRs belong to DExD/H box RNA helicases that are known to be the core cytosolic receptors involved in the recognition of viral RNAs. In mammals, three members in the RLR family have been observed, retinoic acid-inducible gene 1 (RIG-1 or DEAD box polypeptide 58, DDX58), melanoma differentiation-associated gene 5 (Mda5, interferon induced with helicase domain 1, IFIH1, or Helicard), and laboratory of genetics and physiology 2 (LGP2 or DExH box polypeptide 58, DHX58) (13) The RLRs observed in mammals are found to be conservatively present in teleost fish. In fact, all three members, RIG-1, Mda5 and LGP2, have been identified in a range of fish species (14).

The recent advances made in the field of fish immunology over the past few decades, specifically on the knowledge of RLRs in teleost fish, has led the way for to a better understanding of the fish immune system, as well as the diversity and evolution of antiviral immunity in vertebrates. Thus, in this review, we focus on the recent discoveries in relation to PRRs, focusing on Mda5 in particular, which had been identified in several fish species, including model fish species such as zebrafish (Danio rerio), and some economically important fish species such as rainbow trout (Oncorhynchus mykiss) and Japanese flounder (Paralichthys olivaceus).

### GENERAL STRUCTURE OF MDA5 AND ITS ORTHOLOGS IN FISH

Mda5, together with RIG-1, are cytoplasmic sensors of dsRNA comprising of four discrete domains; two caspase recruitment domains (CARDs) at the N-terminal region, a DEAD/DEAH box helicase domain (DEXDc), a regulatory domain (RD) and a helicase C-terminal domain (HELICc) (15, 16) as shown in **Figures 1A,B**. As observed in humans, Mda5 and RIG-1 proteins in fish are closely related proteins, having structural similarities of 23 and 35% amino acid (aa) identity in their N-terminal tandem CARD and C-terminal helicase domains, respectively (17).

Mda5 (and RIG-1) in fish consists of protein domains that are similar to their mammalian counterparts. The first fish ortholog of Mda5 was reported in pufferfish (Fugu rubripes) in 2008, through the use of bioinformatic analyses of available whole genome sequences (18). Since the first characterization of Mda5 in fish, studies focusing on the analysis of this important PRR have increased significantly, partly because of the recent advances made in bioinformatics.

The Mda5 gene has now been cloned and characterized for a number of fish species. Differences in the aa length is quite noticeable when comparing the cloned Mda5 among fish species. As shown in **Figures 1A,B**, this difference can be attributed to the regions of low compositional complexity along the whole sequence (represented as pink boxes in the diagram). To further elucidate this, the six essential domains of Mda5 were analyzed through the use of the Simple Modular Architecture Research Tool (SMART). As predicted by the tool, the domains were found at different position along the Mda5 sequence in different teleost species, moreover, the aa length of the respective domains also differ (**Table 1**). These subtle differences in the sequence of Mda5 between different teleost species does not appear to interfere with the function of the protein. Though they differ in ORF length and number of aa residues (see **Table 2**), analysis of their protein residues show that there is a close phylogenetic relationship between the Mda5 from different fish species and they all have a significant similarity with other vertebrate Mda5 (**Figure 1C**).

### MDA5 AND ITS INVOLVEMENT IN THE INNATE IMMUNE RESPONSE

In 2002, Mda5 was initially discovered as an interferon-inducible putative RNA helicase with double-stranded RNA-dependent ATPase activity and melanoma growth-suppressive properties in human melanoma cells (19), and then in 2004, Mda5 was reported to play a major role in an intracellular signal transduction pathway, resulting in the activation of the IFNβ promoter, and V proteins of paramyxoviruses were shown to interact with Mda5 to block its activity (20). Subsequent studies have indicated that Mda5 is capable of recognizing a viral infection and transmitting a signal by CARD (21, 22). It was also established that Mda5 could sense single stranded RNA with 5' triphosphate and could selectively recognize long dsRNAs (>3 kb) (18, 23), which includes dsRNA replication intermediates of positive-sense RNA viruses, the genome of dsRNA viruses and polyinosinic: polycytidylic acid (poly I:C). The ability of Mda5 to recognize these dsRNA-related molecules induces the secretion of type I IFN, in particular (24). There had been underlying issues regarding the specificity of ligands that Mda5 and RIG-1 recognized, suggesting an overlap in the mechanism of action between these two RLRs. RIG-1 can be activated by diverse positive- and negative-strand RNA viruses including influenza, Rift Valley fever, measles, Ebola, vesicular stomatitis virus (VSV),

*(*BAJ14020.1*)*, *Mus musculus* (NP\_082111.2), *Sus scrofa* (AWH63112.1), *Felis catus (*BAX03651.1), *Macaca mulatta (*ABI33114.1*)*, *Homo sapiens* (AAG34368.1).

and hepatitis C viruses (25). The minimal requirement to activate RIG-1 is a blunt-ended base-paired RNA 10–20 bp long with a 5' triphosphate and free mismatches near the blunt end (denoted 5' pppbpRNA, since this could arise from ssRNA with complementary ends or dsRNA) (26–28). It is also reported that much longer dsRNAs (>200 bp), including poly (I:C), which does not necessarily bear a 5'ppp-end or blunt-ended, can also induce IFN via RIG-1 (23). Thus, it is noteworthy to mention that although poly (I:C) (a synthetic dsRNA commonly used to represent a viral ligand in Mda5 and RIG-1 studies) can induce IFN production through Mda5 and RIG-1 activation, these two RLRs are capable of distinguishing this ligand according to the size wherein, Mda5 tends to recognize long poly (I:C) and RIG-1 specifically reacts with short poly (I:C). Although the mechanism of how these RLRs discriminate between differences in length is not yet understood, we think that it has to do with the "uncoiling" of the dsRNAs and how much ATPase activity is involved for the Mda5 or RIG-1 to be activated.

The CARD and helicase domains of the Mda5 are the domains directly involved in the initiation of the signaling pathway, and triggering the innate immune response. The helicase domain binds to dsRNA leading to the activation of the CARD domains. After interacting with PAMPs, the Mda5 CARDs are exposed and they form a complex with the CARD domain of the mitochondrial protein IFN- promoter stimulator-1 [IPS-1, also known as mitochondrial antiviral signaling (MAVS), virus-induced signaling adapter (VISA), and CARDIF], which is located on the outer membrane of the mitochondria (39– 42). This is then followed by the recruitment of tumor necrosis factor(TNF)-receptor associated factor 3 (TRAF3) and activation of TRAF family member-associated NF-κB-activator binding kinase-1 (TBK1) and inducible IκB kinase (IKKǫ) (40, 41, 43). The activation of these kinases result in the phosphorylation of interferon regulatory factor 3 and 7 (IRF3/7), the phosphorylated IRF3/7 forms a dimer and translocates to the nucleus to activate the type-I IFN promoter (44). Ultimately, these processes initiate



\**start aa-end aa.*

TABLE 2 | Mda5 orthologs in different teleost species.


\**Open Reading Frame (ORF).*

the expression of IFN and pro-inflammatory cytokines. The expression of IFN triggers the release of antiviral effectors, such as IFN-stimulated gene (isg) 15, myxovirus resistance gene (mx), 2', 5'-oligoadenylate synthetase (OAS)-directed ribonuclease L (RNASEL) pathway and protein kinase R (PKR), which in turn enhances the IFN-mediated antiviral response (45) (**Figure 2**).

### MDA5 IN VIVO AND IN VITRO EXPRESSION IN TELEOST

Synthetic RNAs, IFNs and viruses are known to induce the expression of Mda5 in mammals (21, 46). Studies have shown that fish Mda5 is also capable of responding, both in vivo or in

FIGURE 2 | Proposed Schematic diagram of Mda5 signaling pathway in Teleost based on a Mammalian Model. The activation of Mda5 is initiated by the presence of long (+) dsRNA released after viral infection or bacterial nucleic acid, that leads to the phosphorylation of interferon regulatory factor 3 and 7 (IRF3/ IRF7), then to the activation of type I IFN promoter and finally to the expression of type I IFNs and other interferon-stimulated genes (ISGs). CARD, Caspase activation and recruitment domain; RD, regulatory domain; MAVS, mitochondrial-antiviral signaling protein; TRAF, TNF (tumor necrosis factor)-receptor associated factor; IKKǫ, inhibitor of nuclear factor Kappa-B kinase subunit epsilon; TBK1, TANK-binding kinase 1; P, signifies phosphorylation; ISRE, interferon-sensitive response element.

vitro, to stimulation by synthetic double-stranded RNA (dsRNA), poly(I:C) (30, 32, 35) and to viral infections (29, 33, 34).

The CiMda5 transcripts in grass carp (Ctenopharyngodon idella) have been observed to increase in expression after infection with grass carp reovirus (GCRV) in vivo, especially in the spleen and liver (29). When expression of rainbow trout Mda5 was examined in vitro using rainbow trout gonad (RTG-2) and spleen (RTS-11) cell lines after stimulating the cells with poly(I:C), Mda5 transcripts were observed to increase in both cell lines, but this stimulation was greater in the RTS-11 cells. Intracellular poly(I:C) stimulation also caused a significant increase in Mda5 expression. The expression of Mda5 in RTG-2 cells could also be stimulated using synthesized IFNs (16). Expression of Japanese flounder Mda5 was evaluated in vitro using whole kidney leukocytes (KL) and peripheral blood leukocytes (PBL) with poly(I:C) stimulation, and also in vivo with viral hemorrhagic septicemia virus (VHSV), with significant up-regulation of Mda5 transcripts noted both in vitro and in vivo (30).

The studies outlined above have clearly shown that Mda5 is able to be stimulated, both appropriately and efficiently, by synthetic stimulators such as poly(I:C) and viral infections in either fish or in fish cell lines, and have laid the ground work for subsequent studies examining the Mda5 response to viral infections in other fish species. As summarized in **Table 3**, fish Mda5 can strongly be up-regulated in different teleost species through the use of different stimulants, either with a virus or poly(I:C) as observed in different cell lines and organs.

Mammalian Mda5 is established as a viral PAMP-recognizing PRR of different ssRNA, dsRNA viruses as well as poly(I:C), which is a synthetic analog of dsRNA virus. In the case of teleost Mda5, this PRR has been implicated in the stimulation of the immune response against viral antigens, probably by serving as a sensor. However, in the study performed by Ohtani et al. (30), in which they used a synthetic bacterial analog, Lipopolysaccharide (LPS), representing stimulation by Gram-negative bacteria, their results showed up-regulated Mda5 expression after LPS stimulation, suggesting that Mda5 might not be involved exclusively in recognizing viral PAMPS, but they are also capable of indirectly distinguishing bacterial PAMPs. Several studies concurrently showed that LPS or bacterial challenge resulted in up-regulation of fish Mda5, such as in channel catfish (Ictalurus punctatus) challenged with Edwardsiella ictaluri (31), in common carp (Cyprinus carpio) after Aeromonas hydrophila challenge (2) and in black carp (Mylopharyngodon piceus) MPF cells after LPS stimulation (36).

The most recent studies on fish Mda5 further verify that this PRR is indeed not only involved in virus detection, but also has the ability to initiate the RIG-1/Mda5 pathway during bacterial infection. The expression level of Asian seabass Mda5, AsMda5, in response to bacteria was elucidated by infecting juvenile fish with either Vibrio alginolyticus (Gram-negative bacterium) or Staphylococcus aureus (Gram-positive). When Sahul India seabass kidney (SISK) cell line was exposed to LPS, a sustained level of AsMda5 up-regulation was obtained several hours after stimulation, but the levels of expression obtained were not as high as those seen in fish stimulated with LPS in vivo. Poly(I:C) injected fish, on the other hand, produced much higher levels of AsMda5 expression than fish injected with bacterial LPS (37). In tilapia (Oreochromis niloticus), a Streptococcus agalactiae (Grampositive bacterium) infection caused an increase in OnMda5 transcript levels in the intestine, kidney, gills and blood at different time points (38). Together, the results highlighted above for the various fish species indicate that fish Mda5 is not only involved in antiviral immune responses, but also bacterialtriggered immune responses, although the mode of action of Mda5 stimulation by bacteria has yet to be determined.

Bacterial ligands are recognized by a different group of PRRs, nucleotide-binding oligomerization domain (NOD)-like receptors (NLRs) and some toll-like receptors (TLRs), which had been widely observed among vertebrates. NLRs can cooperate with TLRs and regulate inflammatory and apoptotic responses. Different mammalian TLR families had been elucidated and most of them have also been found in teleost fish including two additional fish-specific TLR family members (47, 48). Although these fish orthologs have already been demonstrated in different species, the role of these TLRs in the recognition of ligands from bacteria is now the focus of intense studies. NLRs, specifically, NOD1 and NOD2, recognize peptidoglycan components common to both Gram-positive and Gram-negative bacteria. Both proteins drive activation of mitogen-activated protein kinase (MAPK) and nuclear factor κ-light chainenhancer of activated B cells (NF-κB) pathways, leading to proinflammatory cytokine production (49, 50). There has not been a direct link between these receptors, except for the fact that when these receptors are activated, adaptor proteins (i.e., MyD88, MAVS) trigger a downstream cascade leading to the release of inflammatory genes needed in the fight against pathogens (51). The studies mentioned above, on the overexpression of Mda5 after bacterial challenge, do not provide any additional information on the mechanism involved or if it is definite that Mda5 recognized these bacterial ligands. We, therefore, hypothesize that maybe the presence of excess number of bacterial nucleic acids (i.e., small RNAs) which were indirectly sensed by Mda5 led to the observed overexpression of this PRR. We also believe that if they had included RIG-1 in their analysis, an up-regulation of RIG-1 could also have been observed, but the study focused more on Mda5, and they did not delve into TLR/NLR markers. The limiting factor of previous studies is that lack of expression of any TLR/NLR as a comparison, as this would further explain if these two totally different group of PRRs perform overlapping immune response mechanisms in fish.

### ANTIVIRAL AND ANTIBACTERIAL FUNCTIONS OF MDA5

The essential role of Mda5 in antiviral responses had been suggested by the existence of paramyxovirus proteins. The highly conserved cysteine-rich C-terminal domain of the V proteins of a wide variety of paramyxoviruses binds to Mda5 products. As shown from reporter assays, Mda5 stimulates the basal activity of the IFN-β promoter, and over-expression of Mda5 enhances the activation of IFN-β in response to intracellular dsRNA. It was also shown that Mda5 can activate both NF-κB and IRF-3, suggesting that Mda5 plays a pivotal role in the upstream activation of these transcription factors in response to dsRNA, however, these activities were repressed by co-expression of the V proteins (20). The V protein of the Sendai virus, Hendra virus, simian virus 5, human parainfluenza virus 2, and mumps virus selectively abrogates Mda5 function, highlighting the ingenious mechanisms of initiating antiviral immune responses and the action of virus-encoded inhibitors (17, 20).

It was demonstrated using knockout mice that Mda5 plays a crucial role in type I IFN responses by dendritic cells (DCs) and macrophages, when the mice were stimulated with poly(I:C). Specifically, Mda5-deficient mice showed abrogated production of IFN-α and IFN-β in bone marrow-derived DCs and macrophages, and that this PRR is functionally dominant over TLR3 for type I IFN responses to poly(I:C) in vitro and in vivo. Furthermore, mice without Mda5 activity succumbed sooner to infection by encephalomyocarditis virus (ECMV), confirming the essential role that Mda5 has in the host's resistance to ECMV in vivo (21).

In comparison to mammals, innate immunity in teleost fish is poorly understood. It is therefore important to establish how PAMPs recognize PRRs in fish. Earlier reports on fish PRRs,


TABLE 3 | Up-regulation of Mda5 by different viral stimulants in other teleost species.

especially Mda5, focused on molecular characterization instead of function, and it is only in the past decade that studies have focused on the role of these molecules in the innate immunity of fish.

When the effect of grass carp Mda5 (CiMda5) on the production of various IFNs in C. idella kidney (CIK) cells infected with grass carp reovirus (GCRV) was examined, CiMda5 was shown to induce an extensive IFN response in the infected cells by facilitating total phosphorylation of IRF3 and IRF7, enhancing the heterodimerization of IRF3 and IRF7 and the homodimerization of IRF7. The homodimer IRF7 broadly induces the production of IFN-1 in response to GCRV infection, suggesting that CiMda5 has a crucial role in the cytosolic pathway for the induction of IFN genes in response to GCRV (52). Meanwhile, over-expression of Mda5 in Hirame natural embryo (HINAE) cells resulted in a decreased cytopathic effect in cells infected with VHSV, hirame rhabdovirus (HIRRV) and infectious pancreatic necrosis virus (IPNV). The observed reduction in VHSV titers indicate that Japanese flounder Mda5 inhibits the replication of ssRNA viruses (VHSV and HIRRV) as well as dsRNA viruses (IPNV). Moreover, the expression level of type I IFN, Mx and ISG15 genes in Mda5-overexpressing HINAE cells, infected with VHSV, was significantly higher than in noninfected cells. These results demonstrate the ability of Japanese flounder Mda5 to enhance the antiviral activity of the fish, mediated by the activation of type I IFN and IFN-stimulated genes (30). In the zebrafish cell line ZF4, overexpression of the two splice variants of Mda5 (Mda5a and Mda5b) significantly induced type I interferon promoter activity and promoted protection against SVCV infection in transfected cells (5). In another zebrafish experiment, overexpression of Mda5 in zebrafish liver (ZFL) cells had a 2.5 × 10<sup>6</sup> -fold reduction in viral burden after infected with snakehead rhabdovirus (SHRV) demonstrating that Mda5 overexpression increases resistance to SHRV (1). Overexpression of orange spotted grouper (Epinephelus coioides) Mda5 triggered an increase in the expression levels in IFN and IFN-stimulated response element (ISRE) promoter in a dose-dependent manner (400 and 800 ng ml−<sup>1</sup> ) and also enhanced the expression of IRF3, IRF7, and TRAF6 (TNF receptor-associated factor 6) and some proinflammatory factors including, tumor necrosis factor (TNF-α), interleukin 6 (IL-6) and IL-8 at different time points during SGIV and RGNNV infection (34).

The role of Mda5 in the fish innate immunity by the induction of IFN-mediated immune response after viral infection has been well-elucidated. In addition, studies have demonstrated that Mda5 can also be triggered by bacterial stimulation. However, the studies carried out on fish, showing the ability of bacterial stimulants in up-regulating Mda5 expression, did not specifically discuss the mechanism behind this occurrence, but it has also been observed in other animals. In fact, there are studies in mammals which show that the RIG-1/Mda5 pathway, thought primarily to detect viruses, is also involved in the innate immune response to intracellular bacteria e.g., Legionella pneumophila, a Gram-negative bacterium (53) and Listeria monocytogenes, a Gram-positive bacterium (54). Listeria monocytogenes releases nucleic acids during the infection that are recognized by the cytosolic sensors RIG-1, Mda5, and stimulator of interferon genes (STING), thus resulting in the expression of IFN-β and an inflammasome response (54).

The involvement of Mda5 in the innate immune response in fish is limited to the results presented above, and it has not been confirmed that Mda5 acts as a receptor for viral and/or bacterial ligands. Instead, we can only generalize that the overexpression of Mda5 can lead to the protection of fish, therefore, investigating the "ligand-receptor" interaction of fish Mda5 could give us a better insight if this PRR has an equivalent function to that observed in higher vertebrates.

### INTERACTION OF MDA5 WITH LGP2

Our understanding of fish immunology had increased greatly over the past few decades, including the discovery of orthologous genes for mammalian RIG-1, Mda5, and LGP2 (14). The functional characteristics of these RLRs have been investigated in a range of teleost species, including model fish species such as zebrafish, and these genes appear to have similar functions to those present in mammals [5, 53].

All three RLRs share homologous core structural domains, including a DExD/H box helicase domain, a helicase Cterminal domain and a C-terminal domain (CTD), however, while RIG-1 and Mda5 have CARDs, LGP2 lacks them (55), and this absence of CARDs in LGP2 makes it unable to induce signaling alone. This is consistent with the inability of LGP2 to intrinsically activate the IFN-β promoter in transient overexpression experiments (56). Thus, it is difficult to determine its exact role in RLR-mediated signaling (57), and until now, the function of LGP2 in antiviral signaling has been controversial.

LGP2 has been identified as a negative regulator of IFN response, for example when triggered by Sendai virus, Newcastle disease virus (17, 56, 58, 59) or poly (I:C) (58). However, mice studies have shown that LGP2 can have either a positive or a negative role in IFN induction (60). The action of LGP2 is believed to synergize with that of Mda5, but not RIG-1, to boost IFN signaling at low levels of LGP2 expression. On the other hand, at higher levels of LGP2 expression, it acts as an inhibitor of RIG-1 and Mda5 signaling (61–63). In recent studies, mammalian LGP2 has been shown to be involved in Mda5 filament formation and Mda5-mediated viral RNA recognition (63, 64). Other studies have confirmed that LGP2 synergizes with Mda5: (1) to elevate IFN- transcription in vivo, for example during an encephalomyocarditis virus infection or poly (I:C) stimulation (62); (2) to facilitate viral RNA recognition through its ATPase domain (65) and (3) to sense Sendai virus infection for IFN-1 induction along with the loss of RIG-1, as determined in Chinese tree shrew (66).

Some reports have shown that teleost LGP2 is a negative regulator of antiviral immunity when overexpressed in vitro. For example, overexpression of crucian carp (Carassius carassius) LGP2 reduced the activity of IFN promoters, mediated by RIG-1 and Mda5 (67), and down-regulation of antiviral immune genes like Mda5, was also observed in cells overexpressing the grouper LGP2 (68). In zebrafish, LGP2 negatively regulates the IFN response mediated by poly (I:C) by blocking some of the important signaling factors, including RIG-1 and Mda5, but not IRF3/7 (69). It also appears that the antithetical function of LGP2 in antiviral immunity depends on LGP2 expression levels, similar to that is observed with mammalian LGP2. Zebrafish LGP2 functions as a positive regulator of IFN signaling during the early phase of virus infection; during this time RIG-1 and Mda5 are expressed at low levels, while during latter phases of the infection, LGP2 adopts a negative role. However, the maximum stimulatory effect of zebrafish LGP2 is lower than levels of Mda5 and RIG-1 expression (70). These results have also been corroborated for grass carp LGP2. During the resting state and early phase of grass carp reovirus (GCRV) infection, synthesis and phosphorylation of IRF3/7, and mRNA levels and promoter activities of IFNs and NF-κBs are inhibited, at a time when grass carp LGP2 is overexpressed. Luciferase assay have shown that grass carp LGP2 binds to RIG-1 and Mda5 with diverse domain preference, and this binding is independent of the GCRV infection. Another interesting result showed that grass carp LGP2 inhibits K63and K48-linked RIG-1 and Mda5 ubiquitination, resulting in suppression of protein degradation. These results indicate that LGP2 has a role as a suppressor in RLR signaling pathways, which is important in maintaining cellular homeostasis during the resting state and early phase of GCRV infections (71).

On the other hand, black carp LGP2 (bcLGP2) was clearly shown to have a synergistic effect with bcMda5 using reporter assays, in which, both the induction of zebrafish IFN3 and fathead minnow IFN (eIFN), mediated by bcMda5 and bcLGP2, were much higher than that obtained by bcMda5 alone, and was higher than the combined effect of bcMDa5 or bcLGP2 alone. The synergistic function between bcLGP2 and bcMda5 reflects bcLGP2 effect on the antiviral activity of the host. Epithelioma papulosum cyprini (EPC) cells, overexpressing both bcLGP2 and bcMda5, showed a decrease in CPE development and viral titer during infection with GCRV or SVCV, in contrast with cells expressing either bcMda5 or bcLGP2 alone (36). In a study focusing on rainbow trout Mda5 and LGP2, it was evident that both RLRs are capable of binding to poly (I:C), triggering IFN production. Also, Mda5 expression is not affected by the overexpression of LGP2 in transfected cells, and vice versa, implying that these RLRs function in parallel as positive regulators for IFN production (16). Although these results help to substantiate the synergy between LGP2 and Mda5, the mechanism behind their interaction remains unclear. A proposed mechanism of action showing the antithetical role of LGP2 with Mda5 is suggested in **Figure 3**.

### DOES MDA5 HAVE A MAJOR ROLE IN ACANTHOPTERYGIANS?

The three RLRs, RIG-1, Mda5, and LGP2, are represented in a number of teleost species, however, one intriguing discovery is the absence of RIG-1 in some members of Acanthopterygii. Presently, RIG-1 has only been identified in crucian carp (67), grass carp (72), common carp (73), zebrafish (74, 75), channel catfish (31), Atlantic salmon and EPC (70). Despite efforts to identify RIG-1 fish orthologs, it has not been possible to identify RIG-1 in Japanese pufferfish (Takifugu rubripes) and green spotted puffer fish (Tetraodon nigroviridis) (15, 18), medaka (Oryzias latipes), and three-spined stickleback (Gasterosteus aculeatus) (15), and rainbow trout (16), gilt-head sea bream (Sparus aurata) and European bass (Dicentrarchus labrax) (76). In addition, in silico data mining of Japanese flounder, Nile tilapia and orange-spotted grouper genomes also showed the absence of RIG-1 in the genome of these teleost species (77).

Orthologs of Mda5 and LGP2 seem to be common to all teleost families, unlike RIG-1, which is only found in more primitive fish species, such as those in classes Ostariophysi, Protacanthopterygii, and Paracanthopterygii (77). It is believed that Mda5 might have emerged before RIG-1 and that the domain arrangements of the genes evolved independently by domain grafting rather than a simple gene duplication event (18).

Furthermore, the presence of the three RLRs in very ancient fish, Sarcopterygii, imply the loss of RIG-1 after the divergence of the Acanthopterygii from the Paracanthopterygii

the up-regulation of secreted type I IFNs and ISGs. (77). Interestingly, most of the fish species reared for aquaculture

do not possess the RIG-1 ortholog, and all of them belong to the

efficient and up-regulated expression of IRF3 and IRF7, leading ultimately, to

superorder Acanthopterygii. With the absence of RIG-1 in some of the Acanthopterygians, their RLRs have possibly evolved differently, wherein their Mda5 as well as LGP2 perform most of the pivotal role in antiviral sensing. Knowing that LGP2 and Mda5 have the capability to work synergistically, it is important to help establish if these two RLRs function in place of RIG-1 in these fish species, and if not, what is the equivalent gene that performs the role of RIG-1? For instance, in chicken, another organism that lacks RIG-1, studies found that chicken Mda5 compensates for the lack of RIG-1 by preferentially sensing shorter dsRNA synthetic poly (I:C) instead of long dsRNAs (78). Another example of mammal that does not have RIG-1 is the Chinese tree shrew. It was revealed that the loss of RIG-1 brought positive selection signals to tree shrew Mda5(tMda5) and LGP2(tLGP2). Data showed that tMda5 alone or tMda5/tLGP2 could replace RIG-1 as a sensor for RNA viruses that trigger IFN production (66). It is believed that this replacement is enhanced due to the interaction of tMda5 with tMITA (Mediator of IRF3 activation), which interacts with RIG-1 resulting in a cascade of antiviral signaling (79). This information tells us that even in the absence of RIG-1, the innate immune system has a way of compensating for the loss of some molecules by relying on other functional molecules, probably homologs, although in fish, the compensatory effect of Mda5 in teleost lacking RIG-1 has not yet been verified, and further investigation is essential to establish this. A study was performed on the RLRs, Mda5, and LGP2 of rainbow trout (Oncorhynchus mykiss), focusing on the parallel function of these two RLRs that synergistically increase in the production of IFNs (16), but whether this was in compensation for the lack of RIG-1 in this fish was not specifically discussed.

### CONCLUDING REMARKS AND FUTURE PERSPECTIVES

The teleost immune system may not be as elaborate as that of its mammalian counterparts, but they have an intricate innate immune system that is on a par with the complex immune system present in mammals. Since the aquatic environment in which fish live is very different to that of mammals, where they are in close contact with pathogens, it is important that their innate immune system offers a first line of defense against invading pathogens. The PRRs in teleosts, similar to that present in mammals, are capable of sensing pathogens and inducing antiviral and/or antibacterial responses. Knowing the role that Mda5 plays during an infection will help give a clearer insight of how the teleost immune system works. Mda5, together with other RLRs, are able to sense pathogens and, in turn, activate downstream processes in the fish's immune response, ultimately, preventing them from succumbing to the infection.

Although, Mda5 was described for several different teleost species in this review, the mechanism of action of Mda5 still needs further elucidation. As discussed, Mda5 expression directs the recruitment of the downstream adaptor MAVS from the mitochondria, then associates with signaling molecules like TBKI, TRAF3, and MITA, which in turn facilitate the activation and phosphorylation of IRF3 and IRF7, leading to their translocation into the nucleus for the induction of type I IFNs and ISGs. These downstream molecules have been identified in various teleost species as a result of stimulation with poly(I:C) and LPS, as well as viral and bacterial infections, however, the extent to which Mda5 regulates the whole process of initiating innate immunity in fish has yet to be established and whether Mda5 works in cooperation with other PRRs thereby suggesting a network of immune molecules instead of a single, linear pathway.

Further studies are needed to establish how these PRRs function within the teleost immune system, for example:

1. RLRs that recognize viral ligands include RIG-1, Mda5, and LGP2. RIG-1 and Mda5 recognize distinct viral dsRNA. LGP2, on the other hand, recognizes the same viral ligand as RIG-1. The molecular signaling mechanisms of RIG-1 and Mda5 are known to share some common features and LGP2 has been found to be a co-stimulatory molecule for Mda5. It appears that these RLRs have overlapping mechanism of action upon virus invasion. It is therefore important to further characterize their function to be able to differentiate them from one another, especially with respect to the downstream signaling cascade they initiate during an antiviral response.


### REFERENCES


group of teleosts are needed to establish if these two RLRs are able to sense all types of viral RNAs in fish lacking RIG-1.

In summary, the fish innate immune system is not as simple as often described, and although it is less complex than the mammalian immune system, it has evolved many similar defense mechanisms that are present in terrestrial organisms. Future elucidation of the regulatory mechanism of Mda5 during pathogen infection is required for a more comprehensive understanding of the role of this and other PRRs in the immune response of fish.

### AUTHOR CONTRIBUTIONS

JL designed and wrote the draft. KT contributed by reading and giving several discussion. TJ organized and finalized the paper.

### ACKNOWLEDGMENTS

This research was supported by a Korea Research Foundation grant funded by the Korean Government (NRF-2018R1A2B2005505).


innate immune sensor, laboratory of genetics and physiology 2 (LGP2). J Biol Chem. (2013) 288:938–46. doi: 10.1074/jbc.M112.424416


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

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

# Functional Characterization of Dark Sleeper (Odontobutis obscura) TBK1 on IFN Regulation

Jian Chen1†, Zhuo Cong Li 2,3†, Long Feng Lu2,3, Pei Li <sup>1</sup> , Xi-Yin Li 2,3 and Shun Li 2,3,4 \*

*<sup>1</sup> Fisheries Research Institute, Wuhan Academy of Agricultural Sciences, Wuhan, China, <sup>2</sup> State Key Laboratory of Freshwater Ecology and Biotechnology, Institute of Hydrobiology, Chinese Academy of Sciences, Wuhan, China, <sup>3</sup> University of Chinese Academy of Sciences, Beijing, China, <sup>4</sup> Key Laboratory of Aquaculture Disease Control, Ministry of Agriculture, Wuhan, China*

In East Asia, the dark sleeper, *Odontobutis obscura* (*O. obscura*) is a crucial commercial species of freshwater fish; however, its molecular biology research is still undeveloped, including its innate immune system, which is pivotal to antiviral responses. In this study, we cloned and identified the characterization and kinase function of dark sleeper TANK-binding kinase 1 (TBK1), supplementing the evidence of the conservation of this classical factor in fish. First, the ORF of *Odontobutis obscurus* (*O. obscura*) TBK1 (OdTBK1) was cloned from liver tissue by RACE-PCR. Subsequent nucleic acid and amino acid sequence analysis suggested that OdTBK1 is homologous with other fish TBK1, and the N-terminal Serine/Threonine protein kinases catalytic domain (S\_TKc) and C-terminal coiled coil domain (CCD) are conserved. Subsequently, the cellular distribution demonstrated that OdTBK1 was located in the cytoplasm region. With regard to the identification of functions, OdTBK1 activated several interferon (IFN) promoters' activity and induced downstream IFN-stimulated genes (ISGs) expression. In a canonical manner, wild-type OdTBK1 significantly phosphorylated interferon regulatory factor 3 (IRF3) but failed when the N-terminal region was truncated. Furthermore, overexpression of OdTBK1 decreased viral proliferation remarkably. Collectively, these data systematically analyzed the characterization and function of OdTBK1, initiating the study of the innate antiviral response of dark sleeper.

#### Keywords: TBK1, IFN, RLRs, antiviral, Odontobutis obscura

### HIGHLIGHTS


### INTRODUCTION

A variety of pattern recognition receptors (PRRs) mediate a host's innate immune response to pathogen invasion; six families have been identified, including Toll-like receptors (TLRs), C-type lectins (CTLs), NOD-like receptors (NLRs), RIG-I-like receptors (RLRs), AIM2-like receptors (ALRs), and OAS-like receptors (OLRs). Among them, RLRs are cytoplasmic sensors of RNA (1). Upon activation, they signal to the mitochondrial antiviral signaling protein (MAVS, also called

### Edited by:

*Stephanie DeWitte-Orr, Wilfrid Laurier University, Canada*

### Reviewed by:

*Sarah J. Poynter, University of Waterloo, Canada Nguyen (Nathan) T. K. Vo, McMaster University, Canada*

\*Correspondence:

*Shun Li bob@ihb.ac.cn*

*†These authors have contributed equally to this work*

#### Specialty section:

*This article was submitted to Comparative Immunology, a section of the journal Frontiers in Immunology*

Received: *15 November 2018* Accepted: *16 April 2019* Published: *03 May 2019*

#### Citation:

*Chen J, Li ZC, Lu LF, Li P, Li X-Y and Li S (2019) Functional Characterization of Dark Sleeper (Odontobutis obscura) TBK1 on IFN Regulation. Front. Immunol. 10:985. doi: 10.3389/fimmu.2019.00985* IPS-1, Cardif, or VISA) (2–4) to form active MAVS polymers and subsequently recruit tumor necrosis factor (TNF) receptorassociated factor (TRAF) family ubiquitin E3 ligases to synthesize polyubiquitin chains, which activates IKK and TANK-binding kinase 1 (TBK1). The recruited kinases then phosphorylate MAVS at its conserved pLxIS motif, and, at the same time, the potential transcription factors interferon regulatory factor (IRF) 3 and IRF7 are phosphorylated by IκB kinase IKKε or TBK1. Thereafter, phosphorylated IRF3 and IRF7 form dimers that accumulate in the nuclei where they bind to target sequences to activate the IFN-β promoter to initiate interferon (IFN) secretion (5).

TBK1 contains an N-terminal kinase, an ubiquitin like domain and two C-terminal coiled-coil domains (6). Recruiting of multiple TBK1 dimers to signaling complexes enables activation-loop swapping of locally clustered TBK1 and results in TBK1 auto-phosphorylation (7). In addition, TBK1 is a serine/threonine protein kinase of the IKK kinase family involved in innate immunity to viral infection by inducing type I IFNs and mediating TANK's ability to activate NF-κB (6, 8).

Upon stimulation by virus structural components, pattern recognition receptors (PRRs) recruit TLR/IL-1R domaincontaining adaptor proteins, inducing IFN-β (TRIF), MAVS, and a stimulator of IFN genes (STING; also called MITA, MPYS, and ERIS) to activate TBK1 (9–11). Activated TBK1 phosphorylates IRF3/7, triggers dimerization and nuclear translocation, and leads to formation of active transcriptional complexes that bind to IFN stimulation response elements (ISRE), activating type I IFN gene expression (12). Knockdown assays and gene-targeting studies have shown that TBK1 is essential for type I IFN production in TLR3 and RLR signaling pathways (13–15).

TBK1 has been identified in a variety of fishes. The TBK1 of black carp (Mylopharyngodon piceus), named bcTBK1, shows antiviral activity against both Grass Carp Reovirus (GCRV) and Spring Viremia of Carp Virus (SVCV) (16). TBK1 in the large yellow croaker (Larimichthys crocea) (named LcTBK1) can interact with an E3 ubiquitin ligase LcNrdp1 to defend against Cryptocaryon irritans infection (17). Upon LPS stimulation, TBK1 from grass carp (Ctenopharyngodon idella) (CiTBK1) triggers IFN response in an IRF3/IRF7-independent manner (18). In addition, Danio rerio TBK1 (DrTBK1) or Carassius auratus TBK1 (CaTBK1) can cause the phosphorylation of IRF3/7 (19, 20). Nevertheless, in teleost fish, the identification of the molecular characteristics and mechanisms of TBK1 are still limited.

The dark sleeper (Odontobutis obscurus), an important commercial fish species, is widely distributed in the river systems of southeast China (21, 22). In recent years, because of its delicious taste and high nutritional value, the dark sleeper has become a very promising aquaculture species in China. However, some viruses seriously threaten its growth. Better understanding of antiviral immune mechanisms may contribute to the development of management strategies for disease control.

In this study, we report the characterization of Odontobutis obscurus (O. obscura) TBK1 (OdTBK1). Our findings demonstrate that OdTBK1 is effective for zebrafish and grass carp IFN promoters and significantly induces IFN-stimulated gene (ISG) expression. It can also promote the phosphorylation of IRF3 both in Danio rerio (D. rerio) and O. obscura. OdTBK1 plays a critical role in the antiviral immune defense of fish in promoting the IFN response.

## MATERIALS AND METHODS

### Cells and Viruses

Epithelioma papulosum cyprinid (EPC, now IDed as fathead minnow) and Grass carp ovary (GCO) cells were maintained at 28◦C, 5% CO<sup>2</sup> in medium 199 (Invitrogen) supplemented with 15% fetal bovine serum (FBS, Invitrogen) (23). Human embryonic kidney (HEK) 293T cells were grown at 37◦C, 5% CO<sup>2</sup> in DMEM medium (Invitrogen) supplemented with 15% FBS. Spring Viremia of Carp Virus (SVCV-741, 10<sup>9</sup> TCID50/ml), a negative-sense ssRNA virus, was propagated in EPC cells until cytopathic effect (CPE) were observed, then the cultured media with cells was harvested and stored at −80◦C until use.

### Fish

Healthy dark sleeper (weighing 40 ± 20 g) purchased from a fish farm were maintained and acclimated to re-circulating tanks (1,000-L, 28 ± 1 ◦C) containing filtered and oxygenated water for at least 2 weeks before experiments. All animal experiments were approved by the Committee on the Ethics of Animal Experiments of the Chinese Academy of Sciences.

### Amplification of OdTBK1 and Plasmid Construction

Rapid amplification of cDNA ends (RACE) was carried out using the 5′ RACE system (Invitrogen) and BD SMARTTM RACE cDNA amplification kit (BD Biosciences Clontech) according to the manufacturer's instruction. The first strand cDNA synthesis and RACE were performed on liver-derived RNA. To obtain the 3′ unknown region, primer pairs OdTBK1F4a/APT and OdTBK1F5a/AP (**Table 1**), were used for the primary PCR and the nested PCR, respectively. The amplified PCR product was cloned and sequenced as described above. Similarly, the 5′ end of OdTBK1 was obtained by nested PCR using primer pairs OdTBK1R2a/APG and OdTBK1R3a/AP (**Table 1**). The full-length cDNA sequence was confirmed by sequencing the PCR product amplified by primers OdTBK1-FP and OdTBK1- RP (**Table 1**) within the predicted 5′ and 3′ untranslated regions, respectively. The open reading frame (ORF) of OdTBK1 was subcloned into pcDNA3.1(+) (Invitrogen), pCMV-Myc (Clontech) and pCMV-Tag2C (Clontech), respectively. For subcellular localization, the ORF of OdTBK1 was inserted into pEGFP-N3 (Clontech) vector. The ORFs of zebrafish IRF3 (NM\_001143904) was subcloned into pCMV-HA (Clontech) and pCMV-Myc vector. The primers used to amplify the Nterminal absent OdTBK1 sequence were OdTBK1-1N-FP and OdTBK1-RP (**Table 1**). It was also subcloned into pCMV-Tag2C and pCMV-Myc vector. The fmIFN promoter was obtained from NCBI database (HE856618.1) and cloned into pGL3-Basic luciferase reporter vector (Promega). The plasmids containing DrIFNφ1pro-Luc, DrIFNφ3pro-Luc, gcIFN1, and ISRE-Luc in pGL3-Basic luciferase reporter vectors were constructed as


described previously (24, 25). All constructs were confirmed by DNA sequencing.

### Bioinformatics Analysis

The phylogenetic tree was constructed with the Neighbor-joining method (NJ) by MEGA7 program which was bootstrapped 500 times. All gene sequences used in this study were derived from GenBank. Multiple alignments were accomplished using GENE.DOC.

### Transient Transfection and Virus Infection

Transient transfections were performed in EPC cells or GCO cells seeded in 6-well or 24-well plates by using X-tremeGENE HP DNA Transfection Reagent (Roche) according to the manufacturer's protocol. And the transfection efficiency of EPC cells or GCO cells is nearly 50%. To confirm the antiviral response of OdTBK1, EPC cells were seeded in 24-well plates overnight and transfected with 0.5 µg OdTBK1-Myc or the pCMV-Myc vector separately. At 24 h post-transfection, the EPC cells were infected with SVCV at a multiplicity of infection (MOI = 0.01) and incubated at 28◦C. The supernatant aliquots were then harvested to detect the virus titers at 48 h post infection, using the standard 50% tissue culture infection dose (TCID50) method. The cell monolayers were washed with PBS, fixed with 4% paraformaldehyde (PFA) for 1 h, and stained with 0.05% crystal violet overnight, and observed for the cytopathic effect (CPE). For virus titration, 200 µl of culture medium were collected at 48 h post-infection, and used for plaque assay. The supernatants were subjected to 3-fold serial dilutions and then added (100 µl) onto a monolayer of EPC cells cultured in a 96 well plate. After 48 or 72 h, the medium was removed and the cells were washed with PBS, fixed by 4% PFA and stained with 1% crystal violet. The virus titer was expressed as 50% tissue culture infective dose (TCID50/ml). Results are representative of three independent experiments.

### Luciferase Activity Assay

EPC cells were seeded in 24-well plates, and 24 h later co-transfected with 250 ng luciferase reporter plasmid (DrIFNφ1pro-Luc, DrIFNφ3pro-Luc, ISRE-Luc, or fmIFN-Luc) and 250 ng OdTBK1-Myc or pCMV-Myc and 50 ng Renilla luciferase internal control vector (pRL-TK, Promega). Empty vector pCMV-Myc was used to maintain equivalent amounts of DNA in each well. GCO cells were used to detect ISRE-Luc and gcIFN1-Luc. At 24 h post-transfection, the cells were washed in PBS and lysed for measuring luciferase activity by Dual-Luciferase Reporter Assay System, according to the manufacturer's instructions (Promega). Firefly luciferase activities were normalized on the basis of Renilla luciferase activity. The results were representative of more than three independent experiments, each performed in triplicate.

### Fluorescence Microscopy

EPC cells were plated onto coverslips in 6-well plates and transfected with indicated plasmids for 24 h. The experimental group was stimulated with the viral analog polyinosinicpolycytidylic acid (poly I:C) 12 h before the photo was taken. Then the cells were washed twice with PBS and fixed with 4% PFA for 1 h. After draining the fixative, the cells were stained with DAPI (C1006, Beyotime) for 5 min in dark at room temperature. Finally, the coverslips were washed and observed with a Leica confocal microscope under a ×63 oil immersion objective (SP8; Leica Microsystems).

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

The total RNA from brain, pituitarium, eye, gill, heart, liver, muscle, spleen, head kidney, and ovary were extracted from three apparently healthy dark sleepers to detect the expression of OdTBK1.

Total RNAs of EPC cells and tissues were extracted by Trizol reagent (Invitrogen). RNase-free DNase is used specifically for RNA purification by removing all contaminating genomic DNA. The first-strand cDNA was synthesized by using a GoScript Reverse Transcription System (Promega) according to the manufacturer's instructions. Quantitative real-time PCR (qPCR) was performed with Fast SYBR Green master mix (BioRad) on a CFX96 Real-Time System (BioRad). PCR conditions were as follows: 95◦C for 5 min, then 40 cycles of 95◦C for 20 s, 60◦C for 20 s, 72◦C for 20 s. and the β-actin primers were used as internal control. The specificity of the PCR amplification for all primer sets was verified from the dissociation curves. The identity of each PCR products was confirmed by dideoxy-mediated chain termination sequencing at Wuhan TSINGKE Biological Technology Inc. The relative fold changes were calculated by comparison to the corresponding controls using the 2−11Ct method. Three independent experiments were conducted for statistical analysis. The total RNA extracted from the tissue of liver also treated as cells.

### Co-IP Assay

For transient-transfection and Co-IP experiments, HEK 293T cells were used instead of EPC cells due to the superhigh transfection efficiency of HEK 293T cells. Cells seeded into 10 cm<sup>2</sup> dishes overnight were transfected with a total of 10 µg the indicated plasmids. At 24 h post transfection, the medium was removed carefully, and the cell monolayer was washed twice with 10 ml ice-cold phosphate-buffered saline (PBS). The cells were then lysed in 1.0 ml radioimmunoprecipitation (RIPA) lysis buffer (1% NP-40, 50 mM Tris-HCl [pH 7.5], 150 mM NaCl, 1 mM EDTA, 1 mM NaF, 1 mM sodium orthovanadate [Na3VO4], 1 mM phenylmethylsulfonyl fluoride [PMSF], 0.25% sodium deoxycholate) containing a protease inhibitor cocktail (Sigma-Aldrich) at 4◦C for 1 h on a rocker platform. The cellular debris was removed by centrifugation at 12,000 × g for 15 min at 4◦C. The supernatant was transferred to a fresh tube and incubated with 20 µl anti-Flag affinity gel (Sigma-Aldrich) overnight at 4◦C with constant agitation. These samples were further analyzed by Immunoblotting (IB). Immunoprecipitated proteins were collected by centrifugation at 5,000 × g for 1 min at 4◦C, washed three times with lysis buffer, and resuspended in 100 µl 5 × SDS sample buffer. The immunoprecipitates and whole-cell lysates were analyzed by IB with the indicated Abs.

### In vitro Protein Dephosphorylation Assay

Transfected HEK 293T cells were lysed as described above, except that the phosphatase inhibitors (Na3VO<sup>4</sup> and EDTA) were omitted from the lysis buffer. Protein dephosphorylation was carried out in 100 µl reaction mixtures consisting of 100 µg of cell protein and 10 U of CIP (Sigma-Aldrich). The reaction mixtures were incubated at 37◦C for 1 h, followed by Western blotting analysis.

### Immunoblot Analysis

Whole cells were lysed in radioimmunoprecipitation (RIPA) lysis buffer [1% NP-40, 50 mM Tris-HCl (pH 7.5), 150 mM NaCl, 1 mM EDTA, 1 mM NaF, 1 mM sodium orthovanadate, 1 mM phenyl-methylsulfonyl fluoride, and 0.25% sodium deoxycholate] containing protease inhibitor mixture (Sigma-Aldrich). The Bradford method was used for estimation of protein concentration of cell lysates. The equivalent amount of proteins (10 µg) in different groups were separated by 10% SDS-PAGE and transferred to PVDF membrane (Bio-Rad). The membranes were blocked for 1 h at room temperature in TBST buffer (25 mM Tris-HCl, 150 mM NaCl, 0.1% Tween-20, pH 7.5) containing 5% nonfat dry milk, probed with indicated primary antibodies (Abs) at an appropriate dilution overnight at 4◦C, washed three times with TBST and then incubated with

FIGURE 1 | Phylogenetic tree of various TBK1 members. The tree was depicted on the overall sequences by neighbor-joining method. The 0.1 scale indicates the genetic distance. Accession numbers of TBK1 amino acid sequences are as follows: *Odontobutis obscura* (dark sleeper), MK637832; *Oplegnathus fasciatus* (Stone bream), AHX37216.1; *Paralichthys olivaceus* (bastard halibut), AGL54166.1; *Larimichthys crocea* (large yellow croaker), AKM77645.1; *Salmo salar* (Atlantic salmon), AEA42006.1; *Gadus morhua* (Atlantic cod), ADL60136.1; *Danio rerio* (zebra fish), NP\_001038213.2; *Cyprinus carpio* (koi), ADZ55455.1; *Xenopus tropicalis* (Western clawed frog), NP\_001135652.1; *Crassostrea gigas* (Thunberg), APX43288.1; *Cuculus canorus* (Cuckoo), XP\_009554581.1; *Mus musculus* (Mouse), AAF05990.1; *Sus scrofa* (Pig), NP\_001098762.1; *Bos taurus* (Bovine), NP\_001179684.1; *Homo sapiens* (Human), AAF05989.1.


protein kinases catalytic domain (S\_TKc) was showed in the black box, C-terminal coiled coil domain (CCD) was showed in the gray box.

secondary Abs for 1 h at room temperature. After additional three washes with TBST, the membranes were stained with Immobilon TM Western Chemiluminescent HRP Substrate (Millipore) and detected using an Image Quant LAS4000 system (GE Healthcare). Abs were diluted as follows: anti-Myc (Santa Cruz Biotechnology) at 1:2000, HRP-conjugated anti-mouse IgG (Thermo Scientific) at 1:5000. The results were the representative of three independent experiments.

### Statistics Analysis

The results are expressed as mean ± SEM. Data were analyzed using a Student's unpaired t-tests. A p < 0.05 was considered to be statistically significant.

### RESULTS

### Molecular Cloning and Phylogenetic Analysis of OdTBK1

The coding sequence (CDS) of OdTBK1 was obtained by RACE-PCR. The full-length ORF of OdTBK1 is 2,172 bp, encoding 723 amino acids through initial sequence analysis of OdTBK1 prediction. The SMART program predicted that there were Serine/Threonine protein kinases catalytic domain (S\_TKc) in N terminus (9-306) and coiled coil domain (CCD) in C terminus (678-704). To investigate the homology of OdTBK1, phylogenetic analyses of TBK1 with several species were conducted by MEGA7 based on the CDS sequences; these members could be divided into mammals, amphibians, mollusks, and fish (**Figure 1**). OdTBK1 showed high homology with the TBK1 of L. crocea. Additionally, OdTBK1 and the TBK1 homologs of D. rerio and G. morhua shared similar positions on the homology. Meanwhile, multiple alignments of OdTBK1 to the corresponding sequences of other species showed that the Nterminal kinases catalytic domains of TBK from the teleost were conserved (**Figure 2**).

### Tissue-Specific Expression Pattern of OdTBK1

To investigate the mRNA level of OdTBK1 in vivo, total RNA was isolated from brain, pituitarium, eye, gill, heart, liver, muscle, spleen, kidney, and ovary of the black sleeper separately. The tissue distribution pattern of OdTBK1 was monitored by qRT-PCR. Transcripts of OdTBK1 were detected in all ten tissues: the highest level of OdTBK1 was observed in pituitarium, while the lowest level of OdTBK1 appeared in the head kidney. Generally, the expression levels of OdTBK1 were basically the same across tissues (**Figure 3**). These data demonstrated that OdTBK1 was ubiquitously expressed in the tissues of the dark sleeper.

### OdTBK1-Mediated IFNs and ISRE Activation

TBK1 is a crucial factor involved in type I IFN response. To clarify whether OdTBK1 activates IFN production, the luciferase report gene assay was employed in the following manner. Cotransfection of the OdTBK1 and IFN promoters occurred in EPC cells, and the activities of these promoters were measured

24 h later. According to the results, OdTBK1 significantly upregulated the activation of the DrIFNφ1 promoter up to 7.6 fold compared with the empty vector control group (**Figure 4A**). Similarly, the activation of the DrIFNφ3 promoter increased by 6.3-fold compared with the control group (**Figure 4B**). The activation of the fmIFN promoter was also measured, and it was significantly induced about 20-fold compared with the empty vector control group (**Figure 4C**). Then, ISRE was co-transfected with OdTBK1 in GCO cells to monitor the capacity of OdTBK1 in another fish cell line. Consistent with the result above, ISRE was significantly enhanced 24-fold by OdTBK1 overexpression (**Figure 4D**). Finally, GcIFN1 was co-transfected with OdTBK1 in the same way, and its activity was also upregulated about 16-fold (**Figure 4E**). These results demonstrated that OdTBK1 conservatively activated the IFN promoters from different fishes.

### Cytoplasm Localization of OdTBK1

Fluorescence microscopy was used to investigate the subcellular localization of OdTBK1 with or without I:C stimulation or SVCV infection. Poly I:C, an RNA virus mimic, can induce IFN expression significantly. EPC cells transfected with EGFP-N3 were used as a control group. The fluorescent signals of EGFP-N3 were distributed in the cytosol and the nucleus with or without poly I:C stimulation or SVCV infection. Compared with the controls, without stimulation, the green signals of OdTBK1 were detected in the cytosol, with none in the nucleus. After poly I:C stimulation or SVCV infection, green fluorescent signals were also detected only in the cytosol (**Figure 5**). These data suggested that OdTBK1 was localized in the cytoplasm.

FIGURE 4 | (A–E) Activation of DrIFNφ1, DrIFNφ3 and ISRE promoters by overexpression of OdTBK1. EPC cells were seeded on 24-well-plates overnight and co-transfected with pCMV-Myc or OdTBK1-Myc and DrIFNφ1pro-Luc (A), DrIFNφ3pro-Luc (B), fmIFNpro-Luc (C). GCO cells were treated at the same way and co-transfected with pCMV-Myc or OdTBK1-Myc and gcIFN1pro-Luc (D), ISRE-Luc (E). The luciferase assays were performed 24 h after transfection. The promoter activity is presented as relative light units normalized to *Renilla* luciferase activity. Error bars represents the means ± SEM (*n* = 3), and the experiments were repeated three time with similar results. Asterisks indicate significant differences from control (\**p* < 0.05).

FIGURE 5 | Subcellular localization of OdTBK1. EPC cells seeded onto microscopy cover glass in 6-well plates were transfected with 2 µg EGFP-OdTBK1 or the empty vector. After 24 h, the cells were untreated (null) or transfected with 2 µg poly (I:C) or infected with 4 µl SVCV for 12 h, then the cells were fixed and subjected to confocal microscopy analysis. Green staining represents the OdTBK1 protein signal, and blue staining indicates the nucleus region (original magnification, ×63; oil immersion objective). Bar, 10µm. All experiments were repeated at least three times, with similar results.

### OdTBK1 Acted as a Kinase to Phosphorylate IRF3

As mentioned earlier, IRF3 is a crucial downstream signaling molecule of TBK1. After co-transfection of zebrafish IRF3 (DrIRF3) and OdTBK1, cells were lysed for immunoblotting. Compared with the signal band of DrIRF3 in the control group, band shifts occurred through stimulation with OdTBK1 (**Figure 6A**). To confirm whether the shifted bands represented phosphorylated IRF3, a dephosphorylation assay was performed in vitro. The shifted bands partially disappeared after being treated with calf intestinal phosphatase (CIP), demonstrating that DrIRF3 can be phosphorylated by OdTBK1 (**Figure 6B**). To further characterize this process, a truncated OdTBK1 mutant was generated to identify the functional domain that regulates the phosphorylation of DrIRF3. As shown in **Figure 6C**, the mutant lacking the kinase domain (OdTBK1-1N) failed to phosphorylate DrIRF3, suggesting that this domain was indispensable for its kinase activity (**Figure 6C**). In addition, to validate the association between OdTBK1 and DrIRF3, we transfected expression plasmids for DrIRF3-Myc and OdTBK1-Flag into HEK 293T cells and performed Co-IP assays. The results indicate that OdTBK1 interacted with DrIRF3 (**Figure 6D**) and was involved in phosphorylation of DrIRF3 via its functional kinase domain.

### OdTBK1 Induced Downstream ISGs Expression

OdTBK1 significantly phosphorylates IRF3, which is a crucial transcriptional factor of IFN, to activate ISGs expression.

FIGURE 6 | IRF3 was phosphorylated by OdTBK1. (A) Overexpression of OdTBK1 induces the phosphorylation of DrIRF3 in a dose-dependent manner. HEK 293T cells were seeded in six-well plates overnight and co-transfected with 1 µg DrIRF3-HA (attachment of a HA tag to the N terminus of DrIRF3) and 1 µg empty vector, or OdTBK1-Myc (0.25, 0.5 and 1 µg, respectively) for 24 h. The cell lysates were subjected to IB with the anti-HA, anti-Myc, and anti-β-actin Abs. (B) OdTBK1 mediates the phosphorylation of DrIRF3. HEK 293T cells were seeded in six-well plates overnight and co-transfected with 1 µg DrIRF3-HA and 1 µg empty vector or OdTBK1-Myc for 24 h. The cell lysates (100 µg) were treated with or without CIP (10 U) for 40 min at 37◦C. The lysates were then subjected to IB with the anti-HA, anti-Myc, and anti-β-actin Abs. (C) IRF3 was phosphorylated by the N terminus of OdTBK1. HEK 293T cells were seeded in six-well plates overnight and transfected with the indicated plasmids (1 µg each) for 24 h. The cell lysates were subjected to IB with the anti-Myc and anti-β-actin Abs. (D) OdTBK1 associates with DrIRF3. HEK 293T cells seeded into 10-cm<sup>2</sup> dishes were transfected with the indicated plasmids (10 µg each). After 24 h, cell lysates were immunoprecipitated (IP) with anti-Flag affinity gel. The immunoprecipitates and cell lysates were then analyzed by IB with anti-Myc and anti-Flag Abs, respectively. All experiments were repeated at least three times with similar results.

Therefore, the regulation activity of OdTBK1 was monitored. EPC cells were transfected with OdTBK1 and total RNA was extracted for reverse transcription. qRT-PCR was employed to detect several genes involved in IFN-mediated pathways, such as ifn, mavs, and vig1 (26). The transcription levels of these ISGs increased significantly compared with the control group.

The mRNA level of ifn was increased by 61-fold compared with the control group, and that of vig1 was increased by 15-fold (**Figures 7A–C**). These data indicated that OdTBK1 was capable of activating ISG transcription.

### Antiviral Activity of OdTBK1

Finally, to investigate the role of OdTBK1 in host antiviral immunity, EPC cells were transfected with OdTBK1 and the pCMV-Myc group was used as a control group. These two groups were separately infected with identical titers of SVCV at 24 h post-transfection. At 36 h after the infection, apparent CPE was observed in the control cells, while the CPE of the OdTBK1 group was obviously reduced (**Figure 8A**). Measurement of the virus titers of the OdTBK1 and control groups suggested that it was reduced over 30-fold in the OdTBK1-overexpressing cells compared with the control cells (**Figure 8B**). The data above demonstrated that OdTBK1 enhanced the antiviral ability of cells against the virus, which indicated that OdTBK1 participated as a crucial factor in the host antiviral innate immune response.

### DISCUSSION

As a key positive IFN regulator, TBK1 has been extensively studied in mammals, whereas studies of teleost TBK1 are rare. In this study, we describe a fish TBK1 isolated from O. obscura and its kinase function in inducing IFN expression and phosphorylating IRF3. These data demonstrate that TBK1 is conserved in both lower and higher vertebrates.

In our study, phylogentic analysis of the TBK1 amino acid sequences revealed that dark sleeper TBK1 is closely related to that of other fish and is likely an ortholog of mammalian TBK1s. This is consistent with TBK1 being the crucial kinase in IFN response, an essential and pivotal part of antiviral defense. In addition, other RLR factors such as MAVS, MITA, and IRF3 are conserved and functional in fish. Generally, IFN production activated by the RLR axis is necessary for both fish and mammal antiviral processes.

The tissue expression pattern indicated that OdTBK1 was highest in the pituitary, but relatively low in the spleen and head kidney. This is interesting because the pituitary participates in the formation of the hypothalamic–pituitary–interrenal axis for regulating the metabolism, immune response, and growth of fish generally (27, 28). The lower expression level of OdTBK1 in the head kidney and spleen is probably because OdTBK1 is not necessary in immune-relevant tissues of healthy fish. Also, OdTBK1 nay not be involved in adaptive immune processes.

In mammals, the induction of IFN and downstream ISGs symbolizes antiviral post infection response (29, 30). Similar to this, our study revealed that OdTBK1 exhibits a powerful effect on zebrafish and grass carp IFN promoter activation. Overexpression of OdTBK1 up-regulates the expression of fmifn and fmvig1, but not of fmmavs. MAVS belongs to the IFNrelated gene and recruits TBK1/IKKε by TRAFs, which means that MAVS is not the TBK1 downstream effector protein (31). Furthermore, fish MAVS has been reported to contribute to IFN antiviral immunity upstream of TBK1 and IRF3/7 (32). These findings explain why TBK1 cannot directly up-regulate the expression of fmmavs.

OdTBK1 was located in the cytoplasm with or without poly I:C and SVCV, indicating that TBK1 activates downstream signal molecules from the cytoplasm and that the cytoplasm is where TBK1 is regulated by upstream molecules. After being stimulated by TBK1, dimerized and phosphorylated IRF3 transfers from the cytoplasm to the nucleus, where it binds to ISRE motifs to initiate the transcription of target genes, including IFN and other ISGs (33). As a critical kinase involved in antiviral immunity, TBK1 activity can be regulated in a variety of ways, such as phosphorylation, ubiquitination, and kinase activity modulation. These modifications occur mainly in the cytoplasm. For instance, E3 ubiquitin ligase TRAF3 mediates lysine 63 (K63)-linked polyubiquitination of TBK1 and facilitates its activation in the cytoplasm (34). These data support the cytoplasmic localization of TBK1 as responsible for its activation and function.

Research has indicated that OdTBK1 exhibited kinase activity that induced the phosphorylation of DrIRF3. We have characterized OdIRF3, but do not yet know whether OdIRF3 can be phosphorylated by OdTBK1. The protein kinase domain of OdTBK1 was indispensable for its kinase activity; the dominant negative mutant TBK1-K38M of human or crucian carp is able to effectively block this activity (20, 33). Future studies are needed to construct the mutant OdTBK1-K38M to determine conservation of the kinase functional site.

In summary, we have identified and characterized OdTBK1 and verified the partial function of TBK1 in host antiviral innate immunity. More research is still needed to understand the molecular mechanisms behind the biological functions and regulation of the TBK1-mediated signaling pathway and to gain insight into the potential role of OdTBK1 in fish.

### AUTHOR CONTRIBUTIONS

SL conceived and designed the experiments. SL, JC, ZL, LL, PL, and X-YL performed the experiments and analyzed the data.

### REFERENCES


SL, JC, and ZC wrote the manuscript. All authors reviewed the manuscript.

### FUNDING

This work was supported by National Natural Science Foundation of China Grants 31502200 (SL).

### ACKNOWLEDGMENTS

We thank Dr. Fang Zhou (Institute of Hydrobiology, Chinese Academy of Sciences) for assistance with confocal microscopy analysis and Dr. Feng Xiong (China Zebrafish Resource Center, Institute of Hydrobiology, Chinese Academy of Sciences) for RNA sample extraction.


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

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

# Discovery of All Three Types in Cartilaginous Fishes Enables Phylogenetic Resolution of the Origins and Evolution of Interferons

Anthony K. Redmond1,2,3 \*, Jun Zou1,4,5, Christopher J. Secombes 1,4, Daniel J. Macqueen1,6 and Helen Dooley 1,7,8 \*

*<sup>1</sup> School of Biological Sciences, University of Aberdeen, Aberdeen, United Kingdom, <sup>2</sup> Centre for Genome-Enabled Biology and Medicine, University of Aberdeen, Aberdeen, United Kingdom, <sup>3</sup> Smurfit Institute of Genetics, Trinity College Dublin, University of Dublin, Dublin, Ireland, <sup>4</sup> Scottish Fish Immunology Research Centre, Institute of Biological and Environmental Sciences, University of Aberdeen, Aberdeen, United Kingdom, <sup>5</sup> Key Laboratory of Exploration and Utilization of Aquatic Genetic Resources, Ministry of Education, Shanghai Ocean University, Shanghai, China, <sup>6</sup> The Roslin Institute and Royal (Dick) School of Veterinary Studies, The University of Edinburgh, Edinburgh, United Kingdom, <sup>7</sup> Department of Microbiology and Immunology, University of Maryland School of Medicine, Baltimore, MD, United States, <sup>8</sup> Institute of Marine and Environmental Technology, Baltimore, MD, United States*

### Edited by:

*Leon Grayfer, George Washington University, United States*

#### Reviewed by:

*Jean-Pierre Levraud, Institut Pasteur, France Katherine Buckley, Carnegie Mellon University, United States*

#### \*Correspondence:

*Anthony K. Redmond anthony.k.redmond@gmail.com Helen Dooley hdooley@som.umaryland.edu*

#### Specialty section:

*This article was submitted to Comparative Immunology, a section of the journal Frontiers in Immunology*

Received: *24 April 2019* Accepted: *21 June 2019* Published: *12 July 2019*

#### Citation:

*Redmond AK, Zou J, Secombes CJ, Macqueen DJ and Dooley H (2019) Discovery of All Three Types in Cartilaginous Fishes Enables Phylogenetic Resolution of the Origins and Evolution of Interferons. Front. Immunol. 10:1558. doi: 10.3389/fimmu.2019.01558* Interferons orchestrate host antiviral responses in jawed vertebrates. They are categorized into three classes; IFN1 and IFN3 are the primary antiviral cytokine lineages, while IFN2 responds to a broader variety of pathogens. The evolutionary relationships within and between these three classes have proven difficult to resolve. Here, we reassess interferon evolution, considering key phylogenetic pitfalls including taxon sampling, alignment quality, model adequacy, and outgroup choice. We reveal that cartilaginous fishes, and hence the jawed vertebrate ancestor, possess(ed) orthologs of all three interferon classes. We show that IFN3 groups sister to IFN1, resolve the origins of the human IFN3 lineages, and find that intronless IFN3s emerged at least three times. IFN2 genes are highly conserved, except for IFN-γ-rel, which we confirm resulted from a teleost-specific duplication. Our analyses show that IFN1 phylogeny is highly sensitive to phylogenetic error. By accounting for this, we describe a new backbone IFN1 phylogeny that implies several IFN1 genes existed in the jawed vertebrate ancestor. One of these is represented by the intronless IFN1s of tetrapods, including mammalian-like repertoires of reptile IFN1s and a subset of amphibian IFN1s, in addition to newly-identified intron-containing shark IFN1 genes. IFN-f, previously only found in teleosts, likely represents another ancestral jawed vertebrate IFN1 family member, suggesting the current classification of fish IFN1s into two groups based on the number of cysteines may need revision. The providence of the remaining fish IFN1s and the coelacanth IFN1s proved difficult to resolve, but they may also be ancestral jawed vertebrate IFN1 lineages. Finally, a large group of amphibian-specific IFN1s falls sister to all other IFN1s and was likely also present in the jawed vertebrate ancestor. Our results verify that intronless IFN1s have evolved multiple times in amphibians and indicate that no one-to-one orthology exists between mammal and reptile IFN1s. Our data also imply that diversification of the multiple IFN1s present in the jawed vertebrate ancestor has occurred through a rapid birth-death process, consistent with functional maintenance over a 450-million-year host-pathogen arms race. In summary, this study reveals a new model of interferon evolution important to our understanding of jawed vertebrate antiviral immunity.

Keywords: interferon, phylogenetics, evolution, antiviral immunity, cytokine, retrotransposition, jawed vertebrate, shark

### INTRODUCTION

Antiviral immunity in jawed vertebrates is directed by interferons released by host cells in response to viral pathogens (1, 2). Interferons are members of the class II α-helical cytokines along with interleukin (IL)-10,−19,−20,−22,−24, and −26 (hereafter called the IL-10 family), and are categorized into three classes, denoted as type I [e.g., IFN-α, β, κ, etc. in human [amongst others] and chicken], II (IFN-γ), and III (IFN-λs; known as IL-28A, IL-28B and IL29 in humans) interferons (hereafter called IFN1, IFN2, and IFN3), based on their receptors, genomic location, and sequence/structural homology (2, 3). Roles beyond antiviral immunity have recently come to light for interferons, and IFN2 has been shown to contribute mainly toward defense against bacterial (especially mycobacteria), parasitic and fungal pathogens, leaving IFN1 and IFN3 as the main antiviral cytokines (1, 2, 4–7).

The evolutionary relationships between the three interferon classes, as well as intra-class evolutionary histories, have received considerable attention, but have proven difficult to resolve. The origins of the IFN3 lineage are particularly contentious. While some early studies suggested that the IFN3 and IFN1 lineages diverged in tetrapods, with teleost fishes possessing IFN1/3 like molecules (**Figure 1A**) (8–10), other studies suggested that teleosts possessed either IFN3 (11, 12) or IFN1 orthologs (13). Later work, incorporating protein structures, showed that the teleost molecules were indeed IFN1s (14), and suggested that IFN3s likely emerged early in vertebrate evolution following whole genome duplication events (15–17). This scenario was supported by the discovery of IFN3 receptor homologs (along with those of IFN1 and IFN2) in cartilaginous fishes (18, 19). However, other structure-based studies concluded that IFN3 is either part of the IL-10 family (specifically IL-22 or IL-19) (**Figure 1B**) (3, 20–22), or sister to IFN1s (**Figure 1C**) (15). A recent study has also revived the idea that IFN3 may have emerged from within IFN1 in the tetrapod ancestor (**Figure 1A**) (23). Crucially, as no study has yet identified either the root of the class II α-helical cytokine family or orthologs of IFN3 genes outside tetrapods, none of these three hypotheses has been firmly rejected.

Evolutionary histories within each interferon class also remain unclear. For example, IFN2 is typically considered the most conserved interferon, however a tandem duplicate (IFN-γ-rel) has been found in some teleost fish-species (24), and phylogenetic analyses have failed to clarify whether this an ancient jawed vertebrate gene lost in other lineages (24–26), or teleost-specific (27). Multiple IFN3 genes often exist in individual species, but IFN3s are thought to be tetrapod-specific (28). However, very few studies have specifically focused on IFN3 evolution across vertebrates (28). The evolution of IFN1 genes, while better studied, also appears to be the most complicated. IFN1 genes are often present as lineage-specific clusters; for example, with the exception of IFN-κ, the IFN1 molecules of humans are evidently not directly orthologous with those of chickens (29– 31). Clusters of lineage-specific IFN1s have also been observed in teleost fishes (32), classified as belonging to fish-specific group 1 or group 2 based on cysteine patterns (having two and four conserved cysteines, respectively) in the mature peptide sequence (9), and in amphibians (23). In fact, some phylogenetic analyses place all IFN1 sequences from mammals, teleosts, and amphibians into lineage-specific clades (9, 33–35), supporting a scenario where IFN1s evolve through concerted evolution (36, 37). This would imply that high-turnover, lineage-specific gene gain and loss events, and/or gene conversion are major driving forces of IFN1 evolution (37, 38). This is consistent with functional data, where individual genes appear to be specialized for defense against specific viruses. However, some studies have found phylogenetic relationships between IFN1s that are more difficult to interpret (23).

Poor resolution of interferon phylogenies hinders our ability to infer the history of evolutionary events including retro(trans)position, intron gains and losses, and changes in disulphide bridge structure. Amniote IFN1s are intronless and are classically thought to have arisen as a result of a retro(trans)position early in amniote evolution, as fish and amphibian interferons were found to contain four introns (2, 9, 19, 30, 33, 39). Recent studies have revealed that both introncontaining and intronless IFN1 genes also exist in amphibians, leading to two competing hypotheses to explain the origins of intronless amphibian IFN1 genes; (i) they arose from the same event as the amniote intronless IFN1s (23, 35), or (ii) they arose during independent retro(trans)position events (40). Intronless IFN3 genes also exist in mammals and amphibians (23), but whether they resulted from a single event or not remains to be tested. Similarly, two- and four-cysteine containing IFN1s exist in mammals and teleosts but it is thought that two-cysteine containing IFN1s emerged independently in each lineage; with intronless mammal and intron-containing teleost two-cysteine containing IFNs having lost a different cysteine pair (and hence disulphide bridge) from an ancestral four-cysteine containing IFN1 (19). Better resolution of IFN1 and IFN3 evolution could help determine both the frequency and timing of emergence of such features.

The primary amino acid sequence of interferons are short and rapidly evolving, both characteristics expected to promote phylogenetic error (41). Short alignments may have insufficient phylogenetic information to infer relationships between sequences and are more prone to stochastic errors.

On the other hand, rapidly evolving sequences can be difficult to align and may induce systematic errors, resulting in long branch attraction (LBA) (42). Homoplasy (i.e., convergence due to hidden substitutions) is the best studied cause of LBA, and has previously been acknowledged as a concern when inferring immune gene phylogenies (41). Fortunately, it can often be counteracted by breaking long branches with additional taxa (43–45), applying site-heterogeneous models of evolution (42, 46), removing fast-evolving sites (47), and/or identifying the best outgroup (48–51), in addition to using outgroupfree methods (41, 52–55) to root the tree. Compositional heterogeneity, resulting from differing codon usage preferences among sequences under comparison can also lead to LBA through non-phylogenetic similarity between lineages (56, 57), and can be remedied by applying time-heterogeneous models of evolution (58, 59), or removing compositionally biased sites or sequences (56, 57, 60, 61). Other sources of systematic error have been identified (e.g., heterotachy, heteropecilly, nonindependence of sites), but are either less well studied or thought to have a less important effect on tree topology (62). Attempting to account for multiple sources of error, or applying several errorattenuating strategies at once is thought to improve phylogenetic accuracy (49, 63, 64), and this has proven successful for immune genes in the past (41, 51).

Here, taking account of important phylogenetic considerations overlooked in past studies, we infer the origins and evolutionary history of interferons using a dataset that incorporates unprecedented sampling of both species and interferon diversity. Our findings offer a substantially overhauled model of interferon evolution and provide insights into the varied issues that hinder such studies, which have broader implications for immune gene phylogenetic analysis.

### MATERIALS AND METHODS

### Homolog Identification and Characterization

TBLASTN (65) searches were carried out against a denselysampled set of genomes spanning chordate phylogeny (**Table S1**). An e-value cut-off of 10 was used in all searches, and sequences with either >75% identity compared to the query sequence [a set of known phylogenetically diverse IFNs were used for all searches, while known sequences (**Table S2**) from closely related species were also applied on an ad hoc basis], and/or with a top BLASTP (66) hit against an interferon in the NCBI non-redundant protein database, were retained for further analysis. To increase taxon sampling of cartilaginous fishes beyond elephant shark [until recently the only cartilaginous fish species with a sequenced genome (18, 67)], transcriptome datasets for the small spotted-catshark were also analyzed from this lineage (41, 68, 69). Gene predictions were performed where a protein sequence was not already available, with the FGENESH+ webserver, using parameters for the closest related species available, and using either the blast hit or query as the homologous sequence (70). Structural homology prediction was performed through the Phyre2 protein structure prediction webserver using the "intensive" search option (71). Assessment of evolutionary conservation of sites necessary for IFN-λ3-receptor binding was achieved through visual comparison of multiple sequence alignment.

### Multiple Sequence Alignment

All multiple sequence alignments were generated using PRANK, which has been shown to improve inference of insertions and deletions compared to other alignment approaches (72). This should help avoid alignment of non-homologous sites, reducing the potential for phylogenetic error. Manual curation was also performed (e.g., positions with no homologous amino acid in other classes were removed when examining inter-class relationships). Due to the rapidly evolving nature of IFN1s a set of high-quality, known sequences (**Table S2**) was used to build a base alignment before adding additional sequences from transcriptome and draft genome datasets, which may be truncated and more error prone, to the IFN1 dataset. Prior to analyzing this dataset, the PRANK alignment process was bootstrapped using GUIDANCE, which identifies sites that are not consistently aligned (73, 74). Site alignments present in <93% of the GUIDANCE replicates were then removed to avoid use of unreliably aligned sites in phylogenetics (73, 74). The "—add" function of MAFFT was then used with the L-INS-i approach to add new sequences to this high-quality core alignment (75). Alignment positions present in only a single species were then removed. See **Data S1** for all multiple sequence alignment files.

### Phylogenetic Analyses

All maximum likelihood phylogenetic analyses and model selection were performed in IQ-tree v1.6.7 (76). The Bayesian information criterion was used for model selection using IQtree's ModelFinder (77), and 1,000 ultrafast bootstrap replicates were generated to provide branch support values (78). IQ-tree was also used to detect compositionally biased sequences, using the built-in χ 2 test (71).

Outgroup-free rooted phylogenetic analyses were performed using a relaxed clock model, that permits root inference while accommodating rate variation among different tree branches (52). We have previously applied this approach to root other fast-evolving immune gene families (41, 51, 54, 79, 80), and it appears to work consistently for such datasets, except in the face of extreme rate asymmetry (41). This analysis was performed in BEAST v1.8.3 (81) applying an uncorrelated lognormal relaxed molecular clock model (52), a Yule speciation prior (82, 83), and the best-fit amino acid substitution model (as inferred with IQ-tree). Two Markov chain Monte Carlo (MCMC) chains were run until effective samples sizes (>200) and convergence were sufficient, as assessed in Tracer v1.6 (84). Maximum clade credibility trees were generated in RootAnnotator (85).

Bayesian phylogenetic analyses incorporating outgroups were performed in PhyloBayes v4.1b (86), which also permits testing of site-heterogeneous models. Two MCMC chains were run for each analysis, until convergence was reached and effective sample sizes were sufficient for all statistics. This was assessed using the bpcomp (maxdiff < 0.3) and tracecomp (effsize > 50, and rel\_diff < 0.3) programs within the PhyloBayes package (86).

### Site-Heterogeneous Models, Cross-Validation, and Posterior Predictive Analyses

Site-heterogeneous models typically allow for better detection of homoplasy by accommodating site-specific evolutionary constraints in phylogenomic datasets (42). Such models have been applied with the objective of generating more reliable immune gene phylogenies (87, 88), and have recently been shown to be capable of better explaining the site-specific evolutionary processes of aligned immune gene datasets (41). However, such models do not always provide a better fit for short alignments, and their relative fit cannot always be compared to standard models with the commonly applied information criteria. As such we used 10-fold cross-validation, as implemented in PhyloBayes (42, 86), to compare the relative fit of a range of site-heterogeneous mixture models to the bestfitting standard model for the IFN1 dataset [JTT+Ŵ (89, 90)]. The models tested included the infinite mixture model CAT (46), empirical derivations of CAT (C10/20/30/40/50/60) with limited numbers of site-categories, intended for gene family phylogenies (91), as well as an alternative site-heterogeneous model, WLSR5 (92), and a three-matrix substitution model, UL3, that loosely accommodates evolutionary process differences between structural features (91). Cross-validation relies on randomly partitioning the alignment into equal sized subsamples (10 here, as the analysis is 10-fold), before one of these subsamples is used for validation to test the model, while the rest are combined as a training set. This process is then repeated, using each of the other subsamples as the validation dataset, and then the average results are used for comparison against other models. We ran each individual chain (i.e., one chain for each of the 10 training sets for each model) for 1,000 points, using the first 100 as burn-in (i.e., 10 chains for each model tested).

In addition to assessing relative model fit, posterior predictive simulations (PPS) were also performed in PhyloBayes to determine if the model applied could adequately describe the real data for the tested statistic (42, 86). This approach consists of generating simulated data under the model in question, for comparison against the observed (i.e., real) data. Here, PPS was used to investigate the ability of models to account for homoplasy and compositional heterogeneity across lineages in the IFN1 dataset (42, 56, 57, 60). The compositional heterogeneity test was used to generate a second IFN1 dataset by identifying and removing sequences that deviate significantly from the assumption of homogeneity, measured at Z-scores < −2 and >2 (the default in PhyloBayes, which is slightly more inclusive than P < 0.05 cut-off). All PPS analyses were specifically performed under JTT+Ŵ, as this model should be the most susceptible to error compared to the tested mixtures, so we viewed this as more conservative. Finally, we also tested a time-heterogeneous (59), and a site- and time-heterogeneous (58) model for the IFN1 dataset, but these analyses failed to reach convergence despite running continuously for more than three months each.

### Testing Exacerbation of Potential Errors in Interferon Phylogenetics Analysis

Multiple approaches were tested to induce phylogenetic error (93) in the rapidly-evolving IFN1 family to better explain the discrepancy in the results of past studies, as well as the difficulty in inferring IFN1 evolutionary history. This included applying a more distantly related outgroup in place of the closest related outgroup to root the tree (50, 51, 94), inferring the phylogeny under less well-fitting substitution models (50, 51, 54, 93, 95), including sequences that introduce significant compositional bias in the analysis (63, 64), as well as sequence removal to lengthen target branches and increase the potential for LBA (93, 96).

## RESULTS

### A Cartilaginous Fish IFN-λ

Reciprocal BLAST searches of a multi-tissue small-spotted catshark transcriptome (41) revealed a putative cartilaginous fish IFN3 sequence. Characterization of this sequence by multiple sequence alignment against human interferon sequences and IL-10 (as a representative of the broader IL-10 family) support this assignment (**Figure 2A** and **Figure S1**). Additionally, the signature disulfide bridge-forming Cys pair were present at the C-terminus of this sequence (**Figure 2A** and **Figure S1**) (21, 22). Interestingly, the most important receptor binding sites of human IFN-λ are poorly conserved in this sequence (22); including Phe158, which is vital for human IFN-λ3-receptor interaction (22). However, this may not preclude antiviral functionality of catshark IFN3, considering that this residue is

FIGURE 2 | (A) Analysis of key residues in the catshark IFN-λ sequence compared to a set of other interferons and IL-10. Sequences are represented as cartoon bars, which are relatively scaled according to amino acid sequence length. Arrows denote the end of the signal peptide region, while disulphide bridges are shown as connected regions underneath each cartoon bar, with "C" in the C-terminal region being an unpaired Cys from the characteristic C-terminal disulphide bridge of IFN3s (22). Above the bar the most important residues for IFN-λ3 receptor binding are shown (22). Residues filled in black are conserved, whereas residues filled white are not well conserved, and gray-filled residues involve conserved replacements (e.g., K→ R). The bar over the VXXQ motif of catshark IFN-λ indicates that this is not aligned perfectly to human IFN3s, while the star indicates that mutation of this residue abolishes binding in human IFN-λ3 (22). See Figure S1 for full alignment. (B) Relaxed clock (uncorrelated lognormal) rooted class II α-helical cytokine family phylogeny under JTT + I + τ and a Yule speciation prior. The tree is rooted at the best supported root position. Root posterior probabilities (RPP) are shown for branches with a non-negligible probability (i.e., posterior probability <0.05) of being the root. Posterior probabilities are also shown for key nodes, and clades representing individual family members, or the entire IL-10 family have been collapsed to emphasize deep relationships within the family.

also not conserved in human IFN-λ4 (**Figure 2A** and **Figure S1**), which appears to be capable of binding the IFN3 receptor (97). Further, our preliminary analyses suggest that catshark IFN-λ is involved in antiviral defense (unpublished data). Submission to the Phyre2 protein structure prediction server, an approach which has previously been employed to aid orthology assignment of fast-evolving immune genes in cartilaginous fishes (98), also indicated a best-structural match of the putative catshark sequence to mammalian IFN-λs (**Figure S2**). Finally, phylogenetic analysis (see next section), verified this assignment; indeed, catshark IFN-λ forms a clade with tetrapod IFN3s, to the exclusion of other class II α-helical cytokines, with maximal support (posterior probability [PP] = 1.00; **Figure 2B**, **Figure S3**). The existence of a cartilaginous fish IFN3 allows us to unequivocally reject the hypothesis that IFN3s emerged by duplication from IFN1s in the tetrapod ancestor (**Figure 1A**) (8, 10, 23).

### Deep Relationships Within the Class II α-Helical Cytokine Family

To understand the evolutionary relationships between the interferon classes and other class II α-helical cytokines, and identify the closest outgroups to best infer within-class interferon relationships, we performed a phylogenetic analysis of the full class II α-helical cytokine family. A relaxed clock model was used to root this tree, as the deeper phylogenetic origins of the family, and thus potential outgroups, are not known. Our analysis supports a sister group relationship between IFN1 and IFN3 (PP = 0.91; **Figure 2B** and **Figure S3**), while on the other side of the tree root (root posterior probability [RPP] = 0.82; **Figure 2B** and **Figure S3**), a monophyletic IL-10 family is sister to IFN2 (PP = 0.9; **Figure 2B** and **Figure S3**). These findings reject the hypothesis that IFN3 is part of the IL-10 family (14, 21, 22) (**Figure 1B**), while the root placement suggests that the deepest divergence in the class II α-helical cytokines separates the main antiviral interferons from the rest of the family, consistent with the model of class II α-helical cytokine evolution proposed by Siupka et al. (15). A second root position placing IFN1 as sister to the rest of the family could not be rejected however, although this was only very weakly supported (PP = 0.05; yielding a 16:1 weighting in favor of the best root) (**Figure 2B**). These results concur with the conclusion that the IFN1/3s previously identified outside tetrapods are in fact true IFN1s, consistent with their structural and functional features (14). Further, by supporting a sister group relationship between IFN1 and IFN3 (**Figure 1C**), our findings indicate that IFN1 and IFN3 can be used as reciprocal outgroups in phylogenetic analyses, enabling outgroup-rooted IFN1 and IFN3 phylogenies without the inclusion of more distant, and potentially-biasing, outgroups like IFN2 and/or the IL-10 family (which by the same rationale can be applied as outgroups for each other).

### IFN2 Evolution Indicates That IFN-γ-Rel Is Teleost-Specific

IFN2 is the most structurally conserved of the interferon classes and is thought to have the simplest evolutionary history. Despite being present in single copy across most of vertebrate phylogeny, an additional gene, IFN-γ-rel (24), is present in tandem to IFNγ in some teleosts. Phylogenies in some previous studies suggest IFN-γ-rel could be an ancient lineage, lost from other vertebrates (24–26). Not all phylogenies support this however (27), and it has also been suggested that IFN-γ-rel arose through duplication of IFN-γ during teleost evolution (19). Here, we tested this using a PRANK alignment of IFN-γ sequences spanning jawed vertebrate phylogeny, as well as the best-fitting substitution model, and the most closely related outgroup, the IL-10 family. This phylogenetic analysis maximally supported IFN-γ-rel as sister to teleost IFN-γ (Ultrafast Bootstrap [UFBOOT] = 100%) (**Figure 3**). Together with its absence outside of teleosts, this indicates that IFN-γ-rel resulted from a teleost-specific tandem gene duplication.

### Divergence of Human IFN-λs and Convergent Intron Loss in IFN3 Evolution

In light of the newly discovered catshark IFN-λ, and identification of the IFN1 family as the closest outgroup, we reassessed the evolutionary history of the IFN3s (**Figure 4** and **Figure S4**). Along with a selection of known tetrapod IFNs,

FIGURE 4 | IFN3 phylogeny. Maximum likelihood consensus tree of the IFN3 genes under JTT+τ with IFN1s as outgroup. Clades are collapsed into major lineages and all ultrafast bootstrap support values are shown for non-collapsed portions of the tree.

BLAST analyses of genomes spanning vertebrate phylogeny revealed putative new amphibian and reptile IFN3 sequences, but consistent with previous studies failed to identify coelacanth or teleost IFN3s. Phylogenetic analysis of this dataset revealed that catshark IFN-λ falls sister to all tetrapod IFN3s (UFBOOT = 91%) (**Figure 4** and **Figure S4**), while within tetrapods, amphibians and amniotes form separate sister clades (UFBOOT = 88%) (**Figure 4** and **Figure S4**). Our analyses verified the presence of intronless IFN3s in amphibians (23). Strikingly however, we found that the intronless IFN3s of amphibians and mammals emerged independently, and that intronless IFN3s have evolved at least twice during amphibian evolution (UFBOOT = 100%) (**Figure 4** and **Figure S4**), and hence at least three times throughout vertebrate evolution. Within amniotes, our results are largely consistent with those of Chen et al. (28), as we find that reptiles have at least two IFN3 lineages. In our analyses these lineages form clades with mammalian IFN-λ4 (though reptiles are not monophyletic) (UFBOOT = 97%) and mammalian IL-28/29 (UFBOOT = 82%), suggesting that they are orthologous, and that the IL-28/29 and IFN-λ4 lineages split in the amniote ancestor (**Figure 4** and **Figure S4**). The reptile IL-28/29-like gene appears to have been duplicated in the ancestor of archelosaurians (turtles, birds, and crocodiles) (UFBOOT = 97%), while the human IL-28 and IL-29 lineages appear to have been duplicated in placental mammals (UFBOOT = 98%), with IL-28A and IL-28B later splitting during primate evolution (**Figure 4** and **Figure S4**).

### Accounting for Phylogenetic Errors to Generate a Reliable IFN1 Tree

The evolutionary history of IFN1s has been studied intensively and many very different tree topologies generated. However, previous studies have not intentionally accounted for any of the major known sources of phylogenetic error. To help counter this we first applied two data-centric approaches designed to combat phylogenetic errors (43–45, 48–50, 94, 96). First, we applied only the closely related IFN3 as an outgroup, and increased taxon sampling by identifying new IFN1s from a dense sample of genomes across vertebrate phylogeny. This revealed hundreds of new IFN1 sequences (**Table S2**), which we subsampled prior to phylogenetic analyses, keeping only sequences above 100 amino acids in length (except for cartilaginous fishes and Japanese eel, where sequences of 50 or more amino acids were retained, given the paucity of data available for these species and their important evolutionary placement within jawed vertebrates and teleosts), and removing highly similar sequences within species from densely sampled lineages to reduce computational burden without negatively affecting deeper nodes in the tree.

Next, we tested the utility of site-heterogeneous phylogenetic mixture models to resolve the IFN1 phylogeny; such models have been shown to offer an improved fit to many datasets (42, 49, 95), including immune genes (41), as well as being more resistant to LBA artifacts (42) by accounting for evolutionary process variation among sites. As such we ran PhyloBayes analyses under the best-fitting standard model, JTT+Ŵ, as well as under a variety of site-heterogeneous mixture models (46, 91, 92, 99). Unfortunately, none of these runs reached convergence. However, this is not an uncommon occurrence in PhyloBayes analyses of difficult datasets, and can be remedied by identifying and removing error causing sequences or branches (49).

Interrogation of the MCMC chains for the JTT+Ŵ analyses revealed that effective sample sizes were sufficient, but that individual runs had become "stuck" at different log-likelihoods (the lesser of which must represent a local optimum), and at different tree topologies (i.e., PhyloBayes bpcomp "maxdiff = 1"). Trying to resolve this objectively, while also reducing systematic error, we used PPS to detect and remove sequences that deviated from the assumption of compositional heterogeneity (60). We then re-ran the phylogenetic analyses using this reduced "compositionally homogeneous" dataset (CHOM) and found that runs now converged for JTT+Ŵ (**Figure 5A** and **Figure S5**), as well as all the tested mixture models. Before examining the resultant tree topologies however, we sought to gain further understanding of the difference between analyzing the full and CHOM datasets, and to determine if the site-heterogeneous models might provide a better fit than JTT+Ŵ.

Interrogation of the compositional heterogeneity PPS results for the full dataset showed consistency between chains, with almost fully overlapping sets of sequences identified as biased in both chains, implying that sequence removal (i.e., identification of compositional bias) was not affected by lack of convergence (**Figure 5B**; **Table S3**), and therefore should be reliable.

An additional consideration is that removal of sequences may serve to lengthen branches in the phylogenetic tree, reducing the ability to detect hidden substitutions and increasing the potential of LBA artifacts. To determine if this was an issue we performed PPS analyses to test whether JTT+Ŵ could adequately accommodate homoplasy in both the full and CHOM datasets (42). While a greater level of homoplasy was both observed and predicted in the full dataset, this was adequately predicted by JTT+Ŵ (**Figure 5C**), implying that it should not be a major source of error (including topological inconsistency between chains), in either dataset.

Finally, using Bayesian cross-validation (42) we determined that JTT+Ŵ was in fact better-fitting than any of the tested mixture models (**Figure 5D**), perhaps due to the short alignment length, and so only this tree was used to make evolutionary inferences.

### Birth-Death Evolution of Multiple IFN1 Genes Since the Jawed Vertebrate Ancestor

Phylogenetic analyses of the CHOM dataset under JTT+Ŵ (**Figure 5A** and **Figure S5**) suggest a new paradigm for IFN1 evolution. The resultant tree indicates that duplication and loss events have occurred frequently since the origins of IFN1s (**Figure 5A** and **Figure S5**). This fits a rapid birth–death model of evolution (100, 101), as proposed for salmonid IFN1s (32), rather than the concerted evolution model (i.e., IFN1 expansions and contractions are confined to specific lineages) implied by many previous phylogenies. Multiple features of the tree topology support this scenario, including the presence of an amphibian lineage (red star in **Figure 5A**) that falls sister to all other IFN1s (PP = 0.65), intimating this lineage existed in the jawed vertebrate ancestor and has since been lost in other jawed vertebrates (**Figure 5A**; **Figure S5**). In addition, we identified a new IFN1 locus in elephant shark (**Table S2**), the intron-containing genes of which form the sister group to a clade (PP = 0.82) containing all amniote intronless interferons (PP = 0.74), suggesting that these genes are orthologous and that other jawed vertebrate classes have lost orthologs of this gene (**Figure 5A** and **Figure S5**). This means that the intron-containing IFN1s of teleosts and amphibians which gave rise to the idea that the retrotransposition event occurred in the amniote ancestor are in fact paralogous to this lineage, and as such are not informative on this point. Within the intronless amniote interferon clade, reptiles possess lineagespecific expansions comparable to those seen in mammals, and may retain ancient amniote IFN1s lost in mammals (**Figure 5A**). Elsewhere in the tree, we find that IFN-f, previously found only in teleosts, is likely also present in amphibians (PP = 0.74), and possibly cartilaginous fishes (PP = 0.53), where a lineagespecific expansion has occurred (green star in **Figure 5A**). Thus IFN-f appears to be an ancient jawed vertebrate IFN1 lineage, secondarily lost in amniotes. Importantly, this suggests that the ray-finned fish IFN1s may not be monophyletic and that the two groups defined by cysteine structures (9) may not have a phylogenetic basis (**Figure 5A** and **Figure S5**). Despite this, non-IFN-f ray-finned fish IFN1s group together (PP = 0.59), though, due to a polytomy, they do not appear to be identifiably orthologous to IFN1s from any other jawed vertebrate lineage (**Figure 5A** and **Figure S5**). The relationships of coelacanth IFN1s are similarly unresolved (**Figure 5A** and **Figure S5**). This result seems most consistent with both lineages representing lineage-specific expansions of ancient jawed vertebrate IFN1 genes lost in other jawed vertebrates, though the support for any

FIGURE 5 | Phylogenetic investigation of IFN1 evolution. (A) Bayesian consensus tree of the CHOM IFN1 dataset under JTT+Ŵ. (B) Posterior predictive simulations (PPS) showed that ∼50 IFN sequences introduced significant potential for compositional bias, which were removed to minimize branching artifacts (i.e., forming the CHOM dataset used for part (A). (C) PPS also shows that JTT adequately predicts homoplasy in both the full and CHOM datasets. (D) Model selection via 10-fold Bayesian cross-validation indicates that the site-homogenous JTT model fits the data better than a range of site-heterogeneous mixture models. (E) IFN1 topology is highly sensitive to both dataset bias and methodological error: (i) the full (i.e., compositionally heterogeneous) dataset places the cartilaginous fish group otherwise identified as sister to amniote IFNs in a monophyletic group with the amphibian sequences that form the sister group to all other IFN1s (see also Figures S6, S7), while (ii) and (iii) show that less well fitting models (see also Figure S8) and distant outgroup taxa (full tree in Figure S9) result in evolutionarily irreconcilable root placement.

scenario is low. Taken together, these results imply that several distinct IFN1 genes existed in the jawed vertebrate ancestor and have undergone rapid birth-death evolution since, meaning that ancient interferon genes are sometimes retained in only one or very few extant descendant taxa, while at the same time lineage-specific interferon expansions and contractions are common.

### Sensitivity of IFN1 Family to Phylogenetic Error

Major differences were observed between our results and those of previous studies. While we believe this is a result of improved methodology, we attempted to formally test this by performing experiments designed to exacerbate error potential in phylogenetic analyses (93). First, given that sequences displaying compositional bias contributed to non-convergence of PhyloBayes analyses, we instead built the full IFN1 phylogeny using alternative software (BEAST). This produced a similar topology to that obtained for the PhyloBayes CHOM analysis, however, the cartilaginous fish lineage that fell sister to the intronless amniote IFNs in the CHOM PhyloBayes analysis was instead placed sister to the amphibian sequences that fell sister to all other IFN1s (PP = 0.5) (**Figure 5E** and **Figure S6**). As the CHOM dataset does not deviate from the assumption of compositionally homogeneity, we considered this result to be an error induced by compositional bias. To further explore stability of this cartilaginous fish lineage in the CHOM dataset, we pruned sequences contributing to nearby branches to lengthen this branch, but this did not perturb its placement in the tree (**Figure S7**). Second, we examined tree topologies generated under the less well-fitting mixture models (**Figure S8**). Even for the second best-fitting model, UL3, this resulted in major issues with root placement (**Figure 5E** and **Figure S8**), suggesting an extremely non-parsimonious evolutionary scenario. Third, a similar outcome was observed when more distantly related IFN2 was applied as the outgroup instead of IFN3 (**Figure 5E** and **Figure S9**). Collectively, these results suggest that IFN1 phylogeny is highly sensitive to methodological and sampling errors.

### Intronless IFN1s Emerged in the Tetrapod Ancestor and Multiple Times in Amphibians

Since performing our IFN1 analyses, recent studies have identified new sequences not present in our dataset that may be relevant to IFN1 evolution. However, given the large compute time of performing all our PhyloBayes analyses (i.e., including cross-validation and PPS), it was not practical to rerun these with the addition of the new sequences (102). Instead, we decided upon a reasonable compromise; given that the best-fit model in PhyloBayes was a standard site-homogeneous model, we added the relevant sequences to our alignment and ran this under JTT+Ŵ in IQ-tree, with compositionally biased sequences (as identified by the χ 2 test implemented in IQ-tree) removed.

This dataset (hereafter called EXT) included the recently identified fish IFN-h (103), as well as additional IFNs (both intron-containing and intronless) from amphibians. The EXT phylogenetic tree (**Figure 6A** and **Figure S10**) is generally consistent with that of the CHOM analysis, except that non-IFN-f ray-finned fish IFN1s, coelacanth IFN1s, and IFN-f form a weakly supported clade (UFBOOT = 52%) rather than a polytomy (i.e., PP < 0.5 in PhyloBayes analyses) (**Figures 5A**, **6A**; **Figures S5, S10**). Within this clade, non-IFNf ray-finned fish IFN1s fall sister to IFN-f (UFBOOT = 79%) with the coelacanth IFN1s being sister to both (**Figure 6A** and **Figure S10**). Because the IFN-f clade includes cartilaginous fish (UFBOOT = 79%) and amphibian (UFBOOT = 86%) sequences, this is consistent with non-IFN-f ray-finned fish IFN1s being the only surviving lineage of an interferon gene that was present in the jawed vertebrate ancestor (**Figure 6A** and **Figure S10**). A similar evolutionary scenario can thus be applied to coelacanth IFN1s, but support for this is weaker (UFBOOT = 52%) (**Figure 6A** and **Figure S10**). The newly included IFN-h falls within the clade of non IFN-f teleost IFN1s (UFBOOT = 93%), and as such does not alter the backbone IFN1 phylogeny (**Figure 6A** and **Figure S10**). Similarly, despite being placed differently in past analyses (23, 35), we find that almost all of the recently identified amphibian IFN1s (23, 35) fall into the clade of amphibian sequences (UFBOOT = 100%) that is sister to all other IFN1s (UFBOOT = 88%), in the CHOM analysis (**Figures 5A**, **6A**; **Figures S5, S10**). Within this clade, the deepest split falls between intronless Xenopus IFN1s, and a clade containing intron-containing Xenopus and Nanorana parkeri sequences, as well as intronless N. parkeri sequences, confirming the recently discovered independent origins of intronless IFN1s in these species (40) (**Figure 6B** and **Figure S11**).

Strikingly, a small number of intronless amphibian IFN1s were nested within the mammal and reptile IFN1 clade, falling sister to a clade containing only reptile sequences (UFBOOT = 89%) (**Figure 6A** and **Figure S10**). This suggests that orthologs of amniote intronless IFN1s are present in amphibians and arose in the ancestor of tetrapods. Within this intronless tetrapod clade, two additional ancient reptile lineages are also present, one of which forms the sister group to all mammalian IFN1s (**Figure 6A** and **Figure S10**), while the other forms the sister to all other intronless IFN1s (i.e., both former reptile clades, and their mammalian and amphibian intronless counterparts) (UFBOOT ≥ 74%) (**Figure 6A** and **Figure S10**). This is consistent with a birth-death model of evolution, where reptiles have retained genes from three ancient intronless lineages that were present in the ancestor of tetrapods, but with amphibians and mammals retaining only one of these each, before the onset of independent lineage-specific diversifications. Intriguingly, an amphibian interferon containing a single intron falls sister to the group of cartilaginous fish IFN1s that were sister to the intronless amniote IFN1s in the CHOM analysis (UFBOOT = 84%) and together they fall sister to the intronless tetrapod IFN1s (UFBOOT = 47%). If accurate, this suggests that these cartilaginous fish genes are paralogous rather than orthologous to mammalian IFN1s, as both clades contain amphibians, further increasing the number of IFN1s likely present in the jawed vertebrate ancestor (**Figure 6A** and **Figure S10**).

lone *Anguilla japonica* sequence is truncated, perhaps explaining its absence from the clade. (B) Maximum likelihood consensus tree, under JTT+τ , of the amphibian

## No One-to-One Orthology Relationships Between Mammal and Reptile IFN1s

sister group of all other IFN1s showing two independent origins of intronless IFN1s within this clade.

It has long been recognized that the IFN-α and IFN-β genes of human and chicken are not orthologous (29). In contrast, the recently discovered chicken IFN-κ is purportedly an ortholog of mammalian IFN-κ (31). Interestingly, our CHOM and EXT IFN1 datasets, which greatly expanded taxon sampling in reptiles, failed to find evidence for orthology between IFN-κ genes of mammals and reptiles, but did not include the lineage containing chicken IFN-α because this was compositionally biased (**Figures S5, S10**; **Table S3**). Similarly, a lone amphibian sequence containing a single intron grouped together with the cartilaginous fish sequences that fall sister to the tetrapod intronless interferon clade. As this sequence would, more parsimoniously, be expected to group with the intronless IFN1s we performed more focused phylogenetic analyses to examine this finding. Our analyses included the cartilaginous fish and amphibian sequences that fell sister to this group in the EXT analysis (**Figure 6A** and **Figure S10**), but not more distantly related IFN1s to avoid biases introduced by distant outgroups. We also reinstated sequences, including chicken IFN-α, that were excluded from CHOM and EXT due to compositional bias. Interestingly, in this instance the amphibian sequence sister to cartilaginous fish in CHOM and EXT grouped with the intronless IFN1s of other amphibians (UFBOOT = 68%), away from the cartilaginous fish sequences (UFBOOT = 100%). This, far more parsimonious scenario, verifies the cartilaginous fish sequences as orthologs of the intronless tetrapod IFNs (**Figure 7A** and **Figure S12**). No evidence for orthology between any mammalian and reptile IFN1s was observed in this analysis. If rooted with the cartilaginous fish sequences, the results are also consistent with reptile genomes harboring ancient tetrapod intronless interferon lineages lost in mammals (**Figure 7A** and **Figure S12**). Finally, an unrooted analysis (i.e., excluding cartilaginous fish and amphibian sequences) recovered independent mammal and reptile clans, further supporting the lack of orthology between and reptile and mammalian IFNs (**Figure 7B** and **Figure S12**).

### Group 1, but Not Group 2, Ray-Finned Fish IFN1s Are Monophyletic

Our IFN1 phylogenies consistently showed that IFN-f is not a member of the ray-finned fish-specific IFN1s (**Figures 5A**, **6A**; **Figures S5, S10**). This suggests that IFN1 classification based on conserved cysteine pairs may not have a phylogenetic basis. For example, group 2 IFNs (IFN-b, IFN-c, and IFN-f) do not form a clade despite all having two conserved cysteine pairs in the mature peptide. To better explore this, we performed a focused phylogenetic analysis (**Figure 8** and **Figure S13**) of the remaining ray-finned fish-specific IFN1s that formed a clade in our CHOM and EXT analyses, using IFN-f as an outgroup. This placed the root between the remaining group 2 and group 1 members, in agreement with past hypotheses of fish IFN1 evolution, except for IFN-f (UFBOOT ≥ 77%) (**Figure 8** and **Figure S13**). The group 2 members IFN-b and IFN-c, fell sister to each other (UFBOOT = 95%), while within group 1, IFN-a and IFN-h form a sister group (UFBOOT = 59%), with IFN-d (UFBOOT = 45%) and IFN-e (UFBOOT = 77%) forming successive sister groups (**Figure 8** and **Figure S13**). Thus, our phylogenetic analyses reject

the monophyly of group 2 (two pairs of conserved cysteines), due to the independent origins of IFN-f, but not of group 1 (one pair of conserved cysteines) ray-finned fish IFN1s.

### DISCUSSION

The origins and evolutionary relationships between, and within, interferon subtypes have proven difficult to resolve. Here, with greatly increased taxon sampling and careful application of alignment and phylogenetic methodology, we overhaul our current understanding of the origins and relationships of the three IFN classes. Our findings also provide a significant step forward compared to previous work in understanding the mode and tempo of intra-class IFN evolution.

A notable study finding was our identification of a cartilaginous fish IFN3 gene, revealing that both IFN3 ligands and receptors existed in the jawed vertebrate ancestor, helping to resolve the deep relationships within the class II α-helical cytokines. We found that the four major lineages of this gene superfamily (i.e., IFN1, IFN2, IFN3, and the IL-10 family) diverged by multiple gene duplications [or genome duplication (15)] in quick succession in the ancestor of jawed vertebrates. We also revealed that the antiviral interferons, IFN1 and IFN3, are likely sister groups, with IFN2 being sister to the IL-10 family, similar to the model proposed by Siupka et al. (15). These results

complete IFN1 phylogenies (Figures 5A, 6A; Figures S5, S10), and all other lineages, and IFN-f has been reassigned to group 3 for this reason.

reject both of the other proposed hypotheses of IFN3 origins; (i) that tetrapod IFN3 genes evolved from IFN1s (8, 10, 23) and (ii) that IFN3 is a member of the IL-10 family (which is based on structural homology) (14, 21, 22). Structural similarity between IFN3 and the IL-10 family can be explained if these features were ancestral within the class II α-helical cytokines and secondarily lost in the IFN1 and IFN2 lineages. Importantly, unraveling the early evolution of class II α-helical cytokines also allowed us to objectively choose the best outgroups to test ingroup relationships for each of the IFN classes for the first time. This, along with other improvements in phylogenetic approach, made it possible for us to resolve some of the discrepancies noted in previous studies.

Our findings corroborate the conserved nature of IFN2 genes (which are not predominantly antiviral interferons) compared to IFN1 and IFN3 (24, 27, 104). By incorporating the closest outgroup, including cartilaginous fish IFN-γ (18), and better accounting for insertions and deletions at the alignment stage (72), we found strong support for teleost-specific origins of IFNγ-rel by tandem duplication as proposed previously (19). Thus we can now reject the possibility that this represents an ancestral jawed vertebrate gene that was lost in other groups (24–26). Applying a similar approach to IFN3 evolution, we were able to delineate the evolution of the major IFN3 gene lineages found in humans for the first time; with the IL-28/29 ancestor diverging from IFN-λ4 in the amniote ancestor, and the IL-28 and IL-29 lineages splitting in the ancestor of placental mammals.

Our results confirm that inferring the evolutionary relationships between IFN1 family members is difficult. IFN1 phylogeny is highly sensitive to several confounding factors, including model inadequacy, distant outgroups, and limited taxon sampling. The short length and rapid evolution of IFN1s may also have driven stochastic errors and resulted in some weakly supported branches in our phylogenetic trees. Importantly, we observed consistency in our analyses that were designed to minimize systematic error (i.e., applying best-fit models and outgroups, and exclusion of compositionally biased sequences), both of which are factors that may be indicative of accuracy, even in the face of weak support (105, 106). By accounting for phylogenetic error, and considering consistency across our datasets, we reconstructed a strongly supported scenario of IFN1 evolution where several IFN1 genes existed in the jawed vertebrate ancestor. These genes subsequently underwent extensive lineage-specific gene duplication and loss events. Central to this finding is our unprecedented taxon sampling, which allowed us to identify ancestral jawed vertebrate genes that have become very taxonomically confined due to multiple loss events. Our data imply that while IFN1s often undergo lineage-specific expansions, they can also be lost many times in parallel, generating extreme cases of "elusive" genes (i.e., genes which are difficult to detect because of recurrent loss or biases in generating assembled genomes) (107) and hidden paralogy (i.e., where differential loss results in paralogs presenting as orthologs) (108, 109). A key example of this is the discovery of intron-containing cartilaginous fish orthologs of intronless tetrapod IFN1s, which revealed that intron-containing IFN1s of ray-finned fishes are paralogous, rather than orthologous, to the intronless tetrapod IFN1s. This means that the retrotransposition event giving rise to intronless tetrapod IFN1s may have occurred as early as in the ancestor of bony fishes (indicating loss of intronless IFN1s from ray-finned fishes and coelacanth), or as late as in the most recent common ancestor of extant tetrapods (indicating loss of intron-containing IFN1s from ray-finned fishes and coelacanth). Either way, this lineage, which is remarkably expanded in amniotes, has been lost from teleosts and coelacanth. Together these findings imply that IFN1 molecules, like some other immune genes, evolve via a rapid birth-death evolutionary process, and have done so at least since the jawed vertebrate ancestor (100, 101, 110). This is consistent with a scenario where IFN1 genes have maintained their antiviral function for over 450 million years by evolving rapidly, in terms of both substitutions and gene gain and loss, due to the host pathogen arms race with viruses.

Analyses focused on the evolution of ray-finned fish IFN1s revealed that their group 1 (one conserved cysteine pair), but not group 2 (two conserved cysteine pairs), interferons are monophyletic. Our findings suggest that group 2 should be split into two groups. The first consisting of IFN-b and IFNc (together these form the sister group to group 1), for which we suggest the group 2 name be retained. And the second, consisting only of ray-finned fish IFN-f (although IFN-f appears to be present in at least amphibians and cartilaginous fishes also), which we propose be referred to as group 3. Interestingly, group 1 and group 2 IFN1s use different interferon receptors in zebrafish (11, 111), however zebrafish lack IFN-f, and as such it may be that IFN-f (now group 3) may have a different receptor to both group 1 and group 2. If this proved to be the case, analyses of receptor use may also help verify the assignment of amphibian and cartilaginous fish IFN-f. Importantly, although ray-finned fish group 1 and group 2 IFN1s are sister to each other, and seem to be derived from an ancestral jawed vertebrate IFN1 that has been lost in all other species, our results suggest that the ancestor of both groups possessed two conserved cysteine pairs. Based on the presence of the two conserved cysteine pairs across the IFN1 CHOM and EXT trees, our results are also consistent with the ancestral IFN1 possessing two disulphide bridges and four introns (9, 19, 32).

Similarly focusing on amniote IFN1 evolution we found that several intronless IFN1 genes existed in the tetrapod ancestor, with extensive IFN1 repertoires present in extant reptiles. In fact, as more ancestral tetrapod IFN1s appear to have been retained in reptiles, they evidently have even greater IFN1 diversity than mammals. Our analyses incorporated a greater breadth of mammals and reptiles than previous studies, including aquatic and/or semi-aquatic lineages, and had a more appropriate outgroup, but do not support one-to-one orthology of any mammalian or reptile IFN1s. This confirms non-orthology between human and chicken IFN-α and IFN-β (29, 112), while rejecting orthology of chicken and mammal IFN-κ (31).

Emergence of intronless interferons is more common in the IFN1 and IFN3 families than previously thought, consistent with intronless interferons bestowing an evolutionary advantage over those harboring introns (39). Our results suggest that both of the models (19, 23, 35, 40) put forth previously for the origins of amphibian intronless IFN1s are correct, with some emerging multiple times independently within amphibians, and others resulting from the same event that gave rise to amniote IFN1s. Strikingly, we also found that intronless amphibian IFN3s have emerged at least twice and independently from those of mammals on both occasions. Interestingly, amphibians also possess by far the most diverse set of IFN1s, including those which form part of the intronless tetrapod IFN1 group, the IFN-f group, and those in the sister group to all other IFN1s. Given this highly diverse repertoire of antiviral IFN1s and propensity for retrotransposition (or at least gross loss of introns), it is tempting to speculate a link to their morphology (e.g., permeable skin involved in terrestrial cutaneous respiration) or developmental life-history (e.g., aquatic tadpoles undergo metamorphosis to become terrestrial adults), especially as unique interferon responses have been observed between their distinctive stages of life (112, 113).

Lastly, our study indicates that a new nomenclature system is required to describe IFN1s to avoid relying on awkward (as applied here) or inaccurate descriptions. We have not attempted to formulate one here, as it is likely to be a substantial undertaking and will require input and agreement from several parties.

### DATA AVAILABILITY

All datasets generated and analyzed for the study are included in the manuscript and the **Supplementary Files**.

### AUTHOR CONTRIBUTIONS

AR, JZ, and HD conceived the study. AR performed sequence similarity searches, designed and performed phylogenetic analyses, and drafted the manuscript and figures. JZ and HD performed IFN1 searches for cartilaginous fishes. All authors contributed to and approved study design and the final manuscript.

### ACKNOWLEDGMENTS

PhyloBayes analyses were performed using the University of Aberdeen's Maxwell high performance computing cluster. AR was supported by a University of Aberdeen Center for Genome-Enabled Biology and Medicine Ph.D. studentship. DM received support from BBSRC Institutional Strategic Programme funding

### REFERENCES


(grant number: BBS/E/D/20002172). Silhouettes in **Figure 7** were obtained from http://phylopic.org; all of which are public domain except for the anole lizard silhouette, which was created by Ghedo and T. Michael Keesey (license: https://creativecommons.org/ licenses/by-sa/3.0/).

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2019.01558/full#supplementary-material

structure in vertebrates. J Virol. (2011) 85:8181–7. doi: 10.1128/jvi.0 0521-11


anole lizard (Anolis carolinensis). Dev Comp Immunol. (2016) 61:208–24. doi: 10.1016/j.dci.2016.04.004


over Myriochelata using mitogenomics. Mol Phylogenet Evol. (2008) 48:103– 11. doi: 10.1016/j.ympev.2008.03.033


liver and brain reveals molecular level conservation of vertebrate pancreas function. BMC Genomics. (2014) 15:1074. doi: 10.1186/1471-2164-15-1074


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

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

# Effects of Cell Differentiation on the Phagocytic Activities of IgM<sup>+</sup> B Cells in a Teleost Fish

Liting Wu<sup>1</sup> , Linghe Kong<sup>1</sup> , Yanjian Yang<sup>1</sup> , Xia Bian<sup>1</sup> , Siwei Wu<sup>1</sup> , Bingxi Li <sup>1</sup> , Xiaoxue Yin<sup>1</sup> , Liangliang Mu<sup>1</sup> , Jun Li <sup>2</sup> and Jianmin Ye<sup>1</sup> \*

*<sup>1</sup> Guangdong Provincial Key Laboratory for Healthy and Safe Aquaculture, Institute of Modern Aquaculture Science and Engineering, School of Life Sciences, South China Normal University, Guangzhou, China, <sup>2</sup> School of Biological Sciences, Lake Superior State University, Sault Ste. Marie, MI, United States*

Teleost B cells have phagocytic activities for ingesting particulate antigens, such as bacteria, in addition to the functional secretion of immunoglobulins (Igs). In the present study, the phagocytic activities of IgM<sup>+</sup> B cells under various differentiational conditions residing in peripheral blood leukocytes were investigated in a teleost fish Nile tilapia (*Oreochromis niloticus*). The IgM<sup>+</sup> B cells were recognized as IgMlo or IgMhi subsets based on their membrane IgM (mIgM) levels. The mIgM, secreted IgM (sIgM), major histocompatibility complex class II and reactive oxygen species were detected. Expressions of transcription factors (Pax5 and Blimp-1) and B cell signaling molecules (CD79a, CD79b, BLNK, and LYN) suggested that IgMlo B cells were resembling as plasma-like cells and IgMhi resembling as naïve/mature B cells, respectively. Analysis of phagocytic activities demonstrated that both IgMlo and IgMhi B cells have a similar phagocytic ability (phagocytosis percentage); however, the phagocytic capacity [phagocytic index and the mean fluorescence intensity (MFI)] of IgMhi B cells was significantly higher than that of IgMlo B cells. Taken together, the results indicated that B cell differentiation may cause the decrease of phagocytic capacity but not phagocytic ability of phagocytic IgM<sup>+</sup> B cells in teleost. The finding may provide an evolutionary evidence for understanding the greater specialization of the B cell in more sophisticated adaptive humoral immunity, by decreasing phagocytic activity in order to contribute its function more specifically into antibody-secreting.

#### Edited by:

*Stephanie DeWitte-Orr, Wilfrid Laurier University, Canada*

### Reviewed by:

*Tamiru Alkie, Wilfrid Laurier University, Canada Yong-An Zhang, Huazhong Agricultural University, China*

> \*Correspondence: *Jianmin Ye jmye@m.scnu.edu.cn*

#### Specialty section:

*This article was submitted to Comparative Immunology, a section of the journal Frontiers in Immunology*

Received: *04 December 2018* Accepted: *02 September 2019* Published: *19 September 2019*

#### Citation:

*Wu L, Kong L, Yang Y, Bian X, Wu S, Li B, Yin X, Mu L, Li J and Ye J (2019) Effects of Cell Differentiation on the Phagocytic Activities of IgM*<sup>+</sup> *B Cells in a Teleost Fish. Front. Immunol. 10:2225. doi: 10.3389/fimmu.2019.02225* Keywords: IgM<sup>+</sup> B cells, Oreochromis niloticus, phagocytosis, phagocytic capacity, phagocytic ability

### INTRODUCTION

Phagocytosis is an important defense against pathogen infection in innate immunity, which is a bridge linking innate immunity and adaptive immunity as well (1). Its characteristics mainly include the large particulates endocytosis (diameter >0.5µm), actin polymerization and remodeling, phagolysosome formation (2). Macrophages, monocytes, granulocytes, and dendritic cells are generally considered as professional phagocytes. B cells are antibody (Ab)-producing cells, which can produce immunoglobulins (Igs) to specifically neutralize certain antigens, seemed unable to phagocytose antigenic particles in mammals (3). However, teleost fish B cells, as well as B cells of amphibians (Xenopus laevis), were firstly demonstrated to be phagocytic and microbicidal, and their phagocytic function has been characterized similarly as that of professional phagocytes (4). Thereafter, the function of B cell phagocytic ability in teleost has also been identified in many fish species, such as rainbow trout (Oncorhynchus mykiss), lumpfish (Cyclopterus lumpus L.), Atlantic cod (Gadus morhua L.), halfsmooth tongue sole (Cynoglossus semilaevis), channel catfish (Ictalurus punctatus), and turbot (Scophthalmus maximus) (5–9). The study of phagocytosis in teleost fish B cells might facilitate the exploration of new immune therapy to infection diseases in aquaculture.

In mammals, pre-B cell or B cell lines have been demonstrated to be able to differentiate into macrophages (10, 11). Moreover, the phagocytic activities of peritoneal B1 cells rather than B2-B cells were also identified in mice (12). It has been reported that mouse CD5<sup>+</sup> B cells (named as B1 cells) can differentiate into cells with classical macrophage morphology (13), and the B1 cells have also shown potential innate immunity with production of over 80% of natural serum antibodies (14, 15). B lymphocytes in teleost fish have been considered as functional equivalent to mammalian B1 cells (16, 17), and the recent identification of a homolog of CD5 in rainbow trout provides further solid evidence that the phenotypes and functions of IgM<sup>+</sup> B cells are similar to mammalian B1 cells (18). Unlike mammalian B1 cells with low affinity and wide reactivity to antigens, teleost B cells differentiate into Ab-secreting cells in response to antigenic stimulation (19). The B cell subsets, differentiate from naïve B cells (expression of membrane-bound Ab) into plasmablasts and then, terminally differentiate into plasma cells (including short-lived plasma cells and long-lived plasma cells) with a stronger Ab-secreting ability and capacity (19–21).

So far, four subsets of B cells have been identified in teleost fish, named IgM+/IgD+, IgM−/IgD+, IgM+/IgD−, and IgT<sup>+</sup> B cells (5, 7, 22–24). Among them, IgM<sup>+</sup> B cells are the predominant B cells, with several folds higher mRNA transcriptions of IgM isotype in the peripheral blood leukocytes (PBLs) of teleost fish like salmon and rainbow trout (25, 26), which are specialized in systemic immunity (9, 27). Both IgM<sup>+</sup> and IgT<sup>+</sup> B cells have been found to be phagocytic (4, 22); however, the phagocytic activity and other related roles of B cells under various differentiational conditions (including naïve/mature B cells, activated B cells, plasmablasts, and plasma cells) are poorly understood.

Although B cell developmental pathways in teleost fishes are poorly understood as many essential molecular markers are not yet available; however, some very conservative mammalian transcription factors, like paired box-5 (Pax5) and B lymphocyteinduced protein-1 (Blimp-1), have been studied in rainbow trout (28–30). Pax5 is expressed from the pre-B cell through mature B cells, downregulated during terminal differentiation and absent at the plasma cells in rainbow trout (28). In contrary to Pax5, Blimp-1 is a master regulator of cell differentiation including both terminal B cell and macrophage differentiation (29). Blimp-1 shifts Ig expression from the membrane to the secreted form, leading to increases in secreted Ig (sIg) expression in activated B cells with simultaneous reduction of mIg (29). A recent study demonstrated that the more differentiation of teleost B cells, the less expression membrane IgM (mIgM), and the related B cells, with low or high mIgM expression levels, could be divided into IgMlo or IgMhi subtypes, respectively (30). What's more, the transcription levels of various B cell signaling molecules changed in the process of B cell differentiation, such as CD79a, CD79b, BLNK, and LYN, which were down-regulated during B cell maturation (31, 32).

In this report, we aimed to gain better understanding of B cell phagocytosis in teleost under various differentiational status, in particular for both IgMhi and IgMlo B cell subsets, which are resembling as naïve/mature B cells and plasma-like B cells, respectively. For this purpose, the phagocytic efficiency of Nile tilapia (Oreochromis niloticus) IgM<sup>+</sup> B cells ingesting fluorescent microspheres and killed pathogen Streptococcus agalactiae (S. agalactiae) were elucidated through flow cytometric analysis, respectively. Our results suggested that only the phagocytic capacity, but not phagocytic ability was significantly affected by the differentiation condition of teleost B cells. This finding shed new lights on the avenue to better understand the functional roles of B cells in the evolutionary process from teleost fish to mammalian species.

### MATERIALS AND METHODS

### Fish

Healthy Nile tilapias (Oreochromis niloticus) with mean weight of 750 ± 50 g were obtained from Guangdong Tilapia Breeding farm (Guangdong, China). All fish were maintained in 300 L tanks in the laboratory with recirculating pre-treated (biologically filtered, dechlorinated, chemically balanced, and UV-treated) fresh water. Water temperature was maintained at 28 ± 2 ◦C, and photoperiod was adjusted to match seasonal change (33, 34). All fish experimental procedures were reviewed and the ethics were approved by the University Animal Care and Use Committee of the South China Normal University.

### Isolation of Peripheral Blood Leukocytes

The total peripheral blood leukocytes were prepared according to the previous procedure with some modifications (20, 28). Nile tilapia were anesthetized in water containing ∼0.04% MS-222 (Aladdin, China). Blood samples were collected by venipuncture from the caudal vein with a heparinized syringe [250 µL heparin (0.1 g/L) in 2.5 mL syringe]. Each fish was collected 5 mL blood and then placed in sterile 10 mL tube. After centrifuged at 500 × g for 15 min, the plasma was removed. The blood cells were resuspended gently in four times RPMI-1640 medium (Gibco, USA), supplemented with 100 I.U./mL penicillin G, 100µg/mL streptomycin, 10 units/ mL heparin and 5% fetal bovine serum (FBS) (Gibco, USA), and then put on ice. The diluted blood cells were then layered upon an equal volume of Histopaque 1077 (Sigma, USA) in 50 mL conical centrifuge tubes slowly, 500 × g centrifuged for 40 min at 4◦C. Leukocytes were collected from the interface layer, and then washed three times with RPMI-1640 by centrifugation. Cell viability was determined by 0.4% (Sigma, USA) trypan blue staining, and finally peripheral blood leukocytes (PBLs) were resuspended to a concentration of 1 × 10<sup>7</sup> cells/mL in RPMI-1640 containing 10% FBS.

### Immunofluorescence Staining of PBLs

For immunofluorescence staining, the PBLs were suspended with phosphate buffer saline (PBS, pH = 7.4) and then

TABLE 1 | Primes used for qPCR in this study.

incubated with mouse anti-IgM monoclonal antibody (mAb, IgG1 type from Balb/c mice) labeled by Alexa Fluor 647 (AF647) (35, 36) for 1 h at room temperature (RT). After washing three times with PBS, the cells were incubated with 1µg/mL of DAPI (Sigma, USA) for 10 min. Then the cells were washed again with PBS and subjected to microscopy observation (Zeiss, Germany). As a negative control, an isotype mouse IgG was also applied for the above staining procedure.

### Flow Cytometry (FACS)

The isolated PBLs were incubated with AF647-labeled mouse anti-Nile tilapia IgM mAb (1 mg/mL, 1:2000 dilution) at RT for 1 h (35, 36). After washing with PBS, cells were resuspended in RPMI-1640 contained 5% FBS and subjected to FACS analysis with a BD Arial III flow cytometer (BD, USA) and 50,000 cells were recorded in each sample. PBLs incubated without any antibody or with a normal isotype mouse IgG (Thermo, USA) were also used as blank or negative controls. Further data analysis was performed using FlowJo X.

### Cell Sorting

PBLs were incubated with mouse anti-Nile tilapia IgM mAb (35, 36) as described above and only the gated lymphocytelike cells were selected for sorting in a BD FACS Aria III flow cytometer based on the low forward scatter (FSC) and sideward scatter (SSC) profiles (to exclude the granulocytes). According to the different fluorescence intensity, IgM−, IgMhi , IgMlo, and total IgM<sup>+</sup> B cells were collected. The purity of various sorted cell populations was analyzed (**Figure 2A**). The sorted cells showing a higher purity level (>95%) were collected in Trizol reagent (Vazyme, China) and immediately frozen by liquid nitrogen, and then stored at −80◦C for further isolation of total RNAs.

### Gene Expression Analysis

Total RNA was extracted using Trizol reagent kit (Vazyme, China) according to the manufacture's instruction, and their quality and quantity was determined by Nanodrop 2000 assay (Thermo, USA). The cDNAs were synthesized from the purified RNA and then diluted 10-fold, and stored at −80◦C for further quantitative real time PCR analysis (qPCR). For characterization of various B cell subsets, the transcription levels of membrane IgM (mIgM), secreted IgM (sIgM), major histocompatibility complex class IIβ (MHC IIβ) (37), transcription factors (Pax5 and Blimp-1), and B cell signaling molecules (CD79a, CD79b, BLNK, and LYN) were investigated using the 7500 Real Time PCR System (Applied Biosystem, USA) with the SYBR green dye method in a total of 20 µL volume containing 10 µL of 2 × SYBR mix (Yeasen, China), 2 µL forward primer and 2 µL reverse primer, 3 µL of diluted cDNA, 3 µL double distilled H2O. The β-actin (Accession No. KJ126772.1) gene was used as internal control with primers showed in **Table 1**. Gene-specific primers are listed in **Table 1**. The qPCR was carried out with the following program: 95◦C


*F, forward; R: reverse.*

for 3 min, followed by 40 cycles of 95◦C for 15 s, 60◦C for 1 min.

### The Comparation of sIgM Secreted From IgMhi and IgMlo Cells at Protein Level

In order to compare the antibody secreting abilities of IgMhi and IgMlo B cells at protein level, a previous published ELISA was explored (36). Briefly, IgMhi and IgMlo B cells (1 × 10<sup>6</sup> cells per sample) were sorted (as the description of "Cell Sorting") and cultured in RPMI-1640 medium (Gibco, USA), supplemented with 100 I.U./mL penicillin G, 100µg/mL streptomycin, and 10% FBS (Gibco, USA) for 24 h. The culture supernatants were collected and sorted in −20◦C. Microtiter plate (Corning, USA) was coated with 2µg/mL mouse anti-tilapia IgM mAb (100 µL) at RT for 1 h. Then, the plate was washed three times with 1 × TTBS (contained 0.1% Tween 20), and blocked with 200 µL 0.5% BSA-TTBS for 1 h at RT. Plate was then washed three times with 1 × TTBS. Cell supernatant was added to well, 100 µL per well and incubated for 1 h at RT. Followed by three times washes (1 × TTBS), 100 µL of biotinylated mouse antitilapia Ig mAb (0.25µg/mL) (36) in blocking buffer were then added, and incubated for 1 h at RT. After three times washes, 100 µL of streptavidin-HRP (0.5µg/mL) (Southern Biotech, USA) was added, and incubated for 1 h at RT. After washed three times, 2,2′ -Azinobis-(3-ethylbenzthiazoline-6-sulphonate) (ABTS) (Sigma, USA) of 100 µL was added for 10 minuses, and the optical density (O.D.) was measured using a microplate reader (Thermo, USA) at 405 nm. The ratio of the O.D. value of IgMlo sample to IgMhi sample were taken as the relative expression of sIgM at protein level.

### Reactive Oxygen Species (ROS) Measurement

ROS levels of IgMhi and IgMlo cells were determined using ROS assay kit (Beyotime, China). Cells (1 × 10<sup>6</sup> ) were obtained as the description of "Isolation of Peripheral Blood Leukocytes," were treated with 10µM DCFH-DA dissolved in PBS (1 mL) at 25◦C for 1 h. The control sample was the same amount cells without fluorescent dyes. The fluorescence intensity was monitored with excitation wavelength at 488 nm and emission wavelength at 530 nm (38). The mean fluorescence intensity (MFI) was analyzed using FlowJo X.

### Transmission Electron Microscopy (TEM)

FACS sorted cell fractions (IgM−, IgMlo, and IgMhi cells) were fixed in 2.5% glutaraldehyde in PBS (pH 7.4) and prepared for TEM analysis according to a previous method (7). Cell images were observed and recorded by using the transmission electron microscope Tecnai (FEI, USA).

### Assessment of Phagocytosis

A total of 12 individual fish (750 ± 50 g) were used to investigate the phagocytic activity of IgM<sup>+</sup> lymphocytes in Nile tilapia. PBLs were adjusted to 1 × 10<sup>7</sup> cells/mL using RPMI-1640 medium supplemented with 5% FBS. PBLs were incubated with 0.5 and 1µm Fluoresbrite <sup>R</sup> YG carboxylate microspheres (YG beads) (Polysciences Inc., USA) at a 1:20 (cells: beads) ratio for 4 h at 25◦C, respectively (4, 9). For phagocytosis of bacteria, the predominant pathogens S. agalactiae was used here. The inoculation, bacterial counting, inactivation and fluorescein isothiocyanate (FITC; Sigma, USA) labeled modes of S. agalactiae were performed as described by our previous reports (34, 39). The ratio of cells vs. bacteria for phagocytosis was 1:20 for 4 h at 25◦C as well. After incubation, the cells were collected and centrifuged at 100 × g for 10 min at 4◦C to remove excess beads. Then the cells were resuspended in 1 mL PBS containing 5% FBS, and incubated with anti-IgM mAb labeled with AF 647 (1 mg/mL, 1:2000 dilution) as described above (35). After three times washes with PBS, the phagocytic activities of PBLs from 14 fish were independently analyzed by using BD Arial III flow cytometer (BD, USA). PBLs incubated without any antibody or with a normal isotype mouse IgG (Thermo, USA) were also included as blank or negative controls. Phagocytic activities of IgM<sup>+</sup> cells were expressed as phagocytic ability (% of total phagocytic cells that ingested one or more beads) and phagocytic capacity (the proportion of phagocytic cells that had ingested one, two or three or more beads, respectively), as well as the MFI (6, 7, 40). Data analyses were performed using FlowJo X.

### Statistical Analysis

Statistical analysis was carried out by using SPSS 17.0 software (SPSS, USA). Data were analyzed with analysis of variance (ANOVA) followed by two-tailed Student's t-test when the ANOVA indicated that the variances of both groups differed significantly. No significant difference (n.s.) means p > 0.05 and significant difference was defined as <sup>∗</sup>p < 0.05, ∗∗p < 0.01, and ∗∗∗p < 0.001.

### RESULTS

## IgM<sup>+</sup> B Cells in PBLs

In order to investigate the IgM<sup>+</sup> B cells in the PBLs of Nile tilapia, we firstly gated the lymphocyte-like cell population based on their lower FSC and SSC patterns (**Figure 1A**, upper left panel). Among the gated cell population, a total of 34.6% cells were detected as IgM<sup>+</sup> B cells, and the histogram also showed two distinct populations, which were designed as IgMhi (21.3%) and IgMlo B cells (13.3%), respectively (**Figure 1A**, upper right panels). To further identify these two subsets, we analyzed the cell size (according to the FSC parameter) difference between IgMlo and IgMhi cells, which exhibited a significant larger size in IgMlo than IgMhi cells (**Figure 1A**, lower panels). Under the fluorescence microscope, an obvious variety of red fluorescence intensity colocalized around the surface of IgM<sup>+</sup> B cells were observed which represented the negative, low or high expressions of IgM on the surface of lymphocytes (**Figure 1B**). What's more, the transcription level of mIgM was detected in IgM−, IgMlo, and IgMhi cells as well, which expressed highest in IgMhi cells but lowest in IgM<sup>−</sup> cells (**Figure 1C**).

### Cell Sorting and the Ultrastructure of the Sorted Cells

For further characterization of these two different IgM<sup>+</sup> B subpopulations, IgMhi and IgMlo B cells, as well as IgM<sup>−</sup> cells were sorted out from PBLs according to the difference of fluorescence intensity (**Figure 2A**, left panel). The purity of sorted IgMhi and IgMlo cells was double checked and a high sorting effectiveness was validated (**Figure 2A**, right panel). In addition, the ultrastructure of the sorted cells was observed under the transmission electron microscope, both IgMlo and IgMhi cells showed typical lymphocyte features (e.g., large nucleus, thin cytoplasm), and contained a few cytoplasmic vacuoles, however, more abundant cytoplasmic vacuoles were observed in the IgM<sup>−</sup> cells (**Figure 2B**).

### Profiles of mIgM, sIgM, MHC IIβ, and ROS on IgM<sup>+</sup> B Cell Subpopulations

In order to identify the difference between these two subpopulations, the expressions of mIgM, sIgM, and MHC IIβ by qPCR with specific primers were analyzed (**Table 1**). In comparison to the IgMlo cells, the expression of mIgM in IgMhi B cells was significantly higher than that in the IgMlo subpopulation, and the ratio of IgMhi/IgMlo was 6.2 (**Figure 3A**). In contrast, the level of sIgM expressed more than three times higher in IgMlo B cell subpopulation than that in IgMhi B cells (**Figure 3B**), both at transcription level and protein level. The MHC IIβ displayed significantly higher on IgMhi B cells than that on IgMlo B cells (**Figure 3C**). DCFH-DA-dependent analysis showed that intracellular ROS levels were significant higher in IgMhi than IgMlo B cells (**Figure 3D**).

### Expressions of Transcription Factors and B Cell Signaling Molecules

To further distinguish IgMhi and IgMlo B cell populations, we analyzed the expression profiles of two key B cell transcription

factors, Pax5 and Blimp-1, which are related to plasma cell differentiation, as well as a couple of B cell signaling molecules, like CD79a, CD79b, BLNK, and LYN with specific primers (**Table 1**). The expression profiles for these molecules exhibited significant difference between IgMhi B cells and IgMlo B cells (**Figures 4**, **5**). Transcription of Pax5 was detected significantly high in IgMhi B cells, but low in IgMlo or PBLs (**Figure 4A**); whereas, the expression level of Blimp-1 was significantly high in IgMlo subpopulation than that in IgMhi subpopulation (**Figure 4B**). The expression profiles of B cell signaling molecules (CD79a, CD79b, BLNK, and LYN) showed similar patterns in these two subsets, where they all expressed higher levels in IgMhi

B cells and lower in IgMlo subset (**Figure 5**). Therefore, the IgMhi cells and IgMlo cells may represent naïve/mature B cells and plasma-like B cells, respectively.

### Phagocytic Activity of IgMhi and IgMlo B Cells

To evaluate the phagocytic activity of IgMhi and IgMlo B cells, the PBLs were incubated with 0.5/1µm YG fluorescent beads and killed S. agalactiae, respectively. Then we analyzed the IgM<sup>+</sup> cells by FACS. The scatter plot and the histogram of IgM<sup>+</sup> cell (IgMhi and IgMlo) phagocytized fluorescent microspheres or killed pathogen were shown in **Figure 6A**. When calculated the phagocytic percentage (phagocytic ability) in IgMlo and IgMhi B cells, it performed that the mean percentage of phagocytosis of both IgMlo and IgMhi B cells exhibited similar phagocytic ability to ingest 0.5/1µm YG fluorescent beads and killed S. agalactiae (**Figure 6B**). For further examination of the phagocytic capacity of IgMhi and IgMlo, the MFI of FITC in phagocytic IgMhi and IgMlo subsets were evaluated. As shown in **Figure 6C**, in comparison to IgMlo B cells, IgMhi B cells showed significantly higher MFI. It was clear that there existed different peak of FITC intensity in the histogram of phagocytic IgM<sup>+</sup> B cells when ingested 0.5µm as shown in the **Figure 6A**. The peaks were divided into 1, 2, and 3 of FITC intensity in the histogram of phagocytic IgM<sup>+</sup> B cells, which represented the IgM<sup>+</sup> cells ingesting one bead, two beads, and three or more beads, respectively (**Figure 7**, left panel). The results indicated that the percentage of IgMhi B cells ingesting two or three and more beads were significantly higher than that of IgMlo B cells, but the percentage of IgMlo cells ingested one bead was higher than IgMhi cells (**Figure 7**, right panel).

## DISCUSSION

The analysis of teleost B cell subpopulation has been limited because of the lack of antibodies. Using established mAbs against mIgM to distinguish different B cell subsets based on the different fluorescence intensity (41). In this study, we presented the identification of two IgM<sup>+</sup> B cell subpopulations, IgMlo, and IgMhi, from PBLs in Nile tilapia based on mIgM expression on cell surface. The IgMhi and IgMlo cells were resembled to be naïve/mature B cell and plasma-like B cell, respectively, according to their expression levels of mIgM, sIgM, MHC IIβ, Pax5, Blimp-1, CD79a, CD79b, BLNK, and LYN molecules. The IgMlo and IgMhi B cells had no significant difference in their phagocytic ability, but the phagocytic capacity in IgMhi cells was significantly higher than that in IgMlo B cells. These results collectively indicated that B cell differentiation may cause the decrease of phagocytic

cells was set. The mean relative expressions of sIgM on IgMlo cells compared to IgMhi cells (at transcription level and protein level) were detected (B). The ROS (D) on IgMhi and IgMlo B cell subpopulations were detected and showed as the mean DCF fluorescence. Results were shown as mean ± SD (*n* = 12, three independent experiments and four individual fish per trial). Statistical differences were evaluated by one-way ANOVA followed by two-tailed Student's *t*-test. Statistically significant difference was defined as: \*means *p* < 0.05; \*\*\*means *p* < 0.001.

capacity but have no effect on their phagocytic ability in teleost fish.

In our present study, IgM<sup>+</sup> B cells were demonstrated as the most dominant leukocytes (about 35%) in the PBLs of Nile tilapia (**Figure 1A**, upper panel), which is comparable to other fish species such as rainbow trout (4, 27). What's more, the FACS analysis clearly revealed that there are two distinct IgM<sup>+</sup> B subpopulations, IgMlo and IgMhi B cells, which were also clearly visualized under the microscopy for the cells with different fluorescence intensity colocalized on their surface (**Figure 1B**). In mammals, the expressions of mIgM in differentiated B cell subsets were different, representing higher in naïve B cells but low in plasma cells, and such differentiation and maturation of mammalian B cells was coordinated by the transcription factors of Pax5 and Blimp-1 (42–44). Similar findings of B cell differentiation and their relation to the IgM maker and transcription factors have also been demonstrated in rainbow trout (28). We firstly isolated both IgMlo and IgMhi

B cells through cell sorting (**Figure 2A**) and analyzed by TEM (**Figure 2B**). The ultrastructural images characterized that IgMlo and IgMhi cells had higher nucleus to cytoplasm ratio than in IgM<sup>−</sup> cells, which were similar to the typical features of IgM<sup>+</sup> and IgM<sup>−</sup> cells in rainbow trout (4, 22). The IgM<sup>+</sup> (IgMhi and IgMlo) B cells were characterized by a large round nucleus, a thin cytoplasm, and a varying number of small dendrites extending from the cells. Significant difference existed between IgMhi and IgMlo B cells in size, which characterized as that the IgMlo was larger than the IgMhi B cell. It was consistent with the scatter plots results of the IgMhi and IgMlo cells in **Figure 1B**. Moreover, significant higher transcriptions of mIgM were only identified in IgMhi B cells, while sIgM only from IgMlo B cells which indicated the differentiation process of naïve/mature B cells (higher expression of mIgM) (**Figure 3A**) to IgM-secreting plasma cells (higher expression of sIgM) (**Figure 3B**). The high level of surface MHC IIβ on IgMhi B cells decreased to IgMlo B cells (**Figure 3C**), which indicated that the antigen presenting capacity decreased during the process of B cell differentiation. Plasma cells lose their ability to present antigen (45), and results in increased IgM secretion levels (41). A proprietary enzyme system of phagocytes in professional phagocytes is responsible for the production of ROS during respiratory bursts to kill invasive pathogenic microorganisms, and lower ROS level is determined in a mature phagocyte (46). ROS were generated during aerobic metabolism and played important roles as chemical mediators in normal cell growth, differentiation, programmed cell death and senescence (47). Higher level of ROS in IgMhi B cells than that in IgMlo B cells (**Figure 3D**) indicated that the function in IgMhi B cells may be different with IgMlo B cells. Furthermore, IgMhi B cells were demonstrated to express lower level of Blimp-1 and higher level of Pax5, while IgMlo B cells exhibited higher expression of Blimp-1 and lower expression of Pax5 (**Figure 4**). These results were in consistent with the findings of the differentiating status of B cells in mammals (48) and in rainbow trout (30), where IgMhi resembled as naïve B cell phenotype and IgMlo as Ab-secreting cell phenotype. The phenomenon of the larger size in IgMlo B cells than in IgMhi B cells (**Figure 1A**, lower panel; **Figure 2B**) was consistent with the studies in rainbow trout (41) and in mammals (49). It may imply that in teleost fish, plasma-like cells (IgMlo B cells) build up a large amount of the protein-synthesizing machinery than naïve/mature B cells (IgMhi B cells) as in mammal plasma cells (49). In addition, down-regulation of B cell signaling molecules found in different tilapia IgM<sup>+</sup> B cell subsets (**Figure 5**) was also in agreement with mammalian B cells (31, 50), which provides further evidence for the IgMhi and IgMlo B cells represented different differentiating stages of Nile tilapia B cells.

Phagocytic B cells have been identified in various teleost species, and their contribution to fish innate immunity against various pathogens has also been widely accepted (4–9). In responding to in vitro stimulations through TLRs (from TLR1 to TLR8), mammalian B1 cells would proliferate and differentiate into Ig-secreting cells (51, 52). Phagocytic fish B cells have been considered as equivalents to mammalian B1 cells because they shared more similarities including the phagocytic function (53); however, until now, it is poorly understood whether cell differentiation has any effect on phagocytic activity in the differentiated B cells in mammals (B1 cells), or in teleost. In the current study, we found that the phagocytic ability (uptake rate

in IgM<sup>+</sup> B cells) had no significant difference between IgMhi and IgMlo B cells (**Figure 6B**), but the phagocytic capacity in IgMhi was higher than that in IgMlo B cells (**Figures 6C**, **7**). Since IgMlo and IgMhi B cells were B cell subpopulations at different differentiating stages, the findings of the lower of phagocytic capacity in IgMlo than that in IgMhi B cells might indicate the decrease of B cell phagocytic capacity in the process of B cell differentiation in teleost. With the exploration of the phagocytic activity changing during B cell differentiation in teleost, we can better understand the roles of B cell against infection and in bridging innate and adaptive immunity.

The appearance of the phagocytic IgM<sup>+</sup> B cells in teleost supports the idea of an evolutionarily relationship between B cells and macrophages, in which B cells might have evolved from ancient phagocytic cells (1). However, the main activity of fish B cells is considered to produce high content of natural serum IgM molecules in unimmunized fish (54–56), especially when the B cells differentiate into plasma cells (19, 20). According to the current findings, we propose a model of change in B cell function during the B cell differentiation in teleost (**Figure 8**). We hypothesize that when the B cells differentiate into Absecreting cells (plasma cells), the phagocytic capacity decreases so as to contribute these B cells more specialized to provide efficient Abs in more sophisticated adaptive humoral immunity. In mammals, besides B1 B cells, there are other B cell subsets, including marginal zone B cells and follicular B2 cells. B2 B cells can produce Abs with high affinity and specificity to Tdependent antigens but are unable to be phagocytic (57). These B2 B cells would be able to differentiate into plasma cells under the stimulus of pokeweed mitogen. All these findings may indicate that the B cells, in the process of evolution, decrease other functions, such as phagocytosis, in order to specialize its

function for Ab-secreting. The decrease of phagocytic capacity in teleost B cell during B cell differentiation, from naïve B cells to terminal differentiating plasma cells (possessing strong Ab-secreting ability), might support evolutionary relationship of mammalian B1 B cells and B2 B cells, and provide more evidence for understanding the greater specialization of these B cells in more sophisticated adaptive humoral immunity in mammals.

### AUTHOR CONTRIBUTIONS

LW performed most of the experimental work. LK, YY, and SW assisted LW in the preparation of sample and cell sorting. XB and XY provided primers and performed all transcriptional analysis. BL and LM assisted data analysis and graphing. JL reviewed and polished the manuscript. JY and LW designed the experiments and wrote the main body of the paper.

### FUNDING

This study was supported by National Natural Science Foundation of China (31972818; 31528019; 31472302). JL was partially supported with the Pearl River Scholarship from Guangdong Province.

### REFERENCES


ability to differentiate into dendritic like cells. PLoS ONE. (2012) 7:e49260. doi: 10.1371/journal.pone.0049260


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

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

# Effects of Live Attenuated Vaccine and Wild Type Strains of *Edwardsiella ictaluri* on Phagocytosis, Bacterial Killing, and Survival of Catfish B Cells

Adef O. Kordon<sup>1</sup> , Safak Kalindamar <sup>2</sup> , Kara Majors <sup>1</sup> , Hossam Abdelhamed<sup>1</sup> , Wei Tan<sup>1</sup> , Attila Karsi <sup>1</sup> and Lesya M. Pinchuk <sup>1</sup> \*

*<sup>1</sup> Department of Basic Sciences, College of Veterinary Medicine, Mississippi State University, Mississippi State, MS, United States, <sup>2</sup> Department of Molecular Biology and Genetics, Faculty of Art and Sciences, Ordu University, Ordu, Turkey*

*Edwardsiella ictaluri,* a Gram-negative facultative intracellular pathogen, is the causative

#### *Edited by:*

*Tiehui Wang, University of Aberdeen, United Kingdom*

#### *Reviewed by:*

*Tamiru Alkie, Wilfrid Laurier University, Canada Leon Grayfer, George Washington University, United States*

> *\*Correspondence: Lesya M. Pinchuk pinchuk@cvm.msstate.edu*

#### *Specialty section:*

*This article was submitted to Comparative Immunology, a section of the journal Frontiers in Immunology*

*Received: 31 October 2018 Accepted: 23 September 2019 Published: 09 October 2019*

#### *Citation:*

*Kordon AO, Kalindamar S, Majors K, Abdelhamed H, Tan W, Karsi A and Pinchuk LM (2019) Effects of Live Attenuated Vaccine and Wild Type Strains of Edwardsiella ictaluri on Phagocytosis, Bacterial Killing, and Survival of Catfish B Cells. Front. Immunol. 10:2383. doi: 10.3389/fimmu.2019.02383* agent of enteric septicemia of catfish (ESC). The innate functions of B cells have been demonstrated in several teleost fish, including zebrafish, rainbow trout, and channel catfish. Recently, our group has developed several protective *E. ictaluri* live attenuated vaccines (LAVs). However, the innate role of catfish B cells to phagocytose and destroy *E. ictaluri* wild-type (WT) and live attenuated vaccine (LAV) strains has not been evaluated. In this study, we assessed the efficacy of *E. ictaluri* WT and two LAVs on phagocytosis, microbial killing, and survival of catfish anterior kidney (AK) B cells. Initially, we documented active uptake of *E. ictaluri* WT and two LAVs in B cells by flow cytometry and light microscopy. Then, we observed the *E. ictaluri* strains-induced phagosome and/or phagolysosome formation in the cytoplasm of catfish magnetically sorted IgM<sup>+</sup> B cells. Furthermore, we demonstrated that AK B cells were able to destroy the internalized *E. ictaluri* WT and LAV strains efficiently. Finally, we documented early and late apoptotic/necrotic manifestations induced by *E. ictaluri* in catfish AK B cells. In conclusion, our results suggest that both LAVs and WT strain initiate similar innate immune responses such as active phagocytic uptake, induced bactericidal activity as well as promote early and late apoptotic changes in catfish B cells. Our data suggest that phagocytic and microbicidal B cells may serve as professional APCs in initiation of protective adaptive immune responses against ESC in channel catfish.

Keywords: *Edwardsiella ictaluri,* live attenuated vaccines, catfish B cells, phagocytosis, bacterial killing, apoptosis

### INTRODUCTION

The primary function of B cells in the humoral branch of adaptive immunity is to secrete antibodies of increasing affinity and maintain an immunological memory (1). In recent years, it has been determined that B cells can be subdivided into different subsets with distinct morphology, phenotypes, and functional features and also contribute to innate immune responses. The recent discovery that live Salmonella typhimurium was engulfed by primary human B cells via B cell receptor (BCR)-dependent manner broke the long-held paradigm that B cells were unable to uptake large particulate antigens (2). Two subsets of mammalian B cells, marginal zone (MZ) and B-1 B cells, were classified as "innate B lymphocytes" based on their developmental, phenotypic, and functional characteristics contributing to innate immune responses, such as phagocytosis (3, 4). Importantly, phagocytic B cells from the peritoneal cavity were able to ingest bacteria, produce mature phagolysosomes, destroy the ingested bacteria and present the bacterial antigens to CD4<sup>+</sup> T cells (5).

The first evidence on B cell phagocytosis in rainbow trout was reported by Li et al. (6). Like mammalian B-1 cells, B cells in teleost fish were able to engulf particles and kill the internalized pathogens (6, 7). However, teleost B cells were present in all systemic compartments including blood, spleen, and anterior kidney (AK) and representing 60% of all B cells. In contrast, phagocytic B cells in mammals were mainly found in the peritoneal cavity and represented a 30–40% of total B cell numbers (5, 6, 8–10). The ability of B cells to uptake soluble, particulate and bacterial antigens by phagocytosis has been documented in zebrafish and Atlantic salmon (11, 12). Furthermore, the phagocyting B cells that possessed phagolysosomes were described in rainbow trout suggesting their essential role in bacterial killing (6, 13). Additionally, B cells in Atlantic cod had higher phagocytic capacity to uptake fluorescent beads compared to neutrophils (12). Also, contrary to other teleost fish, the large amount of phagocytic B cells has also been found in catfish blood (6, 14).

Edwardsiella ictaluri is a Gram-negative facultative intracellular fish pathogen that causes enteric septicemia of catfish (ESC), which is one of the most devastating diseases in the US catfish industry (15–18). A live E. ictaluri vaccine (Aquavac-ESC) against ESC was developed by Klesius and Shoemaker (19), and this vaccine protected juvenile catfish (19). Then, immersion studies demonstrated that E. ictaluri LAVs stimulated protective immunity in catfish fry, fingerlings, and eyed catfish eggs (20–23). Recently, a live attenuated E. ictaluri isolate (S97-773) was developed by Wise, and oral vaccination with this isolate protected catfish fingerlings (24).

Edwardsiella ictaluri can survive and replicate in channel catfish macrophages, and E. ictaluri LAVs induced cellmediated immunity to protect catfish against ESC (25–27). Also, catfish vaccinated with LAVs triggered humoral immune responses which augmented the bacterial killing activity of macrophages (25–27).

Recently, we demonstrated the phagocytic and killing properties of catfish peritoneal macrophages induced by two novel E. ictaluri LAV strains (Ei1evpB and ESC-NDKL1) developed in our laboratory which provided significant protection against ESC in both catfish fry and fingerlings (27– 31). Ei1evpB was constructed by in-frame deletion of the evpB gene, one of the main components of type six secretion system (T6SS) (14). ESC-NDKL1 (1gcvP1sdhC1frdA) was constructed by in-frame deletion of three genes in the tricarboxylic acid cycle (sdhC and frdA) and one-carbon metabolism (gcvP) (30, 31). However, the roles of these LAVs on the phagocytosis and intracellular killing properties in catfish B cells were still unexplored. Therefore, the purpose of this study was to assess the ability of channel catfish AK B cells to phagocytose and kill LAV and WT strains of E. ictaluri. Increased phagocytic and killing ability of catfish B cells will delineate the role of B cells in innate immune responses in E. ictaluri infection.

### MATERIALS AND METHODS

### Animals

Specific pathogen free (SPF) channel catfish were obtained from the fish hatchery at the College of Veterinary Medicine, Mississippi State University. All fish experiments were carried out based on a protocol approved by the Mississippi State University Institutional Animal Care and Use Committee (IACUC). Fish were maintained at 25–28◦C throughout the experiments. To sedate and euthanize the catfish, tricaine methanesulfonate (MS-222, Western, Chemical, Inc.) was used. Samples were obtained as described below.

### Bacterial Strains and Opsonization

Bacterial strains for this study are listed in **Table 1**. E. ictaluri 93– 146 wild-type (WT) and two LAVs strains were cultured in BHI agar or broth (Difco, Sparks, MD, United States), and incubated at 30◦C for overnight. Two LAVs and WT strains were labeled with bioluminescence by transferring pAKgfplux1 from an E. coli donor strain (SM10λpir) by conjugation as described previously (33). Ampicillin (Amp: 100 mg/ml), and colistin sulfate (Col: 12.5 mg/ml, Sigma–Aldrich, St. Louis, MN, United States) were added to media when they are required. E. ictaluri WT was incubated in the presence of 10% normal catfish serum for 30 min at room temperature.

### Cell Preparation

Channel catfish (150–200 g) were used in this study. Anterior kidneys (AK) were dissected from 5 catfish and placed in a sterile culture dish that contained Phosphate-buffered saline (PBS). Tissues were pooled and crashed by using sterile forceps and passed through cells dissociation sieves (Sigma, St. Louis, MO) to obtain a single-cell suspension of AK. After that, cells were resuspended and washed in PBS. Cell suspensions were layered on Histopaque 1077 (Sigma) and centrifuged at 500 g for 30 min to obtain enriched white mononuclear cells (WMCs). Following centrifugation, WMCs were collected from the interface and washed three times in PBS at 500 g for 10 min. Cells were counted and assessed for viability by using a hemocytometer and trypan blue exclusion. Finally, cells were used for phagocytosis assessment.

TABLE 1 | Bacterial strains and plasmids.


**Abbreviations:** E. ictaluri, Edwardsiella ictaluri; ESC, Enteric septicemia of catfish; LAV, Live attenuated vaccine; WT, Wild-type; AK, Anterior kidney; SPF, Specific pathogen free; BHI, Brain heart infusion; APCs, Antigen-presenting cells.

### Phagocytosis and Flow Cytometry

Mononuclear white blood cells from AK were resuspended in L-15 medium (ThermoFisher Scientific) as described previously (11) and stained with primary monoclonal antibodies (mAbs, clone 9E1) to catfish IgM<sup>+</sup> B cell-specific marker at 4◦C (34, 35) followed by the addition of isotype-specific fluorochrome (R-PE) conjugate (Mouse F (ab) 2 IgG (H+L). (R-PE), R&D Systems, Inc.). After the staining procedure, cells were washed three times with PBS (all the steps were performed at 4◦C in the dark).

Following the staining, AK WMCs were resuspended in L-15 medium and 2 × 10<sup>6</sup> cells per well were transferred into 6 well plates (FisherScientific, Pittsburgh, PA, United States). Green Fluorescence Protein (GFP) transformed bacterial strains were added in 1:50 ratio to each well and incubated at 30◦C in the dark for 30 min to determine the phagocytic ability of catfish B cells.

Catfish WMCs and lymphocytes were gated based on their relative size and granularity by using forward and side scatters, FSC and SSC, respectively (**Figure 1A**). After setting a gate on IgM<sup>+</sup> cells (**Figure 1B**), phagocytic B cells were determined based on the intensity of GFP fluorescence (**Figure 2**). The percentage of phagocytic B cells was determined by NovoCyte Flow Cytometry (ACEA Biosciences, Inc.) using two-color analysis with Dot Plot Quadrant statistics. Samples were analyzed using FlowJo 7.6.4 Software (Tree Star Inc.).

### Cell Sorting

IgM<sup>+</sup> B cells were positively selected by magnetic sorting from AK WMC populations as described previously with minor modifications (11). Briefly, WMCs were obtained from AK by using Histopaque 1077 separation, resuspended in L-15 medium and passed through the pre-separation filters (Miltenyi Biotec) to remove cell clumps. Then, mAbs specific to channel catfish IgM<sup>+</sup> B cells were used to positively identify catfish B cells (34, 35). Followed by incubation on ice for 30 min, cells were washed and resuspended in MACS buffer (Miltenyi Biotec), and anti-mouse IgG (H+L)-magnetic microbeads (Miltenyi Biotec) were added to cell suspensions, incubated at 4◦C for 15 min in the dark, washed and transferred onto a LS separation column (Miltenyi Biotec), according to manufacturer instructions. The purity of the resulting low size/low granularity IgM<sup>+</sup> populations was 85–94% by flow cytometry (**Figures 1C,D**). After magnetic separation, positively selected B cells were cultured in the L-15 medium at 28◦C overnight to detach the magnetic microbeads. Followed by Histopaque 1077 separation and assessment of viability of the resulting B cell populations by trypan blue exclusion (1– 3%), the purity of the sorted IgM<sup>+</sup> B cells was determined by a FACSCalibur Flow Cytometer (Becton Dickinson), and highly enriched by sorting IgM<sup>+</sup> B cells were used to assess the bacterial killing ability, morphology and apoptotic changes.

### Cytospin and Light Microscopy

Highly enriched IgM<sup>+</sup> and IgM<sup>−</sup> cell populations were incubated in the presence of E. ictaluri strains to characterize their morphology and phagocytic capacity. Following the incubation, cells were harvested and washed in PBS. Then, the cytospins were prepared at 500 rpm for 1 min by using a Cyto-Tek

FIGURE 1 | Separation of catfish B cells by flow cytometry. (A) Assessment of lymphocytes based on their size and granularity. (B) Identification of B cells based on the intensity of B cell-specific staining with mAbs specific to channel catfish IgM<sup>+</sup> B cells. (C) Assessment of highly enriched magnetically sorted IgM<sup>+</sup> cell population based on their size and granularity. (D) Identification of B cells based on the intensity of catfish B cell-specific staining. One of three representative experiments.

centrifuge machine. All samples were fixed on the slides, and Giemsa staining procedure (May-Grunwald Procedure) was applied to observe the morphology of IgM<sup>+</sup> cell population. The Wright's stain (Hemacolor, Merck) was applied to observe the IgM<sup>−</sup> cell population morphology as described previously (36). Samples were analyzed with an Olympus BX60 microscope (OlympusU-TV1 X) and photographed by using Infinity software (Lumenera Corporation).

### Bacterial Killing Assay

The bacterial killing assay was performed as described previously with some minor modifications (26, 37). Briefly, sorted B cells were resuspended with L-15 medium supplemented with 10% FBS, 1% L-glutamine, and 1.5% HEPES buffer and 0.5 × 10<sup>6</sup> cells per well were transferred to 96-well plates (Evergreen Scientific). Then, WT and two LAVs strains were added to cell suspension in 1:20 ratio that did not affect the viability of B cell populations by trypan blue exclusion (1–5%) for 48 h in culture compared to uninfected controls. Plates were centrifuged at 1,500 rpm for 5 min at room temperature to compact cells and bacteria and incubated at 30◦C for 30 min. After that, plates were centrifuged at 2,000 rpm for 7–10 min to remove the supernatant. Next, the pellet was resuspended with L-15 medium supplemented with 10% FBS, 1% L-glutamine, 1.5% HEPES buffer, and 100µg/ml gentamicin (Gibco, Life Technologies, Grand Island, NY, United States) to kill extracellular bacteria, and incubated at 30◦C for 1 h. Following the killing of extracellular bacteria, plates were washed in PBS and resuspended in L-15 medium containing 10µg/ml gentamicin (time 0) for 48 h incubation with 5% CO<sup>2</sup> at 30◦C in the black 96-well plates (Fisher Scientific) to determine the number of the live intracellular E. ictaluri in catfish B cells. Statistical analysis was acquired with the results obtained from Cytation 5 Cell Imaging Multi-Mode Reader (BioTek).

### Apoptosis Assay

The apoptosis assay was performed as described previously with minor modifications (38). Positively selected B cells were resuspended in L-15 medium and 2 × 10<sup>6</sup> cells per well were transferred to 24-well plates (Tissue Culture Plate, CELLTREAT). Then, WT and LAVs E. ictaluri strains were added to the plates in 1:50 ratio and incubated at 30◦C in the dark for 30 min and 3 h to detect early and late apoptosis in catfish B cells. After incubation, cells were collected and washed with cold PBS by centrifugation at 4◦C. Apoptosis in catfish B cells was assessed by using Annexin V-FITC Apoptosis Kit according to manufacturer's instructions (BioVision, Inc., Mountain View, CA). Briefly, cells were resuspended in 1x Binding buffer and incubated with Annexin-V-FITC and propidium iodide (PI) for 5 min at room temperatures in the dark. Samples were analyzed by NovoCyte Flow Cytometry using two-color analyses with Dot Plot Quadrant Statistics. Also, staurosporine (10µM, Sigma) treated catfish B cells were used as positive control for apoptosis.

### Statistical Analysis

One-way ANOVA with PROC GLM procedure in SAS (v 9.4, SAS Institute, Inc., Cary, NC) was used for the bioluminescence intensity of GFP-labeled E. ictaluri strains of antigen uptake and bacterial killing assay. Separate models were fit for each time point. The fixed effect for the bioluminescence intensity of GFP-labeled E. ictaluri strains was treatment. LSMEANS statements with TUKEY adjustment was used to evaluate significant differences among treatments for each model. An alpha level of 0.05 was used to determine statistical significance. The distribution of the residuals was evaluated for each model to make sure the assumptions of normality and homoscedasticity for the statistical method had been met.

Linear models with PROC MIXED in SAS for Windows 9.4 were applied for the percentage of live cells, early apoptotic cells, late apoptotic cells, and necrotic cells. Fixed effects for each outcome (live cells, early apoptotic cells, late apoptotic cells, and necrotic cells) of apoptosis assay were treatment, hour and their interaction. In the case of a significant interaction term, differences in least squares means between 0.5 and 3 h for each of the treatments and also between treatments for each hour were calculated by using an LSMESTIMATE statement. Also, the simulate adjustment for multiple comparisons was used in the case of the significant terms.

The level of significance for all tests was set at P < 0.05. The distribution of the residuals was evaluated for each model to make sure the assumptions of normality and homoscedasticity for the statistical method had been met.

### RESULTS

### Active Phagocytic Uptake of *E. ictaluri* WT and LAVs Strains in AK B Cells

In this study, we determined the active uptake of E. ictaluri WT and two LAVs strains in catfish B cells (**Figure 2A**). Separated from PBMC by two-color flow cytometry IgM<sup>+</sup> B cells actively endocytosed E. ictaluri LAV and WT strains (**Figure 2A**), and the phagocytic intensity levels of the LAV and WT strains of E. ictaluri in catfish AK B cells of three biological replicas did not differ significantly (P > 0.05, data not shown). Furthermore, we used an additional protocol for active bacterial uptake assessment in the AK B cells exposed to the WT E. ictaluri strain opsonized with normal catfish serum (**Figure 2B**). Separated by flow cytometry B cells actively endocytosed opsonized bacterial strain (**Figure 2B**). The data represent one of three biological replicas from the AK–derived mononuclear cells combined from five fish.

To confirm active phagocytic uptake in catfish B cells by flow cytometric approach, we assessed E. ictaluri WT and two LAVs strains uptake in highly purified magnetically sorted IgM<sup>+</sup> B cells by light microscopy (**Figure 3**). In addition to the apparent intracellular bacterial uptake of LAVs and WT E. ictaluri, the phagosome and/or phagolysosome formation was evident in the cytoplasm of catfish B cells (**Figure 3**). To characterize the suggested contrasting morphology and phagocytic capacity of the unlabeled (IgM−) cell populations, we assessed their incorporation of E. ictaluri WT strain by light microscopy (**Figure 3**). The IgM<sup>−</sup> cell populations showed typical macrophage morphology such as larger size and cytoplasm presence with dramatically increased numbers of the engulfed bacteria compared to the sorted IgM<sup>+</sup> B cells (**Figure 3**).

### Killing of *E. ictaluri* and LAVs Strains by Catfish B Cells

After determining the levels of phagocytosis of E. ictaluri and LAVs strains in catfish B cells, we examined how effective catfish B cells were at destroying the ingested bacteria at 30◦C by applying more sensitive Cell Imaging technology (**Figure 4**). IgM<sup>+</sup> B cells were positively selected by magnetic sorting from AK WMC populations (**Figures 1A–D**). WMCs and lymphocytes were gated based on their relative size and granularity by using forward and side scatters, FSC and SSC, respectively (**Figure 1A**). After setting a gate on IgM<sup>+</sup> cells (**Figure 1B**), mAbs specific to channel catfish IgM<sup>+</sup> B cells were used to positively identify catfish B cells (**Figure 1C**), The purity of the resulting IgM<sup>+</sup> populations was 85–94% by flow cytometry (**Figure 1D**). The initial intensity of the intracellular bacterial luminescence (time 0) did not show significant differences in B cells challenged with Ei1evpB and WT strains. However, the WT strain intensity was significantly higher compared to the luminescence intensity of ESC-NDKL1 in B cells (**Figure 4**). Negative control B cells not exposed to bacteria showed the background low levels of luminescence intensity (**Figure 4**). Significant differences in the intensity of luminescence between all challenges were documented at 1 h post-incubation (**Figure 4**). For example, the intensity of Ei1evpB luminescence in B cells at this time was significantly higher than the luminescence of WT and ESC-NDKL1 strains (**Figure 5**). Moreover, the luminescence of WT strain in B cells was significantly higher than ESC-NDKL1 luminescence (**Figure 4**). Interestingly, the luminescence of both LAVs and WT strains in catfish B cells significantly decreased at 2 h; however, there were significant differences between the challenges and negative control B cells not exposed to bacteria (**Figure 4**). The luminescence intensity of all E. ictaluri strains showed time-dependent gradual decreases, and there were no significant differences in the luminescence of ESC-NDKL1 and negative control group at 36 h in catfish B cells (**Figure 4**). However, the intensity of WT and Ei1evpB luminescence at 36 h was still significantly higher compared to the group challenged with ESC-NDKL1 and negative control group (**Figure 4**). Finally, there were significant decreases in the luminescence of Ei1evpB after 2 and 3 h of incubation compared to non-significant changes in the luminescence of ESC-NDKL1

B cells (93.9% purity, left column). Phagocytosis of *E. ictaluri* in IgM<sup>−</sup> macrophages (right column). (B) Active uptake of *E. ictaluri* LAV and WT strains in highly purified catfish B cells (93.9% IgM<sup>+</sup> cells) by light microscopy. Solid arrows indicate the engulfed bacterial strains by catfish AK B cells (*Ei*1*evpB*, ESC-NDKL1, and WT shown in left, middle and right columns, respectively). Arrows with dots indicate phagosomes and/or phagolysosomes in the cytoplasm of catfish B cells. 100 × magnification, scale bar 20 micrometers.

and WT in catfish B cells at these time points (P < 0.001). Timedependent significant decreases in bacterial luminescence were documented for both LAVs and WT strains, however more than 50% of the initial Ei1evpB uptake was eliminated at 4 h, ESC-NDKL1 uptake at 3 h and the WT uptake at 5 h post exposure (**Supplemental Figure 1**). Our data showed that catfish B cells were capable of killing E. ictaluri WT and both LAVs strains; however, they were more efficient at killing of ESC-NDKL1 than destroying of WT and Ei1evpB strains.

### Early and Late Apoptotic Changes in Catfish B Cells Exposed to WT *E. ictaluri* and LAVs

Early and late apoptotic changes (one of three biological replicas of AK-derived B cells combined from five fish in each experimental group) in B cells exposed to LAV and WT E. ictaluri strains have been assessed at 30 min (**Figure 5A**) and 3 h post-exposure (**Figure 5B**). Statistical analysis of early and late

replicas of the AK– derived B cells combined from five fish ±SD in each experimental group.

FIGURE 5 | The effects of *E. ictaluri* WT and LAVs strains on catfish AK B cell apoptosis. Early and late apoptotic changes in catfish magnetically sorted B cells exposed to *Ei*1*evpB*, ESC-NDKL1, and WT *E. ictaluri* strains at 30 min (A) and 3 h (B). Necrotic cells (Q1, PI+/Annexin V-); late apoptotic cells (Q2, PI+/Annexin V+); early apoptotic cells (Q3, Annexin V+/PI-); live cells (Q4, PI-/Annexin V-). The data represent one of three biological replicas of AK-derived B cells combined from five fish in each experimental group.

apoptotic, and necrotic changes in catfish AK B cells based on the three biological replicas is shown in **Figures 6A–D**. Interestingly, there were no significant differences in the percentages of live cells between B cells exposed to WT and LAV strains at both 30 min and 3 h incubation times, however as expected, the percentages of live B cells treated with staurosporine (positive control) were significantly decreased compared to other groups (**Figure 6A**). Also, there were significant differences in the percentages of live B cells between 30 min and 3 h incubation times with decreased percentages of live cells at 3 h post-treatment (**Figure 6A**). Furthermore, there were no significant differences in the percentages of early apoptotic cells between treatments exposed to LAVs and WT strains at 30 min; however, staurosporine induced significantly higher levels of early apoptosis at 30 min compared to other groups (**Figure 6B**). In addition, there were no significant differences in the percentages of early apoptotic cells between the groups exposed to staurosporine, WT and ESC-NDKL1 strains at 3 h post-treatment (**Figure 6B**). In contrast, Ei1evpB caused significantly less early apoptotic changes than staurosporine at 3 h at this time (**Figure 6B**). Moreover, there were no significant differences in the percentages of early apoptotic cells between 30 min and 3 h incubation times (**Figure 6B**). The percentages of late apoptotic cells at 3 h incubation time significantly increased in all treatments except negative control (**Figure 6C**). There was no significant difference in the percentages of late apoptotic cells between treatments at 30 min. However, the percentages of late apoptotic cells in the groups exposed to WT and LAV strains were significantly higher than in the group treated with staurosporine at 3 h (**Figures 6C,D**). There were no significant differences in the percentages of necrotic cells between treatments at 30 min; however, the percentages of necrotic cells at 3 h significantly increased in the group exposed to staurosporine only (**Figure 6D**).

### DISCUSSION

B cells possess the ability to capture antigens, process into peptides, and load the peptides onto MHC class II molecules for their presentation to CD4<sup>+</sup> T cells (39). Several studies reported that B cells in teleost fish served as professional antigen presenting cells (APCs). For instance, zebrafish B cells were able to present both soluble and particulate antigens to prime naïve CD4<sup>+</sup> T cells. Also, this study showed that the expression of MHC class II molecules and co-stimulatory molecules (CD86 and CD83) was upregulated in B cells during the presentation of antigens. Several studies have demonstrated that B cells in teleost fish were important APCs that activated T cells and initiated adaptive immunity in vivo and in vitro (11, 40, 41). Phagocytosis and intracellular killing activities are crucial properties of all professional APCs. Therefore, the current research aimed to determine the phagocytic and bactericidal activity of catfish AK B cells in the uptake of E. ictaluri WT and two LAV strains developed in our laboratory. In teleost fish, AK has a vital function as a hemopoietic organ, which produces all blood elements (42–45). Several studies reported that AK is one of the target organs in early of E. ictaluri infection. For instance, leukocytes including E. ictaluri were observed in the AK of channel catfish at 48 h post-infection (46). Also, bioluminescent E. ictaluri dispersion was found in the catfish AK at 15 min after intraperitoneal injection (47). Recently, the E. ictaluri-induced necrosis was demonstrated in the hemopoietic tissue part of catfish AK (48).

We documented phagocytic capability mediated through nonopsonic and opsonic receptors and bacterial killing activity of B cells in channel catfish against intracellular pathogens in vitro confirming the data obtained in the previous studies. First, we demonstrated that catfish B cells were able to engulf E. ictaluri WT and two LAV strains. Also, we documented phagocytic uptake of the WT strain opsonized with normal catfish serum. Notably, numerical but not significant increases in the intensity of phagocytic uptake have been documented by flow cytometry in the AK B cells exposed to Ei1evpB and WT strains compared to their counterparts challenged with ESC-NDKL1 LAV. Similarly, our previous data revealed that the intensity of Ei1evpB LAV phagocytic uptake was significantly higher compared to the ESC-NDKL1 strain in catfish peritoneal macrophages (27). However, significant differences in bacterial uptake between two LAVs and WT strains were evident in the bacterial killing assay after removal of all the attached bacterial cells. In addition, bacterial killing of the Ei1evpB LAV was significantly increased compared to bacterial killing of ESC-NDKL1 and WT strains in catfish B cells after 2 and 3 h of incubation. Phagocytic ability of B cells has been shown in several teleost fish, such as zebrafish, rainbow trout, Atlantic salmon and Atlantic cod (6, 11, 12). In addition, B-1 and MZ subsets of B cells in mammals were capable of uptake of pathogens (3). Moreover, a recent study in rainbow trout demonstrated that fish IgM<sup>+</sup> B cells share common phenotypic and functional characteristics of mammalian B1 cells (49).

Professional phagocytic cells, such as macrophages, recognized and engulfed pathogens into vesicles known as phagosomes that fuse with lysosomes to form phagolysosomes (50, 51). Ingested pathogens were destroyed and killed in phagolysosomes by enzymes and antimicrobial substances, such as NO (52). Like professional phagocytes, murine B-1 cells have been shown to mature their phagosomes into phagolysosomes (5). In this study, we observed phagosome and/or phagolysosome formation in the cytoplasm of catfish B cells by light microscopy. However, the size of phagosome and/or phagolysosome in AK macrophages was larger than those found in B cells. Moreover, ingested bacteria numbers in the phagosomes of AK macrophages were virtually higher than the numbers of bacteria in the B cell phagosomes confirming our recent report on WT and LAV E. ictaluri strains detected in the phagosomes of peritoneal macrophages in catfish (27). Furthermore, our data showed that catfish B cells were capable of destroying WT and LAV E. ictaluri strains. Kinetics of bacterial killing in catfish B cells were distinct than the kinetics described in peritoneal macrophages. The numbers of ingested bacteria decreased significantly in peritoneal macrophages at 10 h incubation period (27). However, the luminescence of internalized E. ictaluri significantly declined in catfish B cells after 2 h of incubation that correlated with increased numbers of early and late apoptotic and necrotic AK B cells between 30 min and 3 h of incubation. Although the microbicidal capacity of B cells was limited compared to peritoneal macrophages in catfish, the rapid destruction of bacterial antigens in B cells could provide early activation cues to specific T cells against ESC. Similar to mammals, internalization of particles by phagocytic B cells in teleost fish induced the formation of phagolysosomes with the fusion of lysosomes to phagosomes and exhibited the capability to kill ingested particles (6, 13). Another study in rainbow trout showed that sorted IgM<sup>+</sup> and IgT<sup>+</sup> B cells were able to kill internalized bacteria, Escherichia coli (9). Also, phagolysosome formation in mammalian B cells supported the killing activity of phagocytic B cells. B cells from the peritoneal cavity of mice engulfed Staphylococcus aureus (S. aureus), and uptake of bacteria led to the formation of phagolysosomes followed by activation the degradation pathways to kill the ingested bacteria (53).

Apoptosis, the process of programmed cell death, is identified by distinct morphological changes and biochemical modifications, such as protein cleavage and DNA fragmentation (54–56). Apoptosis is a crucial component of numerous processes including development, normal cell turnover, and the immune system, and this process occurs during normal development and aging (54). Also, apoptosis occurs in response to diverse physiological and pathophysiological stimuli and diseases (57, 58). In this study, we applied an apoptosis assay to detect early and late apoptotic changes in catfish B cells exposed to E. ictaluri WT and two LAV strains. Our data showed that E. ictaluri strains caused early and late apoptosis in catfish B cells. It was demonstrated previously that a few numbers of mouse B cells from peritoneal cavity incubated with S. aureus underwent apoptosis (53). Moreover, Trypanosoma brucei induced the loss of IgM<sup>+</sup> B cell population in mice by causing apoptosis (59). In addition, Mycoplasma bovis induced apoptosis of lymphocytes in the bovine model (60).

In conclusion, our study demonstrated that efficacious E. ictaluri LAVs facilitate the phagocytic activity and effective killing of internalized bacteria in channel catfish B cells. We also documented enhanced phagocytic and microbicidal ability in catfish B cells exposed to the Ei1evpB LAV compared to the ESC-NDKL1 counterpart. For the first time, we documented the presence of phagosome and/or phagolysosome formations and the engulfed bacteria in AK B cells of catfish. These results suggest that both LAVs exploit similar to WT E. ictaluri innate immunological mechanisms such as active phagocytic uptake, bactericidal activity and promote early and late apoptotic changes in catfish B cells. However, further research is needed to assess the role of B cells in professional antigen presentation of bacterial-derived peptides to specific T cells and activation of protective adaptive immune responses against ESC in channel catfish. Although identification of non-opsonic receptors recognizing and binding to chemical structures on the surface of bacteria involved in the phagocytic uptake was beyond of the scope of this study, further research should address the molecular mechanism of receptor-mediated endocytosis of E. ictaluri as well as other intracellular pathogens in catfish B cells.

### DATA AVAILABILITY STATEMENT

All datasets generated for this study are included in the manuscript/**Supplementary Files**.

### ETHICS STATEMENT

All fish experiments were carried out based on a protocol approved by the Mississippi State University Institutional Animal Care and Use Committee.

### AUTHOR CONTRIBUTIONS

LP and AK conceived and designed the experiments and provided the original idea of the study, and also contributed reagents, materials, and tools. AOK, SK, KM, HA, and WT performed the experiments. AOK wrote the first draft of the

### REFERENCES


manuscript and was involved in all aspects of the study. All authors were involved in critical interpretation of the data, manuscript revision, and final version approval.

### FUNDING

The funding for this research was provided by Agriculture and Food Research Initiative competitive grant no. 2016-67015-24909 from the USDA National Institute of Food and Agriculture. The funding for the flow cytometry core facility was provided by NIH COBRE P20GM103646 grant.

### ACKNOWLEDGMENTS

The authors thank the Flow Cytometry Core Facility at the College of Veterinary Medicine. Bioluminescence imaging was supported by USDA-ARS Biophotonics Initiative #58-6402-3-018. We acknowledge the assistance of Sinan Kordon in preparation of the figures.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2019.02383/full#supplementary-material

Supplemental Figure 1 | Kinetics of the engulfed *E. ictaluri* LAV and WT strains killing in catfish B cells. Letters show significant differences between the time points (*P* < 0.0001). One way ANOVA with PROC GLM procedure in SAS for 9.4 was used for bacterial bioluminescence intensity. Initial uptake of bacteria at time 0 was considered 100%. The data represent the mean of four biological replicas of the AK– derived B cells combined from five fish ±SD in each experimental group.


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

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